FEATURE pubs.acs.org/est
Greening Coal: Breakthroughs and Challenges in Carbon Capture and Storage Philip H. Stauffer,*,† Gordon N. Keating,† Richard S. Middleton,† Hari S. Viswanathan,† Kathryn A. Berchtold,‡ Rajinder P. Singh,‡ Rajesh J. Pawar,† and Anthony Mancino§ †
Earth and Environmental Sciences Division (EES), Los Alamos. National Laboratory (LANL) Materials Chemistry Division at LANL § International Research, Analysis, and Technical Development Division at LANL ‡
’ INTRODUCTION Carbon capture and storage (CCS) is a critical technology for reducing emissions of greenhouse gases (GHGs) to the atmosphere. CCS is being considered as one piece of a strategy for stabilizing atmospheric CO2 concentrations.1 This plan requires that globally, billions of tonnes of carbon dioxide (GtCO2) each year must be captured, concentrated, and stored to keep it out of the atmosphere for hundreds to thousands of years. The nearterm approach is to capture and compress CO2 from stationary industrial sources (e.g., coal and natural gas burning power plants) and transport it through pipelines for injection and long-term storage in geologic reservoirs (e.g., depleted oil/gas fields and deep saline aquifers). In 2009, U.S. coal power plants generated 307 gigawatts of electricity (GWe) and produced 2.4 GtCO2 out of total U.S. emissions of 6 GtCO2.2 The existing fleet of coal-fired power plants will continue to be a major source of electricity for the next 20 years, with estimated production capacity increasing to 400 GWe.3 In addition, electrical generation in China has expanded rapidly in recent years, nearing the size of the U.S. fleet,4 and three-quarters of China’s power plants burn r 2011 American Chemical Society
coal.5 Given the persistence of this global capacity for coal combustion for the next few decades, CCS represents a bridging technology that will allow us to continue to generate electricity in existing power plants while we transition to a low-carbon energy future. CCS technology must be deployed at a massive scale to have a meaningful impact on reducing industrial CO2 emissions to the atmosphere. This could require the U.S. to capture on the order of 1 GtCO2/yr from hundreds of typical coal-burning power plants and to construct dedicated pipelines to handle a CO2 volume 25% greater than the U.S.’ 2009 daily oil consumption.6 Additionally, this volume of CO2 will require finding extensive geologic formations in low risk environments to store between 1 - 3 km3 of supercritical CO2 each year. This paper explores the science and technology related to CO2 capture, geologic storage, and system-wide integration. We emphasize strategies and technologies suitable for making CCS a reality in the near future, with particular focus on retrofitting existing coal-fired power plants to capture and compress CO2 for geologic storage. Projects involving coal combustion retrofits currently represent an area of particular focus for implementing CCS in the power industry in the next decade. Examples include turbine retrofit and capture of 1 million tonnes per year (MtCO2/yr) at the coal-fired Boundary Dam plant (SaskPower) in Saskatchewan,7 oxy-fuel combustion retrofit and capture of 1.3 MtCO2/yr in the FutureGen 2.0 plant (Ameren) in Illinois,8 and postcombustion capture of 3,000 tCO2/yr at the Gaobeidian power plant in Shanghai.9
’ CO2 CAPTURE Power plants are responsible for greater than one-third of the CO2 emissions worldwide and are a prime focus of global CCS efforts.10 Capturing CO2 from the mixed-gas streams produced during power generation is a first and critical step for CCS. Three strategies for incorporating capture into power generation scenarios are of primary focus today: post-, pre- and oxy-combustion capture (Figure 1). Postcombustion capture systems are designed to separate CO2 from pulverized coal (PC) derived flue gas. PC flue gas contains 10 13% CO2 with a balance of N2, steam, and other impurities (SOx, NOx, heavy metals).11 Oxy-combustion power plants are modified versions of conventional PC plants using Published: September 09, 2011 8597
dx.doi.org/10.1021/es200510f | Environ. Sci. Technol. 2011, 45, 8597–8604
Environmental Science & Technology
FEATURE
Figure 1. Post-, oxy-, and precombustion concepts and separation system integration into power plants.
oxygen (O2) diluted with recycled flue gas instead of air to combust coal into steam and high purity CO2. Precombustion systems are designed to separate CO 2 from synthesis gas (“syngas”) prior to electricity/hydrogen (H2) production and are applicable to new, more efficient, integrated gasification combined cycle (IGCC) plants. Syngas from the coal gasifier is primarily comprised of H2and carbon monoxide (CO). A watergas-shift reactor is added in the capture process to convert CO to CO2, thus facilitating capture while producing additional hydrogen. Power generation with capture using oxy- and precombustion processes has 10 37% higher net efficiency than that of a new air-fired PC plant without CO2 capture and allows more flexibility for future improvements in design.12 The U.S. Department of Energy (DOE) CCS target is to achieve 90% CO2 capture while limiting the increase in cost of electricity (COE) to 35 and 10%, respectively, for plants implementing postcombustion and pre/oxy-combustion capture. 13 The energy consumption and losses associated with CO2 capture using today’s technologies represent an unacceptably high proportion (>75%) of the total cost of CCS (capture/compression, transport, storage). The DOE cost targets require significant improvements in large scale deployable capture technologies. Although all of these technologies will
be vital in the long term, current U.S. and global power production industries are dominated by PC-based plants. Therefore, postcombustion capture is the most likely to have the largest impact on total CO2 emissions reductions over the next few decades. The U.S. Energy Information Administration estimates that, in 2030, 78% of the CO 2 emissions resulting from U.S. electricity generation will still be derived from the current fleet of PC plants.14 Thus, to effect change in the near term, we believe that these PC emissions must be addressed to a large extent through postcombustion separation and capture retrofits to those plants. CO2 separation technologies are readily available and have been used industrially for nearly 60 years. These technologies are based on chemical solvents (e.g., monoethanolamine, MEA) and physical solvents (e.g., glycol or methanol). A regeneration step is used to reclaim the solvent for reuse. Although technically suited for CO2 capture applications, implementation is hindered by exorbitant operating costs due to high energy penalties for solvent regeneration and material and environmental costs due to solvent attrition. In one study, the estimated cost of implementing these existing capture technologies in the operational U.S. power plants will increase the COE by 330%.13 8598
dx.doi.org/10.1021/es200510f |Environ. Sci. Technol. 2011, 45, 8597–8604
Environmental Science & Technology The development of new and innovative capture systems for PC application is imperative. Fortunately, increasing research budgets have enabled scientists and engineers worldwide to make progress in this burgeoning field. Emerging CO2 separation technologies under various stages of development include solvents, sorbents, membranes, and biologically mediated separation systems. A brief summary describing the important aspects of these emerging technologies related to CO2 capture from coalderived power production is provided below. Technology development goals for improved solvents include the realization of low-cost, noncorrosive, stable, low-toxicity materials with high CO2 capacity, rapid mass transfer kinetics, low regeneration energy, and high impurity tolerance.12,13,15 17 Aqueous ammonia (AA) and ionic liquid (IL) based materials are forerunners in current solvent research, development, and demonstration (RD&D) efforts. Systems such as these provide an opportunity for use of less corrosive, more stable solvents with chemically tunable mass transfer rates and capacities, thereby addressing some of the limitations of the more conventional amine-based materials. AA technologies are under pilot-plant and midscale demonstration while IL based solvents are currently at the laboratory development stage. Solid sorbents work by adsorbing gaseous CO2 onto a surface, followed by temperature or pressure driven desorption. CO2 interacts with the sorbent chemically (e.g., immobilized amines and carbonates) or physically (e.g., high surface area metal organic frameworks and zeolites).12,13,15 17 Since these CO2sorbent interactions are weaker than those between CO2 and chemical solvents, less heat is typically required for regeneration and CO2 release.15 Preliminary cost analyses indicate 15% improvement in COE using sorbents as compared to MEA capture.18 Process and materials optimization remain challenges for this technology. As with sorbents, materials cost, stability, CO2 capacity and mass transfer optimization are critical to commercial viability. Additionally, movement of large volumes of solids (e.g., fluidized beds), and the mechanical and thermal robustness required for such process schemes, provide additional process, and materials challenges. Membranes are currently top contenders among new postcombustion CO2 separation technologies due to their low energy consumption, lack of moving parts, and modular design opportunities. However, they incur extra cost due to flue gas compression required to create the driving force for transport, post cleaning, and/or multistage operation; membranes must also be stable in the presence of flue gas contaminants and high temperatures. Both organic and inorganic membrane approaches are being pursued. Development is currently focused on achieving high CO2 throughput with adequate selectivity using lowcost materials. The RD&D performance of organic membranes has the potential to reduce capture costs to as low as $23/tCO2, significantly lower than the $54/tCO2 cost using existing industrial amine based separation technologies.19,20 Further development and pilot-scale efforts are ongoing to fully quantify and increase the estimated cost saving of using membranes for CO2 capture. New efforts employing mechanically robust roomtemperature IL based membranes hold promise for realizing unprecedented CO2 throughputs and capture costs of <$20/ tCO2.21,22 As we transition to a younger and more efficient power production fleet, oxy- and precombustion technologies will play a greater role. To be prepared for this transition, we must continue to work now to develop technologies that will support
FEATURE
the implementation of these next-generation fossil fuel-driven plants. Oxy-combustion provides an opportunity for near-complete capture; however, this process requires a large-scale air separation unit to produce high-purity O2. Although cryogenic air separation is a well-established method for O2 production, it is both capital and energy intensive.23 Several novel air separation technologies currently under development have the potential to reduce this cost. Engineering studies indicate that these new technologies can reduce the power associated with O2 production by 70 80%.12 By example, incorporation of an ion transport membrane for O2production is estimated to reduce total plant costs by more than $130/kW over the same plant with cryogenic separation technologies.24 IGCC with precombustion capture has the potential to lower power production costs over conventional PC-based power production with capture. Plant efficiencies for IGCC power plants with capture are estimated to be 6% greater than those of a conventional PC power plant with capture.25 The extent of the predicted efficiency gains and associated cost differentials is dependent on numerous factors including coal type, plant design, and plant location.25 27 However, neither the technology advantages of IGCC power production nor the DOE CCS targets are fully realized using current, commercially available CO2 separation technologies. Globally, advanced precombustion capture technologies based on solid sorbents, membranes, and advanced solvents are currently all at demonstration stages from laboratory to full scale. High temperature membranes and sorbents with separation conditions (temperature and pressure) matched to IGCC process conditions are emerging as promising routes to efficient H2/CO2 separation.13,16,17 In addition to gains from capture technologies, advances continue to be made in other process areas such as turbines and O2 production. The incorporation of these capture and other process enhancements into advanced IGCC plant designs shows promise for additional efficiency gains in excess of 9% over that of the baseline IGCC case referenced above. Realization of these developing technologies at the commercial scale would thus enable realization of the DOE goals of 90% capture with a <10% increase in COE.24
’ CO2 STORAGE Once captured and compressed, CO2 must be stored or sequestered for hundreds or thousands of years. Several storage options are available, but storing or sequestering CO2 deep in the earth is the only technology that has been established on scales large enough to be useful in our quest to divert billions of tonnes of CO2 from the atmosphere each year.28 Geologic storage has already been demonstrated, without major incident, by the oil industry for nearly 40 years. The oil industry uses CO2 to lower the viscosity of subsurface oil and thus enhance oil recovery (EOR) in depleted oil fields. Not all the CO2 injected is recovered, thus CO2 EOR projects are de facto CO2 storage sites. For example, the SACROC facility in West Texas is estimated to have stored approximately 55 MtCO2 since the early 1970s.29 Additionally, more than 7500 km of CO2 pipelines currently operating in the U.S. demonstrate that the technology to move large quantities of CO2 is safe.30 However, the scale of EOR in the U.S. is 2 orders of magnitude smaller than the total U.S. CO2 emissions.31 Over the past decade, governments around the world have teamed with industrial partners to initiate the first large pilot tests (>1 MtCO2/yr) of geologic sequestration in or near oil/gas 8599
dx.doi.org/10.1021/es200510f |Environ. Sci. Technol. 2011, 45, 8597–8604
Environmental Science & Technology operations (e.g., In Salah and Sleipner). Geologic sequestration of CO2 has initially targeted oil and gas reservoirs because they have existing infrastructure, operators have intimate knowledge of reservoir performance, and naturally occurring CO2 being produced as a waste stream near these sites provides a ready source for injection.32 However, the total storage in existing oil and gas formations is relatively small, and the proximity of these reservoirs to CO2 sources is not optimal. For example, CO2 from sources in the Illinois Basin greatly exceed the capacity of depleted oil and gas reservoirs in this region, leading current research in the direction of using deep saline aquifers to provide the required storage volumes.33 Deep saline aquifers are nearly ubiquitous worldwide, with an estimated potential storage volume up to 10 000 GtCO2. In addition, deep saline waters typically do not meet drinkingwater standards (10 000 ppm TDS) and thus will not be negatively impacted by CO2 mixing.34 Deep saline injection is currently planned for the Mt. Simon sandstone in the Illinois Basin in the U.S. as part of the FutureGen 2.0 project. Other proposed geologic storage formations include off-shore sediments, basalt flows, and unmineable coal seams. To understand how CO2 is trapped in the subsurface, numerous coupled processes must be considered. Primary trapping mechanisms include structural trapping (separate phase, buoyant CO2 is trapped by a geologic structure), residual trapping (discontinuous CO2 bubbles are trapped after the bulk of the CO2 flows through a system), solubility trapping (CO2 dissolves into brine or oil), and mineral trapping (CO2 leads to formation of minerals in the pore space of the rocks). Structural and residual trapping are likely to be the dominant processes on a scale of tens to hundreds of years, while mineralization and convective mixing operate over longer time scales.35,36 The physical processes controlling these mechanisms depend on many parameters, including temperature, pressure, salinity, pore structure, and capillary forces. Several of these same coupled processes can negatively impact geologic sequestration systems. Geo-mechanical impacts can be caused by changes in pressure and temperature. First, lowered effective confining stresses caused by increased pore pressure can allow existing faults to reactivate and cause CO2 to migrate upward. Second, thermal contraction caused by changes in temperature as cool CO2 is injected into hot rock can damage rocks around the injection location and cause wellbores to fail. Injected CO2 may also affect the chemistry of deep saline reservoirs, resulting in the formation of minerals that can reduce permeability or dissolution of minerals that can increase permeability. Near injection wells, crystalline salt forms in rock pores due to evaporation of water into supercritical CO2 and can lead to reduced injectivity of CO2. To incorporate this level of complexity, state-of-the-art process level simulators now include thermo-hydro-chemical-mechanical (THCM) processes. Although subsets of these coupled processes have been investigated previously for other applications, mechanistic coupling of more than two coupled processes has not been investigated extensively and represents an exciting research frontier.37 Input to these numerical models comes from experiments, pilot studies, and natural analogue sites where CO2 has been stored in geologic reservoirs for tens of thousands of years.38 For example, measurements of element solubility in the presence of CO2 in the laboratory and in the field provide modelers with targets to ensure that models can reproduce field data.39,40 Calculations of complex coupled processes provide the
FEATURE
basis for reduced-complexity algorithms that are being used in Monte Carlo risk assessment modeling of sequestration systems with potentially dozens of uncertain parameters.31,41,42 Teams around the world are working to assess risk for geologic storage sites. Such risk assessments are vital to the success of geologic sequestration and provide regulators and citizens with confidence in site selection and performance goals.43 Performance metrics for risk include aspects of health and human safety, environmental degradation, economic impacts, and project planning. In addition to risk assessment undertaken by industry (e.g., CO2 Capture Project44), three major efforts are being coordinated by the International Energy Agency, the U.S. National Risk Assessment Program and the Canadian International Performance Assessment Centre for Geologic Storage of CO2. Primary risks identified by these groups include leakage of brine and CO2 from the storage reservoir into overlying aquifers or other subsurface resources (e.g., contamination of oil fields), induced seismicity triggering damaging earthquakes (e.g., a recently abandoned geothermal project in Switzerland45), lack of injectivity that severely limits the capacity of proposed storage reservoirs, leakage back into the atmosphere on a scale large enough to negate the sequestration effort (greater than 0.01% per year46), and point-source leaks to the surface that could impact human health. These risks are characterized by a combination of high consequences with generally low probability. Many of these risks can be reduced through a combination of site selection, reservoir pressure management, site data collection, and longterm monitoring with contingency plans.47 50 Recent developments from analysis of the deep, permeable sedimentary basins of western Wyoming highlight the interplay between the complex costs and benefits associated with CCS. For example, the Rock Springs Uplift (RSU) is currently being considered as a target reservoir for large scale CO2 sequestration (approximately 26 GtCO2 capacity51), and modeling results show that reservoir management must be considered as an integral part of a total sequestration project to minimize risk of leakage. During reservoir management, brine can be removed (produced) from the reservoir to reduce pressure. The volume of brine production required to reduce seismic and leakage risks to near zero is approximately equal to the volume of injected CO2.47 For the Jim Bridger power plant near the RSU, nearly a cubic kilometer of CO2 could be injected over 50 years, leading to a cubic kilometer of produced brine.52 Produced brines at the RSU will arrive at the surface at temperatures in excess of 100 °C and are a potentially valuable thermal energy resource. The brine can also be desalinated; representing a potential source of freshwater in the arid western U.S. Coproduction of brine is likely to be an attractive resource associated with CO2 injection in many other parts of the world, including the Ordos Basin in China, currently under study by the joint U.S.-China Advanced Coal Technology Consortium.
’ SCALING UP: MAKING CCS A REALITY National goals for CO2 management will require CCS technology to be deployed in a very different manner than existing energy infrastructure. Oil, natural gas, and electricity transmission infrastructures expanded in an ad hoc and incremental fashion over many decades, primarily in response to emerging markets, resource discoveries, and population growth. In contrast, CCS will require a coherent development strategy that considers the intersection of policy, science, and industry. 8600
dx.doi.org/10.1021/es200510f |Environ. Sci. Technol. 2011, 45, 8597–8604
Environmental Science & Technology
FEATURE
Figure 2. (a) SimCCS infrastructure required to transport/store 80 MtCO2/yr from an oil shale industry. (b) Costs to transport/store CO2 and network length for 26 different CO2 management scenarios (5 130 MtCO2/yr).
Further, CCS technology will have to be applied over the next two or three decades to have a significant impact. Investing in infrastructure at this scale will necessitate careful and comprehensive planning. Specifically, decisions will need to be made regarding where, how much, and with what technology to capture CO2; where and how much CO2 to store in geologic reservoirs; where and what capacity pipelines to construct; and how to allocate CO2 between numerous CO2 sources and sinks. Thus, CCS integration may best develop at the basin scale to make best use of natural resources, clustering of industry and population centers, and topography.53 These integrated networks will realize economies of scale and proper infrastructure placement that will keep costs down and minimize environmental issues such as residential exposure to CO2 hazards. Numerous models have been developed to understand how an integrated CCS system—CO2 capture, transport, and storage infrastructure—could and should be deployed. Early CCS integration models focused on straightforward source-sink matching, typically connecting CO2 sources directly to their closest geologic reservoir or sink. These models relied on simplifying assumptions, including geographically impervious straight pipelines,54 that all CO2 must be captured from a source regardless of system-wide economics,55 and that pipelines cannot be networked
to aggregate CO2 flows.56 These early approaches paved the way for more comprehensive CCS models that take into account detailed economics coupled with spatially realistic pipeline networks. For example, a carbon management decision support system, SimCCS,57 simultaneously optimizes the financial investments, operational costs, system capacities, and geospatial construction of CCS infrastructure, as well as routing and networking pipelines across a real-world cost surface. Coupling reservoir performance and risk assessment models (e.g., CO2-PENS,31,41) with system optimization models (e.g., 53,57) provides a more coherent understanding of the effects of reservoir capacity and costs as they relate to optimal pipeline network design (Figure 2), and the propagation of uncertainty through the CCS system.6 Following the more comprehensive approach taken by SimCCS, infrastructure models have evolved to address other sophisticated and critical aspects of infrastructure modeling. These state-of-the-art modeling techniques illustrate how CCS infrastructure may evolve in response to a range of policy decisions. For example, Morbee et al.58 extended the SimCCS optimization to allow infrastructure to be constructed gradually through time as the amount of CO2 to manage varies (e.g., cap-and-trade). Kuby et al.59 contrast optimal CCS systems deployed in response to a CO2 tax versus a cap-and-trade environment. The effect of introducing CO2 8601
dx.doi.org/10.1021/es200510f |Environ. Sci. Technol. 2011, 45, 8597–8604
Environmental Science & Technology certificates or permits on CCS technology in Europe has been explored by Mendelevitch et al.:60 higher priced certificates result in greater adoption of CCS technology and reduce CO2 emissions. CCS deployment in The Netherlands has been mapped out by Van den Broek et al.61 using a GIS-based optimization energy model; study results could help policy makers reduce CO2 emissions by as much as 50% of 1990 levels by 2050. These next-generation models will be used by industry to make informed decisions about managing CO2. For example, the Alberta tar sand oil industry, projected to produce as much as 108 MtCO2/yr by 2020,62 is planning an integrated network to aggregate CO2 from multiple locations into one or more CO2 reservoirs.63 This network could reduce Alberta’s CO2 emissions by 35 MtCO2/yr by the mid 2020s and perhaps as much as 100 MtCO2/yr in the long term. An analysis of the CO2 management requirements for a mature U.S. shale oil industry, producing 1.5 million barrels of oil a day, concludes that an integrated pipeline network is the only feasible approach.53 Currently, the cost of CO2 capture ca. $54/tCO2,19,20 amounts to the greatest share of the cost of CCS integration, compared to the cost of transport and geologic storage (<$10/ tCO2 53,64). However, as new capture technologies come online and CO2 transport and storage networks are developed at the regional scale, these component costs are expected to converge. Although CCS development at the national scale is unlikely to be optimally designed, there are emerging examples of large-scale infrastructure integration. For instance, Southern Company is planning to implement carbon management across its fleet of coal power plants, using existing natural gas pipeline rights-ofway to design CO2 transport and storage infrastructure in advance of national carbon pricing (or tax) policies. This CCS integration is part of a business model of vertical integration within the utility.65 In addition, Denbury is converting natural gas lines and building new pipelines for CO2 to expand connectivity among CO2 sources (natural and anthropogenic) and EOR projects along the U.S. Gulf coast.66
’ CONCLUDING REMARKS CCS is a critical technology for reducing near-term CO2 emissions as economies transition to a low-carbon energy future. Here, we have identified examples of key scientific and technological breakthroughs and challenges that are driving the capture and storage of CO2, focusing on retrofits of coal combustion plants to highlight significant early implementation. We have also highlighted next-generation approaches that model how concerns of science, policy, and industry can be simultaneously addressed to make CCS a reality. The last 10 years have seen significant improvements in all areas of CCS science, technology, and modeling, while industry investment is building a base of operational knowledge. The development of new technologies could reduce the cost of CO2 capture to <$20/tCO2 within a decade (e.g., membrane technology). Current large geologic sequestration projects are now targeting injection rates of greater than 1 MtCO2/yr, linked to industrial CO2 sources. Even in the absence of federal carbon policy, companies like Southern Company, Denbury, SaskPower, and Ameren are building integrated CCS infrastructure for managing natural and anthropogenic CO2 in large part from coal-fired power plants. The early CCS projects that focus on EOR benefits and public-private demonstration plants illustrate the benefits of integrated systems and economies of scale. At the same time, these projects highlight the precarious nature of costly
FEATURE
technology investments in the absence of national carbon economic policies. Independent of the progress on the scientific and engineering research side of this work, strong political will and public education will be required if we are to realize CCS on a scale that can neutralize anthropogenic impacts to the Earth’s climate. Major hurdles to success remain in all areas of CCS. In capture, the most promising new technologies are being demonstrated at the bench scale, and reaching target DOE costs for capture will require scaling these technologies up first to the pilot then the industrial scale. Scale is also a major issue for storage; for CCS to have a meaningful impact, the volumes of injected CO2 will be considerably larger than anything previously attempted. For example, we will need to confirm that our current models of rock failure and fluid flow in fractures are appropriate when very large areas of the subsurface become pressurized with CO2. We must address issues of liability and pore space ownership, where a growing patchwork of individual state regulations could limit the interest of industry to invest in CCS projects. Finally, scaling up CCS from pilot projects to a set of integrated networks nationwide will require vision and planning. Lessons learned during early industrial-scale demonstrations can guide the development of CCS systems—capture technologies, dedicated pipelines, and storage reservoirs—that are capable of reliably and cost-effectively reducing CO2 emissions on a meaningful scale.
’ AUTHOR INFORMATION Corresponding Author
*E-mail: stauff
[email protected].
’ BIOGRAPHY Philip H. Stauffer, Hari S. Viswanathan, and Rajesh J. Pawar are researchers at Los Alamos. National Laboratory (LANL) in the Earth and Environmental Sciences Division (EES). Their research on carbon sequestration emphasizes reservoir simulation, including fully coupled chemical/stress/flow models, and risk analysis. Richard S. Middleton and Gordon N. Keating, also of EES at LANL, focus on systems level models linking infrastructure and policy for CCS projects. Kathryn A. Berchtold and Rajinder P. Singh research carbon capture technologies with a focus on innovative membranes and are in the Materials Chemistry Division at LANL. Anthony Mancino is an expert illustrator also working at LANL in the International Research, Analysis, and Technical Development Division. ’ ACKNOWLEDGMENT This work was supported by the U.S. Department of Energy, Office of Fossil Energy and the State of Wyoming. Finally, the insight of four anonymous reviewers helped to strengthen the paper. ’ REFERENCES (1) Pacala, S.; Socolow, R. Stabilization wedges: Solving the climate problem for the next 50 years with current technologies. Science 2004, 305 (5686), 968–972. (2) U.S. carbon dioxide emissions in 2009: A retrospective review. http://www.eia.doe.gov/oiaf/environment/emissions/carbon/ (accessed September 20, 2011). (3) Annual Energy Outlook 2008, Revised to Include the Impact of H.R. 6, Energy Independence and Security Act of 2007; DOE Energy Information Administration: Washington, DC, 2008. 8602
dx.doi.org/10.1021/es200510f |Environ. Sci. Technol. 2011, 45, 8597–8604
Environmental Science & Technology (4) International Electricity Installed Capacity Data. http://www. eia.doe.gov/emeu/international/electricitycapacity.html (accessed September 20, 2011). (5) China’s Power Sector Reforms: Where to Next?; International Energy Agency, 2006; www.iea.org/textbase/nppdf/free/2006/chinapower.pdf (accessed September 20, 2011). (6) Keating, G. N.; Middleton, R. S.; Viswanathan, H. S.; Stauffer, P. H.; Pawar, R. J. How storage uncertainty will drive CCS infrastructure. Energy Procedia 2011, 4, 2393 2400; DOI: 10.1016/j.egypro.2011.02.132. (7) International Test Centre (ITC) for CO2 Capture, IEAGHG RD&D Projects Database. http://www.co2captureandstorage.info/ project_specific.php?project_id=55 (accessed September 20, 2011). (8) FutureGen Industrial Alliance Announces Carbon Storage Site Selection Process for FutureGen 2.0. http://www.netl.doe.gov/technologies/coalpower/futuregen/index.html (accessed September 20, 2011). (9) Gaobeidian Power Plant, IEAGHG RD&D Projects Database. http://www.co2captureandstorage.info/project_specific.php?project_ id=186 (accessed September 20, 2011). (10) Coal-Fired Power Plants in the United States:Examination of the Costs of Retrofitting with CO2 Capture Technology, Revision 3; National Energy Technology Laboratory, 2011; p 56, http://www.netl.doe.gov/ energy-analyses/pubs/GIS_CCS_retrofit.pdf (accessed September 20, 2011). (11) Existing Plants, Emissions and Capture—Setting CO2 Program Goals; National Energy Technology Laboratory, 2009; p 41, http:// www.netl.doe.gov/technologies/coalpower/ewr/co2/pubs/EPEC% 20CO2%20Program%20Goals%20Final%20Draft%20v40409.pdf (accessed September 20, 2011). (12) DOE/NETL Advanced Carbon Dioxide Capture R&D Program: Technology Update, May 2011 ed.; National Energy Technology Laboratory, 2011; p 118, http://www.netl.doe.gov/technologies/coalpower/ ewr/pubs/CO2Handbook/ (accessed September 20, 2011). (13) DOE/NETL Carbon Dioxide Capture and Storage RD&D Roadmap; National Energy Technology Laboratory, 2010; p 78, http:// www.netl.doe.gov/technologies/carbon_seq/refshelf/CCSRoadmap.pdf (accessed September 20, 2011). (14) Annual Energy Outlook. http://www.eia.doe.gov/oiaf/archive/aeo10/index.html (accessed September 20, 2011). (15) Ciferno, J. P.; Fout, T. E.; Jones, A. P.; Murphy, J. T. Capturing carbon from existing coal-fired power plants. Chem. Eng. Prog. 2009, 105 (4), 33–41. (16) Figueroa, J. D.; Fout, T.; Plasynski, S.; McIlvried, H.; Srivastava, R. D. Advances in CO2 capture technology—The US Department of Energy’s Carbon Sequestration Program. Int. J. Greenhouse Gas Control 2008, 2 (1), 9 20; DOI: 10.1016/s1750-5836(07)00094-1. (17) CO2 Capture Technology Meeting; http://www.netl.doe.gov/ publications/proceedings/10/co2capture/index.html. (18) Nelson, T.; Coleman, L.; Anderson, M.; Herr, J.; Pavani, M. The dry carbonate process: Carbon dioxide recovery from power plant flue gas, In CO2 Capture Technology for Existing Plants, NETL R&D Meeting, Pittsburgh, PA, 2009. (19) Merkel, T. C.; Lin, H. Q.; Wei, X. T.; Baker, R. Power plant post-combustion carbon dioxide capture: An opportunity for membranes. J. Membr. Sci. 2010, 359 (1 2), 126-139; DOI: 10.1016/ j.memsci.2009.10.041. (20) Carbon Dioxide Capture from Existing Coal-Fired Power Plants. 401/110907; DOE/NETL, 2007; p 229, http://www.netl.doe. gov/energy-analyses/pubs/CO2%20Retrofit%20From%20Existing% 20Plants%20Revised%20November%202007.pdf (accessed September 20, 2011). (21) Berchtold, K. A.; Noble, R. D.; Gin, D. L.; Singh, R. P.; Del Sisto, R.; Bhown, A. Achieving a 10,000 GPU permeance for post-combustion carbon capture with gelled ionic liquid-based membranes. In 2010 NETL CO 2 Capture Technology Meeting, Pittsburgh, 2010. (22) Voss, B. A.; Bara, J. E.; Gin, D. L.; Noble, R. D. Physically gelled ionic liquids: Solid membrane materials with liquidlike CO2 gas transport. Chem. Mater. 2009, 21, 3027.
FEATURE
(23) Cost and Performance for Low-Rank Pulverized Coal Oxycombustion Energy Plants: National Energy Technology Laboratory, 2010; p 442, http://www.netl.doe.gov/energy-analyses/pubs/LRPC_Oxycmbst_ 093010.pdf (accessed September 20, 2011). (24) Current and Future Technologies for Gasification-Based Power Generation, Vol. 2: A Pathway Study Focused on Carbon Capture Advanced Power Systems R&D Using Bituminous Coal, Revision 1; National Energy Technology Laboratory, 2010; p 81, http://www.netl.doe.gov/ energy-analyses/pubs/AdvancedPowerSystemsPathwayVol2.pdf (accessed September 20, 2011). (25) Cost and Performance Baseline for Fossil Energy Plants Vol. 1: Bituminous Coal and Natural Gas to Electricity, Revision 2, November 2010; National Energy Technology Laboratory, 2010; p 626, http:// www.netl.doe.gov/energy-analyses/pubs/BitBase_FinRep_Rev2.pdf (accessed September 20, 2011). (26) Chen, C.; Rubin, E. S. CO2 control technology effects on IGCC plant performance and cost. Energy Pol 2009, 37 (3), 915 924; DOI: 10.1016/j.enpol.2008.09.093. (27) Rubin, E. S.; Chen, C.; Rao, A. B. Cost and performance of fossil fuel power plants with CO2 capture and storage. Energy Policy 2007, 35, 4444 4454; DOI: 10.1016/j.enpol.2007.03.009. (28) Firoozabadi, A.; Cheng, P. Prospects for Subsurface CO2 Sequestration. AIChE J. 2010, 56 (6), 1398 1405; DOI: 10.1002/aic.12287. (29) Han, W. S.; McPherson, B. J.; Lichtner, P. C.; Wang, F. P. Evaluation of trapping mechanisms in geologic CO2 sequestration: Case study of SACROC northern platform, a 35-year CO2 injection site. Am. J. Sci. 2010, 310 (4), 282 324; DOI: 10.2475/04.2010.03. (30) Pipeline and Hazardous Materials Safety Administration’s (PHSMA) National Pipeline Mapping System; U.S. Department of Transportation, November 18, 2009. (31) Stauffer, P. H.; Viswanathan, H. S.; Pawar, R. J.; Guthrie, G. D. A system model for geologic sequestration of carbon dioxide. Environ. Sci. Technol. 2009, 43 (3), 565 570; DOI: 10.1021/es800403w. (32) Michael, K.; Golab, A.; Shulakova, V.; Ennis-King, J.; Allinson, G.; Sharma, S.; Aiken, T. Geological storage of CO2 in saline aquifers— A review of the experience from existing storage operations. Int. J. Greenhouse Gas Control 2010, 4 (4), 659 667; DOI: 10.1016/j.ijggc. 2009.12.011. (33) Zhou, Q. L.; Birkholzer, J. T.; Mehnert, E.; Lin, Y. F.; Zhang, K. Modeling basin- and plume-scale processes of CO2 storage for full-scale deployment. Ground Water 2010, 48 (4), 494 514; DOI: 10.1111/ j.1745-6584.2009.00657.x. (34) Bruant, R. G.; Guswa, A. J.; Celia, M. A.; Peters, C. A. Safe storage of CO2 in deep saline aquifers. Environ. Sci. Technol. 2002, 36 (11), 240A–245A. (35) Newell, D. L.; Kaszuba, J. P.; Viswanathan, H. S.; Pawar, R. J.; Carpenter, T. Significance of carbonate buffers in natural waters reacting with supercritical CO2: implications for monitoring, measuring and verification (MMV) of geologic carbon sequestration. Geophys. Res. Lett. 2008, 35, L23403 (5 pp.); DOI: 10.1029/2008gl035615. (36) Rapaka, S.; Pawar, R. J.; Stauffer, P. H.; Zhang, D. X.; Chen, S. Y. Onset of convection over a transient base-state in anisotropic and layered porous media. J. Fluid Mech. 2009, 641, 227 244; DOI: 10.1017/s0022112009991479. (37) Gessner, K.; Kuhn, M.; Rath, V.; Kosack, C.; Blumenthal, M.; Clauser, C. Coupled process models as a tool for analysing hydrothermal systems. Surv. Geophys. 2009, 30 (3), 133 162; DOI: 10.1007/s10712009-9067-1. (38) Gilfillan, S. M. V.; Lollar, B. S.; Holland, G.; Blagburn, D.; Stevens, S.; Schoell, M.; Cassidy, M.; Ding, Z. J.; Zhou, Z.; LacrampeCouloume, G.; Ballentine, C. J. Solubility trapping in formation water as dominant CO2 sink in natural gas fields. Nature 2009, 458 (7238), 614 618; DOI: 10.1038/nature07852. (39) Little, M. G.; Jackson, R. B. Potential impacts of leakage from deep CO2 geosequestration on overlying freshwater aquifers. Environ. Sci. Technol. 2010, 44 (23), 9225 9232; DOI: 10.1021/es102235w. (40) Kharaka, Y. K.; Thordsen, J. J.; Kakouros, E.; Ambats, G.; Herkelrath, W. N.; Beers, S. R.; Birkholzer, J. T.; Apps, J. A.; Spycher, N. F.; 8603
dx.doi.org/10.1021/es200510f |Environ. Sci. Technol. 2011, 45, 8597–8604
Environmental Science & Technology Zheng, L. E.; Trautz, R. C.; Rauch, H. W.; Gullickson, K. S. Changes in the chemistry of shallow groundwater related to the 2008 injection of CO2 at the ZERT field site, Bozeman, Montana. Environ. Earth Sci. 2010, 60 (2), 273 284; DOI: 10.1007/s12665-009-0401-1. (41) Viswanathan, H. S.; Pawar, R. J.; Stauffer, P. H.; Kaszuba, J. P.; Carey, J. W.; Olsen, S. C.; Keating, G. N.; Kavetski, D.; Guthrie, G. D. Development of a hybrid process and system model for the assessment of wellbore leakage at a geologic CO2 sequestration site. Environ. Sci. Technol. 2008, 42 (19), 7280 7286; DOI: 10.1021/es800417x. (42) Oldenburg, C. M.; Bryant, S. L.; Nicot, J. P. Certification framework based on effective trapping for geologic carbon sequestration. International Journal of Greenhouse Gas Control 2009, 3 (4), 444-457; DOI:10.1016/j.ijggc.2009.02.009. (43) Bachu, S. CO2 storage in geological media: Role, means, status and barriers to deployment. Prog. Energy Combust. Sci. 2008, 34 (2), 254 273; DOI: 10.1016/j.pecs.2007.10.001. (44) CO2 Capture Project. http://www.co2captureproject.org/ (accessed September 20, 2011). (45) Giardini, D. Geothermal quake risks must be faced. Nature 2009, 462 (7275), 848 849; DOI: 10.1038/462848a. (46) Hepple, R. P.; Benson, S. M. Geologic storage of carbon dioxide as a climate change mitigation strategy: performance requirements and the implications of surface seepage. Environ. Geol. 2005, 47 (4), 576 585; DOI: 10.1007/s00254-004-1181-2. (47) Buscheck, T. A.; Sun, Y.; Hao, Y. Combining brine extraction, desalination, and residual brine reinjection with CO2 storage in saline formations: Implications for pressure management, capacity, and risk mitigation. Energy Procedia 2011, 4, 4283–4290. (48) Seto, C. J.; McRae, G. J. Reducing risk in basin scale CO2 sequestration: A Framework for integrated monitoring design. Environ. Sci. Technol. 2011, 45 (3), 845 859; DOI: 10.1021/es102240w. (49) Espinoza, D. N.; Kim, S. H.; Santamarina, J. C. CO2 Geological Storage - Geotechnical Implications. Ksce Journal of Civil Engineering 2011, 15 (4), 707 719; DOI: 10.1007/s12205-011-0011-9. (50) Keeling, R. F.; Manning, A. C.; Dubey, M. K. The atmospheric signature of carbon capture and storage. Philosophical Transactions of the Royal Society a-Mathematical Physical and Engineering Sciences 2011, 369 (1943), 2113 2132; DOI: 10.1098/rsta.2011.0016. (51) The Rock Springs Uplift: An outstanding geological CO2 sequestration site in southwest Wyoming; Wyoming State Geological Survey; 2007; 31 p.; http://sales.wsgs.uwyo.edu/catalog/product_info.php? products_id=23. (52) An Integrated Strategy for Carbon Management Combining Geological Co2 Sequestration, Displaced Fluid Production, And Water Treatment, WSGS-2009-CGRD-08; Wyoming Geological Survey, 2009. (53) Keating, G. N.; Middleton, R. S.; Stauffer, P. H.; Viswanathan, H. S.; Letellier, B. C.; Pasqualini, D.; Pawar, R. J.; Wolfsberg, A. V. Mesoscale carbon sequestration site screening and CCS infrastructure analysis. Environ. Sci. Technol. 2011, 45 (1), 215 222; DOI: 10.1021/ es101470m. (54) Assess Carbon Dioxide Capture Options for Illinois Basin Carbon Dioxide Sources; Illinois Geological Survey: Champaign, IL, 2005; p 46, http://sequestration.org/publish/phase1_seq_optimization_topical_ rpt.pdf (accessed September 20, 2011). (55) Dooley, J. J.; Dahowski, R. T.; Davidson, C. L. The potential for increased atmospheric CO2 emissions and accelerated consumption of deep geologic CO2 storage resources resulting from large-scale deployment of a CCS-enabled unconventional fossil fuels industry in the U.S. In. J. Greenhouse Gas Control 2009, 3, 720 730; DOI: 10.1016/j.ijggc.2009.08.004. (56) MIT CO2 pipeline transport and cost model. http://e40-hjhserver1.mit.edu/energylab/wikka.php?wakka=MIT (accessed September 20, 2011). (57) Middleton, R. S.; Bielicki, J. M. A scalable infrastructure model for carbon capture and storage: SimCCS. Energy Policy 2009, 37, 1052 1060; DOI: 10.1016/j.enpol.2008.09.049 (accessed September 20, 2011). (58) The Evolution of the Extent and the Investment Requirements of a Trans-European CO2 Transport Network; European Commission Joint
FEATURE
Research Centre, 2010; http://publications.jrc.ec.europa.eu/repository/ handle/111111111/15100. (accessed September 20, 2011). (59) Kuby, M.; Bielicki, J. M.; Middleton, R. S. Optimal spatial deployment of carbon dioxide capture and storage given a price on carbon dioxide. Int. Reg. Sci. Rev. 2011, 34 (3), 285 305; DOI: 10.1177/ 0160017610397191. (60) CO2 Highways for Europe: Modelling a Carbon Capture, Transport and Storage Infrastructure for Europe; Centre for European Policy Studies, 2010; p 23, http://www.ceps.eu (accessed September 20, 2011). (61) van den Broek, M.; Brederode, E.; Ramirez, A.; Kramers, L.; van der Kuip, M.; Wildenborg, T.; Turkenburg, W.; Faaij, A. Designing a cost-effective CO2 storage infrastructure using a GIS based linear optimization energy model. Environ. Modell. Software 2010, 25 (12), 1754 1768; DOI: 10.1016/j.envsoft.2010.06.015. (62) Schindler, D. Tar sands need solid science. Nature 2010, 468 (7323), 499–501. (63) Carbon Dioxide Capture and Storage: A Canadian Clean Energy Opportunity, 2009; p 20, http://www.ico2n.com/wp-content/ uploads/2010/07/ICO2N-Report_09_final2.pdf (accessed). (64) Eccles, J. K.; Pratson, L.; Newell, R. G.; Jackson, R. B. Physical and economic potential of geological CO2 storage in saline aquifers. Environ. Sci. Technol. 2009, 43 (6), 1962 1969; DOI: 10.1021/es801572e. (65) Esposito, R. A.; Monroe, L. S.; Friedman, J. S. Deployment models for commercialized carbon capture and storage. Environ. Sci. Technol. 2011, 45, 139 146; DOI: 10.1021/es101441a. (66) Operations: CO2 pipelines. http://www.denbury.com/index. php?id=18 (accessed September 20, 2011).
8604
dx.doi.org/10.1021/es200510f |Environ. Sci. Technol. 2011, 45, 8597–8604
VIEWPOINT pubs.acs.org/est
Chromium Contamination Accident in China: Viewing Environment Policy of China Yang Gao* and Jun Xia Key Laboratory of Water Cycle and Related Land Surface Processes, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China
C
hromium (Cr) is the second most abundant inorganic groundwater contaminant at hazardous waste sites. Due to the difference in the toxic nature of both the forms of Cr, the discharge of Cr (VI) into surface water is regulated to below 0.05 mg L 1 by the U.S. Environmental Protection Agency, whereas the total Cr (including Cr(III), Cr(VI) and its other forms) is regulated to below 2 mg L 1.1 Since Cr(VI) is known to be very mobile and hazardous to human health through inhalation, skin contact, and ingestion, being highly toxic, carcinogenic and mutagenic to living organisms, even when present in very low concentrations in water. Cr compounds released into the environment have been increasing continuously as a result of industrial activities and technological developments, posing a significant threat to the environment and public health because of their toxicity, accumulation in the food chain and persistence in nature.2,3 Cr(VI) contaminates through various industrial processes such as electroplating, metal finishing, leather, mining, petroleum refining, wood preservation, corrosion inhibition in power plants and nuclear facilities, manufacturing of pigments, dyes, textiles, carpets, magnetic tapes, jet aircrafts, refractory materials, etc. In spite of the fact that the industry is constantly looking for efficient and innovative routes for chromium recovery or detoxification, large amounts of such soil and wastewater r 2011 American Chemical Society
are still released to the environment, throughout the world, every year. China is the largest amount of Cr slag produced country, and there are about 25 enterprises producing Cr salts, of which the annual production capacities reach 329,000 tons and the annual discharges of Cr slag reach 450 000 tons. The storage of untreated Cr slag in past years has exceeded 400 million tons.4 In China, there are strict environmental regulations on Cr wastes treatment, which require all of Cr wastes must be through harmless treatment before released to environment and every disposal process on Cr wastes must be registered and can be traced. However, on August 13th, 2011, serious Cr slag contamination accident occurred in Qujing in Yunan province. The reason on this accident was due to the amount of 5000 tons of untreated Cr slag, produced by Lvliang Chemical Industry Co., Ltd., was illegally dumped into mountains and Nanpan River located at the source of Zhujian River (Figure 1). The contamination accident lead to the over standard of Cr (VI) concentration detected in Nanpan River was about 2000 times. As government involved in the investigation, they found that there were about 288 400 tons Cr slag piled at the back door of the chemical industry without any other treatment, which there was only one meter distance to Nanpan River. From 1989 to 2005, the storage of Cr slag in the chemical industry has increased to 288400 tons, whereas this industry took no any other detoxification treatment in order to saving treatment cost. Therefore, two drivers were employed by the chemical industry to dump the Cr slag into nearby mountains and reservoirs. With Cr pollution enlarges and media reports, local environment protection department and police arrested the two drivers and close the chemical industry rapidly. However, the effects of Cr slag on ecology, environment and health will last for many years. The Xinrong village is the nearest village to the chemical industry, which is also called “dead village”. According to the villagers’ description, there were at least six to seven people living in this village who died of cancer. Meanwhile, in this village, a total of 77 cows and sheep suddenly died due to drinking Cr polluted water. Although the truths on Cr slag contamination were gradually disclosed, it also leads to a crisis of public confidence. According to the investigation results on the accident, we found that on June 12, 2011, the environmental protection department in Qujing has received an accusation on Cr slag pollution from this chemical industry, whereas the local government opened the processing results on this accident until August 13, 2011. We have enough Received: August 24, 2011 Published: September 19, 2011 8605
dx.doi.org/10.1021/es203101f | Environ. Sci. Technol. 2011, 45, 8605–8606
Environmental Science & Technology
VIEWPOINT
Dr. Chiyuan Miao in Beijing Normal University to give us references and enlightenment. Opinions in the paper only reflect the personal views of the authors. This research was supported by Knowledge Innovation Program of the Chinese Academy of Sciences (KZCX2-YW-QN408), the National Natural Science Foundation of China (40901098) and the National Basic Research Program of China (B000800901).
’ AUTHOR INFORMATION Corresponding Author
*Phone/fax: +86-10-64889011; e-mail:
[email protected];
[email protected]. Figure 1. The river polluted by Cr slag produced by Lvliang Chemical Industry Co., Ltd. and the water color has become yellow.
reason to believe that if there was no media report, the local government would never open the truth to people. The event also reflect that the ecology, environment, and human health are insignificant when compared to economic development in eye's local governments in China, and local officials do not hesitate to suppress the truth in order to hold their official position. In recent years, Chinese people have gradually become accustomed to the significant contamination accidents and resulting deaths. It is very very worrisome that people and government sacrifice environment, life, and health in order to obtain economic development. When public awareness is ignored by government, environmental protection in China is only a mirage. Why does the local government conceal the truth without considering environment and public safety? Currently, government officials are rewarded for actions that increase economic growth, and not only is there no reward for environmental protection, there are indirect penalties when efforts to protect the environment slow economic growth. Instead, officials must be rewarded for efforts to harmonize socioeconomic development with protection of the environment; in some of the most severely endangered areas of the country, it may even be necessary to shift the balance wholly toward rewarding environmental protection. This will create a favorable sociopolitical environment in which officials feel comfortable making and enforcing policies and laws that are oriented toward sustainable development. China’s progress toward environmental sustainability remains difficult, with negative factors currently outweighing positive ones. Given its increasing global economic and environmental importance, China must clearly accept a more active leadership role in steering the world into a sustainable future. The current situation is such that China’s overall environmental conditions will continue to grow worse before they improve. China’s efforts have the potential to enrich the world and show other nations how to achieve sustainability through its institutional, scientific, and technological innovations. If these plans are carefully implemented, we have every reason to look forward to more sustainable prospects for China.
’ REFERENCES (1) Baral, A.; Engelken, R. D. Chromium-based regulations and greening in metal finishing industries in the USA. Environ. Sci. Policy 2002, 5, 121–133. (2) Gavrilescu, M. Removal of heavy metals from the environment by biosorption a review. Eng. Life Sci. 2004, 4, 219–232. (3) Gao, Y.; Miao, C. Y.; Mao, L.; Zhou, P.; Jin, Z. G.; Shi, W. J. Improvement of phytoextraction and antioxidative defense in Solanum nigrum L. under cadmium stress by application of cadmium-resistant strain and citric acid. J. Hazard. Mater. 2010, 181, 771–777. (4) Chromium Slays Pollution Control and Environmental Protection Technical Specifications; State Ministry of Environmental Protection (MEP), 2007.
’ ACKNOWLEDGMENT We are grateful to Dr. Yafeng Wang in State Key Laboratory of Urban, Dr. Shixiong Cao in Beijing Forestry University and Regional Ecology, Research Center for Eco-Environmental Sciences and 8606
dx.doi.org/10.1021/es203101f |Environ. Sci. Technol. 2011, 45, 8605–8606
VIEWPOINT pubs.acs.org/est
China’s Cities Need to Grow in a More Compact Way Jingzhu Zhao,*,† Yu Song,† Lina Tang,† Longyu Shi,† and Guofan Shao†,‡ †
Key Lab of Urban Environment and Health, Institute of Urban Environment, Chinese Academy of Sciences, 1799 Jimei Road, Xiamen 361021, China ‡ Department of Forestry and Natural Resources, Purdue University, 715 West State Street, West Lafayette, Indiana 47907, United States
S
ince the 1990s, China has undergone astonishing economic growth, and its extensive urbanization has become one of the most striking development signatures anywhere in the world. With the world’s largest urban population, China’s urbanization experience provides meaningful insights on future urban development in other countries.1 Because China has an extremely low per-capita land area, constructing compact urban forms is regarded as one of the most important considerations for China’s sustainable city development.2 This is because, in general, a compact city can help better utilize public transport services, strengthen urban community multifunctionality, increase the quality of city life, reduce energy consumption, and improve the urban environment.3 The question is, however, are China’s cities developing toward a more compact form? We assessed the compactness of 35 major cities in China in 2000 and 2007 by using a normalized compactness index (NCI) (0 < NCI < 1), which is defined as a ratio of Thinh et al.’s compactness (T) for a given city over the maximum compactness (Tmax) for a hypothetical round-shape city.4 NCI is independent of the dimension of a target city, and the greater its value, the more compact a city. The 35 major cities in this paper include 26 provincial or autonomous-region capital cities, four municipalities, and five cities specifically designated in the state plan (or state-plan cities) (Figure 1), the same cities as studied by Zhao.2 The geographic extents of the 35 cities ranged from 34 to r 2011 American Chemical Society
550 km2 in 2000 and 65 to 1289 km2 in 2007. The populations of urban-district residents in the 35 cities ranged from 0.57 to 11.36 million in 2000, and from 0.75 to 17.69 million in 2007. In 2000, NCI values ranged from 0.048 (Guangzhou) to 0.257 (Jinan) (Figure 1). The mean of NCI values for the 35 cities was 0.131 ( 0.060 (standard deviation). The skewed distribution of NCI values in 2000 showed that the majority of cities in China were relatively noncompact. Among the 12 cities with the lowest NCI values, 8 were low-population cities. Beijing’s NCI value was 0.112 and Shanghai’s was 0.130, and their NCI values ranked 26th and 18th, respectively. In 2007, NCI values ranged from 0.052 (Nanjing) to 0.335 (Shijiazhuang) (Figure 1). The mean of NCI values for the 35 cities was 0.145 ( 0.066 (standard deviation). The skewed distribution of NCI values in 2007 showed that the majority of China’s cities were still relatively noncompact. Low- and high-population cities were almost evenly distributed within each NCI class, with the exception of the largest NCI class. Beijing’s NCI value increased to 0.131 whereas Shanghai’s remained unchanged, and their NCI values ranked 20th and 21st, respectively. Between 2000 and 2007, the change in NCI values varied tremendously among the 35 cities (Figure 1). Although there was a slight increase in the mean of NCI values, 12 cities experienced negative changes in NCI values. This was particularly the case for larger cities. Among the 12 cities with the lowest NCI values in 2000, 11 cities experienced a positive change in NCI value, including 8 smaller cities (Figure 1). Four out of the five stateplan cities saw their NCI values increase. The state-plan cities are the pioneers of China’s economic reform and recent urban planning in those cities seems to have helped them develop toward greater compactness. For example, Xiamen is a rising state-plan city5 and was ranked the most sustainable city in China.2 The less-populated cities are likely to have more open areas within the urban landscape, where land is suitable for development. Filling up the urban gaps automatically increases city compactness. As population continues to grow, urban development has to take place beyond the existing city limit and urban sprawl may jeopardize the city’s compactness. Beijing, Shanghai, and Chongqing, three mega cities and municipalities, are good examples of “unlimited” urban sprawl without enhanced compactness. This fact is critically important because many local cities tend to follow the urban-growth models of Beijing and Shanghai in their urbanization processes. It is worth noting that China has experienced an increase in per-capita urban area in the past decade, indicating Published: September 20, 2011 8607
dx.doi.org/10.1021/es203138c | Environ. Sci. Technol. 2011, 45, 8607–8608
Environmental Science & Technology
VIEWPOINT
Figure 1. Normalized compactness index (NCI) values for 35 major cities and the number of cities (N) within each of five equal-interval classes in 2000 and 2007. Red dots represent provincial capital cities, blue dots represent mean municipalities, and green dots are the state-plan cities. Smaller dots are cities with less than two million people (median) in 2000.
that land use has become less efficient over time and that there is a need for compact-city development in China. Expansion in urban living space and increase in city compactness both can help improve life quality in China. However, an acceleration in the compactness of China’s cities is needed in the future to ensure that land is used economically and sustainably.
(5) Zhao, J.; Dai, D. B.; Lin, T.; Tang, L. N. Rapid urbanization, ecological effects and sustainable city construction in Xiamen. Int. J. Sustainable Dev. World Ecol. 2010, 17 (4), 271–272.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +86-592-6190999; e-mail:
[email protected].
’ ACKNOWLEDGMENT This research was supported by the Chinese Academy of Sciences (KZCX-2-YW-453; 08I4021D10). ’ REFERENCES (1) Normile, D. China’s living laboratory in urbanization. Science 2008, 319 (5864), 740–743. (2) Zhao, J. Towards Sustainable Cities in China: Analysis and Assessment of Some Chinese Cities in 2008; Springer: New York, 2011. (3) Dempsey, N. Revisiting the compact city. Built Environ. 2010, 36 (1), 5–8. (4) Thinh, N. X.; Arlt, G.; Heber, B.; Hennersdorf, J.; Lehmann, I. Evaluation of urban land-use structures with a view to sustainable development. Environ. Impact Assess. Rev. 2002, 22 (5), 475–492. 8608
dx.doi.org/10.1021/es203138c |Environ. Sci. Technol. 2011, 45, 8607–8608
VIEWPOINT pubs.acs.org/est
Efficient Production of Solar Fuel Using Existing Large Scale Production Technologies Wim Haije*,†,‡ and Hans Geerlings†,§ †
Department of Chemical Engineering, Delft University of Technology, Julianalaan 136, 2628 BL Delft, The Netherlands Energy research Centre of The Netherlands, PO BOX 1, 1755 ZG, Petten, The Netherlands § Shell Global Solutions International B.V., Grasweg 31, 1031 HW Amsterdam, The Netherlands ‡
through hydrogenation of CO2. The CO2 is captured at fossil fuel based power plants or directly from the air. To produce liquid hydrocarbon fuels from solar hydrogen and CO2, the reverse water-gas shift reaction (RWGS): H2 þ CO2 T CO þ H2 O ΔH0 ¼ þ 41kJ=mol
’ INTRODUCTION Global energy consumption will at least double during this century. Today about 80% of the primary energy used is provided by fossil fuels. To avoid anthropogenic climate change, it is essential that the world’s energy system is gradually transformed from fossil energy based to a solar based one. In such a solar world, efficient energy storage is prerequisite to deal with the intermittency of this abundant source. Storage in the form of chemical energy, that is, solar fuels, provides a natural route to deal with this problem. An obvious solar fuel would be hydrogen, which can be produced efficiently from solar energy through direct splitting of water or through combination of solar electricity production with electrolysis of water.1 Unfortunately, hydrogen has a major disadvantage as its transport and storage is not trivial. No such storage problem exists for alcohols or liquid hydrocarbons. Especially liquid hydrocarbons are preferred solar fuels in view of their unprecedented combination of gravimetric and volumetric energy density and the worldwide existing markets and infrastructure.
combined with commercial cobalt based FischerTropsch technology is the obvious route. Figure 1 gives an artist impression of this solar fuels scheme. With the exception of the RWGS process the whole process train is already available on a (sub)commercial scale today. The RWGS is an endothermic equilibrium reaction, and the equilibrium lies on the product side only for temperatures above 800 °C. The conversion at lower temperatures, 150250 °C can be enhanced greatly, up to 100%, if a reaction product, that is, water, is selectively removed. At laboratory scale Carvill et al.2 showed that even pure CO could be produced at 250 °C in a sorbent enhanced RWGS (SERWGS) process. In the current solar fuels process it suffices if the SERWGS step produces a mixture of CO and hydrogen (synthesis gas). The inset in figure 1 shows a typical experimental result. In this case CO2 and hydrogen are fed to a fixed bed reactor, which contains a mixture of a low temperature WGS catalyst (S€udChemie) and a water sorbent (for our experiments molsieve 4A (Merck)). Initially essentially complete conversion of CO2 is obtained. After saturation of the sorbent, breakthrough of CO2 occurs and the conversion converges toward chemical equilibrium. In practice the SERWGS process is also executed in a fixed bed reactor, which contains a mixture of a watergas shift catalyst and a sorbent. Saturation of the sorbent with water necessitates a number of parallel reactors: one reactor is producing synthesis gas, while the others are in various stages of regeneration. This configuration is similar to that of the sorbent enhanced water-gas shift process, SEWGS, currently under development.3 No major hurdles are therefore foreseen for further development of the SERWGS process. The heat released by the sorption reaction is sufficient to drive the RWGS reaction. The heat released by the exothermic FischerTropsch process (160 kJ/mol) can subsequently be used to fully regenerate the water sorbent (55 kJ/mol), optimizing the sustainability of the process. To obtain an estimate of the efficiency from solar energy to liquid fuels one can multiply the efficiency of the various subprocesses. State-of-the-art photovoltaic panels achieve efficiencies of
’ SOLUTION Though hydrogen is not the ideal solar fuel, it is a very attractive intermediate as liquid hydrocarbons can be produced
Received: September 9, 2011 Accepted: September 12, 2011 Published: September 21, 2011
r 2011 American Chemical Society
8609
dx.doi.org/10.1021/es203160k | Environ. Sci. Technol. 2011, 45, 8609–8610
Environmental Science & Technology
VIEWPOINT
Figure 1. An artist impression of a solar fuels process. Liquid hydrocarbon fuels are produced from solar hydrogen and CO2. Inset shows experimental conversion of CO2 as a function of time in the SERWGS reaction (p = 1 bar, T = 300 °C, H2/CO2 feed ratio = 4, GHSV= 170 h1, equilibrium conversion: horizontal dashed line).
19%. The efficiency of high pressure electrolysis is about 85%, which implies that solar hydrogen can be produced with efficiencies around 16%. Including a mismatch between photovoltage and electrolysis potential this number reduces to 13%.1 The efficiency of the SERWGS and FischerTropsch steps are estimated to be of the order of 90% and 70% respectively. This yields an overall efficiency from solar energy to liquid fuels of about 8.2%. This number is not very sensitive to inclusion of the energy penalty connected with CO2 capture. A realistic energy penalty of 150 kJ/mol of CO2 reduces the efficiency of solar fuel synthesis to 7.7%. This number compares favorably with efficiencies reported for a number of alternative routes to produce solar fuels including advanced biomass conversion, artificial photosynthesis and thermochemical production.4 Many of these technologies, though promising in the long term, are still in a rather early stage of their development and the efficiencies from solar energy to chemical energy are still low, typically below 1%.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected] .
’ REFERENCES (1) Blankenship, R. E.; et al. Science 2011, 332, 805–809. (2) Carvill, B. T; Hufton, J. R.; Anand, M.; Sircar, S. Sorptionenhanced reaction process. AIChE J. 1996, 42, 2765. (3) E. R. Van Selow et al., Pilot-scale development of the sorption enhanced water gas shift process. In Carbon dioxide Capture for Storage in Deep Geologic Formations; E. Li, Ed.; Berks CPL Press, 2009, p 157. (4) Somnath, R. C.; Varghese, O. K.; Paulose, M.; Grimes; Craig, A. Toward solar fuels: Photocatalytic conversion of carbon dioxide to hydrocarbons. ACS Nano 2010, 4 (3), 1259–1278.
’ SUMMARY Solar fuels can be produced efficiently already today using existing technologies. Liquid hydrocarbon fuels production from CO2 and water with efficiencies approaching 10%from solar energy to fuelis feasible. This is about an order of magnitude higher than alternative technologies currently under development. The main challenge for further research is to reduce costs and further improve efficiency through optimized process schemes. The main challenge for further research is to reduce costs and further improve efficiency through optimized process schemes and PV systems. 8610
dx.doi.org/10.1021/es203160k |Environ. Sci. Technol. 2011, 45, 8609–8610
VIEWPOINT pubs.acs.org/est
Management Tools for Managing Vapor Intrusion Zachary A. Collier,† John T. Vogel,† Stephen G. Zemba,‡ Elizabeth A. Ferguson,† and Igor Linkov*,†,§ †
Environmental Laboratory, Engineer Research and Development Center, U.S. Army Corps of Engineers, 3909 Halls Ferry Road, Vicksburg, Mississippi 39180, United States ‡ Cambridge Environmental, Inc., 58 Charles Street, Cambridge, Massachusetts 02141, United States
S
triking the correct balance between collecting information to support remedial and mitigation strategy versus selecting an appropriate course of action is an important problem for all contaminated sites. It is especially important on sites affected by vapor intrusion (VI), which describes the migration of volatile chemicals from soil and groundwater into the indoor air of buildings, impacting human health and leading to economic losses. One quarter of the contaminated sites within the United States contain the contaminants and conditions favorable for VI, equating to over 100 000 sites.1 The United States Navy reported that VI costs associated with just 19 contaminated sites totaled over $6.9 million.2 Formerly closed sites are being reexamined and remedial investigations are being prolonged indefinitely as our understanding of vapor intrusion evolves. Stakeholders are forced to come to grips with the difficulties of resolving uncertainties and fitting vapor intrusion into regulatory programs, but no universally accepted guidance or programmatic approach has evolved for VI management. U.S. EPA’s 2002 guidance3 remains in draft form and raises more questions than it answers, and has prompted states to develop varying approaches and industry groups to seek uniformity. Most of the current schemes calculate absolute levels of risk (e.g., cancer risk) associated with each chemical and then assess risk reduction through proposed remedial or abatement strategies. However, numerous r 2011 American Chemical Society
and significant uncertainties are involved. Simply identifying VI chemicals in indoor air is challenging because levels of concern are often near detection limits and the same chemicals are frequently present from other sources. Lacking definitive direct evidence, remedial and mitigation decisions for VI are often based on inferred data and risk estimates using highly variable, site-specific concentration measurements compared to conservative benchmarks derived through simplified models. The natural tendency toward risk aversion results in extensive data collection and overly conservative evaluations, driving remedial costs up and impeding clear-cut, long-term solutions at contaminated sites. Formal multicriteria decision analysis tools (MCDA)4 could be used to guide sampling and provide transparency in remedial decision making (Figure 1). MCDA objectively handles information and analyzes it through structured criteria and metrics, allowing integration of technical data and expert judgment. In the case of VI, multiple lines of evidence of contamination and resulting risks can be available, but not all of them are of equal importance. For instance, groundwater measurements may imply overly conservative results, and indoor air measurements can be skewed by background sources from ambient air and indoor chemicals.5 Furthermore, the location and method of sampling can have a direct impact on data quality (e.g., subslab sampling is more reliable than near-slab sampling due to spatial variability at sites). MCDA allows the decision maker to discount lines of evidence which are less reliable or conclusive, effectively capturing the uncertainty stemming from various sampling locations, methods, and technologies. In addition to its ability to integrate heterogeneous information, MCDA can also assist in evaluating value trade-offs in VI sampling and management decisions. Intrusive alternatives such as excavation of contaminated soil and treatment of groundwater may be effective, but are costly and burdensome to stakeholders who may prefer mitigation strategies such as the installation of vapor barriers or subslab depressurization systems. Short-term and long-term alternatives can be compared as well, such as the preemptive installation of a subslab depressurization system whether or not a VI threat is confirmed. MCDA allows potential management alternatives with dissimilar criteria arising from competing objectives (e.g., minimize cost, minimize environmental impacts, maximize human health) to be compared without the need to reduce all aspects to a single unit (e.g., monetize all criteria). A decision model is ideal for implementing an adaptive VI management strategy that can integrate data collection, examine Received: September 10, 2011 Accepted: September 13, 2011 Published: September 28, 2011 8611
dx.doi.org/10.1021/es203179w | Environ. Sci. Technol. 2011, 45, 8611–8612
Environmental Science & Technology
VIEWPOINT
Figure 1. Conceptual model of MCDA applied to vapor intrusion decisions.
multiple lines of evidence, evaluate uncertainties, and review and revise action over time. For example, by comparing monitoring data against expected results over time, mitigation measures can be adjusted accordingly. By combining enhanced decision-analytical techniques with adaptive management, one can strike an optimal balance between the reduction of uncertainty through additional data collection and the costs of doing so by assigning utility scores to monitoring information. VI is a complex environmental management issue, with potential environmental, human health, economic, and legal consequences. We advocate a shift from attempts to make risk-based remedial decisions by assessing absolute risk associated with VI to comparative risk-based assessments of alternative VI management strategies. A decision-analytical framework for VI response would allow stakeholders to make more informed management decisions and to allocate limited financial resources in a more efficient manner.
(2) McAlary, T.; Ettinger, R.; Johnson, P.; Eklund, B.; Hayes, H.; Chadwick, D. B.; Rivera-Duarte, I. Review of Best Practices, Knowledge and Data Gaps, And Research Opportunities for the U.S. Department of Navy Vapor Intrusion Focus Areas, Technical Report 1982; SSC Pacific: San Diego, CA, 2009. (3) OSWER Draft Guidance for Evaluating the Vapor Intrusion to Indoor Air Pathway from Groundwater and Soils (Subsurface Vapor Intrusion Guidance), EPA530-D-02-004; U.S. Environmental Protection Agency Office of Solid Waste and Emergency Response: Washington, DC, 2002. (4) Linkov, I.; Moberg, E. Multi-Criteria Decision Analysis: Environmental Applications and Case Studies; CRC Press: Boca Raton, FL, 2011. (5) Vapor Intrusion Pathway: A Practical Guidance; Interstate Technology Regulatory Council: Washington, DC, 2007; http://www.itrcweb.org/Documents/VI-1.pdf (accessed June 29, 2011).
’ AUTHOR INFORMATION Corresponding Author
*Phone: 617 233 9869; e-mail:
[email protected]. Present Addresses §
696 Virginia Road, Concord, Massachusetts, 01742, United States.
’ ACKNOWLEDGMENT Permission was granted by the Chief of Engineers to publish this information. The views and opinions expressed in this article are those of the individual authors and not those of the U.S. Army, or other sponsors. ’ REFERENCES (1) Colbert, K. L.; Palazzo, J. E. Vapor intrusion: Liability determination protects profits and minimizes risk. Real Estate Finance 2008, 24, 17–22. 8612
dx.doi.org/10.1021/es203179w |Environ. Sci. Technol. 2011, 45, 8611–8612
CRITICAL REVIEW pubs.acs.org/est
Hexabromocyclododecane: Current Understanding of Chemistry, Environmental Fate and Toxicology and Implications for Global Management Christopher H. Marvin,†,* Gregg T. Tomy,‡,§,* James M. Armitage,||,^ Jon A. Arnot,|| Lynn McCarty,# Adrian Covaci,3 and Vince Palace‡,§ †
Water Science and Technology Directorate, Environment Canada, Burlington, Ontario L7R 4A6 Canada Fisheries & Oceans Canada, Winnipeg, Manitoba R3T 2N6 Canada § Faculty of Environment and Geography, University of Manitoba, Winnipeg, Manitoba, R3T 2N2 Canada University of Toronto Scarborough, Ontario, M1C 1A4, Canada ^ Department of Occupational Medicine, Aarhus University Hospital, Aarhus, Denmark # LS McCarty Scientific Research and Consulting, Ontario, Canada 3 Toxicological Centre, University of Antwerp, 2610 Wilrijk, Belgium
)
‡
bS Supporting Information ABSTRACT: Hexabromocyclododecane (HBCD) is a globally produced brominated flame retardant (BFR) used primarily as an additive FR in polystyrene and textile products and has been the subject of intensified research, monitoring and regulatory interest over the past decade. HBCD is currently being evaluated under the Stockholm Convention on Persistent Organic Pollutants. HBCD is hydrophobic (i.e., has low water solubility) and thus partitions to organic phases in the aquatic environment (e.g., lipids, suspended solids). It is ubiquitous in the global environment with monitoring data generally exhibiting the expected relationship between proximity to known sources and levels; however, temporal trends are not consistent. Estimated degradation half-lives, together with data in abiotic compartments and long-range transport potential indicate HBCD may be sufficiently persistent and distributed to be of global concern. The detection of HBCD in biota in the Arctic and in source regions and available bioaccumulation data also support the case for regulatory scrutiny. Toxicity testing has detected reproductive, developmental and behavioral effects in animals where exposures are sufficient. Recent toxicological advances include a better mechanistic understanding of how HBCD can interfere with the hypothalamic-pituitary-thyroid axis, affect normal development, and impact the central nervous system; however, levels in biota in remote locations are below known effects thresholds. For many regulatory criteria, there are substantial uncertainties that reduce confidence in evaluations and thereby confound management decision-making based on currently available information.
1. INTRODUCTION The compound 1,2,5,6,9,10-hexabromocyclododecane (HBCD, C12H18Br6) has been produced since the 1960s and is today the most used cycloaliphatic additive brominated flame retardant (BFR). HBCD is employed primarily in the building industry where it is incorporated typically at <3% by weight into extruded or expanded polystyrene foam materials. Secondary uses include upholstered furniture, automobile interior textiles, car cushions, and electric and electronic equipment. Recent global annual production estimates for HBCD are not available, in 2001 the total production volume was 16 700 t behind only tetrabromobisphenol-A (>130 000t) and decabromodiphenyl ether (56 100 t); The EU-wide consumption of HBCD in 2007 was 11 000 t (http://echa.europa.eu/doc/consultations/ recommendations/tech_reports/tech_rep_hbcdd.pdf)1. r 2011 American Chemical Society
Technical HBCD (t-HBCD), synthesized by bromine addition to 1,5,9-cyclododecatriene, is dominated by three diastereoisomers: α, β, and γ. The relative amounts of these isomers in the t-HBCD ultimately depends on the manufacturer, but the γ-isomer accounts for ca. 70% of the total with the α- and β-isomer contributing ca. 10 and 6%, respectively. Although quantitative data are not available, there are trace amounts of other diastereoisomers (δ and ε) present in t-HBCD.2 In 2009, HBCD underwent screening-level risk assessments by two international regulatory bodies to determine if it meets criteria of a persistent organic pollutant (POP): (i) United Received: May 6, 2011 Accepted: September 13, 2011 Revised: September 2, 2011 Published: September 14, 2011 8613
dx.doi.org/10.1021/es201548c | Environ. Sci. Technol. 2011, 45, 8613–8623
Environmental Science & Technology
CRITICAL REVIEW
Table 1. Summary of Persistence (P), Bioaccumulation (B), and Long-Range Transport Potential (LRTP) Screening Level Criteria and Estimated Values for HBCD. Ranges reflect variability in empirical data or model uncertainty. See SI for additional details.a criterion/characteristic
regulatory threshold
HBCD
persistence (P) half-life in air (days) half-life in water, (days)
>2 >60
0.4 5.2b no obs. deg (60 130b)
half-life in sediment, (days)
>180
1.1 128c
aerobic, 20 C
11 128
anaerobic, 20 C
1.1 115 >180
6.9 no obs. degc
aerobic, 20 C
63 no obs. deg
anaerobic, 20 C detected in remote region
(yes/no)
6.9 yes
half-life in soil, (days)
120 (12 1200)b
overall persistence (Pov, days) bioaccumulation (B) log KOW log BCF (L kg Log BAF (L kg
1 1
5
5.4
wet weight)
3.7
3.9 4.3
wet weight)
3.7
BMF (lipid-normalized) TMF (lipid-normalized)
5.8
3.7 6.1 0.1 11 0.3 2.2
long-range transport potential (LRTP) persistence in air (days)
>2
vapor pressure (Pa)
<1000
yes
detected in remote region
(yes/no)
yes
CTD (km)
inconclusive
600 (200 1500)
a
KOW = octanol-water partition coefficient; BCF = bioconcentration factor; BAF = bioaccumulation factor; BMF = biomagnification factor; TMF = trophic magnification factor; CTD = characteristic travel distance (distance at which air concentration is 1/e of initial value). b Based on model estimates rather than empirical data. c Tests conducted at lower, more environmentally relevant test concentrations (i.e., ug/kg dwt vs mg/kg dwt) yielded the lower-bound estimated degradation halflives. no obs. deg denotes that degradation rate constants was not quantified; See SI Section S2.
Nations Environment Programme (UNEP) Stockholm Convention on POPs (http://chm.pops.int/Convention/POPsReviewCommittee/POPRCMeetings/POPRC6/tabid/713/mctl/ ViewDetails/EventModID/871/EventID/86/xmid/2887/language/ en-GB/Default.aspx), and (ii) Protocol on POPs of the United Nations Economic Commission for Europe Convention on Long-Range Transboundary Air Pollution (UNECE-POPLRTAP, http://live.unece.org/env/lrtap/welcome.html). Although the screening level criteria of persistence (P), bioaccumulation (B) and toxicity (T) are not harmonized between the two treaties, HBCD was judged as fulfilling the P, B, and T criteria as defined under Annex D of the United Nations Environment Program (UNEP) Stockholm Convention and Executive Body Decision 1998/2 of the UNECE-POP-LRTAP. This paper examines the quality and quantity of information for HBCD in the broad categories of persistence, bioaccumulation, toxicity, and long-range transport. These data are reviewed
in the context of possible global action and the impact of uncertainty due to data gaps and poor and/or conflicting data, on decision-making confidence in full, rather than screening-level, risk assessment activities. While two reviews of HBCD were completed in the last 10 years, at the time of writing of this review, 263 additional papers have appeared in the peerreviewed literature since 2006.3,4 New information on analytical approaches and challenges to measuring enantiomers and enantiomer-specific environmental processes, detection of the meso-forms (δ and ε) in the environment, temporal trends in biota and abiotic environmental compartments, human exposure, recent work on toxicology and the usefulness of (animal) models to help interpret monitoring data are discussed.
2. PHYSICAL CHEMICAL PROPERTIES 2.1. Chemical Identity. The CAS Registry contains two numbers representing the undefined mixtures of commercial or t-HBCD: 25637-99-4 (without numbering for the position of the bromine substitution pattern) and 3194-55-6 for 1,2,5,6,9,10HBCD.5 The stereochemistry of HBCD is complex including 16 stereoisomers3 and HBCD is subject to isomerization during product synthesis and in the environment.6 9 Most of the reported chemical properties and toxicity tests to date are for t-HBCD, a mixture of isomers; however, as discussed in this review the individual isomers are shown to have characteristic properties related to their fate and behavior in the environment and their potential for toxic effects. 2.2. Chemical Properties. Measured and modeled physicalchemical properties considered to be of primary relevance for evaluating the environmental chemistry and toxicology of t-HBCD and the three main diastereoisomers (α-, β-, and γ-HBCD) were compiled and critically reviewed. These data include molar mass (M; g 3 mol 1), melting point (TM; K), water solubility (S; mol 3 L 1), vapor pressure (P; Pa), Henry’s law constant (H; Pa 3 m3 3 mol 1), and the dimensionless equilibrium partition coefficients between air and water (KAW), octanol and water (KOW), and octanol and air (KOA). Recommended methods for establishing thermodynamically consistent, and thus more reliable, solubility and partitioning properties were applied to obtain final adjusted values (FAVs).10 The details of the data compilation, review and calculation of FAVs is documented in the Supporting Information (SI) (Section S-1). Isomer-specific measurements are recommended for P, H, and KOA to further reduce uncertainty in these parameters.
3. PERSISTENCE (P), BIOACCUMULATION (B), AND LONG-RANGE TRANSPORT POTENTIAL (LRTP) Screening level criteria used by the Stockholm Convention (SC)11 and UN-ECE Protocol9 to categorize compounds according to P, B, and LRT potential are presented in Table 1 along with empirical and model-derived estimates (SI Sections S-2 to S-7). Issues and remaining research questions associated with categorization of HBCD are discussed. 3.1. Persistence. Empirical data characterizing persistence of HBCD are available for water, sediments, soil, and wastewater sludge12 15 and model-derived estimates are available for air and water.16 Key remaining uncertainties regarding persistence are related to (i) the paucity of empirical data, (ii) reliability of laboratory test methods and interpretation of data, and (iii) the extrapolation of laboratory-based data to the field. 8614
dx.doi.org/10.1021/es201548c |Environ. Sci. Technol. 2011, 45, 8613–8623
Environmental Science & Technology No empirical data are available for characterizing the reactivity HBCD in the atmosphere and estimates rely on model-derived output (AOPWIN v1.9217). The range in half-life in air presented in Table 1 (0.4 5.2 days) reflects different assumptions regarding the concentration of OH radicals (minor influence) and estimated uncertainty in the second-order rate constant for OH radical attack (main factor) (SI Section S-2). Empirical data on atmospheric degradation pathways and rate constants would reduce uncertainty in the determination of HBCD persistence. Results from one laboratory study (OECD 301D closed bottle test) and model-derived estimates are available to characterize biodegradation of HBCD in water (SI Section S.2).12 Biodegradation was reported to be negligible over the 28-day laboratory test; however, the reported initial water concentration was 7.7 mg L−1, that is, 2 3 orders of magnitude higher than the water solubility limit (SI Section S.1). As % degradation is calculated as the ratio of biochemical oxygen demand and theoretical oxygen demand, it is possible that biodegradation appeared negligible, because only a small fraction of the total HBCD added was freely dissolved (bioavailable). Model-derived estimates for biodegradation of HBCD in water are based on BIOWIN12 output and the estimation approaches proposed by Arnot et al.17 and Aronson et al.18 These results (60 130 days) indicate HBCD exceeds the criterion for persistence in water. However, it is difficult to assess the uncertainty inherent to these model-based approaches. The majority of empirical data available to characterize the persistence of HBCD are from studies using sediments, wastewater sludge and soil.13 15 Degradation of the parent compound in both sediments and soils proceeds faster in anoxic conditions and in the presence of microorganisms.13 Ranges in values for sediments and soils (Table 1) reflect variability in degradation rate constants at different initial HBCD concentrations (μg/kg vs mg/kg), with observed half-lives 1 2 orders of magnitude higher at mg/kg concentrations.14 One possible explanation is mass transfer limitations into microorganisms bias biodegradation kinetics at elevated substrate concentrations; however, half-lives were still well described by first-order kinetics. Based on the data presented in Table 1, HBCD does not satisfy the persistence criterion in sediments, but may do so for soils. Arguments to disregard rate constant data derived from tests conducted at elevated concentrations have merit, but need substantiation. Additional biodegradation experiments conducted across a gradient of initial concentrations would be useful. A remaining question relevant to assessment of persistence is isomer-specific susceptibility to degradation. Davis et al.14 reported no statistically significant differences in degradation halflives between α-, β-, and γ-HBCD, whereas Gerecke et al.15 reported a nearly 2-fold lower susceptibility to degradation for αHBCD. Photolysis has been hypothesized as an important cause for HBCD isomerization in dust.6 Results revealed a rapid photolytically mediated shift from γ-HBCD to α-HBCD, and slower degradation via elimination of HBr. Calculated half-lives (t1/2) showed decay in ∑HBCDs concentration was faster in light-exposed samples (t1/2 = 12 weeks), than in light-shielded dust (t1/2 = 26 weeks). Overall, a key uncertainty is the degree laboratory-based degradation data reflect the behavior of HBCD in the environment. For example, HBCD depth profiles in sediment cores suggest slower degradation kinetics than laboratory-derived estimates.19 However, since degradation of HBCD is known to be enhanced by microbial activity, presence at depth in sediment
CRITICAL REVIEW
could reflect absence of viable microbial communities rather than inhibited biodegradation due to other factors (e.g., temperature). These apparent discrepancies are also relevant for risk management (e.g., allowable uses, disposal practices) and therefore further investigation is warranted. 3.2. Overall Persistence (POV). Overall persistence (POV) is a complementary metric for assessing environmental persistence19 that incorporates distribution in the environment in addition to degradation kinetics in each medium. The OECD Tool was used to calculate POV for HBCD and benchmark chemicals.11,20 Based on the median parameter set, the POV of HBCD is approximately 120 d; however, values range from 12 to 1200 days, reflecting the assumed uncertainty in input variables, particularly in the degradation half-lives assumed for water and soil (SI Section S.3). The POVs included in the benchmarking exercise range from 70 to 4700 (known POPs) and 10 to 1300 days (non-POPs), respectively. The overlap in the ranges of POV between categories introduces some ambiguity, but the calculated POVs for HBCD tend to be lower than model output for other POPs. However, the POV based on the upper bound degradation half-lives for HBCD is approximately 1 order of magnitude higher, reinforcing the need to better characterize HBCD degradation rate constants, particularly in soil. 3.3. Bioaccumulation. The FAVs for log KOW of HBCD isomers range from 5.4 to 5.8, thus exceeding the B POP criterion. In the absence of biotransformation, neutral organic chemicals with log KOW of this magnitude are expected to be bioaccumulative in both aquatic and air-breathing organisms.21 The range of empirical bioconcentration factors (BCFs) 22,23 for technical HBCD and bioaccumulation factors (BAFs) for the isomers 20 22 exceed the B criteria (SI Section S-4). For example, estimated BAFs (L kg 1, wet weight) for α-HBCD in Lake Winnipeg (Canada) range from approximately 50 000 125 000.20 Average field-derived BAFs reported in Harrad et al. for lucustrine fish from England are 5900, 1300, 810, and 2100 for α-, β-, γ-, and ∑HBCDs, respectively.21 Note that these BAFs are (i) based on lipid-weight concentrations in muscle tissue and (ii) in units of L g 1. The corresponding average wet-weight BAFs in units appropriate for comparison to regulatory criteria (i.e., L kg 1 wet weight), estimated based on the assumptions detailed in Harrad et al.,21 range from approximately 41 000(γ-) 295 000(α-). Lipid-normalized biomagnification factors (BMFs) show variability, depending on the isomer and organisms considered (SI Section S.4). For example, the reported BMFs for α-, β-, and γ-HBCD for goldeye/mussels inhabiting Lake Winnipeg are 8.2, 1.0, and 0.3, respectively, but 1.9, 5.0, and 2.9 respectively for burbot/mussels.20 In a Lake Ontario study, BMFs for α- and γ-HBCD are 4.8 and 7.5, 1.0 and 1.5, and 1.1 and 0.8 for trout/ alewife, trout/smelt, and trout/sculpin, respectively.23 BMFs for the same predator/prey relationship are not always consistent among studies. For example, Sørmo et al.24 reported a lipidnormalized BMF of approximately 11 (total HBCD) for ringed seal/polar cod sampled from Svalbard, whereas data reported in Tomy et al.25 for the western Arctic yield a BMF of 0.1. Ranges in trophic magnification factors (TMFs) are more constrained, but still vary by isomer and food web and are not always statistically significant. Law et al.20,26 reported TMFs for α-, β-, γ-, and ∑HBCD in a freshwater aquatic food web (1,4, 1.3, 2.2, and 1.8 respectively), but only the TMF for γ- and ∑HBCD were significant. In contrast, Wu et al. reported that only the TMF for α-HBCD (2.2, p < 0.05) was statistically significant in organisms in a freshwater pond in China.27 8615
dx.doi.org/10.1021/es201548c |Environ. Sci. Technol. 2011, 45, 8613–8623
Environmental Science & Technology Additional studies to better understand variability in BMFs and TMFs would be valuable. However, one trend consistent across studies is the shift in isomer profile with trophic level that appears to favor accumulation of α-HBCD over the β- and γisomers.4,20,26,28 This phenomenon possibly results from a combination of differences in (i) solubility and partitioning behavior (i.e., bioavailability) (ii) uptake and depuration kinetics (including biotransformation) as well as (iii) preferential in vivo bioisomerization of β- and γ- isomers to α-HBCD.4,27 An additional complexity is the extent enantiomer-specificity influences these processes.28 For example, while HBCD chiral signatures were racemic in water and sediment, data revealed enantiomeric enrichment of ( )α-HBCD and (+)γ-HBCD in fish.20,27 The potential role of solubility, phase distribution, chemical uptake efficiencies, biotransformation, and bioisomerization were assessed as determinants of isomer-specific bioaccumulation potential (SI Sections S.5 and S.6). These novel analyses support the hypothesis that biotransformation and bioisomerization are key processes resulting in food web attenuation of γ-HBCD. Overall, as lipid-normalized BMFs and TMFs >1 indicate that a substance is bioaccumulative,29 the available empirical data generally support the conclusion that HBCD fulfills the B criteria. 3.4. Long-Range Transport (LRT) Potential. Under international conventions, LRT potential is characterized based on persistence in air, vapor pressure and detection in remote regions. HBCD fulfills two of three screening criteria unequivocally; however, model-based metrics characterizing LRT potential should also be considered. As with POV, it is useful to compare model output against other compounds (POPs and non-POPs), as was recently done for characteristic travel distance (CTD) and Arctic contamination potential (eACP10)15 (SI Section S.7). Since substantial overlap for both CTD and eACP10 between benchmark categories was found, a definitive categorization of HBCD with respect to other compounds is not possible. The assumed uncertainty in HBCD property values contributes to this ambiguity. With respect to LRT potential, this analysis further demonstrates the need to better characterize the fate of HBCD, particularly atmospheric degradation and gasparticle partitioning.16 In characterizing chemicals with respect to LRT potential, it is also important to recognize the presence of chemicals in remote environments is a function not only of LRT potential but also emission rate.30,31 Information on emission strength and mode of entry are required to interpret remote monitoring data.
4. ANALYTICAL CHEMISTRY A trend toward liquid chromatography-tandem mass spectrometry (LC-MS/MS) for determination of HBCD has been observed in recent years.32 Sample preparation methods, including extraction, cleanup, and fractionation have been described for determination of HBCDs in dust33 human milk,34 biological samples,35 and sewage sludge.36 Increased attention has been also given to methods to determine HBCDs in consumer products and rate of transfer into the indoor environment.37,38 A detailed overview of these methods is given in SI Section S.9). The LC-MS/MS methods are routinely applied39 and may include comprehensive suites of other BFRs, such as tetrabromobisphenol A (TBBPA).35,40 However, LC-MS analyses using electrospray ionization (ESI) are prone to ion suppression resulting in decreased sensitivity. This can be avoided by thorough
CRITICAL REVIEW
sample cleanup and by using 13C- and 2H-labeled HBCDs to compensate for variations in response. Alternatively, other ionization techniques, such as atmospheric pressure photoionization (APPI) and atmospheric pressure chemical ionization (APCI), reportedly less influenced by ion suppression, were assessed for analysis of HBCD together with other BFRs.41,42 A limited number of intercalibration studies have been performed at a low participation rate. Haug et al.43 indicated laboratories determined ∑HBCD concentrations in marine samples with satisfactory quality (RSD <35%). However, analyses of samples with low contamination (<2 ng g−1 lipid weight) needed improvement. There are currently no reference materials with certified values for HBCDs, but indicative values are available for some matrices. In any case, certification of HBCDs in relevant materials is imperative. The α-, β-, and γ-HBCD diastereomers are represented by a corresponding pair of enantiomers, which have been isolated using enantioselective HPLC and confirmed by X-ray diffraction.44 Since enantiomers can have vastly different biological properties/activities and degradation rates, enantioselective analyses are necessary to fully understand the environmental relevance of HBCD. Extraction and clean up methods for determination of HBCD enantiomers are similar to those for the diastereomers, but a chiral LC phase is required for separation. Most methods use a permethylated β-cyclodextrin chiral stationary phase with an elution order of ( )α-HBCD, (-)β-HBCD, (+)α-HBCD, (+)β-HBCD, (+)γ-HBCD, and ( )γ-HBCD.44 An important issue is the apparent nonracemic nature of LC-MS/MS chromatograms of HBCD standards.39,45 Matrix effects, stationary phase and mobile phase gradient are causal factors.39,45 Recent work by Guerra et al. demonstrated influence of temperature in the ESI source.46 Most importantly, measurement of enantiomeric fractions (EFs) of HBCD, that is, ratios of enantiomers, must be corrected using labeled surrogates.39 A comparison of corrected vs uncorrected HBCD EFs in fish demonstrated potential for false conclusions regarding enantioselective bioaccumulation of HBCD when using uncorrected EF values for α-HBCD.39 However, the same study also indicated enantioselective bioaccumulation of the ( )γ-HBCD enantiomer. The most recent reports indicate calculation of corrected EFs is now common in investigations of enantiospecific occurrence/accumulation of HBCD.21,34,46
5. OCCURRENCE, SPATIAL DISTRIBUTION, AND TEMPORAL TRENDS HBCD has been detected widely in biota and abiotic matrices in the northern hemisphere, exhibiting a strong gradient with proximity to main historical production/use regions (ref 3, SI Section S.8, Table S22). The paucity of reported concentrations in background/remote soils is the most substantial data gap. Isomeric profiles of HBCD in abiotic matrices (sediment, 47 sewage sludge,36 and air 33) are generally similar to the technical mixture. The meso forms of HBCD have been identified in technical mixtures.2,44 Recently, the δ-HBCD meso form was quantified in 43% of fish samples (1.0 11% ∑HBCDs) from English lakes,21 which exceeds abundance in a commercial HBCD formulation (0.5% ΣHBCDs). Its absence from water and sediment samples suggests formation via bioisomerization. Its presence in fish also raises the question whether its suspected presence in piscivorous birds16 arises from dietary intake and/or occurs via bioisomerization. 8616
dx.doi.org/10.1021/es201548c |Environ. Sci. Technol. 2011, 45, 8613–8623
Environmental Science & Technology
CRITICAL REVIEW
Figure 1. Reported or calculated mean concentrations of HBCD (total, μg kg 1 lipid-normalized) in [A] ivory gull eggs (Canadian Arctic archipelago 103), [B] peregrine falcon eggs (South Greenland46), [C] herring gull, puffin and kittiwake eggs (Norway104), [D] whole body lake trout (Lake Ontario48), [E] harbor porpoise blubber (United Kingdom105,106) and, [F] human milk (Sweden50) from different locations in the northern hemisphere over time. Note that each plot has its own scale on the y-axis.
Aside from the aforementioned studies, we find no evidence of widespread detections of the meso forms of HBCD in any environmental compartments. Concentrations of HBCD in biota from different locations in the northern hemisphere over time are shown in Figure 1. Temporal trends do not show a uniform pattern.16,48 In some studies,47,49 concentrations may have stabilized or begun to decrease over the past decade, whereas other studies indicate concentrations are increasing in other species, including humans.50,51 However, human data from urban areas likely indicates local or regional changes in production and/or usage and does not reflect trends in remote areas. Overall, lack of consistent trends in animal tissues convolutes determination of persistence or LRT. Analysis of temporal trend data in biota would be facilitated by (i) information on production/use of HBCD; (ii) temporal trends in exposure media; and (iii) information on stability/metabolism of HBCD isomers.
6. HUMAN EXPOSURE TO HBCDS Humans are exposed to HBCD via multiple sources including food, dust, air, and consumer products.4,22,52 54 Exposure may be dermal or oral, and can result from inhalation of vapor and particles.53 In the work environment, direct dermal exposure and inhalation of fine particles are of particular concern. Industrial workers at plants producing EPS with HBCD had HBCD levels up to 856 ng g−1 lw serum in their blood.55 Serum levels in nonoccupationally exposed individuals (<1 ng g−1 lw) were typically much lower54 and indirect exposure via the environment or products becomes in this case of primary concern. HBCD in indoor dust samples ranged from <5 to 130 200 ng g−1 (median 230 ng g−1 56). Abdallah and Harrad57 found HBCD in household air (median 180 pg m 3), household dust (median 1300 ng g−1), office dust (median 760 ng g−1), and car dust (median 13 000 ng g−1). Levels of human dietary exposure vary globally and regionally.58,59 Surveys in Europe and the US reveal levels in the range of <0.01 5 ng g−1 w/w.59 Fatty foods of animal origin such
as meat and fish are likely a major source of dietary human exposure and depends on consumption trends.58,60 Among dietary samples, the highest HBCD concentrations were reported for fish.51 Accordingly in Norway, where fish is an important part of the diet, intake of fish has been found to closely correlate with serum HBCD levels.51,61 The presence of HBCD in vegetables and vegetable oils may arise from use of sewage sludge as fertilizer.62 Stereoisomeric patterns in food samples suggest differences depending on food type.58,59 Although diet is a major exposure route in Europe, U.S. and China,4,58,63 indoor air, and in particular dust, can be important routes of exposure in both adults and toddlers.64,65 For a toddler of 10 kg who ingests an estimated 200 mg dust day−1 with HBCD contamination at the 95th percentile, intake via dust may exceed by 10 times the levels received via diet alone.64 In the study of Roosens et al.,59 daily exposure from food and dust was similar in magnitude for adults, and HBCD concentrations in serum only correlated significantly with estimates of exposure via dust. Exposure to dust appears to be an important exposure route because exposure remains more constant over time, compared to periodic intake of contaminated food.59 As in the case of dietary exposure, there is a large international variation in exposure via pathways like dust and air.21 As a result of continuous exposure via diet and via homes, offices and cars,57 HBCD was found in human adipose tissue66 and blood.55,59,67 Exposure occurs in early development as HBCD is transferred across the placenta to the fetus67 and transferred via breast milk. HBCD has been detected in breast milk from Europe,34,50,68,69 Asia,58,70 72 Russia,73 and the U.S.63 Hence, exposure to HBCD occurs at critical stages of development, both during pregnancy and postnatally via breast milk. HBCD in breast milk appears to mirror market consumption of HBCD.70 In breast milk from Japanese women (age 25 29), HBCD was not detected in any sample collected between 1973 and 1983, but increased from 1988 onward. In the period 8617
dx.doi.org/10.1021/es201548c |Environ. Sci. Technol. 2011, 45, 8613–8623
Environmental Science & Technology 1988 2006, α-HBCD was detected in all pooled milk samples with levels ranging from 0.4 to 1.9 ng g−1 lw. The levels reported from this Japanese study (1 4 ng g−1 lw) are higher than reported for women from Norway where HBCD was detected in only 1/10 samples at an average concentration of 0.13 ng g−1 lw.69 Temporal trends show an increase in Swedish milk up to 2002 after which a leveling occurs.50 While the extent of oral absorption of HBCD in humans is largely unknown, estimates of uptake via breast milk ranges from 50 to 100%.74 According to the EU risk assessment,74 intake of HBCD via breast milk is 1.5 ng kg−1 bw day−1 for 0 3 month olds and 5.6 ng kg−1 bw day−1 in 3 12 month old babies. However, Eljarrat et al.34 calculated the intake to be 175 ng kg−1 bw day−1 for 1 month olds in northern Spain. This is 12-times higher than the estimated daily intake (EDI) for 0 3 month old infants as determined in the EU risk assessment74 and 25 1500 times higher than the EDI for European adults.34,75 A Flemish dietary study suggests that the age group between 1 and 3 years is the highest exposed with an EDI for ∑HBCD of 7 ng kg−1 bw day−1 day. Newborns and adults are less exposed with EDIs of 3 and 1 ng kg−1 bw day−1 respectively.75 In all instances children appear to be more exposed than adults. Although α-HBCD, followed by γ- and β-HBCD, is the predominant diastereoisomer in all biota, including humans,74 profiles in human tissues are not consistent.34,58,59,63 External exposure (time, dose and stereoisomeric pattern), toxicokinetics, biotransformation, and time of sampling may all be important. In the study of Roosens et al.,59 γ-HBCD isomer dominated in food, whereas α-HBCD dominated in dust and was the sole isomer in serum. Although exposure via dust ingestion correlated with concentrations in serum, no correlation was evident with dietary exposure. While no enantioselective enrichment was detected in either dust or diet, substantial enrichment of ( )α-HBCD was observed in serum. The enrichment of the ( )α-HBCD enantiomer in humans appears to be due to in vivo enantioselective metabolism/excretion rather than ingestion of dust or diet.
7. TOXICITY Since the last reviews on HBCD toxicology,4,76 78 there have been significant advances in understanding mechanisms of toxicity. These include a better understanding of effects on thyroid function, brain development and neuron function, and reproduction and development. Recent studies have investigated the importance of oxidative stress and initiation of apoptotic cell death for mediating cellular toxicity (SI Section 10). Figure 2 graphically presents relationships between concentrations of HBCD in tissues or exposure media and disruption of thyroid function, reproductive system, and nerve function and development in classes of vertebrates. SI Table S.23 in Section S.10 provides brief descriptions of the studies cited in Figure 2. 7.1. Thyroid Toxicity. Thyroid effects are a consistent response of exposure to HBCD. Thyroid hyperplasia and lower circulating concentrations of thyroxine (T4) were determined in repeated oral dosing studies with rats.77 Studies have confirmed HBCDs potential to disrupt the thyroid axis in in vivo and in vitro animal models, including mammals,79 88 fish 89 91 and birds.92,93 The specific cellular and molecular mechanisms by which HBCD affects the thyroid axis reveal differences in HBCD accumulation and effects between males and females. Interpretation of data from animal models remains critical as no studies examined potential toxic effects in vivo in humans.
CRITICAL REVIEW
Figure 2. Lowest observed effects levels identified in studies examining effects of HBCD on thyroid end points (black circle), reproduction (red square), nerve function (blue hexagon) and oxidative stress (green circle) in biological systems where in vivo or in vitro HBCD exposure has been quantified (closed symbols), or where dietary HBCD exposure was reported (open symbols plotted against right-hand y-axis). Numbers in figure indicate referenced study.107−114
Mechanisms for HBCD’s effects on thyroid activity have been investigated. It appears likely HBCD affects the thyroid axis by altering expression of biotransformation enzymes. Thyroidal effects of other BFRs, including polybrominated diphenyl ethers, are partially mediated by alterations in CYP1A1 expression; similar mechanisms may occur for HBCD. However, despite at least one report to the contrary,94 HBCD does not appear to induce CYP1A82,95 and may inhibit activity of its regulated proteins.89,96 There is accumulating evidence that HBCD induces CYP2B and CYP3A through interaction with CAR/PXR receptors and analogues, and that inductions are key to explaining HBCD’s thyroid effects. Germer et al.,82 suggested HBCD effects in rats may be mediated via both the CAR and PXR receptors. However, extrapolation to potential effects in humans should be performed with caution because of the different affinity of the receptors for agonists in these two models.82 Other studies examined the ability of HBCD to induce thyroid effects through interaction with PXR and CAR receptors using in vitro exposure systems. Fery et al.,81 investigated HBCD in inducing CYP3A1 enzymes in rat and human hepatoma cells, and concluded HBCD is a PXR agonist which may account for disrupted thyroid activity in several animal models. Moreover, this induction pattern could explain why females are more affected in terms of thyroid hormone reductions than males.86 Concentrations required to elicit effects were several orders of magnitude higher than those in humans. Crump et al.,92 examined the effect of HBCD on the chicken xenobiotic-sensing orphan receptor (CXR). HBCD did not affect gene expression of CXR in chicken embryo hepatocytes, but the phase I metabolizing enzymes CYP2H1 and CYP3A37 were up-regulated by α-HBCD or t-HBCD, suggesting genes of the CXR were activated without activation of the receptor. This result, as well as the study by Germer et al.,82 shows CYP3 induction is one of the most sensitive end points reported in higher animals. 8618
dx.doi.org/10.1021/es201548c |Environ. Sci. Technol. 2011, 45, 8613–8623
Environmental Science & Technology Elevated rates of phase II conjugation activity have been demonstrated in HBCD-exposed organisms. Using tissues from the same rats as Germer et al.,82 Van der Ven et al.,86 reported an increase in hepatic T4-UGT activity accompanied by lower circulating total T4, increased pituitary weight, and increased thyroid weight and epithelial cell heights, indicative of glandular activation. These effects were noted only in females. Canton et al.,79 showed UGT1A1 was up-regulated in males, but down-regulated in females, which provides a mechanistic hypothesis for greater accumulation and slower elimination of HBCD by females. Elevated glucuronidation rates were demonstrated in fish fed HBCD.89 Evidence indicates circulating thyroid hormones may be depleted in HBCD-exposed organisms, which would increase peripheral tissue uptake of thyroid hormones.97 In addition to affecting the thyroid axis by increasing hormone turnover rates, HBCD affects interactions between thyroid hormones and their receptors. Using human cervical carcinoma cells Yamada-Okabe et al.87 (Figure 2: Study No. 6) showed HBCD activated the thyroid receptor in the presence of T3. Affinity for the receptor was noted even in the absence of T3 (ref 98, Figure 2 Study No. 7). Hyperthyroidal effects were observed in the T-screen assay where HBCD induced growth of the rat pituitary tumor GH3 cell line in the absence of T3, and to a greater extent in the presence of T3 (ref 99, Figure 2 Study No. 14). Schriks et al. (ref 88 Figure 2 Study No. 9) investigated the ability of HBCD to affect GH3 cells cultured from the rat pituitary. In the absence of T3, HBCD had no effect on GH3 cell proliferation indicating that HBCD did not directly interact with the thyroid receptor. However, in the presence of T3, HBCD potentiated GH3 cell proliferation. The inhibitory effect of HBCD in the presence and absence of T3 (ref 85, Figure 2 Study No. 10) has been examined on regression of tadpole tail tips in an ex vivo culture system. Collectively, these studies indicate the agonistic activity of HBCD toward T3-mediated effects. 7.2. Toxicity Concerns at Current Environmental Concentrations/Exposures. The studies reviewed indicate HBCD is not acutely toxic to terrestrial organisms, aquatic organisms and lower trophic organisms including algae (SI Section 10.1). There are insufficient data to assess the carcinogenic potential of HBCD, but carcinogenic initiation is not likely to occur via mutagenesis. Oral, dermal and inhalation toxicity is low in mammalian models.77 Acute toxicity of HBCD is not a general concern for aquatic organisms owing to the compound’s limited solubility (2 3 μg L 1 at 25 C).16 Several studies report no toxicity to algae, aquatic invertebrates or fish at nominal exposure concentrations that exceed the limit of HBCD solubility.77 However, several chronic effects arising from exposure to HBCD are consistently noted, with similar responses reported in vivo and/or in vitro using animal models or cell lines, respectively. Activation of the PXR/CAR complex, altered thyroid and lipid metabolism, potential reproductive effects, nervous system damage and oxidative stress, have all been observed. Deriving a sense of the potential for HBCD to affect end points in biota from urban, rural and remote locations is hampered by several factors. Few toxicological studies have examined tissue concentrations to relate to their measured end points. Nominal doses or exposures are given but toxicological bioavailability is not quantified. Concentrations of HBCD reported for biota from the environment are often determined as
CRITICAL REVIEW
“whole animal” or in tissues (e.g., muscle, adipose tissue), but which may have less relevance for determining effects (e.g., liver, thyroid, nervous tissue). Units of HBCD reported differ between studies and often cannot be compared without conversion, introducing a measure of uncertainty. Despite these limitations, there are available toxicological and environmental data on which to generally assess the potential for biological effects in wild fish, birds and mammals. Among the most sensitive end points are effects on the mammalian nervous system. Thresholds from in vitro exposures100,101 and in vivo exposures102 note lowest observable effects at low μM concentrations (approximately 1000 20 000 ng g−1). Thyroid function alterations appear to be the most universally noted effect of HBCD across biological levels of organization. In vitro cell culture work indicates HBCD can bind to thyroid receptors,84,87 affect T3 mediated cellular events84,85,88,97 and gene expression of thyroid binding proteins92 and increase expression of genes that can increase turnover rates of thyroid hormones.81 In vivo thyroid axis effects have also been noted at liver tissue concentrations of >40 000 ng g−1 in mammals86 and as low as ∼2600 ng g−1 in fish.89 Consideration of these collective data, with a primary focus on observed whole organism effects, against environmental concentrations noted in Figure 2 suggest that adverse effect thresholds may be exceeded in biota from some point source and source locations, but that biota in remote locations are below known adverse effect thresholds. Given the wide variety of experimental HBCD toxicity data available, development of a weight of evidence scheme for differentiating between measurable changes and toxicologically significant adverse effects would facilitate technical input into regulatory assessments. A framework for POP risk characterization has recently been proposed including a case study for HBCD.115
8. RECOMMENDATIONS In assessing the overall state of HBCD science globally, we make the following recommendations for future assessments: 1 Comprehensive studies on bioisomerization and biotransformation of HBCD diastereomers in various species are needed. 2 Studies to better understand the apparent variability in BMFs and TMFs are required. 3 Studies of the isomer-specific susceptibility of HBCD to degradation are required to improve our understanding of the isomer-specific fate of HBCD, which in turn will provide insights into behavior of this compound in biota. 4 Investigate global variation in dietary exposure and exposure through air and dust. 5 Additional laboratory intercalibration studies are required for HBCD. 6 There is a requirement for reference materials with certified values for HBCDs. 7 A formal decision-making framework for weighing and classifying technical evidence is needed to aid in distinguishing between measurable effects and toxicologically significant adverse effects. ’ ASSOCIATED CONTENT
bS
Supporting Information. Detailed information regarding criteria used in the assessment of HBCD persistence, bioaccumulation
8619
dx.doi.org/10.1021/es201548c |Environ. Sci. Technol. 2011, 45, 8613–8623
Environmental Science & Technology toxicity and long-range transport. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 905-319-6919 (C. M. T.); 204-983-5167 (G. T. T.). E-mail:
[email protected] (C. M. T.); gregg.tomy@ dfo-mpo.gc.ca (G. T. T.).
’ ACKNOWLEDGMENT Equal contribution to writing of the review by the authors is noted. ’ REFERENCES (1) Janak, K.; Covaci, A.; Voorspoels, S.; Becher, G. Hexabromocyclododecane in marine species from the western Scheldt estuary: Diastereoisomer- and enantiomer-specific accumulation. Environ. Sci. Technol. 2005, 39, 1987–1994. (2) Arsenault, G.; Konstantinov, A.; Marvin, C. H.; MacInnis, G.; McAlees, A.; McCrindle, R.; Riddell, N.; Tomy, G. T.; Yeo, B. Synthesis of the two minor isomers, δ and ε 1,2,5,6,9,10-hexabromocyclododecane, present in commercial hexabromocyclododecane. Chemosphere 2007, 68, 887–892. (3) Law, R. J.; Kohler, M.; Heeb, N. V.; Gerecke, A. C.; Schmid, P. P.; Voorspoels, S.; Covaci, A.; Becher, G.; Janak, K.; Thomsen, C. Hexabromocyclododecane challenges scientist and regulators. Environ. Sci. Technol. 2005, 39, 281A–287A. (4) Covaci, A.; Gerecke, A. C.; Law, R. J.; Voorspoels, S.; Kohler, M.; Heeb, N. V.; Leslie, H. A.; Allchin, C. R.; de Boer, J. Hexabromocyclododecane (HBCDs) in the environment and humans—A review. Environ. Sci. Technol. 2006, 40, 3679–3688. (5) Becher, G. The stereochemistry of 1,2,5,6,9,10-hexabromocyclododecane and its graphic representation. Chemosphere 2005, 58, 989–991. (6) Harrad, S. J.; Abdallah, M. A.-E.; Covaci, A. Causes of variability in concentrations and diasteromer patterns of hexabromocyclododecanes in indoor dust. Environ. Int. 2009, 35, 573–579. (7) Heeb, N. V.; Schweizer, W. B.; Mattrel, P.; Haag, W. R.; Gerecke, A. C.; Schmid, P. P.; Zennegg, M.; Vonmont, H. Regio- and stereoselective isomerization of hexabromocyclododecanes (HBCDs):kinetics and mechanisms of γ to α-HBCD isomerization. Chemosphere 2008, 73, 1201–1210. (8) Koppen, R.; Becker, P. R.; Jung, C.; Nehls, I. On thermally induced isomerization of hexabromocyclododecane stereoisomers. Chemosphere. 2008, 71, 656–662. (9) Law, K.; Palace, V. P.; Halldorson, T. H. J.; Danell, R. W.; Wautier, K.; Evans, R. E.; Alaee, M.; Marvin, C. H.; Tomy, G. T. Dietary accumulation of hexabromocyclododecane diastereoisomers in juvenile rainbow trout (Oncorhynchus mykiss) I: bioaccumulation parameters and evidence of bioisomerization. Environ. Toxicol. Chem. 2006, 25, 1757–1761. (10) Schenker, U.; MacLeod, M.; Scheringer, M.; Hungerbuhler, K. Improving data quality for environmental fate models: A least-squares adjustment procedure for harmonizing physicochemical properties of organic compounds. Environ. Sci. Technol. 2005, 39, 8434–8441. (11) Final act of the conference of plenipotentiaries on the Stockholm Convention on persistent organic pollutants; United Nations Environment Programme: Geneva, Switzerland, 2001. (12) Schaefer, E.; Haberlein, D. Hexabromocyclododecane (HBCD): Closed Bottle Test; 439E-102; Wildlife International Ltd.: Easton, MD, 1996. (13) Davis, J. W.; Gonsior, S. J.; Marty, G.; Ariano, J. M. The transformation of hexabromocyclododecane in aerobic and anaerobic soils and aquatic sediments. Water Res. 2005, 39, 1075–1084.
CRITICAL REVIEW
(14) Davis, J. W.; Gonsior, S.; Markham, D. A.; Friederich, U.; Hunziker, R. W.; Ariano, J. Biodegradation and product identifiaction of [14C]hexabromocyclododecane in wastewater sludge and freshwater aquatic sediment. Environ. Sci. Technol. 2006, 40, 5395–5401. (15) Gerecke, A. C.; Giger, W.; Hartmann, P. C.; Heeb, N. V.; Kohler, H.-P. E.; Schmid, P. P.; Zennegg, M.; Kohler, M. Anaerobic degradation of brominated flame retardants in sewage sludge. Chemosphere 2006, 64, 311–317. (16) Arnot, J.; McCarty, L.; Armitage, J. M.; Toose-Reid, L.; Wania, F.; Cousins, I. T. An Evaluation of Hexabromocyclododecane (HBCD) for Persistent Organic Pollutant Properties and the Potential for Adverse Effects in the Environment; European Brominated Flame Retardant Industry Panel (EBFRIP): Brussels, 2009. (17) Arnot, J.; Gouin, T.; Mackay, D. Practical Methods for Estimating Environmental Biodegradation Rates, CEMN2005003; Canadian Environmental Modeling Network: Peterborough, Ontario, 2005. (18) Aronson, D.; Boethling, R.; Howard, P.; Stiteler, W. Estimating biodegradation half-lives for use in chemical screening. Chemosphere 2006, 63 (11), 1953–1960. (19) Hexabromocyclododecane As a Possible Global POP; Nordic Chemical Group: Copenhagen, Norway, 2008. (20) Law, K.; Halldorson, T. H. J.; Danell, R. W.; Stern, G. A.; Gerwutz, S.; Alaee, M.; Marvin, C.; Whittle, D. M.; Tomy, G. T. Bioaccumulation and trophic transfer of some brominated flame retardants in a Lake Winnipeg (Canada) food web. Environ. Toxicol. Chem. 2006, 25, 2177–2186. (21) Harrad, S. J.; Abdallah, M. A.-E.; Rose, N. L.; Turner, S. D.; Davidson, T. A. Current-use brominated flame retardants in water, sediment, and fish from English Lakes. Environ. Sci. Technol. 2009, 43, 9077–9083. (22) Harrad, S.; de Wit, C. A.; Abdallah, M. A. E.; Bergh, C.; Bjorklund, J. A.; Covaci, A.; Darnerud, P. O.; de Boer, J.; Diamond, M.; Huber, S.; Leonards, P.; Mandalakis, M.; Oestman, C.; Haug, L. S.; Thomsen, C.; Webster, T. F. Indoor contamination with hexabromocyclododecanes, polybrominated diphenyl ethers, and perfluoroalkyl compounds: An important exposure pathway for people? Environ. Sci. Technol. 2010, 44 (9), 3221–3231. (23) Tomy, G. T.; Budakowski, W. R.; Halldorson, T. H. J.; Whittle, D. M.; Keir, M. J.; MacInnis, G.; Alaee, M. Biomagnification of α- and γhexabromocyclododecane (HBCD) isomers in a Lake Ontario food web. Environ. Sci. Technol. 2004, 38, 2298–2303. (24) Sørmo, E. G.; Salmer, M. P.; Jenssen, B.; Hop, H.; Baek, K.; Kovacs, K. M.; Lydersen, C.; Falk-Petersen, S.; Gabrielsen, G. W.; Lie, E.; Skaare, J. U. Biomagnification of polybrominated diphenyl ether and hexabromocyclododecane flame retardants in the polar bear food chain in Svalbard, Norway. Environ. Toxicol. Chem. 2006, 25, 2502–2511. (25) Tomy, G. T.; Pleskach, K.; Ferguson, S. H.; Hare, J.; Stern, G. A.; MacInnis, G.; Marvin, C. H.; Loseto, L. L. Trophodynamics of some PFCs and BFRs in a Western Canadian Arctic Marine Food Web. Environ. Sci. Technol. 2009, 43, 4076–4081. (26) Law, K.; Halldorson, T. H. J.; Danell, R. W.; Stern, G. A.; Gewurtz, S.; Alaee, M.; Marvin, C. H.; Tomy, G. T. Erratum: Bioaccumulation and trophic transfer of some brominated flame retardants in a Lake Winnipeg (Canada) food-web. Environ. Toxicol. Chem. 2007, 26, 190. (27) Wu, J. P.; Guan, Y. T.; Zhang, Y.; Luo, X. J.; Zhi, H.; Chen, S. J.; Mai, B. X. Trophodynamics of hexabromocyclododecanes and several other non-PBDE brominated flame retardants in a freshwater food web. Environ. Sci. Technol. 2010, 44, 5490–5495. (28) Tomy, G. T.; Pleskach, K.; Oswald, T.; Halldorson, T. H. J.; Helm, P. A.; Marvin, C. H.; MacInnis, G. Enantioselective bioaccumulation of hexabromocyclododecane and congener-specific accumulation of brominated diphenyl ethers in an Eastern Canadian Arctic marine food web. Environ. Sci. Technol. 2008, 42, 3634–3639. (29) Gobas, F. A. P. C.; De Wolf, W.; Burkhard, L. P.; Verbruggen, E.; Plotzke, K. Revisiting bioaccumulation criteria for POPs and PBT assessments. Integr. Environ. Assess. Manage. 2009, 5, 624–637. (30) Klecka, G. M.; Muir, D. C. G.; Dohmen, P.; Eisenreich, S. J.; Gobas, F. A. P. C.; Jones, K. C.; Mackay, D.; Tarazona, J. V.; van Wijk, D. 8620
dx.doi.org/10.1021/es201548c |Environ. Sci. Technol. 2011, 45, 8613–8623
Environmental Science & Technology Introduction to special series: Science-based guidance and framework for the evaluation and identification of PBTs and POPs. Integr. Environ. Assess. Manage. 2009, 5, 535–538. (31) Wu, R. S. S.; Chan, A. K. Y.; Richardson, B. J.; Au, D. W. T.; Fang, J. K. H.; Lam, P. K. S.; Giesy, J. P. Measuring and monitoring persistent organic pollutants in the context of risk assessment. Mar. Pollut. Bul. 2008, 57 (6 12), 236–244. (32) Covaci, A.; Voorspoels, S.; Ramos, L.; Neels, H.; Blust, R. Recent developments in the analysis of brominated flame retardants and brominated natural compounds. J. Chromatogr., A 2007, 1153, 145–171. (33) Abdallah, M. A.-E.; Harrad, S. J.; Covaci, A. Hexabromcyclododecanes and tetrabromobisphenol-A in indoor air and dust in Birmingham, UK: Implications for human exposure. Environ. Sci. Technol. 2008, 42, 6855–6861. (34) Eljarrat, E.; Guerra, P.; Martínez, E.; Farre, M.; Alvarez, J. G.; Lopez-Teijon, M.; Barcelo, D. Hexabromocyclododecane in human breast milk: levels and enantiomeric patterns. Environ. Sci. Technol. 2009, 43, 1940–1946. (35) Frederiksen, M.; Vorkamp, K.; Bossi, R.; Riget, F.; Dam, M.; Svensmark, B. Method development for simultaneous analysis of HBCD, TBBPA, and dimethyl-TBBPA in marine biota from Greenland and the Faroe Islands. Int. J. Environ. Anal. Chem. 2007, 87, 1095–1109. (36) Garcia-Valcarcel, A. I.; Tadeo, J. L. Determination of hexabromocyclododecane isomers in sewage sludge by LC-MS/MS. J. Sep. Sci. 2009, 32, 3890–3897. (37) Vilaplana, F.; Karlsson, P.; Ribes-Greus, A.; Ivarsson, P.; Karlsson, S. Analysis of brominated flame retardants in styrenic polymers - Comparison of the extraction efficiency of ultrasonication, microwave-assisted extraction and pressurised liquid extraction. J. Chromatogr., A 2008, 1196 1197, 139–146. (38) Kajiwara, N.; Sueoka, M.; Ohiwa, T.; Takigami, H. Determination of flame-retardant hexabromocyclododecane diastereomers in textiles. Chemosphere 2009, 74, 1485–1489. (39) Marvin, C. H.; MacInnis, G.; Alaee, M.; Arsenault, G.; Tomy, G. T. Factors influencing enantiomeric fractions of hexabromocyclododecane measured using liquid chromatography/tandem mass spectrometry. Rapid Commun. Mass Spectrom. 2007, 21, 1925–1930. (40) Guerra, P.; De la Torre, A.; Martinez, M. A.; Eljarrat, E.; Barcelo, D. Identification and trace level determination of brominated flame retardants by liquid chromatography/quadrupole linear ion trap mass spectrometry. Rapid Commun. Mass Spectrom. 2008, 22, 916–924. (41) Zhou, S. N.; Reiner, E. J.; Marvin, C.; Helm, P.; Riddell, N.; Dorman, F.; Misselwitz, M.; Shen, L.; Crozier, P.; MacPherson, K.; Brindle, I. D. Development of liquid chromatography atmospheric pressure chemical ionization tandem mass spectrometry for analysis of halogenated flame retardants in wastewater. Anal. Bioanal. Chem. 2010, 396, 1311–1320. (42) Zhou, S. N.; Reiner, E. J.; Marvin, C.; Kolic, T.; Riddell, N.; Helm, P.; Dorman, F.; Misselwitz, M.; Brindle, I. D. Liquid chromatography-atmospheric pressure photoionization tandem mass spectrometry for analysis of 36 halogenated flame retardants in fish. J. Chromatogr., A 2010, 1217, 633–641. (43) Haug, L. S.; Thomsen, C.; Liane, V. H.; Becher, G. Comparison of GC and LC determinations of hexabromocyclododecane in biological samples—Results from two interlaboratory comparison studies. Chemosphere 2008, 71, 1087–1092. (44) Heeb, N. V.; Schweizer, W. B.; Mattrel, P.; Haag, R.; Gerecke, A. C.; Kohler, M.; Schmid, P.; Zennegg, M.; Wolfensberger, M. Solidstate conformations and absolute configurations of (+) and ( )alpha-, beta-, and gamma-hexabromocyclododecanes (HBCDs). Chemosphere 2007, 68, 940–950. (45) Dodder, N. G.; Peck, A. M.; Kucklick, J. R.; Sander, l. C. Analysis of hexabromocyclododecane diastereomers and enantiomers by liquid chromatography-tandem mass spectrometry: chromatographic selectivity and ionization matrix effects. J. Chromatogr., A 2006, 1135, 36–42. (46) Guerra, P.; Eljarrat, E.; Barcelo, D. Enantiomeric specific determinatino of hexabromocyclododecane by liquid chromatography-quadrupole
CRITICAL REVIEW
linear ion trap mass spectrometry in sediment samples. J. Chromatogr., A 2008, 1203, 81–87. (47) Vorkamp, K.; Thomsen, M.; Falk, K.; Leslie, H.; Møller, S.; Sørensen, P. B. Temporal development of brominated flame retardants in Peregrine Flacon (Falco peregrinus) eggs from south Greenland (1986 2003). Environ. Sci. Technol. 2005, 39, 8199–8206. (48) Vorkamp, K.; Riget, F. F.; Bossi, R.; Dietz, R. Temporal trends of hexabromocyclododecane, polybrominated diphenyl ethers and polychlorinated biphenyls in ringed seals from East Greenland. Environ. Sci. Technol. 2011, 45, 1243–1249. (49) Ismail, N.; Gewurtz, S.; Pleskach, K.; Whittle, D. M.; Helm, P. A.; Marvin, C. H.; Tomy, G. T. Brominated and chlorinated flame retardants in Lake Ontario Lake trout between 1979 and 2004 and possible influences of food web changes. Environ. Toxicol. Chem. 2009, 28, 910–920. (50) F€angstr€ om, B.; Athanasiadou, I.; Odsj€o, T.; Noren, K. Bergman, Å. Temporal trends of polybrominated diphenyl ethers and hexabromocyclododecane in milk form Stockholm mothers, 1980 2004. Mol. Nutr. Food Res. 2008, 52, 187–193. (51) Knutsen, H. K.; Kvalem, H. E.; Thomsen, C.; Froshaug, M.; Haugen, M.; Becher, G.; Alexander, J.; Meltzer, H. M. Dietary exposure to brominated flame retardants correlates with male blood levels in a selected group of Norwegians with a wide range of seafood consumption. Mol. Nutr. Food Res. 2008, 52, 217–227. (52) Harrad, S.; Goosey, E.; Desborough, J.; Abdallah, M. A. E.; Roosens, L.; Covaci, A. Dust from UK primary school classrooms and daycare centers: The significance of dust as a pathway of exposure of young UK children to brominated flame retardants and polychlorinated biphenyls. Environ. Sci. Technol. 2010, 44, 4198–4202. (53) European Commission Risk Assessment Hexabromocyclododecane, EINECS No.: 247-148-4; Scientific Committee on Health and Environmental Risks: Brussels, Belgium, 2008. (54) Arctic Pollution; Arctic Monitoring Assessment Program: Oslo, Norway, 2009. (55) Thomsen, C.; Molander, P.; Daae, H. L.; Janak, K.; Froshaug, M.; Liane, V. H.; Thorud, S.; Becher, G.; Dybing, E. Occupational exposure to hexabromocyclododecane at an industrial plant. Environ. Sci. Technol. 2007, 41, 5210–5216. (56) Stapleton, H. M.; Allen, J. G.; Kelly, S. M.; Konstantinov, A.; Klosterhaus, S.; Watkins, D.; Mcclean, M. D.; Webster, T. F. Alternate and new brominated flame retardants detected in U.S. house dust. Environ. Sci. Technol. 2008, 42, 6910–6916. (57) Abdallah, M. A.-E.; Harrad, S. J. Personal exposure to HBCDs and its degradation products via ingestion of indoor dust. Environ. Internat. 2009, 35, 870–876. (58) Shi, Z.-X.; Wu, Y.-N.; Li, J.-G.; Zhao, Y.-F.; Feng, J.-F. Dietary exposure assessment of Chinese adults and nursing infants to tetrabromobisphenol-A and hexabromocyclododecanes: occurrence measurements in foods and human milk. Environ. Sci. Technol. 2009, 43, 4314–4319. (59) Roosens, L.; Abdallah, M. A. E.; Harrad, S.; Neels, H.; Covaci, A. Exposure to hexabromocyclododecanes (HBCDs) via dust ingestion, but not diet, correlates with concentrations in human serum: Preliminary results. Environ. Health Perspect. 2009, 117, 1707–1712. (60) Driffield, M.; Harmer, N.; Bradley, E.; Fernandes, A. R.; Rose, M.; Mortimer, D.; Dicks, P. Determination of brominated flame retardants in food by LC-MS/MS: diastereoisomer-specific hexabromocyclododecane and tetrabromobisphenol A. Food Addit. Contam. 2008, 25, 895–903. (61) Thomsen, C.; Knutsen, H. K.; Liane, V. H.; Froshaug, M.; Kvalem, H. E.; Haugen, M.; Meltzer, H. M.; Alexander, J.; Becher, G. Consumption of fish from a contaminated lake strongly affects the concentrations of polybrominated diphenyl ethers and hexabromocyclododecane in serum. Mol. Nutr. Food Res. 2008, 52, 228–237. (62) Kupper, T.; de Alencastro, L. F.; Gatsigazi, R.; Furrer, R.; Grandjean, D.; Tarradellas, J. Concentrations and specific loads of brominated flame retardants in sewage sludge. Chemosphere 2008, 71, 1173–1180. 8621
dx.doi.org/10.1021/es201548c |Environ. Sci. Technol. 2011, 45, 8613–8623
Environmental Science & Technology (63) Schecter, A.; Harris, T. R.; Shah, N.; Musurnbal, A.; Papke, O. Brominated flame retardants in U.S. food. Mol. Nutr. Food Res. 2008, 52, 266–272. (64) Abdallah, M. A. E.; Harrad, S.; Covaci, A. Hexabromocyclododecanes and tetrabromobisphenol-A in indoor air and dust in Birmingham, UK: Implications for human exposure. Environ. Sci. Technol. 2008, 42, 6855–6861. (65) Abdallah, M. A.-E.; Harrad, S. J.; Ibarra, C.; Diamond, M.; Melymuk, L.; Robson, M.; Covaci, A. Hexabromocyclododecanes in indoor dust from Canada, the United Kingdom, and the United States. Environ. Sci. Technol. 2008, 42, 459–464. (66) Antignac, J. P.; Cariou, R.; Maume, D.; Marchand, P.; Monteau, F.; Zalko, D.; Berrebi, A.; Cravedi, J. P.; Andre, F.; Le Bizec, B. Exposure assessment of fetus and newborn to brominated flame retardants in France: preliminary data. Mol. Nutr. Food Res. 2008, 52, 258–265. (67) Meijer, L.; Weiss, J.; Van Velzen, M.; Brouwer, A.; Bergman, A.; Sauerf, P. J. J. Serum concentrations of neutral and phenolic organohalogens in pregnant women and some of their infants in the Netherlands. Environ. Sci. Technol. 2008, 42, 3428–3433. (68) Lignell, S.; Aune, M.; Darnerud, P. O.; Cnattingius, S.; Glynn, A. Persistent organochlorine and organobromine compounds in mother’s milk from Sweden 1996 2006: compound-specific temporal trends. Environ. Res. 2009, 109, 760–767. (69) Polder, A.; Thomsen, C.; Lindstr€om, G.; LØken, K. B.; Skaare, J. U. Levels and temporal trends of chlorinated pesticides, polychlorinated biphenyls and brominated flame retardants in individual human breast milk samples from Northern and Southern Norway. Chemosphere 2008, 73, 14–23. (70) Kakimoto, K.; Akutsu, K.; Konishi, Y.; Tanaka, Y. Time trend of hexabromocyclododecane in the breast milk of Japanese women. Chemosphere 2008, 71, 1110–1114. (71) Malarvannan, G.; Kunisue, T.; Isobe, T.; Sudaryanto, A.; Takahashi, S.; Prudente, M.; Subramanian, A.; Tanabe, S. Organohalogen compounds in human breast milk from mothers living in Payatas and Malate, the Philippines: Levels, accumulation kinetics and infant health risk. Environ. Pollut. 2009, 157, 1924–1932. (72) Tue, N. M.; Sudaryanto, A.; Tu, B. M.; Isobe, T.; Takahashi, S.; Pham, H. V.; Tanabe, S. Accumulation of polychlorinated biphenyls and brominated flame retardants in breast milk from women living in Vietnamese e-waste recycling sites. Sci. Total Environ. 2010, 408, 2155–2162. (73) Polder, A.; Gabrielsen, G. W.; Odland, J. O.; Savinova, T. N.; Tkachev, A.; Loken, K. B.; Skaare, J. U. Spatial and temporal changes of chlorinated pesticides, PCBs, dioxins (PCDDs/PCDFs) and brominated flame retardants in human breast milk from Northern Russia. Sci. Total Environ. 2008, 391, 41–54. (74) Risk Assessment: Hexabromocyclododecane; Swedish National Chemicals Inspectorate, 2003. (75) Roosens, L.; Cornelis, C.; D’Hollander, W.; Bervoets, L.; Reynders, H.; Van Campenhout, K.; Van Den Heuvel, R.; Neels, H.; Covaci, A. Exposure of the Flemish population to brominated flame retardants: Model and risk assessment. Environ. Int. 2010, 36, 368–376. (76) Birnbaum, L. S.; Staskal, D. F. Brominated flame retardants: cause for concern? Environ. Health Perspect. 2004, 112, 9–17. (77) Data Summary and test plan for hexabromocyclododecane (HBCD); AR201-13459A; American Chemistry Council: Arlington, VA, 2001. (78) Darnerud, P. O. Toxic effects of brominated flame retardants in man and in wildlife. Environ. Int. 2003, 29, 583–624. (79) Canton, R. F.; Peijnenburg, A. A. C. M.; Hoogenboom, R. L. A. P.; Piersma, A. H.; van der Ven, L. T. M.; van den Berg, M.; Heneweer, M. Subacute effects of hexabromocyclododecane (HBCD) on hepatic gene expression profiles in rats. Toxicol. Appl. Pharmacol. 2008, 231, 267–272. (80) Ema, M.; Fujii, S.; Hirata-Koizumi, M.; Matsumoto, M. Twogeneration reproductive toxicity study of the flame retardant hexabromocyclododecane in rats. Reprod. Toxicol. 2008, 25, 335–351. (81) Fery, Y.; Buschauer, I.; Salzig, C.; Lang, P.; Schrenk, D. Technical pentabromodiphenyl ether and hexabromocyclododecane
CRITICAL REVIEW
as activators of the pregnane-X-receptor (PXR). Toxicology 2009, 264, 45–51. (82) Germer, S.; Piersma, A. H.; van der Ven, L.; Kamyschnikow, A.; Fery, Y.; Schmitz, H. J.; Schrenk, D. Subacute effects of the brominated flame retardants hexabromocyclododecane and tetrabromobisphenol A on hepatic cytochrome P450 levels in rats. Toxicology 2006, 218, 229–236. (83) Saegusa, Y.; Fujimoto, H.; Woo, G. H.; Inoue, K.; Takahashi, M.; Mitsumori, K.; Hirose, M.; Nishikawa, A.; Shibutani, M. m. t. a. j. Developmental toxicity of brominated flame retardants, tetrabromobisphenol A and 1,2,5,6,9,10-hexabromocyclododecane, in rat offspring after maternal exposure from mid-gestation through lactation. Reprod. Toxicol. 2009, 28, 456–467. (84) Schriks, M.; Roessig, J. M.; Murk, A. J.; Furlow, J. D. Thyroid hormone receptor isoform selectivity of thyroid hormone disrupting compounds quantified with an in vitro reporter gene assay. Environ. Toxicol. Pharmacol. 2007, 23, 302–307. (85) Schriks, M.; Zvinavashe, E.; Furlow, J. D.; Murk, A. J. Disruption of thyroid hormone-mediated Xenopus laevis tadpole tail tip regression by hexabromocyclododecane (HBCD) and 2,2 ’,3,3 ’,4,4 ’, 5,5 ’,6-nona brominated diphenyl ether (BDE206). Chemosphere 2006, 65, 1904–1908. (86) van der Ven, L. T. M.; van de Kuil, T.; Leonards, P. E. G.; Slob, W.; Lilienthal, H.; Litens, S.; Herlin, M.; Hakansson, H.; Canton, R. F.; van den Berg, M.; Visser, T. J.; van Loveren, H.; Vos, J. G.; Piersma, A. H. Endocrine effects of hexabromocyclododecane (HBCD) in a onegeneration reproduction study in Wistar rats. Toxicol. Lett. 2009, 185, 51–62. (87) Yamada-Okabe, T.; Sakai, H.; Kashima, Y.; Yamada-Okabe, H. Modulation at a cellular level of the thyroid hormone receptor-mediated gene expression by 1,2,5,6,9,10-hexabromocyclododecane (HBCD), 4,4 ’diiodobiphenyl (DIB), and nitrofen (NIP). Toxicol. Lett. 2005, 155, 127–133. (88) Schriks, M.; Vrabie, C. M.; Gutleb, A. C.; Faassen, E. J.; Rietjens, I. M. C. M.; Murk, A. J. T-screen to quantify functional potentiating, antagonistic and thyroid hormone-like activities of poly halogenated aromatic hydrocarbons (PHAHs). Toxicol. In Vitro. 2006, 20, 490–498. (89) Palace, V. P.; Pleskach, K.; Halldorson, T. H. J.; Danell, R. W.; Wautier, K.; Evans, B.; Alaee, M.; Marvin, C. H.; Tomy, G. T. Biotransformation enzymes and thyroid axis disruption in juvenile rainbow trout (Oncorhynchus mykiss) exposed to hexabromocyclododecane diastereoisomers. Environ. Sci. Technol. 2008. (90) Lower, N.; Moore, A. The impact of a brominated flame retardant on smoltification and olfactory function in Atlantic salmon (Salmo salar L.) smolts. Mar Freshwater Behav. Physiol. 2007, 40, 267–284. (91) Palace, V.; Park, B.; Pleskach, K.; Gemmill, B.; Tomy, G. Altered thyroxine metabolism in rainbow trout (Oncorhynchus mykiss) exposed to hexabromocyclododecane (HBCD). Chemosphere 2010, 80, 165–169. (92) Crump, D.; Chiu, S.; Egloff, C.; Kennedy, S. W. Effects of Hexabromocyclododecane and polybrominated diphenyl ethers on mRNA expression in chicken (Gallus domesticus) hepatocytes. Toxicol. Sci. 2008, 106, 479–487. (93) Marteinson, S. C.; Kimmins, S.; Letcher, R. J.; Palace, V. P.; Bird, D. M.; Ritchie, I. J.; Fernie, K. J. Diet exposure to technical hexabromocyclododecane (HBCD) affects testes and circulating testosterone levels in American Kestrels (Falco sparverius). Environ. Sci. Technol. 2011. (94) Zhang, X. L.; Yang, F. X.; Xu, C.; Liu, W. P.; Wen, S.; Xu, Y. Cytotoxicity evaluation of three pairs of hexabromocyclododecane (HBCD) enantiomers on Hep G2 cell. Toxicol. In Vitro. 2008, 22, 1520–1527. (95) Kuiper, R. V.; Canton, R. F.; Leonards, P. E. G.; Jenssen, B. M.; Dubbeldam, M.; Wester, P. W.; van den Berg, M.; Vos, J. G.; Vethaak, A. D. Long-term exposure of European flounder (Platichthys flesus) to the flame-retardants tetrabromobisphenol A (TBBPA) and hexabromocyclododecane (HBCD). Ecotoxicol. Environ. Saf. 2007, 67, 349–360. 8622
dx.doi.org/10.1021/es201548c |Environ. Sci. Technol. 2011, 45, 8613–8623
Environmental Science & Technology (96) Ronisz, D.; Finne, E. F.; Karlsson, H.; Forlin, L. Effects of the brominated flame retardants hexabromocyclododecane (HBCDD), and tetrabromobisphenol A (TBBPA), on hepatic enzymes and other biomarkers in juvenile rainbow trout and feral eelpout. Aquat. Toxicol. 2004, 69, 229–245. (97) Kamstra, J. H.; Du, T.; Visser, T. J.; van der Ven, T. M.; Murk, A. J.; Hamers, T. In vitro induction of organic anion transporting polypeptides (OATPs) as a possible mechanism for thyroid hormone disruption by hexabromocyclododecane (HBCD); Fourth International BFR Workshop: The Netherlands, 2007. (98) Sakai, H.; Yamada-Okabe, T.; Kashima, Y.; Matsui, M.; Aono, T.; Aoyagi, M.; Hasegawa, J. Effects of brominated flame retardants on transcruiptional activation mediated by thyroid hormone receptor. Organohalogen Compd. 2003, 61, 215–218. (99) Hamers, T.; Kamstra, J. H.; Sonneveld, E.; Murk, A. J.; Kester, M. H. A.; Andersson, P. L.; Legler, J.; Brouwer, A. In vitro profiling of the endocrine-disrupting potency of brominated flame retardants. Toxicol. Sci. 2006, 92, 157–173. (100) Mariussen, E.; Fonnum, F. The effect of brominated flame retardants on neurotransmitter uptake into rat brain synaptosomes and vesicles. Neurochem. Internat. 2003, 43, 533–542. (101) Dingemans, M. M. L.; Heusinkveld, H. J.; de Groot, A.; Bergman, Å.; van den Berg, M.; Westerlink, R. H. S. Hexabromocyclododecane inhibits depolarization-induced increase in intracellular calcium levels and neurotransmitter release in PC12 cells. Toxicol. Sci. 2009, 107, 490–497. (102) Lilienthal, H.; van der Ven, L. T. M.; Piersma, A. H.; Vos, J. G. Effects of the brominated flame retardant hexabromocyclododecane (HBCD) on dopamine-dependent behavior and brainstem auditory evoked potentials in a one-generation reproduction study in Wistar rats. Toxicol. Lett. 2009, 185, 63–72. (103) Braune, B. M.; Mallory, M. L.; Gilchrist, H. G.; Letcher, R. J.; Drouillard, K. G. Levels and trends of organochlorines and brominated flame retardants in Ivory Gull eggs from the Canadian Arctic, 1976 to 2004. Sci. Total Environ. 2007, 378, 403–417. (104) Knudsen, L. B.; Gabrielsen, G. W.; Verreault, J.; Barrett, R.; Skaare, J. U.; Polder, A.; Lie, E. Temporal trends of brominated flame retardants, cyclododeca-1,5,9-triene and mercury in eggs of four seabird species from Northern Norway and Svalbard; 942/2005; Norwegian Polar Institute: Tromso, Norway, 2005. (105) Law, R. J.; Bersuder, P.; Allchin, C. R.; Barry, J. Levels of the flame retardants hexabromocyclododecane and tetrabromobisphenol A in the blubber of Harbor porpoises (Phocoena phocoena) stranded or bycaught in the U.K., with evidence for an increase in HBCD concentrations in recent years. Environ. Sci. Technol. 2008, 40, 2177–2183. (106) Law, R. J.; Bersuder, P.; Barry, J.; Wilford, B. H.; Allchin, C. R.; Jepson, P. D. A significant downturn in levels of hexabromocyclododecane in the blubber of harbor porpoises (Phococena phocoena) stranded or bycaught in the UK: an update to 2006. Environ. Sci. Technol. 2008, 42, 9104–9109. (107) van der Ven, L. T. M.; Verhoef, A.; van de Kuil, T.; Sloub, W.; Leonards, P. E. G.; Visser, T. J.; Hamers, T.; Herlin, M.; Hakansson, H.; Olausson, H.; Piersma, A. H.; Vos, J. G. A 28-day oral dose toxicity study enhanced to detect endrocine effects of hexabromocyclododecane in Wistar rats. Toxicol. Sci. 2006, 94, 281–292. (108) Aniagu, S. O.; William, T. D.; Chipman, J. K. Changes in gene expression and assessment of DNA methylation in primary human hepatocytes and HepG2 cells exposed to the environmental contaminants hexabromocyclododecane and 17-b estradiol. Toxicol. 2009, 256, 143–151. (109) Reistad, T.; Fonnum, F.; Mariussen, E. Neurotoxicity of the pentabrominated diphenyl ether mixture, DE-71, and hexabromocyclododecane (HBCD) in rat cerebellar granule cells in vitro. Arch. Toxicol. 2006, 80, 785–796. (110) Hu, J.; Liang, Y.; Chen, M.; Wang, X. Assessing the toxicity of TBBPA and HBCD by zebrafish embryo toxicity assay and biomarker analysis. Environ. Toxicol. 2009, 24, 334–342. (111) Zhang, X. L.; Yang, F. X.; Xu, C.; Liu, W. P.; Wen, S.; Xu, Y. Cytotoxicity evaluation of three pairs of hexabromocyclododecane
CRITICAL REVIEW
(HBCD) enantiomers on Hep G2 cell. Toxicol.in Vitro. 2008, 22 (6), 1520–1527. (112) 6-Week Administration Study of 1,2,5,6,9,10-Hexabromocyclododecane for Avian Reproduction Toxicity under Long-Day Conditions Using Japanese Quail, Study No: 08-013; Chemicals Evaluation Office, Policy Planning Division, Environmental Health Department, Ministry of the Environment: Japan, 2009. (113) Deng, J.; Yu, L.; Liu, C.; Yu, K.; Shi, X.; Yeung, L. W. Y.; Lam, P. K. S.; Wu, R. S. S.; Zhou, B. Hexabromocyclododecane-induced developmental toxicity and apoptosis in zebrafish embryos. Aquat. Toxicol. 2009, 93, 29–36. (114) Eriksson, P.; Fischer, C.; Wallin, M.; Jakobsson, E.; Fredriksson, A. Impaired behaviour, learning and memory, in adult mice neonatally exposed to hexabromocyclododecane (HBCDD). Environ. Toxicol. Pharmacol. 2006, 21, 317–322. (115) Arnot, J. A.; Armitage, J. M.; McCarty, L. S.; Wania, F.; Cousins, I. T.; Toose-Reid, L. Toward a consistent evaluative framework for POP risk characterization. Environ. Sci. Technol. 2011, 45, 97–103.
8623
dx.doi.org/10.1021/es201548c |Environ. Sci. Technol. 2011, 45, 8613–8623
POLICY ANALYSIS pubs.acs.org/est
Analysis and Status of Post-Combustion Carbon Dioxide Capture Technologies Abhoyjit S. Bhown* and Brice C. Freeman Electric Power Research Institute (EPRI), 3420 Hillview Avenue, Palo Alto, California 94304, United States
bS Supporting Information ABSTRACT: The Electric Power Research Institute (EPRI) undertook a multiyear effort to understand the landscape of postcombustion CO2 capture technologies globally. In this paper we discuss several central issues facing CO2 capture involving scale, energy, and overall status of development. We argue that the scale of CO2 emissions is sufficiently large to place inherent limits on the types of capture processes that could be deployed broadly. We also discuss the minimum energy usage in terms of a parasitic load on a power plant. Finally, we present summary findings of the landscape of capture technologies using an index of technology readiness levels.
’ INTRODUCTION A number of governmental agencies at the state, federal, and international levels are actively discussing limitations on the emissions of greenhouse gases (GHG), including CO2. The electricity generation industry is among the first group of emission sources being targeted for GHG reductions, including CO2, because coal-fired power plants are the largest stationary point-source emitters of anthropogenic CO2.1 Regulations requiring the electric generation industry to reduce CO2 emissions have already emerged at the state2 and local levels, and are expected at the national level in the United States. Even without explicit legislation, requests to add new coal-fired power plants are being denied at the local permitting level on the basis that they do not include CO2 controls at the onset or have no plans to add them in the near term.3 Indeed in 2009, coal supplied only 44.5% of U.S. electricity, down from 48.2% in 2008 and over 50% in years prior.4 Table 1 shows that in 2008, CO2 emissions from electricity generation in the U.S. accounted for about 40% of anthropogenic CO2 emissions and 34% of the total anthropogenic GHG emissions.5,6 Globally, approximately 31.2 Gt CO2 was emitted in 2008 from fossil fuel combustion and cement, dropping by 1.3% in 2009.7 One option for controlling CO2 emissions is carbon capture and storage (CCS), where CO2 is separated from flue gas and permanently stored in large subsurface geologic reservoirs. In part, due to the absence of national and international regulations, or limited carbon markets, no postcombustion CO2 capture (PCC) systems have been demonstrated at utility-scale on any coal- or gas-fired power plants thus far. Moreover, no PCC technologies are available for order with commercial guarantees for coal-fired power plants, though some companies provide r 2011 American Chemical Society
guarantees for commercial-scale gas-fired power plants today with intentions for coal-fired power plants by 2012.8 Near-term PCC technologies are being developed and demonstrated at subscale and are progressing toward market readiness, but these first-generation capture technologies are energy intensive and, when implemented, will significantly increase the cost of electricity (COE) for the host power plant.9,10 CCS economics typically include the cost of capture and compression, and sometimes transport and storage. Costs are commonly reported in $/tonne CO2 captured or $/tonne CO2 avoided, which are defined elsewhere.9,10 The International Energy Agency (IEA) summarized the findings from recent CCS economic studies from leading institutions and reported the average cost for capture and compression to be $58/tonne CO2 avoided, not including transport and storage, leading to 63% rise in the levelized cost of electricity.11 Recognizing the need to reduce these costs, the U.S. Department of Energy (DOE) set a goal for CO2 mitigation technologies to be widely deployable by 2020 that can achieve less than 35% COE increase with the following: 90% CO2 capture, compression to 140 bar (2000 psi), 160 km (100 miles) transportation, 1525 m (5000 ft) injection, and 100-year storage with measurement, monitoring, and verification.12 Among capture, compression, transport, and storage, CO2 capture is the most expensive, representing 6070% of the total CCS cost, with compression requiring ∼20%, and the balance for transport, Received: December 21, 2010 Accepted: September 9, 2011 Revised: September 7, 2011 Published: September 12, 2011 8624
dx.doi.org/10.1021/es104291d | Environ. Sci. Technol. 2011, 45, 8624–8632
Environmental Science & Technology
POLICY ANALYSIS
Table 1. Anthropogenic Greenhouse Gas Emissions, United States for 20085,6 CO2 emission source
a
Gt CO2
CO2 from electricity generation
2.36
CO2 from non-electricity energy CO2 from other sources
3.37 0.11
total anthropogenic CO2
5.84
total anthropogenic non-CO2 GHGa
1.12
total anthropogenic GHG, CO2(equiv.)
6.96
Based on CO2 equivalent 100-year global warming potential
injection, and monitoring,9,10 and is, as such, the primary focus for CCS cost reduction opportunities.13 Moreover, because capture is energy intensive, capture energy has the largest influence on the CO2 capture cost; therefore, low-energy CO2 capture is the most direct route to achieve lower cost CCS. In an effort to promote the development of low-energy capture technologies, EPRI undertook an effort to identify, vet, and appropriately accelerate emerging PCC technologies that show promise in this area.1417 Over 120 technologies have been evaluated thus far. This technology review campaign provided EPRI with insights on the state of the science and technology of capture systems, some of which are presented in this paper.
’ CHALLENGES FACING CAPTURE SOLUTIONS Central to postcombustion CO2 capture are two challenges: the scale of CO2 emissions and the energy of separation. To discuss these challenges, we focus on coal-fired power plants, which generate nearly 45% of electricity in the U.S.4 Flue gas from coal-fired power plants typically contains 7075% N2, 1015% CO2, 810% H2O, 34% O2, with trace levels of SOx, NOx, and other compounds.10 For this paper, we use 13% CO2 as a typical value. The Energy Information Administration (EIA) reports annual CO2 emissions from each U.S. power plant, and their data show the average CO2 emission in 2009 to be 25.17 t CO2/MWe-day generated or 9187 t CO2/MWe-year generated.18 Note that these values reflect the actual electricity generated, without regard to the capacity factor, which is the ratio of the actual generation to name plate generation during a fixed time period. In part due to the recent financial recession and the shift of power generation away from coal, the U.S. coal fleet had an average capacity factor of 63.8% in 2009, substantially lower than the 72.2% in prior years.4 Scale of Emissions. At 2360 Mt/year, the scale of anthropogenic CO2 emissions from U.S. electricity generation plants is vast. To provide perspective, Table 2 shows the production levels of the 50 largest produced chemicals for 2009 in the U.S. and globally, based on estimates from the latest year for which these figures are available publicly. Our estimates were in some cases extrapolated from 200219 and adjusted to account for changes in production.20,21 Moreover, these data are from the chemical industry, and do not include all possible industries that may produce the same chemicals as well. The accuracy of each chemical listed may therefore vary; nonetheless, even under such approximations, some important observations can be drawn. These observations are important because some capture processes propose to use chemicals to capture power plant flue gas CO2 or produce a saleable product in a once-through process, akin to SO2 capture by limestone to produce gypsum, and thereby
avoid the very large energy penalties to regenerate solvents or sorbents. In a hypothetical case wherein a PCC process uses a commodity chemical to capture CO2 in an equimolar ratio, i.e., one mole of the chemical captures permanently one mole of CO2, then the “GWe-yr at 90% capture” columns in Table 2 show the maximum coal-fired power plant generation in GWe-yr whose flue gas (containing 13% CO2) could be treated at a 90% capture rate before depleting either the U.S. or global supplies of that chemical. For example, a process that uses one mole of ammonia to capture one mole of CO2 in a nonregenerative system will exhaust the domestic supply of ammonia after treating CO2 emitted from only 4.4 GWe-yr of coal-fired power plant generation, equivalent to 2.2% of U.S. generation. A more efficient process that might use one mole of a chemical to capture two or three moles of CO2 would correspondingly decrease chemical intake of the capture process by a factor of two or three. Even so, such changes would not materially affect the fact that current supplies of commodity chemicals are small relative to the 200 GWe-yr of coal-fired generation, representing 63.8% of the 314 GWe installed capacity in the U.S., or the >1000 GWe-yr coalfired generation worldwide. Note also that many of these chemicals are themselves precursors to others chemicals also on the list. For example, nearly 60% of sulfuric acid (no. 1) is used to make phosphoric acid (no. 10) and about 60% of propylene (no. 7) is used to make polypropylene (no. 18). Hence, the annual total of 419 Mt (approximately 8700 Gmoles) of the top 50 chemicals already double-counts a portion of material actually available. Finally, note that in this hypothetical process, it is certainly possible to increase the production of a chemical if it is favorable for CO2 capture, but then the combination of its manufacture, transportation, and its use in the capture process itself must yield an overall net reduction of CO2 in order to have a net reduction in anthropogenic CO2 emissions. However, the chemical industry as a whole, as exemplified in Table 2, yields a net increase in CO2 emission because it involves the use of large quantities of energy or involves chemistry that releases CO2 somewhere in a production chain.22 This analysis leads to two important conclusions. First, any capture process that uses a commodity chemical as a reagent to capture CO2 in a “once-through” manner will quickly exhaust the global supply of that chemical before making a meaningful reduction in CO2 emissions. For the process to be widely applicable, the capture process must employ regenerative chemistry, use membranes for separation, or devise a method for producing a reagent in quantities many times higher than the current manufacturing capacity of the entire chemical industry while ensuring that a net reduction of CO2 still occurs. This argument, therefore, severely limits the potential of “oncethrough” CO2 capture processes. The second conclusion is that any process that uses CO2 as a reactant to produce a chemical or other commodity product will overwhelm global market demand for the produced product. Although once-through capture methods may be technically feasible, their cost estimates often depend on the salability of the produced products to offset an otherwise high-cost CO2 capture process. In such cases, widespread applicability of the process would not be economically feasible because doing so would saturate U.S. or global markets for that chemical by orders of magnitude, effectively dropping the product’s price to near 8625
dx.doi.org/10.1021/es104291d |Environ. Sci. Technol. 2011, 45, 8624–8632
Environmental Science & Technology
POLICY ANALYSIS
Table 2. Approximate Production of Top 50 Chemicals in 20091921 estimated U.S.
estimated global GWe-yr at
Mt
GWe-yr at
90% capture
Gmola
Mt
90% capture
1
sulfuric acid
38.7
394
2.1
199.9
1879
2
nitrogen
32.5
1159
6.2
139.6
4595
24.5
3
ethylene
25.0
781
4.2
112.6
3243
17.3
4
oxygen
23.3
829
4.4
100.0
3287
17.5
5
lime
19.4
347
1.8
283.0
4653
24.8
6
polyethylene (HDPE, LDPE, etc.)
17.0
530
2.8
60.0
1729
9.2
7 8
propylene ammonia, synthetic anhydrous
15.3 13.9
354 818
1.9 4.4
53.0 153.9
1134 8332
6.0 44.3
9
chlorine
12.0
169
0.9
61.2
795
4.2
10
phosphoric acid
11.4
116
0.6
22.0
207
1.1
45b
acetic acid
2.3
38
0.2
8.0
123
0.7
46
propylene oxide
2.1
37
0.2
6.3
100
0.5
47
phenolic resins
2.1
21
0.1
6.8
63
0.3
48
calcium carbonate (precipitated)
2.0
20
0.1
13.0
120
0.6
49 50
butadiene (1.3) nylon resins and fibers
2.0 1.9
36 8
0.2 0.0
10.3 2.3
175 8
0.9 0.0
total
a
Gmola
419
8,681
46
2009 coal-fired generation, GWe-yr4
200
installed coal-fired capacity, GWe4
314
2,412
48,385
10.0
257 >1000+ >1000+
CO2 from electricity
2,400
54,545
∼9600
218,182
CO2 from all sources
6,000
136,364
∼31200
750,000
For polymers, the number of moles represents the moles of repeating units. b Numbers 11-44 are provided in Supporting Information Table S.1.
zero. This argument, therefore, severely limits the beneficial use of CO2 insofar as it involves saleable chemicals. Note that the annual combined mass of the top 50 chemicals manufactured in the U.S. totals approximately 419 Mt whereas the electricity generation sector in the U.S. alone emits nearly six times that amount of CO2 and the U.S. emits about 15 times. Similar ratios are true for global chemical production relative to CO2 emissions. And assuming an equimolar reaction and 90% CO2 capture, the top 50 chemicals combined in the U.S. could only capture CO2 emitted from 46 GWe-yr of coal-fired electricity generation in the U.S., a fraction of the 200 GWe-yr actual generation. Likewise, the global top 50 chemicals could capture CO2 from about 257 GWe-yr of coal-fired electricity, a fraction of the global generation. To further this perspective on scale, Table S.2 in the Supporting Information shows the ten largest CO2 emitters in the U.S. electricity generation sector in 2009.18 These power plants typically consist of multiple boilers, each capable of producing sufficient steam to generate a few hundred MWe of electricity. The ten power plants shown in the table generate a sum of 29 GWe-yr electricity and emit about 192 Mt CO2, almost onehalf of the estimated tonnage of the top 50 chemicals produced the United States. Plant Scherer, the largest of these, annually emits 24.1 Mt of CO2, about the same as all ethylene produced by the U.S. chemical industry. By contrast, CO2 for enhanced oil recovery (EOR) in the U.S., which represents the majority of CO2 utilized in the U.S., is approximately 65 Mt/year.23 This is about equivalent to the emissions from 10 GWe of coal-fired power plants, i.e., the top
four power plants could supply all of the CO2 needed for EOR in the U.S. In terms of storage potential, however, we note that a recent study estimated the U.S. has approximately 3500 GWe-yr of CO2 emissions from coal-fired power plants.23 These data demonstrate the first challenge in CO2 capture: the scale of CO2 emissions, which places inherent limits on potential capture technologies and limits the beneficial use of CO2. We next turn our attention the second major challenge in CO2 capture: the energy required. Energy of Separation. As stated earlier, the energy of regeneration factors strongly in the overall cost of capture. To understand how much improvement is possible from CO2 capture improvements to current state of the art, it is necessary to determine the theoretical minimum amount of energy that is required to perform the separation of CO2 from the remaining constituents in flue gas. In this analysis, we specifically focus on capture and exclude compression since, as noted earlier, capture has the most potential for cost reductions and efficiency improvement. Moreover, the minimum energy needed to compress CO2 from atmospheric pressure to pipeline pressure is about the same as the minimum energy needed to separate CO2 from the flue gas of a typical coal-fired power plant.24 However, while new compressor concepts hope to reduce capital cost, not much efficiency gains can be expected from commercial compressors,24,25 and as we note earlier, compression represents only about 2030% of the energy needed for CCS, while capture represents 7080%, at least for near-term technologies. To separate a mixture such as flue gas into its pure components, the minimum amount of work that must be done is equal 8626
dx.doi.org/10.1021/es104291d |Environ. Sci. Technol. 2011, 45, 8624–8632
Environmental Science & Technology
POLICY ANALYSIS
Figure 1. Isothermal separation of a mixture.
to the change in Gibbs function (free energy) of a reversible process26 Emin ¼ ΔG ΔH TΔS
ð1Þ
where Emin is the minimum energy required, ΔG is the change in Gibbs function, ΔH is the change in enthalpy, T is absolute temperature, and ΔS is the change in entropy. Because no reaction takes place, ΔH is zero, and the minimum work for separation is only the change in entropy of mixing. As a reasonable approximation, we may assume that flue gas is an ideal binary mixture of CO2 and “inerts” representing all non-CO2 constituents, which gives26 Emin ¼ ΔGmix ¼ TΔSmix ¼ RT
∑i xi ln xi
require only 90% CO2 capture, as shown in the Isothermal Separation Unit I of Figure 1b. Flue gas containing 13% CO2 could be separated into two streams: one stream containing 100% CO2 equal to 90% of the CO2 originally in the flue gas, and the other stream containing inerts and the 10% of CO2 not captured. To calculate the entropy of mixing in Unit I shown in Figure 1b, which is the CO2 capture process of interest, we calculate the entropy of mixing of the overall separation of Units I and II into pure components, and subtract the entropy of mixing in Unit II. If xCO2 is the mole fraction of CO2 in flue gas, and η is the fraction of CO2 captured in the first separation, then (1 η)xCO2 is the mole fraction of CO2 fed into Unit II. Equation 2 therefore implies that the entropy change in Unit I (per mole CO2 captured) is given by þ II þ II Emin ¼ ΔGImix ¼ ΔGImix ΔGIImix ¼ RTðΔSImix ΔSIImix Þ
¼ RT½xCO2 ln xCO2 þ ð1 xCO2 Þlnð1 xCO2 Þ
RT f½xCO2 ln xCO2 þ ð1 xCO2 Þlnð1 xCO2 Þ ηxCO2 ½1 ηx½ð1 ηÞxCO2 lnfð1 ηÞxCO2 g
¼
ð2Þ where ΔGmix is the change in the Gibbs function due to mixing per mole of mixture, R is the universal gas constant, T is the absolute temperature, ΔSmix is the change in entropy due to mixing, xi is the mole fraction of component i, and xCO2 is the mole fraction of CO2. Note that eq 2 reflects the entropy change per mole mixture when the mixture is separated completely into pure components as shown in Figure 1a, whereas ΔGmix/xCO2 is the change in Gibbs function per mole CO2. At 40 °C, completely separating a flue gas containing 13% CO2 into pure CO2 and pure “inerts” will take a minimum energy of 1006 J/mol mixture, equivalent to 7738 J/mol CO2 or 0.1758 GJ/tonne CO2. Practical processes have a finite selectivity, and therefore the separated CO2 stream is never 100% pure. It often contains some amount of other flue gas components, e.g., water, oxygen, nitrogen, etc., that transport with CO2. These other components must often be further separated by supplemental methods to achieve a stream of sufficient CO2 purity to meet transportation and storage purity requirements, but for this calculation, we have assumed the final CO2 stream to be 100% pure. Moreover, complete separation of flue gas into pure components is not necessarily required for CO2 capture operations. For example, a PCC application may
þ f1 ð1 ηÞxCO2 glnf1 ð1 ηÞxCO2 gg
ð3Þ
At 40 °C, separating pure CO2 from a flue gas containing 13% CO2 at 90% capture will take a minimum energy of 0.1611 GJ/tonne CO2 captured. We can compare this minimum energy of separation to the electrical energy generated by a coal-fired power plant. The average U.S. coal-fired power plant emits approximately 9187 tonnes CO2 for each (MWe-yr) of net electricity it generates, as described earlier. Expressed differently, this value is equivalent to 3.43 GJ of net electricity generated per tonne of CO2 emitted. For 100% capture, the minimum energy needed to capture each tonne of CO2 shown by eq 2 is 0.1758 GJ, which is 5.12% of the electrical energy generated by the power plant. At 90% capture, eq 3 showed the minimum value as 0.1611 GJ for each tonne of CO2 captured, which is 4.22% of the electrical energy generated by the power plant. Hence, if the only source of energy to separate 13% CO2 in flue gas were the electricity generated by the same coal-fired power plant, the thermodynamic minimum parasitic load would be 5.12% 8627
dx.doi.org/10.1021/es104291d |Environ. Sci. Technol. 2011, 45, 8624–8632
Environmental Science & Technology
POLICY ANALYSIS
Figure 2. Minimum energy penalty.
at 100% capture and 4.22% at 90% capture. Such calculations can be extended to other CO2 concentrations and percent captured via eq 3. The yellow lines in Figure 2 show these results at 40 °C and are directly proportional to the absolute temperature. For typical flue-gas capture conditions of 1115% CO2 and 90% capture from a power plant emitting 25.17 tonnes CO2/MWeday, the minimum energy penalty is about 4.04.5% of the net electricity generated before capture as shown in the lines in Figure 2. Based on generation and emissions data from the EIA, the shaded yellow region in Figure 2 shows the minimum energy as the upper 90th percentile and lower 10th percentile bounds of U.S. power plants assuming their flue gas contained 13% CO2.4 Note that the yellow lines in the figure represent the minimum energy for capture, and specifically do not include any makeup power. From a thermodynamic and separations perspective, this is an important metric to consider. From a policy and electricity generation perspective, however, energy penalty is often reported to be the increase in fuel input on a fixed net power output, i.e., to include makeup power that was lost due to CCS.9,27 Under such a definition, the increase required to deliver a fixed net power is therefore higher than that shown in Figure 2 and is given by the equation27 PL ¼
1 1 1 pl
ð4Þ
where pl is the parasitic load of the capture process itself and PL is the increase in input fuel to provide the same net power after a capture process. The blue lines and blue shaded area in Figure 2 show the result of applying eq 4 to the yellow lines and shaded area in Figure 2, i.e., the minimum increase in fuel requirement for the same net power output. Note again, these values exclude compression. Further note that if energy is extracted from the power plant that is not used to generate electricity, e.g., waste heat, then the parasitic load could be less, but the amount of energy required per tonne CO2 captured does not change. Likewise, if steam from the steam cycle were used to drive the CO2 capture process, which is typically the case in conventional solvent-based capture systems, then the energy penalty could be lower if the capture process can use the steam more efficiently than converting it to electricity that
subsequently drives the capture process. Indeed, all solvent processes are thermally integrated into the power plant to minimize the parasitic load on electricity generation. For comparison, a hypothetical CO2 separation process, not including compression, based on aqueous solutions of monoethanol amine (MEA) imposes approximately a 20% reduction in net plant output,10 or about five times the calculated minimum energy. The work done by multistage compression is about 810% of net plant output,10 or about twice the minimum energy, again highlighting the fact that capture is far more energy inefficient than compression. We believe that in order to overcome these two central challenges in CO2 capture, scale and energy penalty, a trio of expertise working synergistically is needed: chemists to develop new capture chemistry, process engineers concurrently working to design processes around the new chemistry, and power plant personnel concurrently working to integrate the process with a power plant. In addition to the challenge of scale of emissions and energy for capture just discussed, there are additional practical constraints such as limited water availability, limited land availability, and the tolerance of capture solvents to other flue gas constituents. Other Constraints. A detailed engineering study modeling an advanced amine capture system found that the capture equipment required 40 400 m2 (10 acres) of land for a 850 MWe power plant application, whose equipment normally occupies 257 000 m2 (63 acres).28 For some power plants, space for CCS is either not available or not close enough to the stack such that CCS may be impractical or require too many changes to the plant infrastructure. Additionally, some capture processes require nearly twice as much cooling water on a per-MWe basis relative to that used by the host power plant, which may not be available in some plant locations or may come at a very high cost.29 Finally, flue gas contains heavy metals, SOx, NOx, and a myriad of other compounds, which the capture process must tolerate or additional pretreatment equipment must be added, which will increase both the cost and energy penalty of the process. These issues are power-plant specific, but nearly all plants will have some limiting factors that will influence the type or size of capture equipment options. Capture processes which are less demanding with regard to these constraints (water, land, flue gas pretreatment) will have a greater potential to be widely applicable.
’ STATUS OF TECHNOLOGY DEVELOPMENT In our work,1417 we actively sought process developers, reviewed major literature articles, and identified postcombustion capture processes that we could technically evaluate. Technologies in very early stages or early science were largely omitted, leaving approximately 120 postcombustion CO2 capture technologies where we could identify the underlying separation method, determine approximate mass and energy balances, potential constraints, and for most, a state of development. Virtually all capture processes used an absorbent (solvent), adsorbent, membrane, converted CO2 to a mineral, or employed biofixation. As part of the evaluation, 95 of the 120 technologies were assigned a state of development status using a Technology Readiness Level (TRL) taxonomy. Originally developed by the NASA for conveying the readiness of a technology for space applications,30 the TRL system has since been adopted by EPRI and other public and private agencies.31,32 The TRL system divides the development process into nine discrete levels from the conceptual stage at TRL of 1 to full-scale tested commercial products TRL of 9. 8628
dx.doi.org/10.1021/es104291d |Environ. Sci. Technol. 2011, 45, 8624–8632
Environmental Science & Technology Each TRL step is described in Table S.3 (see Supporting Information). We assigned each capture process a TRL level at the time of its review. Figure 3 shows a histogram of the results from EPRI’s study. The majority of PCC technologies are absorption-based processes (60%), followed by membrane (14%), mineralization (14%), and adsorption (12%). This result is perhaps reflective of the fact that absorption technologies are widely used already in the broader chemical process and oil/gas production industries, and hence operational confidence of the basic process is relatively high. Most of the technologies reviewed are in the earlier stages of development with a few moving from the laboratory to pilot operations. Among the capture technologies which showed the most promise for lower cost (low energetics) operation, the majority rank in the TRL 2 to 4 level, indicating that they will require significant time to achieve commercial status. Although the capture technologies EPRI reviewed did not include many in the TRL 1 tier, this does not reflect the population of capture processes. This is attributable to the fact that the first TRL phase is very short-lived and the subject technology is not frequently disclosed publically or it represents basic science that is not yet focused on a capture process per se. Capture technologies are generally disclosed by the developer only after they have cleared TRL 2. In terms of timing, Figure S.1 (see Supporting Information) represents EPRI’s translation of Alstom’s projections for their Chilled Ammonia Process33 into TRL taxonomy, under a wellfunded and aggressive development schedule. Even under these conditions, the total development cycle is nearly 1015 years. Table 3 shows the general state of capture development for the three major types of capture methods. The usage of the technology in the chemical process industry and operational confidence are obviously correlated. For absorption-based processes, the major source of the energy penalty is from the thermal regeneration of the solvent. The energy penalty for adsorbent-based processes uses thermal and/or vacuum regeneration, while membrane processes generally rely on a vacuum applied to the permeate-side to separate CO2. Though we considered mineralization processes, they are considered “once-through” processes and are therefore subject to the constraints of supply of the reagents and disposal of products as discussed earlier. The following section reviews each of the three primary capture methods and describes the current state and themes of development for each type. Absorption. Absorption refers to the uptake of CO2 into the bulk phase of another material—for example, dissolving CO2 molecules into a liquid solution. This contrasts to adsorption, where CO2 molecules adhere to the surface of a solid particle. Both are used widely in the chemical, petrochemical, and other industries, with absorption being more common than adsorption. The absorption cycle is based on temperature dependent acidbase reaction wherein the CO2, which is acidic, reacts with a basic solution at flue gas temperatures (4060 °C). After reacting with CO2, the “loaded” solution is directed to a regeneration vessel where the solvent is heated to reverse the reaction thus liberating gaseous CO2 which is then collected, dried, compressed, and transported to a storage reservoir. State of Development. Among the three major separation methods considered for CO2 capture, absorption is the most mature. Absorption processes are common in the chemical process industries, and there is significant commercial experience
POLICY ANALYSIS
Figure 3. Distribution of post-combustion CO2 capture technology types by TRL ranking.
with their operation including for CCS applications and treating natural gas. This stems from the fact that absorption-based processes are generally less expensive for large-scale separations, easier to operate, and are more robust than other processes. Indeed, all the near-term CO2 capture processes are solventbased, as are the majority of the less-developed capture processes. One challenge for CO2 capture is to develop solvents that regenerate using minimal energy. This challenge is difficult because low regeneration energy solutions typically have a low heat of absorption, which translates to poor CO2 absorption characteristics. The near-term CO2 capture technologies impose an estimated 2530% parasitic load on the net power output (including CO2 compression to pipeline requirements) of a typical coal-fired power plant in order to capture 90% of emitted CO2.10,28 Though absorption technologies continue to improve, these advances offer only incremental gains. Still, because of widespread commercial use of absorption processes, even such incremental advances are commercially significant. Current research efforts in absorption focus on lowering the heat of regeneration. A common point of reference for performance is 30 wt % aqueous MEA, which, depending on variances in process design, requires approximately 3.6 GJ of energy for each tonne of CO2 captured, imposing parasitic load of approximately 20% on a coal-fired power plant in the absence of compression; with compression, this value increases to 30%.10 Regeneration occurs at approximately 120 °C under slight pressure, with a number of factors determining the optimal operating condition for a given plant. There are multiple approaches to lowering the regeneration energy, all based on modifying the solvent. The first approach is to increase the solvents’ concentration above 30 wt % so as to reduce the working volume of solvent needed to capture CO2 and then to add corrosion inhibitors and other additives to mitigate the adverse effects of operating at higher concentrations. This results in lower sensible heating requirements, less water evaporation in the reboiler, and consequently lower regeneration energy. Fluor’s Econamine FG+ is one example of a process that has reduced regeneration energy by increasing the solvent concentration, adding corrosion inhibitors, and optimizing the process integration.34 A second approach is to use different solvents that are still based on aqueous alkanomines but either have lower capacity and/or lower reaction kinetics. Examples include diethanolamine (DEA), methyldiethanolamine (MDEA), various secondary amines, various tertiary amines, and numerous blends of amines. Additional differentiation comes from process optimization and thermal integration into the plant. Cansolv is one example of a 8629
dx.doi.org/10.1021/es104291d |Environ. Sci. Technol. 2011, 45, 8624–8632
Environmental Science & Technology
POLICY ANALYSIS
Table 3. State of Post-Combustion CO2 Capture Development absorbent
membrane
commercial usage in CPIa
high
moderate
operational confidence
high
high, but complex
low to moderate
primary source of
solvent regeneration
sorbent regeneration
compression on feed and/
energy penalty development trends a
adsorbent
(thermal) new chemistry, thermal integration
low/niche
(thermal/vacuum)
or vacuum on permeate
new chemistry, process configuration
new membrane, process configuration
CPI = Chemical process industries.
process using blends of amines.35 In contrast to alkanomines, Alstom’s chilled ammonia process (CAP) uses aqueous ammonium salts (ammonium carbonate) to capture CO2 as ammonium bicarbonate, and aims to make use of waste heat and to regenerate at elevated temperatures and pressures to reduce downstream compression.36 In general, this collection of aqueous solvents represents the near-term CO2 capture technologies. Remaining approaches to absorption, which have a longer path to commercialization, involve changing the absorption chemistry itself or using promoters (catalysts) to enhance the rate of absorption. Examples include computationally designed solvents, the use of “fast” amines such as piperazine37 to improve reaction kinetics, and the use of industrial enzymes (carbonic anhydrase)38 as a catalyst to increase the kinetics of aqueous CO2 reactions. Though amine-based solvents are more common, other solvents with affinity for CO2 such as ionic liquids39 and siloxanes40 are also being investigated. Most of these are at conceptual or early lab scale testing. These also aim to reduce the energy of regeneration, while some offer the additional advantage of potentially reducing the gasliquid contactor volume. The other challenge is to produce low cost solvents. All solvents degrade during operation and must be replaced, imparting both disposal cost and replacement cost for the operator. Some solvents with the best performance also happen to be the most expensive to produce. The overall operational economics will ultimately dictate the extent to which higher-performing, highercost solvents offer a lower cost of capture compared to lowerperforming, lower-cost solvents. Adsorption. Adsorption refers to uptake of CO2 molecules onto the surface of a solid sorbent, to which they adhere via weaker van der Waals forces (physisorption) or stronger covalent bonding (chemisorption). This contrasts to absorption, where CO2 molecules dissolve into the bulk of the material itself. Adsorption processes can be implemented several different ways, with the adsorbent used in packed beds or fluidized beds. In a packed bed, adsorbent is loaded into a column, flue gas flows through the void spaces between the adsorbent particles, and the CO2 adsorbs onto the particle surfaces. In fluidized beds, flue gas flows upward through a column at higher velocities such that the adsorbent particles are suspended in the gas flow. In both approaches, the adsorbent selectively adsorbs more CO2 relative to the other constituents passing through the column. The adsorbent is typically regenerated by changing the pressure (lower pressure) and/or temperature (higher temperate) to liberate the adsorbed CO2. In a packed bed, this is accomplished by switching the flue gas flow to a second column, and heating the CO2-laden adsorbent in the first column to drive off the CO2. In a fluidized bed, the sorbent is typically conveyed continuously to the regenerator column and then back to the absorber column. State of Development. Adsorption-based separation processes are used at large-scale in chemical process industry
applications, but they are less common than absorption-based systems. Because adsorption can be used in different process configurations, both adsorbent properties and process design can strongly influence the effectiveness of separation. Consequently, developing adsorption processes for very large-scale CO2 capture would require development of both adsorbent materials and corresponding processes.41 Adsorbents being developed for CO2 capture exhibit a variety of origins, characteristics, and chemistries. Low-cost, moistureinsensitive carbonaceous adsorbents are being investigated at the University of Wyoming.42 Metal oxide frameworks (MOFs), which are nanoporous materials synthesized via the self-assembly of metal or metal oxide vertices interconnected by rigid organic linker molecules, are being investigated at a variety of universities such as University of California at Los Angeles,43 Berkeley44 and elsewhere.45 Other types of materials are also being investigated at early stages. None of these have yet advanced beyond lab-scale testing, with large-scale production and accompanying process development largely unknown. In addition, it is not yet known whether the benefits of using novel adsorbents will be sufficient to offset their anticipated higher cost. Hence, some researchers have taken the approach of using less sophisticated and inexpensive adsorbents, and employing them in novel process designs to capture CO2. Research groups pursuing this strategy include RTI in North Carolina,46 and The Ohio State University.47 Many of these are at bench scales. Membranes. Membranes separate CO2 from flue gas because the transport speed of a gas constituent through the membrane is proportional to its permeability, defined as the product of its solubility and diffusivity in the membrane material. If CO2 has a higher permeability than other constituents of flue gas, then CO2 preferentially permeates it. In some cases, chemical agents that selectively react with CO2 are added to the membrane to facilitate CO2 transport. Under typical conditions, CO2 will transport across a membrane only if the partial pressure of CO2 is higher on the side of the membrane that contacts the flue gas than on the other. A partial pressure gradient of CO2 can be obtained by pressurizing the flue gas on one side of the membrane, applying a vacuum on the other side of the membrane, or both. In effect, applying this pressure differential provides the energy for separation. State of Development. Membrane separation processes are used much less frequently in chemical process industrial applications compared to either absorption or adsorption. With few exceptions such as reverse osmosis, large-scale membranes are not commonly available or used in the chemical process industries. However, there is a commercial polymeric membrane process from UOP (Separex) that removes CO2 from natural gas, in applications with a gas flow rate equivalent to that from the flue gas from a 300-MWe power plant.48 Though this polymer is not suitable for separating CO2 from flue gas, it does demonstrate 8630
dx.doi.org/10.1021/es104291d |Environ. Sci. Technol. 2011, 45, 8624–8632
Environmental Science & Technology that the general scale of membrane systems is within the range required for coal-fired power plant applications. Coal-fired power plants present some unique challenges for membranes. Particulates can deposit on the membrane surface, decreasing its permeability or damaging it physically over time.49 Another issue is that membranes deployed at power plant scale require a very large membrane area. For example, a 600-MWe power plant could require more than 1 million square meters of membrane which must be packaged into modular containers. Large modules today are typically 3546 cm diameter (1418 in.), 0.91.2 m long (34 ft), and contain a membrane area of 50100 m2 (5381076 ft2) that is spirally wound around a central tube. Thousands of such membrane modules connected in a serial and parallel piping network would be required to treat flue gas emissions from a typical power plant, presenting engineering design challenges to ensure proper and efficient gas distribution since flue gas is at relatively low pressures. Despite these challenges, membranes are attractive because they can be arranged in a relatively small footprint (about 0.4 ha [1.0 acre] for a 600-MWe power plant) and potentially impose a small energy penalty.50 Some membranes will also permeate the water present in flue gas, which offers the possibility to “harvest” water that may be used by the host power plant. Current membrane development efforts are at laboratory or bench scales, with much effort directed to developing new materials.
’ DISCUSSION CO2 capture presents two central challenges that we have attempted to quantify and provide perspectives: scale and energy. The scale of CO2 emissions is several multiples larger that the commodity chemicals made by the chemical industry. This places limits on capture processes that use or generate commodity chemicals, suggesting that regenerative chemistry or membranebased separation, and not “once-through” chemistry to capture CO2, is needed in order for a capture process to be widely deployed. As explained earlier, the minimum parasitic load to capture 90% of the CO2 from a typical coal-fired power plant is about 44.5% of the net electrical output as measured from the same plant without capture. Actual processes, such as capture by MEA without compression, require about five times this minimum energy with careful thermal integration into the power plant. In addition CO2 capture processes face a variety of additional challenges including limited land space, limited water availability, tolerance to other flue gas constituents, and other local power plant criteria, such as major O&M intervals and turn down ability. The Technology Readiness Level index can be a useful tool to ascertain the landscape of CO2 capture technologies. The majority of capture processes continue to focus on absorption, followed by adsorption, membranes, and other separation methods. This is due to the fact that absorption is more widely practiced in the chemical process industries and has practical advantages at large scales. However, unless radically different capture chemistries are investigated, only incremental advances in absorption solvents and processes are likely to ensue. Adsorption and membrane technologies both require process development along with material development, and will require more development than absorption. Hybrid systems are largely unexplored. A final and critical point is that in our analysis, the current landscape of CO2 capture technology development involves
POLICY ANALYSIS
three groups working largely independently of each other: chemists who design and synthesize appropriate separation materials, process engineers who can design separation processes around those materials, and power plant engineers who can integrate the process into a power plant. We strongly believe that the emergence of breakthrough technologies will require close interdisciplinary collaboration among these groups, who further need to have a technical understanding of how each depends on the other.
’ ASSOCIATED CONTENT
bS
Supporting Information. Full, expanded tables and figures that have been cited in the text. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT We thank Dr. George Offen for his feedback and comments on this manuscript, Adam Berger for preparation of key figures, and Katarina Fustar for preparation of the manuscript. We also acknowledge and thank the reviewers for the very helpful and thorough feedback on the manuscript. ’ REFERENCES (1) Center for Global Development, Science Daily, November 15, 2007; http://www.sciencedaily.com/releases/2007/11/071114163448.htm. (2) California’s Global Warming Solutions Act of 2006. Assembly Bill 32. (3) $45.3 Billion in US Coal-Fired Power Plants Cancelled in 2007; Resource Media, January 8, 2009. (4) U.S. Energy Information Administration (EIA). Electric Power Annual 2009; DOE/EIA-0348 (2009); Washington, DC, 2011. (5) U.S. Energy Information Agency (EIA). Emissions of Greenhouse Gases in the United States, 2010; http://www.eia.doe.gov/oiaf/ 1605/ggrpt/carbon.html (accessed June 25, 2011). (6) U.S. Energy Information Agency (EIA). Inventory of U.S. Greenhouse Gas Emissions and Sinks: 19902008; http://www.epa.gov/climatechange/emissions/usinventoryreport.html (accessed June 25, 2011). (7) Friedlingstein, P.; Houghton, R. A.; Marland, G.; Hackler, J.; Boden, T. A.; Conway, T. J.; Canadell, J. G.; Raupach, M. R.; Ciais, P.; Le Quere, C. Update on CO2 Emissions. Nat. Geosci. 2010, 3, 811–812. (8) Holton, S.; Yonekawa, S.; Irvin, N. MHI’s KM-CDR PostCombustion CO2 Capture Process Overview and Coal-Fired 500 ton per day Demonstration Project Update, Tenth Annual Carbon Capture & Sequestration Conference, Pittsburgh, PA, May 4, 2011. (9) Metz, B., Davidson, O., Coninck, H. C. d., Loos, M., Meyer, L. A., Eds. IPCC Special Report on Carbon Dioxide Capture and Storage; Prepared by Working Group III of the Intergovenmental Panel on Climate Change; Cambridge University Press: Cambridge, UK, 2005. (10) Carbon Dioxide Capture from Existing Coal-Fired Power Plants; DOE/NETL-401/110907; National Energy Technology Laboratory, November 2007. (11) Finkeenrath, M. Cost and Performance of Carbon Dioxide Capture from Power Generation; International Energy Agency, 2011. (12) Ciferno, J. Overview of DOE/NETL CO2 Capture R&D Program, Presentation at the Annual NETL CO2 Capture Technology for Existing Plants R&D Meeting, Pittsburgh, PA, March 2009. 8631
dx.doi.org/10.1021/es104291d |Environ. Sci. Technol. 2011, 45, 8624–8632
Environmental Science & Technology (13) Booras, G.; Holt, N.; Marasigan, J. Updated Cost and Performance Estimates for Clean Coal Technologies Including CO2 Capture; Product ID 1014223; EPRI: Palo Alto, CA, 2008. (14) Freeman, B.; Rhudy, R. Assessment of CO2 Capture Options Currently Under Development; PID 1012796; EPRI: Palo Alto, CA, 2007. (15) Bhown, A. S.; Freeman, B. Program on Technology Innovation: Post-Combustion CO2 Capture Technology Development; PID 1016995; EPRI: Palo Alto, CA, 2008. (16) Bhown, A. S.; Freeman, B. Post-Combustion CO2 Capture Technology Development; PID 1017644; EPRI: Palo Alto, CA, 2009. (17) Bhown, A. S.; Freeman, B. Post-Combustion CO2 Capture Technology Development; PID 1019812; EPRI: Palo Alto, CA, 2010. (18) U.S. Department of Energy, Energy Information Administration (EIA). 2010 November EIA-923 Monthly Time Series File, Sources EIA-923 and EIA-860; http://www.eia.gov/cneaf/electricity/page/ eia906_920.html (accessed June 25, 2011). (19) Guide to the Business of Chemistry; American Chemistry Council: Washington, DC, 2003. (20) American Chemistry Council. Production Index, 2008, 2009, 2010. (21) C&EN Staff. Gains In Chemical Output Level Off. Chem. Eng. News 2008, 86, 61–70. (22) Worell, E.; Phylipsen, D.; Einstein, D.; Martin, N. Energy Use and Energy Intensity of the U.S. Chemical Industry; Lawrence Berkeley National Laboratory: Berkeley, CA, 2010; http://escholarship.org/uc/ item/2925w8g6;jsessionid=8B5AC06A6916F20CA2C9D4F3894CB980#page-1 (accessed June 25, 2011). (23) Role of Enhanced Oil Recovery in Accelerating the Deployment of Carbon Capture and Sequestration, A Massachusetts Institute of Technology Energy Initiative and Bureau of Economic Geology at the University of Texas at Austin Symposium, (Austin), TX, July 23, 2010. (24) McGlashan, N. R.; Marquis, A. J. Availability Analysis of PostCombustion Carbon Capture Systems: Minimum Work Input. Proc. Inst. Mech. Eng., Part C 2007, 221, 1057–1065. (25) Koopman, A.; Bahr, D. The Impact of CO2 Compressor Characteristics and Integration in Post Combustion Carbon Capture Carbon Sequestration Comparative Economic Analysis. In Proceedings of ASME Turbo Expo 2010, June 1418, 2010, Glasgow, UK. (26) Rao, Y. V. C. Postulational and Statistical Thermodynamics; Allied Publishers: New Delhi, 1994. (27) Rubin, E.; Chen, C.; Rao, A. Cost and Performance of Fossil Fuel Power Plants with CO2 Capture and Storage. Energy Policy 2007, 35, 4444–4454. (28) Dillon, D.; Chu, R.; Choi, G.; Harper, W.; Sugita, S. An Engineering and Economic Assessment of Post-Combustion CO2 Capture for Ultra-Supercritical Pulverized Coal Power Plant Applications, Phase II Task 3 Final Report; PID 1017515; EPRI: Palo Alto, CA, 2010. (29) Cost and Performance Baseline for Fossil Energy Plants. Volume 1: Bituminous Coal and Natural Gas to Electricity, Revision 2; DOE/2010/ 1397; National Energy Technology Laboratory, 2010. (30) Mankins, J. Technology Readiness Levels: A White Paper; NASA, April 1995. (31) Nolte, W. Technology Readiness Level Calculator; Air Force Research Laboratory, October 2003. (32) U.S. Department of Energy. Technology Readiness Assessment Guide; DOE G 413.3-4; October 12, 2009. (33) Ericson, A. The Role of CCS in the Decarbonization of Power. Presentation at 8th Annual Conference on Carbon Capture and Sequestration, Pittsburgh, PA, May 5, 2009. (34) Simonson, T.; Parish W. A. CCS Demonstration Plant (FE0003311). Presentation at 2010 NETL CO2 Capture Technology Meeting, Pittsburgh, PA, September 2010. (35) Hakka, L.; Ouimet, M. Method of Recovery of CO2 from Gas Streams. US Patent 7,056,482, June 6, 2006. (36) Kozak, F.; Petig, A.; Morris, E.; Rhudy, R.; Thimsen, D. Chilled Ammonia Process for CO2 Capture. Energy Procedia 2009, 1, 1419–1426. (37) Freeman, S.; Dugas, R.; Van Wagener, D. H.; Nguyen, T.; Rochelle, G. T. Carbon Dioxide Capture with Concentrated, Aqueous Piperazine. Int. J. Greenhouse Gas Control 2010, 4 (2), 119–124.
POLICY ANALYSIS
(38) Voyer, N.; Fradette, S. An Efficient Technology Platform for the Capture and Regeneration of CO2 Exploiting the Catalytic Power of Carbonic Anhydrase. Seventh Annual Conference on Carbon Capture & Sequestration, Pittsburgh, PA, May 58, 2008. (39) Bara, J.; Carlisle, T. K.; Gabriel, C. J.; Camper, D.; Finotello, A.; Gin, D. L.; Noble, R. D. Guide to CO2 Separations in ImidazoliumBased Room Temperature Ionic Liquids. Ind. Eng. Chem. Res. 2009, 48 (6), 2739–2751. (40) Perry, R.; Grocela-Rocha, T.; O’Brien, M.; Genovese, S.; Wood, B.; Soloveichik, G.; Rubinsztajn, M.; Kniajanski, S.; Draper, S.; Enick, R.; Johnson, J. K.; Xie, H.; Tapriyal, D. Aminosilicone Solvents for CO2 Capture. ChemSusChem 2010, 3 (8), 919–930. (41) Sjostrom, S.; Krutka, H. Evaluation of Solid Sorbents as a Retrofit Technology for CO2 Capture. Fuel 2010, 89 (6), 1298–1306. (42) Radosz, M.; Hu, X.; Krutkramelis, K.; Shen, Y. Flue Gas Carbon Capture on Carbonaceous Sorbents: Toward a Low-Cost Multifunctional Carbon Filter for Green Energy Producers. Indus. Eng. Chem. Res. 2008, 47 (10), 3783–3794. (43) Deng, H.; Doonan, C.; Furukawa, H.; Ferreira, R.; Towne, J.; Knobler, C.; Wang, B.; Yaghi, O. Multiple Functional Groups of Varying Ratios in Metal-Organic Frameworks. Science 2010, 327 (5967), 846–850. (44) Demessence, A.; D’Alessandro, D.; Foo, M.; Long, J. Strong CO2 Binding in a Water-Stable, Triazolae-Bridged Metal-Organic Framework Functionalized with Ethylenediamine. J. Am. Chem. Soc. 2009, 131 (25), 8784–8786. (45) Keskin, S.; van Heest, T.; Sholl, D. Can Metal-organic Framework Materials Play a Useful Role in Large-Scale Carbon Dioxide Separations? ChemSusChem 2010, 3 (8), 879–891. (46) Liang, Y.; Harrison, P; Gupta, R. P.; Green, A.; McMichael, W. J. Carbon Dioxide Capture Using Dry Sodium-Based Sorbents. Energy Fuels 2004, 18 (2), 569–575. (47) Gupta, H.; Fan, L. S. Carbonation-Calcination Cycle Using High reactivity Calcium Oxide for Carbon Dioxide Separation from Flue Gas. Ind. Eng. Chem. Res. 2002, 41 (16), 4035–4042. (48) Baker, R. W.; Lokhandwala, K. Natural Gas Processing with Membranes: An Overview. Ind. Eng. Chem. Res. 2008, 47 (7), 2109–212. (49) Pakala, N.; Bhown, A. Effect of Particulate Matter on Mass Transfer through Microporous Hollow Fiber Membranes. J. Membr. Sci. 1996, 111 (1), 71–79. (50) Merkel, T.; Lin, H.; Wei, X.; Baker, R. Power Plant PostCombustion Carbon Dioxide Capture: An Opportunity for Membranes. J. Membr. Sci. 2010, 359 (12), 126–139.
8632
dx.doi.org/10.1021/es104291d |Environ. Sci. Technol. 2011, 45, 8624–8632
POLICY ANALYSIS pubs.acs.org/est
Alternative “Global Warming” Metrics in Life Cycle Assessment: A Case Study with Existing Transportation Data Glen P. Peters,*,† Borgar Aamaas,† Marianne T. Lund,† Christian Solli,‡ and Jan S. Fuglestvedt† † ‡
Center for International Climate and Environmental Research Oslo (CICERO), PB 1129 Blindern, 0318 Oslo, Norway Environmental Systems Analysis (Miljøsystemanalyse, MiSA), Beddingen 14, 7014 Trondheim, Norway
bS Supporting Information ABSTRACT: The Life Cycle Assessment (LCA) impact category “global warming” compares emissions of long-lived greenhouse gases (LLGHGs) using Global Warming Potential (GWP) with a 100-year time-horizon as specified in the Kyoto Protocol. Two weaknesses of this approach are (1) the exclusion of shortlived climate forcers (SLCFs) and biophysical factors despite their established importance, and (2) the use of a particular emission metric (GWP) with a choice of specific time-horizons (20, 100, and 500 years). The GWP and the three time-horizons were based on an illustrative example with value judgments and vague interpretations. Here we illustrate, using LCA data of the transportation sector, the importance of SLCFs relative to LLGHGs, different emission metrics, and different treatments of time. We find that both the inclusion of SLCFs and the choice of emission metric can alter results and thereby change mitigation priorities. The explicit inclusion of time, both for emissions and impacts, can remove value-laden assumptions and provide additional information for impact assessments. We believe that our results show that a debate is needed in the LCA community on the impact category “global warming” covering which emissions to include, the emission metric(s) to use, and the treatment of time.
1. INTRODUCTION The Kyoto Protocol is directed toward the mitigation of the long-lived greenhouse gases (LLGHGs) covering CO2, CH4, N2O, and the fluorinated gases.1 It is well recognized, however, that climate is affected by more than the LLGHGs.2 Both warming and cooling of climate are caused by short-lived climate forcers (SLCFs) such as ozone, black carbon, and sulfate2 and biophysical factors such as land surface change.3 Mitigation studies that focus only on the LLGHGs may lead to different mitigation choices compared to when SLCFs2,48 and biophysical factors3,911 are included. The Kyoto Protocol additionally compares emissions using a particular metric with a specific treatment of time.1 Many studies show that mitigation choices can differ depending on the climate impact considered and the treatment of time.4,1218 Our aim in this paper is to demonstrate how the results of a Life Cycle Assessment (LCA) can differ when SLCFs, different emission metrics, and different treatments of time are used. A challenge with multicomponent impact assessment methods is the need to make different emissions comparable.13,17,19 Key challenges are the end-point of the metric (emissions, concentration, radiative forcing, temperature, damages, etc.) and the treatment of time. The standard practice in the LCA impact category “global warming” is to compare emissions using Global Warming Potential (GWP) with a fixed time-horizon independent of when the emissions occur. Within the LCA community, there has been little documented reflection on the use of different end-points20 or on different treatments of time,21 though time is well recognized as r 2011 American Chemical Society
an issue in biofuel discussions.2224 Several recent non-LCA studies have included SLCFs with a variety of emission metrics, particularly in the transport sector.17,2530 One study found, for example, that using GWPs international shipping causes a cooling effect for 300 years,25 but in terms of temperature change the cooling persists for less than 40 years.26 Among the LLGHGs, CH4 can be considered as a SLCFs and consequently a change in metric can have a significant effect on its relative weighting with CO2;31 Brazil has suggested the GWP should be replaced in climate policy.32 The growing literature on alternative emission metrics suggests there is a case to reconsider the LCA impact category “global warming”, particularly, what emissions to include, which end-points to use, and how to treat time. LCA typically utilizes the 100-year Global Warming Potential (GWP100)1 to compare the LLGHGs, and SLCFs and biophysical factors are excluded from the impact assessment. Some impact assessment methods apply different time-horizons.33 As stated in the IPCC’s First Assessment Report, the GWP and the three depicted time-horizons (20, 100, and 500 years) are illustrative examples,34 are somewhat arbitrary,34,35 and have vague interpretations.13,34,35 The GWP does not have a clear physical interpretation in terms of climate impact36 and is believed to be inconsistent with the wording Received: February 23, 2011 Accepted: September 2, 2011 Revised: May 31, 2011 Published: September 21, 2011 8633
dx.doi.org/10.1021/es200627s | Environ. Sci. Technol. 2011, 45, 8633–8641
Environmental Science & Technology
POLICY ANALYSIS
Figure 1. Comparison of LLGHGs versus SLCFs for different emission metrics. The CO2-eq in each set of bars are different as they depend on the metric. “Synthetic” represents the mainly halogenated hydrocarbons in the Kyoto and Montreal Protocols.
of the UNFCCC.37,38 Its application in climate policy could be seen as an “inadvertent consensus” of the scientific community.35 The continued use of the GWP100 in LCA probably reflects its perceived policy acceptance, as many scientific studies highlight weaknesses of the GWP concept.12,13,17,19,35,36,3942 Even though the selection of a metric necessarily requires value judgments,13,19 it is arguably desirable to minimize the number of value judgments and to make the metric sufficiently consistent with policy goals. The treatment of time is a key issue in the evaluation of mitigation options and consequently the development of emission metrics. Two aspects of time are important: (1) the selection of the time-horizon over which to compare climate impacts, and (2) the time at which emissions and climate impacts occur. First, the timehorizon can be used as the time over which to compare climate impacts usually by converting the emissions to the “equivalent” emission of a reference gas, usually CO2. In the particular case of the GWP, different time-horizons additionally represent different climate impacts;34 short time-horizons relate to rates of climate change, while long time-horizons may relate to sea level change.34 Second, and distinct from the time-horizon, is when the emissions occur and when the climate impacts are evaluated;4 for example, the emissions occur from 1990 to 2010 and the climate impacts are evaluated in 2050 and 2100. The timing of emissions and climate impacts are particularly important for prioritizing mitigation options.12,1416,18 Consequently, mitigation studies generally consider emission scenarios and impacts as a function of time. This use of time is uncommon in LCA, but LCA studies of biofuels have shown that the explicit inclusion of time makes biofuels less favorable to gasoline.2124 Given the importance of time for impact assessment and the relatively rudimentary treatment of time for “global warming” in LCA, there is arguably scope to reassess the treatment of time in LCA. Our aim in this paper is to invigorate debate in the LCA community on the impact category “global warming” by illustrating
with the transport sector the importance of SLCFs, alternative emission metrics, and different treatments of time. We begin by giving a brief overview of the most common emission metrics, before presenting and discussing LCA results of different transport modes for “global warming” using a variety of emission metrics with different treatments of time.
2. EMISSION METRICS The field of emission metrics has existed since the early 1990s with several reviews available.13,17,19,35,36 The purpose of an emission metric is to compare the climate impact of different emissions.13 A simple comparison by weight (e.g., kilograms) does not consider that the components have substantially different radiative efficiencies and atmospheric lifetimes.2,13 A general formulation of an emission metric can be expressed as2,43 Z ∞ AMi ¼ ½ðIðΔCrþi ðtÞÞ IðΔCr ðtÞÞÞgðtÞdt ð1Þ 0
where I(ΔCi(t)) is a function describing the “impact” of a change in climate, ΔC, at time t, with a discount function, g(t), and compared to a reference system, r, on which the perturbation occurs, i. The discount function need not be an exponential discounting, but may be used to represent a fixed time-horizon using a step-function (such as in most integrated metrics) or instantaneous evaluation using a Dirac delta function (such as in most end-point metrics). Metrics can be compared for different perturbations, AMi and AMj, or an emission index can be derived by normalizing with a reference component (e.g., Mi = AMi/AMj). 2.1. Absolute Metrics. 2.1.1. Radiative Forcing (RF). Radiative forcing is often used to compare the radiative imbalance of the Earth between two fixed times.2 It is also possible to consider the radiative forcing, RFi, for a time evolving emission, Ei(t). In this case, RFi must consider the atmospheric residence time of the different emissions 8634
dx.doi.org/10.1021/es200627s |Environ. Sci. Technol. 2011, 45, 8633–8641
Environmental Science & Technology
POLICY ANALYSIS
Figure 2. Radiative forcing for emissions occurring over the life-cycle. “Synthetic” represents the mainly halogenated hydrocarbons in the Kyoto and Montreal Protocols.
and this can be expressed with a convolution Z t RFi ðtÞ ¼ ai Ei ðsÞR r, i ðt sÞds
ð2Þ
0
where ai is the radiative efficiency (radiative forcing per unit mass) and Ri is the Impulse Response Function for a unit pulse of component i into the reference atmosphere Rr, i ðtÞ ¼
K
∑ bk expðt=τkÞ
k¼1
ð3Þ
and represents the fraction of atmospheric component i that remains in the atmosphere after time t relative to the reference system, r.4446 Most components have a single decay time (K = 1), though CO2 decays over multiple time-scales and has a component remaining in the atmosphere indefinitely.2 The lifetimes and radiative efficiencies used in the IPCC assessment reports are based on a reference system with constant current atmospheric conditions,2 though other choices exist.45 In the context of eq 1, the impact I is the radiative forcing at the top of the atmosphere (RFi) and the metric is evaluated at the end-point which removes the integral. 2.1.2. Integrated Radiative Forcing (iRF). The iRF is the timeintegrated RFi Z t iRFi ðtÞ ¼ RFi ðsÞds ð4Þ 0
With reference to eq 1, the impact I is radiative forcing and the discount function g is a step function ( 1 t e TH gðtÞ ¼ ð5Þ 0 t > TH where TH is the time-horizon. The iRF is rarely used as an absolute emission metric and is usually normalized.
2.1.3. Global Temperature Change (ΔT). The Global Temperature Change, ΔT, is an end-point metric and includes a climate model at the expense of increased uncertainty,16,31,42 Z t ΔTi ðtÞ ¼ R Fi ðsÞRT ðt sÞds ð6Þ 0
where RT is the Impulse Response Function to an instantaneous unit pulse of radiative forcing [c.f., 47]. As for RF, ΔT is an endpoint concept removing the integral in eq 1. As a way of analogy, RT mimics a physical discount function: RT makes the metric “forget” the forcing at early times as energy moves into the deep ocean, while in effect RT = 1 for the iRF and hence iRF “remembers” the forcing at early times. 2.2. Normalized Metrics. It is common to use a normalized metric that converts a component i into the equivalent climate impact of a reference gas, usually CO2, Mi ðTHÞ ¼
AMi ðTHÞ AMCO2 ðTHÞ
ð7Þ
where AM represents iRF or ΔT and leads to the Global Warming Potential (GWP) or Global Temperature change Potential (GTP). The normalized metrics, Mi, are evaluated for a given timehorizon, TH, representing the time over which to compare the climate impact of the two pulses.34 Multiplying a normalized metric with the emission of component i converts Ei into the equivalent emissions of CO2 (Mi Ei) that ideally would lead to the same climate impact (CO2-equivalent emissions). The GWP is by far the most common metric and is used in the Kyoto Protocol,1 however there have been many critiques of the GWP concept.12,13,17,31,35,36,40 The GTP is the next most discussed emission metric.16,17,30,31,39,4749 2.3. Other Metrics. Many other metrics exist, though we will not elaborate on them here.19 We have considered metrics which 8635
dx.doi.org/10.1021/es200627s |Environ. Sci. Technol. 2011, 45, 8633–8641
Environmental Science & Technology move along the causeeffect chain (ΔCf RF f ΔT). Integrated Global Temperature Change (iΔT) could be considered to be a physical metric closer to damages, but we find that it is numerically similar to the GWP in normalized form (see Supporting Information), and hence, we only present results for the GWP. In addition to the common pulse-based metrics, it is also possible to have metrics based on sustained emissions.31
3. TREATMENT OF TIME There are two main treatments of time in relation to metrics. Absolute metrics compare the climate impact of different emissions over time, whereas normalized metrics compare the climate impact relative to a reference gas. 3.1. Time in an Absolute Metric. The metrics RF, iRF, ΔT, and iΔT can be compared for emission profiles as a function of time allowing a comparison of the climate impact in absolute quantities (RF in W/m2, iRF in W/m2yr, ΔT in K, and iΔT in Kyr). This approach is used, for example, to compare historic emissions or emission scenarios in terms of a given climate impact.4 For emission scenarios, it is necessary to know the start and end time of the emissions and the evaluation time (TE) to quantify the climate impact;4 for example, emissions starting in 2010 and finishing in 2050, with the climate impact evaluated in 2100. This use of time is rare in LCA, but common in attribution and mitigation studies. 3.2. Time in a Normalized Metric. The GWP, GTP, and iGTP are emission indices that convert the emission of a given component Ei into the CO2-equivalent emission for the given time-horizon TH (CO2-eq = Mi Ei). The need for a timehorizon is partly due to the long-term properties of CO2.36 After an initial release into the atmosphere, about 20% of the CO2 remains permanently in the atmosphere while other components decay to zero.50 Thus, as time progresses CO2 will always dominate over other species requiring a value-based cutoff (time-horizon) and this was one of the key challenges developing emission metrics for climate.36 Together with the time-horizon, a normalized metric allows an emission of one component to be compared with that of the reference gas. The existence of the GWP may be the reason that the Kyoto Protocol is a multigas treaty.35 It is less clear, however, that normalized metrics are as relevant in applications like LCA; emissions scenarios, for example, use absolute metrics. The TH is usually fixed, but it has been suggested that the TH could be a function of time giving more weight to SLCFs as a target year is approached.12,16,31.34 There is no reason why the same TH should be used for different metrics as different metrics measure different climate impacts which may operate on different time scales [c.f., 17]. 4. DATA AND METHODS We focus on four transportation modes from the Ecoinvent LCA database:51 (1) regular diesel car based on the 2010 European fleet average (Ecoinvent: “Transport, passenger car, diesel, fleet average 2010/RER U”), (2) regular bus (Ecoinvent: “Transport, regular bus/CH U”), (3) long-distance train with Swiss electricity mix (Ecoinvent: “Transport, long-distance train, SBB mix/CH U”), and (4) passenger aircraft traveling in Europe (Ecoinvent: “Transport, aircraft, passenger, Europe/RER U”). We chose these four transport modes to highlight different mixes of LLGHGs and SLCFs in the life-cycle. The diesel car emits more SLCFs in operation compared to a gasoline car; the bus has
POLICY ANALYSIS
a higher efficiency per person kilometer; the train has virtually no operation emissions due to the Swiss electricity mix; the aircraft has significant SLCFs during operation. The emissions were taken directly from the EcoInvent database, with the exception of black carbon (BC) and organic carbon (OC) which required further processing (see Supporting Information). We consider different methods of distributing emissions along the life-cycle. First, we assume all emissions occur in year one as a pulse (equivalent to emissions over time with a fixed time-horizon) as is conventional in LCA. Second, we assume the emissions occur according to the life-cycle of the technologies (i.e., construction in year one, operation and maintenance distributed over the life of the transport vehicle, and disposal in the final year). The operation period for cars and buses is taken as 10 years, planes 30 years, and trains 40 years. We calculate the GWP and GTP values analytically using radiative efficiencies and lifetimes from previous studies.2,17 The ΔT values for emissions scenarios are calculated using convolutions between the emissions and analytical metric expressions (see the Supporting Information).
5. RESULTS The results focus on various aspects of LLGHGs, SLCFs, emission metrics, and the treatment of time in LCA using the four transportation modes. Our aim is to highlight how results change relative to standard LCA and not to analyze the transportation results in a policy context. We use 20 and 100 year time-horizons for GWP and 20 and 50 years for GTP [c.f., 17]. The GWP and GTP represent different climate impacts, and there is no reason that different climate impacts should use the same time-horizon. For the GTP, we use 50 years instead of 100 years based on estimates of when business-as-usual scenarios are likely to exceed the two-degree policy target.52 5.1. Importance of LLGHGs versus SLCFs for Different Emission Metrics. Figure 1 shows the CO2-equivalent emissions
for LLGHGs and SLCFs for the GWP and GTP. A standard LCA uses GWP100 including only the LLGHGs in the Kyoto Protocol (CO2, CH4, N2O, and the fluorinated gases). In the case of car, bus, and train, there is a mix of SLCFs, some of which are cooling and some of which are warming. These occur in construction, operation, and disposal. The case of aircraft is quite different and LCA results vary depending on the metric. Emissions from aircraft operation leads to the formation of contrails and cirrus clouds, which have large but uncertain radiative efficiencies.53 The short lifetime of the contrails and cirrus means the effect is short-lived, as can be seen by the large CO2-eq values for GWP20 and small values for GTP50. Overall for the car, bus, and train, the choice of metric does not significantly change the results of a standard LCA because LLGHGs dominate. However, for the case of aircraft the results are highly metric dependent, as is the case for shipping.26,54 The Supporting Information shows Figure 1 with the emissions allocated to life-cycle stages. 5.2. Treatment of Time in LCA. Generally, LCA assumes the same fixed TH independent of when the emissions occur (i.e., if the emissions occur in year t, the TH is the same as if the emissions occurred in year t + Δt); this is the same as assuming all emissions occur in the first year. When including both LLGHGs and SLCFs, the mix of emissions may be different depending on the life-stage and thus the timing of emissions may become more important for the net climate impact. We discuss two aspects of time that could be used in LCA: first, emissions as a function of 8636
dx.doi.org/10.1021/es200627s |Environ. Sci. Technol. 2011, 45, 8633–8641
Environmental Science & Technology
POLICY ANALYSIS
Figure 3. Values of ΔT for emissions occurring over the life-cycle. “Synthetic” represents the mainly halogenated hydrocarbons in the Kyoto and Montreal Protocols.
time using absolute metrics, and second, normalized metrics that convert emissions into CO2-eq. 5.2.1. Emissions as a Function of Time. Figure 2 shows the radiative forcing (RF) as the emissions occur over the life-cycle. The use of the RF allows a comparison of each component in the same units (W/m2) and gives an indication of when the emissions occur in the life cycle; the construction, operation, and disposal phases are clearly shown. Abrupt changes can be seen for the construction and disposal phases due to emissions of SLCFs with large radiative efficiencies but short atmospheric residence times. In terms of total contributions to the climate impact, the construction and disposal phases are small as the total emissions are relatively small and the SLCFs decay quickly (except for the case of the train which has low emissions in operation, Supporting Information). In the operation phase, the SLCFs give a near constant forcing, as the emissions decay rapidly, but are quickly replaced converging to a constant concentration. The increasing components in the use phase relate to the LLGHGs which accumulate in the atmosphere due to slow decay rates and then persist in the atmosphere after the lifetime of the transport mode. Overall, the use of radiative forcing as a metric allows the different components to be compared in the same unit providing additional insight on the importance of timing and of construction and disposal. Figure 3 shows the ΔT for the transportation modes, where it is seen that the SLCFs are quickly removed from the atmosphere after the emissions cease, leading to a decreased climate impact. In all transportation modes, the total net ΔT converges to the ΔT of CO2 only as the SLCFs are progressively forgotten and the other LLGHGs only have small contributions. For example, in the case of aircraft the SLCFs dominate the net ΔT up until the end of the life of the aircraft, in which case CO2 takes over as the dominant component due to its long atmospheric lifetime. This
is distinct to the case of the iRF (Supporting Information), where the SLCFs are relatively more important as the iRF integration remembers the emissions at earlier times (compare with Figure 1). It is standard in LCA to have a fixed time-horizon which implies that all the emissions occur in year one, Figure 4. Comparing Figure 3 (scenario) and Figure 4 (pulse) shows quite different treatments of time (c.f., 21). The peak temperature (ΔT) is larger and appears earlier in the pulse since all the emissions occur at once, while the scenario distributes the emissions over time. The scenario also shows the different life-cycle phases clearly. The explicit representation of time shows the relationship between the timing of emissions and impacts, and provides more information for the end-user of the results. 5.2.2. CO2-Equivalent Emissions Using the Time-Horizon. The treatment of time in the scenario, Figure 3, is quite different from standard LCA. In Figure 3, the emissions occur as a function of time, but the evaluation is also a function of time (TE, time of evaluation). If TE is chosen as 100 years, for example, then the emissions in year 20 are still evaluated at 100 years, which means the emissions are only resident in the atmosphere for 80 years. If the emissions are to be evaluated after being resident in the atmosphere for the same amount of time (analogous to a fixed time-horizon), then they are treated as a hypothetical pulse emission (Figure 4). It is possible to have a variable time-horizon (analogous to Figure 3), TH = TE t, which means that the relative impact of different emissions varies over time.12,16,18 The consequence of this is that the SLCFs will have a larger effect on temperature relative to CO2 as the time t approaches TE as TH gets progressively smaller.12,16,18 Figure 5 shows the CO2-eq emissions for the different transport modes using a fixed time-horizon and a variable time-horizon (the normalized metric versions of Figure 3 and Figure 4). We took TE = 50 to highlight the differences as it is closer to the 8637
dx.doi.org/10.1021/es200627s |Environ. Sci. Technol. 2011, 45, 8633–8641
Environmental Science & Technology
POLICY ANALYSIS
Figure 4. Values of ΔT for emissions occurring as a pulse for the entire life-cycle. “Synthetic” represents the mainly halogenated hydrocarbons in the Kyoto and Montreal Protocols.
Figure 5. CO2-eq emissions using GWPs and GTPs for emissions occurring as a function of time using a fixed TH and a variable TH. The TE of 50 years was taken as it is closer to the lifetime of the transportation modes and hence highlights the differences.
lifetime of the transportation modes. In the case of the car and to a lesser extend the bus, the difference between the lifetime of the transport mode and TE is 40 years, which is much longer than the
lifetime of the SLCFs and hence the differences between a fixed and variable TH are small. In the case of the train (disposal phase only) and air transport, the difference between the lifetime of the 8638
dx.doi.org/10.1021/es200627s |Environ. Sci. Technol. 2011, 45, 8633–8641
Environmental Science & Technology transport mode and the TE (10 years for a train, and 20 years for a plane) is closer to the lifetime of the SLCFs (but still significantly different). In these cases, the variable TH makes the emissions of the SLCFs relatively more important as t approaches TE (since TH gets smaller). A difference between the results for a fixed and variable TH will occur if (a) SLCFs are important, and (b) the SLCFs are emitted sufficiently close to TE. Given the only modest variations shown here, it is arguable whether the extra complexity of the variable time-horizon is justified.
6. DISCUSSION Our results show that SLCFs are important, particularly for short time-horizons, when the emissions occur close to the evaluation times, or for integrated metrics relative to end-point metrics. The importance of SLCFs and time will also depend on the product under investigation; in our examples, SLCFs are important in all transport modes, but particularly important for aircraft. Thus, it is important to include SLCFs in impact assessments to avoid potentially misleading results.48,17,2527,29 If SLCFs are included in impact assessments of “global warming”, then the choice of metric and treatment of time becomes particularly relevant.13,16,17,19,31,39,43 These results suggest that a more lively discussion of metrics for “global warming” is needed, particularly considering the latest research.17,19,39,55 Our results show that time can be more transparently included in assessments removing value-laden assumptions on the choice of time. Performing calculations depending on the time the emissions occur, “dynamic LCA”,21 provides more information and therefore creates valuable insights to impact assessment. In addition, showing how the climate impact for a given emissions scenario varies over time (RF, iRF, ΔT, iΔT), as opposed to converting to a CO2-eq emissions, may provide a different approach to time as conventionally used in LCA. Particularly if emissions are treated as a function of time, it is perhaps appropriate to compare climate impacts as a function of time as well, as is often done in climate studies.4,26,56,57 We only provide simple discussions and comparisons of emission metrics to highlight key issues in metric choice, but many other issues need to be considered (c.f., 19). Different emission metrics have different uncertainties that generally increase for end-point metrics (compared to integrated metrics) or when the climate impact requires additional modeling.42 Likewise, climate impacts due to SLCFs are usually more uncertain than LLGHGs2 and are usually regionally dependent.58 For long-term climate impacts, CO2 dominates and the impact may be irreversible,50,59 though SLCFs can have persistent climate impacts60 and both LLGHGs and SLCFs may induce feedbacks.55 Cumulative emissions of CO2 may dominate peak warming,52 but SLCFs can be particularly important for rates of change.61 These types of issues need to be considered when deciding which metric(s) to apply in different situations. We see a need for further discussions in the LCA community on the choice of emission metrics and treatment of time in the impact category “global warming”. There is a need to harmonize across impact assessments what components to include, covering LLGHGs, SLCFs, and biophysical factors. There is a need for further discussion on which emission metric(s) to use in different impact assessments, where the goal and scope may define different time-scales. Time is a cross-cutting theme with value laden aspects, but a more transparent representation of time can remove some of the value laden judgments. As highlighted
POLICY ANALYSIS
before, multidisciplinary research and discussion on emission metrics,35 for example between the LCA and climate communities, can build understanding and strengthen applications in a wide-range of relevant fields.
’ ASSOCIATED CONTENT
bS
Supporting Information. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT This research was partly funded by the Norwegian Research Council Project “Transport and Environment Measures and Policies (TEMPO)”. We thank Rolf Hagman (Institute of Transport Economics, Norway) for useful comments. ’ REFERENCES (1) UNFCCC. The Kyoto Protocol to the United Nations Convention on Climate Change; United Nations: New York, 1997. (2) Forster, P.; Ramaswamy, V.; Artaxo, P.; Berntsen, T.; Betts, R.; Fahey, D. W.; Haywood, J.; Lean, J.; Lowe, D. C.; Myhre, G.; Nganga, J.; Prinn, R.; Raga, G.; Schulz, M.; Dorland, R. V. Changes in Atmospheric Constituents and in Radiative Forcing. In Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change; Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K. B., Tignor, M., Miller, H. L., Eds. Cambridge University Press: Cambridge, U.K. and New York, 2007. (3) Jackson, R. B.; Randerson, J. T.; Canadell, J. G.; Anderson, R. G.; Avissar, R.; Baldocchi, D. B.; Bonan, G. B.; Caldiera, K.; Diffenbaugh, N. S.; Field, C. B.; Hungate, B. A.; Jobbagy, E. G.; Kueppers, L. M.; Nosetto, M. D.; Pataki, D. E. Protecting climate with forests. Environ. Res. Lett. 2008, 3, 044006. (4) den Elzen, M.; Fuglestvedt, J. S.; H€ohne, N.; Trudinger, C.; Lowe, J.; Matthews, B.; Romstad, B.; Pires de Campos, C.; Andronova, N. Analysing a countries’ contribution to climate change: Scientific and policy-related choices. Environ. Sci. Policy 2005, 8, 614–636. (5) Shindell, D.; Lamarque, H.-F.; Unger, N.; Koch, D.; Faluvegi, G.; Bauer, S.; Teich, H. Climate forcing and air quality change due to regional emissions reductions by economic sector. Atmos. Chem. Phys. 2008, 8, 7101–7113. (6) Unger, N.; Shindell, D. T.; Wang, J. S. Climate forcing by the onroad transportation and power generation sectors. Atmos. Environ. 2009, 43, 3077–3085. (7) Unger, N.; Bond, T. C.; Wang, J. S.; Koch, D. M.; Menon, S.; Shindell, D. T.; Bauer, S. Attribution of climate forcing to economic sectors. Proc. Natl. Acad. Sci., U.S.A. 2009, 107, 3382–3387. (8) Unger, N. Short-lived non-CO2 pollutants and climate policy: Fair trade? Environ. Sci. Technol. 2010, 44, 5332–5333. (9) Betts, R. A. Offset of the potential carbon sink from boreal forestation by decreases in surface albedo. Nature 2000, 408, 187–190. (10) Pielke, R.; Marland, G.; Betts, R. A.; Chase, T. N.; Eastman, J. L.; Niles, J. O.; Niyogi, D. D. S.; Running, S. W. The influence of landuse change and landscape dynamics on the climate system: Relevance to climate-change policy beyond the radiative effect of greenhouse gases. Philos. Trans. R. Soc., A 2002, 360, 1705–1719. (11) Bala, G.; Caldeira, K.; Wickett, M.; Phillips, T. J.; Lobell, D. B.; Delire, C.; Mirin, A. Combined climate and carbon-cycle effects of largescale deforestation. Proc. Natl. Acad. Sci., U.S.A. 2007, 104, 6550–6555. 8639
dx.doi.org/10.1021/es200627s |Environ. Sci. Technol. 2011, 45, 8633–8641
Environmental Science & Technology (12) Manne, A. S.; Richels, R. G. An alternative approach to establishing trade-offs among greenhouse gases. Nature 2001, 410 (6829), 675–677. (13) Fuglestvedt, J. S.; Berntsen, T. K.; Godal, O.; Sausen, R.; Shine, K. P.; Skodvin, T. Metrics of climate change: Assessing radiative forcing and emission indices. Climatic Change 2003, 58, 267–331. (14) O’Neill, B. C.; Oppenheimer, M. Climate change impacts are sensitive to the concentration stabilization path. Proc. Natl. Acad. Sci., U.S.A. 2004, 101 (47), 16411–16416. (15) Rive, N.; Torvanger, A.; Berntsen, T.; Kallbekken, S. To what extent can a long-term temperature target guide near-term climate change commitments? Climatic Change 2007, 82 (3), 373–391. (16) Shine, K. P.; Berntsen, T.; Fuglestvedt, J. S.; Stuber, N.; Skeie, R. B. Comparing the climate effect of emissions of short and long lived climate agents. Philos. Trans. R. Soc., A 2007, 365, 1903–1914. (17) Fuglestvedt, J. S.; Shine, K. P.; Berntsen, T.; Cook, J.; Lee, D. S.; Stenke, A.; Skeie, R. B.; Velders, G. J. M.; Waitz, I. A. Transport impacts on atmosphere and climate: Metrics. Atmos. Environ. 2010, 44, 4648–4677. (18) Berntsen, T.; Tanaka, K.; Fuglestvedt, J. Does black carbon abatement hamper CO2 abatement? Climatic Change 2010, 103 (3), 627–633. (19) Tanaka, K.; Peters, G. P.; Fuglestvedt, J. S. Multi-component climate policy: Why do emission metrics matter? Carbon Manage. 2010, 1, 191–197. (20) De Schryver, A. M.; Brakkee, K. W.; Goedkoop, M. J.; Huijbregts, M. A. J. Characterization Factors for Global Warming in Life Cycle Assessment Based on Damages to Humans and Ecosystems. Environ. Sci. Technol. 2009, 43 (6), 1689–1695. (21) Levasseur, A.; Lesage, P.; Margni, M.; Deschnes, L.; Samson, R. Considering Time in LCA: Dynamic LCA and Its Application to Global Warming Impact Assessments. Environ. Sci. Technol. 2010, 44, 3169–3174. (22) Kendall, A.; Chang, B.; Sharpe, B. Accounting for TimeDependent Effects in Biofuel Life Cycle Greenhouse Gas Emissions Calculations. Environ. Sci. Technol. 2009, 43, 7142–7147. (23) O’Hare, M.; Plevin, R. J.; Martin, J. I.; Jones, A. D.; Kendall, A.; Hopson, E. Proper accounting for time increases crop-based biofuels’ greenhouse gas deficit versus petroleum. Environ. Res. Lett. 2009, 4, 024001. (24) Cherubini, F.; Peters, G. P.; Berntsen, T.; Strømman, A. H.; Hertwich, E. CO2 emissions from biomass combustion for bioenergy: Atmospheric decay and contribution to global warming. GCB Bioenergy 2011, 3, 413–426. (25) Fuglestvedt, J.; Berntsen, T.; Myhre, G.; Rypdal, K.; Skeie, R. B. Climate forcing from the transport sectors. Proc. Natl. Acad. Sci., U.S.A. 2008, 105 (2), 454–458. (26) Berntsen, T.; Fuglestvedt, J. S. Global temperature responses to current emissions from the transport sectors. Proc. Natl. Acad. Sci., U.S.A. 2008, 105, 19154–19159. (27) Skeie, R. B.; Fuglestvedt, J. S.; Berntsen, T.; Lund, M. T.; Myhre, G.; Rypdal, K. Global temperature change from the transport sectors: Historical development and future scenarios. Atmos. Environ. 2009, 43, 6260–6270. (28) Sausen, R. Transport impacts on atmosphere and climate. Atmos. Environ. 2010, 44, 4646–4647. (29) Borken-Kleefeld, J.; Berntsen, T.; Fuglestvedt, J. Specific Climate Impact of Passenger and Freight Transport. Environ. Sci. Technol. 2010, 44, 5700–5706. (30) Myhre, G.; Shine, K. P.; R€adel, G.; Gauss, M.; Isaksen, I. S. A.; Tang, Q.; Prather, M. J.; Williams, J. E.; van Velthoven, P.; Dessens, O.; Koffi, B.; Szopa, S.; Hoor, P.; Grewe, V.; Borken-Kleefeld, J.; Berntsen, T. K.; Fuglestvedt, J. S. Radiative forcing due to changes in ozone and methane caused by the transport sector. Atmos. Environ. 2011, 45 (2), 387–394. (31) Shine, K. P.; Fuglestvedt, J. S.; Hailemariam, K.; Stuber, N. Alternatives to the global warming potential for comparing climate impacts of emissions of greenhouse gases. Climatic Change 2005, 68, 281–302.
POLICY ANALYSIS
(32) Brazil. Second National Communication of Brazil to the United Nations Framework Convention on Climate Change; Ministry of Science and Tecnology, General-Coordination on Global Climate Change: Brasilia, 2010. (33) Goedkoop, M. J.; Heijungs, R.; Huijbregts, M.; de Schryver, A.; Struijs, J.; van Zelm, R. ReCiPe 2008, A life cycle impact assessment method which comprises harmonised category indicators at the midpoint and the endpoint level; Report I: Characterisation; Ministry of Housing, Spatial Planning and Environment (VROM): The Hague, the Netherlands, 2009. (34) IPCC. Climate Change: The IPCC Scientific Assessment; Cambridge University Press: Cambridge, United Kingdom, 1990. (35) Shine, K. P. The global warming potential - The need for an interdisciplinary retrail. Climatic Change 2009, 96, 467–472. (36) O’Neill, B. C. The Jury is Still Out on Global Warming Potentials. Climatic Change 2000, 44 (4), 427–443. (37) Tol, R. S. J.; Berntsen, T. K.; O’Neil, B. C.; Fuglestvedt, J. S.; Shine, K. P.; Balkanski, Y.; Makra, L. Metrics for Aggregating the Climate Effect of Different Emissions: A Unifying Framework; Economic and Social Reseach Institute: Dublin, Ireland, 2008. (38) O’Neill, B. A Unifying Framework for Metrics for Aggregating the Climate Effect of Different Emissions. In IPCC Expert Meeting on the Science of Alternative Metrics, Plattner, G.-K.; Stocker, T. F.; Midgley, P.; Tignor, M., Eds. IPCC Working Group I Technical Support Unit: Bern, Switzerland, 2009. (39) IPCC. IPCC Expert Meeting on the Science of Alternative Metrics; IPCC Working Group I Technical Support Unit: Bern, Switzerland, 2009. (40) Johansson, D.; Persson, U.; Azar, C. The Cost of Using Global Warming Potentials: Analysing the Trade off Between CO2, CH4 and N2O. Climatic Change 2006, 77 (3), 291–309. (41) Manning, M.; Reisinger, A. Broader perspectives for comparing different greenhouse gases. Proc. R. Soc., A 2011, 369, 1891–1905. (42) Reisinger, A.; Meinshausen, M.; Manning, M.; Bodeker, G. Uncertainties of global warming metrics: CO2 and CH4. Geophys. Res. Lett. 2010, 37 (14), L14707. (43) Kandlikar, M. Indices for comparing greenhouse gas emissions: integrating science and economics. Energy Econ. 1996, 18, 265–281. (44) Maier-Reimer, E.; Hasselmann, K. Transport and storage of CO2 in the ocean - an inorganic ocean-circulation carbon cycle model. Climate Dynamics 1987, 2, 63–90. (45) Enting, I. G.; Wigley, T. M. L.; Heimann, M. Future Emissions and Concentrations of Carbon Dioxide: Key Ocean/Atmosphere/Land Analyses; Technical Paper no. 31; CSIRO Division of Atmospheric Research: Oslo, 1994. (46) Enting, I. G. Laplace transform analysis of the carbon cycle. Environ. Model. Softw. 2007, 22 (10), 1488–1497. (47) Boucher, O.; Reddy, M. S. Climate trade-off between black carbon and carbon dioxide emissions. Energy Policy 2008, 36, 193–200. (48) Boucher, O.; Friedlingstein, P.; Collins, B.; Shine, K. P. The indirect global warming potential and global temperature change potential due to methane oxidation. Environ. Res. Lett. 2009, 4, 044007. (49) Collins, W. J.; Sitch, S.; Boucher, O. How vegetation impacts affect climate metrics for ozone precursors. J. Geophys. Res. 2010, 115 (D23), D23308. (50) Solomon, S.; Plattner, G.-K.; Knutti, R.; Friedlingstein, P. Irreversible climate change due to carbon dioxide emissions. Proc. Natl. Acad. Sci., U.S.A. 2009, 106 (6), 1704–1709. (51) Spielmann, M.; Bauer, C.; Dones, R.; Tuchschmid, M. Transport Services; Swiss Centre for Life Cycle Inventories: D€ubendorf, 2007. (52) Meinshausen, M.; Meinshausen, N.; Hare, W.; Raper, S. C. B.; Frieler, K.; Knutti, R.; Frame, D. J.; Allen, M. R. Greenhouse-gas emission targets for limiting global warming to 2 °C. Nature 2009, 458 (7242), 1158–1162. (53) Lee, D. S.; Pitari, G.; Grewe, V.; Gierens, K.; Penner, J. E.; Petzold, A.; Prather, M. J.; Schumann, U.; Bais, A.; Berntsen, T.; Iachetti, D.; Lim, L. L.; Sausen, R. Transport impacts on atmosphere and climate: Aviation. Atmos. Environ. 2010, 44 (37), 4678–4734. 8640
dx.doi.org/10.1021/es200627s |Environ. Sci. Technol. 2011, 45, 8633–8641
Environmental Science & Technology
POLICY ANALYSIS
(54) Eyring, V.; Isaksen, I. S. A.; Berntsen, T.; Collins, W. J.; Corbett, J. J.; Endresen, O.; Grainger, R. G.; Moldanova, J.; Schlager, H.; Stevenson, D. S. Transport impacts on atmosphere and climate: Shipping. Atmos. Environ. 2010, 44 (37), 4735–4771. (55) Gillett, N. P.; Matthews, H. D. Accounting for carbon cycle feedbacks in a comparison of the global warming effects of greenhouse gases. Environ. Res. Lett. 2010, 5 (3), 034011. (56) H€ohne, N.; Blum, H.; Fuglestvedt, J. S.; Skeie, R. B.; Kurosawa, A.; Hu, G.; Lowe, J.; Gohar, L.; Matthews, B.; de Salles, A. C. N.; Ellermann, C. Contributions of individual countries’ emissions to climate change and their uncertainty. Climatic Change 2010, 106, 359–391. (57) Prather, M. J.; Penner, J.; Fuglestvedt, J. S.; Raper, S.; de Campos, C. P.; Jain, A.; van Aardenne, J.; Lal, M.; Wagner, F.; Kurosawa, A.; Skeie, R. B.; Lowe, J.; Stott, P.; H€ohne, N. Tracking uncertainties in the causal chain from human activities to climate. Geophys. Res. Lett. 2009, 36. (58) Berntsen, T.; Fuglestvedt, J. S.; Myhre, G.; Stordal, F.; Berglen, T. F. Abatement of greenhouse gases: Does location matter? Climatic Change 2006, 74, 377–411. (59) Fr€olicher, T.; Joos, F. Reversible and irreversible impacts of greenhouse gas emissions in multi-century projections with the NCAR global coupled carbon cycle-climate model. Climate Dynamics 2010, 35 (7), 1439–1459. (60) Solomon, S.; Daniel, J. S.; Sanford, T. J.; Murphy, D. M.; Plattner, G.-K.; Knutti, R.; Friedlingstein, P. Persistence of climate changes due to a range of greenhouse gases. Proc. Natl. Acad. Sci. 2010, 107 (43), 18354–18359. (61) Penner, J. E.; Prather, M. J.; Isaksen, I. S. A.; Fuglestvedt, J. S.; Klimont, Z.; Stevenson, D. S. Short-lived uncertainty? Nat. Geosci. 2010, 3 (9), 587–588.
8641
dx.doi.org/10.1021/es200627s |Environ. Sci. Technol. 2011, 45, 8633–8641
ARTICLE pubs.acs.org/est
A 3500-Year Record of Hg and Pb Contamination in a Mediterranean Sedimentary Archive (The Pierre Blanche Lagoon, France) F. Elbaz-Poulichet,*,† L. Dezileau,‡ R. Freydier,† D. Cossa,§ and P. Sabatier‡,|| †
)
Laboratoire Hydrosciences UMR 5569, CNRS, Universites Montpellier I & II, IRD, Place Eugene Bataillon, CC MSE, 34095 Montpellier cedex 5, France ‡ Laboratoire Geosciences Montpellier UMR 5243, CNRS, Universite Montpellier II, Place Eugene Bataillon, CC 60, 34095 Montpellier cedex 5, France § IFREMER, Centre de Mediterranee, BP 330, 83507 La Seyne-sur-mer, France Laboratoire des Environnements Dynamiques et des Territoires de Montagne UMR 5204, Universite de Savoie, CISM, Campus Scientifique, 73376 Le Bourget du Lac cedex, France
bS Supporting Information ABSTRACT: A sediment core encompassing 3500 years of continuous sedimentation has been collected from a coastal lagoon located on the southwestern French Mediterranean coast. Lead concentrations and stable isotopes show that the sediments have recorded the three major periods of Pb pollution: the Etruscan Greek Roman period (650 BC to AD 50), the medieval period (AD 650 to AD 1450), and the modern period (from around AD 1850 to the present). These periods were separated by low pollution periods during the Dark Ages (between AD 50 and 650) and during the 16th century. From the end of the 19th century to the 1960s, Pb pollution increased exponentially. Coal combustion was the major source of Pb in the lagoon in the second half of the 20th century. Both the decrease in coal consumption and the ban on leaded gasoline resulted in a decrease in Pb pollution by a factor of 1.5 between 1973 and 1995. From 1991, sewage treatment plants and incinerators could be the major source of Pb. The average baseline Hg concentration from 1525 BC to AD 900 was 0.017 ( 0.003 μg g 1 (n = 54). The Hg concentrations profile shows three major peaks: in AD 1150, AD 1660, and AD 1969, with the concentrations being respectively 8, 5, and 34 times higher than the baseline levels. The medieval peak (AD 1150) is attributed the medical use of Hg in the town of Montpellier and/or the burning of soil and vegetation. Noticeable Hg pollution was also detected during the 17th century in relation to gold and silver amalgamation in Europe. From the end of the 19th century, Hg concentrations increased exponentially until 1969. This modern pollution is attributed to the burning of coal.
’ INTRODUCTION The history of atmospheric Pb pollution for 3000 years in Europe is well established by studies of Greenland and Arctic ice cores,1,2 lake sediments,3,4 and peat deposits.5 7 The pollution reached a peak during the Roman period, about 2000 years ago. After a decline during the Dark Ages (AD 400 900), Pb deposition increased again during the Medieval period (around AD 1000). Beginning with the Industrial Revolution (from about AD 1850), the increase in atmospheric pollution was exponential until around AD 1970. Since then it has decreased. For mercury, several studies of peat bog and lake sediments from northwestern Spain,8 Greenland and Denmark,6 the Swiss Jura,9 and southern England10 date atmospheric Hg pollution back to the Middle Ages. As for Pb, the studies show an increase of atmospheric Hg during the last two centuries that is generally related to coal combustion. The influence of these long-term Pb and Hg emmissions on the pollution of the marine environment has not been closely studied, although this might contribute to a better understanding r 2011 American Chemical Society
of present-day oceanic metal cycling. These historical inputs are expected to be high in the Mediterranean basin which hosts some of the oldest and largest metal mines in the world. The Iberian Pyrite Belt (Spain)11 and the Hg mines of Almaden (Spain) were already exploited by the Romans. The history of Pb contamination in the Gulf of Lions over the last century has been reconstructed by Miralles et al.12 Lead contamination in ancient Mediterranean harbors during the Bronze Age and Roman period has been reported.13 In Alexandria harbor, anthropogenic Pb contamination was traced back to the Bronze Age (900 BC) and Roman period.14 Lastly, Greco-Roman Pb was identified in Marseille harbor.15 For Hg, data are scarcer than for Pb, and to our knowledge no long-term record of Hg contamination in marine environment has been published. Received: February 14, 2011 Accepted: August 31, 2011 Revised: August 30, 2011 Published: August 31, 2011 8642
dx.doi.org/10.1021/es2004599 | Environ. Sci. Technol. 2011, 45, 8642–8647
Environmental Science & Technology
ARTICLE
Figure 1. Map of the Pierre-Blanche Lagoon showing the location of the PB06 core.
To reconstruct the long-term history of Pb and Hg inputs into the Mediterranean coastal environment, the concentrations of these elements were measured in a sediment core collected in the Pierre Blanche lagoon (France). This lagoon is well suited to investigate the chronology of Pb and Hg inputs to sediments, because sufficient information is available on sedimentary conditions, rates of sedimentation, and paleostorm events.16,17 The sediment analyzed in the present study encompassed 3500 years of continuous sedimentation from 1500 BC to AD 2006. The isotopic composition of Pb was used to better constrain the sources of Pb contamination.
’ MATERIALS AND METHODS Description of the Pierre-Blanche Lagoon. The Pierre-Blanche (P-B) lagoon is part of the Palavasian lagoonal complex (Figure 1). It has a maximum water depth of approximately 1 m. The lagoon exchanges water with the Rhone-Sete Canal, built in the 18th century, and receives fresh water inputs from a 700 km2 drainage basin, which includes those of the Lez and Mosson rivers. These two rivers drain the urbanized area of Montpellier (380 000 inhabitants with suburbs) and are the main sources of sediments to the lagoon, together with the landward transport of sand and silt materials during storm events.17 The P-B lagoon suffers commonly from severe eutrophication.18 Complementary information on hydrology and soil occupation is given in the Supporting Information (SI). Sample Collection. A 7.9-m-long piston core (PB06) was collected in the P-B lagoon (Figure 1) in March 2006 using the UWITEC coring platform. For this study, the sediments between the surface and a depth of 4 m were subsampled. Chemical Analysis. Details for sample preparation and the chemical analysis are given in the SI. The Particulate Organic Carbon (POC) was analyzed according to ref 19. Aluminum, Pb
concentrations, and Pb isotopes were determined after total digestion of the sediments using an ICP-MS, X Series II (Thermo Fisher Scientific), equipped with a CCT (Collision Cell Technology) chamber. Certified reference material from the Canadian National Research Council, i.e., MESS-3 (marine sediment), was used to check analytical accuracy and precision. Measured concentrations agree with recommended values to within (5% (Al) and 2% (Pb). The precision was better than (4%. For Pb isotope ratios, the external precision was assessed using eight individual measurements of MESS-3 reference. The following values were obtained: 206 Pb/207Pb = 1.2353 ( 0.0014 (1σ), 208Pb/207Pb = 2.4964 ( 0.0024 (1σ), and 208Pb/206Pb = 2.0212 ( 0.0033 (1σ). Total Hg concentrations in the sediment were determined using an automatic mercury analyzer (AMA-254, Altec) according to Cossa et al.20 Chronology frAmework. The chronology of core PB06 is based on 137Cs and 210Pb depth profiles and AMS 14C dates. All details concerning the dating procedure, sedimentation rates, and fluxes for this core are given in Sabatier et al.16 and in the SI. The sedimentation rates varies between 0.051 mm yr 1 and 0.268 mm yr 1, with the highest values being measured between AD 1930 and AD 2006. Between AD 950 and AD 1000, the shift from a protected lagoonal environment to a closed lagoon changed the 14C reservoir age, resulting in a relatively poor time resolution (around 100 yr) during this period. At the sediment surface, the excess 210Pb profile indicated the presence of a 3.5-cm thick mixed layer related to bioturbation and resuspension. The time resolution was thus estimated at 10 yr for the last 100 years.
’ RESULTS Lead Concentrations. The lead concentration data are listed in Table S1. Natural Pb is usually associated with the fine-grained 8643
dx.doi.org/10.1021/es2004599 |Environ. Sci. Technol. 2011, 45, 8642–8647
Environmental Science & Technology
ARTICLE
Figure 2. Al, POC, Pb/Al, 206Pb/207Pb, and Hg concentrations in the sediment profile from the Pierre-Blanche lagoon. The dotted line is the local background concentration. The variations in coal production and coal consumption in France (data Charbonnage de France) and Pb emissions from leaded gasoline57 are also shown.
fraction mainly constituted of organic matter and clay mineral on the alumino silicate fraction, for which Al is a good tracer. The affinity of natural Pb for this fraction in the Pierre-Blanche lagoon is reflected by the good correlation (R = 0.92) between Al and Pb in deep sediments (depth <320 cm, date <690 BC). To correct Pb concentrations for the dilution of this fraction by shell debris, normalization to Al, a common practice in marine geochemical studies,21 was adopted. The average local Pb/Al background ratio (depth <320, n = 24) was equal to 3.7 10 8 ( 0.2 10 8. Higher Pb/Al ratios were generally observed between 310 and 250 cm depth (650 BC AD 60), 220 107 cm depth (AD 650 1450), and between 9 and 10 cm depth (AD 1960 1972). The highest Pb/Al ratios, 16.4 10 8 and 15.8 10 8, were found, respectively, in the 0.5-cm depth sample (AD 2004) and between 10 and 16 cm depth (AD 1959 1972). These values, higher by a factor of about 4 than the background ratio, are in the range of those reported in the surface sediments of the nearby Thau Lagoon.22 Lower Pb/Al peaks (5.8 10 8, 6.2 10 8, 6.8 10 8) were also found, respectively, at 150 cm (AD 1150), 205 cm (AD 800), and 274 cm depths (200 BC). Isotopic Composition of Pb. The distribution of the 206 Pb/207Pb ratios with depth mirror that of Pb/Al ratios, with the isotope ratios generally decreasing when the Pb/Al ratios increase (Figure 2). At depths >315 cm (prior to 650 BC), the average 206Pb/207Pb ratio was equal to 1.2011 ( 0.0015 (n = 24), a value which can be considered as the background level. The ratios decreased to a minimum of 1.1938 at a depth of 280 cm (280 BC). The layers comprised between depths of 107 and 205 cm exhibited values down to 1.1898. Above 50 cm (AD 1800), the 206 Pb/207Pb ratios decreased sharply and leveled off at around 1.1728 ( 0.0028 between 14 cm depth (AD 1940) and the 1 cm depth (AD 2002).
Mercury Concentrations. The Hg concentrations are given in Table S1. Mercury, Al, and POC (Figure 2) were not found to be correlated in the P-B lagoon, and normalization of Hg concentrations to Al or POC, as proposed by various authors (e.g., refs 23 and 24) to minimize the variations related to changes in sedimentary clay and organic matter content, did not drastically modify the Hg profile. For depths >190 cm (i.e., prior to around AD 900), Hg concentrations were relatively low (mean = 0.017 ( 0.003 μg g 1, n = 54) compared to crustal values,25 and displayed minor variations. From 190 cm depth (∼AD 900), concentrations increased sharply, peaking at 0.13 μg g 1 at 153 cm depth (AD 1150). A second lower maximum (0.082 μg g 1) was observed at 80 cm (∼AD 1650). Above 30 cm depth, concentrations sharply increased, with a broad peak at 0.58 μg g 1 being observed between 10 and 15 cm depth (AD 1945 1968).
’ DISCUSSION History of Pb Contamination. The Al-normalized Pb concentration and the 206Pb/207Pb profiles (Figure 2) show that the P-B sediments have recorded the three major periods of Pb pollution in Europe outlined by Renberg:26 the Etruscan Greek Roman period (650 BC to AD 50), the medieval period (AD 650 to 1450), and the modern period (from about AD 1850 to the present). These periods are separated by low pollution periods during the Dark Ages (between AD 50 and 650) and during the 16th century. During the Etruscan Greek Roman period Pb isotopes (Figure 3) suggest a possible contamination of P-B sediments by Pb originating from the region of Murcia in Spain (Cartagena mines) in agreement with archeological studies27 and the isotopic composition of Roman ingots found in the vicinity of the P-B lagoon.28,29 8644
dx.doi.org/10.1021/es2004599 |Environ. Sci. Technol. 2011, 45, 8642–8647
Environmental Science & Technology
ARTICLE
Figure 3. Plot of 206Pb/207Pb ratios vs 208Pb/206Pb ratios in preindustrial (before AD 1800) Pierre-Blanche sediments (a) and in modern sediments (after AD 1800) (b). Pre-650 BC samples: black circles; samples from the Etruscan Greek Roman period (650 BC to AD 50): triangles; medieval samples (AD 640 to 1450): full squares; modern samples (AD 1800 to 2002): open squares. Sources: Carthagena mine,52 Spanish ingots,28 Harz mineralization,53 Mont Lozere mines,54 Malines mines,55 Melle mine,56 Rio Tinto,52 coal,57,58 effluent from local sewage treatment plants,34 ash from a local incinerator, and leaded gasoline.35
The medieval pollution comprises two pollution phases with a maximum Pb concentration around AD 800 and a second, lower peak around AD 1150. The studies in southern Europe have generally not detected a peak of Pb pollution before AD 10007,8,30 with the exception of Kylander et al.31 in a high-resolution peat core in northwestern Spain. In that study, the peak of Pb pollution is observed around AD 900, which is consistent with the observations in the P-B lagoon, taking into account the great uncertainty ((100 yr) on dating due to variations of 14C reservoir age in the P-B lagoon in the medieval period.16 During the medieval period the Pb isotope composition differs from that of the Estruscan Greek Roman period (Figure 3). It reflects inputs of mines from southern France. Several mines from the southern border of the Massif Central were very active during this period.32 From AD 1250, the pollution starts to decrease in the P-B lagoon. The beginning of this decline coincides with the abandoning of the silver mines of southern France,32 and continues with the economic crisis and the Black Plague, which affected the region during the 14th century. Concentrations remain around geochemical background levels until the middle of the 16th century. From the end of the 19th century the Pb concentrations increase whereas the isotope composition moves toward less radiogenic Pb sources, in agreement with other regional records.26 From the end of the 19th century, Pb concentrations increase exponentially and the 206Pb/207Pb ratios show a general decrease until 1950, indicating variations in the contamination sources. The isotope composition of Pb (Figure 3) indicates that the Pb in gasoline always represented a minor contribution to Pb in the P-B lagoon even in the 1970s, when its use reached maximum. The similarity between coal production and/or coal consumption in France and the Pb profile in the lagoon (Figure 2), suggests that the use of coal as the main source of energy largely contributed to Pb contamination of the P-B lagoon until the 1970s. Although the decline in coal consumption slowed contamination, the decrease in Pb concentrations did not occur until the phasing out of Pb additives to gasoline, which began in 1974 in several European countries including France. Between 1972 and 1995, the Pb concentrations decreased by a factor of 1.5 in relation to changes in the composition of gasoline. Similar results
have been obtained in peat bogs from Denmark6 and the Swiss Jura mountains.33 However, unlike these studies, our data did not reveal an increase in 206Pb/207Pb as expected after the banning of Pb additives to gasoline. More recent Pb sources to the lagoon include effluents of sewage treatment plants34 and to a lesser extent those of regional incinerators.35 The increase of Pb concentrations observed after 1991 may be related to an increased contribution of these sources. History of Hg Contamination. Fitzgerald et al.36 provided convincing evidence that Hg accumulating in lake sediments is not significantly affected by diagenetic processes. Furthermore, the histories of Hg deposition derived from dated sediment cores agree well with the known histories of inputs.37 In oxic coastal sediments Hg is likely to be recycled prior to incorporation in the permanent sediment records; however, little postdepositional redistribution has been observed.38,39 In addition, the P-B record shares some features with atmospheric deposition records in peat from NW Spain,8 the Swiss Jura,9 and in the Diss Mere lake (southern England).10 All this pleads in favor of good preservation of the P-B Hg record. The Hg concentration profile (Figure 2) shows three major peaks: in AD 1150, in AD 1660, and between AD 1945 and AD 1969, with concentrations, respectively, 8, 5, and 34 times higher than the baseline level (0.017 ( 0.003 μg g 1). In the Middle Ages, significant Hg contamination is observed. As in other studies8 10 this could be related to local inputs due to the widespread use of Hg salts as medication against the plague and leprosy.40 At that time the town of Montpellier was famous for its Faculty of Medicine founded in AD 1150 and the use of Hg is attested by the discovery in the 19th century of metallic Hg and Hg salts in cellars of medieval houses.41 However, as suggested by Givelet et al.,42 deforestation and vegetation burning could also have contributed significantly to the medieval pollution of the P-B lagoon. These practices, which have been shown to mobilize the Hg currently stored in soils and vegetation and to increase the quantity of Hg in the water,43,44 were widespread during the medieval period in Europe.45 Like lead, Hg contamination of the lagoon decreased during the 14th century, due to the deep economic depression. Mercury concentration increased again during the 16th century and peaked 8645
dx.doi.org/10.1021/es2004599 |Environ. Sci. Technol. 2011, 45, 8642–8647
Environmental Science & Technology in the middle of the 17th century. Similar observations in northwest Spain were attributed to the intense activity of the Almaden mine.8 This period corresponds to the development of the Hg amalgamation process for the recovery of Ag and Au.46 As a consequence, the production of the European Hg mines (Almaden in Spain and Idrija in Slovenia) nearly doubled between AD 1550 and AD 165047 in order to supply European and American Au and Ag mines. However, although increasing, European Hg production did not peak during the 17th century.47 Therefore the peak of pollution observed in the P-B lagoon more probably reflects Hg uses. Mining activities in Central and South America, as well as in Eastern Europe, cannot account for the peak of contamination, because these activities did not result in global atmospheric contamination.48 For this reason, the Hg peak observed in the P-B lagoon is attributed to the expansion of gold and silver amalgamation in Europe and in the vicinity of the study area, where many auriferous rivers were exploited during the 16th and 17th centuries, a period that is sometimes considered to be the golden age of goldwashing in France. In the 19th century this type of gold production decreased sharply, with most of the gold being obtained from gold mines.49 From the end of the 19th century until 1960 there was an exponential increase in the concentration of Hg. During the 20th century, the variations in Hg concentrations characterized by maximum concentrations between AD 1945 AD and 1969 AD and a decline from 1969 on well correlated with coal production and consumption in France (Figure 2). This correlation and the synchronicity of Hg and Pb enrichments, suggest that coal burning, which has been found to be the major source of atmospheric Hg deposited in peat bogs from remote areas in the second half of the 20th century,6,9,50 was also the major source of Hg in the P-B lagoon. This study shows that for at least 1000 years the P-B lagoon has received significant Hg inputs related to human activity. Other records in Mediterranean coastal areas would be necessary to define the respective importance of local and regional Hg sources, since this is fundamental for a better understanding of the Hg cycle in the Mediterranean Sea.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional experimental details, data, and references. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT This study was financed by the ECLICA project (INSU, ACIFNS) and the French National Research Agency (EXTREMA project). Thanks to the Laboratoire de Mesure 14C (LMC14) on ARTEMIS in the CEA institute at Saclay for the 14C analyses (ECLICA and INTEMPERIES projects). ’ REFERENCES (1) Murozumi, M.; Chow, T. J.; Patterson, C. Chemical concentrations of pollutant lead aerosols, terrestrial dusts and sea salts in Greenland and Antarctic snow strata. Geochim. Cosmochim. Acta 1969, 33 (10), 1247–1294.
ARTICLE
(2) Hong, S.; Candelone, J. P.; Patterson, C. C.; Boutron, C. F. Greenland ice evidence of hemispheric lead pollution two millennia ago by Greek and Roman civilizations. Science 1994, 265, 1841–1843. (3) Renberg, I.; Persson, M.; Emteryd, O. Pre-industrial atmospheric lead contamination detected in Swedish lake sediments. Nature 1994, 368, 323–326. (4) Br€annval, M.-L.; Bindler, R.; Emteryd, O.; Renberg, I. Vertical distribution of atmospheric pollution lead in northern Europe: A summary from Swedish lake sediments. J. Paleolimnol. 2001, 25, 421–435. (5) Shotyk, W.; Weiss, D.; Appleby, P. G.; Cheburkin, A. K.; Frei, R.; Gloor, M.; Kramers, J. D.; Reese, S.; Van der Knaap, W. O., History of atmopsheric lead deposition since 12 370 14C yr BP from a peat bog, Jura mountains, Switzerland. Science 1998, 281. (6) Shotyk, W.; Goodsite, M. E.; Roos-Barraclough, F.; Frei, R.; Heinemeier, J.; Asmund, G.; Lohse, C.; Hansen, T. S. Anthropogenic contributions to atmospheric Hg, Pb and As accumulation recorded in peat cores from southern Greenland and Denmark dated using the 14C bomb pulse curve. Geochim. Cosmochim. Acta 2003, 67, 3991–4011. (7) Baron, S.; Lavoie, M.; Ploquin, A.; Carignan, D.; Pulido, M.; de Beaulieu, J.-L. Records of metal workshops in peat deposits; history and environmental impact on the Mont Lozere Massif. Environ. Sci. Technol. 2005, 39, 5131–5140. (8) Martinez-Cortizas, A.; Pontevedra-Plombal, X.; Garcia-Rodeja, E.; Novoa-Munoz, J. C.; Shoyk, W. Mercury in a Spanish peat bog archive of climate change and atmospheric deposition. Science 1999, 284, 939–942. (9) Roos-Barraclough, F.; Shotyk, W. Millenial-scale records of atmospheric mercury deposition obtained from ombrothrophic and minerotrophic peatlands in the Swiss Jura mountains. Environ. Sci. Technol. 2003, 37, 235–244. (10) Yang, H. Historical mercury contamination in sediments and catchment soils of Diss Mere, UK. Environ. Pollut. 2010, 158 (7), 2504–2510. (11) Leblanc, M.; Morales, J. A.; Borrego, J.; Elbaz-Poulichet, F. A 4500 years old mining pollution in Spain. Econ. Geol. 2000, 95, 655–662. (12) Miralles, J.; Veron, A. J.; Radakovitch, O.; Deschamps, P.; Tremblay, T.; Hamelin, B. Atmospheric lead fallout over the last century recorded in Gulf of Lions sediments (Mediterranean Sea). Mar. Pollut. Bull. 2006, 52 (11), 1364–1371. (13) Marriner, N.; Morhange, C. Geoscience of ancient Mediterranean harbours. Earth-Sci. Rev. 2007, 80, 137–194. (14) Veron, A.; Goiran, J. P.; Morhange, C.; Marriner, N.; Empereur, J. Y. Pollutant lead reveals the pre-Hellenistic occupation and ancient growth of Alexandria, Egypt. Geophys. Res. Lett. 2006, 33. (15) Le Roux, G.; Veron, A.; Morhange, C. Lead pollution in the ancient harbour of Marseilles. Mediterranee 2005, 1 2, 31–35. (16) Sabatier, P.; Dezileau, L.; Blanchemanche, P.; Siani, G.; Condomines, M.; Bentaleb, I.; Piques, G. Holocene variations of radiocarbon reservoir ages in a Mediterranean lagoonal system. Radiocarbon 2010, 52 (1), 91–102. (17) Sabatier, P.; Dezileau, L.; Condomines, M.; Briqueu, L.; Colin, C.; Bouchette, F.; Le Duff, M.; Blanchemanche, P. Reconstruction of paleostorm events in a coastal lagoon (Herault, South of France). Mar. Geol. 2008, 251 (3 4), 224–232. (18) IFREMER. Reseau de suivi lagunaire du Languedoc-Roussillon: Bilan des resultats 2007; 2008; p 363. (19) Etcheber, H.; Relexans, J.-C.; Beliard, M.; Weber, O.; Buscail, R.; Heussner, S. Distribution and quality of sedimentary organic matter on the Aquitanian margin (Bay of Biscay). Deep Sea Res. Part II 1999, 40 (10), 2249–2288. (20) Cossa, D.; Coquery, M.; Nakhle, K.; Claisse, D. Dosage du mercure total et du monomethylmercure dans les organismes et les sediments marins; IFREMER: Brest, 2002; p 28. (21) Van der Weijden, C. Pitfalls of normalization of marine geochemical data using a common divisor. Mar. Geol. 2002, 184, 167–187. (22) Kawakami, S.; Seidel, J. L.; Elbaz-Poulichet, F.; Achterberg, E. Trace metal biogeochemistry in the Mediterranean Thau Lagoon, a shellfish area. J. Coastal Res. 2008, 24, 194–202. 8646
dx.doi.org/10.1021/es2004599 |Environ. Sci. Technol. 2011, 45, 8642–8647
Environmental Science & Technology (23) Biester, H.; Bindler, R.; Martinez-Cortizas, A.; Engstrom, D. R. Modelling the past deposition of mercury using natural archives. Environ. Sci. Technol. 2007, 33, 4391–4395. (24) Hung, J. J.; Lu, C. C.; Huh, C. A.; Liu, J. T. Geochemical controls on distributions and speciation of As and Hg in sediments along the Gaoping (Kaoping) Estuary-Canyon system off southwestern Taiwan. J. Mar. Syst. 2009, 76 (4), 479–495. (25) Taylor, S. R.; McLennan, S. M. The Continental Crust: Its Composition and Evolution; Blackwell Scientific: Oxford, 1985; p 312. (26) Renberg, I.; Bindler, R.; Br€annval, M.-L. Using the historical and atmospheric lead-deposition record as a chronological marker in sediment deposits in Europe. The Holocene 2001, 11, 511–516. (27) Laubenheimer-Leenhardt, F. Recherches sur les lingots de Cu et Pb d’epoque romaine dans les regions de Languedoc-Roussillon et Provence-Corse. Revue Archeologique de Narbonnaise 1973, Supplement 3, 215 pp. (28) Labonne, M.; Ben Othman, D.; Luck, J.-M. Recent and past anthropogenic impact on a Mediterranean lagoon: Lead isotope constraints from mussel shells. Appl. Geochem. 1998, 13, 885–892. (29) Trincherini, P. R.; Domergue, C.; Manteca, I.; Nesta, A.; Quarati, P. The identification of lead ingots from the Roman mines of Cartagena (Murcia, Spain): The role of lead isotope analysis. J. Roman Archaeol. 2009, 22, 123–145. (30) Alfonso, S.; Grousset, F.; Masse, L.; Tastet, J.-P. European lead isotope signal recorded from 6000 to 300 years BP in coastal marshes (SW France). Atmos. Environ. 2001, 35, 3595–3605. (31) Kylander, M. E.; Weiss, D. J.; Martinez Cortizas, A.; Spiro, B.; Garcia-Sanchez, R.; Coles, B. J. Refining the pre-industrial atmospheric Pb isotope evolution curve in Europe using an 8000 year old peat core from NW Spain. Earth Planet. Sci. Lett. 2005, 240 (2), 467–485. (32) Bailly-Ma^itre, M. C. L’Argent. Du minerai au pouvoir dans la France medievale; A & J. Picard: Paris, 2002; p 212. (33) Shotyk, W.; Weiss, D.; Kramers, J. D.; Frei, R.; Cheburkin, A. K.; Gloor, M.; Reese, S. Geochemistry of the peat bog at Etang de Gruere, Jura mountain, Switzerland, and its record of atmospheric Pb and lithogenic trace metals (Sc, Ti, Y, Zr, and REE) since 12370 14C yr BP. Geochim. Cosmochim. Acta 2001, 65, 2337–2360. (34) Monna, F.; Ben Othman, D.; Luck, J. M. Pb isotopes and Pb, Zn and Cu concentrations in the rivers feeding a coastal pond (Thau, southern France): Constraints on the origin and flux of metals. Sci. Total Environ. 1995, 166, 19–34. (35) Monna, F.; Lancelot, J.; Croudace, I.; Cundy, A.; Lewis, J. Pb isotopic composition of airborne particulate material from France and the southern United Kingdom: Implications for Pb pollution sources in urban areas. Environ. Sci. Technol. 1997, 31, 2277–2286. (36) Fitzgerald, W.; Engstrom, D.; Mason, R. P.; Nater, E. A. The case for atmospheric mercury contamination in remote areas. Environ. Sci. Technol. 1998, 22 (1), 1–7. (37) Lockhart, W. L.; Macdonald, R. W.; Outridge, P. M.; Wilkinson, P.; DeLaronde, J. B.; Rudd, J. W. M. Tests of the fidelity of lake sediment core records of mercury deposition to known histories of mercury contamination. Sci. Total Environ. 2000, 260 (1 3), 171–180. (38) Gobeil, C.; Cossa, D. Mercury in sediments and sediments pore water in Laurentian Trough. Can. J. Fish. Aquat. Sci. 1993, 50, 1794–1800. (39) Hurley, J. P.; Krabbendorf, D. P.; Aiken, G.; Olson, M. L.; Cleckner, L. B., System controls on water column total and methyl mercury in the northern Everglades. In Fourth International Conference on Mercury as a Global Pollutant, Hamburg, Germany, 1996; p 74. (40) Rasmussen, K. L.; Boldsen, J. L.; Kristensen, H. K.; Skytte, L.; Hansen, K. L.; Molholm, L.; Grootes, P. M.; Nadeau, M.-J.; Eriksen, K. M. Mercury levels in Danish Medieval human bones. J. Archaeol. Sci. 2008, 35 (8), 2295–2306. (41) Anonymous. Sur le gisement de mercure natif, au milieu des marnes tertiaires qui composent une partie du sol sur lequel Montpellier se trouve b^atie. 1934, Compte-Rendu des seances de la Societe Geologique de France, (IV), 367. (42) Givelet, N.; Ross-Barraclough, F.; Shotyk, W. Predominant anthropogenic sources and rates of atmospheric mercury accumulation
ARTICLE
in southern Ontario recorded by peat cores from three bogs: comparison with natural “background” values (past 8000 years). J. Environ. Monit. 2003, 5, 935. (43) Garcia, E. R.; Carignan, D.; Lean, R. S. Seasonal and interannual variations in methyl mercury concentrations in zooplankton from boreal lakes impacted by deforestation or natural forest fires. Environ. Monit. Assess. 2007, 131, 1–11. (44) Mainville, N.; Webb, J.; Lucotte, M.; Davidson, R.; Betancourt, O.; Cueva, E.; Mergler, D. Decrease of soil fertility and release of mercury following deforestation in the Andean Amazon, Napo River Valley, Ecuador. Sci. Total Environ. 2006, 368 (1), 88–98. (45) Williams, M. Dark ages and dark areas: Global deforestation in the deep past. J. Hist. Geogr. 2000, 26 (1), 28–46. (46) Agricola, G. De Re Metallica; Dover Publications: New York, 1950; p 640. (47) Hylander, L. D.; Meili, M. 500 years of mercury production: Global annual inventory by region until 2000 and associated emissions. Sci. Total Environ. 2003, 304, 13–27. (48) Lamborg, C.; Fitzgerald, W.; Damman, A.; Benoit, J.; Balcom, P.; Engstrom, D., Modern and historic atmospheric mercury fluxes in both hemispheres: Global and regional cycling implications. Global Biogeochem. Cycles 2002, 16, 1104; doi: 10.1029/2001GB001847. (49) Guiollard, P. C. Guide pratique du chercheur d’or en France; BRGM Editions, 2004; p 127. (50) Shotyk, W.; Goodsite, M.; Roos-Barraclough, F.; Givelet, N.; Le Roux, G.; Weiss, D.; Cheburkin, A.; Knudsen, K.; Heinemeier, J.; van Der Knapp, W.; Norton, S.; Lohse, C. Accumulation rates and predominant atmospheric sources of natural and anthropogenic Hg and Pb on the Faroe Island. Geochim. Cosmochim. Acta 2005, 69, 1–17. (51) Ferrand, J.-L. Etude isotopique du cycle geochimique du plomb anthropique et naturel en milieu marin et c^otier. Ph-D dissertation, AixMarseille, 1996. (52) Stos-Gale, Z.; Gale, N. H.; Hiughton, J.; Speakman, R. Lead isotope data from the isotrace laboratory, Oxford: Archaeometry data base 1, ores from the western Mediterranean. Archaeometry 1995, 37, 407–415. (53) Monna, F.; Hamer, K.; Leveque, J.; Sauer, M. Pb isotopes as a reliable marker of early mining and smelting in the Northern Harz province (Lower Saxony, Germany). J. Geochem. Explor. 2000, 68 (3), 201–210. (54) Baron, S.; Carignan, J.; Ploquin, A. Dispersion of heavy metals (metalloids) in soils from 800-year-old pollution (Mont-Lozere, France). Environ. Sci. Technol. 2006, 40, 5319–5326. (55) Le Guen, M.; Orgeval, J. J.; Lancelot, J. Lead isotope behaviour in a polyphased Pb-Zn ore deposit: Les Malines (Cevennes, France). Mineralia Deposita 1991, 26, 180–188. (56) Tereygeol, F. Production and circulation of silver and secondary products (lead and glass) from Frankish royal silver mines at Melle (VIIth-Xth century). In Post-Roman towns and trade in Europe, Byzantium and the near East, Henning, J., Ed., 2005; Vol. 1, pp 123 134. (57) Diaz-Somoano, M.; Kylander, M. E.; Lopez-Anton, M. A.; Suarez-Ruiz, I.; Martinez-Tarazona, M. R.; Ferrat, M.; Kober, B.; Weiss, D. J. Stable Lead Isotope Compositions In Selected Coals From Around The World And Implications For Present Day Aerosol Source Tracing. Environ. Sci. Technol. 2009, 43 (4), 1078–1085. (58) Farmer, J. G.; Eades, L. J.; Graham, M. C. The lead content and isotopic composition of British coals and their implications for past and present releases of lead to the UK environment. Environ. Geochem. Health 1999, 21 (3), 257–272.
8647
dx.doi.org/10.1021/es2004599 |Environ. Sci. Technol. 2011, 45, 8642–8647
ARTICLE pubs.acs.org/est
Measurements of Isoprene-Derived Organosulfates in Ambient Aerosols by Aerosol Time-of-Flight Mass Spectrometry—Part 2: Temporal Variability and Formation Mechanisms Lindsay E. Hatch,† Jessie M. Creamean, † Andrew P. Ault,†,3 Jason D. Surratt,‡,4 Man Nin Chan,§ John H. Seinfeld,§,|| Eric S. Edgerton,^ Yongxuan Su,† and Kimberly A. Prather†,#,* †
Department of Chemistry and Biochemistry, University of California, San Diego, La Jolla, California 92093, United States Department of Chemistry, California Institute of Technology, Pasadena, California 91125, United States § Division of Engineering and Applied Science, California Institute of Technology, Pasadena, California 91125, United States Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, California 91125, United States ^ Atmospheric Research & Analysis, Inc., Cary, North Carolina 27513, United States # Scripps Institution of Oceanography, University of California, San Diego, La Jolla, California 92093, United States
)
‡
bS Supporting Information ABSTRACT: Organosulfate species have recently gained attention for their potentially significant contribution to secondary organic aerosol (SOA); however, their temporal behavior in the ambient atmosphere has not been probed in detail. In this work, organosulfates derived from isoprene were observed in single particle mass spectra in Atlanta, GA during the 2002 Aerosol Nucleation and Characterization Experiment (ANARChE) and the 2008 August Mini-Intensive Gas and Aerosol Study (AMIGAS). Real-time measurements revealed that the highest organosulfate concentrations occurred at night under a stable boundary layer, suggesting gas-to-particle partitioning and subsequent aqueous-phase processing of the organic precursors played key roles in their formation. Further analysis of the diurnal profile suggests possible contributions from multiple production mechanisms, including acid-catalysis and radical-initiation. This work highlights the potential for additional SOA formation pathways in biogenically influenced urban regions to enhance the organic aerosol burden.
1. INTRODUCTION Organic species can significantly influence the health and climate impacts of atmospheric particulate matter due to their potential toxicity and effect on hygroscopicity.1 Organic aerosols can arise from both primary (directly emitted) and secondary (formed in the atmosphere) sources.2 Despite extensive research in recent years,2 models often under-predict the contribution of secondary organic species to aerosol mass3 as a result of the remaining low scientific understanding of organic aerosol formation pathways. Biogenic volatile organic compounds (BVOCs) can play a significant role in the formation of secondary organic aerosols (SOA) due to their high reactivity toward atmospheric oxidants. Indeed, radiocarbon dating has attributed a dominant fraction (∼65%) of SOA to biogenic sources in heavily forested regions.4 Isoprene (2-methyl-1,3-butadiene, C5H8) is the most abundant BVOC and has been implicated in SOA formation (ref 5 and references therein). Further, in urban environments, r 2011 American Chemical Society
BVOCs can interact with anthropogenic pollutants resulting in an enhancement of the organic aerosol mass.6,7 Anthropogenic emissions can influence biogenic SOA in a number of ways,8 including by introducing new chemical pathways to produce unique, low-volatility chemical species. Herein, we focus on a specific class of compounds, organosulfates, which have been demonstrated to form via reaction between BVOCs and particulate sulfate derived predominantly from industrial SO2 emissions.9 Several mechanisms have been proposed from laboratory investigations to explain the formation of organosulfate compounds, including acid-catalyzed alcohol sulfate-esterification,10 reaction with sulfate radicals (SO4 3 , HSO4 3 ),11 13 or acid-catalyzed Received: April 9, 2011 Accepted: August 2, 2011 Revised: July 11, 2011 Published: September 09, 2011 8648
dx.doi.org/10.1021/es2011836 | Environ. Sci. Technol. 2011, 45, 8648–8655
Environmental Science & Technology
ARTICLE
.Scheme 1. Organosulfate Formation via IEPOX Intermediate (adapted from ref 14)
epoxide hydrolysis.14,15 The esterification mechanism has been shown to be kinetically limited and not likely to occur in the ambient atmosphere.16 Radical-initiated formation has been proposed following the observation of organosulfate production in smog chamber reactions under irradiated conditions, which ceased in the dark.11,17 Several studies have implicated the role of epoxide intermediates in leading to SOA compounds,18 20 and in particular to organosulfate species.14,21 The atmospheric fate of the isoprene-derived epoxide (IEPOX) largely depends on aerosol pH and liquid water content,22 but has been shown to produce 2-methyltetrols and, in the presence of acidic sulfate aerosol, organosulfate species, as shown in Scheme 1. In Part 1 of this study, we described observations of the sizeresolved mixing state of organosulfates during the 2002 Aerosol Nucleation and Characterization Experiment (ANARChE) and the 2008 August Mini-Intensive Gas and Aerosol Study (AMIGAS) in Atlanta, GA.23 The online mass spectrometry analysis employed in these studies provided the high time resolution necessary to monitor rapid changes in aerosol composition. Herein, we describe the temporal trends of organosulfate species and the resulting insights into the possible formation mechanisms occurring in the ambient atmosphere. To our knowledge, this is the first study to investigate in real-time the temporal variability of ambient particle-phase organosulfate species.
2. EXPERIMENTAL SECTION 2.1. Field Measurements. Aerosol sampling was performed during two field studies at the Jefferson Street Southeastern Aerosol Research and Characterization (SEARCH) Network site (33.8 N, 84.4 W), a mixed residential industrial location about 4 km northwest of metro Atlanta.24 Data are presented in Eastern Standard Time. Real-time ATOFMS measurements were made of single particles during ANARChE from August 4 11, 2002 and during AMIGAS from August 22-September 10, 2008. Additional details of ATOFMS data collection and analysis are provided elsewhere.23,25 Briefly, ATOFMS detected the size and chemical composition of individual particles from 50 350 nm during ANARChE and 200 3000 nm during AMIGAS. ATOFMS collects both positive and negative spectra for each particle; for the data sets considered here, some negative spectra displayed incorrect mass calibration in the mass range of interest (>100 m/z) due to instability in the mass spectrometer voltages. The manual correction method applied to the data is described in the Supporting Information. Further, as described by Hatch et al.,23 organosulfates were predominantly detected in submicrometer particles; therefore, only the submicrometer particles (0.2 1 μm) with dual-polarity spectra collected during AMIGAS are included in the detailed analysis herein. All particles with dualpolarity spectra were included from ANARChE since these measurements were restricted to particles <350 nm.
Co-located instrumentation monitored meteorology (wind speed/direction, precipitation, temperature, relative humidity, solar radiation), and gas-phase pollutant levels (NOx, CO, O3, SO2, HNO3).24 Continuous PM2.5 samplers also monitored total carbon levels.26 During AMIGAS, gas-phase water-soluble organic carbon (WSOC) measurements were obtained by a mistchamber.27
3. RESULTS AND DISCUSSION As discussed by Hatch et al.,23 ATOFMS detected isoprenederived organosulfates at m/z 139, 153, 155, 199, and 215 during both ANARChE and AMIGAS. Identification of these species was verified through ATOFMS analysis of chemical standards and comparison of ambient ATOFMS data to accurate mass analysis obtained from co-located filter samples analyzed off-line by high resolution mass spectrometry.23 For brevity, this manuscript focuses on the marker at m/z 215 since it was the most abundant isoprene-derived organosulfate compound observed in the ATOFMS data by a significant margin (∼4) and the other markers displayed similar temporal profiles, as shown in the Supporting Information for the AMIGAS campaign. While the formation mechanism of m/z 215 will be explored below, for simplicity, it will be referred to as the IEPOXderived organosulfate since both IEPOX and the corresponding organosulfate were detected in Atlanta during AMIGAS.28 3.1. Temporal Trends of Organosulfate Species. Because both ANARChE and AMIGAS were conducted during summer, comparison of the data sets serves to provide verification of observations and improved understanding of the general behavior of organosulfate species. However, it should be noted at the outset that this work focuses on the qualitative trends of organosulfates during these Atlanta field campaigns. The ATOFMS mass spectral data have not been scaled to correct for potential instrumental differences (e.g., MS tuning) that could influence the absolute peak intensity measured during each campaign. Therefore, only the relative trends in organosulfate mass spectral peak area observed in the two data sets will be compared herein (e.g., similarities in diurnal cycle); differences in the peak magnitude should not be used to infer differences in organosulfate concentrations between the two studies. Using the high time resolution of real-time, single-particle data, the peak area of the IEPOX-derived organosulfate at m/z 215 was shown to fluctuate significantly throughout the ANARChE and AMIGAS campaigns (Figure 1). Much of this variation can be attributed to changes in meteorology. Both study periods exhibited several days characteristic of a diurnal cycle, as evidenced by regular patterns in meteorological (e.g., temperature, relative humidity) and gas-phase pollutant data (e.g., ozone) (ANARChE: 8/4 8/5/02, 8/8 8/11/02; AMIGAS: 8/29 8/31/08, 9/3/08, 9/6 9/10/08). Additionally, during AMIGAS, two tropical cyclones (TCs; Gustav, 8/24 9/4/0829 and Hannah, 8/26 9/7/0830) influenced the southeastern United States. The time periods that TCs directly impacted 8649
dx.doi.org/10.1021/es2011836 |Environ. Sci. Technol. 2011, 45, 8648–8655
Environmental Science & Technology
ARTICLE
Figure 1. Hourly averaged temporal variation in the absolute area of m/z 215, solar radiation, wind speed and direction, precipitation, temperature, and relative humidity. Black bars indicate time periods influenced by tropical cyclones (TCs). Data from both ANARChE and AMIGAS are included, separated by the dashed line.
the Atlanta region are inferred from the prevailing southeasterly winds and fast wind speeds (i.e., 8/23 8/27/08, 8/31 9/2/08, 9/4 9/5/08; Figure 1), and included a rain episode from 8/24 8/26/08. Tropical storm Bertha was also present in the Gulf of Mexico during ANARChE (8/4 8/9/02)31 and may have slightly influenced the Atlanta region, as indicated by a shift in meteorology on 8/6 8/8/02 (Figure 1). The high temporal resolution of ATOFMS measurements provided unique observations of the behavior of organosulfate species during both the calm days and the meteorological events, as described below. 3.1.1. Diurnal Trends. The non-TC days during ANARChE (8/4 8/5/02, 8/8 8/11/02) and AMIGAS (8/29 8/31/08, 9/3/08, 9/6 9/10/08) provided insight into the typical diurnal cycle of organosulfate species in Atlanta. Consistently during both campaigns, the highest abundance of organosulfate species occurred at night during periods with low wind speeds (Figure 1). The stagnant wind conditions and buildup of NOx and CO (Figure 2) indicate that a stable boundary layer was present overnight. Radiosonde data were obtained from the NOAA Integrated Global Radiosonde Archive32 to investigate the boundary layer stability; vertical temperature profiles are included in Figure 3. These soundings were taken at 12Z (7am EST), but are expected to be representative of the nocturnal temperature profile since breakup of inversions in Atlanta has been observed to occur after 9 am.33 The radiosonde data confirmed the presence of nocturnal temperature inversions during these periods (Figure 3) yielding a very shallow mixed layer and increased pollutant levels (e.g., NOx and CO; Figure 2). The nights characterized by the strongest inversions, as indicated by the temperature difference between the top and bottom of the inversion layer, are associated with the highest levels of organosulfates (e.g., 9/6/08, 9/8/08, Figures 1 and 3). Also of note, an
inversion was not present on 9/10/08 (Figure 3), which could explain the lower abundance of organosulfates observed on this night during AMIGAS (Figure 1). These soundings and the diurnal trends observed by ATOFMS provide strong evidence for the role of a stable boundary layer in the formation of organosulfates in Atlanta. The night-time meteorology also displayed a characteristic decrease in temperature and increase in relative humidity (RH), providing favorable conditions for condensation of water and semivolatile species. While IEPOX is not expected to partition appreciably to the organic phase,28 Eddingsaas et al.22 predicted an increase in IEPOX uptake with increasing aerosol water content, as would be expected under the high nocturnal RH conditions. It should also be noted that the presence of water can influence the ionization of species from a particle in the ATOFMS, and in particular has been shown to suppress negative ion formation.34 Since organosulfate species were detected as negative ions and in higher abundance at night (higher RH and therefore higher aerosol water content), it is believed that the observed diurnal trend is not a measurement artifact, but rather, perhaps represents a lower bound on the nighttime organosulfate abundance relative to other species in the particles. The peak area of m/z 215 strongly tracks NOx (NO2 + NO) concentration during both ANARChE and AMIGAS (Figure 2). Interestingly, IEPOX has been shown to form under low-NOx conditions,14 making the correlation between the IEPOX-derived organosulfate and NOx somewhat unexpected. Several possible explanations exist for this relationship: (1) Particulate organosulfate species and NOx may have increased overnight independently of each other, driven predominantly by the lowering of the boundary layer. In this scenario, the organosulfate precursors may have formed during the day under low-NOx 8650
dx.doi.org/10.1021/es2011836 |Environ. Sci. Technol. 2011, 45, 8648–8655
Environmental Science & Technology
ARTICLE
Figure 2. Hourly averaged temporal variation in the m/z 215 absolute peak area, gas-phase water-soluble organic carbon (WSOC), total PM2.5 carbon mass concentration (Total Carbon), NOx, CO, O3, and SO2. Black bars indicate time periods influenced by tropical cyclones (TCs). Data from both ANARChE and AMIGAS are included in the figure, separated by the dashed line.
conditions, followed by condensation of the semivolatile oxidized VOCs and organosulfate formation overnight. The overnight increase in NOx would then be largely uninvolved in the organosulfate mechanism. (2) It is also possible that nocturnal nitrogen oxide chemistry played a role in the formation of organosulfates by raising aerosol acidity, an important parameter in the formation of organosulfates.35 Overnight, NO2 reacts with ozone to produce NO3 3 , which can subsequently react with NO2 to form N2O5; heterogeneous hydrolysis of N2O5 yields HNO3(aq) (ref 36 and references therein). Figure 4 shows the temporal trends of NO2, HNO3(g), particulate nitrate, and m/z 215. The rise in NO2 concentration overnight closely tracks the ATOFMS nitrate signal (m/z 62), indicating the possible conversion of NOx into HNO3 by the reaction outlined above. Further, Brown et al. have demonstrated an increase in N2O5 uptake to aerosols with high sulfate content,37 as is typical of Atlanta aerosol. While measurements of NO3 3 and N2O5 were not available for these studies, previous work has shown that N2O5 is the dominant species under high NO2 conditions,36 as observed overnight in Atlanta (Figure 4). In addition, we note that gas-phase nitric acid decreases as night falls (Figure 4), consistent with the partitioning of HNO3 to the aerosol phase. Since the condensation of nitric acid and increase in particulate nitrate coincide with the rise in organosulfate abundance, it is possible that the reactive nitrogen species played a role in the formation of organosulfate by raising the aerosol acidity and thus enhancing the reactive uptake of organosulfate precursors. Although, on the basis of these measurements, we cannot definitively establish a causal link between reactive nitrogen chemistry and organosulfate formation versus a dominant role of the nocturnal boundary layer, we note that both scenarios (1) and (2) involve partitioning/reactive uptake of the organosulfate precursors overnight. (3) A third possible explanation for the correlation between organosulfates
and NOx involves the indirect formation of organosulfates via an organonitrate intermediate. Organonitrates can arise from reaction of isoprene with OH under high-NOx conditions during the day19 or at night with nitrate radical;38 particulate nitrate can also act as a nucleophile toward epoxides similar to sulfate in Scheme 1.39 Tertiary organonitrates have been found to undergo rapid substitution by water or sulfate, leading to the formation of polyols and organosulfates.39 Indeed, nitrooxy-organosulfate species derived from isoprene have been previously observed in Atlanta by Surratt et al.,40 wherein one or two nitrate groups are in the place of hydroxyl groups in the product compound of Scheme 1. The hydrolysis of such compounds would produce the same m/z 215 ion described herein. In summary, the apparent correlation between organosulfates and reactive nitrogen species could be driven by a combination of effects (1) (3), or by some other mechanism, warranting further exploration. By peaking mostly at night, organosulfate species are largely anticorrelated with O3, which forms by photochemical processes during the day. However, the highest night-time concentrations of organosulfates generally follow days with high O3 levels (e.g., 8/8 8/9/02 and 9/3 9/4/08; Figure 2). This relationship indicates that there was likely a high degree of photochemical isoprene oxidation contributing to the enhanced O3 levels41 and leading to the daytime formation of the organosulfate precursors, which then became concentrated under the low nocturnal boundary layer. These observations are supported by the watersoluble organic carbon (WSOC) measurements from AMIGAS (Figure 2). Gas-phase WSOC levels increased overnight, consistent with a low mixing height concentrating these species (likely including the organosulfate precursors) in the atmosphere, which likely enhanced the gas-to-particle partitioning. Somewhat surprisingly, a clear trend between organosulfates and SO2 was not observed: the large SO2 spikes did not appear 8651
dx.doi.org/10.1021/es2011836 |Environ. Sci. Technol. 2011, 45, 8648–8655
Environmental Science & Technology
ARTICLE
Figure 3. Vertical temperature profiles during ANARChE (left) and AMIGAS (right) obtained from the NOAA Integrated Global Radiosonde Archive.32 Only non-TC days are shown for AMIGAS to avoid graph clutter.
Figure 4. Hourly averaged temporal trends in the absolute peak areas of m/z 215 and m/z 62 (NO3 ), HNO3(g), and NO2 during ANARChE (top) and AMIGAS (bottom). For graph clarity, only the period with highest organosulfate abundance (9/2 9/10/08) is shown for AMIGAS.
to significantly enhance the levels of organosulfates. However, a small afternoon increase in organosulfate abundance is noted on the afternoon of 8/5/02 during an SO2 plume. Similarly, during AMIGAS, an SO2 spike on the afternoon of 9/6/08 led to one of the only days with higher organosulfate levels during the day than the following night.23 It is possible that the semivolatile nature of IEPOX prevented extensive partitioning during the day, even if there was an increase in aerosol acidity induced by SO2 emissions.
The correlations of organosulfates with meteorological and gas-phase pollutant data described above suggest that gas-toparticle partitioning may be the key step in determining the concentration of particle-phase organosulfate species. As the boundary layer height and temperature decrease and RH increases at night, there is likely enhanced partitioning of the isoprene oxidation products (i.e., IEPOX in the case of the m/z 215 marker) followed by particle-phase reactions resulting in organosulfate formation. The relatively long gas-phase IEPOX 8652
dx.doi.org/10.1021/es2011836 |Environ. Sci. Technol. 2011, 45, 8648–8655
Environmental Science & Technology
ARTICLE
Figure 5. Hourly averaged temporal trends of m/z 215 and 195 absolute peak areas normalized by CO concentration during ANARChE (top) and AMIGAS (bottom). For graph clarity, only the period with highest organosulfate abundance (9/2 9/10/08) is shown for AMIGAS. The ozone traces are overlaid to highlight the correlation with photochemical activity.
lifetime of 18 22 h42 is consistent with this proposed delay between afternoon photochemical production of IEPOX and the nighttime formation of organosulfates. We note that IEPOX and other organosulfate precursors could have formed locally or advected from other locations during the day; indeed it is likely that the organosulfate precursors form regionally, given the similarities between the IEPOX concentration in Atlanta versus a rural Georgia site.28 3.1.2. Tropical Cyclone Influence. The lowest abundance of m/z 215 occurred during the rainy, TC-influenced period of AMIGAS (8/24 8/27/08) when the aerosols and pollutants were scavenged by precipitation or advected from the area, as evidenced by the low levels of gas-phase pollutants and total particulate carbon (Figure 2). This period and the additional TC periods (9/1 9/2/08, 9/4 9/5/08) were characterized by high wind speeds and predominantly southeasterly winds. During these storms, solar radiation and O3 concentration decreased (Figures 1 and 2), consistent with a decrease in photochemical activity, resulting in reduced formation of isoprene oxidation products, and therefore lower organosulfate abundance. During ANARChE, the levels of organosulfates were lowest from 8/6 8/8/02. A shift in meteorology is clear from the wind speed and relative humidity data (Figure 1); a decrease in barometric pressure was also observed on 8/6/02 (not shown). As mentioned above, TC Bertha present in the Gulf of Mexico may have induced the shift in meteorological patterns, though its potential influence on Atlanta has not been verified. Additionally, a deeper temperature inversion was present on 8/6/02 and a negligible one existed on 8/7/02 (Figure 3); this change in boundary layer height could have also contributed to the lower organosulfate abundance on these nights. 3.2. Possible Formation Mechanisms of Organosulfates. Since boundary layer stability appears to play a strong role in
the absolute levels of particulate organosulfate species in Atlanta, the possible chemical mechanisms were further investigated by normalizing the organosulfate peak area by CO concentration. CO has been used as an inert chemical tracer to separate the effects of vertical mixing and chemical reactions on pollutant concentrations.43 The CO-normalized temporal profiles are shown in Figure 5. During the ANARChE campaign, the normalized IEPOX-derived organosulfate levels exhibited a maximum in the afternoon on the days not influenced by a TC (8/4 8/5/02, 8/9 8/11/02), coinciding with high ozone concentrations indicative of high photochemical activity. The concurrent increase in the ATOFMS sulfuric acid marker (m/z 195, HSO4 3 H2SO4 )44 is consistent with photochemical production from SO2 and possibly indicates the importance of particle acidity in the formation of organosulfates. However, previous studies have also proposed a radicalinitiated organosulfate formation mechanism via reaction between isoprene and sulfate/bisulfate radical giving rise to the m/z 215 organosulfate.11,17 As these radical reactions occur under irradiated conditions, potential contribution of this mechanism to the daytime organosulfate concentration cannot be ruled out. It is noteworthy that organosulfate compounds have been observed with the same molecular formulas via both the radical-initiated and acid-catalyzed mechanisms10,11 and thus these ATOFMS measurements cannot differentiate the afternoon contributions from these two potential pathways. It is therefore possible that the afternoon increase in organosulfate levels is due to either the radical formation mechanism or acidcatalyzed reactive uptake (or both). As described above, the highest absolute levels of organosulfates were detected overnight, primarily driven by the low boundary layer height. While the precursors may have formed elsewhere, it is likely that the organosulfates formed locally, as 8653
dx.doi.org/10.1021/es2011836 |Environ. Sci. Technol. 2011, 45, 8648–8655
Environmental Science & Technology opposed to advection of organosulfate-containing particles to Atlanta, given the very low wind speed (<0.5 m/s) during these periods. The radical-initiated mechanism cannot explain the production of organosulfates overnight due to the lack of solar radiation required to form the sulfate radicals. Therefore, nocturnal organosulfate formation can likely be attributed to acid-catalyzed reactive uptake of the precursors. During AMIGAS both the particle size and temporal trends (Figures S2 and S3 of the Supporting Information) demonstrate a relationship of increasing IEPOX-derived organosulfate with increasing m/z 195, an indicator of acidity. While the trends from ANARChE and AMIGAS are not in full agreement on the correlation between organosulfates and sulfuric acid (discussed in the Supporting Information), it is important to note that nitric acid possibly contributed to the nocturnal aerosol acidity, as described above, with consistent trends during both studies (Figure 4). Further, particulate IEPOX was detected during AMIGAS, including night-time samples,28 and its role in forming organosulfate species has been confirmed by smog chamber studies,14 thus providing additional support for the hypothesized reactive-uptake of IEPOX in leading to the overnight formation of organosulfates. In summary, the results from ANARChE and AMIGAS indicate that several formation mechanisms can contribute to the production of particulate organosulfate species in Atlanta. While further studies are necessary to better elucidate the conditions that favor a given formation pathway in the ambient atmosphere, the ATOFMS observations suggest that organosulfates form by the following atmospheric processes: During the daytime, isoprene is photochemically oxidized to IEPOX, with moderate organosulfate production from either a radicalinitiated mechanism or acid-catalyzed reactive uptake. Since IEPOX is semivolatile and reactive uptake increases with aerosol water content,22 it likely partitions to the aerosol phase to a greater extent as the RH increases and the temperature and boundary layer height decrease at night. Upon condensation, the acid-catalyzed mechanism likely dominates with subsequent particle-phase formation of organosulfate species. These results highlight the important roles of aqueous-phase processing and boundary layer stability in SOA formation.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional information regarding the data correction method, temporal profiles of other organosulfate markers (Figure S1), and discussion of the correlation between aerosol acidity and organosulfate formation (Figures S2 S4). This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected]. Present Addresses 3
Department of Chemistry, University of Iowa, Iowa City, Iowa 52242, United States. 4 Department of Environmental Sciences and Engineering, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599, United States.
ARTICLE
’ ACKNOWLEDGMENT We gratefully acknowledge Stephanie Shaw and Eladio Knipping (EPRI) for coordinating the AMIGAS campaign. The authors thank the rest of the Prather group for extensive support throughout the studies. Additional assistance in Atlanta from Jerry Brown (ARA, Inc.) was appreciated during AMIGAS. Water-soluble organic carbon measurements were provided by Xiaolu Zhang and Prof. Rodney Weber (Georgia Institute of Technology). Michele Sipin is acknowledged for assisting with data collection during ANARChE. The authors thank anonymous reviewers for insightful comments. ANARChE was supported by the University of Rochester EPA PM Center, Grant R827354. AMIGAS was funded by the Electric Power Research Institute. L.H. has been funded by a National Science Foundation Graduate Research Fellowship (2008-2011). ’ REFERENCES (1) Jacobson, M. C.; Hansson, H. C.; Noone, K. J.; Charlson, R. J. Organic atmospheric aerosols: Review and state of the science. Rev. Geophys. 2000, 38 (2), 267–294. (2) Hallquist, M.; Wenger, J. C.; Baltensperger, U.; Rudich, Y.; Simpson, D.; Claeys, M.; Dommen, J.; Donahue, N. M.; George, C.; Goldstein, A. H.; Hamilton, J. F.; Herrmann, H.; Hoffmann, T.; Iinuma, Y.; Jang, M.; Jenkin, M. E.; Jimenez, J. L.; Kiendler-Scharr, A.; Maenhaut, W.; McFiggans, G.; Mentel, T. F.; Monod, A.; Prevot, A. S. H.; Seinfeld, J. H.; Surratt, J. D.; Szmigielski, R.; Wildt, J. The formation, properties and impact of secondary organic aerosol: current and emerging issues. Atmos. Chem. Phys. 2009, 9 (14), 5155–5236. (3) Volkamer, R.; Martini, F. S.; Molina, L. T.; Salcedo, D.; Jimenez, J. L.; Molina, M. J. A missing sink for gas-phase glyoxal in Mexico City: Formation of secondary organic aerosol. Geophys. Res. Lett. 2007, 34, (19). (4) Lewis, C. W.; Klouda, G. A.; Ellenson, W. D. Radiocarbon measurement of the biogenic contribution to summertime PM-2.5 ambient aerosol in Nashville, TN. Atmos. Environ. 2004, 38 (35), 6053–6061. (5) Carlton, A. G.; Wiedinmyer, C.; Kroll, J. H. A review of Secondary Organic Aerosol (SOA) formation from isoprene. Atmos. Chem. Phys. 2009, 9 (14), 4987–5005. (6) Goldstein, A. H.; Koven, C. D.; Heald, C. L.; Fung, I. Y. Biogenic carbon and anthropogenic pollutants combine to form a cooling haze over the southeastern United States. Proc. Natl. Acad. Sci. U.S.A. 2009, 106 (22), 8835–8840. (7) Carlton, A. G.; Pinder, R. W.; Bhave, P. V.; Pouliot, G. A. To what extent can biogenic SOA be controlled? Environ Sci. Technol. 2010, 44 (9), 3376–3380. (8) Hoyle, C. R.; Boy, M.; Donahue, N. M.; Fry, J. L.; Glasius, M.; Guenther, A.; Hallar, A. G.; Hartz, K. H.; Petters, M. D.; Petaja, T.; Rosenoern, T.; Sullivan, A. P. A review of the anthropogenic influence on biogenic secondary organic aerosol. Atmos. Chem. Phys. 2011, 11 (1), 321–343. (9) Hewitt, C. N. The atmospheric chemistry of sulphur and nitrogen in power station plumes. Atmos. Environ. 2001, 35 (7), 1155–1170. (10) Surratt, J. D.; Kroll, J. H.; Kleindienst, T. E.; Edney, E. O.; Claeys, M.; Sorooshian, A.; Ng, N. L.; Offenberg, J. H.; Lewandowski, M.; Jaoui, M.; Flagan, R. C.; Seinfeld, J. H. Evidence for organosulfates in secondary organic aerosol. Environ. Sci. Technol. 2007, 41 (2), 517–527. (11) Noziere, B.; Ekstr€om, S.; Alsberg, T.; Holmstr€om, S. Radicalinitiated formation of organosulfates and surfactants in atmospheric aerosols. Geophys. Res. Lett. 2010, 37. (12) Perri, M. J.; Lim, Y. B.; Seitzinger, S. P.; Turpin, B. J. Organosulfates from glycolaldehyde in aqueous aerosols and clouds: Laboratory studies. Atmos. Environ. 2010, 44 (21 22), 2658–2664. (13) Rudzi nski, K. J. Degradation of isoprene in the presence of sulphoxy radical anions. J. Atmos. Chem. 2004, 48 (2), 191–216. 8654
dx.doi.org/10.1021/es2011836 |Environ. Sci. Technol. 2011, 45, 8648–8655
Environmental Science & Technology (14) Surratt, J. D.; Chan, A. W. H.; Eddingsaas, N. C.; Chan, M. N.; Loza, C. L.; Kwan, A. J.; Hersey, S. P.; Flagan, R. C.; Wennberg, P. O.; Seinfeld, J. H. Reactive intermediates revealed in secondary organic aerosol formation from isoprene. Proc. Natl. Acad. Sci. U.S.A. 2010, 107 (15), 6640–6645. (15) Iinuma, Y.; B€oge, O.; Kahnt, A.; Herrmann, H. Laboratory chamber studies on the formation of organosulfates from reactive uptake of monoterpene oxides. Phys. Chem. Chem. Phys. 2009, 11 (36), 7985–7997. (16) Minerath, E. C.; Casale, M. T.; Elrod, M. J. Kinetics feasibility study of alcohol sulfate esterification reactions in tropospheric aerosols. Environ. Sci. Technol. 2008, 42 (12), 4410–4415. (17) Galloway, M. M.; Chhabra, P. S.; Chan, A. W. H.; Surratt, J. D.; Flagan, R. C.; Seinfeld, J. H.; Keutsch, F. N. Glyoxal uptake on ammonium sulphate seed aerosol: reaction products and reversibility of uptake under dark and irradiated conditions. Atmos. Chem. Phys. 2009, 9 (10), 3331–3345. (18) Minerath, E. C.; Schultz, M. P.; Elrod, M. J. Kinetics of the reactions of isoprene-derived epoxides in model tropospheric aerosol solutions. Environ. Sci. Technol. 2009, 43 (21), 8133–8139. (19) Paulot, F.; Crounse, J. D.; Kjaergaard, H. G.; Kroll, J. H.; Seinfeld, J. H.; Wennberg, P. O. Isoprene photooxidation: New insights into the production of acids and organic nitrates. Atmos. Chem. Phys. 2009, 9 (4), 1479–1501. (20) Wang, W.; Kourtchev, I.; Graham, B.; Cafmeyer, J.; Maenhaut, W.; Claeys, M. Characterization of oxygenated derivatives of isoprene related to 2-methyltetrols in Amazonian aerosols using trimethylsilylation and gas chromatography/ion trap mass spectrometry. Rapid Commun. Mass Spectrom. 2005, 19 (10), 1343–1351. (21) Iinuma, Y.; Muller, C.; Berndt, T.; B€oge, O.; Claeys, M.; Herrmann, H. Evidence for the existence of organosulfates from betapinene ozonolysis in ambient secondary organic aerosol. Environ. Sci. Technol. 2007, 41 (19), 6678–6683. (22) Eddingsaas, N. C.; VanderVelde, D. G.; Wennberg, P. O. Kinetics and products of the acid-catalyzed ring-opening of atmospherically relevant butyl epoxy alcohols. J. Phys. Chem. A 2010, 114 (31), 8106–8113. (23) Hatch, L. E.; Creamean, J. M.; Ault, A. P.; Surratt, J. D.; Chan, M. N.; Seinfeld, J. H.; Edgerton, E. S.; Su, Y.; Prather, K. A. Measurements of isoprene-derived organosulfates in ambient aerosols by aerosol time-of-flight mass spectometry—Part 1: Single particle atmospheric observations in Atlanta. Environ. Sci. Technol. 2011, 45 (12), 5105–5111. (24) Hansen, D. A.; Edgerton, E. S.; Hartsell, B. E.; Jansen, J. J.; Kandasamy, N.; Hidy, G. M.; Blanchard, C. L. The southeastern aerosol research and characterization study: Part 1—Overview. J. Air Waste Manage. Assoc. 2003, 53 (12), 1460–1471. (25) Gard, E.; Mayer, J. E.; Morrical, B. D.; Dienes, T.; Fergenson, D. P.; Prather, K. A. Real-time analysis of individual atmospheric aerosol particles: Design and performance of a portable ATOFMS. Anal. Chem. 1997, 69 (20), 4083–4091. (26) Edgerton, E. S.; Hartsell, B. E.; Saylor, R. D.; Jansen, J. J.; Hansen, D. A.; Hidy, G. M. The southeastern aerosol research and characterization study, part 3: Continuous measurements of fine particulate matter mass and composition. J. Air Waste Manage. Assoc. 2006, 56 (9), 1325–1341. (27) Hennigan, C. J.; Bergin, M. H.; Russell, A. G.; Nenes, A.; Weber, R. J. Gas/particle partitioning of water-soluble organic aerosol in Atlanta. Atmos. Chem. Phys. 2009, 9 (11), 3613–3628. (28) Chan, M. N.; Surratt, J. D.; Claeys, M.; Edgerton, E. S.; Tanner, R. L.; Shaw, S. L.; Zheng, M.; Knipping, E. M.; Eddingsaas, N. C.; Wennberg, P. O.; Seinfeld, J. H. Characterization and quantification of isoprene-derived epoxydiols in ambient aerosol in the southeastern United States. Environ. Sci. Technol. 2010, 44 (12), 4590–4596. (29) Beven, J. L.; Kimberlain, T. B. National Hurricane Center tropical cyclone report: Hurricane Gustav. 2009. (30) Brown, D. P.; Kimberlain, T. B. National Hurricane Center tropical cyclone report: Hurricane Hanna. 2008. (31) Beven, J. L. National Hurricane Center tropical cyclone report: Tropical Storm Bertha. 2002.
ARTICLE
(32) Durre, I.; Vose, R. S.; Wuertz, D. B. Overview of the Integrated Global Radiosonde Archive. J. Clim. 2006, 19 (1), 53–68. (33) Duncan, B. N.; Stelson, A. W.; Kiang, C. S. Estimated contribution of power-plants to ambient nitrogen-oxides measured in Atlanta, GA in August 1992. Atmos. Environ. 1995, 29 (21), 3043–3054. (34) Neubauer, K. R.; Johnston, M. V.; Wexler, A. S. Humidity effects on the mass spectra of single aerosol particles. Atmos. Environ. 1998, 32 (14 15), 2521–2529. (35) Surratt, J. D.; Lewandowski, M.; Offenberg, J. H.; Jaoui, M.; Kleindienst, T. E.; Edney, E. O.; Seinfeld, J. H. Effect of acidity on secondary organic aerosol formation from isoprene. Environ. Sci. Technol. 2007, 41 (15), 5363–5369. (36) Chang, W. L.; Bhave, P. V.; Brown, S. S.; Riemer, N.; Stutz, J.; Dabdub, D. Heterogeneous Atmospheric chemistry, ambient measurements, and model calculations of N2O5: A review. Aerosol Sci. Technol. 2011, 45 (6), 665–695. (37) Brown, S. S.; Ryerson, T. B.; Wollny, A. G.; Brock, C. A.; Peltier, R.; Sullivan, A. P.; Weber, R. J.; Dube, W. P.; Trainer, M.; Meagher, J. F.; Fehsenfeld, F. C.; Ravishankara, A. R. Variability in nocturnal nitrogen oxide processing and its role in regional air quality. Science 2006, 311 (5757), 67–70. (38) Rollins, A. W.; Kiendler-Scharr, A.; Fry, J. L.; Brauers, T.; Brown, S. S.; Dorn, H. P.; Dube, W. P.; Fuchs, H.; Mensah, A.; Mentel, T. F.; Rohrer, F.; Tillmann, R.; Wegener, R.; Wooldridge, P. J.; Cohen, R. C. Isoprene oxidation by nitrate radical: Alkyl nitrate and secondary organic aerosol yields. Atmos. Chem. Phys. 2009, 9 (18), 6685–6703. (39) Darer, A. I.; Cole-Filipiak, N. C.; O’Connor, A. E.; Elrod, M. J. Formation and stability of atmospherically relevant isoprene-derived organosulfates and organonitrates. Environ. Sci. Technol. 2011, 45 (5), 1895–1902. (40) Surratt, J. D.; Gomez-Gonzalez, Y.; Chan, A. W. H.; Vermeylen, R.; Shahgholi, M.; Kleindienst, T. E.; Edney, E. O.; Offenberg, J. H.; Lewandowski, M.; Jaoui, M.; Maenhaut, W.; Claeys, M.; Flagan, R. C.; Seinfeld, J. H. Organosulfate formation in biogenic secondary organic aerosol. J. Phys. Chem. A 2008, 112 (36), 8345–8378. (41) Chameides, W. L.; Lindsay, R. W.; Richardson, J.; Kiang, C. S. The role of biogenic hydrocarbons in urban photochemical smog— Atlanta as a case-study. Science 1988, 241 (4872), 1473–1475. (42) Paulot, F.; Crounse, J. D.; Kjaergaard, H. G.; Kurten, A.; St Clair, J. M.; Seinfeld, J. H.; Wennberg, P. O. Unexpected epoxide formation in the gas-phase photooxidation of isoprene. Science 2009, 325 (5941), 730–733. (43) Blanchard, C. L.; Hidy, G. M.; Tanenbaum, S.; Edgerton, E. S. NMOC, ozone, and organic aerosol in the southeastern United States, 1999 2007: 3. Origins of organic aerosol in Atlanta, Georgia, and surrounding areas. Atmos. Environ. 2011, 45 (6), 1291–1302. (44) Pratt, K. A.; Prather, K. A. Aircraft measurements of vertical profiles of aerosol mixing states. J. Geophys. Res. 2010, 115.
8655
dx.doi.org/10.1021/es2011836 |Environ. Sci. Technol. 2011, 45, 8648–8655
ARTICLE pubs.acs.org/est
Natural 37Ar Concentrations in Soil Air: Implications for Monitoring Underground Nuclear Explosions Robin A. Riedmann*,† and Roland Purtschert† †
Climate- & Environmental Physics, Physics Institute, University of Berne, Switzerland, Sidlerstrasse 5, CH-3012 Berne, Switzerland
bS Supporting Information ABSTRACT: For on-site inspections (OSI) under the Comprehensive NuclearTest-Ban Treaty (CTBT) measurement of the noble gas 37Ar is considered an important technique. 37Ar is produced underground by neutron activation of Calcium by the reaction 40Ca(n,α)37Ar. The naturally occurring equilibrium 37 Ar concentration balance in soil air is a function of an exponentially decreasing production rate from cosmic ray neutrons with increasing soil depth, diffusive transport in the soil air, and radioactive decay (T1/2: 35 days). In this paper for the first time, measurements of natural 37Ar activities in soil air are presented. The highest activities of ∼100 mBq m3 air are 2 orders of magnitude larger than in the atmosphere and are found in 1.52.5 m depth. At depths >8 m 37Ar activities are <20 mBq m3 air. After identifying the main 37Ar production and gas transport factors the expected global activity range distribution of 37Ar in shallow subsoil (0.7 m below the surface) was estimated. In high altitude soils, with large amounts of Calcium and with low gas permeability, 37Ar activities may reach values up to 1 Bq m3.
’ INTRODUCTION The verification of the CTBT (Comprehensive Nuclear-Test-Ban Treaty) relies on analytical tools for the identification and localization of clandestine nuclear test explosions. Subsurface tests produce radioactive noble gas isotopes which may migrate through the underground and eventually reach the shallow soil zone and the atmosphere. Measurements of these isotopes in quantities above natural background levels are therefore strong indicators of a nuclear test. The radioactive noble gas isotope 37Ar is a definitive and unambiguous indicator of an underground nuclear explosion (UNE). In the atmosphere 37Ar is produced by neutron tion of 40 Ar, 40Ar(n,4n)37Ar, and in the lithosphere due to neutron activation of Calcium by the reaction 40Ca(n,α)37Ar. The anthropogenic 37Ar background is very low unless there are recurring neutron fluxes in the vicinity of the targets 40Ar or 40Ca. The relatively long half-life of 35 days compared to short-lived Xe isotopes (131 mXe, 133Xe, and 133 mXe) enhances the likelihood that detectable concentrations reach the shallow soil zone. In order to distinguish between natural and artificially elevated 37Ar concentrations it is crucial to determine the location-specific natural 37 Ar activity range in soils and rocks and its controlling factors. In this article we will present measured natural 37Ar activities in soil air. ’ MATERIALS AND METHODS Soil-air samples have been extracted at different locations and various depths below the surface (see Table 1). In Switzerland, bore holes of 7 cm diameter were drilled. The locations mainly had been chosen according to soil thickness, water table, and Calcium content.1 Existing bore holes were accessed in Beziers and Fontainebleu, France,2,3 M€unchen, Germany,4 and in Arizona, USA.5 r 2011 American Chemical Society
A filter made from metallic sinter material was mounted onto a copper tube of 0.2 cm inner diameter and positioned at the respective sampling depth. The sampling interval was filled with 10 cm gravel and sealed with a betonite layer of at least 20 cm. The remaining depth of the bore hole was filled up with the originally excavated material. A sampling site 10 km northeast of Zurich has been set up, and two bore holes (several meters spacing between them) with three sampling depths for each were installed. Each sampling depth was filled and sealed as described above. Soilair samples were taken after a recovery period of several weeks to allow for a re-equilibration with the surrounding soil matrix. Soil gas was then extracted with a pump at a flow rate of 0.40.8 L min1. The total sample volume of 40100 L was compressed in 20 dm3 steel flasks. For a gas-filled porosity of 0.10.3 and a homogeneous soil the radius of the gas-filled pore space of a soil sphere drained is in a range ∼3262 cm. This implies an uncertainty in the sampling depth of ∼50 cm and a corresponding minimal sampling depth in order to prevent contamination with ambient air. O2 and CO2 were continuously measured in the field with a portable multigas analyzer (airTOX, Fresenius Umwelttechnik). 222 Rn was monitored with a commercial Radon gas monitor (ALPHAGUARD Genitron Instruments, Frankfurt, Germany).6 In the laboratory 200700 cm3 STP Argon were purified from the soil-air samples in a GC based separation line. The specific 37 Ar activity is measured by ultralow-level β-counting (internal proportional counting of Auger electrons) in an underground laboratory.7 Received: April 12, 2011 Accepted: August 31, 2011 Revised: July 6, 2011 Published: August 31, 2011 8656
dx.doi.org/10.1021/es201192u | Environ. Sci. Technol. 2011, 45, 8656–8664
Environmental Science & Technology
ARTICLE
Table 1. Sampling Locations, Measured Natural 37Ar Activities in Soils, and Calculated Minimum Detectable Activities name Baltenswil
coordinates 47.43°N 8.64°E
altitude
calcium content
depth
CO2
Rn
(m asl)
(%)
(m)
(%)
(kBq m3)
65
2.1
570
4.1
46.52°N 7.31°E
830
13.4
Ar
MDA (mBq m3)
-
-
77.9 ( 9.0
5.5
4.2 ( 0.7
-
119.3 ( 29.0
24.2
-
-
108.0 ( 11.5
8.4
4.7 ( 0.5
-
59.9 ( 13.6
12.2
5.7
-
-
68.9 ( 19.1
19.7
7.5
2.1 ( 0.2
-
46.7 ( 22.2
29.0
4.5 ( 0.3
-
35.9 ( 5.1 24.2 ( 4.3
3.9 4.0
-
-
9.9 ( 3.0
3.7
-
-
13.2
4.4
1.2 ( 0.3
-
19.1 ( 3.8
3.3
2.0 ( 0.1
66.2 ( 7.5
27.9 ( 3.4
2.6
5.8 ( 0.2
61.4 ( 13.9
24.6 ( 3.9
2.9
9.6 Belp 1
37
(mBq m3)
1.2
Belp 2
46.52°N 7.31°E
830
12.4
2.4
1.2 ( 0.5
-
38.0 ( 4.2
2.6
Burgdorf
47.03°N 7.38°E
540
7.4
1.7
2.0 ( 0.2 3.6 ( 0.4
20.6 ( 1.0 -
28.6 ( 4.3 105.4 ( 12.9
3.5 7.9
4.4 ( 0.2
25.7 ( 1.9
64.9 ( 6.9
4.9
M€ohlin
47.34°N 7.53°E
340
1.4
4.0
5.8 ( 1.0
-
35.7 ( 5.1
4.1
5.5 ( 0.8
-
18.3 ( 4.7
4.3
4.9 ( 0.1
87.0 ( 14.1
28.6 ( 7.9
6.6
0.4 ( 0.2
-
<5.0
5.0
0.8 ( 0.2
27.9 ( 3.0
<3.1
3.1
0.6 ( 0.4 -
10.9 ( 3.2 -
7.9 ( 2.6 <11.4
2.8 11.4
Tiefenau Weissenstein Beziers
Fontainebleu M€unchen
Arizona
46.58°N 7.28°E 47.16°N 7.29°E 43.23°N 3.12°E
48.24°N 2.42°E 48.08°N 11.35°E
33.17°N 111.04°W
540 980 80
110 580
1300
1.1
1.4
26.7 1.0
0.6 2.3 4.8
2.2 ( 0.1
-
3.1 ( 1.9
2.9
15.8
1.7 ( 0.1
-
<6.2
6.2
2.5
<0.2
-
<1
1.0
10.7
3.1
-
<1.4
1.4
90.0
2.0
-
-
16.0 ( 4.0
-
3.5
-
-
22.0 ( 4.5
-
-
9.0 4.0
1.1
-
3.0 ( 1.0 36.7 ( 11.7
-
130.0
1.1
-
<0.6
0.6
1.0
In oder to determine the Calcium content, soil samples from the respective filter depth were dried for several hours at 105 °C and ground to fine powder. The soil was then digested in aqua regia at 90 °C.8 The elemental composition was analyzed by ICP/MS at “Activation Laboratories” in Ontario, Canada.8 37 Ar (T1/2 = 35.04 d)9 decays by electron capture (EC) with a decay energy of 2.8 keV.10 The energy spectrum is recorded with a 7-bit Multi-Channel Analyzer (MCA) with a linear energy range of 035 keV. The specific 37Ar activity [mBq m3 air] is calculated from the net peak count rate P [cpm] according to P 1 f1 f2 c VAr 3 Y 3 3 3 λ37 A r 3 tm f1 ¼ 1 expðλ37 A r 3 tm Þ f2 ¼ expðλ37 A r 3 tlag Þ ½37 Ar ¼
ð1Þ
between sampling date and start of the measurement (tlag) and during the measurement (tm), respectively. The constant c converts to mBq m3 air [c = 107/60 (mBq m3 air)/(cpm cm3Ar)]. It is assumed that CO2 is stochiometrically produced by O2 consumption as a result of aerob respiration and therefore CO2 + O2 = 20.8% at all times and for all sampling depths . This entails an argon concentration of 1% in the soil gas. The very low specific activity of 37Ar (natural 37Ar/Ar∼1020) requires a minimal background count rate (BG [cpm]). In the underground laboratory of the University of Berne the BG is in the order of 3 103 cpm for the 16 cm3 counter and 2 102 cpm for the 100 cm3 counter for typical filling pressures of 1022 bar and 422 bar, respectively. 39Ar is the main background source apart from the intrinsic activity of the counters. The modern 39Ar activity11 of 0.11 dpm/LAr contributes to 3040% of the BG depending on the Argon volume. The statistical 2σ error of P in cpm is given by12,13 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2σP ¼ pffiffi 3 P þ 2BG t
where VAr is the Argon volume, Y is the counting efficiency (∼0.70), and f1 and f2 are decay corrections for the time lag 8657
ð2Þ
dx.doi.org/10.1021/es201192u |Environ. Sci. Technol. 2011, 45, 8656–8664
Environmental Science & Technology
ARTICLE
Figure 1. Measured natural Ar activities in soils for different locations in Switzerland, France, Germany, and USA. Highest 37Ar activities are observed in 1.52.5 m soil depth. 37
where t is the counting time. Detection of a peak is claimed with 95% confidence level when13 P > PDL ¼ 1:6449 3 2 3 σP ¼ 0
pffiffiffi ¼ 3:29 3 2 3
rffiffiffiffiffiffiffi BG t
ð3Þ
’ DISCUSSION
where PDL is the count rate of the detection limit in cpm. The Minimum Detection Activity (MDA) in mBq m3 air is finally calculated by substituting eq 3 in eq 113 f1 f2 MDA ¼ 4:65 3 c 3 3 3 VAr 3 Y
rffiffiffiffiffiffiffi BG t
Ar production rate Rtg [atoms m3 s1] can be calculated according to Z Emax σðEÞΦðEÞdE ð5Þ Rtg ¼ Ntg Production. The
ð4Þ
The resulting MDAs (Table 1) are in the range 0.629.0 mBq m3 air. For example, the MDA for a sample measured in the 16 cm3 counter with 230 cm3 STP of Argon, a BG of 1.7 103 cpm, a counting time of 28,400 min (f2 = 1.21), and a lag time of 4 days (f1 = 1.08) amounts to 1.7 mBq m3 air. The total 2σ uncertainty of the activity measurement [37Ar] including the 5% uncertainty of the counting yield is calculated by error propagation of eq 1.
’ RESULTS AND DISCUSSION Measurements. Measured natural
Figure 2. Reaction cross-section for the 40Ca(n,α)37Ar reaction. For thermal neutrons a 1/v law (where v is the neutron velocity) was used. In the epithermal region the cross-section is not well-known but probably smallest, σ < 105 barn. The cross section for 1 MeV < En < 20 MeV is well documented, and the values extracted from JENDL 3.3 in ref 29 were used. At energies En > 20 MeV the literature on the cross-section is scarce, and the values given in ref 30 were used. The reaction crosssection is largest for energies 1 MeV < En < 20 MeV where most literature from experimental data is given. The gray dotted line gives the relative neutron spectrum of ref 23 at 430 m asl used here, normalized to the maximum flux at 24 MeV.
37
Ar activities in soil air range over 2 orders of magnitude between <3.1 to 120 mBq m3 air (Figure 1). The lowest values agree, within the uncertainties, with the atmospheric 37Ar background activity of 1.2 ( 0.5 mBq m3 air measured in the period 20032010 and with previously published data (0.330.42 mBq m3 air,14 0.63.8 mBq m3 air15). Highest 37Ar activities are typically observed in the depth range 1.52.5 m. Repeated measurements at the same location and depth vary within a factor <2 and mainly in shallow depths (Figure 1). This probably reflects seasonal variations of the soil permeability. With increasing soilwater content due to precipitation the gas transport and porosity decrease which generally leads to higher 37Ar activities.
37
0
where Ntg is the concentration of the target nuclides [atoms m3], σ(E) is the energy-dependent reaction cross-section [barn = 1024 cm2], and Φ(E) is the differential neutron flux in eV [neutrons eV1 cm2 s1]. At production and decay equilibrium Rtg equals the activity concentration [37Ar]. The main 37Ar production channel in the atmosphere is the 40 Ar(n,4n)37Ar spallation reaction. With eq 5 calculated tropospheric 37Ar activities assuming a well-mixed air column range between 0.4 and 1.2 mBq m3 air.16,17 Atmospheric neutrons originate from interactions of primary galactic cosmic rays (GCRs) with Oxygen and Nitrogen nuclei in the air.18 Several studies have been published in which measured spectra at various altitudes were presented.1923 As a reference flux we used the data from ref 23, measured in Japan, at 430 m above sea level and 2 m above ground. A peak at thermal energies can be observed, reflecting neutron thermalization through the atmosphere and backscattering from the earth.23 A second peak can be observed around a few MeV (evaporation peak), and a third peak is found for energies around 100 MeV (Figure 2). The differential neutron flux and thus the 37 Ar production rates vary with location (latitude, altitude, overhead shielding) and time (season, solar activity, and atmospheric pressure).24,25 The amplitude of neutron flux variations induced by changing solar activity (∼25%) in the 11year solar cycle decreases with increasing atmospheric depth.2628 The thermal peak <1 eV remains almost constant in time.21 The neutron flux maximum around 100 MeV was shown to vary by 8658
dx.doi.org/10.1021/es201192u |Environ. Sci. Technol. 2011, 45, 8656–8664
Environmental Science & Technology
ARTICLE
Figure 3. GCR scalings depending on the latitude, altitude, and depth below the surface. The scaling of the latitude is shown for the meridian and at sea level. The geographic coordinates are converted into geomagnetic inclination, and the scalings are therefore not symmetric. Latitudinal scalings vary by a factor <2. GCR scalings increase by a factor >20 from 0 to 4 km altitude. Depth dependent GCR neutron scalings decreases exponentially with an attenuation length of 148 g cm2 for an average dry terrestrial composition. The muon flux is scaled to its relative importance to produce 37Ar at the surface (<4% of the total 37Ar production)42 and decreases exponentially with an attenuation length of 247 g cm2.33.
about 30% from day to day.21 These temporal variations in cosmic ray intensity are not the focus of this paper and are therefore neglected in the following. Differences in cosmic-ray intensity at an arbitrary sampling site compared to the reference site are accounted for by multiplying the reference flux by a scaling factor.31,32 More recent scaling models31,33 are significantly more sophisticated than earlier ones.32,34 Here the model of ref 33 was used, which assumes a neutron energy spectrum independent of altitude. Comparison of the global cosmic ray scaling according to ref 33 with values measured with neutron monitors reported in ref 35 shows a difference of 3.6% for locations below 800 m asl and 9.9% for locations above 800 m asl. Although for exposure dating this may be unacceptably large, it is a reasonable approximation for the 37 Ar production in the vadose zone. The meridionally averaged latitudinal and the altitudinal scaling factors are shown in Figure 3. The latitudinal scaling decreases by a factor of 2 between the poles and the equator. The asymmetric shape of latitudinal scaling is due to the influence of the nondipole contributions to the geomagnetic field. Cosmic ray intensities decrease by a factor >20 with decreasing altitude from 4 km to sea level. For constant composition, at the soil surface the cosmic ray flux Φ(z) and thus the 37Ar production rate decreases exponentially with increasing depth z (g cm2), e.g.3638 z ΦðzÞ ¼ Φð0Þ 3 exp ð6Þ Λ The attenuation length Λ depends on the bulk rock or soil density, its elemental composition, and on the gas- and waterfilled porosity. For a density F of 2 g cm3 and an average dry terrestrial composition (47.3% O, 2.5% Na, 4.0% Mg, 6% Al, 29% Si, 5.0% Ca, and 6.0% Fe, 0.2% H) an attenuation length Λ of 148 g cm2 was reported.39 This estimate was used in the calculations here (Figure 3). In the lithosphere 37Ar is produced by neutron activation of Calcium 40Ca(n,α)37Ar and Potassium 39K(n,p2n)37Ar. However,
because the reaction cross-section for the Ca reaction is a factor of 6 higher40 and because Ca/K ∼100 for the investigated soils, Calcium dominates the 37Ar production (the average Ca/K ratio in the continental lithosphere is 1.23.941). The differential reaction cross sections for the 40Ca(n,α)37Ar reaction shown in Figure 2 were used for the production calculations. The contribution of the muon reaction 39K(μ,2n)37Ar39 and spallation with K, Fe, and Ti are an order of magnitude lower than the Ca reaction (Figure 3).40 Muon induced reactions are therefore neglected in the following. Only a fraction of the produced 37Ar will emanate from the site of production into the air-filled pore volume θg.43 The emanation coefficient ε is defined as the ratio of the measured concentration in the pores (Cair) to the maximum possible concentration in the mineral grains Crock Cair ¼ Crock 3 ε 3
1 θg θg
ð7Þ
In recoil emanation the energy of the 40Ca(n,α)37Ar reaction is partly transferred as kinetic energy to the daughter atom (37Ar) that is displaced in the mineral over a distance in the order of 0.010.7 μm.9,4446 The daughter atom escapes the solid phase if it is located within recoil distance from the solid boundary. The emanation coefficient therefore increases with increasing specific surface area. Atoms which have not escaped from the particle by recoil may still enter the pore space by diffusion.47 However, the effective diffusion coefficient Dc,eff of Argon in crystal lattice is <1026 m2 s1.48,49 The diffusion path length L over the 37Ar lifetime 1/λ 50 is L = (Dc,eff/λ)0.5 < 0.2 nm. The emanation by diffusion is therefore negligible compared with recoil emanation. The emanation coefficient of 37Ar in water-saturated aquifers is estimated to be in the range 0.122.2%.43 In the gas-filled pore space, the emanation coefficient is assumed to be slightly higher, 27%, due to the larger specific surface area of 37Ar in the vadose zone compared to the saturated zone of an aquifer. 8659
dx.doi.org/10.1021/es201192u |Environ. Sci. Technol. 2011, 45, 8656–8664
Environmental Science & Technology
ARTICLE
Table 2. Parameters Used for the Barometric Pumping Included in the Simplified Model To Interprete Measured 37 Ar Activities at the Sampling Site 10 km Northeast of Zurich
Figure 4. Measured natural 37Ar activities in a soil profile in Switzerland. The relatively large gas volumes extracted (4070 L) imply an uncertainty of the sampling depth of 50 cm in radius. 37Ar activities are highest at ∼2 m depth and decrease with increasing depth. The analytical solution of eq 11 was done with generalized D = Dg,eff (gray band) and with D = 2 Dg,eff (cyan band). The calculated diffusion profile agrees with the measured natural 37Ar activities.
Transport.
37
Ar concentration gradients at the soilatmosphere interface (Figure 3) are averaged out by diffusive and advective-dispersive gas transport processes.51 The effective molecular diffusion coefficient Dg,eff [m2 s1] was calculated from refs 52 and 53 Dg, eff ¼ Dg, 0 3 τ 3 θg
ð8Þ
where τ is the dimensionless tortuosity 0 < τ <1, and Dg,0 is the molecular diffusion of Argon in air at ambient temperature (Ar: 1.8 105 m2 s1).54,55 Diffusion of 37Ar in the liquid phase is 4 orders of magnitude slower than in air,56 and partitioning of 37Ar from the gas phase into the liquid phase is negligible.57 Advection is driven by pressure gradients caused by cyclic atmospheric pressure variations because of high- (pmax) and lowpressure (pmin) fronts.58 The mean gas flow velocity |V(z)| at depth z during a pressure cycle with period T is given by jV ðzÞj =
Δp 3 2 ðZ zÞ p0 3 T 3
ð9Þ
were p0 is the mean atmospheric pressure, Δp = (pmaxpmin) is the pressure perturbation, and Z is the thickness of the porous layer.58 Close to the surface (at depths zp < Δp/p0•Z) this vertical air movement ventilates the soil column more effectively than by diffusion alone. The up and down movement of air in depths greater than zp causes mechanical dispersion that can be included in the effective diffusion coefficient D51,5860 D ¼ Dg, eff þ α 3 jV j
ð10Þ
where α is the mechanical dispersivity [m].58 The steady state 37Ar concentration balance in soil air as a function of depth can now be expressed by the following onedimensional transport equation RðzÞ ¼ D 3
∂2 C g þ λCg ðzÞ ∂z2
ð11Þ
thickness of unsaturated zone
Z
1015 m
mean pressure perturbation dispersivity75
Δp α
020 mb 0.10.15 m
period76
T
03 days
emanation coefficient
ε
0.040.07
where R(z) = Rtg Cair F is the 37Ar production rate at depth z in the gas-filled pore space [Bq m3 air], Cg is the 37Ar concentration in the gas phase [atoms m3 air], and λ is the 37Ar decay constant [s1]. The boundary conditions are C(z = 0) = C0 and C(z = ∞) = C∞, where C0 and C∞ are the atmospheric concentration and the concentration at inifinite depth, respectively. The latter is determined by the subsurface nucleogenic 37Ar production.61 The 37Ar activity at depth z is calculated analytically from eq 11 (Figure 4). 37 Ar Depth Profiles. In Figure 4 a single bore hole 37Ar depth profile is shown. The sampling site is located approximately 10 km northeast of Zurich, 570 m above sea level. The soil consists of fine luvisol (para-brown soil) in 04 m depth and of silty sandy ground moraine in 613 m depth.62,63 37Ar samples were taken between 210 m depth at different points in time throughout the year. The measured 37Ar activities are interpreted by means of the simplified model represented by eq 11 (Figure 4). The neutron flux at the soil surface was calculated according to the scaling model,33 and the neutron flux was attenuated at depth z according to eq 6. A temporally averaged gas-filled porosity of 0.170.236467 and a tortuosity in the range 0.10.26871 were assumed. These estimates result in an effective diffusion coefficient Dg,eff of 1.88.1 107 m2 s1. In a first run barometric pumping was neglected (D = Dg,eff). An emanation coefficient in the range 0.020.06 was chosen. In a second run the effect of barometric pumping was included by means of eqs 9 and 10 and the assumptions given in Table 2, leading to a D = 1.42.3 Dg,eff and a ventilated zone zp of 0.20.3 m. In general, the modeled diffusion profiles agree well with the measured 37Ar activities (Figure 4). The highest 37Ar activities in the calculated depth profile are found in 1.52.0 m. The qualitative agreement between the measured and modeled 37Ar concentrations at this sampling location (but also for all 37Ar activities shown in Figure 1) provides evidence for the permissibility of the underlying basic assumptions of eq 11: The effects of barometric pumping (mechanical dispersion and the ventilation effect) are within the uncertainties of our measurements. However at other locations with different parameters than here (Table 2) and other geological settings72 barometric pumping may become a relevant process. Natural 37Ar activities in soils are largely determined (i) by depth-dependent 37Ar production from neutron activation of Calcium with cosmic ray neutrons, (ii) by radioactive decay, and (iii) by diffusion (or dispersion) in the shallow soil-column. Temporal variations (Figure 4) are probably linked with soilwater content variations which affect both tortuosity and porosity and thus Dg,eff. Variance Analysis. For a more general assessment of 37Ar concentrations in soil air taking into account the entire data set in Table 1, with typically only one sampling depth per location, 37Ar activities were correlated with four crucial parameters p14: 8660
dx.doi.org/10.1021/es201192u |Environ. Sci. Technol. 2011, 45, 8656–8664
Environmental Science & Technology Altitude, sampling depth, Calcium content of the soil matrix, and CO2 concentration. The latter is, like 37Ar, produced in the subsurface but by completely independent production mechanisms due to aerob metabolism of soil biota and plants. However, CO2 is affected by the same transport as 37Ar. A high correlation between CO2 and 37Ar therefore implies that the concentrations are mainly determined by the effective diffusion coefficient. Despite the sparse amount of data a simple variance analysis was carried out. The remaining dimensionless variance of all 37Ar concentrations after normalizing with parameter pi and the
Figure 5. Variance analysis of the measured natural 37Ar activities in soils. The variance is given as σ̅ 2i = σ2/median2. The numbers above each bar are the numbers of normalized 37Ar activities available in Table 1 for the calculation. The error bars indicate the uncertainty of the 37 Ar activities.
ARTICLE
median is given as σ̅ 2 i ¼
σ2 ð37 Ar=pi Þ median2 ð37 Ar=pi Þ
ð12Þ
A high σ̅ 2i indicates a low correlation with parameter pi and vice versa. σ̅ 2i were calculated individually for three depth ranges: 01.75 m, 1.754.5 m, and for depths >4.5 m and the results are shown in Figure 5. In the shallow depth range (<1.75 m) the highest correlations (lowest σ̅ 2i) can be observed for CO2 and sampling depth. This is consistent with the conclusion of the previous section that 37Ar concentrations at the soilatmosphere interface are mainly diffusion-dispersion controlled. Productionrelevant factors (altitude and Ca concentration) explain only a small part of the variance. At depths >1.75 m σ̅ 2i for the sampling depth and CO2 increases and σ̅ 2i for the altitude decreases compared with depths <1.75 m. At depths >4.5 m 37Ar is most sensitive (lowest σ̅ 2i) to altitude and Ca content and insensitive to CO2. The importance of diffusion decreases with increasing depth due to decreasing 37Ar concentration gradients with depth. σ̅ 2i of the sampling depth decreases with increasing depth due to (i) the 37 Ar peak maximum at depths 1.52.5 m and (ii) decreasing importance of the neutron flux from cosmic rays compared with the lithogenic neutron flux. Altitude is most important in the mid-depth range (1.754.5 m) where σ̅ 2i is larger than in <1.75 m depth and where the neutron flux from cosmic rays is still larger than the lithogenic neutron flux. The importance of Calcium content generally increases with increasing depth since 37 Ar transport and cosmic ray neutrons decrease with increasing depth. Estimation of the Global 37Ar Production in Shallow Subsoils. The main controlling factors of 37Ar activities identified above can be used to constrain the range of natural 37Ar activities in soils to be expected on a global scale. For this purpose
Figure 6. Global scaling of 37Ar production at 70 cm soil depth, normalized to the site of the 37Ar depth profile in Switzerland (47.26° N, 8.39° E, 570 m asl) and 1.45% Ca. The resolution is 100 100 . GCR scaling was done according to ref 33, the geomagnetic inclination was calculated with the IGRF-10 model, the altitude was taken from NOAA, announcement 88-MGG-02, and the calcium content was extracted from the Harmonised World Soil Database (HWSD). 8661
dx.doi.org/10.1021/es201192u |Environ. Sci. Technol. 2011, 45, 8656–8664
Environmental Science & Technology the location, Ca concentration, and activities of the site where the soil-air profile was measured (see Figure 4) are used as reference (46.95° N, 7.45° E, 570 m asl, [CaHWSD] = 1.45%), and the 37Ar activities are extrapolated based on the global distribution of the main controlling factors. Global scalings of 37Ar production in the shallow subsoil were derived using the GCR flux scaled to the respective geomagnetic coordinates and altitude and using the Ca content in the subsoil of the respective location. The altitude was extracted from the geographical topography available at the homepage of the National Geophysical Data Center, NOAA US Department of Commerce.73 The subsoil (i.e., 0.70 mwe depth) Ca content was extracted from the Harmonised World Soil Database (HWSD). The data were compiled by the Land Use Change and Agriculture Program of IIASA.74 The resolution of the grid map is 100 100 . In the equatorial region this equals approximately 18.6 18.6 km or 344 km2. Both the altitude and the Calcium content may vary by up to 2 orders of magnitude in such a grid cell. Thus, this map is not specifically designed to predict the 37Ar production of a single specific site. However, the soil type and the subsoil Ca content in the database are a useful approximation and a reference point of the region. The higher the scaling factor, the higher the probability for a high subsoil 37Ar concentration. Scaling factors in Figure 6 vary by 3 orders of magnitude, and the highest scalings are an order of magnitude larger than in the reference site. In other words, if 1000 samples are extracted and analyzed in a grid cell (100 100 ) with the highest scaling, one will find on average 10 times higher 37Ar activities than in the reference area where ∼100 mBq m3 were measured. Thus, in these areas, in 24 m depth a natural 37Ar background of up to 1000 mBq m3 is to be expected. At depths >8 m, where cosmic ray neutrons have decreased by 3 orders of magnitude, natural 37Ar activities are likely less than 50 mBq m3. In general, the highest 37Ar activities in the subsoil are expected around 30°60° northern latitude and around 20° 40° southern latitude. The highest scalings are almost always a combination of mountainous areas (high altitudes) and high average Calcium content. Intermediate scalings occur in semiarid areas. Almost no 37Ar production in 70 cm soil depth is expected above 60° N (boreal zone, Tundra) and in the equatorial regions (tropical rain forests) as there is almost no Calcium is these soils.
’ ASSOCIATED CONTENT
bS
Supporting Information. Tables S1 and S2 and Figures S1S12. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT The project was funded and supported by the Comprehensive Test Ban Treaty Organisation CTBTO. We thank J€org R€uedi and Joachim Elsig for providing preliminary data and interpretation. H.-H. Loosli contributed to the measurements of 37Ar in the
ARTICLE
atmosphere. The authors also wish to thank Charles Carrigan for fruitful discussions and comments.
’ REFERENCES (1) Walthert, L.; Zimmermann, S.; Blaser, P.; Luster, J. Waldb€oden der Schweiz; Hep Verlag: 2004. (2) Corcho, J. A.; Purtschert, R.; Barbecot, F.; Chabault, C.; Rueedi, J.; Schneider, V.; Aeschbach-Hertig, W.; Kipfer, R.; Loosli, H. H. Constraining the age distribution of highly mixed groundwater using 39 Ar: A multiple environment: tracer (3H/3He, 85Kr, 39Ar, and 14C) study in the semiconfined Fontainebleau Sands Aquifer (France). Water Resour. Res. 2007, 43, W03427. (3) Gillon, M.; Barbecot, F.; Gibert, E.; Corcho Alvarado, J. A.; Marlin, C.; Massault, M. Open to closed system transition traced through the TDIC isotopic signature at the aquifer recharge stage, implications for groundwater 14C dating. Geochim. Cosmochim. Acta 2009, 73, 6488–6501. (4) European Lysimeter Platform. http://www.lysimeter.at/HP_EuLP/ web/germany/de43.html (accessed month day, year). (5) R€uedi, J. Estimating groundwater recharge and dynamics in arid regions using environmental tracers: application on Jullemeden Basin (Niger). Ph.D. Dissertation, University of Bern, 2002. (6) Lehmann, B. E.; Lehmann, M.; Neftel, A.; Tarakanov, V. Radon222 monitoring of soil diffusivity. Geophys. Res. Lett. 2000, 27, 3917– 3920. (7) Forster, M.; Loosli, H. H. In Current techniques for measuring the activity of Ar-37 and Ar-39 in the environment, Isotopes of Noble Gases as Tracers in Environmental Studies, International Atomic Energy Agency: Vienna, 1989. (8) Actlabs trace element geochemistry, “Ultratrace 1”.http:// www.actlabs.com/page.aspx?page=502&app=226&cat1=549&tp=12&lk= no&menu=64 (accessed month day, year). (9) Renne, P. R.; Norman, E. B. Determination of the half-life of 37Ar by mass spectrometry. Phys. Rev. C: Nucl. Phys. 2001, 63. (10) Barsanov, V. I.; Dzhanelidze, A. A.; Zlokazov, S. B.; Kotelnikov, N. A.; Markov, S. Y.; Selin, V. V.; Shakirov, Z. N.; Abdurashitov, D. N.; Veretenkin, E. P.; Gavrin, V. N.; Gorbachev, V. V.; Ibragimova, T. V.; Kalikhov, A. V.; Mirmov, I. N.; Shikhin, A. A.; Yants, V. E.; Khomyakov, Y. S.; Cleveland, B. T. Artificial neutrino source based on the Ar-37 isotope. Phys. At. Nucl. 2007, 70, 300–310. (11) Loosli, H. H. A dating method with 39Ar. Earth Planet. Sci. Lett. 1983, 63, 51–62. (12) Jenkins, R.; Gilfrich, J. V. Figures-of-Meritc, their philosophy, design and use. X-Ray Spectrom. 1992, 21, 263–269. (13) Gedcke, D. A. AN59: How counting statistics controls detection limits and peak precision, 2001. (14) Loosli, H. H.; Oeschger, H.; Studer, R.; Wahlen, M.; Wiest, W. In Atmospheric concentrations and mixing of Argon-37, 1973. (15) G€aggeler, H. W. Radioactivity in the Atmosphere. Radiochim. Acta 1995, 70, 345–353. (16) Lal, D.; Peters, B. Cosmic Ray Produced Radioactivity on the Earth; Springer-Verlag: 1967. (17) Oeschger, H.; Houtermans, H. J.; Loosli, H. H.; Wahlen, M. In The Constancy of Cosmic Radiation from Isotope Studies in Meteorites an on the Earth, Nobel Symposium 12, Radiocarbon Variations and Absolute Chronology, 1970. (18) Zanini, A.; Ongaro, C.; Durisi, E.; Visca, E.; DeAgostini, S.; Fasolo, F.; Pelliccioni, M.; Saavedra, O. In Differential Neutron Flux in Atmosphere at Various Geophysical Conditions, 28th International Cosmic Ray Conference, 2003. (19) Florek, M.; Masarik, J.; Szarka, I.; Nikodemova, D.; Hrabovcova, A. Natural neutron fluence rate and the equivalent dose in localities with different elevation and latitude. Radiat. Prot. Dosim. 1996, 67, 187–192. (20) Goldhagen, P.; Reginatto, M.; Kniss, T.; Wilson, J. W.; Singleterry, R. C.; Jones, I. W.; Van Steveninck, W. Measurement of the energy spectrum of cosmic-ray induced neutrons aboard an ER-2 high-altitude airplane. Nucl. Instrum. Methods Phys. Res., Sect. A 2002, 476, 42–51. 8662
dx.doi.org/10.1021/es201192u |Environ. Sci. Technol. 2011, 45, 8656–8664
Environmental Science & Technology (21) Nakamura, T.; Nunomiya, T.; Abe, S.; Terunuma, K.; Suzuki, H. Sequential measurements of cosmic-ray neutron spectrum and dose rate at sea level in Sendai, Japan. J. Nucl. Sci. Technol. 2005, 42, 843–853. (22) Kowatari, M.; Nagaoka, K.; Satoh, S.; Ohta, Y.; Abukawa, J.; Tachimori, S.; Nakamura, T. Evaluation of the altitude variation of the cosmic-ray induced environmental neutrons in the Mt. Fuji area. J. Nucl. Sci. Technol. 2005, 42, 495–502. (23) Kowatari, M.; Ohta, Y.; Satoh, S.; Nagaoka, K.; Abukawa, J.; Nakamura, T. Evaluation of geomagnetic latitude dependence of the cosmic-ray induced environmental neutrons in Japan. J. Nucl. Sci. Technol. 2007, 44, 114–120. (24) Croft, S.; Bourva, L. C. A. The specific total and coincidence cosmic-ray-induced neutron production rates in materials. Nucl. Instrum. Methods Phys. Res., Sect. A 2003, 505, 536–539. (25) Moser, M. R.; Ryan, J. M.; Desorgher, L.; Fl€uckiger, E. O. In Atmospheric Neutron Measurements in the 10170 MeV Range, 29th International Cosmic Ray Conference, Pune, 2005. (26) Mishra, R. K.; Agarwal, R.; Tiwari, S. Solar cycle variation of cosmic ray intensity along with interplanetary and solar wind plasma parameters. Latv. J. Phys. Tech. Sci. 2008, 45, 63–68. (27) Fluckiger, E. O.; Butikofer, R. Swiss neutron monitors and cosmic ray research at Jungfraujoch. Adv. Space Res. 2009, 44, 1155– 1159. (28) Kudela, K.; Langer, R. Cosmic ray measurements in High Tatra mountains: 19572007. Adv. Space Res. 2009, 44, 1166–1172. (29) Shibata, K. Calculation of Neutron Nuclear Data on Calcium Isotopes for JENDL-4. J. Nucl. Sci. Technol. 2007, 44, 10–20. (30) Reedy, R. C.; Arnold, J. R. Interaction of solar and galactic cosmic-ray particles with the moon. J. Geophys. Res. 1972, 77, 537–555. (31) Desilets, D.; Zreda, M. Spatial and temporal distribution of secondary cosmic-ray nucleon intensities and applications to in situ cosmogenic dating. Earth Planet. Sci. Lett. 2003, 206, 21–42. (32) Lal, D. Cosmic ray labeling of erosion surfaces: in situ nuclide production rates and erosion models. Earth Planet. Sci. Lett. 1991, 104, 424–439. (33) Dunai, T. J. Scaling factors for production rates of in situ produced cosmogenic nuclides: a critical reevaluation. Earth Planet. Sci. Lett. 2000, 176, 157–169. (34) Stone, J. O. Air pressure and cosmogenic isotope production. J. Geophys. Res. 2000, 105, 230 753–7230 759. (35) Ziegler, J. F. Terretrial cosmic ray intensities. IBM J. Res. Dev. 1998, 42, 117–138. (36) Gosse, J. C.; Phillips, F. M. Terrestrial in situ cosmogenic nuclides: theory and application. Quat. Sci. Rev. 2001, 20, 1475–1560. (37) Phillips, F. M.; Stone, W. D.; Fabryka-Martin, J. T. An improved approach to calculating low-energy cosmic-ray neutron fluxes near the land/atmosphere interface. Chem. Geol. 2001, 175, 689–701. (38) Lal, D. In situ-produced cosmogenic isotopes in terrestrial rocks. Annu. Rev. Earth Planet. Sci. 1988, 16, 355–388. (39) Masarik, J.; Reedy, R. C. Terrestrial cosmogenic-nuclide production systematics calculated from numerical simulations. Earth Planet. Sci. Lett. 1995, 136, 381–395. (40) Hohenberg, C. M.; Marti, K.; Podosek, F. A.; Reedy, R. C.; Shirck, J. R. In Proceedings of the Lunar Planetary Science Conference, 1978; Vol. 9, pp 23112344. (41) Weaver, B. L.; Tarney, J. Major trace-element composition of the continental lithosphere. Phys. Chem. Earth 1984, 15, 39–68. (42) Fabryka-Martin, J. T. Production of radionuclides in the earth and their hydrogeologic significance, with emphasis on chlorine-36 and iodine-129. Ph.D. Dissertation, University of Arizona, 1988. (43) Lehmann, B. E.; Loosli, H. H. Isotopes formed by underground production. In Applied isotope hydrology, a case study in Northern Switzerland; Pearson, F. J., Balderer, W., Loosli, H. H., Lehmann, B. E., Matter, A., Peters, T., Schmassmann, H., Gautschi, A., Eds.; Nagra, 1991. (44) Jourdan, F.; Matzel, J. P.; Renne, P. R. 39Ar and 37Ar recoil loss during neutron irradiation of sanidine and plagioclase. Geochim. Cosmochim. Acta 2007, 71, 2791–2808.
ARTICLE
(45) Onstott, T. C.; Miller, M. L.; Ewing, R. C.; Arnold, G. W.; Walsh, D. S. Recoil refinements: Implications for the 40Ar/39Ar dating technique. Geochim. Cosmochim. Acta 1995, 59, 1821–1834. (46) Renne, P. R.; Knight, K. B.; Nomade, S.; Leung, K. N.; Lou, T. P. Application of deuteron-deuteron (D-D) fusion neutrons to Ar40/Ar-39 geochronology. Appl. Radiat. Isot. 2005, 62, 25–32. (47) Fl€ugge, S.; Zimens, K. E. Die Bestimmung von Korngr€ossen und von Diffusionskonstanten aus dem Emanierverm€ogen - Die Theorie der Emaniermethode. Phys. Chem. B 1939, 42, 179–220. (48) Farley, K. A. Helium diffusion from apatite: General behavior as illustated by Durango fluoroapatatite. J. Geophys. Res. 2000, 105, 2903–2914. (49) Watson, E. B.; Cherniak, D. J. Lattice diffusion of Ar in quartz, with constraints on Ar solubility and evidence of nanopores. Geochim. Cosmochim. Acta 2003, 67, 2043–2062. (50) Crank, J. The Mathematics of Diffusion; Oxford Science Publications: 1998. (51) Sharma, P.; Poulsen, T. G. Gas dispersion and immobile gas content in granular porous media: effect of particle size nonuniformity. Soil Sci. 2010, 175, 426–431. (52) Cook, P. G.; Solomon, D. K. Transport of atmospheric trace gases to the water table; implications for groundwater dating with chlorofluorocarbons and krypton 85. Water Resour. Res. 1995, 31, 263–270. (53) Tick, R. G.; McColl, C. M.; Yolcubal, I.; Brusseau, M. L. Gasphase diffusive tracer test for the in-situ measurement of tortuosity in the vadose zone. Water, Air, Soil Pollut. 2007, 184, 355–362. (54) Fuller, E. N.; Ensley, K.; Giddings, J. C. Diffusion of halogenated hydrocarbons in helium. J. Phys. Chem. 1969, 73, 3679. (55) Schaefer, C. E.; Arands, R. R.; van de Sloot, H. A.; Kosson, D. S. Modeling of the gaseous diffusion coefficient through unsaturated soil systems. J. Contam. Hydrol. 1997, 29, 1–21. (56) Lehmann, B. E.; Purtschert, R. Radioisotope dynamics - the origin and fate of nuclides in groundwater. Appl. Geochem. 1997, 12, 727–738. (57) Wanninkhof, R. H.; Mulholland-Patrick, J.; Elwood, J. W. Gas exchange rates for a first-order stream determined with deliberate and natural tracers. Water Resour. Res. 1990, 26, 1621–1630. (58) Auer, L. H.; Rosenberg, N. D.; Birdsell, K. H.; Whitney, E. M. The effects of barometric pumping on contaminant transport. J. Contam. Hydrol. 1996, 24, 145–166. (59) Gidda, T.; Cann, D.; Stiver, W. H.; Zytner, R. G. Airflow dispersion in unsaturated soil. J. Contam. Hydrol. 2006, 82, 118–132. (60) Costanza-Robinson, M. S.; Brusseau, M. L. Gas phase advection and dispersion in unsaturated porous media. Water Resour. Res. 2002, 38, WR000895. (61) Feige, Y.; Oltman, B. G.; Kastner, J. Production Rates of Neutrons in Soils Due to Natural Radioactivity. J. Geophys. Res. 1968, 73, 3135–3142. (62) Stoll, S.; Hendricks Frannsen, H. J.; Butts, M.; Kinzelbach, W. Analysis of the impact of climate change on groundwater related hydrological fluxes: a multi-model approach including different downscaling methods. Hydrol. Earth Syst. Sci. 2011, 15, 21–38. (63) Onnis, G. A. Interpreting multiple environmental tracer data with a groundwater model in a perialpine catchment. Ph.D. Dissertation, Swiss Federeal Institute of Technology, Zurich, 2008. (64) Johnson, P. C.; Bruce, C.; Johnson, R. L.; Kemblowski, M. W. In situ measurement of effective vapor-phase porous media diffusion coefficients. Environ. Sci. Technol. 1998, 32, 3405–3409. (65) Hers, I.; Zapf-Gilje, R.; Li, L.; Atwater, J. Measurement of in situ gas-phas diffusion coefficients. Environ. Technol. 2000, 21, 631–640. (66) Werner, D.; H€ohener, P. In situ method to measure effective and sorption-affected gas-phase diffusion coefficients in soils. Environ. Sci. Technol. 2003, 37, 2502–2510. (67) Werner, D.; Grathwohl, P.; Werner, D. Review of field methods for the determination of the tortuiosity and effective gas-phase diffusiviity in the vadose zone. Vadose Zone J. 2004, 3, 1240–1248. (68) Moldrup, P.; Olesen, T.; Komatsu, T.; Schjonning, P.; Rolston, D. E. Tortuosity, diffusivity, and permeability in the soil liquid and gaseous phases. Soil Sci. Soc. Am. J. 2001, 65, 613–623. 8663
dx.doi.org/10.1021/es201192u |Environ. Sci. Technol. 2011, 45, 8656–8664
Environmental Science & Technology
ARTICLE
(69) Althaus, R. Datierung und Charakterisierung von Grundwasservorkommen in der Schweiz und im s€uddeutschen Raum anhand von 85 Kr-Messungen und weiteren isotopenphysikalischen Methoden. Ph.D. Dissertation, University of Bern, 2009. (70) Kristensen, A. H.; Thorbjorn, A.; Jensen, M. P.; Pedersen, M.; Moldrup, P. Gas-phase diffusivity and tortuosity of structured soils. J. Contam. Hydrol. 2010, 115, 26–33. (71) Tuli, A.; Hopmans, J. W.; Rolston, D. E.; Moldrup, P. Comparison of air and water permeability between disturbed and undisturbed soils. Soil Sci. Soc. Am. J. 2005, 69, 1361–1371. (72) Carrigan, C. R.; Heinle, R. A.; Hudson, G. B.; Nitao, J. J.; Zucca, J. J. Baromotric gas transport along faults and its application to nuclear test-ban monitoring, Lawrence Livermore National Laboratory, Livermore, CA, 94550, 1997. (73) ETOPO5 5-minute gridded elevation data.http://www.ngdc. noaa.gov/mgg/global/etopo5.HTML (accessed month day, year). (74) Harmonized World Soil Database. http://www.iiasa.ac.at/ Research/LUC/External-World-soil-database/HTML (accessed month day, year). (75) Gelhar, L. W.; Welty, C.; Rehfeldt, K. R. A critical review of data on field-scale dispersion in aquifers. Water Resour. Res. 1992, 28, 1955– 1974. (76) Amplitude and period of atmospheric pressure changes in Switzerland during a storm event.http://www.meteoschweiz.admin. ch/web/de/wetter/wetterereignisse/stuermische_kaltfront_am_12_ oktober_2009.Par.0010.DownloadFile.tmp/gross.jpg (accessed month day, year).
8664
dx.doi.org/10.1021/es201192u |Environ. Sci. Technol. 2011, 45, 8656–8664
ARTICLE pubs.acs.org/est
Biomagnification of Perfluorinated Compounds in a Remote Terrestrial Food Chain: LichenCaribouWolf Claudia E. M€uller,‡,§ Amila O. De Silva,^ Jeff Small,^ Mary Williamson,^ Xiaowa Wang,^ Adam Morris,† Sharon Katz,),|| Mary Gamberg,# and Derek C. G. Muir*,^ ‡
Laboratory for Analytical Chemistry, Empa, D€ubendorf, Switzerland Institute for Chemical- and Bioengineering, ETH Z€urich, Z€urich, Switzerland ^ Aquatic Ecosystem Protection Research Division, Environment Canada, Burlington, ON Canada † School of Environmental Sciences, University of Guelph, Guelph ON Canada ) Aurora Research Institute, Inuvik, NT Canada # Gamberg Consulting, Whitehorse, YK Canada §
bS Supporting Information ABSTRACT: The biomagnification behavior of perfluorinated carboxylates (PFCAs) and perfluorinated sulfonates (PFSAs) was studied in terrestrial food webs consisting of lichen and plants, caribou, and wolves from two remote northern areas in Canada. Six PFCAs with eight to thirteen carbons and perfluorooctane sulfonate (PFOS) were regularly detected in all species. Lowest concentrations were found for vegetation (0.020.26 ng/g wet weight (ww) sum (Σ) PFCAs and 0.0020.038 ng/g ww PFOS). Wolf liver showed highest concentrations (1018 ng/g ww ΣPFCAs and 1.41.7 ng/g ww PFOS) followed by caribou liver (610 ng/g ww ΣPFCAs and 0.72.2 ng/g ww PFOS). Biomagnification factors were highly tissue and substance specific. Therefore, individual whole body concentrations were calculated and used for biomagnification and trophic magnification assessment. Trophic magnification factors (TMF) were highest for PFCAs with nine to eleven carbons (TMF = 2.22.9) as well as PFOS (TMF = 2.32.6) and all but perfluorooctanoate were significantly biomagnified. The relationship of PFCA and PFSA TMFs with the chain length in the terrestrial food chain was similar to previous studies for Arctic marine mammal food web, but the absolute values of TMFs were around two times lower for this study than in the marine environment. This study demonstrates that challenges remain for applying the TMF approach to studies of biomagnification of PFCAs and PFSAs, especially for terrestrial animals.
’ INTRODUCTION Perfluorinated compounds (PFCs) including perfluorinated carboxylates (PFCAs) and perfluorinated sulfonates (PFSAs) and their precursors, are a group of anthropogenic substances ubiquitously found in the environment, in industrialized regions as well as remote areas.17 Although they have been produced since the 1950s for industrial applications and consumer products, their environmental presence was discovered only recently, in the early millennium.7 Since then, PFCs have been found in all environmental compartments, as well as in humans from around the world.17 PFCAs and PFSAs are environmentally recalcitrant and have a bioaccumulation and long-range transport (LRT) potential similar to other persistent organic pollutants (POPs). The PFCAs and PFSAs are ionic substances and nonvolatile and therefore, their transport behavior differs from many classical POPs. Atmospheric LRT as anions is considered less significant compared with two alternative LRT processes: (i) transport by r 2011 American Chemical Society
oceanic currents810 or (ii) distribution of volatile precursor substances such as fluorotelomer alcohols (FTOHs) and perfluorosulfonamides via the atmosphere with subsequent degradation to PFCAs and PFSAs.1113 The relative importance of these two processes is still under debate. However, for remote terrestrial environments atmospheric transport of precursors is more likely. PFCAs with eight to twelve carbons and PFSAs bioaccumulate and biomagnify in aquatic food webs, as has been shown in several marine food web studies.13,1417 These anionic PFCs are proteinophilic and generally found at highest concentrations in blood and liver in contrast to lipophilic pollutants such as polychlorinated biphenyls (PCBs), which enrich mainly in lipid tissue. Some of the highest concentrations Received: April 20, 2011 Accepted: August 26, 2011 Revised: August 24, 2011 Published: September 09, 2011 8665
dx.doi.org/10.1021/es201353v | Environ. Sci. Technol. 2011, 45, 8665–8673
Environmental Science & Technology of PFCs in wildlife have been found for perfluorooctane sulfonate (PFOS) in polar bear liver (up to 3100 ng/g wet weight, ww).4 Terrestrial animals from remote locations have also considerable PFOS concentrations, for example, the arctic fox, an opportunistic feeder, had liver concentrations of 250 ng/g ww while the herbivorous caribou had liver concentrations around 3 ng/g ww.4,18 While poorly metabolized chemicals with high octanol air partition coefficients KOA (g106) and medium octanol water partition coefficients KOW (1010 g KOW g 102) are expected to biomagnify in air-breathers in terrestrial food webs,19,20 little is known about accumulation of proteinophilic compounds in the same food webs. Therefore, to address this question we investigated the biomagnification of PFCs in terrestrial food webs in two areas in the Canadian Arctic in this study. These food webs included vegetation (plants and lichens), barren ground caribou (rangifer tarandus groenlandicus) and wolves (canis lupus). Investigation of this remote, terrestrial food chain has certain advantages: First of all, the lichencaribou wolf food chain is well documented and in particular caribou have been studied intensively due to their economic and social importance for indigenous people in the Canadian Arctic.21 It is relatively simple, as caribou feed mostly on lichen (in summer the diet also consists of willow, sedges and grasses) and wolves living near barren-ground caribou herds almost exclusively feed on them. It is therefore potentially easier to assess diet-consumer relationships than for more complex aquatic food webs. A further advantage is that the PFC input into this remote environment is solely via the atmosphere and that local sources are absent. The objectives of this study were therefore to (i) assess the input of PFCAs and PFSAs in the arctic terrestrial environment, (ii) study their biomagnification potential in terrestrial mammals and (iii) compare aquatic, marine mammalian and terrestrial biomagnification. To our knowledge, this is the first study on PFC biomagnification in a remote terrestrial food chain.
’ MATERIALS AND METHODS Chemicals. Information on chemicals (standards of PFCAs, PFSAs, and isotopically labeled surrogates as well as reagents) is summarized in the Supporting Information SI. Samples. Liver, muscle, and kidney samples from two caribou herds were collected; from the Porcupine herd in northern Yukon Territory and the Bathurst herd in the Northwest Territories (NWT)/western Nunavut (see map in SI Figure S1). Wolf, lichen (cladonia mitis/rangiferina and flavocetraria nivalis/ cucullata), and plant samples were collected in the same region as the caribou. Plant samples included cottongrass (eriophorum vaginatum), aquatic sedge (carex aquatilis), willow (salix pulchra), moss (rythidium rugosum), and mushrooms (unknown species). Liver and muscle samples were collected from the sampled wolves. The sampling procedure, time, storage, and sample types are described in more detail in the SI. Extraction and Cleanup. Vegetation samples were analyzed using a method similar to that described in Powley et al. (details see SI).22 In brief, plant matter was extracted three times with methanol, the combined extract was concentrated and subjected to further clean up with a carbon solid phase extraction (SPE) cartridge. The liver, kidney, and muscle samples were analyzed according to a modified method from Powley et al.22 The homogenized tissue was extracted twice with methanol or with acetonitrile and
ARTICLE
the combined extract was concentrated to dryness. The reconstituted extract (in methanol) was then cleaned with a carbon cartridge. Instrumental Analysis. All samples were analyzed for PFCAs from perfluorohexanoate PFHxA to perfluorotetradecanoate PFTeA, as well as the PFSAs, perfluorohexane sulfonate PFHxS, and PFOS. The analyses were performed by liquid chromatography with negative electrospray tandem mass spectrometry (LC-MS/MS). Analytes were detected using an API 4000 Q Trap (Applied Biosystems, Carlsbad, CA) after chromatographic separation with an Agilent 1100 LC (injection volume = 40 μL, flow rate = 300 μL/min). Chromatography was performed using an ACE C18 column (50 mm 2.1 mm, 3 μm particle size, Aberdeen, U.K.), preceded by a C18 guard column (4.0 2.0 mm, Phenomenex, Torrance, CA) and the column oven was set to 30 C. Samples were quantified with a six point calibration curve and isotopic dilution method. Quality Assurance/Quality Control (QA/QC). Reproducibility of the analytical method was assessed by triplicate analysis of a sample of every species. Reproducibility was between 3 and 15% for all species and substances except for perfluorotridecanoate (PFTrA) which had relative standard deviation of up to 50%, due to very low concentrations. Recoveries were determined with spike and recovery, as well as by internal standard comparisons, and ranged from 26% to 46% for plants and from 65%129% for all other sample types (see SI Table S3). PFTeA showed very high recovery (170%) for wolf liver, probably because of matrix enhancement. PFTeA results were therefore not reported. Method blanks were conducted for every batch of six to ten samples. All samples were blank-subtracted. The method detection limit (MDL) was defined as either as three times the standard deviation of the blank samples or, if the blanks had no detectable contamination, as the instrumental detection limit (IDL). MDLs and average blank levels are shown in the SI (Table S4 and Table S5). Stable Isotope Analysis. Determination of carbon and nitrogen stable isotope ratios (δ13C and δ 15N) were conducted for all species studied here to investigate the diet relationship in this food chain. Further information is provided in the SI. Biomagnification and Trophic Magnification Factor Calculations. Biomagnification factors (BMF = Cconsumer/Cdiet) were calculated based on two different approaches: (i) on single tissue concentrations and (ii) on an estimated whole body concentration for caribou and wolf (see eq 1). Cwhole body ¼
∑ Ctissue n f tissue n
n¼1
ð1Þ
CTissue is the concentration of the specific tissue and fTissue is the mass fraction of this tissue in the whole body. Tissue fractions were obtained from post-mortem examinations of selected investigated animals. Every individual animal was calculated separately on the basis of the specific tissue concentration. Animals with only one tissue were excluded from this calculation. Concentrations of tissues, which were not measured here, were estimated on the basis of literature values.1 It was assumed that the PFCA and PFSA concentration in blood and lungs were half that of liver and the carcass was assumed to have half the concentration found in muscle tissue. Bones were excluded from the whole body calculation because PFCs are assumed to not enrich in this media and bones are not part of the diet of wolves. When average concentrations were below MDL, they were substituted with MDL/2 in order to get numerical values for 8666
dx.doi.org/10.1021/es201353v |Environ. Sci. Technol. 2011, 45, 8665–8673
Environmental Science & Technology
ARTICLE
whole body estimates. Tissue fractions for caribou and wolf are listed in SI Table S6. Average concentrations were used for BMF calculations. Trophic magnification factor (TMF) was calculated according to Jardine et al.23 The TMF provides information on the average change in contaminant concentration per relative trophic level and is calculated using the natural logarithm of the concentration (Cww) of individual organisms versus their trophic level (TL): ln Cww ¼ a þ ðb TLÞ
ð2Þ
a is the intercept and b the slope of the linear equation. Two alternate calculation methods were tested: Cww for the second and third trophic level was either the concentration in a specific tissue (liver or muscle) or the whole body concentration from eq 1. Additionally, lichen alone or all vegetation samples were used as the primary producer. The trophic level can be derived from the stable isotope ratio of nitrogen δ15N for the consumer (wolf or caribou) and first trophic level lichen with the relationship TL ¼ 1 þ ðδ15 Nconsumer δ15 Nlichen Þ=Δ15 N
ð3Þ
where 1 is the assumed trophic level of lichen and Δ15N is the trophic enrichment factor constant. 24 Generally a factor of 3.4 % is assumed for biomagnification assessments. For this study however, a factor of 3.8 % was chosen, because caribou have been shown to have higher enrichment factors. This is based on earlier observations of Ben-David et al., who reported an enrichment of 3.8 % and higher for barren ground caribou.25 The TMF can then be calculated based on the slope b of eq 2. TMF ¼ eb
ð4Þ
Statistical Analysis. All statistical tests were done using SYSTAT (version 10, SPSS Inc.). The two tailed t test was used to test for differences in concentrations between species and locations. Slopes of TMF regressions were tested for significance at P = 0.05. Additionally, correlations between age and concentrations of caribou and wolf samples (both muscle and liver) were tested.
’ RESULTS AND DISCUSSION Concentrations and Profiles of PFCs. An overview of PFCA and PFOS concentrations in both food chains, the Porcupine herd food chain in Yukon and the Bathurst herd food chain in NWT, can be found in Table 1 (more information in SI Table S7). Of the 11 PFCs analyzed in this study only PFCAs with eight to thirteen carbons and PFOS were regularly detected. Highest concentrations were measured in wolf liver (10 and 18 ng/g ww sum (Σ) PFCAs for Yukon and NWT, respectively), followed by caribou liver (6 and 10 ng/g ww ΣPFCAs, respectively). Considerably lower concentrations were found in vegetation samples, often close to the MDL. Vegetation. Lowest average concentrations were found in cottongrass in Yukon (e.g., from <MDL for tridecanoate (PFTrA) to 0.013 ( 0.009 ng/g ww for perfluorooctanoate (PFOA)). Higher concentrations were found in lichen (e.g., 0.08 ( 0.01 and 0.10 ( 0.02 ng/g ww for perfluorononanoate (PFNA) in Yukon and NWT, respectively), willow (e.g., 0.19 ( 0.10 ng/g ww PFOA in Yukon), and mushroom (e.g., 0.19 ( 0.08 ng/g ww PFOA). While all concentrations are very low, the PFCA congener composition differed among types of vegetation (Figure 1).
All plants (grass, sedge, willow, and moss) are dominated by PFOA, while both lichen species showed a dominance of the odd carbon chain lengths (C8 < C9 and C12 < C13, C10 < C11 only for lichen in Yukon). A weaker version of this pattern can also be seen in cottongrass and willow from NWT (C10 < C11, C12 < C13), but not in Yukon. This is most likely due to the fact that the vegetation samples had generally lower concentrations in Yukon and PFCs were more often below detection limits. PFOS concentrations in all vegetation samples are lower than for PFCAs, from <MDL to 0.062 ng/g. The difference in PFOA concentrations between plants and lichen could possibly be explained by the different uptake mechanisms: Plants such as willows and cotton grass obtain nutrients and water via root systems in the soil. Lichen, on the other hand, do not have roots but absorb nutrients directly from precipitation. Therefore, PFCA levels in lichen should represent the direct input from the atmosphere. Uptake from soil is a more indirect pathway, where different discrimination mechanisms can play a role (preferential soil adsorption of longer chain PFCs,26 possible preferential root permeation of shorter PFCs), leading to a PFCA composition different from the atmosphere. In general, the source of atmospheric PFCAs and PFSAs is uncertain. In urban areas, precipitation samples often have high value of PFOS and PFOA, in more remote areas, this dominance seems to be weakened.27,28 Measurements of PFCs in rain collected at Snare Rapids (NWT), a sampling site west of the Bathurst herd range, showed relatively high PFNA concentrations followed by PFOA (see Figure 1; details in SI Table S11).29 The other odd chain PFCAs are, however, very low (C10 > C11, C12 > C13). Precipitation and particulate air samples from more northern locations in the Arctic show a pattern with strong dominance of PFOA and low contribution of odd chain PFCAs.30,31 Furthermore, PFOS concentrations are much higher than PFOA or PFNA in these Arctic samples, whereas rain from the Bathurst region and the lichen samples have a low contribution of PFOS to total PFCs (a set of literature values are presented in SI Table S11). Caribou. Highest PFC liver concentrations were found for PFNA (2.2 ( 0.2 and 3.2 ( 0.4 ng/g ww for the Porcupine and Bathurst herds, respectively) followed by perfluorodecanoate (PFDA, 1.9 ( 0.1 and 2.2 ( 0.2 ng/g ww) and perfluoroundecanoate (PFUnA, 1.7 ( 0.1 and 3.2 ( 0.2 ng/g ww). Similar to the vegetation results, the PFOS content was lower (0.67 ( 0.13 ng/g ww) in the Porcupine caribou. The Bathurst caribou on the other hand had PFOS concentrations (2.2 ( 0.3 ng/g ww) similar to that corresponding to the most prominent PFCAs. These results are in agreement with those reported in a study of PFCs in caribou from several communities in Nunavut in which PFCA and PFOS liver concentrations were in the same range (PFNA 2.0 ( 1.7 and PFOS 2.7 ( 2.3 ng/g ww).18 Mean PFC concentrations in caribou were higher in the Bathurst samples, however this difference was statistically significant (p < 0.05) in liver only for PFOS, PFNA, and PFUnA (SI Figure S4). This might be due to differences in distance from source regions in North America. The study area in northern Yukon is more remote from these source regions than the Bathurst area in the NWT. In comparison to liver, PFC concentrations in muscle and kidney were 10 to 20 times lower, which shows the extreme preference of PFCAs and PFSAs for liver. PFCAs and PFOS bind preferentially to certain proteins in the blood and liver, leading to enrichment in these tissues.3234 However, when the respective 8667
dx.doi.org/10.1021/es201353v |Environ. Sci. Technol. 2011, 45, 8665–8673
a
8668
Porcupine
Porcupine
Porcupine
Bathurst
Bathurst
Porcupine Porcupine
Bathurst
Bathurst
muscle
liver
kidney
muscle
liver
muscle liver
muscle
liver
5
10
6 6
7
9
10
10
7
5 3
3
3
9
10
7
8
10
n
Fla: flavocetratia. Cla: cladonia
Bathurst Bathurst
Bathurst
lichen (Cla)a
moss mushrooms
Porcupine
cotton grass
Bathurst
Porcupine
willow
Bathurst
Porcupine
lichen (Cla)a
cotton grass
Porcupine
lichen (Fla)a
willow
herd
sample
2007
2007
2010 2010
2008
2008
2007
2007
2007
2009 2009
2009
2009
2009
2008
2008
2008
2008
year
0.056 ( 0.021
0:19 ( 0:10
1:9 ( 0:1 0:02 ( 0:01
0.064 ( 0.008 2:2 ( 0:2 0:10 ( 0:02
0:022 ( 0:008
0.093 ( 0.009 3:2 ( 0:4 0:44 ( 0:06 4:7 ( 0:9 1:1 ( 0:1 7:4 ( 1:3
0.024 ( 0.006 0:11 ( 0:01 0.083 ( 0.013 0.23 ( 0.08 0.06 ( 0.01 0.10 ( 0.03
<0.01
<0.5
0:033 ( 0:007
0.058 ( 0.018 0.012 ( 0.007
0.024 ( 0.004 0:19 ( 0:08
0.005 ( 0.001
0.038 ( 0.012
3:2 ( 0:2
0:21 ( 0:03
0:11 ( 0:02 2:0 ( 0:2
wolf
2:2 ( 0:2
0.075 ( 0.009
caribou
0.006 ( 0.002 <0.005
0.005 ( 0.003
0.021 ( 0.001
0.067 ( 0.039
0.034 ( 0.005
0.099 ( 0.017
0.003 ( 0.003
0.010 ( 0.010
0.021 ( 0.010
PFUnA
6.4 ( 1.2
0.57 ( 0.14
0.15 ( 0.02 2.5 ( 0.4
3.2 ( 0.2
0.18 ( 0.01
0.02 ( 0.01
1.7 ( 0.1
0.070 ( 0.010
0.020 ( 0.009 0.008 ( 0.003
0.020 ( 0.004
0.013 ( 0.002
0.041 ( 0.005
0.003 ( 0.002
0.003 ( 0.002
0.049 ( 0.010
0.038 ( 0.014
vegetation 0.025 ( 0.013
PFDA
0.047 ( 0.015
<0.003
<0.004
0.083 ( 0.010
0.025 ( 0.010 0.013 ( 0.009
0.056 ( 0.018
PFNA
0.023 ( 0.009
PFOA
concentration ( SE (ng/g ww)
0:26 ( 0:15 0:55 ( 0:10
0:72 ( 0:14
0.005 ( 0.003 0:35 ( 0:04
0:49 ( 0:03
0.052 ( 0.006
<0.01
<0.5
0.031 ( 0.008
0.010 ( 0.003 <0.005
0.009 ( 0.001
0.008 ( 0.002
0.010 ( 0.003
<0.004
<0.004
0.018 ( 0.010
0.014 ( 0.010
PFTrA
0.039 ( 0.021
0.017 ( 0.003 0:42 ( 0:05
0:66 ( 0:05
0.031 ( 0.015
<0.01
<0.5
<0.01
0.004 ( 0.001 0.016 ( 0.008
0.002 ( 0.002
<0.002
0.006 ( 0.001
0.005 ( 0.002
0.003 ( 0.004
0.010 ( 0.010
0.016 ( 0.008
PFDoA
1.7 ( 0.1
0.14 ( 0.02
0.13 ( 0.02 1.4 ( 0.3
2.2 ( 0.3
0.076 ( 0.019
0.020 ( 0.003
0.67 ( 0.13
0.028 ( 0.023
0.008 ( 0.002 0.038 ( 0.006
0.014 ( 0.007
0.018 ( 0.003
0.021 ( 0.003
0.002 ( 0.002
0.062 ( 0.029
0.013 ( 0.006
0.014 ( 0.008
PFOS
18 ( 2.9
2.3 ( 0.5
0.8 ( 0.1 10 ( 1.7
9.8 ( 0.9
0.45 ( 0.06
0.15 ( 0.03
5.8 ( 0.4
0.22 ( 0.04
0.12 ( 0.04 0.22 ( 0.09
0.12 ( 0.04
0.08 ( 0.01
0.22 ( 0.07
0.02 ( 0.02
0.26 ( 0.14
0.21 ( 0.06
0.17 ( 0.07
Σ PFCAs
20
2.4
0.93 12
12
0.53
0.17
6.5
0.25
0.13 0.26
0.13
0.10
0.24
0.03
0.33
0.22
0.19
Σ PFCs
Table 1. Overview of Sample Type, Size, Year of Collection, and Mean Concentration ( Standard Error (ng/g ww) of PFCAs and PFOS for the Porcupine (Northern Yukon) and Bathurst (Northwest Territories/Nunavut) Caribou Herd Food Chains
Environmental Science & Technology ARTICLE
dx.doi.org/10.1021/es201353v |Environ. Sci. Technol. 2011, 45, 8665–8673
Environmental Science & Technology
Figure 1. PFCA homologue composition of species studied here and selected comparisons from literature sources. (a) Wolf and caribou homologue distribution shown here are from muscle samples. Reference values are form (b) Martin et al.,4 (c) Scott et al.,29 and (d) Stock et al.31
body fractions of the tissues are considered, the distribution changes dramatically. Muscle tissue contains 6090% of the body burden (concentration tissue tissue fraction on total body weight, see SI Figure S3). Liver has only a minor contribution and kidney loses importance with less than 1% of the total burden. This shows that studying only liver concentration for biomagnification will greatly alter results, as muscle (and other) tissues are equally important dietary items for wolves. Concentrations of PFCs in caribou muscle and liver were not significantly correlated with age except for PFOA in Bathurst caribou muscle (SI Table S8). Weak or nonsignificant correlations of PFCs with age have also been found in ringed seals35 and dolphins.1 Wolf. The top predator of this food chain, the wolf, displayed a similar PFC distribution as caribou. The PFCA and PFSA concentrations present in wolf liver were 5 to 15 times higher compared to muscle and the contribution of liver to the whole body burden was about 10 to 40%. Highest concentrations were observed in PFNA (4.7 ( 0.9 and 7.4 ( 1.3 ng/g ww for Yukon and Bathurst, respectively) and PFUnA (2.5 ( 0.4 and 6.4 ( 1.2 ng/g ww, respectively). Again, concentrations were higher for the Bathurst wolves, however, this was significant only for PFDA and PFUnA. Concentrations of PFCs in Porcupine wolf liver and muscle were not significantly correlated (P > 0.05) with age, though a slight increase with age was seen for some PFCs (SI Table S9). To more thoroughly investigate the relationship between age and concentration, more individuals would be needed for both wolves and caribou. In comparison to other predators, the PFC concentrations found in wolf are relatively low. Concentrations of up to 250 ng/g PFOS and 22 ng/g for PFNA were found in Arctic fox.4
ARTICLE
This terrestrial animal is an opportunistic predator and its diet can contain also marine animals, which could lead to higher contaminant levels. Wolves on the other hand mainly feed on herbivorous animals such as caribou, moose and sometimes smaller animals. In polar bear livers, PFOS concentrations up to 3100 ng/g ww and PFNA levels up to 190 ng/g ww have been found.3 These much higher levels can be partially explained by the longer food chain (polar bears are on trophic level four/five) hence the biomagnification pathway is longer. Additionally, there is more PFC input into the marine food web, from atmospheric, as well as oceanic transport. Looking at all species studied here two observations become apparent: (i) PFC concentrations in tissues increase with tropic level (see Table 1). For most PFCs, levels are highest in wolf, followed by caribou and then vegetation. (ii) The PFCA homologue pattern for lichen, caribou, and wolf are very similar (see Figure 1) with a dominance of the odd carbon chain PFCAs. Cotton grass, willow, and other plant have a dominance of PFOA. Stable Isotopes and Food Web Structure. The stable isotope analyses were conducted for all species studied here to investigate the dietary relationship in this food chain. An overview of the δ13C and δ15N results of this study is shown in the SI (Figure S2). δ13C vary widely in photosynthetic organisms and 13 C is only moderately enriched along the food chain (12%).23 Lichen, caribou, and wolf had similar δ13C values implying that the caribou were mainly feeding on lichen and the wolves mainly on caribou. On the other hand, the δ15N difference between lichen and caribou is rather large (78%) compared to usually assumed Δ15N of 3.43.8%. It is however known that there are much larger Δ15N values for herbivores feeding on a nitrogen poor diet.25,36 For example, a controlled feed study on white tailed deer showed high shifts of 4.93 ( 0.74%.36 Additionally, the isotope composition can change seasonally with diet supply, for example, fasting in winter can increase the δ15N.37 A linear combination of food sources was calculated for caribou with the use of the IsoSource model from Phillips et al.38 The results indicated that lichen, cotton grass, and mushroom are the main food sources (see SI Table S12). However, this analysis also showed that all plant samples had basically too low δ13C values (more than 2% lower than caribou) and numerical results could only be obtained when an enrichment of 2% was assumed. It is therefore possible that these plants did not play a major role in the caribou diet. We analyzed most of the major plants that are thought to contribute to the diet but cannot rule out the possibility that some other plant or mushroom species are important. It has to be mentioned that vegetation and animal species were collected in different seasons, which may introduce additional variation and hamper food source interpretation. Both lichen and all vegetation were thus used for magnification factor calculations to study the potential influences of vegetation other than lichen. For BMF calculations, combined vegetation concentrations were calculated based on the food source contribution results of the IsoSource model. Biomagnification Factors. Biomagnification factors (BMFs) were calculated to evaluate the biomagnification behavior between diet and consumer. Usually, the lipid normalized concentrations are used for these calculations to eliminate lipid content variability between organs/organisms. This is however not possible for PFCs, because they are rather proteinophilic than lipophilic. In literature, different approaches have been taken to calculate representative BMFs for PFCs: using either specific tissue concentrations (mostly blood or liver),17 whole body concentrations,1 8669
dx.doi.org/10.1021/es201353v |Environ. Sci. Technol. 2011, 45, 8665–8673
Environmental Science & Technology
ARTICLE
Table 2. Biomagnification Factors (BMFs) and Trophic Magnification Factors (TMFs) for PFCs for this Study and Reference Studiesa
cariboumuscle/lichen
Porcupine
caribouliver/lichen
Bathurst Porcupine
wolfmuscle/cariboumuscle wolfliver/caribouliver
PFOA
PFNA
PFDA
PFUnA
PFDoA
0.9 ( 0.4
1.2 ( 0.3
biomagnification factors (BMFs) 1.3 ( 0.6 1.9 ( 0.4 1.9 ( 0.9
0.9 ( 0.2 40 ( 12
1.1 ( 0.7 75 ( 25
PFTrA
PFOS
2.1 ( 0.9
2.0 ( 1.8
4.3 ( 0.6 46 ( 12
5.2 ( 2.8 16 ( 4.9
5.0 ( 1.7 17 ( 7
3.6 ( 1.0 4.0 ( 3.7
110 ( 28
47 ( 15
3.1 ( 0.9
1.2 ( 0.9
4.9 ( 2.9
1.8 ( 0.5
Bathurst
11 ( 1.2
32 ( 7.0
33 ( 19
78 ( 11
Porcupine
3.8 ( 1.5
6.9 ( 1.2
3.3 ( 0.8
2.1 ( 0.4
Bathurst
2.6 ( 0.8
12.4 ( 1.7
2.8 ( 0.6
3.2 ( 0.8
Porcupine
0.0
2.1 ( 0.5
1.0 ( 0.1
1.4 ( 0.3
Bathurst
0.9 ( 0.3
2.3 ( 0.5
1.4 ( 0.1
2.0 ( 0.4
1.1 ( 0.2
1.1 ( 0.2
0.8 ( 0.1
4.5 ( 3.8 2.1 ( 0.6
caribouwhole/lichen
Porcupine
1.4 ( 0.4
2.8 ( 0.7
6.1 ( 2.2
3.8 ( 0.9
2.9 ( 1.1
2.4 ( 1.0
4.8 ( 2.3
caribouwhole/vegetation
Bathurst Porcupine
2.6 ( 0.5 1.8 ( 0.7
2.7 ( 0.6 8.5 ( 2.6
2.9 ( 1.7 12.4 ( 5.5
8.2 ( 1.1 9.8 ( 3.2
11 ( 3 4.5 ( 1.7
6.9 ( 2.2 7.1 ( 4.7
9.1 ( 1.6 9.1 ( 4.9
Bathurst
0.3 ( 0.1
5.3 ( 1.0
7.2 ( 3.8
14.5 ( 1.8
8(3
9.0 ( 2.5
7.9 ( 1.6
wolfwhole/caribouwhole
Porcupine
2.4 ( 0.6
3.8 ( 0.6
1.7 ( 0.2
2.0 ( 0.3
1.2 ( 0.1
0.8 ( 0.2
3.3 ( 1.1
3.2 ( 1.5
Bathurst
2.1 ( 0.5
5.4 ( 0.8
2.1 ( 0.2
2.8 ( 0.5
1.4 ( 0.4
dolphinwhole/seatroutwhole
Houde et al.1
1.8
2.1
2.4
2.5
0.6
1.2 ( 0.2
seatroutwhole/pinfishwhole
Houde et al.1
7.2
1.5
3.7
0.9
0.1
laketrout/prey
Martin et al.3
0.4
2.3
2.7
3.4
1.6
2.5
2.9
polar bearliver/ sealliver
Martin et al.4
8.6
33
22
18
10
14
177
wolfliver/caribouliver/lichen
Porcupine
2.4 ( 0.1
6.7 ( 0.5
7.1 ( 0.4
6.6 ( 0.4
4.1 ( 0.1
3.7 ( 0.1
6.7 ( 0.3
Bathurst
2.2 ( 0.1
4.5 ( 0.2
5.1 ( 0.3
6.1 ( 0.3
5.2 ( 0.3
4.2 ( 0.3
5.2 ( 0.4
Porcupine
1.3 ( 0.1
2.7 ( 0.2
2.6 ( 0.1
2.5 ( 0.1
1.4 ( 0.1
1.4 ( 0.1
2.6 ( 0.1
Bathurst
1.3 ( 0.1
2.2 ( 0.1
2.3 ( 0.1
2.8 ( 0.1
2.2 ( 0.1
2.0 ( 0.1
2.4 ( 0.1
Porcupine
1.1 ( 0.1
2.0 ( 0.2
2.3 ( 0.1
2.2 ( 0.2
1.3 ( 0.0
1.4 ( 0.0
2.2 ( 0.1
Bathurst
1.3 ( 0.1
1.9 ( 0.1
2.3 ( 0.2
2.9 ( 0.3
2.0 ( 0.1
1.8 ( 0.1
2.3 ( 0.1
piscivorous food web
Kelly et al.15
0.4
0.6
0.6
1.1
1.0
0.4
0.5
arctic marine mammalian food web dolphin food web
Kelly et al.15 Houde et al.1
2 6
4 2
5 2
5 2
3 1
2
11 2
0.9 4.6
trophic magnification factors (TMFs)
wolfwhole/caribouwhole/lichen wolfwhole/caribouwhole/vegetation
a
BMF ( standard error; TMF ( 95% confidence interval (CI). Upper and lower CI did not differ after rounding to one decimal place.
or normalize to protein content of the tissue.15 We tested the first two approaches. Tissue specific BMFs varied considerably (see Table 2 and SI Table S13); highest values were found for lichen to caribou liver BMFs from 11 for PFOA to 110 for perfluorododecanoate PFDoA (Bathurst caribou). PFCs are only moderately magnified in muscle tissue of caribou from 0.9 to 2.4 for PFOA to 1.9 to 5.2 for PFDoA. In general, biomagnification seemed to be dependent on fluorinated alkyl chain length: BMFs were highest for PFNA/PFOS to PFUnA for caribou liver, whereas PFUnA to PFTrA had highest biomagnifications for caribou muscle. This shows that single compartment BMFs can obscure the true biomagnification behavoiur. The difference between these two tissues might be explained with different protein compositions. Additionally, the different metabolic rate of muscle and liver could have an influence. Liver tissue has a faster turnover rate than muscle tissue and liver concentrations could therefore represent a shorter term contaminant situation. As tissue specific BMFs differ considerably, BMFs on whole body provide a more realistic estimate of biomagnification behavior (see Table 2). Lichen to caribou BMFs range from 1.4 ( 0.1 (PFOA) to 6.1 ( 2.2 (PFDA) for the Porcupine caribou and from 2.6 ( 0.5 (PFOA) to 11 ( 3 (PFDoA) for the
Bathurst caribou. It can be seen that the BMFs differ considerably between the two caribou herds. However, if combined vegetation based on weighted species average is used as diet for the caribou, then the two herds are more uniform. BMFs were lowest for PFOA with 0.3 to 1.8 and highest for PFDA and PFUnA with 12.4 and 14.5, respectively. In late fall/early winter, caribou are thought to consume also shrubs and mushroom, and the animals studied here were all collected in this period. The studied willow and also cotton grass showed higher long-chain PFCA concentrations in the Bathurst than in Porcupine region, which could explain the difference in BMFs. Wolves seem to biomagnify all investigated PFCs. Caribou to wolf BMFs (whole body) are highest for PFNA (3.8 ( 0.6 and 5.4 ( 0.8) but also PFOA has a BMF above one (2.4 ( 0.6 and 2.1 ( 0.5). There are two BMF values which do not follow the general trend. PFOS BMF was low for the Bathurst food chain with 1.2 ( 0.2 and PFTrA BMF was high with 3.2 ( 1.5. The high PFTrA BMF is probably because concentrations are rather low for both species and tissues and the error is consequently rather high. The low PFOS BMF is reflective of PFOS concentrations being in the same range for Bathurst wolf and caribou. Except for PFOA, caribou seem to biomagnify PFCs to a greater extent than wolves. PFNA to PFTrA values are two to four times 8670
dx.doi.org/10.1021/es201353v |Environ. Sci. Technol. 2011, 45, 8665–8673
Environmental Science & Technology
ARTICLE
Figure 2. TMF values ( standard errors for the Porcupine herd (black dots) and the Bathurst herd (white dots) food chain are shown on the left. The two plots on the right show literature TMFs for (a) an arctic marine mammal,15 (b) a dolphin,1 and (c) a piscivorous food web.15 Fluorinated carbon chain length 7 equals PFOA, 8 equals PFNA and PFOS, 9 equals PFDA etc.
higher for caribou, which might be explained by higher food assimilation efficiency of caribou. Previous (aquatic) food web studies have found that biomagnification is dependent on the chain length of PFCAs and PFSAs: highest BMFs are mostly found for PFNA to PFUnA and PFOS, but the magnitude of BMF varies (see Table 2). The only other published mammal to mammal BMF is for seal to polar bear on a liver concentration basis.4 About 10 times higher BMFs were observed for this food chain but this comparison is limited since PFC body distribution might vary for the organisms. A whole body BMF comparison would be more significant. Dolphin from the Charleston SC area had BMFs (seatrout as the diet) with similar magnitude (∼2) and chain length dependence (maximum BMF for PFNA to PFUnA) as did wolves from this study. Piscivorous food webs do not show consistent BMFs. Lake trout magnify all PFCAs and PFOS but not PFOA,3 which could mean that the less proteinophilic PFCA39,40 is not enriched in the water breathing organisms in contrast to the air breathing mammals. In general, it can be said that PFCAs from PFNA to PFDoA, as well as PFOS, biomagnify in the lichencaribouwolf food chain and that PFOA shows only moderate enrichment for caribou. The comparison of wolfcaribou and cariboulichen BMFs shows that the biomagnification for these two species has different chain length dependence demonstrating that PFC biomagnification is highly species-specific and may be dependent on many factors including protein composition. 33,39,41 Calculation of whole body concentrations may help to understand biomagnification of PFCs. However, there are certain disadvantages to this method: it is very complex and laborious to obtain all information needed to calculate whole body concentrations for larger animals. Additionally, concentrations or body composition have to estimated, when not all information is available, which introduces uncertainties. Trophic Magnification. Determining the biomagnification of PFCs in a whole food web with TMFs is equally as challenging as determining BMFs as it is not possible to use lipid normalized concentrations. In this study, the estimated whole body concentration in caribou and wolf was used for TMF calculation and the first trophic level was assumed to be lichen or all vegetation (Table 2). Additionally, a separate TMF calculation was done using liver. Statistically significant relationships between trophic level and logarithmic concentrations (ng/g ww) were found for
PFOS and almost all PFCAs, except PFOA (see SI Table S14). TMFs ranged between 1 and 3 for both food chains with highest values for PFNA to PFUnA. PFOS had very similar TMF values compared to its carboxylate counterpart PFNA (e.g., 2.7 ( 0.2 for PFNA and 2.6 ( 0.1 for PFOS for the Porcupine herd). This indicates that the biomagnification process is mainly dependent on the fluorinated chain and not on the functional group. Both food webs show very similar TMFs, the only major difference can be seen for PFDoA and PFTrA. The Bathurst food web has higher TMFs (2.2 ( 0.1for PFDoA and 2.0 ( 0.1 for PFTrA, respectively) than the Porcupine herd (1.4 for both PFDoA and PFTrA). This is because PFCA levels are slightly higher for caribou in the Bathurst herd. Other vegetation could potentially influence this difference (e.g., higher concentrations of PFDoA/ PFTrA in plants from the Porcupine herd range compared to the Bathurst range). The incorporation of all other vegetation into TMF calculation does, however, not change the TMF dramatically. All values are in the same range as with lichen only. Not surprisingly, when liver is used as the basis of TMF calculations, higher concentrations in caribou and wolf liver give rise to a greater slope for the linear regression of concentration versus trophic level. TMFs are consistently two to three times higher and the overall trend remains with highest enrichment for PFNA to PFUnA. This shows that liver-based TMFs will overestimate the biomagnification and that whole body calculations can help to more accurately approximate actual biomagnification behavior. This is particularly important for PFOA, because according to its whole body TMF, it is not statistically biomagnified, whereas liver TMFs would indicate that PFOA is bioaccumulative. A similar relationship with the chain length can be observed for TMFs from the Arctic marine mammal food web, but the absolute values are around two times lower for this study (Figure 2).15 PFOS shows a much stronger biomagnification behavior in the marine food web. Studies of the lake trout food web in Lake Ontario show a range of TMFs for PFOS from around 13.8.3,42 Respiratory elimination of PFCs is important in water breathing organisms, as PFCAs and PFOS are relatively water-soluble and hydrophilic.15 Air breathing animals cannot eliminate anionic PFCs over the lungs, as these PFCs are not volatile. Especially in food webs near source regions, nonvolatile precursors such as perfluoroalkyl phosphates may play an 8671
dx.doi.org/10.1021/es201353v |Environ. Sci. Technol. 2011, 45, 8665–8673
Environmental Science & Technology additional role. These precursors enter the surface waters via wastewater treatment plants and could be metabolized by higher organisms, thereby increasing the concentration of PFCAs and PFSAs in higher trophic levels and leading to overestimates of BMF and TMF in source areas.43 On the other hand, it is possible that semivolatile precursors of PFOS, such as the perfluorosulfonamido alcohols, which have been detected in Arctic air,31,44 as well as nonvolatile precursors such as perfluoroalkyl phosphates transported on particles, could be present in vegetation and then metabolized by caribou. However, no measurements of these precursors in vegetation have been made to our knowledge. It is therefore difficult to draw a clear conclusion on whether terrestrial/marine mammalian differs from piscivorous magnification. Overall Implications. This study shows that even in remote areas PFCs are present in terrestrial animals and vegetation because of atmospheric transport and deposition. PFCAs and PFOS have clear biomagnification potential similar to marine mammalian food webs. This study also shows that challenges remain for applying the TMF approach to studies of terrestrial biomagnification of PFCAs and PFSAs and indeed other persistent and bioaccumulative organics. Seasonal variation in δ13C and δ15N in vegetation will be important in temperate and arctic climates as will dietary variation in herbivores. Selection of the species used to calculate trophic levels and of the appropriate trophic enrichment factor are important factors which need further study in terrestrial food webs.
’ ASSOCIATED CONTENT
bS
Supporting Information. More information on analytical methods, sampling locations, standard chemicals, data sets on all concentrations, TMF calculations, and further details are shown in Figures S1S6 and Tables S1S14.This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +1-905-319-6921. Fax: +1-905-336-6430. E-mail: derek.
[email protected]. )
Present Addresses
CBET, Waterloo ON, Canada.
’ ACKNOWLEDGMENT The authors thank Yukon and NWT hunters and trappers who provided caribou and wolf samples for the project, and Bruno Croft and Dean Cluff (Dept of Environment and Natural Resources, Government of the NWT, Yellowknife) who organized those collections in NWT. We also thank members of the Tr’ondek Hwech’in First Nation, Aurora Research Institute and Don Reid (Wildlife Conservation Society) who assisted with plant collections. We are also grateful to Keith Hobson (Environment Canada, Saskatoon) for his help with the stable isotope data interpretation. Funding for the project was provided by Northern Contaminants Program (Indian and Northern Affairs Canada) and Chemical Management Plan (Environment Canada). The first author was financially supported by the Swiss Federal Office for the Environment (FOEN).
ARTICLE
’ REFERENCES (1) Houde, M.; Bujas, T. A. D.; Small, J.; Wells, R. S.; Fair, P. A.; Bossart, G. D.; Solomon, K. R.; Muir, D. C. G. Biomagnification of perfluoroalkyl compounds in the bottlenose dolphin (Tursiops truncatus) food web. Environ. Sci. Technol. 2006, 40 (13), 4138–4144. (2) Tomy, G. T.; Budakowski, W.; Halldorson, T.; Helm, P. A.; Stern, G. A.; Friesen, K.; Pepper, K.; Tittlemier, S. A.; Fisk, A. T. Fluorinated organic compounds in an Eastern arctic marine food web. Environ. Sci. Technol. 2004, 38 (24), 6475–6481. (3) Martin, J. W.; Whittle, D. M.; Muir, D. C. G.; Mabury, S. A. Perfluoroalkyl contaminants in a food web from Lake Ontario. Environ. Sci. Technol. 2004, 38 (20), 5379–5385. (4) Martin, J. W.; Smithwick, M. M.; Braune, B. M.; Hoekstra, P. F.; Muir, D. C. G.; Mabury, S. A. Identification of long-chain perfluorinated acids in biota from the Canadian Arctic. Environ. Sci. Technol. 2004, 38 (2), 373–380. (5) Olsen, G. W.; Church, T. R.; Miller, J. P.; Burris, J. M.; Hansen, K. J.; Lundberg, J. K.; Armitage, J. B.; Herron, R. M.; Medhdizadehkashi, Z.; Nobiletti, J. B.; O’Neil, E. M.; Mandel, J. H.; Zobel, L. R. Perfluorooctanesulfonate and other fluorochemicals in the serum of American Red Cross adult blood donors. Environ. Health Persp. 2003, 111 (16), 1892–1901. (6) Kannan, K.; Koistinen, J.; Beckmen, K.; Evans, T.; Gorzelany, J. F.; Hansen, K. J.; Jones, P. D.; Helle, E.; Nyman, M.; Giesy, J. P. Accumulation of perfluorooctane sulfonate in marine mammals. Environ. Sci. Technol. 2001, 35 (8), 1593–1598. (7) Giesy, J. P.; Kannan, K. Global distribution of perfluorooctane sulfonate in wildlife. Environ. Sci. Technol. 2001, 35 (7), 1339–1342. (8) Ahrens, L.; Barber, J. L.; Xie, Z.; Ebinghaus, R. Longitudinal and latitudinal distribution of perfluoroalkyl compounds in the surface water of the atlantic ocean. Environ. Sci. Technol. 2009, 43 (9), 3122–3127. (9) Wei, S.; Chen, L. Q.; Taniyasu, S.; So, M. K.; Murphy, M. B.; Yamashita, N.; Yeung, L. W. Y.; Lam, P. K. S. Distribution of perfluorinated compounds in surface seawaters between Asia and Antarctica. Mar. Pollut. Bull. 2007, 54 (11), 1813–1818. (10) Yamashita, N.; Taniyasu, S.; Petrick, G.; Wei, S.; Gamo, T.; Lam, P. K. S.; Kannan, K. Perfluorinated acids as novel chemical tracers of global circulation of ocean waters. Chemosphere 2008, 70 (7), 1247–1255. (11) D’Eon, J. C.; Hurley, M. D.; Wallington, T. J.; Mabury, S. A. Atmospheric chemistry of N-methyl perfluorobutane sulfonamidoethanol, C4F9SO2N(CH3)CH2CH 3 2OH: Kinetics and mechanism of reaction with OH. Environ. Sci. Technol. 2006, 40 (6), 1862–1868. (12) Ellis, D. A.; Martin, J. W.; De Silva, A. O.; Mabury, S. A.; Hurley, M. D.; Sulbaek Andersen, M. P.; Wallington, T. J. Degradation of fluorotelomer alcohols: A likely atmospheric source of perfluorinated carboxylic acids. Environ. Sci. Technol. 2004, 38 (12), 3316–21. (13) Martin, J. W.; Ellis, D. A.; Mabury, S. A.; Hurley, M. D.; Wallington, T. J. Atmospheric chemistry of perfluoroalkanesulfonamides: Kinetic and product studies of the OH radical and Cl atom initiated oxidation of N-ethyl perfluorobutanesulfonamide. Environ. Sci. Technol. 2006, 40 (3), 864–872. (14) Haukas, M.; Berger, U.; Hop, H.; Gulliksen, B.; Gabrielsen, G. W. Bioaccumulation of per- and polyfluorinated alkyl substances(PFAS) in selected species from the Barents Sea food web. Environ. Pollut. 2007, 148 (1), 360–371. (15) Kelly, B. C.; Ikonomou, M. G.; Blair, J. D.; Surridge, B.; Hoover, D.; Grace, R.; Gobas, F. A. P. C. Perfluoroalkyl contaminants in an arctic marine food web: Trophic magnification and wildlife exposure. Environ. Sci. Technol. 2009, 43 (11), 4037–4043. (16) Powley, C. R.; George, S. W.; Russell, M. H.; Hoke, R. A.; Buck, R. C. Polyfluorinated chemicals in a spatially and temporally integrated food web in the Western Arctic. Chemosphere 2008, 70 (4), 664–672. (17) Tomy, G. T.; Pleskach, K.; Ferguson, S. H.; Hare, J.; Stern, G.; MacInnis, G.; Marvin, C. H.; Loseto, L. Trophodynamics of some PFCs and BFRs in a Western Canadian arctic marine food web. Environ. Sci. Technol. 2009, 43 (11), 4076–4081. 8672
dx.doi.org/10.1021/es201353v |Environ. Sci. Technol. 2011, 45, 8665–8673
Environmental Science & Technology (18) Ostertag, S. K.; Tague, B. A.; Humphries, M. M.; Tittlemier, S. A.; Chan, H. M. Estimated dietary exposure to fluorinated compounds from traditional foods among Inuit in Nunavut, Canada. Chemosphere 2009, 75 (9), 1165–1172. (19) Czub, G.; Wania, F.; McLachlan, M. S. Combining long-range transport and bioaccumulation considerations to identify potential arctic contaminants. Environ. Sci. Technol. 2008, 42 (10), 3704–3709. (20) Kelly, B. C.; Ikonomou, M. G.; Blair, J. D.; Morin, A. E.; Gobas, F. A. P. C. Food web-specific biomagnification of persistent organic pollutants. Science 2007, 317 (5835), 236–239. (21) Hansen, J. C.; Van Oostdam, J.; Gilman, A.; Odland, J.; Vaktskjold, A.; Dudarev, A. AMAP Assessment 2009: Human Health in the Arctic. 2009; AMAP: Oslo, 2009. (22) Powley, C. R.; George, S. W.; Ryan, T. W.; Buck, R. C. Matrix effect-free analytical methods for determination of perfluorinated carboxylic acids in environmental matrixes. Anal. Chem. 2005, 77 (19), 6353–6358. (23) Jardine, T. D.; Kidd, K. A.; Fisk, A. T. Applications, considerations, and sources of uncertainty when using stable isotope analysis in ecotoxicology. Environ. Sci. Technol. 2006, 40 (24), 7501–7511. (24) Borga, K.; Kidd, K. A.; Muir, D. C. G.; Berglund, O.; Conder, J. M.; Gobas, F. A. P. C.; Kucklick, J.; Malm, O.; Powell, D. E. Trophic magnification factors: Considerations of ecology, ecosystems and study design. Environ. Assess. Manag. 2011, (DOI: 10.1002/ieam.244). (25) Ben-David, M.; Shochat, E.; Adams, L. G. Utility of stable isotope analysis in studying foraging ecology of herbivores: Examples from moose and caribou. Alces 2001, 37 (2), 421–434. (26) Higgins, C. P.; Luthy, R. G. Sorption of perfluorinated surfactants on sediments. Environ. Sci. Technol. 2006, 40 (23), 7251–7256. (27) Dreyer, A.; Matthias, V.; Weinberg, I.; Ebinghaus, R. Wet deposition of poly- and perfluorinated compounds in Northern Germany. Environ. Pollut. 2010, 158 (5), 1221–1227. (28) Scott, B. F.; Spencer, C.; Mabury, S. A.; Muir, D. C. G. Poly and perfluorinated carboxylates in north American precipitation. Environ. Sci. Technol. 2006, 40 (23), 7167–7174. (29) Scott, B. F.; Spencer, C.; de Silva, A. O.; Muir, D. C. G., PFAs in precipitation at Snare Rapids NT. Unpublished data. Aquatic Ecosystem Protection Research Division, Environment Canada, Burlington ON. (30) Young, C. J.; Furdui, V. I.; Franklin, J.; Koerner, R. M.; Muir, D. C. G.; Mabury, S. A. Perfluorinated Acids in Arctic Snow: New Evidence for Atmospheric Formation. Environ. Sci. Technol. 2007, 41 (10), 3455–3461. (31) Stock, N. L.; Furdui, V. I.; Muir, D. C. G.; Mabury, S. A. Perfluoroalkyl Contaminants in the Canadian Arctic: Evidence of Atmospheric Transport and Local Contamination. Environ. Sci. Technol. 2007, 41 (10), 3529–3536. (32) Han, X.; Snow, T. A.; Kemper, R. A.; Jepson, G. W. Binding of perfluorooctanoic acid to rat and human plasma proteins. Chem. Res. Toxicol. 2003, 16 (6), 775–781. (33) Jones, P. D.; Hu, W.; De Coen, W.; Newsted, J. L.; Giesy, J. P. Binding of perfluorinated fatty acids to serum proteins. Environ. Toxicol. Chem. 2003, 22 (11), 2639–2649. (34) Martin, J. W.; Mabury, S. A.; Solomon, K. R.; Muir, D. C. G. Bioconcentration and tissue distribution of perfluorinated acids in rainbow trout(Oncorhynchus mykiss). Environ. Toxicol. Chem. 2003, 22 (1), 196–204. (35) Butt, C. M.; Mabury, S. A.; Kwan, M.; Wang, X.; Muir, D. C. G. Spatial trends of perfluoroalkyl compounds in ringed seals(Phoca hispida) from the Canadian arctic. Environ. Toxicol. Chem. 2008, 27 (3), 542–553. (36) Darr, R. L.; Hewitt, D. G. Stable isotope trophic shifts in whitetailed deer. J. Wildlife Manage 2008, 72, 1525–1531. (37) Hobson, K. A.; Alisauskas, R. T.; Clark, R. G. Stable-nitrogen iostope enrichment in avain tissues due to fasting and nutritional stress: Implications for isotopic analyses of diet. The Condor 1993, 95, 388–394. (38) Phillips, D. L.; Gregg, J. W. Source partitioning using stable isotopes: Coping with too many sources. Oecologia 2003, 136 (2), 261–269. (39) Bischel, H. N.; MacManus-Spencer, L. A.; Luthy, R. G. Noncovalent interactions of long-chain perfluoroalkyl acids with serum albumin. Environ. Sci. Technol. 2010, 44 (13), 5263–5269.
ARTICLE
(40) MacManus-Spencer, L. A.; Tse, M. L.; Hebert, P. C.; Bischel, H. N.; Luthy, R. G. Binding of perfluorocarboxylates to serum albumin: A comparison of analytical methods. Anal. Chem. 2009, 82 (3), 974–981. (41) Luebker, D. J.; Hansen, K. J.; Bass, N. M.; Butenhoff, J. L.; Seacat, A. M. Interactions of flurochemicals with rat liver fatty acidbinding protein. Toxicology 2002, 176 (3), 175–185. (42) Houde, M.; Czub, G.; Small, J. M.; Backus, S.; Wang, X.; Alaee, M.; Muir, D. C. G. Fractionation and bioaccumulation of perfluorooctane sulfonate (PFOS) isomers in a Lake Ontario food web. Environ. Sci. Technol. 2008, 42 (24), 9397–9403. (43) Lee, H.; D’eon, J.; Mabury, S. A. Biodegradation of polyfluoroalkyl phosphates as a source of perfluorinated acids to the environment. Environ. Sci. Technol. 2010, 44 (9), 3305–3310. (44) Shoeib, M.; Harner, T.; Vlahos, P. Perfluorinated chemicals in the Arctic atmosphere. Environ. Sci. Technol. 2006, 40 (24), 7577–7583.
8673
dx.doi.org/10.1021/es201353v |Environ. Sci. Technol. 2011, 45, 8665–8673
ARTICLE pubs.acs.org/est
Major Issues Regarding the Efficiency of Monitoring Programs for Nitrate Contaminated Groundwater T. Y. Stigter,*,† A. M. M. Carvalho Dill,‡ and L. Ribeiro† † ‡
Geo-Systems Centre/CVRM - Instituto Superior Tecnico, Av. Rovisco Pais, 1049-001 Lisbon, Portugal Geo-Systems Centre/CVRM - Universidade do Algarve, Campus de Gambelas, 8005-139 Faro, Portugal ABSTRACT: Major issues regarding the efficiency of monitoring programs for nitrate contaminated groundwater are analyzed in this paper: (i) representativeness of monitoring networks; (ii) correct interpretation of the monitoring data and resulting time series and trends; and (iii) differentiation among the different sources of nitrates in groundwater. Following an overview of the nitrate contamination problem and possible solutions, as well as some of the difficulties found, a relatively straightforward method for assessing monitoring network representativity is presented, namely interpolation standard error assessment. It is shown how nitrate-concentration time series resulting from periodic observations can be corrected with a conservative tracer, in order to avoid misinterpretation and confirm or correct apparent trends. Finally, coupled 15N and 18O isotope signatures of nitrate (NO3) in groundwater are used to differentiate among nitrogen (N) sources, to ensure correct targeting of restoration measures. The case study regards a Nitrate Vulnerable Zone in the south of Portugal, designated in compliance with the European Nitrates Directive, where coastal discharge of nutrient-rich groundwater threatens the good qualitative and ecological status of the Ria Formosa coastal lagoon. Results show that mineral fertilizer is the main source of N in groundwater, and that increases in N load can be masked by dilution phenomena.
’ INTRODUCTION Groundwater contamination by nitrate (NO3, hereafter referred to as NO3) has been recognized for decades as a major concern for mankind. Despite the existence of a large number of point and diffuse sources of nitrogen (N), including septic tanks, sewage discharge, cattle feedlots, oxidation of soil organic N and atmospheric deposition, the excessive application of mineral and organic fertilizers in agriculture is the principal source of N in groundwater on a regional scale.1,2 In Europe, at a large scale agriculture contributes to 5080% of the total N load on the aquatic environment.3 Protecting and restoring the NO3contaminated groundwater resources in the world not only concerns their direct use (i.e., consumption), but also the impact that groundwater discharge has on the qualitative and ecological status of surface water bodies, particularly coastal estuaries and lagoons. This is because ratios of dissolved inorganic N over phosphorus (P) in contaminated aquifers are usually much higher than the Redfield ratio that defines the nutrient requirements by phytoplankton,4 contrary to what occurs in natural coastal waters, which are often N-limited ecosystems.58 There is much reported evidence that in the past few decades increased inputs of N to coastal waters have led to a global increase in eutrophication, resulting in widespread hypoxia and anoxia, habitat degradation, and increased occurrence of harmful algal blooms.916 Goals to reduce the volume of NO3 transport in groundwater can be approached by controlling the sources or by enhancing the r 2011 American Chemical Society
sinks. As NO3 neither forms insoluble minerals that can precipitate, nor is significantly adsorbed, its removal from groundwater predominantly occurs through denitrification.17 However, this process is largely restricted to anaerobic environments, which are often difficult to obtain in shallow and even in deep aquifer systems.18,19 The control of the N source is therefore considered more effective, but depends on the possibility to implement such measures, which in turn depends on the type of source. Point sources are generally easier to control, and progress has been made in Europe and the United States in reducing nutrient pollution from wastewater sources.20 Reducing nonpoint source pollution has proven much more difficult.10,20,21 In the European Union the EC Nitrates Directive was drawn up with the specific purpose to reduce water pollution caused by NO3 from agricultural sources and prevent further such pollution. EU member states had to identify waters affected by NO3 pollution, designate Nitrate Vulnerable Zones (NVZs), implement monitoring programs, and establish action programs and codes of good agricultural practice containing mandatory measures concerning the land application and storage of mineral and organic fertilizers.22,23 Despite the progress made in recent years, many areas still require NVZ designation, particularly with regard Received: May 26, 2011 Accepted: August 24, 2011 Revised: August 5, 2011 Published: August 24, 2011 8674
dx.doi.org/10.1021/es201798g | Environ. Sci. Technol. 2011, 45, 8674–8682
Environmental Science & Technology
ARTICLE
to eutrophication and shallow groundwater contamination criteria.24 Moreover, one-third of the monitoring stations located in the 15 countries that formed the EU before 2004 showed an upward NO3 trend in 20042007.25,26 The most promising groundwater restoration measures are targeted toward (i) an increase in nutrient efficiency and (ii) an improvement in nutrient balance in the application of fertilizers, with adequate timings based on crop requirements.21,2731 Overfertilization is a major problem in many areas, since farmers believe it guarantees maximum crop yield and quality, even though adding fertilizer N follows the classical crop response curve for yield-limiting nutrients as presented by Dibb,32 i.e., it significantly increases crop yield only up to a certain point, after which nutrient efficiency drops rapidly. Farmers also largely underestimate the N available from other sources, such as residues from previous crop cycles, soil, or irrigation water.20 As a result, in the U.S. about one-third of the N applied as fertilizer to the crops is lost to the environment.27 The lack of balanced nutrition may also be an important factor, as the presence of other macronutrient elements such as potassium and phosphorus in adequate proportions in soil and fertilizer is essential for N recovery by the crops (e.g., refs 3337). Comprehensive water monitoring networks lie at the basis of policy implementations, and significant progress in the EU26 has made spatial and temporal trend analysis possible. However, it is important to understand if these networks are sufficiently representative, an issue that is often neglected. Moreover, trends in observed NO3 concentrations can be misleading and it is important to separate variations caused by actual N sinks/ sources from those that are a consequence of dilution/concentration events. For instance, a decrease in dilution potential caused by a reduction in aquifer recharge, expected to occur in many regions due to climate change (e.g., 38,39), may increase NO3 concentrations even when the N load is reduced. Finally, in areas with more than one N source, differentiation is essential to support adequate policy implementation (e.g., 18,23). This paper will deal with these three issues of monitoring: (i) representativity, (ii) result interpretation, and (iii) source differentiation, supported by a case study in the south of Portugal.
’ METHODS Structural Analysis of the Spatial Distribution of NO3. Regarding the spatial distribution of NO3 concentrations, its structural analysis can be performed by calculating the data’s experimental variogram 2γ*(h), the arithmetic mean of the squared difference in value between pairs of data a distance and direction h apart:41,42
2γðhÞ ¼
1 nh ðCx Cxþh Þ2i nh i ¼ 1
∑
errors may account for a certain fraction of the total data variance (nugget effect) visible at the origin. Different spatial orientations should be analyzed to detect the presence of anisotropy. A theoretical model, frequently spherical in geostatistical applications to hydrogeology when no trend exists, is fitted to the experimental variogram. The obtained model parameter values are subsequently introduced into a kriging interpolation algorithm, based on the minimum error variance concept. Ordinary kriging assumes a constant but unknown mean. A large advantage of kriging is that it provides a measure of the estimation error, as the corresponding standard deviations of the estimation error are calculated, also known as standard errors (SE). A larger SE denotes a lower reliability of estimation. At short distances from sampled locations the SE is largely determined by the nugget effect, and as the distance increases, the error depends on the distribution of the sample network, as this determines the length of the search radius with respect to the range of influence in the experimental variogram. A large range of influence, particularly when combined with a high sill and a low nugget effect guarantees a trustworthy interpolation, since the variability among the data and thus also the spread of the SE is still relatively small at the maximum search radius. As a thumb rule, a SE larger than the original sample standard deviation (SD) denotes an unreliable estimate.42 The spatial distribution of the SE is an indicator of the representativeness of implemented monitoring networks. NO3 Time Series Analysis. NO3 often is a very stable and conservative contaminant, particularly in aerobic conditions. A conservative tracer can be used to separate N sources/sinks from dilution/concentration events. In many cases the conservative tracer can be chloride (Cl, hereafter referred to as Cl), but only if no significant sources of this ion exist in the aquifer, so that its variations truly reflect groundwater concentration (due to evapotranspiration) or dilution (due to mixing with fresh water). Where Cl is used in fertilizers, or where significant Cl sources exist, e.g., due to seawater intrusion or evaporate dissolution, its use as a tracer is pointless. To apply the correction to observed NO3 concentration time series, first a concentration factor between subsequent observation dates (time points) has to be calculated: mClðj, ti Þ ð2Þ Fj, ti ¼ mClðj, ti1 Þ where Fj,ti is the concentration factor in observation well j at time point ti, mCl(j,ti ) the Cl concentration in well j at time point ti, and mCl(j,ti-1) is the Cl concentration (in the same units) in well j at a previous time point ti-1. Values of Fj,ti > 1 indicate that concentration took place, whereas values <1 indicate dilution. The corresponding “conservative” NO3 concentration of sample j for each time step i can be calculated as mcNO3ðj, t Þ ¼ Fj, ti mcNO3ðj, t
ð1Þ
where nh is the number of pairs of samples a distance h apart, and Cx and Cx+h are the concentrations at location x and x + h of sample pair i. In the case of an irregular sample grid a tolerance on the specification h is placed,42 the size of which depends on the variable’s spatial structure and the density of the sample network. Plotting the values of 2γ*(h) against h, to obtain the experimental function of the (semi)variogram, reveals the distance (range of influence) and associated value (sill) at which the data become independent of each other. Small-scale variability and measurement
i
i1 Þ
ð3Þ
where mcNO3(j,ti ) and mcNO3(j,ti-1) are the conservative NO3 concentrations in observation well j at time points ti and ti-1, respectively. In a time series of well j starting at t0, the conservative NO3 concentration for the first time point is calculated as follows: mcNO3ðj, t Þ ¼ Fj, t1 mNO3ðj, t Þ 1
0
ð4Þ
where mNO3(j,t0) is the observed NO3 concentration at t0 and Fj,t1 the concentration factor at time point t1. From this point forward, subsequent conservative NO3 concentrations can be calculated 8675
dx.doi.org/10.1021/es201798g |Environ. Sci. Technol. 2011, 45, 8674–8682
Environmental Science & Technology
ARTICLE
using eq 3. The sink or source of N could then be simply calculated by subtracting conservative from observed NO3 (corrected for the N/NO3 molecular weight ratio), but only if dilution/concentration events were considered to occur before addition/removal of N. This could result in an overestimation of N sources and sinks (as this implies that these are not influenced by dilution/concentration). It is more realistic to consider simultaneous occurrence of these phenomena. The sink or source of N is then calculated using the following equation: ! pffiffiffiffiffiffiffi 14 mNO3ðj, ti Þ pffiffiffiffiffiffiffi mNO3ðj, t Þ Fj, ti ð5Þ mNðj, ti Þ ¼ i1 62 Fj, ti where mN(j,ti ) is the net N source/sink for observation well j at time point i and 14/62 is the N/NO3 molecular weight ratio). By using the square-root of Fj,ti the concentration factor is allocated evenly to the observed NO3 concentrations of two subsequent time steps. Finally, the cumulative N sink/source for well j at time point i = n is calculated as 0 1 n
∑ mN i¼1
ðj, ti Þ
¼
pffiffiffiffiffiffiffi B B mNðj, ti Þ ffi þ Fj, ti Brffiffiffiffiffiffiffiffiffiffiffiffiffi n @ Q Fj, ti
n
∑ mN i¼1
C C C ðj, ti1 Þ A
ð6Þ
i¼1
Equation 6 accounts for the different concentration factors that occur between subsequent observation time points. The calculations allow the analysis of the variation of net losses or gains of N for each combination of observation well and date (time point). Negative values represent a sink at the location of the well, i.e., an actual decrease of N that cannot be explained by dilution, whereas positive values indicate a source, i.e., an increase in N that is not caused by concentration. Concentration and dilution can also be represented by mixing with water of higher and lower Cl content, respectively. If NO3/Cl ratios of mixing waters are similar, then mixing will result in mcNO3= mNO3, whereas mixing with higher or lower NO3/Cl water will result in the calculation of a N source or sink, respectively. Differentiation among the Different NO3 Sources. Stable isotope ratios of N (15N/14N) have been widely applied to distinguish between various sources of NO3 in groundwater (e.g., 4345). These sources include mineral and organic fertilizers, septic tanks, sewage discharge, cattle feed lots, oxidation of soil organics, and atmospheric deposition. Later on, the oxygen (O) isotope ratios in NO3 (18O/16O) were combined with those of N (e.g., 4648) to increase the differentiation potential for N sources and biochemical processes affecting the fate of N in soil and groundwater, such as volatilization, nitrification, or denitrification.43,49 N isotope ratios are reported in per mil (%) relative to N2 in atmospheric air, using the delta (δ) notation: δ15 N ¼
ð15 N=14 NÞsample ð15 N=14 NÞair ð15 N=14 NÞair
1000%
The δ18O values of NO3 are reported in % relative to the standard V-SMOW (Vienna Standard Mean Ocean Water): δ18 O ¼
ð18 O=16 OÞsample ð18 O=16 OÞSMOW ð18 O=16 OÞSMOW
1000%
Mineral N fertilizers are typically much lighter in 15N (δ15N 4 to 4%) than organic waste such as manure and septic tanks (δ15N 10% 20%).43,49,50 Regarding δ18O, mineral NO3 fertilizers are
significantly more positive than most other sources, including NO3 derived from the nitrification of mineral ammonium (NH4+, hereafter referred to as NH4) fertilizers. The isotopic composition of oxygen in chemical NO3 fertilizers is similar to air oxygen, which is about 23.5%,51 whereas microbial nitrification results in δ18O values of 2 to 6%.52 Denitrification can cause enrichment in heavy isotopes of N and O in the residual NO3.49 Study Area. Largely affected by agricultural practices that caused NO3 contamination, the NVZ of Faro in the South of Portugal was designated in 1997 and enlarged in 2003, comprising an area of 98 km2. Groundwater in the upper aquifer, which consists of Miocene sand and Plio-Quaternary sand and gravel, reveals the highest NO3 concentrations,18,40,53,54 whereas the lower, semiconfined aquifer, built up of Miocene sandy limestones, is less affected. The area is bordered to the south by the Ria Formosa lagoon, a highly valuable and sensitive wetland, and a designated RAMSAR and Natura 2000 site. The impact of nutrient-rich groundwater discharge on the qualitative and ecological status of the lagoon is a subject of ongoing research. In quantitative terms, the total N load is estimated to be significant, larger than stream runoff and of the same order of magnitude as effluent discharge from treatment plants.55 Average annual rainfall in the area is around 600 mm, and groundwater recharge mainly takes place toward the north, where Jurassic limestones crop out.40,53 Irrigated citrus orchards are the prevailing crop type.40 Greenhouse crops that once dominated the southern part of the area are currently losing importance, partly being substituted by citrus culture. An average amount of 150300 kg/ha 3 yr of fertilizer N is applied for citrus trees, whereas for crops such as tomato and melon, both grown in the study area, quantities are considered per crop cycle ((half a year) and equal 150200 kg/ha and 50100 kg/ha, respectively.56 NPK fertilizers, ammonium sulfate, and urea are the main used fertilizers, with fewer applications of nitrate fertilizers. Cl containing fertilizers are not used. Data on dissolved NO3 and Cl concentrations in groundwater were obtained from the official monitoring program operated by the Algarve River Basin Administration. This official entity samples groundwater on a biannual basis. Some monitoring wells had to be excluded from the study, either due to a lack of data (wells that were only used for monitoring for short periods) or because they monitor deeper aquifers (below 100 m) with a completely different hydrochemical signature. Care was also taken to avoid the use of wells near the coast that contain additional Cl sources from local seawater mixing and evaporite dissolution.18 Here, the additional source would make it useless to perform dilution/concentration calculations. The main input of Cl is from dry and wet atmospheric deposition, with an average estimated yearly input of 80100 kg/ha.57 Local point-source inputs from septic tanks can provide an additional 510 kg/ha of Cl (as well as N57). Besides constituting only 510% of atmospheric input, wastewater has higher NO3/Cl ratios, so that when calculating conservative NO3 based on Cl concentrations, the interference of “septic tank Cl” will be negligible. A total of 28 monitoring wells were used for the study. In addition, five shallow wells and a deep well tapping the sandy limestone aquifer were sampled in June 2009 for analysis of NO3 and its isotopes 18O and 15N, after pumping sufficiently long to ensure aquifer condition representativity. Electrical conductivity (EC), temperature, alkalinity (field titration), and NO3 concentration (test strips Merckoquant) were measured in the field. Samples were filtered and for cation analysis acidified to a pH below 2. The samples 8676
dx.doi.org/10.1021/es201798g |Environ. Sci. Technol. 2011, 45, 8674–8682
Environmental Science & Technology
ARTICLE
Figure 1. (top) Spatial distribution of median NO3 concentrations in 19972000 and in 20072010; also shown is the location of the monitoring wells, the wells sampled for nitrate isotopes (+ shallow wells, O deep well) and the streams (dashed lines); groundwater flow in the area is N to S; (bottom) corresponding maps of kriging error standard deviations, plotted as % of original data standard deviation.
were sent to two different laboratories, the Universidade de Tras-osMontes e Alto Douro in Vila Real for major ion analysis and the Stable Isotope Laboratory of the University of East Anglia in Norwich for 18 O and 15N analysis, following described procedures.58,59
’ RESULTS AND DISCUSSION Monitoring Network Representativity. Figure 1 presents maps of the spatial distribution of median NO3 concentrations and related standard errors (SE) of interpolation, for the NVZ of Faro in 19972000 and 20072010. The SE has been plotted as percent of the original data standard deviation (SD). Near the monitoring points the SE is smallest, and it increases with distance, according to the variographic configuration of the data. Areas are left blank either where there are insufficient data within the search radius of the interpolator or where interpolation is not reliable, based on Clark’s thumb rule,42 i.e., a SE below the SD of the original data.42 For both periods the maps reveal those areas where interpolation is least reliable. However, particularly in the case of 20072010, the application of the SE = SD criterion still results in some interpolated areas located far away from the observation points, giving the impression that they are more a result of extrapolation rather than interpolation. The SE limit of 75% of the original SD seems to provide a more accurate idea of the network representativity. It can be observed that in 19972000 the monitoring program was clearly insufficient. In compliance with the Nitrates Directive,
several monitoring wells were included, but the recent situation still reveals several gaps on the SE map, either due to the absence of observation points or because the wells are too deep to represent the upper aquifers. It can therefore be concluded that relatively large areas within the designated NVZ are currently not (sufficiently) monitored. On the other hand, the central area of the NVZ may have an excess of monitoring points that could be reduced to avoid increasing the costs of the monitoring program too much, thereby optimizing the network efficiency. A correct decision should be supported by the analysis of the contamination level time series, as reducing the number of observation points in the most contaminated or most dynamic areas may be inappropriate. Interpretation of Monitoring Results. When comparing the past and present situation, separated by a decade, a first look tells us that the total area of NO3 concentrations above 100 mg/L (yellow and darker) has been slightly reduced. One is tempted to think this might be a consequence of the application of the Action Program implemented in compliance with the Nitrates Directive. Figure 2 shows the NO3 time series for eight observation wells (marked in Figure 1). Indeed many of the wells reveal a decreasing or stabilizing trend, although all wells currently still exceed the threshold value of 50 mg/L. To understand if agricultural practices are in fact changing and consequently reducing the N load on groundwater, it is necessary to eliminate concentration and dilution effects, using a conservative tracer. For this purpose, eqs 26 were applied to each of the monitoring well time series in the study area, revealing the variation of conservative mNO3 (mcNO3, eq 3), and accumulated N 8677
dx.doi.org/10.1021/es201798g |Environ. Sci. Technol. 2011, 45, 8674–8682
Environmental Science & Technology
ARTICLE
Figure 2. Time series of observed NO3 concentration and conservative NO3 concentration (“C” suffix, dashed lines) of eight monitoring wells in the NVZ of Faro, in the northern (top) and southern sector (bottom); also shown is the oscillation of the water table monitored in a well nearby.
(ΣmN, eq 6) at each location. The results are shown for the eight previously selected wells in Figures 2 and 3, respectively. Negative values of ΣmN represent a sink, associated with crop uptake, denitrification, or lateral transport, whereas positive values represent a source, either from vertical (leaching) or lateral transport. Oscillations in mcNO3, which basically reflect dilution and concentration events, seem to respond to seasonal rechargedischarge phenomena, though they are not always consistent and certain time lags are involved. The effect of water level on mcNO3 is more evident: in years with high water levels (such as 20012003) mcNO3 tends to decrease (this effect is enhanced in well 240, which is located near an influent stream), whereas in dry years (lower water table) mcNO3 tends to increase, revealing the concentration effect associated with the lack of recharge. The latter phenomenon will be enhanced in the future, as climate change is expected to significantly lower aquifer recharge in the Mediterranean region (e.g., 38,39). Seasonal trends in ΣmN can also be detected in some wells and are different from those observed for mcNO3. High ΣmN values positively correlate to water level, as NO3 leaching is promoted in autumn and winter, when drainage is high and evapotranspiration is low.60,61 Moreover, as the water table rises, groundwater mixes with percolating water with higher concentrations, causing a “layering” effect that results in the increase in ΣmN. As the water table drops again, mixing with resident groundwater occurs, further promoted by pumping, thereby eliminating this effect. For wells 153, 190, and 241 located more to the north the trend of ΣmN follows the decreasing trend in observed mNO3, whereas mcNO3 shows a relatively stable long-term trend (Figure 2). The net loss of N in this area is mainly a result of lateral outflow of groundwater with higher NO3 concentrations
and replacement by less contaminated groundwater (from vertical and lateral inflow). The importance of improved (i.e., more balanced and efficient) fertilization practices is not clear, as in some cases the decreasing trend starts before the official implementation of the first Action Program in 2001. A similar phenomenon was observed in Denmark,62 largely due to a reduction from wastewater point sources in the rural areas, both of domestic and animal origin. Here, this is not the case, but it is known that the application of the Action Program in the field suffered considerable delay and resistance. Moreover, as groundwater velocities in the upper aquifers of the study site are low,40,53 a time lag is expected between the implementation of measures and the effect on groundwater quality. A possible alternative explanation may be a change in land use. It is not unusual to find horticulture fields that have been abandoned (for socio-economic reasons) or transformed into citrus culture. Of the four wells located more toward the south (lower plot of Figures 2 and 3) two (156, 260) indicate a clearly positive ΣmN at the end of 2010, i.e., a N source, whereas for well 242 the trend was reversed in 2006. Trends of increasing ΣmN are linked to decreasing mcNO3 (dashed lines), revealing the existence of a dilution effect that in some cases masks the net increase in N. The latter is most likely caused by leaching of excess fertilizer and movement of groundwater with higher NO3 concentrations toward the south. Dilution occurred due to an increase in natural recharge in the period 20012004, illustrated by the rise in groundwater head (Figure 2). The shutdown of several municipal wells in 1999, when the public water supply source changed to surface water supplied by large reservoirs, contributed to an increase in groundwater discharge. Increased irrigation using 8678
dx.doi.org/10.1021/es201798g |Environ. Sci. Technol. 2011, 45, 8674–8682
Environmental Science & Technology
ARTICLE
Figure 3. Time series of cumulative nitrogen of monitoring wells in the NVZ of Faro, in the northern (top) and southern sector (bottom); the thick lines represent polynomial fits to the series.
deeper wells that tap more pristine groundwater may also have contributed to a dilution effect.40,53 Now that for each well the accumulated N sink/source (ΣmN) for the 10-year period has been quantified and separated from dilution/concentration effects, ΣmN can be mapped spatially and compared to that of the observed difference in mNO3, i.e., ΔmNO3, over the same period. The purpose is to see if the image changes significantly, i.e., if areas of significant N sources (or sinks) that were masked by dilution (or concentration) are exposed. ΣmN was calculated based on median values for three-year periods, i.e., 19972000 vs 20072010, to eliminate seasonal anomalies. Ordinary kriging was used to produce the results, following variographic analysis and model fitting. Figure 4 presents the results in mg/L N (NO3N) for both maps, and two main observations can be made. First, the map of ΣmN overall presents less negative and more positive delta values than the observed ΔmN map, which means that in some areas actual N sinks are less significant, in some areas they are nonexistent, and in other areas N sources are more significant. Second, there is a relatively large, well-structured area in the southern sector of the NVZ where the N load on groundwater of the upper aquifer is increasing, which was not visible on the observed ΔmN map, as it was masked by dilution phenomena. The correction of obtained N data is particularly useful in areas where dilution and concentration phenomena are significant and may mask the true fate of N in groundwater. One may argue that observed concentrations are the only ones that matter for policymakers. However, it is important to understand the evolution of the actual N load, so as to anticipate what will happen when concentration/dilution phenomena are ceased or reversed. Differentiation of NO3 Sources. Observing the spatial distribution of NO3 concentrations in Figure 1 in more detail, it is clear that the southern and central areas of the NVZ are most contaminated, with concentrations reaching up to 300 mg/L. Though the excessive application of N fertilizers is undoubtedly
the trigger, irrigation with locally extracted groundwater induces a groundwater cycle that gradually increases the NO3 concentrations along with the overall groundwater salinity.40 The results of the NO3 isotope analyses are shown graphically in Figure 5a and b. The δ15N values range from 4.6 to 6.7%, whereas δ18O values range from 4.8 to 6.0%. Results for both isotopes are characterized by a narrow range, indicating little spatial variation (see Figure 1 for location of sampled wells). Figure 5a compares these values with typical ranges found in the literature for the main processes that could explain the high NO3 concentrations found in the study area, namely: (i) leaching of fertilizer NO3; (ii) nitrification of fertilizer NH4 and subsequent leaching; and (iii) leaching of water from septic tanks. The δ18O signature apparently reveals that direct leaching of mineral NO3 fertilizers has not occurred. This does not automatically rule out contamination from NO3 fertilizers, since microbial immobilizationmineralization processes of N in soils can largely alter the δ18O signal of the leachate.63 Notwithstanding, the findings are in agreement with surveys held among the farmers in the region that showed that the main applied fertilizers contain N in the form of NH4.57 The low ranges of both isotopes are indeed consistent with substantial inputs of such fertilizers, though the observed δ15N range of the six samples, +4.6% to +6.7%, is slightly higher than the reported range. Ammonia volatilization43,49,64 and nitrification of reduced fertilizers and soil organic N49,50,64 are particularly important phenomena that can occur and lead to further groundwater enrichment in 15N (indicated by the green dashed lines in Figure 5b). Mixing with N from septic waste could also explain part of the shift toward the right of the diagram, as well as the positive correlation of δ15N with mNO3. Higher NO3 concentrations in the study have previously been related to point source contamination from septic tanks.40,54,57 For instance, considering conservative mixing between two end members: (1) sample with NO3 = 200 mg/L and 8679
dx.doi.org/10.1021/es201798g |Environ. Sci. Technol. 2011, 45, 8674–8682
Environmental Science & Technology
ARTICLE
Figure 4. Spatial distribution of difference in observed N concentrations (left, ΔmN) and of accumulated N corrected for dilution/concentration (right, ΣmN) between the periods 19972000 and 20072010.
Figure 5. (a) δ18O versus δ15N for nitrate concentrations of six groundwater samples in the NVZ of Faro; ranges of values obtained from literature;43,44,46,49,50 dashed green lines indicate potential effects of nitrification or volatilization; (b) δ15N versus mNO3 for the same samples, indicating conservative mixing between sample with lowest δ15N and wastewater with a δ15N of 10% (dashed line) and 20% (continuous line).
δ15N = 4.6%; (2) untreated wastewater with NO3 = 376 mg/L (85 mg/L N-NO3, according to a “strong wastewater” composition65) and δ15N between 10 and 20%, the contribution of wastewater in the sample of δ15N = 6.7% would be between 8 and 25% (mixing is represented by the curves in Figure 5b). When considering δ15N = 0% and NO3 = 200 mg/L for a “fertilizer” end member, then the sample with the lowest δ15N would have a 19% “septic waste” component. The obtained results in the case study illustrate the importance of representative monitoring and correct data interpretation, which can largely increase the efficiency as well as the rate and duration of success of restoration policies. Unfortunately, due to a lack of time and/or resources, monitoring and data interpretation by the official agencies are often performed hastily and without deeper analysis. Together, the applied methods may contribute to overcoming the resistance to improve agricultural practices and implement other restoration measures. The differentiation of the possible N sources in groundwater in the present case shows that wastewater can provide an important pointsource contribution, though fertilizer-N is the dominant source. This information is very relevant for policy-makers, providing concrete answers as to where the problems lie and where and how they can be controlled, as well as for farmers, since it
demonstrates that the leaching of N from fertilizers in this case is the main contaminant source. Controlling the nutrient source from fertilizers and optimizing their use on a global scale are massive challenges, but undoubtedly among the most efficient measures to reduce their load on groundwater.
’ AUTHOR INFORMATION Corresponding Author
*Tel: +351 21 8417396; E-mail:
[email protected].
’ ACKNOWLEDGMENT We acknowledge the FCT Fundac-~ao para a Ci^encia e a Tecnologia (POCTI/AMB/57432/2004) for supporting the research. We further thank the three reviewers for their extremely useful comments that contributed to improving the manuscript. T.Y.S. holds a research position under the Ci^encia 2007 program financed by the FCT. ’ REFERENCES (1) Nolan, B. T. Relating nitrogen sources and aquifer susceptibility to nitrate in shallow groundwaters of the United States. Ground Water 2001, 39, 290–299. 8680
dx.doi.org/10.1021/es201798g |Environ. Sci. Technol. 2011, 45, 8674–8682
Environmental Science & Technology (2) B€ohlke, J. K. Groundwater recharge and agricultural contamination. Hydrogeol. J. 2002, 10, 153–179. (3) EEA. Source Apportionment of Nitrogen and Phosphorus Inputs into the Aquatic Environment; Report 7/2005; European Environmental Agency: Copenhagen, Denmark, 2005. (4) Slomp, C. P.; Van Cappellen, P. Nutrient inputs to the coastal ocean through submarine groundwater discharge: Controls and potential impact. J. Hydrol. 2004, 295, 64–86. (5) Howarth, R. W. Nutrient limitation of net primary production in marine ecosystems. Ann. Rev. Ecol. Syst. 1988, 19, 89–110. (6) Nixon, S. W. Coastal marine eutrophication: A definition, social causes, and future concerns. Ophelia 1995, 41, 199–219. (7) Nixon, S. W.; Ammerman, J. W.; Atkinson, L. P.; Berounsky, V. M.; Billen, G.; Boicourt, W. C.; Boynton, W. R.; Church, T. M.; Ditoro, D. M.; Elmgren, R.; Garber, J. H.; Giblin, A. E.; Jahnke, R. A.; Owens, N. J. P.; Pilson, M. E. Q.; Seitzinger, S. P. The fate of nitrogen and phosphorus at the land-sea margin of the North Atlantic Ocean. Biogeochemistry 1996, 35, 141–180. (8) Howarth, R. W.; Marino, R. Nitrogen as the limiting nutrient for eutrophication in coastal marine ecosystems: Evolving views over three decades. Limnol. Oceanogr. 2006, 51 (1, part 2), 364–376. (9) Valiela, I.; Costa, J.; Foreman, K. H.; Teal, J. M.; Howes, B; Aubrey, D. Transport of groundwater-borne nutrients from watersheds and their effects on coastal waters. Biogeochemistry 1990, 10, 177–197. (10) Boesch, D. F.; Brinsfield, R. B.; Magnien, R. E. Chesapeake Bay eutrophication: Scientific understanding, ecosystem restoration, and challenges for agriculture. J. Environ. Qual. 2001, 30, 303–320. (11) Gobler, C. J.; Sanudo-Wilhelmy, S. A. Temporal variability of groundwater seepage and brown tide blooms in a Long Island embayment. Mar. Ecol.-Prog. Ser. 2001, 217, 299–309. (12) Lee, Y. W.; Kim, G. Linking groundwater-borne nutrients and dinoflagellate red-tide outbreaks in the southern sea of Korea using a Ra tracer. Estuar. Coast. Shelf Sci. 2007, 71 (12), 309–317. (13) Costa, O. S., Jr.; Nimmo, M.; Attrill, M. J. Coastal nutrification in Brazil: A review of the role of nutrient excess on coral reef demise. J. S. Am. Earth Sci. 2008, 25, 257–270. (14) Howarth, R. W. Coastal nitrogen pollution: A review of sources and trends globally and regionally. Harmful Algae 2008, 8, 14–20. (15) Herrera-Silveira, J. A.; Morales-Ojeda, S. M. Evaluation of the health status of a coastal ecosystem in southeast Mexico: Assessment of water quality, phytoplankton and submerged aquatic vegetation. Mar. Pollut. Bull. 2009, 60 (13), 72–86. (16) Lee, Y. W.; Kim, G.; Lim, W. A.; Hwang, D. W. A relationship between submarine groundwater-borne nutrients traced by Ra isotopes and the intensity of dinoflagellate red-tides occurring in the southern sea of Korea. Limnol. Oceanogr. 2010, 55 (1), 1–10. (17) Appelo, C. A. J.; Postma, D. Geochemistry, Groundwater and Pollution, 2nd ed.; A.A.Balkema Publishers: Leiden, The Netherlands, 2005; 649 pp. (18) Stigter, T. Y.; Ribeiro, L.; Carvalho Dill, A. M. M. Application of a groundwater quality index as an assessment and communication tool in agro-environmental policies Two Portuguese case studies. J. Hydrol. 2006, 327, 578–591. (19) Hamilton, P. A.; Helsel, D. R. Effects of agriculture on groundwater quality in five regions of the United States. Ground Water 1995, 33, 217–226. (20) Howarth, R.; Ramakrishna, K. (coordinating lead authors). Nutrient Management. In Ecosystems and Human Well-Being: Policy Responses (Vol. 3 Findings of the Responses Working Group), Millennium Ecosystem Assessment; Island Press, 2005; Chapter 9, pp 295311. (21) Boesch, D. F. Challenges and opportunities for science in reducing nutrient over-enrichment of coastal ecosystems. Estuaries 2002, 25 (4b), 886–900. (22) EC. Council Directive 91/676/EEC of 12 December 1991 Concerning the Protection of Waters against Pollution Caused by Nitrates from Agricultural Sources; OJ L 375, 31.12.1991; pp 1-8.
ARTICLE
(23) Goodchild, R. G. EU Policies for the reduction of nitrogen in water: the example of the Nitrates Directive. Environ. Pollut. 1998, 102 (S1), 737–740. (24) CECCommission of the European Communities. Report from the Commission to the Council and the European Parliament on Implementation of Council Directive 91/676/EEC Concerning the Protection of Waters Against Pollution Caused by Nitrates from Agricultural Sources for the Period 20002003; COM(2007)120; Brussels, 2007; 11 pp. (25) CEC. Commission Staff Working Document on Implementation of Council Directive 91/676/EEC Concerning the Protection of Waters Against Pollution Caused by Nitrates from Agricultural Sources Based on Member State Reports for the period 20042007; SEC(2010)118; Brussels, 2010; 41 pp. (26) CEC. Report from the Commission to the Council and the European Parliament on Implementation of Council Directive 91/676/ EEC Concerning the Protection of Waters Against Pollution Caused by Nitrates from Agricultural Sources Based on Member State Reports for the period 20042007; COM(2010)47; Brussels, 2010; 11 pp. (27) NRCNational Research Council. Clean Coastal Waters: Understanding and Reducing the Effects of Nutrient Pollution; National Academy Press: Washington, DC, 2000. (28) Lin, B.; Jiakang, L. Jin, J. Outlook on Fertilizer Use in China in Next Century. In Proceedings of the Agro 99 International Symposium on Plant Protection and Plant Nutrition, Beijing, 1999. (29) Fixen, P. E.; West, F. B. Nitrogen fertilizers: Meeting contemporary challenges. Ambio 2002, 31, 169–176. (30) Li, X.; Hu, C.; Delgado, J. A.; Zhang, Y.; Ouyang, Z. Increased nitrogen use efficiencies as a key mitigation alternative to reduce nitrate leaching in north China plain. Agric. Water Manage. 2007, 89, 137–147. (31) Mathers, N. J.; Nash, D. M.; Gangaiya, P. Nitrogen and phosphorus exports from high rainfall zone cropping in Australia: Issues and opportunities for research. J. Environ. Qual. 2007, 36, 1551–1562. (32) Dibb, D. W. The mysteries (myths) of nutrient use efficiency. Better Crops 2000, 84, 3–5. (33) Zhu, Y. Fertilization and eutrophication of surface water. In Proceedings of Symposium on Fertilization and Environment; China Agricultural Scientech Press: Beijing, China, 1994; pp 4044. (34) Schlegel, A. J.; Havlin, J. L. Corn response to long-term nitrogen and phosphorus fertilization. J. Prod. Agric. 1995, 8, 181–185. (35) Johnson, J. W.; Murrell, T. S.; Reetz, H. F., Jr. Balanced fertility management: A key to nutrient use efficiency. Better Crops 1997, 81, 3–5. (36) Jin, S. Effect of water management and fertilization techniques on N utilization rate. In Abstracts of Symposium on N Cycling in Agriculture and the Environment; Xiamen, Fujian, 2001. (37) Wang, Y.; Wang, E.; Wang, D.; Huang, S.; Ma, Y.; Smith, C. J.; Wang, L. Crop productivity and nutrient use efficiency as affected by long-term fertilization in North China Plain. Nutr. Cycl. Agroecosys. 2010, 86, 105–119. (38) Giorgi, F. Climate change hot-spots. Geophys. Res. Lett. 2006, 33, L08707. (39) Stigter, T.; Ribeiro, L.; Samper, J.; Fakir, Y.; Pisani, B.; Li, Y.; Nunes, J. P.; Tome, S.; Oliveira, R.; Hugman, R.; Monteiro, J. P.; Silva, A. C. F.; Tavares, P. C. F.; Shapouri, M, Cancela da Fonseca, L.; El Mandour, A.; Yacoubi-Khebiza, M.; El Himer, H.. Assessing and Managing the Impact of Climate Change on Coastal Groundwater Resources and Dependent Ecosystems Final Report; CIRCLE-Med Project, Instituto Superior Tecnico: Lisbon; 187 pp. (40) Stigter, T. Y.; Ribeiro, L.; Carvalho Dill, A. M. M. Evaluation of an intrinsic and a specific vulnerability assessment method in comparison with groundwater salinisation and nitrate contamination levels in two agricultural regions in the south of Portugal. Hydrogeol. J. 2006, 14, 79–99. (41) Journel, A. G. Geostatistics for the Environmental Sciences, An Introduction; Project CR 811893; Applied Earth Sciences Department, Stanford University: Stanford, CA, 1987. (42) Clark, I. Practical Geostatistics; Applied Science Publishers Ltd: London, 1979. 8681
dx.doi.org/10.1021/es201798g |Environ. Sci. Technol. 2011, 45, 8674–8682
Environmental Science & Technology (43) Kreitler, C. W. Nitrogen-isotope ratio studies of soils and groundwater nitrate from alluvial fan aquifers in Texas. J. Hydrol. 1979, 42, 147–170. (44) Kaplan, N.; Magaritz, M. A nitrogen isotope study of the sources of nitrate contamination in ground water of the Pleistocene coastal plain aquifer, Israel. Water Resour. 1986, 20, 131–153. (45) Komor, S. C.; Anderson, H. W. Nitrogen isotopes as indicators of nitrate sources in Minnesota Sand-Plain aquifers. Ground Water 1993, 31 (2), 260–270. (46) Aravena, R.; Evans, M. L.; Cherry, J. A. Stable isotopes of oxygen and nitrogen in source identification of nitrate from septic systems. Ground Water 1993, 31, 180–186. (47) Wassenaar, L. I. Evaluation of the origin and fate of nitrate in the Abbotsford Aquifer using the isotopes of 15N and 18O in NO3. Appl. Geochem. 1995, 10, 391–405. (48) Liu, C.-Q.; Li, S.-L.; Lang, Y.-C.; Xiao, H.-Y. Using δ15N-and δ18O-values to identify nitrate sources in karst ground water, Guiyang, Southwest China. Environ. Sci. Technol. 2006, 40 (22), 6928–6933. (49) Kendall, C. Tracing Nitrogen Sources and Cycling in Catchments. In Isotope Tracers in Catchment Hydrology; Kendall, C., McDonnell, J. J., Eds.; Elsevier Science B.V.: Amsterdam, 1998; pp 519576. (50) Heaton, T. H. E. Isotopic studies of nitrogen pollution in the hydrosphere and atmosphere: A review. Chem. Geol. (Isot. Geosci. Sect.) 1986, 60, 87–102. (51) Kroopnick, P. M.; Craig, H. Atmospheric oxygen: isotopic composition and solubility fractionation. Science 1972, 175, 54–55. (52) Amberger, A.; Schmidt, H. L. Nat€urliche Isotopengehalte von Nitrat als Indikatoren f€ur dessen Herkunft. Geochim. Cosmochim. Acta 1987, 51, 2699–2705. (53) Stigter, T. Y.; Van Ooijen, S. P. J.; Post, V. E. A.; Appelo, C. A. J.; Carvalho Dill, A. M. M. A hydrogeological and hydrochemical explanation of the groundwater composition under irrigated land in a Mediterranean environment, Algarve, Portugal. J. Hydrol. 1998, 208, 262–279. (54) Stigter, T. Y.; Ribeiro, L.; Carvalho Dill, A. M. M. Building factorial regression models to explain and predict nitrate concentrations in groundwater under agricultural land. J. Hydrol. 2008, 358, 42–56. (55) Stigter, T.; Carvalho Dill, A.; Malta, E.-j.; Santos, R. Nutrient sources for green macroalgae in the Ria Formosa lagoon assessing the role of groundwater. In: Ribeiro, L.; Stigter, T. Y.; Chambel, A.; Condesso de Melo, M. T.; Monteiro, J. P.; Medeiros, A., Eds.; Selected Papers on Hydrogeology - Groundwater and Ecosystems; Taylor & Francis, Abingdon, Oxford, UK, Accepted. (56) Quelhas dos Santos, J. Fertilizac-~ao - Fundamentos da utilizac-~ao dos adubos e correctivos: Francisco Lyon de Castro, Publ. EuropaAmerica, Lda.: Mem Martins, Portugal, 1991; 441 pp, In Portuguese. (57) Stigter, T. Y.; Carvalho Dill, A. M. M. Estudo geologico e hidrogeoquímico das regi~oes abrangidas pelo projecto; Report of Project Interreg II Portugal-Spain “The effect of the intensive use of fertilisers and pesticides on the quality of soil and groundwater”; Universidade do Algarve: Faro, Portugal, 2001,; 67 pp, In Portuguese. (58) Sigman, D. M.; Casciotti, K. L.; Andreani, M.; Barford, C.; Galanter, M.; B€ohlke, J. K. A bacterial method for the nitrogen isotopic analysis of nitrate in seawater and freshwater. Anal. Chem. 2001, 73, 4145–4153. (59) Casciotti, K. L.; Sigman, D. M.; Hastings, M. G.; B€ ohlke, J. K.; Hilkert, A. Measurement of the oxygen isotopic composition of nitrate in seawater and freshwater using the denitrifier method. Anal. Chem. 2002, 74, 4905–4912. (60) Di, H. J.; Cameron, K. C. Nitrate leaching in temperate agroecosystems: sources, factors and mitigating strategies. Nutr. Cycl. Agroecosyst. 2002, 64, 237–256. (61) Andrade, A. I. A. S. S.; Stigter, T. Y. Multi-method assessment of nitrate and pesticide contamination in shallow alluvial groundwater as a function of hydrogeological setting and land use. Agric. Water Manage. 2009, 96, 1751–1765. (62) Hansen, B.; Thorling, L.; Dalgaard, T; Erlandsen, M. Trend reversal of nitrate in Danish groundwater a reflection of agricultural
ARTICLE
practices and nitrogen surpluses since 1950. Environ. Sci. Technol. 2011, 45 (1), 228–234. (63) Mengis, M.; Walther, U.; Bernasconi, S. M.; Wehrli, B. Limitations of using δ18O for the source identification of nitrate in agricultural soils. Environ. Sci. Technol. 2001, 35 (9), 1840–1844. (64) Scanlon, B. R.; Reedy, R. C.; Bronson, K. F. Impacts of land use change on nitrogen cycling archived in semiarid unsaturated zone nitrate profiles, Southern High Plains, Texas. Environ. Sci. Technol. 2008, 42 (20), 7566–7572. (65) Metcalf and Eddy. Wastewater Engineering. Treatment, Disposal, Reuse, 3rd ed.; McGraw-Hill, Inc.: Singapore, 1991.
8682
dx.doi.org/10.1021/es201798g |Environ. Sci. Technol. 2011, 45, 8674–8682
ARTICLE pubs.acs.org/est
Assessing the Severity of Rainfall-Derived Infiltration and Inflow and Sewer Deterioration Based on the Flux Stability of Sewage Markers Jessica M. Shelton,† Lavane Kim,† Jiasong Fang,§ Chittaranjan Ray,†,‡ and Tao Yan*,† †
Department of Civil and Environmental Engineering, ‡Water Resources Research Center, University of Hawaii at Manoa, Honolulu, Hawaii, 96822 United States § Department of Natural Sciences, Hawaii Pacific University, Honolulu, Hawaii, 96813 United States ABSTRACT: This study investigated the flux stability of select chemical and biological sewage markers, including caffeine, total nitrogen (TN), total suspended solids (TSS), E. coli, and enterococci, and their suitability in assessing the severity of rainfall-derived infiltration and inflow (RDII) in a residential sewershed. To quantify and compare marker flux stability, concentrations of the candidate markers in two dry-weather periods were determined and the one-day lag autocorrelation coefficients (r) of their mass fluxes were calculated. TN (r = 0.820.88) exhibited higher and more consistent flux stability than TSS (r = 0.490.82), caffeine (r = 0.560.58), E. coli (r = 0.360.87), and enterococci (by culture; r = 0.400.52), all of which except enterococci by qPCR (r = 0.100.21) showed significant autocorrelation. Sewage flows and marker concentrations were also monitored in two wet-weather periods, and the severity of RDII (RRDII) were calculated using either flow measurements or marker concentrations independently. Corresponding to its outstanding flux stability, RRDII values estimated by TN predicted all severe RDII instances and gave the highest and most consistent correlation (r = 0.740.78) among the different sewage markers. Overall, the study illustrated the feasibility of using the flux stability of sewage markers in assessing the severity of RDII and thereby deterioration levels in sewer systems.
’ INTRODUCTION Well-maintained wastewater collection systems are important for proper conveyance of sanitary sewage to treatment facilities prior to environmental discharge. However, the ca. 600 000 miles of sewer pipes in the United States are eroding with age, use, and neglect.13 It is estimated that approximately 75% of sewer systems are functioning at less than half of their design capacities,4,5 and about 23 00075 000 sanitary sewer overflows (SSOs) occur every year in the United States as a result of deteriorated sewer systems.6,7 To adequately protect the environment and public health, significant sewer rehabilitation and replacement works are urgently needed, which would cost approximately $270 billion through the year 2019 according to a 2002 U.S. EPA estimate.3 Efficient implementation of such massive rehabilitation and replacement efforts would require a thorough assessment of sewer conditions to rank the different sewer sections based on the level of deterioration.4 Prioritization of high-risk areas would allow utilities to focus their efforts and proactively treat problems, avoiding the higher costs associated with emergency repairs. Ranking the sewers in terms of deterioration severity and potential for abrupt failure also allows for better asset management.8 However, most current inspection methods are not suitable for large-scale sewer condition assessment at the sewershed or sewer system level. Sewer inspection is predominantly conducted by physical methods, with the most commonly used being CCTV examination.9 These methods are often too detail-focused and r 2011 American Chemical Society
labor-intensive to achieve high-throughput system-wide condition assessment.8,10 One potential approach for system-wide sewer deterioration severity assessment is to use rainfall-derived infiltration and inflow (RDII) as an indicator of sewer structure deterioration.8 RDII is a symptom of sewer deterioration and is also a contributing cause of overloading the system, SSOs, and reduced efficiency of treatment facilities.4,11 Because RDII increases with increased severity of sewer deterioration, given the same precipitation conditions, the levels of RDII are directly related to sewer deterioration levels in different sewer sections.7 However, determining RDII by direct flow measurement is limited by the technical difficulties of obtaining accurate flow measurements12 and the high costs associated with simultaneous measurement of sewage flow at numerous locations in a sewer system, which is required for system-wide assessment efforts. Alternatively, RDII may be quantified through chemical and bacteriological analyses, which are usually less costly to perform and can be easily automated for higher throughputs than by direct flow measurements. Among the countless chemicals and organisms present in sanitary sewage, some may exhibit stable diurnal fluxes (i.e., having relatively constant total mass over a Received: June 7, 2011 Accepted: September 8, 2011 Revised: September 8, 2011 Published: September 08, 2011 8683
dx.doi.org/10.1021/es2019115 | Environ. Sci. Technol. 2011, 45, 8683–8690
Environmental Science & Technology given time period in any day) due to their specific sources and inputs from households in a residential community. Such markers may include fecal bacteria such as Escherichia coli and enterococci,13,14 caffeine,15,16 total nitrogen,17 and total suspended solids (TSS).17 For the sewage markers with stable fluxes, sewage flow dilution caused by RDII would proportionally decrease their concentrations; conversely, the concentration changes can be quantified to determine the RDII level. Because of the high-throughput analysis techniques available for these sewage markers, numerous sewer sections in a sewer system can be sampled and analyzed simultaneously under the same weather conditions, and the resulting RDII levels can be used to infer sewer deterioration severity. The overall goal of this study was to develop a marker-based approach to determine RDII levels for assessing sewer deterioration severity at the system-wide level. It was hypothesized that certain sewage markers exhibit stable diurnal fluxes and their concentration changes under wet-weather conditions are a function of the severity of RDII and sewer deterioration. Three specific tasks were carried out in this study to achieve the goal. First, concentration dynamics of select candidate sewage markers, including TSS, total nitrogen (TN), caffeine, E. coli, and enterococci, were investigated under dry-weather conditions in the Manoa sewershed (a residential sewershed on the Island of Oahu, Hawaii). Second, the fluxes of the candidate markers were determined for the eight 3-h time periods of a day under dry-weather conditions, and the markers with stable diurnal fluxes were identified. Finally, sewage flow and sewage marker concentrations under wet-weather conditions were monitored. The severity levels of RDII were estimated by sewage marker concentrations, which were then compared with those determined by direct flow measurement.
’ EXPERIMENTAL SECTION Sewershed and Sampling Site. Manoa sewershed serves the predominantly residential Manoa valley (Figure 1) which includes the University of Hawaii. Sanitary sewage from Manoa valley flows through two parallel sewer pipes through the University of Hawaii-Manoa campus. The sampling location (sewer manhole 368092) is located on the campus grounds but is just upstream of wastewater discharges from the university facilities (Figure 1), which avoids fluctuations and atypical inputs caused by university generated activities. The sewer pipe to be sampled is 15 in. in diameter and accessible through a 6.6 ft deep manhole. Data for the sewershed geography and the sewer pipe network were obtained from Hawaii Statewide Geographic Information System (http://hawaii.gov/dbedt/gis/ download.htm) and processed using ArcGIS 9.3 software (ESRI; Redlands, CA). Flow Monitoring and Sewage Sample Collection. Flow measurements at the sampling site were conducted using an ISCO 2150 area velocity flow meter (Teledyne, Lincoln, NE), and data acquisition and analysis were accomplished using the FlowLink 5 software (Teledyne). Sewage samples were collected using a Sigma 900 automatic field sampler (Hach Company, Loveland, CO). The area velocity sensor of the flow meter and inlet tube of the autosampler were submerged into sewage flow and mounted to the bottom of the sewer pipe using an ISCO street level manhole installation apparatus (Teledyne). Flow monitoring and sewage sample collection were conducted in two dry-weather (DW) periods and two wet-weather
ARTICLE
Figure 1. Sewer pipe network in the Manoa sewershed and the location of Manoa watershed on the island of Oahu, Hawaii. The sampling manhole location is indicated by a dark star.
(WW) periods. The two DW periods were Aug. 25Sept. 1, 2010 and Feb. 1017, 2011, which flank the two WW periods which were Sept. 30Oct. 7 and Nov. 310, 2010. Flow measurements were taken every 15 min, and the flows were averaged over 3-h periods to coincide with the 3-h sewage sampling interval. Sewage samples (150 mL) were taken every 15 min and every 12 samples were combined into a 3-h composite sample. For each sampling period, the flow measurements and sewage sampling continued for seven days, resulting in eight samples per day for a total of 56. The sewage samples within the autosampler were kept at approximately 4 °C using ice packs, and transported back to the laboratory every day for immediate processing. Sample Analysis. The sewage samples were analyzed for various parameters, including pH, salinity, TSS, TN, caffeine, E. coli, and enterococci. pH was measured using a Denver Instruments pH meter (Denver Instruments, Bohemia, NY), and salinity was measured using an Orion 150A Plus conductivity meter (Thermo Scientific, Beverley, MA). TSS was measured by filtering sewage samples (50 mL) through 1.1-μm glass fiber filters (VWR Grade 693; West Chester, PA) and determining dry weight difference prior to and after filtration according to Standard Methods 2540D.18 TN, Caffeine measurement, and bacterial enumeration (either by cultivation or by real-time PCR (qPCR)) are described in detail below. TN Quantification. To measure TN in sewage samples, 30 mL of each sample was first filtered through a 0.45-μm cellulose-ester membrane filter, and the filtrates were acidified to pH 2 or less using 1 N HCl before being analyzed based on an oxidative combustion-chemiluminescence method using a Shimadzu TOC/TN analyzer (Shimadzu, Columbia, MD). TN in the sample was converted into nitrogen monoxide by combustion 8684
dx.doi.org/10.1021/es2019115 |Environ. Sci. Technol. 2011, 45, 8683–8690
Environmental Science & Technology
ARTICLE
and then to nitrogen dioxide by reacting with the oxidizer ozone, and changes in chemiluminescence were detected by the instrument, which were compared against external calibration standards to determine TN concentrations. Caffeine Quantification. A SPE-GCMS procedure developed by the U.S. Geological Survey 19 was modified to quantify caffeine in sewage samples. One liter of each sewage sample was first prefiltered through 0.7-μm glass-fiber filters to remove suspended solids. The filtrates were then amended with 60 g of NaCl to increase the ionic strength and stabilize the solution, and 100 μL of 13C3-caffeine at 20 ng/μL in methanol was added to each 1-L sample as the recovery surrogate. Samples were acidified to pH 4.3 using sodium acetate buffer (30 g/L of acetic acid and 15 g/L of sodium acetate). Solid phase extraction (SPE) of the acidified samples was performed using HLB Oasis cartridges (Waters, Milford, MA) with a sample flow-through rate of ca. 15 mL per min. The sample bottles were rinsed with potassium phosphate buffer (30 g/L of K2HPO4 and 20 g/L of KH2PO4, pH 7), which was also passed through the SPE cartridge to neutralize in preparation for elution. The cartridges were airdried, and then eluted with six passes of 3 mL of the solvent mixture (hexanes:dichloromethane; 50:50). The eluents were evaporated using a N-Evap nitrogen evaporator (Thompson, Clear Brook, VA) to approximately 0.4 mL. After amendment with 20 μL of 100 ng/μL of Phenanthrene-d10 in dichloromethane as an internal quantification standard, the samples were evaporated to the final volume of 0.4 mL for subsequent GC-MS quantification. All samples were analyzed on an Agilent 7890A gas chromatograph (GC) interfaced with an Agilent 5975C Mass Selective Detector and an Agilent 7683B autoinjector. Analytical separation of the analytes was accomplished using a 30 m 0.25 mm (i.d.) DB-5 fused-silica capillary column (J&W Scientific, Folsom, CA). Column temperature was programmed from 50 (held for 1 min) to 190 at 40 °C min1, then to 220 at 3 °C min1, and finally to 300 at 30 °C min1. Selective ion monitoring (SIM) for target analytes (Table 1) was conducted for quantification, and a full scan run was performed to confirm the SIM analysis results. Compounds were identified based on mass spectra. Recovery efficiencies for the analytes were calculated based on the quantity of the surrogate standard recovered. Concentration of each compound was calculated based on the GC-MS response relative to that of the internal standard after correction for recovery efficiency, which averaged 79%. Any samples lower than these amounts were discarded and not used during the analysis. In all cases, samples were assumed to be free from contamination if samples that contained only solvents and reagents (i.e., blanks) produced chromatograms free of peaks. Bacterial Enumeration. E. coli and enterococci in sewage samples were enumerated following the standard cultivationbased modified mTEC agar method20 and mEI agar method,21 respectively. Briefly, 10-fold serial dilutions of each sewage sample were made, and 20 mL of appropriate dilutions (usually the 106 and 105 dilutions for E. coli and the 104 and 103 dilutions for Table 1. GC/MS-SIM Acquisition Parameters for Caffeine Analysis target compound
quantification ion
confirmatory ion
caffeine
194
109, 82
13
197
110
phenanthrene-d10
188
C3-caffeine
enterococci) were filtered through sterile 0.45-μm cellulose-ester membrane filters using vacuum filtration and placed on the modified mTEC agar plates (for E. coli) and the mEI agar plates (for enterococci). The modified mTEC agar plates were incubated at 37 °C for 2 h before being transferred to 44 °C for overnight incubation, and the mEI plates were incubated at 42 °C for 24 h before colony forming units (CFU) were counted. qPCR Quantification. Enterococci in sewage samples were also quantified using a qPCR method.22 Briefly, sewage samples were filtered through 0.45-μm cellulose ester membrane filters and subjected to DNA extraction using the UltraClean soil DNA isolation kit (Mo-Bio; Carlsbad, CA). The qPCR reactions include 1X TaqMan universal PCR master mix (ABI; Carlsbad, CA), 8 μM primers ECST784F and ENC854R, 4 μM TaqMan probe GPL813TQ, 0.2% BSA (by weight), and 1 μL of DNA template. The thermal cycler conditions included an initial denaturing at 95 °C for 10 min and subsequent 50 cycles of 95 °C for 15 s and 60 °C for 1 min. Absolute quantification of qPCR was achieved with five-point external calibration standard curves constructed using stationary-phase cells of Enterococcus faecalis ATCC 29212. Calibration curves were established for each batch of qPCR reactions, with r2 ranging from 0.962 to 0.998, and the concentrations were represented as calibrator cell equivalent per mL (CCE/mL). Data Analysis. The mass flux of sewage markers (mass/time) was calculated by multiplying concentrations of the sewage markers in the composite sewage sample by the average sewage flow within the sampling period. Bacterial measurements (in either CFU or CCE) were log transformed in the flux calculation based on the assumption of log-normal distribution. Linear correlations between time-series data were tested using Pearson’s coefficients (r) using Microsoft Excel with a statistical add-in (Statistixl, Australia). Autocorrelation coefficients of timeseries data were calculated based on a 1-day lag to indicate the daily reproducibility of the signal patterns of marker fluxes. Average marker fluxes within each of the 3-h time intervals and their standard deviations across different sampling days were calculated to indicate flux stability. Average marker flux differences and their 95% confidence intervals between the two dry-weather sampling periods, which flanks the wet-weather sampling periods, were calculated and plotted to indicate secular variations of marker fluxes. The default significance level in statistical tests was P e 0.05 unless otherwise stated. Precipitation data from the weather station at Lyon Arboretum operated by the National Ocean and Atmospheric Administration (NOAA) and streamflow data from the U.S. Geological Survey Kanewai station (USGS 16242500) were used in verifying the weather conditions during sampling periods. Stable Flux Model. A sewage marker with stable flux should have a flux pattern reproducible each day and a limited secular variation within the study period. For such markers, fluxes under dry weather conditions should be equal to fluxes under wetweather conditions (eq 1), where Cdry is the average marker concentration under dry-weather conditions, Qdry is the average sewage flow under dry-weather conditions, Cwet is the wetweather marker concentration under the influence of RDII, and Qwet is the wet-weather sewage flow with RDII contribution. The wet-weather sewage flow (Qwet) can be expressed as the summation of a dry-weather average flow component (Qdry) and the RDII contribution (QRDII) (eq 2), and the ratio between QRDII and Qdry is defined as RRDII (eq 3), which is an indicator of the severity of RDII and sewer deterioration. Independently, 8685
dx.doi.org/10.1021/es2019115 |Environ. Sci. Technol. 2011, 45, 8683–8690
Environmental Science & Technology
ARTICLE
RRDII can also be expressed as a function of sewage marker concentrations (eq 3). Cdry Qdry ¼ Cwet Qwet
ð1Þ
Qwet ¼ Qdry þ QRDII
ð2Þ
RRDII ¼
Cdry Cwet QRDII ¼ Qdry Cwet
ð3Þ
’ RESULTS AND DISCUSSION Concentration Dynamics of Sewage Markers. Sewage flow and the five candidate sewage markers, including caffeine, TN, TSS, E. coli, and enterococci, were monitored in two DW periods that were devoid of significant precipitation in the Manoa sewershed (Figure 2). As expected, sewage flow in the DW periods exhibited a typical diurnal fluctuation with two peaks daily (Figure 2A), and a high level of similarity in sewage flow patterns between the two DW periods was detected (Pearson’s r = 0.80). The autocorrelation coefficients for sewage flow in the two DW periods were 0.71 and 0.83, indicating that the sewage flow under dry-weather conditions exhibited a strong daily pattern. The DW periods 1 and 2 had flow ranges of 16.552.3 L/s and 14.738.4 L/s, respectively. Concentration dynamics of the candidate sewage markers exhibited daily fluctuation patterns with different levels of reproducibility between the two DW periods (Figure 2BG). Among the physical and chemical candidate markers, TN concentration in sewage samples appeared to be the most stable, with a single major peak present in the beginning of each day (Figure 2C). The concentrations of caffeine (Figure 2B) and TSS (Figure 2D) also showed reproducible patterns, but with less regularity than that of TN. The observed concentration ranges for caffeine, TN, and TSS within the two DW periods were 5.0103.4 μg/L, 9.234.2 mg/L, and 18456 mg/L, respectively. Concentrations of the bacterial marker E. coli and enterococci in sewage samples were determined by cultivation-based membrane filtration methods (Figure 2EF). A DNA-based qPCR method, which affords higher throughput, was also evaluated for quantifying the enterococci marker in sewage samples (Figure 2G). Among the bacterial markers, enterococci by cultivation (Figure 2F) exhibited better diurnal repeatability than enterococci by qPCR (Figure 2G) and E. coli by culture (Figure 2E). The observed concentration ranges for enterococci (by cultivation), enterococci (by qPCR), and E. coli (by culture) were 4.96.3 log CFU/100 mL, 3.47.3 log CCE/100 mL, and 6.38.9 log CFU/100 mL, respectively. Overall, the bacterial markers exhibited larger fluctuation in concentration than the physicochemical sewage markers. Flux Stability of Sewage Markers. Mass fluxes of the candidate sewage markers in the two DW sampling periods were calculated to identify those markers with a stable flux. Autocorrelation coefficients of marker fluxes with a one-day lag indicate the reproducibility of flux patterns across different sampling days (Table 2). Among the candidate markers, TN flux exhibited the highest and most consistent regularity in both DW periods (autocorrelation r = 0.820.88). Significant autocorrelations were also observed for the fluxes of TSS, caffeine,
Figure 2. Dry-weather sewage flow (A) and concentration dynamics of caffeine (B), TN (C), TSS (D), E. coli by culture (E), enterococci by culture (F), and enterococci by qPCR (G) in Manoa sewershed. The two sampling weeks were Aug. 26Sept. 2, 2010 (b) and Feb. 1017, 2011 (O). The bright and shaded areas represent daytimes (8 a.m.8 p.m.) and nighttimes (8 p.m.8 a.m.).
Table 2. Autocorrelation Coefficients of Sewage Flow and Marker Fluxes in the Dry-Weather Sampling Periods correlation coefficients r (P value) parameter (units)
8686
DW period 1
DW period 2
flow (L/s)
0.71 (<0.001)
0.83 (<0.001)
TSS flux (g/s)
0.49 (<0.001)
0.82 (<0.001)
caffeine flux (mg/s) TN flux (g/s)
0.58 (<0.001) 0.83 (<0.001)
0.56 (<0.001) 0.88 (<0.001)
enterococci flux (CFU/s)
0.40 (0.005)
enterococci flux (CCE/s)
0.21 (0.15)
E. coli flux (CFU/s)
0.36 (0.01)
0.52 (<0.001) 0.10 (0.51) 0.87 (<0.001)
dx.doi.org/10.1021/es2019115 |Environ. Sci. Technol. 2011, 45, 8683–8690
Environmental Science & Technology
ARTICLE
Table 3. Mean Values and Standard Deviations (In Parentheses) of Sewage Marker Fluxes within the Dry-Weather Sampling Periods caffeine (mg/s)
TN (g/s)
TSS (g/s)
enterococci (log CFU/s)
enterococci (log CEE/s)
E. coli (log CFU/s)
P1 (5 a.m.8 a.m.)
2.16 (0.95)
1.00 (0.12)
8.95 (3.69)
8.46 (0.21)
7.99 (0.44)
8.87 (1.16)
P2 (8 a.m.11 a.m.) P3 (11 a.m.2 p.m.)
2.4 (0.73) 1.80 (0.56)
0.48 (0.12) 0.38 (0.07)
6.04 (2.11) 4.46 (1.91)
8.39 (0.16) 8.02 (0.25)
8.08 (0.61) 7.97 (0.61)
8.90 (1.25) 8.81 (1.22)
P4 (2 p.m.5 p.m.)
1.44 (0.55)
0.44 (0.08)
4.72 (1.28)
8.06 (0.24)
8.02 (0.44)
8.85 (1.28)
P5 (5 p.m.8 p.m.)
1.78 (0.67)
0.45 (0.06)
5.43 (1.38)
8.00 (0.28)
7.84 (0.42)
8.75 (1.03)
P6 (8 p.m.11 p.m.)
1.24 (0.67)
0.48 (0.07)
3.75 (2.41)
7.96 (0.22)
7.72 (0.44)
8.59 (0.94)
P7 (11 p.m.2 a.m.)
0.34 (0.25)
0.28 (0.06)
1.36 (0.53)
7.67 (0.25)
7.52 (0.46)
8.26 (0.81)
P8 (2 a.m.5 a.m.)
0.49 (0.27)
0.39 (0.10)
2.16 (0.85)
8.02 (0.25)
7.38 (2.17)
8.41 (0.99)
time intervals
enterococci (by cultivation), and E. coli, but not for enterococci (by qPCR) (Table 2). In addition to the reproducibility of flux patterns, the means and standard deviations of marker flux values among replicate sampling days within the dry-weather sampling periods were also calculated (Table 3). Among the three physicochemical markers, TN exhibited the smallest standard deviations from mean values in all eight daily sampling intervals, indicating small variances among replicate sampling days. Caffeine and TSS flux values showed higher variances than TN, which is in agreement with their less regular flux patterns (Table 2). The bacterial markers all had small standard deviations from the mean values. However, the mean values did not differ significantly from each other among the eight time intervals, corroborating with the lack of regular flux patterns among the bacterial markers (Table 2). Secular variances of marker fluxes between the two dry-weather sampling periods, which flank the whole sampling efforts, were also determined by calculating the means flux differences and 95% confidence intervals (Figure 3). Only E. coli had significant secular variances, as indicated by the clear separation of flux mean difference and zero (Figure 3F); the other markers did not show consistent secular variances among the eight daily sampling time intervals (Figure 3AE). The different levels of flux stability exhibited by sewage markers are attributable to their different sources, which are governed by the daily routines of the residents of a residential sewershed. The primary sources of caffeine in a sewershed are associated with the preparation and consumption of coffee by the residents, who in general follow regular daily routines but may also deviate occasionally. The remarkable flux stability of TN can be explained by the fact a major component of TN in sewage is human urea, the production and secretion of which is controlled by the metabolic activities of human bodies and follow regular daily patterns. Therefore, the TN flux in sanitary sewage can be expected to be very stable and may even be independent of seasonal changes. The bacterial markers selected in this study are the E. coli and enterococci, which inhabit human lower intestines and are commonly used as fecal indicators. The bacterial fluxes are much less stable than the physicochemical markers (such as TN). This may be due to large variations in fecal loading. In addition, E. coli and enterococci may die off under different environmental stresses present in the sewage systems at different rates,23,24 and they may also actively replicate given appropriate environmental conditions.25,26 It is worth noting that the flux of enterococci based on DNA-based qPCR quantification was the only unstable marker without significant autocorrelation. The accuracy of
Figure 3. Mean flux differences and the 95% confidence intervals (error bars) between the two dry-weather sampling periods: TSS (A), caffeine (B), TN (C), enterococci by cultivation (D), enterococci by qPCR(E), and E. coli (F).
qPCR may be affected by varying DNA recoveries during extraction,27 varying 23S rRNA gene copies of enterococci species,28,29 and the presence of PCR inhibitory compounds.30,31 Nevertheless, the qPCR approach, as an emerging technology that affords the highest throughput (e.g., 384 samples can be processed simultaneously in several hours), still present a potentially valuable tool for sewer condition assessment after the said pitfalls are overcome. 8687
dx.doi.org/10.1021/es2019115 |Environ. Sci. Technol. 2011, 45, 8683–8690
Environmental Science & Technology
ARTICLE
Table 4. Comparison between RRDII Determined by Flow Measurements and by Sewage Marker Concentrations Pearson’s coefficients r (P value) sewage markers
WW period 1
WW period 2
TSS
0.28 (0.04)
0.64 (<0.001)
caffeine
0.27 (0.06)
0.69 (<0.001)
TN enterococci
0.78 (<0.001) 0.09 (0.51)
0.74 (<0.001) 0.24 (0.08)
Rainfall-Derived Infiltration and Inflow. The candidate markers with stable fluxes, including caffeine, TN, TSS, and enterococci were used as sewage markers in wet weather monitoring. Sewage flows and marker concentrations in the Manoa sewershed were monitored in two WW sampling periods when significant rainfall occurred. The occurrence of RDII was shown by the significantly reduced autocorrelation coefficients in sewage flow (r = 0.36 and 0.41) due to the RDII contribution. The severity level of RDII (RRDII) over time in the Manoa sewershed was calculated based either on flow measurements or on sewage marker concentrations (eq 3). To determine the effectiveness of using sewage markers in estimating the severity level of RDII, the RRDII values estimated by the different sewage markers were compared with those estimated by direct flow measurements using correlation analysis (Table 4). The RRDII values calculated from TN concentrations exhibited the best correlation with the RRDII values obtained from direct flow measurement (r = 0.740.82), which corroborates with the fact that TN exhibited the most stable flux among all sewage markers. Figure 4 further illustrates the correspondence between the RRDII values calculated by flow measurements and the RRDII values derived from TN concentration changes over time. In particular, all the high severity instances of RDII during periods of active precipitation, which were indicated by the elevated flows in the Manoa Stream, were successfully detected by the changes in TN concentration. In predicting RRDII, TSS and caffeine are less consistent than TN (Table 3). Although both TSS and caffeine predicted RRDII in significant and strong correlation with those determined by direct flow measurement in WW period 2 (r = 0.640.69), their performance in WW period 1 was much poorer (r = 0.270.28). The RRDII values predicted by enterococci correlated poorly with the values obtained by flow measurement in both WW periods. The varying performances of different sewage markers in predicting RRDII were related to the different levels of stability in marker fluxes. Additionally, eqs 13 assume no external sources of sewage markers under wet-weather conditions, which may bear different reliabilities for different markers. For example, the inflow portion of the RDII flow contribution may bring additional TSS into the sewage flow and enterococci in Hawaii soil32,33 may also enter the sewer system. Marker-Based Sewer Condition Assessment. The present study represents the first attempt to examine the flux stability of sewage markers in a residential sewershed. Sewage marker fluxes are governed by the daily routines of the residents, which collectively follow fairly regular daily patterns. The present study observed flux stability, which entails regular daily patterns and limited secular variances, for TN, TSS, caffeine, and enterococci in a tropical sewershed. Similar flux stability can also be expected of residential sewersheds in temperate regions, although limited
Figure 4. Severity levels of I/I as determined by flow measurements (b) and concentration of TN in sewage samples (O) and flow in the Manoa Stream (bar) in the two wet weather sampling weeks: Sept. 30Oct. 7, 2011 (A) and Nov. 310, 2011 (B).
secular variations may only be available in shorter time periods. Although industrial, commercial, and business sewage contributions can significantly affected flux stability, these “hotspots” of irregularity can be easily identified and excluded in study design, and thus should not affect the applicability of the proposed approach. Because it is practically impossible to identify the locations and timings of marker inputs, the study sewershed was treated as a blackbox, and marker fluxes were defined at the sewershed outlet (i.e., the sampling location). This treatment effectively neglected the impact of different marker residence times in the sewershed and can potentially reduce the predictive power of the sewage markers. The impact of this necessary simplification, however, can be mediated by defining small study sewer sections and relatively long sampling time intervals. For example, the present study used a 3-h sampling time interval for determining marker fluxes, which is significantly longer than the maximum collection time of the study sewershed, thus rendering the different marker residence times insignificant. The utility of sewage marker flux stability was demonstrated in estimating the severity of RDII under wet-weather conditions. Among the candidate sewage markers examined, TN exhibited the most stable flux and its concentration changes during wetweather conditions were strongly related to RDII severity levels. Based on the average flux values established under dry-weather conditions, the TN concentration changes accurately predicted the timing of severe RDII periods in the WW sampling periods (Figure 4). It should be noted that although the RRDII values, either calculated by flow measurement or based on TN concentration, changed over time in similar patterns, their absolute values did not coincide with each other all the time. Although sewage flow measurements can be inaccurate,12 predictions based on marker 8688
dx.doi.org/10.1021/es2019115 |Environ. Sci. Technol. 2011, 45, 8683–8690
Environmental Science & Technology concentrations are certainly affected by marker flux fluctuations (Table 3). Even for the most stable TN, predictions based on marker concentrations can still significantly deviate from the estimations based solely on flow measurements (Figure 4). For example, in the later part of the WW period 2, where a prolonged precipitation occurred, RRDII values calculated from TN concentrations significantly deviated from the values calculated by direct flow measurement, which implicates the relevance of local hydrometerological conditions and history in the marker-based RDII prediction. Theoretically, the severity level of RDII (RRDII) is a function of sewer deterioration severity and local precipitation conditions. The present study, as an important initial step, examined RRDII in the same sewershed at different times, which was validated by the correlation between RRDII and precipitation conditions (Figure 4). To develop the sewage marker concept for the assessment of sewer deterioration, subsequent studies shall focus on assessing RDII severity levels at different sewer sections under the same precipitation conditions at the same time. This would eventually normalize the effect of local hydrometerological conditions on RDII severity levels, and significantly improve the quantitative prediction power of the sewage marker-based RDII severity assessment approach.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 808-956-6024; fax: 808-956-5014; e-mail: taoyan@ hawaii.edu; mail: University of Hawaii at Manoa, Department of Civil and Environmental Engineering, 2540 Dole Street, Holmes Hall 383, Honolulu, HI 96822.
’ ACKNOWLEDGMENT We thank Ms. Bunnie Yoneyama and Mr. Joseph Lichwa for their technical support in the laboratory, and UH Manoa campus facility management for supporting the sampling efforts. This material is based upon work supported by the National Science Foundation under Grant 0964260. ’ REFERENCES (1) Wirahadikusumah, R.; Abrahama, D. M.; Iseley, T.; Prasanth, R. K. Assessment technologies for sewer system rehabilitation. Autom. Constr. 1998, 7 (4), 259–270. (2) ASCE. 2009 Report Card for America’s Infrastructure; American Society of Civil Engineers: Reston, VA, 2009. (3) USEPA. Clean Water and Drinking Water Infrastructure Gap Analysis; USEPA-816-R-02-020; Washington, DC, 2002. (4) USEPA. Innovation and Research for Wastewater Infrastructure for the 21st Century, Research Plan; EPA-ORD-NRMRL-CI-08-03-02; Washington, DC, 2007. (5) ASCE. State and Local Public Works Needs; Civil Engineering Research Foundation: Reston, VA, 1994. (6) USEPA. Sanitary Sewer Overflows and Peak Flows; http:// cfpub.epa.gov/npdes/home.cfm?program_id=4, 2011. (7) Lai, D.; Vallabhaneni, S.; Chan, C.; Brugess, E. H.; Field, R. I., A toolbox for sanitary sewer overflow analysis and planning (ssoap) and applications. In WEF 2007 Collection Systems Conference, Portland, OR, May 1316, 2007; Water Environment Federation: Alexandria, VA, 2007. (8) USEPA. Condition Assessment of Wastewater Collection Systems, State of Technology Review; EPA-600-R-09-049; Washington, DC, 2009. (9) Tafuri, A. N.; Selvakumar, A. Wastewater collection system infrastructure research needs in the USA. Urban Water 2002, 4, 21–29.
ARTICLE
(10) Guo, W.; Soibelman, L; Garrett, J., Jr. Automated defect detection for sewer pipeline inspection and condition assessment. Autom. Constr. 2009, 18 (5), 587–596. (11) USEPA. National Conference on Sanitary Sewer Overflows (SSOs), Washington DC, April 24-25, 1995; Washington, DC, 1995. (12) USEPA. Review of Sewer Design Criteria and RDII Prediction Methods; EPA/600-R-08/010; National Risk Management Research Laboratory: Cincinnati, OH, 2008. (13) Glassmeyer, S. T.; Furlong, E. T.; Koplin, D. W.; Cahill, J. D.; Zaugg, S. D.; Werner, S. L.; Meyer, M. T.; Kryak, D. D. Transport of Chemical and Microbial Compounds from Known Wastewater Discharges: Potential for Use as Indicators of Human Fecal Contamination. Environ. Sci. Technol. 2005, 39 (14), 5157–5169. (14) USEPA. Bacteriological Ambient Water Quality Criteria for Marine and Freshwater Recreational Waters; PB86-158-045; Springfield, VA, 1986. (15) Beurge, I. J.; Poigner, T.; Muller, M. D.; Buser, H. Caffeine, an anthropogenic marker for wastewater contamination of surface waters. Environ. Sci. Technol. 2003, 37 (4), 691–700. (16) Barone, J. J.; Roberts, H. R. Caffeine Consumption. Food Chem. Toxicol. 1996, 34 (1), 119–129. (17) Metcalf and Eddy. Wastewater Engineering, Treatment and Reuse, 4th ed.; McGraw-Hill: New York, 2003. (18) Eaton, A. D.; Clesceri, L. S.; Rice, E. W.; Greenberg, A. E.; Franson, M. A. H. Standard Methods for the Examination of Water and Wastewater, 21st ed.; American Public Health Association: Washington, DC, 2005. (19) Zaugg, S. D.; Smith, S. G.; Schroeder, M. P.; Barber, L. B.; Burkhardt, M. R. Methods of Analysis by the U.S. Geological Survey National Water Quality Laboratory - Determination of Wastewater Compounds by Polystyrene-Divinylbenzene Solid-Phase Exctraction and Capilary-Column Gas Chromatography/Mass Spectometry; U.S. Geological Survey Water-Resources Investigations Report 01-4186, 2006. (20) USEPA. Method 1603: Escherichia coli (E. coli) in Water by Membrane Filtration Using modified membrane Thermotolerant Escherichia coli Agar (Modified mTEC); Office of Water: Washington, DC, 2002. (21) USEPA. Method 1600: Enterococci in Water by membrane Filtration Using membrane-Enterococcus Indoxyl-b-D-Glucoside Agar (mEI); Office of Water: Washington, DC, 2002. (22) Haugland, R. A.; Siefring, S. C.; Wymer, L. J.; Brenner, K. P.; Dufour, A. P. Comparison of Enterococcus measurements in freshwater at two recreational beaches by quantitative polymerase chain reaction and membrane filter culture analysis. Water Res. 2005, 39 (4), 559–568. (23) Davies, C. M.; Long, J. A.; Donald, M.; Ashbolt, N. J. Survival of fecal microorganisms in marine and freshwater sediments. Appl. Environ. Microbiol. 1995, 61 (5), 1888–1896. (24) Feng, F.; Goto, D.; Yan, T. Effects of autochthonous microbial community on the die-off of fecal indicators in tropical beach sand. FEMS Microbiol. Ecol. 2010, 74 (1), 214–225. (25) Yamahara, K. M.; Walters, S. P.; Boehm, A. B. Growth of enterococci in unaltered, unseeded beach sands subjected to tidal wetting. Appl. Environ. Microbiol. 2009, 75 (6), 1517–1524. (26) Ishii, S.; Ksoll, W. B.; Hicks, R. E.; Sadowsky, M. J. Presence and growth of naturalized Escherichia coli in temperate soils from lake superior watersheds. Appl. Environ. Microbiol. 2006, 72 (1), 612–621. (27) Dauphin, L. A.; Stephens, K. W.; Eufinger, S. C.; Bowen, M. D. Comparison of five commercial DNA extraction kits for the recovery of Yersinia pestis DNA from bacterial suspensions and spiked environmental samples. J. Appl. Microbiol. 2010, 108 (1), 163–172. (28) Palmer, K. L.; Carniol, K.; Manson, J. M.; Heiman, D.; Shea, T.; Young, S.; Zeng, Q.; Gevers, D.; Feldgarden, M.; Birren, B.; Gilmore, M. S. High-quality draft genome sequences of 28 Enterococcus sp. isolates. J. Bacteriol. 2010, 192 (9), 2469–2470. (29) Patra, G.; Fouet, A.; Vaissaire, J.; Guesdon, J.-L.; Mock, M. Variation in rRNA operon number as revealed by ribotyping of Bacillus anthracis strains. Res. Microbiol. 2002, 153 (3), 139–148. 8689
dx.doi.org/10.1021/es2019115 |Environ. Sci. Technol. 2011, 45, 8683–8690
Environmental Science & Technology
ARTICLE
(30) Volkmann, H.; Schwartz, T.; Kirchen, S.; Stofer, C.; Obst, U. Evaluation of inhibition and cross-reaction effects on real-time PCR applied to the total DNA of wastewater samples for the quantification of bacterial antibiotic resistance genes and taxon-specific targets. Mol. Cell. Probes 2006, 21 (2), 125–133. (31) Koonjul, P. K.; Brandt, W. F.; Farrant, J. M.; Lindsey, G. G. Inclusion of polyvinylpyrrolidone in the polymerase chain reaction reverses the inhibitory effects of polyphenolic contamination of RNA. Nucleic Acids Res. 1999, 27 (3), 915–916. (32) Fujioka, R. S.; Tenno, K.; Kansako, S. Naturally occurring faecal coliforms and faecal streptococci in Hawaii’s freshwater streams. Toxic. Assess. 1988, 3, 613–630. (33) Goto, D. K.; Yan, T. Genotypic Diversity of Escherichia coli in the Water and Soil of Tropical Watersheds in Hawaii. Appl. Environ. Microbiol. 2011, 77 (12), 3988–3997.
8690
dx.doi.org/10.1021/es2019115 |Environ. Sci. Technol. 2011, 45, 8683–8690
ARTICLE pubs.acs.org/est
Vehicle and Driving Characteristics That Influence In-Cabin Particle Number Concentrations Neelakshi Hudda,† Evangelia Kostenidou,† Constantinos Sioutas,† Ralph J. Delfino,‡ and Scott A. Fruin*,§ †
Department of Civil and Environmental Engineering, University of Southern California, Los Angeles, California 90089, United States Department of Epidemiology, School of Medicine, University of California, Irvine, California 92617, United States § Keck School of Medicine, Environmental Health Division, University of Southern California, Los Angeles, California 90089, United States ‡
bS Supporting Information ABSTRACT: In-transit microenvironments experience elevated levels of vehicle-related pollutants such as ultrafine particles. However, in-vehicle particle number concentrations are frequently lower than on-road concentrations due to particle losses inside vehicles. Particle concentration reduction occurs due to a complicated interplay between a vehicle’s air-exchange rate (AER), which determines particle influx rate, and particle losses due to surfaces and the in-cabin air filter. Accurate determination of inside-to-outside particle concentration ratios is best made under realistic aerodynamic and AER conditions because these ratios and AER are determined by vehicle speed and ventilation preference, in addition to vehicle characteristics such as age. In this study, 6 vehicles were tested at 76 combinations of driving speeds, ventilation conditions (i.e., outside air or recirculation), and fan settings. Under recirculation conditions, particle number attenuation (number reduction for 10 1000 nm particles) averaged 0.83 ( 0.13 and was strongly negatively correlated with increasing AER, which in turn depended on speed and the age of the vehicle. Under outside air conditions, attenuation averaged 0.33 ( 0.10 and primarily decreased at higher fan settings that increased AER. In general, in-cabin particle number reductions did not vary strongly with particle size, and cabin filters exhibited low removal efficiencies.
’ INTRODUCTION The proximity of vehicles to relatively undiluted emissions from other vehicles on freeways and busy roadways leads to elevated pollutant concentrations in vehicle cabins compared to other indoor environments. Thus, a significant and disproportionate share of total personal exposure can occur during driving, especially for pollutants emitted mostly by vehicles, such as ultrafine particles (UFP, particles smaller than 100 nm). Fruin et al.1 calculated that in Los Angeles, 33 45% of UFP exposure occurs during driving, taking other microenvironmental concentrations and time activity patterns into account but ignoring in-vehicle particle losses. In suburban locations of less traffic, Wallace and Ott2 estimated the in-vehicle microenvironment contributes 17% to total UFP exposure. Despite its importance as a route of exposure, the contribution of the in-transit vehicular microenvironment remains largely uncharacterized, in part due to the difficulty of characterizing the large differences in rate of air turnover in the vehicle cabin, that is, the air-exchange rate (AER), which drives particle influx rates and varies not only from vehicle to vehicle but also across different driving conditions. To better characterize AERs, Fruin et al.3 tested 59 vehicles and reported that AER under recirculation r 2011 American Chemical Society
ventilation conditions can be reliably predicted on the basis of the vehicle’s age, mileage, driving speed, and manufacturer (r2 = 0.7). For the eight vehicles tested under outside air conditions, they found AERs to be an order of magnitude higher than at recirculation settings and reported strong positive correlation between AER and ventilation fan setting. In a similar study, Knibbs et al.4 reported AER values at higher fan settings under outside air conditions to be 73% higher than at the lowest fan setting for six vehicles. Knibbs et al.4 also observed that AER is influenced by driving speed for older vehicles. In-vehicle UFP concentrations result from an interaction between particle influx and removal rates, which in turn depend upon multiple factors. Besides being strongly influenced by AER, the physical characteristics of the vehicle, particle size, and incabin filter efficiency have been found to affect removal rates.5,7 Any accurate determination of the relative influence of each factor requires experiments in which they are systemically varied Received: June 15, 2011 Accepted: August 30, 2011 Revised: August 13, 2011 Published: September 19, 2011 8691
dx.doi.org/10.1021/es202025m | Environ. Sci. Technol. 2011, 45, 8691–8697
Environmental Science & Technology under real driving (aerodynamic) conditions, else AERs are not realistic. Several recent studies5,6 addressing in-vehicle particle losses have used artificial air movement or have relied on measurements in a controlled laboratory environment not representative of realworld driving. Furthermore, these studies have suffered from small sample sizes, ranging from only one to three vehicles.5,6 Of the few studies that have used real driving conditions, Pui et al.8 and Qi et al.9 demonstrated a dramatic reduction in incabin particle number concentrations under recirculation conditions in two vehicles, although AERs were not reported. Zhu et al.10 observed particle losses of about 85% at recirculation settings in three vehicles, but AER was not measured and variable speeds during the tests would have resulted in variable AERs because AER is a strong function of speed.3 The study of Zhu et al.10 was the only one identified that made size-resolved particle concentration measurements, up to 217 nm. They reported the in-vehicle to roadway concentration ratios to be primarily dependent on particle size and vehicle characteristics. The most useful study from an exposure assessment perspective has been by Knibbs et al.,11 who measured the inside-tooutside UFP concentration ratios in five vehicles and reported a high correlation between these ratios and AER (r2 = 0.81). They reported an average particle reduction of 0.69 during recirculation settings at low fan setting and 0.08 at outside air intake, but did not characterize the losses by particle size or test specific removal mechanisms such as cabin filtration. The goal of this study was to quantify particle reduction ratios due to changes in (1) ventilation settings (i.e., recirculation or outside air), (2) measured AER, (3) fan settings during outside air ventilation conditions, (4) driving speed, (5) filter condition, and (6) easily obtainable vehicle characteristics such as age and mileage (which strongly affect AER under recirculation settings) and to determine the relative importance of each of these variables under real driving conditions. It is the first study to combine measurements of AER and particle losses as a function of particle size.
’ MATERIALS AND METHODS Vehicle Selection and Conditions Tested. Six vehicles were selected such that AERs under recirculation conditions spanned the interquartile range of AERs (4.5 13 h 1, median 7.4 h 1) measured in our previous study.3 The previous study used a 59 vehicle selection chosen to be representative of the California fleet in terms of age, mileage, and manufacturer.3 Two older vehicles (a 1999 Ford Contour and a 2001 Ford Escort) covered the high range of AER (8 19 h 1 when mobile) and were more than 10 years old when tested. The four newer vehicles (a 2010 Toyota Prius, a 2010 Scion Xb, a 2009 Toyota Matrix, and a 2009 Honda Civic) were 3 16 months old and covered the low to medium ranges of AER (3 9 h 1 when mobile). Relevant vehicle characteristics (Table S1) and AER under RC conditions are summarized (Table S2) in the Supporting Information (SI). All six vehicles were evaluated under both recirculation (RC) and outside air intake (OA) ventilation conditions. At each ventilation condition, particle concentration measurements were conducted in both stationary and mobile modes at both medium and high fan speed settings. AERs at medium and full fan settings are identified in Table S3 of the SI. At RC setting, particle losses and AER were measured under 42 conditions, that is, six vehicles at seven AERs each (different AERs resulting from different combinations of ventilation fan setting and driving speed). 34
ARTICLE
conditions were evaluated under OA conditions. Air conditioning (AC) was kept on during all experiments, except for those at which the fan was off (6 of a total of 76 conditions evaluated). Particle Concentration Measurements. Particle number concentration measurements were made using a condensation particle counter (CPC, TSI Inc. model 3007, detection size range 10 1000 nm), and size-segregated number concentration measurements were made using a Scanning Mobility Particle Sizer with 300 s scanning rate (240 s up-scan and 60 s down-scan) (SMPS, TSI Inc. Differential Mobility Analyzer model 3080 and model 3022a CPC). The measured mobility diameter (Dp) range was 14 750 nm. However, data are presented only for the 14 400 nm range because above 400 nm relatively few particles were counted and the resulting concentrations had large uncertainties. Furthermore, because roadway particles between 14 and 25 nm are volatile, concentrations in this size range are exceptionally variable on roadway environments. Reported results for this size range should be interpreted with caution. Experiments were conducted at 0, 20, and 35 mi h 1 (0, 32, and 56 km h 1), with speed being recorded each second by a Garmin GPSMAP unit 76CSC. Experiments at 20 and 35 mi h 1 were conducted while driving at constant speed around the Rose Bowl Stadium in Pasadena, CA, a 5.5 km long loop with little vehicular traffic. Freeway speeds were not evaluated in this study due to the rapidly changing traffic and particle number and size distributions typically present on Los Angeles freeways.1,12,13 All in-vehicle measurements were made with windows fully closed. Before and after the in-vehicle measurement period, outside vehicle concentrations were measured for 10 20 min with windows fully open to allow the outside air to pass freely through the vehicle. The concentrations with fully open windows were assumed to equal the roadway concentrations. Earlier in the study, it was verified that open window conditions allow accurate measurement of roadway concentrations by comparing simultaneous measurements with two CPCs, one measuring concentrations with a 1 m long inlet sampling air right outside the vehicle and the other CPC sampling in the middle of the backseat of the vehicle. Ambient particle number concentrations and size distributions were also measured using a stationary monitor at a position central to the run loop. All ambient concentrations were stable to within 10% before, during, and after a run. The attenuation factor (AF), defined as 1 minus the inside-tooutside (I/O) particle concentration ratio, was calculated for all measurements, as a measure of particle removal rather than particle persistence within the vehicle cabin. (I/O can be readily calculated from AF by subtracting AF from 1.0.) AF was calculated after concentrations were stable in the vehicle over 15 min or more of sampling. AFs reported for specific size ranges are equivalent to the average of mobility-diameter-specific AFs within the range. Air Exchange Rate Measurements. During both mobile and stationary conditions, air exchange occurs between the vehicle cabin and the outside environment through leaks in the body of the vehicle (door seals, window cracks, etc.) and through the ventilation system, when it is set to draw outside air into the cabin. This air exchange continually replenishes the vehicle cabin with pollutants/particles from the outside environment and, hence, is an essential measurement in any study of in-vehicle exposure. AERs were determined at RC conditions as part of a previous study3 using CO2 as a tracer gas and two occupants as a stable source of CO2 generation. Buildup rates of CO2 concentration were used to determine the CO2 emission rates, and equilibrium 8692
dx.doi.org/10.1021/es202025m |Environ. Sci. Technol. 2011, 45, 8691–8697
Environmental Science & Technology
ARTICLE
Figure 1. AF dependence on AER for 25 400 nm particles under recirculation (RC) and outside air (OA) ventilation settings.
CO2 concentrations at fixed speeds then allowed calculation of the AER at that speed. Under OA conditions, AERs are much higher, requiring higher starting CO2 concentrations that cannot be readily reached by occupants. As it was observed that speeds up to 35 mi h 1 play an insignificant role in affecting AER compared to fan setting at OA conditions,3 only stationary tests were necessary to characterize AER, and CO2 from a pressurized cylinder was used to produce the necessary high starting concentrations. The AER determination procedure is described in detail in Fruin et al.3 Tables S2 and S3 of the SI list the AERs for the vehicles at various conditions tested.
’ RESULTS AND DISCUSSION Effect of AER on AF. Under RC ventilation conditions, 42 AERs across the six vehicles tested varied from <1.0 to 19 h 1. At the OA setting, measured AERs varied from 20 to 145 h 1. Any significant increase in AER resulted in an increased particle influx rate and lowered the AFs. This is demonstrated most distinctively in the difference between the AFs at RC and OA ventilation settings in Figure 1, which presents AF results under both ventilation conditions and illustrates the strong dependence of AF on AER. On average for the six vehicles tested, a switch in ventilation condition from RC to OA increased AERs by an order of magnitude and decreased AF by a factor of 2.6 ( 0.14. Any further increase in AER at each setting decreased AF less dramatically at OA conditions compared to RC conditions. Across the size range (25 400 nm), an increase in AER lowered AF, but the effect was strongest for particles above 200 nm. Under RC conditions, the r2 between AF and AER was 0.80, indicating that AER is the most significant determinant of AF at RC settings. (Table S4 in the SI lists correlation coefficients at different speeds.) The average AF at RC was 0.83 ( 0.13. In contrast, at OA setting, the AF averaged only 0.33 ( 0.10. The average r2 between AF and AER at OA setting was 0.75 (r2 values for all but one vehicle were g0.9). The only other study to measure AF and AER, that of Knibbs et al.,11 also used a CPC 3007 and reported an r2 of 0.81 between AF and AER, irrespective of
the ventilation setting. Additionally, measured AERs in Knibbs et al.4 and the corresponding predictions of the AER model developed in Fruin et al.3 for the same vehicles agreed well, with an r2 of 0.83. Effect of Vehicle Speed and Age on AER and AF. Under RC conditions, an increase in speed increased AER and decreased AF. On average, a 10 mi h 1 increase in speed resulted in a 1.65 h 1 increase in AER and a 0.035 decrease in AF. Speed affected AER more for older vehicles (+2.4 h 1/10 mi h 1) compared to the newer vehicles (+1.2 h 1/10 mi h 1), similar to the results reported by Knibbs et al.4 As a result, AF, which depends strongly on AER, decreased with speed at twice the rate for older vehicles ( 0.05/10 mi h 1, Pearson r2 = 0.78) than for newer vehicles ( 0.025/10 mi h 1, Pearson r2 = 0.20). Despite these differences by vehicle age, an overall strong correlation was observed between AER and speed as well as AF and AER at RC settings. In Fruin et al.3 an r2 equivalent of 0.92 was calculated between AER and speed for a much larger fleet of vehicles and for speeds up to 70 mi h 1 using a Generalized Estimating Equation (GEE).16 (GEE techniques account for correlated measurements within a vehicle; e.g, a tight vehicle with lower AER will consistently have a higher AF across each speed compared to a leakier vehicle with higher AER.) In this study, also using GEE techniques, 82% of the variation in AER could be accounted for by speed (p value = 2.5 10 9). Furthermore, nearly all variation (r2 = 0.98, p value = 6.9 10 7) in AF at RC setting was explained by changes in AER. Given the consistent and strong correlations between AER and speed (as shown in ref 3 and this study) and between AF and AER at RC setting (as shown in ref 11 and this study), these results can be expected to extrapolate well to the higher AERs of vehicles traveling at higher freeway speeds (65 70 mi h 1) from the speeds tested in this study, that is, e35 mi h 1. In contrast, under OA conditions, no definitive association between speed and AF could be discerned for the six vehicles tested. Because AER at OA is driven by mechanical ventilation rather than speed-driven pressure differences outside the vehicle shell, this lack of association is not surprising. Knibbs et al.4 have 8693
dx.doi.org/10.1021/es202025m |Environ. Sci. Technol. 2011, 45, 8691–8697
Environmental Science & Technology
ARTICLE
Figure 2. Agreement between measured AF during variable speed driving and model-predicted AF.
previously shown that at OA conditions, for the newer four of the total six vehicles tested, the association between AER and speed was weak [linear regression on all four vehicles: AER = 0.15 speed (mi h 1) + 51, r2 = 0.06]. For the oldest two (10 and 19 year old) of the six total vehicles tested, the relationship was stronger [AER = 0.59 speed (mi h 1) + 52, r2 = 0.52]. This suggests that for older vehicles the AER at OA may increase at speeds higher than those tested in this study, which should lead to a reduction in AF. However, for the two oldest vehicles in this study (10 years of age), any increase in AER with speeds between 1 and 35 mi h 1 did not cause a noticeable change in AF. Most real-world driving involves constantly changing speeds due to traffic conditions and widely varying roadway particle number concentrations. To illustrate how well the steady speed/ AER and relatively stable roadway concentration condition results apply to such changing speed and concentration conditions, measurements were made during two 15 20 min trips on a freeway (I-110) and an arterial road (Figueroa Street, downtown Los Angeles) in a 2009 Toyota Matrix. No attempt was made to maintain steady speeds during these runs, and average speeds were 55 and 27 mi h 1 on the freeway and arterial road, respectively. The ventilation setting was set to OA, leading to moderately high AER (35 h 1) allowing the rapid influx of roadway concentration into the vehicle cabin and reflecting the unsteady roadway environment, as shown in Figure 2. It should be noted that under low AER conditions, in-cabin concentrations are fairly steady due to limited influx of particles. The average AF calculated from average in-cabin and roadway particle number concentrations agreed almost perfectly with the AF predicted from the AF versus speed regression equation based on measurements at steady speeds. Effect of Particle Size on AF. Other than AER, particle size itself can be expected to play a role in attenuation, because particle infiltration, surface deposition, and filtration efficiency are functions of particle size.14,15,17 AFs for the five size ranges plotted in Figure 3 show the lowest attenuation for particles in the size range 200 400 nm. Similar values for the least attenuated particle size have been reported in other indoor environments.14 The highest AFs were consistently associated with the UFP size
range. This is consistent with the high losses expected for smaller particles due to higher diffusion rates.17 AFs under various conditions tested are summarized in section S9 of the SI (Tables S6 S9). It should be noted that AFs for the size ranges <25 and >400 nm have relatively high uncertainty. For a given ventilation and speed combination, particle sizespecific AFs were found to be similar across size for newer vehicles and only moderately different across size for older vehicles. At RC, the AFs for 100 200 and 200 400 nm were, respectively, 0.04 (5%) and 0.07 (8%) lower than the average AF (0.84 ( 0.09) for ultrafine range. At OA, the differences in sizespecific AF were more accentuated than at RC. The AFs for both 100 200 and 200 400 nm were 0.06 (30%) lower than the average AF (0.25 ( 0.12) for ultrafine range. As can be observed in Figure 3, the difference in AF across size was much less than the difference in AF across ventilation conditions. Effect of Ventilation Fan Setting on AF. Figure 4 shows that increasing the ventilation fan setting from medium to full lowers the AF at both RC and OA, but this reduction was far more pronounced at OA than at RC. At RC, the attenuation reduction is somewhat counterintuitive, because increasing fan setting might be thought to increase particle removal via greater rates of airflow through the in-cabin filter, when present. Also, the deposition of particles on cabin surfaces increases with an increase in in-cabin air velocity.5 However, fan setting has been previously demonstrated to increase AER for older vehicles,3,4 likely due to leaks into the ventilation system, and filtration efficiency decreases at greater air flow velocities associated with higher fan settings.9 Apparently, these mostly offset the particle reductions expected due to greater volumes of air being filtered at high fan settings under RC conditions. However, higher fan settings resulted in increased losses for particles smaller than 50 nm, perhaps due to the increased turbulence in the ventilation system at higher fan settings. At OA, an increase in fan speed settings strongly increased AER. The median difference in AF at full and medium fan settings at OA was 42% (interquartile range 14 97%), and AERs at full fan setting were about 65% higher than at medium fan. Thus, ventilation fan setting is a key predictor of AF under OA ventilation conditions. 8694
dx.doi.org/10.1021/es202025m |Environ. Sci. Technol. 2011, 45, 8691–8697
Environmental Science & Technology
ARTICLE
Figure 3. Size range specific AF at three speeds and two ventilation conditions tested. The dashed lines join values from the same vehicle.
Figure 4. Comparison of AF at different speeds and fan settings.
Ventilation fan speed also affects the time needed to reach stable, in-vehicle particle concentrations. Figure S1 in the SI (section S.3) illustrates the decay of total particle number concentration with time in the Honda Civic for the first 10 min of the experiment in the vehicle. The rate of removal of particles increased moderately with an increase in fan speed but resulted in lower attenuation factors. This observation was consistent across all six vehicles tested. This aspect of AF may be important during short vehicle trips. Effect of Cabin Air Filter and Loading on AF. To determine the effect of cabin filters on particle attenuation, AF measurements were conducted under several filter conditions, no filter, used (loaded), and new, at both OA and RC setting in three stationary vehicles. The Ford Contour’s in-use filter had been operational for 36 months at the time of testing and was heavily loaded. The Honda Civic’s filter had been in operation for ∼14 months at the
Figure 5. Attenuation of particles by filter condition or absence under OA conditions in a 2010 Toyota Prius.
time of testing and was moderately loaded, and the Prius’s filter had been in use for ∼3 months and was lightly loaded. All new filters were standard replacement cabin air filters bought from an auto parts store (all either STP or Purolator brand). The presence or absence of the filter, or its loading, was observed to have only a small effect on AF. Used, loaded filters provided only moderately higher or comparable AFs to a new filter.16 The results in Figure 5 for a Toyota Prius are typical and show that overall particle loss is not significantly affected by the presence of a new filter or even a loaded filter (section S5 in the SI) at OA settings. Without any filter, AF was only moderately lower (maximum AF difference observed was 0.1). This implies that the particle attenuation due to filtration is a small fraction of the total attenuation and that most of the attenuation is probably due to turbulent surface deposition in the ventilation 8695
dx.doi.org/10.1021/es202025m |Environ. Sci. Technol. 2011, 45, 8691–8697
Environmental Science & Technology system or vehicle surface itself. Pui et al.8 reported 19% particle loss in a Toyota Camry in the absence of filter and suggested intrinsic losses in the ventilation system as an attenuation mechanism. In our tests 23, 40, and 40% particle losses were observed for the Honda Civic, Toyota Prius, and Ford Contour with no filter, respectively. AFs based on total particle number concentration are reported in Table S5 of the SI for three vehicles by filter presence and condition. At RC settings, the AF in the absence of a filter was on average only 5% lower than with a filter in place. Although the presence and condition of the filter were insignificant to AF, they somewhat affected the time that the system required to achieve maximum attainable attenuation (section S6 in the SI). Similar increases in the time required to reach maximum attenuation have been reported by Pui et al..8 At OA condition, the AF in the filter’s absence was on average about 20% lower than in that the filter’s presence (<0.1 AF unit). Furthermore, the effect of several different filter loadings as characterized by in-vehicle pressure drop for the same vehicle was also investigated at OA. Increased loading and the resulting reduced pore size and flow rates were found to increase AF for UFP, but no significant changes were observed for particles exceeding 100 nm in size. Results are discussed in section S7 of the SI.
’ IMPLICATIONS FOR IN-VEHICLE PARTICLE MODELS Some recent studies6,7 attribute all particle losses in the ventilation system to filtration and incorporate them into models by using filter efficiency as the removal mechanism. If the presence of cabin filters actually plays a minor role in the attenuation of particles inside the vehicles, as observed in our measurements, a much larger component of attenuation occurs due to losses onto cabin and ventilation system surfaces. These different loss mechanisms should therefore be differentiated in models because they likely differ under different ventilation and driving conditions. Suggested modified equations are presented in section S8 in the SI. Furthermore, any quantitative modeling should account for the intrusion flow of outside air into the ventilation system at RC. Knibbs et al.4,11 and Fruin et al.3 report that at RC increases in fan setting increase the AER. They also reported that this increase seems to increase with the age of the vehicle and could vary considerably by vehicle. A <15% increase on average was reported by Fruin et al.3 when the fan setting was changed from medium to full, but this became as large as 40% for the older vehicles. An equation modified to include this leak as a variable is also presented in section S8 of the SI. ’ IMPLICATIONS FOR EXPOSURE ASSESSMENT Predicting particle exposure inside vehicles requires determining ventilation setting first and foremost (i.e., OA or RC), due to its large impact on AER. Under OA conditions, fan setting is the most dominant variable, and AF was approximately 0.4 and fairly independent of speed. Under RC conditions, AF has a large range and varies from 0.5 to 1.0, depending on AER, which can be predicted by speed and vehicle age and mileage.3 Under openwindow conditions, AF approaches 0. Difficult to obtain information, such as the state of the in-cabin filter loading, does not appear to be a crucial factor in assessing AF and the resulting in-vehicle particle exposures. It also does not appear that changes in on-road size distribution have a large impact on AF. Figure 6 exhibits the AF differences for four widely
ARTICLE
Figure 6. Impact of change in particle size distribution on number concentration weighed attenuation factors (AF).
different hypothetical size distributions having number concentration mode <25 nm (fresh vehicle exhaust plume), 25 50 nm (on-road diluted plumes), 50 100 nm (aged vehicle emissions), and 100 200 nm (aged aerosol observed as urban background). The largest difference in AF occurs between aged and fresh aerosol, changing the overall AF by no more than 0.1 at OA and by <0.05 at RC. Considering that on-road particle size number concentrations are dominated by ultrafine particles,13 AF measurements based only on total number concentration (e.g., a total particle count from CPC) would be expected to (and were observed to) produce AF measurements nearly identical to an average AF resulting from a number-weighted average AF across multiple size ranges. Therefore, all of the variables needed to estimate AF within 10% or less can be obtained through questionnaires given to vehicle owners. Finally, for the case of short duration trips, the equilibration conditions reported in this study for AF may not be reached. The particle attenuation reported here was attained typically within 5 10 min at OA conditions and within 15 20 min at RC conditions (see section S10 of the SI). The benefit derived from higher reduction in particle exposure with the use of RC setting may be reduced if short trips are taken repeatedly. The dynamic nature of particle attenuation should be considered in the assessment of short trip exposures.
’ ASSOCIATED CONTENT
bS
Supporting Information. The vehicles tested are listed (S1) along with AERs at the two ventilation settings (Table S2 and S3). The effects of driving speed (S3), ventilation fan setting (S4), and filter’s presence as well as state on AF are shown (S5, S6, and S7). Furthermore, suggestions as to how to incorporate the major findings of the study into ultrafine particle modeling in vehicle cabins are discussed (S8). Size range specific AF are listed (S9), and progression of particle loss is discussed (S10). This material is available free of charge via the Internet at http:// pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected]. 8696
dx.doi.org/10.1021/es202025m |Environ. Sci. Technol. 2011, 45, 8691–8697
Environmental Science & Technology
ARTICLE
’ DISCLOSURE The statements and conclusions in this paper are those of the authors and not necessarily those of the California Air Resources Board. ’ ACKNOWLEDGMENT This work was funded by the California Air Resources Board through Contract 07-310. We acknowledge Sandrah P. Eckel at the USC Division of Biostatistics for statistical advice. ’ REFERENCES (1) Fruin, S. A.; Westerdahl, D.; Sax, T.; Fine, P. M.; Sioutas, C. Measurements and predictors of in-vehicle ultrafine particle concentrations and associated pollutants on freeways and arterial roads in Los Angeles. Atmos. Environ. 2008, 42, 207–219. (2) Wallace, L.; Ott, W. Personal exposure to ultrafine particles. J. Exposure Sci. Environ. Epidemiol. 2011, 21, 20–30. (3) Fruin, S. A.; Hudda, N.; Sioutas, C.; Delfino, R. A predictive model for vehicle air exchange rates based on a large and representative sample. Environ. Sci. Technol. 2011, 45, 3569–3575. (4) Knibbs, L. D.; de Dear, R. J.; Atkinson, S. E. Field study of air change and flow rate in six automobiles. Indoor Air 2009, 303–313. (5) Gong, L.; Xu, B.; Zhu, Y. Ultrafine particles deposition inside passenger vehicles. Aerosol Sci. Technol. 2009, 43, 544–553. (6) Xu, B.; Zhu, Y. Quantitative analysis of the parameters affecting in-cabin to on-roadway (I/O) ultrafine particle concentration ratios. Aerosol Sci. Technol. 2009, 43, 400–410. (7) Xu, B.; Shusen, L.; Junjie, L.; Zhu, Y. Effects of vehicle cabin filter efficiency on ultrafine particle concentration ratios measured in-cabin and on-roadway. Aerosol Sci. Technol. 2011, 45, 215–224. (8) Pui, D. Y. H.; Qi, C.; Stanley, N.; Oberdorster, G.; Maynard, A. Recirculating air filtration significantly reduces exposure to airborne nanoparticles. Environ. Health Perspect. 2008, 116, 863–866. (9) Qi, C.; Stanley, N.; Pui, D. Y. H.; Kuehn, T. H. Laboratory and on-road evaluations of cabin air filters using number and surface area concentration monitors. Environ. Sci. Technol. 2008, 42, 4128–4132. (10) Zhu, Y.; Eiguren-Fernandez, A.; Hinds, W. C.; Miguel, A. In-Cabin commuter exposure to ultrafine particles on Los Angeles freeways. Environ. Sci. Technol. 2007, 41, 2138–2145. (11) Knibbs, L. D.; de Dear, R. J.; Morawska, L. Effect of cabin ventilation rate on ultrafine particle exposure inside automobiles. Environ. Sci. Technol. 2010, 44, 3546–3551. (12) Westerdahl, D.; Fruin, S.; Sax, T.; Fine, P. M.; Sioutas, C. Mobile platform measurements of ultrafine particles and associated pollutant concentrations on freeways and residential streets in Los Angeles. Atmos. Environ. 2005, 39, 3597–3610. (13) Morawska, L.; Thomas, S.; Bofinger, N.; Wainwright, D.; Neale, D. Comprehensive characterization of aerosols in a subtropical urban atmosphere: particle size distribution and correlation with gaseous pollutants. Atmos. Environ. 1998, 32, 2467–2478. (14) Liu, D. L.; Nazaroff, W. W. Particle penetration through building cracks. Aerosol Sci. Technol. 2003, 37, 565–573. (15) Nazaroff, W. W. Indoor particle dynamics. Indoor Air 2004, 14, 175–183. (16) Liang, K.; Zeger, S. Longitudinal data analysis using generalized linear models. Biometrika 1986, 73, 13–22. (17) Hinds, W. C. Aerosol Technology: Properties, Behavior, and Measurement of Airborne Particles; Wiley: New York, 1999.
8697
dx.doi.org/10.1021/es202025m |Environ. Sci. Technol. 2011, 45, 8691–8697
ARTICLE pubs.acs.org/est
Discontinued and Alternative Brominated Flame Retardants in the Atmosphere and Precipitation from the Great Lakes Basin Amina Salamova and Ronald A. Hites* School of Public and Environmental Affairs, Indiana University, Bloomington, Indiana 47405, United States
bS Supporting Information ABSTRACT: Air (vapor and particle) and precipitation samples were collected at five sites (two urban, one rural, and two remote) around the Great Lakes during 20052009 as a part of the Integrated Atmospheric Deposition Network (IADN). The concentrations of polybrominated diphenyl ethers (PBDEs), decabromodiphenylethane (DBDPE), hexabromobenzene (HBB), pentabromoethylbenzene (PBEB), and 1,2-bis(2,4,6-tribromophenoxy)ethane (BTBPE) were measured in these samples. The highest concentrations of these compounds were generally observed at the two urban sites—Chicago and Cleveland—with a few exceptions: The remote site at Eagle Harbor had particularly high levels of PBEB in all three phases, and the rural Sturgeon Point site had the highest HBB concentrations in the vapor phase. The sources of HBB and PBEB to these sites are unknown. A multiple linear regression model was applied to the concentrations of these compounds in the vapor phase, particle phase, precipitation, and the three phases combined. This regression resulted in overall (three phases combined) halving times for total PBDE concentrations of 6.3 ( 1.1 years. The overall halving times for HBB and BTBPE were 9.5 ( 4.6 years and 9.8 ( 2.8 years, respectively. For PBEB and DBDPE, the regression was not statistically significant for the combined phases, indicating that the atmospheric concentrations of these compounds have not changed between 2005 and 2009.
’ INTRODUCTION The rapid development of the polymer industry over the last 100 years eventually led to requirements that some of these materials be flame resistant. For example, polyurethane foam used in furniture must be able to withstand an open flame for 12 s.1 These flame resistant properties are imparted to the polymeric materials by adding percent level amounts of flame retardants. Over the last several decades, this has been a growth market for the chemical industry, and there are now more than 175 chemicals classified as flame retardants. Of these, at least 75 are brominated flame retardants (BRFs).2 Among them, the polybrominated diphenyl ethers (PBDEs) have been widely used and are persistent and accumulate in the environment.3,4 As a result, the use of the penta-, octa-, and deca-BDE commercial mixtures was restricted in the European Union,5,6 the production and use of the penta- and octa-mixtures in the United States was voluntarily phased out in 2004, and the production, import, and sale of the deca-BDE mixture in the United States will be discontinued by the end of 2013.7 These restrictions have led to an increased market demand for nonregulated flame retardants—or at least, flame retardants that are not in the news. These include decabromodiphenylethane (DBDPE) and 1,2-bis(2,4,6-tribromophenoxy)ethane (BTBPE), both of which have been marketed as alternatives to various PBDE r 2011 American Chemical Society
formulations. In addition, “older” chemicals are sometimes being reintroduced to the market. These include hexabromobenzene (HBB) and pentabromoethylbenzene (PBEB), compounds that have apparently been manufactured for several decades. Although little is known about their uses or production volumes, the environmental presence of several alternative flame retardants has been reported recently.813 Interestingly, de Wit et al.10 have detected several alternative flame retardants in the Arctic, which suggests long-range atmospheric transport of some of these chemicals. The goal of this study is to provide insights on the long-term spatial and temporal distribution patterns of PBDEs, DBDPE, HBB, PBEB, and BTBPE in the atmosphere (vapor and particle phases) and in precipitation at the five United States Integrated Atmospheric Deposition Network (IADN) sites located in the Great Lakes basin. To elucidate temporal trends in these chemicals’ concentrations, we used a harmonic regression approach14 that includes local population and distance from an assumed source to model the measured concentrations in each Received: June 15, 2011 Accepted: September 7, 2011 Revised: August 18, 2011 Published: September 26, 2011 8698
dx.doi.org/10.1021/es2020378 | Environ. Sci. Technol. 2011, 45, 8698–8706
Environmental Science & Technology
ARTICLE
phase (vapor, particle, and precipitation) separately and in all the phases combined at all sites.
instrumental analysis and quality control and quality assurance procedures are provided in the Supporting Information.
’ EXPERIMENTAL SECTION The locations of the United States IADN sampling sites are shown in the Supporting Information (SI) (Figure S1). The two urban sites are in Chicago, IL (41.8344°N, 87.6247°W) and Cleveland, OH (41.4921°N, 81.6785°W). A rural site is located at Sturgeon Point, NY (42.6931°N, 79.0550°W). The two remote sites are at Sleeping Bear Dunes, MI (44.7611°N, 86.0586°W) and Eagle Harbor, MI (47.4631°N, 88.1497°W). The IADN Web site provides detailed information on air sampling procedures and site operations (www.msc.ec.gc.ca/iadn). The samples discussed here were collected during the period of January 1, 2005 to December 31, 2009. A modified Anderson high-volume air sampler (General Metal Works, model GS2310) was used to collect air samples for 24 h every 12 days at a flow rate giving a total sample volume of about 820 m3. The vapor phase was collected on Amberlite XAD-2 resin (Supelco, Bellefonte, PA; 2060 mesh) held in a stainless steel cartridge, and particles were collected on Whatman quartz fiber filters (QM-A, 20.3 25.4 cm). Details of the sampling procedures and site operations can be found elsewhere.15 Precipitation samples were collected using MIC automated wet-only samplers (MIC Co., Thornhill, ON). Each sampler consists of a 46 46 cm stainless steel funnel connected to a 30 cm long by 1.5 cm i.d. glass column (ACE Glass, Vineland, NJ) packed with XAD-2 resin. The sampler is normally covered; it opens when a precipitation event is sensed by a conductivity grid located outside of the sampler. Precipitation flows through the funnel and the XAD-2 column into a large carboy used to measure the total precipitation volume. Both particulate and dissolved organic phase compounds are collected by the XAD-2 column. The funnel and the interior of the sampler are kept at 15 ( 5 °C to melt snow collected in the sampler and to prevent the XAD column from freezing. Details on the performance of these samplers are provided elsewhere.16 Precipitation events are integrated over each calendar month. A detailed description of the sample treatment and chemical analysis procedures for the air and precipitation samples has been given elsewhere.17 In summary, the samples were Soxhlet extracted for 24 h with a 1:1 acetone hexane mixture. Prior to extraction, a recovery standard was spiked into the sample; this standard included known amounts of BDE-77, BDE-166, and 12 13C -BDE-209. The extract was reduced in volume by rotary evaporation, the solvent was exchanged to hexane, and fractionated on a column containing 3.5% w/w water deactivated silica gel (3.0% w/w for precipitation samples). This column was eluted with 25 mL of hexane (fraction 1) and 25 mL of a 1:1 hexane dichloromethane mixture (fraction 2). After N2 blow down, the samples were spiked with the quantitation internal standard, consisting of a known amount of BDE-118. For chemical analysis, the samples were further concentrated by N2 blow down to ∼100 μL. The samples were analyzed for 34 PBDE congeners (7, 10, 15, 17, 28, 30, 47, 66, 85, 99, 100, 119, 126, 138140, 153, 154, 156, 169, 180, 183, 184, 191, 196, 197, 201, and 203209), and for decabromodiphenylethane (DBDPE), hexabromobenzene (HBB), pentabromoethylbenzene (PBEB), and 1,2-bis(2,4,6-tribromophenoxy)ethane (BTBPE) on an Agilent 6890 series gas chromatograph coupled to an Agilent 5973 mass spectrometer using helium as the carrier gas. Details on the
’ RESULTS AND DISCUSSION Polybrominated Diphenyl Ethers. PBDEs were found in the gas, particle, and precipitation phases. Figure 1 shows the spatial distributions of total PBDE (∑PBDE), BDE-47, 99, and 209 concentrations in the vapor, particle, and precipitation phases at the five U.S. IADN sites. Table 1 gives the arithmetic mean and standard errors of these concentrations along with ANOVA results (applied to log-transformed concentrations) and percent detected. The highest ∑PBDE concentrations were detected at the two urban sites (Chicago and Cleveland) in all three phases. The mean ∑PBDE concentrations in the vapor and particle phases at the rural Sturgeon Point (SP) site and at the remote Sleeping Bear Dunes (SB) site were lower but statistically indistinguishable from each another. The lowest ∑PBDE concentrations in the vapor and particle phases were measured at the remote Eagle Harbor (EH) site. In precipitation, there were no statistical differences among the ∑PBDE concentrations measured at the three rural and remote sites. Among the studies that have measured PBDE concentrations in the atmosphere only a few have reported concentrations for vapor and particle phases separately; this limits direct comparison with our results. In addition, not all of these previous studies have included BDE-209 in the reported ∑PBDE concentrations. Nevertheless, the spatial distribution pattern in ∑PBDE concentrations observed here was similar to the one measured in a previous study from our laboratory, where the ∑PBDE concentrations (sum of vapor plus particle phase concentrations) were shown to be strongly correlated with the human population density at a given site.18 In this previous study, ∑PBDE concentrations varied from a low of 5.8 ( 0.4 pg/m3 at the remote Eagle Harbor site to a high of 87 ( 8 pg/m3 at the urban Cleveland site. The levels of ∑PBDE concentrations (sum of vapor plus particle phase concentrations) in air samples collected in 19971999 at the same Great Lakes sites, except Cleveland, were ∼50 pg/m3 for Chicago and 515 pg/m3 at the other rural and remote sites.19 All of these levels were similar to our results. Salamova and Hites20 reported average ∑PBDE concentrations of 34 ( 3.5 pg/m3 and 23 ( 2.3 pg/m3 in Chicago and 26 ( 2.6 pg/m3 and 58 ( 6.9 pg/m3 in Cleveland in the vapor and particle phases, respectively, in air samples collected in 20032007 from the same Great Lakes sites. The levels at the rural and remote sites were in the range of 24 pg/m3 for the vapor phase and 25 pg/m3 for particles. ∑PBDE concentrations in precipitation ranged from ∼65 ng/L at the urban sites to ∼1 ng/L at the rural and remote sites. These levels were similar to our values. Our current values are also in the range of reported ∑PBDE concentrations in precipitation samples collected at the same Great Lakes sites in 20052006.21 On the global scale, our vapor and particle ∑PBDE concentrations for urban sites are similar to those reported at urban and industrial sites in Izmir, Turkey (929 pg/m3 in the gas phase and 2762 pg/m3 in particles),22 at a site near a sanitary landfill in Ottawa, Canada (∼20 pg/m3),23 and after a dust storm event in Kuwait (4754 pg/m3).24 The atmospheric ∑PBDE concentrations reported for rural sites in China are significantly higher than our values (210690 pg/m 3 ); 25 however, the values reported in Chinese precipitation samples are within the concentration range measured in this study (962 ng/L).25 8699
dx.doi.org/10.1021/es2020378 |Environ. Sci. Technol. 2011, 45, 8698–8706
Environmental Science & Technology
ARTICLE
Figure 1. Concentrations of ∑PBDE, BDE-47, BDE-99, and BDE-209 in the atmospheric vapor phase, particle phase, and precipitation; the thin black lines represent the median, and the thick red lines represent the arithmetic mean; the boxes represent the 25th and 75th percentiles; the whiskers represent the 5th and 95th percentiles. Site abbreviations: EH Eagle Harbor, CH Chicago, SB Sleeping Bear Dunes, CL Cleveland, SP Sturgeon Point.
Table 1. Arithmetic Mean ( Standard Error of ∑PBDE, BDE-47, BDE-99, and BDE-209 Concentrations in Vapor Phase (pg/m3), Particle Phase (pg/m3), and Precipitation (ng/L) at the Five U.S. IADN Sitesa Eagle Harbor conc.
*
Chicago
% det
conc.
*
Sleeping Bear Dunes % det
conc.
*
% det
Cleveland conc.
*
Sturgeon Point
% det
conc.
*
% det
∑PBDE
vapor
2.6 ( 0.4
c
100
35 ( 4
a
100
4.8 ( 0.7
b
100
25 ( 2
a
100
7.2 ( 1.6
b
100
BDE-47
vapor
1.1 ( 0.2
c
64
18 ( 1.8
a
94
1.7 ( 0.3
bc
83
14 ( 1
a
98
1.5 ( 0.2
b
73
BDE-99 BDE-209
vapor vapor
1.2 ( 0.3 0.5 ( 0.1
b c
59 65
6.4 ( 1.2 3.4 ( 0.9
a a
95 51
1.4 ( 0.3 0.8 ( 0.3
b c
80 66
4.8 ( 0.6 1.8 ( 0.5
a b
86 34
1.2 ( 0.4 0.7 ( 0.3
b c
72 76
∑PBDE
particle
3.2 ( 0.6
d
100
25 ( 3
a
100
4.3 ( 0.7
c
100
62 ( 13
a
100
8.2 ( 0.7
bc
100
BDE-47
particle
0.7 ( 0.2
b
42
4.5 ( 0.5
a
88
0.5 ( 0.1
b
68
4.5 ( 0.7
a
58
0.7 ( 0.1
b
74
BDE-99
particle
1.4 ( 0.5
b
23
3.9 ( 0.8
a
88
1.0 ( 0.1
b
80
7.3 ( 1.5
a
50
1.2 ( 0.2
b
70
BDE-209
particle
1.3 ( 0.4
d
64
13 ( 2
b
96
2.5 ( 0.9
c
69
56 ( 15
a
86
1.9 ( 0.2
c
81
∑PBDE BDE-47
precip precip
2.1 ( 0.4 0.9 ( 0.4
cd bc
100 100
42 ( 8 20 ( 4
a a
100 100
5.2 ( 1.2 2.5 ( 0.6
c b
100 100
6.4 ( 1.5 0.9 ( 0.2
b b
100 100
1.5 ( 0.3 1.3 ( 0.8
d c
100 100
BDE-99
precip
0.2 ( 0.02
b
100
8.7 ( 1.8
a
100
1.0 ( 0.4
b
100
0.7 ( 0.2
b
100
0.1 ( 0.02
c
100
BDE-209
precip
0.4 ( 0.1
b
97
2.1 ( 0.3
a
100
0.5 ( 0.1
b
100
4.1 ( 1.3
a
98
0.6 ( 0.1
b
100
a
ANOVA results for the log-transformed concentrations are given in the columns headed by an asterisk; the concentrations are not significantly different for those locations sharing the same letter. Percent of detects for each compound in each phase is also shown.
The ∑PBDE concentrations (excluding BDE-209) measured in the atmosphere over the Southern and Atlantic Oceans are similar to those reported here for the remote sites (1.1 pg/m3 for the gas phase and 0.3 pg/m3 for particles).8 ∑PBDE concentrations
reported in precipitation samples from the Baltic Sea26 and from Sweden27,28 are also similar to our values. BDE-47, -99, and -209 were the main components of the heavily used PBDE commercial mixtures, and as a result, these 8700
dx.doi.org/10.1021/es2020378 |Environ. Sci. Technol. 2011, 45, 8698–8706
Environmental Science & Technology
ARTICLE
congeners comprise up to 99% of the measured ∑PBDE concentrations (see SI Figure S2). Similar to the ∑PBDE spatial distribution pattern, the highest levels of these individual congeners were detected at the urban sites of Chicago and Cleveland, and the levels at the other three sites were generally indistinguishable from one another. Interestingly, the concentrations of BDE-209 in the particle phase were particularly high at Cleveland (56 ( 15 pg/m3). These elevated levels of BDE-209 in the particle phase at Cleveland have been shown to be associated with local sources.18 The levels of BDE-47, 99, and 209 detected in this study are in the range of the PBDE congener concentrations measured in air and precipitation samples collected in 20032007 at the same Great Lakes sites.20 Overall, BDE-47 is the relatively most abundant congener in the vapor phase, and BDE-209 is the relatively most abundant congener in the particle phase. This is true at all five sites except Sturgeon Point (see SI Figure S2). In precipitation, BDE-47 is the most abundant PBDE congener at all the sites, except Cleveland, where BDE-209 is the most abundant PBDE congener. This pattern is different from the pattern observed in precipitation samples in previous studies from our laboratory.20,21 In these previous studies, BDE-209 was the relatively most abundant congener in precipitation at all the sites except Chicago, where BDE-47 was the most abundant congener. PBDE Temporal Trends. Following the approach of Venier and Hites,14 the concentration data for all the sites were combined together for each phase. The natural logarithms of these concentrations were fitted using the following harmonic regression equation: lnðCÞ ¼ a0 þ a1 sinðztÞ þ a2 cosðztÞ þ a3 log2 ðpopÞ þ a4 logðdistÞ þ a5 t þ ε
ð1Þ
where C is the concentration in the gas phase, particle phase, or precipitation, t is time expressed in Julian days starting from January 1, 2005, z = 2π/365.25 (which fixes the periodicity at one year), pop is the number of people within a 25 km radius of the sampling site,14 dist is the distance (in km) of the sampling site from Cleveland, a0 is an intercept which rationalizes the units, a1 and a2 describe the seasonal variations in the concentration with time, a3 describes the change in concentration as a function of population, a4 is the coefficient describing the change of concentration as a function of distance from Cleveland, a5 is a first-order rate constant (in days1), and ε is the regression residual. Cleveland was assumed to be the source location because of relatively high BDE209 concentrations previously measured at this site.18 The calculations for this multiple regression were done using Minitab 16, which returned the coefficients, their standard errors, and the sum-ofsquares associated with each term. The details on the derivation of this expression and the calculations of the maximum concentration date and halving time can be found elsewhere.14 This regression was applied to ∑PBDE, BDE-47, -99, and -209 concentrations, and the coefficients and their standard errors from the regression for these compounds, as well as the dates of maximum concentration and halving times are reported in SI Table S1. We then normalized the data to the same scale by subtracting the variations caused by local human population, seasonal factors, distance from an assumed source, and the intercept. This correction gives what are known as “partial residuals” lnðCÞ a0 a1 sinðztÞ a2 cosðztÞ a3 log2 ðpopÞ a4 logðdistÞ ¼ ε0
ð2Þ
Table 2. Halving and Doubling (Noted with a Negative Sign) Times (In Years) and Their Standard Errors for ∑PBDE, BDE-47, BDE-99, BDE-209, DBDPE, HBB, PBEB, and BTBPE Concentrations in the Vapor, Particle, And Precipitation Phases, And in the Three Phases Combined compound
a
vapor
particles
precipitation
combined phases
∑PBDE
9.9 ( 4.2
6.6 ( 1.8
4.7 ( 2.4
6.3 ( 1.1
BDE-47 BDE-99
6.3 (1.8 9.8 ( 4.6
6.9 ( 2.2 8.8 ( 3.7
4.8 ( 2.2 1.9 ( 0.3
6.1 ( 0.3 5.1 ( 0.9
BDE-209
NSa
NS
7.2 ( 3.4
NS
DBDPE
NS
NS
NS
NS
HBB
5.9 ( 3.0
3.2 ( 0.6
2.2 ( 0.5
9.5 ( 4.6
PBEB
7.7 ( 3.5
4.3 ( 1.2
5.5 ( 2.5
NS
BTBPE
NS
13 ( 6.5
3.8 ( 1.0
9.8 ( 2.8
NS: not significant at P < 0.05.
In this equation, ε0 is the partial residual, and a0, a1, a2, a3, and a4 are the coefficients from eq 1. This step was applied to each chemical in each phase separately; the results for all three phases were combined together; and a new time-dependent linear regression was fitted to these partial residuals ε0 ¼ a00 þ a05 t 0
ð3Þ 0
where a5 is an overall first-order rate constant, and a0 is an intercept. This process yields rate constants that can be used to calculate the halving times of each chemical in each phase separately and in all three phases combined. The additional step of combining the data for all three phases together allows us to calculate a weighted average of the halving times using the number of data points in each phase as weights. Table 2 (top) presents the halving times and their standard errors for each phase and for the three phases combined using this approach. The combined phase regressions are shown in Figure 2. The regressions for ∑PBDE, BDE-47, and BDE-99 concentrations are all highly significant (P < 0.0001). Overall, ∑PBDE concentrations are declining in the atmosphere of the Great Lakes with halving times of ∼6 years. These results are generally slower than halving times reported previously; Venier and Hites reported declining trends for ∑PBDE atmospheric concentrations (sum of vapor plus particle phase concentrations) at each IADN site with halving times of ∼4 years.18 The overall halving time of ∑PBDE concentrations is shorter than that of ∑PCBs (∼17 years), ∑PAHs (∼10 years), and ∑DDTs (∼9 years) reported recently for the same sites around the Great Lakes and using the same harmonic regression approach used here.14 This may indicate that the 2004 production restrictions of penta- and octa-PBDEs are having an effect and that the concentrations of PBDEs in the atmosphere are declining faster than those of other persistent organic pollutants. The concentrations of BDE-47 and -99 are also declining with halving times of ∼6 years. Overall, with the exception of the low value of ∼2 years for BDE-99 in precipitation, the average halving times for these individual congeners are the same as for ∑PBDE concentrations. The values reported here for BDE-47 and -99 are somewhat higher than the halving times reported for these compounds in a previous study from our laboratory (∼13 years).18 The previous study used IADN PBDE data from only 20052006, which was right after production of the penta-BDE mixture was discontinued in 2004. Significantly faster halving times 8701
dx.doi.org/10.1021/es2020378 |Environ. Sci. Technol. 2011, 45, 8698–8706
Environmental Science & Technology
Figure 2. Regression of partial residuals versus sampling date for the three phases combined (vapor, red squares; particles, green triangles; and precipitation, yellow circles) for (A) ∑PBDEs (B) BDE-47 (C) BDE-99, and (D) BDE-209.
were reported by Venier and Hites for chlordane concentrations in precipitation in comparison to chlordane’s halving times in vapor and particle phases,14 which is similar to our observation for BDE99. We are not able to explain this finding. The regressions for the BDE-209 concentrations versus time were generally not significant, suggesting that the concentrations of this congener in the environment are not yet decreasing. Perhaps when the restrictions on the production and use of this compound go into effect after 2013, these atmospheric levels will start to decrease. DBDPE, HBB, PBEB, and BTBPE. Figure 3 shows the concentrations of these compounds as box-plots in the vapor, particle, and precipitation phases at the five U.S. IADN sites. Table 3 gives the arithmetic mean and standard errors of these concentrations along with ANOVA results (applied to log-transformed concentrations) and percent detected. DBDPE has been produced and used as an additive flame retardant for more than 20 years.29 Due to its structural resemblance to BDE-209, DBDPE seems to have the same applications as the deca-BDE commercial product and has recently become an alternative to this restricted formulation.30 DBDPE has been detected in indoor29,31 and outdoor air,18,20 precipitation,20 sewage sludge and sediments,29,32 tree bark,20,33,34 household dust,31,35 and biota.13,36,37 A recent international survey on DBDPE in sludge samples found DBDPE in samples from each of the 12 countries included in the survey.38 Clearly, DBDPE is a worldwide pollutant.
ARTICLE
In this study, DBDPE was detected in all three phases, but not as often as most of the other BFRs; see Table 3. The average percent detection for DBDPE was 6% in the vapor phase, 23% in the particle phase, and 61% in precipitation samples. Overall, the highest DBDPE concentrations were detected at the urban Chicago and Cleveland sites. These concentrations were not statistically different among the three nonurban sites in any phase; see Table 3. DBDPE concentrations are in general agreement with those reported in previous studies from our laboratory,18,20 but they are low relative to those of Kierkegaard et al.,29 who reported much higher concentrations in ambient air near an electronics dismantling facility in Stockholm, Sweden (700 pg/m3). SI Figure S3 shows the percentage of DBDPE concentrations relative to those of ∑PBDE at the five U.S. IADN sites in the vapor, particle, and precipitation phases. Interestingly, the ratio of DBDPE concentrations to ∑PBDE concentrations was the highest at the three rural and remote sites in the particle phase with the ratios of 0.75 at Sturgeon Point, 0.37 at Eagle Harbor, and 0.35 at Sleeping Bear Dunes. HBB and PBEB are also flame retardants, but not much is known about their production, use, or environmental fate. In the United States, both of these flame retardants were produced and used in 1970s and 1980s; however, there seems to be no current information on the U.S. production of these compounds. The last list from the EPA Inventory Update Report indicates HBB and PBEB production amounts in 1998 and 1986, respectively; in both cases, the amounts produced were 5250 tons.39 HBB seems to be produced in Japan and China, and PBEB seems to be produced by Albemarle Chemical Corporation in France.30 HBB was used as an additive flame retardant in wood, textiles, electronics, and plastics; PBEB was used in polyurethane foam, wire and cable coatings, and polyester resins.30 PBEB is included in the OSPAR List of Chemicals for Priority Action as a persistent, potentially bioaccumulative, and toxic chemical.40 HBB was first “spotted” in the environment several decades ago in sediment and human samples from Japan, which was probably a result of Japan’s heavy use of HBB at that time.41,42 Similarly, PBEB was identified in Canadian lake trout caught in 1979.43 More recent studies on HBB and PBEB are scarce and scattered. These compounds were found in air samples from the United States,44 Canada,45 the United Kingdom,46 and the Atlantic and Southern oceans;8 sediment and sludge samples from Norway;9 biota samples from the Great Lakes and Arctic;10,13,47 and human blood samples from China.48 In this study, we observed surprising trends in the spatial distributions of both HBB and PBEB concentrations; see Table 3. Both of these compounds were detected in all three phases ∼70% of the time, but their concentrations were generally higher in the vapor phase. Unlike ∑PBDE concentrations, the highest average HBB concentrations (6.4 ( 0.9 pg/m3) were observed at the rural Sturgeon Point site in the vapor phase. The HBB levels were similar to those measured in air samples from Drammen, Norway9 and in the atmosphere over the Atlantic Ocean (0.0411 pg/m3).8 The concentration of HBB at the Sturgeon Point site is ∼610 times higher than the HBB concentrations in Cleveland and Chicago. The average vapor HBB concentration at this site was 80% of that of the ∑PBDE concentration in the vapor phase; see SI Figure S3. This finding is remarkable given that there are no known HBB sources near Sturgeon Point. It has been suggested that possible HBB sources can include release from polymeric flame retardants; however, the estimated release of HBB from these sources was shown to be insufficient to explain 8702
dx.doi.org/10.1021/es2020378 |Environ. Sci. Technol. 2011, 45, 8698–8706
Environmental Science & Technology
ARTICLE
Figure 3. Concentrations of DBDPE, HBB, PBEB, and BTBPE in the atmospheric vapor phase, particle phase, and precipitation; the thin black lines represent the median, and the thick red lines represent the arithmetic mean; the boxes represent the 25th and 75th percentiles; the whiskers represent the 5th and 95th percentiles. Site abbreviations: EH Eagle Harbor, CH Chicago, SB Sleeping Bear Dunes, CL Cleveland, SP Sturgeon Point.
Table 3. Arithmetic Mean ( Standard Error of DBDPE, HBB, PBEB, and BTBPE Concentrations in Vapor Phase (pg/m3), Particle Phase (pg/m3), and Precipitation (ng/L) at the Five U.S. IADN Sitesa Eagle Harbor conc.
*
Chicago
% det
conc.
*
Sleeping Bear Dunes % det
conc.
*
Cleveland
% det
conc.
DBDPE
vapor
0.02 ( 0.01
c
4
4.7 ( 2.3
a
11
0.7 ( 0.2
ab
10
0.10 ( 0.04
HBB
vapor
0.2 ( 0.1
d
40
0.7 ( 0.1
b
75
0.2 ( 0.1
c
67
1.1 ( 0.2
*
Sturgeon Point % det
conc.
*
% det
abc
4
0.02 ( 0.01
bc
2
b
68
6.4 ( 0.9
a
81
PBEB
vapor
1.2 ( 0.2
c
75
0.8 ( 0.1
b
94
0.05 ( 0.01
e
66
1.5 ( 0.1
a
92
0.11 ( 0.01
d
87
BTBPE
vapor
0.2 ( 0.1
c
14
0.8 ( 0.2
a
60
0.3 ( 0.1
bc
40
0.5 ( 0.1
ab
13
0.2 ( 0.02
bc
43
DBDPE
particle
1.2 ( 0.5
b
7
3.6 ( 0.8
b
34
3.2 ( 1.0
b
13
14 ( 7
a
47
2.9 ( 1.1
b
14
HBB
particle
0.2 ( 0.1
c
55
0.4 ( 0.1
a
68
0.2 ( 0.1
c
57
0.3 ( 0.1
a
67
0.2 ( 0.04
b
62
PBEB
particle
0.7 ( 0.1
a
39
0.1 ( 0.02
c
75
0.1 ( 0.03
d
21
0.14 ( 0.03
b
88
0.1 ( 0.03
d
21
BTBPE
particle
0.2 ( 0.03
c
29
1.0 ( 0.2
a
91
0.9 ( 0.3
b
83
0.9 ( 0.1
a
93
0.5 ( 0.1
c
82
DBDPE
precip
0.3 ( 0.1
ab
31
0.8 ( 0.2
a
74
0.4 ( 0.1
ab
63
0.6 ( 0.2
a
71
0.5 ( 0.2
b
65
HBB PBEB
precip precip
0.3 ( 0.1 0.03 ( 0.01
a a
85 79
0.2 ( 0.04 0.03 ( 0.01
ab bc
72 67
0.6 ( 0.2 0.01 ( 0.01
a bc
93 62
0.1 ( 0.02 0.01 ( 0.01
b b
80 87
0.3 ( 0.1 0.01 ( 0.01
ab c
78 74
BTBPE
precip
0.01 ( 0.01
b
51
0.1 ( 0.02
a
94
0.04 ( 0.01
b
80
0.1 ( 0.01
a
89
0.05 ( 0.01
b
76
a
ANOVA results for the log-transformed concentrations are given in the columns headed by an asterisk; the concentrations are not significantly different for those locations sharing the same letter. Percent of detects for each compound in each phase is also shown.
prevailing HBB concentrations in air.45 Other possible sources of HBB include degradation of higher brominated PBDE congeners
or DBDPE in the environment or during e-waste recycling or burning.9,49 The concentrations of HBB in precipitation were 8703
dx.doi.org/10.1021/es2020378 |Environ. Sci. Technol. 2011, 45, 8698–8706
Environmental Science & Technology similar at all five sites, and it was detected in ∼80% of the samples. In the particle phase, the highest HBB concentrations in particle phase were detected at Chicago and Cleveland. The mean concentrations of PBEB in the vapor phase were highest at Cleveland and lowest at Sturgeon Point and Sleeping Bear Dunes. Surprisingly, PBEB concentrations in particles and precipitation were highest at Eagle Harbor—our most remote site. In fact, the PBEB concentrations at Eagle Harbor were ∼45% of those of ∑PBDE (see SI Figure S3). The generally low levels of PBEB in precipitation suggest that this compound is not scavenged by precipitation as effectively as PBDEs. The PBEB levels we observe here are much lower than the single day PBEB measurement in Chicago air (520 pg/m3 in the vapor phase and 29 pg/m3 in particles) reported previously;44 however, these concentrations in Great Lakes air are higher than those measured in air samples in Drammen, Norway, and Egbert, Canada.9,45 With the exception of the one sample in Chicago,44 the vapor and particle phase PBEB concentrations at Eagle Harbor reported in this study are the highest atmospheric concentrations yet reported for this compound. Considering that Eagle Harbor is a very remote site and few people live or work there, this observation is hard to explain. BTBPE (in some papers abbreviated as TBE) is an additive flame retardant produced since the 1970s by Great Lakes Chemical (now a part of Chemtura) and used as a replacement for the discontinued octa-BDE mixture.44 It was a high production volume chemical in the United States with an estimated production of 5005000 tons/year in 1998,44 but its estimated worldwide usage was ∼20 tons in 2000.30 In our samples, BTBPE was present in all three phases, but it was primarily detected in the particle phase and in precipitation. The highest levels of BTBPE were measured at Chicago and Cleveland in all three phases. The levels at the other three sites were, in general, not statistically different from one another in any phase. The average BTBPE concentration at Sleeping Bear Dunes was ∼20% of that of ∑PBDE concentration at this site in the particle phase (see SI Figure S3). The levels of BTBPE measured in this study were similar to those reported in a previous study from our laboratory (0.51.2 pg/m3)18 and to levels measured in Chicago (4.0 pg/m3),44 Bloomington, Indiana (2.8 pg/m3),44 and at a suburban area close to Stockholm (<3 pg/m3).50 DBDPE, HBB, PBEB, and BTBPE Temporal Trends. Temporal trends for these alternative flame retardants were analyzed using the same approach as used for ∑PBDE temporal trend analysis (see eq 1) except the distance term was omitted because there was no known source for these chemicals. This regression was applied to DBDPE, HBB, PBEB, and BTBPE concentrations individually. The coefficients and their standard errors from the regression for these compounds, as well as the dates of maximum concentration and halving times are reported in SI Table S2. The subsequent steps in the temporal trend analysis, including calculation of the partial residuals, were the same as described above for PBDE temporal trends. Table 2 (bottom) summarizes these results, and the combined phase regressions are shown in Figure 4. The regression coefficients for HBB and BTBPE are all significant at P < 0.05, but those for DBDPE and PBEB are not. The temporal trends for these flame retardants are not as clear as for PBDEs. The regression of DBDPE partial residuals versus time was not significant in any phase (see Table 2). This suggests that there is no change over time in these concentrations, and there may be a continuing source of this compound in the environment. HBB concentrations in the vapor and particle
ARTICLE
Figure 4. Regression of partial residuals versus sampling date for the three phases combined (vapor, red squares; particles, green triangles; and precipitation, yellow circles) for (A) DBDPE, (B) HBB, (C) PBEB, and (D) BTBPE.
phases are declining with halving times of 36 years, but these concentrations are increasing in precipitation with a doubling time of ∼2 years. For all three phases combined, atmospheric HBB concentrations are decreasing with a halving time of ∼10 years. PBEB concentrations in the vapor phase and precipitation are also increasing with doubling times of 68 years; however, these concentrations are decreasing in the particle phase with a halving time of ∼4 years. This finding suggests that the mechanism of PBEB removal in the particle phase might be different from that in the vapor phase and in precipitation. The regression of PBEB partial residuals versus time was not statistically significant for the three phases combined, suggesting that the increasing concentrations in the vapor phase and in precipitation are balanced by decreasing concentrations in the particle phase. The regression of BTBPE vapor concentrations vs time was not statistically significant, probably because BTBPE is mostly present in the particle phase and in precipitation. In these phases, BTBPE concentrations are decreasing with halving times of 413 years. Overall for the three phases combined, BTBPE concentrations are decreasing with a halving time of ∼10 years. The overall regressions (of the three phases combined) were statistically significant for only two of the four alternative flame retardants analyzed here, and these regressions suggest that the concentrations of these two compounds (HBB and BTBPE) are declining in the atmosphere with halving times of ∼10 years. 8704
dx.doi.org/10.1021/es2020378 |Environ. Sci. Technol. 2011, 45, 8698–8706
Environmental Science & Technology This value is somewhat higher than the halving time values found for ∑PBDEs and for individual PBDE congeners in this study, but it is similar to halving times reported elsewhere for ∑PCBs, ∑PAHs, and ∑DDTs.14
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional experimental details (including a map of the sampling sites), tables of regression coefficients (and their standard errors) for all analytes in all the phases, PBDE congener distributions, percentage ratio of alternative flame retardant concentrations to ∑PBDE concentrations, and details on the seasonality, population, and distance regression terms are included. This material is free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT We thank Ilora Basu and Team IADN for the operation of the network and Marta Venier for helpful discussions. This work was supported by the Great Lakes National Program Office of the U.S. Environmental Protection Agency (Grant No. GL00E76601, Todd Nettesheim, project manager). ’ REFERENCES (1) State of California Department of Consumer Affairs, Bureau of Electronic and Appliance Repair, Home Furnishings and Thermal Insulation. Technical Bulletin 117. Requirements, test procedure and apparatus for testing the flame retardancy of resilient filling materials used in upholstered furniture. March 2000. http://www.bhfti.ca.gov/ industry/bulletin.shtml (accessed June 12, 2011). (2) Alaee, M.; Arias, P.; Sj€odin, A.; Bergman, Å. An overview of commercially used brominated flame retardants, their applications, their use patterns in different countries/regions and possible modes of release. Environ. Int. 2003, 29, 683–689. (3) de Wit, C. A. An overview of brominated flame retardants in the environment. Chemosphere 2002, 46, 583–624. (4) Hites, R. A. Polybrominated diphenyl ethers in the environment and in people: A meta-analysis of concentrations. Environ. Sci. Technol. 2004, 38, 945–956. (5) Directive 2003/11//EC of the European Parliament and of the Council of 6 February 2003 amending for the 24th time Council directive 76/769/EEC relating to restrictions on the marketing and use of certain dangerous substances and preparations (pentabromodiphenyl ether and octabromodiphenyl ether). Official Journal L 042, 15/02/2003. (6) European Court of Justice. Cases C-14/06 and C-295/06, Judgment of the Court, 1 April 2008, Directive 200/95/EC and Commission Decision 2005/717/EC; 2008. http://curia.europa.eu (accessed September 16, 2011). (7) U.S. Environmental Protection Agency. Deca-BDE Phase-out Initiative. http://www.epa.gov/oppt/existingchemicals/pubs/actionplans/ deccadbe.html (accessed June 12, 2011). (8) Xie, Z.; M€oller, A.; Ahrens, L.; Sturm, R.; Ebinghaus, R. Brominated flame retardants in seawater and atmosphere of the Atlantic Ocean and the Southern Ocean. Environ. Sci. Technol. 2011, 45, 1820–1826. (9) Arp, H. P. H.; Møskeland, T.; Andersson, P. L.; Nyholm, J. R. Presence and partitioning properties of the flame retardants pentabromotoluene, pentabromoethylbenzene and hexabromobenzene near suspected source zones in Norway. J. Environ. Monit. 2011, 13, 505–513.
ARTICLE
(10) de Wit, C. A.; Herzke, D.; Vorkamp, K. Brominated flame retardants in the Arctic environment trends and new candidates. Sci. Total Environ. 2010, 408, 2885–2918. (11) Wu, J. P; Guan, Y. T; Zhang, Y.; Luo, X. J.; Zhi, H.; Chen, S. J.; Mai, B. X. Trophodynamics of hexabromocyclododecanes and several other non-PBDE brominated flame retardants in a freshwater food web. Environ. Sci. Technol. 2010, 44, 5490–5495. (12) Nyholm, J. R.; Asamoah, R. K.; van der Wal, L.; Danielsson, C.; Andersson, P. L. Accumulation of polybrominated diphenyl ethers, hexabromobenzene, and 1,2-dibromo-4-(1,2-dibromoethyl)-cyclohexane in earthworm (Eisenia fetida). Effects of soil type and aging. Environ. Sci. Technol. 2010, 44, 9189–9194. (13) Gauthier, L. T.; Potter, D.; Hebert, C. E.; Letcher, R. J. Temporal trends and spatial distribution of non-polybrominated diphenyl ether flame retardants in the eggs of colonial populations of Great Lakes herring gulls. Environ. Sci. Technol. 2009, 43, 312–317. (14) Venier, M.; Hites, R. A. Time trend analysis of atmospheric POPs concentrations in the Great Lakes Region Since 1990. Environ. Sci. Technol. 2010, 44, 8050–8055. (15) Basu, I.; Bays, J. C. Collection of Air and Precipitation Samples: IADN Project Standard Operating Procedure; Indiana University: Bloomington, IN, 2005; http://www.indiana.edu/∼hiteslab/FieldSOP2005.pdf (16) Carlson, D. L.; Basu, I.; Hites, R. A. Annual variation of pesticide concentrations in Great Lakes precipitation. Environ. Sci. Technol. 2004, 38, 5290–5296. (17) Basu, I. Analysis of PCBs, Pesticides, PAHs and PBDEs in Air and Precipitation Samples; IADN Project Sample Preparation Procedure; Indiana University: Bloomington, IN, 2007; http://www.indiana.edu/ ∼hiteslab/sampleprep07.pdf. (18) Venier, M; Hites, R. A. Flame retardants in the atmosphere near the Great Lakes. Environ. Sci. Technol. 2008, 42, 4745–4751. (19) Strandberg, B.; Dodder, N. G.; Basu, I.; Hites, R. A. Concentrations and spatial variations of polybrominated diphenyl ethers and other organohalogen compounds in Great Lakes air. Environ. Sci. Technol. 2001, 35, 1078–1083. (20) Salamova, A.; Hites, R. A. Evaluation of tree bark as a passive atmospheric sampler for flame retardants, PCBs, and organochlorine pesticides. Environ. Sci. Technol. 2010, 44, 6196–6201. (21) Venier, M.; Hites, R. A. Atmospheric deposition of PBDEs to the Great Lakes featuring a Monte Carlo analysis of errors. Environ. Sci. Technol. 2008, 42, 9058–9064. (22) Cetin, B.; Odabasi, M. Particle-phase dry deposition and air soil gas exchange of polybrominated diphenyl ethers (PBDEs) in Izmir, Turkey. Environ. Sci. Technol. 2007, 41, 4986–4992. (23) St-Amand, A. D.; Mayer, P. M.; Blais, J. M. Seasonal trends in vegetation and atmospheric concentrations of PAHs and PBDEs near a sanitary landfill. Atmos. Environ. 2008, 42, 2948–2958. (24) Gevao, B.; Jaward, F. M.; Macleod, M.; Jones, K. C. Diurnal fluctuations in polybrominated diphenyl ether concentrations during and after a severe dust storm episode in Kuwait City, Kuwait. Environ. Sci. Technol. 2010, 44, 8114–8120. (25) Zhang, B.; Guan, Y.; Li, S.; Zeng, E. Y. Occurrence of polybrominated diphenyl ethers in air and precipitation of the Pearl River delta, South China: Annual washout ratios and depositional rates. Environ. Sci. Technol. 2009, 43, 9142–9147. (26) Ter Schure, A. F. H.; Larsson, P.; Agrell, C.; Boon, J. P. Atmospheric transport of polybrominated diphenyl ethers and polychlorinated biphenyls to the Baltic sea. Environ. Sci. Technol. 2004, 38, 1282–1287. (27) Ter Schure, A. F. H.; Larsson, P. Polybrominated diphenyl ethers in precipitation in Southern Sweden (Skane, Lund). Atmos. Environ. 2002, 36, 4015–4022. (28) Ter Schure, A. F. H.; Agrell, C.; Bokenstrand, A.; Sveder, J.; Larsson, P.; Zegers, B. N. Polybrominated diphenyl ethers at a solid waste incineration plant II: Atmospheric deposition. Atmos. Environ. 2004, 38, 5149–5155. 8705
dx.doi.org/10.1021/es2020378 |Environ. Sci. Technol. 2011, 45, 8698–8706
Environmental Science & Technology (29) Kierkegaard, A.; Bj€orklund, J.; Friden, U. Identification of the flame retardant decabromodiphenyl ethane in the environment. Environ. Sci. Technol. 2004, 38, 3247–3253. (30) Covaci, A.; Harrad, S.; Abdallah, M. A. E.; Ali, N.; Law, R. J.; Herzke, D.; de Wit, C. A. Novel brominated flame retardants: A review of their analysis, environmental fate and behavior. Environ. Intern. 2011, 37, 532–556. (31) Karlsson, M.; Julander, A.; van Bavel, B.; Hardell, L. Levels of brominated flame retardants in blood in relation to levels in household air and dust. Environ. Intern. 2007, 33, 62–69. (32) Ricklund, N.; Kierkegaard, A.; McLachlan, M. S. Levels and potential sources of decabromodiphenyl ethane and decabromodiphenyl ether in lake and marine sediments in Sweden. Environ. Sci. Technol. 2010, 44, 1987–1991. (33) Zhu, L.; Hites, R. A. Brominated flame retardants in tree bark from North America. Environ. Sci. Technol. 2006, 40, 3711–3716. (34) Qiu, X.; Hites, R. A. Dechlorane Plus and other flame retardants in tree bark from the Northeastern United States. Environ. Sci. Technol. 2008, 42, 31–36. (35) Stapleton, H. M.; Allen, J. G.; Kelly, S. M.; Konstantinov, A.; Klosterhaus, S.; Watkins, D.; McClean, M. D.; Webster, T. F. Alternate and new brominated flame retardants detected in U.S. house dust. Environ. Sci. Technol. 2008, 42, 6910–6916. (36) Hu, G. C.; Luo, X. J.; Dai, J. Y.; Zhang, X. L.; Wu, H.; Zhang, C. L.; Guo, W.; Xu, M. Q.; Mai, B. X.; Wei, F. W. Brominated flame retardants, polychlorinated biphenyls, and organochlorine pesticides in captive giant panda (Ailuropoda melanoleuca) and red panda (Ailurus fulgens) from China. Environ. Sci. Technol. 2008, 42, 4704–4709. (37) Luo, X.; Zhang, X.; Liu, J.; Wu, J.; Luo, Y.; Chen, S.; Mai, B.; Yang, Z. Persistent halogenated compounds in water-birds from an e-waste recycling region in South China. Environ. Sci. Technol. 2009, 43, 306–311. (38) Ricklund, N.; Kierkegaard, A.; McLachlan, M. S. An international survey of decabromodiphenyl ethane and decabromodiphenyl ether in sewage sludge samples. Chemosphere 2008, 73, 1799–1804. (39) United States Environmental Protection Agency. Inventory Update Reporting. http://www.epa.gov/iur/tools/data/2002-vol.html (accessed June 11, 2011). (40) OSPAR Convention for the protection of the marine environment of the North-East Atlantic. OSPAR List of chemicals for priority action (revised in 2009). http://www.ospar.org/content/content.asp? menu=00940304440000_000000_000000 (accessed June 11, 2011). (41) Watanabe, I.; Kashimoto, T.; Tatsukawa, R. Hexabromobenzene and its debrominated compounds in river and estuary sediments in Japan. Bull. Environ. Contam. Toxicol. 1986, 36, 778–784. (42) Yamaguchi, Y.; Kawano, M.; Tatsukawa, R.; Moriwaki, S. Hexabromobenzene and its debrominated compounds in human adipose tissues of Japan. Chemosphere 1988, 17, 703–707. (43) Ismail, N.; Gewurtz, S. B.; Pleskach, K.; Whittle, D. M.; Helm, P. A.; Marvin, C. H.; Tomy, G. T. Brominated and chlorinated flame retardants in Lake Ontario, Canada in lake trout (Salvelinus namaycush) between 1979 and 2004 and possible influences of food-web changes. Environ. Toxicol. Chem. 2009, 28, 910–920. (44) Hoh., E.; Zhu, L.; Hites, R. A. Novel flame retardants, 1,2bis(2,3,6-tribromophenoxy)ethane and 2,3,4,5,6-pentabromoethylbenzene, in United States’ environmental samples. Environ. Sci. Technol. 2005, 39, 2472–2477. (45) Gouteux, B.; Alaee, M.; Mabury, S. A.; Pacepavicius, G.; Muir, D. C. G. Polymeric brominated flame retardants: Are they a relevant source of emerging brominated aromatic compounds in the environment? Environ. Sci. Technol. 2008, 42, 9039–9044. (46) Lee, R. G. M.; Thomas, G. O.; Jones, K. C. Atmospheric concentrations of PBDEs in western Europe. Organohalogen Compd. 2002, 58, 193–196. (47) Gauthier, L. T.; Hebert, C. E.; Weseloh, D. V. C.; Letcher, R. J. Current-use flame retardants in the eggs of herring gulls (Larus argentatus) from the Laurentian Great Lakes. Environ. Sci. Technol. 2007, 41, 4561–4567.
ARTICLE
(48) Zhu, L.; Ma, B.; Hites, R. A. Brominated flame retardants in serum from the general population in Northern China. Environ. Sci. Technol. 2009, 43, 6963–6968. (49) Buser, H. R. Polybrominated dibenzofurans and dibenzo-p-dioxins— Thermal reaction products of polybrominated diphenyl ether flame retardants. Environ. Sci. Technol. 1986, 20, 404–408. (50) Sj€odin, A.; Carlsson, H.; Thuresson, K.; Sjolin, S.; Bergman, A.; Ostman, C. Flame retardants in indoor air at an electronics recycling plant and at other work environments. Environ. Sci. Technol. 2001, 35, 448–454.
8706
dx.doi.org/10.1021/es2020378 |Environ. Sci. Technol. 2011, 45, 8698–8706
ARTICLE pubs.acs.org/est
Enrichment of Humic Material with Hydroxybenzene Moieties Intensifies Its Physiological Effects on the Nematode Caenorhabditis elegans Ralph Menzel,*,† Stefanie Menzel,†,‡ Sophie Tiedt,† Georg Kubsch,§ Reinhardt St€oßer,§ Hanno B€ahrs,† Anke Putschew,|| Nadine Saul,† and Christian E. W. Steinberg† †
Department of Biology, Freshwater and Stress Ecology, Humboldt-Universit€at zu Berlin, Sp€athstr. 80/81, 12437 Berlin, Germany Institute of Biochemistry, Freie Universit€at Berlin, Thielallee 63, 14195 Berlin, Germany § Department of Chemistry, Structural Analysis & Environmental Chemistry, Humboldt-Universit€at zu Berlin, Brook-Taylor-Str. 2, 12489 Berlin, Germany Department of Water Quality Control, Technical University Berlin, Str. des 17. Juni 135, 10623 Berlin, Germany
)
‡
bS Supporting Information ABSTRACT: Dissolved humic substances are taken up by organisms and interact on various molecular and biochemical levels. In the nematode Caenorhabditis elegans, such material can promote longevity and increase its reproductive capacity; moreover, the worms tend to stay for longer in humic-enriched environments. Here, we tested the hypothesis that the chemical enrichment of humic substances with hydroxybenzene moieties intensifies these physiological effects. Based on the leonardite humic acid HuminFeed (HF), we followed a polycondensation reaction in which this natural humic substance and a dihydroxybenzene (hydroquinone or benzoquinone) served as reaction partners. Several analytical methods showed the formation of the corresponding copolymers. The chemical modification boosted the antioxidant properties of HF both in vitro and in vivo. Humic substances enriched with hydroxybenzene moieties caused a significantly increased tolerance to thermal stress in C. elegans and extended its lifespan. Exposed nematodes showed delayed linear growth and onset of reproduction and a stronger pumping activity of the pharynx. Thus, treated nematodes act younger than they really are. In this feature the modified HF replicated the biological impact of hydroquinone-homopolymers and various plant polyphenol monomers, thereby supporting the hydroxybenzene moieties of humic substances as major effective structures for the physiological effects observed in C. elegans.
’ INTRODUCTION Humic substances (HS) are complex organic molecules that represent the largest constituent of natural organic matter (NOM) in the soil (∼60%);1 they also make up most (50 80%) of the dissolved NOM in freshwater ecosystems.2 4 HS possess a variety of molecular structures and are considered as key components of both soil and freshwater ecosystems. Humic substances are able to bind a variety of organic and inorganic compounds due to their hydrophobic and mainly aromatic structural features, including metal ions via S-, N-, and carboxyl groups; thereby, they modulate the bioavailability of chemical compounds. 5 9 Besides such indirect effects, there is increasing evidence that humic substances can also be directly effective. Thus, humic substances are taken up by organisms, as has often been demonstrated,10 14 and once internalized they can positively influence the metabolic and signaling pathways involved in plant development by directly acting on specific physiological targets such as root formation and nitrogen metabolism.15 17 Humic substances were also found to act as r 2011 American Chemical Society
a mild source of natural chemical stress and as ‘natural biocides’ in relation to a fish-pathogenic bacterium,18 an aquatic macrophyte,19 a fungus,20 and different species of algae;21 hormone-like effects of humic substances have even been described in fish 22 and amphibians.23 In this study, we used the bacteriovorous nematode Caenorhabditis elegans (Maupas, 1900) as the model organism, which belongs to the most abundant class of metazoan organisms in soils and sediments with a high biodiversity. H€oss et al.24 and Steinberg et al. 25 described the impact of different humic substance isolates on the reproductive capacity of this nematode. While the majority of humic substances tested were found to stimulate reproduction in the nematode, a fulvic acid fraction Received: July 8, 2011 Accepted: September 9, 2011 Revised: September 7, 2011 Published: September 09, 2011 8707
dx.doi.org/10.1021/es2023237 | Environ. Sci. Technol. 2011, 45, 8707–8715
Environmental Science & Technology from a raised peat bog lake decreased reproduction. In a more recent paper, it was demonstrated that C. elegans is gradually attracted by different humic substances;26 moreover, a processed leonardite (HuminFeed, HF) significantly prolonged its lifespan, whereas a polysaccharide-rich NOM isolate of Lake Schwarzer See did not.27 This difference was attributed to the corresponding chemical compositions of the humic substances applied. However, substantial progress in the research on humic substances is still hampered by the lack of a detailed characterization of the material used. Several authors tried to approach this problem with case studies relating the available chemical parameters of various humic substances to biological effects. These data served as the basis of a statistical evaluation to identify potentially effective structures in the sources tested. Overall, the biocide-like impact of specific humic substances, both on the aquatic macrophyte Ceratophyllum demersum19 and on the fish-pathogenic fungus Saprolegnia parasitica,20 could be attributed to humic substances with a higher molecular weight and aromaticity, which probably contain a higher number of stable semiquinone radicals, which may influence the redox homeostasis of an organism. In particular, phenolic and quinonoid moieties have been hypothesized as being highly active.19,20 Indeed, humic substances are redox-active compounds. Evidence has been accumulated to show that they, and particularly their hydroxybenzene moieties, play an important role as electron shuttles in, for example, microbial redox reactions involved in metal reduction and the biodegradation of priority pollutants.28 31 Whether or not these phenolic and quinonoid moieties are also responsible for the rather beneficial impact of humic substances on, for example, the increased reproduction and lifespan of C. elegans is unknown and comprises the central hypothesis of the present study. In this context, it is particularly noticeable that several polyphenolic substances, such as tannic and gallic acid, which are expected to form the building blocks of humic substances,32 are able to significantly extend the lifespan of C. elegans.33,34 In this work, we intended to enrich a selected humic substance with hydroxybenzene moieties via chemical modification rather than perform another correlation analysis using an extensive number of different humic sources. We followed a copolycondensation reaction35 in which natural humic substances and a dihydroxybenzene (hydroquinone or benzoquinone) served as reaction partners in a formaldehyde condensation reaction. Because coalderived humics, mostly consisting of an aromatic core,36 have been considered to be the best candidates for this type of modification,35 we selected HF. Moreover, we added a synthetic, humic-like substance that was synthesized by autoxidation of hydroquinone at a high pH. This humic-like substance consists of aliphatically linked phenolic rings with functional hydroxybenzene groups. The aim of this study was to investigate whether chemically modified or synthetically produced humic substances are able to alter basic physiological parameters in terms of lifespan and thermotolerance, reproductive capacity, pharyngeal pumping activity, and linear growth. Moreover, the property of humic substances as antioxidant to quench free radicals by hydrogen donation and their potential to attract the worms were evaluated.
’ EXPERIMENTAL SECTION Nematodes and Cultivation. The C. elegans wild-type strain Bristol N2 was used throughout this study. The worms were
ARTICLE
maintained on NGM agar37 plates using Escherichia coli OP50 as a food source, according to standard procedures. 38,39 The specified humic substance was added to both the agar and bacterial suspension at concentrations of 0, 0.2, 0.4, and 2.0 mM dissolved organic carbon (DOC). The incubation temperature was always 20 °C. A synchronous culture was generated by rinsing worms from NGM plates and filtering them through a 10 μm membrane (SM 16510/11, Sartorius AG, G€ottingen, Germany), a pore size that retains all but the first-stage juveniles (L1). Humic Substances. Origin. HuminFeed (Humintech GmbH, D€usseldorf, Germany) was made by an alkaline extraction process of highly oxidized lignite. Prior to extraction, the raw lignite material was milled and homogenized. The humate solution was separated from the humin via sedimentation of the latter. Detailed physicochemical analyses in comparison to several other humic sources are available.20 HuminFeed was used for practical reasons and by no means as an advertisement for this product. Chemical Modification. First, formaldehyde polycondensation was conducted between HF and hydroquinone under the conditions taken from Perminova et al.35 HuminFeed (1 g) was dissolved in 2.5 mL of distilled water and further diluted in water to a volume of 50 mL; the pH was adjusted to pH 7. The addition of 500 mg of the monomer hydroquinone (Merck, Darmstadt, Germany) was performed in the presence of catalytic amounts of oxalic acid and 1 g of a 35% solution of formaldehyde. The mixture was stirred for 1 h at 100 °C. The dissolved humic substance was precipitated by the addition of 6 M HCl at pH 1, filtrated, and freeze-dried. This product was called HuminfeedHydroquinone (HF-HQ). For the second modification, 1 g of HF was dissolved in 2.5 mL of distilled water and diluted to 50 mL with 0.1 M KH2PO4 (pH 7). Then, 500 mg of benzoquinone was added, and the mixture was stirred for 1 h at room temperature. Subsequently, the humic substance was precipitated with 6 M HCl at pH 1, filtrated, and freeze-dried. In line with the modification performed, the product was called HF-BQ. Elemental Analysis. Carbon, hydrogen, nitrogen, and sulfur were determined after the combustion of dried samples by a thermal conductivity detector. The metal contents were analyzed by an atomic absorption spectrometer (Perkin-Elmer, MA, USA) after digestion with nitric acid in a high pressure microwave oven (MLS GmbH, Leutkirch im Allg€au, Germany). Synthesis of a Hydroquinone-Homopolymer. Hydroquinone (5 g) was dissolved in 200 mL of 0.2 M Na2CO3 solution and vented with air at 30 L/h for 5 h. The dissolved polymerized hydroquinone was precipitated by the addition of 6 M HCl at pH 1, filtrated, and freeze-dried. This product was called synthetic hydroquinone-homopolymer (sHQ). Size-Exclusion Chromatography (SEC). The size distribution of DOC was determined using size-exclusion chromatography with UV254 detection and online carbon detection. The SEC was carried out using an LC-OCD system (DOC Labor Dr Huber, Karlsruhe, Germany). A Toyopearl HW-50S column was used (250 20 mm, particle size 20 40 μm, Tosoh Bioscience, Tokyo, Japan) for the separation. The compounds were eluted with a 20 mM phosphate buffer (pH 6.5) at a flow rate of 1.5 mL/min. After separation and detection of the UV absorption the mobile phase passed through a Graentzel thin film reactor, where the organic compounds were oxidized after the addition of concentrated phosphoric acid (0.4 mL/min) and irradiation with UV light at 184 nm. The carbon dioxide produced was 8708
dx.doi.org/10.1021/es2023237 |Environ. Sci. Technol. 2011, 45, 8707–8715
Environmental Science & Technology transferred into an infrared detector with nitrogen 5.0 (20 L/h) as carrier gas. For more details, the reader is referred to Sachse et al.40,41 Furthermore, the specific absorbance of the humic substances was determined. The slope of the absorption curves, as measured by the ratios of absorbance at 465 and 665 nm (E4/ E 6 ratio), has traditionally been suggested to be inversely related to the degree of condensation of aromatic groups (aromaticity).42 Electron Paramagnetic Resonance (EPR)-Spectroscopy. The EPR-spectroscopy provides information on spin concentrations in the material, which are indicative of stable free radicals, which, in turn, can react with many chemical structures, including biotic macromolecules. The Cw X-band EPR-spectra of freeze-dried samples were recorded at room temperature using a spectrometer of the type ERS300 (Centre of Construction of Scientific Devices/Magnettech GmbH, Berlin, Germany) at a microwave power (PMW) of 2 mW, as described in Paul et al.43 Spin concentrations (used to determine the number of unpaired spins per gram of carbon) were determined using strong pitch as a reference. Nuclear Magnetic Resonance (NMR)-Spectroscopy. The 1H MAS NMR spectra were recorded on a Bruker AVANCE 400 spectrometer (Larmor frequency: ν1H = 400.1 MHz) using a 4 mm MAS probe with rotors made from ZrO2. The spectra were recorded with a π/2 pulse duration of p1 = 2.5 μs, a recycle delay of 5 s, and a spinning frequency of 10 kHz. Values of the isotropic chemical shifts are given with respect to tetramethylsilane (TMS). The existing background signals of 1H were suppressed by the application of a phase-cycled depth pulse sequence according to Cory et al.44 Lifespan Assay. This assay accurately followed the protocol published recently.45 In order to start the test, freshly grown L4 larvae (P0) were placed onto either control or treatment plates. Then, L4 larvae of the next generation (F1) were transferred to assay plates with 0, 0.2, 0.4, and 2.0 mM DOC of each humic substance tested. The first day of adulthood was defined as day 1. We performed two independent trials, each comprising ten small agar plates (Ø = 35 mm) with overall 150 nematodes per trial. Alive and dead animals were scored daily until all worms had died. Thermotolerance Test. The F1 generation nematodes were cultured as described above starting from L4 on either control or humic substance-containing plates. During the reproductive period, adult nematodes were transferred daily to fresh treatment plates. On the sixth day of adulthood the agar plates were switched to 35 °C for 7 h; subsequently, the dead and alive nematodes were counted. This thermal resistance test was performed with 100 nematodes for each concentration in three independent trials. Reproduction Assay. Again, L4 larvae of the F1 generation were placed onto individual treatment plates (n = 12 per concentration, two independent trials). Nematodes were transferred daily to fresh treatment plates until reproduction ended, typically day 4 of adulthood. The offspring of each individual nematode was counted once grown to the L2 or L3 stage. Growth Alterations. For the growth assay, 20 nematodes of the F1 generation per concentration and trial (n = 4) were utilized. On the first and sixth day of adulthood, respectively, the nematodes were killed by heat exposure, and the length of each individual nematode was measured using a microscope equipped with a graduated eyepiece. Pharyngeal Pumping Frequency. On the third, sixth, and ninth day of adulthood the pharyngeal pumping frequency was
ARTICLE
quantified. Nematodes of the different treatments were randomly selected (n g 10), and the pumping frequency was determined three times over a 15 s time span. Three independent experimental trials were performed for each assay. Attraction Assay. The NGM agar plates were prepared according to our previously published protocol,26 with six alternating spots of bacteria, lacking or supplemented with humic substances. Total numbers of offspring present after 96 h on the three control spots and the three spots supplemented with humic substances were summed and compared as percentages. Four plates were used for each concentration, and all tests were comprised of three independent experimental trials. Antioxidant Capacity. The superoxide anion radical scavenging activities (SOSA) of both pure humic substances and homogenates of C. elegans were analyzed via photoluminescence for the sum of hydrophilic antioxidants (=ascorbic acid equivalents) in a PHOTOCHEM device and the corresponding ACW-Kit (AnalytikJena, Jena, Germany) according to Popov and Lewin.46 This assay measures the ability of an antioxidant to quench free, photochemically generated superoxide anion radicals by hydrogen donation. The sensitive luminometric detection uses luminol as detector substance. Before measurement, pure humic substances were dissolved in distilled water to a final concentration of 500 μg/mL DOC and neutralized with NaOH to pH = 7 (to provide conditions similar to those used for the biological tests). For worm exposure, about 5000 synchronized L1 larvae were treated with humic substances for three days until they reached the young adult stage. The animals were rinsed from the agar plates with ice-cold M9 buffer, washed three times, and homogenized in a Speedmill (AnalytikJena, Germany). Finally, the resulting homogenate was centrifuged at 2000 rpm for 2 min at 4 °C and immediately used for measurement. The antioxidant capacities in the worms were related to their protein content, which was measured according to Bradford,47 using Bradford reagent (Sigma-Aldrich, MO, USA). All measurements were performed in triplicate. Statistical Analysis. Median and mean lifespan and percentage changes (compared to controls) were determined. The statistical significance of alterations in the mean lifespan was calculated using a log-rank test (Bioinformatics at the Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia; http://bioinf.wehi.edu.au/software/russell/logrank). Mean survival rates and percentage changes to the control were calculated for the thermal stress resistance assays. Statistical significance was defined via the chi-square test (SigmaStat 3.5; SPSS Inc., Chicago, IL) and one-way ANOVA (SigmaStat 3.5; SPSS Inc., Chicago, IL) for the daily and total reproductive output, body length, the attraction assay, the pharyngeal pumping frequency, and the antioxidant capacity.
’ RESULTS AND DISCUSSION The objectives of this study were to synthesize and characterize humic substances enriched with hydroxybenzene moieties and to estimate their biological impact on the nematode C. elegans. It was expected that the introduction of additional dihydroxybenzene groups into the structure of the humics would be reflected in changes to the physical and chemical characteristics, for example, spectral and redox properties and, hence, would lead to an enhancement of their beneficial biological properties in exposed nematodes, such as lifespan extension and increased stress resistance. 8709
dx.doi.org/10.1021/es2023237 |Environ. Sci. Technol. 2011, 45, 8707–8715
Environmental Science & Technology
ARTICLE
Figure 1. The superoxide anion radical scavenging activities (SOSA). (A) Mean total of the antioxidant capacity of four different humic substances, measured in vitro as pure substances and compared to gallic acid and tannic acid (n = 3). Please note the logarithmic scale of the Y-axis. (B) SOSA of water-soluble substances in homogenates obtained from humic substance-treated and untreated wild-type nematodes. The tests were repeated three times with a total of approximately 5000 nematodes per treatment in each single trial. The data are shown as equivalents of nM ascorbic acid and are related to mM dissolved organic carbon (A) or the total protein content (B).
Structural Studies of the Chemically Modified Humic Substances. In their pioneering work, Perminova et al.35 sup-
plied substantial evidence confirming the success of the intended hydroxybenzene-enrichment of leonardite humic substance by using monomeric dihydroxybenzenes; our chemical analyses of modified HF material further validated this protocol. The data are provided in the Supporting Information. A first investigation of the elemental composition and atomic ratios showed a principal similarity between the different data sets (Table S1, Supporting Information). However, the lower H/C atomic ratios of 0.88 (HF-HQ) and 0.79 (HF-BQ) in comparison to 0.95 for the starting material (HF) were the first indicator of the higher aromaticity of the modified humic substances. This was accompanied by a decrease in the E4/E6 ratio, which is inversely related to the degree of aromaticity.42 For HF-HQ and HF-BQ, the E4/E6 ratios were about 20% lower than for the original HF (Table S1, Supporting Information). Further evidence for the successful modification was provided by the increase in spins per gram of carbon, as measured by EPR (Table 1, Supporting Information). The highest spin concentrations, indicative of the content of stable free radicals, were found in HF-BQ and HFHQ. The corresponding single lines gave g-values of g = 2.0031 (or 2.0033 for HF-BQ), which were very similar for all of the samples. They can be attributed to organic radicals and not, for example, to metal ions, which give quite different values.48 The measured g-values (i.e., g∼2.003) are known for humic acids and indicate that the organic radicals were most likely part of the aromatic subsystems, whereas higher g-values (i.e., g close to 2.004) are typical for semiquinone-type radicals.49 The increase in the g-value in the case of HF-BQ provides a further indication of the enrichment of original HF with hydroxybenzene moieties. The results of the SEC (Table S1, Figure S1A, Supporting Information) were very similar for HF and both modifications and comprised the typical main peak which is referred to humic substances as classified by DOC fingerprint analysis of various aquatic samples containing humic substances41 (typically between 3500 and 4000 g/mol). Beside that main signal compounds of lower molecular weight were detected (RT < 60 min, typically 1500 2000 g/mol) and, in case of HF and HF-HQ, higher molecular weight compounds (RT < 45 min, typically 6000 7000 g/mol). Because the higher molecular weight compounds were UV-active at 254 nm (data not shown), we assumed that it did not comprise polysaccharides, which are not detectable by UV.
Our fourth sample, sHQ, a polycondensed hydroquinonehomopolymer, produced a qualitatively similar elution pattern, yet the peaks appeared after prolonged retention times (Figure S1A, Supporting Information). Here, the condensation was done by a different mechanism, and the SEC shows that the polymerization of HQ leads to polymers with a smaller molecular size as found after the reaction of HF with HQ, indicating that the polymerization of HF with HQ was indeed successful. This shift in retention time was exactly reproduced using another polycondensed hydroquinone-homopolymer, HS1500 (Figure S1A, Supporting Information), which was commercially produced by the Sopar Pharma GmbH (Mannheim, Germany). The assumption of a higher aromaticity of the modified humic substances, as discussed above, was supported by the two 1H MAS NMR spectra shown in Figure S1B (Supporting Information). Although both modified (HF-HQ) and unmodified (HF) humic substances gave 1H signals at the same position (Table S1, Supporting Information), their intensities were different. The modified HF-HQ had an unambiguous higher intensity in the region of ring protons or double bonds (7 ppm) (Figure S1B, Supporting Information). Together, all of the data provided the conclusion that the polycondensation of HF and the selected dihydroxybenzenes (HQ and BQ) led to the formation of corresponding copolymers and not simply mixtures of humic substances and hydroquinonehomopolymers. Chemical Modification Boosts the Antioxidant Capacity of HF. Humic substances are known to interfere in redox and radical processes, and their antioxidant capacity can be used to evaluate these properties.50 As a final step in the characterization process, we measured the SOSA of HF before and after the chemical modifications. The relative results are presented and correspond to equivalents of ascorbic acid (Figure 1A). Whereas the original HF only possesses a few nM equivalents per mM DOC, enrichment with dihydroxybenzenes increased this capacity by more than 8-fold (HF-BQ) or even 17-fold in the case of HF-HQ. Hence, the chemically modified HFs showed an antioxidant capacity similar to the hydroquinone-homopolymer sHQ. The quenching capacities even approached those we found for polyphenols such as gallic and tannic acid, which were measured as positive controls (Figure 1A). Both substances were selected because the same and related compounds, in particular those resulting from the enzymatic degradation of lignin, 8710
dx.doi.org/10.1021/es2023237 |Environ. Sci. Technol. 2011, 45, 8707–8715
Environmental Science & Technology
ARTICLE
Figure 2. Survival curves of C. elegans in the absence or presence of humic substances, which were added at three different concentrations: 0.2 mM (A), 0.4 mM (B), and 2.0 mM DOC (C). Day 1 refers to the first day of adulthood. The statistical significance of alterations in the mean lifespan was calculated using a log-rank test, P < 0.001 (**). Please note that all individual experimental trials of HS exposure, performed independently, produce statistically significant different results in comparison to the control. (D) On the sixth day of adulthood, the number of dead and alive animals was counted following a 7 h exposure to a thermal stress of 35 °C. The columns present the mean values of three independent experimental trials, in each case comprising 5 20 nematodes. The error bars represent the standard error of the mean (SEM), and differences compared to the control were considered significant at P < 0.05 (*) and P < 0.001 (**).
significantly contribute to the antioxidant capacity of humus.51 The strong rise in SOSA as a result of the chemical modification provided further evidence that both HF-HQ and HF-BQ were successfully enriched with hydroxybenzene moieties. In the next step we wanted to evaluate the capacity of modified humic substances to modulate the redox homeostasis of whole organisms (= the potential to balance reactive oxygen species, ROS). In the lower exposure concentrations (0.2 and 0.4 mM DOC), only 0.2 mM HF-BQ-treated nematodes revealed the presence of significantly enhanced antioxidant capacity (Figure 1B). The SOSA of the nematodes treated with the original HF was found to be slightly reduced in comparison to the control (Figure 1B). Only the highest exposure concentration (2 mM DOC) increased the antioxidant capacity in the organisms. The SOSA values of sHQ and both chemically modified HF humics were significantly increased not only in comparison to the control but also to the original HF-derived samples (P between 0.033 and 0.004). Similar results were obtained for several polyphenol monomers, for example, quercetin and caffeic acid52 as well as tannic and ellagic acid,53 which are all well-known to possess substantial antioxidant properties. Why did the in vivo situation not produce an adequate concentration-effect relationship and why did SOSA even decrease in response to the HF treatment (0.2 and 0.4 mM DOC)? As known redox shuttles, humic substances may also exhibit pro-oxidant characteristics depending on the individual redox status of the humic molecule, the inorganic chemistry of the cell, and the quality of nutrients available. These effects interfere with the intrinsic antioxidant capacity of humic substances and can exceed its impact. This may explain several findings published recently, showing for example a decrease in total antioxidant capacity of the blood of Japanese quails treated with humic acid (600 mg/kg food),54 and
higher amounts of oxidatively damaged DNA indicated by an increase in 8-oxodGuo levels in both human primary fibroblasts55 and maternal mice.56 Moreover, leaf litter leachates have been shown to function as pro-oxidants, as determined in an aquatic macrophyte.57 Treatment with Humic Substances Extend the Lifespan in Wild-Type Nematodes. The lifespan of untreated (control) and humic substance-treated wild-type nematodes was compared. Pure HF treatment resulted in a significantly enhanced lifespan at all three concentrations tested (Figure 2A-C), whereas the lower concentrations (0.4 and 0.2 mM DOC, see Figure 2B and A) proved to be the most effective. Almost the same results were obtained for the HF-BQ treatment (Figure 2A-C). In the case of HF-HQ and the synthetic polycondensed sHQ, each concentration tested increased the mean lifespan in comparison to the other humic substances tested. With more than 22% and 17%, the highest gains in mean lifespan were achieved using 0.4 mM DOC of sHQ and HF-HQ, respectively. Both underlying data sets were significantly different in comparison to the original HF data (Log-rank test; P = 0.0054 and 0.0066, respectively). Moreover, by reaching a lifespan extension of about 20%, only these two modifications reached the degree of impact of polyphenol-monomers, for example quercetin45 and tannic acid,34 which extended the lifespan of C. elegans by 18% and 18 20%, respectively. Nematodes Exposed to Modified Humic Substances Exhibit Enhanced Resistance to Thermal Stress. Because longevity has frequently been found to be accompanied by enhanced thermal tolerance, both in long-lived mutant strains58,59 and in response to polyphenols,34,45,52,60 we preferred in this work the thermal stress resistance test over tests following an exposure to reactive oxygen species or oxidizing reagents. The latter assays 8711
dx.doi.org/10.1021/es2023237 |Environ. Sci. Technol. 2011, 45, 8707–8715
Environmental Science & Technology
ARTICLE
Figure 3. The impact of humic substances on reproduction and growth in length. Total (A) and daily (B) reproductive output was determined from 12 animals per substance and concentration in two independent trials. Day 1 refers to the first day of adulthood. The length of treated and untreated nematodes was measured on both the first (C) and the sixth day (D) of adulthood. In each case, 20 nematodes were used for growth measurements (n = 4). Each bar/point illustrates the average of all associated trials. The error bars represent the standard error of the mean (SEM), and differences compared to the control were considered significant at P < 0.05 (*) and P < 0.001 (**).
often produced contradictory results, as shown for tannic acid,34 rosmarinic acid, and caffeic acid52 as well as an extract of blueberry polyphenols,60 and were not able to enlighten a key effect as longevity is. In order to analyze how humic substances affect the response to thermal stress, nematode survival was monitored following a 7 h exposure to 35 °C. Again, worms treated with humic substances achieved better results than the control animals (Figure 2D). However, only treatment with the chemically modified HFs and sHQ resulted in a significantly enhanced stress tolerance (Figure 2D). Also, HF-HQ and sHQ again caused the highest impact on the nematodes by increasing thermal stress resistance by up to 30% (0.2 mM DOC). So far, humic substances have only been proven to be an inducer of enhanced stress resistance in M. macrocopa, where indigenous humics isolated from coastal lagoons promoted resistance to salt stress.61 Polyphenol monomers, on the other hand, were repeatedly characterized as efficient inducers of stress resistance, in response to both thermal and oxidative stimuli (for C. elegans, see refs 34,45,51, and 52). The finding that only chemically modified humic substances and the hydroquinone-homopolymer sHQ (in the used concentrations) were able to replicate the impact of polyphenol monomers provides a further piece of evidence to show the biological impact of the hydroxybenzene moieties of humics. Exposure to 0.2 mM DOC of HS Slightly Increases Reproduction; Higher Concentrations Reduce Growth in Length. A significant extension of an organism’s lifespan may take place at the expense of reproductive capacity and/or growth.62,63 The lowest concentration of 0.2 mM DOC consistently increased the reproductive capacity of C. elegans for all humic substances tested (Figure 3A). A closer look at the distribution of brood size over the four reproductive days revealed a significant
reduction in the initial reproduction capacity on the second day for the chemically modified HFs and sHQ compared to the control. This reduction, however, was overcompensated on day 3 (HF-BQ and sHQ) and/or day 4 (HF-HQ and sHQ) (Figure 3B). The original HF did not produce this specific effect. When polyphenol monomers were again taken as the reference, tannic, gallic, and ellagic acid were proven to delay reproduction in C. elegans, but they did not modulate the total reproductive capacity.52 However, a set of different humic substances was found to stimulate reproduction in the nematode;24,25 measurements of a potential delay in the onset of reproduction were not performed in previous studies. The size of treated and untreated nematodes was determined on the first (Figure 3C) and sixth (Figure 3D) day of adulthood. On both days, exposure to the highest concentration of humic substances (2.0 mM DOC) resulted in a significant reduction in body length. When the concentration of humics was at 0.4 mM DOC, this effect persisted with HF-HQ, HF-BQ, and sHQ but only on the first day of adulthood. Again, the initial delays were compensated for over time. The increasing content of hydroxybenzene moieties, as mediated by an increased total concentration of humics or by a specific enhancement via chemical modification, seemed to be the trigger. These data again match the findings for the tannic acid52 and catechin polyphenols.64 Indications for an Impact on Eating Behavior. Previous work26 revealed that C. elegans possesses an intrinsic prevalence for humic substances; here, we support this finding. The nematodes were attracted to bacteria spiked with HF-substances or sHQ as indicated by the number of eggs that were laid on the humic-treated bacterial spots (Figure 4A). Differences between the specific treatments became apparent when we compared the concentration-effect relationships. Whereas the preference for 8712
dx.doi.org/10.1021/es2023237 |Environ. Sci. Technol. 2011, 45, 8707–8715
Environmental Science & Technology
Figure 4. Humic substances modulate eating behavior. (A) Attraction assay results: the percentage distribution of F1 offspring on agar plates with six alternating spots of bacteria, lacking or supplemented with humic substances, are shown. Open bars refer to the offspring present on the control spots, the filled bars correspond to offspring present on the spots supplemented with humic substances. (B) The pumping activity of 3 10 (three independent trials) treated and untreated nematodes was monitored on the third, sixth, and ninth day of adulthood (from left to right). Each individual nematode was quantified three times for 15 s. The error bars represent the standard error of the mean (SEM), and differences compared to the control were considered significant at P < 0.05 (*) and P < 0.001 (**).
ARTICLE
pumping frequency on all days, whereas the lower concentration (0.2 mM DOC) as well as the higher concentration (2.0 mM DOC) significantly increased pharynx activity only on the sixth and ninth day of adulthood (Figure 4B). Obviously, the older the nematodes are, the more they can profit from treatment with humic substances. Chemical intervention to modify the original HF did not further enforce the activating property of HF on the pumping frequency of C. elegans. Several polyphenols also share this property with humic substances. Quercetin, rosmarinic, and caffeic acid45 as well as gallic acid and catechin,52 were found to mediate an increase in pharyngeal activity. As with the humics, the relative gain in activity upon polyphenolic exposure peaked in each case on the ninth day of adulthood. It is well-known that the pumping frequency of C. elegans reaches its maximum in the young adult stage, followed by a continuous decrease until death.65 Taking the observed delays in reproduction and growth in length into account, the higher pharyngeal activity reflected a phenotype where the nematodes seemed to look younger than they really were. The enhancement of HF with hydroxybenzene moieties intensified this effect, also seen by the increased thermotolerance and the extended lifespan. In conclusion, chemical modification of humic substances from a leonardite source, in order to enhance its hydroxybenzene moieties, resulted in a gain of beneficial properties. Almost all of the variables measured responded to the modifications and the observed values resembled the findings obtained for the hydroquinone-homopolymer sHQ and polyphenol monomers. The major effect seems to be a delay in aging. Our data support that the hydroxybenzene moieties of humic substances were mainly responsible for the physiological effects observed in the nematode C. elegans.
’ ASSOCIATED CONTENT
bS
Supporting Information. Please find here chemical parameters (Table S1) as well as DOC-fingerprint analyses and 1H MAS NMR spectra (Figure S1) of the humic substances. This material is available free of charge via the Internet at http:// pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
the original HF showed its highest measured value at 2.0 mM DOC, the concentration dropped to 0.4 mM in the case of HFHQ and HF-BQ. Furthermore, sHQ attracted most of the nematodes at the lowest concentration of 0.2 mM DOC, whereas 2.0 mM DOC was not effective. With the polyphenol monomers, the preferential behavior of C. elegans was heterogeneous. Whereas tannic and gallic acid were able to attract the nematodes,52 quercetin, rosmarinic, and caffeic acid did not,45 and ellagic acid even repelled C. elegans.52 Our data suggest that the chemical modifications of HF (or the polymerization of hydroquinone to sHQ) enhanced the fraction of moieties in the HS mediating the attracting properties; a process which led to a decline of the most efficient DOC concentration. The observed preference for humic substances at certain concentrations was also reflected by an increase in pharynx activity, as determined by counting the pumping frequency on the third, sixth, and ninth day of adulthood (Figure 4B). All humic substances tested significantly increased the pumping frequency. The concentration of 0.4 mM DOC increased the
*Phone: +4930 63224241. Fax: +4930 6369446. E-mail: ralph.menzel@ biologie.hu-berlin.de.
’ ACKNOWLEDGMENT This work was supported by the grant STE 673/16-1 awarded by the ‘Deutsche Forschungsgemeinschaft’ (DFG). We are grateful to Dr. Gudrun Scholz (HU Berlin, Institute for Chemistry) for excellent support and generation of parts of the NMR figure. ’ REFERENCES (1) Stevenson, F. J. Organic forms of soil nitrogen. In Humus Chemistry: Genesis, Composition, Reactions; Stevenson, F. J., Ed.; John Wiley & Sons: New York, 1994; pp 59 95. (2) Thurman, E. M. Aquatic Humic Substances. Organic Geochemistry of Natural Waters; Nijhoff-Junk: Dordrecht, 1985. (3) Steinberg, C. E. W.; M€unster, U. Geochemistry and ecological role of humic substances in lakewater. In Humic Substances in Soil, 8713
dx.doi.org/10.1021/es2023237 |Environ. Sci. Technol. 2011, 45, 8707–8715
Environmental Science & Technology Sediment and Water. Geochemistry, Isolation and Characterization; Aiken, G. R., McKnight, D. M., Wershaw, R. L., MacCarthy, P., Eds.; John Wiley & Sons: New York, 1985. (4) Wetzel, R. G. Limnology. Lake and River Ecosystems; Academic Press: San Diego, 2001. (5) Giesy, J. P.; Leversee, G. J.; Williams, D. R. Effects of natural occurring aquatic organic fractions on cadmium toxicity to Simocephalus serrulatus (Daphnidae) and Gambusia affinis (Poeciliidae). Water Res. 1977, 11, 1013–1020. (6) Campbell, J. H.; Evans, R. D. Inorganic and organic ligand binding of lead and cadmium and resultant implications for bioavailability. Sci. Total Environ. 1987, 62, 219–224. (7) Haitzer, M.; H€oss, S.; Traunspurger, W.; Steinberg, C. E. W. Effects of dissolved organic matter (DOM) on the bioconcentration of organic chemicals in aquatic organisms A review. Chemosphere 1998, 37, 1335–1362. (8) Haitzer, M.; Abbt-Braun, G.; Traunspurger, W.; Steinberg, C. E. W. Effects of humic substances of polycyclic aromatic hydrocarbons: Correlations with spectroscopic and chemical properties of humic substances. Environ. Toxicol. Chem. 1999, 18, 2782–2788. (9) Perminova, I. V.; Yashchenko, N. Yu.; Petrosyan, V. S. Relationship between structure and binding affinity of humic substances for polyaromatic hydrocarbons: Relevance of molecular descriptors. Environ. Sci. Technol. 1999, 33, 3781–3787. (10) Ziechmann, W. Huminstoffe und ihre Wirkungen; Spektrum Akademischer Verlag: Heidelberg. 1996. (11) Wang, W. H.; Bray, C. M.; Jones, M. N. The fate of 14C-labelled humic substances in rice cells in cultures. J. Plant Physiol. 1999, 154, 203–211. (12) Nardi, S.; Pizzeghello, D.; Muscolo, A.; Vianello, A. Physiological effects of humic substances on higher plants. Soil Biol. Biochem. 2002, 34, 1527–1536. (13) Steinberg, C. E. W. Ecology of Humic Substances in Freshwaters —Determinants from Geochemistry to Ecological Niches; Springer: Berlin, 2003. (14) Kulikova, N. A.; Perminova, I. V.; Badun, G. A.; Chernysheva, M. G.; Koroleva, O. V.; Tsvetkova, E. A. Estimation of uptake of humic substances from different sources by Escherichia coli cells under optimum and salt stress conditions by use of tritium-labeled humic materials. Appl. Environ. Microbiol. 2010, 76, 6223–6230. (15) Quaggiotti, S.; Ruperti, B.; Pizzeghello, D.; Francioso, O.; Tugnoli, V.; Nardi, S. Effect of low molecular size humic substances on nitrate uptake and expression of genes involved in nitrate transport in maize (Zea mays L.). J. Exp. Bot. 2004, 55, 803–813. (16) Trevisan, S.; Pizzeghello, D.; Ruperti, B.; Francioso, O.; Sassi, A.; Palme, K.; Quaggiotti, S.; Nardi, S. Humic substances induce lateral root formation and expression of the early auxin-responsive IAA19 gene and DR5 synthetic element in Arabidopsis. Plant Biol. 2010, 12, 604–614. (17) Trevisan, S.; Francioso, O.; Quaggiotti, S.; Nardi, S. Humic substances biological activity at the plant-soil interface: From environmental aspects to molecular factors. Plant Signal Behav. 2010, 5, 635–643. (18) Liltved, H.; Landfald, B. Effects of high intensity light on ultravioletirradiated and nonirradiated fish pathogenic bacteria. Water Res. 2000, 34, 481–486. (19) Pflugmacher, S.; Pietsch, C.; Rieger, W.; Steinberg, C. E. W. Influence of structural moieties of dissolved organic matter on the photosynthetic oxygen production of aquatic plants. Sci. Total Environ. 2006, 357, 169–175. (20) Meinelt, T.; Paul, A.; Phan, T. M.; Zwirnmann, E.; Kr€uger, A.; Wienke, A.; Steinberg, C. E. W. Reduction in vegetative growth of the water mold Saprolegnia parasitica (Coker) by humic substances of different qualities. Aquat. Toxicol. 2007, 83, 93–103. (21) Prokhotskaya, L. Yu.; Steinberg, C. E. W. Differential sensitivity of a coccal green alga and a cyanobacterial species to dissolved natural organic matter. Environ. Sci. Pollut. Res. 2007, 14, 11–18. (22) Meinelt, T.; Schreckenbach, K.; Knopf, K.; Wienke, A.; St€uber, A.; Steinberg, C. E. W. Humic substances increase the constitution of swordtail (Xiphophorus helleri). Aquat. Sci. 2004, 66, 239–245.
ARTICLE
(23) Lutz, I.; Zhang, J.; Opitz, R.; Kloas, W.; Xu, Y.; Menzel, R.; Steinberg, C. E. W. Synthetic humic substances act as weak xenoestrogens and affect the thyroid system of Xenopus laevis. Chemosphere 2005, 61, 1183–1188. (24) H€oss, S.; Bergtold, M.; Haitzer, M.; Traunspurger, W.; Steinberg, C. E. W. Refractory dissolved organic matter can influence the reproduction of Caenorhabditis elegans (Nematoda). Freshwater Biol. 2001, 46, 1–10. (25) Steinberg, C. E. W.; H€oss, H.; Br€uggemann, R. Further evidence that humic substances have the potential to modulate the fertility of the nematode Caenorhabditis elegans. Intern. Rev. Hydrobiol. 2002, 87, 121–133. (26) Menzel, R.; St€urzenbaum, S.; B€arenwaldt, A.; Kulas, J; Steinberg, C. E. W. Humic material induces behavioral and global transcriptional responses in the nematode Caenorhabditis elegans. Environ. Sci. Technol. 2005, 39, 8324–8332. (27) Steinberg, C. E. W.; Saul, N.; Pietsch, K.; Meinelt, T.; Rienau, S.; Menzel, R. Dissolved humic substances facilitate fish life in extreme aquatic environments and have the potential to extend lifespan of Caenorhabditis elegans. Ann. Environ. Sci. 2007, 1, 81–90. (28) Aeschbacher, M.; Sander, M.; Schwarzenbach, R. P. Novel electrochemical approach to assess the redox properties of humic substances. Environ. Sci. Technol. 2010, 44, 87–93. (29) Maurer, F.; Christl, I.; Kretzschmar, R. Reduction and reoxidation of humic acid: Influence on spectroscopic properties and proton binding. Environ. Sci. Technol. 2010, 44, 5787–5792. (30) Gu, B.; Chen, J. Enhanced microbial reduction of Cr(VI) and U(VI) by different natural organic matter fractions. Geochim. Cosmochim. Acta 2003, 67, 3575–3582. (31) Nurmi, J. T.; Tratnyek, P. G. Elecrochemical properties of natural organic matter (NOM), fractions of NOM, and model biogeochemical electron shuttles. Environ. Sci. Technol. 2002, 36, 617–624. (32) Leenheer, J. A. Systematic Approaches to Comprehensive Analyses of Natural Organic Matter. Ann. Environ. Sci. 2009, 3, 1–130. (33) Pietsch, K.; Hofmann, S.; Henkel, R.; Saul, N.; Menzel, R.; Steinberg, C. E. W. The plant polyphenol caffeic acid affects life traits differently in the nematode Caenorhabditis elegans and the cladoreran Moina macrocopa. Fresenius Environ. Bull. 2010, 19, 1238–1244. (34) Saul, N.; Pietsch, K.; Menzel, R.; St€urzenbaum, S. R.; Steinberg, C. E. W. The longevity effect of tannic acid in Caenorhabditis elegans: Disposable soma meets hormesis. J. Gerontol. Biol. Sci. 2010, 65, 626–635. (35) Perminova, I. V.; Kovalenko, A. N.; Schmitt-Kopplin, P.; Hatfield, K.; Hertkorn, N.; Belyaeva, E. Y.; Petrosyan, V. S. Design of quinonoid-enriched humic materials with enhanced redox properties. Environ. Sci. Technol. 2005, 39, 8518–8524. (36) Lawson, G. J.; Stewart, D. Coal humic acids. In Humic Substances II. In Search of Structure; Hayes, M. H. B., MacCarthy, P., Malcolm, R. L., Swift, R. S., Eds.; John Wiley & Sons: Chichester, New York, Brisbane, Toronto, Singapore, 1989; p 642. (37) Brenner, S. The genetics of Caenorhabditis elegans. Genetics 1974, 77, 71–94. (38) Sulston, J. E.; Hodgkin, J. Methods. In The Nematode Caenorhabditis elegans; Wood, W. B., Ed.; CSHL Press: Cold Spring Harbor, 1988. (39) Lewis, J. A.; Fleming, J. T. Basic culture methods. In Caenorhabditis elegans: Modern Biological Analysis of an Organism; Epstein, H. F., Shakes, D. C., Eds.; Academic Press: San Diego, 1995. (40) Sachse, A.; Babenzien, D.; Ginzel, G.; Gelbrecht, J.; Steinberg, C. E. W. Characterization of dissolved organic carbon (DOC) of a dystrophic lake and an adjacent fen. Biogeochemistry 2001, 54, 279–296. (41) Sachse, A.; Henrion, R.; Gelbrecht, J.; Steinberg, C. E. W. Classification of dissolved organic carbon (DOC) in rivers systems: influence of catchment characteristics and autochthonous processes. Org. Geochem. 2005, 36, 923–935. (42) Senesi, N.; Miano, T. M.; Provenzano, M. R.; Brunetti, G. Spectroscopic and compositional comparative characterization of I.H.S. S. reference and standard fulvic and humic acids of various origin. Sci. Total Environ. 1989, 81/82, 143–156. 8714
dx.doi.org/10.1021/es2023237 |Environ. Sci. Technol. 2011, 45, 8707–8715
Environmental Science & Technology (43) Paul, A.; St€osser, R.; Zehl, A.; Zwirnmann, E.; Vogt, R.; Steinberg, C. E. W. Nature and abundance of organic radicals in natural organic matter: effect of sample pH and irradiation. Environ. Sci. Technol. 2006, 40, 5897–5903. (44) Cory, D. G.; Ritchey, W. M. Suppression of signals from the probe in Bloch decay spectra. J. Magn. Reson. 1988, 80, 128–132. (45) Pietsch, K.; Saul, N.; Menzel, R.; St€urzenbaum, S. R.; Steinberg, C. E. W. Quercetin mediated lifespan extension in Caenorhabditis elegans is modulated by age-1, daf-2, sek-1 and unc-43. Biogerontology 2009, 10, 565–578. (46) Popov, I.; Lewin, G. Antioxidative homeostasis: Characterization by means of chemiluminescent technique. Methods Enzymol. 1999, 300, 437–456. (47) Bradford, M. M. A rapid and sensitive method for quantification of microgram quantities of protein utilizing principle of protein dye binding. Ann. Biochem. 1976, 72, 248–254. (48) Senesi, N.; Steelink, C. Application of EPR spectroscopy to the study of humic substances. In: Humic substances; Hayes, M. H. B., MacCarty, P., Malcolm, R. L., Swift, R. S., Eds.; John Wiley & Sons: Chichester, New York, Brisbane, Toronto, Singapore, 1989; pp 374 408. (49) Jezierski, A.; Czechowski, F.; Jerzykiewicz, M.; Drozd, J. EPR investigationof structure of humic acids from compost, soil, peat and soft brown coal upon oxidation and metal uptake. Appl. Magn. Reson. 2000, 18, 127–136. (50) Rimmer, D. L.; Smith, A. M. Antioxidants in soil organic matter and in associated plant materials. Eur. J. Soil Sci. 2009, 60, 170–175. (51) Flaig, W.; Beutelspacher, H.; Rietz, E. Chemical composition and physical properties of humic substances. In Soil components; Gielseking, J. E., Ed.; Springer: New York, 1975; Vol. 1, pp 1 211. (52) Pietsch, K.; Saul, N.; St€urzenbaum, S. R.; Menzel, R.; Steinberg, C. E. W. Hormetins, antioxidants and prooxidants: defining quercetincaffeic acid- and rosmarinic acid- mediated life extension in C. elegans. Biogerontology 2011, 12, 329–347. (53) Saul, N.; Pietsch, K.; Menzel, R.; St€urzenbaum, S. R.; Steinberg, C. E. W. The diversity of polyphenol action in Caenorhabditis elegans: Between toxicity and longevity. J. Nat. Prod. 2011, 74, 1713–1720. (54) Ipek, H.; Avci, M.; Iriadam, M.; Kaplan, O.; Denek, N. Effects of humic acid on some hematological parameters, total antioxidant capacity and laying performance in Japanese quails. Eur. Poultry Sci. 2008, 72, 56–60. (55) Cheng, M. L.; Ho, H. Y.; Huang, Y. W.; Lu, F. J.; Chiu, D. T. Humic acid induces oxidative DNA damage, growth retardation, and apoptosis in human primary fibroblasts. Exp. Biol. Med. 2003, 228, 413–423. (56) Hu, C. W.; Yen, C. C.; Huang, Y. L.; Pan, C. H.; Lu, F. J.; Chao, M. R. Oxidatively damaged DNA induced by humic acid and arsenic in maternal and neonatal mice. Chemosphere 2010, 79, 93–99. (57) Nimptsch, J.; Pflugmacher, S. Decomposing leaf litter: the effect of allochthonous degradation products on the antioxidant fitness and photosynthesis of Vesicularia dubyana. Ecotoxicol. Environ. Saf. 2008, 69, 541–545. (58) Lithgow, G. J.; White, T. M.; Melov, S.; Johnson, T. E. Thermotolerance and extended life-span conferred by single-gene mutations and induced by thermal stress. Proc. Natl. Acad. Sci. U. S. A. 1995, 92, 7540–7544. (59) Munoz, M. J.; Riddle, D. L. Positive selection of Caenorhabditis elegans mutants with increased stress resistance and longevity. Genetics 2003, 163, 171–180. (60) Wilson, M. A.; Shukitt-Hale, B.; Kalt, W.; Ingram, D. K.; Joseph, J. A.; Wolkow, C. A. Blueberry polyphenols increase lifespan and thermotolerance in Caenorhabditis elegans. Aging Cell 2006, 5, 59–68. (61) Suhett, A. L.; Steinberg, C. E. W.; Santangelo, J. M.; Bozelli, R. L.; Farjalla, V. L. Natural dissolved humic substances increase lifespan and resistance to salt stress in the cladoceran Moina macrocopa. Environ. Sci. Pollut. Res. 2011, 18, 1004–1014. (62) Kirkwood, T. B. Evolution of ageing. Nature 1977, 270, 301–304.
ARTICLE
(63) Kirkwood, T. B. The nature and causes of ageing. Ciba Found. Symp. 1988, 134, 193–207. (64) Saul, N.; Pietsch, K.; Menzel, R.; St€urzenbaum, S.; Steinberg, C. E. W. Catechin induced longevity in C. elegans: from key regulator genes to disposable soma. Mech. Ageing Dev. 2009, 130, 477–486. (65) Hosono, R.; Sato, Y.; Aizawa, S.; Mitsui, Y. Age-dependent changes in mobility and separation of the nematode Caenorhabditis elegans. Exp. Gerontol. 1980, 15, 285–289.
8715
dx.doi.org/10.1021/es2023237 |Environ. Sci. Technol. 2011, 45, 8707–8715
ARTICLE pubs.acs.org/est
Artificial Sweetener Sucralose in U.S. Drinking Water Systems Douglas B. Mawhinney,†,* Robert B. Young,†,‡ Brett J. Vanderford,† Thomas Borch,‡,§ and Shane A. Snyder|| †
Southern Nevada Water Authority, P.O. Box 99954, Las Vegas, Nevada 89193, United States Department of Soil and Crop Sciences, Colorado State University, Fort Collins, Colorado 80523, United States § Department of Chemistry, Colorado State University, Fort Collins, Colorado 80523, United States Department of Chemical and Environmental Engineering, University of Arizona, Tucson, Arizona 85721, United States
)
‡
bS Supporting Information ABSTRACT:
The artificial sweetener sucralose has recently been shown to be a widespread of contaminant of wastewater, surface water, and groundwater. In order to understand its occurrence in drinking water systems, water samples from 19 United States (U.S.) drinking water treatment plants (DWTPs) serving more than 28 million people were analyzed for sucralose using liquid chromatography tandem mass spectrometry (LC-MS/MS). Sucralose was found to be present in source water of 15 out of 19 DWTPs (47 2900 ng/L), finished water of 13 out of 17 DWTPs (49 2400 ng/L) and distribution system water of 8 out of the 12 DWTPs (48 2400 ng/L) tested. Sucralose was only found to be present in source waters with known wastewater influence and/ or recreational usage, and displayed low removal (12% average) in the DWTPs where finished water was sampled. Further, in the subset of DWTPs with distribution system water sampled, the compound was found to persist regardless of the presence of residual chlorine or chloramines. In order to understand intra-DWTP consistency, sucralose was monitored at one drinking water treatment plant over an 11 month period from March 2010 through January 2011, and averaged 440 ng/L in the source water and 350 ng/L in the finished water. The results of this study confirm that sucralose will function well as an indicator compound for anthropogenic influence on source, finished drinking and distribution system (i.e., tap) water, as well as an indicator compound for the presence of other recalcitrant compounds in finished drinking water in the U.S.
’ INTRODUCTION Artificial sweeteners are a class of organic contaminants recently discovered in wastewater, groundwater, surface water, bank filtrate and drinking water.1 7 These compounds are widely added to foods, personal care products and pharmaceutical formulations, and enter the environment in part due to incomplete removal during wastewater treatment. While artificial sweeteners undergo comprehensive toxicological testing prior to their use in consumer products, their unintended presence in water is still cause for concern. This is because the long-term health effects resulting from chronic exposure to low levels of these compounds and other trace organic contaminants in water, such as pharmaceuticals and personal care products,8,9 are largely unknown. 10,11 However, the true importance of the artificial sweeteners will likely be as indicator compounds for wastewater influence on other water types. 1,3,6 In any case, the quantification of these compounds in various water types, including finished drinking water, is the first step in understanding the significance of this issue. A great deal of analytical work for artificial sweeteners has concerned measuring their concentration in food products. 12 17 However, there are a growing number of studies of artificial r 2011 American Chemical Society
sweetener occurrence in wastewater and surface water. A large volume of this work has been completed in Europe, where acesulfame and sucralose were found to be the most persistent through wastewater treatment and in the environment. 1,3,4 Acesulfame was found to occur at higher levels than sucralose in the waste-, surface, ground, and tap water of Switzerland in Buerge et al.1 and wastewaters and surface waters of Germany in Scheurer et al.3 Both studies proposed acesulfame as an indicator compound for domestic wastewater influence on other water types. The occurrence of sucralose in surface waters has been measured in 27 European countries in Loos et al.,4 as well as smaller studies in Switzerland1 and Germany3 that included groundwater. This compound has also been proposed as an indicator compound of domestic wastewater influence on surface water,3 although acesulfame was found to be the predominant artificial sweetener in these studies. Both compounds were the subject of a recent study of bench-scale and full-scale drinking Received: July 12, 2011 Accepted: August 31, 2011 Revised: August 29, 2011 Published: August 31, 2011 8716
dx.doi.org/10.1021/es202404c | Environ. Sci. Technol. 2011, 45, 8716–8722
Environmental Science & Technology
ARTICLE
Table 1. Summary of Treatment Processes Employed at the DWTPs Included in This Study (Adapted From Ref 9) and the Concentration of Sucralose Determined in the Source, Finished and Distribution System Water for Each Sitea DWTP No.
treatment scheme
influence
source (ng/L)
finished (ng/L)
% removal
distribution (ng/L)
1
C/F f Sed. f Cl2 f Filt. f NH2Cl
ww
2400
2400
0
2400
2 3
ClO2 f C/F f Sed. f Filt. f UV f Cl2 ClO2 f C/F f Sed. f Filt. f UV f Cl2
ww ww
2900 790
1500 1100b (680)
48 14
n/a 1100b (680)
4
Cl2 f C/F f Sed. f Cl2 f Filt. f Cl2
ww
220
210
5
200
5
Cl2 f O3 f C/F f Filt. f Cl2
ww
240
n/a
n/a
n/a
6
Cl2 f O3 f C/F f Filt. f Cl2
ww
240
190
21
180
7
Cl2 f C/F f Sed. f Cl2 f Filt. f NH2Cl
ww
2200
2000
9
n/a
8
Cl2 f C/F f Sed. f Cl2 f Filt. f NH2Cl
ww
1500
1500
0
n/a
9
Cl2 f C/F f Sed. f Cl2 f Filt. f NH2Cl
ww
400
370
8
n/a
10 11
Cl2 f C/F f Sed. f Filt. f NH2Cl Cl2 f C/F f Sed. f Filt. f NH2Cl
n n
<MRL <MRL
<MRL <MRL
n/a n/a
<MRL <MRL
12
Cl2 f NH2Cl
n
<MRL
<MRL
n/a
<MRL
13
Cl2 f O3 f C/F f Sed. f NH2Cl
ww
150
130
13
120
14
Filt. f O3 f NH2Cl
n
<MRL
<MRL
n/a
<MRL
15
Cl2 f C/F f Sed. f Cl2 fFilt. f NH2Cl
ww
100
n/a
n/a
93
16
C/F f Sed. f O3 f dm-Filt. f NH2Cl
r
81
55
32
53
17
Cl2 f C/F f Sed. f Cl2 f Filt. f NH2Cl
r
47
49
(+4%)
48
18 19
O3 f C/F f Sed. f GAC/sand bio-Filt. f NH2Cl O3 f C/F f Sed. f GAC/sand bio-Filt. f NH2Cl
r r
72 62
108 98
(+50%) (+58%)
n/a n/a
C/F: coagulation/flocculation; Sed.: sedimentation; Cl2: free-chlorine (hypochlorite); Filt.; filtration; NH2Cl: chloramine; ClO2: chlorine dioxide: pre-ox.: O3; ozonation; dm-Filt.: dual-media filtration; GAC/sand bio-Filt.; granular activated carbon/sand biofiltration; ww: wastewater; n: none; r: recreational; <MRL: below method reporting limit; n/a: not available for analysis. b The finished water for DWTP-3 is a blend, and the value in parentheses has a blending factor applied. a
water treatment for their removal, where acesulfame was found to somewhat persistent, while sucralose was effectively removed by granular activated carbon in Scheurer et al.7 There is a growing body of work related to the artificial sweetener sucralose in the United States and Canada. Sucralose was measured in the marine and coastal waters of the U.S., as well as wastewater treatment plant effluent and river water subject to its discharge, where concentrations reached 1.9 μg/L in Mead et al.2 Later, sucralose was detected in five out of five wastewater samples from three locations, at a median concentration of 1.5 μg/L in Ferrer and Thurman.5 It was also detected in eight out of 22 surface water samples, in eight out of eight alluvial groundwater samples from two locations, and suggested as a conservative tracer of sewage effluent in surface and groundwater. Most recently, sucralose was demonstrated to be a viable indicator compound for wastewater loading in surface waters, where it was consistently found in wastewater, wastewaterinfluenced surface water and septic samples in Oppenheimer et al.6 Sucralose occurrence has also recently been measured in urban groundwater in Canada, although acesulfame was found more consistently in Van Stempoort et al.18 Sucralose has been found to be persistent through wastewater treatment processes in municipal plants in Torres et al.,19 and some bench scale studies in Soh et al.20 Sucralose, a chlorinated form of sucrose, is an artificial sweetener that is approximately 600 times sweeter than sucrose and is widely used as a food additive.21 It is not extensively adsorbed and metabolized in humans, resulting in the majority being excreted without transformation.22 Sucralose is very soluble in water (282 g/L at 20 °C, Log P = 0.51),23 so it is not expected to be extensively adsorbed to organic solids in the environment. Further, sucralose has been shown to be resistant
to treatment processes used in the production of drinking water at the bench scale, such as oxidation by free chlorine and ozone.20 Because of these properties, it is likely that sucralose may be widespread in drinking waters in the U.S., persisting through traditional drinking water treatment processes. The objective of this work was to quantify the amount of sucralose present in select U.S. drinking water systems. To this end, an archived set of extracts from the study reported in Benotti et al.,9 of the source, finished, and distribution system water from 19 U.S. DWTPs from 2006 to 2007, was analyzed for this compound. The relative ineffectiveness of commonly employed drinking water treatment strategies to remove sucralose is presented. The utility of this compound as an indicator for anthropogenic influence on drinking water systems, and as an indicator for other recalcitrant compounds in finished drinking water, is also discussed. To date, this manuscript represents the first study to quantify the artificial sweetener sucralose in U.S. drinking water systems.
’ EXPERIMENTAL SECTION Sample Collection. The sampling sites consisted of the source, finished and distribution systems of 19 U.S. DWTPs. This is the same sample set analyzed by Benotti et al.9 Only 10 of the 19 DWTPs had sites in the distribution system sampled, and one finished water sample extract from one DWTP was lost during storage and handling. The treatment process employed at each plant is summarized in Table 1. The hydraulic retention time was taken into account when collecting samples so that the data should represent, as closely as possible, the same plug of water moving through the treatment process. 8717
dx.doi.org/10.1021/es202404c |Environ. Sci. Technol. 2011, 45, 8716–8722
Environmental Science & Technology Samples were collected in 1 L silanized, baked amber glass bottles containing the preservative sodium azide and ascorbic acid to quench any residual oxidant. All samples were cooled on ice, shipped overnight and stored at 4 °C in the laboratory for a maximum of 14 days until extraction. Travel blanks were included with each sampling event, and consisted of laboratorygrade water that was shipped with the bottles, and transferred to a sampling bottle during the sampling procedure. All samples were extracted using the SPE method described in Vanderford et al.,24 and resulted in a 1000-fold concentration. Following extraction, the sample extracts in methanol were capped in airtight sample vials and stored at 80 °C until the time of analysis, which totaled 36 48 months depending on sampling and extraction dates. Prior to analysis, extracts were allowed to warm to room temperature and a positive displacement pipet was used to transfer 40 μL to an autosampler vial containing 60 μL of a solution of the isotope-labeled internal standard, sucralose-d6 (Toronto Research Chemicals, Toronto, ON). After thorough mixing using a glass pipet, the sample was diluted 2.5 times to a final composition of 46% MeOH, 54% DI water, and an internal standard concentration of 200 ng/mL. The original extract was capped and returned to the 80 °C freezer, and the prepared extract was stored at 4 °C until analysis. The source waters of the DWTPs included in this study display a range of anthropogenic influence. Only one of the DWTPs utilizes a groundwater source (DWTP-12), while the remaining employ surface water. The source water from DWTPs 1 to 9, 13, and 15 are impacted by wastewater inputs. The source waters for DWTPs 16 to 19 are reservoirs with recreational usage, but no known wastewater inputs. The remaining DWTPs have no known wastewater inputs or recreational usage. Analysis Method. All extracts were analyzed by isotope dilution liquid chromatography tandem mass spectrometry (LC-MS/MS). The chromatography was performed on an Agilent (Santa Clara, CA) 1200 system equipped with a Restek (Bellefonte, PA) Allure Organic Acids column, 150 3.2 mm with 5 μm particles. Mass spectrometric analysis was performed using an AB SCIEX (Foster City, CA) API-4000 QTrap operating with negative electrospray ionization in MRM mode. Sucralose was quantified using the (M-H) precursor ion at m/z 395.0 and the Cl product ion at m/z 35.0 and confirmed using the (M-H) precursor ion at m/z 397.0 and the Cl product ion at m/z 35.0. The internal standard was measured using the (M-H) precursor ion at m/z 403.0 and the Cl product ion at m/z 35.0 in order to avoid significant contribution from the naturally occurring isotopes of sucralose. A product ion spectrum of the precursor ion at m/z 395.0 collected at a collision energy of 22 eV is shown in Figure 1. Further details on the analytical method can be found in the Supporting Information. Representative LC-MS/ MS chromatograms for source water, finished water and travel blank extracts from a sampling event for DWTP-8 can also be found in the Supporting Information (Figure SI-1).
’ RESULTS AND DISCUSSION Analytical Parameters and Data Quality. The percent recovery was determined by analyzing 10 samples of laboratory reagent water spiked with 250 ng/L of sucralose and with 200 ng/L of isotopically labeled internal standard, taken through the entire sample extraction procedure. The absolute recovery of sucralose was determined by external calibration (i.e., no internal standard correction) to be 85 ( 7%. This represents the
ARTICLE
Figure 1. Sucralose product ion spectrum in negative ion mode with a collision energy of 22 V. The molecular structure, empirical formula, and average mass are included. The precursor ion is the (M-H) species at m/z = 395.0 Da.
recovery of the sucralose in the samples through the extraction procedure. Relative recovery was determined by ratioing the sucralose to the internal standard in order to account for any analyte ionization suppression and extraction efficiency, which was calculated to be 99%. Since the internal standard was added postextraction in the procedure used for the archived extracts, this only corrects for ion suppression, but not for the extraction recovery. The results reported in this manuscript are reported as such, with no correction for extraction efficiency. It should be noted that the artificial sweeteners saccharin and acesulfame were also considered for analysis; however, the extraction procedure used in the original study9 resulted in very low recoveries. Therefore, they could not be quantified in these extracts. The minimum detection limit (MDL) was determined from the analysis of eight samples of laboratory reagent water spiked with 10 ng/L of sucralose and 200 ng/L of the internal standard. The samples were subjected to the entire sample preparation procedure and quantified by the LC-MS/MS method using internal calibration. The MDL was calculated from the resulting data by multiplying the standard deviation of these eight replicate measurements by the appropriate Student’s T-value for n 1 degrees of freedom, and determined to be 1.6 ng/L. The low calibrator and reporting limit was set to a factor 6 times higher, or 10 ng/L. A similar procedure has been used previously to determine MDL values for pharmaceutical compounds in water matrices in Vanderford et al.24 Because the analyses were performed on a set of sample extracts archived for 36 48 months at 80 °C, there are concerns with sample integrity during long-term storage. Sucralose has been shown to be a relatively inert and stable molecule, and is known to be very soluble in the storage solvent, methanol.25 There were several duplicate samples collected throughout the study for source, finished drinking and distribution water. The majority differed by less than 10% of the average of the two values, although one distribution duplicate sample had a 15% difference. This error is well within that expected in typical water analysis methods, and indicates acceptable data quality. It also shows that the long-term storage conditions did not cause 8718
dx.doi.org/10.1021/es202404c |Environ. Sci. Technol. 2011, 45, 8716–8722
Environmental Science & Technology significant differences between samples. Additionally, sucralose was not detected in any of the travel blanks, demonstrating the integrity of sample collection and handling procedures. Sucralose Occurrence in Source and Finished Water. Sucralose was detected in the majority of source water samples, and displayed a strong dependence on the relative amount of anthropogenic influence. Those source waters that were impacted by wastewater all tested positive and displayed the highest concentrations, ranging from 100 ng/L (DWTP-15) to 2900 ng/ L (DWTP-2), as shown in Table 1. Source waters that had no known wastewater inputs but were subject to recreational usage, such as boating and swimming, also tested positive but displayed lower concentrations than those impacted by wastewater. This sample group also displayed very similar levels to one another, ranging from 47 (DWTP-17) to 81 ng/L (DWTP-16), with sucralose likely being present from improper disposal of foodstuffs and human waste.22 Those source waters that had no known wastewater inputs or recreational usage did not contain quantifiable levels of sucralose (DWTP-10 to 12, 14). The concentrations reported in this work are comparable with previously published levels of artificial sweeteners 2 6 in surface and groundwater. Sucralose concentrations range higher than those reported in European surface waters,3,4 but are comparable to those measured in American surface and ground waters,2,5,6 likely due to the time period for its approved usage (beginning in 1998 in the U.S. versus 2004 in the E.U.), as well as differences in artificial sweetener usage patterns. For results obtained in U.S. waters, the concentrations reported here range up to 2900 ng/L, compared with 1900 ng/L by Mead et al., 2400 ng/L by Ferrer and Thurman, and 10,000 ng/L by Oppenheimer et al. Interestingly, these results are from surface water samples, with the exception of Ferrer et al., which is from a groundwater sample, demonstrating that this compound can be of concern even for those DWTPs with groundwater sources. All of the DWTPs that used source water that tested positive for sucralose also had quantifiable amounts in the finished water, with the highest level at 2400 ng/L (DWTP-1). In fact, most DWTPs showed very little removal, including those that employed ozone as an oxidant. Counting any increases in concentration between finished and source water as 0% removal, the average removal across the plants with representative samples was 12%. DWTP-2 displayed the highest removal (48%), and employed ClO2 preoxidation and UV treatment prior to chlorination. Although DWTP-3 has significantly less removal (14%), this still suggests that advanced oxidation processes (AOP) should be explored further for the removal of sucralose from drinking water. Persistence of Sucralose in Distribution Systems. The distribution system water of the DWTPs whose finished water tested positive for sucralose also tested positive (Figure 2). Further, within expected experimental error, every distribution system sample tested at the same levels as the finished water. These results indicate that residual free chlorine or chloramines are not effective at removing sucralose during transit from the DWTP to the distribution sampling point. As mentioned in the Experimental Section, these sampling sites were chosen to be relatively far removed from the DWTP, so it can be expected that sucralose persists completely through the distribution system. Every distribution water (i.e., tap water) that was produced from source water with measurable amounts of sucralose, also had sucralose present (Table 1, Figure 2). Therefore, it is
ARTICLE
Figure 2. Comparison of sucralose concentration the source, finished and distribution water samples from seven utilities. These seven DWTPs represent those that had quantifiable amounts of sucralose in the source water, and had at least one finished and one distribution sample available for testing. *DWTP-3 was corrected for relative input to blend.
Figure 3. Sucralose concentrations in the source and finished water of DWTP-5 measured monthly by isotope dilution LC-MS/MS, and it is ratio to the meprobamate concentrations in the same sample sets. The average daily output of the DWTP during this time was 130 million gallons/day.
reasonable to expect that human exposure to sucralose via tap water is widespread in the U.S. Since the exact amount of tap water consumed daily on an individual basis varies, an average consumption amount of 2 L is estimated by the U.S. Environmental Protection Agency (US EPA).26 Using this volume of tap water, the largest exposure to sucralose was from DWTP-1, where an individual would have been exposed to 4.8 μg of sucralose per day. The toxicological relevance of chronic exposure to this amount warrants further studies, but is beyond the scope of this paper. 11-Month Monitoring Program at DWTP-5. In order to test whether the presence of sucralose is consistent over time, the 8719
dx.doi.org/10.1021/es202404c |Environ. Sci. Technol. 2011, 45, 8716–8722
Environmental Science & Technology
ARTICLE
Table 2. Comparison of the Average Concentrations, Median Concentrations and Concentration Range of Sucralose with Compounds That Were Consistently Present for Each Water Type with Either Wastewater (ww) or Recreational Usage (r) Influence from Supporting Information in Benotti et al.9 water type
influence
analyte
average concentration, ng/L
median concentration (ng/L)
concentration range (ng/L)
MRL (ng/L)
source ww
sucralose
1000
400
150 2900
10
ww
carbamazepine
17
4.8
1.9 51
0.50
ww
phenytoin
13
8.7
2.5 29
1.0
ww ww
meprobamate sulfamethoxazole
20 35
10 15
3.2 73 1.6 110
0.25 0.25
ww
atrazine
180
32
1.4 870
0.25
r
sucralose
66
67
47 81
10
r
carbamazepine
1.5
1.7
0.60 1.8
0.50
r
meprobamate
3.0
3.1
1.6 4.3
0.25
r
sulfamethoxazole
1.7
1.6
0.62 3.2
0.25
r
atrazine
310
230
0.50 780
0.25
finished ww
sucralose
1000
1100
130 2400
10
ww
phenytoin
8.7
6.2
1.1 19
1.0
ww
meprobamate
17
13
3.3 42
0.25
ww
atrazine
150
59
0.88 870
0.25
r
sucralose
78
77
49 108
10
r r
meprobamate atrazine
1.4 120
1.5 110
1.0 1.8 0.38 240
0.25 0.25
distribution ww
sucralose
680
190
93 2400
10
ww
phenytoin
5.3
3.6
1.1 15
1.0
ww
meprobamate
14
6.1
2.7 38
0.25
ww
atrazine
186
113
0.94 915
0.25
r r
sucralose meprobamate
51 1.5
51 1.5
48 53 1.3 1.6
10 0.25
r
atrazine
0.47
0.47
0.36 0.58
0.25
concentration in the raw and finished drinking water at DWTP-5 was monitored over the 11-month period from March, 2010 until January, 2011 (Figure 3). This DWTP has an ozone contact time of 24 min and a chlorine contact time of ∼1.5 h. The amount of sucralose was quantified by isotope dilution LC-MS/MS using the method outlined in the Experimental Section, with the exception that the isotope was spiked into the sample prior to extraction, and the methanol extract was analyzed without dilution. This procedure corrects both for losses due to extraction and handling, as well as any ion suppression during analysis. The concentration of sucralose was relatively stable in both sample sets over that time, indicating that the flux of sucralose to that body of water was relatively consistent. There was an increase in concentration between October and November, which is due to seasonal fluctuations in the mixing dynamics of the treated wastewater stream and the reservoir, as determined previously through water quality parameters.27,28 The removal efficiency is consistently low through the treatment process (20% average removal), demonstrating the resistance of the compound to oxidation by both ozone and chlorine. Sucralose as an Indicator Compound. Indicator compounds for drinking water have been defined as anthropogenic
compounds that can be used to represent certain classes of compounds based on their similarities in occurrence and behavior under various environmental and treatment conditions.29 They have also been defined as chemical compounds that denote the presence of different types of water influence, usually wastewater.6 The data presented here confirm that sucralose is a reliable indicator compound for anthropogenic influence on source water across the United States. It is important to note that it is not only present in water that is directly impacted by wastewater, but also water where recreational use is permitted (with no known wastewater inputs). Given the poor removal of sucralose during the various treatment schemes, it is also a viable indicator compound in finished and distribution water for other resistant compounds, such as meprobamate and atrazine. 9,30 In fact, when the concentration of sucralose is ratioed to that of meprobamate for data collected over a 11-month period at DWTP-5, relatively consistent values are obtained for both the source (average ratio = 41, %RSD = 8.4) and finished (average ratio = 46, %RSD = 5.7) water (Figure 3). A comparison of the sucralose data with those from the much broader study by Benotti et al.9 reveals important trends. Table 2 shows the concentrations of the compounds that were 8720
dx.doi.org/10.1021/es202404c |Environ. Sci. Technol. 2011, 45, 8716–8722
Environmental Science & Technology consistently present based on type (source, finished or distribution) and influence (wastewater or recreational usage). Sucralose, atrazine and meprobamate were the only analytes from this compound set that were present in source, finished and distribution water, regardless of the influence type. The source of sucralose and meprobamate is likely from direct human activities, whereas the majority of atrazine present is likely from agricultural application and runoff. Therefore, sucralose and meprobamate should represent better indicators of wastewater and/or direct human input to source water, and subsequently finished drinking and distribution (i.e., tap) waters. While the concentration of sucralose was much higher than the concentration of meprobamate for all of the waters tested, from an analytical standpoint, the concentrations are comparable in relation to the respective method reporting limits (MRLs). A comparison with the meprobamate data presented by Benotti et al.9 is included in the Supporting Information (Table SI-1). Given expected decreases in the input of meprobamate from human usage,31 33 sucralose will likely be the more important indicator compound in the future.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (702) 856-3634; fax: (702) 856-3647; e-mail:
[email protected].
’ ACKNOWLEDGMENT We thank Janie Holady and Josephine Chu for their assistance throughout the project. The authors also thank the Water Research Foundation (Project No. 3085) for funding the initial sample collection and extraction. T.B. was funded by a National Science Foundation (NSF) CAREER Award (EAR 087683). ’ REFERENCES (1) Buerge, I. J.; Buser, H.-R.; Kahle, M.; M€uller, M. D.; Poiger, T. Ubiquitous occurrence of the artificial sweetener acesulfame in the aquatic environment: an ideal chemical marker of domestic wastewater in groundwater. Environ. Sci. Technol. 2009, 43 (12), 4381–4385. (2) Mead, R. N.; Morgan, J. B.; Avery, G. B., Jr; Kieber, R. J.; Kirk, A. M.; Skrabal, S. A.; Willey, J. D. Occurrence of the artificial sweetener sucralose in coastal and marine waters of the United States. Mar. Chem. 2009, 116 (1 4), 13–17. (3) Scheurer, M.; Brauch, H.-J.; Lange, F. T. Analysis and occurrence of seven artificial sweeteners in German waste water and surface water and in soil aquifer treatment (SAT). Anal. Bioanal. Chem. 2009, 394 (6), 1585–1594. (4) Loos, R.; Gawlik, B. M.; Boettcher, K.; Locoro, G.; Contini, S.; Bidoglio, G. Sucralose screening in European surface waters using a solid-phase extraction-liquid chromatography-triple quadrupole mass spectrometry method. J. Chromatogr., A 2009, 1216 (7), 1126–1131. (5) Ferrer, I.; Thurman, E. M. Analysis of sucralose and other sweeteners in water and beverage samples by liquid chromatography/time-offlight mass spectrometry. J. Chromatogr., A 2010, 1217 (25), 4127–4134. (6) Oppenheimer, J.; Eaton, A.; Badruzzaman, M.; Haghani, A. W.; Jacangelo, J. G. Occurrence and suitability of sucralose as an indicator compound of wastewater loading to surface waters in urbanized regions. Water Res. 2011, 45 (13), 4019–4027.
ARTICLE
(7) Scheurer, M.; Storck, F. R.; Brauch, H.-J.; Lange, F. T. Performance of conventional multi-barrier drinking water treatment plants for the removal of four artificial sweeteners. Water Res. 2010, 44 (12), 3573–3584. (8) Kolpin, D. W.; Furlong, E. T.; Meyer, M. T.; Thurman, E. M.; Zaugg, S. D.; Barber, L. B.; Buxton, H. T. Pharmaceuticals, hormones, and other organic wastewater contaminants in US streams, 1999 2000: A national reconnaissance. Environ. Sci. Technol. 2002, 36 (6), 1202–1211. (9) Benotti, M. J.; Trenholm, R. A.; Vanderford, B. J.; Holady, J. C.; Stanford, B. D.; Snyder, S. A. Pharmaceuticals and endocrine disrupting compounds in U.S. drinking water. Environ. Sci. Technol. 2008, 43 (3), 597–603. (10) Snyder, S.; Westerhoff, P.; Yoon, Y.; Sedlak, D. L. Pharmacauticals, personal care products, and endocrine disruptors in water: implications for the water industry. Environ. Eng. Sci. 2003, 20 (5), 449–469. (11) Jones, O. A.; Lester, J. N.; Voulvoulis, N. Pharmaceuticals: a threat to drinking water? Trends Biotechnol. 2005, 23 (4), 163–167. (12) Quinlan, M. E.; Jenner, M. F. Analysis and stability of the sweetener sucralose in beverages. J. Food Sci. 1990, 55 (1), 244–246. € (13) Demiralay, E. C-.; Ozkan, G.; Guzel-Seydim, Z. Isocratic separation of some food additives by reversed phase liquid chromatography. Chromatographia 2006, 63 (1), 91–96. (14) García-Jimenez, J. F.; Valencia, M. C.; Capitan-Vallvey, L. F. Simultaneous determination of antioxidants, preservatives and sweetener additives in food and cosmetics by flow injection analysis coupled to a monolithic column. Anal. Chim. Acta 2007, 594 (2), 226–233. (15) Sheridan, R. K., T. Determination of cyclamate in foods by ultraperformance liquid chromatography/tandem mass spectrometry. J. AOAC Int. 2008, 91 (5), 1095–1102. (16) Wasik, A.; McCourt, J.; Buchgraber, M. Simultaneous determination of nine intense sweeteners in foodstuffs by high performance liquid chromatography and evaporative light scattering detection— Development and single-laboratory validation. J. Chromatogr., A 2007, 1157 (1 2), 187–196. (17) Yang, D.-j.; Chen, B. Simultaneous determination of nonnutritive sweeteners in foods by HPLC/ESI-MS. J. Agr. Food Chem. 2009, 57 (8), 3022–3027. (18) Van Stempvoort, D. R.; Roy, J. W.; Brown, S. J.; Bickerton, G. Artificial sweeteners as potential tracers in groundwater in urban environments. J. Hydrol. 2011, 401 (1 2), 126–133. (19) Torres, C. I.; Ramakrishna, S.; Chiu, C.-A.; Nelson, K. G.; Westerhoff, P.; Krajmalnik-Brown, R. Fate of sucralose during wastewater treatment. Environ. Eng. Sci. 2011, 28 (5), 325–331. (20) Soh, L.; Connors, K. A.; Brooks, B. W.; Zimmerman, J. Fate of sucralose through environmental and water treatment processes and impact on plant indicator species. Environ. Sci. Technol. 2011, 45 (4), 1363–1369. (21) Knight, I. The Development and applications of sucralose, a new high-intensity sweetener. Can. J. Physiol. Pharmacol. 1994, 72 (4), 435–439. (22) Roberts, A.; Renwick, A. G.; Sims, J.; Snodin, D. J. Sucralose metabolism and pharmacokinetics in man. Food Chem. Toxicol. 2000, 38 (Supplement 2), 31–41. (23) Jenner, M. R.; Smithson, A. Physicochemical properties of the sweetener sucralose. J. Food Sci. 1989, 54 (6), 1646–1649. (24) Vanderford, B. J.; Snyder, S. Analysis of pharmaceuticals in water by isotope dilution liquid chromatography/tandem mass spectrometry. Environ. Sci. Technol. 2006, 40, 7312–7320. (25) Li, X.; Du, Z.; Huang, X.; Yuan, W.; Ying, H. Solubility of sucralose in different solvents from (283.15 to 333.15) K. J. Chem. Eng. Data 2010, 55 (7), 2600–2602. (26) Estimated Per Capita Water Ingestion and Body Weight in the United States an Update, EPA-822-R-00-001; U.S. Environmental Protection Agency: Washington, DC, 2004. (27) LaBounty, J. F.; Burns, N. M. Characterization of Boulder Basin, Lake Mead, Nevada-Arizona, USA—Based on analysis of 34 limnological parameters. Lake Reservoir Manage. 2005, 21 (3), 277–307. 8721
dx.doi.org/10.1021/es202404c |Environ. Sci. Technol. 2011, 45, 8716–8722
Environmental Science & Technology
ARTICLE
(28) LaBounty, J. F.; Horn, M. J. The influence of drainage from the Las Vegas Valley on the limnology of Boulder Basin, Lake Mead, Arizona-Nevada. Lake Reservoir Manage. 1997, 13 (2), 95–108. (29) Dickenson, E. R. V.; Drewes, J. E.; Sedlak, D. L.; Wert, E. C.; Snyder, S. A. Applying surrogates and indicators to assess removal efficiency of trace organic chemicals during chemical oxidation of wastewaters. Environ. Sci. Technol. 2009, 43 (16), 6242–6247. (30) Westerhoff, P.; Yoon, Y.; Snyder, S.; Wert, E. Fate of endocrinedisruptor, pharmaceutical, and personal care product chemicals during simulated drinking water treatment processes. Environ. Sci. Technol. 2005, 39 (17), 6649–6663. (31) Littrel, R. A.; Hayes, L. R.; Stillner, V. Carisoprodol (Soma): a new and cautious perspective on an old agent. South. Med. J. 1993, 86 (7), 753–756. (32) Harden, R. N.; Argoff, C. A review of three commonly prescribed skeletal muscle relaxants. J. Back Musculoskelet. 2000, 15 (2 3), 63–66. (33) Olsen, H.; Koppang, E.; Alvan, G.; Morland, J. Carisoprodol elimination in humans. Ther. Drug Monit. 1994, 16 (4), 337–340.
8722
dx.doi.org/10.1021/es202404c |Environ. Sci. Technol. 2011, 45, 8716–8722
ARTICLE pubs.acs.org/est
Occurrence of Iodinated X-ray Contrast Media and Their Biotransformation Products in the Urban Water Cycle Jennifer Lynne Kormos, Manoj Schulz, and Thomas A. Ternes* Federal Institute of Hydrology (BfG), Am Mainzer Tor 1, D-56068, Koblenz, Germany
bS Supporting Information ABSTRACT: A LC tandem MS method was developed for the simultaneous determination of four iodinated X-ray contrast media (ICM) and 46 ICM biotransformation products (TPs) in raw and treated wastewater, surface water, groundwater, and drinking water. Recoveries ranged from 70% to 130%, and limits of quantification (LOQ) varied between 1 ng/L and 3 ng/L for surface water, groundwater and drinking water, and between 10 ng/L and 30 ng/L for wastewater. In a conventional wastewater treatment plant, iohexol, iomeprol, and iopromide were transformed to >80%, while iopamidol was transformed to 35%. In total, 26 TPs were detected above their LOQ in WWTP effluents. A significant change in the pattern of ICM TPs was observed after bank filtration and groundwater infiltration under aerobic conditions. Predominately, these TPs are formed at the end of the microbial transformation pathways in batch experiments with soil and sediment. These polar ICM TPs, such as iohexol TP599, iomeprol TP643, iopromide TP701A, and iopromide TP643, were not or only partially removed during drinking water treatment. As a consequence, several ICM TPs were detected in drinking water, at concentration levels exceeding 100 ng/L, with a maximum of 500 ng/L for iomeprol TP687.
’ INTRODUCTION Iodinated X-ray contrast media (ICM) are used for the imaging of internal body structures (i.e., organs, blood vessels, and soft tissues) during diagnostic examinations.1 ICM are reported to be the most widely used pharmaceuticals for intravascular administration and most frequently used in hospitals.1,2 They are applied at high doses (i.e., up to 200 g/application) and are eliminated nonmetabolized in the urine within 24 h.3 Most ICM are derivatives of 2,4,6-triiodobenzoic acid and are classified as ionic or nonionic depending on the functional moieties at their side chains. For instance, the ionic ICM, diatrizoate, is negatively charged at neutral pH due to its carboxylic moiety (pKa = 3.4), while the nonionic ICM (i.e., iohexol, iopamidol, and iomeprol) contain functional moieties which are uncharged at neutral pH.4,5 It has been reported in several studies that these hydrophilic and metabolically stable ICM are not effectively removed in conventional wastewater treatment plants (WWTPs).68 The elevated concentrations of ICM in surface water, groundwater, and even treated water of drinking water treatment plants (DWTPs) can be explained by their persistence.2,913 However, recent research has shown that certain oxidation processes (i.e., ozonation, advanced oxidation processes) as well as biological processes of WWTPs with elevated sludge retention times (SRT) of >12 d are capable of transforming nonionic ICM.1417 Nevertheless, elimination did not result in mineralization of the parent ICM. In most cases, oxidation products or biotransformation products (TPs) were formed.16,1822 Aerobic r 2011 American Chemical Society
biotransformation reduced the length of the ICM side chains, while electrochemical and photochemical processes may lead to deiodinated products.17,23 Recently, 46 TPs of four nonionic ICM have been identified which were grouped into different phases depending on the sequence of their formation.2022 So far, minimal efforts have focused on the presence of these newly identified TPs in the aquatic environment and drinking water supplies. The main objective of this study was to investigate the occurrence and fate of the newly identified TPs of four nonionic ICM (iohexol, iomeprol, iopamidol, and iopromide) in different aqueous matrixes using an optimized LC tandem MS analytical method. In particular, this study focuses on changes occurring in the composition and patterns of these TPs in the urban water cycle from WWTPs via surface water to groundwater and drinking water.
’ EXPERIMENTAL SECTION Description of Sampling Locations. Wastewater Treatment Plant (WWTP). Five-day composite samples were collected from a
conventional German WWTP during dry weather conditions in December 2009. This WWTP serves 285 000 population equivalents (PE), has a SRT and hydraulic retention time (HRT) of Received: December 18, 2010 Accepted: August 11, 2011 Revised: August 1, 2011 Published: August 31, 2011 8723
dx.doi.org/10.1021/es2018187 | Environ. Sci. Technol. 2011, 45, 8723–8732
Environmental Science & Technology
ARTICLE
Figure 1. Flow scheme of a conventional German WWTP consisting of mechanical and biological treatment. Sampling points 1 and 2 refer to the influent 1 and 2; sampling point 3 refers to prebiological treatment; sampling point 4 refers to postbiological treatment; and sampling point 5 refers to final effluent.
16 d and 60 h, respectively, and an average wastewater inflow of approximately 35 500 m3/d. It receives two different wastewater streams from the surrounding area, consisting of domestic sewage from various districts of the city as well as hospital wastewaters (i.e., University hospital). This WWTP applies mechanical treatment, biological phosphate removal, denitrification, nitrification, secondary sedimentation, and filtration. The mechanical treatment consists of different sized screens, a grit removal tank (sand filtration tank), and a primary sedimentation tank. After primary sedimentation, the wastewater enters a biological phosphate elimination tank, followed by an activated sludge tank containing denitrification and nitrification. The effluent enters a secondary sedimentation tank and a two-layer filter (consisting of anthracite and sand) before being discharged into the receiving water body. Figure 1 provides a schematic diagram of this WWTP and the sampling sites. Samples were taken from both influent streams (sampling point 1 and 2), prior to denitrification and nitrification (sampling point 3), after secondary sedimentation (sampling point 4), and after filtration (sampling point 5). Rhine River Water and Riverbank Wells. On November 18th, 2009, three riverbank wells close to the Rhine River were sampled at Urmitz, a small town in Germany close to Koblenz. The distance from the wells to the river ranged from 36 to 150 m, and the water level depths varied between 8.53 and 9.29 m. In addition, four-week composite samples were collected along the Rhine River at Koblenz, a distance of 590.3 km upstream from Lake Constance, from September to November 2009. Water quality parameters of the individual wells are summarized in Table S1 of the Supporting Information. The river bank and soil surrounding the wells consisted of crushed rock and gravel. The average infiltration velocity from the Rhine River into groundwater was approximately 13 m/d. DWTP Treatment Processes. Water was collected from four German DWTPs (DWTP14) after individual treatment processes. The raw water source for DWTP1 is a mixture of groundwater and bank filtrate, which is then directed to a granular activated carbon (GAC) filtration system before being distributed. DWTP2 receives its raw water from bank filtration of surface water, artificial groundwater recharge, and natural groundwater. Aeration and sand filtration are the main treatment processes which are capable of removing contaminants. DWTP3 uses river water as a raw water source and is treated by flocculation with Fe(III)chloride, followed by sedimentation, ozonation, multilayer filtration, and then GAC filtration. Samples were collected from the raw water source as well as after
ozonation and after GAC filtration. DWTP4, applying a similar multibarrier approach as DWTP3, uses a combination of river water and groundwater as raw water sources. After the river water undergoes flocculation with Fe(III)chloride, sedimentation, sand filtration, and GAC filtration, it is mixed with groundwater and is treated by aeration slow sand filtration and chlorine dioxide disinfection. Samples were collected from the river, after GAC filtration of the river water, after mixing of the treated water and groundwater, and finally after the addition of chlorine dioxide. Sample Preparation and Extraction for ICM TPs. The aqueous samples were filtered through glass fiber filters (Schleicher and Schuell, Dassel, Germany), acidified to pH 2.6 by adding sulfuric acid (3.5 M H2SO4), and stored at 4 °C prior to extraction. Two solid phase extraction (SPE) cartridges, C18 cartridge (3 mL, 200 mg, J.T. Baker) placed on top of Bakerbond SDB-1 (3 mL, 200 mg, J.T. Baker), were used for cleanup and extraction, respectively. Both SPE cartridges were conditioned with 4 1 mL of methanol and 4 1 mL of groundwater adjusted to pH 2.6 with 3.5 M H2SO4. For raw and treated wastewater, 100 and 200 mL were extracted, respectively, and 500 mL or 1 L was extracted for surface water, groundwater, and drinking water. Prior to extraction, all samples were spiked with 10 μL of a surrogate solution (20 ng/μL). The surrogate standard solution consisted of iohexol-d5, iomeprol-d3, iopamidol-d3, diatrizoate-d6, and desmethoxyiopromide (DMI). DMI was kindly provided by Bayer Schering Pharma (Berlin, Germany), diatrizoate-d6 was purchased from Campro Scientific (Berlin, Germany), and iohexol-d5, iomeprol-d3, and iopamidol-d3 were purchased from Toronto Research Chemicals (North York, Canada). After enrichment by SPE, the cartridges were washed with 4 mL of Milli-Q water adjusted to pH 2.6 by adding 3.5 M H2SO4. The C18 cartridges were disposed of, and the SDB-1 cartridges were dried under a gentle stream of nitrogen gas. The analytes bound on the sorbents in the SDB-1 cartridges were eluted with 4 2 mL of methanol, evaporated to 100 μL using nitrogen gas, and reconstituted with 900 μL of Milli-Q water. The extracted samples were stored at 4 °C until measured by LC tandem MS. LC ESI(+) Tandem MS Detection of ICM and TPs. An Agilent 1200 Series HPLC system consisting of an autosampler, binary pump, degasser (Waldbronn, Germany), and a column oven (MayLab Analytical Instruments, Austria) was used. Chromatographic separation of the analytes was achieved on two coupled Chromolith Performance RP-18e columns (4.6 mm 100 mm, Merck, Darmstadt, Germany) equipped with a Chromolith 8724
dx.doi.org/10.1021/es2018187 |Environ. Sci. Technol. 2011, 45, 8723–8732
Environmental Science & Technology
ARTICLE
Table 1. Recoveries (%) and 95% Confidence Intervals (n = 3) for ICM and Selected TPs in the Investigated Environmental Matrixes Analyte Iohexol Iohexol TP687A Iohexol TP657 Iohexol TP599 Iomeprol Iomeprol TP701
WWTP influenta
WWTP effluenta
River waterb
Groundwaterb
Drinking waterc
91 ( 25
120 ( 21
71, 89 (n = 2)
106 ( 14
96 ( 31
95 ( 11 106 ( 21
119 ( 6 117 ( 15
78 ( 12 90 ( 10
75 ( 12 105 ( 4
84 ( 10 120 ( 7 136 ( 18
83 ( 49
123 ( 36
82 ( 5
85 ( 2
123 ( 158d
92 ( 21
80 ( 9
113 ( 14
89 ( 41
94 ( 29
108 ( 22
121 ( 26
106 ( 29
96 ( 27 120 ( 40
Iomeprol TP643
90 ( 27
104 ( 15
84 ( 27
104 ( 46
Iomeprol TP629
105 ( 17
112 ( 9
74 ( 8
82 ( 16
83 ( 26
Iopamidol
111 ( 13
100 ( 9
117 ( 18
113 ( 20
128 ( 38
Iopamidol TP791 Iopamidol TP761
87 ( 4 112 ( 19
93 ( 5 93 ( 21
90 ( 17 88 ( 14
93 ( 6 97 ( 10
87; 86 (n = 2) 141 ( 29
Iopamidol TP745
94 ( 4
99 ( 3
88 ( 14
78 ( 6
109 ( 25
Iopromide
79 ( 13
89 ( 14
119 ( 33
105 ( 1
111 ( 15
Iopromide TP819
78 ( 4
91 ( 2
87 ( 21
41 ( 14
111 ( 15
Iopromide TP31B
86 ( 12
97 ( 3
115 ( 24
84 ( 11
101 ( 12
Iopromide TP729A
121 ( 14
112 ( 15
137 ( 13
110 ( 19
120 ( 16
Iopromide TP759
82 ( 10
95 ( 10
102 ( 3
93 ( 15
123 ( 18
Iopromide 701A Iopromide 701B
81 ( 4 78 ( 9
92 ( 3 86 ( 4
100 ( 15 114 ( 16
94 ( 13 101 ( 16
111 ( 16 105 ( 11
Iopromide TP643
83 ( 11
99 ( 10
76 ( 13
72 ( 18
71 ( 13
Spiking concentration of 50 μg/L for ICM concentrations higher than 10 μg/L (in those cases direct injection was used without SPE). b Spiking concentration of 0.1 μg/L. c Spiking concentration of 0.05 μg/L. d Spiking concentration was too low (background concentrations were high compared to spiking concentration). a
RP-18e guard column (4.6 mm 5 mm, Merck, Darmstadt, Germany). A sample aliquot of 50 μL was injected into the LC tandem MS, and the analytes were eluted from the column using two mobile phases, 95% Milli-Q water, 5% methanol, and 0.5% formic acid (A) and 99.5% methanol and 0.5% formic acid (B). The gradient elution program was as follows: 02 min, 100% A; 17 min, 90% A; 17.120 min, 100% A. Iopromide and its TPs were measured in a different HPLC run using an isocratic elution program which consisted of 90% A and 10% B for 20 min. A flow rate of 0.8 mL/min and column oven temperature of 50 °C was used for the measurements of ICM and TPs. The HPLC system was coupled to a 4000 Q Trap MS system (Applied Biosystems/MDS Sciex, Darmstadt, Germany) consisting of an electrospray ionization (ESI) source (operated in positive ionization mode). The source-dependent parameters were optimized for the ICM and their TPs. These parameters are summarized in Table S2 of the Supporting Information. Two mass transitions were optimized for the parent ICM and each TP for identification and confirmation purposes in MRM mode. The MRM transitions and compound-dependent parameters for the parent ICM and TPs as well as the surrogate standards are summarized in Table S3S7 of the Supporting Information. Method Validation. Recoveries for the analytical method were determined by spiking the analytes and surrogate standards into raw and treated wastewater, surface water, groundwater, and drinking water. The spiking concentrations were based on background concentrations of the parent ICM previously detected in different matrix types. For instance, raw and treated wastewater were spiked with concentrations of 0.1 and 1 μg/L, while surface water, groundwater, and drinking water were spiked at 0.05 and 0.1 μg/L. The recoveries were calculated according to eq 1, where Cspiked is the analyte concentration in the spiked sample,
Cnonspiked is the analyte concentration in the nonspiked sample, and Cinitial is the spiking concentration. Recovery ½% ¼
ðCspiked Cnonspiked Þ 100 ðCinitial Þ
ð1Þ
Variations from the mean values (n = 3) are given as 95% confidence intervals. Calibration samples were prepared by spiking 10 μL of a surrogate standard solution (20 ng/μL), the standard solutions of iohexol, iomeprol, and iopamidol, and nine isolated TP standards (iohexol TP687A, TP657, TP599; iomeprol TP701, TP643, TP629; iopamidol TP791, TP761, TP745) as well as iopromide and its seven isolated TPs (TP819, TP759, TP731B, TP729A, TP701A, TP701B, TP643) into a final volume of 1 mL of Milli-Q water. A detailed explanation of the isolation of TPs and the preparation of the standard solutions can be found elsewhere.20,22 The 12 point calibration curve ranged from 1 to 3000 ng/L. Linear and quadratic regressions were applied to the calibration curves with a weighing factor of 1/x. The ICM TPs which could not be isolated in sufficient quantities so far were quantified using either a calibration curve of the parent compound or an isolated TP showing the same detected MS fragment ions. Since there was no guarantee that this procedure was always accurate, these TPs were marked with an asterisk indicating that the data is semiquantitative. A summary of the analytes as well as the surrogate standards used for quantification (semiquantification) of each analyte are summarized in Table S8S9 of the Supporting Information. In addition, samples obtained from the soilwater batch systems spiked with ICM (i.e., 1 g/L) according to Schulz et al.20 and Kormos et al.21 were analyzed to verify the correctness of the retention times of ICM TPs having similar MRM transitions. 8725
dx.doi.org/10.1021/es2018187 |Environ. Sci. Technol. 2011, 45, 8723–8732
Environmental Science & Technology
ARTICLE
Figure 2. Concentration (nmol/L) of iopromide and its TPs (a) and iomeprol and its TPs (b) detected prior to denitrification and nitrification (sampling point 3), after denitrification and nitrification (sampling point 4), and in the final effluent (sampling point 5) at the WWTP.
In brief, samples (10 μL) from the batch systems were taken at specific time intervals and spiked with 10 μL of the surrogate standard (20 ng/μL) solution and diluted to 1 mL with Milli-Q water prior to LC tandem MS detection. The limits of quantification (LOQ) were defined as a signal/ noise (S/N) ratio of >10, which was individually determined in each environmental sample and calibration sample for each analyte. In any case, the LOQ values were always higher than the lowest calibration point.
’ RESULTS AND DISCUSSION Method Validation for ICM and TPs in Aqueous Matrixes. A LC tandem MS method was developed to determine the
occurrence of four nonionic ICM (iohexol, iomeprol, iopamidol, and iopromide) and 46 ICM TPs of the nonionic ICM in various aqueous matrixes. The recoveries for the parent ICM and selected TPs are summarized in Table 1. In general, the recoveries varied between 70 and 130% with 95% confidence intervals (n = 3) less than 30%. The LOQ for analytes detected in surface water, groundwater, and drinking water ranged from 1 to 3 ng/L. The LOQ in raw and treated wastewater ranged from 10 to 30 ng/L due to elevated background matrix composition and lower sample volume of 100 and 200 mL, respectively. The analytical methods were able to quantify ICM and 46 ICM TPs in a variety of aqueous matrixes with sufficient accuracy and sensitivity. Transformation of ICM and TPs in Municipal WWTPs. In order to facilitate the description of the results for so many TPs, 8726
dx.doi.org/10.1021/es2018187 |Environ. Sci. Technol. 2011, 45, 8723–8732
7.1 ( 1.2
1400 ( 150
(mmol/d)
0.69 ( 0.24 (10 ( 3) 12 ( 4
2.1 ( 0.3
(28 ( 4)
36 ( 5
nmol/L
8727
4.3; 4.3
(130; 130) 150; 150
nmol/L
(g/d) (mmol/d)
336 ( 48
325 ( 65
150; 150
90 ( 9
(mmol/d)
Eliminationi %
97 ( 19
(68 ( 13)
2.9 ( 0.6
2.0 ( 0.4
(36 ( 7) 52 ( 10
1.5 ( 0.3
1.1 ( 0.2
16 ( 3
(12 ( 3)
0.42 ( 0.07
0.30 ( 0.05
4.4 ( 0.6
(3.2 ( 0.4)
0.26 ( 0.03
0.19 ( 0.02
5.8 ( 1.1
0.32 ( 0.06 (4.2 ( 0.8)
0.23 ( 0.04
Phase III TPsd
84 ( 5
580 ( 140
(450 ( 110)
17 ( 4
13 ( 3
(410 ( 110) 530 ( 17
16 ( 0.5
12 ( 0.4
3500 ( 80
(2700 ( 60)
91 ( 2
71 ( 2
123 ( 4
(95 ( 3)
7.3 ( 0.3
5.7 ( 0.2
3100 ( 300
168 ( 16 (2400 ( 230)
131 ( 12
Iomeprol
129 ( 34
462 ( 80
(300 ( 50)
14 ( 2 (110 ( 30)
8(2 3.8 ( 1.0
(170 ( 60) 256 ( 92
7.6 ( 2.7
5(2
61 ( 8
(40 ( 5)
1.6 ( 0.2
1.0 ( 0.1
13 ( 5
(9 ( 3)
0.79 ( 0.30
0.5 ( 0.2
19 ( 4
1.1 ( 0.2 (13 ( 3)
0.7 ( 0.1
Phase II TPsf
3.1 ( 0.9
(95 ( 37) 122 ( 46
3.6 ( 1.4
3(1
58 ( 19
(45 ( 15)
1.5 ( 0.5
1.2 ( 0.4
4.5 ( 0.4
(3.5 ( 0.3)
0.27 ( 0.02
0.21 ( 0.02
14 ( 4
0.74 ( 0.19 (11 ( 3)
0.6 ( 0.2
Phase I TPse
90 ( 4
48 ( 3
(40 ( 3)
1.4 ( 0.1
1.2 ( 0.1
(31 ( 7) 37 ( 8
1.1 ( 0.2
0.9 ( 0.2
500 ( 14
(410 ( 10)
13 ( 0.3
11 ( 0.3
76 ( 8
(62 ( 6)
4.5 ( 0.5
3.7 ( 0.4
404 ( 42
22 ( 2 (330 ( 30)
18 ( 2
Iohexol
<0.01
<0.01
<0.01
<0.01
<0.03
Phase I TPs
20 ( 7
(13 ( 5)
0.59 ( 0.20
0.4 ( 0.1
(8 ( 3) 12 ( 4
0.34 ( 0.12
0.23 ( 0.08
6.4 ( 1.8
0.17 ( 0.05
4.1 ( 1.2
0.11 ( 0.03
13 ( 3
(8 ( 2)
0.77 ( 0.18
0.5 ( 0.1
3.6 ( 1.0
0.19 ( 0.05 (2 ( 1)
0.13 ( 0.03
Phase II TPsg
35 ( 14
675 ( 122
(530 ( 95)
20 ( 4
16 ( 3
(460 ( 20) 587 ( 27
17 ( 1
13 ( 1
1040 ( 70
(810 ( 60)
27 ( 2
21 ( 1
142 ( 10
(110 ( 8)
8.4 ( 0.6
6.6 ( 0.5
827 ( 78
45 ( 4 (640 ( 60)
35 ( 3
Iopamidol
3.0; 3.4
(2.4; 2.7)
0.09; 0.10
0.07; 0.08
(2.3 ( 0.7) 2.9 ( 0.9
0.09 ( 0.03
0.07 ( 0.02
<0.01
<0.01
<0.03
TPs
a Concentrations detected below the LOQ are given as
(260 ( 40)
(260 ( 50)
(120; 120)
(g/d)
8(1 9.9 ( 1.4
(250 ( 40) 323 ( 47
(270 ( 50) 335 ( 65 8(2
9.5 ( 1.4
9.6 ( 1.9
3.4; 3.4
4.3; 4.3
μg/L
nmol/L
Final Effluent
3.7; 3.8
μg/L
Postbiological treatment
41 ( 6
(32 ( 3)
7(1
76 ( 6
1500 ( 98
9.9 ( 1.9
(62 ( 8)
(1200 ( 80)
(g/d)
(mmol/d)
1.1 ( 0.2
0.82 ( 0.1
2.3 ( 0.8
(1.7 ( 0.6)
0.14 ( 0.05
0.10 ( 0.04
0.64 ( 0.16
0.04 ( 0.01 (0.5 ( 0.1)
0.03 ( 0.01
Phase II TPsc
8(2
1.6 ( 0.1 2.0 ( 0.2
31 ( 2
nmol/L
40 ( 3
Prebiological treatmenth μg/L
(mmol/d)
(g/d)
0.6 ( 0.2
1.7 ( 0.2
μg/L
Influent 2
0.31 ( 0.05 0.39 ( 0.06 (5.7 ( 0.9)
60 ( 6
76 ( 8 (1100 ( 120)
Phase I TPsb
μg/L
Iopromide
nmol/L (g/d)
Influent 1
WWTP 1
Sampling/Analyte
Table 2. Concentrations [μg/L] of ICM and Their TPs in the WWTP as well as Loads [g/d] and Elimination [%] of the Parent ICM and Their TPs during Wastewater Treatmenta
Environmental Science & Technology ARTICLE
dx.doi.org/10.1021/es2018187 |Environ. Sci. Technol. 2011, 45, 8723–8732
Environmental Science & Technology
ARTICLE
we defined for iopromide, iomeprol, and iohexol “phases” in which a group of TPs are summarized on the basis of the occurrence and degradation periods in the batch experiments described.20,22 The TP formation of iopromide was divided into three phases, while for iomeprol and iohexol only two phases were defined. Phase I refers to TPs formed at the beginning of the X-ray contrast media degradation. Those TPs which are still persistent after 100 d (degraded to less than 50% in relation to the maximum TP concentration) belong to last phase distinguished (e.g., phase III for iopromide and phase II for iomeprol and iohexol). Only for iopromide, we defined phase II comprising TPs which are degraded by more than 90% until day 100 and are formed not before day 15. In contrast to the other ICM TPs, the iopamidol TPs were not classified into different phases because Table 3. Mass Balances [%] of the Parent ICM and TPs during Treatment at the WWTP and after Infiltration of Rhine Water Biological wastewater
Rhine river water
Analytes
treatment [%]
infiltration well 1, 2, 3 [%]
Iopromide + TPs Iomeprol + TPs
54 32
64, 85, 78 95, 107, 101
Iohexol + TPs
14
97, 114, 106
Iopamidol + TPs
65
21, 14, 17
of the slower biodegradation of this nonionic ICM and the lack of information available to group the TPs into different phases. To elucidate whether similar transformation processes are occurring during wastewater treatment, four nonionic ICM and their 46 TPs were monitored in a WWTP. Five-day composite samples were collected from the WWTP at five sampling sites. ICM concentrations in influent stream 1 (sampling point 1), which is connected to a University hospital, ranged from 18 μg/L (iohexol) to 131 μg/L (iomeprol) and from 3.7 μg/L (iohexol) to 6.6 μg/L (iomeprol) in influent stream 2 (Figure 2 and Table 2). In total, a load of approximately 2 kg/d (285 000 PE) of nonionic ICM and their TPs were discharged into the river. Transformation of the nonionic ICM occurred mainly during biological treatment, while the other processes were of minor importance. The elimination of iohexol, iomeprol, and iopromide were >80% indicating high biological efficiency at this WWTP (Table 2). This is likely due to an elevated HRT of 60 h and an elevated SRT of 16 d. In the literature, the removal rates for ICM varied from no removal68 to more than 80%,13 indicating that a minimal SRT of approximately >8 d is required.11 Batt et al.19 reported that a SRT of 12 to 14 d was responsible for the high removal efficiency of iopromide during activated sludge treatment. In contrast to the other nonionic ICM, iopamidol was transformed in the WWTP to only 35% during biological treatment. This relatively low transformation efficiency is consistent with
Figure 3. Concentrations [nmol/L] of iopromide and TPs (a); iomeprol and TPs (b); iohexol and TPs (c); and iopamidol and TPs (d) in Rhine River water and bank filtrates (well 13). 8728
dx.doi.org/10.1021/es2018187 |Environ. Sci. Technol. 2011, 45, 8723–8732
14 ( 4.3 6.5 ( 1.0 9.6; 16 <3 144; 91 <1 <3 45 ( 12 <1 <1 <3
For duplicates, the results of the two samples were separated by semicolons. Concentrations detected below the LOQ are given as < LOQ value. b Specific water throughput of approximately 190 m3/kg GAC. c Specific water throughput of 40 m3/kg GAC. a
84; 50 73; 47 7; 20 91 ( 1 44 ( 15 9.5 ( 1.9 3.1 ( 1.7 2.9 ( 1.1 2.9 ( 1.1 25 ( 12 11 ( 5.1 <3 121 ( 27 122; 184 119 ( 44 13 ( 3.3 7.6; 6.3 <1 100 ( 22 38 ( 5.7 <3 24 ( 6 36 ( 13 59 ( 21 14 ( 3 5.4 ( 2.7 <1 43 ( 6 11 ( 5 <3
37; 48 15; 14 <1
116; 155 58; 72 41; 63 160 ( 48 89 ( 27 20 ( 7 3.1 ( 0.8 22; 28 20 ( 3 43 ( 5.4 21 ( 3.0 5(1 217 ( 54 204; 167 192 ( 45 57 ( 18 18; 27 10; 16 150 ( 28 83 ( 12 34 ( 6 33 ( 8 27, 49 23 ( 6 32 ( 6 13; 23 4(1 83 ( 11 41 ( 5 5(0
87 ( 32 77; 92 35 ( 8
34 ( 7 120 ( 40 118 ( 24 1230 ( 100 270 ( 68 272 ( 29 70 ( 9 14 ( 3 14 ( 1 57 ( 12 <3 <3 2460 ( 210 1407 ( 266 1660 ( 190 860 ( 120 8(1 7(4 1450 ( 160 23 ( 15 11 ( 2 200 ( 29 627 ( 87 510 ( 24 190 ( 33 20 ( 1 18 ( 1 120 ( 7 17 ( 4 21 ( 12
160 ( 33 17 ( 3 18 ( 2
24 ( 9 <1 9; 11 57 <1 39 ( 10 6; 9 27; 25 24; 23 7; 22 70 ( 19 11; 12 57; 59 55; 51 14; 27 <3 <3 <3 <3 <3 250 ( 80 6; 5 200; 230 200; 180 150; 370 <1 <1 <1 <1 <1 <3 <3 <3 <3 <3 99; 110 62 ( 4 106; 120 94; 88 17; 40 10; 10 <1 <1 <1 <1 <3 <3 <3 <3 <3
2; 2 <1 3; 2 <1 <1
ARTICLE
DWTP 1 bank filtrate groundwater mixed water after GAC-1b filtration after GAC-2c filtration DWTP 2 lake water bank filtrate after sand filtration DWTP 3 river water after ozonation after GAC filtration DWTP 4 river water after GAC filtration infiltrated groundwater and treated river water after slow sand filtration
TPs iopamidol phase II TPs iohexol phase II TPs phase I TPs iomeprol phase III TPs phase II TPs phase I TPs iopromide sampling site
Table 4. Concentrations [ng/L] and 95% Confidence Intervals (n = 3) of Iopromide, Iomeprol, Iohexol, Iopamidol, and Their TPs in Water Resources and after Certain Drinking Water Treatment Processes Collected at Four DWTPs in Germanya
Environmental Science & Technology
its slow biotransformation rates observed in aerobic soilwater and sedimentwater batch experiments.22 For iopromide, all 12 TPs identified in aerobic batch experiments were found at the WWTP, while 10 of 15 TPs and 3 of 10 TPs were found above the LOQ for iomeprol and iohexol, respectively. For iopamidol, only one of eight TPs was detected in the WWTP effluent. Presumably due to its branched side chains, iopamidol is more reluctant toward microbial transformation in municipal WWTPs than the other nonionic ICM. In total, 26 of 46 TPs formed in contact with soil and sediment during batch experiments were identified in the WWTP. Figure 2 illustrates the dominant TPs of iomeprol and iopromide formed during biological treatment. Concentrations as high as 9.1 nmol/L (5.7 μg/L) for iomeprol TP629 were detected. However, in addition to the phase I TPs, such as iomeprol TP805 and TP791 and iopromide TP817A and TP805A, phase II and III TPs (i.e., iomeprol TP629 and iopromide TP701A) were also found in the WWTP (Figure 2). This might be caused by higher transformation rates compared to batch experiments. However, it has to be noted that certain TPs are already present in the WWTP influent. This indicates that biotransformation has already been initiated prior to entering WWTPs. Transformation of ICMs after Rhine Water Infiltration into Groundwater. Figure 3 shows the change in the pattern of ICM and TPs after water infiltration into groundwater along the Rhine River at Urmitz. In the river samples, both the parent ICM and their TPs were present. However, iopromide, iomeprol, and iohexol concentrations were reduced by >80% during bank infiltration, while concentrations of the TPs formed at the end of the biotransformation pathways (e.g., iopromide phase III TPs, iomeprol phase II TPs) increased significantly even in the well closest to the Rhine. Maximum concentrations of individual TPs after river water infiltration were found for iopromide TP701A and iopmeprol TP643 and iohexol TP599 with up to 0.55 nmol/L (0.35 μg/L) for iomeprol TP 643 (see Supporting Information). The reduced concentrations of iopromide, iomeprol, and iohexol in the groundwater after infiltration are accompanied with a significant increase of the respective TP concentrations. The same pattern was not observed for iopamidol and its TPs. The concentration of iopamidol was reduced after bank filtration from 0.5 to 0.1 nmol/L, but the TP concentrations did not increase. A comparison of the TP pattern with the results of the batch experiments20,22 can imply the following conclusions. (i) The transformation of the nonionic ICM during river water infiltration seems to be faster than in the watersoil and watersediment batch systems. For example, iopromide was already transformed to phase III TPs within 20 days when considering an infiltration velocity of 13 m/d, while iopromide took more than 100 days in the batch experiments to attain a similar pattern. (ii) The most persistent TPs of iopromide, iohexol, and iomeprol, formed in the batch systems at the end of the biotransformation pathways, were also present in groundwater after infiltration. However, the precursor TPs (e.g., iopromide phase I and II TPs) were missing. (iii) Iopamidol biotransformation during river water infiltration cannot be sufficiently explained by the reactions occurring in the batch systems, since it is likely that processes are occurring which have not been identified in watersoil and watersediments batch systems. (iv) The mass balance of iopromide, iohexol, and iomeprol are almost complete (see Table 3), indicating that the biotransformation pathways observed in the watersoil and watersediment batch systems can be transferred to situations along the Rhine River near Koblenz. 8729
dx.doi.org/10.1021/es2018187 |Environ. Sci. Technol. 2011, 45, 8723–8732
Environmental Science & Technology
ARTICLE
Table 5. Concentrations [ng/L] and 95% Confidence Intervals (n = 3) of Four Nonionic ICM and Their TPs in Drinking Water Collected at Four Water Treatment Facilities in Germanya Iohexol
Iohexol
Iohexol
Iomeprol
Iomeprol
Iomeprol
Iomeprol
Iomeprol
Iomeprol
TP745
TP657
TP599
Iomeprol
TP761b
TP745b
TP701
TP687b
TP643
TP629
<3 <3
<3 8(1
8(1 7(1
23 ( 1 7(0
<3 11 ( 2
<1 7(4
<3 380 ( 60
7(1 120 ( 20
<1 500 ( 60
170 ( 40 250 ( 20
2(1 400 ( 40
DWTP3 5 ( 2
23 ( 10
<3
8(2
34 ( 6
<3
12 ( 4
<1
40 ( 4
20 ( 3
119 ( 37
<1
13 ( 1
8(2
<3
<3
<3
21 ( 2
35 ( 12
53 ( 12
Iopamidol
Iopamidol
Iopamidol
TP791
TP773b
TP761
Iohexol DWTP1 DWTP2 DWTP4
<1
Iopromide
9.4; 8.1
Iopromide
Iopromide
Iopromide
Iopromide
Iopromide
Iopromide
TP805Ab
TP759
TP729A
TP701A
TP701B
TP643
Iopamidol
<3
<3
<3
<3
40 ( 11
<1
19 ( 4
20; 15
<1
<1
<1
DWTP2 21 ( 12
18 ( 1
260 ( 11
12 ( 1
170 ( 4
40 ( 8
40 ( 1
270 ( 30
42 ( 4
30 ( 8
29 ( 9
5(0 <3
4(1 <10
2(0 3(2
<10 <10
10 ( 3 22 ( 6
<3 7(2
11 ( 3 13 ( 3
20 ( 7 7(1
<3 <3
<3 <3
8; 6 <3
DWTP1 DWRP3 DWTP4
a Concentrations detected below the LOQ are given as
Mass Balances of ICM and TPs in the WWTP and after Rhine Water Infiltration. Mass balances [%] were calculated
Fate and Occurrence of ICM and TPs during Different Treatment Processes at DWTPs. The results from the DWTPs
according to eq 2 for biological treatment at the WWTP and for the infiltration of Rhine water at Urmitz, into groundwater. Parent ICMafter and parent ICMbefore are the loads (mmol/d) of the parent ICM before and after biological treatment at the WWTP or the average concentration (nmol/L) in the Rhine water at Koblenz from September to November 2009 and in the groundwater after infiltration. TPsafter and TPsbefore refer to the sum of TP loads (mmol/d) of one particular ICM before and after biological treatment at the WWTP or the sum of TPs concentration (nmol/L) of one particular ICM in river water and groundwater.
suggest that common drinking water treatment processes, such as bank filtration, GAC filtration, ozonation, slow sand filtration, or flocculation, do not effectively remove ICM TPs from drinking water supplies. Four DWTPs were monitored to investigate the removal of ICM and their TPs via different drinking water treatment processes. The results are summarized in Table 4 and Table S10, Supporting Information. After bank filtration, the concentrations of the parent ICM were significantly diminished (Figure S1, Supporting Information, DWTP2); however, the concentration of iopromide and iopamidol TPs formed at the end of the microbial transformation pathway increased. For example, the biotransformation of iopromide resulted in an increase of iopromide phase III TPs from 201 ( 29 to 627 ( 87 ng/L at DWTP2, and the sum of iomeprol phase II TPs was as high as 1400 ( 300 ng/L in bank filtrates and 1700 ( 200 ng/L in drinking water. Iohexol TP657 and TP599, iomeprol TP687, TP643, and TP629, and iopromide TP701A and TP643 were the dominant ICM TPs present in drinking water (Table 5). The removal efficiency of these polar ICM TPs by GAC (applied at DWTP1, 3 and 4) was very limited. The removal of the nonionic ICM by GAC was similar to results reported by Seitz et al.12 and Snyder et al.24 GAC filtration has shown to be a very efficient for antiepileptics, antiphlogistics, and antibiotics as well as other organic pollutants, such as pesticides and estrogens.2427 However, most polar ICM TPs of this study formed at the end of biotransformation pathways (i.e., iopromide phase III TPs, iomeprol phase II TPs, and iohexol phase II TPs) were not removed. However, it was found that less preloaded GAC filters (specific water throughput of 40 m3/kg at DWTP1) can partly remove certain TPs (Table 4). Ozonation, only used at DWTP3, slightly reduced the concentrations of ICM and TPs. As previously reported in another study,16 ozonation causes the formation of stable ICM oxidation products which were not monitored in our study. Slow sand filtration used at DWTP2 and DWTP4 showed no significant removal of ICM and their TPs. Flocculation used at DWTP3 and DWTP4 was not individually investigated, but the overall removal exhibited its inefficiency.
Mass balance½% ¼
ðParent ICMafter þ TPsafter Þ 100 ðParent ICMbefore þ TPsbefore Þ
ð2Þ The sum of iopromide and its TPs in WWTP effluent corresponds to about 54% of the sum of iopromide and TPs prior to biological treatment at the WWTP (Table 3). Hence, 46% of biotransformed iopromide cannot be explained by formation of the TPs identified. Lower mass balances of 32% and 14% were determined for iomeprol and iohexol, respectively. For iohexol, only phase II TPs (i.e., TP687A, TP657, and TP599) were detected. The incomplete mass balances might indicate that in municipal WWTPs other TPs are formed and other transformation pathways have to be considered compared to what was observed in watersoil and watersediments batch systems. Currently, it is unknown whether the anaerobic/anoxic conditions occurring in WWTPs are responsible for the incomplete mass balances (refer to Table 3) or if other aerobic TPs are formed which have not been identified so far. The calculated mass balances after infiltration of river water were more complete than those obtained for the WWTP (Table 3). They varied between 64 and 114% for iopromide, iomeprol, and iohexol. It can be assumed that the majority of the parent ICM led to the formation of known TPs. The mass balance of iopamidol was only found between 14 and 21%, which suggests that further TPs are formed.
8730
dx.doi.org/10.1021/es2018187 |Environ. Sci. Technol. 2011, 45, 8723–8732
Environmental Science & Technology As a consequence of the limited removal during drinking water treatment, the four nonionic ICM and their TPs were found in finished drinking water of selected German DWTPs. In general, the concentrations varied from 1 to 50 ng/L (Table 5). However, elevated concentrations of up to 500 ng/L (iomeprol TP687) were detected. Currently, it is unclear whether these ICM TPs have a toxicological impact on human health. However, it cannot be excluded that ICM TPs may react with strong oxidants (i.e., chlorine) during drinking water disinfection and form oxidative byproducts16 as well as iodinated disinfection byproducts30,31 which are known to be highly toxic.
’ ASSOCIATED CONTENT
bS
Supporting Information. Details about the selected sampling locations and the LC tandem MS method. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Telephone: +49 261 1306 5560. Fax: +49 261 1306 5363. E-mail:
[email protected].
’ ACKNOWLEDGMENT We would like to thank the personnel at the WWTPs and DWTPs for participating in this study and providing samples. In addition, the authors acknowledge D. A. Urban (BfG) for developing the LC tandem MS method, M. P. Schl€usener and G. Fink (BfG) for their assistance during surface water sampling, and C. Prasse for reviewing the manuscript. Financial support was provided by the Marie Curie Research Training Network KEYBIOEFFECTS (MRTN-CT-2006-035695) and EU-Project NEPTUNE (Project no. 036845), which are both funded by the European Commission within the Sixth Framework Programme and are gratefully acknowledged. In addition, we would like to thank Bayer Schering Pharma (Berlin, Germany) for supplying the surrogate standard (DMI). ’ REFERENCES (1) Christiansen, C. X-ray contrast mediaan overview. Toxicology 2005, 209 (2), 185–187. (2) Hirsch, R.; Ternes, T. A.; Lindart, A.; Haberer, K.; Wilken, R.-D. A sensitive method for the determination of iodine containing diagnostic agents in aqueous matrices using LC-electrospray-tandem-MS detection. Fresenius J. Anal. Chem. 2000, 366 (8), 835–841. (3) Steger-Hartmann, T.; L€ange, R.; Schweinfurth, H.; Tschampel, M.; Rehmann, I. Investigations into the environmental fate and effects of iopromide (ultravist), a widely used iodinated X-ray contrast medium. Water Res. 2002, 36 (1), 266–274. (4) Perez, S.; Barcelo, D. Fate and occurrence of X-ray contrast media in the environment. Anal. Bioanal. Chem. 2007, 387 (4), 1235–1246. (5) Busetti, F.; Linge, K. L.; Blythe, J. W.; Heitz, A. Rapid analysis of iodinated X-ray contrast media in secondary and tertiary wastewater by direct injection liquid chromatography-tandem mass spectrometry. J. Chromatogr., A 2008, 1213 (2), 200–208. (6) Ternes, T. A.; Hirsch, R. Occurrence and behavior of X-ray contrast media in sewage facilities and the aquatic environment. Environ. Sci. Technol. 2000, 34 (13), 2741–2748. (7) Carballa, M.; Omil, F.; Lema, J. M.; Llompart, M.; García-Jares, C.; Rodriguez, I.; Gomez, M.; Ternes, T. Behavior of pharmaceuticals, cosmetics and hormones in a sewage treatment plant. Water Res. 2004, 38 (12), 2918–2926.
ARTICLE
(8) Clara, M.; Strenn, B.; Gans, O.; Martinez, E.; Kreuzinger, N.; Kroiss, H. Removal of selected pharmaceuticals, fragrances and endocrine disrupting compounds in a membrane bioreactor and conventional wastewater treatment plants. Water Res. 2005, 39 (19), 4797–4807. (9) Putschew, A.; Wischnack, S.; Jekel, M. Occurrence of triiodinated X-ray contrast agents in the aquatic environment. Sci. Total Environ. 2000, 255 (1), 129–134. (10) Schittko, S.; Putschew, A.; Jekel, M. Bank filtration: a suitable process for the removal of iodinated X-ray contrast media? Water Sci. Technol. 2004, 50 (5), 261–268. (11) Joss, A.; Zabczynski, S.; G€obel, A.; Hoffmann, B.; L€ offler, D.; McArdell, C. S.; Ternes, T. A.; Thomsen, A.; Siegrist, H. Biological degradation of pharmaceuticals in municipal wastewater treatment: proposing a classification scheme. Water Res. 2006, 40 (8), 1686–1696. (12) Seitz, W.; Jiang, J.-Q.; Weber, W. H.; Lloyd, B. J.; Maier, M.; Maier, D. Removal of iodinated X-ray contrast media during drinking water treatment. Environ. Chem. 2006, 3 (1), 35–39. (13) Ternes, T. A.; Bonerz, M.; Hermann, N.; Teiser, B.; Andersen, H. R. Irrigation of treated wastewater in Braunschweig, Germany: An option to remove pharmaceuticals and musk fragrances. Chemosphere 2007, 66 (5), 894–904. (14) Doll, T. E.; Frimmel, F. H. Kinetic study of photocatalytic degradation of carbamazepine, clofibric acid, iomeprol and iopromide assisted by different TiO2 materialsdetermination of intermediates and reaction pathways. Water Res. 2004, 38 (4), 955–964. (15) Perez, S.; Eichhorn, P.; Celiz, M. D.; Aga, D. S. Structural characterization of metabolites of the x-ray contrast agent iopromide in activated sludge using ion trap mass spectrometry. Anal. Chem. 2006, 78 (6), 1866–1874. (16) Seitz, W.; Jiang, J.-Q.; Schulz, W.; Weber, W. H.; Maier, D.; Maier, M. Formation of oxidation by-products of the iodinated X-ray contrast medium iomeprol during ozonation. Chemosphere 2008, 70 (7), 1238–1246. (17) Zwiener, C.; Glauner, T.; Sturm, J.; W€orner, M.; Frimmel, F. H. Electrochemical reduction of the iodinated contrast medium iomeprol: iodine mass balance and identification of transformation products. Anal. Bioanal. Chem. 2009, 395 (6), 1885–1892. (18) Kalsch, W. Biodegradation of the iodinated X-ray contrast media diatrizoate and iopromide. Sci. Total Environ. 1999, 225 (12), 143–153. (19) Batt, A. L.; Kim, S.; Aga, D. Enhanced biodegradation of iopromide and trimethoprim in nitrifying sludge. Environ. Sci. Technol. 2006, 40 (23), 7367–7373. (20) Schulz, M.; L€offler, D.; Wagner, M.; Ternes, T. A. Transformation of the X-ray contrast medium iopromide in soil and biological wastewater treatment. Environ. Sci. Technol. 2008, 42 (19), 7207–7217. (21) Kormos, J. L.; Schulz, M.; Wagner, M.; Ternes, T. A. Multistep approach for the structural identification of biotransformation products of iodinated X-ray contrast media by liquid chromatography/hybrid triple quadrupole linear ion trap mass spectrometry and 1H and 13C nuclear magnetic resonance. Anal. Chem. 2009, 81 (22), 9216–9224. (22) Kormos, J. L.; Schulz, M.; Kohler, H.-P. E.; Ternes, T. A. Biotransformation of selected iodinated X-ray contrast media and characterization of microbial transformation pathways. Environ. Sci. Technol. 2010, 44, 4998–5007. (23) Perez, S.; Eichhorn, P.; Ceballos, V.; Barcelo, D. Elucidation of phototransformation reactions of the X-ray contrast medium iopromide under simulated solar radiation using UPLC-ESI-QqTOF-MS. J. Mass Spectrom. 2009, 44 (9), 1308–1317. (24) Snyder, S. A.; Adham, S.; Redding, A. M.; Cannon, F. S.; DeCarolis, J.; Oppenheimer, J.; Wert, E. C.; Yoon, Y. Role of membranes and activated carbon in the removal of endocrine disruptors and pharmaceuticals. Desalination 2007, 202 (13), 156–181. (25) Ternes, T. A.; Meisenheimer, M.; McDowell, D.; Sacher, F.; Brauch, H.-J.; Haist-Gulde, B.; Preuss, G.; Wilme, U.; Zulei-Seibert, N. Removal of pharmaceuticals during drinking water treatment. Environ. Sci. Technol. 2002, 36 (17), 3855–3863. (26) Stackelberg, P. E.; Gibs, J.; Furlong, E. T.; Meyer, M. T.; Zaugg, S. D.; Lippincott, R. L. Efficiency of conventional drinking-watertreatment processes in removal of pharmaceuticals and other organic compounds. Sci. Total Environ. 2007, 377 (23), 255–272. 8731
dx.doi.org/10.1021/es2018187 |Environ. Sci. Technol. 2011, 45, 8723–8732
Environmental Science & Technology
ARTICLE
(27) Vieno, N. M.; H€arkki, H.; Tuhkanen, T.; Kronberg, L. Occurrence of pharmaceuticals in river water and their elimination in a pilotscale drinking water treatment plant. Environ. Sci. Technol. 2007, 41 (14), 5077–5084. (28) Lesjean, B.; Rosenberger, S.; Laabs, C.; Jekel, M.; Gnirss, R.; Amy, G. Correlation between membrane fouling and soluble/colloidal organic substances in membrane bioreactors for municipal wastewater treatment. Water Sci. Technol. 2005, 51 (67), 1–8. (29) Xiong, Y.; Liu, Y. Biological control of microbial attachment: a promising alternative for mitigating membrane biofouling. Appl. Microbiol. Biotechnol. 2010, 86 (3), 825–837. (30) Richardson, S. D.; Fasano, F.; Ellington, J. J.; xCrumley, F. G.; Buettner, K. M.; Evans, J. J.; Blount, B. C.; Silva, L. K.; Waite, T. J.; Luther, G. W.; McKague, A. B.; Miltner, R. J.; Wagner, E. D.; Plewa, M. J. Occurrence and mammalian cell toxicity of iodinated disinfection byproducts in drinking water. Environ. Sci. Technol. 2008, 42 (22), 8330–8338. (31) Duirk, S. E.; Lindell, C.; Cornelison, C. C.; Kormos, J.; Ternes, T. A.; Attene-Ramos, M.; Osiol, J.; Wagner, E. D.; Plewa, M. J.; Richardson, S. D. Formation of toxic iodinated disinfection by-products from compounds used in medical imaging. Environ. Sci. Technol. 2011, 45, 6845–6854.
8732
dx.doi.org/10.1021/es2018187 |Environ. Sci. Technol. 2011, 45, 8723–8732
ARTICLE pubs.acs.org/est
Quantifying Differences in the Impact of Variable Chemistry on Equilibrium Uranium(VI) Adsorption Properties of Aquifer Sediments Deborah L. Stoliker,†,* Douglas B. Kent,† and John M. Zachara‡ † ‡
U.S. Geological Survey, 345 Middlefield Rd. MS 496, Menlo Park, California 94025, United States Pacific Northwest National Laboratory, Richland, Washington, United States
bS Supporting Information ABSTRACT: Uranium adsorptiondesorption on sediment samples collected from the Hanford 300-Area, Richland, WA varied extensively over a range of field-relevant chemical conditions, complicating assessment of possible differences in equilibrium adsorption properties. Adsorption equilibrium was achieved in 5001000 h although dissolved uranium concentrations increased over thousands of hours owing to changes in aqueous chemical composition driven by sediment-water reactions. A nonelectrostatic surface complexation reaction, >SOH + UO22+ + 2CO32- = >SOUO2(CO3HCO3)2, provided the best fit to experimental data for each sediment sample resulting in a range of conditional equilibrium constants (logKc) from 21.49 to 21.76. Potential differences in uranium adsorption properties could be assessed in plots based on the generalized massaction expressions yielding linear trends displaced vertically by differences in logKc values. Using this approach, logKc values for seven sediment samples were not significantly different. However, a significant difference in adsorption properties between one sediment sample and the fines (<0.063 mm) of another could be demonstrated despite the fines requiring a different reaction stoichiometry. Estimates of logKc uncertainty were improved by capturing all data points within experimental errors. The massaction expression plots demonstrate that applying models outside the range of conditions used in model calibration greatly increases potential errors.
’ INTRODUCTION Obtaining reliable simulations of contaminant transport in groundwater requires not only understanding contaminant adsorption properties of aquifer sediments but whether, and the extent to which, these properties vary spatially.1 For contaminants whose sorption is controlled by simple partitioning, spatial variability is often assessed by determining empirical constants such as distribution coefficients (Kd).1,2 The extent of adsorption of many inorganic contaminants varies with chemical conditions and, therefore, any parameters used to quantify the extent of adsorption must account for chemical variability.3 This complicates assessment of spatial variations in adsorption properties of the sediments themselves. Previous work to assess spatial variability has focused on relating differences in contaminant adsorption to measurable sediment properties such as extractable metal concentrations, surface area, mineralogy, etc.48 The surface complexation modeling (SCM) approach has been shown to be a simple and effective approach for describing the impact of variable chemical conditions on metal and metalloid ion adsorption on soils and sediments.913 Adsorption is represented as the minimum number of chemical reactions between the adsorbing solute(s), surface sites, and other species known to influence the extent of adsorption over a relevant range of chemical conditions.14 The SCM approach should also be capable of assessing differences in adsorption properties of aquifer sediments by evaluating similarities or differences in r 2011 American Chemical Society
the conditional equilibrium constant (Kc) if uncertainty can be determined adequately. Uranium(VI) adsorption is well suited for an examination of the utility of the SCM approach to assess differences in adsorption properties of site-specific sediments.9 The SCM approach has been used in reactive transport simulations of U(VI) and results are sensitive to the logKc value describing U(VI) adsorption.11,1518 In the alkaline pH range, U(VI) forms a variety of aqueous complexes with carbonate and calcium whose predominance changes with changing pH, and carbonate and calcium concentrations.1822 Uranium(VI)carbonato surface complexes are known to form at some mineral surfaces.23 Thus, the extent of U(VI) adsorption varies considerably with variable carbonate concentration, pH, and calcium concentration.20,2426 This chemical variability greatly complicates distinguishing differences in adsorption properties of sediments. The objective of this paper was to demonstrate a new approach for quantifying differences in U(VI) adsorption properties of sediment samples. To accomplish this, the time required to achieve equilibrium adsorption was determined despite longterm increases in uranium concentrations. Large differences Received: January 3, 2011 Accepted: August 19, 2011 Revised: August 18, 2011 Published: September 16, 2011 8733
dx.doi.org/10.1021/es202677v | Environ. Sci. Technol. 2011, 45, 8733–8740
Environmental Science & Technology
ARTICLE
Table 1. Elevation, Surface Area, Total and Adsorbed Uranium, and LogKc Based on Reaction 1 for IFRC Sediments and the Weighted Average Used in Modeling “All Data”a elevation
surface area <2 mm
surface area finesb
(m above sea level)
2
(m /g)
2
(m /g)
fines
(nmol/g)
(nmol/g)
logK
stdev g
226 227
105.8107.3 106.1107.3
11.7 13.4
14.6 18.4
21.1 10.2
12.2 26.2
2.5 11.5
21.59 ( 0.18 21.76 ( 0.19
0.05 0.06
230
106.4107.3
13.3
16.4
18.1
46.5
18.3
21.68 ( 0.12
0.08
231
106.5107.3
12.1
19.7
6.1
10.8
3.1
21.76 ( 0.24
0.05
331
106.7107.6
11.7
16.5
10.1
16.1
8.2
21.57 ( 0.13
0.07
332
106.3107.3
11.6
13.9
12.5
21.2
8.5
21.49 ( 0.12
0.05
SWC
102.6108.0
14.1
24.0
27.2
12.2
4.4h
21.61 ( 0.19
0.05
all datai
102.6108.0
12.6j
18.2j
15.8j
18.9j
7.5j
21.64 ( 0.36
0.03
sample
a
weight % of b,c
total Ud
adsorbed Ue
logKc FITEQL c,f
For errors in surface area and uranium measurements see SI Table S2. Fines refer to the <0.063 mm size fraction. c Determined by wet sieving. d Determined by gamma-spectrometry on <2 mm sediment samples. e Determined by bicarbonate extraction on <2 mm sediment samples. f Reaction 1, logKc uncertainty calculated by manually fitting data, see text. g Reaction 1, standard deviation calculated by FITEQL propagation of errors. h SWC <0.063 mm Uads =9.76 nmol/g. i 226, 227, 230, 231, 331, 332, and SWC <2 mm data modeled together. j Weighted average based on experimental data used to obtain fit.
in U(VI) adsorption, owing to variable chemical conditions, were accounted for by examining trends in experimental U(VI) adsorption between different samples. Uncertainties in equilibrium adsorption model parameters, here, the conditional equilibrium constant for the adsorption reaction, were quantified. Uncertainties resulting from imperfect descriptions of the impact of variable chemical conditions were accounted for in addition to uncertainties propagating from random analytical errors.
’ MATERIALS AND METHODS Site and Sample Collection. Six individual wells located across the ∼2100 m2 Hanford 300-Area Integrated Field Research Challenge (IFRC) site were selected for study due to their spatial distribution. Close neighbor wells or cluster wells (226 and 227; 230 and 231; 331 and 332) are spaced approximately 2 m apart while the clusters are approximately 30 m from each other (see site map in Supporting Information (SI), Figure S1). Grab samples of sediment were collected during the drilling of wells, discretized to 2 foot depth intervals, air-dried, and dry sieved to <2 mm. A continuous core from well 25 was examined to define the seasonally saturated smear-zone in which higher levels of adsorbed uranium accumulate due to repeated wetting and drying cycles.2729 Once the smear-zone elevation was defined, material from this elevation range was isolated for each of the six individual wells mentioned above (See Table 1 for specific elevations). For five of the six individual wells (226, 230, 231, 331, and 332), in addition to the smear-zone sample, a sample was also collected from the deeper permanently saturated aquifer (97103 m above sea level) to establish background total and adsorbed uranium concentrations. Finally, the <2 mm size fraction sediment collected from the smear-zone of 31 wells across the site (See SI Figure S1) was mixed together into a site-wide composite (SWC). Sediment Characterization. Each smear-zone sediment sample was wet-sieved to determine particle size distribution. Samples were also dry sieved to isolate the <0.063 mm size fraction (herein referred to as “fines”) for further study. Surface areas were measured, in duplicate, on both the original <2 mm and dry sieved fines of each sample using N2 adsorption at 77.3 K (Micromeritics Tristar 3000). Samples (∼2 g) were degassed at 105 °C for 35 days and the 5-point BET method applied to
b
adsorption curves with C parameters between 100 and 300.30 X-ray diffractograms were collected on the fines using a Phillips X-ray diffractometer between 4 and 70 degrees 2 theta. Total uranium was measured via nondestructive gammaspectrometry. 238U was determined by measurement of the 234 Th daughter 63 keV gamma ray emission line on each <2 mm sample, assuming secular equilibrium in the sample.31 Adsorbed uranium was determined via bicarbonate extraction. The <2 mm sediment (50 g/L) and fines of SWC (15 g/L) were reacted with a solution of 2.8 mM Na2CO3 and 14.4 mM NaHCO3 at pH 9.5 and 20 meq/L alkalinity. Experimental studies on Hanford and other sediments have shown that determinations of adsorbed uranium by bicarbonate extraction are equal to those determined by 233U isotopic exchange.20,32 Duplicate reaction bottles were mixed on an orbital shaker table and intermittently subsampled by filtering (0.45 μm, PVDF) 3 mL of supernatant after settling. Samples were collected up to 9400 h (56 weeks) and U(VI) concentrations measured using a Chemchek KPA-11 kinetic phosphorescence analyzer (KPA). Major solution cations were measured via inductively coupled plasma-optical emission spectroscopy (ICP-OES) using a Thermo Scientific iCAP 6500. Citrate bicarbonate dithionite (CBD) extractions were performed in duplicate on both the <2 mm (5 g) and fines (2 g) of each sample per the modified method detailed by Loppert and Inskeep,1996 at 80 °C and pH 7.33 The <2 mm size fraction was also ground to pass through a 0.063 mm sieve (2 g) and extracted with CBD. Ammonium oxalate extractions in the dark (AOD) were carried out as described in Chao and Zhou, 1983 in duplicate on the <2 mm samples (5 g) and fines (1 g) at ambient temperature and pH 3.34 Aqueous iron concentrations were measured by ICP-OES. Equilibrium Adsorption Experiments. A series of five artificial groundwaters (AGWs) were designed to mimic the 300Area groundwater during its seasonal fluctuations due to mixing of river water and groundwater. These waters have been described elsewhere and compositions are given in SI Table S1.18 Briefly, solution calcium (0.5 mM), magnesium (0.5 mM), and potassium (0.15 mM) concentrations were maintained while alkalinity (NaHCO3) was varied between 0.5 and 6 meq/L in equilibrium with the atmosphere (pH 7.88.1). Sodium salts (nitrate and sulfate) were added as necessary to maintain an ionic strength of 10 mM. The <2 mm size fraction of each material as well as 8734
dx.doi.org/10.1021/es202677v |Environ. Sci. Technol. 2011, 45, 8733–8740
Environmental Science & Technology the SWC fines were mixed with AGW in polycarbonate centrifuge tubes on an end-overend rotator at ambient temperature. All experiments were performed in duplicate at either 100 g/L (AGW 1, 2, 3) or 1000 g/L (AGW 3, 4, 5) and tubes sacrificesampled at various time points up to 7000 h (42 weeks). While experiments began at atmospheric pCO2, no attempt was made to control the pCO2 in the head space of each tube. At the time of sampling, the lid was quickly removed and replaced with one outfitted with a pH electrode to determine suspension pH while minimizing changes in pCO2. The tube was then centrifuged, supernatant filtered, and the solution uranium and cation concentrations measured. Alkalinity was determined on the filtered solution by titration with standardized sulfuric acid (Fisher Scientific). Modeling. Conditional equilibrium constants (Kc) for surface complexation adsorption reactions with no electrostatic term were computed from the experimental data by nonlinear, leastsquares optimization using FITEQL.35 The adsorbed U(VI) concentration at equilibrium was calculated from the original adsorbed U(VI) concentration for each sample, determined using bicarbonate extraction (Table 1), minus the dissolved concentration. The site concentration was calculated from the solid liquid ratio, specific surface area (Table 1), and an assumed site density of 3.84 μmol/m2.3 In the generalized composite approach to surface complexation modeling, an average, generic surface site (SOH) is used to represent adsorption onto the complex assemblage of minerals in field samples. Carbonate concentrations were computed from measured pH and alkalinity values. These values, along with the concentrations of calcium, magnesium, sodium, potassium, and sulfate, were used to compute aqueous U(VI) speciation in the fitting. Nitrate concentrations were calculated to achieve charge balance. Experimental errors used in the fitting procedure were 10% for H+, 3% for carbonate, 10% for adsorbed U(VI), 3% for dissolved U(VI), 10% for site concentration, and 5% for all other dissolved concentrations. The basis for each error assumption is described in SI Table S2. Aqueous U(VI) speciation was computed using previously reported thermodynamic data.15 Experimental data collected at 2000 h (12 weeks) of equilibration were used in the fitting. Each individual sample was modeled separately and the 2000 h data for all samples were then modeled together to produce what is referred to as the “all data” model.
’ RESULTS AND DISCUSSION Sediment Characterization. Surface areas of the <2 mm sediment varied between 11.5 and 14.1 m2/g. The average specific surface area, calculated by averaging surface area values associated with each data point used in the “all data” fitting procedure, was 12.6 m2/g (Table 1). The surface area of the fines (<0.063 mm) ranged from 13.9 to 24.0 m2/g and constituted 627% by weight of the <2 mm material. The fine-grained material accounts for 1026% of the surface area of each individual well sample and 46% of the surface area of the site-wide composite (SWC). Total uranium varied between 11 and 47 nmol/g for the smear-zone sediment samples (Table 1). The highest and lowest total uranium concentrations occurred in sediment collected 2 m apart in one of the three well clusters (230 and 231). Similar variations over short spatial distances were observed for another well cluster (226 and 227) while the third showed less variation between samples (331 and 332). The deep saturated-
ARTICLE
zone samples contained much less total uranium with an average of 7.6 nmol/g, near the average crustal abundance of uranium of 812 nmol/g. Previous spectroscopic evaluations of sediment from the Hanford formation of the 300-Area near this field site have shown only the presence of oxidized hexavalent uranium U(VI).29,36,37 Therefore, uranium described herein is assumed to be U(VI). Adsorbed U(VI) ranged from 2.5 to 18.3 nmol/g accounting for 2061% of the total uranium in each sample. During bicarbonate extraction, dissolved concentrations of uranium increased rapidly for the first 24 h and then more slowly out to 1700 h after which concentrations remained steady. The slow progression toward complete desorption is consistent with previous findings of mass-transfer diffusion limitations on contaminant release at this site.15,16,38,39 A positive linear correlation exists between total and adsorbed uranium. Similar to the total uranium concentrations, the highest and lowest adsorbed uranium concentrations occurred in sediment collected 2 m apart (230 and 231, see SI Figure S2). No discernible pattern is evident between extractable uranium and sample location. The spatial variations in adsorbed uranium (and total uranium) observed in the smearzone were not evident in the deep saturated-zone samples. For the deep saturated-zone samples, the average adsorbed uranium was 1.1 nmol/g; representing between 4.3 and 12.6% of the total uranium. Adsorbed uranium in samples from the deep saturated-zone may represent natural background levels of adsorbed uranium similar to other field sites such as Rifle, CO at 1.4 nmol/g and Naturita, CO at 0.87 nmol/g.7,26 Mineralogy and iron extractions of sediment samples suggest similar solid-phase compositions across the field site. XRD of fines showed each sample was comprised primarily of quartz and plagioclase with moderate contributions from pyroxene and minor (<5%) abundance of poorly crystalline clays composed of illite, kaolinite, and extremely poorly crystalline smectite. CBD extractions of the <2 mm smear-zone sediments yielded between 3.6 and 6.2 μmol/m2 iron while AOD released between 5.5 and 10.3 μmol/m2 (SI Table S3). Extractions of the fines resulted in approximately two times more iron released by each method when normalized to surface area. Details of the limitations of these extractions can be found in the SI. Though small differences in extractable iron were observed between samples, no correlations could be found between adsorbed uranium or surface area and the iron extracted by either AOD or CBD. In addition, no significant correlation could be found between the amount of adsorbed uranium on a sediment and its surface area or weight percentage of fines. Equilibrium Adsorption Experiments. Desorption experiments were carried out at 100 g/L for groundwaters with low alkalinities (0.52 meq/L). However, at higher alkalinities (>2 meq/L), 100% of the adsorbed uranium was extracted at this suspension density within 700 h. Therefore, higher alkalinity experiments were conducted at 1000 g/L. At low suspension densities (100 g/L), dissolved uranium concentrations increased rapidly for the first 24 h and then increased at a slower rate for duration of the experiments (Figure 1a). In these systems, between 52 and 94% of the adsorbed uranium was released after 2000 h. Changes in pH, alkalinity, and calcium concentrations were observed throughout the duration of the experiments (see Figure 1a). The initial chemical changes (2472 h) can be attributed to dissolution of aquifer salts, which precipitated when the sediment was dried, and ion exchange reactions.16 However, later chemical changes are likely due to a combination of slow 8735
dx.doi.org/10.1021/es202677v |Environ. Sci. Technol. 2011, 45, 8733–8740
Environmental Science & Technology
ARTICLE
concentrations fluctuated initially and stabilized after ∼500 h. While none of the AGWs were oversaturated with respect to calcite, once placed in contact with the sediments, solutions were found to be oversaturated as soon as the first sampling point at 24 h. For example, in the SWC AGW 2 system at 100 g/L shown in Figure 1a, calcite ion activity product increased from 0.04 to 0.29 over 2,000 h of reaction and from 0.32 to 0.79 for SWC in AGW 4 at 1000 g/L. These values are similar to those reported elsewhere for U(VI) contaminated calcareous sediments.7,20 It is critically important to allow for this oversaturation in modeling uranium adsorption as calcium readily forms solution complexes with uranium and carbonate and strongly impacts solid/solution partitioning of uranium.7 Modeling the Impact of Variable Chemistry. Our objective was to develop an SCM that accounted for the impact of variable chemical conditions on equilibrium U(VI) adsorption so that the U(VI) adsorption properties of individual sediment samples could be compared. The optimal SCM was determined by fitting experimentally measured U(VI) equilibrium adsorption desorption on different sediment samples across the range of chemical conditions applicable to the study area. Nine reactions (SI Table S4) differing in stoichiometric relationships between U(VI), carbonate, and H+ were tested to find which reaction (or combination of reactions) achieved the best possible description of the impact of variable chemical conditions on U(VI) adsorption on these sediment samples (see SI for a description of the assessment of the quality of fit for each reaction). The best fit for all <2 mm sediment samples was achieved with Reaction 1 while the SWC fines were best fit by Reaction 2. > SOH þ UO2 2þ þ 2CO3 2 ¼> SOUO2 ðCO3 HCO3 Þ2
ð1Þ > SOH þ UO2 2þ þ 2CO3 2 ¼ > SOUO2 ðCO3 Þ2 3 þ Hþ
ð2Þ Figure 1. Measured dissolved uranium concentrations, “free” carbonate concentrations in the form of CO 3 2 and modeled equilibrium aqueous uranium concentrations; dissolved calcium concentrations and pH plotted on the secondary y-axis for sample SWC <2 mm in AGW-2, 100 g/L (a) and AGW-4, 1000 g/L (b). Error bars for measured variables represent differences between experimental duplicates while error bars for modeled [U] are based on logK c uncertainty (Table 1).
incongruent alumino-silicate mineral dissolution, evidenced by the continued release of silicon, and possible aerobic respiration.40 While organic content in these sediments is low (0.05 0.2% by weight), the impact of microbial activity, observed previously in column experiments, is suggested by increases in alkalinity.18 Therefore, at 100 g/L, aqueous chemical conditions influencing uranium(VI) speciation changed steadily throughout the experiments. Less extensive changes in solution chemistry were observed in experiments conducted at higher suspension density (1000 g/L). Uranium concentrations increased rapidly during the initial 24 h of the experiments, slowly increased out to about 500 h, and then remained relatively constant for the duration of the experiments (Figure 1b). At higher suspension densities, 3248% of the adsorbed uranium was released. Alkalinity, pH and calcium
For each individual data set, model-calculated and measured Kd values agreed within experimental uncertainties for 80% of the data points. Data points where model-calculated and measured adsorption fell out of the range of experimental errors showed no discernible trend with chemical conditions. The model captures the decreasing trends in U(VI) adsorption with increasing alkalinity (Figure 2). This trend is significant as lower alkalinity Columbia River water mixes with groundwater of higher alkalinity during high river stages.27,28 As will be shown, scatter in the measured and model-calculated U(VI) adsorption results primarily from variable chemical conditions other than alkalinity. For all samples, attempts to fit a second reaction to the data failed as long as Reaction 1 (or Reaction 2 for the SWC fines) was included, indicating that this reaction by itself accounted for most of the chemical variability observed in the data. Similarly, no improvement in fits was obtained using single-reaction two-site models, where a portion of the total sites were allowed to bind more strongly than the remainder of the sites.24,41 Bond et al., 2008 found that two reactions (Reactions 7 and 8 in SI Table S4) provided the best fit for samples from the vadose-zone of the 300Area.20 This sample set included sediments containing uranium mineral precipitates and showing clear signs of alteration, likely resulting from the highly acidic and alkaline solutions discharged to the shallow pits in this area.20,36,37,42 Examination of the impact of the assumed site density on the conditional constant 8736
dx.doi.org/10.1021/es202677v |Environ. Sci. Technol. 2011, 45, 8733–8740
Environmental Science & Technology
Figure 2. U(VI) adsorption, expressed as the ratio of adsorbed (mol/kg) and dissolved (mol/L) U(VI) concentrations, on IFRC samples. Modelcalculated U(VI) adsorption used the best-fit SCM for “all data” based on Reaction 1.
revealed that they are inversely proportional until the site density approaches the initial adsorbed U(VI) concentration and the assumed site density does not influence which reaction provides the best fit (Table S5, Figure S3, and SI discussion). Given the low adsorbed U(VI) concentrations applicable to the study site (<1.4 nmol/m2), U(VI) adsorption on sediments may be dominated by a small percentage of total sites with high affinity for U(VI). In that case, a lower logKc value might be required at higher adsorbed U(VI) concentrations. The proportionality between site density and Kc shows that the approach used here to determine differences in U(VI) binding by sediments cannot distinguish between differences in the intrinsic U(VI) binding properties of adsorption sites and small differences in site density between samples. In order to understand the factors controlling continual changes in uranium concentrations observed during the experiments, the SCM for each sample was used to compute the equilibrium uranium concentration at each time point based on the observed chemical conditions. Uranium adsorption equilibrium was achieved within 500 h for SWC 100 g/L experiments as indicated by the agreement between calculated equilibrium solution concentrations and measured concentrations (Figure 1a). Subsequently, uranium concentrations remained at adsorption equilibrium while chemical conditions (e.g., calcium concentrations) evolved in a manner that increased the equilibrium U(VI) concentration. In the SWC 1000 g/L experiment, uranium concentrations were above those expected for adsorptive equilibrium for the first 500 h of the experiment (Figure 1b). At this high solid-to-liquid ratio, uranium concentrations were evidently affected by irreversible processes likely associated with the history of the sample.27,28,16 Adsorptive equilibrium was achieved within 1000 h, after which time uranium concentrations remained at equilibrium for the duration of the experiment. Similar results were obtained for experiments conducted with the other sediment samples at both suspension densities. Standard deviations estimated by FITEQL account for the impact of analytical errors on fitted conditional equilibrium con-
ARTICLE
stants. An additional source of uncertainty derives from possible incomplete description of the impact of variable chemical conditions provided by the reaction or reactions chosen to fit the data.14 This additional source of uncertainty in the logKc values was estimated by adjusting the logKc value up and down until the “logKc envelope” captured all experimental data points (Table 1). For all samples, the logKc envelope uncertainty was larger than the FITEQL standard deviation. This is particularly clear for the “all data” fit. The standard deviation computed from propagation of experimental errors (0.03) decreased in proportion to the square root of the increase in data points used in the fit.12 In contrast, the logKc envelope uncertainty increased (0.36) as one would expect when fitting data from sediment samples with apparent differences in logKc values. Comparing Adsorption Properties of Sediment Samples. In order to determine whether there are significant differences in U(VI) adsorption properties of sediment samples one must first minimize variability and scatter in measured and model-calculated adsorbed U(VI) concentrations owing to aqueous chemical conditions (e.g., Figure 2). This can be achieved by rearranging mass-action expressions for the adsorption reactions as linear functions of the log of the activity of the aqueous species that control U(VI) speciation: UO22+, CO32-, and H+. For Reaction 1 the expressions are as follows: log
½SOUO2 ðCO3 HCO3 Þ2 ¼ logðR 1 CO3 Þ ½SOHðUO2 2þ Þ ¼ log K c þ 2logðCO3 2 Þ
log
log
ð3Þ
½SOUO2 ðCO3 HCO3 Þ2 ¼ logðR 1 UO2 Þ ½SOHðCO3 2 Þ2 ¼ log K c þ logðUO2 2þ Þ
ð4Þ
½SOUO2 ðCO3 HCO3 Þ2 ¼ logðR 1 HÞ ¼ log K c ½SOHðUO2 2þ ÞðCO3 2 Þ2
ð5Þ
where square brackets symbolize concentrations (surface species, in mol/L) and parentheses symbolize activity (aqueous species). Plots of the experimental data, expressed as the quotients on the left-hand side of eqs 3 and 4 against log carbonate and log uranyl activity, should yield linear trends with slopes of 2 and one, respectively. Hydrogen ions do not appear in Reaction 1 so a plot of the mass-action expression on the left-hand side of eq 5 vs pH should yield horizontal trends. Samples with significantly different adsorption properties will yield parallel, linear trends separated by differences in their logKc values. Analogous expressions can be written based on other reactions to be examined; these would have different slopes based on the stoichiometric coefficients for the applicable reaction. To the extent that the chosen reaction does not provide an adequate description of the impact of variable chemistry on adsorption, the slope of the linear trend in the experimental data will deviate from the stoichiometric coefficient. Plots of the experimental data for sediment samples 227, 231, 332, and SWC using eqs 35 show linear trends for each sample (SI Figure S4). The overlap of data points within experimental error across the range of log carbonate, uranyl, and H+ conditions studied indicate that these samples do not have significantly different logKc values. Slopes of the log(R1CO3) vs log(CO3) trendlines (SI Figure S4a) for 227, 231, 33,2 and 8737
dx.doi.org/10.1021/es202677v |Environ. Sci. Technol. 2011, 45, 8733–8740
Environmental Science & Technology
ARTICLE
Figure 3. Experimental data plotted for 332 and the SWC Fines (<0.063 mm) for Reaction 1 in terms of (a) logR1CO3 vs CO3, (b) logR1UO2 vs UO2, (c) logR1pH vs pH and for Reaction 2 in terms of (d) logR2CO3 vs CO3, (e) logR2UO2 vs UO2, (f) logR2pH vs pH. The best-fit conditional equilibrium constant (logKc) is marked by a solid line for each sample. The uncertainty in logKc is bounded by the dotted or dashed lines calculated by adjusting the parameter in the positive and negative direction until calculated U(VI) adsorbed concentrations agree with all measured values. Vertical and horizontal error bars represent propagation of error values assumed in modeling (SI Table S2).
SWC are 2.0, 2.2, 2.0, and 1.8, respectively, as compared to the theoretical slope of 2. Similarly, slopes of log(R1UO2) vs log(UO22+) trendlines (SI Figure S4b) are 1.0, 0.9, 1.0, and 1.1, respectively (theoretical slope equals 1); slopes of log(R1pH) vs pH trendlines (SI Figure S4c) are 0.1, 0.4, 0.1, and 0.5, respectively, (theoretical slope equals 0). Thus, Reaction 1 captures most of the impact of the variability in chemistry on uranium adsorption for these samples. Adsorption properties of sediment sample 332, which exhibited the lowest U(VI) adsorption logKc (Table 1), were compared with those of the SWC fines using mass-action expressions for Reaction 1, the best fit for sample 332, and Reaction 2, the best fit for the fines (SI Table S4). In all massaction plots, trends in the experimental data for the SWC fines fall above those for sample 332 in excess of experimental errors over the range of chemical conditions examined (Figure 3). Also, trends calculated from logKc values for the SWC fines exceed those for sample 332 even when the maximum and minimum logKc values (logKc envelope) for the SWC fines and sample 332 and are considered. Larger uncertainty values are necessary to capture all data points when the surface complexation model does not describe the impact of variable chemistry well. For sample 332, the uncertainty in logKc using Reaction 1 is 0.12 but increases to 0.25 with Reaction 2. Conversely, the SWC fines uncertainty for Reaction 1 is 0.45 but decreases to 0.12 with Reaction 2. Using either reaction, the SWC fines have a statistically significant higher U(VI) adsorption affinity than sediment sample 332 (Tables 1 and SI S4). Experimental trends overlapped within experimental errors along the range of chemical conditions studied for all other pairs of sediment samples from the site (SI Figure S4). Thus, no significant difference in U(VI)
adsorption properties between these sediment samples could be established. Ratios of Kc values for the SWC fines to sample 332 are 4 and 3 for Reaction 1 and 2, respectively (SI Table S4). The maximum difference in Kc values among the IFRC <2 mm sediment samples for Reaction 1 is a factor of 2 (Table 1). Based on this set of samples, the minimum difference in Kc values needed to establish a significant difference in U(VI) adsorption properties over and above differences in U(VI) adsorption owing to variable chemical conditions is a factor of 3 to 4. For comparison, adsorbed U(VI) concentrations vary by a factor of 30 over the range of chemical conditions studied (Figure 2). The implications of the lack of differences in logKc values to U(VI) transport at the site will be important to understand. Sediments examined in this investigation were collected over relatively coarse vertical intervals (Table 1). Examinations conducted at a finer scale may reveal greater variability. Implications. Some additional observations about logKc uncertainty can be made using this mass-action based analysis. Disparities between slopes of the experimental trends and the theoretical slopes increase the uncertainty in the logKc value. This is especially clear in plots of log(R1pH) and log(R2pH) vs pH, where the experimental trends suggest a slope of 0.1 for sample 332 and a slope of 0.7 for the SWC fines (Figure 3c and f with theoretical slopes of 0 and 1, respectively). Where such disparities exist, applying the SCM outside of the range of chemical conditions used for model calibration can result in errors in computed adsorbed U(VI) that exceed those that can be estimated from the experimental data. A reliable comparison of U(VI) adsorption properties of sediment samples can only be made if the range of chemical conditions over which the 8738
dx.doi.org/10.1021/es202677v |Environ. Sci. Technol. 2011, 45, 8733–8740
Environmental Science & Technology properties are determined coincide. The appropriate range of chemical conditions for such comparisons and for assessment of SCM applicability is best defined in terms of the species that comprise aqueous and surface complexes: free UO22+, H+, and CO32. Using these species will take into account variations in concentrations of other solutes like calcium and magnesium that form aqueous U(VI) species. The lack of significant differences in logKc values of sediment samples collected across the Hanford IFRC site may be related to the lack of mineralogical differences observed. Spectroscopic studies suggest that dominant U(VI) surface complexes form on quartz and phyllosilicate minerals.43 Determining differences in these ubiquitous minerals across a heterogeneous field site could prove exceedingly difficult. In a study carried out in a quartz-sand aquifer, a significant correlation was found between surface area normalized Pb2+ adsorption (0.71.2 μmol/m2) and extractable iron and aluminum with similar results for Zn 2+ (0.03 0.07 μmol/m2).5 This is expected given the high adsorption densities and importance of aluminum-substituted goethite coatings on quartz surfaces in controlling adsorption properties in the absence of organic or inorganic carbon.5,44 Under these conditions, traditional chemical extractions may provide a reliable estimate of variability in the abundance of the constituents and, thereby, their absorptive capacity and variability. At the Hanford IFRC site, with carbonate mineral complexation of uranium and the diffuse distribution of clays and clay coatings throughout the aquifer, characterizing spatial variability of adsorption properties of uranium requires more thorough study.
’ ASSOCIATED CONTENT
bS
Supporting Information. Discussion of iron extractions, model fitting, and assumed site density impacts on logKc. Tables S1S5: AGW composition, SCM reactions and assumed errors, extractable iron data and site density parameters. Figures S1S4: Site map, adsorbed uranium depth profile, assumed site density impacts of logKc and additional mass-action plots. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 650-329-4529; fax: 650-329-4545; e-mail: dlstoliker@ usgs.gov.
’ ACKNOWLEDGMENT Funding for this work was provided by the U.S. Department of Energy Office of Science, Subsurface Biogeochemical Research program, through the Hanford 300-Area IFRC. Additional funding came from the U.S. Geological Survey through the Hydrologic Research and Development program. Technical assistance from Chris Fuller in the measurement of total uranium is gratefully acknowledged. We thank Tracie Conrad and James Hein for XRD analyses. Discussions with Roy Haggerty, Chongxuan Liu, and Jun Yin along with reviews by Gary Curtis, James Davis, and four anonymous reviewers improved the quality of this manuscript. ’ REFERENCES (1) Allen-King, R. M.; Divine, D. P.; Robin, M. J. L.; Aldredge, J. R.; Gaylord, D. R. Spatial distributions of perchloroethylene reactive
ARTICLE
transport parameters in the Borden Aquifer. Water Resour. Res. 2006, 42, W01413, DOI: 10.1029/2005WR003977. (2) Rajaram, H. Time and scale dependent effective retardation factors in heterogeneous aquifers. Adv. Water Resour. 1997, 20, 217–229. (3) Davis, J. A.; Kent, D. B. Surface complexation modeling in aqueous geochemistry. In Mineral-Water Interface Geochemistry Vol 23; Hochella, M. F., White, A. F., Eds.; Mineralogical Society of America: Washington, DC, 1990; pp 177260. (4) Davis, J. A.; Fuller, C. C.; Coston, J. A.; Hess, K. M.; Dixon, E. Spatial Heterogeneity of Geochemical and Hydrologic Parameters Affecting Metal Transport in Ground Water; U.S. Environmental Protection Agency: Washington, DC, 1993; EPA/600/S-93/006. (5) Fuller, C. C.; Davis, J. A; Coston, J. A.; Dixon, E. Characterization of metal adsorption variability in a sand and gravel aquifer, Cape Cod, Massachusetts, U.S.A. J. Contam. Hydrol. 1996, 22, 165–187. (6) Hull, L. C.; Grossman, C.; Fjeld, R. A.; Coates, J. T.; Elzerman, A. W. Hybrid empirical-theoretical approach to modeling uranium adsorption. Appl. Geochem. 2004, 19, 721–736. (7) Hyun, S. P.; Fox, P. M.; Davis, J. A.; Campbell, K. M.; Hayes, K. F.; Long, P. E. Surface complexation modeling of U(VI) adsorption by aquifer sediments from a former mill tailing site at Rifle, Colorado. Environ. Sci. Technol. 2009, 43, 9368–9373. (8) Sharif, M. S. U; Davis, R. K.; Steele, K. F.; Kin, B.; Hays, P. D.; Kreese, T. M.; Fazio, J. A. Surface complexation modeling for predicting solid phase arsenic concentrations in the sediments of the Mississippi River Valley alluvial aquifer, Arkansas, USA. Appl. Geochem. 2011, 26, 496–504. (9) Miller, A. W.; Rodriguez, D. R.; Honeyman, D. B. Upscaling sorption/desorption processes in reactive transport models to describe metal/radionuclide transport: A critical review. Environ. Sci. Technol. 2010, 44, 7996–8007. (10) Curtis, G. P.; Davis, J. A.; Naftz, D. L. Simulation of reactive transport of uranium (VI) in groundwater with variable chemical conditions. Water Resour. Res. 2006, 42, W04404, DOI: 10.1029/ 2005WR003979. (11) Ma, C.; Zheng, C.; Prommer, H.; Greskowiak, J.; Liu, C.; Zachara J. A field-scale reactive transport model for U(VI) migration influenced by coupled multirate mass transfer and surface complexation reactions. Water Resour. Res. 2010, 46, W05509; DOI: 10.1029/ 2009WRR008168. (12) Dzombak, D. A.; Morel, F. M. M. Surface Complexation Modeling: Hydrous Ferric Oxide; Wiley-Interscience: New York, 1990. (13) Westall, J. C.; Jones, J. D.; Turner, G. D.; Zachara, J. M. Models for association of metal ions with heterogeneous environmental sorbents. 1. Complexation of Co(II) by leonardite humic acid as a function of pH and NaClO4 concentration. Environ. Sci. Technol. 1995, 29, 951–959. (14) Davis, J. A.; Coston, J. A.; Kent, D. B.; Fuller, C. C. Application of the surface complexation concept to complex mineral assemblages. Environ. Sci. Technol. 1998, 32, 2820–2828. (15) Liu, C.; Zachara, J. M.; Qafoku, N. P.; Wang, Z. Scale-dependent desorption of uranium from contaminated subsurface sediments. Water Resour. Res. 2008, 44, W08413, DOI: 10.1029/2007WR006478. (16) Liu, C.; Shi, S.; Zachara, J. M. Kinetics of uranium(VI) desorption from contaminated sediments: Effect of geochemical conditions and model evaluation. Environ. Sci. Technol. 2009, 43, 6560– 6566. (17) Greskowiak, J.; Prommer, H.; Liu, C.; Post, V. E. A.; Ma, R.; Zheng, C.; Zachara, J. M. Comparison of parameter sensitivities between laboratory and field-scale model of uranium transport in a dual domain, distributed rate reactive system. Water Resour. Res., 2010, 46, W05509, DOI: 10.1029/2009WR008781. (18) Yin, J.; Haggerty, R.; Stoliker, D. L.; Kent, D. B.; Istok, J. D.; Greskowiak, J.; Zachara J. M. Transient groundwater chemistry near a river: Effects on U(VI) transport in laboratory column experiments. Water Resour. Res. 2011, 47, DOI: 10.1029/2010WR009369. (19) Dong, W.; Brooks, S. C. Determination of the formation constants of ternary complexes or uranyl and carbonate with alkaline 8739
dx.doi.org/10.1021/es202677v |Environ. Sci. Technol. 2011, 45, 8733–8740
Environmental Science & Technology earth metals (Mg2+, Ca2+, Sr2+, and Ba2+) using anion exchange method. Environ. Sci. Technol. 2006, 40, 4689–4695. (20) Bond, D. L.; Davis, J. A.; Zachara, J. M. Uranium(VI) release from contaminated vadose zone sediments: estimation of potential contributions for dissolution and desorption. In Adsorption of Metals by Geomedia II: Variables, Mechanisms and Model Applications; Barnett, M. O., Kent, D. B., Eds.; Elsevier: Amsterdam, 2008; pp 375416. (21) Fox, P. M.; Davis, J. A.; Zachara, J. M. The effect of calcium on aqueous uranium(VI) speciation and adsorption to ferrihydrite and quartz. Geochim. Cosmochim. Acta 2006, 70, 1379–1387. (22) Stewart, B. D.; Mayes, M. D.; Fendorf, S. Impact of uranylcalcium-carbonato complexes on uranium(VI) adsorption to synthetic and natural sediments. Environ. Sci. Technol. 2010, 44, 928–934. (23) Bargar, J. R.; Reitmeyer, R.; Lenhart, J. J.; Davis, J. A. Characterization of U(VI)-carbonato ternary complexes on hematite: EXAFS and electrophoretic mobility measurements. Geochim. Cosmochim. Acta 2000, 64, 2737–2749. (24) Waite, T. D.; Davis, J. A.; Payne, T. E.; Waychunas, G. A.; Xu, N. Uranium (VI) adsorption to ferrihydrite: Application of a surface complexation model. Geochim. Cosmochim. Acta 1994, 58, 5465–5478. (25) Barnett, M. O. Adsorption and transport of U(VI) in subsurface media. Soil Sci. Soc. Am. J. 2000, 64, 908–917. (26) Davis, J. A.; Meece, D. E.; Kohler, M.; Curtis, G. P. Approaches to surface complexation modeling of uranium(VI) adsorption on aquifer sediments. Geochim. Cosmochim. Acta 2004, 68, 3621–3641. (27) Zachara, J. M.; Davis, J. A.; Liu, C.; McKinley, J. P.; Qafoku, N.; Wellman, D. M.; Yabusaki, S. Uranium Geochemistry in Vadose Zone and Aquifer Sediments From the 300-Area Uranium Plume; United States Department of Energy: Washington, DC, 2005; PNNL-15121. (28) McKinley, J. P.; Zachara, J. M.; Wan, J.; McCready, D. E.; Heald, S. M. Geochemical controls on contaminant uranium in vadose Hanford formation sediments at the 200 Area and 300 Area, Hanford Site, Washington. Vadose Zone J. 2007, 6, 1004–1017. (29) Singer, D. M.; Zachara, J. M.; Brown, G. E. Uranium speciation as a function of depth in contaminated Hanford sediments a microXFR, micro-XRD, and micro- and bulk-XAFS study. Environ. Sci. Technol. 2009, 43, 630–636. (30) Gregg, S. J.; Sing, K. S. W. Adsorption, Surface Area, and Porosity; Academic Press: New York, 1982. (31) Davis, J. A.; Curtis, G. P. Application of Surface Complexation Modeling to Describe Uranium (VI) Adsorption and Retardation at the Uranium Mill Tailings Site at Naturita, Colorado; U.S. Nuclear Regulatory Commission: Rockville, MD, 2003; NUREG/CR-6708;. (32) Kohler, M.; Curtis, G. P.; Meece, D. E.; Davis, J. A. Methods for estimating adsorbed uranium(VI) and distribution coefficients of contaminated sediments. Environ. Sci. Technol. 2004, 38, 240–247. (33) Loeppert, R. H.; Inskeep, W. P. Iron. In Methods of Soil Analysis, Part 3, Chemical Methods, SSSA Book Series No. 5; Sparks, D. L., Ed.; Soil Science Society of America: Madison, WI, 1996; pp 639664. (34) Chao, T. T.; Zhou, L. Extraction techniques for selective dissolution of amorphous iron oxides from soils and sediments. Soil Sci. Soc. Am. J. 1983, 47, 225–232. (35) Herbelin, A. L.; Westall, J. C. FITEQL: A Computer Program for the Determination of Chemical Equilibrium Constants from Experimental Data; Chemistry Department, Oregon State University: Corvallis, OR, 1999. (36) Catalano., J. G.; McKinley, J. P.; Zachara, J. M.; Heald, S. M.; Smith, S. C.; Brown, G. E. Changes in uranium speciation through a depth sequence of contaminated Hanford sediments. Environ. Sci. Technol. 2006, 40, 2517–2524. (37) Arai, Y.; Marcus, M. A.; Tamura, N.; Davis, J. A.; Zachara, J. M. Spectroscopic evidence for uranium bearing precipitates in vadose zone sediments at the Hanford 300-area site. Environ. Sci. Technol. 2007, 41, 4633–4639. (38) Shang, J.; Liu, C.; Wang, Z.; Zachara, J. M. Effect of grain size on uranium(VI) surface complexation kinetics and adsorption additivity. Environ. Sci. Technol. 2011, 45, 6025–6031.
ARTICLE
(39) Qafoku, N. P.; Zachara, J. M.; Liu, C.; Gassman, P. L.; Qafoku, O. S.; Smith, S. C. Kinetic desorption and sorption of U(VI) during reactive transport in a contaminated Hanford sediment. Environ. Sci. Technol. 2005, 39, 3157–3165. (40) Blum, A. E.; Stillings, L. L. Feldspar dissolution kinetics. In Rev. Mineral. Geochem. Vol. 31; White, A. F., Brantley, S. L., Eds.; Mineralogical Society of America: Washington, DC, 1995; pp 291351. (41) Kohler, M.; Curtis, G. P.; Kent, D. B.; Davis, J. A. Experimental investigation and modeling of uranium(VI) transport under variable chemical conditions. Water Resour. Res. 1996, 32, 3539–3551. (42) Stubbs, J. E.; Veblen, L. A.; Elbert, D. C.; Zachara, J. M.; Davis, J. A.; Veblen, D. R. Newly recognized hosts for uranium in the Hanford Site vadose zone. Geochim. Cosmochim. Acta 2009, 73, 1563–1576. (43) Wang, Z.; Zachara, J. M.; Boily, J.; Xia, Y.; Resch., T.; Moore, D. A.; Liu, C. Determining individual mineral contributions to U(VI) adsorption in a contamination aquifer sediment: a fluorescence spectroscopy study. Geochim. Cosmochim. Acta 2011, 75, 2965–2979. (44) Zhang, S.; Kent, D. B.; Elbert, D. C.; Shi, Z.; Davis, J. A.; Veblen, D. R. Mineralogy, morphology, and textural relationships in coasting on quartz grains in sediments in a quartz-sand aquifer. J. Contam. Hydrol. 2011, 124, 57–67.
8740
dx.doi.org/10.1021/es202677v |Environ. Sci. Technol. 2011, 45, 8733–8740
ARTICLE pubs.acs.org/est
Dissolved Organic Carbon Enhances the Mass Transfer of Hydrophobic Organic Compounds from Nonaqueous Phase Liquids (NAPLs) into the Aqueous Phase Kilian E. C. Smith,*,† Martin Thullner, Lukas Y. Wick, and Hauke Harms UFZ Helmholtz Centre for Environmental Research, Department of Environmental Microbiology, Leipzig, Germany
bS Supporting Information ABSTRACT: The hypothesis that dissolved organic carbon (DOC) enhances the mass transfer of hydrophobic organic compounds from nonaqueous phase liquids (NAPLs) into the aqueous phase above that attributable to dissolved molecular diffusion alone was tested. In controlled experiments, mass transfer rates of five NAPLphase PAHs (log KOW 4.155.39) into the aqueous phase containing different concentrations of DOC were measured. Mass transfer rates were increased by up to a factor of 4 in the presence of DOC, with the greatest enhancement being observed for more hydrophobic compounds and highest DOC concentrations. These increases could not be explained by dissolved molecular diffusion alone, and point to a parallel DOC-mediated diffusive pathway. The nature of the DOC-mediated diffusion pathway as a function of the DOC concentration and PAH sorption behavior to the DOC was investigated using diffusion-based models. The DOC-enhanced mass transfer of NAPL-phase hydrophobic compounds into the aqueous phase has important implications for their bioremediation as well as bioconcentration and toxicity.
’ INTRODUCTION Hydrophobic organic compounds (HOCs) are often found as mixtures in the form of nonaqueous phase liquids (NAPLs), such as hydrocarbon oil spills or industrial NAPLs contaminating terrestrial and aquatic environments. Due to their hydrophobic nature, the HOCs preferentially remain in the NAPL, with slow mass transfer into the aqueous phase.13 Because water-dissolved HOCs play a key role in diffusive uptake into organisms,46 this has contrasting implications for the biota inhabiting NAPL-contaminated environments. Microorganisms using HOCs as a source of carbon and energy face a large reservoir of inaccessible food in the NAPL, often reflected in slow bioremediation of NAPL contaminated sites.7 However, HOC bioconcentration can also lead to toxicity.8 Here, low aqueous concentrations and thus reduced bioconcentration might ameliorate detrimental effects of the NAPL pollution. In NAPL-polluted aquifers, mass transfer into the aqueous phase also determines groundwater contamination, with consequences for human health and the effectiveness of water-based remediation technologies.7,9 Often, process dynamics mean that equilibrium partitioning between the NAPL and aqueous compartments is not reached, and aqueous concentrations are determined by mass exchange kinetics.10,11 Therefore, to understand NAPL toxicity and (bio)remediation, evaluation of the key controlling factors is required. There are many influences on NAPL to aqueous phase mass transport, including mixing10 and interfacial area,9 NAPL viscosity and surface film development,3,12 and the properties of the component HOCs.1,2 Furthermore, biodegrading organisms can lead to r 2011 American Chemical Society
altered mass fluxes. Microbial adaptations such as biofilm formation,13 chemotaxis,14 or surfactant production,15 can increase mass fluxes compared to abiotic dissolution alone. Biofilms can also provide an additional diffusive barrier reducing mass transfer.16 Mobile “colloid-like” phases can contribute to diffusive mass exchange processes between surfaces and the bulk aqueous phase,1718 a phenomenon termed enhanced or facilitated diffusion. For example, surfactants micelles can increase, decrease, or negligibly impact mass transfer from NAPLs into the aqueous phase depending on the surfactant, NAPL properties and architecture, and pore-water flow.2022 Enhanced diffusion has also been studied in the context of other HOC exchange fluxes, including polymerwater exchange,2326 aquatic bioaccumulation,27 and biodegradation.28 The nature of the sorbing phases leading to enhanced diffusion is varied, including nanoparticles,23 proteins,26 and dissolved organic carbon (DOC).24,25,27,28 The ubiquity of DOC and its propensity to sorb HOCs raises the relevant question as to whether DOC can increase mass transfer of HOCs out of a NAPL and into the surrounding aqueous phase, and what the implications are for biodegradation and toxicity. The important role of DOC in enhancing metal mass fluxes and availability to organisms is well recognized, and the theoretical basis describing the relevant processes exists.19,18,29,30 Received: January 23, 2011 Accepted: August 31, 2011 Published: August 31, 2011 8741
dx.doi.org/10.1021/es202983k | Environ. Sci. Technol. 2011, 45, 8741–8747
Environmental Science & Technology This study was conceived to test the hypothesis that DOC increases mass fluxes of HOCs from a NAPL into the aqueous phase above those attributable to dissolved diffusion alone. As representative HOCs, five PAHs (log KOW 4.155.39) were dissolved in 2,2,4,4,6,8,8-heptamethylnonane (HMN) as a model NAPL. PAH mass transfer into saline aqueous medium containing different concentrations of DOC were measured over time using equilibrium negligible depletion-solid phase microextraction (nd-SPME). This study was part of a larger effort looking at contamination in the marine environment, hence the use of saline medium. The temporal development in dissolved and DOC-sorbed PAH concentrations were interpreted using diffusion models, accounting for any contribution of DOC-enhanced diffusion.
’ MATERIALS AND METHODS The complete experimental section is given in the Supporting Information (SI) and only a summary is provided below. Experimental Setup. Mass transfer experiments were carried out in the setup shown in Figure S1 with 2 L of saline medium (see Table S1) containing 0, 0.0226, or 0.226 g DOC L1. The glass tube (250 mm long, 25 mm id) was inserted through the lid to 3 cm below the water surface and experiments started by adding 1 mL of HMN containing PAHs, each at approximately 1 g L1 (see Table S2), to the inside of the glass tube so that it floated on the surface of the water to give a defined and constant surface area. The bottles were kept at 22 °C under stirring conditions sufficient to keep the aqueous phase homogeneous, while avoiding breaking of the NAPL/water interface. For each DOC concentration, duplicate experiments were conducted on different dates. Sampling and nd-SPME Analysis. Equilibrium nd-SPME with internal standards was used to measure the total, DOC-sorbed, and dissolved aqueous PAH concentrations.28,31 At intervals (see Table S3), 10-mL aqueous samples were taken and deuterated PAHs were added as internal standards together with a 0.5-cm length of polydimethylsiloxane (PDMS)-coated SPME fiber. These were shaken for 24 h to allow equilibrium to be reached (see Figure S2), before placing the fibers in 100 μL of toluene containing 1 μg of acenaphthylene-d8 as an injection standard to re-extract the PDMS sorbed PAHs.
Total, DOC-Sorbed and Dissolved PAH Concentrations. Dissolved concentrations in the DOC samples were calculated using a 4-point external calibration performed in triplicate (n = 12). PAH concentrations in the PDMS were plotted against those in the saline medium (Figure S3), with the slopes giving the PDMS: saline medium partitioning ratios KPDMS/d (L L1) (Table S4) for converting the PDMS sorbed PAH concentrations into dissolved concentrations. Quantifying the PDMS sorbed PAHs relative to the PDMS sorbed PAH internal standards gives the total aqueous concentrations, and does not require calibration of the SPME method.31 In the experiments with saline medium, total concentrations are equivalent to the dissolved concentrations. In experiments with DOC, these represent the dissolved plus sorbed concentrations. DOC-sorbed concentrations are calculated by difference, without the necessity of assuming a complete mass balance. HMN Concentrations. The initial concentrations of PAHs in the HMN were analyzed by dissolving 50 μL of the HMN in 950 μL of toluene and adding 1 μg of each deuterated PAH in toluene as internal standards. Analysis and Quantification. GC-MS analyses were performed using an HP 6890 Series GC (Hewlett-Packard, California, USA), with a 30-m HP5MS capillary column (0.25 mm i.d.
ARTICLE
and 0.25 μm film thickness) and 2 m uncoated and deactivated HP retention gap (0.53 mm i.d.) in selected ion monitoring mode (SIM). PAHs were quantified using the HP Environmental Data Analysis software. Before adding the HMN containing the PAHs to start an experiment, a medium sample was taken from each setup as a blank. These blanks amounted to less than 2% of the final concentrations, and were subtracted from the aqueous concentrations measured during the experiments. Data were fitted by the least-squares method using Graphpad Prizm 5 (San Diego, CA, USA), giving the respective rate constants and equilibrium concentrations. Fitting the Data and Modeling of Enhanced Diffusion. A complete description of the modeling approaches is given in the SI, with only a summary given below. Mass transfer of PAHs from the NAPL into the aqueous compartment leads to an increase in the total concentration ctot (μg L1) with time t (h). In the presence of DOC, ctot is given as the sum of dissolved cd (μg L1) and DOC sorbed cs (μg L1) concentrations ctot ¼ cd þ cs ¼ cd þ ½DOC 3 KDOC=d 3 cd ¼ cd 3 ð1 þ ½DOC 3 KDOC=d Þ
ð1Þ
where [DOC] (kg DOC L1) is the aqueous concentration of the DOC and KDOC/d (L kg DOC 1) is the partitioning ratio between DOC sorbed and the dissolved phases. In the absence of DOC, ctot is equal to cd, and the rate of change in aqueous concentration ∂ctot/∂t = ∂cd/∂t (μg L1 h1) can be described using the two-film model2 cd ðtÞ ¼ ceq, d 3 ð1 expð kd 3 tÞÞ
ð2Þ
where kd (h1) is the mass transfer rate constant and ceq,d (μg L1) is the dissolved concentration in equilibrium with the NAPL phase. Fitting eq 2 to the measured cd(t) allows ceq,d and kd to be determined. The dominant resistance to the mass transfer lies in the aqueous phase,2 and kd is determined by diffusion within an aqueous diffusive boundary layer (BL). Provided the NAPL surface area, aqueous volume, and mixing remain constant, ceq,d and kd is given by kd ¼
A A Dd md ¼ 3 3 V V δd
ð3Þ
where A (equal to 4.91 104 m2) is the NAPL/aqueous interfacial area, V (equal to 2 L) is the volume of the aqueous phase, md (m h1) is the mass transfer velocity, Dd (m2 h1) is the molecular diffusion coefficient, and δd (m) is the thickness of the molecular diffusive BL of the PAH in question. In the experiments with DOC, an equation analogous to eq 2 can be fitted to the measured ctot(t) to determine a compound mass transfer rate constant ktot (h1). ctot ðtÞ ¼ ceq, tot 3 ð1 expð ktot 3 tÞÞ
ð4Þ
where ceq,tot (μg L1) is the total aqueous concentration in equilibrium with the NAPL. DOC might lead to two scenarios: (i) it does not influence dissolution, and molecular diffusion alone determines the changes in aqueous concentrations or (ii) DOC mediates an additional diffusive pathway across the BL. In the latter case, the diffusion of the dissolved and sorbed PAHs can be described by an average 8742
dx.doi.org/10.1021/es202983k |Environ. Sci. Technol. 2011, 45, 8741–8747
Environmental Science & Technology
ARTICLE
diffusion coefficient D (m2 h1), weighted for the dissolved and sorbed fractions.19,18 D̅ ¼
Dd þ Ds 3 ξ 3 ½DOC 3 KDOC 1 þ ½DOC 3 KDOC
ð5Þ
where Ds (m2 h1) is the diffusion coefficient of the DOC-sorbed PAH and ξ (unitless) is a lability factor describing that portion of the sorbed PAH that can be considered labile. This ranges between 0 (completely nonlabile) and 1 (fully labile) (see SI for more details). The value of ktot derived from fitting eq 4 to the experimental data can, analogously to eq 3, be defined in terms of D and δ̅ ,32 ktot ¼
A A D̅ mtot ¼ 3 3 V V δ̅
ð6Þ
where mtot (m h1) is the compound mass transfer velocity and δ̅ (m) is the thickness of the joint diffusive BL. The value of δ̅ is intermediate between δd and the thickness of the diffusive BL for the DOC-sorbed PAHs δs (m),33 and can be calculated as described in the SI. By substituting eq 5 into eq 6, an expression for the compound mass transfer rate constant ktot is obtained ktot ¼
kd þ ks 3 ξ 3 ½DOC 3 KDOC=d 1 þ ½DOC 3 KDOC=d
ð7Þ
where the mass transfer rate constants kd* (= A/V 3 Dd/δ̅ ) and ks* (= A/V 3 Ds/δ̅ ) are defined with respect to the thickness of the joint diffusion layer δ̅ . For the situation where DOC does not contribute to the diffusive transport, ktot in eq 7 reduces to ktot, dissolved ¼
kd 1 þ ½DOC 3 KDOC=d
ð8Þ
where ktot,dissolved (h1) is the mass transfer rate constant when only dissolved molecular diffusion contributes to the process. All the parameters for calculating ktot,dissolved using eq 8 are independently measured. Fitting the PAH concentration profiles from the saline medium set-ups give kd (defined in terms of Dd and δd), [DOC] is known, and KDOC/d is determined from the nd-SPME measurements. Therefore, when DOC does not play a role in the mass transfer process, values of ktot derived from fitting eq 4 to ctot(t) should be identical to values ktot,dissolved calculated using eq 8. In the case of DOC-enhanced diffusion, fitted values of ktot will be higher due to an additional contribution mediated by the DOC. When DOC contributes to the diffusive transport ktot is given by the full eq 7. Apart from the product ks* 3 ξ, the parameters in eq 7 are independently known (Dd and Ds from the literature, δ̅ is calculated as described in the SI, [DOC] is known, and KDOC/d is measured). Therefore, the value of ktot obtained from fitting eq 4 to ctot(t) allows eq 7 to be solved for ks* 3 ξ. The product ks* 3 ξ contains information about the diffusion coefficient of the DOCsorbed PAH, the thickness of δ̅ , and the lability the sorbing DOC aggregates. For complete lability of the sorbed PAH this reduces to ks*.
’ RESULTS AND DISCUSSION Are the Assumptions Behind the Modeling Approaches Met? The correct interpretation and comparison of the kinetic
rate constants from the various experiments requires a homogeneous
aqueous phase of constant volume, the HMN interface area and PAH concentrations are constant, and partitioning can be described by a single KDOC/d value. Mixing conditions were sufficiently vigorous to ensure a homogeneous aqueous phase (see duplicate experiment data in Figure S4). The cumulative volume removed was between 5 and 8% for the different experiments, and thus V can be taken as constant. The dissolved aqueous concentrations immediately adjacent to the HMN interface are assumed to be in instantaneous equilibrium with the HMN. These cannot be measured directly and are inferred from the bulk aqueous concentrations at system equilibrium. Significant depletion of the HMN due to mass transfer into the aqueous phase would render this approach invalid.1 Maximum depletion of the HMN due to mass transfer into the aqueous phase was lower than 12%, with the exception of acenaphthene, fluorene, and phenanthrene in the highest DOC experiment where depletion was still less that 20% (Table S5). Therefore, the aqueous interface concentrations were taken as constant and equal to the bulk dissolved concentration at equilibrium. The validity of the assumption that PAH partitioning is well-described by a single KDOC/d value is shown in Figure S5, where sorbed PAH concentrations are plotted against Cd for the DOC 0.226 g L1 experiments. In agreement with the literature for this DOC type,34 sorption was linear over the range in experimental dissolved concentrations implying that the DOC concentration and sorbing properties remained unchanged. The total and dissolved concentrations measured using nd-SPME were used to calculate KDOC/d for both DOC concentrations, and the mean values are shown in Figure S6 and tabulated in Table S6. KDOC/d values increased with PAH hydrophobicity, falling within the range given in the literature.34 For the less hydrophobic acenaphthene to phenanthrene, sorbed concentrations in the DOC 0.0226 g L1 treatment were too low to allow KDOC/d to be reliably determined. KDOC/d values were slightly higher at DOC 0.226 g L1 compared to the low DOC treatment by a factor of 1.9 and 3.6 for fluoranthene and pyrene, respectively. Because enhanced diffusion could only be discerned at the higher DOC concentration, the values from the DOC 0.226 g L1 experiments have been used for the calculations below. Mass Transfer into Saline Medium. Figure S4 shows the increase in saline medium PAH concentrations for the duplicate experiments over time. Good agreement attests to the reproducibility of the setup and measurement protocols. PAH dissolved concentrations increased before leveling out within the experiment duration as a partitioning equilibrium between the HMN and aqueous phases was reached. The values of ceq,d and kd were obtained by fitting eq 2 to the cd(t) data, and are summarized in Table S7 together with the values of md and δd calculated using eq 3. Ceq,d decreased with increasing PAH hydrophobicity because of the increased partitioning to the HMN. This is evident from Figure S7 where log KHMN/d for the different PAHs are plotted against log KOW, increasing with PAH hydrophobicity (log KHMN/d = 0.772 log KOW + 1.020; r2 = 0.98). The fitted mass transfer rate constants kd were similar for all PAHs, ranging by less than a factor of 1.5 from 0.0114 h1 for pyrene to 0.0173 h1 for fluorene (Figure 2). These are directly reflected in the values of md which varied between 4.66 102 m h1 for pyrene to 7.03 102 m h1 for fluorene. The δBL values ranged between 31 and 43 μm (Table S7), as expected given the identical hydrodynamic conditions and similarity in the molecular diffusion coefficients. Although the magnitudes of kd (and therefore also md) and δBL are system dependent, similar trends between HOCs have also been observed in other studies.2,35 8743
dx.doi.org/10.1021/es202983k |Environ. Sci. Technol. 2011, 45, 8741–8747
Environmental Science & Technology
ARTICLE
Figure 1. Mass transfer of PAHs in HMN into saline medium containing DOC concentrations of 0.0226 and 0.226 g L1. The dashed lines show the best fit of eq 4 to ctot(t).
Mass Transfer into Saline Medium with DOC. Figure 1 shows the temporal change in ctot for the experiments with 0.0226 and 0.226 g DOC L1 (Figure S8 shows ctot and cd). With increased sorption to the DOC, it takes longer for the system to attain partitioning equilibrium, and indeed for the heavier PAHs this was not reached within the duration of the experiment. The cd values were similar to or lower than ctot (Figure S8) with the difference increasing with either increasing PAH hydrophobicity and DOC concentration because of increased sorption. For acenaphthene to phenanthrene in the low 0.0226 g L1 DOC treatment, there was a small positive bias in cd, with these being higher than ctot by a factor of up to 1.5. This might lie in either DOC altering the activity coefficient of the PAHs in the aqueous environment relative to that of the pure saline medium used for calibration,36 or perhaps the manufacturer-supplied PDMS volume per unit length of fiber was too low for this batch of fibers. In the case of significant variations in coating thickness along the fiber length, the short lengths needed to avoid dissolved phase depletion may also have contributed. Important to note is that all rate parameters were determined from fitting the ctot data, and cd was used only to calculate KDOC/d. The ctot(t) data were fitted using eq 4 to obtain ktot, with the values shown in Figure 2 (and tabulated in Table S8). The values of ktot are generally lower than those of kd, with the difference increasing with PAH hydrophobicity. For the DOC 0.0226 g L1 experiments, ktot ranged from 0.00295 h1 for pyrene to 0.01750 h1 for acenaphthene. For the DOC 0.226 g L1 experiments, these were comparable, ranging from 0.00334 h 1 for pyrene to
Figure 2. Dissolved (kd, h1) and total (ktot, h1) mass transfer rate constants together with their standard errors, obtained by fitting eqs 2 and 4 to cd(t) and ctot(t), respectively.
0.01217 h1 acenaphthene. For both DOC treatments, ktot decreased with increasing PAH hydrophobicity. Contribution of DOC-Enhanced Diffusion. Figure 3 shows the ratio of ktot (from fitting eq 4 to the measured ctot(t)) to ktot,dissolved (calculated using eq 8) for both DOC concentrations. The calculated values of ktot,dissolved assume that DOC increases the capacity of the aqueous phase for the PAHs but does not 8744
dx.doi.org/10.1021/es202983k |Environ. Sci. Technol. 2011, 45, 8741–8747
Environmental Science & Technology
Figure 3. Ratio of ktot (from fitting eq 4 to ctot(t)) to ktot,dissolved (calculated using eq 8) for the low and high DOC treatments. The values of ktot,dissolved assume that DOC acts as a simple sorbing phase and does not mediate an additional diffusive flux. A ratio of ktot/ktot,dissolved higher than one (indicated by the dashed line) implies that DOC-enhanced diffusion contributes to ktot. Standard errors were calculated via error propagation.
Figure 4. Dissolved (kd*, h1) and DOC-mediated (ks* 3 ξ, h1) mass transfer rate parameters for the DOC 0.226 g L1 concentration. The standard errors of ks* 3 ξ were calculated via error propagation. For acenapthene and fluorene, ktot and kd* overlapped and ks* 3 ξ and could not be calculated (nc).
mediate an additional diffusive flux, such that ktot is determined solely by kd. Therefore, a ktot/ktot,dissolved ratio higher than one implies an additional contribution to ktot over and above that from kd. For the lower DOC concentration, the ktot/ktot,dissolved ratios for all PAHs are close to one, indicating that enhanced diffusion due to the DOC is not discernible under these conditions. For the higher DOC concentration experiments, the ktot/ ktot,dissolved ratios range from 0.99 for acenapthene up to 3.8 for fluoranthene. For the more hydrophobic PAHs the values higher than one imply that dissolved diffusion alone cannot explain the changes in ctot, and that DOC-enhanced diffusion contributes to ktot. Increases of a similar magnitude have been observed in other studies investigating DOC-enhanced diffusion of HOCs. For example, relative to the case with pure water, mass transfer rate constants measured for polymer exchange increased by a factor
ARTICLE
of up to 8 for these same PAHs in the presence of a DOC concentration of 0.383 g L1.25 Given the minimal effects of DOC at a concentration of 0.0226 g L1, the focus of the following discussion is on the data from the high DOC treatment. Equation 7 was used to calculate ks* 3 ξ(tabulated in Table S8), with these shown in Figure 4 together with the values of kd* calculated assuming Ds is 1 order of magnitude below Dd.33,37 For acenapthene and fluorene, ktot and kd* overlapped and ks* 3 ξ could not be calculated. Calculated values of ks* 3 ξ were 0.00028, 0.00194, and 0.00172 for phenanthrene, fluoranthene, and pyrene, respectively. The mass transfer rate constant for the DOC-mediated pathway ks* contains the parameters A, V, Ds, and δ̅ . These are known or can be calculated, and it is possible to independently arrive at a setup value for ks*. Therefore, it should be possible to calculate ξ from the ks* 3 ξ values in Table S8. Calculated values of ξ (under the assumption of Ds = Dd/10) were 0.13, 0.91, and 0.84 for phenanthrene, fluoranthene, and pyrene, respectively. Although perhaps indicative of significant DOC-sorbed PAH lability these values should be treated with caution given the uncertainty in Ds. The assumption of a single literature value for Ds is a gross oversimplification since DOC is a dynamic and complex system of aggregates composed of differently sized interacting macromolecules.38 Therefore, depending on the aqueous environment, and how this affects the DOC aggregation behavior and size, sorption propensity and lability, specific fractions will differently contribute to the observed mass transport. It is recommended that in future experiments of this type, more concerted efforts be directed to obtaining information on the DOC size fractions and their sorption characteristics to further refine the values for Ds. In this way, it will be possible to use experimental data more successfully in conjunction with the existing theory19,18 to determine, for example, ξ. Initial PAH Mass Transfer Rates. The magnitude of the rate constants kd or ktot indicates how fast the system approaches a partitioning equilibrium. Therefore, ktot decreases with increasing DOC concentration and PAH hydrophobicity because an increased mass of compound needs to transfer into the aqueous phase before equilibrium is reached. However, the mass transfer rate ∂ctot/∂t is given by the product of ktot and the pertaining concentration gradient (Ceq,tot Ctot). In the field, HOCs entering the aqueous compartment via mass transfer out of the NAPL phase are subject to biodegradation or advective transport. Therefore, the scenario where the aqueous concentrations are kept low, and thus ∂ctot/∂t is maintained close to its maximum, is particularly relevant in terms of understanding biodegradation or remediation using flushing technologies. Figure 5 compares the initial PAH mass transfer rates for the experiments with saline medium and the two DOC treatments. These were calculated using the fitted values of kd and cd(eq) or ktot and ctot(eq) (from Tables S7 and S8), and assuming that cd and ct in eqs 2 and 4 are zero. These represent the initial (and indeed maximum possible) values of the dissolution rate ∂ctot/∂t. The lower DOC concentration of 0.0226 g L1 had only a minor effect on the initial mass transfer rates, with similar rates being observed in the presence of DOC. However, for the higher DOC concentration of 0.226 g L1 higher initial mass transfer rates were observed for the more hydrophobic PAHs. Relative to the case of saline medium only, initial rates were increased by factors of 1.4, 3.7, and 3.3 for phenanthrene, fluoranthene, and pyrene, respectively. Therefore, high concentrations of DOC increase mass transfer rates of more 8745
dx.doi.org/10.1021/es202983k |Environ. Sci. Technol. 2011, 45, 8741–8747
Environmental Science & Technology
Figure 5. Initial mass transfer rates for DOC concentrations of 0, 0.0226, and 0.226 g L1. These were calculated using eqs 2 and 4 and setting cd or ctot to zero. Standard errors were calculated using error propagation. For the DOC 0.226 g L1 concentration, the factor increases in the initial mass transfer rates relative to those measured in the absence of DOC are shown.
hydrophobic NAPL constituents such as PAHs out a NAPL and into the aqueous phase. Implications for Biodegradation and Toxicity. Figure S9 shows simulated profiles in ctot and cd for acenaphthene and pyrene as representative low and high hydrophobicity HOCs. When the DOC concentration is sufficiently high and there is significant sorption of the HOC, then DOC increases both the mass transfer rates but also the final amounts in the aqueous compartment. Both dissolved and DOC-sorbed HOCs are accessible to the degradation activities of microorganisms28,39 and this could lead to increased biodegradation. However, in the case of bioconcentration and toxicity, final levels reached in the organisms are determined by the dissolved concentrations.8 Despite the increase in HOC masses that are transferred into the aqueous phase with DOC, there is little difference in the time profiles of cd (Figure S9). Therefore, increased mass transfer rates in the presence of DOC may lead to enhanced biodegradation of NAPL-phase HOCs but have a small influence on their final bioconcentration and toxicity. On the other hand, DOC might enhance bioconcentration and toxicity for mobile organisms with short-time exposure to HOC hotspots. Relevance of DOC Enhanced Diffusion. The relevance of a DOC-mediated diffusion pathway has to be kept in mind. The lower DOC 0.0226 g L1 concentration is within the range found in marine algal blooms40 or marine sediment porewaters,41 and did not significantly enhance ∂ctot/∂t. A measurable increase was only observed at the highest DOC concentration of 0.226 g L1, and here only for the more hydrophobic PAHs tested in this study. However, such high DOC concentrations are confined to rather specific environments such as wastewater treatment plants or oil-processing waters containing added xanthan gums. Nevertheless, a potential application for DOC-enhanced transfer is in engineered systems where artificially high concentrations can be applied. One example could be using DOC solutions for pumpand-treat NAPL bioremediation technologies, with the low cost and low toxicity of the DOC also being advantageous. The flushing of a soil contaminated with diesel with a 0.8 g L1 humic acid solution increased both the solubilization and biodegradation of naphthalenes leading to shorter remediation times.42
ARTICLE
Figure S10 shows the calculated enhancement in ktot relative to ktot,dissolved for a series of hypothetical HOCs possessing log KDOC/d values between 1 and 7. These were calculated using the mean kd, kd* and ks* 3 ξ values measured in this study. Given that the importance of DOC-enhanced diffusion increases with DOC concentration and compound sorption (as indicated by eq 7), for “super” hydrophobic compounds this pathway could both dominate and increase mass transfer rates by orders of magnitude, even at environmental DOC concentrations. Future enhanced diffusion investigations might focus on such compounds, which incidentally are the more difficult compounds to (bio)remediate. This study was conducted using DOC of terrestrial origin, and the marine backdrop meant the experiments were conducted in a saline aqueous environment. In the wider context of DOCenhanced diffusion, the relevant properties are reversible sorption of the HOCs and the DOC mobility in the aqueous phase.17 These apply to other types of DOC in different environments including soils and groundwater, and thus DOC-enhanced diffusion will also likely play a role here. However, to move to the next level of describing DOC-enhanced diffusion in terms of fundamental system and chemical properties will require additional study, with emphasis on the nature of the DOC (size distribution, HOC sorption propensity, sorption/desorption kinetics, lability, etc.) but also for HOCs covering a more diverse range in molecular sizes and structures.
’ ASSOCIATED CONTENT
bS
Supporting Information. Full versions of the materials and methods and modeling description; additional figures and tables showing measurement data from this study. This information is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +45 87158463; fax: + 45 46301114; e-mail:
[email protected]. Present Addresses †
Department of Environmental Science, Faculty of Science and Technology, Aarhus University, Frederiksborgvej 399, 4000 Roskilde, Denmark.
’ ACKNOWLEDGMENT This work was funded by the European Community Project FACEiT [STREP Grant 018391 (GOCE)]. Skilled technical help by Birgit W€urz and Jana Reichenbach is gratefully acknowledged. The disposable SPME fibers were kindly provided by Philipp Mayer. ’ REFERENCES (1) Schluep, M.; G€alli, R.; Imboden, D. M.; Zeyer, J. Dynamic equilibrium dissolution of complex nonaqueous phase liquid mixtures into the aqueous phase. Environ. Toxicol. Chem. 2002, 21, 1350–1358. (2) Schluep, M.; Imboden, D. M.; Galli, R.; Zeyer, J. Mechanisms affecting the dissolution of nonaqueous phase liquids into the aqueous phase in slow-stirring batch systems. Environ. Toxicol. Chem. 2001, 20, 459–466. (3) Ortiz, E.; Kraatz, M.; Luthy, R. G. Organic phase resistance to dissolution of polycyclic aromatic hydrocarbon compounds. Environ. Sci. Technol. 1999, 33, 235–242. 8746
dx.doi.org/10.1021/es202983k |Environ. Sci. Technol. 2011, 45, 8741–8747
Environmental Science & Technology (4) Wick, L. Y.; Colangelo, T.; Harms, H. Kinetics of mass transferlimited bacterial growth on solid PAHs. Environ. Sci. Technol. 2001, 35, 354–361. (5) van Loosdrecht, M. C. M.; Lyklema, J.; Norde, W.; Zehnder, A. J. B. Influence of interfaces on microbial activity. Microbiol. Rev. 1990, 54, 75–87. (6) Bayen, S.; ter Laak, T. L.; Buffle, J.; Hermens, J. L. M. Dynamic exposure of organisms and passive samplers to hydrophobic chemicals. Environ. Sci. Technol. 2009, 43, 2206–2215. (7) Stroo, H. F.; Unger, M.; Ward, C. H.; Kavanaugh, M. C.; Vogel, C.; Leeson, A.; Marqusee, J. A.; Smith, B. P. Remediating chlorinated solvent source zones. Environ. Sci. Technol. 2003, 37, 224A–230A. (8) Escher, B. I.; Hermens, J. L. M. Internal exposure: Linking bioavailability to effects. Environ. Sci. Technol. 2004, 38, 455A–462A. (9) Soga, K.; Page, J. W. E.; Illangasekare, T. H. A review of NAPL source zone remediation efficiency and the mass flux approach. J. Hazard. Mater. 2004, 110, 13–27. (10) Pennell, K. D.; Abriola, L. M.; Weber, W. J. Surfactant-enhanced solubilization of residual dodecane in soil columns. 1. Experimental investigation. Environ. Sci. Technol. 1993, 27, 2332–2340. (11) Pennell, K. D.; Jin, M. Q.; Abriola, L. M.; Pope, G. A. Surfactant enhanced remediation of soil columns contaminated by residual tetrachloroethylene. J. Contam. Hydrol. 1994, 16, 35–53. (12) Ghoshal, S.; Pasion, C.; Alshafie, M. Reduction of benzene and naphthalene mass transfer from crude oils by aging-induced interfacial films. Environ. Sci. Technol. 2004, 38, 2102–2110. (13) Johnsen, A. R.; Karlson, U. Evaluation of bacterial strategies to promote the bioavailability of polycyclic aromatic hydrocarbons. Appl. Microbiol. Biotechnol. 2004, 63, 452–459. (14) Law, A. M. J.; Aitken, M. D. Bacterial chemotaxis to naphthalene desorbing from a nonaqueous liquid. Appl. Environ. Microbiol. 2003, 69, 5968–5973. (15) García-Junco, M.; Gomez-Lahoz, C.; Niqui-Arroyo, J. L.; OrtegaCalvo, J. J. Biosurfactant- and biodegradation-enhanced partitioning of polycyclic aromatic hydrocarbons from nonaqueous-phase liquids. Environ. Sci. Technol. 2003, 37, 2988–2996. (16) Mukherji, S.; Weber, W. J. Mass transfer effects on microbial uptake of naphthalene from complex NAPLs (vol 60, pg 750, 1998). Biotechnol. Bioeng. 2001, 75, 130–141. (17) Higuchi, W. I. Effects of interacting colloids on transport rates. J. Pharm. Sci. 1964, 53, 532. (18) de Jong, H. G.; van Leeuwen, H. P.; Holub, K. Voltammetry of metal-complex systems with different diffusion-coefficients of the species involved. Part i. Analytical approaches to the limiting current for the general-case including association/dissociation kinetics. J. Electroanal. Chem. 1987, 234, 1–16. (19) de Jong, H. G.; van Leeuwen, H. P. Voltammetry of metal-complex systems with different diffusion-coefficients of the species involved. Part ii. Behavior of the limiting current and its dependence on association dissociation/kinetics and lability. J. Electroanal. Chem. 1987, 234, 17–29. (20) Johnson, J. C.; Sun, S. B.; Jaffe, P. R. Surfactant enhanced perchloroethylene dissolution in porous media: The effect on mass transfer rate coefficients. Environ. Sci. Technol. 1999, 33, 1286–1292. (21) Prak, D. J. L.; Abriola, L. M.; Weber, W. J.; Bocskay, K. A.; Pennell, K. D. Solubilization rates of n-alkanes in micellar solutions of nonionic surfactants. Environ. Sci. Technol. 2000, 34, 476–482. (22) Mayer, A. S.; Zhong, L. R.; Pope, G. A. Measurement of masstransfer rates for surfactant-enhanced solubilization of nonaqueous phase liquids. Environ. Sci. Technol. 1999, 33, 2965–2972. (23) Benhabib, K.; Town, R. M.; van Leeuwen, H. P. Dynamic speciation analysis of atrazine in aqueous latex nanoparticle dispersions using solid phase microextraction (SPME). Langmuir 2009, 25, 3381–3386. (24) ter Laak, T. L.; Van Eijkeren, J. C. H.; Busser, F. J. M.; Van Leeuwen, H. P.; Hermens, J. L. M. Facilitated transport of polychlorinated biphenyls and polybrominated diphenyl ethers by dissolved organic matter. Environ. Sci. Technol. 2009, 43, 1379–1385. (25) Mayer, P.; Fernqvist, M. M.; Christensen, P. S.; Karlson, U.; Trapp, S. Enhanced diffusion of polycyclic aromatic hydrocarbons in
ARTICLE
artificial and natural aqueous solutions. Environ. Sci. Technol. 2007, 41, 6148–6155. (26) Kramer, N. I.; van Eijkeren, J. C. H.; Hermens, J. L. M. Influence of albumin on sorption kinetics in solid-phase microextraction: Consequences for chemical analyses and uptake processes. Anal. Chem. 2007, 79, 6941–6948. (27) ter Laak, T. L.; ter Bekke, M. A.; Hermens, J. L. M. Dissolved organic matter enhances transport of PAHs to aquatic organisms. Environ. Sci. Technol. 2009, 43, 7212–7217. (28) Smith, K. E. C.; Thullner, M.; Wick, L. Y.; Harms, H. Sorption to humic acids enhances polycyclic aromatic hydrocarbon biodegradation. Environ. Sci. Technol. 2009, 43, 7205–7211. (29) van Leeuwen, H. P.; Buffle, J. Chemodynamics of aquatic metal complexes: From small ligands to colloids. Environ. Sci. Technol. 2009, 43, 7175–7183. (30) Buffle, J.; Wilkinson, K. J.; van Leeuwen, H. P. Chemodynamics and bioavailability in natural waters. Environ. Sci. Technol. 2009, 43, 7170–7174. (31) Poerschmann, J.; Zhang, Z.; Kopinke, F.; Pawliszyn, J. Solid phase microextraction for determining the distribution of chemicals in aqueous matrices. Anal. Chem. 1997, 69, 597–600. (32) van Leeuwen, H. P. Metal speciation dynamics and bioavailability: Inert and labile complexes. Environ. Sci. Technol. 1999, 33, 3743–3748. (33) van Leeuwen, H.; Cleven, R.; Buffle, J. Voltammetric techniques for complexation measurements in natural aquatic media - role of the size of macromolecular ligands and dissociation kinetics of complexes. Pure Appl. Chem. 1989, 61, 255–274. (34) Burkhard, L. P. Estimating dissolved organic carbon partition coefficients for nonionic organic chemicals. Environ. Sci. Technol. 2000, 34, 4663–4668. (35) Mukherji, S.; Peters, C. A.; Weber, W. J. Mass transfer of polynuclear aromatic hydrocarbons from complex DNAPL mixtures. Environ. Sci. Technol. 1997, 31, 416–423. (36) MacKenzie, K.; Georgi, A.; Kumke, M.; Kopinke, F. D. Sorption of pyrene to dissolved humic substances and related model polymers. 2. Solid-phase microextraction (SPME) and fluorescence quenching technique (FQT) as analytical methods. Environ. Sci. Technol. 2002, 36, 4403–4409. (37) Pinheiro, J. P.; Mota, A. M.; Goncalves, M. L. S.; van Leeuwen, H. P. The pH effect in the diffusion coefficient of humic matter: Influence in speciation studies using voltammetric techniques. Colloids Surf., A 1998, 137, 165–170. (38) Sutton, R.; Sposito, G. Molecular structure in soil humic substances: The new view. Environ. Sci. Technol. 2005, 39, 9009–9015. (39) Haftka, J. J. H.; Parsons, J. R.; Govers, H. A. J.; Ortega-Calvo, J. Enhanced kinetics of solid-phase microextraction and biodegradation of polycyclic aromatic hydrocarbons in the presence of dissolved organic matter. Environ. Toxicol. Chem. 2008, 27, 1526–1532. (40) Simjouw, J.; Mulholland, M. R.; Minor, E. C. Changes in dissolved organic matter characteristics in Chincoteague Bay during a bloom of the pelagophyte Aureococcus anophagefferens. Estuaries 2004, 27, 986–998. (41) Deflandre, B.; Gagne, J. Estimation of dissolved organic carbon (DOC) concentrations in nanoliter samples using UV spectroscopy. Water Res. 2001, 35, 3057–3062. (42) van Stempvoort, D. R.; Lesage, S.; Novakowski, K. S.; Millar, K.; Brown, S.; Lawrence, J. R. Humic acid enhanced remediation of an emplaced diesel source in groundwater. 1. Laboratory-based pilot scale test. J. Contam. Hydrol. 2002, 54, 249–276.
8747
dx.doi.org/10.1021/es202983k |Environ. Sci. Technol. 2011, 45, 8741–8747
ARTICLE pubs.acs.org/est
Oxidative Dissolution of Biogenic Uraninite in Groundwater at Old Rifle, CO Kate M. Campbell,†,* Harish Veeramani,‡,# Kai-Uwe Ulrich,§,3 Lisa Y. Blue,§,O Daniel E. Giammar,§ Rizlan Bernier-Latmani,‡ Joanne E. Stubbs,||,( Elena Suvorova,‡ Steve Yabusaki,^ Juan S. Lezama-Pacheco,|| Apurva Mehta,|| Philip E. Long,^ and John R. Bargar|| †
U.S. Geological Survey, 3215 Marine Sreet, Boulder, Colorado 80303, United States cole Polytechnique Federale de Lausanne, Lausanne, CH 1015, Switzerland Environmental Microbiology Laboratory, E § Washington University in Saint Louis, Saint Louis, Missouri 63130, United States Stanford Synchrotron Radiation Lightsource, 2575 Sand Hill Road, Menlo Park, California 94025, United States ^ Pacific Northwest National Laboratory, Richland, Washington 99352, United States
)
‡
bS Supporting Information ABSTRACT: Reductive bioremediation is currently being explored as a possible strategy for uranium-contaminated aquifers such as the Old Rifle site (Colorado). The stability of U(IV) phases under oxidizing conditions is key to the performance of this procedure. An in situ method was developed to study oxidative dissolution of biogenic uraninite (UO2), a desirable U(VI) bioreduction product, in the Old Rifle, CO, aquifer under different variable oxygen conditions. Overall uranium loss rates were 50100 times slower than laboratory rates. After accounting for molecular diffusion through the sample holders, a reactive transport model using laboratory dissolution rates was able to predict overall uranium loss. The presence of biomass further retarded diffusion and oxidation rates. These results confirm the importance of diffusion in controlling in-aquifer U(IV) oxidation rates. Upon retrieval, uraninite was found to be free of U(VI), indicating dissolution occurred via oxidation and removal of surface atoms. Interaction of groundwater solutes such as Ca2+ or silicate with uraninite surfaces also may retard in-aquifer U loss rates. These results indicate that the prolonged stability of U(IV) species in aquifers is strongly influenced by permeability, the presence of bacterial cells and cell exudates, and groundwater geochemistry.
’ INTRODUCTION The legacy of uranium (U) ore extraction and processing has left many sites with U-contaminated groundwater, even after extensive tailings removal and remediation projects. Reductive bioremedation is currently being explored as a possible strategy for sites with excess groundwater uranium concentrations. Bioremediation at the Old Rifle site (Colorado) has been performed by amending the aquifer with acetate to stimulate the growth of natural metal- and sulfate-reducing microbial communities, and the reduction of U(VI) to U(IV).13 Owing to the relatively low solubility of U(IV), this process substantially decreases the concentration of dissolved U in the groundwater.1,4,5 Uraninite (UO2(s)) is one of the most desirable products, and its longevity under reducing conditions is evidenced by its importance in low-temperature sedimentary uranium ore deposits,6,7 and recent work by our group (unpublished results) shows it may occur in aquifers following bioreduction. Metal- and sulfate-reducing bacteria are capable of enzymatically reducing U(VI) to uraninite (e.g., refs 810), as well as to other forms of U(IV), such as U(IV) adsorbed to biomass and minerals.1114 Of these products, uraninite is the only well-characterized material for r 2011 American Chemical Society
which thermodynamic and kinetic constants are known. It therefore provides the best available proxy for other forms of U(IV) and can be used to constrain U(IV) stability in sediments. Biogenic uraninite is almost universally reported to be a nanoparticulate (25 nm) solid, although it tends to form larger agglomerates15,16 and exhibits structural similarity even when produced by a phylogenetically diverse group of organisms.10 Whereas the oxidative dissolution of biogenic uraninite has been studied in the laboratory (e.g., refs 1721), its behavior under aquifer conditions has not been investigated. The rates of uraninite oxidation in groundwater are expected to differ from those measured under well-stirred laboratory conditions because of the sensitivity of uraninite reaction rates to the presence of various trace groundwater solutes22 and to diffusionlimited conditions in aquifer pore spaces.23 Developing a quantitative understanding of the effect of chemical conditions and kinetics of U(IV) oxidation under aquifer conditions is therefore important to Received: February 11, 2011 Accepted: September 12, 2011 Revised: August 26, 2011 Published: September 12, 2011 8748
dx.doi.org/10.1021/es200482f | Environ. Sci. Technol. 2011, 45, 8748–8754
Environmental Science & Technology understanding the performance of stimulated reductive bioremediation and developing appropriate site management practices. This study was designed to elucidate the effect of groundwater geochemistry, particularly variable dissolved oxygen (DO), on oxidative dissolution of uraninite. Of potential oxidants (e.g., DO, nitrate, manganese oxides, iron oxides, and natural organic matter17,21,2427), DO is particularly important because of its relative abundance. At circumneutral pH in the presence of dissolved inorganic carbon (DIC), oxidation of biogenic uraninite by DO proceeded by oxidation of U(IV) atoms, followed by rapid removal of surface U(VI) from the surface by bicarbonate complexation.19 The same study showed that the solubility and dissolution rates of biogenic uraninite are similar to those of coarse-grained UO2.00.18,19 The objectives of this study were to extend this previous work into aquifer conditions where uraninite would be exposed to ambient DO, bicarbonate, Ca2+, silicate, sulfate, and other species. We deployed biogenic uraninite into wells at the Old Rifle site using novel membrane-walled cells that allowed for diffusive solute exchange with groundwater but prevented dispersive loss of uraninite particles and invasion of bacteria. Two wells with contrasting groundwater chemistry were utilized to probe the impact of the master variable, DO: one being a typical oxic background well, and the other located in a zone of natural bioreduction, characterized by suboxic conditions, higher Fe(II) concentrations, and evidence of U(IV) in nearby sediments.
’ MATERIALS AND METHODS Rifle, CO Field Site. The shallow, unconfined aquifer has residual U contamination from U ore processing. The groundwater composition has been reported previously.1,28 Generally, the aquifer has low but measurable amounts of DO, has ∼8 mM sulfate, and Ca2+ and HCO3 are typically near equilibrium with calcite solubility. Groundwater U concentrations are 0.41.8 μM, and the primary aqueous species are U(VI)-carbonate and Ca U(VI)-carbonate ternary species at typical groundwater pH (pH 7.27.4). Nitrate concentrations are low (<20 μM), and it is unlikely to compete with DO as a potential oxidant. Wells used for uraninite incubations were chosen to bracket a range of DO concentrations. The wells used for this study were B-02, a background well, and P-103, located in a naturally bioreduced zone. The groundwater at B-02 is typical of the Rifle site. In contrast, P-103 is naturally suboxic to anoxic with reduced U(IV) phases present in the surrounding sediment. Neither B-02 nor P-103 has been exposed to acetate amendment. Biogenic Uraninite Synthesis. Biogenic uraninite installed as cleaned oxide was precipitated by Shewanella oneidensis strain MR-1 to facilitate comparison to previous studies.29,30 The solid was sequentially washed with 1 M NaOH, hexane and 100 mM NaHCO3 to remove biomass and residual U(VI). Biogenic uraninite to be installed with its host biomass intact was produced under similar conditions by a Shewanella isolate from Rifle groundwater (as required by the Colorado Environmental Protection Agency), but was not washed with NaOH, bicarbonate, or hexane so that the biomass-uraninite association was preserved. The cleaned uraninite produced by the isolate was structurally comparable to the uraninite produced by MR-1 (data not shown). In addition, this material contained a small fraction of monomeric U(IV) complexes associated with the biomass.13 A uraninite slurry (either with biomass (P103-BIOMASS-slurry, B02-BIOMASS-slurry) or without biomass (P103-CLN-slurry,
ARTICLE
B02-CLN-slurry)) was placed directly into a sample tube for deployment, used to create a UO2-doped gel puck for mass balance determination (described below), or archived for characterization and dissolution rate measurements. The biomass associated with the uraninite was not expected to grow under deployment conditions due to the absence of electron donor. Biogenic Uraninite-Doped Gel Pucks for Mass Balance. Polyacrylamide gel pucks31 doped with uraninite were prepared to evaluate mass loss rates during oxidative dissolution; mass loss could not be evaluated from slurry samples. The gels acted as an inert matrix that did not dissolve nor substantially affect the uraninite reactivity.31,32 Biogenic uraninite was homogeneously distributed within the gel puck, with an average variation of 10%. One half of each puck was retained to determine total U loading per unit weight of each gel. A fraction of the other half was cut into slabs (approximately 2 cm 3 mm 2 mm) and transferred to a sample cell (described below) for deployment. Scanning electron microscopy (SEM) measurements on the gel pucks show the uraninite to be present in the pucks in loose agglomerations of individual nanoparticles, similar to particles not embedded in gel (Supporting Information (SI) Figure S1). Upon recovery from the wells, the gel fractions were weighed and equilibrated with 5 mL of concentrated HNO3. The solution was then analyzed for U by ICPOES (Perkin-Elmer, Plasma 2000). Uranium loss from the gel pucks was determined by calculating the difference in U concentration in gel pucks before and after deployment. Gels were doped with washed uraninite (P103CLN-gel, B02-CLN-gel) or with biomass-associated uraninite (P103-BIOMASS-gel, B01-BIOMASS-gel). Permeable Sample Cells, Assembly, And Deployment. Permeable sample cells were developed to maintain biogenic uraninite in contact with groundwater. Cells were constructed from 2 mL polyethylene tubes with three vertical slots along the body of the tube (∼50% of surface area removed). Cellulose ester dialysis membrane (Spectrum Spectra/Por, 10 000 Da) was bonded to the tube to cover the slots. The tubes were loaded in the field in a N2-filled anaerobic chamber with either a suspension of biogenic uraninite (∼50 to 300 mg of uraninite per tube, for spectroscopy) or uraninite-doped polyacrylamide gel pucks (∼510 mg uraninite per tube, for mass balance). The maximum diffusive path within the gel pucks in tubes was <3 mm (SI Figure S2). Diffusion of KNO3 through the membrane on a tube was used to determine the overall mass transfer coefficient for the tubes (1.08 104 cm/s); a tube loaded with 0.1 M KNO3 fully equilibrated with 100 mL of surrounding solution within two to three days, whereas diffusion of water and solutes through a gel puck is on the order of hours for a 23 mm gel.31,32 The tubes were secured in a plastic holder (SI Figure S2); the entire assembly was deployed in the well about 5 m below the ground surface and at least 1 m below the water table for the duration of the experiment. The entire well below the water table was screened. Uraninite was deployed on February 11, 2009 and recovered on May 5, 2009 (83 days of reaction). A second set of gel puck samples was deployed on July 16, 2009 and recovered on October 26, 2009 (102 days of reaction). Both NaOH-washed uraninite and biomassassociated uraninite slurries were deployed in July, whereas only NaOH-washed uraninite was deployed in February (SI Table S1). Characterization of Biogenic Uraninite. Lattice parameter, crystallite size, local structure, and the composition of uraninite recovered from the slurry-filled tubes after in situ reaction was determined by synchrotron powder X-ray diffraction (XRD), 8749
dx.doi.org/10.1021/es200482f |Environ. Sci. Technol. 2011, 45, 8748–8754
Environmental Science & Technology X-ray absorption near edge structure (XANES), extended X-ray absorption fine structure (EXAFS), and chemical digestion. XRD data were collected at SSRL beamline 72 in transmission geometry on a 6-circle Huber diffractometer using an energy dispersive Vortex detector placed behind 1 mrad Soller slits. X-ray wavelengths (0.7646 - 0.7647 Å) were calibrated using a LaB6 standard. A subset of each archived and deployed uraninite slurry sample was dried anaerobically, gently disaggregated, and loaded into 0.3 mm diameter borosilicate glass capillaries and sealed with epoxy. Data were collected from 1.0 to 14.5 Å1 Q, 0.02 Å1 step size. The recorded count rates were normalized to those of the incident beam. Lattice parameters and nominal crystallite sizes were determined by Rietveld refinement using the Generalized Structure Analysis System of Larson and Von Dreele33 and the user interface EXPGUI.34 For XANES and EXAFS analysis, a subset of deployed slurry sample (∼10 mg) was either dried and diluted with boron nitride or used wet without dilution, loaded into Al X-ray cells (1 mm thick) with Kapton windows and stored in the anaerobic chamber until immediately prior to analysis. The samples were mounted in a liquid N2 cryostat. U LIII-edge transmission spectra were collected using a typical beam size of 0.2 to 0.5 mm at SSRL beamlines 112 and 41, using detuned Si (220) double-crystal monochromators. EXAFS spectra were processed using SIXPACK.35 Backscattering phase and amplitude functions required for fitting of spectra were obtained from FEFF8.36 Chemical digestions of an aliquot of deployed slurry sample were performed in sequence: separation of the dissolved phase, 1 mM HCl wash (adsorbed phase), aqua regia digest (digest). A subset of each archived sample of uraninite was digested in aqua regia for comparison. The fractions were analyzed for a suite of 26 elements by ICP-MS (Agilent 7500ce). Electron Microscopy. Samples for electron microscopy were prepared from gel pucks by drying the pucks in an anaerobic chamber atmosphere at 30 C, cutting thin sections of the resin, and placing sections on a copper grid. Bright field transmission electron microscopy (BFTEM), high-resolution TEM (HRTEM) and selected area electron diffraction (SAED) were performed with a FEI CM300UT/FEG microscope (Eindhoven, Netherlands). Phase identification was derived from SAED patterns and Fourier transforms of the HRTEM images using the JEMS software.37 Direct chemical information was obtained by X-ray energy dispersive spectroscopy (EDS) mapping. Dissolution Rate of UO2 in Rifle Artificial Ground water. The dissolution rate of NaOH-washed biogenic uraninite was measured in Rifle artificial groundwater and ultrapure water using stirred continuous-flow reactors (CFR) previously used for the quantification of UO2 dissolution rates in solutions of simpler compositions.19,20 Duplicate reactors (12.6 mL) were loaded with ∼1 g/L suspensions of UO2 and sealed with 0.2 μm polycarbonate filter membranes (Millipore). Influent solutions were prepared with compositions to simulate Rifle groundwater (Rifle Artificial Groundwater, RAGW, SI Table S2). These solutions were prepared at pH 8.5 and contained 1 mM dissolved inorganic carbon as well as Na+, K+, Ca2+, Mg2+, Cl, and SO42-. A pH of 8.5 was selected to enable future comparisons of the rates with those previously measured.20 Influent solutions were sparged with gas to provide the desired DO concentration. For the first stage of the experiment, the influents were sparged with a 5%/95% H2/N2 mixture in the presence of a Pd catalyst to remove residual oxygen; CO2 loss was not significant. In the second stage, the solution was sparged with a gas mixture of 1%
ARTICLE
O2 and 99% N2, and the final stage used air as the sparging gas (8.6 mg/L O2). Each stage of the experiment was operated for a time equivalent to 40 residence times or longer. The reactor effluents were sampled, filtered (0.020 μm, Whatman, alumina membranes), acidified with 1% HNO3, and analyzed by ICP-MS for dissolved U. The effluent was regularly monitored for the flow-rate, pH, and DO as measured with an in-line electrode (Microelectrodes, Inc.). Dissolution rates normalized to surface area (Rn, mol m2 min1) were calculated from the steady-state effluent concentration of U ([U]ss, mol/L), the hydraulic residence time (tres, ∼6 min for this study), the specific surface area of the biogenic uraninite (SSA, 50.1 m2/g), and the solids loading of the reactor ([solid], g/L) (eq 118). Rn ¼
½Uss tres 3 SSA 3 ½solid
ð1Þ
The steady-state effluent concentration was determined by finding the best fit of a simple model for CFR behavior to the experimental data. Intrinsic dissolution rate constants were then calculated from the dissolution rates using an approach used previously.18 Reactive Transport Model. A multicomponent reactive transport model was developed to predict diffusion-limited biogenic uraninite dissolution in membrane tubes under groundwater conditions present in well B-02. Key elements of the model include a 6 mm diameter zone of 6 mg of uraninite centered in the 8 mm diameter membrane cell with advection of ambient solution outside the membrane. Symmetry was invoked allowing modeling of half the domain with a no-flux boundary condition along the line of symmetry. Diffusion was assumed to be the only active transport process in the tube and membrane. The diffusion coefficient of the membrane was adjusted to reproduce the results of the membrane tube KNO3 diffusion experiment. The initial aqueous condition within the cylinder was pH 7 solution without U or DO. A pH 7 solution in equilibrium with an oxygen partial pressure of 0.00243 atm and no U was assumed to be flowing tangentially by the membrane (SI Figure S3). The reaction network employed H+, UO22+ (as a proxy for dissolved U(VI)), and O2(aq) as primary species and solved with the following reactions and stability constants: OH- þ Hþ ¼ H2 O ðlog K ¼ 13:991Þ
ð2Þ
O2 ðgÞ ¼ O2 ðaqÞ ðlog K ¼ 2:893Þ
ð3Þ
UO2ðsÞ þ 4Hþ ¼ U4þ þ 2H2 O ðlog K ¼ 1:5Þ
ð4Þ
U4þ þ H2 O þ
1 O2ðgÞ ¼ UO2 2þ þ 2Hþ ðlog K ¼ 32:4999Þ 2
ð5Þ A transition state rate law was employed for uraninite; no other solid phases were needed in the model. An intrinsic rate of 109.5 mol/m2/s and surface area 50.1 m2/g was used for uraninite.19 The research code used for these calculations is based on finite column discretization using a Cartesian grid. A global implicit solution of the transport and reactions was used with a backward Euler numerical scheme. 8750
dx.doi.org/10.1021/es200482f |Environ. Sci. Technol. 2011, 45, 8748–8754
Environmental Science & Technology
’ RESULTS AND DISCUSSION Dissolution Rate of Biogenic Uraninite in RAGW. The dissolution rate of biogenic uraninite increased with increasing DO in the RAGW/CFR experiments (Figure 1). Because the solutions contained 1 mM DIC and 5 mM Ca, the U(VI) produced by uraninite oxidation is mobilized from the surface through formation of soluble CaU(VI)-carbonate complexes.19 The major cations and anions present in the artificial groundwater, especially DIC, increased the rate of uraninite dissolution approximately 3-fold at high pO2 compared to ultrapure water. Groundwater Composition. Groundwater compositions at wells B-02 (oxic) and P-103 (suboxic/anoxic) are listed in SI Table S3. The primary difference between these wells was DO concentration; B-02 had 5- to 10-fold higher DO than P-103 during the 83-day experiment, and 2- to 3-fold higher during the 102-day experiment. Variation in DO values was driven by summer peak streamflow in the Colorado River, which seasonally elevates the water table and traps DO. P-103 had consistently higher Fe(II) concentrations (1527 μM compared to 49 μM in B-02). The major ion chemistry, pH, and alkalinity of the wells were otherwise similar. Thus, contrasting chemical behavior for uraninite recovered from these wells should be attributed primarily to differences in DO.
Figure 1. Uranium dissolution rates (Rn) for NaOH-washed biogenic uraninite dissolution in ultrapure water (DIW, 0.) and Rifle artificial groundwater (RAGW, 9) at different levels of dissolved oxygen at pH ≈ 8.5 and with 1 mM DIC in the CFR experiment. Also shown are overall U loss rates from gel puck samples deployed in well B-02 (b, B02-CLNgel, 102-day reaction). P103-CLN-gel (83-day and102-day reaction) showed negligible loss and was therefore not included. Error bars are mostly smaller than symbol size.
ARTICLE
Uraninite Mass Balance in Field-Reacted Gel Pucks. During the 102-day deployment, more dissolution was observed in the NaOH-cleaned uraninite in B-02 (55% lost) compared to P-103 (no substantial loss) (Figure 2). The decrease in U loss cannot be attributed to the presence of monomeric U(IV), which oxidizes faster than uraninite (unpublished results). The presence of biomass, however, reduced the amount of uraninite lost to only 11% in both P-103 and B-02 (Figure 2). Since the biomass should not have been metabolically active or provided a significant redox buffer, we conclude that its impact was due to retarded diffusion of U and DO. Molecular-Scale Structure and Oxidation State of Biogenic Uraninite. XANES spectra for the archived and deployed uraninite slurry samples (NaOH-cleaned and biomass-associated) are consistent with that of biogenic uraninite (SI Figure S4), indicating that U(VI) products did not accumulate during in-well exposure. X-ray diffraction data and Rietveld refinement results are given in Table 1 and fits are shown in SI Figure S5. The lattice parameters refined to 5.464, 5.464, and 5.468 Å for the unreacted archive sample, B-02-CLN-slurry and P-103-CLN-slurry samples, respectively. These values are in agreement with one another and with previously published lattice parameters for both biogenic and abiotic uraninite (5.46 5.47 Å).30 In comparison, the lattice constant for U4O9 is 5.44 Å. The crystallite sizes estimated from Rietveld refinement (∼2.1 nm) do not vary significantly among the reacted and archive samples, nor do the calculated strains (∼3%), which are interpreted as semiquantitative measures of strain gradients between crystallite cores and surfaces.30 The fit derived crystallite size compares well to the TEM-derived estimates of 1.5 nm (SI Figure S6). These observations indicate that the uraninite unit cell structure and size were not substantially altered by interaction with the groundwater during the incubation period. No other phases (e.g., U(VI) minerals) were observed in any postreacted slurry sample. The EXAFS, Fourier transforms (FTs, Figure 3), and fit results (SI Table S4) from the NaOH-cleaned slurry samples are similar to one-another. The observed UU distances, and coordination numbers (CNs) are consistent with biogenic uraninite.10,30 Because crystallite size does not vary significantly, the increase in the 3.8 Å (R+dR) UU FT frequency following submersion in the Rifle aquifer indicates that the particles became more ordered after reaction in groundwater, implying increased stability. Mechanism of Uraninite Corrosion. Corrosion of biogenic uraninite (oxidative dissolution) follows one of two different
Figure 2. U losses from gel pucks deployed in wells after 102 days compared to an aliquot of unreacted gel puck. Left panel (A) is NaOH-washed biogenic uraninite (B02-CLN-Gel and P103-CLN-Gel), and right panel (B) is uraninite including biomass from rifle isolate (B02- BIOMASS -Gel and P103- BIOMASS -Gel). 8751
dx.doi.org/10.1021/es200482f |Environ. Sci. Technol. 2011, 45, 8748–8754
Environmental Science & Technology
ARTICLE
Table 1. Results of Rietveld Refinement of Powder X-ray Diffraction Data on Unreacted Samples and NaOH-Cleaned Slurry Samples Deployed for 102 Days at Riflea sample
a (Å)
crystallite size (nm)
strain (%)
Rwp
R(F2)
unreacted biogenic uraninite (control)
5.464(1)
2.07(5)
2.8(4)
0.032
0.016
B02-CLN-slurry P103-CLN-Slurry
5.464(1) 5.468(1)
2.15(5) 2.09(4)
3.1(4) 2.7(3)
0.034 0.031
0.016 0.016
Estimated standard deviation (ESD) in final digit is shown in parentheses. Unreacted control is the parent material, which was not incubated in Rifle wells. These are model-dependent, statistical ESDs. In the cases of crystallite size and strain gradient, they are based on a simple, empirical peak shape profile that assumes all peak broadening arises from crystallite size and strain gradient. The actual ESDs are likely several times larger. Rwp is the crystallographic R-factor for the whole pattern and R(F2) is the R-factor for the Bragg peaks. a
Figure 3. EXAFS (left) and corresponding Fourier Transforms (right) for biogenic uraninites incubated at the Old Rifle aquifer in membrane tubes (samples B02-CLN-Slurry and P103-CLN-Slurry, 102-day reaction, and unreacted archive sample). Dotted lines represent fits to the data.
processes.18,22 Under conditions of high DO, low dissolved bicarbonate, and/or diffusion limited conditions, oxidation of surface U(IV) atoms can be fast relative to dissolution of the oxidized surface species, resulting in accumulation of U(VI) solids. In contrast, if bicarbonate is present in relative abundance, DO is moderate or low in concentration, and diffusion is not limiting, then oxidized surface U(VI) atoms can be removed more rapidly than they can accumulate. There was no evidence for accumulation of U(VI) or UO2+x in the field-reacted samples, in accordance with the presence of excess bicarbonate. Moreover, reactive transport modeling predicted nonoxidative uraninite dissolution to be insignificant. Therefore, we conclude that the oxidized surface U atoms were removed by dissolved bicarbonate, and uraninite surface atom oxidation was the rate-limiting step. Effect of DO on Uraninite Oxidation under Ambient Groundwater Conditions. In well B-02, the rate of dissolution is approximately 2 orders of magnitude slower than what we would predict based on the RAGW CFR results (Figure 1). Possible explanations for this observation include: (a) kinetic limitation of diffusion across the cell membranes, (b) passivation of uraninite by adsorption, (c) structural incorporation of groundwater solutes not present in RAGW, and/or (d) cementation of uraninite nanoparticles (i.e., mechanical isolation of uraninite from solution). The differing pH between the CFR (8.5) and Rifle groundwater (7.2) should not significantly affect the mechanism or rates of oxidation because uranyl-carbonato
complexes are the dominant aqueous U(VI) species at both pH values.19 A reactive transport model was used to estimate the affect of transmembrane diffusion of DO and U(VI) on the overall loss rates under the groundwater conditions present in well B-02. Model results indicate that 52% of the uraninite (∼3 mg of uraninite in the model) should be lost over the 102-day experiment. This compares well with the amount lost from the gel pucks (without biomass), estimated to be 50% (35 mg of uraninite). This comparison and model sensitivity analysis (SI) suggests that diffusion is a key control on the rate of uraninite dissolution in the system (explanation (a)). Since the prediction uses a laboratory-derived dissolution rate for uraninite, the in-gel U loss rate (no diffusive barriers) is overall consistent with that derived from published laboratory values.19 In an aquifer sedimentary environment, diffusion is complex and controlled by a number of factors, including pore size and shape, particle size, porosity, and diffusivity. A spectrum of diffusion limited conditions are expected. The experiment presented here is expected to be analogous to a condition intermediate in the range of pore-scale environments present in aquifers; the prolonged stability of uraninite in wells may also apply to diffusion-limiting sediments. In high conductivity flow regimes, the effect of diffusion on oxidative dissolution would be expected to be relatively minor. However, in fine-grained sediments (such as those common in naturally bioreduced zones), the presence of diffusive barriers should be expected to strongly retard U oxidation. 8752
dx.doi.org/10.1021/es200482f |Environ. Sci. Technol. 2011, 45, 8748–8754
Environmental Science & Technology Precipitation of iron sulfides and calcite and accumulation of biomass during biostimulation is expected to decrease sediment permeability, which would provide additional diffusive barriers in the subsurface. Silicate and Ca have previously been shown to substantially inhibit uraninite corrosion when adsorbed on the surface of synthetic uraninite at concentrations similar to Rifle groundwater.22,38 Dissolved silicate was absent in RAGW but was present at ∼0.3 mM in the groundwater of the present study. The reaction of DO with Fe2+(aq) should preclude Fe(II) inhibiting dissolution. Additionally, there was no difference in extractable Fe between the archive sample, P103-CLN-slurry, and B02-CLN-slurry, indicating that Fe(II) oxidation did not supplant uraninite oxidation. Therefore, it is possible that the presence of Ca, Si, or other solute-uraninite interactions could have contributed to the observed slow uraninite loss rates (explanation (b)). The chemical extraction results show that Ca was substantially enriched in the strongly bound fraction compared to the unreacted uraninite (0.20.3 mol Ca/mol U, SI Table S5). Ca was approximately 10-fold more concentrated in the strongly bound fraction than in the adsorbed phase. EDS mapping also showed Ca (as well as Si and P) to be associated with biogenic uraninite in gel pucks from after reaction in well B-02 (SI Figures S8 and S9). There were no substantial differences in chemical extraction results between wells B-02 and P-103. Incorporation of Ca into uraninite structural sites beyond a dopant concentration of approximately 1% (0.01 mol Ca/mol U) would have caused a distortion and weakening of the EXAFS UU shell, as seen for Mn(II),29 but was not observed in the present data set. No evidence of crystalline calcite or other phases besides uraninite was detected by XRD or HRTEM (SI Figure S5), nor was there evidence that uraninite nanoparticles were cemented together by such phases. Therefore, Ca must have been present as an adsorbed species and/or an amorphous solid, and explanations (c) and (d) can be neglected. Environmental Implications. It has been hypothesized that the oxidation of biogenic uraninite in aquifers would proceed faster than synthetic or bulk UO2 because of its nanoparticulate nature.9 The present study has shown that the stability of biogenic uraninite with respect to oxidative dissolution is complex and is likely to be strongly influenced by the groundwater geochemistry and association with diffusional barriers. In particular, diffusion limitations to the rates of supply and removal of DO and U(VI) to/from uraninite, as well as association with biomass and adsorbed solutes, are expected to substantially prolong the lifespan of biogenic uraninite, even when the groundwater is oxic. These same factors are expected to retard oxidation of other forms of U(IV) in aquifers. The slow oxidation/release of uranium observed in this study, even under oxic conditions, suggests than uranium may be sustainably released from either naturally or artificially reduced sediments over long periods of time. This general model offers a potential mechanism to explain plume persistence at contaminated sites such as Old Rifle that contain sediment-hosted naturally reduced uranium.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional methods, SEM/TEM images of uraninite embedded in gel pucks, constituents of RAGW, model schematic and results, XRD data and fitting results, XANES spectra, EXAFS fits, chemical extraction data,
ARTICLE
STEM, EDS mapping and spectra. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (303) 541-3035; fax (303) 541-3084; e-mail: kcampbell@ usgs.gov. Present Addresses #
Department of Geosciences, Virginia Polytechnic and State University, Blacksburg, Virginia 24060 3 BGD Soil and Groundwater Laboratory, 01219 Dresden, Germany O Covalent Research Technologies, LLC, Lexington, Kentucky 40507 ( Consortium for Advanced Radiation Sources, University of Chicago, Chicago, Illinois 60637
’ ACKNOWLEDGMENT We thank Richard Dayvault, David Traub, Carol Morris, Aaron Gooch, Joe Rogers, Michael Hay, and Tanya Gallegos. Funding for this project was provided by a DOE-OBER grant to SLAC (work package 2009-SLAC-10006), grant DE-FG0206ER64227 to EPFL, Swiss NSF grants 20021-113784 and 200020-126921/1, and USGS/NRC postdoctoral program. Portions of this research were carried out at the Stanford Synchrotron Radiation Light Source, a national user facility operated by Stanford University on behalf of the U.S. Department of Energy Office of Basic Energy Sciences. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government. ’ REFERENCES (1) Anderson, R. T.; Vrionis, H. A.; Ortiz-Bernad, I.; Resch, C. T.; Long, P. E.; Dayvault, R.; Karp, K.; Marutzky, S.; Metzler, D. R.; Peacock, A.; White, D. C.; Lowe, M.; Lovley, D. R. Stimulating the in situ activity of Geobacter species to remove uranium from the groundwater of a uranium-contaminated aquifer. Appl. Environ. Microbiol. 2003, 69 (10), 5884–5891. (2) Finneran, K. T.; Anderson, R. T.; Nevin, K. P.; Lovley, D. R. Potential for bioremediation of uranium-contaminated aquifers with microbial U(VI) reduction. Soil Sediment Contam. 2002, 11 (3), 339–357. (3) Lovley, D. R.; Phillips, E. J. P. Reduction of uranium by Desulfovibrio desulfuricans. Appl. Environ. Microbiol. 1992, 58 (3), 850–856. (4) Lovley, D. R.; Phillips, E. J. P. Bioremediation of uranium contamination with enzymatic uranium reduction. Environ. Sci. Technol. 1992, 26 (11), 2228–2234. (5) Wu, W. M.; Carley, J.; Gentry, T.; Ginder-Vogel, M. A.; Fienen, M.; Mehlhorn, T.; Yan, H.; Caroll, S.; Pace, M. N.; Nyman, J.; Luo, J.; Gentile, M. E.; Fields, M. W.; Hickey, R. F.; Gu, B.; Watson, D.; Cirpka, O. A.; Zhou, J.; Fendorf, S.; Kitanidis, P. K.; Jardine, P. M.; Criddle, C. S. Pilot-scale in situ bioremedation of uranium in a highly contaminated aquifer. 2. Reduction of U(VI) and geochemical control of U(VI) bioavailability. Environ. Sci. Technol. 2006, 40 (12), 3986–3995. (6) Bargar, J. R.; Bernier-Latmani, R.; Giammar, D. E.; Tebo, B. M. Biogenic uraninite nanoparticles and their importance for uranium remediation. Elements 2008, 4 (6), 407–412. (7) Plant, J. A.; Simpson, P. R.; Smith, B.; Windley, B. F. Uranium ore deposits - Products of the radioactive earth. Rev. Mineral. and Geochem. 1999, 38, 254–319. (8) Lovley, D. R.; Phillips, E. J. P.; Gorby, Y. A.; Landa, E. R. Microbial reduction of uranium. Nature 1991, 350 (6317), 413–416. (9) Wall, J. D.; Krumholz, L. R. Uranium reduction. Annu. Rev. Microbiol. 2006, 60, 149–166. 8753
dx.doi.org/10.1021/es200482f |Environ. Sci. Technol. 2011, 45, 8748–8754
Environmental Science & Technology (10) Sharp, J. O.; Schofield, E. J.; Veeramani, H.; Suvorova, E. I.; Kennedy, D. W.; Marshall, M. J.; Mehta, A.; Bargar, J. R.; Bernier-Latmani, R. Structural similarities between biogenic uraninites produced by phylogenetically and metabolically diverse bacteria. Environ. Sci. Technol. 2009, 43 (21), 8295–8301. (11) Khijniak, T. V.; Slobodkin, A. I.; Coker, V.; Renshaw, J. C.; Livens, F. R.; Bonch-Osmolovskaya, E. A.; Birkeland, N. K.; MedvedevaLyalikova, N. N.; Lloyd, J. R. Reduction of uranium(VI) phosphate during growth of the thermophilic bacterium Thermoterrabacterium ferrireducens. Appl. Environ. Microbiol. 2005, 71 (10), 6423–6426. (12) Fletcher, K. E.; Boyanov, M. I.; Thomas, S. H.; Wu, Q.; Kemner, K. M.; L€offler, F. E. U(VI) reduction to mononuclear U(IV) by desulfitobacterium species. Environ. Sci. Technol. 2010, 44 (12), 4705–4709. (13) Bernier-Latmani, R.; Veeramani, H.; Vecchia, E. D.; Junier, P.; Lezama-Pacheco, J. S.; Suvorova, E.; Sharp, J. O.; Wigginton, N. S.; Bargar, J. R. Non-uraninite products of microbial U(VI) reduction. Environ. Sci. Technol. 2010, 44, 5104–5111. (14) Suzuki, Y.; Banfield, J. F. Geomicrobiology of uranium. Rev. Mineral. Geochem. 1999, 38 (1), 393–432. (15) Suzuki, Y.; Kelly, S. D.; Kemner, K. M.; Banfield, J. F. Nanometresize products of uranium bioreduction. Nature 2002, 419 (6903), 134. (16) Burgos, W. D.; McDonough, J. T.; Senko, J. M.; Zhang, G.; Dohnalkova, A. C.; Kelly, S. D.; Gorby, Y.; Kemner, K. M. Characterization of uraninite nanoparticles produced by Shewanella oneidensis MR-1. Geochim. Cosmochim. Acta 2008, 72 (20), 4901–4915. (17) Moon, H. S.; Komlos, J.; Jaffe, P. R. Biogenic U(IV) oxidation by dissolved oxygen and nitrate in sediment after prolonged U(VI)/Fe(III)/SO42- reduction. J. Contam. Hydrol. 2009, 105 (12), 18–27. (18) Ulrich, K. U.; Ilton, E. S.; Veeramani, H.; Sharp, J. O.; BernierLatmani, R.; Schofield, E. J.; Bargar, J. R.; Giammar, D. E. Comparative dissolution kinetics of biogenic and chemogenic uraninite under oxidizing conditions in the presence of carbonate. Geochim. Cosmochim. Acta 2009, 73 (20), 6065–6083. (19) Ulrich, K. U.; Singh, A.; Schofield, E. J.; Bargar, J. R.; Veeramani, H.; Sharp, J. O.; Bernier-Latmani, R.; Giammar, D. E. Dissolution of biogenic and synthetic UO2 under varied reducing conditions. Environ. Sci. Technol. 2008, 42 (15), 5600–5606. (20) Komlos, J.; Peacock, A.; Kukkadapu, R. K.; Jaffe, P. R. Longterm dynamics of uranium reduction/reoxidation under low sulfate conditions. Geochim. Cosmochim. Acta 2008, 72 (15), 3603–3615. (21) Senko, J. M.; Kelly, S. D.; Dohnalkova, A. C.; McDonough, J. T.; Kemner, K. M.; Burgos, W. D. The effect of U(VI) bioreduction kinetics on subsequent reoxidation of biogenic U(IV). Geochim. Cosmochim. Acta 2007, 71 (19), 4644–4654. (22) Shoesmith, D. W. Fuel corrosion processes under waste disposal conditions. J. Nucl. Mater. 2000, 282 (1), 1–31. (23) Brusseau, M. L.; Rao, P. S. C. Modeling solute transport in structured soils: A review. Geoderma 1990, 46, 169–192. (24) Wu, W. M.; Carley, J.; Luo, J.; Ginder-Vogel, M. A.; Cardenas, E.; Leigh, M. B.; Hwang, C.; Kelly, S. D.; Ruan, C.; Wu, L.; Van Nostrand, J.; Gentry, T.; Lowe, K.; Mehlhorn, T.; Carroll, S.; Luo, W.; Fields, M. W.; Gu, B.; Watson, D.; Kemner, K. M.; Marsh, T.; Tiedje, J.; Zhou, J.; Fendorf, S.; Kitanidis, P. K.; Jardine, P. M.; Criddle, C. S. In situ bioreduction of uranium (VI) to submicromolar levels and reoxidation by dissolved oxygen. Environ. Sci. Technol. 2007, 41 (16), 5716–5723. (25) Ginder-Vogel, M.; Criddle, C. S.; Fendorf, S. Thermodynamic constraints on the oxidation of biogenic UO2 by Fe(III) (hydr)oxides. Environ. Sci. Technol. 2006, 40 (11), 3544–3550. (26) Gu, B.; Yan, H.; Zhou, P.; Watson, D. B.; Park, M.; Istok, J. Natural humics impact uranium bioreduction and oxidation. Environ. Sci. Technol. 2005, 39 (14), 5268–5275. (27) Sani, R. K.; Peyton, B. M.; Dohnalkova, A.; Amonette, J. E. Reoxidation of reduced uranium with iron(III) (hydr)oxides under sulfate-reducing conditions. Environ. Sci. Technol. 2005, 39 (7), 2059–2066. (28) Yabusaki, S. B.; Fang, Y.; Long, P. E.; Resch, C. T.; Peacock, A. D.; Komlos, J.; Jaffe, P. R.; Morrison, S. J.; Dayvault, R. D.; White, D. C.; Anderson, R. T. Uranium removal from groundwater via in situ
ARTICLE
biostimulation: Field-scale modeling of transport and biological processes. J. Contam. Hydrol. 2007, 93 (14), 216–235. (29) Veeramani, H.; Schofield, E. J.; Sharp, J. O.; Suvorova, E. I.; Ulrich, K. U.; Mehta, A.; Giammar, D. E.; Bargar, J. R.; Bernier-Latmani, R. Effect of Mn(II) on the structure and reactivity of biogenic uraninite. Environ. Sci. Technol. 2009, 43 (17), 6541–6547. (30) Schofield, E. J.; Veeramani, H.; Sharp, J. O.; Suvorova, E.; Bernier-Latmani, R.; Mehta, A.; Stahlman, J.; Webb, S. M.; Clark, D. L.; Conradson, S. D.; Ilton, E. S.; Bargar, J. R. Structure of biogenic uraninite produced by Shewanella oneidensis strain MR-1. Environ. Sci. Technol. 2008, 42 (21), 7898–7904. (31) Campbell, K. M.; Root, R.; O’Day, P. A.; Hering, J. G. A gel probe equilibrium sampler for measureing arsenic porewater profiles and sorption gradients in sediments: I. Laboratory development. Environ. Sci. Technol. 2007, 42, 504–510. (32) Kneebone, P. E.; O’Day, P. A.; Jones, N.; Hering, J. G. Deposition and fate of arsenic in iron- and arsenic-enriched reservoir sediments. Environ. Sci. Technol. 2002, 36 (3), 381–386. (33) Larson, A. C.; Von Dreele, R. B. General Structure Analysis System (GSAS), LANL Report No. LAUR 96-748, 2004, . (34) Toby, B. H. EXPGUI, a graphical user interface for GSAS. J. Appl. Crystallogr. 2001, 34, 210–213. (35) Webb, S. M. SIXPack: A graphical user interface for XAS analysis using IFEFFIT. Phys. Scri. 2005, T115, 011–1014. (36) Rehr, J. J. A., R. C.; Zabinsky, S. I. High-order multiplescattering calculations of X-ray absorption fine structure. Phys. Rev. Lett. 1992, 69 (23), 3397–3400. (37) Stadelmann, P., http://cimewww.epfl.ch/people/stadelmann/ jemswebsite/jems.html (accessed May 2010). (38) Santos, B. G.; No€el, J. J.; Shoesmith, D. W. The influence of silicate on the development of acidity in corrosion product deposits on SIMFUEL (UO2). Corros. Sci. 2006, 48 (11), 3852–3868.
8754
dx.doi.org/10.1021/es200482f |Environ. Sci. Technol. 2011, 45, 8748–8754
ARTICLE pubs.acs.org/est
Hormone Discharges from a Midwest Tile-Drained Agroecosystem Receiving Animal Wastes Heather E. Gall,† Stephen A. Sassman,‡ Linda S. Lee,*,‡ and Chad T. Jafvert† †
School of Civil Engineering and ‡Department of Agronomy, Purdue University, West Lafayette, Indiana 47907, United States
bS Supporting Information ABSTRACT: Manure is increasingly being viewed as a threat to aquatic ecosystems due to the introduction of natural and synthetic hormones from land application to agricultural fields. In the Midwestern United States, where most agricultural fields are tile-drained, there is little known about hormone release from fields receiving animal wastes. To this end, seven sampling stations (four in subsurface tile drains and three in the receiving ditch network) were installed at a Midwest farm where various types of animal wastes (beef, dairy, and poultry lagoon effluent, dairy solids, and subsurface injection of swine manure) are applied to agricultural fields. Water flow was continuously monitored and samples were collected for hormone analysis during storm events and baseline flow for a 15 month study period. The compounds analyzed included the natural hormones 17α- and 17β-estradiol, estrone, estriol, testosterone, and androstenedione and the synthetic androgens 17α- and 17β-trenbolone and trendione. Hormones were detected in at least 64% of the samples collected at each station, with estrone being detected the most frequently and estriol the least. Testosterone and androstendione were detected more frequently than synthetic androgens, which were detected in fewer than 15% of samples. Hormone concentrations in subsurface tile drains increased during effluent irrigation and storm events. Hormones also appeared to persist over the winter, with increased concentrations coinciding with early thaws and snowmelt from fields amended with manure solids. The highest concentration of synthetic androgens (168 ng/L) observed coincided with a snowmelt. The highest concentrations of hormones in the ditch waters (87 ng/L for total estrogens and 52 ng/L for natural androgens) were observed in June, which coincides with the early life stage development period of many aquatic species in the Midwest.
’ INTRODUCTION Estrogenic and androgenic compounds have been detected in surface waters around the world.1 3 Humans and livestock are important sources of these compounds with major inputs to the environment including discharge from wastewater treatment plants, combined sewer overflows, and the land application of biosolids and animal wastes. The increasing size of concentrated (or confined) animal feeding operations (CAFOs) has led to manure generation at a higher mass per unit area.4 Additionally, the amount of estrogens introduced into the environment from land application of animal wastes has been estimated to be >200 times the amount introduced from biosolids applications,5,6 thereby increasing the potential of CAFOs to be a significant source of hormones to the environment. Hormones associated with livestock are introduced into the environment when animal wastes are applied to agricultural fields as a nutrient source. The type and amount of hormones in these wastes vary by animal, reproductive stage, and treatment with growth-promoting compounds. Cattle excrete the majority of hormones in feces, whereas poultry and swine excrete the majority of estrogens in urine.7 17α-Estradiol (17α-E2) constitutes approximately 60% of estrogens in cattle feces,1 whereas 17β-E2 and the E2 metabolite estrone (E1) comprise the majority of estrogens in poultry and swine excretions, respectively.7 r 2011 American Chemical Society
Although few studies have focused on the excretion of natural androgens by livestock, Lange et al.5 estimated that livestock in the United States excrete 4.4 tonnes of androgens each year, with laying hens and cattle (calves and bulls) as the largest sources. Cattle receiving growth-promoting ear implants containing 17β-trenbolone acetate (TBA) excrete the synthetic androgens 17β-trenbolone (17β-TB), trendione (TND), and 17α-trenbolone (17α-TB), with the majority being excreted within the first 5 weeks after implant;8 thus, their input into the environment is not as consistent as that of natural hormones. Various laboratory and field studies have been conducted to assess the potential impact of CAFOs on nearby waterways. In laboratory studies, the parent hormones (e.g., E2, TB, and testosterone) have exhibited relatively short half-lives in aerobic soils and manure-amended soils on the order of a few days,9 leading to hypotheses that hormone discharge to surface water and groundwaters should be minimal. However, Kjær et al.10 observed hormone concentrations in tile drainage up to 11 months after subsurface injection of liquid swine manure during a 1 year Received: April 5, 2011 Accepted: August 30, 2011 Revised: August 29, 2011 Published: August 30, 2011 8755
dx.doi.org/10.1021/es2011435 | Environ. Sci. Technol. 2011, 45, 8755–8764
Environmental Science & Technology study period. They suggested that soil temperature fluctuations and preferential flow through soil macropores may play important roles in explaining the different observations between field and laboratory studies. Hormones are known to cause endocrine disruption in sensitive aquatic organisms at low nanograms per liter levels, although the lowest observable effect level (LOEL) varies by species and compound.11 Definitive links between CAFO-originated hormone concentrations and altered aquatic species are complicated by the presence of other contaminants and environmental conditions; however, hormone concentrations in surface waters affected by CAFOs have been shown to be above some of the reported LOELs.1,9 Also noteworthy is that whereas the 17β isomers of E2 and TB induce effects at much lower concentrations (<100 times lower) than the 17α isomers or other metabolites in mammalian toxicological studies, similar trends are not observed for aquatic species.11 15 For example, 17α-E2 was found to be only 8 30 times less potent than 17β-E2 to medaka fish and fathead minnows, 17α- and 17β-TB were found to have similar reproductive effects on fathead minnow,14,16 and E1 was found to skew sex ratios toward females and induce vitellogenin production in zebrafish at concentrations similar to or lower than those of 17β-E2.11 Despite the recognized negative effects of hormones on fish and other aquatic organisms, the fate and transport of hormones in agroecosystems remain poorly understood, with many studies to date being limited to experimental plots under simulated rainfall.4,17 This study focused on hormone release from subsurface tile-drained agricultural fields at a Midwest U.S. farm where various types of animal wastes including dry manure solids, liquid manure, and animal lagoon effluent are applied to fields. Hormone concentrations were monitored in tile drains and the ditch network receiving tile drainage during storm events, base flow, effluent irrigation, and thawing/snowmelt events. The hormones monitored included 17α- and 17β-E2, E1, estriol (E3), testosterone (TST), androstenedione (AND), 17α- and 17β-TB, and TND. Hormone structures and selected chemical properties are listed in Table SI-1 of the Supporting Information (SI).
’ MATERIALS AND METHODS Study Site. This study was conducted in north central Indiana at Purdue University’s Animal Science Research and Education Center (ASREC), which is a working farm, an EPA-designated CAFO, and includes approximately 600 ha of tile-drained cropland. Soils at the site are predominantly silty clay loams and silt loams with loess and glacial till as the soil parent materials. Due to the presence of these poorly drained soils, perforated subsurface tile drains ranging in diameter from 10 to 61 cm were installed in the early 1990s approximately 1 m below the soil surface at 8 40 m spacings. Site maps and additional details are provided in the SI. Animal production at ASREC includes beef, dairy, poultry, sheep, swine, and Ossabaw swine units (see SI for details). Beef cattle each received a Revalor-S hormone implant containing 28 mg of 17β-E2 and 140 mg of TBA. Animal wastes are collected and stored on-site. Primary storage includes belowground pits that are washed into above-ground lagoon systems,18 above-ground storage units of liquid slurry from liquid/ solid separators, and above-ground stacking of bedding/manure wastes. Wastes are land-applied via solids broadcasting (dewatered bedding/manure wastes), pivot irrigation (effluent from on-site
ARTICLE
lagoons), or subsurface injection (liquid slurry from above-ground storage units). Further details are provided in the SI. Monitoring stations were installed to measure discharge and collect water samples at four tile drains (stations D1 D4) and three locations (stations S1 S3) in the ditch network at ASREC (SI, Figures SI-1 and SI-2). Each station consisted of a Campbell Scientific CR1000 datalogger, a Campbell Scientific 107 water temperature probe, a flow and/or water level sensor, a Teledyne ISCO automated sampler, and a Campbell Scientific radio and antenna enabling two-way wireless communication. Tile drains monitored by D1 and D2 are 30.5 cm in diameter, and those monitored by D3 and D4 are 61 cm in diameter. The flow rate in each tile was measured with a Marsh-McBirney Flo-Tote 3 and Flo-Station. Water levels in the ditches were monitored with a Campbell Scientific shaft encoder pulley system, and rating curves were developed to calculate flow rates. Stations were located to capture each major animal waste application practice according to ASREC’s manure management plan: (i) beef and dairy effluent (D1 and S1); (ii) beef and dairy effluent and annual applications of dairy solids (D3 and S2); and (iii) poultry and swine effluent (D4 and S3). For fields drained by D2, beef and dairy manure solids had been routinely applied up to 2007, but not during our study period. Although the major waste sources did align with our station plan, additional sources were occasionally applied to some fields. Details regarding all applications are provided in the SI (Tables SI-2 SI-8). After the start of our study, piping between lagoon systems (referred to as an interconnect system) was installed as an additional safety measure to better control lagoon heights. In addition, two valves in this interconnect system were identified to leak, thereby causing unintentional transport from north to south lagoon systems. This led to the movement of some beef wastes to the dairy, swine, and poultry lagoon systems. Within the ditch network, S1 and S2 monitored Marshall Ditch and S3 monitored Box Ditch (SI, Figure SI-1). S1’s drainage area encompassed the tile-drained area monitored by D1. S2 was located downstream of S1 and received drainage from areas monitored by stations D1, D2, D3, and S1. S3’s drainage area encompassed the area monitored by station D4. Station drainage area details and total amounts of animal wastes applied are provided in Table SI-2 of the SI. Data were collected at some stations for over 2 years; however, the work presented here is focused on the data collected from January 2009 through March 2010, prior to the commencement of spring animal waste applications. In addition to the monitoring stations at ASREC, we monitored subsurface tile drainage from a control plot at Purdue’s Water Quality Field Station (WQFS), which immediately neighbors the east boundary of ASREC. This control plot (E30) has received only commercial N fertilizer for over 10 years and has never had any animal wastes intentionally applied to the field. We monitored plot E30 as a control plot from August 14, 2009, to May 16, 2010, during our ASREC study. Details are provided in the SI. Sampling Methodology and Analysis. Sampling Methods. Samples were collected in 1 L polyethylene bottles using Teledyne ISCO samplers controlled by dataloggers programmed to trigger samples at time-paced intervals during base flow and at flow-paced intervals over hydrographs. Each time a sample was triggered, a 1 L sample was collected. Samples on the rising limb of hydrographs were collected at preprogrammed flow rate thresholds to appropriately capture the rise. D1 and S1 dataloggers were programmed such that real-time flow data were used to 8756
dx.doi.org/10.1021/es2011435 |Environ. Sci. Technol. 2011, 45, 8755–8764
n = 190
8757
(0.17, 0.58)
TST
natural androgens
total
(0.22, 0.74)
17α-TB
(1.9, 6.4)
TND
synthetic androgens 17β-TB (0.47, 1.6)
total
(0.21, 0.64)
n = 183
n = 589
2.4 31%, 10%
11%, 5.4%
12.7
13.4 13%, 1.6%
36.8
1.1%, 1.1% n = 183
13%, 6.8%
6.8%, 5.6% n = 589
1.4
n = 142
n = 423
8.9
15%, 0%
3.5%, 0.7%
NA
n = 183
n = 589
28.5
13.1 0.5%, 0.5%
34.0
4.6%, 1.4%
27.1 27%, 23%
n = 187
n = 581
68.7 64%, 36%
1.1%, 0%
NA
1.4%, 0.7%
3.5
n = 589
E3
26.7 13%, 8.4%
17%, 7%
(0.09, 0.29)
17α-E2
25%, 10% n = 190
48%, 19% n = 589 51.8
1.3
n = 190
18.1
n = 589
4.2 34%, 11%
9.5
33%, 19%
(0.05, 0.16)
E1
(0.06, 0.21)
17β-E2
estrogens
% n > LOD, % n > LOQ n
% n > LOD, % n > LOQ n
(LOD, LOQ) (ng/L)
D2 max (ng/L)
max (ng/L)
hormone
D1 D4
n = 508
16%, 7%
3.7
7.7%, 3.0%
30.1
1.4%, 1.0% n = 508
22.7
n = 351
4.0%, 0.3%
12.2
n = 508
4.7%, 2.0%
13.6
53.5 69%, 40%
n = 522
1.5%, 0%
NA
n = 531
24%, 14%
n = 518
15%, 7.9%
16.1
10%, 6.7%
32.7
5.7%, 4.9% n = 518
18.7
n = 332
5.7%, 1.5%
12.1
n = 518
3.7%, 1.4%
18.6
42.8 88%, 64%
n = 537
2.6%, 1.1%
19.6
n = 552
18%, 11%
13.3
78%, 41% n = 552
46%, 20% n = 531 31.1
n = 552 33.5
50%, 29%
16.3
% n > LOD, % n > LOQ n
max (ng/L)
n = 531 25.6
39%, 20%
16.4
% n > LOD, % n > LOQ n
max (ng/L)
D3
n = 576
26%, 14%
50.5
7.6%, 3.8%
54.3
3.3%, 2.6% n = 576
9.7
n = 402
3.5%, 0.2%
6.5
n = 576
4.2%, 2.6%
53.6
23.6 69%, 43%
n = 590
4.1%, 1.4%
7.8
n = 595
14%, 6.7%
14.1
62%, 36% n = 595
n = 595 23.1
39%, 23%
9.3
% n > LOD, % n > LOQ n
max (ng/L)
S1 S3
n = 545
26%, 15%
n = 345
21%, 10%
11.7
10%, 7.5%
9.0%, 4.4%
15.4
1.4%,1.4% n = 345 35.6
11.7
n = 236
13%, 8.5%
35.3
n = 345
2.6%, 0.6%
3.3
12.5 95%, 66%
n = 368
2.4%, 0%
NA
n = 372
23%, 9.7%
6.1
91%, 55% n = 372
n = 372 9.0
58%, 30%
12.1
% n > LOD, % n > LOQ n
max (ng/L)
3.7%, 2.8% n = 506 168
19.1
n = 377
3.7%, 0.3%
6.5
n = 506
5.5%, 3.9%
162
87.0 91%, 60%
n = 674
2.5%, 1.6%
12.4
n = 683
29%, 11%
26.9
89%, 52% n = 683
n = 683 40.0
43%, 22%
20.9
% n > LOD, % n > LOQ n
max (ng/L)
S2
Table 1. Data Summary Including Maximum Observed Concentrations, Percent of Total Samples Collected That Were Greater than the LOD and LOQ, and Total Number of Samples Collected (n)a
Environmental Science & Technology ARTICLE
dx.doi.org/10.1021/es2011435 |Environ. Sci. Technol. 2011, 45, 8755–8764
52%, 28% 50%, 22% 35%, 15% 39%, 16% 30%, 14% 31%, 10% = 183 22%, 9.5%
The summary is for samples collected from January 2009 through March 2010 (prior to the commencement of spring 2010 effluent irrigation). Note that due to equipment errors, data at D4 and S3 are reported only through December 2009. NA, maximum concentration not applicable; all concentrations were below the LOQ.
a
n = 345 11.9 16.2
n = 545
22.8
51.7
n = 518
8.2 2.4
n = 183
14.2 total
n = 589
n = 508
n = 576
5.0
43%, 19% 37%, 11%
9.4 16.6 12.5
21%, 9.1%
8.2
17%, 3.6%
ARTICLE
19%, 9%
2.1
21%, 3.8%
1.6 AND
14%, 4.8%
n n (ng/L)
(0.12, 0.41)
n
% n > LOD, % n > LOQ
n
n
% n > LOD, % n > LOQ
n
% n > LOD, % n > LOQ
n
% n > LOD, % n > LOQ % n > LOD, % n > LOQ (LOD, LOQ)
S1
% n > LOD, % n > LOQ
% n > LOD, % n > LOQ
S3 S2
max (ng/L) max (ng/L)
D4
max (ng/L)
D3
max (ng/L)
D2
max (ng/L)
D1
max (ng/L) hormone
Table 1. Continued
max (ng/L)
Environmental Science & Technology
predict the hydrograph recession and the collection times at flow-paced intervals for the remaining samples to ensure that not all bottles were filled before the end of the recession.19 A variation of this methodology was employed at the remaining stations with preprogrammed flow-paced intervals determined by the value of the peak flow rate such that sufficient samples would be collected during the recession for both small and large hydrographs. Sample Preparation. Sample preparation is detailed in the SI. Briefly, water samples (1 L) were refrigerated immediately upon receipt in the laboratory for typically less than 36 h but no longer than 72 h prior to solid-phase extraction. Samples were weighed, filtered, amended with deuterated standards (6.25 ng of 17β-E216,16,17-d3 and 5 ng of TST-16,16,17-d3 dissolved in 0.5 mL of methanol), and preconcentrated by solid-phase extraction immediately or stored at 4 °C in the dark for typically less than 36 h but no longer than 72 h prior to further processing. Loaded cartridges were stored at 20 °C for up to 4 months prior to washing and eluting analytes with methanol. The eluant was evaporated to dryness, and residues were reconstituted in methanol (0.5 mL). Samples were analyzed using high-performance reverse-phase liquid chromatography tandem electrospray ionization mass spectrometry (HPLC-ESI-MS/MS). HPLC/MS/MS Analysis. LC/MS/MS analysis was performed using a Shimadzu HPLC system coupled to a Sciex API-3000 triple-quadrupole operated in multiple reaction monitoring mode. Column and mobile phase gradient details for estrogens and androgens are summarized in Tables SI-10 and SI-11 of the SI, respectively. Retention times, precursor and product ions monitored, and the method limits of detection (LOD) and quantitation (LOQ) for aqueous samples are summarized in Table SI-12 of the SI. Method LOD and LOQ values for each hormone are also provided in Table 1 for easy reference. Other analyses performed and detailed in the SI include sample matrix effects on HPLC/ESIMS/MS response to hormones, extraction recovery of hormones, hormone sorption to ISCO polyethylene collection bottles, and hormone stability in field samples.
’ RESULTS AND DISCUSSION Hormone Recovery, Matrix Effects, and Stability. Hormone concentrations were corrected for recoveries and matrix effects using deuterated internal standards added prior to extraction and assuming similar extraction efficiencies based on similar hydrophobicities (SI, Table SI-1) and that signal suppression for estrogens and androgens could be reasonably approximated by 17β-E2-16,16,17-d3 and TST-16,16,17-d3, respectively. Internal standard recoveries with matrix corrections were in the range expected for large field studies with >78% of the recoveries being between 50 and 150% with an average recovery in this ranges of 91.1 ( 18.3% for 17β-E2-16,16,17-d3 and 73 ( 19.6% for TST16,16,17-d3 (SI, Table SI-14). Any samples with an internal standard recovery outside the 50 150% range were included in the data analysis, but not modified by internal standard recoveries. Hormone-specific recoveries (SI, Table SI-13) and matrix effects were assessed as detailed in the SI (Figures SI-3 and SI-4). Potential errors in reported concentrations using the internal standards for recovery and matrix effects are generally within 20% (SI, Table SI-13). Hormones are subject to sorption and microbial degradation from the time of ISCO collection to the loading of the aqueous samples onto the solid-phase cartridges. Sorption to polyethylene 8758
dx.doi.org/10.1021/es2011435 |Environ. Sci. Technol. 2011, 45, 8755–8764
Environmental Science & Technology
ARTICLE
Figure 1. Hyetograph, hydrographs, and total estrogen chemographs (E1 + E2 + E3) for the study period at D1, D3, and S2. Animal waste applications are shown across the top of each panel. For graphical clarity, several high total estrogen concentrations are shown using a broken-axis notation.
bottles was assessed at 4 °C over a 72 h period (detailed in the SI). Concentrations were found to vary <5% at 0, 24, 48, and 72 h with no statistical differences at the 95% confidence level. The potential loss of hormones during sample processing was evaluated by monitoring hormone-amended S2 ditch water incubated under representative conditions, which included unfiltered and filtered stream water at 4 °C and ∼23 °C. Data from stream water amended with hormones in the laboratory suggest that prior to filtering, there is a substantial degradation potential for some of the hormones within the first 24 h (SI, Figure SI-5 and SI-6), especially the natural androgens, of which up to 50% was gone within the first 24 h at 23 °C. If degradation rates estimated from the laboratoryfortified ditch water were directly applicable to field samples, then at near-steady-state conditions in the ditch network, concentrations for samples collected at later times would be expected to be higher than those from earlier collection times prior to sample pickup. However, this was not the case even for samples collected over a ∼72 h period in the summer months (air temperatures of 15 30 °C). We suspect that the aerobic microbial degradation rates measured in the laboratory experiments were greatly elevated relative to the field due to aeration of the ditch water during homogenization immediately prior to hormone addition and potentially the concomitant addition of methanol (hormone carrier), which may serve as a readily available microbial food source.20,21 Even with the
expectation that degradation in our actual site samples is considerably slower than observed in well-mixed laboratoryamended ditch water, the hormone concentrations reported from subsurface tile drain and ditch network samples are likely still underestimated. Hormone Discharge Dynamics. After land application, biogeochemical and hydrologic processes control the subsequent transport of hormones. Seasonal differences in rainfall intensity and amount, temperature, and waste management strategies confound the ability to discern between environmental and anthropogenic influences on hormone dynamics. In addition, deviations from the anticipated management plan at our study site and the intentional and unintentional routing of wastes between animal-specific lagoon systems prevent an explicit delineation of hormone release between waste types in this study. Furthermore, data interpretation of specific hormones may be biased as a result of degradation during sample collection/processing. To minimize under-representation of the hormone concentrations and associated biases, hormone levels were primarily assessed in terms of total estrogens, synthetic androgens, and natural androgens. Given the greater stability of E1 (metabolite of E2 isomers) and TND (metabolite of TB isomers), errors in total estrogens (17α-E2, 17β-E2, E1, and E3) and total synthetic androgens (17α-TB, 17β-TB, and TND), respectively, will be much smaller than the error associated with 8759
dx.doi.org/10.1021/es2011435 |Environ. Sci. Technol. 2011, 45, 8755–8764
Environmental Science & Technology
ARTICLE
Figure 2. Hyetograph, hydrographs, and total estrogen and total natural androgen chemographs at D3 and D4. Effluent irrigation events (65.5 m3/ha) are shown as dashed vertical lines (gray, lagoon effluent; black, retention pond effluent).
each individual hormone. Concentrations for the natural androgens (TST and AND) are likely the most underestimated because both TST and its metabolite AND degraded the most rapidly in the laboratory assessment. Using this approach, data analysis focused on trends observed during storm events, release during snowmelt events, and the influence of effluent irrigation on hormone concentrations in subsurface tile drainage. Although concentrations are likely underestimated in this study, the general trends are expected to remain the same. Given the large data set, various monitoring stations, and range of waste application types, the resulting summarization of general hormone discharge dynamics is likely to be representative of many tiledrained fields receiving animal waste applications. Hormone Concentration Summary. The most to least frequently detected estrogens at each sampling location were E1, 17β-E2, 17α-E2, and E3, with E3 detected in <5% of samples (Table 1). Additionally, natural androgens (TST and AND) were detected more frequently than synthetic androgens (TB and TND), which were detected in <15% of the samples. Overall, hormones were detected in at least 64% of samples collected at each station that received animal waste applications during the study period. Ditch water total estrogen (E1 + E2 + E3) concentrations were highest in the spring and summer, with the maximum (87 ng/L) observed on June 1, 2009 at S2 during a 6 cm rainfall event that occurred 3 days after fields were irrigated with dairy effluent. On average, androgen concentrations were highest during the fall and winter (SI, Figures SI-7 and SI-8). The highest concentration of synthetic androgens (168 ng/L) was
observed in relation to a snowmelt. However, the highest total natural androgen concentration (52 ng/L) was observed on June 25, 2009, during dairy effluent irrigation. The May June time frame coincides with early life stage development period, a sensitive time for gonadal development and sexual differentiation, for many aquatic species.9 Hormone concentrations measured in the samples collected from August 14, 2009 to May 16, 2010 at the WQFS control plots are summarized in Table SI-16 of the SI. Almost all estrogen and androgen concentrations were below the LOQ with the exceptions of 17β-E2 and E1. 17β-E2 was observed in two samples with a maximum concentration of 3.13 ng/L in January 2010. E1 was observed in five samples with a maximum value of 0.38 ng/L. Hormone Discharge during Storm Hydrographs. The influence of precipitation events on the tile drain and ditch network hydrographs is dependent on storm intensity and duration, evapotranspiration, and antecedent soil moisture conditions. During the summer months, increased evapotranspiration due to rapid crop growth and warm temperatures led to lower antecedent moisture conditions and lower tile drain and ditch flow rates relative to the winter and spring. Flow and total estrogen data collected during the study period are shown in Figure 1 for D1, D3, and S2 along with timing of animal waste applications and rainfall. Additional figures for the remaining monitoring stations and the androgens (natural and synthetic) data are provided in the SI (Figures SI-7 SI-12). Note that values in all of the hyetographs represent total daily rainfall (not intensity), 8760
dx.doi.org/10.1021/es2011435 |Environ. Sci. Technol. 2011, 45, 8755–8764
Environmental Science & Technology
ARTICLE
Figure 3. Hyetograph, hydrographs, and total estrogens (E1 + E2 + E3) chemographs for S1, D3, and S2 during the winter and early spring prior to the commencement of spring effluent irrigation. The timing of dairy solids applications (32.5 m3/ha) at fields monitored by D3 and S2 are shown as dotted vertical lines. S1 received dairy and beef effluent irrigation in late November 2009 (see Table SI-6 of the Supporting Information) and is shown for comparison. Due to a sampler error, samples were missed on the recession limb of the storm hydrograph on December 23 27, 2009. Snow records indicate that snowmelt coincided with rainfall on December 22 24, 2009 (∼5 cm), and February 19 20, 2010 (∼8 cm). Snowmelt (∼20 cm) occurred without rainfall on January 11 16, 2010.
and hormone concentrations below the LOD are plotted as 0 ng/L in all of the chemographs. During storm hydrographs, hormone concentrations generally increased as exemplified at D1, D3, and S2 in Figure 1 for total estrogens for the study period (January 2009 March 2010). Peak concentrations were often highest during the first storm event following an animal waste application with lower concentrations observed in subsequent events before additional applications (e.g., April 2009 for D1 in Figure 1). Hormone chemographs also generally paralleled hydrographs with hormone concentrations increasing along the rising hydrograph limb, peaking near the hydrograph peak, and decreasing along the recession limb in tile drains and ditches (Figure 1 and further exemplified for D1 and D4 in Figure SI-10 of the SI). These chemograph hydrograph similarities were the most pronounced during the first storm hydrograph following an application regardless of the magnitude of the peak storm hydrograph flow rate. The steep rising limb of tile drain hydrographs is due primarily to macropore flow, which is known to transport land-applied chemicals to receiving ditches, especially in the first two rain events after application.22 Transport of hormones sorbed to soils or associated with manure solids (see Table SI-1 of the SI) are also highly subject to surface runoff during high-intensity rain events. Although surface runoff was not measured directly, it was inferred to occur when flow in the smaller tile drains (e.g., D1) reached full capacity and
the area-normalized flow rates were substantially higher in the ditches than in the larger tile drains (e.g., D3). Intense rains occurred several times in May and June 2009, leading to fullcapacity tile flow in D1 and surface runoff to the ditch network (e.g., D1 and S2 in Figure 1 and detailed in Figure SI-11 of the SI). During the first two storm events in June after dairy effluent irrigations, total estrogens (Figure 1 and Figure SI-11 of the SI), natural androgens (SI, Figure SI-7), and synthetic androgens (SI, Figure SI-8) in ditch water increased to levels above those observed in the tile drains with concentrations highest at S2 (downstream of S1). Surface runoff is typically high in suspended solids, to which hormones may be sorbed. Hormone concentrations at D2, which drains an area that had not received manure applications since 2007, also increased during these events (SI, Figure SI-12). Rapid Transport following Effluent Irrigation. Effluent irrigation is typically used during late spring and summer while crops are growing and evapotranspiration rates are high; however, to minimize the potential of lagoon overflow during periods of snowmelt and heavy spring rainfall, effluent was also frequently applied in March and November (see Tables SI-3 SI-8 of the SI). In general, hormone concentrations following effluent irrigation increased during rainfall events as exemplified in Figures 1 and 2. Additionally, hormone concentrations occasionally were observed to increase during and shortly after effluent irrigation events that were not associated with rainfall. For example, hormone concentrations 8761
dx.doi.org/10.1021/es2011435 |Environ. Sci. Technol. 2011, 45, 8755–8764
Environmental Science & Technology increased on March 18 and 27 28, 2009 at D3 following dairy effluent irrigation, July 12 13, 2009 at D3 following beef effluent irrigation, and during several poultry effluent irrigation events during July and August 2009 at D4 (July 12 13, 17, and 27; August 6 and 10) (Figure 2). Elevated concentrations, albeit low nanograms per liter, were also observed at D1 during July 2009 irrigations (Figure 1). These trends were more pronounced when effluent irrigations occurred shortly after rainfall, which increased antecedent soil moisture. Higher antecedent soil moisture conditions have been correlated to enhanced macropore flow of chemicals,22,23 although not consistently.24 Effluent irrigation also influences soil moisture conditions, as each irrigation (65.5 m3/ha) was the equivalent (in terms of moisture) to ∼6.6 mm of rainfall. In some cases, hormone concentrations in tile drainage were higher during effluent irrigation than during rainfall events (e.g., March 26 28 at D3; July 12 13 at D3; July 12 13 and 17 at D4). Hormone concentrations in tile drains immediately following effluent irrigation appear to be indicative of preferential flow through an established macropore network. The latter is consistent with subsurface tile drainage studies in which tracers in irrigation water reached the tile drain within 1 h after irrigation regardless of their sorption characteristics.22 Although no direct measurements of preferential flow were made at the study site (e.g., tracer studies), preferential flow is known to occur at similar subsurface tile-drained fields and has been observed within tracer studies at experimental tile-drained plots at the Purdue WQFS immediately adjacent to ASREC.25 These observations of rapid solute transport to subsurface tile drains are consistent with observations at several other field studies.26 29 Notably, Lapen et al.27 observed “application-induced discharge” (as opposed to “precipitation-induced discharge”) of pharmaceuticals and personal care products to tile drains following land application of biosolids with concentrations increasing within minutes after application. They attributed this rapid transport to flow through networks in the soil that directly connect to the tile drains (i.e., preferential flow pathways). Transport through such networks reduces the reactive time with soil particles, potentially reducing sorption and degradation30 and increasing the potential importance of preferential flow to water quality implications with regard to hormones. Hormone Preservation and Discharge during Cold Months. According to best management practices, solid manure applications should occur when soil temperatures drop below 10 °C to minimize the potential for nutrient loss;31 however, colder temperatures also can preserve manure-borne hormones. When temperatures rose in early February 2009 (>13 °C) and caused a snowmelt (∼5 cm of snow was on the ground at this time), total estrogen concentrations increased to 12 ng/L at D3 (Figure 1), for which the last waste application had been dairy solids in September 2008. Estrogen concentrations also increased during a large rain event (total rainfall of 6.5 cm over a period of 4 days) in early March 2009 at both D3 and S2 prior to the commencement of spring effluent irrigation (Figure 1). Total synthetic hormones also increased at D3 and S2 during this event, with S2 reaching a maximum value of ∼170 ng/L (SI, Figure SI-8). Fields drained by D1 and S1 did not receive solid applications but were irrigated multiple times with dairy and beef effluent in fall 2008 (SI, Tables SI-3 and SI-6). Additionally, estrogen concentrations increased at D1 during the February snowmelt and early March rain event, although concentrations were higher at D3 and S2
ARTICLE
(Figure 1). Total synthetic hormone concentrations also increased at D1, D3, and S1 during the February event (SI, Figure SI-8). These observations suggest that hormones are preserved in the field during the winter months and that fall applications of solids lead to greater winter and early spring export of hormones than fall effluent irrigation. Similar trends were observed at D3 and S2 in winter 2009 and early spring 2010 (Figure 3). Fields monitored by D3 and S2 received several applications of dairy solids later in the year in 2009 than in 2008, with one application occurring in early January 2010 when ∼5 cm of snow was on the ground (SI, Tables SI-4 and SI-7). Fields monitored by D1 and S1 received one application of dairy solids in early October 2009 and multiple applications of dairy and beef effluent (SI, Tables SI-3 and SI-6). Total estrogen concentrations increased during each storm event following these applications (Figure 3) through March prior to the commencement of spring effluent irrigation. Hormone concentrations increased to higher values at D3 and S2 than at S1 during the rain event on January 22 28, 2010, likely due to the recent dairy solids applications. The apparent preservation of hormones exemplified three times at the site (early and late 2009 and early 2010) suggests that such winter and early spring dynamics play a significant role in hormone export and can be expected at other subsurface tile-drained sites. Implications and Study Limitations. Subsurface tile drains are well-known to change the pore structure within soil profiles, dramatically altering the natural hydrology and expediting the transport of water and solutes through the soil profile, into the tile drains, and ultimately into nearby surface water bodies.22 However, the role these systems play in hormone discharge following land application of animal wastes is not well-known. Rapid increases in hormone concentrations observed in tile drains following effluent irrigation suggest aqueous and particulate-borne hormones are rapidly transported to subsurface tile drains through an established macropore network consistent with tracer irrigation studies.22 During storm events, the rise and fall of hormone concentrations in tile drainage generally followed hydrograph trends. When smaller tile drains flowed full and areanormalized ditch flow rates increased significantly compared to flow in the larger tile drains, elevated hormone concentrations in the ditch network relative to tile drainage suggested hormone transport via surface runoff. Peak hormone concentrations in the ditches occurred in June shortly after effluent irrigation, coinciding with a sensitive early life stage development period for many aquatic species. Cold temperatures during the winter months appeared to preserve hormones from late fall animal waste applications and resulted in increased hormone export during storm events via tile drainage to the ditch network in the early spring. This increase in hormone concentrations during storm events continued for as long as 4 months after fall animal waste applications. Winter rain events and snowmelt increased exported hormone concentrations three times during the 15 month monitoring period (early and late 2009 and early 2010), suggesting that hormone export through such winter and early spring dynamics may be expected at similarly managed subsurface tile-drained sites. The ASREC site presented a unique opportunity to evaluate the discharge of hormones following various animal waste applications occurring at similar field study sites, with each animal waste type applied multiple times during the study period. Although the study did add to our understanding of the potential contribution of hormones from animal-derived effluent and solid 8762
dx.doi.org/10.1021/es2011435 |Environ. Sci. Technol. 2011, 45, 8755–8764
Environmental Science & Technology wastes applied to subsurface tile-drained fields, the management complexity of the site, including multiple types of wastes applied to a single drainage area, limited explicit interpretation of the data collected. Sufficient characterization of the waste being applied was also limited due to the challenge of obtaining waste samples in a regular and timely manner given the application frequency and more pressing responsibilities of farm personnel. In addition, identifying some level of sample preservation that did not interfere with hormone analysis would have helped to minimize underestimation of the discharged hormone levels. Finally, real-time monitoring of soil moisture and several key surface runoff collection points would have improved the utilization of the data set toward recommending improved animal waste management strategies.
’ ASSOCIATED CONTENT
bS
Supporting Information. Study site details and figures, additional information regarding sample analysis, and additional figures depicting hormone chemograph behavior. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (765) 494-8612; fax (765) 496-2926; e-mail: lslee@ purdue.edu.
’ ACKNOWLEDGMENT Funding for this research was provided by the U.S. Environmental Protection Agency Science to Achieve Results (STAR) program (RD833417). We thank Steve Smith and Jeff Fields (Purdue’s ASREC staff) for their collaboration and willingness to share manure management plans and other farm data; Dr. Byron Jenkinson, Craig Stinson, Glenn Hardebeck, Joseph Rivera, and Christopher Campbell for their help in the field; Tirzah Brown, Zach Meyers, and Michael Mashtare for their help in the laboratory; and Dr. Dev Niyogi and Kenneth Scherringa from the Indiana State Climate Office at Purdue University for providing us with snowfall records. Additionally, we thank four anonymous reviewers for their valuable suggestions for improving this paper. ’ REFERENCES (1) Kolok, A. S.; Sellin, M. K. The Environmental Impact of GrowthPromoting Compounds Employed by the United States Beef Cattle Industry: History, Current Knowledge, And Future Directions; Springer: New York, 2008. (2) Matthiessen, P.; Arnold, D.; Johnson, A. C.; Pepper, T. J.; Pottinger, T. G.; Pulman, K. G. T. Contamination of headwater streams in the United Kingdom by oestrogenic hormones from livestock farms. Sci. Total Environ. 2006, 367, 616–630. (3) Duong, C. N.; Ra, J. S.; Cho, J.; Kim, S. D.; Choi, H. K.; Park, J.; Kim, K. W.; Inam, E.; Kim, S. D. Estrogenic chemicals and estrogenicity in river waters of South Korea and seven Asian countries. Chemosphere 2010, 78, 286–293. (4) Carmosini, N.; Lee, L. S. Sorption and degradation of selected pharmaceuticals in soil and manure. In Fate of Pharmaceuticals in the Environment and in Water Treatment Systems; Aga, D. S., Ed.; CRC Press: Boca Raton, FL, 2008. (5) Lange, I. G.; Daxenberger, A.; Schiffer, B.; Witters, H.; Ibarreta, D.; Meyer, H. D. Sex hormones originating from different livestock
ARTICLE
production systems: fate and potential endocrine disrupting activity in the environment. Anal. Chim. Acta 2002, 473, 27–37. (6) U.S. EPA. Biosolids Generation, Use and Disposal in the United States; EPA530-R99-009; Office of Solid Waste: Washington, DC, 1999. (7) Hanselman, T. A.; Graetz, D. A.; Wilkie, A. C. Manure-borne estrogens as potential environmental contaminants: a review. Environ. Sci. Technol. 2003, 37, 5471–5478. (8) Khan, B. The Environmental Fate of Anabolic Steroid Trenbolone Acetate; Dissertation, Purdue University, West Lafayette, IN, 2009. (9) Lee, L. S.; Carmosini, N.; Sassman, S. A.; Dion, H. M.; Sepulveda, M. S. Agricultural contributions of antimicrobials and hormones on soil and water quality. Adv. Agron. 2007, 93, 1–67. (10) Kjær, J.; Olsen, P.; Bach, K.; Barlebo, H. C.; Ingerslev, F.; Hansen, M.; Sørensen, B. H. Leaching of estrogenic hormones from manure-treated structured soils. Environ. Sci. Technol. 2007, 41, 3911–3917. (11) Leet, J. K.; Gall, H. E.; Sepulveda, M. S. A review of studies on androgen and estrogen exposure in fish early life stages: effects on gene and hormonal control of sexual differentiation. J. Appl. Toxicol. 2011, 31, 379–398. (12) Shappell, N. W.; Elder, K. H.; West, M. Estrogenicity and nutrient concentration of surface waters surrounding a large confinement dairy operation using best management practices for land application of animal waste. Environ. Sci. Technol. 2010, 44, 2365–2371. (13) Holbech, H.; Kinnberg, K.; Petersen, G. I.; Jackson, P.; Hylland, K.; Norrgren, L.; Bjerregaard, P. Detection of endocrine disrupters: evaluation of a fish sexual development test (FSDT). Comp. Biochem. Phys. C 2006, 144, 57–66. (14) Jensen, K.; Makynen, E. A.; Kahl, M. D.; Ankley, G. T. Effects of the feedlot contaminant 17α-trenbolone on reproductive endocrinology of the fathead minnow. Environ. Sci. Technol. 2006, 40, 3112–3117. (15) Huang, C.; Zhang, Z.; Wu, S.; Zao, Y.; Hu, J. In vitro and in vivo estrogenic effects of 17α-estradiol in medaka (Oryzias latipes). Chemosphere 2010, 80, 608–612. (16) Ankley, G. T.; Jense, K. M.; Makynen, E. A.; Kahl, M. D.; Korte, J. J.; Hornung, M. W.; Henry, T. R.; Denny, J. S.; Leino, R. L.; Wilson, V. S.; Cardon, M. C.; Hartig, P. C. Effects of the androgenic growth promoter 17β-trenbolone on fecundity and reporoductive endocrinology of the fathead minnow. Environ. Toxicol. Chem. 2003, 22, 1350– 1360. (17) Dutta, S.; Inamdar, S.; Tso, J.; Aga, D. S.; Sims, J. T. Free and conjugated estrogen exports in surface-runoff from poultry litter-amdended soil. J. Environ. Qual. 2010, 39, 1688–1698. (18) Huang, X.; Lee, L. S. Effects of dissolved organic matter from animal waste effluent on chlorpyrifos sorption by soilds. J. Environ. Qual. 2001, 30, 1258–1265. (19) Gall, H. E.; Jafvert, C. T.; Jenkinson, B. Integrating hydrograph modeling with real-time monitoring to generate hydrograph-specific sampling schemes. J. Hydrol. 2010, 393, 331–340. (20) Lim, T. H.; Gin, K. Y. H.; Chow, S. S.; Chen, Y. H.; Reinhard, M.; Tay, J. H. Potential for 17β-estradiol and estrone degradation in a recharge aquifer system. J. Environ. Eng. ASCE 2007, 133, 819– 826. (21) Ying, G.; Toze, S.; Hanna, J.; Yu, X.; Dillon, P. J.; Kookana, R. S. Decay of endocrine-disrupting chemicals in aerobic and anoxic groundwater. Water Res. 2008, 42, 1133–1141. (22) Kladivko, E. J.; Brown, L. C.; Baker, J. L. Pesticide transport to subsurface tile-drains in humid regions of North America. Crit. Rev. Environ. Sci. Technol. 2001, 31, 1–62. (23) Hardie, M. A.; Cotching, W. E.; Doyle, R. B.; Holz, G. K.; Lisson, S.; Mattern, K. Effect of antecedent moisture on preferential flow in a texture-contrast soil. J. Hydrol. 2010, 398, 191–201. (24) Geohring, L. D.; McHugh, O. V.; Walter, M. T.; Steenhuis, T. S.; Akhtar, M. S.; Walter, M. F. Phosphorus transport into subsurface drains by macropores after manure applications: implications for best management practices. Soil Sci. 2001, 166, 896–909. (25) Haws, N. W. Integrated Flow and Transport Processes in SubSurfafce Drained Agricultural Fields; Dissertation, Purdue University, West Lafayette, IN, 2003. 8763
dx.doi.org/10.1021/es2011435 |Environ. Sci. Technol. 2011, 45, 8755–8764
Environmental Science & Technology
ARTICLE
(26) Villholth, K. G.; Jense, K. M.; Fredericia, J. Flow and transport processes in a macroporous subsurface-drained glacial till soil I: field investigations. J. Hydrol. 1998, 207, 98–120. (27) Lapen, D. R.; Topp, E.; Metcalfe, C. D.; Li, H.; Edwards, M.; Gottschall, N.; Bolton, P.; Curnoe, W.; Payne, M.; Beck, A. Pharmaceutical and personal care products in tile drainage following land application of municipal biosolids. Sci. Total Environ. 2008, 50–65. (28) Shipitalo, M. J.; Gibbs, F. Potential for earthworm burrow to transmit injected animal wastes to tile drains. Soil Sci. Soc. Am. J. 2000, 64, 2103–2109. (29) Steenhuis, T. S.; Staubitz, W.; Andreini, M. C. Preferential movement of pesticides and tracers in agricultural soils. J. Irrig. Drain. Eng. ASCE 1990, 116, 50–56. (30) Jensen, M. B.; Jorgensen, P. R.; Hansen, H. C. B.; Nielsen, N. E. Biopore mediated subsurface transport of dissolved orthophosphate. J. Environ. Qual. 1998, 27, 1130–1137. (31) Frankenberger, J. R.; Jones, D. D.; Gould, C.; Jacobs, L. Best Environmental Management Practices: Farm Animal Production; Purdue University Cooperative Extension Service and Michigan State Universtion Extension Program: West Lafayette, IN, 2004.
8764
dx.doi.org/10.1021/es2011435 |Environ. Sci. Technol. 2011, 45, 8755–8764
ARTICLE pubs.acs.org/est
Uptake of Np(IV) by CSH Phases and Cement Paste: An EXAFS Study Xavier Gaona,*,†,|| Rainer D€ahn,† Jan Tits,† Andreas C. Scheinost,‡,§ and Erich Wieland† †
Laboratory for Waste Management, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland Helmholtz-Zentrum Dresden-Rossendorf, Institute of Radiochemistry, D-01314 Dresden, Germany § Rossendorf Beamline (ROBL) at ESRF, F-38043 Grenoble, France ‡
bS Supporting Information ABSTRACT: Nuclear waste disposal concepts developed worldwide foresee the use of cementitious materials for the immobilization of long-lived intermediate level waste (ILW). This waste form may contain significant amounts of neptunium237, which is expected to be present as Np(IV) under the reducing conditions encountered after the closure of the repository. Predicting the release of Np(IV) from the cementitious near field of an ILW repository requires a sufficiently detailed understanding of its interaction with the main sorbing components of hardened cement paste (HCP). In this study, the uptake of Np(IV) by calcium silicate hydrates (CSH) and HCP has been investigated using extended X-ray absorption fine structure (EXAFS) spectroscopy. The EXAFS studies on Np(IV)-doped CSH and HCP samples reveal that Np(IV) is predominantly incorporated in the structure of CSH phases having different Ca:Si ratios. The two main species identified correspond to Np(IV) in CSH with a Ca:Si mol ratio of 1.65 as in fresh cement and with a Ca:Si mol ratio of 0.75 as in highly degraded cement. The local structure of Np(IV) changes with the Ca:Si mol ratio and does not depend on pH. Furthermore, Np(IV) shows the same coordination environment in CSH and HCP samples. This study shows that CSH phases are responsible for the Np(IV) uptake by cementitious materials and further that incorporation in the interlayer of the CSH structure is the dominant uptake mechanism.
1. INTRODUCTION Cementitious materials will be used for construction of the engineered barrier in repositories for long-lived intermediate level radioactive waste (ILW). The cementitious near field in underground repositories is subject to chemical alteration processes caused by the interaction of groundwater with hardened cement paste (HCP). In particular, calcium silicate hydrates (CSH), which are the most abundant cement mineral in HCP (∼50 wt %), transform from Ca-rich (Ca:Si (C:S) mol ratio ∼1.8, pH ∼12.5) to Ca-depleted (C:S mol ratio ∼0.7, pH ∼10) phases during the course of cement degradation. Calcium silicate hydrates are very important sinks for di-, tri-, and hexavalent radionuclides13 and therefore they are highly relevant with a view to the safe disposal of radioactive waste in cement-based repositories. The behavior of neptunium in the cementitious near field is of particular importance because of its long half-life (237Np, t1/2 = 2.14 106 years), its radiotoxicity, and its redox sensitivity. Under the reducing conditions of a repository4 and in line with the thermodynamic data available for actinides,5 neptunium is expected to predominantly form Np(IV) species. Despite the relevance for safety considerations, there are very few experimental studies on r 2011 American Chemical Society
the interaction of Np(IV) with cementitious materials. Zhao et al.6 observed the partial reduction of Np(V) to Np(IV) during the uptake by cement paste, although it appears that the reduction mechanism did not affect the overall partitioning between the aqueous phase and cementitious material. Considering its analogy with Th(IV), Np(IV) is expected to be significantly retarded by HCP and CSH,7 although little is known on the mechanisms controlling the uptake process. The present study aims at developing a mechanistic understanding of Np(IV) uptake by cementitious materials. Given the amorphous character of the solid phases considered in this work and the relatively low concentrations of Np used, extended X-ray absorption fine structure (EXAFS) spectroscopy was considered to be the most suitable technique to study the local structure of Np in Np(IV)-doped CSH and HCP samples. The former cement phase has been selected because it is a major constituent of HCP and an expected sink of Np(IV) in cement paste. Received: April 15, 2011 Accepted: August 31, 2011 Revised: August 8, 2011 Published: August 31, 2011 8765
dx.doi.org/10.1021/es2012897 | Environ. Sci. Technol. 2011, 45, 8765–8771
Environmental Science & Technology
ARTICLE
Table 1. Np(IV) Sorption Samples of CSH and HCP Materials Prepared for EXAFS Measurements CNa2S2O4 [M]
pHexp
EH [V]
CNp dry [ppm]
HCP, ACWgg
5 103
13.3
0.76 ( 0.05
2150
2150
|
CSH, C:S = 1.65
5 103
12.5
0.68 ( 0.05
8590
859
sequence of cement degradation
CSH, C:S = 1.07
5 103
11.9
0.62 ( 0.05
8590
859
CSH, C:S = 0.75
5 103
10.2
0.54 ( 0.05
8590
859
11.9
0.12 ( 0.05
8590
859
test sample: effect of Na2S2O4
5 103
13.3
0.65 ( 0.05
8590
859
test sample: effect of pH
sample
CSH, C:S = 1.07 CSH, C:S = 1.07 ACW
2. EXPERIMENTAL SECTION All solutions and suspensions were prepared from reagentgrade chemicals in Milli-Q water, previously degassed with N2. All experiments were carried out in a glovebox under a N2 atmosphere (CO2 and O2 e 2 ppm).
2.1. Preparation of CSH Phases and Cement Paste As Sorbing Materials. CSH phases with C:S mol ratios 0.75,
1.07, and 1.65 were synthesized either in Milli-Q water or in an artificial cement pore water (ACW) using a procedure described elsewhere.2 The ACW was prepared using 0.18 M KOH and 0.114 M NaOH, which resulted in a pH of 13.3. AEROSIL 300 (SiO2) (Degussa-Huls AG, Baar Switzerland) was mixed with CaO (obtained by firing CaCO3 at 1000 °C) in polyallomere centrifuge tubes to obtain the target C:S mol ratios of 0.75, 1.07, and 1.65. The selected C:S ratios represent the entire CSH sequence during the course of cement degradation. To these mixtures, ACW or Milli-Q water was added to prepare samples with a solid to liquid ratio (S:L) of 2.5 g 3 L1. Sodium dithionite (Na2S2O4) 5 103 M was added to all suspensions (except one, see below) to establish the very reducing conditions stabilizing Np(IV) (EH defined by the border of water reduction). The suspensions were mixed, homogenized with a shear mixer, and equilibrated for at least 2 weeks on an end-over-end shaker.2 A sulfate-resistant Portland cement (CEM I 52.5 N HTS) provided by Lafarge (France), was used to prepare the HCP samples. A modified ACW (ACWgg) with pH = 13.3 and simulating the equilibrium composition of pore water with fresh HCP8 was added to the solid material to achieve the final S:L of 5 g 3 L1. 2.2. Preparation of Sorption Samples. A neptunium(IV) stock solution was prepared in a glovebox by the electrolytic reduction of Np(V) as reported elsewhere.9 The purity of the +IV redox state was confirmed by UVvis. The concentration of the Np(IV) stock solution was 4.5 103 M (or 28 kBq 3 mL1) in a 1.1 M HCl matrix. Sorption samples were prepared in 40 mL polyallomere centrifuge tubes (Beckmann Instruments Inc.). An aliquot of the Np(IV) tracer (0.6 mL) was added to 30 mL CSH suspensions with 5 103 M Na2S2O4. Only 0.3 mL of the Np(IV) tracer was added to 30 mL of HCP suspension. Two additional test samples were prepared to determine the effect of Na2S2O4 and pH in the sorption process: (a) CSH with C:S = 1.07 but prepared in the absence of Na2S2O4, and (b) CSH with C:S = 1.07 synthesized in ACW at pH = 13.3 rather than Milli-Q (pH = 12.0). Both the concentration of Np in the different sorbing materials (CNp dry, mass balance assuming 100% of Np sorbed and dry weight of the sorbing material) and the concentration expected in the wet synchrotron samples (CNp wet, estimated assuming 90% water content in the CSH paste3) are listed in Table 1 along with the measured pH and EH values.
CNp wet [ppm]
comments
jV
The centrifuge tubes were shaken end-over-end for 2 months; pH and EH were monitored during this period of time, revealing no significant variations. After equilibration, synchrotron samples were prepared according to the protocol described below. Finally, a complete phase separation of the solid and liquid phase was carried out by centrifugation (1 h at 90 000g) on an L7-35 ultracentrifuge (Beckman Instruments Inc.). Triplicate samples were withdrawn from the supernatant solution after centrifugation, and analyzed by liquid scintillation counting (LSC) with alpha/beta discrimination to determine the Np(IV) activity in solution. The latter measurements were used to estimate the distribution ratio (Rd) of Np(IV). 2.3. Synchrotron-Based Spectroscopic Investigations. Sample Preparation. Np(IV) is very sensitive to oxidation in contact with air. Therefore, special precautions were taken to avoid/ minimize the access of oxygen to the samples, even inside the glovebox. After equilibration, the separation of the sorbing material from the supernatant solution was achieved either by filtration with 220 nm Millipore filters or by centrifugation at 15 000 rpm for 5 min (outside the glovebox) and separation of the supernatant solution (inside the glovebox). After completion of the separation step, the wet material was placed in a polyethylene (PE) sample holder, sealed with Kapton tape (first containment), and thermo-sealed with a PE cap (second containment). Gamma activity measurements of the samples (mandatory according to European transport regulations) were carried out with the PE holders completely immersed in dry ice (78 °C) to minimize the diffusion of O2 into the sample. Finally, the samples were transferred to a N2-filled Dewar (Voyager 12, Air Liquide - DMC, France). The complete process between phase separation and final storage in the N2-filled Dewar lasted less than 2 h. The N2-filled Dewar was also used to transport the samples to the synchrotron facility. EXAFS Data Collection. EXAFS experiments were conducted at the Rossendorf Beamline (ROBL) at the European Synchrotron Radiation Facility, Grenoble, France (ESRF). Spectra were recorded across the neptunium LIII-edge (17610 eV). The beamline is equipped with a water-cooled Si(111) double crystal monochromator between Pt-coated mirrors for beam collimation and rejection of higher order harmonics in the monochromatic beam. The monochromator energy was calibrated by assigning the first inflection point of the K-absorption edge of a Y foil to 17038 eV. Several fluorescence spectra (3 to 9) were acquired using a 13-element high-purity Ge solid-state detector and averaged to improve the signal-to-noise ratio. The risk of O2 diffusion into the samples was minimized by maintaining the sample temperature at 15 K using a closed-cycle He-cryostat. The transfer of the samples from the N2-Dewar to the Hecryostat occurred within less than 5 min, which limited temperature rise. 8766
dx.doi.org/10.1021/es2012897 |Environ. Sci. Technol. 2011, 45, 8765–8771
Environmental Science & Technology
ARTICLE
Figure 1. (a) Np LIII-edge XANES spectra, (b) k3-weighted EXAFS spectra (experimental (solid line) and reconstruction by 2 components (dots) with the residual of the LC fits), and (c) FT for Np(IV) sorbed on CSH 0.75, 1.07, 1.07 (no Na2S2O4), 1.07 ACW, HCP, and CSH 1.65. Fit in CSH 1.07, 1.07 (no Na2S2O4), 1.07 ACW, and HCP calculated by ITFA for 2.5 e k [Å1] e 9.5.
To assess the kinetics of Np(IV) oxidation in the presence of O2, a few samples were kept at room temperature and air atmosphere after the EXAFS measurements. The oxidation of Np(IV) was observed within 100 h after air exposure using EXAFS. EXAFS Data Reduction and Analysis. EXAFS data reduction and analysis were performed using the SIXPACK/IFEFFIT software following standard procedures.10 The SAMVIEW program was used for dead-time corrections and to average the raw spectra. Reduction and modeling of the EXAFS data were performed using the ATHENA/ARTEMIS package.11 After background subtraction, the energy was converted to photoelectron wave vector units (Å1) by assigning the ionization energy of the neptunium LIII-edge (17610 eV), E0, to the first inflection point of the absorption edge. Radial structure functions (RSFs) were obtained by Fourier transforming k3-weighted χ(k) functions between 2.0 and 14.0 Å1 using a Kaiser-Bessel window function. The EXAFS spectra were analyzed in a two-step approach. The first step consisted of a statistical analysis of the data based on principal component analysis (PCA) and iterative target-transformation factor analysis (ITFA). Principal component analysis is used to determine whether a set of spectra can be represented as linear combinations of a smaller number of independent component spectra. Three PCA programs were used in this study: PCA tools included in the Sixpack and Berkeley Lab interfaces,12 as well as the factor analysis code developed by A. Rossberg.13 All codes led to very similar results, although calculations shown in this work reflect only the outcome using the latter. The criterion applied to determine the minimum of independent components sufficient to reproduce the set of experimental spectra was the indicator value of Malinowsky (IND).14 The calculation of the IND factor was performed in k3χ(k) spaces for both 2 e k [Å1] e 7.5 and 2.5 e k [Å1] e 9.5. The use of ITFA allows component spectra with a physical meaning to be extracted. Components were converted to a new basis by linear combination aiming at the reproduction of the original spectra as a sum of the new components with coefficients (weights) between 0 and 1.12,13 The spectra with a coefficient of ∼1 indicate the presence of a single component. They were considered to correspond to the spectra of the main species in the studied systems.
Structural information was obtained by applying a multishell fit approach on the EXAFS data of the single components identified by PCA and ITFA. Due to lacking XRD structures for CaSi-bearing Np(IV) compounds, theoretical single scattering paths (SS) for the fit approach were calculated with FEFF8.415 using the structure of britholite16 (a Th(IV)SiCa compound) as a model compound. The paths involving Th(IV) had been recalculated as Np(IV) paths using the ATOMS routine in ARTEMIS. The relevant SS paths included NpO, NpSi, and NpCa. Single scattering paths for NpNp and NpC were initially considered in the fitting process, but disregarded in the final fit. The total number of independent fitting parameters was 10. The amplitude reduction factor, S02, was fixed to 1.0 for all fits. Uncertainties associated with the structural parameters were directly calculated using ARTEMIS along with the R-factor, which is an indicator of the quality of the fit.
3. RESULTS AND DISCUSSION Transport and storage of Np(IV) samples in liquid N2 were found to preserve the original oxidation state. In contrast, exposure of the samples to room temperature and presence of air caused the oxidation of Np(IV) to Np(V) within a few hours even in the presence of a protecting double containment (see further details in the Supporting Information). No significant differences in the EXAFS spectra were observed for Np(IV) samples prepared in presence and absence of Na2S2O4 (Figure 1). This observation indicates a strong stabilization of Np(IV) by CSH in a glovebox under anoxic conditions, which is further reflected by the very high Rd values determined (105106 L 3 kg1, see also Tits et al.17). The comparison of all k3χ(k) spectra, together with the calculated IND values and the ITFA, showed that two components are sufficient to reconstruct the spectra. Factor loadings of the spectra of CSH 0.75 and 1.65 suggested that these samples represent the two end-member species. Therefore, these spectra were considered to represent 100% of components 1 or 2, respectively, in the iterative target test step. With this constraint, the percentage of components 1 and 2 could be determined in the remaining spectra (Table 2). Summed species values between 8767
dx.doi.org/10.1021/es2012897 |Environ. Sci. Technol. 2011, 45, 8765–8771
Environmental Science & Technology
ARTICLE
Figure 2. Experimental and shell-fitted (a) k3-weighted Np LIII-edge EXAFS spectra and (b) corresponding Fourier transforms (modulus and imaginary parts) for Np(IV) sorbed on CSH 0.75 and 1.65. Solid lines represent experimental data; dashed lines represent fits in k3-weighted EXAFS spectra and Fourier transforms.
94% and 100% were obtained for these EXAFS spectra (CSH 1.0, CSH 1.0 ACW, CSH 1.0 no Na2S2O4, HCP). This indicates either the uncertainty associated with the iterative target transformation or suggests the presence of very small amounts of additional Np(IV) species. This uncertainty also applies to the species defined as main components (CSH 0.75 and CSH 1.65). Further details on the PCA and the ITFA are provided as Supporting Information. The results from linear combination (LC) analyses are illustrated in Figure 1, showing XANES and k3-weighted EXAFS spectra and the corresponding Fourier transform (FT) of all Np(IV) samples. The experimental EXAFS spectra can be properly reconstructed by two components, as indicated by the small error (dotted line). The two end-members can be clearly identified by the presence of the Si shell in CSH 0.75 at R + ΔR = 2.88 Å and the Ca shell in CSH 1.65 at R + ΔR = 3.45 Å (Figure 1c). This observation is in agreement with the structures of CSH phases with a molecular environment richer in Si in CSH 0.75 but richer in Ca in CSH 1.65.18 The coordination environments of Np(IV) in CSH 1.07 and CSH 1.07 ACW were found to be very similar (Table 2). Both sorbing materials have the same C:S mol ratio, but they differ in the pH values of the equilibrium solutions (11.9 and 13.3). This observation indicates that the coordination environment of Np(IV) taken up by CSH is not affected by pH. This finding is further in accordance with the aqueous speciation of Np(IV) in the pH range 10 e pH e 13.3, which is dominated by Np(OH)4(aq). Furthermore, a similar molecular environment was observed for Np(IV) in HCP and CSH 1.65 (Table 2). This finding indicates that CSH is the main sorbing phase for Np(IV) in HCP, as CSH with C:S mol ratio >1.5 is known to dominate in fresh HCP.19
Table 2. Composition of Np(IV) Samples in CSH 0.75, 1.07, 1.07 (no Na2S2O4), 1.07 ACW, 1.65, and HCP as Calculated by ITFA (Calculations Were Performed in k3χ(k) Space for 2.5 e k [Å1] e 9.5) sample
a
component 1 component 2 sum
Np(IV)CSH 0.75
1.00a
0.00
Np(IV)CSH 1.07
0.27
0.70
Np(IV)CSH 1.07 (no Na2S2O4)
0.16
0.80
0.96
Np(IV)CSH 1.07 ACW
0.35
0.59
0.94
Np(IV)HCP
0.00
1.00
1.00
Np(IV)CSH 1.65
0.00
1.00a
0.97
Values fixed in the ITFA procedure.
The structural parameters of the two components were determined using the EXAFS spectra of the Np(IV)CSH 0.75 and Np(IV)CSH 1.65 samples. Fitting was carried out in R-space with 2.5 e k [Å1] e 9.5 and 1.0 e R [Å] e 4.4. These k- and R-ranges resulted in 15 independent parameters, as calculated with the Nyquist criterion. Experimental and theoretical k3-weighted Np LIII-edge EXAFS spectra as well as the corresponding Fourier transform magnitude (FT) are shown in Figure 2, while the resulting structural parameters are listed in Table 3. About eight oxygen atoms were found at a distance of 2.31 ( 0.01 Å in the first coordination shell of Np (Table 3), which is consistent with solid (or sorbed) Np(IV) species. Note that the NpO distances are significantly shorter than those reported for the aquo ion (2.37 Å to 2.41 Å2022). Nevertheless, the distances agree well with NpO distances attributed to hydrolyzed Np(IV) species sorbed to fulvic acids.22 In contrast to Denecke et al.,22 no split in the oxygen shell could be observed for Np(IV) sorbed in 8768
dx.doi.org/10.1021/es2012897 |Environ. Sci. Technol. 2011, 45, 8765–8771
Environmental Science & Technology
ARTICLE
Table 3. Structural Parameters Determined from the Np(IV)CSH 0.75 and Np(IV)CSH 1.65 EXAFS Data (10 Free Parameters Were Fitted) sample Np(IV)CSH 0.75
Np(IV)CSH 1.65
a
NpO
NpSi
NpCa
quality of the fit
NO = 7.7 + 1.2
NSi = 4.6 ( l.l
NCa = 8.0 ( 3.0
ΔE = 6.6 ( 1.1
rNp-O = 2.31 ( 0.01 Å σ2 = 0.010 ( 0.002 Å2
rNp-Si = 3.63 ( 0.02 Å σ2 = 0.010 ( 0.008 Å2
rNp-Ca = 4.19 ( 0.04 Å σ2 = 0.02 Å2a
R-factor =0.001
NO = 8.3 ( 1.2
NSi = 2.9 ( l.l
NCa = 12.7 ( 2.8
ΔE = 4.5 ( 0.8
rNp-O = 2.31 ( 0.01 Å
rNp-Si = 3.60 ( 0.03 Å
rNp-Ca = 4.18 ( 0.04 Å
R-factor = 0.002
σ2 = 0.010 ( 0.002 Å2
σ2 = 0.010 Å2a
σ2 = 0.019 ( 0.009 Å2
Values fixed in the fit.
Figure 3. Candidates for Np(IV)CSH end members, according to Np structural parameters determined in this work and in agreement with CSH structures reported elsewhere18 for low (a) and high (b) C:S ratios.
CSH, probably because of the lower resolution in our EXAFS data (0.22 Å). Fits involving second and third shells were successfully achieved by considering Si and Ca backscatterers. The fits were not improved when introducing additional NpO and/or NpC backscattering pairs. Polynuclear Np surface complexes or precipitates could be ruled out, since NpNp backscattering contributions could not be identified by Fourier back transformation of the FT peaks at longer distances (Figure S5 in the Supporting Information). The coordination numbers with Si (NSi) of 4.6 ( 1.1 and 2.9 ( 1.1 obtained for both Np(IV)CSH 0.75 and Np(IV)CSH 1.65, respectively, suggest an incorporation mechanism in the CSH structure. The observation of larger NSi for Np(IV)CSH 0.75 than for Np(IV)CSH 1.65 is consistent with the incorporation of Np(IV) in the interlayer of the CSH structure as the number of Si atoms is larger in CSH 0.75. Large NCa and DebyeWaller factors for the NpCa backscattering pairs were determined for both Np(IV)CSH 0.75 and 1.65. Because of the strong correlation between NCa and σ2, we consider likely an overestimation of the NCa values reported in Table 3. The use of an increased k-range of 2.5 e k [Å1] e 12.5 allowed to fit the data with two NpCa shells at 4.08 and 4.25 Å, while the R-factors changed for the worse (R = 0.07 0.09). Importantly, the changes in the structural parameters (Table 3), which are visible in the Ca amplitudes of the FT in Figure 2b, indicate a Np(IV) coordination environment richer in Ca for Np(IV) incorporated in CSH 1.65. This is again consistent with the structure of CSH phases with high C:S mol ratios. The bridging Si tetrahedra prevailing in the interlayer structure of CSH 0.75 are completely removed and the vacancies are occupied by Ca in CSH 1.65.23 Structures for Np(IV)CSH 0.75 and Np(IV)CSH 1.65, which are consistent with the
experimental observations made in this work and structures proposed earlier by Garbev et al.18 for nanocrystalline CSH phases, are provided in Figure 3. Hence, the structure proposed in Figure 3b indicates the substitution of Ca by Np in the interlayer. Although no direct comparison with the data in Garbev et al.18 is possible, the interlayer distance provided in the latter study for CSH with C:S 1.25 (11.7 Å) in combination with a tobermorite-like structure where all bridging Si tetrahedra have been removed (Figure 8c in Garbev et al.18) allows the quantification of certain ancillary metrical information. The distances resulting for Ca in the interlayer (CaO: 2.72.8 Å; CaSi: 4.3 Å) are longer than those determined in this work for Np, although a slight decrease in the interlayer distance by ∼1 Å would result in more consistent values (CaO: 2.3 Å; CaSi: 3.73.8 Å; compared to NpO: 2.3 Å ; NpSi: 3.6 Å, Table 3). Structural differences arising from the type of CSH phases used in this work and in Garbev et al.,18 as well as differences in the electronic properties of Np(IV) and Ca, may be responsible for this disagreement. Despite the feasibility of the structures proposed, the issue of charge balance in the uptake process remains uncertain. Garbev et al.18 determined the presence of Ca with N = 6 in the interlayer of CSH with C:S = 1.25. Four of these coordination positions were occupied by O-atoms of the main Si-layer (similarly to the sorbed species proposed for Np(IV) in Figure 3b), whereas two additional coordinating positions were bonded to H2O. In the case of Np(IV), these latter positions might correspond to two OH-groups (as hydrolysis of the Np4+ cation), therefore resulting in a sorbed species with a charge similar to Ca2+. Other mechanisms may involve coupled one-for-two substitution (1Np4+ S 2Ca2+) where one of the Ca-sites would remain vacant, or protonationdeprotonation reactions of the silanol groups exposed toward the interlayer of CSH. To this point, 8769
dx.doi.org/10.1021/es2012897 |Environ. Sci. Technol. 2011, 45, 8765–8771
Environmental Science & Technology
ARTICLE
cement degradation. In combination with wet chemistry data currently available for Np(IV)CSH and Th(IV)CSH systems,17 the results suggest that tetravalent actinides will be strongly retarded in an evolving cement-based repository during the complete sequence of cement degradation.
’ ASSOCIATED CONTENT
bS
Supporting Information. Two tables and five figures. This material is available free of charge via the Internet at http:// pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
the EXAFS data reported in this work do not allow an unambiguous identification of the involved charge compensation mechanisms. The Np(IV)Si distances determined in this study were compared with the distances between tetravalent actinides and Si reported for Pu(IV), U(IV), and Th(IV) with the aim of checking the consistency of the distances and gaining further insights into the type of uptake mechanism of Np(IV) in CSH phases. Bond distances of the An(IV) with other elements tend to increase with increasing crystal radii (as well as the ionic radii) from Pu(IV) to Th(IV). For example, this trend is reflected in the An(IV)O bond distance both for aqueous species and solid compounds (Figure 4). Furthermore, the An(IV)Si distance increases from Pu(IV) to Th(IV) although An(IV) and Si are not directly bonded in silicate compounds. This trend is illustrated by comparing the An(IV)Si distances in mono- and bicoordinated silicates (Figure 4) as reported in XRD studies on An(IV)SiO4(cr) compounds2426 and an EXAFS study on Th(IV) sorption on silica under acidic conditions.27 The Np(IV)Si distances determined in this study (3.57 3.65 Å) are significantly shorter than those expected for monocoordinated Np(IV)silicates (∼ 3.8 Å) (Figure 4). This observation suggests that bicoordinated silicate-type bonding is involved in the uptake of Np(IV) by CSH. Note that EXAFS data fitting only provides an average Np(IV)Si distance, which involves both mono- and bicoordinated Np(IV)silicate structures. Considering the structure of CSH (see Figure 3 and data in Garbev et al.18), it is conceivable to rationalize the observed Np(IV)Si distances in terms of mono- and bicoordinated silicate bonding in the Si-rich environment of the CSH interlayer. Hence, along with the high NSi and NCa, the short distances observed for Np(IV)Si coordination indicate an uptake mechanism where Np(IV) is incorporated in the interlayer of the CSH structure. The results obtained in this study allow a detailed mechanistic understanding of the uptake of Np(IV) by cementitious materials to be presented. The predominant role of CSH phases in the uptake process has been confirmed for the different stages of the
Present Addresses
)
Figure 4. An(IV)O and An(IV)Si bond distances as reported in the literature for aqueous, sorbed, and solid species of Pu(IV), Np(IV), U(IV), and Th(IV). 5 ThSiO4(cr), XRD;24 7 USiO4(cr), XRD;25 2 PuSiO4(cr), XRD;26 —0— Th sorbed on silica, EXAFS;27 —f— Np(IV) incorporated in CSH phases, EXAFS this work; ! An4+ OH2 aquo ions, EXAFS (compiled in Neck and Kim28 from different EXAFS data sources); shaded " ThO2(cr), XRD;29 left solid " UO2(cr), XRD;30 " NpO2(cr), XRD;31 ` PuO2(cr), XRD.32 Crystal radii as reported elsewhere33,34 for An(IV) with N = 9.
Institut f€ur Nukleare Entsorgung, Karlsruhe Institute of Technology, P.O. Box 3640, D-76021 Karlsruhe, Germany.
’ ACKNOWLEDGMENT The staff of the ROBL beamline at ESRF (especially A. Rossberg and D. Banerjee) is thanked for the experimental assistance with the EXAFS measurements. Thanks are extended to D. Kunz for assistance during the measuring campaigns. The research leading to these results has received funding from the European Union’s European Atomic Energy Community’s (Euratom) Seventh Framework Programme FP7/2007-2011 under grant agreement 212287 (RECOSY project) and from the National Cooperative for the Disposal of Radioactive Waste (NAGRA), Switzerland. X. G. acknowledges the Generalitat de Catalunya for his postdoctoral grant (“Beatriu de Pinos” program). ’ REFERENCES (1) Tits, J.; Stumpf, T.; Rabung, T.; Wieland, E.; Fanghanel, T. Uptake of Cm(III) and Eu(III) by calcium silicate hydrates: A solution chemistry and time-resolved laser fluorescence spectroscopy study. Environ. Sci. Technol. 2003, 37 (16), 3568–3573. (2) Tits, J.; Wieland, E.; M€uller, C. J.; Landesman, C.; Bradbury, M. H. Strontium binding by calcium silicate hydrates. J. Colloid Interface Sci. 2006, 300 (1), 78–87. (3) Harfouche, M.; Wieland, E.; Dahn, R.; Fujita, T.; Tits, J.; Kunz, D.; Tsukamoto, M. EXAFS study of U(VI) uptake by calcium silicate hydrates. J. Colloid Interface Sci. 2006, 303 (1), 195–204. (4) Berner, U. Project Opalinus Clay: Radionuclide Concentration Limits in the Cementitious Near-Field of an ILW Repository; PSI report 02-26; Paul Scherrer Institut: Villigen, Switzerland, 2003. (5) Guillaumont, R.; Fangh€anel, J.; Neck, V.; Fuger, J.; Palmer, D. A.; Grenthe, I.; Rand, M. H. Chemical Thermodynamics, Vol. 5, Update on the Chemical Thermodynamics of Uranium, Neptunium, Plutonium, Americium and Technetium; Elsevier, North Holland: Amsterdam, 2003. (6) Zhao, P.; Allen, P. G.; Sylwester, E. R.; Viani, B. E. The partitioning of uranium and neptunium onto hydrothermally altered concrete. Radiochim. Acta 2000, 88 (911), 729–736. (7) Wieland, E.; Van Loon, L. R. Cementitious Near-Field Sorption Database for Performance Assessment of an ILW Repository in Opalinus Clay; Nagra Technical Report NTB 02-20; Nagra: Wettingen, Switzerland, 2002. (8) Wieland, E.; Tits, J.; Ulrich, A.; Bradbury, M. H. Experimental evidence for solubility limitation of the aqueous Ni(II) concentration and isotopic exchange of Ni-63 in cementitious systems. Radiochim. Acta 2006, 94 (1), 29–36. 8770
dx.doi.org/10.1021/es2012897 |Environ. Sci. Technol. 2011, 45, 8765–8771
Environmental Science & Technology (9) Denecke, M. A.; Dardenne, K.; Marquardt, C. M. Np(IV)/ Np(V) valence determinations from Np L3 edge XANES/EXAFS. Talanta 2005, 65 (4), 1008–1014. (10) Webb, S. M. SIXpack: A graphical user interface for XAS analysis using IFEFFIT. Phys. Scr. 2005, T115, 1011–1014. (11) Ravel, B.; Newville, M. ATHENA, ARTEMIS, HEPHAESTUS: Data analysis for X-ray absorption spectroscopy using IFEFFIT. J. Synchrotron Radiat. 2005, 12, 537–541. (12) Marcus, M. The PCA (Principal Components Analysis) Program. Manual of the PCA software from the 10.3.2 beamline at Berkeley Lab, 2005. (13) Rossberg, A.; Reich, T.; Bernhard, G. Complexation of uranium(VI) with protocatechuic acid - application of iterative transformation factor analysis to EXAFS spectroscopy. Anal. Bioanal. Chem. 2003, 376 (5), 631–638. (14) Malinowski, E. R. Determination of number of factors and experimental error in a data matrix. Anal. Chem. 1977, 49 (4), 612–617. (15) Rehr, J. J.; Albers, R. C. Theoretical approaches to x-ray absorption fine structure. Rev. Mod. Phys. 2000, 72 (3), 621–654. (16) Oberti, R.; Ottolini, L.; della Ventura, G.; Parodi, G. C. On the symmetry and crystal chemistry of britholite: New structural and microanalytical data. Am. Mineral. 2001, 86 (9), 1066–1075. (17) Tits, J.; Gaona, X.; Laube, A.; Wieland, E. Influence of the redox state on the neptunium sorption by cementitious materials. In Proceedings of ReCosy 3rd Annual Workshop; Altmaier, M., Duro, L., Grive, M., Montoya, V., Eds.; Balaruc-les-Bains, France, 2011. (18) Garbev, K.; Bornefeld, M.; Beuchle, G.; Stemmermann, P. Cell dimensions and composition of nanocrystalline calcium silicate hydrate solid solutions. Part 2: X-ray and thermogravimetry study. J. Am. Ceram. Soc. 2008, 91 (9), 3015–3023. (19) Taylor, H. F. W. Cement Chemistry, 2nd ed.; T. Telford: London, 1997. (20) Antonio, M. R.; Soderholm, L.; Williams, C. W.; Blaudeau, J. P.; Bursten, B. E. Neptunium redox speciation. Radiochim. Acta 2001, 89 (1), 17–25. (21) Hennig, C.; Ikeda-Ohno, A.; Tsushima, S.; Scheinost, A. C. The sulfate coordination of Np(IV), Np(V), and Np(VI) in aqueous solution. Inorg. Chem. 2009, 48 (12), 5350–5360. (22) Denecke, M. A.; Marquardt, C. M.; Rothe, J.; Dardenne, K.; Jensen, M. P., XAFS study of actinide coordination structure in Np(IV)fulvates. J. Nucl. Sci. Technol. 2002, Suppl. 3. (23) Kulik, D. Improving the structural consistency of C-S-H solid solution thermodynamic models. Cem. Concr. Res. 2011, 41 (5), 477–495. (24) Taylor, M.; Ewing, R. C. Crystal-structures of ThSiO4 polymorphs - huttonite and thorite. Acta Crystallogr. B 1978, 34 (Apr), 1074–1079. (25) Fuchs, L. H.; Gebert, E. X-Ray studies of synthetic coffinite, thorite and uranothorites. Am. Mineral. 1958, 43 (34), 243–248. (26) Keller, C. Untersuchungen €uber die Germanate und Silikate des Typs ABO4 der vierwertigen Elemente Thorium bis Americium. Nukleon 1963, 5, 41–48. (27) Osthols, E.; Manceau, A.; Farges, F.; Charlet, L. Adsorption of thorium on amorphous silica: An EXAFS study. J. Colloid Interface Sci. 1997, 194 (1), 10–21. (28) Neck, V.; Kim, J. I. An electrostatic approach for the prediction of actinide complexation constants with inorganic ligands-application to carbonate complexes. Radiochim. Acta 2000, 88 (911), 815–822. (29) Morss, L. R.; Richardson, J. W.; Williams, C. W.; Lander, G. H.; Lawson, A. C.; Edelstein, N. M.; Shalimoff, G. V. Powder neutrondiffraction and magnetic-susceptibility of 248CmO2. J. Less-Common Met. 1989, 156, 273–289. (30) Cooper, M. J. The analysis of powder diffraction data. Acta Crystallogr. A 1982, 38 (Mar), 264–269. (31) Zachariasen, W. H. Crystal chemical studies of the 5f-series of elements 0.12. new compounds representing known structure types. Acta Crystallogr. 1949, 2 (6), 388–390. (32) Belin, R. C.; Valenza, P. J.; Reynaud, M. A.; Raison, P. E. New hermetic sample holder for radioactive materials fitting to Siemens
ARTICLE
D5000 and Bruker D8 X-ray diffractometers: Application to the Rietveld analysis of plutonium dioxide. J. Appl. Crystallogr. 2004, 37, 1034–1037. (33) Shannon, R. D. Revised Effective Ionic-Radii and Systematic Studies of Interatomic Distances in Halides and Chalcogenides. Acta Crystallogr. A 1976, 32 (Sep1), 751–767. (34) Choppin, G. R.; Rizkalla, E. N. Solution Chemistry of Actinides and Lanthanides. In Handbook on the Physics and Chemistry of Rare Earths; Gschneidner Jr., K. A., Eyring, L., Eds.; North Holland: Amsterdam, 1994.
8771
dx.doi.org/10.1021/es2012897 |Environ. Sci. Technol. 2011, 45, 8765–8771
ARTICLE pubs.acs.org/est
Enhanced Reductive Dechlorination of Polychlorinated Biphenyl Impacted Sediment by Bioaugmentation with a Dehalorespiring Bacterium Rayford B. Payne,† Harold D. May,‡ and Kevin R. Sowers†,* †
Department of Marine Biotechnology, Institute of Marine and Environmental Technology, University of Maryland Baltimore County, Baltimore, Maryland 21202 ‡ Marine Biomedicine and Environmental Science Center, Department of Microbiology and Immunology, Medical University of South Carolina, Charleston, South Carolina 29425
bS Supporting Information ABSTRACT: Anaerobic reductive dehalogenation of commercial PCBs such as Aroclor 1260 has a critical role of transforming highly chlorinated congeners to less chlorinated congeners that are then susceptible to aerobic degradation. The efficacy of bioaugmentation with the dehalorespiring bacterium Dehalobium chlorocoercia DF1 was tested in 2-L laboratory mesocosms containing sediment contaminated with weathered Aroclor 1260 (1.3 ppm) from Baltimore Harbor, MD. Total penta- and higher chlorinated PCBs decreased by approximately 56% (by mass) in bioaugmented mesocosms after 120 days compared with no activity observed in unamended controls. Bioaugmentation with DF-1 enhanced the dechlorination of doubly flanked chlorines and stimulated the dechlorination of single flanked chlorines as a result of an apparent synergistic effect on the indigenous population. Addition of granulated activated carbon had a slight stimulatory effect indicating that anaerobic reductive dechlorination of PCBs at low concentrations was not inhibited by a high background of inorganic carbon that could affect bioavailability. The total number of dehalorespiring bacteria was reduced by approximately half after 60 days. However, a steady state level was maintained that was greater than the indigenous population of putative dehalorespiring bacteria in untreated sediments and DF1 was maintained within the indigenous population after 120 days. The results of this study demonstrate that bioaugmentation with dehalorespiring bacteria has a stimulatory effect on the dechlorination of weathered PCBs and supports the feasibility of using in situ bioaugmentation as an environmentally less invasive and lower cost alternate to dredging for treatment of PCB impacted sediments.
’ INTRODUCTION Polychlorinated biphenyls (PCBs) are persistent organic pollutants that are globally dispersed as a result of cycling between air, water, and soil.1 3 PCBs are hydrophobic and bioaccumulate throughout the food chain by absorption in the fatty tissue of animals and humans where they have been reported to act as endocrine disrupters4 and possible carcinogens.5 The most widely applied technologies for treating PCB-impacted sites are removal by dredging and subsequent transfer to hazardous waste landfills or in situ containment in subaqueous sites by capping.6 In addition to being cost prohibitive for treating large areas of contamination in rivers, lakes and coastal sediments, these technologies are disruptive to environmentally sensitive areas, such as marshes and wetlands. Development of an in situ treatment system that utilizes microbial activity, bioaugmentation, could potentially reduce high costs and negative environmental impacts associated with dredging and capping. There have been reports on the potential of aerobic bioaugmentation with bacteria, fungi and plants, but these processes have limited capacity to attack highly chlorinated congeners often r 2011 American Chemical Society
found in PCB impacted sites.7 Microbial reductive dechlorination of higher chlorinated PCB congeners has the potential to complement these processes, but there have been very few studies to date describing the use of anaerobic bioaugmentation to stimulate in situ treatment of PCB impacted sediments. Wu and Weigel8 and Bedard et al.9 observed only slight or no stimulation of weathered Aroclor 1260 in Housatanic River sediments bioaugmented with PCB enriched soil and sediment slurries. Natarajan at al10 reported dechlorination of weathered Aroclor 1242 and 1248 in microcosms inoculated with microbial granules from a UASB digestor. However, the possibility could not be ruled out that the observed activity resulted from stimulation of indigenous PCB dechlorinating bacteria by hydrogen generated from fermentation of wood powder added an electron donor rather than the result of bioaugmentation with dehalorespiring bacteria. Received: May 6, 2011 Accepted: September 8, 2011 Revised: September 1, 2011 Published: September 08, 2011 8772
dx.doi.org/10.1021/es201553c | Environ. Sci. Technol. 2011, 45, 8772–8779
Environmental Science & Technology More recently Dehalococcoides spp. and related species within the Chloroflexi have been identified that are capable of reductively dechlorinating PCBs by utilizing PCBs as terminal electron acceptors, a process termed dehalorespiration, and several strains with specific dechlorinating capabilities have been reported.11 14 The commercial PCB mixture Aroclor 1260 was reported to be significantly dechlorinated by a consortium consisting of one or more phylotypes enriched from sediment microcosms,15sediment-free microcosms11 and by an individual species, Dehalococcoides sp. CBDB1.16 A recent report showed that Dehalobium chlorocoercia DF-1, bacterium o-17, phylotypes SF1 and DEH10, each with different PCB congener specificities for dechlorination, would alter the overall pathway for dechlorination of spiked Aroclor 1260 differently depending on which microorganisms were added to sediment microcosms.17 Bedard et al.11 observed in a sediment-free enrichment culture that a critical mass of cells was required before reductive dechlorination of spiked Aroclor 1260 was detected and proposed that low indigenous numbers of dehalorespiring bacteria might explain why substantial attenuation of PCBs is rarely observed in the environment. Although these combined reports suggest that bioaugmentation could be a potential strategy for treatment of PCB-impacted sediments, several questions need to be addressed to determine if in situ treatment of weathered PCBs by bioaugmentation is possible. Bedard9 demonstrated that bioaugmentation with sediment slurries enriched for PCB dechlorinating bacteria combined with PCB 116 as a primer stimulated dechlorination of relatively high levels (50 ppm) of weathered Aroclor 1260 from Housatonic River sediment. However, the effectiveness of bioaugmentation for dechlorinating low levels of PCBs most commonly associated with weathered PCB contaminated sites where both kinetic and availability issues might affect the effectiveness of in situ PCB transformation is currently unknown.18,19 A recent report by May et al.20 showed that bioaugmentation with DF-1 stimulated the reductive dechlorination of weathered Aroclor 1260 (4.6 ppm) in contaminated soil microcosms, and Krumins,21 found that the addition of D. ethenogenes and pentachloronitrobenzene stimulated the dechlorination of weathered Aroclors 1248, 1254, and 1260 (2.1 ppm) in sediment microcosms. Combined, these results suggest that using bioaugmentation for low levels of weathered PCBs is feasible. Another unknown is whether dehalorespiring microorganisms used for in situ bioaugmentation can successfully compete with indigenous populations and be sufficiently sustainable to have a significant impact on reduction of higher chlorinated PCB congeners. In this report, we tested the efficacy of using bioaugmentation with the dehalorespiring bacterium D. chlorocoercia DF-1 to stimulate the reductive dechlorination of weathered Aroclor 1260 in sediment mesocosms that simulate in situ conditions. Specifically, we examined the effects of bioaugmentation on Aroclor transformation over the course of 120 days in the presence of activated carbon to assess any potential effects of bioavailability and monitored the microbial population to assess the sustainability of DF-1 within the background of the indigenous bacterial community.
’ MATERIALS AND METHODS Media and Growth Conditions. Dehalobium chlorocoercia DF1 was grown anaerobically in estuarine mineral medium (ECl) as described previously.20,22 Sodium formate (10 mM) was added as the electron donor and PCB 61 (2,3,4,5-PCB) was added in acetone (0.1% v/v) at a final concentration of 173 μM. Desulfovibrio sp.
ARTICLE
extract (1% v/v) was added as a growth factor and titanium(III) nitrilotriacetate (0.5 mM) was added as a chemical reductant.20,23 D. chlorocoercia was routinely grown in 50 mL of medium in 160-ml serum bottles sealed with 20-mm Teflon-coated butyl stoppers (West Pharmaceutical, Inc.). All cultures were incubated statically at 30 °C in the dark. Growth was monitored by gas chromatographic analysis of PCB 61 dechlorination to PCB 23 (2,3,5-PCB) and by quantitative RT-PCR of 16S rRNA gene copies (described below). Mesocosm Experiments. Mesocosms were prepared in glass 2.2 L TLC tanks (Fisher Scientific). PCB impacted sediments were sampled on 14 May 2009 from the Northwest Branch of Baltimore Harbor (BH) with a petite Ponar grab sampler at 39°16.8_N, 76°36.2_W and stored in the dark under nitrogen for 19 days at 4 °C prior to use. Sediment was pooled and homogenized anaerobically by stirring in an anaerobic glovebag. Two liters of sediment were added to each mesocosm tank with 2 cm of indigenous water above the sediment surface. A glass plate covered each mesocosm to minimize evaporation with a 1 cm gap on one end for air exchange. Water lost because of evaporation was periodically replenished with deionized water to maintain the osmolarity of the harbor water. The bioaugmentation inoculum was prepared using ten 50 mL cultures of DF1 grown until 50% of PCB 61 was dechlorinated. The cultures were transferred into 250 mL Oak Ridge bottles in an anaerobic glovebox and sealed under nitrogen carbon dioxide (4:1). The bottles were centrifuged at 22,000g for 30 min, decanted, and the pellets were pooled in 50 mL of sterile ECl medium. The concentration of pooled DF1 was approximately 5 107 16S rRNA gene copies per ml. Spent medium supernatant was prepared by passing DF-1 culture supernatant through a 0.22 μm filter (Millipore, www.millipore.com) to remove residual cells. Mesocosms were amended with one of four treatments in an anaerobic glovebox: (1) 20 mL of concentrated DF1 (5 105 cells gram 1 sediment); (2) 20 mL of spent cell-free growth medium; (3) 20 mL of concentrated DF1 adsorbed to 25 g of granulated activated carbon for 1 h; or (4) 20 mL of spent cell-free growth medium adsorbed to 25 g activated carbon (CAS# 7440-44-0, type TOG-NDS 80 325, Calgon Carbon Corp.) for 1 h. The total GAC added in an individual mesocosm (4.4% sediment dry weight) was similar to concentrations (3 5%) used in situ to sequester PCBs from benthic organisms.24 No exogenous electron donor was added. Mesocosms were homogenized after addition of the amendments by stirring with a Teflon spoon, then removed from the anaerobic glovebox and incubated at 23 °C in the dark. Mesocosms were sampled by taking six cm deep cores using a five ml syringe barrel with the end cut off. Triplicate cores were sampled for each time point using a random sampling grid and homogenized prior to analysis for PCBs and DNA as described below. DNA Extraction. DNA was extracted by adding 0.25 g of sediment from each sample core to a PowerBead microfuge tube of a Power Soil DNA Isolation Kit (MOBIO Laboratories, Inc.). The PowerBead tubes were mixed by hand prior to 30 s of bead beating at speed “4.5” using a FastPrep120 (Q-Biogene, 8 CA). Total DNA was then isolated from the PowerBead tubes according to the manufacturer’s directions. DNA was eluted in 100 μL of TE buffer and quantified with a NanoDrop 1000 Spectrophotometer (ThermoScientific). Extracted DNA samples had an A260/280 ratio of g1.6 and an A260/230 ratio of g2.0. All DNA samples were diluted to 2 ng/μL in TE buffer. Enumeration of PCB Dehalorespiring Bacteria by Quantitative PCR. The quantification of putative dechlorinating 8773
dx.doi.org/10.1021/es201553c |Environ. Sci. Technol. 2011, 45, 8772–8779
Environmental Science & Technology Chloroflexi in each subcore was performed by quantitative PCR (qPCR) using iQ SYBR green supermix (Bio-Rad Laboratories) and primers specific for the 16S rRNA gene of a deep branching, putative dechlorinating clade within the Chloroflexi (348F/884R).13 Each 25-μL reaction volume contained 1 iQ SYBR green supermix, 500 nM forward and reverse primers and 1 μL of sample DNA. PCR amplification and detection were conducted in an iCycler (Bio-Rad Laboratories). qPCR conditions were as follows: initial denaturation for 15 min at 95 °C followed by 35 cycles of denaturation for 30 s at 95 °C, annealing for 30 s at 61 °C, elongation for 30 s at 72 °C, with a final extension step of 5 min at 72 °C. One copy of the gene per genome was assumed based on the genomes of Dehalococcoides ethenogenes strain 195 and Dehalococcoides sp, strain CBDB1.25,26 qPCR data were analyzed with MJ Opticon Monitor Analysis Software v3.1 and compared to a standard curve of purified DF1 348F/884R 16S rRNA gene product. The standard curve consisted of duplicate dilutions over 6 orders of magnitude. The specificity of qPCR amplification was verified by melting curve analysis, followed by gel electrophoresis. Amplification efficiencies of dilutions of gel purified DF1 16S rRNA gene PCR product used as standards were 87% ( 7% (r2 = 0.99), and amplification efficiencies of DF1 cells and environmental samples were almost identical to those of the standard curve (86 ( 7%). The observed departure of amplification efficiencies from 100% was possibly due to the consensus nature of primer 884R, which has 1 mismatch with the DF-1 16S rRNA gene, combined with the stringency of the PCR reaction conditions. Community Analysis of PCB Dechlorinating Bacteria by Denaturing HPLC. Denaturing HPLC (DHPLC) analyses were performed using a WAVE 3500 HT system (Transgenomic, Inc.) as described previously,27 except that the instrument was equipped with a florescence detector (excitation 490 nm, emission 520 nm). The primer set 348F/884R was used for PCR amplification of 16S rRNA genes from bacteria within the Chloroflexi.13 DHPLC fractions were sequenced as described previously.27 A phylogenetic tree was created with Tree Builder (http://rdp.cme.msu.edu/index.jsp). PCB Extraction. Sediment samples were extracted using an Accelerated Solvent Extractor (Dionex), following EPA method 3545. Approximately 5 g wet weight of sediment was dried with pelletized diatomaceous earth (Dionex) at room temperature in a desiccator containing CaCl2 3 2H2O. The dried sediment (1 g) was transferred to an 11 mL stainless steel extraction cell containing 0.6 g of Cu and 2.4 g of fluorosil between two cellulose filters on the bottom of the cell and the remaining cell volume was filled with anhydrous Na2SO4. To correct for extraction efficiency, 10 μL of a 400 μg L 1 solution of PCB 166 in hexane was pipetted on top of the Na2SO4. The sample containing the surrogate was extracted with approximately 20 mL of pesticide grade hexane (Acros Organics) at 100 °C and purged with 13.8 MPa nitrogen. The sample was evaporated to a final volume of 1 mL at 30 °C under nitrogen using a N-EVAP 111 nitrogen evaporator (Organomation Associates). Before PCB analysis, 10 μL of PCB 30 and PCB 204 (400 μg L 1 each in acetone) was added to the sample as internal standards. PCB Analysis. PCB congeners were analyzed using a HewlettPackard 6890 series II gas chromatograph (GC) with a DB-1 capillary column (60 m by 0.25 mm by 0.25 μm; JW Scientific) and a 63Ni electron capture detector by a modified method of EPA 8082. PCB congeners in a mixture containing 250 μg L 1 Aroclor 1232, 180 μg L 1 Aroclor 1248, and 180 μg L 1 Aroclor
ARTICLE
1262 were quantified with a 10-point calibration curve using PCB 30 and PCB 204 as internal standards. Individual congeners and respective concentrations were obtained from Mullins et al.28 Fifty-five additional congeners not present in the Aroclor mixture that were potential dechlorination products were added to the calibration table containing the Aroclor congeners. The additional congeners were quantified with 10-point calibration curves at concentrations of 2, 5, 10, 20, and 40 μg L 1 (in duplicate) for the low range calibration and 40, 100, 200, 400, and 800 μg L 1 (in duplicate) for the high range calibration. Using this protocol 173 congeners were resolved in 130 individual peaks (excluding internal standards PCB 30 and PCB 204 and surrogate PCB166) (Supporting Information Table S1). The final concentration of individual congeners in samples was corrected for the recovery efficiency of the surrogate (82 ( 6%), which was reduced below 100% by the high content of carbon black in Baltimore Harbor sediments30 and the activated carbon amendment. Co-eluting peaks were indicated as multiple congeners. Total chlorines per biphenyl was calculated as the product of the average number of chlorines and molar concentration of each congener divided by the sum of the total molar concentration of all congeners.29 When coeluting peaks were different homologues, equal amounts of each homologue were assumed.29 All PCBs (99% purity) were purchased from AccuStandard.
’ RESULTS Characteristics of Mesoscosm Sediment. Sediment used in mesocosms was black in color with a slightly gelatinous texture. Relatively high amounts of PAHs, trans-nonachlor and heavy metals have been reported near this site as recently as 1997.31 Total PCB concentration in the pooled sediment samples was 1.3 ( 0.15 ppm with a mean of 4.76 chlorines per biphenyl. The congener profile was consistent with the first reported survey of PCBs in BH by Morgan and Sommer32 showing that Aroclor 1260 occurred predominantly with smaller amounts of Aroclor 1254 (Supporting Information Table S1). However, less chlorinated congeners were detected that are potential weathered products of Aroclor 1260 (Supporting Information Table S1). After bioaugmentation the sediments appeared to be anaerobic 1 2 cm below the surface based on black coloration and the distinct odor of sulfide. In mesocosms without GAC the sediment surface developed a gray coloration; with GAC the sediment surface developed a distinct orange-brown coloration. The surface of the mesocosm was disrupted within 1 week by holes (about 0.2 cm in diameter) caused by the activity of benthic worms but the activity was less apparent by day 60 (Supporting Information Figure S1). Effects of Treatments on Reductive Dechlorination of Weathered PCBs. BH sediment mesocosms were bioaugmented with dehalorespiring D. chlorocoercia DF-1 to a final concentration of approximately 5 105 cells g 1 sediment and PCBs were monitored over 120 days. In mesocosms bioaugmented directly with DF1, there was a net decrease in the total amounts of septaand octa-PCBs (Figure 1C). In mesocosms bioaugmented with DF-1 with GAC, which reduces the bioavaiability of PCBs,24 there was a net decrease in the total amounts of penta- through octa-PCBs and a net increase in the total amount of tetra- and triPCBs over the course of 120 days (Figure 1D). In contrast, there was no obvious change in homologue distribution in nonbioaugmented mesocoms treated with spent cell-free medium with or without GAC. Bioaugmentation both directly and with GAC as 8774
dx.doi.org/10.1021/es201553c |Environ. Sci. Technol. 2011, 45, 8772–8779
Environmental Science & Technology
ARTICLE
Figure 1. PCB analysis by homologue at day 0 (white bars) and day 120 (black bars) after treatment with filter sterilized spent growth medium (A), sterilized spent growth medium and GAC (B), concentrated DF1 in growth medium inoculated directly into the sediment (C) and concentrated DF1 adsorbed onto GAC (D). Each bar represents the mean and standard deviation of three replicates samples.
Figure 2. Changes in Cl per biphenyl in mesocosms over time after treatment with sterilized spent growth medium (O), sterilized spent growth medium and GAC (0),DF1 inoculated directly into sediment (b), and DF1 adsorbed onto GAC inoculated into sediment (9). Each datum point represents the mean and standard deviation of three replicates samples.
an adsorption substrate resulted in 0.6 0.7 mol of Cl per biphenyl dechlorinated, respectively, after 120 days (Figure 2). A greater rate of dechlorination (0.0067 Cl/biphenyl/day) was observed after bioaugmentation with GAC compared with direct injection (0.0041 Cl/biphenyl/day). A t-test (two-sample assuming unequal variances, α = 0.05) showed a significant difference in dechlorination rates between the bioaugmentation treatments with a P-value of 0.041 (df = 4). Single congener analysis of the mesocosm bioaugmented with DF1 adsorbed to GAC resulted in a significant decrease in higher chlorinated congeners and corresponding increase in lesser chlorinated congeners (Figure 3 and Supporting Information Table S1). The total sum of predicted substrates and products (Figure 3) was 1.26 ( 0.307 10 9 mol per gram sediment at day 0 and 1.02 ( 0.639 10 9 mol g 1 sediment at day 120. Predicted substrate congeners decreased from 1.00 ( 0.228 to 0.291 ( 0.282 10 9 mol per gram sediment and product congeners increased from 0.256 ( 0.0786 to 0.729 ( 0.356 10 9 mol per gram sediment. The inability to obtain a zero mass balance might
have resulted from the absence of possible dechlorination products in the GC method (e.g., PCBs 187, 109, 92, 96, and 84 among others, Supporting Information Tables S1 and S2), volatilization of less chlorinated congeners or aerobic degradation by indigenous microorganisms. Strikingly, PCB 194 decreased from about 23 to 8 ppb and PCB 133, the predicted product of sequential double flanked para reductive dechlorination of PCB 194 (via PCB 172) was found to increase from about 1 to 12 ppb over 120 days (Figure 3 and Supporting Information Table S2). Potential dechlorination products of double flanked reductive dechlorination of PCB 208 (PCB 179 or PCB 136), PCB 195 (PCB 187), PCB 190 (PCB 163), PCB 180 (PCB 153 or PCB 141), PCB 174 (PCB135 or PCB 149) and PCB 170 (PCB 87) were detected after 120 days; however, these products did not accumulate to substantial amounts. Furthermore, the less chlorinated congeners that did accumulate were not products resulting from dechlorination of double flanked chlorines (PCB 57, PCB 53, PCB 52, PCB 51, PCB 49 and PCB 32, Supporting Information Table S2). Since DF-1 is reported to reductively dechlorinate only doubly flanked chlorines 20 the dechlorination products likely resulted from enhanced activity by the indigenous population of dehalorespiring bacteria. Addition of cell-free medium used to grow DF-1 did not stimulate PCB dechlorination, indicating that the enhanced activity by indigenous microorganisms did not result from “priming” by residual PCBs or biostimulation by the medium. Sustainability of D. chlorocoercia DF1 after Bioaugmentation. To determine whether DF-1 was sustainable in the presence of relatively low PCB concentrations and the indigenous microbial community, putative dehalorespiring microorganisms were enumerated during the experiment based on the number of 16S rRNA gene copies g 1 dry sediment. In both bioaugmented mesocosms the initial number of putative dehalorespiring bacteria was 2-fold higher than in untreated mesocosms (1.3 106 compared to about 6.0 105 copies per gram, respectively) indicating that added DF-1 accounted for approximately half the total population of putative dehalorespiring bacteria 8775
dx.doi.org/10.1021/es201553c |Environ. Sci. Technol. 2011, 45, 8772–8779
Environmental Science & Technology
ARTICLE
Figure 3. Congeners showing greatest change in mesocosms receiving both GAC and DF1 at day 0 (0) and day 120 (9). An asterisk (*) indicates congeners that could be dechlorination substrates of DF1, or products from double flanked dechlorination by DF-1. Each bar represents the mean and standard deviation of three replicates samples.
Figure 4. Enumeration of putative dechlorinating Chloroflexi normalized to 16S rRNA gene copies/gram sediment at day 0 (open), day 60 (striped), and day 120 (solid). Treatments included spent medium only, spent medium and GAC, amendment with DF-1 by direct injection, and amendment with DF-1 adsorbed to GAC. Each bar represents the mean and standard deviation of three replicates samples.
in those mesocosms. A Student’s t-Test (α = 0.05) showed a significant difference between initial 16S rRNA gene copy numbers between treatments with a two-tail P-value of 0.01 (df = 3). The total 16S rRNA copy numbers in treated mesocosms decreased by 60 days before reaching an apparent steady state for the 120 day incubation period (Figure 4). However, for bioaugmentation treatments by both direct injection and on GAC substrate, the total number of putative dehalorespiring microorganisms remained nearly 2-fold higher in the bioaugmented mesocosms compared with the untreated mesocosms (8.0 105 compared to about 4.5 105 copies per gram at day 120, respectively). Detection of putative dechlorinating bacteria in untreated inactive mesocosms suggests that only a small proportion of the indigenous bacteria within this clade of the Chloroflexi were capable of dechlorinating PCBs. The results indicate that greater numbers of dehalorespiring microorganisms were maintained in bioaugmented mesocosms during active dechlorination of the weathered Aroclor compared with nonbioaugmented controls. Seven predominant phylotypes were detected by DHPLC in the BH sediment mesocosms. The community phylotypes were generally similar between mesocosms over 120 days, with the exception of DF1, which was only detected in bioaugmented mesocosms (Figure 5). The putative DF1 fraction was collected, sequenced, and found to be 100% identical to DF1 (Supporting Information Figure S2 and Table S3). One phylotype present at time zero (BH 4) was 100% identical to phylotype DEH10, a
Figure 5. DHPLC community analysis of putative dechlorinating Chloroflexi 16S rRNA genes in mesocosms at day 0 (bottom trace) and day 120 (top trace): A, addition of spent growth media alone; B, addition of spent growth media and GAC; C, addition of DF1; D, addition of DF1and GAC. Peak eluting at about 6.3 min (labeled 8/DF1 with arrow) in C and D confirmed DF1 16S rRNA gene by sequencing. Other phylotypes of putative dechlorinating Chloroflexi labeled 1, 2, 3, 4, 5, 6, or 7 are described in Supporting Information Table S2.
PCB dechlorinating bacterium previously detected BH sediment microcosms,13,15 but no other previously reported BH phylotypes were detected.
’ DISCUSSION There have been several reports on aerobic bioremediation by microbes and plants, but these approaches are only effective for treatment of less chlorinated congeners.7 Although highly chlorinated congeners commonly associated with commercial Aroclors such as Aroclor 1254, 1258, 1260, and 1262 can be reductively dechlorinated to less chlorinated congeners that serve as substrates 8776
dx.doi.org/10.1021/es201553c |Environ. Sci. Technol. 2011, 45, 8772–8779
Environmental Science & Technology for complete degradation by aerobic bacteria, concerns over several factors have called into question whether bioaugmentation by dehalorespiring bacteria is an effective strategy for in situ treatment of PCB impacted sediments. Bioavailability. Previous studies on the kinetics of reductive dechlorination in sediment microcosms spiked with Aroclor 124218 or 124819 reported that activity was limited at concentrations below 37 40 ppm. However, Magar et al.33 reported evidence of reductive dechlorination at PCB concentrations as low as 1 2 ppm in Lake Hartwell over a period of 11 years, which suggested that dehalorespiring bacteria slowly dechlorinated low levels of weathered Arcolor over time. In the current report we show unequivocally that bioaugmentation with the dehalorespiring bacterium DF-1 successfully stimulates the reductive dechlorination of sediment containing 1.3 ppm of weathered Aroclor 1260. Furthermore, addition of GAC, which increases the partition coefficient of PCBs between the sediment matrix and aqueous phase, had no inhibitory effect on dechlorination activity in bioaugmented mesocosms. GAC has been reported to reduce the bioavailability of PCBs to macroorganisms that ingest the particles;34 however, the slight stimulatory effect of GAC observed in the bioaugmented mesocosm suggests that PCB dehalorespiring bacteria have mechanisms that enable them to utilize weathered PCBs sequestered in the inorganic carbon fraction. The true threshold of bioavailability for PCB dechlorinating bacteria is not currently known, however, the results clearly demonstrate that bioaugmentation can stimulate reductive dechlorination of low levels of weathered PCBs in the 1 2 ppm range even with activated carbon present to sequester PCBs from any rapid desorption pool of sediment-associated PCB. Although the less chlorinated PCBs generated by bioaugmentation are more bioavailable to benthic organisms, they are generally less toxic because they are noncoplanar or have at least two adjacent unsubstituted carbon atoms that would make them subject to rapid metabolic degradation.35 Sustainability. One factor that is critical for successful bioaugmentation of weathered PCB-impacted sediments is the ability of the inoculated biocatalyst to compete with the indigenous community of nondehalorespiring bacteria for electron donors and nutrients. In the current study DF-1 was detected in bioaugmented sediment mesocosms after 120 days indicating that non-native DF1 could coexist with the indigenous sediment community. The total numbers of putative dehalorespiring bacteria decreased by approximately half after 90 days to a steady state of (7 8) 105 cells per gram of sediment, but community analysis showed that DF-1 was sustained as a predominant member of the putative dehalorespiring community. Although the total numbers of putative dehalorespiring phylotypes was 1 2 orders of magnitude lower than observed in prior reports of Aroclor 1260 dechlorinating microcosms11,15 neither exogenous electron donors nor electron acceptors were added in our study; therefore the lower steady state numbers likely reflect lowers concentrations of indigenous electron donors or acceptors available in the sediment. The observation that dehalorespiring bacteria were sustained in the sediment mesocosms without an exogenous electron donor is consistent with the ability of dehalorespiring bacteria to outcompete hydrogenotrophic sulfate reducers, acetogens, and methanogens in the presence of limited hydrogen concentration.36,37 Interestingly, the indigenous phylotypes of putative dehalorespiring bacteria, with the exception of phylotype DEH10 had not been reported previously in Baltimore Harbor sediment Aroclor 1260 enrichment microcosms. 15
ARTICLE
One possible explanation is that a different population of dehalorespiring bacteria was enriched with high concentrations of spiked Aroclor 1260 in the prior studies and the population in the current study is more relevant at in situ concentrations. The results show that DF-1 was able to coexist successfully with the indigenous microbial population even with the lower background concentrations of electron donor. Another significant observation was the detection of PCB congeners that did not result from dechlorination of doubly flanked chlorines. These dechlorination reactions were not observed in control mesocosms with spent cell-free medium indicating that the activity observed was neither due to a “priming” effect from residual PCB 61 or 23 in the inoculum nor a nutrient effect from carryover of formate or cell byproducts in the medium. This phenomenon has been reported previously. May et al. 20 observed enhanced dechlorination of congeners without flanked chorines in Aroclor 1260 impacted soil microcosms inoculated with DF-1. Krumins et al. 21 observed in PCB-impacted Anacostia River sediments bioaugmented with D. ethenogenes strain 195 stimulated PCB dechlorination and although this species was not sustained enhanced activity continued, which was attributed to the indigenous dehalorespiring community. One possible explanation for this effect is “priming” by the accumulation of dechlorination products from the initial activity of high numbers of DF-1 inoculated into the sediments. Addition of PCB congeners and analogs is known to stimulate the reductive dechlorination of PCBs in lab studies 21,38,39 and in field tests.40 Overall the results indicate that in addition to initiating and directly dechlorinating weathered PCBs, bioaugmentation with DF-1 had a synergistic effect on the indigenous dehalorespiring community by an as yet unknown mechanism that contributed further to the dechlorination process. Bioaugmentation. There have been recent attempts to test the effects of bioaugmentation with pure cultures of dehalorespiring bacteria. Krumins et al.21 reported enhanced reductive dechlorination of weathered PCBs (∼2 ppm) in sediment microcosms after bioaugmentation with D. ethenogenes strain 195 and May et al.20 reported enhanced dechlorination of Aroclor impacted soil (4.6 ppm) after bioaugmentation with DF-1. In contrast to the prior studies where bioaugmentation stimulated the reductive dechlorination of PCB by 0.2 Cl/ biphenyl after 415 days and 0.35 Cl/biphenyl after 145 days, respectively, bioaugmentation in the current study stimulated the reductive dechlorination by 0.7 Cl/biphenyl after only 120 days. Furthermore, bioaugmentation results in the current study stimulated 56% by mass reduction of penta- through nonachlorobiphenyls to lesser-chlorinated congeners (primarily tetrachlorobiphenyls), which are susceptible to aerobic degradation, with no detectable activity in untreated controls. The discrepancies in the rates and extent of dechlorination could possibly occur due to a number of factors including available nutrients, presence of inhibitory cocontaminants, the dehalorespiring strain used and the growth state and numbers of cells used for bioaugmentation. Distribution of cells in the current study was effective either by direct injection or on GAC particles. The ability to use a solid substrate such as GAC for inoculation of cells offers a possible solution for dispersing cells in the field. The results of this study support the potential feasibility of using bioaugmentation for treatment of PCB impacted sediments. We showed that bioavailability does not prevent bioaugmentation from treating low levels of weathered PCBs in sediment mesocosms and that GAC actually enhanced the overall process. 8777
dx.doi.org/10.1021/es201553c |Environ. Sci. Technol. 2011, 45, 8772–8779
Environmental Science & Technology Furthermore, DF-1, a nonindigenous species, was sustained throughout the dechlorination process and had a positive synergistic effect on the indigenous dehalorespiring population that contributed to the process. Although DF-1 was used successfully to bioaugment reductive dechlorination of weathered Aroclor 1260 in Baltimore Harbor sediments, discrepancies between the current and prior studies highlight the importance of testing inoculum with sediments from each site. All PCB bioaugmentation studies including the current study have been conducted on a laboratory scale and field studies will be required ultimately to validate the approach for in situ treatment of contaminated sites. However, the overall results of this mesocosm-based study provide compelling evidence to support further testing and development of bioaugmentation with dehalorespiring bacteria as an environmentally less invasive and lower cost alternative for in situ treatment of PCB impacted sediments.
’ ASSOCIATED CONTENT
bS
Supporting Information. Image of laboratory mesocosms, phylogenetic tree of Chloroflexi identified in mesocosms, congeners detected in mesocosms, inferred pathways of PCB dechlorination in bioaugmented mesocosm, and identification of phylotypes in mesocosms. This information is available free of charge via the Internet at http://pubs.acs.org/.
’ AUTHOR INFORMATION Corresponding Author
*Postal address: Department of Marine Biotechnology, Institute of Marine and Environmental Technology, 701 E. Pratt St., Baltimore, MD 21202. Telephone: (410) 234-8878. Fax: (410) 234-8896. E-mail:
[email protected].
’ ACKNOWLEDGMENT This work was supported by the National Institute of Environmental Health Science Superfund Research Program (5R01ES016197-02) and U.S. Department of Defense, Strategic Environmental Research, and Development Program (ER-1502). ’ REFERENCES (1) Alder, A. C.; H€aggblom, M. M.; Oppenheimer, S. R.; Young, L. Y. Reductive dechlorination of polychlorinated biphenyls in anaerobic sediments. Environ. Sci. Technol. 1993, 27, 530–538. (2) Quensen, J. F., III; Boyd, S. A.; Tiedje, J. M. Dechlorination of four commercial polychlorinated biphenyl mixtures (Aroclors) by anaerobic microorganisms from sediments. Appl. Environ. Microbiol. 1990, 56, 2360–2369. (3) Versar, I. PCBs in the United States: Industrial Use and Environmental Distribution, EPA Report 560/6-76-005; U.S. Environmental Protection Agency: Washington, DC, 1976. (4) Crisp, T. M.; Clegg, E. D.; Cooper, R. L.; Wood, W. P.; Anderson, D. G.; Baetcke, K. P.; Hoffmann, J. L.; Morrow, M. S.; Rodier, D. J.; Schaeffer, J. E.; Touart, L. W.; Zeeman, M. G.; Patel, Y. M. Environmental endocrine disruption: an effects assessment and analysis. Environ. Health Perspect. 1998, 106, 11–56. (5) Safe, S. Toxicology, structure function relationship, and human and environmental health impacts of polychlorinated biphenyls: progress and problems. Environ. Health Perspect. 1993, 100, 259–68. (6) EPA Contaminated Sediment Remediation Guidance for Hazardous Waste Sites, EPA-540-R-05-012. Final Report; U.S. Environmental Protection Agency: Washington, DC, 2005.
ARTICLE
(7) Abraham, W.-R.; Nogales, B.; Golyshin, P. N.; Pieper, D. H.; Timmis, K. N. Polychlorinated biphenyl-degrading microbial communities in soils and sediments. Curr. Opin. Microbiol. 2002, 5, 246–253. (8) Wu, Q.; Wiegel, J. Two anaerobic polychlorinated biphenyldehalogenating enrichments that exhibit different para-dechlorination specificities. Appl. Environ. Microbiol. 1997, 63 (12), 4826–32. (9) Bedard, D. L.; VanDort, H. M.; May, R. J.; Smullen, L. A. Enrichment of microorganisms that sequentially meta, para-dechlorinate the residue of Aroclor 1260 in Housatonic River sediment. Environ. Sci. Technol. 1997, 31 (11), 3308–3313. (10) Natarajan, M. R.; Nye, J.; Wu, W.-M.; Wang, H.; Jain, M. K. Reductive dechlorination of PCB contaminated Raisin River sediments by anaerobic microbial granules. Biotechnol. Bioeng. 1997, 55, 181–190. (11) Bedard, D. L.; Ritalahti, K. M.; Loffler, F. E. The Dehalococcoides population in sediment-free mixed cultures metabolically dechlorinates the commercial polychlorinated biphenyl mixture Aroclor 1260. Appl. Environ. Microbiol. 2007, 73 (8), 2513–21. (12) Cutter, L. A.; Watts, J. E. M.; Sowers, K. R.; May, H. D. Identification of a microorganism that links its growth to the reductive dechlorination of 2,3,5,6-chlorobiphenyl. Environ. Microbiol. 2001, 3 (11), 699–709. (13) Fagervold, S. K.; Watts, J. E. M.; May, H. D.; Sowers, K. R. Sequential reductive dechlorination of meta-chlorinated polychlorinated biphenyl congeners in sediment microcosms by two different Chloroflexi phylotypes. Appl. Environ. Microbiol. 2005, 71 (12), 8085–8090. (14) Wu, Q.; Watts, J. E. M.; Sowers, K. R.; May, H. D. Identification of a bacterium that specifically catalyzes the reductive dechlorination of polychlorinated biphenyls with doubly flanked chlorines. Appl. Environ. Microbiol. 2002, 68, 807–812. (15) Fagervold, S. K.; May, H. D.; Sowers, K. R. Microbial reductive dechlorination of Aroclor 1260 in Baltimore Harbor sediment microcosms is catalyzed by three phylotypes within the phylum Chloroflexi. Appl. Environ. Microbiol. 2007, 73 (9), 3009–18. (16) Adrian, L.; Dudkova, V.; Demnerova, K.; Bedard, D. L. Dehalococcoides sp. strain CBDB1 extensively dechlorinates the commercial polychlorinated biphenyl mixture Aroclor 1260. Appl. Environ. Microbiol. 2009, 75 (13), 4516–4524. (17) Fagervold, S. K.; Watts, J. E. M.; May, H. D.; Sowers, K. R. Effects of bioaugmentation on indigenous PCB dechlorinating activity in sediment microcosms. Wat. Res. 2011, 45, 3899–3907. (18) Fish, K. M. Influence of Aroclor 1242 concentration on polychlorinated biphenyl biotransformations in Hudson River test tube microcosms. Appl. Environ. Microbiol. 1996, 62 (8), 3014–3016. (19) Rhee, G. Y.; Sokol, R. C.; Bethoney, C. M.; Cho, Y. C.; Frohnhoefer, R. C.; Erkkila, T. Kinetics of polychlorinated biphenyl dechlorination and growth of dechlorinating microorganisms. Environ. Toxicol. Chem. 2001, 20, 721–726. (20) May, H. D.; Miller, G. S.; Kjellerup, B. V.; Sowers, K. R. Dehalorespiration with polychlorinated biphenyls by an anaerobic ultramicrobacterium. Appl. Environ. Microbiol. 2008, 74 (7), 2089–2094. (21) Krumins, V.; Park, J. W.; Son, E. K.; Rodenburg, L. A.; Kerkhof, L. J.; Haggblom, M. M.; Fennell, D. E. PCB dechlorination enhancement in Anacostia River sediment microcosms. Wat. Res. 2009, 43 (18), 4549–4558. (22) Berkaw, M.; Sowers, K. R.; May, H. D. Anaerobic ortho dechlorination of polychlorinated biphenyls by estuarine sediments from Baltimore Harbor. Appl. Environ. Microbiol. 1996, 62 (7), 2534–2539. (23) Moench, T. T.; Zeikus, J. G. An improved preparation method for a titanium(III) media reductant. J. Microbiol. Methods 1983, 1, 199–202. (24) Ghosh, U.; Luthy, R. G.; Cornelissen, G.; Werner, D.; Manzie, C. A. In-situ sorbent amendments: A new direction in contaminated sediment management. Environ. Sci. Technol. 2011, 45 (4), 1163–1168. (25) Seshadri, R.; Adrian, L.; Fouts, D. E.; Eisen, J. A.; Phillippy, A. M.; Methe, B.; Ward, N. L.; Nelson, W. C.; Deboy, R. T.; Khoudry, H. M.; Kolonay, J. F.; Dodson, R. J.; Daugherty, S. C.; Brinkac, L. M.; Sullivan, S. A.; Madupu, R.; Nelson, K. E.; Kang, K. H.; Impraim, M.; Tran, K.; Robinson, J. M.; Forberger, H. A.; Fraser, C. M.; Zinder, S. H.; 8778
dx.doi.org/10.1021/es201553c |Environ. Sci. Technol. 2011, 45, 8772–8779
Environmental Science & Technology
ARTICLE
Heidelberg, J. F. Genome sequence of the PCE-dechlorinating bacterium Dehalococcoides etheneogenes. Science 2005, 307, 105–108. (26) Kube, M.; Beck, A.; Zinder, S. H.; Kuhl, H.; Reinhardt, R.; Adrian, L. Genome sequence of the chlorinated compound-respiring bacterium Dehalococcoides species strain CBDB1. Nat. Biotechnol. 2005, 23, 1269–73. (27) Kjellerup, B. V.; Sun, X.; Ghosh, U.; May, H. D.; Sowers, K. R. Site-specific microbial communities in three PCB-impacted sediments are associated with different in situ dechlorinating activities. Environ. Microbiol. 2008, 10, 1296–1309. (28) Mullin, M. D.; Pochini, C. M.; McCrindle, S.; Romkes, M.; Safe, S. H.; Safe, L. M. High resolution PCB analysis: synthesis and chromatographic properties of all 209 PCB congeners. Environ. Sci. Technol. 1984, 18, 468–476. (29) Cho, Y. C.; Kwon, O. S.; Sokol, R. C.; Bethoney, C. M.; Rhee, G. Y. Microbial PCB dechlorination in dredged sediments and the effect of moisture. Chemosphere 2001, 43 (8), 1119–26. (30) Grossman, A.; Ghosh, U. Measurement of activated carbon and other black carbons in sediments. Chemosphere 2009, 75 (4), 469–475. (31) Baker, J.; Mason, R.; Cornwell, J.; Ashley, J.; Halka, J.; Hill, J. Spatial Mapping of Sedimentary Contaminants in the Baltimore Harbor/ Patapsco River/Back River System. UMCES CBL Reference Series, 97-142; University of Maryland Center for Environmental Science, Chesapeake Biological Laboratory, Solomons, MD: August, 1997. (32) Morgan, W. P.; Sommer, S. E. Polychlorinated biphenyls in Baltimore Harbor sediments. Bull. Environ. Contam. Toxicol. 1979, 22, 413–419. (33) Magar, V. S.; Brenner, R. C.; Johnson, G. W.; Quensen, J. F., 3rd Long-term recovery of PCB-contaminated sediments at the Lake Hartwell superfund site: PCB dechlorination. 2. Rates and extent. Environ. Sci. Technol. 2005, 39 (10), 3548–54. (34) Sun, X.; Ghosh, U. The effect of activated carbon on partitioning, desorption, and biouptake of native PCBs in four freshwater sediments. Environ. Toxicol. Chem. 2008, 27, 2287–2295. (35) Safe, S., Polyhalogenated aromatics: uptake, disposition and metabolism. In Halogenated Biphenyls, Naphthalenes, Dibenzodioxins and Related Products, 2nd ed.; Kimbrough, R. D., Jensen, A. A., Eds.; ElsevierNorth Holland: Amsterdam, 1989; pp 51 69. (36) Fennell, D. E.; Gossett, J. M. Modeling the production of and competition for hydrogen in a dechlorinating culture. Environ. Sci. Technol. 1998, 32 (16), 2450–2460. (37) Loffler, F. E.; Tiedje, J. M.; Sanford, R. A. Fraction of electrons consumed in electron acceptor reduction and hydrogen thresholds as indicators of halorespiratory physiology. Appl. Environ. Microbiol. 1999, 65 (9), 4049–4056. (38) Bedard, D. L.; VanDort, H.; Deweerd, K. A. Brominated biphenyls prime extensive microbial reductive dehalogenation of Aroclor 1260 in Housatonic River sediment. Appl. Environ. Microbiol. 1998, 64 (5), 1786–1795. (39) Deweerd, K. A.; Bedard, D. L. Use of halogenated benzoates and other halogenated aromatic compounds to stimulate the microbial dechlorination of PCBs. Environ. Sci. Technol. 1999, 33 (12), 2057–2063. (40) Bedard, D. L.; Quensen, J. F. Microbial reductive dechlorination of polychlorinated biphenyls. In Microbial Transformation and Degradation of Toxic Organic Chemicals; Young, L. Y., Cerniglia, C. E., Eds.; A. John Wiley: New York, 1995; pp 127 216.
8779
dx.doi.org/10.1021/es201553c |Environ. Sci. Technol. 2011, 45, 8772–8779
ARTICLE pubs.acs.org/est
Transverse Mixing Enhancement due to Bacterial Random Motility in Porous Microfluidic Devices Rajveer Singh and Mira Stone Olson* Department of Civil, Architectural, and Environmental Engineering, Drexel University, 3141 Chestnut Street, Philadelphia, Pennsylvania 19104, United States
bS Supporting Information ABSTRACT: Bacterial swimming in groundwater may create flow disturbances in the surrounding microenvironment thereby enhancing contaminant mixing. Porous microfluidic devices (MFDs) were fabricated in three different pore geometry designs: uniform grain size with large pore throats (MFD-I), nonuniform grain size with restricted pore space (MFD-II), and uniform grain size with small pore throats (MFD-III). Escherichia coli HCB33 was used to assess the effect of bacterial random motility on transverse mixing of a tracer, fluorescent labeled dextran, under three experimental conditions in which motile bacteria, nonmotile bacteria, and plain buffer suspensions were flown through the MFDs at four different flow rates. Mixing was quantified in terms of the best-fit effective transverse dispersion coefficient ((Dcy)eff). A mixing enhancement index (MEI) was defined as the ratio of the (Dcy)eff of tracer in experiments with motile bacteria and without bacteria. Motile bacteria caused a maximum 56 fold increase in MEI in MFD-II, a nearly 4-fold increase in MFD-I, and very little observed change in MFD-III. The apparent transverse dispersivities (αapp) of MFD-II and MFD-I increased by 3 and 2.3 times, respectively, with no change in MFD-III. These observations indicate that both pore throat size and pore arrangement are critical factors for contaminant mixing in porous media.
1. INTRODUCTION The presence of soil grains introduces tortuosity in groundwater flow paths, resulting in velocity gradients that may enhance transverse contaminant mixing.1 For situations in which contaminant mixing is controlled by molecular diffusion, the transverse mixing zone width may be as small as on the order of millimeters or centimeters.2 Transverse contaminant mixing is critical at the fringes of contaminant plumes, which form sharp concentration gradients as a result of dilution with surrounding groundwater.29 Under steady-state flow conditions, transverse dispersion may enhance biodegradation of contaminant plumes by (i) replenishing depleted electron acceptors in the plume zone,7 (ii) diluting contaminant concentrations, thereby creating a potentially more favorable degradation environment,1013 and (iii) promoting activities such as chemotaxis through formation of chemical gradients at plume fringes. Many soil-inhabiting bacteria such as Escherichia coli and Pseudomonas putida are capable of swimming in aqueous solutions via flagella attached to their surfaces. A typical swimming mechanism of flagellated bacteria in aqueous media is characterized as a series of alternating run and tumbling motions, which are propelled by 68 rotating helical filaments (flagella) protruding from the cell surfaces.14 The rotation of flagella is governed by a flagellar motor at their base. Counterclockwise rotation of motors and hence flagella tends to form a coordinated bundle behind the cell body and pushes the cell to move forward r 2011 American Chemical Society
smoothly. When one or more motor changes its rotational direction, the flagella bundle unravels, and the cell tumbles chaotically and reorients itself to move in another direction.14,15 Flagellar motors have a unique feature in that they alternate direction between clockwise to counterclockwise in a random manner, and therefore, overall bacterial motion results in a 3D random walk in the surrounding fluid16 that resembles molecular diffusion of gases.14 This chaotic movement of bacteria creates a natural mechanism for mixing in their microsurroundings and may be exploited to engineer an enhancement of overall mixing,16 which has been demonstrated in several studies.1622 In the past decade, the novel idea of chemical mixing enhancement due to bacterial flagellar rotation has been explored in different applications. Mixing enhancement due to freely swimming E. coli in aqueous suspensions is reported as a result of bacterial random motility19 and chemotaxis,20 where effective diffusion coefficients are observed to increase linearly with increases in bacterial and attractant concentrations, respectively. Kim and Breuer16 reported a greater than 30-fold increase in the effective diffusion coefficient of a passive tracer as a result of the random rotation of the flagella of a Serratia marcescens carpet Received: May 19, 2011 Accepted: August 30, 2011 Revised: August 29, 2011 Published: August 30, 2011 8780
dx.doi.org/10.1021/es201706w | Environ. Sci. Technol. 2011, 45, 8780–8787
Environmental Science & Technology
ARTICLE
Figure 1. Schematic representing the bilayer microfluidic device (MFD) used for testing contaminant mixing enhancement due to bacterial motility in porous media. An isometric-view (top) showing the porous channel (between the glass and first PDMS layers) simulating a very thin layer of a twodimensional aquifer slice connected to an overlying channel (between two PDMS layers) simulating leakage from an underground storage tank. A vertical cross-sectional view through section AA' (bottom) shows the closed channel opening though which different injectates flow through the device. Cross sections representing sampling locations at 4.9, 9.8, 19.6, 29.2, and 36.75 mm downstream from the tracer inlet point into the porous channel are also shown (top). A representative zoomed-in view at a cross section is also shown on the right. Specific details of different devices are presented in Table 1.
attached to the surface of a microfluidic system. Al-Fandi et al.17 studied mixing due to engineered tethered E. coli cells and found that mixing enhancement was the result of cell body rotation around a single shortened flagellum attached to a surface. In other studies, diffusion enhancement of fluorescent tracer particles18 and micrometer-scale bead 22 was observed due to bacterial carpets and freely swimming bacteria, respectively. Mixing enhancement as a result of bacterial locomotion in fluids in some of these studies16,19,22 is attributed to superdiffusion, a nonconventional diffusion process that accounts for bacteriabacteria interactions in aqueous media and results in chemical mixing that is faster than would be predicted with a convential Fickain diffusion model where concentration gradient is considered as the only driving force. The central focus of this paper is on mixing enhancement in porous media due to bacterial motility, which to the best of our knowledge has not yet been explored. This study compares the apparent transverse dispersivity in microfluidic devices with varying pore geometries in chemical transport experiments with motile bacteria, nonmotile bacteria, and without bacteria. Polydimethylsiloxane (PDMS) microfluidic devices (MFDs) have been used in many scientific and industrial fluid manipulation applications and may be used to measure mixing and dispersion at the submillimeter scale.23 Numerous groundwater studies8,9,2427
Figure 2. Images showing zoomed-in pore geometry details of the three pore designs used in this study: a regular pattern of uniform grain size (MFDI), a regular pattern of two different grain sizes (MFD-II), and a regular pattern of uniform grain size with equivalent porosity of MFD-II (MFD-III). Bottom right side picture shows a macroscopic image taken at 63 magnification inside a pore body of MFD-II. The larger grain diameter in all devices (dark black circles in A, B, and C) is 0.300 mm, and smaller grain diameter in MFD-II (smaller black dots in B) is 0.100 mm. Pore throat spacing in MFD-I and MFD-II is 0.050 mm and in MFD-III is 0.032 mm.
have successfully used these devices for quantification of pore scale parameters covering various aspects of groundwater remediation 8781
dx.doi.org/10.1021/es201706w |Environ. Sci. Technol. 2011, 45, 8780–8787
Environmental Science & Technology
ARTICLE
Figure 3. Typical fluorescein isothiocyanate labeled dextran (FITCdextran) dispersion images at cross sections 9.8 and 29.4 mm downstream from the tracer inlet point in the porous channel at Darcy velocity 0.22 mm/s. The overlaid profiles in each image indicate the fluorescent intensity variation across the channel between two horizontal rows of simulated soil grains.
research such as colloidal24,25,28,29 and bacterial26,27,30 transport, contaminant mixing,9 and biomass growth3133 in porous media. In this study, we employ a novel two-dimensional bilayer porous microfluidic device (MFD) that simulates formation of a plume in an aquifer due to leakage of an underground storage tank for quantification of mixing enhancement due to bacterial motility in porous media. Mixing enhancement is quantified by comparing the effective dispersion coefficient of a model contaminant in experiments in the presence and absence of suspended motile bacteria. The effects of pore geometry and permeability of porous media on contaminant mixing are also analyzed by comparing the effective transverse diffusion coefficients and apparent transverse dispersivities in three microfluidic devices at four separate flow velocities.
2. MATERIALS AND METHODS 2.1. Design and Fabrication of Microfluidic Devices. The microfluidic device designed and fabricated for this study simulates an environmentally relevant scenario of groundwater contamination due to leakage from an underground storage tank (UST). A schematic of the device is shown in Figure 1, which shows the two PDMS layers, one on top of the other, bonded to a cover glass. The lower PDMS layer, with imprinted porous media, simulates a very thin layer (∼0.022 mm) of a two-dimensional aquifer and forms a confined porous channel with the cover glass. The nonporous channel on the top, formed between two PDMS layers, simulates continuous leakage of contaminant from an UST and is connected to the lower porous channel at its downstream end through an orifice at the center of the porous channel. Further design details of the device are shown in Figure 1. The MFD was fabricated using standard photolithography and soft lithography techniques,34,35 the details of which can be found in the Supporting Information. In summary, masks were created using AutoCAD2008 to simulate the homogeneous two-dimensional porous media (see Figure 2 and pore geometries section for details) and printed on a high resolution transparency that served as a mask for the photolithography process. The mask designs were transferred to a 3 in. silicon wafer coated with a thin layer (∼0.022 mm) of negative photoresist SU-8 by exposing it to UV light to form the final mold that was subsequently used in the soft lithography process. A 9:1 mixture of 184-Sylgaurd elastomer base and curing agent was mixed and poured onto the mold and heat cured. The solidified PDMS layer was peeled off from the mold and separated into the porous and nonporous channels. The porous channel was bonded with a cleaned glass microscopic slide, and the nonporous channel was bonded to the top of the porous channel to form the final device.
2.2. Pore Geometries. Three homogeneous pore geometries representing three pore structures used in this study are shown in Figure 2. The first porous media design consists of a regular pattern of uniform dark circles of diameter 0.300 mm (representing soil grains) spaced 0.350 mm apart and therefore leaving a pore throat spacing of 0.050 mm between them (Figure 2A). A second design consists of a similar pattern of dark circles, along with one smaller circle of 0.100 mm right in the middle of the pore space resulting from the four larger circles (Figure 2B). These two designs in Figure 2A and B result in porosities of 42.3% and 35.9%, respectively. To study the effect of pore geometry and porosity on contaminant mixing enhancement, a third pore geometry design was selected with a pore grain design similar to the first design and equivalent porosity to the second pore design. This porous media design resulted in pore throat spacing of 0.032 mm (Figure 3C). For future discussion, the microfluidic devices fabricated using the first, second, and third designs will be referred to as MFD-I, MFD-II, and MFD-III, respectively. Rough surfaces of the simulated soil grains are evident from Figure 2D and offer a better representation of soil grains. The KozenyCarman equation36 was employed to estimate the permeability (k) of the devices, as reported in Table 1:
k¼
Ø2s D2p
ε3 150 ð1 εÞ2
ð1Þ
where Øs is sphericity of the grains (Øs = 1 for this study), Dp is grain diameter, and ε is porosity. 2.3. Cell Culture Preparation and Chemicals. Escherichia coli HCB33 (wild type strain) cells were obtained from Dr. Min Jun Kim, Drexel University. Cells from a 80 °C frozen glycerol stock were grown on Luria broth (10 g of tryptone, 5 g of yeast extract, 10 g of NaCl per liter of DI water) in a shaking incubator at 200 rpm (VWR incubating orbital shaker) at 33 °C for 4 h. Cells were harvested in their exponential growth phase at an optical density at 590 nm (OD590) of ∼0.81.0 (Spectronic 20 GENESYS spectrophotometer). Cells were washed with 10% random motility buffer (RMB) (11.2 g of K2HPO4, 4.8 g of KH2PO4, and 0.029 g of EDTA per liter of DI water; pH ∼ 7.0) by centrifugation and resuspended in buffer to a final concentration of OD590 = 0.60 (∼1.5 109 cells/mL). For experiments in which nonmotile bacteria were required, carbonylcyanide-p-trifluoromethoxyphenyl hydrazone (FCCP) was added (0.001% w/v) to 10% RMB to de-energize the flagellum rotary motor and render the cells nonmotile.19 A relatively high molecular weight diffusive tracer, fluorescein isothiocyanate labeled dextran (FITCdextran) (Molecular Probe, Invitrogen), average molecular weight 70 000, was used in this study for quantification of mixing enhancement. Kim and Breuer20 8782
dx.doi.org/10.1021/es201706w |Environ. Sci. Technol. 2011, 45, 8780–8787
Environmental Science & Technology
ARTICLE
Table 1. Summary of Apparent Dispersivity (αapp) Values for Three Devices under Different Experimental Conditions Uniform grain size media
Nonuniform grain size media
Uniform grain size low porosity media
(MFD-I), ε = 42.3%, k = 1.36 104 mm2,
(MFD-II), ε = 35.9%, k = 0.44 104a mm2,
(MFD-III), ε = 35.9%, k = 0.68 104 mm2,
(D0)eff = 4.86 106 mm2/s
(D0)eff = 4.13 106 mm2/s
(D0)eff = 4.13 106 mm2/s
Motile bacteria
0.0048
0.0038
0.0023
No bacteria
0.0021
0.0013
0.0020
Nonmotile bacteria
0.0028
Experimental condition
a
Value is calculated based on the geometric mean grain size of the pore design shown in Figure 2.
recommend the use of higher molecular weight tracers for determination of effective diffusion coefficients to avoid a dramatic concentration deflection common with smaller tracer molecules. 2.4. Experimental Set Up. To determine the effect of bacterial random motility on transverse mixing in porous media, three types of contaminant transport experiments were performed by injecting three different injectates into the porous bottom channel (through port A, Figure 1) of the MFDs: (1) motile bacterial suspension in 10% RMB, (2) nonmotile bacterial suspension in 10% RMB, and (3) 10% RMB only. A low concentration (0.02% (w/v)) of FITCdextran solution in 10% RMB was simultaneously injected through the top channel (port B, Figure 1) in all three type of experiments. A CCD camera (AxioCam, Carl Zeiss) mounted on a stereoscope (Zeiss Stereo Discovery V8 equipped with GFP-470/525 filter, 3.2 magnification) was used to capture images at 80 ms exposure time at transverse cross sections 4.9, 9.8, 19.6, 29.4, and 36.75 mm downstream from the point of FITCdextran injection into porous channel (Figure 1) in all three experiments. 2.5. Microfluidic Device Operation. A multiple loading syringe pump (kd Scientific, Model 200 Series) was used for injecting two injectates, depending on the type of experiment, into the two ports of the MFD. The top channel was connected to a 1 mL syringe (BD, internal diameter 4.77 mm), and the bottom channel was connected to a 10 mL syringe (BD, internal diameter 14.50 mm), resulting in a volumetric flow rate ratio of 9.2:1 in the bottom/top channels. Polyethylene tubing (INTRAMEDIC Polyethylene Tubing, Clay Adams) was used to attach the punched inlets of the channels to their respective syringes via syringe needles (BD 21G1). The high volumetric flow rate ratio was chosen to mimic groundwater and contaminant volumetric flow ratios in actual groundwater aquifers. Experiments were performed at Darcy velocities (calculated by dividing volumetric flow rates by the total cross-sectional area of the flow channel) of 0.11, 0.22, 0.44, and 0.88 mm/s, a range chosen from similar groundwater studies in micromodels.9,27 Equilibrium flow conditions were ensured by allowing at least two pore volumes of fluid to flow through the device before taking measurements. 2.6. Mathematical Modeling. 2.6.1. Transverse Dispersion Coefficient. The two-dimensional advection-dispersion equation that governs transport of a tracer in porous media may be given by the following mathematical expression: Rc
∂C ∂ðCÞ ∂2 C ∂2 C ¼ vf þ ðDcx Þeff 2 þ ðDcy Þeff 2 ∂t ∂x ∂x ∂y
ð2Þ
where Rc is retardation factor, C is tracer concentration, t is time, x and y are longitudinal and transverse directions to flow
direction, vf is average linear flow velocity, and (Dcx)eff and (Dcy)eff are effective longitudinal and transverse dispersion coefficients, respectively. For a conservative tracer Rc = 1. At steady-state flow conditions, the longitudinal concentration gradient is small; therefore, diffusive mixing in the longitudinal direction is also small in comparison to the advective transport term and can be neglected.27 The tracer concentration was linearly related to fluorescent intensity, I, of the FITCdextran and could therefore be substituted into eq 2. Under these assumptions, eq 2 is reduced to the following: vf
∂ðIÞ ∂2 I ¼ ðDcy Þeff 2 ∂x ∂y
ð3Þ
The analytical solution to eq 3 was derived from 1-D solute diffusion from a confined region of width h at the tracer inlet into the porous channel and is given by the following:37 9 8 h h > > = < y þ y I 1 2 2 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi p p ð4Þ erf ¼ þ erf I0 2> 2 ðDcy Þeff x=vf > ; : 2 ðDcy Þeff x=vf where I0 is the peak fluorescent intensity. Experimentally obtained images were analyzed for the fluorescent intensity distribution at cross sections 4.9, 9.8, 19.6, 29.4, and 36.75 mm downstream from the FITCdextran inlet point into the porous bottom channel using an image processing program (Image J) developed by the National Institute of Health. For the sake of consistency, the intensity profiles obtained at each cross section were normalized with respect to the corresponding peak intensity value, as opposed to the peak intensity at the inlet. It was not possible to measure the maximum intensity (I0) near the tracer inlet (Figure 1) due to light interference from the tracer inlet channel. To obtain the effective transverse dispersion coefficients, normalized steady state transverse intensity profiles were fitted to the analytical solution given by eq 4. Due to symmetry of the device, the intensity profiles can be cut in half by a vertical plane at y = 0, and therefore, only half of the porous channel region (y g 0) was chosen for modeling the fluorescent intensity profiles. The tracer plume width, h, was measured immediately downstream (x = 1.05 mm) from the point of injection of the tracer into the porous bottom channel and was assumed to be constant at that cross section (x = 1.05 mm) due to continuous injection of FITCdextran (I[x e 1.05 mm, h/2 e y e h/2, t g tequib] = I0), where tequib is the time required to attain equilibrium flow conditions. The total width of the device (6.3 mm) was approximately three times the maximum plume width observed during the experiments, and therefore, channel boundaries were approximated as infinite distance, zero concentration boundary conditions. 8783
dx.doi.org/10.1021/es201706w |Environ. Sci. Technol. 2011, 45, 8780–8787
Environmental Science & Technology
ARTICLE
Figure 4. Markers represent FITCdextran normalized intensity profiles in MFD-I at a cross section 29.2 mm downstream from the inlet in experiments with motile bacteria (left plot), immobilized bacteria (middle plot), and no bacteria (right plot). Each marker represents pixel intensity averaged over 10 pixels, and the curves represent corresponding modeled profiles for the respective effective transverse dispersion coefficients obtained as the global average of triplicate experiments.
2.6.2. Apparent Transverse Dispersivity. The effective transverse dispersion coefficient in porous media is given by the following: ðDcy Þeff ¼ ðD0 Þeff þ αapp vf
ð5Þ
where (D0)eff is the effective diffusion coefficient and αapp is the apparent transverse dispersivity of the porous medium. Both (D0)eff and αapp are lumped parameters that account for the effect of the presence of bacteria and porous media in the system. The effective diffusion coefficient of the tracer can be given by the following:3840 ε ðD0 Þeff ¼ D0 ð6Þ τ where ε and τ are porosity and tortuosity of the porous medium and D0 is the molecular diffusion coefficient of the tracer. The porosity (ε) values for different porous media are reported in Table 1. The tortuosity term incorporates the effects of the available pore space and the transport mechanism of the tracers 39 and an assumed value of 2 will be used in this study based on values used in similar groundwater studies.2,27 The molecular diffusion coefficient of the tracer FITCdextran is D0 = 2.3 105 mm2/s.41
3. RESULTS 3.1. Comparison of FITCDextran Intensity Profiles. A fluorescent plume along the length of the device was apparent as FITCdextran entered and flowed through the porous bottom channel. At steady-state flow conditions, images of the plume were recorded at the predefined cross sections (Figure 1). Figure 3 shows representative images at cross sections 9.8 and 29.4 mm downstream from the FITCdextran inlet in MFD-I under different experimental conditions. Transverse dispersion of the plume is evident from the increase in the width of fluorescent intensity profiles from 9.8 mm (row 1, Figure 3) to 29.4 mm (row 2, Figure 3) in all three experimental conditions (columns 13). Comparison of concentration profiles under different experimental conditions (across different columns) shows wider profile widths in experiments with motile bacteria (column 1, Figure 3) than in experiments with no bacteria (column 2, Figure 3) at respective cross sections. To further verify that the effect was due to bacterial motility, control experiments were performed in which the motile bacteria were replaced by similar concentrations of nonmotile bacteria in a third set of experiments. The widths of the fluorescent intensity
Figure 5. Variation of effective transverse dispersion coefficient, (Dcv)eff, with Darcy velocity in MFD-I under three different experimental conditions. Error bars represent standard error values of nine best fit profiles at three cross sections of the device in triplicate experiments. The lines represent results of the dispersion model (eq 6) with the best-fit model parameter values given in Table 1.
profiles in experiments with nonmotile bacteria (column 3, Figure 3) are not significantly different from the corresponding profile widths in experiments with no bacteria (column 2, Figure 3). These observations indicate that the wider profile widths in experiments with motile bacteria are the result of the random motility of E. coli HBC33 in porous media. 3.2. Effective Transverse Dispersion Coefficients. Fluorescent intensity distributions were analyzed at cross sections 4.9, 9.8, 19.6, 29.4, and 36.75 mm downstream from the tracer inlet point in the porous bottom channel of the MFDs. The experimentally obtained representative normalized intensity profiles for MFD-I 29.4 mm downstream from the inlet under various experimental conditions are shown in Figure 4. The best-fit dispersion coefficient values at each cross section of the device were obtained by fitting the model (eq 4) to the corresponding experimentally obtained normalized intensity profiles using leastsquares regression analysis. Average transverse dispersion coefficients (ATDC) for each cross section were calculated by averaging the best-fit values of triplicate experiments. The average best-fit ATDC values for two upstream cross sections (4.9 and 9.8 mm) were abnormally high (due to reasons explained in the Supporting Information) and therefore were not included in the effective transverse dispersion coefficients, (Dcy)eff, calculations. (Dcy)eff for the device under different experimental conditions was calculated by averaging the ATDC values at cross sections 8784
dx.doi.org/10.1021/es201706w |Environ. Sci. Technol. 2011, 45, 8780–8787
Environmental Science & Technology 19.6, 29.4, and 36.75 mm downstream from the inlet, where these values were nearly constant (Figure S1 of the Supporting Information). The model-predicted (eq 4) profiles for the effective transverse dispersion coefficients for MFD-I at a cross section 29.4 mm downstream from the inlet under various experimental conditions are shown in Figure 4 along with the corresponding experimentally obtained profiles. The best-fit curves to the experimentally obtained effective transverse dispersion coefficient data for MFD-I are presented in Figure 5 under three experimental conditions. 3.3. Apparent Transverse Dispersivities. Effective diffusion coefficients for three porous media geometries were calculated (eq 6), and the values obtained are reported in Table 1. Apparent transverse dispersivity values under each experimental condition in each MFD were calculated using eq 5 given the effective diffusion coefficient values along with the best-fit effective transverse dispersion coefficients, (Dcy)eff, obtained in the previous section. Least-squares regression analysis was used to determine the best-fit values of the effective diffusion coefficient, (D0)eff, and the apparent dispersivity, αapp, for triplicate experiments. Apparent dispersivity values for MFD-I under three experimental conditions are reported in Table 1. Compared with experiments run with no bacteria, it can be observed that the presence of motile bacteria results in a 2.5-fold increase in the apparent transverse dispersivity of the device. The presence of nonmotile bacteria also resulted in a marginal (1.3 times) increase in the apparent transverse dispersivity of the device (Table 1). 3.4. Effect of Pore Geometry on Average and Effective Transverse Dispersion Coefficients. The ATDC values for MFD-II at Darcy velocities 0.11 and 0.22 mm/s at various cross sections are shown in Figure S1 of the Supporting Information. Similar to MFD-I, ATDC values for MFD-II at cross sections closer to the inlet (at 4.9 and 9.8 mm) are significantly higher than the values for the remaining three downstream cross sections, where ATDC values were nearly constant. Similar observations were made for MFD-III as well, and effective transverse dispersion coefficient values for MFD-II and MFDIII were calculated by averaging ATDC values at three downstream cross sections, as in the case of MFD-I. Apparent transverse dispersivities and effective diffusion coefficient values were also calculated for MFD-II and MFD-III as described for MFD-I, and the resulting values are presented in Table 1. A 3-fold increase in the apparent transverse dispersivity was observed in experiments with motile bacteria as opposed to no bacteria in MFD-II, while no such increase was observed in MFD-III.
4. DISCUSSION Hydrodynamic dispersion results in an increase in the variance of a tracer in porous media as it moves downstream from its injection point. This phenomenon is responsible for the spreading of fluorescent intensity profiles from cross sections 9.8 mm (column 1, Figure 3) to 29.4 mm (column 2, Figure 3) downstream of the inlet in all three experimental conditions. Wider profile widths were observed at each cross section in the presence of motile E. coli HCB33 (row 1, Figure 3) as compared to experiments with no bacteria (row 2, Figure 3). However, the profile widths at each cross section in the experiments with nonmotile bacteria (row 3, Figure 3) were similar to those in experiments with no bacteria. Best-fit effective dispersion coefficient values display a similar trend, with largest values for the
ARTICLE
Figure 6. Effect of flow velocity on mixing enhancement due to bacterial random motility in different MFDs. Error bars indicate standard errors.
experiments with motile bacteria and approximately similar values for experiments with nonmotile bacteria and no bacteria (Figure 6). These two observations indicate that random motility of E. coli HCB33 does enhance mixing of FITCdextran in the porous MFDs. The marginal differences in the effective dispersion coefficient values in experiments with nonmotile bacteria and no bacteria (Figure 5) may be attributed to the following factors. Even though the nonmotile bacteria may not have contributed in mixing enhancement due to their motility, frequent collision with the porous media structures may have resulted in their deflection across the flow lines and thereby resulted in flow disturbances. The other reasons could be the pore space occupied by the nonmotile bacteria (bacterial concentration used in this study accounts for approximately ∼0.36% of the total available pore space (calculated based on the values reported by Kim and Breuer19 in similar studies) or the non-Fickian (or anomalous) diffusion of the tracer due to bacterial swimming in porous media. In general, the best-fit models for the effective dispersion coefficients capture the shapes of experimentally obtained profiles well (Figure 4). However, the minor disparity between model predictions and experimentally obtained values at the lower tails of the curves (Figure 4) may be attributed to nonFickian diffusion of tracer due to bacterial motility42,43 or to experimental limitations, including nonideal inlet conditions and/or pump fluctuations. Another potential source of error between experimental data and modeled predictions may have been introduced when intensity profiles at downstream cross sections were normalized with respect to their local maximum intensity rather than the peak intensity at the inlet. Although the peak intensity values at all cross sections were analyzed and no significant change was observed among the various cross sections, it is still possible that the normalization process was a source of error. While dispersivity is generally considered an intrinsic property of the porous medium, different longitudinal44,45 and transverse27 dispersivity values have been reported for the same porous medium based on the dispersant used. In their colloidal transport study in micromodels, Auset and Keller24 have emphasized the dependence of dispersivity on the size of the dispersant in addition to properties of the porous medium. Results from this study also show three different values of apparent transverse dispersivity for similar porous media under three different experimental conditions (with bacteria, with nonmotile bacteria, and without bacteria) when the same 8785
dx.doi.org/10.1021/es201706w |Environ. Sci. Technol. 2011, 45, 8780–8787
Environmental Science & Technology
ARTICLE
dispersant (FITCdextran) was used. Therefore, we suggest that dispersivity may be considered as a system-specific parameter, incorporating effects of the porous medium and dispersant and characteristics of the transporting fluid. 4.1. Mixing Enhancement Index. A mixing enhancement index (MEI) was defined to quantify the extent of mixing enhancement and was quantified in terms of the ratio of the effective transverse dispersion coefficient values in experiments with and without motile bacteria: Mixing enhancement index ðMEIÞ ¼
ððDcy Þeff Þwith bacteria ððDcy Þeff Þno bacteria
ð7Þ
The effect of flow velocity on the MEI is shown in Figure 6. For MFD-I and MFD-II, MEI shows an increasing trend initially and finally decreases at the highest flow velocity used in this study. Mixing enhancement in the presence of motile bacteria in these devices may be attributed to the free swimming of bacteria across flow streamlines creating flow disturbances in their surrounding microenvironment. A common understanding may suggest that increasing Darcy velocity should adversely affect bacterial random motility and thereby the contaminant mixing index. This belief would be further strengthened by the fact that higher flow velocities will cause greater shear rates on motile bacteria in porous MFDs. However, Marcos46 showed in his study of the effect of shear flow on the motility of the marine bacterium Pseudomonas haloplanktis, a bacterium with similar shape, size, and swimming properties to E. coli HCB 33, that bacteria are capable of overcoming vortices of moderate strength. Experimental results of Marcos46 showed bacteria crossing flow streamlines under moderate shear flow conditions as well as aligning themselves across the flow stream lines when vortices were switched from low to moderate strength. Lanning et al.26 have also reported similar transverse movement of E. coli HCB1 across flow stream lines at Darcy velocities greater than those used in this study (∼1 mm/s) as a result of chemotactic migration. Thus, it can be argued that the flow velocity range of 0.110.44 mm/s used in this study may cause low to moderate shear flow conditions and thereby result in an increase in MEI (Figure 6). The decrease in the MEI values in MFD-I and MFDII at the highest Darcy velocity (0.88 mm/s) suggests that, at stronger shear flow rates, bacteria not only follow the flow streamline but also align their body axis with the flow lines.46 In contrast to the other two devices, the observed MEI in MFD-III remained approximately constant for all Darcy velocities tested. The reason for this difference may be attributed to the pore throat spacing of MFD-III, because all other features of this device are similar to MFD-I. Typical mean bacterial run lengths for E. coli HCB1, a bacteria similar to E. coli HCB33 used in this study, are ∼28 μm (calculated based on the values reported by Dong et al.44), which is comparable to the pore throat size in MFD-III (32 μm). However, the pore throat size in MFD-I (50 μm) is significantly larger and would cause less interfere with bacterial free swimming paths as compared to MFD-III. Therefore, we believe that the ratio of pore throat spacing to the bacterial run length may be a critical parameter for enhanced mixing, though this speculation should be verified with single cell experiments. Comparing across the three devices, MFD-II resulted in the highest MEIs for almost all flow velocities, which may be attributed to its specific pore geometry; the smaller
grain in the middle of the pore, which was not present in the other two pore geometries, may have acted as an additional agent of mixing in this device. Results from this study contribute to a better understanding of mass transfer mechanisms in porous media and provide a framework for incorporating mixing enhancement due to bacterial motility in bioremediation models. In most bioremediation studies at the laboratory and field scales, contaminant transport is predicted based on tracer tests as a surrogate for contaminants that do not involve bacteria.27 Such contaminant transport predictions may be erroneous as results from this study show that motile bacteria may provide significant contaminant mixing in porous media. Incorporating contaminant mixing due to bacterial motility in field scale bioremediation strategies such as monitored natural attenuation may be useful for more accurate predictions of remediation time frames. Further research may look beyond the effect of bacterial random motility on contaminant mixing in porous media to include other motility mechanisms such as chemotaxis, which is more likely to prevail and may dominate mixing enhancement in certain bioremediation scenarios.
’ ASSOCIATED CONTENT
bS
Supporting Information. Supporting Information is available detailing the microfabrication procedure used in this study and providing further explanation of the effective transverse dispersivity calculations. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (215) 895-2987; fax: (215) 895-1363; e-mail: mira.s.
[email protected].
’ ACKNOWLEDGMENT The authors gratefully acknowledge funding for this research, which was provided by the National Science Foundation through award EAR 0911429. We also thank three anonymous reviewers whose insightful comments helped to greatly improve this manuscript. ’ REFERENCES (1) Ajdari, A.; Bontoux, N.; Stone, H. A. Hydrodynamic dispersion in shallow microchannels: the effect of cross-sectional shape. Anal. Chem. 2005, 78 (2), 387–392. (2) Acharya, R. C.; Valocchi, A. J.; Werth, C. J.; Willingham, T. W. Pore-scale simulation of dispersion and reaction along a transverse mixing zone in two-dimensional porous media. Water Resour. Res. 2007, 43, (10). (3) Cirpka, O. A.; Frind, E. O.; Helmig, R. Numerical simulation of biodegradation controlled by transverse mixing. J. Contam. Hydrol. 1999, 40 (2), 159–182. (4) Cirpka, O. A.; Olsson, A.; Ju, Q. S.; Rahman, M. A.; Grathwohl, P. Determination of transverse dispersion coefficients from reactive plume lengths. Ground Water 2006, 44 (2), 212–221. (5) Huang, W. E.; Oswald, S. E.; Lerner, D. N.; Smith, C. C.; Zheng, C. M. Dissolved oxygen imaging in a porous medium to investigate biodegradation in a plume with limited electron acceptor supply. Environ. Sci. Technol. 2003, 37 (9), 1905–1911. 8786
dx.doi.org/10.1021/es201706w |Environ. Sci. Technol. 2011, 45, 8780–8787
Environmental Science & Technology (6) Knutson, C.; Valocchi, A.; Werth, C. Comparison of continuum and pore-scale models of nutrient biodegradation under transverse mixing conditions. Adv. Water Resour. 2007, 30 (67), 1421–1431. (7) Rahman, M. A.; Jose, S. C.; Nowak, W.; Cirpka, O. A. Experiments on vertical transverse mixing in a large-scale heterogeneous model aquifer. J. Contam. Hydrol. 2005, 80 (34), 130–148. (8) Werth, C. J.; Cirpka, O. A.; Grathwohl, P. Enhanced mixing and reaction through flow focusing in heterogeneous porous media. Water Resour. Res. 2006, 42, 12. (9) Willingham, T. W.; Werth, C. J.; Valocchi, A. J. Evaluation of the effects of porous media structure on mixing-controlled reactions using pore-scale modeling and micromodel experiments. Environ. Sci. Technol. 2008, 42 (9), 3185–3193. (10) Singh, R.; Olson, M. S. Kinetics of trichloroethylene and toluene toxicity to Pseudomonas putida F1. Environ. Toxicol. Chem. 2010, 29 (1), 56–63. (11) Abuhamed, T.; Bayraktar, E.; Mehmetoglu, T.; Mehmetoglu, U. Kinetics model for growth of Pseudomonas putida F1 during benzene, toluene and phenol biodegradation. Process Biochem. 2004, 39 (8), 983–988. (12) Kim, D. J.; Choi, J. W.; Choi, N. C.; Mahendran, B.; Lee, C. E. Modeling of growth kinetics for Pseudomonas spp. during benzene degradation. Appl. Microbiol. Biotechnol. 2005, 69 (4), 456–462. (13) Ray, S.; Peters, C. A. Changes in microbiological metabolism under chemical stress. Chemosphere 2008, 71 (3), 474–483. (14) Ford, R. M.; Harvey, R. W. Role of chemotaxis in the transport of bacteria through saturated porous media. Adv. Water Resour. 2007, 30 (67), 1608–1617. (15) Turner, L.; Ryu, W. S.; Berg, H. C. Real-time imaging of fluorescent flagellar filaments. J. Bacteriol. 2000, 182 (10), 2793–2801. (16) Kim, M. J.; Breuer, K. S. Use of bacterial carpets to enhance mixing in microfluidic systems. J. Fluids Eng. 2007, 129 (3), 319–324. (17) Al-Fandi, M.; Jaradat, M. A. K.; Fandi, K.; Beech, J. P.; Tegenfeldt, J. O.; Yih, T. C. Nano-engineered living bacterial motors for active microfluidic mixing. IET Nanobiotechnol. 2010, 4 (3), 61–71. (18) Darnton, N.; Turner, L.; Breuer, K.; Berg, H. C. Moving fluid with bacterial carpets. Biophys. J. 2004, 86 (3), 1863–1870. (19) Kim, M. J.; Breuer, K. S. Enhanced diffusion due to motile bacteria. Phys. Fluids 2004, 16 (9), L78–L81. (20) Kim, M. J.; Breuer, K. S. Controlled mixing in microfluidic systems using bacterial chemotaxis. Anal. Chem. 2007, 79 (3), 955–959. (21) Sokolov, A.; Goldstein, R. E.; Feldchtein, F. I.; Aranson, I. S. Enhanced mixing and spatial instability in concentrated bacterial suspensions. Phys. Rev. E 2009, 80, (3). (22) Wu, X. L.; Libchaber, A. Particle diffusion in a quasi-twodimensional bacterial bath. Phys. Rev. Lett. 2000, 84 (13), 3017–3020. (23) Stone, H. A.; Stroock, A. D.; Ajdari, A. Engineering flows in small devices: microfluidics toward a lab-on-a-chip. Annu. Rev. Fluid Mech. 2004, 36, 381–411. (24) Auset, M.; Keller, A. A. Pore-scale processes that control dispersion of colloids in saturated porous media. Water Resour. Res. 2004, 40, (3). (25) Auset, M.; Keller, A. A.; Brissaud, F.; Lazarova, V. Intermittent filtration of bacteria and colloids in porous media. Water Resour. Res. 2005, 41, (9). (26) Lanning, L. M.; Ford, R. M.; Long, T. Bacterial chemotaxis transverse to axial flow in a microfluidic channel. Biotechnol. Bioeng. 2008, 100 (4), 653–663. (27) Long, T.; Ford, R. M. Enhanced transverse migration of bacteria by chemotaxis in a porous T-sensor. Environ. Sci. Technol. 2009, 43 (5), 1546–1552. (28) Wan, J. M.; Wilson, J. L. Visualization of the role of the gaswater interface on the fate and transport of colloids in porous-media. Water Resour. Res. 1994, 30 (1), 11–23. (29) Smith, J.; Gao, B.; Funabashi, H.; Tran, T. N.; Luo, D.; Ahner, B. A.; Steenhuis, T. S.; Hay, A. G.; Walter, M. T. Pore-scale quantification of colloid transport in saturated porous media. Environ. Sci. Technol. 2008, 42 (2), 517–523.
ARTICLE
(30) Lanning, L. M.; Ford, R. M. Glass micromodel study of bacterial dispersion in spatially periodic porous networks. Biotechnol. Bioeng. 2002, 78 (5), 556–566. (31) Dupin, H. J.; McCarty, P. L. Impact of colony morphologies and disinfection on biological clogging in porous media. Environ. Sci. Technol. 2000, 34 (8), 1513–1520. (32) Nambi, I. M.; Werth, C. J.; Sanford, R. A.; Valocchi, A. J. Porescale analysis of anaerobic halorespiring bacterial growth along the transverse mixing zone of an etched silicon pore network. Environ. Sci. Technol. 2003, 37 (24), 5617–5624. (33) Stewart, T. L.; Fogler, H. S. Biomass plug development and propagation in porous media. Biotechnol. Bioeng. 2001, 72 (3), 353–363. (34) Whitesides, G. M.; Stroock, A. D. Flexible methods for microfluidics. Phys. Today 2001, 54 (6), 42–48. (35) Duffy, D. C.; McDonald, J. C.; Schueller, O. J. A.; Whitesides, G. M. Rapid prototyping of microfluidic systems in poly(dimethylsiloxane). Anal. Chem. 1998, 70 (23), 4974–4984. (36) McCabe, W. L.; Smith, J. C.; Harriot, P. Unit Operations of Chemical Engineering, 7th ed.; McGraw-Hill: New York, 2005. (37) Crank, J. The Mathematics of Diffusion, 2nd ed.; Clarendon Press: Oxford, U.K., 1975. (38) Sherwood, J. L.; Sung, J. C.; Ford, R. M.; Fernandez, E. J.; Maneval, J. E.; Smith, J. A. Analysis of bacterial random motility in a porous medium using magnetic resonance imaging and immunomagnetic labeling. Environ. Sci. Technol. 2003, 37 (4), 781–785. (39) Olson, M. S.; Ford, R. M.; Smith, J. A.; Fernandez, E. J. Analysis of column tortuosity for MnCl2 and bacterial diffusion using magnetic resonance imaging. Environ. Sci. Technol. 2005, 39 (1), 149–154. (40) Geankoplis, C. J. Transport Processes and Unit Operations, 3rd ed.; PTR Prentice Hall: Engelwood Cliffs, N.J., 1993; p xiii, p. 921. (41) Periasamy, N.; Verkman, A. S. Analysis of fluorophore diffusion by continuous distributions of diffusion coefficients: application to photobleaching measurements of multicomponent and anomalous diffusion. Biophys. J. 1998, 75 (1), 557–67. (42) Berkowitz, B.; Cortis, A.; Dentz, M.; Scher, H. Modeling nonFickian transport in geological formations as a continuous time random walk. Rev. Geophys. 2006, 44 (2), RG2003. (43) Bijeljic, B.; Rubin, S.; Scher, H.; Berkowitz, B. Non-Fickian transport in porous media with bimodal structural heterogeneity. J. Contam. Hydrol. 2011, 120121, 213–221. (44) Dong, H.; Rothmel, R.; Onstott, T. C.; Fuller, M. E.; DeFlaun, M. F.; Streger, S. H.; Dunlap, R.; Fletcher, M. Simultaneous transport of two bacterial strains in intact cores from Oyster, Virginia: biological effects and numerical modeling. Appl. Environ. Microbiol. 2002, 68 (5), 2120–2132. (45) Hornberger, G. M.; Mills, A. L.; Herman, J. S. Bacterial transport in porous media: evaluation of a model using laboratory observations. Water Resour. Res. 1992, 28 (3), 915–923. (46) Marcos, R. S. Microorganisms in vortices: a microfluidic setup. Limnol. Oceanogr.: Methods 2006, 4, 392–398.
8787
dx.doi.org/10.1021/es201706w |Environ. Sci. Technol. 2011, 45, 8780–8787
ARTICLE pubs.acs.org/est
Floc Volume Effects in Suspensions and Its Relevance for Wastewater Engineering Jochen Henkel,* Barbara Siembida-L€osch, and Martin Wagner Technische Universit€at Darmstadt, Institut IWAR, Section for Wastewater Technologies, Petersenstrasse 13, 64287 Darmstadt, Germany ABSTRACT: The aim of this paper is to better understand oxygen transfer reduction caused by floc suspensions. We demonstrate that the overall floc volume significantly influences oxygen transfer depletion. Submerged fine bubble and coarse bubble diffusers are affected in the same way by this phenomenon. The mixed liquor suspended solids concentration (MLSS concentration) is not an appropriate parameter for describing or relating phenomena that are caused by the overall floc volume in activated sludge (e.g., oxygen transfer depression and sludge sedimentation characteristics). A better correlation is achieved by using the mixed liquor volatile suspended solids concentration (MLVSS concentration). To characterize the effects of the overall floc volume in suspensions whose MLVSS concentration cannot be determined (e.g., inorganic iron hydroxide flocs), a new method—the hydrostatic floc volume (HFV)—that approximates the overall floc volume in floc suspensions is introduced. Application of this method demonstrates that oxygen transfer depression caused by iron hydroxide flocs and activated sludge flocs is similar.
’ INTRODUCTION The sedimentation behavior of activated sludge continues to be an active research area.14 The main driver for these investigations is to understand and optimize the settling performance of activated sludge in the clarifier, which is essential for reliable operation of activated sludge plants. The standard procedure applied, in practice, to judge the sedimentation behavior is the integrated sludge volume index (SVI). It relates the sludge volume determined after 30 min to the MLSS concentration. Variations of this method, mainly to overcome the bridging problems with filamentous sludge or at higher sludge concentrations, have been developed, including the dilution (DSVI) or the stirred method (SSVI). Bridging is caused by the interaction of the flocs. Generally, steric interactions, which hinder separation of the free water from the bound water of the flocs, occur more frequently with increasing floc number.5,6 Recently, our group was able to demonstrate that the overall floc volume influences oxygen transfer in floc suspensions.7 This was achieved using iron hydroxide flocs and relating oxygen transfer depression (α-factor) to the overall floc volume. In contrast to the SVI concept, the purpose of the floc volume concept is not to estimate the sedimentation behavior of activated sludge but to use a simple method for estimating the overall volume of all flocs in the system. In an earlier study we hypothesized that the better correlation of oxygen transfer depression and the MLVSS concentration, compared to the MLSS concentration of different activated sludge sources, was obtained because the MLVSS concentration regulates the free water content and better correlates with the overall floc volume in activated sludge.8 One explanation is that r 2011 American Chemical Society
the major part of the activated sludge floc consists of water (>90%), which is bound to organic matter as extracellular polymeric substances (EPS), which again contribute to more than 50% of the organic dried matter of sludge.912 Consequently, it is the amount of bounded water that governs the mass and volume of the entire floc. The MLSS concentration, in contrast to the MLVSS concentration, also includes inorganic materials, such as silt, clay, and sand, which escape removal in the primary settling tank and which, compared to the organic matter, have low water binding capabilities but contribute significantly to the dry solid content (∼1525%). This inorganic content varies, depending on the wastewater influent composition and the primary settling tank performance, which led to the wide range of values for α-factors in the past, if the values were related to the MLSS concentration of different wastewater sources.8 We also hypothesized that oxygen transfer depression caused by iron hydroxide flocs is similar to the mechanism by which activated sludge flocs influence oxygen transfer.7 However, at that time, because a methodology for approximating the overall floc volume of activated sludge was lacking, we were not able to verify this hypothesis. This paper aims to fill these knowledge gaps to better understand oxygen transfer reduction caused by activated sludge flocs and by iron hydroxide flocs. A new method is introduced called Received: May 24, 2011 Accepted: August 18, 2011 Revised: August 14, 2011 Published: September 14, 2011 8788
dx.doi.org/10.1021/es201772w | Environ. Sci. Technol. 2011, 45, 8788–8793
Environmental Science & Technology
ARTICLE
the hydrostatic floc volume (HFV) that approximates the overall floc volume in activated sludge and is able to compare results obtained in floc suspensions. Furthermore, we show that the MLSS concentration is not an appropriate parameter to correlate and compare phenomena that are caused by the floc volume.
’ MATERIALS AND METHODS Generally, the experiments in this study are a continuation of the experiments described in previous studies.7,8 In the current work, two pilot-scale membrane bioreactors fed with municipal wastewater were used for activated sludge sampling. Oxygen transfer measurements with iron hydroxide flocs and wastewater sludge were performed in a separate lab-scale column. Hydrostatic Floc Volume (HFV). In a mixture of rigid particles and a liquid, the total volume of the particles, in terms of the total volume of the suspension (solid holdup), can be easily determined by the bulk volume of the particles. However, this relationship cannot be used for elastic flocs that incorporate a great deal of water, as is the case for iron hydroxide flocs and activated sludge flocs. The only method which generates an estimate of floc volume is the sludge volume index (SVI, APHA 2710 D).13 It determines the sludge volume after 30 min of sedimentation and relates it to the suspended solids concentration. It is used to characterize the sedimentation characteristics of activated sludge in the clarifier. However, because sedimentation still continues after 30 min, it is not an appropriate parameter to determine the overall floc volume. Consequently, a method had to be developed that enables approximation of the overall floc volume in these suspensions. The following procedure was applied. A 1 L sample of activated sludge or iron hydroxide suspension was taken and stored in a settling column (7 cm diameter). The sedimentation process of the sample continued until the volume of the settled flocs remained constant. In the case of activated sludge, the flocs started to float because of microbial gas production, and therefore, the sludge respiration had to be suppressed. This was achieved using cyanide, as described by Dobbs et al.,14 at a dose of 0.1 g of cyanide per 1 g of biomass dry content. The final volume of the suspension is called the hydrostatic floc volume (HFV). An example of activated sludge sedimentation and definition of the HFV is summarized in Table 1. It can be seen that after approximately 45 h the suspension reached its terminal volume, which is the HFV. Oxygen Transfer Measurements and α-Factor Calculation. The α-factor is the relationship between the oxygen transfer
coefficient (kLa) in the investigated medium (in our case activated sludge and iron hydroxide suspension) and the oxygen transfer coefficient in clean water (eq 1). α-factor ¼
kL amedium kL acleanw
ð1Þ
The α-factor describes how much better or worse oxygen diffuses into the medium compared to clean water and plays an important role in estimating the required standard oxygen transfer rate (SOTR) in activated sludge, which is the key parameter in submerged diffused aeration systems. To determine oxygen transfer coefficients, the desorption method according to Wagner et al.15 was applied (see also Henkel et al.7,8). Four oxygen sensors (iRAS automation GmbH, Bad Klosterlausnitz, Germany) recorded the change in oxygen concentration in the suspension at constant air flow rate and
Table 1. Example of Floc Sedimentation (MLSS = 7.2 g/L; MLVSS = 6 g/L) elapsed time
volume ratio (mL/L)
0h
1000
2 h 15 min 3 h 15 min
380 340
20 h 20 min
260
44 h 55 min
250 (HFV)
141 h
250
Figure 1. Volumetric mass transfer coefficient during iron hydroxide experiments (fine bubble aeration).
solid concentration. The air flow rate, which was measured with a thermal flow sensor TA10 (Hoentzsch GmbH, Waiblingen, Germany), was then changed and the procedure repeated at the same solid concentration. Three air flow rates were chosen for each test (1, 2, 3 m3/h). Subsequently, the three average kLa values were plotted against the superficial gas velocity (SGV). If applicable, a linear trend line was plotted and calculated. This procedure was repeated for every solid concentration and the clean water test. An example is given in Figure 1. Finally, the α-factor was calculated by dividing the trend line equation at a specific solid concentration and specific air flow rate (2 m3air/(diffuser 3 h)) by the equation obtained during the clean water test. This procedure was applied because it was impossible to repeat exactly the same air flow rate for each test series. Humidity, air pressure, air temperature, and the air blower itself are subject to daily variations that influence the air flow rate, which has to be transformed to standard conditions. Lab-Scale Experiments. The bubble column used in this experiment was 1.30 m high and 0.43 m in diameter. The total water volume in each test was 100 L. In contrast to the iron hydroxide experiments performed by Henkel et al.,7 this time 4.5 kg of ferric chloride was mixed with 250 L of potable water in a separate vessel and bentonite was not added. The pH was then adjusted to 8 with 1 M sodium hydroxide, which substantially increased the salt concentration. As the salt content also influences oxygen transfer, the supernatant of the iron hydroxide flocs was replaced by potable water several times until the salt content reached a concentration equivalent to that of potable water. The final salt content varied between 600 and 850 mg/L. After sedimentation of the iron hydroxide flocs, 150 L of the supernatant was stored separately. 8789
dx.doi.org/10.1021/es201772w |Environ. Sci. Technol. 2011, 45, 8788–8793
Environmental Science & Technology
ARTICLE
Table 2. Hydrodynamic Specifications during Lab-Scale Oxygen Transfer Experiments diffuser
number of
riser surface,
superficial gas velocity
specific slit aeration
specific diffuser aeration rate,
type
orifices, n
m2
(SGV), cm3Air/(cm2Surf. 3 s)
rate, cm3Air/(slit 3 s)
m3Air/(diffuser 3 h)
disc
8000
0.15
0.150.55
0.030.10
13
tube
10
0.15
0.150.55
2585
13
Figure 2. Relationship between sludge volumes (dilution method) and MLSS concentration.
Figure 3. Relationship between sludge volumes and MLVSS concentration.
The oxygen transfer experiments started at the highest iron hydroxide concentration (26.1 g/L). After each experiment, a fixed amount of iron hydroxide flocs was removed from the system and replaced with the stored supernatant. kLa and the α-factor were determined as described in section. A fine bubble and a coarse bubble aeration device were tested to investigate whether iron hydroxide flocs show the same behavior as observed previously with activated sludge.8 The specifications of the setup are listed in Table 2. Finally, the same experiments were performed with activated sludge and the fine bubble aeration device. The activated sludge was taken from a membrane bioreactor fed with real wastewater from a municipal wastewater plant (DarmstadtEberstadt, Germany). To achieve a range of sludge concentrations, the sludge was diluted with permeate from the membrane bioreactor. The highest sludge concentration tested was 17.5 g/L. Sedimentation Experiments. Once the hydrostatic floc volume procedure was established, it was applied during operation of two pilot-scale membrane bioreactors. The membrane bioreactors had a water volume of 1 m3 (1.7 m water depth), and the wastewater flow was 100 L/h. The organic load was 0.9 kg of COD/d. The F/M ratio depended on the actual MLVSS concentration in the tank but ranged from 0.45 to 0.02 kg BOD5/ (kg MLVSS 3 d). Additionally, the sludge volume was determined after 30 min without dilution to compare the results.13
different kinds of sludges showing different settling behavior at the same MLSS concentration could show similar settling behavior at the same MLVSS concentration if the overall floc volume is the main reason for bridging. Figure 2 depicts the sludge volume, determined with the dilution method, as a function of the MLSS concentration of three sludges. The data for the greywater sludge was generated during the 2 year operation of two membrane bioreactors fed with synthetic greywater.7,8 The data for the wastewater sludges were generated during the operation of two separate membrane bioreactors fed with real wastewater. A linear relationship can be observed in all three cases, although the variation of the sludge volume at a certain MLSS concentration was quite large. Whereas the wastewater sludges showed a similar development with increasing MLSS concentration, the greywater sludge volume increased more slowly with increasing MLSS concentration. Calculation of the SVI indicates that the sedimentation characteristic of membrane bioreactor wastewater sludge at a specific MLSS concentration was worse (for example, SVI 83 mL/g at MLSS of 12 g/L) than that of the greywater sludge (SVI ≈ 33 mL/g at MLSS of 12 g/L). Microscopic investigation of the sludges revealed that neither filamentous bacteria nor bulking sludge were present in the MBR greywater sludge and the MBR wastewater sludge 2010. During colder periods, especially in January and February, the MBR wastewater sludge 2009 did contain filamentous bacteria. The major difference between sludges was the organic content, which was around 38% and 82% for the greywater sludge and the wastewater sludge, respectively. The nontypical low ignition loss of greywater sludge was caused by the presence of a large amount of abrasive materials, used in toothpaste, and zeolites, used in washing agents, in the greywater influent. If the same data are plotted against MLVSS concentration (Figure 3), each type of sludge showed a much closer relationship than before. We concluded that the overall floc volumes of greywater and wastewater sludge at a certain MLVSS concentration were
’ RESULTS AND DISCUSSION Sludge Volume Index. The sludge volume index provides a rough estimate of how well sludge settles in the clarifier. It is calculated from the ratio of sludge volume after 30 min to MLSS concentration. As noted, the dilution method and/or stirring method have been applied to determine sludge volume after 30 min, when sludge settling properties are hindered by bridging between flocs.5,16 We hypothesized that the MLVSS concentration better correlates to the overall floc volume in activated sludge. One conjecture generated by this hypothesis was that
8790
dx.doi.org/10.1021/es201772w |Environ. Sci. Technol. 2011, 45, 8788–8793
Environmental Science & Technology
ARTICLE
Figure 4. Sludge volume after 30 min of sedimentation (blue triangles) and hydrostatic floc volume (yellow rectangles).
similar, which resulted in similar sedimentation characteristic, as observed in this study, and similar oxygen transfer behaviors of different kinds of sludges observed in our previous study.8 This is another indicator for the close relationship of MLVSS concentration to the overall floc volume in activated sludge. Hydrostatic Floc Volume. As noted in the former section, the sludge volume method, especially at high suspended solids concentrations, delivers a relatively wide range of sludge volumes at a certain MLSS or MLVSS concentration. Additionally, sludge volumes larger than 1000 mL/L are an artifact of the dilution method, caused by multiplication of the dilution applied to the sludge, rather than an approximation of the overall floc volume present in the sludge. Because the MLVSS concentration may only be used to approximate the overall floc volume in biological floc aggregates and we wanted to compare activated sludge flocs with iron hydroxide flocs (inorganic flocs), we developed a new method (see Materials and Methods) to approximate the overall floc volume in floc suspensions that does not destroy the floc structure and is easy to apply. We emphasize that it is not an alternative to the SVI method, which aims at characterizing activated sludge sedimentation behavior in the clarifier. Figure 4 shows the results of sludge volume determination according to the APHA method,13 without stirring or dilution, and the HFV method, related to the MLVSS concentration. The volume ratio increased with increasing MLVSS concentration in both cases. A linear correlation was observed for the HFV method up to a floc volume of about 500 mL/L at an MLVSS concentration of 11 g/L. The trend line follows the equation HFV ¼ 43:85 3 MLVSSð( 22Þ
ð2Þ
Compared to the sludge volume method with dilution presented in the section Sludge Volume Index, the HFV method still delivers reasonable values at elevated MLVSS concentrations and, additionally, shows smaller deviations. The values achieved with the sludge volume method without dilution demonstrate the effect of bridging that occurs at sludge volumes higher than 300 mL/L. The steric interferences of the flocs hinder sedimentation and result in a sharp increase in sludge volumes. It seems as if the HFV method resolves this bridging effect and correlates well with the MLVSS concentration. It indicates that this method may be used to correlate the overall floc volume of activated sludge. Oxygen Transfer at Various Floc Concentrations. In a previous study, we demonstrated that the MLVSS concentration affects oxygen transfer depression (α-factor) of submerged fine
Figure 5. α-Factor of iron hydroxide flocs and activated sludge flocs related to the concentration of suspended solids.
Figure 6. α-Factor of iron hydroxide flocs and activated sludge flocs related to the hydrostatic floc volume.
bubble and coarse bubble aeration systems in the same way.8 To test the hypothesis that the mechanism of oxygen transfer depression caused by activated sludge flocs and iron hydroxide flocs is similar, we exposed iron hydroxide flocs to fine and coarse bubble aeration systems. Figure 5 illustrates the results of this experiment. Additionally, one set of tests was performed with activated sludge in the same column to compare the results. Figure 5 shows that the α-factor in the experiments with iron hydroxide decreased linearly with increasing MLSS concentration, irrespective of the aeration system (coarse bubble aeration or fine bubble aeration). This result is identical to the data derived from activated sludge in the earlier studies. However, even though the decrease is also linear for activated sludge flocs, the slope, compared to the iron hydroxide flocs, is different. Hence, it could be concluded that activated sludge flocs reduce oxygen transfer more strongly than iron hydroxide flocs. As mentioned above, one reason why we developed the HFV method was the theory that the MLSS concentration did not correlate with the overall floc volume of various suspensions. In Figure 6 we compare the α-factors from Figure 5 to the HFV. In contrast to the results shown in Figure 5, all suspensions now show similar trends. This result is remarkable because hydroxide flocs and activated sludge flocs are very different in 8791
dx.doi.org/10.1021/es201772w |Environ. Sci. Technol. 2011, 45, 8788–8793
Environmental Science & Technology their water binding capacities and densities.17,18 The results support the hypothesis that the overall floc volume is the actual driver for oxygen transfer depression in suspensions that contain flocs and that the MLSS concentration insufficiently correlates with the overall floc volume. These results also suggest that the HFV is a useful parameter for comparing phenomena that can be attributed to the floc volume, such as gas transfer. In oxygen transfer studies activated sludge has been considered to be a pseudohomogeneous medium with non-Newtonian pseudoplastic fluid properties.19 One way of characterizing these fluids is to measure their apparent viscosity. Although the apparent viscosity of activated sludge has often been assumed to be responsible for the decrease in the α-factor, supporting systematic studies are scarce. Some authors correlated the apparent viscosity to the α-factor of different sludges but only with limited success, especially if the results of these studies are compared with each other.20,21 It is debatable whether apparent viscosity is a meaningful parameter in terms of oxygen transfer in floc suspensions. A promising solution to this dilemma would be to replace the idea of a pseudohomogeneous medium with the new concept of a suspension consisting of flocs that interact with the air bubble. Interpretations of how the floc influences oxygen transfer into the free water fraction could then incorporate results that come from investigations of solid suspensions in the field of chemical engineering. A parameter used to compare oxygen transfer in solid suspensions is the solid holdup, which relates the overall volume occupied by the solids to the overall volume of the suspension. With increasing solid holdup (εS), a decrease in oxygen transfer is commonly observed.2224 Mena et al.25 summarize eight ways in which the gasliquid system can be affected by solids. Transferring these observations to activated sludge leads to the proposition that the following phenomena may influence oxygen transfer. 1 The increased floc volume decreases the interfacial area between the bubble and the liquid phase, since it makes contact with the bubble surface and hinders transfer to the liquid phase. 2 The tendency of bubbles to coalesce is favored by attachment of small, hydrophobic flocs, which leads to larger bubbles with a lower specific surface area. 3 Turbulence in the bubble wake is diminished due to accumulation of flocs in the wake area. 4 With increasing floc number, the possibility of collision during bubble formation at the orifices increases, which reduces the bubble formation frequency and leads to larger bubbles at the orifice. 5 The decrease in the free water content leads to an increase in gas holdup, related to the liquid phase at the same air flow rate. This shifts the critical superficial gas velocity, which is responsible for the change from the homogeneous flow regime to the heterogeneous flow regime, to lower values. 6 The increased floc volume also increases the probability of collision between the flocs themselves and increases the apparent viscosity. This again leads to a decrease in the liquid velocity at the same air flow rate, which results in formation of larger bubbles at the orifice and enhances the probability of bubble coalescence. All these effects would lead to a decrease of the volumetric oxygen transfer coefficient kLa in activated sludge. Since suppression of the α-factor for fine bubble and coarse bubble aeration systems was similar and bubble formation and bubble rise
ARTICLE
characteristics of coarse bubbles are not affected by the liquid properties,26 only two phenomena remain that could explain the similar behavior. The first is the reduction of turbulence in the bubble wake area; the second describes suppression of the oxygen transfer from the bubble to the liquid phase. However, further investigations are required to determine which mechanism is responsible for this phenomenon.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +1 303 273 3871, fax: +1 303 273 3413; e-mail:
[email protected].
’ ACKNOWLEDGMENT The authors wish to thank Prof. Cornel, our department head, for his time and inspiring discussions, the German Federal Ministry of Education & Research, and the Faudi Foundation for their financial support. Ms. Anita Curt, our laboratory assistant at the wastewater treatment plant in Eberstadt, deserves special thanks for the enormous number of samples she processed during this work. ’ REFERENCES (1) Chen, Y.; Yang, H.; Gu, G. Effect of acid and surfactant treatment on activated sludge dewatering and settling. Water Res. 2001, 35 (11), 2615–2620. (2) Li, X. Y.; Yang, S. F. Influence of loosely bound extracellular polymeric substances (EPS) on the flocculation, sedimentation and dewaterability of activated sludge. Water Res. 2007, 41 (5), 1022–1030. (3) Schuler, A. J.; Jang, H. Causes of variable biomass density and its effects on settleability in full-scale biological wastewater treatment systems. Environ. Sci. Technol. 2007, 41 (5), 1675–1681. (4) Wilen, B.-M.; Lumley, D.; Mattsson, A.; Mino, T. Relationship between floc composition and flocculation and settling properties studied at a full scale activated sludge plant. Water Res. 2008, 42 (16), 4404–4418. (5) Dick, R. I.; Vesilind, P. A. The Sludge Volume Index: What Is It? Water Pollut. Control Fed. 1969, 41 (7), 1285–1291. (6) Bye, C. M.; Dold, P. L. Sludge volume index settleability measures: Effect of solids characteristics and test parameters. Water Environ. Res. 1998, 70 (1), 87–93. (7) Henkel, J.; Cornel, P.; Wagner, M. Free Water Content and Sludge Retention Time: Impact on Oxygen Transfer in Activated Sludge. Environ. Sci. Technol. 2009, 43 (22), 8561–8565. (8) Henkel, J.; Lemac, M.; Wagner, M.; Cornel, P. Oxygen transfer in membrane bioreactors treating synthetic greywater. Water Res. 2009, 43 (6), 1711–1719. (9) Vaxelaire, J.; Cezac, P. Moisture distribution in activated sludges: a review. Water Res. 2004, 38 (9), 2215–2230. (10) Smollen, M. Moisture retention characteristics and volume reduction of municipal sludges. Water SA 1988, 14 (1), 25–28. (11) Frolund, B.; Palmgren, R.; Keiding, K.; Nielsen, P. H. Extraction of extracellular polymers from activated sludge using a cation exchange resin. Water Res. 1996, 30 (8), 1749–1758. (12) Wilen, B. M.; Jin, B.; Lant, P. The influence of key chemical constituents in activated sludge on surface and flocculating properties. Water Res. 2003, 37 (9), 2127–2139. (13) APHA, Standard methods for the examination of water and wastewater; American Public Health Association, American Water Works Association, Water Pollution Control Federation: Washington DC, 2005. 8792
dx.doi.org/10.1021/es201772w |Environ. Sci. Technol. 2011, 45, 8788–8793
Environmental Science & Technology
ARTICLE
(14) Dobbs, R. A.; Shan, Y. G.; Wang, L. P.; Govind, R. Sorption on Wastewater solids - Elimination of biological activity. Water Environ. Res. 1995, 67 (3), 327–329. (15) Wagner, M. R.; Johannes P€opel, H.; Kalte, P. Pure oxygen desorption method - A new and cost-effective method for the determination of oxygen transfer rates in clean water. Water Sci. Technol. 1998, 38 (3), 103–109. € (16) Stobbe, G. Uber das Verhalten von belebtem Schlamm in aufsteigender Wasserbewegung; Ver€offentlichungen des Institutes f€ur Siedlungswasserwirtschaft der Technischen Hochschule Hannover: Hannover, Germany, 1964; Vol. 18. (17) Colin, F.; Gazbar, S. Distribution of water in sludges in relation to their mechanical dewatering. Water Res. 1995, 29 (8), 2000–2005. (18) Wu, C. C.; Huang, C.; Lee, D. J. Bound water content and water binding strength on sludge flocs. Water Res. 1998, 32 (3), 900–904. (19) Yang, G. Q.; Du, B.; Fan, L. S. Bubble formation and dynamics in gas-liquid-solid fluidization- A review. Chem. Eng. Sci. 2007, 62 (12), 2–27. (20) Krampe, J.; Krauth, K. Oxygen transfer into activated sludge with high MLSS concentrations. Water Sci. Technol. 2003, 47 (11), 297–303. (21) Krause, S. Untersuchungen zum Energiebedarf von Membranbelebungsanlagen. Dissertation; Technische Universit€at Darmstadt, Darmstadt, 2005. (22) Schumpe, A.; Fang, L. K.; Deckwer, W.-D. Stoff€ubergang Gas/ Fl€ussigkeit in Suspensions-Blasens€aulen. Chem. Ing. Tech. 1984, 56 (12), 924–926. (23) Krishna, R.; deSwart, J. W. A.; Ellenberger, J.; Martina, G. B.; Maretto, C. Gas holdup in slurry bubble columns: Effect of column diameter and slurry concentrations. AIChE J. 1997, 43 (2), 311–316. (24) Freitas, C.; Teixeira, J. A. Oxygen mass transfer in a high solids loading three-phase internal-loop airlift reactor. Chem. Eng. J. 2001, 84 (1), 57–61. (25) Mena, P. C.; Ruzicka, M. C.; Rocha, F. A.; Teixeira, J. A.; Drahos, J. Effect of solids on homogeneous-heterogeneous flow regime transition in bubble columns. Chem. Eng. Sci. 2005, 60 (22), 6013–6026. (26) Kumar, R.; Kuloor, N. R., The formation of bubbles and drops. In Advances in chemical engineering; Drew, T., Cokelet, G., Hoopes, J., Vermeulen, T., Eds. Academic Press: New York, 1970; Vol. 8, pp 255368.
8793
dx.doi.org/10.1021/es201772w |Environ. Sci. Technol. 2011, 45, 8788–8793
ARTICLE pubs.acs.org/est
In Situ Profiling of Microbial Communities in Full-Scale Aerobic Sequencing Batch Reactors Treating Winery Waste in Australia Simon J. McIlroy, Lachlan B. M. Speirs, Joseph Tucci, and Robert J. Seviour* Biotechnology Research Centre, Department of Pharmacy and Applied Science, La Trobe University, Bendigo, Victoria 3552, Australia ABSTRACT: On-site aerobic sequencing batch reactor (SBR) treatment plants are implemented in many Australian wineries to treat the large volumes of associated wastewater they generate. Yet very little is known about their microbiology. This paper represents the first attempt to analyze the communities of three such systems sampled during both vintage and nonvintage operational periods using molecular methods. Alphaproteobacterial tetrad forming organisms (TFO) related to members of the genus Defluviicoccus and Amaricoccus dominated all three systems in both operational periods. Candidatus ‘Alysiosphaera europaea’ and Zoogloea were codominant in two communities. Production of high levels of exocellular capsular material by Zoogloea and Amaricoccus is thought to explain the poor settleability of solids in one of these plants. The dominance of these organisms is thought to result from the high COD to N/P ratios that characterize winery wastes, and it is suggested that manipulating this ratio with nutrient dosing may help control the problems they cause.
’ INTRODUCTION The wine industry is a major global industry with over 25 billion liters of wine produced in 2008 alone (www.oiv.org). It has been estimated that for each liter of wine 12.5 L of wastewater is produced,1,2 although this can be as high as 14 L.3 Consequently, large amounts of wastewater are generated at various stages during the wine production period (vintage) as well as from bottling and cleaning of equipment and the maintenance of cellar hygiene in nonvintage periods.1,4,5 A wide range of treatment technologies has been applied around the world for dealing with this wastewater, involving both aerobic and anaerobic plant configurations, where one of the main selection criteria used is their operating costs (reviewed by refs 4 and 5). Of the aerobic configurations, which are more efficient for treatment, sequencing batch reactors are used in many Australian wineries (J. Constable, personal communication) because of their simplicity and suitability for coping with the high seasonal variations in winery waste composition and generated loads.1,6 The chemical nature of winery wastewater differs substantially from that of domestic sewage. In particular, it possesses high BOD/ COD:N/P ratios, which, while fluctuating considerably over the course of the year, for COD:P ratios are reported to be in excess of 100:1.1,6 Its pH is also generally lower than domestic sewage being below 4.5.1 These marked differences reflect the high levels of sugars and organic acids in winery wastewaters, and in many plants P and N supplementation of the wastewater is undertaken to facilitate its biological treatment (J. Constable, personal communication). The high carbon load poses serious potential environmental threats, and so again driven by cost considerations, many of the larger wineries treat their own wastes on site instead of discharging them into domestic sewerage systems. Molecular rRNA-based methods have allowed us to begin to resolve the biodiversity of the microbial communities in plants of r 2011 American Chemical Society
many different configurations treating wastewater from domestic and industrial sources.7 Furthermore, they have provided the tools to reveal the in situ functions of many of these populations, especially those involved in P and N removal.810 It is quite clear that the microbial community that develops in a treatment process reflects the selective pressures imposed by both the plant operational conditions and the nature of the wastewater being treated.11 As mentioned above, winery wastewaters are highly distinctive in their nature, which poses challenges for their successful biological treatment. Despite this, very few attempts have been made to understand the microbiology involved, and these have been restricted largely to recognizing dominant morphotypes from microscopic examination of the biomass (e.g., ref 12) or from studies using culture dependent methods (e.g., refs 2 and 13). Both are restricted in the levels of information they can provide on community biodiversity. This situation is surprising, because many of the aerobic SBR treatment plants in Australia suffer from operational problems like poor sludge settleability, which are likely to be microbiological in origin. Consequently, we applied molecular techniques to profile communities of three aerobic SBR systems treating the waste in different wine-producing regions in Australia. The data presented here show for the first time that these are each dominated by bacterial populations reported elsewhere in habitats largely characterized by high COD:P ratios, and we propose that these bacteria may cause the operational problems encountered there. Received: May 31, 2011 Accepted: August 29, 2011 Revised: August 29, 2011 Published: August 29, 2011 8794
dx.doi.org/10.1021/es2018576 | Environ. Sci. Technol. 2011, 45, 8794–8803
Environmental Science & Technology
ARTICLE
Table 1. Plant Operating Parameters Plant A item
a
nonvintage
typical crush (Tonne)
N/A
SBR volume (ML)
1.1
Plant B vintage
30 000
nonvintage N/A
Plant C vintage
20 000
4.6
nonvintage N/A
vintage 1000
0.2
daily flow (kL)
30
280
350
550
4
20
total cycle time (h)
4
12
12
12
12
12
aerate (min)a
30
560
500
400
600
565
settle (min) decant (min)
60 10
60 60
60 40
60 70
105 7
60 40
feed (min)
10
60
130
190
8
55
MLSS (g L1)
5000
900011 000
6000
4480
2400
5600
SRT (days)
Not dewatered
2030
4060
2030
1520
810
COD
1000
800010 000
50007000
50007000
10 400
16 000
BOD
500
60008000
30004000
40006000
6700
10 000
DO (end of aeration)
12
12
34
34
7.1
5.8
pH, feed pH, SBR
57 8
45 7.5
45 6.77.3
45 6.87.3
5.7 7.1
4.0 8.1
temp. (°C)
1214
3035
1214
2830
11
24.6
nutrient dosage (kg week1)
none
urea, 10; DAPb 5
none
urea, 140
urea, 10; DAP,b 2
urea, 20; DAP,b 5
To reduce operating costs, these reactors are not always aerated during the feed steps or periods of the aeration stages. b Diammonium phosphate.
’ EXPERIMENTAL SECTION Samples. Samples were taken from three full-scale sequencing batch reactors (SBR) treating winery waste in New South Wales (plant A) in nonvintage and South Australia (plant B) and Tasmania (plant C) in both vintage (during wine production) (April 2011) and nonvintage (July 2010) operating periods. Operating parameters for each plant are given in Table 1. Typical influent total P values were 510 mg/L for plants A and B and 530 mg/L for plant C. Fluorescence in Situ Hybridization (FISH). FISH was carried out using PFA (paraformaldehyde) fixed biomass samples as described by Daims et al.14 The FISH probes listed in Table 2 were used and hybridizations performed at 46 °C for 1.5 h unless otherwise stated. All probes were purchased from ProOligo (Sigma-Aldrich, Australia). Slide wells were mounted in Vectashield (Vector Laboratories) and viewed on a Nikon E800 epifluorescence microscope. To increase cell wall permeability to some of the FISH probes, pretreatments with lysozyme, achromopeptidase, and mild acid hydrolysis as detailed in the protocol of Kragelund et al.15 were applied to biomass samples. Appropriate controls for biomass autofluorescence using the non-EUB probes were included in all analyses. Histochemical Staining. Intracellular polyphosphate granules were detected by applying 4,6-diamidino-2-phenylindole (DAPI) solution (50 μg mL1) for 10 min to air-dried biomass smears as described by Kawaharasaki et al.46 Biomass was also stained for the presence of intracellular polyhydroxyalkanoate (PHA) granules with Nile Blue A (100 mg L1 in ethanol) at 55 °C for 10 min, based on the method of Ostle and Holt47 as described in Ahn et al.48 The India ink negative capsule and Neisser staining methods were carried out as described by Jenkins et al.49 16S rRNA Gene Clone Library Preparation from Plant A Biomass. Biomass DNA was extracted using three different methods in attempts to minimize possible extraction biases
associated with each: the sodium trichloroacetate-based method of McIlroy et al.,50 the potassium xanthogenate-based method of Tillett and Neilan,51 and the FastDNA SPIN Kit (Qbiogene, Melbourne, Australia). For each of the DNA extracts, 16S rRNA genes were PCR amplified in quadruplicate with the 27f/1492r and 27f/1525r primer sets,52 with annealing temperatures of 42 and 52 °C, respectively, and the resulting products pooled and stored at 70 °C until required. All other PCR, cloning, and sequencing details are those described by Nittami et al.29 Partial sequence reads (>500 bp) were organized into OTUs based on shared 99% sequence similarities and a ‘complete’ 16S rRNA sequence obtained for a representative from each OTU of interest. 16S rRNA Sequencing of Cultured Isolates. Attempts were made to culture some of the dominant morphotypes from plant C by streaking out the biomass onto solid media. Both GS and R2A agar were used as described by Maszenan et al.53 Colonies were screened microscopically, and those with the desired morphotype (see Results) were subcultured repeatedly until Gram staining suggested they were pure. A single colony was then used for 16S rRNA sequencing as above, except the DNA was extracted from cells as described by McIlroy et al.,54 and PCR products were sequenced directly by AGRF (Brisbane, Australia).
’ RESULTS Application of the EUBmix FISH probe set indicated that the majority of the organisms present in all three winery wastewater treatment plant biomass samples taken during both vintage and nonvintage operational periods were Bacteria. Large yeast-like cells, fungal hyphae, and Protozoa were seen too, particularly in the plant C waste sample, but always in low abundance. Almost all bacterial cells stained strongly for PHA inclusions, while a smaller number (approximately <20% of the total) also stained positively for small polyphosphate inclusions (data not shown). 8795
dx.doi.org/10.1021/es2018576 |Environ. Sci. Technol. 2011, 45, 8794–8803
Environmental Science & Technology
ARTICLE
Table 2. List of Oligonucleotide FISH Probes Used in This Study probe name b
target
[FA]a (%)
ref
EUB338-I EUB338-IIb
most Bacteria Planctomycetales
2050 2050
16 17
EUB338-IIIb
Verrucomicrobiales
2050
17
NON-EUB
control probe complementary to EUB338
n/a
18
BET42a
Betaproteobacteria
35
19
PAO462bc
Candidatus ‘Accumulibacter phosphatis’
35
20
PAO651c
Candidatus ‘Accumulibacter phosphatis’
35
21
PAO846bc ZRA23a
Candidatus ‘Accumulibacter phosphatis’ most members of the Zoogloea lineage
35 35
20 22
AZA645
Azoarcus group
20
23
SNA
Sphaerotilus natans
45
24
GAM42a
Gammaproteobacteria
35
19
GB_G1d
Candidatus ‘Competibacter phosphatis’, group G1
35
25
GB_G2d
Candidatus ‘Competibacter phosphatis’, group G2
35
25
ALF968
Alphaproteobacteria, except Rickettsiales
20
26
TFO_DF218e TFO_DF618e
cluster I Defluviicoccus-related TFO cluster I Defluviicoccus-related TFO
35 35
27 27
DF988f
cluster II Defluviicoccus-related TFO
35
28
DF1020f
cluster II Defluviicoccus-related TFO
35
28
DF198
cluster III Defluviicoccus-related ‘Nostocoida limicola’
35
29
DF1004
subgroup within cluster III Defluviicoccus-related ‘Nostocoida limicola’
35
29
DF1013
subgroup within cluster III Defluviicoccus-related ‘Nostocoida limicola’
35
29
DF181Ag
clone K42 within cluster IV Defluviicoccus-related TFO
30
30
DF181Bg AMAR839
some cluster IV Defluviicoccus-related TFO Amaricoccus spp.
30 20
30 31
Noli-644
Candidatus ‘Alysiosphaera europaea’
35
32
HGC69a
Actinobacteria (high G+C Gram-positive bacteria)
25
33
LGC354Ah
most Firmicutes (Gram-positive bacteria with low G+C content)
35
34
LGC354Bh
most Firmicutes (Gram-positive bacteria with low G+C content)
35
34
LGC354Ch
most Firmicutes (Gram-positive bacteria with low G+C content)
35
34
CFX1223I
phylum Chloroflexi (green nonsulfur bacteria)
35
35
GNSB-941I CHL1851
phylum Chloroflexi (green nonsulfur bacteria) type 1851 filamentous bacterium
35 35
36 37
EU251238
type 1851 filamentous bacterium
35
38
CFX197
type 0092 filamentous bacterium
40
39
CFX223
type 0092 filamentous bacterium
35
39
CFX67aj
type 0914 filamentous bacterium
35
40
CFX67bj
type 0914 filamentous bacterium
35
40
T08030654
type 0803 filamentous bacterium
30
41
T0803ind-0642 PLA46
type 0803 filamentous bacterium Planctomycetales
30 30
41 42
CF319ak
most Flavobacteria, some Bacteroidetes, some Sphingobacteria
35
43
CF319bk
some Flavobacteria and Sphingobacteria
35
43
CFB719
most members of the class Bacteriodetes, some Flavobacteria and Sphingobacteria
30
44
CFB286
most members of the genus Tannerella and the genus Prevotella of the class Bacteriodetes
50
44
HHY
Haliscomenobacter hydrossis
20
24
TM7905
candidate TM7 division
20
45
a
[FA] = formamide concentration in hybridization buffer. Multiple probes applied in equimolar amounts to cover respective groups. b EUBmix. c PAOmix. d GBmix. e DF1mix. f DF2mix. g DF4mix. h LGCmix. I CFXmix. j CFX67mix. k CF319mix. n/a = not applicable. Competitor and helper probes were applied as suggested in the original papers for each probe.
In situ structures for each plant community were analyzed by FISH, and the results are summarized in Table 3. FISH Analysis of Plant A Biomass. This sample taken from the treatment plant during the nonvintage operational period
was dominated by Gram-negative, Neisser-positive filaments with the distinctive ‘Nostocoida limicola’ type II morphology (Figure 1ac) extending from the floc surface and participating in some interfloc bridging. Large inclusions, which stained 8796
dx.doi.org/10.1021/es2018576 |Environ. Sci. Technol. 2011, 45, 8794–8803
Environmental Science & Technology
ARTICLE
Table 3. Relative Abundances of Selected FISH-Probed Bacterial Populations of Interest in Each Winery Plant Biomass Samplea
a
Quantification estimate criteria: ++++ = dominant (>50% of biovolume); +++ = very abundant (approx. 2550% of biovolume); ++ = common (approx. 525% of biovolume); + = some to few (approx. <5% of biovolume); = none; NV = nonvintage; V = vintage.
strongly with Nile Blue A, were observed in most of the individual cells, consistent with them being intracellular PHA granules (Figure 1c). A preincubation with lysozyme and achromopeptidase and an extended overnight FISH hybridization period was required to obtain a positive FISH signal from this filament. Positive signals were obtained with the ALF968 probe, covering the Alphaproteobacteria, the DF988 probe, targeting cluster II members of the genus Defluviicoccus, and the Noli-644 probe, targeting Candidatus ‘Alysiosphaera europaea’. Certainly its morphology most closely resembled that of Candidatus ‘A. europaea’, which also stains positively with the Neisser stain and for PHA inclusions and appears commonly in communities in plants treating industrial wastes in Europe as a probable cause of bulking.55,56 To clarify these FISH data and resolve the identity of this filament, a 16S rRNA gene clone library was generated from the plant A biomass. Clones were then screened for sequences possessing either the Noli-644 and/or the DF988 FISH probe target sites. Of the 88 clones sequenced, 19 were members of the Alphaproteobacteria with 16 of these being >97% similar to Candidatus ‘A. europaea’ (W-OTU1 and W-OTU2) (Figure 3). These sequences had only a single terminal mismatch to that of
the Noli-644 probe target (Table 4), suggesting this FISH probe is the one binding to this filament in situ. They also contained a potential binding site for the DF988 probe, with 3 terminal mismatches and a missing internal base (Table 4). The positions of the terminal mismatches were similar to those allowing nonspecific binding of the DF988 FISH probe to the same position in the 16S rRNA sequence of cluster III Defluviicoccus members29 as shown in Table 4 and have a lower theoretical ΔΔG°overall (calculated for the perfectly matched sequence with mathFISH software57). The ability of FISH probes to bind to target sites despite there being missing bases in the target site and forming a bulge over these missing bases has been demonstrated previously.58 However improbable the DF988 probe binding to Candidatus ‘A. europaea’ despite four mismatches in its target site might seem, inclusion in the FISH protocol of an unlabeled competitor probe with the perfect match to this putative probe target sequence, as shown in Table 4, eliminated all the above-background level fluorescence (data not shown). FISH data also suggested that cluster I TFO members of the alphaproteobacterial Defluviicoccus vanus-related group were 8797
dx.doi.org/10.1021/es2018576 |Environ. Sci. Technol. 2011, 45, 8794–8803
Environmental Science & Technology
ARTICLE
Figure 1. Micrographs of SBR-activated sludge biomass from (ad) plant A and (e and f) plant B. (a) Neisser stain of the dominant filament in plant A; (b) phase contrast images with corresponding composite FISH image with the EUBmix (Fluos = green) and DF988 (CY3 = red) probe sets (green + red = yellow); (c) phase contrast image with corresponding fluorescence image after Nile Blue A staining (PHA granules fluoresce red); (d) phase contrast image with corresponding composite FISH image with the EUBmix (green) and DF218 (red) probe sets; (e) phase contrast image with corresponding composite FISH image with the ALF968 (green) and DF218 (red) probe sets; (f) phase contrast image with corresponding composite FISH image with the EUBmix (green) and DF988 (red) probe sets. Scale bars indicate 10 μm.
codominant with this filament in the plant A community, being present mainly as tight spherical clustered cells within the flocs (Figure 1d). These populations were detected in the clone library (W-OTU3) (Figure 3). Members of cluster II Defluviicoccus TFO were also present in the flocs but at much lower abundances. Unfortunately, analysis of the corresponding bacterial community
during the vintage period was not possible as the plant was decommissioned at the end of the nonvintage period. FISH Analysis of Plant B. The microbial community composition of the biomass taken from this plant in both vintage and nonvintage was very similar, with both being dominated overwhelmingly by TFO cells responding to the alphaproteobacterial 8798
dx.doi.org/10.1021/es2018576 |Environ. Sci. Technol. 2011, 45, 8794–8803
Environmental Science & Technology
ARTICLE
Figure 2. Micrographs of SBR-activated sludge biomass from plant C: (a) phase contrast image with the corresponding composite FISH image with the ALF968 (Fluos = green) and AMAR 839 (CY3 = red) probes (green + red = yellow); (b) phase contrast image with the corresponding fluorescence image after Nile Blue A staining (PHA granules fluoresce red); (c) phase contrast image with the corresponding fluorescence image after DAPI staining (PolyP granules fluoresce light yellow); (d) phase contrast image with India ink capsule staining; (e) phase contrast images with the corresponding composite FISH image with the EUBmix (green), ZRA23a (red), and Beta42a (CY5 = blue) probe sets (green + red + blue = white). Scale bars indicate 10 μm.
probe ALF968. The vast majority of these were identified as members of clusters I and II of Defluviicoccus-related organisms. Cluster I members, the more dominant of the two, occurred in large irregular clusters at times exceeding 200 μm in diameter (Figure 1e). Cluster II members were present as smaller clusters of TFO scattered throughout the biomass (Figure 1f). Thin Neisser-negative, PHA-positive filaments with a morphology similar to that of Haliscomenobacter hydrossis but not hybridizing with the HHY probe were seen commonly protruding from the flocs. These hybridized with the Beta42a probe but were too thin to be Sphaerotilus natans, a betaproteobacterial filament implicated in bulking in activated sludge systems, and they also failed to respond to the SNA probe designed to target it. Filamentous Chloroflexi were also present, although in much lower abundances, and some of these responded to the FISH probes targeting the Eikelboom morphotypes type 080341 and type 1851,37 with the former the more abundant of the two. Little
difference in the community composition was observed between samples taken during the nonvintage and vintage periods, except that cluster I Defluviicoccus cell clusters were generally smaller and more loosely associated in the vintage community. FISH Analysis of Plant C Community. This community from the nonvintage period was again dominated by alphaproteobacterial TFO arranged in large sheets (Figure 2ad). However, these did not respond to any of the Defluviicoccus subgroup FISH probes but instead fluoresced with the AMAR839 probe targeting members of the genus Amaricoccus (Figure 2a). The intensity of the fluorescence signal with this probe varied substantially between individual cells and was improved considerably after a lysozyme prehybridization treatment. These AMAR839-positive cells also stained brightly for PHA inclusions (Figure 2b), and DAPI staining indicated that many possessed small polyP inclusion bodies (Figure 2c). Clusters were surrounded by substantial capsular material (Figure 2d). This community was codominated 8799
dx.doi.org/10.1021/es2018576 |Environ. Sci. Technol. 2011, 45, 8794–8803
Environmental Science & Technology by betaproteobacterial coccibacilli arranged in large microcolonies. These were identified after FISH with the ZRA23a probe as members of the genus Zoogloea (Figure 2e), and staining showed the cells were again heavily capsulated (Figure 2d). A small number of Chloroflexi filaments, most of which responded to the probe targeting type 0803 (T08030654), and H. hydrossis were also detected by FISH, but filamentous bacteria were generally rare. In the plant sample taken from the vintage period, both Amaricoccus TFO and Zoogloea were still present but now in much lower abundance (estimated visually to be about a 50% reduction in both). During the nonvintage period plant C suffered from poor solids settleability or bulking. Yet the biomass possessed insufficient numbers of bacterial filaments to explain this, and interfloc bridging was absent.49 Furthermore, the few fungal filaments seen (Figure 2a) were not associated with the flocs but suspended in the mixed liquor. Instead, the flocs contained heavily capsulated Amaricoccus TFO and Zoogloea (Figure 2d), and it seems much
Figure 3. Maximum likelihood tree of 16S rRNA gene sequences obtained in this study (represented in bold) (all sequences were at least 1200 bp long). The JF957136 sequence represents isolates W-TFO1, W-TFO2, and W-TFO3. Parsimony bootstrap values were calculated as percentages of 1000 analysis and are only indicated for values g 75%: (O) bootstrap value g 75%; (b) bootstrap value g 95%. Scale bar corresponds to 0.1 substitutions per nucleotide position. Brackets define Defluviicoccus-related clusters. AS = Activated sludge. OTU = Operational taxonomic unit.
ARTICLE
more likely that these two heavily capsulated populations were both responsible for this episode of nonfilamentous ‘viscous’ bulking.49 Identification of Isolates from Plant C Biomass. Four of the 100 isolates cultured onto GS medium from plant C were Gramnegative tetrads. The 16S rRNA gene sequence of the first three of these isolates (designated W-TFO1, W-TFO2, and W-TFO3: Acc. No. JF957136) were identical and 97.3% similar to Amaricoccus kaplicensis (Figure 3), also isolated from activated sludge.53 The 16S rRNA sequence of the fourth of these isolates (W-TFO4: Acc. No. JF957137) was 98.8% similarity to the other three and 98.0% similar to A. kaplicensis (Figure 3). All isolates were pleomorphic, with large variation in their cell sizes, typical of Amaricoccus.59,60 These also stained positively for aerobic polyphosphate and PHA storage and produced an extensive capsule. Unlike other Amaricoccus isolates,53 these did not grow on R2A or PYC agar. Their ability to accumulate polyphosphate is also different to that reported for the other Amaricoccus isolates.53
’ DISCUSSION This paper describes for the first time the community composition of winery wastewater treatment plants using cultureindependent rRNA-based methods. It also provides an explanation for the dominance of certain populations and a possible solution to the operational problems they appear to cause. The FISH data from three SBRs treating winery wastewater unexpectedly showed each community was dominated by members of the Alphaproteobacteria. The dominating TFO populations were identified as members of clusters I and II Defluviicoccus-related organisms and possibly new Amaricoccus spp., Candidatus ‘A. europaea’ and Zoogloea were also codominant in the biomass from plants A and C, respectively. Organisms with a similar TFO morphology to those seen in these plants occur commonly in activated sludge communities analyzed by microscopy.61 TFO have also been reported to dominate the biomass treating wastes from wineries,12 although these TFO were not identified nor were they in the study of Liu et al.,62 where they dominated overwhelmingly the community of an EBPR laboratory reactor supplied with a high C:P feed. Bacteria growing as tetrads and representing a wide phylogenetic diversity are known to occur in activated sludge communities.61 However, the TFO of Liu et al.62 are considered likely to be members of the genus Defluviicoccus, since the alphaproteobacterial TFO in the community of a reactor set up to operate in exactly the same way were identified as cluster I members of this genus.58,63 Furthermore, bearing in mind the earlier reports
Table 4. Mismatches in Target Sites Between FISH Probes and Selected 16S rRNA Sequences
*
E. coli position numbering. Base mismatches to the Noli-644 and DF988 probes are highlighted and italicized. 8800
dx.doi.org/10.1021/es2018576 |Environ. Sci. Technol. 2011, 45, 8794–8803
Environmental Science & Technology concerning TFO distribution, it is probably more than a coincidence that the only cultured member of this genus, Defluviicoccus vanus, was isolated from an activated sludge plant treating high carbohydrate brewery wastes.60 In fact, members of this genus are seen frequently in large numbers in anaerobic:aerobic EBPR plant communities operating with low P removal capacity. They are thought to outcompete the polyphosphate-accumulating organisms (PAO) under P-limiting conditions, producing extensive capsular material and storing instead glycogen, thus earning the description of glycogen-accumulating organisms (GAO).11,64 Yet they are not restricted to plants of this configuration, and have been detected in high abundance in communities of plants dealing with paper mill waste, which is also characterized by its high COD:P ratio.30,6567 Furthermore, extremely high Defluviicoccus enrichments (up to 95%) have been achieved in laboratoryscale reactor communities fed high COD:P feeds,6870 adding further evidence to support the view that these conditions, so typical of these winery wastes, support their excessive proliferation. The ecology of Amaricoccus spp. is less clear than that of Defluviicoccus, yet they seem to share several similar ecological traits. Originally isolated from a laboratory-scale EBPR reactor fed glucose and not acetate and showing low EBPR capacity,71 these TFO ‘G-Bacteria’ were identified as members of a new genus Amaricoccus.53 Four species have been described so far with Amaricoccus tamworthensis, like D. vanus, being isolated from an activated sludge plant treating carbohydrate-rich brewery malting wastes. Amaricoccus spp. were also detected by FISH in high abundance in the communities of plants treating paper mill wastes.72 Yet their impact on EBPR capacity is likely to be less detrimental than that of Defluviicoccus. Pure cultures of A. kaplicensis have shown no ability to assimilate substrates under anaerobic conditions,59 meaning that they should not pose a threat to the polyphosphate-accumulating organisms (PAO) in the anaerobic zones of EBPR plants. Why Candidatus ‘A. europaea’ should dominate the community of plant A is unclear, since little is known of its ecophysiology.55 However, Levantesi et al.56 reported its presence in large amounts especially in activated sludge plants treating paper mill and potato wastes, where again the raw plant influent in both would be distinguished by its very high COD:N/P ratio. All the Alphaproteobacteria in these winery wastes appear to store considerable intracellular PHA stores in these winery wastes, and Defluviicoccus, Amaricoccus, and Zoogloea (especially the latter two) also synthesize substantial exocellular capsular material (Figure 2d). These metabolic features are typical responses of bacterial populations to conditions of unbalanced growth73 which would arise from high COD:P/N ratio feeds. Zoogloea populations are commonly seen in activated sludge plants,22 and pure culture studies74 suggest exocellular polysaccharide production by them is encouraged under N-limiting conditions. Therefore, it seems likely that Defluviicoccus, Amaricoccus, ‘Alysiosphaera’, and Zoogloea share a similar ecological niche in that they proliferate under nutrient-limiting high C:N/P feed conditions, and this operational feature probably explains their dominance in these winery wastewater treatment systems. The high relative abundances of floc-associated Amaricoccus and Zoogloea aggregates (Figure 2) may also explain the operational solids separation problems experienced by plant C. The nonfilamentous viscous bulking49 experienced by the plant during the nonvintage period of operation is considered to be caused by them, encouraged to proliferate excessively by the high COD:P/N
ARTICLE
ratio of the feed. Heavily encapsulated TFO bacteria were also thought to be responsible for nonfilamentous bulking in a plant treating winery waste in Hungary.12 This solids separation problem observed in plant C did not occur in vintage, where the Amaricoccus population decreased substantially. The reason for this decline in relative abundance of Amaricoccus is unclear. Clearly more studies of this kind are needed with winery plants to see if the microbiological trends recorded here occur globally. The data from this study suggest that winery wastewater treatment plants and others treating carbohydrate-rich waste may contain a rich reservoir of further biodiversity of Defluviicoccus and Amaricoccus members. They also suggest that episodes of nonfilamentous viscous bulking, invariably linked in the past to Zoogloea, may in fact be caused by other bacterial populations also favored by N- and/or P-limiting conditions like Amaricoccus. If so, bulking control in these plants may be controlled readily by careful P and N supplementation of the influent wastes to allow balanced growth of the communities and to discourage the excessive proliferation of these potentially troublesome bacterial populations.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +61 3 5444 7459. Fax: +61 3 5444 7476. E-mail: r.seviour@ latrobe.edu.au.
’ ACKNOWLEDGMENT The authors would like to thank the plant operators, Robert Morris, Andrew Johns, Grant Kohlhagen, and Nichola Seymour, for supplying samples and plant operation information. Thanks also go to John Constable (JJC Engineering Pty. Ltd) for his helpful advice and information. S.M. and L.S. were supported by a Post Graduate Writing Up Award from La Trobe University and APA Ph.D. scholarship, respectively. Photograph of a stream used in the TOC art was taken by Kristy McIlroy and used with permission. ’ REFERENCES (1) Brito, A. G.; Peixoto, J.; Oliveira, J. M.; Oliveira, J. A.; Costa, C.; Nogueira, R.; Rodrigues, A. Brewery and winery wastewater treatment: some focal points of design and operation. In Utilization of by-products and treatment of waste in the food industry; Oreopoulou, V., Russ, W., Eds.; Springer: New York, 2007; Vol. 3, pp 109131. (2) Eusebio, A.; Petruccioli, M.; Lageiro, M.; Federici, F.; Duarte, J. C. Microbial characterisation of activated sludge in jet-loop bioreactors treating winery wastewaters. J. Ind. Microbiol. Biotechnol. 2004, 31 (1), 29–34. (3) Eusebi, A. L.; Nardelli, P.; Gatti, G.; Battistoni, P.; Cecchi, F. From conventional activated sludge to alternate oxic/anoxic process: the optimization of winery wastewater treatment. Water Sci. Technol. 2009, 60 (4), 1041–1048. (4) Arvanitoyannis, I. S.; Ladas, D.; Mavromatis, A. Wine waste treatment methodology. Int. J. Food Sci. Technol. 2006, 41 (10), 1117–1151. (5) Frost, P.; Kumar, A.; Correll, R.; Quayle, W.; Kookana, R.; Christen, E.; Oemcke, D. Current practices for winery wastewater management and its reuse: an Australian industry survey. Wine Ind. J. 2007, 22 (1), 40–46. (6) Torrijos, M.; Moletta, R. Winery wastewater depollution by sequencing batch reactor. Water Sci. Technol. 1997, 35 (1), 249–257. (7) Seviour, R. J.; Nielsen, P. H. Microbial Ecology of Activated Sludge; IWA Publishing: London, 2010. 8801
dx.doi.org/10.1021/es2018576 |Environ. Sci. Technol. 2011, 45, 8794–8803
Environmental Science & Technology (8) He, S.; McMahon, K. D. Microbiology of Candidatus Accumulibacter in activated sludge. Microb. Biotechnol. In press, doi:10.1111/ j.1751-7915.2011.00248.x. (9) McMahon, K. D.; He, S.; Oehmen, A. The microbiology of phosphorus removal. In Microbial ecology of activated sludge; Seviour, R. J., Nielsen, P. H., Eds.; IWA Publishing: London, 2010; pp 281319. (10) Daims, H.; Wagner, M. The microbiology of nitrogen removal. In Microbial ecology of activated sludge; Seviour, R. J., Nielsen, P. H., Eds.; IWA Publishing: London, 2010; pp 259280. (11) Oehmen, A.; Lemos, P. C.; Carvalho, G.; Yuan, Z.; Keller, J.; Blackall, L. L.; Reis, M. A. M. Advances in enhanced biological phosphorus removal: from micro to macro scale. Water Res. 2007, 41 (11), 2271–2300. (12) Jobbagy, A.; Literathy, B.; Tardy, G. Implementation of glycogen accumulating bacteria in treating nutrient-deficient wastewater. Water Sci. Technol. 2002, 46 (12), 185–190. (13) Gonzalez, J. M.; Jurado, V.; Laiz, L.; Zimmermann, J.; Hermosin, B.; Saiz-Jimenez, C. Pectinatus portalensis non. sp., a relatively fastgrowing, coccoidal, novel Pectinatus species isolated from a wastewater treatment plant. Antonie van Leeuwenhoek 2004, 86 (3), 241–248. (14) Daims, H.; Stoecker, K.; Wagner, M. Fluorescence in situ hybridization for the detection of prokaryotes. In Molecular Microbial Ecology; Osborn, A. M., Smith, C. J., Eds.; Taylor & Francis: New York, 2005; pp 213239. (15) Kragelund, C.; Remesova, Z.; Nielsen, J.; Thomsen, T.; Eales, K.; Seviour, R.; Wanner, J.; Nielsen, P. Ecophysiology of mycolic acidcontaining Actinobacteria (Mycolata) in activated sludge foams. FEMS Microbiol. Ecol. 2007, 61 (1), 174–184. (16) Amann, R. I.; Binder, B. J.; Olson, R. J.; Chisolm, S. W.; Devereux, R.; Stahl, D. A. Combination of 16S rRNA-targeted oligonucleotide probes with flow cytometry for analyzing mixed microbial populations. Appl. Environ. Microbiol. 1990, 56 (6), 1919–1925. (17) Daims, H.; Br€uhl, A.; Amann, R.; Schleifer, K.; Wagner, M. The domain-specific probe EUB338 is insufficient for the detection of all Bacteria: development and evaluation of a more comprehensive probe set. Syst. Appl. Microbiol. 1999, 22 (3), 434–444. (18) Wallner, G.; Amann, R.; Beisker, W. Optimizing fluorescent in situ hybridization with rRNA-targeted oligonucleotide probes for flow cytometric identification of microorganisms. Cytometry 1993, 14 (2), 136–143. (19) Manz, W.; Amann, R.; Ludwig, W.; Wagner, M.; Scheifer, K.-H. Phylogenetic oligodeoxynucleotide probes for the major subclasses of Proteobacteria: problems and solutions. Syst. Appl. Microbiol. 1992, 15 (4), 593–600. (20) Zilles, J.; Peccia, J.; Kim, M.; Hung, C.; Noguera, D. Involvement of Rhodocyclus-related organisms in phosphorus removal in fullscale wastewater treatment plants. Appl. Environ. Microbiol. 2002, 68 (6), 2763–2769. (21) Crocetti, G. R.; Hugenholtz, P.; Bond, P. L.; Schuler, A.; Keller, J.; Jenkins, D.; Blackall, L. L. Identification of polyphosphate-accumulating organisms and design of 16S rRNA-directed probes for their detection and quantitation. Appl. Environ. Microbiol. 2000, 66 (3), 1175–1182. (22) Rossello-Mora, R. A.; Wagner, M. The abundance of Zoogloea ramigera in sewage treatment plants. Appl. Environ. Microbiol. 1995, 61 (2), 702–707. (23) Hess, A.; Zarda, B.; Hahn, D.; Haner, A.; Stax, D.; Hohener, P.; Zeyer, J. In situ analysis of denitrifying toluene- and m-xylene-degrading bacteria in a diesel fuel-contaminated laboratory aquifer column. Appl. Environ. Microbiol. 1997, 63 (6), 2136–41. (24) Wagner, M.; Amann, R.; K€ampfer, P.; Assmus, B.; Hartmann, A.; Hutzler, P.; Springer, N.; Schleifer, K.-H. Identification and in situ detection of gram-negative filamentous bacteria in activated sludge. Syst. Appl. Microbiol. 1994, 17 (3), 405–417. (25) Kong, Y.; Ong, S. L.; Ng, W. J.; Liu, W.-T. Diversity and distribution of a deeply branched novel proteobacterial group found in anaerobic-aerobic activated sludge processes. Environ. Microbiol. 2002, 4 (11), 753–757. (26) Neef, A.; Witzenberger, R.; K€ampfer, P. Detection of sphingomonads and in situ identification in activated sludge using 16S rRNA-
ARTICLE
targeted oligonucleotide probes. J. Ind. Microbiol. Biotechnol. 1999, 23 (45), 261–267. (27) Wong, M.-T.; Tan, F. M.; Ng, W. J.; Liu, W.-T. Identification and occurrence of tetrad-forming Alphaproteobacteria in anaerobic-aerobic activated sludge processes. Microbiology 2004, 150 (11), 3741–3748. (28) Meyer, R. L.; Saunders, A. M.; Blackall, L. L. Putative glycogenaccumulating organisms belonging to the Alphaproteobacteria identified through rRNA-based stable isotope probing. Microbiology 2006, 152 (2), 419–429. (29) Nittami, T.; McIlroy, S.; Seviour, E. M.; Schroeder, S.; Seviour, R. J. Candidatus Monilibacter spp., common bulking filaments in activated sludge, are members of Cluster III Defluviicoccus. Syst. Appl. Microbiol. 2009, 32 (7), 480–489. (30) McIlroy, S.; Seviour, R. J. Elucidating further phylogenetic diversity among the Defluviicoccus-related glycogen-accumulating organisms in activated sludge. Environ. Microbiol. Rep. 2009, 1 (6), 563–568. (31) Maszenan, A.-M.; Seviour, R. J.; Patel, B. K.; Wanner, J. A fluorescently-labelled r-RNA targeted oligonucleotide probe for the in situ detection of G-bacteria of the genus Amaricoccus in activated sludge. J. Appl. Microbiol. 2000, 88 (5), 826–835. (32) Snaidr, I.; Beimfohr, C.; Levantesi, C.; Rossetti, S.; van der Waarde, J.; Geurkink, B.; Eikelboom, D.; Lemaitre, M.; Tandoi, V. Phylogenetic analysis and in situ identification of “Nostocoida limicola”like filamentous bacteria in activated sludge from industrial wastewater treatment plants. Water Sci. Technol. 2002, 46 (12), 99–104. (33) Roller, C.; Wagner, M.; Amann, R.; Ludwig, W.; Schleifer, K.-H. In situ probing of Gram-positive bacteria with high DNA G + C content using 23S rRNA-targeted oligonucleotides. Microbiology 1994, 140 (10), 2849–2858. (34) Meier, H.; Amann, R.; Ludwig, W.; Schleifer, K. Specific oligonucleotide probes for in situ detection of a major group of Grampositive bacteria with low DNA G + C content. Syst. Appl. Microbiol. 1999, 22 (3), 186–196. (35) Bj€ornsson, L.; Hugenholtz, P.; Tyson, G. W.; Blackall, L. L. Filamentous Chloroflexi (green non-sulfur bacteria) are abundant in wastewater treatment processes with biological nutrient removal. Microbiology 2002, 148 (8), 2309–2318. (36) Gich, F.; Garcia-Gil, J.; Overmann, J. Previously unknown and phylogenetically diverse members of the green nonsulfur bacteria are indigenous to freshwater lakes. Arch. Microbiol. 2001, 177 (1), 1–10. (37) Beer, M.; Seviour, E. M.; Kong, Y.; Cunningham, M.; Blackall, L. L.; Seviour, R. J. Phylogeny of the filamentous bacterium Eikelboom Type 1851, and design and application of a 16S rRNA targeted oligonucleotide probe for its fluorescence in situ identification in activated sludge. FEMS Microbiol. Lett. 2002, 207 (2), 179–183. (38) Kragelund, C.; Levantesi, C.; Borger, A.; Thelen, K.; Eikelboom, D.; Tandoi, V.; Kong, Y.; van der Waarde, J.; Krooneman, J.; Rossetti, S.; Thomsen, T.; Nielsen, P. Identity, abundance and ecophysiology of filamentous Chloroflexi species present in activated sludge treatment plants. FEMS Microbiol Ecol 2007, 59 (3), 671–682. (39) Speirs, L.; Nittami, T.; McIlroy, S.; Schroeder, S.; Seviour, R. J. Filamentous bacterium Eikelboom type 0092 in activated sludge plants in Australia is a member of the phylum Chloroflexi. Appl. Environ. Microbiol. 2009, 75 (8), 2446–52. (40) Speirs, L. B. M.; McIlroy, S. J.; Petrovski, S.; Seviour, R. J. The activated sludge bulking filament Eikelboom morphotype 0914 is a member of the Chloroflexi. Environ. Microbiol. Rep. 2011, 3 (2), 159–165. (41) Kragelund, C.; Thomsen, T. R.; Mielczarek, A. T.; Nielsen, P. H. Eikelboom’s morphotype 0803 in activated sludge belongs to the genus Caldilinea in the phylum Chloroflexi. FEMS Microbiol. Ecol. 2011, 76 (3), 451–62. (42) Neef, A.; Amann, R.; Schlesner, H.; Schleifer, K.-H. Monitoring a widespread bacterial group: in situ detection of Planctomycetes with 16S rRNA-targeted probes. Microbiology 1998, 144 (12), 3257–3266. (43) Manz, W.; Amann, R.; Ludwig, W.; Vancanneyt, M.; Schleifer, K. Application of a suite of 16S rRNA-specific oligonucleotide probes designed to investigate bacteria of the phylum Cytophaga-Flavobacter-Bacteroides in the natural environment. Microbiology 1996, 142 (5), 1097–1106. 8802
dx.doi.org/10.1021/es2018576 |Environ. Sci. Technol. 2011, 45, 8794–8803
Environmental Science & Technology (44) Weller, R.; Gl€ockner, F.; Amann, R. 16S rRNA-targeted oligonucleotide probes for the in situ detection of members of the phylum Cytophaga-Flavobacterium-Bacteroides. Syst. Appl. Microbiol. 2000, 23 (1), 107–114. (45) Hugenholtz, P.; Tyson, G. W.; Webb, R. I.; Wagner, A. M.; Blackall, L. L. Investigation of candidate division TM7, a recently recognized major lineage of the domain bacteria with no known pureculture representatives. Appl. Environ. Microbiol. 2001, 67 (1), 411–419. (46) Kawaharasakia, M.; Tanakab, H.; Kanagawaa, T.; Nakamuraa, K. In situ identification of polyphosphate-accumulating bacteria in activated sludge by dual staining with rRNA-targeted oligonucleotide probes and 40 ,6-diamidino-2-phenylindol (DAPI) at a polyphosphateprobing concentration. Water Res. 1999, 33 (1), 257–265. (47) Ostle, A.; Holt, J. Nile blue A as a fluorescent stain for poly-betahydroxybutyrate. Appl. Environ. Microbiol. 1982, 44 (1), 238–241. (48) Ahn, J.; Schroeder, S.; Beer, M.; McIlroy, S.; Bayly, R. C.; May, J. W.; Vasiliadis, G.; Seviour, R. J. Ecology of the microbial community removing phosphate from wastewater under continuously aerobic conditions in a sequencing batch reactor. Appl. Environ. Microbiol. 2007, 73 (7), 2257–2270. (49) Jenkins, D.; Richard, M. G.; Daigger, G. T. Manual on the causes and control of activated sludge bulking, foaming and other solids separation problems, 3rd ed.; CRC Press LLC: London, England, 2004. (50) McIlroy, S.; Porter, K.; Seviour, R. J.; Tillett, D. Simple and safe method for simultaneous isolation of microbial RNA and DNA from problematic populations. Appl. Environ. Microbiol. 2008, 74 (21), 6806–6807. (51) Tillett, D.; Neilan, B. A. Xanthogenate nucleic acid isolation from cultured and environmental Cyanobacteria. J. Phycol. 2000, 36 (1), 251–258. (52) Lane, D. 16S/23S rRNA sequencing. In Modern microbial methods: Nucleic acid techniques in bacterial systematics; Stackebrandt, E., Goodfellow, M., Eds.; John Wiley & Sons: England, 1991; pp 115175. (53) Maszenan, A.; Seviour, R.; Patel, B.; Rees, G.; McDougall, B. Amaricoccus gen. nov., a Gram-negative coccus occurring in regular packages or tetrads, isolated from activated sludge biomass, and descriptions of Amaricoccus veronensis sp. nov., Amaricoccus tamworthensis sp. nov., Amaricoccus macauensis sp. nov., and Amaricoccus kaplicensis sp. nov. Int. J. Syst. Bacteriol. 1997, 47 (3), 727–734. (54) McIlroy, S.; Hoefel, D.; Schroeder, S.; Ahn, J.; Tillett, D.; Saint, C.; Seviour, R. FACS enrichment and identification of floc-associated alphaproteobacterial tetrad-forming organisms in an activated sludge community. FEMS Microbiol. Lett. 2008, 285 (1), 130–135. (55) Kragelund, C.; Kong, Y.; van der Waarde, J.; Thelen, K.; Eikelboom, D.; Tandoi, V.; Thomsen, T.; Nielsen, P. Ecophysiology of different filamentous Alphaproteobacteria in industrial wastewater treatment plants. Microbiology 2006, 152 (10), 3003–3012. (56) Levantesi, C.; Beimfohr, C.; Geurkink, B.; Rossetti, S.; Thelen, K.; Krooneman, J.; Snaidr, J.; van der Waarde, J.; Tandoi, V. Filamentous Alphaproteobacteria associated with bulking in industrial wastewater treatment plants. Syst. Appl. Microbiol. 2004, 27 (6), 716–727. (57) Yilmaz, L. S.; Parnerkar, S.; Noguera, D. R. mathFISH, a web tool that uses thermodynamics-based mathematical models for in silico evaluation of oligonucleotide probes for fluorescence in situ hybridization. Appl. Environ. Microbiol. 2011, 77 (3), 1118–22. (58) McIlroy, S. J.; Tillett, D.; Petrovski, S.; Seviour, R. J. Non-target sites with single nucleotide insertions or deletions are frequently found in 16S rRNA sequences and can lead to false positives in fluorescence in situ hybridization (FISH). Environ. Microbiol. 2011, 13 (1), 38–47. (59) Falvo, A.; Levantesi, C.; Rossetti, S.; Seviour, R.; Tandoi, V. Synthesis of intracellular storage polymers by Amaricoccus kaplicensis, a tetrad forming bacterium present in activated sludge. J. Appl. Microbiol. 2001, 91 (2), 299–305. (60) Maszenan, A.; Seviour, R.; Patel, B.; Janssen, P.; Wanner, J. Defluvicoccus vanus gen. nov., sp. nov., a novel Gram-negative coccus/ coccobacillus in the Alphaproteobacteria from activated sludge. Int. J. Syst. Evol. Microbiol. 2005, 55 (5), 2105–2111. (61) Seviour, R. J.; Maszenan, A. M.; Soddell, J. A.; Tandoi, V.; Patel, B. K. C.; Kong, Y.; Schumann, P. Microbiology of the ’G-bacteria’ in activated sludge. Environ. Microbiol. 2000, 2 (6), 581–593.
ARTICLE
(62) Liu, W. T.; Mino, T.; Nakamura, K.; Matsuo, T. Glycogen accumulating population and its anaerobic substrate uptake in anaerobic-aerobic activated sludge without biological phosphorus removal. Water Res. 1996, 30 (1), 75–82. (63) Kong, Y. H.; Beer, M.; Rees, G. N.; Seviour, R. J. Functional analysis of microbial communities in aerobic-anaerobic sequencing batch reactors fed with different phosphorus/carbon (P/C) ratios. Microbiology 2002, 148 (8), 2299–2307. (64) Mino, T.; Liu, W.-T.; Kurisu, F.; Matsuo, T. Modelling glycogen storage and denitrification capability of microorganisms in enhanced biological phosphate removal processes. Water Sci. Technol. 1995, 31 (2), 25–34. (65) McIlroy, S. J.; Nittami, T.; Seviour, E. M.; Seviour, R. J. Filamentous members of cluster III Defluviicoccus have the in situ phenotype expected of a glycogen accumulating organism in activated sludge. FEMS Microbiol. Ecol. 2010, 74 (1), 248–256. (66) Pisco, A. R.; Bengtsson, S.; Werker, A.; Reis, M. A. M.; Lemos, P. C. Community structure evolution and enrichment of glycogenaccumulating organisms producing polyhydroxyalkanoates from fermented molasses. Appl. Environ. Microbiol. 2009, 75 (14), 4676–86. (67) Bengtsson, S.; Werker, A.; Welander, T. Production of polyhydroxyalkanoates by glycogen accumulating organisms treating a paper mill wastewater. Water Sci. Technol. 2008, 58 (2), 323–330. (68) Burow, L. C.; Mabbett, A. N.; Borras, L.; Blackall, L. L. Anaerobic central metabolic pathways active during polyhydroxyalkanoate production in uncultured cluster 1 Defluviicoccus enriched in activated sludge communities. FEMS Microbiol. Lett. 2009, 298 (1), 79–84. (69) Oehmen, A.; Zeng, R.; Saunders, A.; Blackall, L.; Keller, J.; Yuan, Z. Anaerobic and aerobic metabolism of glycogen-accumulating organisms selected with propionate as the sole carbon source. Microbiology 2006, 152 (9), 2767–2778. (70) Dai, Y.; Yuan, Z.; Wang, X.; Oehmen, A.; Keller, J. Anaerobic metabolism of Defluviicoccus vanus related glycogen accumulating organisms (GAOs) with acetate and propionate as carbon sources. Water Res. 2007, 41 (9), 1885–1896. (71) Cech, J. S.; Hartman, P. Competition between polyphosphate and polysaccharide accumulating bacteria in enhanced biological phosphate removal systems. Water Res. 1993, 27 (7), 1219–1225. (72) McIlroy, S. J. Phylogenetic diversity and ecophysiology of alphaproteobacterial glycogen accumulating organisms in enhanced biological phosphorus removal activated sludge systems. Ph.D. Thesis, La Trobe University, Bendigo, 2010. (73) Kessler, B.; Witholt, B. Factors involved in the regulatory network of polyhydroxyalkanoate metabolism. J. Biotechnol. 2001, 86 (2), 97–104. (74) Norberg, A. B.; Enfors, S.-O. Production of extracellular polysaccharide by Zoogloea ramigera. Appl. Environ. Microbiol. 1982, 44 (5), 1231–1237.
8803
dx.doi.org/10.1021/es2018576 |Environ. Sci. Technol. 2011, 45, 8794–8803
ARTICLE pubs.acs.org/est
Large Shift in Source of Fine Sediment in the Upper Mississippi River Patrick Belmont,*,†,‡ Karen B. Gran,‡,§ Shawn P. Schottler,|| Peter R. Wilcock,‡,^ Stephanie S. Day,‡,# Carrie Jennings,#,r J. Wesley Lauer,O Enrica Viparelli,‡,[ Jane K. Willenbring,‡,z,+ Daniel R. Engstrom,|| and Gary Parker‡,[ †
Department of Watershed Science, Utah State University, Logan, Utah 84332, United States National Center for Earth-surface Dynamics, University of Minnesota, Minneapolis, Minnesota 55414, United States § Department of Geological Sciences, University of Minnesota-Duluth, Duluth, Minnesota 55812, United States St. Croix Watershed Research Station, Science Museum of Minnesota, Marine on St. Croix, Minnesota 55047, United States ^ Department of Geography and Environmental Engineering, Johns Hopkins University, Baltimore, Maryland 21218, United States # Geology and Geophysics, University of Minnesota-Twin Cities, Minneapolis, Minnesota 55455, United States r Minnesota Geological Survey, St. Paul, Minnesota 55114, United States O Department of Civil and Environmental Engineering, Seattle University, Seattle, Washington 98122, United States [ Department of Civil & Environmental Engineering and Department of Geology, University of Illinois, Urbana, Illinois 61801, United States z Department of Earth and Environmental Science, University of Pennsylvania, Philadelphia, Pennsylvania 19104, United States + Helmholtz Centre Potsdam, GFZ German Research Centre for Geosciences, Section 3.4, Earth Surface Geochemistry, D-14473 Potsdam, Germany
)
‡
bS Supporting Information ABSTRACT: Although sediment is a natural constituent of rivers, excess loading to rivers and streams is a leading cause of impairment and biodiversity loss. Remedial actions require identification of the sources and mechanisms of sediment supply. This task is complicated by the scale and complexity of large watersheds as well as changes in climate and land use that alter the drivers of sediment supply. Previous studies in Lake Pepin, a natural lake on the Mississippi River, indicate that sediment supply to the lake has increased 10-fold over the past 150 years. Herein we combine geochemical fingerprinting and a suite of geomorphic change detection techniques with a sediment mass balance for a tributary watershed to demonstrate that, although the sediment loading remains very large, the dominant source of sediment has shifted from agricultural soil erosion to accelerated erosion of stream banks and bluffs, driven by increased river discharge. Such hydrologic amplification of natural erosion processes calls for a new approach to watershed sediment modeling that explicitly accounts for channel and floodplain dynamics that amplify or dampen landscape processes. Further, this finding illustrates a new challenge in remediating nonpoint sediment pollution and indicates that management efforts must expand from soil erosion to factors contributing to increased water runoff.
’ INTRODUCTION Sediment and turbidity are leading causes of impairment in U.S. rivers and streams1,2 and remedial action requires identification of the sources and mechanisms of sediment supply. Despite extraordinary efforts, sediment remains one of the most difficult nonpoint-source pollutants to quantify for several reasons.37 Erosion is typically episodic and highly localized. Erosion mechanisms are strongly nonlinear, and their rates are contingent on multiple factors including climate, geology, and land use history.8,9 Eroded sediment may exit the watershed quickly or be stored for very long periods.10,11 Finally, the accuracy of current methods for estimating sediment r 2011 American Chemical Society
yield from agricultural watersheds has been challenged4,8,12 because most estimates are based on empirical models of soil erosion and require a scalar reduction factor to estimate sediment yield as a fraction of erosion. Few studies provide evidence to constrain this reduction factor, and available observations indicate diverse and highly nonlinear scaling with drainage area.7,8 Received: June 4, 2011 Accepted: August 31, 2011 Revised: August 30, 2011 Published: August 31, 2011 8804
dx.doi.org/10.1021/es2019109 | Environ. Sci. Technol. 2011, 45, 8804–8810
Environmental Science & Technology
Figure 1. Inset A shows the Lake Pepin (LP) watershed within the upper Midwest, composed of the Minnesota (MNR), upper Mississippi (MISS), and St. Croix River (SC) watersheds. Inset B shows lidar topography data for the incised portion of the Le Sueur watershed (LS), including the locations of gaging stations on all three main branches.
Accurate identification of sediment sources and erosion rates are needed to understand and manage the landscape sediment routing system13 and related biogeochemical processes.14 The reliability of sediment source estimates can be improved by using multiple, overlapping methods of measurement within the strong constraint of a mass balance, or sediment budget.15 A sediment budget is a useful tool for evaluating landscape change and sediment yield.10,1619 The scope and accuracy of sediment budgets depend strongly on the availability of information for earlier conditions in a watershed. For example, historical information such as photos, maps, and field studies have been used to provide reliable information on previous conditions in order to close a sediment budget over a time period long enough to average over stochastic temporal variability. This approach is strengthened by a suite of new research tools that allow precise dating of land surfaces, geochemical identification of sediment provenance, and high-resolution measurement of topography using airborne and terrestrial lidar.20,21 The waters of the Upper Mississippi River (UMR) and its major tributaries have been listed as impaired for turbidity by the U.S. Environmental Protection Agency. Turbidity, eutrophication, and sedimentation have been identified as urgent problems for Lake Pepin, a natural lake on the Mississippi River of exceptional recreational and popular importance (Figure 1). Coring records examining the past 500 years indicate that sedimentation rates in Lake Pepin may have increased by as much as an order of magnitude over the past 150 years.22 Of the sediment delivered to Lake Pepin, past and present, 80% to 90% derives from the Minnesota River Basin (MRB), despite the fact that the MRB comprises only a third of the drainage area.22,23 The relatively high sediment yield of the MRB stems from a combination of Quaternary landscape history and human land and water management. Land cover in the basin has shifted from poorly drained tall-grass prairie and wetlands24 to 78% row crop agriculture25 over the 150 yr period of increased sedimentation, suggesting that the change in land use underlies the increase.26 The 2880 km2 Le Sueur River watershed produces the highest sediment yield (73.5 Mg/km2) of any Minnesota River tributary, accounting for as much as 30% of the Minnesota River sediment load.26 The Le Sueur landscape is naturally primed for rapid geomorphic change and large sediment supply. The geologic substrate is a 60 m thick package of semiconsolidated, but soft, fine-grained (67% silt and clay, 33% sand, <1% gravel and boulders) tills and glaciofluvial sands.27 Glacial Lake Agassiz
ARTICLE
Figure 2. Longitudinal profiles of the Le Sueur River (red) and its two main tributaries, the Maple (blue) and Cobb (green) rivers. The locations of the knickpoint, approximately 40 km from the mouth in all three branches, as well as the knickzone, or reach of active incision, are indicated.
drained catastrophically through the proto-Minnesota River 13,400 years ago, incising the Minnesota River valley up to 70 m near the mouth of the Le Sueur.28,29 This incision was experienced by the Le Sueur as a drop in local base level causing a knickpoint, or a sharp increase in channel gradient, at the mouth of the Le Sueur. The knickpoint has propagated 40 km up the Le Sueur River network,30,31 leading to rapid vertical incision (∼5 m/ka) producing valleys with steep river gradients (0.002 m/m), actively eroding bluffs, and ravines. We refer to this expanding, incised reach of the channel network as the knickzone, and the upstream propagating head of the knickzone where the slope discontinuity is observed as the knickpoint (Figure 2). Prior to settlement, the landscape was dominated by tall-grass prairie and wetlands.24 Above the knickpoint the channel network was poorly connected with many areas that drained internally to wetlands and ponds within the watershed. Beginning in the 19th century, an expanding network of surface ditches and subsurface conduit has drained the wetlands and uplands, allowing development of agriculture.32 Beginning in the 1940s, changes in technology and markets led to consolidation of smaller farms into larger operations. These changes are perceived to have initially resulted in more intense and severe agricultural practices, which have more recently evolved into more careful and precise agricultural management.33 However, little quantitative information has been collected to constrain the magnitude of these effects on historical erosion and sediment delivery. Presently, row crops cover as much as 92% of the uplands in the basin.25 The investment required to reduce sediment loading and other nonpoint-source water quality problems is enormous. As an indication, a substantial down payment was made in 2008 when Minnesotans passed a state constitutional amendment that will raise over $3.5 billion in tax revenue over 25 years for the purpose of protecting and restoring water, wildlife, and cultural resources.34 Effective use of the portion of these funds dedicated to reducing sediment pollution will require accurate identification of sources and mechanisms of sediment supply. This issue is controversial, particularly in determining the role of historic and contemporary agricultural practice versus natural erosion processes.35 The stakes are high; management choices will have political, social, and financial implications as well as environmental consequences. In this study, we combine multiple, independent lines of evidence to evaluate sediment sources to the UMR. We use geochemical fingerprinting from Lake Pepin sediment cores to interpret erosional history at the landscape scale and to record 8805
dx.doi.org/10.1021/es2019109 |Environ. Sci. Technol. 2011, 45, 8804–8810
Environmental Science & Technology shifts in the proportion of sediment derived from soil erosion vs near-channel sources. For the purposes of this paper we use the term ‘soil erosion’ to refer to removal of sediment from the vast and relatively flat upland terrestrial surface via sheet wash, shallow gully erosion, and wind erosion. The term ‘near-channel erosion’ is used to refer to processes of channelized fluvial erosion including incision and undercutting of bluffs, banks, and ravines. We also use a sediment budget for the Le Sueur River, a primary contributor of sediment to the Minnesota River, to identify sediment source location and mechanism. We compare the sediment budget constrained for the time period 20002010 with the same averaged over the entire Holocene to provide context for the modern rates under current land use and environmental conditions. The sediment budgets are established for silt and clay (<62 μm), which is the primary contributor to turbidity and deposition in Lake Pepin. Key to our approach is the use of multiple, redundant sources of information to constrain each component of the budget, including aerial lidar analyses, repeat terrestrial lidar scans, geochemical fingerprinting, radiocarbon and optically stimulated luminescence dating, air photo analyses, field surveys, and extensive water and sediment gaging. Details on methods and development of the sediment budgets are presented in the Supporting Information.
’ METHODS Sediment Budgets. The Holocene sediment budget was primarily constrained through analyses of high resolution (1 m) lidar topography data and dating of strath terraces. We hand-digitized polygons of the incised valley and ravines independently. The upper extent of the modern knickzone was defined by the location of discontinuities in loglog plots of local channel gradient versus contributing drainage area for each of the three main tributaries using the Stream Profiler Tool (available for free download from www.geomorphtools.org). Over 600 strath terrace surfaces were mapped from aerial lidar and field surveys. Optically stimulated luminescence and Carbon-14 analyses of terrace alluvium confirmed the timing of initial incision and constrained rates of incision throughout the Holocene. To compute the mass of material in the incised valley and ravines we divided the valley polygons into 3 km sections and assumed a flat surface with an elevation consistent with surrounding topography. Volumes of sediment removed were converted into a mass using a bulk density of 1.8 g/cm3 and 67% silt and clay and were divided by 13,400 years to obtain the long term average rate of mass flux from the knickzone.29 Each component of the 20002010 sediment budget was constrained using multiple lines of information. Annual sediment loads were computed by the Minnesota Pollution Control Agency for all seven gaging stations in the Le Sueur River basin using US Army Corps of Engineers FLUX program. Flow and sediment data were available for all years (20002010) for the gage at the mouth of the watershed. Average 20002010 loads were computed for each of the tributary gages based on available data (39 years depending on the tributary), with missing tributary data estimated relative to the gage at the mouth of the watershed (see the Supporting Information). Bluff erosion rates were constrained over a decadal time scale from air photo analyses. A total of 451 bluffs were identified from air photos, and their crests were manually digitized for multiple years between 1938 and 2005. Bluff toe retreat was independently estimated by measuring channel meander migration rate
ARTICLE
near the toe using the Planform Statistics Tool available online from the National Center for Earth-surface Dynamics Stream Restoration Toolbox (http://www.nced.umn.edu/content/ tools-and-data). Additionally, bluff-related sediment sources were constrained by three years of repeat terrestrial laser scanning covering a wide range of hydrologic years including intermediate, dry, and wet years for 2008, 2009, and 2010, respectively. An Optech ILRIS-36D, ER Terrestrial Lidar Scanner from the Lidar Lab at Western State College of Colorado was used to scan 12 bluffs at resolutions varying from 1700 to 10,000 points per m2. Polyworks metrology software (InnovMetric, Quebec City, Canada) was used to align scans, convert point cloud data to TINs, and digitally remove vegetation. An extensive error analysis was conducted to constrain error due to instrumentation, alignment, TIN creation, erroneous points, and assumptions made differencing scans from multiple years.36 Volumetric bluff erosion rates were computed as the product of bluff height, length, and retreat rate and converted to mass flux using a bulk density of 1.8 g/cm3 and 67% silt and clay. Net local bank erosion rates were computed using the method of Lauer and Parker,37 which computes net, local sediment flux from bank erosion as a function of a measured meander migration rate and the difference in elevation between opposing channel banks. To perform this calculation we used the Planform Statistics Toolbox. Meander migration rates were measured from historic air photo analysis (19382005), discretizing the channel every 20 m, and bank elevations were extracted from 1 m aerial lidar (source: Blue Earth County Environmental Services) using a 10 m buffer on each bank. Bank contributions from channel widening were computed by digitizing banks from historic air photos (19372009), identified by vegetation line, for multiple years and over 14 reaches, each approximately 10 meander bends in length. Width change was converted to volumetric change by computing average depth from a basin-specific downstream hydraulic geometry relationship and assuming that depth remains constant as the channels widen at the observed rate. Volumetric bank erosion rates were converted to mass flux using a bulk density of 1.5 g/cm3 and 50% silt and clay. We used airborne lidar to map ravine locations throughout Blue Earth County and compared historical air photos from 1938 and 2005 to identify recent changes in ravine tip locations. Ravine loads were measured using autosamplers to capture storm events for three monitoring seasons on up to four ravines. Loads were extrapolated to other ravines throughout the watershed based on incised area measured from aerial lidar. See the Supporting Information for data and additional explanation of methods. Geochemical Fingerprinting. We conducted sediment fingerprinting38,39 using naturally occurring radiogenic tracers Beryllium-10 (10Be), Lead-210 (210Pb), and Cesium-137 (137Cs) measured in suspended sediment samples collected at multiple gages within the Le Sueur watershed as well as from Lake Pepin sediment cores. In brief, 10Be and 210Pb are produced in the atmosphere, delivered via rainfall and dry deposition, and are adsorbed to the outside of soil particles within the top 150 and 5 cm of the soil profile, respectively.40,41 Thus, both tracers exhibit relatively high concentrations in sediment eroded from the soil surface and low concentrations in bluff material, which has experienced relatively little exposure to the atmosphere. Cesium-137 was delivered from atmospheric deposition primarily from nuclear testing in the 1950s and 1960s,41,42 and the concentration is also high in upland soils and low in near-channel (bluff, ravine, bank) sediment. 8806
dx.doi.org/10.1021/es2019109 |Environ. Sci. Technol. 2011, 45, 8804–8810
Environmental Science & Technology Because these tracers are radioactive and have very different half-lives (1.4 million yr, 22.3 yr, and 30 yr for 10Be, 210Pb, and 137 Cs, respectively) their concentrations change during transport and storage in the channel-floodplain network.4346 Specifically, 10 Be decay is negligible over the time scales in which channel transport and floodplain storage occur. However, some additional 10Be is added to sediment during floodplain storage. In contrast, 210Pb and 137Cs decay over floodplain residence times (assumed to be 102103 yr), and therefore the upland signature of these tracers is diluted to a degree that depends on the rate of channel floodplain exchange. In this way, 10Be fingerprinting provides an upper constraint and 210Pb and 137Cs provide minimum constraints on the percentage of sediment derived from uplands. Determining a unique solution for sediment apportionment based on combined 10Be, 210Pb, and 137Cs data is not possible at this time for a number of reasons, including number of samples available, insufficient constraints on the variability of atmospheric delivery rates, and the lack of independent constraints on channel-floodplain exchange activity. Therefore, we compute source apportionment for suspended sediment samples based on 10Be results using a two end-member unmixing model (bluffs: [10Be ] = 0.07 ((.01) 108 atoms g1; uplands: [10Be ] = 2.0 ((.36) 108 atoms g1). Suspended sediment samples were also analyzed for 210Pb and 137Cs by alpha and gamma spectroscopy. We collected a total of 28 suspended sediment samples from the Le Sueur River and tributaries during the 2009 field season. Sample activity for 210Pb was corrected for direct deposition, and sediment apportionment for both 210Pb and 137Cs were computed based on an extensive data set from 30 reference lakes to constrain the upland tracer signature, described in detail by Schottler et al.47 See the Supporting Information for data and additional explanation of methods. Additional raw data and GIS shapefiles are available from the corresponding author upon request.
’ RESULTS AND DISCUSSION The well-dated incision history and low-gradient postglacial terrain of the Le Sueur watershed29 allow us to establish an average sediment budget for the Holocene. We used aerial lidar combined with optically stimulated luminescence dating48 to compute the volume of material excavated from the incised valley and ravines over the Holocene (Figure 3, top panel). Valley excavation produced 60,000 Mg/yr, with 80% from bluff erosion (Bl), bank erosion (Ba), and vertical channel incision (C) and 20% from ravine incision (R). A small amount of deposition (Fp) within the incising valley, equivalent to 5000 Mg/yr, is recorded in strath terraces.29 Sediment evacuation modeling based on dated terraces suggests that sediment yield was relatively steady over the Holocene.48 Prior to Euro-American settlement in the early 19th century, sediment supply from the poorly drained, low-gradient uplands of dense perennial grasses and wetlands was assumed negligible relative to that from the incising valley. We developed a sediment budget for the period 20002010 (Figure 3, bottom panel), relying on estimated sediment loads at seven gaging stations, including one at the mouth of the watershed and two on each of the three main branches, with one just upstream from their confluence and the other just above the knickpoint (red triangles in Figure 1). Bluff erosion (Bl) dominates the sediment budget, contributing 26,000 Mg/yr above the knickpoint and 107,000 Mg/yr within the knickzone, with erosion rates more than double the Holocene average.
ARTICLE
Figure 3. Fine sediment budgets (clay + silt) for the Le Sueur River, averaged over Holocene time and over 20002010. Values given in 103 Mg/yr. Sediment sources include Bluffs (Bl), Bank erosion resulting from channel migration (BaM) and widening (BaW), Channel incision (C), Ravines (R), and Uplands (U). Superscripts indicate methods used to constrain each flux (1: Gaging data, 2: Geochemical tracers, 3: Aerial lidar analysis, 4: Terrestrial lidar scans, 5: Air photo analysis, 6: Numerical modeling, 7: Field surveys, 8: OSL and 14C dating).
Ravines comprise only 0.3% of the watershed area but contribute 5% of all sediment, including 1000 Mg/yr above the knickpoint and 12,000 Mg/yr within the knickzone. Based on sediment fingerprinting results, agricultural uplands contribute 45,000 Mg/yr above the knickpoint and add an additional 23,000 Mg/yr within the knick zone. Based on air photo measurements of channel changes since 1938, net bank erosion 37 from meander migration (BaM) and channel widening (BaW) contribute 6000 and 13,000 Mg/yr above the knickpoint, respectively, and 10,000 and 14,000 Mg/yr within the knickzone. Channel incision within the knickzone contributes an additional 4000 Mg/yr. Given the history of agricultural development in the watershed and the legacy of significant valley bottom deposition in areas with a similar history,10,11 the possibility that the modern sediment budget includes erosion of valley bottom legacy sediment must be carefully considered. Precisely constraining floodplain storage (legacy or otherwise) remains a difficult task in large watersheds. Yet, the importance of floodplain storage should not be understated in development of a watershed sediment budget.49 In this study, four separate observations, taken together, support the finding that historic and modern floodplain storage is neither large nor changing considerably under current conditions. First and foremost, floodplain storage is expected to be small in incising systems. Within the knickzone, floodplains are relatively narrow30 and continued, rapid vertical incision limits opportunity for vertical floodplain accretion. Second, we have observed floodplain inundation during multiple events and have mapped small pockets of deposition following a high flow event in September 2010. Specifically, we documented 020 cm of localized deposition, typically less than 10% silt and clay, though this grain size fractionation may have been influenced by conditions specific to this event. In any case, our observations indicate that the channel is not unable to access floodplains but also that floodplain storage is not a large number in any part of 8807
dx.doi.org/10.1021/es2019109 |Environ. Sci. Technol. 2011, 45, 8804–8810
Environmental Science & Technology the channel network. Third, we do not observe large differences in height between eroding and depositing banks above the knickzone, which would otherwise indicate a larger amount of storage having occurred in the past and an ongoing process of net floodplain degradation. Lastly, some overbank deposition occurs behind large woody debris jams above the knickpoint. These debris jams are transient features that appear to form and degrade over annual to decadal time scales. Deposition behind the debris jams appears to be reasonably balanced by bank scour as the channel migrates around the jams. Assuming the frequency and magnitude of these woody debris jams has not changed significantly over recent decades, net sediment storage associated with them has also not changed. While floodplain storage is the leastwell constrained number in our modern sediment budget, these observations all suggest that floodplain storage is neither large nor changing significantly from recent decades. Significant year-to-year variability exists in sediment loading, which translates to variability in contributions from each of the sources. The suspended sediment load for the entire watershed during the driest year (2009) of our monitoring period was a mere 29,000 Mg in contrast to the wettest year (2010), which was 543,000 Mg. Years 2007 and 2008 exhibited intermediate flow, with annual watershed suspended sediment loads of 135,000 and 136,000 Mg, respectively. It was simply fortuitous that the years over which we intensively measured fluxes and erosion rates covered the full spectrum of hydrologic conditions. Defining uncertainty in sediment budgets is a persistent problem. Budgets are typically assembled using a wide range of information of different types, each with their own uncertainty. These include sediment sources from topographic differencing, sediment flux measured at stream gages, and estimates of the proportion of sediment yield derived from terrestrial vs nearchannel sources. Because of the very different nature of these data (sediment supply averaged over large space and time scales, sediment concentration calculated from individual stream samples and extrapolated over time series of flow, and source proportions based on geochemical analyses of individual soil and sediment samples), a formal uncertainty analysis is difficult to define and interpret. Further, the strong constraint of mass balance and the plausible requirement that approximations hold across similar locations and time periods, add certainty to the combined sediment budget that cannot be combined with the uncertainty of individual components in any obvious way. Despite the lack of a formal uncertainty analysis, plausible conclusions can be drawn by invoking sediment mass conservation over multiple time and space scales. Values for each component of the sediment budget, drawn from a plausible range based on the uncertainty in each component, are subject to the collective constraint of mass conservation. For example, a sediment source may be judged to be minor if even exceptionally large values (within the measured range of uncertainty) do not produce a significant fraction of the overall mass balance. Further, the plausible range of uncertainty in any individual component may be constrained if exceptionally large or small values of that component require implausible combinations of other components in order to balance the budget. In the end, a weight-of-evidence approach, constrained by the use of multiple lines of evidence, is used to support conclusions regarding sediment sources, fluxes, and sinks. Here we have closed our budget for the 20002010 time period by applying a single parameter to adjust all predicted source fluxes as a fraction of their respective uncertainty (see the Supporting Information). This approach provides an objective
ARTICLE
Figure 4. Depth profile of Lake Pepin sedimentary record showing sedimentation rate across the entire lake bottom (bottom axis) and concentrations of radionuclide sediment tracers (top axes). Upland soil erosion delivers sediment enriched in both tracers; bluffs and ravines deliver tracer-deficient sediment.
means to balance the budget in a manner that is sensitive to the estimated uncertainty associated with each of the measurements. Based on similar topographic history and land use, the sediment supply observed over the past decade on the Le Sueur is likely representative of other tributaries to the Minnesota River, which collectively account for the bulk of sediment contributed to the UMR and Lake Pepin.23 Sedimentation rates in Lake Pepin (Figure 4) have increased from an apparent ‘background’ rate of ∼80,000 Mg/year prior to 1830 to as high as 850,000 Mg/yr between 1950 and 2008.22 We analyzed natural atmospheric fallout nuclides 210Pb and 10 Be in Lake Pepin sediment cores as tracers that document the relative proportion of fine sediment derived from uplands versus near-channel sources over the past 500 years. Upland sources have high concentrations of both tracers, whereas bluffs and ravines have correspondingly low concentrations (40,41 see the Supporting Information). Both tracers show similar changes over time (Figure 4, red dots for 10Be, orange Xs for 210Pb). The low 10 Be concentration 500 years ago indicates very little upland soil erosion relative to bluff erosion. During the mid-20th century, a sharp increase in 10Be concentrations indicates a pulse of soil erosion from agricultural fields. In the past three decades, both 10 Be and 210Pb document a strong shift back toward nearchannel sources, consistent with our 20002010 sediment budget in the Le Sueur watershed. Because 10Be and 210Pb have very different half-lives (210Pb 22 years; 10Be 1.4 million years), the decrease in both nuclide concentrations indicates a real decrease in field sources relative to bluffs and ravines, rather than recent activation of legacy field sediment stored in the valley bottom, which would be depleted in 210Pb, but not 10Be. Additionally, measured concentrations from source areas confirm that 10Be concentration in upland soils remains high and therefore the decrease in 10Be concentration is not related to prior erosion of nuclide-rich soil, which has been documented in 8808
dx.doi.org/10.1021/es2019109 |Environ. Sci. Technol. 2011, 45, 8804–8810
Environmental Science & Technology parts of the Chesapeake Bay watershed.50 A 210Pb apportionment model47 indicates that the proportion of sediment derived from fields was high in the 1940s-1960s (3.2 pCi g1 ∼ 65% field source), remained relatively high through the 1980s (59% field), and then decreased by the mid-1990s (32% to 35% field). Erosion rates of near-channel sources (particularly bluff and streambanks) are sensitive to changes in river discharge. Two recent hydrologic changes, both related to human activity, may be acting to increase discharge and amplify erosion of nearchannel sources. First, climate records indicate that mean precipitation has increased in Minnesota, along with an increase in the frequency and magnitude of extreme events.51,52 Climate models predict a continued upward trend in the 21st century across the midwestern U.S., primarily in the form of more frequent heavy rains.53 A second important driver is the extensive modification of the channel network with agricultural ditches (25% of the modern network) and subsurface tile drains, particularly over the past 3040 years.54 These modifications have increased both the effective drainage area and the efficiency of drainage. Additionally, tile drains are likely increasing infiltration capacity and thereby reducing surface runoff, which would reduce fluvial soil erosion at the expense of increasing flow in the river, consistent with the apparent decrease in upland soil delivery and increase in near-channel erosion observed over the past few decades. However, at present we are unable to deconvolve the influence of tile drains from other land use and climatic factors that are also contributing to changes in hydrology. The Lake Pepin sediment cores indicate that the rate of sedimentation in the past decade has remained large, even as the sediment supply has shifted from upland to near-channel sources (Figure 4). The Le Sueur 20002010 sediment budget (Figure 3, bottom panel) corroborates a dominant near-channel source of recent sediment supply. The combination of Holocene and modern sediment budgets and geochemical fingerprinting of Lake Pepin sediment cores provide strong evidence that the dominant source of persistent large sediment loads has shifted from agricultural soil erosion to amplified near-channel erosion, driven by a combination of changing precipitation and a vastly altered drainage network. Our results indicate that 70% of sediment contributed to the Le Sueur River during 20002010 is derived from the channel network (bluffs, banks, ravines, and channel incision), which comprises less than 1% of the landscape. This finding underscores the need for development of combined watershed erosion and sediment routing models that account for channel adjustments and changes in channel-floodplain storage in response to changes in flow and the amount and type of sediment supply. Such models are imperative for understanding sediment as a water quality metric, especially under nonstationary climate conditions, and for predicting the effectiveness of remedial actions. Such models will be strengthened by increasing availability of high resolution topography data that includes channel bathymetry, providing opportunities for development of 2D and 3D models of flow, erosion, and deposition. Incorporating theory for production and decay of radiogenic tracers associated with the sediment would provide a platform for hypothesis testing and model validation that would allow us to move beyond conventional empirical approaches for prediction of watershed sediment yield. The research, management, and policy challenges posed by these findings are imposing: both the source and driving
ARTICLE
mechanism of elevated sediment loads are changing. Effective remediation must now accommodate both erosion and runoff controls. Land and water resource management must develop watershed-scale solutions to mitigate systemic hydrologic amplification of natural erosion processes.
’ ASSOCIATED CONTENT
bS
Supporting Information. All supporting data used to constrain the Holocene and 20002010 sediment budgets, including synthesis of sediment excavation over Holocene time, a summary of modern sediment gaging data on main channels and ravines, air photo analysis of channel widening and channel migration, explanation of sediment fingerprinting sample processing and results, and summary data for air photo and repeat terrestrial lidar analysis of bluff erosion. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (435) 881-7697. E-mail:
[email protected].
’ ACKNOWLEDGMENT This work was funded by the National Center for Earthsurface Dynamics NSF EAR 0120914, Minnesota Pollution Control Agency, and Minnesota Department of Agriculture. Additional support was provided by the Utah State University Agricultural Experiment Station (paper #8288), the University of Minnesota Limnological Research Center, and Minnesota Supercomputing Institute. Aerial lidar data were provided by the Environmental Services Department of Blue Earth County, Minnesota. J.K.W. thanks the Alexander von Humboldt foundation for a postdoctoral fellowship. We also thank three anonymous reviewers for constructive comments on the manuscript and Julia Marquard, Noah Finnegan, Caitlyn Echterling, Jenny Graves, Khalif Maalim, Luam Azmera, Ashley Thomas, Bridget Wolfe, Mandy Kinnick, and James Ley for their assistance on the project. ’ REFERENCES (1) US Environmental Protection Agency. Water Quality Assessment and TMDL Information. 2011. http://iaspub.epa.gov/waters10/ attains_nation_cy.control (accessed March 6, 2011). (2) Palmer, M. A.; et al. Linkages between aquatic sediment biota and life above sediments as potential drivers of biodiversity and ecological processes. BioScience 2000, 50, 1062–1075. (3) Walling, D. E. The sediment delivery problem. J. Hydrol. 1983, 65, 209–37. (4) Trimble, S. W.; Crosson, P. U.S. Soil Erosion Rates Myth and Reality. Science 2000, 289, 248–250. (5) Langland, M. J.; Moyer, D. L.; Blomquist, J. Changes in streamflow, concentrations, and loads in selected nontidal basins in the Chesapeake Bay Watershed, 19852006. U.S. Geol. Surv. Open File Rep. 2007, 1372, 76. (6) Smith, S. M. C.; Belmont, P.; Wilcock, P. R. Closing the gap between watershed modeling, sediment budgeting, and stream restoration. In Stream Restoration in Dynamic Systems: Scientific Approaches, Analyses, and Tools. Simon, A., Bennett, S. J., Castro, J., Thorne, C., Eds.; AGU: Washington, D. C., 2011 (doi:10.1029/2011GM001085). (7) Montgomery, D. R. Soil erosion and agricultural sustainability. Proc. Natl. Acad. Sci. 2007, 104, 13268–13272. (8) De Vente, J.; Poesen, J.; Arabkhedri, M.; Verstraeten, G. The sediment delivery problem revisited. Prog. Phys. Geog. 2007, 31, 155–178. 8809
dx.doi.org/10.1021/es2019109 |Environ. Sci. Technol. 2011, 45, 8804–8810
Environmental Science & Technology (9) Phillips, J. D. Evolutionary geomorphology: thresholds and nonlinearity in landform response to environmental change. Hydrol. Earth Syst. Sci. 2006, 10, 731–742. (10) Trimble, S. W. Decreased rates of alluvial sediment storage in the Coon Creek Basin, Wisconsin, 197593. Science 1999, 285, 1244–1246. (11) Walter, R.; Merritts, D. Natural streams and the legacy of waterpowered milling. Science 2008, 319, 299–304. (12) Nearing, M. A.; et al. Measurements and Models of Soil Loss Rates. Science 2000, 290, 1300b. (13) Allen, P. A. From landscapes into geological history. Nature 2008, 451, 274–276. (14) Quinton, J. N.; Govers, G.; Van Oost, K.; Bardgett, R. D. The impact of agricultural soil erosion on biogeochemical cycling. Nat. Geosci. 2010, 3, 311–314. (15) Reid, L. M.; Dunne, T. Rapid Evaluation of Sediment Budgets; Catena Verlag, Reiskirchen, Germany, 1996; 164 pp. (16) Gilbert, G. K. Hydraulic-mining debris in the Sierra Nevada. U.S. Geol. Survey Prof. Paper 1917, 105, 154. (17) Dietrich, W. E.; Dunne, T. Sediment budget for a small catchment in mountainous terrain. Z. Geomorphol., Suppl. 1978, 29, 191–206. (18) Costa, J. E.; Cleaves, E. T. The Piedmont landscape of Maryland: A new look at an old problem. Earth Surf. Processes Landforms 1984, 9, 59–74. (19) Trimble, S. W. Man-induced soil erosion on the southern Piedmont. Soil and Water Conservation Society: Ankeny, IA, 2008. (20) Snyder, N. P. Studying stream morphology with airborne laser elevation data. EOS, Trans., Am. Geophys. Union 2009, 90 (6), 45-46. doi:10.1029/2009EO060001. (21) Wheaton, J. M.; Brasington, J.; Darby, S. E.; Sear, D. Accounting for Uncertainty in DEMs from Repeat Topographic Surveys: Improved Sediment Budgets. Earth Surf. Processes Landforms 2010, 35 (2), 136156. DOI: 10.1002/esp.1886. (22) Engstrom, D. R.; Almendinger, J. E.; Wolin, J. A. Historical changes in sediment and phosphorus loading to the upper Mississippi River: mass-balance reconstructions from the sediments of Lake Pepin. J. Paleolimnol. 2009, 41, 563–588. (23) Kelley, D. W.; Brachfeld, S. A.; Nater, E. A.; Wright, H. E., Jr. Historical changes in sediment and phosphorus loading to the upper Mississippi River: mass-balance reconstructions from the sediments of Lake Pepin. J. Paleolimnol. 2006, 35, 193–206. (24) Minnesota County Biological Survey (MCBS), Minnesota Department of Natural Resources. Native Plant Communities and Rare Species of The Minnesota River Valley Counties, Minnesota Department of Natural Resources. September 2007. http://www.mndnr.gov/eco/ mcbs/index.html (accessed March 3, 2011). (25) Musser, K.; Kudelka, S.; Moore, , R. Minnesota River Basin Trends Report. 2009; 66 pp. http://mrbdc.wrc.mnsu.edu/mnbasin/ trends/index.html (accessed March 10, 2011). (26) Wilcock, P. Synthesis Report for Minnesota River Sediment Colloquium. Report for MPCA, 2009. http://www.lakepepinlegacyalliance. org/SedSynth_FinalDraft-formatted.pdf (accessed March 13, 2011). (27) Jennings, C. E. Open File Report 10-03, Minnesota Geological Survey, map, report and digital files. 2010. ftp://mgssun6.mngs.umn.edu/ pub4/ofr10_03/ (accessed March 13, 2011). (28) Clayton, L.; Moran, S. R. Chronology of late-Wisconsinan glaciations in middle North America. Quat. Sci. Rev. 1982, 1, 55–82. (29) Gran, K. B.; et al. Management and restoration of fluvial systems with broad historical changes and human impacts. GSA Sp. Paper 2009, 451, 119–130. (30) Belmont, P. Floodplain width adjustments in response to rapid base level fall and knickpoint migration. Geomorphology 2011, 128, 92–102. (31) Gardner, T. W. Experimental study of knickpoint and longitudinal profile evolution in cohesive, homogenous material. GSA Bull. 1983, 94, 664–672. (32) Fry, P. A River runs through it: Cultures, economies and landscapes within the Minnesota River Basin to 1900. Ph.D. Dissertation. University of Minnesota. 1999; 288 pp.
ARTICLE
(33) Andersson, O.; Crow, T. R.; Lietz, S. M.; Sterns, F. Transformation of a landscape in the upper mid-west, USA: The history of the lower St. Croix river valley, 1830 to present. Landscape and Urban Planning 1996, 35 (4), 247–267. (34) Additional information available on Clean Water Funds at http://www.pca.state.mn.us/ (accessed August 1, 2011). (35) Star-Tribune, Seize momentum on water quality. MinneapolisSt. Paul Star Tribune Editorial, December 26, 2010. (36) Day, S. S.; Gran, K. B.; Belmont, P.; Wawrzyniec, T. Change detection on bluffs using terrestrial laser scanning technology. Revision submitted to Earth Surf. Processes Landforms, in review. (37) Lauer, J. W.; Parker, G. Modeling framework for sediment deposition, storage, and evacuation in the floodplain of a meandering river, Part I: Theory. Water Resour. Res. 2008, 44, W04425. (38) Collins, A. L.; Walling, D. E. Selecting fingerprint properties for discriminating potential suspended sediment sources in river basins. J. Hydrol. 2002, 261, 218–244. (39) Davis, C. M.; Fox, J. F. Sediment Fingerprinting: Review of the method and future improvements for allocating nonpoint source pollution. J. Environ. Eng. 2009, 135 (7), 490–504. (40) Willenbring, J. K.; von Blanckenburg, F. Meteoric cosmogenic Beryllium-10 adsorbed to river sediment and soil: applications for Earthsurface dynamics. Earth Sci. Rev. 2010, 98, 105–122. (41) Quaternary Geochronology: Methods and Applications; Noller, J. S., Soweres, J. M., Lettis, W. R., Eds.; American Geophysical Union: Washington, DC, AGU Reference Shelf, 2002; Vol. 4, 582 p. (42) Owens, P. N.; Walling, D. E.; He, Q. The behavior of bombderived Caesium-137 fallout in catchment soils. J. Environ. Radioact. 1996, 32 (3), 169–191. (43) Bonniwell, E. C.; Matisoff, G.; Whiting, P. J. Determining the times and distances of particle transit in a mountain stream using fallout radionuclides. Geomorphology 1999, 27, 75–92. (44) Matisoff, G.; Bonniwell, E. C.; Whiting, P. J. Radionuclides as indicators of sediment transport in agricultural watersheds that drain to Lake Erie. J. Environ. Qual. 2002, 31, 62–72. (45) Lauer, J. W.; Willenbring, J. Steady state reach-scale theory for radioactive tracer concentration in a simple channel/floodplain system. J. Geophys. Res. Earth Surf. 2010, 115, F04018. (46) Viparelli, E.; Lauer, J. W.; Belmont, P.; Parker, G. A Numerical Model to Develop Long-term Sediment Budgets Using Isotopic Sediment Fingerprints. Computers and Geosciences Special issue on Modeling for Environmental Change, in review. (47) Schottler, S. P.; Engstrom, D. R.; Blumentritt, D. Fingerprinting sources of sediment in large agricultural river systems. Report for MPCA. 2010. http://www.smm.org/static/science/pdf/scwrs-2010fingerprinting.pdf (accessed March 12, 2011. (48) Belmont, P.; Gran, K. B.; Jennings, C.; Wittkop, C. Holocene Landscape Evolution and Erosional Processes in the Le Sueur River, Central Minnesota. Kirk Bryan Field Trip Guide for 2011 Geological Society of America Annual Meeting. 2011. (49) Walling, D. E.; Owens, P. N.; Leeks, G. J. L. The role of channel and floodplain storage in the suspended sediment budget of the River Ouse, Yorkshire, UK. Geomorphology 1998, 22, 225–242. (50) Valette-Silver, J. N.; Brown, L.; Pavich, M.; Klein, J.; Middleton, R. Detection of erosion events using 10 Be profiles: example of the impact of agriculture on soil erosion in the Chesapeake Bay area USA. Earth Planet. Sci. Lett. 1986, 80, 82–90. (51) Novotny, E. V.; Stefan, H. G. Stream flow in Minnesota: Indicator of climate change? J. Hydrol. 2007, 334, 319–333. (52) Min, S.; Zhang, X.; Zwiers, F. W.; Hegerl, G. C. Human contribution to more-intense precipitation extremes. Nature 2011, 470, 378–381. (53) USGCRP. Global Climate Change Impacts in the United States. Cambridge University Press: New York, 2009. (54) Blann, K. L.; Anderson, J. L.; Sands, G. R.; Vondracek, B. Effects of agricultural drainage on aquatic ecosystems: a review. Crit. Rev. Environ. Sci. Technol. 2009, 39, 909–1001. 8810
dx.doi.org/10.1021/es2019109 |Environ. Sci. Technol. 2011, 45, 8804–8810
ARTICLE pubs.acs.org/est
Fate of Endogenous Steroid Hormones in Steer Feedlots Under Simulated Rainfall-Induced Runoff D. Scott Mansell,† Reid J. Bryson,‡ Thomas Harter,‡ Jackson P. Webster,§ Edward P. Kolodziej,§ and David L. Sedlak*,† †
Department of Civil and Environmental Engineering, University of California, Berkeley, California 94720-1710, United States Department of Land, Air, and Water Resources, University of California, Davis, California 95616-8628, United States § Department of Civil and Environmental Engineering, University of Nevada, Reno, Nevada 89557-0258, United States ‡
bS Supporting Information ABSTRACT: Steroid hormones pose potential risks to fish and other aquatic organisms at extremely low concentrations. To assess the factors affecting the release of endogenous estrogenic and androgenic steroids from feedlots during rainfall, runoff, and soil samples were collected after simulated rainfall on a 14-steer feedlot under different rainfall rates and aging periods and analyzed for six steroid hormones. While only 17α-estradiol, testosterone, and progesterone were detected in fresh manure, 17β-estradiol, estrone, and androstenedione were present in the surficial soil after two weeks. In the feedlot surficial soil, concentrations of 17α-estradiol decreased by approximately 25% accompanied by an equivalent increase in estrone and 17β-estradiol. Aging of the feedlot soils for an additional 7 days had no effect on estrogen and testosterone concentrations, but androstenedione concentrations decreased substantially, and progesterone concentrations increased. Androstenedione and progesterone concentrations in the surficial soil were much higher than could be accounted for by excretion or conversion from testosterone, suggesting that other potential precursors, such as sterols, were converted after excretion. The concentration of androgens and progesterone in the soil were approximately 85% lower after simulated rainfall, but the estrogen concentrations remained approximately constant. The decreased masses could not be accounted for by runoff, suggesting the possibility of rapid microbial transformation upon wetting. All six steroids in the runoff, with the exception of 17β-estradiol, were detected in both the filtered and particle-associated phases at concentrations well above thresholds for biological responses. Runoff from the aged plots contained less 17α-estradiol and testosterone, but more estrone, androstenedione, and progesterone relative to the runoff from the unaged plots, and most of the steroids had a lower particle-associated fraction.
’ INTRODUCTION During the past decade, estrogens, androgens, and progestogens have been detected in surface waters at concentrations high enough to affect fish and other aquatic organisms. In particular, attention has been focused on the feminization of fish in surface waters receiving discharges of steroid-containing municipal wastewater effluent.1 3 Exposure of sensitive species of fish to estrogenic hormone concentrations as low as 1 10 ng/L can result in biological changes, decreased reproductive success and alterations in species composition in receiving waters.1,4 Androgens and progestogens have also been detected in surface waters subjected to municipal wastewater effluent discharges5 7 at concentrations known to elicit behavioral and biochemical responses in fish.7 9 In addition to wastewater effluent, animal husbandry activities in which large numbers of animals are maintained in a small area may be important sources of steroid hormones in some watersheds. Concentrated animal feeding operations,6,10,11 grazing cattle,12 dairies,13 and aquaculture facilities13 all have the potential to r 2011 American Chemical Society
release steroid hormones to surface waters. In watersheds where runoff from feedlots, pastures, or land to which manure has been applied is not sufficiently treated or diluted, steroid hormones could reach concentrations that pose risks to aquatic ecosystems. However, the diffuse or intermittent nature of agricultural runoff coupled with the transport of hormones in association with particles and organic matter make it difficult to predict steroid hormone concentrations in surface waters impacted by agriculture. Previous studies have demonstrated the presence of steroid hormones in runoff and waste storage lagoons at dairies and beef cattle feedlots at concentrations up to several orders of magnitude higher than reported thresholds for biological responses.13 15 However, these wastes often receive some form of treatment before reaching surface waters. The most common forms of treatment Received: June 17, 2011 Accepted: August 24, 2011 Revised: August 20, 2011 Published: September 06, 2011 8811
dx.doi.org/10.1021/es202072f | Environ. Sci. Technol. 2011, 45, 8811–8818
Environmental Science & Technology include application to cropland and pastures, infiltration through soil, composting, and treatment in ponds or wetlands. Irrespective of whether or not runoff is treated prior to release, management decisions require an understanding of the fraction of the steroid hormones excreted by animals that is released to surface waters. In cattle feedlots, steroid hormones in runoff only account for a small fraction of the mass of the excreted steroids.16,17 The steroid hormones that are not transported in runoff are assumed to be transformed by microbes in the soil and waste. Results from lab experiments in which manure and/or soil were stored or incubated under moist, aerobic conditions indicate that most endogenous steroids are rapidly transformed to other biologically active steroids or compounds with little affinity for steroid receptors with half-lives on the order of hours to days.10,18 21 However, some studies indicate that the steroid hormones can persist much longer during manure storage.16 In addition, most steroids exhibit a relatively high affinity for soils, retarding their movement in runoff and groundwater and increasing the importance of transport of particle-associated steroids in surface waters.18,21,22 As a result of the important role of natural attenuation and particle-associated transport in the release of steroid hormones to surface waters, it is important to understand the role of factors such as soil moisture, aging, and phase partitioning on the transformation and transport of steroids in feedlots. Because some transformation products exhibit biological activity, it is also important to identify steroids formed in animal wastes after excretion. To gain insight into these processes, field studies were conducted using rainfall simulators in beef cattle feedlot pens under conditions representative of commercial animal feeding operations.
’ MATERIALS AND METHODS Materials. All reagents except acetonitrile were purchased from Fisher Scientific (Fairlawn, NJ) at the highest possible purity. Acetonitrile (99.8%), steroids, and heptaflourobutyric anhydride were purchased from Sigma Aldrich (St Louis, MO) also at the highest possible purity. Deuterated steroids were purchased from CDN isotopes (Quebec, Canada) except for estrone-2,4,16, 16-d4 which was purchased from Isotec (Miamisburg, OH). Experimental Design. Experiments were conducted during the summer and fall of 2009 at the University of California’s Animal Science Feedlot in Davis, CA. Weather conditions for this period are summarized in the Supporting Information. Fourteen steers were housed in two, 190 m2 pens with 33% shade cover for 14 days prior to initiating the experiments. The pens were scraped to the clay layer and graded at 3% slope prior to the introduction of animals. The animals were implanted with Revalor S (Intervet Inc. Millsboro, DE), an implant which contains 120 mg of trenbolone acetate and 24 mg estradiol, either 15 or 8 days before they were placed in the pens, providing a steady release of steroid hormones to the animals. Data on the concentration of trenbolone and its metabolites in the animal waste and runoff will be reported elsewhere (Webster et al., manuscript in preparation). After 14 days, the cattle were removed from the pens where a new soil layer of mostly dried manure solids (mixed with soil particles from the dense clay layer below) had developed. Rainfall was simulated on eight, 3-m2 subplots within each pen using a rainfall simulator described previously.23,24 Briefly, a 1/2 in.
ARTICLE
nozzle (Spraying Systems Co, Carol Stream, IL) held on an aluminum frame 3 m above the ground provided uniform rainfall over a 2 m 1.5 m plot (see Supporting Information Figures SI1 and SI2). A solenoid valve and timer were used to control the intensity of the rainfall. The simulator provided a drop size, velocity, and uniformity close to that of rainfall.23,24 Reverse osmosis water (Applied Membranes Inc., Vista, CA) used in the simulator was equilibrated overnight in an open 500 gallon (1890 L) container prior to rainfall simulation. The simulator was operated at a rainfall intensity of either 15 mm/h (low rate) or 25 mm/h (high rate). Four plots received the low rainfall rate and four received the high rate within each pen. Experiments in which simulated rainfall was applied immediately after the cattle were removed from the pen were repeated twice. After these two experiments, an additional experiment was conducted on the same pens in which the wastes were aged an additional 7 days after removing the cattle before simulating rainfall. The pens were scraped to the packed clay layer between each experiment. Prior to rainfall application, an aluminum frame was placed in the ground around each plot to capture runoff and channel it into an aluminum pipe where samples were collected. Runoff samples were collected 5, 20, 40, and 60 min following initiation of runoff. The runoff flow rate was measured by timing flow into a 250 mL container. The samples at each time point from four replicate plots were composited. This process was repeated at both rainfall rates in both pens. For each experiment, 10 12, 100 g (wet weight) grab samples of the topsoil and the clay layer were collected in plastic bags and composited prior to and immediately following simulated rainfall. A grab sample of fresh manure from one of the pens was also collected. Runoff samples were transported on ice to UC Berkeley for analysis. All liquid samples were extracted within 48 h of collection. Solids samples were frozen, homogenized at UC Davis, and shipped on ice to UC Berkeley where they were kept frozen until extraction which occurred within 3 months. Chemical Analysis. Total suspended solids (TSS) (2540 D), fraction organic matter of the suspended solids (fom) (2540E), moisture content and fraction organic matter of the soil (2540 G), dissolved organic carbon (DOC) (5310 B), chloride (4500-Cl F), and nitrate (4500-NO3 C) were analyzed using standard methods.25 DOC was analyzed using a Shimadzu TOC 5000A analyzer. Chloride and nitrate concentration were analyzed using a Dionex DX-120 ion chromatograph with an Ionpac AS14A 4 250 mm column. Steroid Hormone Analysis. Steroid hormones were analyzed using solid phase extraction, Florisil cleanup, derivatization, and gas chromatography tandem mass spectrometry (GC/MS/MS) with isotope dilution. A surrogate standard consisting of 100 ng each of 17α-estradiol-2,4-d2, 17β-estradiol-2,4,16,17,17-d5, estrone-2,4,16,16-d4, 4-androsten-3,17-dione-2,2,4,6,6,16,16-d7, testosterone-16,16,17-d3, progesterone-2,2,4,6,6,-17α-21,21,21-d9 and mesterolone in acetonitrile was added to each runoff sample, suspended solids sample, and soil sample. Runoff samples (1 L) were centrifuged at 5000 rpm (4620g) for 10 min to remove suspended solids. The pellet was frozen until extraction. The supernatant was filtered through a 1 μm, glass fiber filter (Millipore, Bellerica, MA), the surrogate standard was added, and the sample was extracted on a Supelco ENVI-18 solid phase extraction cartridge (Sigma Aldrich, St. Louis, MO). The cartridge was eluted with 12 mL methanol and dried at 60 °C in a vacuum oven. The sample was resuspended in 2 mL 95% dichloromethane/5% methanol and passed through a Supelco 8812
dx.doi.org/10.1021/es202072f |Environ. Sci. Technol. 2011, 45, 8811–8818
Environmental Science & Technology LC-Florisil cartridge (Sigma Aldrich, St Louis, MO) to remove polar organic matter. The cartridge was eluted with 20 mL of 95% dichloromethane/5% methanol which was collected, added to the 2 mL of solvent which had already passed through the cartridge, and dried in a vacuum oven. When the solvent remaining was less than 1 mL, the sample was transferred to a 2 mL gas chromatography vial (Grace, Deerfield, IL), dried under vacuum, and resuspended in 200 μL acetonitrile. The extracts were derivatized using 50 μL heptaflourobutyric anhydride followed by heating at 80 °C for 90 min. The samples were then dried under a gentle stream of nitrogen and resuspended in 100 μL isooctane with 1 ng/μL hexachlorobenzene as an internal standard. Steroids were analyzed by GC/MS/MS as described previously.12 The method was modified to include 17α-estradiol-2,4-d2, 4-androsten-3,17-dione-2,2,4,6,6,16,16-d7, and testosterone-16,16,17-d3 as additional surrogate standards. The parent ion used to identify 17α-estradiol-2,4-d2 was 666, and the two daughter ions were 239 and 452. The parent ion used to identify 4-androsten-3,17-dione-2,2,4,6,6,16,16-d7 was 488, and the two daughter ions were 473 and 275. The parent ion used to identify testosterone-16,16,17-d3 was 683, and the two daughter ions were 668 and 320. The parent and daughter ions for all of the measured steroids are listed in Supporting Information Table SI1. For soil-associated steroids, approximately 1 2 g (wet weight) of soil was weighed into a 15 mL centrifuge tube and the surrogate standard was added. A 10-mL aliquot of methanol was then added and the tubes were shaken by hand for 20 s followed by sonication for 15 min. They were then centrifuged at 5000 rpm (3836g) and the supernatant was decanted. A 10 mL aliquot of methanol was added and the extraction process was repeated two more times. 70 mL of deionized water was added to the 30 mL of methanol and the solution was filtered through an AP40 glass fiber filter. The solution was then extracted on a C-18 SPE cartridge followed by Florisil cleanup, derivatization, and GC/MS/MS analysis as described above. Particle-associated steroids were analyzed in a similar manner except that the whole pellet after centrifugation of the runoff was extracted in 500 mL centrifuge bottles using three times the amount of methanol and deionized water. Quality Assurance/Quality Control. To ensure accurate and precise analysis in the complex matrix, blanks and matrix recoveries were included with each set of samples. Steroids were never detected above quantification limits in blanks. For one time point per hydrograph, triplicate samples were collected and analyzed. Two samples were analyzed as duplicates, and the third was used as a matrix spike. Matrix spikes of filtered and suspended phases after separation were amended with 100 ng of 17α-estradiol, 17β-estradiol, estriol, estrone, testosterone, androstenedione, and progesterone in 100 μL acetonitrile prior to analysis. One of every six soil samples also served as matrix spikes. Despite all precautions, the complex matrix and multiple sample handling steps occasionally resulted in incomplete recovery of matrix spikes or altered instrument responses. The mean spike for each compound for each round was calculated from all the individual spikes, and, in approximately 11% of these, exceptionally high or low recoveries were observed, most likely due to the presence of high concentrations of NOM or colloids. To avoid systematic errors, only data from groups of samples where mean spike recoveries between 75 and 140% were observed were used for all compounds except 17α-estradiol. For this compound, the deuterated surrogate standard was not added to any of the runoff
ARTICLE
Figure 1. Mean steroid hormone concentrations (ng/g dry weight) in fresh manure and surficial soil before and after aging 7 days and after simulated rainfall. 17αE2 = 17α-estradiol, 17βE2 = 17β-estradiol, E1 = estrone, T = testosterone, AD = androstenedione, and PR = progesterone. Error bars represent ( standard error.
samples or the first round of soil samples, so 17β-estradiol2,4,16,17,17-d5 was used as a surrogate. As a result, spike recoveries for 17α-estradiol exhibited greater variability and spike recoveries between 75 and 180% were considered to be acceptable. The relative percent difference (RPD) of the duplicate samples was calculated as the difference between the two samples divided by the mean multiplied by 100. The RPD varied with the measured steroid concentration, with higher RPD values at lower concentrations (Supporting Information Figure SI3). Median RPDs were 18%, 21%, and 7% for filtered samples with steroid concentrations above 10 ng/L and suspended solids and soil samples with steroids concentrations above 10 ng/g. However, median RPDs were 17%, 31%, and 21% for filtered, suspended, and soil samples below those concentrations, respectively. The quantification limit was between 1 and 10 ng/L for filtered runoff, 0.5 and 10 ng/g for suspended solids, and 0.5 and 5 ng/g for soil and manure depending on the steroid (see Supporting Information Table SI1 for a summary). For samples with detectable steroid concentrations below the quantification limits, half the limit of quantification was used in calculations.
’ RESULTS The concentrations of steroid hormones in the fresh manure ranged from below detection limit to 15 ng/g (dry weight) (Figure 1). The concentration of testosterone in fresh manure (2 ng/g) agreed with published values for median steroid excretion rates for veal calves17 (2.8 ng/g) which has been used to estimate endogenous steroid excretion by steers.16,26 Progesterone was detected in the fresh manure, but below the limit of quantification. While Arts et al. did not report the presence of progesterone in manure, they did detect the steroid in the urine of veal calves.17 The concentration of 17α-estradiol (15 ng/g) was approximately five times higher than the median reported value in Arts et al. (3 ng/g), presumably because some of the 17β-estradiol in the implants was metabolized to 17α-estradiol by the cattle.27,28 The water content of the fresh manure was 79%. The surficial soil sampled at the initiation of the experiment consisted of manure deposited during the 14-day period in which 8813
dx.doi.org/10.1021/es202072f |Environ. Sci. Technol. 2011, 45, 8811–8818
Environmental Science & Technology
ARTICLE
Table 1. Mean Total Suspended Solid (TSS), Percent Organic, Dissolved Organic Carbon (DOC), and Chloride Concentrations for the Runoff Collected. Average ( Standard Error. For Changes with Time, See Supporting Information Figure SI4 TSS (g/L)
percent organic in TSS (%)
DOC (mg/L)
chloride (mg/L)
From initial plots, high rainfall rate
5.9 ( 0.4
34 ( 1
498 ( 48
155 ( 26
From initial plots, low rainfall rate From aged plots, high rainfall rate
4.6 ( 0.3 3.3 ( 0.2
36 ( 2 31 ( 2
636 ( 70 509 ( 52
215 ( 36 398 ( 45
From aged plots, low rainfall rate
2.4 ( 0.1
29 ( 1
645 ( 55
464 ( 44
the cattle were present mixed with clay and feed by the action of the animals walking in the pen. The top 3 cm of surficial soil above the packed clay layer had a water content between 3 and 6% and was between 19 and 25% organic matter on a dry weight basis. After aging, the organic matter content was between 25 and 32%. The water content of the soil did not change significantly during the 7-day period after the animals were removed, so all drying must have taken place during the 14-day period in the which the cattle were present. All six steroids were detected in the surficial soil. Concentrations of 17α-estradiol and testosterone were lower in the soil than in the manure, but concentrations of all other hormones were substantially higher (Figure 1). In particular, androstenedione and progesterone increased from concentrations near or below detection limits to concentrations up to approximately 60 ng/g. The mean concentration of androstenedione decreased by approximately 70% during the 7-day period after the cattle were removed whereas the mean concentration of progesterone increased by approximately 30% (Figure 1). Samples collected from the clay layer had concentrations of steroid hormones near or below the limit of quantification at all times with concentrations never exceeding 1 ng/g with the exception of androstenedione and progesterone, which were detected at concentrations up to 11 ng/g and 6 ng/g, respectively (Supporting Information Table SI2). Soil samples collected after simulated rainfall showed little or no change in concentration of estrogens (i.e., 17α-estradiol, 17βestradiol, and estrone) while concentrations of androstenedione, progesterone, and testosterone decreased by approximately 85% after simulated rainfall (Figure 1). The water content of the soil increased from 3 to 6% before rainfall to 44% ( 3% after simulated rainfall. In the runoff, concentrations of chloride and dissolved organic carbon decreased with runoff duration during the experiments while suspended solids concentrations remained approximately constant (Supporting Information Figure SI4). The higher rainfall rate resulted in lower chloride and dissolved organic carbon concentrations, but higher suspended solids concentrations (Table 1 and Supporting Information Figure SI4). The organic matter content of the suspended solids was similar to that of the soil (i.e., the suspended solids consisted of 23 to 48% organic matter compared to 19 32% for the soil). While the relationship between the suspended solids concentration and runoff rate was weak (r2 = 0.32), the suspended solids concentrations were higher from plots receiving the higher rainfall rate (p = 0.005). Aging caused a decrease in the average suspended solids concentration in the runoff from approximately 5 g/L to 3 g/L (p = 0.007). Aging also increased the concentration of chloride in the runoff at both rainfall rates. However, this may be because the same pens were used for each experiment and excess salts were not removed when the pens were scraped between experiments.
Figure 2. Mean steroid hormone concentrations in runoff from unaged and aged plots after simulated rainfall. 17αE2 = 17α-estradiol, 17βE2 = 17β-estradiol, E1 = estrone, T = testosterone, AD = androstenedione, and PR = progesterone. Error bars represent ( standard error.
All six steroids were detected consistently in the runoff with concentrations remaining approximately constant throughout the hydrograph and showing no significant correlation with rainfall rate (Supporting Information Figure SI5). Therefore, comparisons among treatments were made by averaging all time points in the hydrograph at both rainfall rates (Figure 2). Concentrations of 17α-estradiol, androstenedione, and progesterone ranged from 50 to 250 ng/L whereas concentrations of 17β-estradiol, estrone, and testosterone were usually less than 50 ng/L. With the exception of 17β-estradiol, a significant portion of each steroid was detected in both the filtered and particleassociated phases. Aging affected both the concentrations and partitioning of each steroid. For most of the steroids, a higher fraction of the steroids were present in the filtered phase of the runoff from the aged plots where TSS concentrations were approximately 45% lower. The runoff from the aged plots contained less 17α-estradiol and testosterone, but more androstenedione, progesterone, and estrone relative to the unaged plots.
’ DISCUSSION Mass Balance. A mass balance estimate for the plots was made using published data on hormone excretion rates,17,26,29 along with measured hormone concentrations in manure, soil, and runoff (see Supporting Information for details). The mass balance calculation suggests that approximately a quarter of the 8814
dx.doi.org/10.1021/es202072f |Environ. Sci. Technol. 2011, 45, 8811–8818
Environmental Science & Technology
Figure 3. Mass balance estimate on steroid hormones. 17αE2 = 17αestradiol, 17βE2 = 17β-estradiol, E1 = estrone, T = testosterone, AD = androstenedione, and PR = progesterone.
17α-estradiol excreted by the cattle was converted into estrone and 17β-estradiol during the 14-day period in which the cattle were present in the pen (Figure 3). Aging of soil after the animals were removed resulted in a small increase in the mass of estrone and 17β-estradiol, and simulated rainfall had little effect on the total mass of estrogens in the soil (Figure 3). The mass of testosterone decreased by approximately 70% during the 14-day period in which the cattle were present. No further loss was observed during the 7-day period after the animals were removed (Figure 3). After rainfall, approximately 5% of the testosterone excreted by the cattle remained in the soil. Removal of testosterone from the plot via runoff accounted for less than 1% of the missing mass (Figure 3). Unlike the estrogens and testosterone, androstenedione and progesterone exhibit a different pattern. Androstenedione was not detected in the fresh manure, but had the highest mass of any of the steroids in the soil after the 14-day period in which the cattle were present. During the 7-day period after the cattle were removed, the mass decreased by approximately 75% (Figure 3). The mass of progesterone increased by roughly an order of
ARTICLE
magnitude while the cattle were present, and increased further during the aging period after the animals were removed (Figure 3). Although 83 97% of the masses of these two steroids were lost after simulated rainfall, only a small fraction of the masses of these two steroids was recovered in the runoff. Estrogens. Under the relatively dry conditions of the feedlot soil, the estrogens were surprisingly stable. While there was an initial decrease in the concentration of 17α-estradiol accompanied by an increase in the concentration of estrone and 17βestradiol during the 14-day period when the cattle were present, the concentrations remained essentially unchanged during the 7-day aging period and in the brief period after water was added in the simulated rainfall (Figure 1). Estrone is a metabolite of 17α-estradiol, and interconversion of 17β-estradiol and 17αestradiol through estrone has been observed in sediments under anaerobic conditions.30 It is also possible that the 17β-estradiol was produced when conjugated forms of the steroid in the urine were deconjugated in the soil. The total concentration of estrogens in the soil was approximately constant throughout the 14 21-day experimental period (Figure 1). Other researchers have reported rapid transformation of 17αestradiol and 17β-estradiol to estrone in dairy manure and waste lagoons.15 17β-estradiol has also been observed to undergo conversion to estrone in dairy waste solids,20 soils,19,21 river water,31 and wastewater treatment plants32 on the time scale of hours to days. However, the storage conditions of the manure appear to play a role in the rate of transformation. For example, Lorenzen et al.33 observed that under some storage conditions, the estrogenicity of manure remained constant or even increased by several orders of magnitude. However, they were unable to identify the conditions responsible for the stability of the compounds. One possible explanation for the enhanced stability of estrogens in the feedlot soils is decreased microbial activity in the relatively dry soils. Microbial activity in soil typically increases with moisture content until aeration becomes insufficient and is lowest under air-dried conditions.34 Transformation rates of 17β-estradiol19 and trenbolone35 in soil also have been reported to decrease steadily as soil moisture is reduced, so it is possible that the dry conditions were responsible for the high stability of the estrogens. However, the estrogen concentrations also remained constant upon wetting of the soil by the simulated rainfall while the androgens and progesterone concentrations decreased substantially (Figure 1). Also, androstenedione and progesterone concentrations changed substantially during the dry, 7-day aging period suggesting there was some microbial activity in the soil (Figure 1). It is possible that other factors, such as location of the steroid within the manure also protected the estrogens from biotransformation. Only a small fraction of the mass of the estrogens was released into the runoff, suggesting that they were not easily removed from the soil (Figure 3). This close association of these steroids with the soil could also have limited their bioavailability, increasing their persistence. Androgens and Progesterone. Although the cattle excreted little androstenedione and progesterone, relatively high concentrations of the steroids were observed in the soil after the 14-day period when the cattle were present on the feedlot (Figure 1). In previous experiments in which manure from pregnant cattle was aged 0 14 days and applied to pasture plots prior to irrigation, progesterone concentrations in the runoff also increased as the manure aged (Supporting Information Figure SI7). Zheng et al. also observed an increase in progesterone concentrations when 8815
dx.doi.org/10.1021/es202072f |Environ. Sci. Technol. 2011, 45, 8811–8818
Environmental Science & Technology dairy manure was aged, with concentrations increasing from below detection limits to approximately 200 ng/g over a 2-week period.15 Androstenedione and progesterone have also been detected concurrently in streams at relatively high concentrations downstream of grazing cattle with concentrations up to 44 ng/L and 27 ng/L, respectively. While only detected in 5% of the samples, when present, progesterone concentrations were usually higher than those of the other steroids measured.12 Androstenedione is produced from testosterone in soils with half-lives on the order of several hours to a few days depending on the soil.18 However, even if all the testosterone excreted by the cattle were converted to androstenedione, it would only account for approximately 10% of the measured concentration. An alternative explanation is production of androstenedione and progesterone by the transformation of sterols in manure. Soil bacteria and E. coli isolated from human feces have been shown to convert cholesterol, sitosterol, and stigmasterol into androstenedione.36 39 Reported yields for these reactions varied from 1% to 39% depending on the sterol, the organism, and the medium used for incubation. Cholesterol and stigmasterol have been measured in cow manure at concentrations of 3770 ( 250 μg/g and 490 ( 20 μg/g (mean ( standard deviation), respectively.40 If the steer manure contained similar concentrations of these sterols and the lowest reported yield for conversion of these two sterols into androstenedione (i.e., 1%) is used, the calculated mass of androstenedione would be between 30 and 250 times higher than those measured in the soils under the conditions observed in our experiments. Cholesterol has been shown to form oxidation products in frozen food even while frozen at 20 C.41 Therefore, it is possible that some of the additional androstenedione and progesterone were formed in the frozen soil and manure during storage. However, it is more likely that they were formed at the ambient temperatures on the pen. Progesterone and androstenedione have also been detected in surface waters subject to discharges from sterol-rich pulp and paper mill effluents where masculinization of fish was observed.42 Lab experiments verified the release of progesterone from pine wood used for pulp followed by conversion to androstenedione by soil bacteria.43,44 While our findings suggest that a similar process may also be occurring in manure-containing soils, additional research is needed to determine which sterols are responsible for progesterone and/or androstenedione formation and to better understand the factors affecting their formation. When simulated rainfall was applied to the plots, soil concentrations of the androgens and progesterone decreased substantially within a very short period. The loss of steroid mass from the soil was not accounted for by the runoff (Figure 3). It is also unlikely that the steroids were leached through the soil; the clay layer under the soil was very dense, and the soil was only wetted to a depth of 2 3 cm. Furthermore, concentrations of steroids detected in the clay layer before and after simulated rainfall were near or below quantification limits (Supporting Information Table SI2). One possible explanation for the rapid loss is stimulation of microbial activity in the brief period between wetting and extraction. For example, in aged feedlot runoff, Havens et al. reported rapid transformation of estrogens and androgens within 3 h of spiking if the spiked samples were not acidified.45 The soil samples from our feedlot were placed on ice immediately after collection and were kept frozen until extraction. Runoff samples were also placed immediately on ice after collection and were extracted within 2 days. While a similar rapid disappearance was not observed for the estrogens, previous
ARTICLE
research has demonstrated that testosterone is more labile than estradiol in soils.18,46 While data on the transformation of androstenedione and progesterone in soil or agricultural runoff are limited, Chang et al. reported higher removal efficiencies for these compounds (96 97%) than for the estrogens (67 80%) in municipal wastewater treatment systems, implying they are more labile in the presence of aerobic bacteria.5 Further research is needed to determine if wetting the soil could induce rapid degradation of the androgens and progesterone, but not the estrogens. Partitioning and Transport of Steroids to Surface Waters. Because the centrifuged and filtered runoff contained dissolved organic carbon concentrations as high as 1400 mg/L, it is likely that a significant portion of the steroids measured in the filtered fraction of the runoff were not truly dissolved, but were bound to colloidal organic matter. Reported partition coefficients for association of steroids with colloidal organic matter are typically between 103 and 106.47 Using these values along with the mean filtered organic carbon concentration of 570 mg/L we estimate that between 37% and 99% of the steroids could have been associated with organic matter that was not removed by centrifugation and filtration. Colloidal organic matter potentially enhances transport,21,48 increases the fraction of steroids in the filtered fraction,49 51 and slows biotransformation of steroids.52 Management practices, such as settling basins and lagoons, which rely on gravitational settling to remove suspended solids from feedlot runoff will remove steroid hormones associated with large particles, but may have little effect on steroids associated with colloidal organic carbon. The partitioning of the steroids between the filtered and particulate phases did not fit equilibrium hydrophobic partitioning models, with or without consideration of colloids. There also was no relationship between the fraction of steroids associated with particles and the concentration of suspended solids as would be predicted from equilibrium partitioning. Equilibrium partitioning models assume a matrix where hydrophobic partitioning yields nearly linear partitioning provided sufficient time is allowed for equilibration. The matrix in this runoff was complex and heterogeneous and may not have reached equilibrium prior to sample collection. Equilibration times for 17β-estradiol and estrone to soils and sediments vary greatly with some researchers reporting equilibrium partitioning after 5 24 h,22 while others report a range from 3 to 14 days.49,51 Previous studies have also indicated that steroid sorption to soils was not correlated with log Kow values.18,53 Thus, equilibrium hydrophobic partitioning models may not accurately predict observations for conditions encountered in feedlots. The concentrations of steroids measured in the runoff exceed reported thresholds for end points related to feminization, masculinization, and interfere with pheromonal responses for fish.1,8,54,55 The estrogenicity of estrone and 17α-estradiol relative to 17β-estradiol depends upon the assay used,54,56 but assuming the lowest and highest values reported (0.08 and 0.8 for 17α-estradiol, 0.01 and 1 for estrone relative to 17β-estradiol, the total estrogenicity expressed in 17β-estradiol equivalents in the runoff was between 1 and 2 orders of magnitude higher than the 1 ng/L threshold for biological response and was not significantly different after the plots were aged for 7 days. Androstenedione concentrations in the runoff from the initial and aged plots were approximately 3 4 times higher than the reported threshold for masculinization of 40 ng/L.8 Androstenedione also elicits odorant responses in goldfish at concentrations as low as 0.3 ng/L,9 and testosterone can elicit pheromonal response in Atlantic salmon 8816
dx.doi.org/10.1021/es202072f |Environ. Sci. Technol. 2011, 45, 8811–8818
Environmental Science & Technology parr at 0.003 ng/L.57 Thus the runoff concentrations are 2 3 orders of magnitude above thresholds for odorant responses in fish. While progesterone has been shown to induce vitellogenin expression in crayfish at concentrations of approximately 30 μg/L,58 no data on threshold concentrations for endocrine disruption or pheromonal activity in aquatic life are available. However, other progestogens, such as 17-α,20-β-dihyroxy-4-pregnen-3-one have been shown to illicit pheromonal and odorant responses in goldfish at 3 ng/L and 0.03 ng/L, respectively.59 Further research is needed to determine if progesterone also poses a threat to aquatic life at similarly low concentrations. In applications where treatment or dilution is insufficient, estrogens and androgens could pose threats to aquatic life. Microbial transformation reactions may reduce the concentrations of some steroid hormones, but aging of wastes under dry conditions will have little effect on the mass of estrogens and may even result in a substantial increase in the mass of androgens or progestogens. As a result, caution must be exercised in the management of waste discharges from animal feeding facilities and pastures.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 510-643-0256; fax: 510-642-7483; e-mail: sedlak@berkeley. edu.
’ ACKNOWLEDGMENT This research was funded by a U.S. EPA Science to Achieve Results (STAR) Award No. R833422. We thank the Dr. Frank Mitloehner, who built and maintained the feedlot pens, and Kim Stackhouse, who managed the cattle operators at the UC Davis Research Feedlot. ’ REFERENCES (1) Purdom, C. E.; Hardiman, P. A.; Bye, V. J.; Eno, N. C.; Tyler, C. R.; Sumpter, J. P. Estrogenic effects of effluents from sewage treatment works. Chem. Ecol. 1994, 8, 275–285. (2) Routledge, E. J.; Sheahan, D.; Desbrow, C.; Brighty, G. C.; Waldock, M.; Sumpter, J. P. Identification of estrogenic chemicals in STW effluent. 2. In vivo responses in trout and roach. Environ. Sci. Technol. 1998, 32, 1559–1565. (3) Jobling, S.; Nolan, M.; Tyler, C. R.; Brighty, G.; Sumpter, J. P. Widespread sexual disruption in wild fish. Environ. Sci. Technol. 1998, 32, 2498–2506. (4) Kidd, K. A.; Blanchfield, P. J.; Mills, K. H.; Palace, V. P.; Evans, R. E.; Lazorchak, J. M.; Flick, R. W. Collapse of a fish population after exposure to a synthetic estrogen. Proc. Natl. Acad. Sci. U. S. A. 2007, 104, 8897–8901. (5) Chang, H.; Wan, Y.; Wu, S. M.; Fan, Z. L.; Hu, J. Y. Occurrence of androgens and progestogens in wastewater treatment plants and receiving river waters: Comparison to estrogens. Water Res. 2011, 45, 732–740. (6) Kolok, A. S.; Snow, D. D.; Kohno, S.; Sellin, M. K.; Guillette, L. J. Occurrence and biological effect of exogenous steroids in the Elkhorn River, Nebraska, USA. Sci. Total Environ. 2007, 388, 104–115.
ARTICLE
(7) Kolodziej, E. P.; Gray, J. L.; Sedlak, D. L. Quantification of steroid hormones with pheromonal properties in municipal wastewater effluent. Environ. Toxicol. Chem. 2003, 22, 2622–2629. (8) Stanko, J. P.; Angus, R. A. In vivo assessment of the capacity of androstenedione to masculinize female mosquitofish (Gambusia affinis) exposed through dietary and static renewal methods. Environ. Toxicol. Chem. 2007, 26, 920–926. (9) Sorenson, P. W., Stacey, N. E., In Advances in Chemical Signals in Vertebrates; Johnston, R. E., Muller-Schwarze, D., Sorenson, P. W., Eds.; Kluwer Academic/Plenum: New York, 1999; pp 15 47. (10) Lee, L. S.; Carmosini, N.; Sassman, S. A.; Dion, H. M.; Sepulveda, M. S. Agricultural contributions of antimicrobials and hormones on soil and water quality. Adv. Agron. 2007, 93, 1–68. (11) Matthiessen, P.; Arnold, D.; Johnson, A. C.; Pepper, T. J.; Pottinger, T. G.; Pulman, K. G. T. Contamination of headwater streams in the United Kingdom by oestrogenic hormones from livestock farms. Sci. Total Environ. 2006, 367, 616–630. (12) Kolodziej, E. P.; Sedlak, D. L. Rangeland grazing as a source of steroid hormones to surface waters. Environ. Sci. Technol. 2007, 41, 3514–3520. (13) Kolodziej, E. P.; Harter, T.; Sedlak, D. L. Dairy wastewater, aquaculture, and spawning fish as sources of steroid hormones in the aquatic environment. Environ. Sci. Technol. 2004, 38, 6377–6384. (14) Durhan, E. J.; Lambright, C. S.; Makynen, E. A.; Lazorchak, J.; Hartig, P. C.; Wilson, V. S.; Gray, L. E.; Ankley, G. T. Identification of metabolites of trenbolone acetate in androgenic runoff from a beef feedlot. Environ. Health Perspect. 2006, 114, 65–68. (15) Zheng, W.; Yates, S. R.; Bradford, S. A. Analysis of steroid hormones in a typical dairy waste disposal system. Environ. Sci. Technol. 2008, 42, 530–535. (16) Lange, I. G.; Daxenberger, A.; Schiffer, B.; Witters, H.; Ibarreta, D.; Meyer, H. H. D. Sex hormones originating from different livestock production systems: Fate and potential disrupting activity in the environment. Anal. Chim. Acta 2002, 473, 27–37. (17) Arts, C. J. M.; Vanbaak, M. J.; Denhartog, J. M. P. Controlsystem for detection of the illegal use of naturally-occurring steroids in calves. J. Chromatogr., Biomed. Appl. 1991, 564, 429–444. (18) Lee, L. S.; Strock, T. J.; Sarmah, A. K.; Rao, P. S. C. Sorption and dissipation of testosterone, estrogens, and their primary transformation products in soils and sediment. Environ. Sci. Technol. 2003, 37, 4098–4105. (19) Colucci, M. S.; Bork, H.; Topp, E. Persistence of estrogenic hormones in agricultural soils: I. 17 beta-estradiol and estrone. J. Environ. Qual. 2001, 30, 2070–2076. (20) Raman, D. R.; Layton, A. C.; Moody, L. B.; Easter, J. P.; Sayler, G. S.; Burns, R. T.; Mullen, M. D. Degradation of estrogens in dairy waste solids: Effects of acidification and temperature. Trans. ASAE 2001, 44, 1881–1888. (21) Casey, F. X. M.; Larsen, G. L.; Hakk, H.; Simunek, J. Fate and transport of 17 beta-estradiol in soil-water systems. Environ. Sci. Technol. 2003, 37, 2400–2409. (22) Casey, F. X. M.; Simunek, J.; Lee, J.; Larsen, G. L.; Hakk, H. Sorption, mobility, and transformation of estrogenic hormones in natural soil. J. Environ. Qual. 2005, 34, 1372–1379. (23) Miller, W. P. A solenoid-operated, variable intensity rainfall simulator. Soil Sci. Soc. Am. J. 1987, 51, 832–834. (24) Humphry, J. B.; Daniel, T. C.; Edwards, D. R.; Sharpley, A. N. A portable rainfall simulator for plot-scale runoff studies. Appl. Eng. Agric. 2002, 18, 199–204. (25) American Public Health Association, A. W. W. A., Water Environment Federation. Standard Methods for the Examination of Water and Wastewater. 19th ed.; American Public Health Association: Washington, DC, 1995. (26) Johnson, A. C.; Williams, R. J.; Matthiessen, P. The potential steroid hormone contribution of farm animals to freshwaters, the United Kingdom as a case study. Sci. Total Environ. 2006, 362, 166–178. (27) Rico, A. G. Metabolism of endogenous and exogenous anabolic agents in cattle. J. Anim. Sci. 1983, 57, 226–232. 8817
dx.doi.org/10.1021/es202072f |Environ. Sci. Technol. 2011, 45, 8811–8818
Environmental Science & Technology (28) Hanselman, T. A.; Graetz, D. A.; Wilkie, A. C. Manure-borne estrogens as potential environmental contaminants: A review. Environ. Sci. Technol. 2003, 37, 5471–5478. (29) Khan, S. J.; Roser, D. J.; Davies, C. M.; Peters, G. M.; Stuetz, R. M.; Tucker, R.; Ashbolt, N. J. Chemical contaminants in feedlot wastes: Concentrations, effects and attenuation. Environ. Int. 2008, 34, 839–859. (30) Czajka, C. P.; Londry, K. L. Anaerobic biotransformation of estrogens. Sci. Total Environ. 2006, 367, 932–941. (31) Jurgens, M. D.; Holthaus, K. I. E.; Johnson, A. C.; Smith, J. J. L.; Hetheridge, M.; Williams, R. J. The potential for estradiol and ethinylestradiol degradation in English rivers. Environ. Toxicol. Chem. 2002, 21, 480–488. (32) Joss, A.; Andersen, H.; Ternes, T.; Richle, P. R.; Siegrist, H. Removal of estrogens in municipal wastewater treatment under aerobic and anaerobic conditions: Consequences for plant optimization. Environ. Sci. Technol. 2004, 38, 3047–3055. (33) Lorenzen, A.; Hendel, J. G.; Conn, K. L.; Bittman, S.; Kwabiah, A. B.; Lazarovitz, G.; Masse, D.; McAllister, T. A.; Topp, E. Survey of hormone activities in municipal biosolids and animal manures. Environ. Toxicol. 2004, 19, 216–225. (34) Miller, R. D.; Johnson, D. D. The effect of soil moisture tension on carbon dioxide evolution, nitrification, and nitrogen mineralization. Soil. Sci. Soc.Am. Proc. 1964, 28, 644–647. (35) Khan, B.; Lee, L. S. Soil temperature and moisture effects on the persistence of synthetic androgen 17 alpha-trenbolone, 17 beta-trenbolone and trendione. Chemosphere 2010, 79, 873–879. (36) Marsheck, W. J.; Kraychy, S.; Muir, R. D. Microbial degradation of sterols. Appl. Microbiol. 1972, 23, 72–&. (37) Nagasawa, M.; Bae, M.; Tamura, G.; Arima, K. , Microbial transformation of sterols 0.2. Cleavage of sterol side chains by microorganisms. Agric. Biol. Chem. 1969, 33, 1644–&. (38) Owen, R. W.; Tenneson, M. E.; Bilton, R. F.; Mason, A. N. The degradation of cholesterol by Escherichia-coli isolated from human feces. Biochem. Soc. Trans. 1978, 6, 377–379. (39) Conner, A. H.; Nagaoka, M.; Rowe, J. W.; Perlman, D. Microbial conversion of tall oil sterols to C19 steroids. Appl. Environ. Microbiol. 1976, 32, 310–311. (40) Tyagi, P.; Edwards, D. R.; Coyne, M. S. Use of sterol and bile acid biomarkers to identify domesticated animal sources of fecal pollution. Water, Air, Soil Pollut. 2008, 187, 263–274. (41) Pie, J. E.; Spahis, K.; Seillan, C. Cholesterol oxidation in meatproducts during cooking and frozen storage. J. Agric. Food Chem. 1991, 39, 250–254. (42) Jenkins, R.; Angus, R. A.; McNatt, H.; Howell, W. M.; Kemppainen, J. A.; Kirk, M.; Wilson, E. M. Identification of androstenedione in a river containing paper mill effluent. Environ. Toxicol. Chem. 2001, 20, 1325–1331. (43) Jenkins, R. L.; Wilson, E. M.; Angus, R. A.; Howell, W. M.; Kirk, M.; Moore, R.; Nance, M.; Brown, A. Production of androgens by microbial transformation of progesterone in vitro: A model for androgen production in rivers receiving paper mill effluent. Environ. Health Perspect. 2004, 112, 1508–1511. (44) Carson, J. D.; Jenkins, R. L.; Wilson, E. M.; Howell, W. M.; Moore, R. Naturally occurring progesterone in Loblolly Pine (Pinus taeda l.): A major steroid precursor of environmental androgens. Environ. Toxicol. Chem. 2008, 27, 1273–1278. (45) Havens, S. M.; Hedman, C. J.; Hemming, J. D. C.; Mieritz, M. G.; Shafer, M. M.; Schauer, J. J. Stability, preservation, and quantification of hormones and estrogenic and androgenic activities in surface water runoff. Environ. Toxicol. Chem. 2010, 29, 2481–2490. (46) Casey, F. X. M.; Hakk, H.; Simunek, J.; Larsen, G. L. Fate and transport of testosterone in agricultural soils. Environ. Sci. Technol. 2004, 38, 790–798. (47) Neale, P. A.; Escher, B. I.; Schafer, A. I. pH dependence of steroid hormone-organic matter interactions at environmental concentrations. Sci. Total Environ. 2009, 407, 1164–1173. (48) Stanford, B. D.; Amoozegar, A.; Weinberg, H. S. The impact of co-contaminants and septic system effluent quality on the transport of
ARTICLE
estrogens and nonylphenols through soil. Water Res. 2010, 44, 1598– 1606. (49) Yu, Z. Q.; Xiao, B. H.; Huang, W. L.; Peng, P. Sorption of steroid estrogens to soils and sediments. Environ. Toxicol. Chem. 2004, 23, 531–539. (50) Stumpe, B.; Marschner, B. Long-term sewage sludge application and wastewater irrigation on the mineralization and sorption of 17 beta-estradiol and testosterone in soils. Sci. Total Environ. 2007, 374, 282–291. (51) Bowman, J. C.; Zhou, J. L.; Readman, J. W. Sediment-water interactions of natural oestrogens under estuarine conditions. Mar. Chem. 2002, 77, 263–276. (52) Stumpe, B.; Marschner, B. Dissolved organic carbon from sewage sludge and manure can affect estrogen sorption and mineralization in soils. Environ. Pollut. 2010, 158, 148–154. (53) Das, B. S.; Lee, L. S.; Rao, P. S. C.; Hultgren, R. P. Sorption and degradation of steroid hormones in soils during transport: Column studies and model evaluation. Environ. Sci. Technol. 2004, 38, 1460– 1470. (54) Shappell, N. W.; Hyndman, K. M.; Bartell, S. E.; Schoenfuss, H. L. Comparative biological effects and potency of 17 alpha- and 17 beta-estradiol in Fathead minnow. Aquat. Toxicol. 2010, 100, 1–8. (55) Adams, M. A.; Teeter, J. H.; Katz, Y.; Johnsen, P. B. Sexpheromones of the sea lamprey (Petromyzon-marinus) - steroid studies. J. Chem. Ecol. 1987, 13, 387–395. (56) Liu, Z. H.; Kanjo, Y.; Mizutani, S. Removal mechanisms for endocrine disrupting compounds (EDCs) in wastewater treatment— Physical means, biodegradation, and chemical advanced oxidation: A review. Sci. Total Environ. 2009, 407, 731–748. (57) Moore, A.; Scott, A. P. Testosterone is a potent odorant in precocious male Atlantic salmon (Salmo-salar l) parr. Philos. Trans. R. Soc. London, Ser. B: Biol. Sci. 1991, 332, 241–244. (58) Coccia, E.; De Lisa, E.; Di Cristo, C.; Di Cosmo, A.; Paolucci, M. Effects of estradiol and progesterone on the reproduction of the freshwater crayfish Cherax albidus. Biol. Bull. 2010, 218, 36–47. (59) Sorensen, P. W.; Hara, T. J.; Stacey, N. E. Extreme olfactory sensitivity of mature and gonadally-regressed goldfish to a potent steroidal pheromone, 17-alpha,20-beta-dihydroxy-4-pregnen-3-one. J. Comp. Physiol., A 1987, 160, 305–313.
8818
dx.doi.org/10.1021/es202072f |Environ. Sci. Technol. 2011, 45, 8811–8818
ARTICLE pubs.acs.org/est
Brominated Flame Retardants in the Atmosphere of E-Waste and Rural Sites in Southern China: Seasonal Variation, Temperature Dependence, and Gas-Particle Partitioning Mi Tian,†,‡ She-Jun Chen,*,† Jing Wang,†,‡ Xiao-Bo Zheng,†,‡ Xiao-Jun Luo,† and Bi-Xian Mai† †
State Key Laboratory of Organic Geochemistry, Guangzhou Institute of Geochemistry, Chinese Academy of Sciences, Guangzhou 510640, China and ‡ Graduate University of Chinese Academy of Sciences, Beijing 100049, China
bS Supporting Information ABSTRACT: The recycling of electrical and electronic waste (e-waste) in developing countries has attracted much attention as a significant source of brominated flame retardants (BFRs). Gaseous and particle-bound BFRs were measured in the atmosphere at e-waste and rural sites in southern China during 20072008. The annual average concentrations in the air were 3260 ( 3370 and 219 ( 192 pg/m3 for polybrominated diphenyl ethers (PBDEs) and were 546 ( 547 and 165 ( 144 pg/m3 for non-PBDE BFRs at the e-waste and rural sites, respectively. PBDEs had unusually high relative concentrations of di- and tribrominated congeners at the e-waste site. The ClausiusClapeyron (CC) plots showed that the gaseous concentrations of less brominated BFRs (di- through hexaBFRs) were strongly controlled by temperature-driven evaporation from contaminated surfaces (e.g., e-waste, soils, and recycled e-waste remains) except for winter. However, weak temperature dependence at the rural site suggests that regional or long-range atmospheric transport was largely responsible for the air concentrations. Gas-particle partitioning (KP) of PBDEs correlated well with the subcooled liquid vapor pressure (PLo) for most sampling events. The varied slopes of log KP versus log PLo plots for the e-waste site (0.59 to 1.29) indicated an influence of ambient temperature and atmospheric particle properties on the partitioning behavior of BFRs. The flat slopes (0.23 to 0.80) for the rural site implied an absorption-dominant partitioning. This paper suggests that e-waste recycling in Asian low-latitude regions is a significant source of less brominated BFRs and has important implications for their global transport from warm to colder climates.
’ INTRODUCTION Over the past two decades, there have been growing concerns about brominated flame retardants (BFRs). Many BFRs exhibit persistence and bioaccumulation properties similar to many persistent organic pollutants (POPs), and they are detected globally.1 Polybrominated diphenyl ethers (PBDEs) are one of the most widely used BFRs, for which three PBDE technical mixtures (penta-, octa-, and deca-BDE) have been or are manufactured. The main components (tetra- through heptaBDEs) of the former two mixtures have been included in the newly listed POPs of the 2009 Stockholm Convention because of their persistence, bioaccumulative capacity, toxicity, and adverse effects.2 More recently, there has been increasing evidence of the environmental occurrence of novel BFRs such as decabromodiphenyl ethane (DBDPE) and 1,2-bis(2,4,6-tribromophenoxy)ethane (BTBPE), in sediment, air, indoor dust, and aquatic and terrestrial species from different locations in the world.3,4 The recycling of electrical and electronic waste (e-waste) in developing countries has attracted significant attention as an r 2011 American Chemical Society
important source of many toxic chemicals. It has been reported that around 80% of the world’s e-waste is exported to mid- and low-latitude developing countries/regions of Asia, and 90% of this flows into China.5 The techniques for recycling e-waste in these regions are usually primitive and hazardous and lead to the emissions of high concentrations of heavy metals, polychlorinated biphenyls (PCBs), polychlorinated dibenzo-p-dioxin/furans (PCDD/Fs), and BFRs in the environment.69 The elevated body burdens of PBDEs, PCBs, and PCDD/Fs of residents and workers from an e-waste site in China suggest a substantial human exposure to these chemicals.10 The atmosphere plays a significant role in transporting BFRs from their sources to rural/remote locations. Studies on atmospheric BFRs have been conducted largely in mid- to high-latitude Received: July 3, 2011 Accepted: September 8, 2011 Revised: September 2, 2011 Published: September 08, 2011 8819
dx.doi.org/10.1021/es202284p | Environ. Sci. Technol. 2011, 45, 8819–8825
Environmental Science & Technology
ARTICLE
regions of North America and Europe,1,4 while data from lowlatitude regions are very limited.11,12 Investigations from these regions are therefore needed, given that most worldwide e-waste is recycled in low-latitude regions where the tropical climate favors the release of BFRs into the atmosphere and their subsequent movement to colder regions of the earth.12,13 It has been known that the air concentrations of semivolatile organic compounds (SOCs) have a strong relationship to ambient temperature.14 The temperature dependence of SOCs in the air is conveniently described by the ClausiusClapeyron (CC) equation, which can be expressed as ln P ¼ m=T þ b
Table 1. Summary of Concentrations (pg/m3) of BFRs in the Atmosphere (Gas + Particle) at the E-Waste and Rural Sites in Southern Chinaa e-waste site BFRs
ð1Þ
where P is the partial pressure (atm), T is the temperature (K), and m and b are fitting parameters. The CC plot has been used to indicate the relative importance of evaporation from local surfaces versus long-range atmospheric transport (LRAT) or advection of background air mass in controlling air concentrations.14,15 Although a few studies have reported on the atmospheric concentrations of BFRs (in particular, PBDEs) from e-waste locations, emission mechanisms into the atmosphere have been poorly investigated.8,16,17 In the present study, we measured the concentrations of PBDEs and several non-PBDE BFRs in the atmosphere at an e-waste recycling site as well as a rural site in southern China. The emission profiles and seasonal variations of BFRs at the two sites were examined. This paper presents the temperature-dependence and gas-particle partitioning of BFRs to provide insights the emission mechanisms into the atmosphere and partitioning behaviors between the gas and particle phases.
’ EXPERIMENTAL SECTION Sample Collection. Details regarding the studied area and sampling technology are provided in the Supporting Information (SI). Briefly, the sampling was conducted at an e-waste site and a rural site in southern China. At the e-waste site, large quantities of e-waste (∼700 000 t) that is mostly imported from overseas is processed within an area of ∼330 km2 using primitive methods (e.g., mechanical shredding, acid processing, and open burning). The rural site (25 km northeast of the e-waste site) is located in an agricultural area, where BFRs in the air are mainly from the e-waste and urban areas (SI Figure S1).9 The average daily temperatures, wind speeds, and relative humidity during the sampling period varied from 7 to 33 °C, from 1.2 to 7.6 m/s, and from 44% to 90%. Air samples were taken by drawing air through glass fiber filters (GFFs) followed by polyurethane foam (PUF) plugs for 24 h, using a high-volume air sampler. A total of 60 pairs of samples (gas and particle) from each site were collected simultaneously on five consecutive days each month from July 2007 to June 2008. Sample Preparation and Instrumental Analysis. The PUF and GFF samples were Soxhlet extracted separately with a mixture of hexane and acetone (1:1) for 48 h. Prior to the extraction, BDE77, BDE181, and 13C-BDE209 (GFFs only) were added as surrogates to monitor the recoveries. The extracts were concentrated to 12 mL and then eluted through a silica/ alumina column with 1:1 hexane-dichloromethane. Following solvent exchange to hexane, BDE118 and BDE128 were added as quantitation standards. The target BFRs were analyzed using a gas chromatograph coupled to a mass spectrometer in electron capture negative ionization mode (GC-ECNI-MS). Di- through hepta-BDEs,
AM
GM
range
AM
GM
di-BDEs
2.324120
399
131
0.4164.4
8.92
4.67
tri-BDEs
3.983000
354
131
0.1529.3
6.05
4.07
tetra-BDEs
8.69762
220
128
0.7115.7
5.82
4.45
penta-BDEs 7.761170
284
160
0.9331.6
7.00
4.99
hexa-BDEs 3.25571 hepta-BDEs 3.12619
131 127
1.7019.1 nd17.8
4.29 5.50
3.60 3.94
octa-BDEs
nd491
154
nona-BDEs
nd767
187
67.8 67.5 94.2 139
nd64.9
10.3
1.6299.1
25.2
deca-BDEs
72.29700 1410
862
18.6804
142
PBDEs
10917900 3260
2080
37.0952
216
PBT
0.19125
PTBX
nd158
PBEB HBB
0.29867 4.47559
BTBPE
4.49398
DBDPE PBBs total a
range
rural site
nd2240 nd466
21.8 8.29 41.0 138 78.7 209 49.7
12019000 3810
5.96 18.3 91.8 165
9.64 0.213.57
0.93
0.76
0.64
nd3.43
0.33
0.09
0.104.80 0.4213.9
0.75 4.49
0.53 3.14
nd28.4
2.97
12.3 88.8 45.3 137 14.4 2490
3.971370 158 0.26.74
0.92
54.71710 384
1.65 80.6 0.50 287
AM: arithmetic mean; GM: geometric mean; nd: not detectable.
pentabromoethylbenzene (PBEB), 2,3,5,6-tetrabromo-p-xylene (pTBX), pentabromotoluene (PBT), hexabromobenzene (HBB), and 2,20 ,4,40 ,5,50 -polybrominated biphenyl (BB153) were separated with a DB-XLB (30 m 0.25 mm i.d., 0.25 μm film thickness) capillary column. For octa- through deca-BDEs, DBDPE, BTBPT, and BB209, a DB-5HT (15 m 0.25 mm i.d., 0.10 μm film thickness) column was used. More information on the procedures is given in the SI. Quality Control. The breakthrough of the gas-phase BFRs was tested twice at each sampling site using a second PUF plug (2.5 cm thick) in series with the first one. Field blanks (clean PUF plugs and GFFs) were prepared identically to that of the real samples, except with no air drawn through. Twenty-four procedural blanks were run sequentially with the samples. Breakthrough less than 10% occurred for several di- through hexaBDEs at the e-waste site. BDE28, 47, 99, 206, 207, 208, and 209 were found in the procedural blanks, but their amounts were all less than 5% of those in the corresponding sample extracts. The concentrations in the samples were blank corrected accordingly. The amounts of PBDEs found in the field blanks were not significantly higher than the procedural blanks. The recoveries of the surrogate standard (mean ( standard deviation) were 94.3 ( 15.2% for BDE77, 87.2 ( 11.3% for BDE181, and 107 ( 18.0% for 13C-BDE209. The recoveries of the target compounds in the six matrix-spiked samples were 67.2%122% (standard deviations <15.4%). The repeatability of analysis was assessed by analyzing five filter replicates (subsampled from loaded GFFs), and the relative standard deviations were within 1.1221.7%. Reported concentrations were not surrogate-recovery corrected. The method detection limits, defined as the mean blank mass plus three standard deviations or a signal five times the noise level, were between 0.06 and 1.15 pg/m3 based on the average volume (350 m3). 8820
dx.doi.org/10.1021/es202284p |Environ. Sci. Technol. 2011, 45, 8819–8825
Environmental Science & Technology
ARTICLE
’ RESULTS AND DISCUSSION Concentrations and Compositions. Table 1 summarizes the concentrations of BFRs in air at the e-waste and rural sites. The air (gaseous and particle phases) concentrations of total BFRs measured at the e-waste site ranged from 120 to 19 000 pg/m3, with an arithmetic mean of 3810 pg/m3. BFRs were also found in the air at the rural site, although the concentrations were 1 order of magnitude lower than those at the e-waste site, ranging from 54.7 to 1710 pg/m3 (with an arithmetic mean of 384 pg/m3). PBDEs constituted the most prevalent group of BFRs in the air, with average contributions of 84% and 61% at the e-waste and rural sites, respectively (SI Figure S2). DBDPE (as a replacement of the technical deca-BDE mixture) exhibited the second highest concentrations at both sites, and its average contribution in the rural air (35%) was significantly higher than at the e-waste site (6.9%) (p = 0.001). The concentrations or percentage contributions of other BFRs were in the order of HBB > pTBX > BTBPE > PBBs > PBEB > PBT for both sites. This BFR profile was similar to the atmospheric deposition profile observed at the same sites in our previous study.9 SI Figure S3 depicts the congener profiles of the atmospheric PBDEs at the two sites. BDE209 was the most dominant congener in almost all of the samples. It contributed 47 ( 17% and 59 ( 16% of the total PBDEs at the e-waste and rural sites, respectively. This is consistent with the fact that the deca-BDE mixture is one of the most frequently used flame retardants across the world, especially in electronic/electric products.3,18 On the other hand, it is surprising to observe high contributions of diand tri-BDE congeners in the air (especially of the e-waste site) because these congeners are not present, or present only in minor amounts, in the technical PBDE mixtures.19 Highest geometric mean concentrations of di- and tri-BDEs of 4120 and 3000 pg/m3, respectively, were found at the e-waste site and were comparable to the concentrations of tetra-, penta-, and nona-BDEs (Table 1). The atmospheric PBDE congener profiles in this study were quite different from the profiles that have typically been observed in the air from other locations, in which di- and tri-BDEs accounted for very small proportions of the PBDEs.2023 In fact, tetra-, penta-, and hexa-BDEs, which are mainly derived from technical penta-BDE mixtures, were also present in significant proportions in the studied area compared to those in the urban air in southern China from an early study.11 This finding provides evidence that e-waste recycling is a significant emission source of less brominated PBDEs (di- through hexa-BDEs) to the environment. These PBDEs (in particular those absent from the technical PBDE products) could originate from photolytical degradation of highly brominated BDEs in the environment. However, they are more likely to originate from pyrolytic degradation during the recycling processes (e-waste burning) due to the shielding effects of carbonaceous aerosols to photodegradation indicated in a previous study.24 Additionally, high ambient temperatures facilitate the volatilization of less brominated congeners (with relatively high vapor pressures) from contaminated environmental compartments, massive e-waste, and recycled e-waste remains stacked in the fields. The total PBDE concentrations (∑36PBDEs) at the e-waste site in this study were lower than those detected in the air from another e-waste site (Guiyu) in China, with average concentrations of 21 500 pg/m3 in 2004 (∑22PBDEs without BDE209) and 8760 pg/m3 in 2005 (∑11PBDEs).16,17 These differences could
Figure 1. Seasonal variations (from July 2007 to June 2008) of the major BFRs in the gaseous phase at the two sites in southern China. E-waste site: A (sum of di-, tri-, and tetra-BDEs, PBT, PEEB, and HBB) and B (sum of penta- and hexa-BDEs). Rural site: C (sum of di- and triBDEs, PBT, PEEB, and HBB) and D (sum of tetra-, penta-, and hexaBDEs).
be ascribed to the distance from the sampling sites to the e-waste recycling facilities and/or the recycling manners. There has been a substantial decrease in the activities of open burning e-waste because of the increasingly stringent restrictions. For instance, the highest concentrations in Guiyu were measured right at a building where there are open burning operations.16 It has been shown in previous studies that the average PBDE concentrations in urban air were on an order of 50150 pg/m3, and those in rural/remote air were generally in the range of 515 pg/m3 in North America, Europe, and Asia.1,20,23,25,26 A few studies have reported higher average concentrations in China (2450 pg/m3 in urban air and 220 pg/m3 in rural air)11,27 and in Ontario, Canada (300 pg/m3 in rural air)28 as well as lower concentrations in the Arctic (18 pg/m3).22,29 These levels are generally lower than the average concentrations of ∑16PBDEs (that have been typically reported in the literature) at the e-waste site (2220 pg/m3) and rural site (192 pg/m3) in the present study (SI Table S2). Atmospheric measurements of non-PBDE BFRs in other studies are very limited. The average concentrations of BTBPE in rural air were 3.4 pg/m3 in the east-central U.S. and 0.51.2 pg/m3 near the Great Lakes (U.S.), which were comparable to that in this study.21,25 The average concentrations of DBDPE in the air near the Great Lakes were 122 pg/m3, much lower than the concentrations in the present study.25 PBEB, with a relatively high concentration of 550 pg/m3, was measured near the Great Lakes.25 Seasonal Variations. The seasonal variations of the total air BFR concentrations during the sampling year (July 2007June 2008) at the e-waste and rural sites are shown in SI Figure S4. The monthly air BFR concentrations (five day averaged concentration) varied from 912 to 10 300 pg/m3 at the e-waste site. Seasonal variations of the total BFR concentrations in the gaseous and particle phases were generally similar (r = 0.85, p = 0.001). The monthly air BFR concentrations at the rural site ranged from 144 to 1160 pg/m3. Different seasonal variations were observed for the gaseous and particle-bound BFRs. 8821
dx.doi.org/10.1021/es202284p |Environ. Sci. Technol. 2011, 45, 8819–8825
Environmental Science & Technology
Figure 2. Seasonal variations (from July 2007 to June 2008) of the major BFRs in the particle phase at the two sites in southern China. E-waste site: A (sum of di-, tri-, and tetra-BDEs, PBT, PEEB, and HBB), B (sum of penta-, hexa-, and hepta-BDEs), and C (sum of octa-, nona-, and deca-, BTBPE, DBDPE). Rural site: D (sum of tri-, tetra-, and pentaBDEs), E (sum of hexa-, hepta-, and octa-BDEs), and F (nona- and decaBDEs and DBDPE).
Seasonal variations of the major gaseous and particle-bound BFRs (BFR concentrations with close variations were combined) were compared in Figures 1 and 2. At both the sites, the concentrations of more volatile BFRs (di-, tri, and tetra-BDEs, PBT, PBEB, and HBB) in the gaseous phase were relatively high in the summer months. Gaseous penta- and hexa-BDEs showed a similar seasonal variation to the more volatile BFRs at the e-waste site; whereas variation of gaseous tetra-, penta-, and hexa-BDEs differed from the more volatile BFRs at the rural site (Figure 1). The particle-bound BFR concentrations showed different seasonal variations at both the sites (Figure 2). It is interesting to find that, at the e-waste site, penta-, hexa-, and hepta-BDEs exhibited variations similar to the less volatile BFRs (octa-, nona-, and decaBDEs, DBDPE, and BTBPE) in warm season (April to September) and to the more volatile BFRs in colder season (October to March). This may suggest an influence of gas-particle partitioning behavior of these BFRs (e.g., lower temperature favors deposition of gaseous BFRs to particles). At the rural site, particle-bound hexa-, hepta-, and octa-BDEs and BTBPE showed a seasonal variation similar to more volatile tri-, tetra-, and penta-BDEs rather than less volatile nonaand deca-BDEs and DBDPE. These less volatile compounds are the dominant ingredients of technical deca-BDE and DBDPE products that are used in large quantities the nearby urban region; while trithrough octa-BDEs and BTBPE are more important pollutants at the e-waste site than in the urban region.9 The results suggest differences in the source and/or behavior of the BFRs in the atmosphere at the two sites.
ARTICLE
The seasonal variations are believed to be affected by the meteorological conditions. However, significant relationships were not found between the air concentrations (individual or total BFRs) and wind speed or relative humidity. Temperature is one of the major factors controlling seasonality of SOCs in air. The influence of ambient temperature on BFRs at the sites is discussed below. Influences of Ambient Temperature. The CC equation was applied to the major gaseous BFRs at the two sites, and the regression results are summarized in SI Table S3. The CC plots show apparent differences in temperature dependence of the BFR concentrations at high and low temperatures at the e-waste site (Figure 3). At high temperatures (1930 °C, average daily temperature), the gaseous concentrations (except for hepta-BDEs) were positively correlated with temperatures (p < 0.001), which explained approximately 44%78% (r2) of the variations. The mean slope of the CC plots in the present study was 20 840 ( 3250, which are steeper than those for PBDEs reported from Europe (9110 ( 3940) and North America (6400 ( 600 for BDE47 and 5300 ( 960 for BDE99).25 To date, the temperature dependence of air concentrations associated with e-waste has not been examined. This finding suggests that gaseous concentrations of BFRs in high-temperature seasons at the e-waste site are strongly controlled by temperature-driven evaporation from contaminated surfaces (e-waste, soils, and recycled e-waste remains as mentioned above) in the local surroundings of this site.14,15 This result also has important implications for the global transport from warm climates to colder climates of these chemicals resulting from e-waste recycling (grass-hopper effect). At low temperatures (819 °C) during the winter, many BFRs showed increasing concentrations with declining temperatures, although the linear relationships were not all statistically significant (SI Table S3). The differences in temperature relationships for SOCs in winter and other seasons have been reported in early studies, although the winter temperatures were quite different.14,15,30 These studies attributed the lack of a temperature dependence of SOCs at lower temperatures to LRAT being the dominant sources of the air concentrations. Despite the correlations with temperature for many BFRs in winter in the present study, we think temperature may explain a small fraction at best of the variability of these BFRs in the air. Their occurrence was also unlikely to be a result of LRAT because of the obvious presence of local sources. Instead, we speculated that the elevated concentrations were due to the increasing e-waste recycling activities in winter of the sampling year, which resulted in an increased emission of pollutants. Similar CC plots for PBDEs have also been observed in Chilton, UK, where the high concentrations in winter resulted from seasonal combustion sources.20 At the rural site, BFR concentrations had significant temperature dependence for di-, tri, and tetra-BDEs and PBT but were weak or lacking in temperature dependence for other BFRs (SI Table S3 and Figure S5). The slopes of the CC plots (7185 ( 2500) were significantly flatter than those observed at the e-waste site. A clear decline in slope with increasing distance from the suspected sources have been observed for PCBs.14 A flat slope or low temperature dependence indicates that regional or longrange atmospheric transport controls atmospheric levels at a sampling site.14,15 The e-waste and urban areas are the main sources of BFRs in the rural air as indicated in our recent finding.9 Gas-Particle Partitioning. Gas-particle partitioning is an important process governing the atmospheric fate of SOCs. 8822
dx.doi.org/10.1021/es202284p |Environ. Sci. Technol. 2011, 45, 8819–8825
Environmental Science & Technology
ARTICLE
Figure 3. ClausiusClapeyron (CC) plots of BFRs at the e-waste site. Data of all samples from September 2007 and a sample from February 2008 were excluded because of the exceptionally low and high concentrations, respectively. The open circles represent data from winter.
The partitioning is usually defined by the partition coefficient (KP) KP ¼ ðF=TSPÞ=A
ð2Þ
where F and A are the particle-bound and gaseous concentrations (pg/m3), respectively, and TSP is the concentration of total suspended particles (μg/m3). Two mechanisms (adsorption onto the particle surface and absorption into the organic matter of particles) have been used to interpret the partitioning, both of which lead to a linear relationship between log KP and log PoL: log KP ¼ mlog PLo þ b
ð3Þ
where PoL is subcooled liquid vapor pressure.31,32 Although it has been suggested that the slope m should be near 1 for true equilibrium partitioning for both adsorption and absorption processes, deviations of m for field data do not necessarily indicate nonequilibrium conditions. Instead, the slopes for a given set of compounds rely largely on sorption properties of particles, ambient temperature, and relative humidity.33 The relationships between log KP and temperature-corrected log PoL were investigated separately for PBDEs from different sampling events at the two sites (Table 2 and SI Figures S67). The slopes of Junge-Pankow plots ranged widely from 0.59 to 1.29 at the e-waste site. The influence of temperature and
relative humidity on the varied slopes was evaluated by investigating the correlations between them. The slopes correlated positively with temperature (p = 0.04) (slopes tended to be flat with increasing temperature) but not with relative humidity (SI Figure S8). While possible blow-off sampling artifact could cause this relationship with temperature,34 a likely interpretation for this is that more BFRs with a high vapor pressure would volatilize into the air as ambient temperature increased, and their subsequent partitioning onto/into particles would make the slopes flatter. This interpretation also implies a nonequilibrium of the partitioning. However, temperatures explained only ∼36% of the total variation of the slopes. Another factor was probably attributed to different sorptive properties of the atmospheric particles as a result of their various origins at this site. The particles could originate from e-waste shredding and burning, resuspension of soil and dust, and secondary aerosols formed in the atmosphere. The observed slopes at the e-waste site were not able to indicate adsorption or absorption mechanisms.33 Pankow calculated that intercept values between 7.3 and 8.9 would specify absorptive partitioning. 31 The intercepts b (4.42 to 7.72) were mostly outside this range suggesting a primary adsorption mechanism. The slopes (0.23 to 0.80) for data from the rural site clearly deviated from 1 (except for those with no linear relationships for three events), indicating that absorption into 8823
dx.doi.org/10.1021/es202284p |Environ. Sci. Technol. 2011, 45, 8819–8825
Environmental Science & Technology
ARTICLE
Table 2. Slopes (m), Intercepts (b), r2, p-Values, and Number of Data Points (n) of the Linear Regression of Log KP (m3/μg) versus Log PoL (Pa) of PBDEs in Different Sampling Eventsa sampling event
e-waste site m ( SE
b
b ( SE
rural site 2
r
p-value
n
m ( SE
b ( SE
r2
p-value
n
July 2007
0.83 ( 0.05
5.80 ( 0.15
0.82
<0.001
56
0.59 ( 0.08
4.80 ( 0.30
0.66
<0.001
26
August 2007
0.94 ( 0.07
6.21 ( 0.21
0.75
<0.001
59
0.80 ( 0.08
5.44 ( 0.26
0.72
<0.001
41
September 2007 October 2007
0.74 ( 0.09 0.98 ( 0.06
4.44 ( 0.37 6.10 ( 0.20
0.69 0.77
<0.001 <0.001
27 70
0.37 ( 0.04
NSc 4.16 ( 0.14
0.60
<0.001
26 49
November 2007
1.21 ( 0.07
6.87 ( 0.25
0.80
<0.001
67
0.23 ( 0.08
3.03 ( 0.29
0.23
0.004
34
December 2007
1.09 ( 0.05
6.62 ( 0.18
0.89
<0.001
59
0.37 ( 0.07
3.96 ( 0.28
0.43
<0.001
35 34
January 2008
1.29 ( 0.07
7.72 ( 0.29
0.83
<0.001
66
0.35 ( 0.11
3.44 ( 0.45
0.33
0.003
February 2008
1.02 ( 0.07
6.11 ( 0.27
0.78
<0.001
60
0.36 ( 0.10
3.46 ( 0.41
0.27
<0.001
March 2008
0.93 ( 0.05
5.31 ( 0.17
0.83
<0.001
66
April 2008
0.59 ( 0.08
4.41 ( 0.31
0.66
<0.001
30
0.24 ( 0.09
3.36 ( 0.36
0.28
0.013
21
May 2008 June 2008
0.92 ( 0.07 1.06 ( 0.06
5.74 ( 0.24 6.58 ( 0.17
0.75 0.84
<0.001 <0.001
55 60
0.39 ( 0.05
3.82 ( 0.18 NS
0.53
<0.001
49 41
NS
40 27
a
Only BFRs that had concentrations above the detection limit in both phases were included in the KP calculations. The PoL values of PBDEs were obtained from the literature.35 b Standard Error. c Not Significant.
organic matter may dominate gas-particle partitioning of BFRs at this site.33 The slopes decreased with increasing temperature, although the relationship was not statistically significant. Volatilization of more volatile BFRs from particles as the temperature increased would result in steeper slopes. The results also suggest a different particle source at the rural site from the e-waste site. Particles from nearby urban region that may be enriched with carbonaceous particles are a significant source at the rural site,9 in agreement with the different seasonal variations in the particlebound BFRs at this site. This study provides insight into the emission mechanisms of BFRs in the air associated with e-waste recycling in southern China. The findings showed that e-waste recycling in southern China is as an important source of less brominated BFRs and their release is strongly temperature-dependent. However, this study has limitations in (i) clearly elucidating the emission mechanisms of highly brominated BFRs that are dominantly particle-bound related to e-waste recycling and (ii) determining the atmospheric particle properties (e.g., aerodynamic size and organic/black carbon content) in the study area, which should be addressed in future research.
’ ASSOCIATED CONTENT
bS
Supporting Information. Detailed sample collection, materials, sample preparation, instrumental analysis, data analysis and additional figures and tables. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +86-20-85290146; fax: +86-20-85290706; e-mail:
[email protected].
’ ACKNOWLEDGMENT This study was financially supported by the National Science Foundation of China (No. 40801199, 40821003, 41073078, and 40632012) and the Earmarked Fund of the State Key Laboratory of Organic Geochemistry (SKLOG2009A04). This is contribution No. IS-1388 from GIGCAS.
’ REFERENCES (1) Hites, R. A. Polybrominated diphenyl ethers in the environment and in people: A meta-analysis of concentrations. Environ. Sci. Technol. 2004, 38, 945–956. (2) United Nations Environment Programme (UNEP). Stockholm Convention text and annexes as amended in 2009. Available at http:// chm.pops.int/Convention/tabid/54/language/en-US/Default.aspx#convtext (accessed 22 November 2010). (3) de Wit, C. A.; Herzke, D.; Vorkamp, K. Brominated flame retardants in the Arctic environment—Trends and new candidates. Sci. Total Environ. 2010, 408, 2885–2918. (4) Covaci, A.; Harrad, S.; Abdallah, M. A.-E.; Ali, N.; Law, R. J.; Herzke, D.; Wit, C. A. d. Novel brominated flame retardants: A review of their analysis, environmental fate and behaviour. Environ. Int. 2011, 37, 532–556. (5) Basel Action Network (BAN) and Silicon Valley Toxics Coalition (SVTC). Exporting Harm: The High-Tech Trashing of Asia; BAN: Seattle, WA; SVTC: San Jose, CA, 2002; www.ban.org/E-waste/technotrashfinalcomp.pdf. (6) Leung, A. O. W.; Luksemburg, W. J.; Wong, A. S.; Wong, M. H. Spatial distribution of polybrominated diphenyl ethers and polychlorinated dibenzo-p-dioxins and dibenzofurans in soil and combusted residue at Guiyu, an electronic waste recycling site in southeast China. Environ. Sci. Technol. 2007, 41, 2730–2737. (7) Ma, J.; Addink, R.; Yun, S. H.; Cheng, J. P.; Wang, W. H.; Kannan, K. Polybrominated dibenzo-p-dioxins/dibenzofurans and polybrominated diphenyl ethers in soil, vegetation, workshop-floor dust, and electronic shredder residue from an electronic waste recycling facility and in soils from a chemical industrial complex in eastern China. Environ. Sci. Technol. 2009, 43, 7350–7356. (8) Muenhor, D.; Harrad, S.; Ali, N.; Covaci, A. Brominated flame retardants (BFRs) in air and dust from electronic waste storage facilities in Thailand. Environ. Int. 2010, 36, 690–698. (9) Tian, M.; Chen, S. J.; Wang, J.; Shi, T.; Luo, X. J.; Mai, B. X. Atmospheric deposition of halogenated flame retardants at urban, e-Waste, and rural locations in southern China. Environ. Sci. Technol. 2011, 45, 4696–4701. (10) Zhang, J. Q.; Jiang, Y. S.; Zhou, J.; Wu, B.; Liang, Y.; Peng, Z. Q.; Fang, D. K.; Liu, B.; Huang, H. Y.; He, C.; Wang, C. L.; Lu, F. N. Elevated body burdens of PBDEs, dioxins, and PCBs on thyroid hormone homeostasis at an electronic waste recycling site in China. Environ. Sci. Technol. 2010, 44, 3956–3962. (11) Chen, L. G.; Mai, B. X.; Bi, X. H.; Chen, S. J.; Wang, X. M.; Ran, Y.; Luo, X. J.; Sheng, G. Y.; Fu, J. M.; Zeng, E. Y. Concentration levels, 8824
dx.doi.org/10.1021/es202284p |Environ. Sci. Technol. 2011, 45, 8819–8825
Environmental Science & Technology compositional profiles, and gas-particle partitioning of polybrominated diphenyl ethers in the atmosphere of an urban city in South China. Environ. Sci. Technol. 2006, 40, 1190–1196. (12) Zhang, G.; Chakraborty, P.; Li, J.; Sampathkumar, P.; Balasubramanian, T.; Kathiresan, K.; Takahashi, S.; Subramanian, A.; Tanabe, S.; Jones, K. C. Passive atmospheric sampling of organochlorine pesticides, polychlorinated biphenyls, and polybrominated diphenyl ethers in urban, rural, and wetland sites along the coastal length of India. Environ. Sci. Technol. 2008, 42, 8218–8223. (13) Wania, F. Assessing the potential of persistent organic chemicals for long-range transport and accumulation in polar regions. Environ. Sci. Technol. 2003, 37, 1344–1351. (14) Wania, F.; Haugen, J. E.; Lei, Y. D.; Mackay, D. Temperature dependence of atmospheric concentrations of semivolatile organic compounds. Environ. Sci. Technol. 1998, 32, 1013–1021. (15) Hoff, R. M.; Brice, K. A.; Halsall, C. J. Nonlinearity in the slopes of ClausiusClapeyron plots for SVOCs. Environ. Sci. Technol. 1998, 32, 1793–1798. (16) Deng, W. J.; Zheng, J. S.; Bi, X. H.; Fu, J. M.; Wong, M. H. Distribution of PBDEs in air particles from an electronic waste recycling site compared with Guangzhou and Hong Kong, South China. Environ. Intern. 2007, 33, 1063–1069. (17) Chen, D. H.; Bi, X. H.; Zhao, J. P.; Chen, L. G.; Tan, J. H.; Mai, B. X.; Sheng, G. Y.; Fu, J. M.; Wong, M. H. Pollution characterization and diurnal variation of PBDEs in the atmosphere of an e-waste dismantling region. Environ. Pollut. 2009, 157, 1051–1057. (18) Birnbaum, L. S.; Staskal, D. F. Brominated flame retardants: Cause for concern? Environ. Health Perspect. 2004, 112, 9–17. (19) La Guardia, M. J.; Hale, R. C.; Harvey, E. Detailed polybrominated diphenyl ether (PBDE) congener composition of the widely used penta-, octa-, and deca-PBDE technical flame-retardant mixtures. Environ. Sci. Technol. 2006, 40, 6247–6254. (20) Lee, R. G. M.; Thomas, G. O.; Jones, K. C. PBDEs in the atmosphere of three locations in western Europe. Environ. Sci. Technol. 2004, 38, 699–706. (21) Hoh, E.; Hites, R. A. Brominated flame retardants in the atmosphere of the East-Central United States. Environ. Sci. Technol. 2005, 39, 7794–7802. (22) Bossi, R.; Skov, H.; Vorkamp, K.; Christensen, J.; Rastogi, S. C.; Egelov, A.; Petersen, D. Atmospheric concentrations of organochlorine pesticides, polybrominated diphenyl ethers and polychloronaphthalenes in Nuuk, South-West Greenland. Atmos. Environ. 2008, 42, 7293–7303. (23) Cetin, B.; Odabasi, M. Air-water exchange and dry deposition of polybrominated diphenyl ethers at a coastal site in Izmir Bay, Turkey. Environ. Sci. Technol. 2007, 41, 785–791. (24) Raff, J. D.; Hites, R. A. Deposition versus photochemical removal of PBDEs from Lake Superior air. Environ. Sci. Technol. 2007, 41, 6725–6731. (25) Venier, M.; Hites, R. A. Flame retardants in the atmosphere near the Great Lakes. Environ. Sci. Technol. 2008, 42, 4745–4751. (26) Su, Y. S.; Hung, H.; Brice, K. A.; Su, K.; Alexandrou, N.; Blanchard, P.; Chan, E.; Sverko, E.; Fellin, P. Air concentrations of polybrominated diphenyl ethers (PBDEs) in 20022004 at a rural site in the Great Lakes. Atmos. Environ. 2009, 43, 6230–6237. (27) Qiu, X. H.; Zhu, T.; Hu, J. X. Polybrominated diphenyl ethers (PBDEs) and other flame retardants in the atmosphere and water from Taihu Lake, East China. Chemosphere 2010, 80, 1207–1212. (28) Gouin, T.; Thomas, G. O.; Cousins, I.; Barber, J.; Mackay, D.; Jones, K. C. Air-surface exchange of polybrominated diphenyl ethers and polychlorinated biphenyls. Environ. Sci. Technol. 2002, 36, 1426–1434. (29) Su, Y.; Hung, H.; Sverko, E.; Fellin, P.; Li, H. Multi-year measurements of polybrominated diphenyl ethers (PBDEs) in the Arctic atmosphere. Atmos. Environ. 2007, 41, 8725–8735. (30) Harrad, S.; Mao, H. J. Atmospheric PCBs and organochlorine pesticides in Birmingham, UK: Concentrations, sources, temporal and seasonal trends. Atmos. Environ. 2004, 38, 1437–1445.
ARTICLE
(31) Pankow, J. F. An absorption model of gas/particle partitioning of organic compounds in the atmosphere. Atmos. Environ. 1994, 28, 185–188. (32) Lohmann, R.; Harner, T.; Thomas, G. O.; Jones, K. C. A comparative study of the gas-particle partitioning of PCDD/Fs, PCBs, and PAHs. Environ. Sci. Technol. 2000, 34, 4943–4951. (33) Goss, K. U.; Schwarzenbach, R. P. Gas/solid and gas/liquid partitioning of organic compounds: Critical evaluation of the interpretation of equilibrium constants. Environ. Sci. Technol. 1998, 32, 2025– 2032. (34) Su, Y.; Lei, Y. D.; Wania, F.; Shoeib, M.; Harner, T. Regressing gas/particle partitioning data for polycyclic aromatic hydrocarbons. Environ. Sci. Technol. 2006, 40, 3558–3564. (35) Wong, A.; Lei, Y. D.; Alaee, M.; Wania, F. Vapor pressures of the polybrominated diphenyl ethers. J. Chem. Eng. Data 2001, 46, 239–242.
8825
dx.doi.org/10.1021/es202284p |Environ. Sci. Technol. 2011, 45, 8819–8825
ARTICLE pubs.acs.org/est
Effect of Aqueous Fe(II) on Arsenate Sorption on Goethite and Hematite Jeffrey G. Catalano,* Yun Luo,† and Bamidele Otemuyiwa Department of Earth and Planetary Sciences, Washington University, St. Louis, Missouri 63130, United States
bS Supporting Information ABSTRACT: Biogeochemical iron cycling often generates systems where aqueous Fe(II) and solid Fe(III) oxides coexist. Reactions between these species result in iron oxide surface and phase transformations, iron isotope fractionation, and redox transformations of many contaminant species. Fe(II)-induced recrystallization of goethite and hematite has recently been shown to cause the repartitioning of Ni(II) at the mineral water interface, with adsorbed Ni incorporating into the iron oxide structure and preincorporated Ni released back into aqueous solution. However, the effect of Fe(II) on the fate and speciation of redox inactive species incompatible with iron oxide structures is unclear. Arsenate sorption to hematite and goethite in the presence of aqueous Fe(II) was studied to determine whether Fe(II) causes substantial changes in the sorption mechanisms of such incompatible species. Sorption isotherms reveal that Fe(II) minimally alters macroscopic arsenate sorption behavior except at circumneutral pH in the presence of elevated concentrations (10 3 M) of Fe(II) and at high arsenate loadings, where a clear signature of precipitation is observed. Powder X-ray diffraction demonstrates that the ferrous arsenate mineral symplesite precipitates under such conditions. Extended X-ray absorption fine structure spectroscopy shows that outside this precipitation regime arsenate surface complexation mechanisms are unaffected by Fe(II). In addition, arsenate was found to suppress Fe(II) sorption through competitive adsorption processes before the onset of symplesite precipitation. This study demonstrates that the sorption of species incompatible with iron oxide structure is not substantially affected by Fe(II) but that such species may potentially interfere with Fe(II)-iron oxide reactions via competitive adsorption.
’ INTRODUCTION Biogeochemical processes in aquatic, soil, groundwater, and sedimentary environments often involve redox transformations of Fe. In such systems aqueous Fe(II) and solid Fe(III) oxide minerals commonly coexist.1,2 Reaction between Fe species in these two oxidation states transforms ferrihydrite into more crystalline iron oxides,3 5 fractionates iron isotopes,6 8 and activates localized growth and dissolution of iron oxide surfaces.9 11 Recent work has demonstrated that Fe(II) oxidatively adsorbs to iron oxides12 15 in a process that involves atom exchange and electron transfer.16 18 Heterogeneous systems containing Fe(III) oxides have been shown to catalyze the reduction of contaminants such as U(VI), Cr(VI), Tc(VII), halogenated hydrocarbons, and nitroaromatic compounds by aqueous Fe(II).19 24 Such systems also appear to catalyze the oxidation of As(III) to As(V),25 explaining the enhanced sorption of As(III) in the presence of Fe(II).26 The mechanism for this process is still being debated as in the presence of magnetite and ferrihydrite O2 is required for oxidation to occur,27 which is attributed to a Fenton-type reaction.28 Less clear is the effect of Fe(II) on the fate of redox-inactive species at iron oxidewater interfaces. Fe(II) has been shown to cause hysteresis in the adsorption of divalent metals onto goethite29 although a later study did not observed a similar effect on hematite.30 We have recently demonstrated that Fe(II) may substantially alter the fate of Ni(II) in the presence of goethite and hematite, causing the incorporation r 2011 American Chemical Society
of adsorbed Ni into the mineral structure and the release of previously incorporated Ni back into solution.31 This process is ultimately driven by iron oxide recrystallization induced by Fe(II). Ni and other divalent metal cations are structurally compatible with iron oxide minerals and are known to substitute at levels of up to a few mol %.32 35 Other contaminant species, especially those that occur as oxoanions, lack this structural compatibility and the effect of Fe(II)-induced iron oxide recrystallization on their fate is unclear. In the present study the sorption of arsenate on hematite and goethite in the presence of aqueous Fe(II) is investigated to explore how such recrystallization affects a structurally incompatible contaminant species. Arsenate was selected as it generally cannot be abiotically reduced by Fe(II),36 with As(III) oxidation the apparently favored reaction direction in such systems.25 Macroscopic signatures of the alteration of arsenate sorption behavior by Fe(II) were examined using sorption isotherms. In addition, changes in arsenate sorption mechanisms were evaluated using X-ray absorption fine structure (XAFS) spectroscopy and powder X-ray diffraction (XRD). Sorption isotherms were also employed to characterize the effect of arsenate on macroscopic Fe(II) sorption behavior. Received: July 15, 2011 Accepted: September 7, 2011 Revised: August 29, 2011 Published: September 07, 2011 8826
dx.doi.org/10.1021/es202445w | Environ. Sci. Technol. 2011, 45, 8826–8833
Environmental Science & Technology
’ MATERIALS AND METHODS Mineral Synthesis. Hematite and goethite were prepared following previously described procedures.37 Goethite was prepared by hydrolysis of a Fe(NO3)3 solution with KOH followed by aging at 70 °C for 3 days. Hematite was prepared by aging an acidic FeCl3 solution at 98 °C for 7 d. Both solids were washed repeatedly in deionized water (>18.2 MΩ cm) to remove excess electrolytes. The minerals were suspended in deionized water at concentrations of 20 35 g L 1 and then transferred into an anaerobic chamber (Coy Laboratory Products) with a 4% H2/96% N2 atmosphere and a Pd catalyst to remove residual O2(g). A secondary gas filtration system was used to help maintain low-O2 conditions and to remove CO2 from the chamber atmosphere. The system consisted of two gas washing bottles in sequence through which the chamber atmosphere was bubbled. The first contained a 15% pyrogallol/ 50% KOH solution to remove O2 and CO2; the second contained deionized water. The mineral suspensions were bubbled with the gas stream from this system overnight to remove dissolved O2 and CO2. After bubbling the dissolved oxygen content was measured colorimetrically (CHEMetrics test kit K-7511); all analyses were below the 1 μg L 1 detection limit. An aliquot was taken from each mineral suspension before use for gravimetric determination of the suspension concentration. Specific surface areas of 13 and 39 m2 g 1 for hematite and goethite, respectively, were determined using N2 BET adsorption isotherms at 77 K (Quantachrome Instruments Autosorb-1). Substantial clumping of the hematite was observed upon air drying prior to the BET measurements so the specific surface area may be underestimated. Scanning electron microscopy (SEM) images (JEOL JSM-7001F FE-SEM) of hematite (Supporting Information (SI) Figure S1) suggest a geometric specific surface area of approximately 22 m2 g 1. Batch Isotherm Measurements. Arsenate and Fe(II) sorption isotherms were measured at pH 4 and 7 on both minerals. Reaction times of 5 days were chosen as Ni incorporation into hematite and goethite in the presence of Fe(II) was found to require such time scales to occur.31 At final volumes all samples contained 4 g L 1 goethite or hematite, 10 2 M NaCl, and a buffer of either 2-(N-morpholino)ethanesulfonic acid (MES) at pH 4 or 4-(2-hydroxyethyl)-1-piperazinepropanesulfonic acid (EPPS) at pH 7 at a concentration of 10 3 M. The As(V) isotherms were measured in the presence of 0, 10 4 M, or 10 3 M FeCl2; the Fe(II) isotherms were determined in the presence of 0 or 5 10 4 M Na2HAsO4. Each data point corresponds to a separate sample; the pH values of the samples in each varied by less than (0.1 pH units from the target pH. All samples were made by dilution of stock solutions prepared in the anaerobic chamber from reagent grade chemicals and deoxygenated deionized water. FeCl2 stock solutions were filtered and acidified to pH 2 using HCl, stored in opaque bottles, and used within 7 days; a test Fe(II) solution showed no signs of oxidation after storage for 60 days in the anaerobic chamber. pH adjustments were made using HCl and NaOH that were stored for at least 90 days in the chamber and had been bubbled following the same procedure as was used for the mineral suspensions. All samples were wrapped in aluminum foil to prevent light exposure and placed on end-overend rotators for the duration of the reaction. At the end of the reaction period the samples were centrifuged in the anaerobic chamber and the supernatant was decanted, filtered (0.2 μm, MCE), and acidified with trace metal grade HNO3. Dissolved As and Fe concentrations were determined by inductively coupled plasma optical emission spectrometry (ICP-OES) using a Perkin-Elmer Optima 7300 DV.
ARTICLE
Aliquots from select samples were collected prior to acidification for analysis of dissolved Fe(II) using the ferrozine method.38 These analyses yielded dissolved Fe(II) concentrations within <1% of the total Fe concentration determined by ICP-OES, less than the expected systematic error in the measurements, indicating that dissolved Fe(III) concentrations were insignificant. Thermodynamic Calculations. Calculations of mineral solubility were made in The Geochemist’s Workbench39 using the Lawrence Livermore National Laboratory thermochemical database.40 This database employs an extended Debye H€uckel activity correction model that is parametrized to be accurate in up to 3 m NaCl solution and approximately 0.5 to 1 m ionic strengths of other electrolytes.41 43 Complexation constants from Langmuir et al.44 and a recent value for the solubility constant of symplesite, Fe3(AsO4)2 3 8H2O,45 were added to the database. The complexation of arsenate by Fe(II) was suppressed for the calculation of symplesite saturation index as this was not accounted for in the original determination of the solubility product.45 Characterization of Arsenic Speciation. A series of samples were prepared for characterization of arsenic speciation (Table 1). All samples were prepared following identical procedures as used in the isotherm measurements. Six samples for each mineral (adsorption samples: G1-G6 and H1-H6) were prepared under conditions where the sorption isotherms suggested that arsenate uptake was dominated by adsorption and not precipitation. The wet mineral paste remaining after the supernatant was decanted at the end of the reaction period was collected for these samples and loaded into a polycarbonate sample holder sealed with Kapton tape. These were then heat sealed in polyethylene bags to preserved anaerobic conditions and then transported to the Advanced Photon Source at Argonne National Laboratory for collection of XAFS spectra. Data collection was conducted at beamline 20-BM (PCN/XSD) using a Si (111) double-crystal monochromator. Fluorescence yield As K-edge XAFS spectra were collected using a 12-element energy dispersive Ge array detector. The harmonic content of the X-ray beam was reduced by detuning the second crystal of the monochromator by 10% and by insertion of a Rh-coated harmonic rejection mirror 1 m before the sample. The incident X-ray beam was also focused both vertically and horizontally to an approximately 700 700 μm size using a toroidal mirror coated with Pt and a 10 nm Al2O3 overcoat; the focusing mirror is located 2 m past the monochromator. Focusing was done to increase the usable flux from the bending magnet (approximately 1011 ph/s) to improve the counting statistics. The X-ray beam energy was calibrated using a Pt metal foil with the first inflection point in the Pt LIII-edge set to 11.564 keV. XAFS data were processed using the Athena46 and SixPack47 interfaces to IFEFFIT.48 Backscattering phase and amplitude functions for structural analysis of the extended X-ray absorption fine structure (EXAFS) spectra were calculated using FEFF 7.0249 from the structure of scorodite, FeAsO4 3 2H2O.50 The amplitude reduction factor, S02, was fixed to 1.0 for fitting. All spectra for each mineral were refined versus a single starting structural model consisting of As O and As Fe single scattering paths and three multiple scattering paths: a triangular As O O path, a collinear As O As O path, and a noncollinear As O As O path. The parameters for these paths were determined from geometric considerations and the parameters for the As O single scattering path and thus did not require the introduction of additional variables. Details regarding the inclusion of these multiple scattering paths have been recently described.51 For the As Fe single scattering path, σ2 (a Debye Waller-type factor based on a Gaussian distribution of interatomic distances) was fixed to 0.004 to reduce correlations during fitting. 8827
dx.doi.org/10.1021/es202445w |Environ. Sci. Technol. 2011, 45, 8826–8833
Environmental Science & Technology
ARTICLE
Table 1. Conditions for the Adsorption and Precipitation Samples Examined by EXAFS Spectroscopy and XRD sample
mineral
pH
4gL
1
G2
4gL
1
goethite
4.07
G3
4gL
1
goethite
G4
4gL
1
G5
4gL
1
G6
4gL
1
goethite
G7
4gL
1
G8 H1
4gL 4gL
1
H2
[Fe2+]init (M)
[AsO43-]init (M)
ΓAs (10-3 mol g 1)
SIa
10
4
0.025
10
4
10
4
0.025
14.4
4.07
10
3
10
4
0.025
12.0
goethite
6.11
0
10
4
0.025
goethite
7.21
10
4
10
4
0.025
12.0
6.03
10
3
10
4
0.025
6.3
goethite
4.00
10
3
10
3
0.107
4.2
goethite 1 hematite
6.97 4.17
10 0
3
10 10
3
0.176 0.025
1.0
4gL
1
4.04
10
4
10
4
0.025
17.2
H3
4gL
1
hematite
4.88
10
3
10
4
0.025
9.2
H4
4gL
1
hematite
7.23
0
10
4
0.025
G1
goethite
hematite
4.03
0
4
H5
4gL
1
hematite
7.79
10
4
10
4
0.025
7.0
H6
4gL
1
hematite
7.13
10
3
10
4
0.025
2.5
H7
4gL
1
hematite
4.05
10
3
10
3
0.080
3.9
H8
4gL
1
hematite
6.97
10
3
10
3
0.153
0.9
Saturation index [log (Q/K)] of symplesite, Fe3(AsO4)2 3 8H2O, at the final pH and concentration conditions; the calculations accounted for activity corrections. a
Additional samples were prepared at higher As(V) concentrations, where the isotherms suggest the formation of a precipitate at pH 7 (precipitation samples: G7, G8, H7, and H8). These samples were examined by powder XRD on a Rigaku Geigerflex D-MAX/A diffractometer using Cu Kα radiation, a graphite diffracted beam monochromator, and a scintillation detector. A step size of 0.04° 2θ was used with a counting time of 3 s per point for full patterns and 30 s per point for narrow scans around select diffraction lines.
’ RESULTS Arsenate Isotherms. In the absence of aqueous Fe(II) at pH 4 and 7 on both hematite and goethite arsenate shows Langmuirian sorption behavior (Figure 1). Linearization of the data confirms that the sorption can be described with a Langmuir isotherm (SI Figure S2). Aqueous Fe(II) does not substantially alter the macroscopic sorption behavior of arsenate except at pH 7 in the presence of 10 3 mol L 1 Fe(II). At this high initial Fe(II) concentration arsenate sorption follows the sorption behavior seen in the absence of Fe(II) at low arsenate concentration but displays a second, rapid rise in sorption at higher concentration (Figure 1). This rise correlates with a drop in the aqueous Fe(II) concentrations (SI Figure S3) and is suggestive of a change in sorption mechanism from adsorption to precipitation. For conditions where the isotherms do not suggest precipitation the effect of aqueous Fe(II) is minimal. Arsenate Sorption Mechanisms. A series of adsorption samples (Table 1) were characterized using XAFS spectroscopy to determine the effect of Fe(II) on the adsorption mechanisms of arsenate. Conditions were selected where the final sample conditions were undersaturated with respect to symplesite based on the isotherm measurements. XANES spectra (SI Figure S4) demonstrate that no arsenate reduction occurs. EXAFS spectra (Figure 2) suggest differences in arsenate surface complexation between hematite and goethite in the absence of Fe(II), with arsenate adsorbed on hematite having a lower number of apparent Fe neighbors at ∼3.3 Å (SI Table S2), consistent with past
Figure 1. Arsenate sorption isotherms on goethite at (A) pH 4 and (B) pH 7 and on hematite at (C) pH 4 and (D) pH 7. Also shown are Langmuir isotherm fits to all data sets lacking a clear precipitation feature.
studies.52 54 The As Fe distances indicate the presence of bridging bidentate inner-sphere surface complexes on both minerals. The number of Fe neighbors at ∼3.3 Å on both minerals is less than 2, the expected value if all of the sorbed arsenate was in a single bridging bidentate surface complex. This could indicate the presence of outer-sphere arsenate surface complexes,55,56 but the Fe coordination number is highly dependent on the choice of σ2 (i.e., use of a larger σ2 would yield a larger Fe coordination number) and the occurrence of such species thus cannot be determined unambiguously. The lower Fe coordination number 8828
dx.doi.org/10.1021/es202445w |Environ. Sci. Technol. 2011, 45, 8826–8833
Environmental Science & Technology
ARTICLE
Figure 2. Data (dotted) and structural fits (solid) to the As K-edge EXAFS spectra (left), Fourier transform magnitudes (center), and real components of the Fourier transforms (right) of the series of adsorption samples. Detailed sample information is provided in Table 1.
on hematite could indicate a larger proportion of outer-sphere species but also could indicate arsenate binding at two distinct surface sites producing slightly different As Fe distances. If these two distances differ by less than the resolution of an EXAFS spectrum then their contributions to the spectrum destructively interfere with each other, giving the appearance of a single Fe neighbor with a lower coordination number. Such a phenomenon has been used in previous studies to explain the EXAFS spectra of arsenate53 and uranyl57 adsorbed to hematite. The number of Fe neighbors and the distance to those neighbors are statistically invariant at pH 4 and 7 for each mineral (Figure 3), suggesting little effect of pH on surface complexation geometry over the range studied. Fe(II) has no effect on arsenate surface complexation mechanism as the coordination number of and distance to the Fe shell is statistically invariant on each mineral over all Fe(II) and pH conditions examined (Figure 3). The sorption isotherms (Figure 1) have features at the highest initial Fe(II) concentration at pH 7 on both minerals that are consistent with the precipitation of arsenate. Thermodynamic calculations indicate that at the end of the reaction time symplesite is slightly supersaturated under such conditions (Table 1). XRD patterns (SI Figure S5) were collected on samples supersaturated
with respect to symplesite at pH 7; samples with identical initial Fe(II) and arsenate concentrations reacted at pH 4 were also examined for comparison. Only goethite or hematite peaks are observed in the pH 4 samples. For the pH 7 samples a small peak appears near 13° 2θ. Narrow scans around this position with longer counting times to improve sensitivity to minor phases (Figure 4) reveals a clear feature at pH 7 that corresponds to the strongest powder XRD line for symplesite.58 XRD measurements were not conducted on the samples examined by XAFS spectroscopy because the lower As concentration would make detection of an As-bearing precipitate difficult. However, the EXAFS spectra (Figure 2) clearly demonstrate that arsenate exists in an adsorbed form that is unaffected by the presence of Fe(II). Further, these spectra clearly differ from the EXAFS spectrum of symplesite.59 Fe(II) Isotherms. Isotherms were collected to explore the response of Fe(II) sorption to the presence of arsenate (Figure 5). In the absence of arsenate these are generally consistent with past isotherms, including minor adsorption at pH 4 after 5 days of reaction, 22,29,60,61 and are well described by Langmuir isotherms (Figure 5). In the presence of arsenate the isotherms appear similar, with sorption slightly increasing (Figure 5, SI Table S3) except at 8829
dx.doi.org/10.1021/es202445w |Environ. Sci. Technol. 2011, 45, 8826–8833
Environmental Science & Technology
ARTICLE
low Fe(II) concentrations at pH 7 where it appears to be suppressed (Figure 5B,D), with a greater equilibrium Fe(II) concentration for the same surface coverage. This is also apparent in the linearized isotherms (SI Figure S6) with data points at low Fe(II) concentration deviating from linear behavior and appearing scattered along the y-axes of the diagrams. This observation does not appear to be an analytical or experimental artifact as arsenate sorption does not cause a detectable Fe signal in ICP-OES in the absence of Fe(II). The response of arsenate concentrations to increasing Fe(II) concentrations (Figure S7) suggests that while the total amount of sorption is little affected by the presence of arsenate the mechanism of sorption may change at pH 7. Arsenate concentrations are largely unaffected by Fe(II) at pH 4 but drop substantially with increasing Fe(II) concentrations at pH 7. The solutions are slightly supersaturated with respect to symplesite under these conditions, suggesting that Fe(II) precipitation is a substantial contribution to the sorption isotherms at pH 7 in the presence of arsenate. Figure 3. EXAFS-derived As Fe interatomic distances (RFe) and the coordination number of second-shell Fe atoms (NFe) for the series of adsorption samples. Also plotted are the mean values for each group (dashed). Detailed sample information is provided in Table 1, and the full fitting parameters are available in SI Table S2.
Figure 4. Powder XRD patterns of the precipitation samples in the region of the symplesite (020) reflection (vertical line). Detailed sample information is provided in Table 1.
’ DISCUSSION Effect of Fe(II) on As(V) Sorption. The effects of Fe(II) on the sorption behavior of arsenate are limited to thermodynamically controlled adsorption and precipitation processes. EXAFS spectroscopy demonstrates that Fe(II) does not cause the incorporation of arsenate into the mineral structure or a change in surface complex geometry. Arsenate adsorption does increase by up to 5 10% in the presence of Fe(II) (Figure 1). This is consistent with Fe(II) adsorption shifting the surface potential to a more positive value, which reduces the As(V) surface complex electrostatic activity coefficient and thus increases the adsorbed concentration. Aqueous Fe(II) concentrations show a slight decrease with increasing arsenate surface coverage (SI Figure S3), also consistent with surface potential changes affecting adsorption. At high Fe(II) concentrations (10 3 M or greater) under circumneutral to alkaline conditions the ferrous arsenate mineral symplesite precipitates, but this is purely a thermodynamic effect as the solutions from which the mineral formed were initially highly supersaturated. The observation of substantial symplesite supersaturation at the end of the reaction time in all samples where isotherms show a precipitation signature suggests that equilibration of a fluid with symplesite requires longer than 5 days or that the thermodynamic data available underpredicts its solubility. The observed symplesite solubility was outside the uncertainty limit of the published solubility product,45 suggesting that the determination of this value may need to be revisited,
Figure 5. Iron sorption isotherms on (A,B) goethite and (C,D) hematite and the corresponding Langmuir isotherm fit to each data set. Suppression of Fe(II) adsorption in the presence of arsenate occurs at low Fe(II) concentrations (B,D). 8830
dx.doi.org/10.1021/es202445w |Environ. Sci. Technol. 2011, 45, 8826–8833
Environmental Science & Technology perhaps with a more comprehensive study that approaches equilibrium from both supersaturated and undersaturated conditions. Arsenate complexation by Fe(II) is relatively weak44 and cannot explain the discrepancy between the observed and calculate solubility of symplesite. Effect of Arsenate on Fe(II) Sorption. Arsenate has a limited effect on Fe(II) sorption behavior although it may alter the mechanisms involved. Arsenate adsorption should substantially shift the surface potential to a more negative value, enhancing Fe(II) adsorption. A minor shift in Fe(II) sorption is generally seen in the presence of arsenate except at pH 7 on hematite, where Fe(II) sorption slightly decreases. It is unclear if the latter observation is real or results from a systematic error such as an inaccurate hematite suspension concentration. The most substantial effect is the inhibition of Fe(II) sorption by arsenate at low Fe(II) concentrations at pH 7. This effect cannot be the result of solution complexation as Fe(II) As(V) complexation is weak.44 However, this may be explained by competitive adsorption with arsenate for a limited number of surface sites. Fe(II) sorption then increases once a threshold concentration is reached where symplesite become saturated and precipitates. Arsenate concentrations stay relatively constant until an Fe(II) concentration of approximately 3 10 5 mol L 1 is reached, at which point concentrations drop precipitously (SI Figure S7). This behavior is consistent with symplesite precipitation if the solubility product is increased by 2 orders of magnitude above the published value.45 Because of mass balance constraints symplesite precipitation can only yield a sorbed Fe(II) concentration of at most 0.1875 mol g 1. The additional Fe(II) sorption (Figure 5) must then results from adsorption to goethite or hematite; symplesite precipitation lowers the adsorbed arsenate concentration, reducing the inhibition of Fe(II) adsorption from competitive adsorption processes. Implications for Environmental Systems. While Fe(II) may substantially alter the speciation and fate of Ni and other redoxinactive elements that are structurally compatible with iron oxide minerals,29,31 it has only a minor effect on the fate of arsenate other than to cause its precipitation as symplesite when conditions are supersaturated. This is consistent with the finding that arsenate reduction is the dominant process that controls As release from iron oxides62 as generation of Fe(II) through reductive dissolution should not appreciably affect arsenate adsorption. In contrast, the present data suggests that arsenate substantially affects Fe(II) adsorption by blocking surface sites. Arsenate is known to inhibit the Fe(II)-catalyzed transformation of ferrihydrite to goethite63 and other strongly sorbing oxoanions have a similar effect.64,65 Arsenate may thus potentially inhibit interfacial electron transfer processes at iron oxide surfaces even though its adsorption is weakly affected by aqueous Fe(II). The present work further highlights the clear relationship between the fate of inorganic, redox-inactive contaminants and their structural compatibility with iron oxides during Fe biogeochemical cycling. Divalent metal cations like Ni commonly substitute into the structure of iron oxides.32,33,35 Despite often being employed in studies of As release,66,67 evidence is lacking that arsenate coprecipitated with iron oxides actually substitutes into crystal structures rather than simply binding to surfaces. The chemically analogous phosphate has been incorporated into hematite68 through a deliberate synthesis process at 100 °C but this has not been demonstrated under conditions relevant to biogeochemical systems. This suggests that arsenate is generally not compatible with the structure of iron oxides and will thus not
ARTICLE
be substantially affected by atom exchange and electron transfer reactions between aqueous Fe(II) and Fe(III) oxides as occurs for divalent metals.
’ ASSOCIATED CONTENT
bS
Supporting Information. Tables describing the fitting parameters for the sorption isotherms and EXAFS spectra, SEM images of the synthesized iron oxides, linearized sorption diagrams, additional elemental concentrations in sorption experiments, XRD patterns over a larger 2θ range, and As K-edge XANES spectra of adsorption samples. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +1-314-935-6015; fax: +1-314-935-7361; e-mail: catalano@ eps.wustl.edu. Present Addresses †
Gemological Institute of America, Carlsbad, California 92008.
’ ACKNOWLEDGMENT This research was supported by the National Science Foundation, Division of Earth Sciences, Geobiology and Low-temperature Geochemistry Program through Grant No. EAR-0818354. PNC/XSD facilities at the Advanced Photon Source, and research at these facilities, are supported by the U.S. Department of Energy (DOE)—Basic Energy Sciences, a Major Resources Support grant from NSERC, the University of Washington, Simon Fraser University, and the Advanced Photon Source. Use of the Advanced Photon Source, an Office of Science User Facility operated for the U.S. DOE Office of Science by Argonne National Laboratory, was supported by the U.S. DOE under Contract No. DE-AC02-06CH11357. ICP-OES analyses were performed at the Nano Research Facility (NRF), a member of the National Nanotechnology Infrastructure Network (NNIN), which is supported by the National Science Foundation under Grant No. ECS-0335765. SEM imaging was conducted at the Center for Materials Innovation at Washington University in St. Louis. Dale Brewe and Steve Heald are thanked for beamline support and Kate Nelson is thanked for assistance operating the ICP-OES. ’ REFERENCES (1) Kappler, A.; Straub, K. L. Geomicrobiological cycling of iron. Rev. Mineral. Geochem. 2005, 59, 85–108. (2) Weber, K. A.; Achenbach, L. A.; Coates, J. D. Microorganisms pumping iron: Anaerobic microbial iron oxidation and reduction. Nat. Rev. Microbiol. 2006, 4, 752–764. (3) Hansel, C. M.; Benner, S. G.; Fendorf, S. Competing Fe(II)induced mineralization pathways of ferrihydrite. Environ. Sci. Technol. 2005, 39, 7147–7153. (4) Hansel, C. M.; Benner, S. G.; Neiss, J.; Dohnalkova, A.; Kukkadapu, R. K.; Fendorf, S. Secondary mineralization pathways induced by dissimilatory iron reduction of ferrihydrite under advective flow. Geochim. Cosmochim. Acta 2003, 67, 2977–2992. (5) Hansel, C. M.; Benner, S. G.; Nico, P.; Fendorf, S. Structural constraints of ferric (hydr)oxides on dissimilatory iron reduction and the fate of Fe(II). Geochim. Cosmochim. Acta 2004, 68, 3217–3229. (6) Crosby, H. A.; Johnson, C. M.; Roden, E. E.; Beard, B. L. Coupled Fe(II)-Fe(III) electron and atom exchange as a mechanism 8831
dx.doi.org/10.1021/es202445w |Environ. Sci. Technol. 2011, 45, 8826–8833
Environmental Science & Technology for Fe isotope fractionation during dissimilatory iron oxide reduction. Environ. Sci. Technol. 2005, 39, 6698–6704. (7) Crosby, H. A.; Roden, E. E.; Johnson, C. M.; Beard, B. L. The mechanisms of iron isotope fractionation produced during dissimilatory Fe(III) reduction by Shewanella putrefaciens and Geobacter sulfurreducens. Geobiology 2007, 5, 169–189. (8) Wu, L. L.; Beard, B. L.; Roden, E. E.; Kennedy, C. B.; Johnson, C. M. Stable Fe isotope fractionations produced by aqueous Fe(II)hematite surface interactions. Geochim. Cosmochim. Acta 2010, 74, 4249–4265. (9) Catalano, J. G.; Fenter, P.; Park, C.; Zhang, Z.; Rosso, K. M. Structure and oxidation state of hematite surfaces reacted with aqueous Fe(II) at acidic and neutral pH. Geochim. Cosmochim. Acta 2010, 74, 1498–1512. (10) Tanwar, K. S.; Petitto, S. C.; Ghose, S. K.; Eng, P. J.; Trainor, T. P. Structural study of Fe(II) adsorption on hematite(1 102). Geochim. Cosmochim. Acta 2008, 72, 3311–3325. (11) Tanwar, K. S.; Petitto, S. C.; Ghose, S. K.; Eng, P. J.; Trainor, T. P. Fe(II) adsorption on hematite (0 0 0 1). Geochim. Cosmochim. Acta 2009, 73, 4346–4365. (12) Larese-Casanova, P.; Scherer, M. M. Fe(II) sorption on hematite: New insights based on spectroscopic measurements. Environ. Sci. Technol. 2007, 41, 471–477. (13) Larese-Casanova, P.; Scherer, M. M. Morin transition suppression in polycrystalline 57Hematite (α-Fe2O3) exposed to 56Fe(II). Hyperfine Interact. 2007, 174, 111–119. (14) Williams, A. G. B.; Scherer, M. M. Spectroscopic evidence for Fe(II)-Fe(III) electron transfer at the iron oxide water interface. Environ. Sci. Technol. 2004, 38, 4782–4790. (15) Pedersen, H. D.; Postma, D.; Jakobsen, R.; Larsen, O. Fast transformation of iron oxyhydroxides by the catalytic action of aqueous Fe(II). Geochim. Cosmochim. Acta 2005, 69, 3967–3977. (16) Handler, R. M.; Beard, B. L.; Johnson, C. M.; Scherer, M. M. Atom exchange between aqueous Fe(II) and goethite: An Fe isotope tracer study. Environ. Sci. Technol. 2009, 43, 1102–1107. (17) Rosso, K. M.; Yanina, S. V.; Gorski, C. A.; Larese-Casanova, P.; Scherer, M. M. Connecting observations of hematite (α-Fe2O3) growth catalyzed by Fe(II). Environ. Sci. Technol. 2010, 44, 61–67. (18) Yanina, S. V.; Rosso, K. M. Linked reactivity at mineral water interfaces through bulk crystal conduction. Science 2008, 320, 218–222. (19) Buerge, I. J.; Hug, S., J. Influence of mineral surfaces on chromium(VI) reduction by iron(II). Environ. Sci. Technol. 1999, 33, 4285–4291. (20) Charlet, L.; Liger, E.; Gerasimo, P. Decontamination of TCEand U-rich waters by granular iron: Role of sorbed Fe(II). J. Environ. Eng. 1998, 124, 25–30. (21) Charlet, L.; Silvester, E.; Liger, E. N-compound reduction and actinide immobilisation in surficial fluids by Fe(II): the surface FeIIIOFeIIOH° species, as major reductant. Chem. Geol. 1998, 151, 85–93. (22) Liger, E.; Charlet, L.; Van Cappellen, P. Surface catalysis of uranium(VI) reduction by iron(II). Geochim. Cosmochim. Acta 1999, 63, 2939–2955. (23) Plymale, A. E.; Fredrickson, J. K.; Zachara, J. M.; Dohnalkova, A. C.; Heald, S. M.; Moore, D. A.; Kennedy, D. W.; Marshall, M. J.; Wang, C. M.; Resch, C. T.; Nachimuthu, P. Competitive reduction of pertechnetate (99TcO4 ) by dissimilatory metal reducing bacteria and biogenic Fe(II). Environ. Sci. Technol. 2011, 45, 951–957. (24) Zachara, J. M.; Heald, S. M.; Jeon, B. H.; Kukkadapu, R. K.; Liu, C. X.; McKinley, J. P.; Dohnalkova, A. C.; Moore, D. A. Reduction of pertechnetate [Tc(VII)] by aqueous Fe(II) and the nature of solid phase redox products. Geochim. Cosmochim. Acta 2007, 71, 2137–2157. (25) Amstaetter, K.; Borch, T.; Larese-Casanova, P.; Kappler, A. Redox transformation of arsenic by Fe(II)-activated goethite (α-FeOOH). Environ. Sci. Technol. 2010, 44, 102–108. (26) Dixit, S.; Hering, J. G. Sorption of Fe(II) and As(III) on goethite in single- and dual-sorbate systems. Chem. Geol. 2006, 228, 6–15.
ARTICLE
(27) Ona-Nguema, G.; Morin, G.; Wang, Y.; Foster, A. L.; Juillot, F.; Calas, G.; Brown, G. E., Jr. XANES evidence for rapid arsenic(III) oxidation at magnetite and ferrihydrite surfaces by dissolved O2 via Fe2+mediated reactions. Environ. Sci. Technol. 2010, 44, 5416–5422. (28) Hug, S. J.; Leupin, O. Iron-catalyzed oxidation of arsenic(III) by oxygen and by hydrogen peroxide: pH-dependent formation of oxidants in the Fenton reaction. Environ. Sci. Technol. 2003, 37, 2734– 2742. (29) Coughlin, B. R.; Stone, A. T. Nonreversible adsorption of divalent metal-ions (MnII, CoII, NiII, CuII, and PbII) onto goethite: Effects of acidification, FeII addition, and picolinic acid addition. Environ. Sci. Technol. 1995, 29, 2445–2455. (30) Jeon, B. H.; Dempsey, B. A.; Burgos, W. D.; Royer, R. A. Sorption kinetics of Fe(II), Zn(II), Co(II), Ni(II), Cd(II), and Fe(II)/ Me(II) onto hematite. Water Res. 2003, 37, 4135–4142. (31) Frierdich, A. J.; Luo, Y.; Catalano, J. G., Trace element cycling through iron oxide minerals during redox-driven dynamic recrystallization. Geology 2011, doi: 10.1130/G32330.1. (32) Cornell, R. M.; Schwertmann, U., The Iron Oxides: Structure, Properties, Reactions, Occurrences, and Uses, 2nd ed.; Wiley-VCH: Weinheim, 2003; p 664. (33) Manceau, A.; Schlegel, M. L.; Musso, M.; Sole, V. A.; Gauthier, C.; Petit, P. E.; Trolard, F. Crystal chemistry of trace elements in natural and synthetic goethite. Geochim. Cosmochim. Acta 2000, 64, 3643–3661. (34) Singh, B.; Gilkes, R. J. Properties and distribution of iron oxides and their association with minor elements in the soils of south-western Australia. J. Soil Sci. 1992, 43, 77–98. (35) Singh, B.; Sherman, D. M.; Gilkes, R. J.; Wells, M.; Mosselmans, J. F. W. Structural chemistry of Fe, Mn, and Ni in synthetic hematites as determined by extended X-ray absorption fine structure spectroscopy. Clays Clay Miner. 2000, 48, 521–527. (36) Kocar, B. D.; Fendorf, S. Thermodynamic constraints on reductive reactions influencing the biogeochemistry of arsenic in soils and sediments. Environ. Sci. Technol. 2009, 43, 4871–4877. (37) Schwertmann, U.; Cornel, R. M., Iron oxides in the laboratory: Preparation and characterization, 2nd ed.; Wiley-VCH: Weinheim, 2000; p 188. (38) Stookey, L. L. Ferrozine—A new spectrophotometric reagent for iron. Anal. Chem. 1970, 42, 779–781. (39) Bethke, C. M. The Geochemist’s Workbench Release 8.0; Hydrogeology Program, University of Illinois: Urbana, IL, 2009. (40) Delany, J. M.; Lundeen, S. R. The LLNL Thermochemical Database; Lawrence Livermore National Laboratory: Livermore, CA, 1990. (41) Helgeson, H. C. Thermodynamics of hydrothermal systems at elevated temperatures and pressures. Am. J. Sci. 1969, 267, 729–804. (42) Helgeson, H. C.; Kirkham, D. H. Theoretical prediction of the thermodynamic behavior of aqueous electrolytes at high pressures and temperatures: I. Summary of the thermodynamic/electrostatic properties of the solvent. Am. J. Sci. 1974, 274, 1089–1198. (43) Helgeson, H. C.; Kirkham, D. H. Theoretical prediction of the thermodynamic behavior of aqueous electrolytes at high pressures and temperatures: II. Debye-H€uckel parameters for activity coefficients and relative partial molal properties. Am. J. Sci. 1974, 274, 1199–1261. (44) Langmuir, D.; Mahoney, J.; Rowson, J. Solubility products of amorphous ferric arsenate and crystalline scorodite (FeAsO4 3 2H2O) and their application to arsenic behavior in buried mine tailings. Geochim. Cosmochim. Acta 2006, 70, 2942–2956. (45) Johnston, R. B.; Singer, P. C. Solubility of symplesite (ferrous arsenate): Implications for reduced groundwaters and other geochemical environments. Soil Sci. Soc. Am. J. 2007, 71, 101–107. (46) Ravel, B.; Newville, M. ATHENA, ARTEMIS, HEPHAESTUS: Data analysis for X-ray absorption spectroscopy using IFEFFIT. J. Synchrotron Rad. 2005, 12, 537–541. (47) Webb, S. M. SIXPack: A graphical user interface for XAS analysis using IFEFFIT. Phys. Scr. 2005, T115, 1011–1014. (48) Newville, M. IFEFFIT: Interactive EXAFS analysis and FEFF fitting. J. Synchrotron Rad. 2001, 8, 322–324. 8832
dx.doi.org/10.1021/es202445w |Environ. Sci. Technol. 2011, 45, 8826–8833
Environmental Science & Technology
ARTICLE
(49) Ankudinov, A. L.; Rehr, J. J. Relativistic calculations of spindependent x-ray-absorption spectra. Phys. Rev. B 1997, 56, R1712– R1715. (50) Xu, Y.; Zhou, G.-P.; Zheng, X.-F. Redetermination of iron(III) arsenate dihydrate. Acta Crystallogr., Sect. E: Struct. Rep. Online 2007, 63, i67–i69. (51) Mikutta, C.; Frommer, J.; Voegelin, A.; Kaegi, R.; Kretzschmar, R. Effect of citrate on the local Fe coordination in ferrihydrite, arsenate binding, and ternary arsenate complex formation. Geochim. Cosmochim. Acta 2010, 74, 5574–5592. (52) Arai, Y.; Sparks, D. L.; Davis, J. A. Effects of dissolved carbonate on arsenate adsorption and surface speciation at the hematite-water interface. Environ. Sci. Technol. 2004, 38, 817–824. (53) Sherman, D. M.; Randall, S. R. Surface complexation of arsenic(V) to iron(III) (hydr)oxides: Structural mechanism from ab initio molecular geometries and EXAFS spectroscopy. Geochim. Cosmochim. Acta 2003, 67, 4223–4230. (54) O’Reilly, S. E.; Strawn, D. G.; Sparks, D. L. Residence time effects on arsenate adsorption/desorption mechanisms on goethite. Soil Sci. Soc. Am. J. 2001, 65, 67–77. (55) Catalano, J. G.; Park, C.; Fenter, P.; Zhang, Z. Simultaneous inner- and outer-sphere arsenate complexation on corundum and hematite. Geochim. Cosmochim. Acta 2008, 72, 1986–2004. (56) Catalano, J. G.; Zhang, Z.; Park, C.; Fenter, P.; Bedzyk, M. J. Bridging arsenate surface complexes on the hematite (012) surface. Geochim. Cosmochim. Acta 2007, 71, 1883–1897. (57) Zeng, H.; Singh, A.; Basak, S.; Ulrich, K.-U.; Sahu, M.; Biswas, P.; Catalano, J. G.; Giammar, D. E. Nanoscale size effects on uranium(VI) adsorption to hematite. Environ. Sci. Technol. 2009, 43, 1373–1378. (58) Mori, H.; Ito, T. The structure of vivianite and symplesite. Acta Crystallogr. 1950, 3, 1–6. (59) J€onsson, J.; Sherman, D. M. Sorption of As(III) and As(V) to siderite, green rust (fougerite) and magnetite: Implications for arsenic release in anoxic groundwaters. Chem. Geol. 2008, 255, 173–181. (60) Jeon, B. H.; Dempsey, B. A.; Burgos, W. D. Kinetics and mechanisms for reactions of Fe(II) with iron(III) oxides. Environ. Sci. Technol. 2003, 37, 3309–3315. (61) Jeon, B. H.; Dempsey, B. A.; Burgos, W. D.; Royer, R. A. Reactions of ferrous iron with hematite. Colloids Surf., A 2001, 191, 41–55. (62) Tufano, K. J.; Reyes, C.; Saltikov, C. W.; Fendorf, S. Reductive processes controlling arsenic retention: Revealing the relative importance of iron and arsenic reduction. Environ. Sci. Technol. 2008, 42, 8283–8289. (63) Masue-Slowey, Y.; Loeppert, R. H.; Fendorf, S. Alteration of ferrihydrite reductive dissolution and transformation by adsorbed As and structural Al: Implications for As retention. Geochim. Cosmochim. Acta 2011, 75, 870–886. (64) Borch, T.; Kretzschmar, R.; Kappler, A.; Van Cappellen, P.; Ginder-Vogel, M.; Voegelin, A.; Campbell, K. Biogeochemical redox processes and their impact on contaminant dynamics. Environ. Sci. Technol. 2010, 44, 15–23. (65) Kukkadapu, R. K.; Zachara, J. M.; Fredrickson, J. K.; Kennedy, D. W. Biotransformation of two-line silica-ferrihydrite by a dissimilatory Fe(III)-reducing bacterium: formation of carbonate green rust in the presence of phosphate. Geochim. Cosmochim. Acta 2004, 68, 2799–2814. (66) Erbs, J. J.; Berquo, T. S.; Reinsch, B. C.; Lowry, G. V.; Banerjee, S. K.; Penn, R. L. Reductive dissolution of arsenic-bearing ferrihydrite. Geochim. Cosmochim. Acta 2010, 74, 3382–3395. (67) Pedersen, H. D.; Postma, D.; Jakobsen, R. Release of arsenic associated with the reduction and transformation of iron oxides. Geochim. Cosmochim. Acta 2006, 70, 4116–4129. (68) Galvez, N.; Barron, V.; Torrent, J. Preparation and properties of hematite with structural phosphorus. Clays Clay Min. 1999, 47, 375–385.
8833
dx.doi.org/10.1021/es202445w |Environ. Sci. Technol. 2011, 45, 8826–8833
ARTICLE pubs.acs.org/est
On the Reversibility of Environmental Contamination with Persistent Organic Pollutants Sung-Deuk Choi† and Frank Wania* Department of Physical and Environmental Sciences, University of Toronto Scarborough, 1265 Military Trail, Toronto, Ontario, Canada, M1C 1A4
bS Supporting Information ABSTRACT: An understanding of the factors that control the time trends of persistent organic pollutants (POPs) in the environment is required to evaluate the effectiveness of emission reductions and to predict future exposure. Using a regional contaminant fate model, CoZMo-POP 2, and a generic bell-shaped emission profile, we simulated time trends of hypothetical chemicals with a range of POP-like partitioning and degradation properties in different compartments of a generic warm temperate environment, with the objective of identifying the processes that may prevent the reversibility of environmental contamination with POPs after the end of primary emissions. Evaporation from soil and water can prevent complete reversibility of POP contamination of the atmosphere after the end of emissions. However, under the selected conditions, only for organic chemicals within a narrow range of volatility, that is, a logarithm of the octanol air equilibrium partition coefficient between 7 and 8, and with atmospheric degradation half-lives in excess of a few month can evaporation from environmental reservoirs sustain atmospheric levels that are within an order of magnitude of those resulting from primary emissions. HCB and αHCH fulfill these criteria, which may explain, why their atmospheric concentrations have remained relatively high decades after their main primary emissions have been largely eliminated. Soil-to-water transfer is found responsible for the lack of reversibility of POP contamination of the aqueous environment after the end of emissions, whereas reversal of water-sediment exchange, although possible, is unlikely to contribute significantly. Differences in the reversibility of contamination in air and water suggests the possibility of changes in the relative importance of various exposure pathways after the end of primary emissions, namely an increase in the importance of the aquatic food chain relative to the agricultural one, especially if the former has a benthic component. Since simulated time trends were strongly dependent on degradation half-lives, partitioning properties and selected environmental input parameters, it should not be surprising, that different field studies often generate highly divergent time trends.
’ INTRODUCTION A key question in the regulation of persistent organic pollutants (POPs) is: If we reduce or completely eliminate emissions, how long will it take before environmental concentrations drop below certain thresholds considered acceptable? Or, in other words, what are the time trends of POP contamination in the environment? A great deal of effort goes into addressing this question through environmental monitoring. Time trends of POPs have been investigated by collecting the same kind of samples over a long period of time. Some monitoring programs now have continuous records of atmospheric POP contamination that exceed a decade. Whereas biota samples such as fish, bird eggs, and mammals have been widely used for time trend monitoring of POPs in freshwater and marine environments, archived soil and herbage samples have served in the reconstruction of POP time trends in the terrestrial environment. The only means of obtaining time trends for period during which no samples had been taken in the past is to rely on natural archives of pollution, such as those in lake and ocean sediments, peat cores, glaciers, ice-caps, and tree bark. Some key studies on the r 2011 American Chemical Society
experimental determination of time trends of POPs in the environment are listed in Table S1 in the Supporting Information (SI). Considering the funds and efforts expended on long-term monitoring programs, rather little effort has been directed toward a theoretical understanding of the time trends of POPs. Yet such an understanding is clearly paramount, if only to answer some of the questions raised by the results of monitoring programs: Why is there often a bewildering variability in time trends between different constituents of a mixture, between different regions, or between different sample types? Is the absence of a declining trend due to ongoing emissions or due to secondary sources, namely the mobilization of chemicals stored in environmental reservoirs? A typical emission time trend for POPs involves a steady increase followed by a steady decline, generally on the time scale Received: May 23, 2011 Accepted: September 9, 2011 Revised: September 6, 2011 Published: September 09, 2011 8834
dx.doi.org/10.1021/es2017544 | Environ. Sci. Technol. 2011, 45, 8834–8841
Environmental Science & Technology of a few decades. This trend reflects the increasing use of a new commercial substance and the realization of its environmental impact, which eventually leads to decreasing or discontinued use. Because of stockpiling and/or the slow release from products in use or after disposal, the decrease in emissions is often much slower than the decrease in production.1,2 The estimated global emissions of many POPs show this general pattern of increase and decrease (Figure S1 and references in the SI), which may be represented generically using a bell-shaped function covering 50 years of emissions. In principle, we can distinguish four generic types of time trends in environmental concentrations in response to such emission profiles: (i) Concentrations rise and drop in concert with the emissions, possibly with a small lag time (fast reversibility), (ii) concentrations rise and drop with emissions, but the maximum in concentrations may be considerably delayed relative to the maximum in emissions, and/or the rate of concentration decline is much slower than the rate of decline in emissions (slow reversibility), (iii) concentrations rise, but do not drop in response to emission reductions (no reversibility), (iv) concentrations rise and initially drop in response to emission reductions, but then stabilize at a constant level or decline at much slower rates (partial reversibility). Mechanistically, the first three types indicate an environmental reservoir that is filled during times of emissions, which is depleted quickly, slowly, or not at all after emissions cease. Time trends of type (iv) on the other hand require a steady supply of contamination to a compartment after primary emissions have ceased. These time trend types are not precisely defined and the boundaries between those categories, especially the first three, are fluid. The latter three time trends can indicate a lack of contamination reversibility, that is, situations where levels in the environment remain high long after emissions have ceased. What constitutes high and long in this context is of course arbitrary, and what degree of lack of reversibility warrants concern depends obviously on the amount of a chemical emitted, its toxicity, and in what part of the environment it resides. It is currently poorly understood how chemical and environmental factors interact in controlling which of these scenarios will materialize for a particular substance in a particular environmental medium and region, even though it has overarching importance for predicting, interpreting, and evaluating the effectiveness of national and international POP control measures. For example if observation indicate partial reversibility for a chemical in a compartment, even though theoretically it should show complete reversibility, this is an indication that there continue to be primary emissions, which in principle could be subject to remedial action. If however, a chemical can be subject to partial reversibility in the absence of primary emissions, further remedial action on primary sources is unlikely to result in faster contamination reversibility. In this study, we set out to use a dynamic multimedia fate model to gain a quantitative understanding of how chemical and environmental characteristics may interact to control long-term trends of POPs in various environmental compartments. Whereas emission time trends and the rates of chemical degradation in various media evidently have an important role in determining concentration time trends, the influence of a chemical’s partitioning properties and environmental characteristics on time trends is much less obvious. Other studies have employed dynamic environmental chemical fate models to interpret time trends of POPs. Most seek to
ARTICLE
reproduce measured time trends for specific substances using realistic emission estimates (e.g., ref 3) and then predict likely future time trends in response to variable control strategies. Hung et al.4 used simulations with a global scale dynamic model to conclude that the rate of decline of PCB air concentrations during the 1990s was mostly driven by declines in primary emission, and predicted that evaporation from soil would control the concentrations of PCBs in the atmosphere after primary emission ceased completely. Wania5 used the same model to show that atmospherically deposited perfluorocarboxylic acids would experience similar time trends in the Arctic marine environment as the same compounds directly emitted to the ocean. Also relying on a global scale model, Schenker et al.6 reconstructed measured time trends for DDT and its degradation products, and predicted rates of recovery for different climate zones using future emission scenarios. Stroebe et al.7 showed that the removal rate long after the end of emissions, at the so-called temporal remote state, is independent of the release pattern but depends on the environmental partitioning that a substance will eventually adopt. Gouin and Wania,8 rather than simulating the time trends of real chemicals, used global model calculations to identify those combinations of chemical properties that result in a delay between the start of a global emission decrease and the begin of a decline in Arctic ecosystem contamination. This study is not concerned with reproducing past or predicting future time trends for specific POPs, but with identifying and characterizing the impact of various factors on the time trends of POP-like chemicals in the environment. The models’ purpose is that of formulating and testing the reasonability of hypotheses, which can hopefully be of use in the interpretation of observed time trends and in informing regulatory decisions. The approach taken in this study: (i) identifies four combinations of partitioning properties that represent the range of distribution characteristics of POPs, (ii) explores how different combinations of degradation rates in atmosphere and surface media influence the concentration time trends of chemicals with such partitioning properties in a variety of media, and (iii) investigates which environmental model input parameters, that is, which environmental processes, impact on these time trends. The study is restricted to investigating the time trends in nonbiotic media, relying on a physical contaminant fate model. Time trends in biological samples are determined by those in the physical environment, but are subject to a whole range of additional influences.9
’ MATERIALS AND METHODS Outline of the Overall Approach. Time trends of the concentrations of hypothetical chemicals in air, water, soil, sediment, and tree foliage were calculated using CoZMoPOP 2, a dynamic multimedia fate and transport model that simulates the fate of neutral organic chemicals in a coastal region or the drainage basin of a large lake.10 The default settings describe a generic warm temperate watershed (8 104 km2, 5% water coverage, 50% forest cover) represented by nine compartments (SI Figure S2). Monthly average air temperatures in the environment range from 0.2 to 26.3 °C, and the annual precipitation rate is 700 mm. All default parameters are provided in the SI (Tables S2S8). Whereas loss by atmospheric advection was not considered, implying that the model domain is surrounded by areas similar to the watershed described within the model, water is advected from the watershed’s fresh water compartment. 8835
dx.doi.org/10.1021/es2017544 |Environ. Sci. Technol. 2011, 45, 8834–8841
Environmental Science & Technology
ARTICLE
Figure 1. Time trends of the concentrations in different environmental media (normalized to the average over the 100 year simulation period) of 64 hypothetical chemicals with the partitioning properties of HCB (AirPOP), α-HCH (WaterPOP), DDE (CenterPOP), and BDE-209 (ParticlePOP) and different combinations of degradation half-lives in air and surface compartments. Panels applying to real POPs are hatched in orange.
Details on model structure and the parametrization of contaminant fate processes can be found elsewhere.10 A generic bellshaped emission scenario with a 25-year increase followed by a 25-year decrease was used for all hypothetical chemicals. The calculations were extended 50 years beyond the end of emissions for a total simulation time of 100 years. This reflects the time period of the human experience of POP emissions and bans and thus the time period for which it is currently possible to obtain measured time trends. From these calculations, a few chemical property combinations were identified, for which concentrations in the exposure-relevant media air and water are not completely reversible. A pathway analysis for these property combinations revealed the processes responsible for a lack of reversibility, and a
sensitivity analysis identified the environmental parameter that have the most influence on the extent of reversibility. Properties of Selected Chemicals. The CoZMoPOP 2 model requires equilibrium partitioning coefficients between air, water and octanol (KOW, KAW, KOA) and internal energies of phase transfer that quantify the temperature dependence of those partitioning coefficients (ΔUOW, ΔUAW, ΔUOA). The 64 hypothetical chemicals are permutations of four different combinations of partitioning properties and 16 different combinations of degradation half-lives in air (HLAir = 10 days, 100 days, 3 years, 30 years) and surface media (HLSurf = 0.1 years, 1 year, 10 years, 100 years). We selected and named four partitioning property combinations that reflect the full diversity of POPs 8836
dx.doi.org/10.1021/es2017544 |Environ. Sci. Technol. 2011, 45, 8834–8841
Environmental Science & Technology (SI Figure S3, Table S9): “CenterPOP” designating chemicals of intermediate volatility with the partitioning properties of dichlorodiphenyldichloroethylene (DDE), “AirPOP” standing for relatively volatile chemicals with the distribution properties of hexachlorobenzene (HCB), “WaterPOP” representing relatively volatile and water-soluble chemicals with the properties of αhexachlorocyclohexane (α-HCH), and “ParticlePOP” as a surrogate for extremely involatile chemicals with the distribution properties of decabrominated diphenyl ether (BDE-209). It should be stressed that the property combinations represent hypothetical placeholders with POP-like properties rather than the real chemicals with those partitioning properties. In order to identify volatility thresholds for atmospheric contamination reversibility, we further calculated the ratio between the air concentration at the end of the simulation and the maximum air concentration during the simulation for the chemical partitioning space defined by the partitioning properties log KOA and log KAW. All possible hypothetical combinations of log KOA between 5 and 10, and log KAW between 5 and 0 were considered. The hypothetical chemicals were assumed to have internal energies of phase transfer of ΔUAW = 60, ΔUOA = 80, and ΔUOW = 20 kJ/mol. Six different HLAir (10 days, 100 days, 1 year, 3 years, 30 years, and 100 years) and a single HLSurf (100 years) were used. All the other environmental and emission parameters were the same as in the other simulations.
’ RESULTS AND DISCUSSION Influence of Degradation Half-lives and Partitioning Properties on Time Trends. The time trends over 100 years
of the 64 hypothetical chemicals AirPOP, WaterPOP, CenterPOP, and ParticlePOP, with the partitioning properties of HCB, α-HCH, DDE, and BDE-209, and different combinations of HLAir and HLSurf are shown in Figure 1. The time trends in different media are normalized and no scale is provided, as the absolute concentration values are arbitrary and the analysis is limited to the shape of the time trend curves. HL combinations that could conceivably apply to real POPs are identified by panels with orange hatching: Degradation half-lives in surface media HLSurf for many POPs are believed to be in the range of decades, if not centuries, although somewhat shorter for BDE-209. Halflives in air HLAir are in the range of 10100 days for most POPs, but in the range of a few years for HCB.11 As the HLSurf increases (from left to right in these figures) and exceeds one year, differences in the time trends in different compartments become apparent. Whereas all of the chemicals in the left two columns show “fast reversibility”, those in the right two columns display all four of the time trend types described in the Introduction. The following can be deduced from Figure 1: (a) Concentration time trends in different media can be decoupled, that is, they can be different in different media if the chemical is sufficiently persistent. In particular, even though we applied the same HL in all surface media, different surface media may display different time trends. (b) Environmental contamination with chemicals with HLSurf of a year or less is completely and rapidly reversible. (c) Contamination of the atmosphere (and by extension the forest canopy) is almost always “fast reversible”, that is, the time trend in air concentrations matches closely the time trend in emissions. The exception are relatively volatile chemicals (WaterPOP and AirPOP) with high persistence in air (HLAir g 100 days); when persistent in
ARTICLE
surface media their time trends indicate slow (HLSurf g 10 years), or partial reversibility (HLSurf = 100 years). This phenomenon is discussed further below. (d) “Slow reversibility” is prevalent for POP concentrations in soils and sediment, if HLSurf is on the order of decades. If the HLSurf are on the order of centuries, the time trends in those media approach the “no reversibility” category. (e) The impact that the value of HLAir has on the time trends increases with increasing HLSurf and increasing volatility of a chemical. For example, virtually no effect of HLAir on the time trends is observed if HLSurf is one year or less, and the ParticlePOPs trends are even the same for different HLAir. The latter is the result of the model’s assumption of negligible degradation in the atmospheric particle phase. (f) For POPs that are very persistent in surface media (HLSurf = 100 years) and are relatively involatile (CenterPOPs and ParticlePOPs), partial reversibility time trends appear in the water compartment, which do not appear in either atmosphere/canopy or soil/sediment. (g) With increasing time after the end of emissions, the trends in the various media appear to converge to a similar rate of decline, which is partly due to assuming the same HLSurf in all surface compartments. It is also in agreement with the concept of the temporal remote state, in which the chemical has the same rate of decline in all media.7 Only the time trends for ParticlePOPs with very high persistence in surface media still diverge, indicating that they are still far from the temporal remote state. If we focus on the panels representing chemicals with long HLSurf (100 years) and relatively short HLAir (10 days) (blue panels in Figure 1), pronounced differences in the time trends for different partitioning coefficient combinations become apparent. The number of time trend types increases with decreasing volatility of the chemical (AirPOP ≈ WaterPOP < CenterPOP < ParticlePOP). AirPOP has two time trend patterns: one for air/ canopies/water and the other for soils/sediments. WaterPOP has a similar trend pattern, although the trend in freshwater is slightly different from those in air/canopies. If we increase HLAir for WaterPOP and AirPOP (moving down from the blue panels), the number of time trend types remains the same but reversibility in air is now partial rather than fast. In the blue panel for CenterPOP, a distinct third trend pattern for freshwater that falls between the patterns for air and soil appears. Lastly, five distinct patterns are discernible in the blue panel for ParticlePOP, that is, differences in the time trends in soils and sediments only become apparent in the most persistent and least volatile chemicals. ParticlePOP shows a slightly increasing trend in sediments and freshwater until the end of the simulation. The time trend of concentrations in both forest canopies is very similar to the air concentration trend (Figure 1), supporting the usefulness of archived vegetation samples for inferring the time trend of POPs in the atmosphere.12 Whereas the time trend in the deciduous canopy is almost exactly the same as in air, that in coniferous canopy shows a small lag that increases with decreasing compound volatility (from no lag for AirPOP to a lag of 5 years for ParticlePOP). Coniferous foliage has a lifetime of five years in the model and, as a result, has a short-term memory of past atmospheric contamination of the least volatile POPs, whose uptake in needles is kinetically limited over the lifetime of the needles. Monitoring of time trends using needles thus has to rely on needles of the same age. 8837
dx.doi.org/10.1021/es2017544 |Environ. Sci. Technol. 2011, 45, 8834–8841
Environmental Science & Technology
Figure 2. Net-fluxes (input-output) to and from air of hypothetical persistent chemicals with the partitioning properties of HCB (AirPOP) and α-HCH (WaterPOP). Data were normalized to the largest net flux observed during the simulation period. Positive and negative values represent net evaporation of chemicals from the surface compartments and net deposition to the surface compartments, respectively.
All POPs in the panels on the right of Figure 1 (indicating high HLSurf) show slow or no reversibility in soils and sediments, which is intuitive because these compartments are major POP storage reservoirs and it takes time to deplete large reservoirs of highly persistent chemicals. On the other hand, the lack of reversibility in air for AirPOP and WaterPOP and in water for all persistent chemical partitioning property combinations in the panels on the right side of Figure 1 is much less intuitive because air and water generally contain very small fractions of the environmental inventory of a POP. We can infer that there must be processes operating that continue to supply chemical to air and water after emissions ceased. An understanding of these processes is paramount if we are to understand the reversibility of organism exposure to POPs because such exposure is ultimately controlled by concentrations in air and water. In the following sections, we therefore identify those processes and discuss their dependence on chemical and environmental properties. What Prevents Fast Reversibility of Air Concentrations? In order to identify the process that supplies persistent substances with the partitioning properties of HCB and α-HCH to the air compartment after emissions have ceased, we calculated net chemical fluxes of persistent AirPOP and WaterPOP to and from the air compartment as a function of time (Figure 2) as well as normalized cumulative mass balances for the atmosphere for the periods of emissions (19502000) and the time period a few years after emissions have ceased (20052050) (SI Figure S5). A positive flux in Figure 2 refers to net evaporation from the surface compartments to the atmosphere. During the entire simulation periods, the forest is a net sink of both persistent AirPOP and WaterPOP from the atmosphere. This is also true for soils and fresh water during the initial period of increasing emissions (Figure 2). Already during the period of declining emissions soil and freshwater begin to act as a net source of AirPOP and WaterPOP to the atmosphere. For AirPOP soil is the major secondary source and the transition from sink to source occurs earlier for soil (approximately 6 years after peak emissions) than for water (approximately 16 years after peak emissions). For WaterPOP water is a more important secondary source compartment and it becomes a net source to the atmosphere about 10 years before soil does. Net-fluxes of AirPOP and WaterPOP with a shorter HLAir of 10 days were also calculated (SI Figure S4). The temporal pattern of net-fluxes from soil and water is similar to those of persistent chemicals with a longer HLAir, but degradation now becomes the dominant loss process for both
ARTICLE
chemicals from air. The air concentrations after the end of emissions are now only a very small fraction of the concentrations prevailing during the time period of primary emissions, that is, atmospheric contamination with such less persistent chemicals is reversible (SI Figure S4). We conclude that soil and water can continue to supply relatively volatile and atmospherically persistent POPs to air even decades after primary emissions have ceased and thus prevent complete reversibility of atmospheric contamination. What surface compartment is more important depends on partitioning properties; whereas soil is the dominant source of AirPOP to air, water is a more important source of WaterPOP. Indeed, there is field evidence that volatilization of relatively volatile POPs can notably affect levels in air. Interannual differences in air concentrations of HCB, HCHs, and lighter PCBs around the Great Lakes could be attributed to variables rates of volatilization from surfaces caused by climate fluctuations.13 Evaporation from cold seas and lakes has been invoked to explain relatively high α-HCH air concentrations in the Arctic,14 in Atlantic Canada and around Lake Superior,15 and in Tibet.16 To evaluate the influence of environmental parameters on the atmospheric concentration time trend of persistent and relatively volatile POPs in air and soil, three soil parameters (soil depth, organic carbon fraction, mass transfer coefficient through air boundary over soil layer) were changed one at a time, by either multiplying or dividing the default value by 10 (SI Figure S6). The value of these soil parameters does not affect the calculated time trend of AirPOP in soil, but those in air are sensitive to the change of soil parameters (SI Figure S6). In particular, levels in air after the end of emission are higher if the soil is shallow and has a low organic content and if gas exchange with the soil is rapid (e.g., under windy conditions) (SI Figure S6). In all cases, the evaporation of AirPOP from soil prevents air contamination reversibility, except when an unreasonably high soil depth shuts down the evaporation flux completely. Generally, the sensitivity analysis for WaterPOP shows similar results, except that the contamination of a shallow or low organic matter soil with WaterPOP is reversible, because of the relatively low storage capacity of such soils for WaterPOP. The time trend in the air above such soils tends to show the same reversibility (SI Figure S6). Clearly, the storage capacity of soil, which is mostly a function of soil depth and organic carbon fraction, can influence the time trends of relatively volatile POPs in the atmosphere. Partitioning Property Thresholds for Reversibility in Atmospheric Contamination with POPs. HCB and α-HCH like chemicals (AirPOP, WaterPOP) can exhibit partial contamination reversibility in air, but chemicals with the partitioning properties of DDE and BDE-209 (CenterPOP, ParticlePOP) do not (Figure 1). In order to establish the chemical property thresholds, beyond which atmospheric contamination, and therefore terrestrial food chain contamination, is reversible, we quantified the lack of contamination reversibility as follows: R ¼ Clast =Cmax
ð1Þ
where Clast is the air concentration at the end of simulation (i.e., the year 2050 in this study), and Cmax is the maximum air concentration during the entire period of simulation, which generally tends to occur during or shortly after the time of maximum emissions. With this definition, it is possible to identify chemicals within the chemical partitioning space that will exhibit reversibility of 8838
dx.doi.org/10.1021/es2017544 |Environ. Sci. Technol. 2011, 45, 8834–8841
Environmental Science & Technology
ARTICLE
Figure 3. Chemical space map of the lack of reversibility in air (R = Clast/Cmax) as a function of log KOA and log KAW and the degradation half-life in air HLAir. Property combinations indicated by blue color indicates reversibility of the contamination in air (i.e., the concentrations in air fifty years after the end of emission are less than 10% of the maximum concentrations during the time period of emission). The degradation half-life in the surface compartment was assumed to be 100 years. The partitioning properties of HCB (1), α-HCH (2), and DDE (3) are indicated by yellow circles.
contamination on the time scale of a human life and those which do not. R-values as a function of the chemical space are displayed in Figure 3. Chemicals with a short HLAir of 10 days always show atmospheric contamination reversibility, if an R threshold of 10% is applied, that is, if we define a lack of reversibility as concentrations fifty years after the end of emission which are on the same order of magnitude as the maximum concentrations at the time of emissions. For substances with a HLAir of 100 days, only the atmospheric contamination of substances with a log KOA between 7 and 8 and a log KAW above 4 is not reversible (R > 10%). As the HLAir increases further, the atmospheric contamination of more and more volatile chemicals fails to be reversible, but the vertical log KOA threshold of 8 remains the same. In conclusion, after primary emissions ceased, evaporation from soils and water bodies can only maintain atmospheric concentrations of POPs at elevated levels (i.e., at at least 10% of maximum levels), if they have fairly high volatility (log KOA < 8) and a long atmospheric degradation half-life (HLair g 100 days). Besides HCB and α-HCH, most POPs fail to fulfill these conditions, because they are either less volatile or more degradable in the atmosphere. Of course the log KOA threshold is dependent on what limit is set for the lack of contamination reversibility. If an R of 0.01 is chosen, that is, concentrations that remain above 1% of the maximum 50 years after the end of emission, a higher log KOA threshold will apply. Interestingly, γ-HCH has a slightly lower KAW, a slightly higher KOA17 and slightly faster OH radical reaction rate constant11 than αHCH, which all conspires to increase its atmospheric contamination
reversibility (i.e., decrease its R). Indeed, after the recent restriction on γ-HCH use globally, its levels in air appear to have declined much more rapidly than those of α-HCH, which was phased out several decades earlier.18 POPs with log KOA higher than 8 and HLAir shorter than 100 days, such as most PCBs, may still evaporate from environmental reservoirs, but the air concentrations will be relatively low compared to those resulting from primary emissions. Or formulated differently, primary PCB emissions are likely to control atmospheric concentrations as long as they occur.4 We should caution that the thresholds in Figure 3 are not just a function of the chemical partitioning properties, but they also depend on environmental characteristics, for example, the retentive capacity of the soil (e.g., fOC), the kinetics of air/surface exchange, and the relative surface coverage of soils and water. For example, in a model environment with higher water surface coverage, the horizontal threshold is lower in the diagram and the vertical threshold is further to the left. What Prevents Fast Reversibility of the Water Concentrations of POPs? All four of the chemical property combinations in the right-hand panels of Figure 1 (long HLSurf) displayed only partial reversibility of contamination in the aqueous environment (in the case of AirPOP and WaterPOP only when long HLAir apply). Considering the general reversibility of atmospheric contamination for the less volatile POPs, the aquatic food web is thus likely the major pathway of human and wildlife exposure for those substances after emissions cease. Time trends of POPs in water, and the factors that maintain relatively high levels in the water after the emissions have stopped, are therefore particularly important to understand. The model considers three input 8839
dx.doi.org/10.1021/es2017544 |Environ. Sci. Technol. 2011, 45, 8834–8841
Environmental Science & Technology
Figure 4. Net-fluxes to and from the fresh water compartment of hypothetical persistent chemicals with the partitioning properties of HCB (AirPOP), α-HCH (WaterPOP), DDE (CenterPOP), and BDE209 (ParticlePOP). Data were normalized to the largest net flux observed during the simulation period.
pathways to the aqueous environment: atmospheric deposition, soil runoff, and release from the sediments. Four processes contribute to the loss of chemical from water: volatilization, degradation, transfer to sediments, and advective outflow. As before, we calculated net chemical fluxes to and from the water compartment as a function of time (Figure 4) as well as normalized cumulative mass balances for the water compartment for the period of emissions (19502000) and the period a few years after emissions have ceased (20052050) (SI Figure S8). A positive flux in Figure 4 refers to an input to water. The dominant pathway pattern for the postemission period (20052050) is from soil to water to air and sediment (SI Figure S8). The major reservoir, soil, continuously supplies chemicals to water, which is a net source to both atmosphere and sediment. In fact, soil-to-water flux becomes the dominant water input process for all four chemicals even before the end of emissions (or, in the case of WaterPOP, already early during the period of declining emissions) (Figure 4). Therefore, the lack of complete reversibility of all four chemicals in water (Figure 1) is due to the continuous supply from soil. The sediment is the ultimate sink of POPs in aquatic system. The only exception is WaterPOP, for which sediment is a net source to water (remobilization from sediment) during the post emission period (SI Figure S8). If a shorter HLAir of 10 days is used for AirPOP (SI Figures S7 and S9), a similar transition is observed in its net sedimentwater flux; net deposition from water to sediment during emissions is followed by net sedimentwater transfer afterward. This result indicates that sediment can be a net source to water after emission ceased for substances for which the atmosphere is an efficient sink, that is, for relatively volatile chemicals with a short or intermediate HLAir. However, even then it is a very small flux and is not responsible for the lack of reversibility in water contamination. To evaluate the influence of environmental parameters on the time trend in water concentrations, the sensitivity to sediment burial rate, particle deposition rate to the sediment, and volume fraction of solids in runoff water was explored (SI Figure S10).
ARTICLE
The influence of these three parameters on the time trends of the four POPs in water was quite limited, especially for AirPOP and WaterPOP. The most influential parameter was the volume fraction of solids in runoff water, which is consistent with particle associated soil-to-water transfer being the major input pathway for POPs to the aqueous environment after the emission ceased. In the case of CenterPOP and ParticlePOP, a 10-fold increase in the soil erosion rate even changes the simulated time profile from partial to no reversibility (SI Figure S10). This result suggests that the time trends of less volatile POPs in different aqueous systems may vary because those systems vary in terms of soilrunoff rates, and to a lesser extent because they vary in terms of sedimentation and burial rates (e.g., oligotrophic versus eutrophic freshwater systems). In particular, we could expect aqueous contamination with POPs to show limited reversibility in high erosion environments, such as some tropical coastal regions. Time trends of the more water-soluble POPs, such as the HCHs, are likely dependent on runoff rates and the residence time of water in a lake or coastal area. As both dissolved and particle-bound substances are supplied to the water compartment by soil runoff, no KOW threshold can be defined for the reversibility of water contamination with POPs. A key outcome of this study is that it should not be surprising at all, that different field studies often generate very divergent time trends. Simulated time trends in this study were easily influenced by degradation half-lives, partitioning properties, and key environmental input parameters. We did not even consider that degradation half-lives differ between different surface media and different climates. Neither did we consider long-range transport even though import of contaminants from distant sources can further add variability to time trends. In other words, we should almost expect large variability in the time trends of POP-like chemicals. Another conclusion is that all but the most persistent POP-like chemicals display reversible contamination. Irrespective of partitioning properties, the simulations showed fast or slowly reversible time trends if the HL in surface media is on the order of a decade. Only if HLsurf was on the order of a century, did the model predict time trends showing no or only partial reversibility. Clearly, if degradation occurs on such a long time scale, other slow loss processes, such as irreversible sequestration in soil, can become relevant, but they are currently not included in the model. We suggest that the following knowledge is required for a quantitative understanding and prediction of the reversibility of POP contamination: (a) half-lives in soils and sediments, whereby it may not be enough to know that it is “long”. At least it needs to be distinguished whether degradation occurs on the time scale of decades or centuries. (b) The extent of soil-to-water transfer, because soil runoff is an important flux to water especially for less volatile POPs that are predominantly associated with particulate organic matter and also for more watersoluble POPs. (c) The residual primary emissions, if any, because controlling processes and pathways are very different before and after emissions have ceased, although possibly less so for less volatile POPs.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional material as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.
8840
dx.doi.org/10.1021/es2017544 |Environ. Sci. Technol. 2011, 45, 8834–8841
Environmental Science & Technology
’ AUTHOR INFORMATION Corresponding Author
*Phone: (416)287-7225; e-mail:
[email protected]. Present Addresses †
Ulsan National Institute of Science and Technology, School of Urban and Environmental Engineering, Ulsan 689798, South Korea
’ ACKNOWLEDGMENT Funding through the Long-range Research Initiative (LRI) of CEFIC is gratefully acknowledged.
ARTICLE
(16) Xiao, H.; Kang, S.; Zhang, Q.; Han, W.; Loewen, M.; Wong, F.; Hung, H.; Lei, Y. D.; Wania, F. Transport of semivolatile organic compounds to the Tibetan Plateau: Monthly resolved air concentrations at Nam Co. J. Geophys. Res. 2010, 115, D16310. (17) Xiao, H.; Li, N.; Wania, F. A compilation, evaluation and selection of physical chemical property data for α, β and γ-hexachlorocyclohexane. J. Chem. Eng. Data 2004, 49, 173–185. (18) Becker, S.; Halsall, C. J.; Tych, W.; Kallenborn, R.; Su, Y.; Hung, H. Long-term trends in atmospheric concentrations of α- and γ-HCH in the Arctic provide insight into the effects of legislation and climatic fluctuations on contaminant levels. Atmos. Environ. 2008, 42, 8225– 8233.
’ REFERENCES (1) Breivik, K.; Sweetman, A.; Pacyna, J. M.; Jones, K. C. Towards a global historical emission inventory for selected PCB congeners—A mass balance approach: 1. Global production and consumption. Sci. Total Environ. 2002, 290, 181–198. (2) Breivik, K.; Sweetman, A.; Pacyna, J. M.; Jones, K. C. Towards a global historical emission inventory for selected PCB congeners—A mass balance approach: 2. Emissions. Sci. Total Environ. 2002, 290, 199–224. (3) Breivik, K.; Wania, F. Evaluating a model of the historical behavior of two hexachlorocyclohexanes in the Baltic Sea environment. Environ. Sci. Technol. 2002, 36, 1014–1023. (4) Hung, H.; Lee, S. C.; Wania, F.; Blanchard, P.; Brice, K. Measuring and simulating atmospheric concentration trends of polychlorinated biphenyls in the Northern Hemisphere. Atmos. Environ. 2005, 39, 6502–6512. (5) Wania, F. A global mass balance analysis of the source of perfluorocarboxylic acids in the Arctic ocean. Environ. Sci. Technol. 2007, 41, 4529–4535. (6) Schenker, U.; Scheringer, M.; Hungerb€uhler, K. Investigating the global fate of DDT: Model evaluation and estimation of future trends. Environ. Sci. Technol. 2008, 42, 1178–1184. (7) Stroebe, M.; Scheringer, M.; Hungerb€uhler, K. Measures of overall persistence and the temporal remote state. Environ. Sci. Technol. 2004, 38, 5665–5673. (8) Gouin, T.; Wania, F. Time trends of Arctic contamination in relation to emission history and chemical persistence and partitioning properties. Environ. Sci. Technol. 2007, 41, 5986–5992. (9) Bignert, A.; Olsson, M.; de Wit, C.; Litzen, K.; Rappe, C.; Reutergardh, L. Biological variation - an important factor to consider in ecotoxicological studies based on environmental samples. Fresenius J. Anal. Chem. 1994, 348, 76–85. (10) Wania, F.; Breivik, K.; Persson, N. J.; McLachlan, M. S. CoZMo-POP 2 - A fugacity-based dynamic multi-compartmental mass balance model of the fate of persistent organic pollutants. Environ. Modell. Software 2006, 21, 868–884. (11) Brubaker, W. W.; Hites, R. A. OH reaction kinetics of gas-phase α- and γ-hexachlorocyclohexane and hexachlorobenzene. Environ. Sci. Technol. 1998, 32, 766–769. (12) Jones, K. C.; Sanders, G.; Wild, S. R.; Burnett, V.; Johnston, A. E. Evidence for decline of PCBs and PAHs in rural vegetation and air in the United Kingdom. Nature 1992, 356, 137–140. (13) Ma, J.; Hung, H.; Blanchard, P. How do climate fluctuations affect persistent organic pollutant distribution in North America? Evidence from a decade of air monitoring. Environ. Sci. Technol. 2004, 38, 2538–2543. (14) Jantunen, L. M.; Bidleman, T. F. Reversal of the air-water gas exchange direction of hexachlorocyclohexanes in the Bering and Chukchi Seas—1993 versus 1988. Environ. Sci. Technol. 1995, 29, 1081–1089. (15) Shen, L.; Wania, F.; Lei, Y. D.; Teixeira, C.; Muir, D. C. G.; Bidleman, T. F. Hexachlorocyclohexanes in the North American atmosphere. Environ. Sci. Technol. 2004, 38, 965–975. 8841
dx.doi.org/10.1021/es2017544 |Environ. Sci. Technol. 2011, 45, 8834–8841
ARTICLE pubs.acs.org/est
Dynamic Multicrop Model to Characterize Impacts of Pesticides in Food Peter Fantke* Institute of Energy Economics and the Rational Use of Energy, University of Stuttgart, Hessbruehlstrasse 49a, 70565 Stuttgart, Germany
Ronnie Juraske Institute of Environmental Engineering, Swiss Federal Institute of Technology Zurich, CH-8093 Zurich, Switzerland
Assumpcio Anton Institute of Agriculture and Food Research and Technology, 08348 Cabrils, Barcelona, Spain
Rainer Friedrich Institute of Energy Economics and the Rational Use of Energy, University of Stuttgart, Hessbruehlstrasse 49a, 70565 Stuttgart, Germany
Olivier Jolliet Environmental Health Sciences, School of Public Health, University of Michigan, 109 S. Observatory, Ann Arbor, Michigan 48109-2029, United States Quantis, EPFL Science Park (PSE-D), CH-1015 Lausanne, Switzerland
bS Supporting Information ABSTRACT: A new dynamic plant uptake model is presented to characterize health impacts of pesticides applied to food crops, based on a flexible set of interconnected compartments. We assess six crops covering a large fraction of the worldwide consumption. Model estimates correspond well with observed pesticide residues for 12 substance-crop combinations, showing residual errors between a factor 1.5 and 19. Human intake fractions, effect and characterization factors are provided for use in life cycle impact assessment for 726 substance-crop combinations and different application times. Intake fractions typically 1 . Human health range from 102 to 108 kgintake kgapplied impacts vary up to 9 orders of magnitude between crops and 10 orders of magnitude between pesticides, stressing the importance of considering interactions between specific crop-environments and pesticides. Time between application and harvest, degradation half-life in plants and residence time in soil are driving the evolution of pesticide masses. We demonstrate that toxicity potentials can be reduced up to 99% by defining adequate pesticide substitutions. Overall, leafy vegetables only contribute to 2% of the vegetal consumption, but due to later application times and higher intake fractions may nevertheless lead to impacts comparable or even higher than via the larger amount of ingested cereals.
’ INTRODUCTION Life cycle impact assessment (LCIA) has been used as a tool to characterize potential toxic impacts on human health and the environment attributable to pesticide use.13 LCIA thereby builds upon substance-specific characterization factors (CF) combining pesticide exposure and toxicity potentials to represent contributions of pesticides to overall human health and environmental impacts. Human health impacts of pesticides, however, are still poorly represented in existing LCIA approaches, since r 2011 American Chemical Society
only effects from diffuse emissions are considered, thereby disregarding ingestion exposure from residues in field crops after direct pesticide application. While in case of diffuse emissions environmental media like air and soil serve as emission target Received: June 11, 2011 Accepted: September 9, 2011 Revised: September 4, 2011 Published: September 09, 2011 8842
dx.doi.org/10.1021/es201989d | Environ. Sci. Technol. 2011, 45, 8842–8849
Environmental Science & Technology
ARTICLE
Table 1. Selected Crops and Represented Archetypes According to a Set of Systematic Criteria Including Share of Crop on Human Consumption of Archetype (ucrop) and Share of Archetype on Total Human Vegetal Consumption (uarchetype) consumption share
knowledge base experimental
crop wheat
archetype cereals
jcrop
jarchetype
residues
68%
24%
medium
characteristics
adaptable models
grass like; grain harvested
Fantke et al.7
residue data Cao16
11
paddy rice
paddy cereals
97%
13%
medium
grass like; paddy water;
Inao et al.
grain harvested tomato
herbaceous fruits
apple
fruit trees
Rein et al. Fantke et al.7
Ishii18
17
15%
26%
high
herbaceous; fruit harvested
Juraske et al.
13%
17%
high
treelike; perennial;
Trapp10
Wang et al.19 8,20
Juraske et al.8,21
and vegetables
potato
leafy vegetables roots and tubers
14% 51%
2% 18%
Xu et al.24
Paraíba
fruit harvested lettuce
Kumar et al.23
22
high
herbaceous; high adsorption;
Kulhanek et al.
Fenoll et al.26,27
medium
leaves harvested herbaceous; root or stem
Juraske et al.9
Juraske et al.9
tuber harvested
Paraíba and Kataguiri28
Abdel-Gawad et al.29
compartments, in case of direct application the cultivated food crop receives the applied mass. As a first attempt to account for effects of direct application, recent studies contrasted residues from direct and diffuse sources for human intake fractions (iF) in fruits and vegetables4,5 and concluded that ingestion of directly treated food crops is the most important human exposure route. As a result, detailed exchange processes between environmental media and vegetation have been introduced in multimedia models designed for LCIA, traditionally considering steady-state conditions. However, for pesticide residues and related impacts, steady-state is usually not reached during the short time period from substance application to ultimate crop harvest, which is why the evolution of residues needs to be assessed dynamically. In addition, pesticide uptake and translocation mechanisms vary considerably between crop species and may indicate significant differences in related health impacts.6 Consequently, differing crop-specific characteristics need to be considered as provided for individual crop species by recently developed plant uptake models.711 Fantke et al.7 compare a wide range of crop-specific uptake models assessing environmental fate of pesticides after direct application and conclude that none of the existing tools is able to contrast various crops consumed by humans. Since this implies a major drawback in characterizing human toxicity for LCIA, the present paper aims at introducing a consistent approach for answering the following questions: (i) How can human intake of pesticides via ingestion of different food crops and related health impacts be characterized and evaluated in a transparent, consistent and concise way? (ii) What is the influence of crop characteristics, substance properties and application times on the dynamic behavior of pesticides in crops and on subsequent human intake? (iii) What are the differences between crop-specific characterization factors from direct pesticide application to different food crops and generic characterization factors from continuous, diffuse emissions to the environment? (iv) How can substitution of pesticides be evaluated and their health impacts compared on a similar functional basis? To answer these questions, we developed a new dynamic assessment model for human health impacts due to uptake of
25
pesticides into multiple crop types (dynamiCROP) based on a transparent matrix algebra framework. We selected six food crops covering a large fraction of the worldwide consumption of vegetal origin, thereby representing the most important crop archetypes. For each archetype, substance-specific human ingestion intake fractions are calculated and evaluated. In addition, the influence of crop and substance characteristics as well as the time between application and harvest on pesticide characterization is discussed. Finally, crop-specific characterization factors are provided— differentiated according to human cancer and noncancer effect information—along with generic characterization factors to also account for diffuse pesticide emissions.
’ MATERIALS AND METHODS Selection of Crops. We introduce six characteristic plant species representing the most relevant crop archetypes with respect to human vegetal food based on a systematic criteria approach. Selection criteria are human consumption quantity, expected pesticide residues in harvest, crop characteristics (cropping practice, plant phenotype, and harvested components), availability of knowledge from other models, and finally availability of experimental data for comparison with modeled residues. Human consumption is analyzed based on global FAO statistics of 159 food crops.12 Ranges of expected residues in food products are estimated based on maximum residue levels (MRL),13 considering MRL data from 70 countries, the European Commission14 and the Codex Alimentarius Commission.15 Table 1 summarizes the criteria analysis and lists selected crop species accounting for 45% of the global vegetal consumption in 2007.12 These crops cover the most important archetypes, that is, cereals (wheat), paddy cereals (rice), herbaceous fruits and vegetables (tomato), fruit trees (apple), leafy vegetables (lettuce), as well as roots and tubers (potato). Based on these archetypes, the model framework can be easily adapted to assess additional species. Assessment Framework. For human impacts, we followed the general LCIA cause-effect chain by linking applied pesticide masses to health impacts via environmental fate, exposure and effects.30 The human-toxicological population impact score, 8843
dx.doi.org/10.1021/es201989d |Environ. Sci. Technol. 2011, 45, 8842–8849
Environmental Science & Technology
ARTICLE
ISi,x (t) [DALY ha1], caused by intake of active ingredient i (hereafter referred to as “pesticide”) applied to crop x that is harvested at time t is expressed as the product of the characterization factor for human toxicity, CFi,x (t) [DALY kg1 applied] with the life cycle inventory output, that is, the total mass of pesticide applied, mapp,i,x [kgapplied ha1]: IStotal, x ¼
n
n
∑ ISi, x ðtÞ ¼ i∑¼ 1 CFi, x ðtÞ mapp, i, x i¼1
ð1Þ
with IStotal,x as total impact score per crop, expressed in DALY (disability adjusted life years as measure for overall population health impacts) per hectare. We determine the characterization factor by multiplying the human effect factor for pesticide i, EFi [DALY kg1 intake], with the total population intake fraction of the pesticide via consumption of crop x, iFi,x (t) [kgintake kg1 applied]: CFi, x ðtÞ ¼ EFi iFi, x ðtÞ ¼ βi DF iFi, x ðtÞ
ð2Þ
Effects Factors. EFi consists of a doseresponse slope factor, βi [incidence risk kg1 intake], and a severity factor DF [DALY incidence1] for distinguishing between cancer and noncancer effects. Severity factors of 11.5 and 2.7 DALY incidence1 for cancer and noncancer effects, respectively, are based on global human health statistics31 and are intended for comparative purposes rather than estimating absolute damages. Slope factors relating potential risks of pesticides in humans to their quantities ingested via food crop consumption are derived from the chronic lifetime dose of pesticide i affecting 50% of a popula1 tion, ED50i [risk kg1 intake person ]. For noncarcinogenic effects, chronic ED50 values are only rarely available. Hence, assuming linear doseresponse relationships, ED50i,s is estimated from no-observed effect levels of exposed animal species s, 1 NOELi,s [mg kg1 applied d ]: βi ¼
cfED50 ¼ cfED50 ED50i NOELi, s cfNOEL BW LT d-to-y cfs cftime mg-to-kg
!1
ð3Þ
BW = 70 kg person1 denotes average body weight, LT = 70 years is the average human lifetime, d-to-y = 365 d yr1 corrects for number of days per year, mg-to-kg = 106 mg kg1 corrects for mg per kg, and cf [] denotes an extrapolation factor. cfED50 = 0.5 accounts for the human response level corresponding to ED50, cfNOEL = 9 is the NOEL-to-ED50 extrapolation factor, cfs corrects for interspecies differences between the studied animal and humans, and cftime accounts for differences in exposure time.31,32 Details on extrapolation factors are provided in Table S3 (Supporting Information (SI)). Intake Fractions. Intake fractions are commonly used in LCIA to express source-to-intake relationships33 and are introduced to describe the mass fractions of applied pesticides that are ultimately taken in by the human population via food ingestion: iFi, x ðtÞ ¼
mintake, i, x ðtÞ ¼ hFi, x ðtÞ PFx mapp, i, x
ð4Þ
with mintake,i,x(t) [kgintake ha1] as mass of pesticide i taken in via ingestion of crop x, hFi,x(t) [kgin harvest kg1 applied] as mass fraction of applied pesticide that is found as residue in crops at harvest time t [days], and finally PFx [] as crop-specific food processing
factor accounting for reduction in pesticide residues between harvest and final consumption. Mass in harvest is a result of the mass balance system of differential equations that is solved analytically by means of matrix algebra implemented in Matlab coupled with a spreadsheet as fully described in Fantke et al.:7 Kt m BðtÞ ¼ e m Bð0Þ
ð5Þ
with m B(t) a column vector of substance masses at time t in the different compartments [kg], and K the square matrix of firstorder rate coefficients [d1], in which diagonal elements represent overall removal rate coefficients in compartments and offdiagonal elements correspond to single intercompartmental transfer rate coefficients. In contrast to other plant uptake models focusing on a single-species environment, the present model accounts for system characteristics of various crops based on a flexible set of interconnected compartments. This allows to directly compare modeled crop-specific residues with experimentally derived concentrations in harvested plants and to contrast model results against official MRLs. Parameters determining the fate processes are physicochemical properties of applied pesticides, system initial and boundary conditions as well as cropspecific characteristics. Crop Characteristics. Initial deposition of pesticides on plant surfaces directly after application depends on the crop-specific leaf area index (LAI) representing leaf growth stage, and furthermore the pesticide capture coefficient being a measure of a crop’s capture efficacy.34 While the latter can be seen as a fixed value, crop growth and related LAI development are timedependent. Logistic crop development is applied for all studied crops, describing an initial exponential growth that finally saturates at some maximum level.7,11 However, evolution of leaf area follows a more complex behavior. Crop-specific LAI curves based on experimental data are applied for wheat,35 paddy rice,36 tomato,37 apple,38 lettuce39 and potato40 as shown in Figure S1 (SI). After a pesticide is applied to a crop, a certain fraction of the initial dose has the potential for drifting from the agricultural site. In most cases, drift mainly depends on application method, pesticide formulation, environmental conditions and crop type.41 Typical crop type- and application method-specific drift values are applied as loss fractions. Finally, in order to evaluate the effect of food processing on the magnitude of pesticide residues in the studied crops, food processing factors for washing, peeling, cooking, juicing, and baking are adopted from Kaushik et al.42 Model Evaluation. Internal model consistency was continuously examined by checking the underlying mass balance, that is, ensuring that the sum of elimination and biodistribution pathways at any time equals the total pesticide mass applied. Second, modeled residues are evaluated by analyzing whether model simulations adequately represent collected experimental data from the literature as given in Table 1. The measure used to estimate model prediction quality compared with experiments is the residual error, also known as standard deviation of the log of residuals between observed and modeled concentrations.43 A residual error of, for example, 0.5 implies a deviation between modeled and experimental data of approximately a factor 10Student’s t0.5. For a Student’s t-value of ∼2, we would then arrive at a factor 10 deviation. Third, we studied model sensitivity to determine the influence of the most important parameters on output variability. In addition, the newly calculated crop-specific characterization factors for direct residues are compared and eventually combined 8844
dx.doi.org/10.1021/es201989d |Environ. Sci. Technol. 2011, 45, 8842–8849
Environmental Science & Technology
ARTICLE
Figure 1. Measured and modeled pesticide residues in plant components harvested for human consumption for each of the six studied crops with residual error (ER) for each substance-crop combination.
with generic characterization factors accounting for diffuse emission pathways as calculated with the USEtox model5 for the set of selected substances.
’ RESULTS Pesticide Residues in Crops. Modeled residues are compared with measured concentrations of eleven different pesticides applied to the six selected crops (Figure 1). Experimentally derived maximum concentrations are reported to range from 29 mg kg1 in apples at the day of application to 0.01 mg kg1 in potato tubers measured 15 days after the tested pesticide was applied, demonstrating a variability of 3 orders of magnitude between crops. Measurements and model estimates correspond well with total crop-specific residual errors ranging between 0.08 (factor 1.5 deviation) for fenitrothion applied to lettuce and 0.64 (factor 19) for propisochlor sprayed on rice with an overall residual error of 0.33 (factor 4.5) over all 12 substance-crop combinations. A higher accuracy of prediction is observed in crops where the final commodity stands in direct contact with the applied pesticide (apple, lettuce, tomato, and wheat). In comparison, crops in which the pesticide has to pass an additional medium like paddy water (rice) or soil (potato) in order to reach the harvested good, on average showed higher uncertainties. Human Intake Fractions. We calculated intake fractions and characterization factors as measures normalized to one unit mass of applied pesticide for 726 potential substance-crop combinations, that is, 121 substances applied to six crops. Intake fractions of a pesticide can vary between 4 and 14 orders of magnitude for fenoxaprop-p and flufenacet, respectively, when applied to different crops at recommended doses and harvested at typical times after applications. In contrast, iF between all pesticides applied to the same crop can vary between three (potato) and sixteen (wheat) orders of magnitude, thereby indicating that substance properties are almost as influential on iF as crop properties.
The lowest intake fraction is found for metam sodium applied to apple with iF = 1.7 1020 kgintake kg1 applied , whereas the highest intake fraction is found for epoxiconazole applied to lettuce with iF = 1.9 101 kgintake kg1 applied. Intake fractions for all potential substance-crop combinations are provided in Table S2 (SI). Substance degradation in plants and the time between pesticide application and crop harvest are known to predominantly determine model sensitivity.7,20 In Figure 2, we present intake fractions as a function of the substance degradation half-life in plants— typically varying between 1 and 10 days—for different times to harvest. For all crops but potato, intake fractions are typically in the range of 102 and 108 kgintake kg1 applied for typical times between application and harvest. Decreasing the half-life in plants results in continuously decreasing iF, with the magnitude of decrease being amplified up to 10 orders of magnitude with higher time lag between application and harvest. In this case, the difference in degradation of pesticides within crops has more time to establish a significant influence. When looking at potato, there is little influence of the degradation half-life in plant. Instead, the residence time in soil is a main factor of influence affecting iF variation in Figure 2. Residence times in soil are more adequate than half-lives to examine iF variation for potato, since they encompass the various removal processes in soil and since the pesticide always has to pass the heterogeneous soil layer, before entering the tuber. In this case, soil characteristics become predominant.9,28 How crop and substance characteristics as well as time to harvest are influencing the variation of intake fractions is contrasted in Figure 3. For the same generic time between application and harvest of 20 days (dark boxes), potato shows the lowest range of intake fractions with a median of 2.7 107 kgintake kg1 applied and less than 3 orders of magnitude variation between 5th and 95th percentiles. Potato is followed by cereals and fruit crops, for which we basically obtain a similar behavior with median iF 8845
dx.doi.org/10.1021/es201989d |Environ. Sci. Technol. 2011, 45, 8842–8849
Environmental Science & Technology
ARTICLE
Figure 2. Intake fractions as a function of degradation half-lives of pesticides (n = 121) in plants for different time periods between substance application and crop harvest (Δt) for each of the six studied crops.
Figure 3. Box and whisker plot of human intake fractions for pesticides directly applied to the six selected crops (n = 121, dynamic assessment for different times between application and harvest Δt) and from diffuse emissions (n = 97, assessed with USEtox, steady state, left box: emissions to air, right box: emissions to soil).
ranging from 2 105 to 2.5 104 kgintake kg1 applied and typical variation ranges from 2 to 4 orders of magnitude between pesticides. In contrast, lettuce as leafy crop shows highest iF with a median of 4.3 104 kgintake kg1 applied and 6 orders of magnitude variation. Another set of influencing factors are the intrinsic crop characteristics, mainly due to losses during application via wind drift, LAI growth over time and food processing after harvest. Drift fractions, however, are not only crop-specific, but also depend on application method, for example, foliar spray or soil injection, whereas loss fractions due to food processing also differ between substances. In practice, times between application and harvest depend on crop species, pest occurrence, weather conditions and a pesticide’s mode of action. However, since for LCIA we are interested in providing best estimates, we also need to distinguish between pesticide target classes for arriving at typical times to harvest. For fungicides and insecticides, officially reported minimum preharvest intervals are selected. Herbicides, in contrast, are usually applied pre-emergent or during early crop growth resulting in relatively long times to harvest between 55 and 150 days as discussed in the SI. Overall, average times to
harvest range between 5 days for fungicides/insecticides applied to tomato and 150 days for herbicides applied to wheat and apple (SI Table S1). Varying application times lead to additional iF variation between pesticides with herbicides showing lower intake fractions and higher variation due to their longer time lags between application and harvest. (Figure 3, gray middle boxplots). For fungicides and insecticides, the later application leads to higher intake fraction, especially for tomato and lettuce, for which application can take place only 510 days before harvest (Figure 3, left box plots). Figure 3 also enables us to compare direct pesticide application modeled dynamically with diffuse emissions calculated by USEtox assuming steady-state conditions and continuous input (see two white box-plots at the right end of Figure 3). With the generic time to harvest of 20 days, all crops except potato show higher iF due to direct application residues compared to iF due to continuous, diffuse emissions. For recommended times to harvest, median iF of herbicides applied to all crops and of fungicides/insecticides applied to cereals decrease below USEtox values. In contrast, for fungicide/insecticide applied shortly before harvest (tomato, lettuce), median intake fractions 8846
dx.doi.org/10.1021/es201989d |Environ. Sci. Technol. 2011, 45, 8842–8849
1.2 108 (4.2 10119.9 106) 2.6 10 ) (8.1 10 1.4 10 (2.4 10 4.7 10 )
6
See text for explanation.
0.89 (0.03101) potato
1.8 10
9
0.2 (0.031.6) lettuce
4.3 10
Figure 4. Human toxicity impact scores of different scenarios expressed in DALY per ha of applied fungicides, insecticides, herbicides and total pesticides applied on wheat, and relative impact scores normalized to scenario no. 1.
a
8.5 105 (1.3 1084.9 103)
6
4.8 10 9.5 10 ) (1.8 10
7
2 14
ARTICLE
8
(6.3 10
11
4.1 10 )
9.0 106 (9.5 10154.2 104)
2
2.2 10 ) (1.4 10
16 4
1.4 10 6.0 10 ) (1.7 10
3
0.86 (0.043.2) apple
9.5 10
2.4 105 (4.8 1096.3 103)
3 20
4.1 104 (2.1 10131.7 102)
5 18
3 5
0.27 (0.024.8) tomato
3.9 103 (1.8 10121.7 102)
3.1 108 (2.5 10153.6 105) 5.7 107 (3.2 10139.7 104) 1.2 107 (5.7 10185.5 105) 1.3 106 (1.0 10123.3 104) cancera: 0.17 (0.0071.6) noncancer: 0.11 (0.0092.5) 0.16 (0.012.9) 0.75 (0.043.8) wheat rice (paddy)
1.0 106 (3.4 10163.9 104) 6.9 106 (2.3 10102.3 103)
IS [DALY ha1] CF [DALY kg1 applied] EF [DALY kg1 intake] iF [kgintake kg1 applied] mapp [kgapplied ha1] crop
Table 2. Median Values with 5th and 95th Percentiles (In Brackets) of Crop-Specific Application Amount mapp, Intake Fraction iF, Characterization Factor CF, CropIndependent Effect Factor EF, and Impact Score IS for 121 Pesticides
Environmental Science & Technology
from diffuse emissions strongly underestimate overall intake by up to 4 orders of magnitude. Finally, in the case of potatoes, residues from direct application of all pesticides remain of minor importance, that is, with lower median iF values, than diffuse emissions. Characterization Factors and Impact Scores. Characterization factors for the set of 121 pesticides applied to the six selected food crops are calculated according to eq 2 to characterize the direct impact per kg of pesticide applied. However, characterization factors for substance-crop combinations not known to be officially authorized shall only be used as reference for further crops. Therefore, overall impacts per ha are only calculated for a subset of 181 substance-crop combinations authorized for use in at least one of the countries listed in the Codex Alimentarius15 with given recommended amounts applied. The whole source to impact pathway from pesticide application to human impacts is summarized in Table 2. Variability is before all due to variation in intake fraction (15 orders of magnitude). Additional limited variability is introduced by human effect factors, more specifically by substance-specific doseresponse slope factors. Information related to cancer effects is given in Rosenbaum et al.5 for less than 20% of the 121 pesticides, which is in line with Huijbregts et al.31 In addition, 77% of substances with available information related to cancer do not show any cancer potential (effect factors set to zero). This indicates that most of today’s pesticides are rather leading to noncancer effects. All in all, effect factors vary by 3.5 orders of magnitude between pesticides. The combination of large variations in intake fractions with lower variations in effect factors leads to an overall variation in characterization factors of almost 17 orders of magnitude (Table 2). Combining this with variability in applied pesticide mass of 4 orders of magnitude tends to reduce the overall variability on the impact score to 13 orders of magnitude, suggesting that some of the pesticides applied at low dose tend to have rather high toxicity potentials. Detailed information is provided in the SI on cancer and noncancer potentials (SI Table S3), crop-specific and generic characterization factors (SI Tables S4 and S5), and finally impact scores (SI Figure S2). Functional Assessment and Pesticide Substitution. When assessing the change in impacts linked to substitution of pesticides, a functional assessment is required, ensuring that the combination and quantities of pesticides applied are able to control a set of distinct pests in an equivalent way. Figure 4 presents an example of how to conduct crop-specific substitution of different pesticides. Pesticide target classes focus 8847
dx.doi.org/10.1021/es201989d |Environ. Sci. Technol. 2011, 45, 8842–8849
Environmental Science & Technology on distinct pest categories, for example, fungi, insects, weeds. Substitution, hence, must be discussed separately within each pesticide target class, for example, insecticides can only be substituted by other insecticides targeting the same insect pests. In scenario no. 1, we exemplarily combined applications of βcyfluthrin and carbaryl on wheat against a set of common wheatdamaging insects (wheat bulb fly, cereal leaf beetle, aphids, and thrips). This insecticide mix is substituted by a combination of the less human health impacting insecticides α-cypermethrin and deltamethrin in scenario no. 3. Individual insecticides and application rates are chosen to ensure a similar ability to control the same unwanted insects in both scenarios. Figure 4 demonstrates that substituting scenario no. 1 by scenario no. 3 reduces the total impact score of applied insecticides by more than 4 orders of magnitude to less than 0.1% of the impact score of scenario no. 1. This approach is similarly applied to fungicides and herbicides, where substituting scenario no. 1 by scenario no. 3 results in impact scores reduced by more than 2 and 6 orders of magnitude, respectively. Scenario no. 2 represents an intermediary situation showing some, but not as much reduction in impact scores for all target classes as scenario no. 3. Background information for this example of pesticide substitution is provided in Table S6 (SI).
’ DISCUSSION Potential and Limitations. The presented approach demonstrates the importance of dynamic pesticide assessment and enables us to distinguish between various food crops. By identifying the combined influence of crop characteristics, application times and substance properties we demonstrate that it is crucial to choose appropriate times to apply pesticides and that diffuse emission pathways may be significant for early application, but strongly underestimate human intake for late application before harvest. For practical implementation, we therefore recommend to use our crop-specific characterization factors to account for direct application residues. We thereby stress that results must always be interpreted as a function of times to harvest, that is, recalculations are required to account for changing the date of pesticide application. To account for impacts from diffuse emissions in addition to direct residues, initial loss fractions to air and soil during application should be multiplied by the USEtox characterization factors for emissions to urban air and agricultural soil, respectively. For typical foliar application, crop-specific loss fractions to air range from 5 to 25% and to soil from 5 to 70%, respectively, where the latter also depends on crop development stage. All presented crop-specific characterization factors are global averages and based on generic values for most underlying parameters, such as human lifetime and body weight. Hence, for a spatial assessment, these parameters need to be adjusted accordingly. The present approach is so far limited to neutral organic substances, since inorganics require a different consideration of their partitioning behavior and ionizable compounds require considering electrochemical interactions for the dissociated species. Differences between Crops. Variation in crop-specific intake is mainly driven by distinct characteristics between crop archetypes, for example, with respect to harvested plant components, from which we can basically classify food crops into roots and tubers, fruits and cereals, as well as leafy vegetables. Overall, leafy vegetables only contribute to 2% of the total human vegetal consumption, but may nevertheless lead to human impacts
ARTICLE
comparable or even higher than via ingestion of cereals. Cereals, on the other hand, contribute to 37% of the human vegetal consumption (including paddy cereals), but substances are usually applied earlier for these crops, leading to lower intake fractions. Highest impacts are expected via consumption of herbaceous crops and fruit trees with usually high intake fractions and consumption, while roots and tubers only contribute little due to very low intake fractions. Pesticide Substitution. Whenever developing substitution scenarios, we strongly recommend considering aspects related to pesticide authorization, since a substance may be authorized for use on particular crops in some countries, but not in others, because of decreased susceptibility of target pests (resistance) to certain pesticides. Furthermore, for considering multiple applications at different application times, eq 1 can be summed up over various applications in addition to summing up over pesticides. However, usually only the latest application plays a predominant role due to increasing reduction of intake fractions with time.9
’ ASSOCIATED CONTENT
bS
Supporting Information. The dynamic multicrop model is available at http://dynamicrop.org. Detailed information on LAI development, application rates, intake fractions, human effect information, characterization factors, impact scores and pesticide substitution is provided. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (+49) 711 685 878 45; fax: (+49) 711 685 878 73; e-mail:
[email protected].
’ ACKNOWLEDGMENT This work is financially supported by the project “Life Cycle Impact Assessment Methods for Improved Sustainability Characterisation of Technologies” (LC-IMPACT), Contract No. 243827, funded by the European Commission under the Seventh Framework Programme. Furthermore, we thank Dr. Rapha€el Charles, Eva Sevigne Itoiz, Anna Kounina, and Prof. Mark Huijbregts for their excellent scientific support. ’ REFERENCES (1) Humbert, S.; Margni, M. D.; Charles, R.; Salazar, O. M. T.; Quir os, A. L.; Jolliet, O. Toxicity assessment of the main pesticides used in Costa Rica. Agr. Ecosyst. Environ. 2007, 118, 183–190. (2) Juraske, R.; Mutel, C. L.; Stoessel, F.; Hellweg, S. Life cycle human toxicity assessment of pesticides: Comparing fruit and vegetable diets in Switzerland and the United States. Chemosphere 2009, 77, 939–945. (3) Margni, M. D.; Rossier, D.; Crettaz, P.; Jolliet, O. Life cycle impact assessment of pesticides on human health and ecosystems. Agr. Ecosyst. Environ. 2002, 93, 379–392. (4) Pennington, D. W.; Margni, M. D.; Ammann, C.; Jolliet, O. Multimedia fate and human intake modeling: Spatial versus nonspatial insights for chemical emissions in Western Europe. Environ. Sci. Technol. 2005, 39, 1119–1128. (5) Rosenbaum, R. K.; Bachmann, T. M.; Gold, L. S.; Huijbregts, M. A. J.; Jolliet, O.; Juraske, R.; Koehler, A.; Larsen, H. F.; MacLeod, M.; Margni, M. D.; McKone, T. E.; Payet, J.; Schuhmacher, M.; van de Meent, D.; Hauschild, M. Z. USEtox - the UNEP-SETAC toxicity 8848
dx.doi.org/10.1021/es201989d |Environ. Sci. Technol. 2011, 45, 8842–8849
Environmental Science & Technology model: Recommended characterisation factors for human toxicity and freshwater ecotoxicity in life cycle impact assessment. Int. J. Life Cycle Assess. 2008, 13, 532–546. (6) Trapp, S.; Kulhanek, A. Human exposure assessment for food— One equation for all crops is not enough. In Phytoremediation and Rhizoremediation; Mackova, M., Dowling, D., Macek, T., Eds.; Springer Press: Dordrecht, 2006; pp 285300. (7) Fantke, P.; Charles, R.; de Alencastro, L. F.; Friedrich, R.; Jolliet, O. Plant uptake of pesticides and human health: Dynamic modeling of residues in wheat and human intake. Chemosphere 2011, DOI: 10.1016/ j.chemosphere.2011.1008.1030. (8) Juraske, R.; Anton, A.; Castells, F.; Huijbregts, M. A. J. Human intake fractions of pesticides via greenhouse tomato consumption: Comparing model estimates with measurements for captan. Chemosphere 2007, 67, 1102–1107. (9) Juraske, R.; Vivas, C. S. M.; Velsquez, A. E.; Santos, G. G.; Moreno, M. B. B.; Gomez, J. D.; Binder, C. R.; Hellweg, S.; Dallos, J. A. G. Pesticide uptake in potatoes: Model and field experiments. Environ. Sci. Technol. 2011, 45, 651–657. (10) Trapp, S. Fruit tree model for uptake of organic compounds from soil and air. SAR QSAR Environ. Res. 2007, 18, 367–387. (11) Rein, A.; Legind, C. N.; Trapp, S. New concepts for dynamic plant uptake models. SAR QSAR Environ. Res. 2011, 22, 191–215. (12) The FAO (Food and Agriculture Organization of the United Nations) Statistical Database; FAO, data retrieved on March 3, 2011. (13) International Maximum Residue Limit Database; United States Department of Agriculture, data retrieved on April 15, 2011. (14) Commission Regulation (EC) No 149/2008 setting maximum residue levels for products covered by Annex I thereto (January 29, 2008); European Commission. (15) Codex Pesticides Residues in Food Online Database, Joint FAO/WHO Food Standards Programme; Codex Alimentarius Commission, 2010. (16) Cao, T. T. Fate and a Prediction Model of Pesticides Residues cole Polytechnique Federale de Lausanne, Evolution in Plants. Thesis, E Lausanne, 2001. (17) Inao, K.; Watanabe, H.; Karpouzas, D. G.; Capri, E. Simulation models of pesticide fate and transport in paddy environment for ecological risk assessment and management. JARQ 2008, 42, 13–21. (18) Ishii, Y. A comparative study of the persistence of organophosphorous and carbamate insecticides in rice plants at harvesting. Bull. Natl. Inst. Agro-Environ. Sci. 2004, 23, 1–14. (19) Wang, S.-L.; Liu, F.-M.; Jin, S.-H.; Jiang, S.-R. Dissipation of propisochlor and residue analysis in rice, soil and water under field conditions. Food Control 2007, 18, 731–735. (20) Juraske, R.; Castells, F.; Vijay, A.; Mu~noz, P.; Anton, A. Uptake and persistence of pesticides in plants: Measurements and model estimates for imidacloprid after foliar and soil application. J. Hazard. Mater. 2009, 165, 683–689. (21) Juraske, R.; Anton, A.; Castells, F. Estimating half-lives of pesticides in/on vegetation for use in multimedia fate and exposure models. Chemosphere 2008, 70, 1748–1755. (22) Paraíba, L. C. Pesticide bioconcentration modelling for fruit trees. Chemosphere 2007, 66, 1468–1475. (23) Kumar, V.; Sood, C.; Jaggi, S.; Ravindranath, S. D.; Bhardwaj, S. P.; Shanker, A. Dissipation behavior of propargite—An acaricide residues in soil, apple (Malus pumila) and tea (Camellia sinensis). Chemosphere 2005, 58, 837–843. (24) Xu, X.-M.; Murray, R. A.; Salazar, J. D.; Hyder, K. The temporal pattern of captan residues on apple leaves and fruit under field conditions in relation to weather and canopy structure. Pest Manage. Sci. 2008, 64, 565–578. (25) Kulhanek, A.; Trapp, S.; Sismilich, M.; Janku, J.; Zimova, M. Crop-specific human exposure assessment for polycyclic aromatic hydrocarbons in Czech soils. Sci. Total Environ. 2005, 339, 71–80. (26) Fenoll, J.; Helln, P.; Camacho, M. d. M.; Lpez, J.; Gonzlez, A.; Lacasa, A.; Flores, P. Dissipation rates of procymidone and azoxystrobin in greenhouse grown lettuce and under cold storage conditions. Int. J. Environ. Anal. Chem. 2008, 88, 737–746.
ARTICLE
(27) Fenoll, J.; Hellín, P.; Lopez, J.; Gonzalez, A.; Lacasa, A.; Flores, P. Dissipation rates of fenitrothion in greenhouse grown lettuce and under cold storage conditions. Int. J. Food Sci. Technol. 2009, 44, 1034–1040. (28) Paraíba, L. C.; Kataguiri, K. Model approach for estimating potato pesticide bioconcentration factor. Chemosphere 2008, 73, 1247–1252. (29) Abdel-Gawad, H.; Afifi, L. M.; Abdel-Hameed, R. M.; Hegazi, B. Distribution and degradation of 14C-ethyl prothiofos in a potato plant and the effect of processing. Phosphorus Sulfur 2008, 183, 2734–2751. (30) Life-Cycle Impact Assessment: Striving Towards Best Practice; Udo de Haes, H. A., Finnveden, G., Goedkoop, M., Hauschild, M. Z., Hertwich, E., Hofstetter, P., Jolliet, O., Kl€opffer, W., Krewitt, W., Lindeijer, E., M€uller-Wenk, R., Olsen, S., Pennington, D. W., Potting, J., Steen, B., Eds.; SETAC Press: Pensacola, FL, 2002. (31) Huijbregts, M. A. J.; Rombouts, L. J. A.; Ragas, A. M. J.; van de Meent, D. Human-toxicological effect and damage factors of carcinogenic and noncarcinogenic chemicals for life cycle impact assessment. Integr. Environ. Assess. Manage. 2005, 1, 181–244. (32) Huijbregts, M.; Hauschild, M.; Jolliet, O.; Margni, M.; McKone, T.; Rosenbaum, R. K.; van de Meent, D. USEtoxTm User Manual, Version 1.01, 2010. (33) Bennett, D. H.; McKone, T. E.; Evans, J. S.; Nazaroff, W. W.; Margni, M. D.; Jolliet, O.; Smith, K. R. Defining intake fraction. Environ. Sci. Technol. 2002, 36, 207A–211A. (34) Gyldenkaerne, S.; Secher, B. J. M.; Nordbo, E. Ground deposit of pesticides in relation to the cereal canopy density. Pestic. Sci. 1999, 55, 1210–1216. (35) Lenz, V. I. S. A process-based crop growth model for assessing global change effects on biomass production and water demand. Ph.D. Disserstation, University of K€oln, K€ oln, 2007. (36) Chen, J.; Lin, H.; Liu, A.; Shao, Y.; Yang, L. A semi-empirical backscattering model for estimation of leaf area index (LAI) of rice in southern China. Int. J. Remote Sens. 2006, 27, 5417–5425. (37) Anton Vallejo, A. Utilizacion del analisis del ciclo de vida en la evaluacion del impacto ambiental en el cultivo bajo invernadero mediterraneo. Ph.D. Dissertation, Univerisitat Politecnica de Catalunya, Barcelona, 2004. (38) Gong, D.; Kang, S.; Zhang, L.; Du, T.; Yao, L. A two-dimensional model of root water uptake for single apple trees and its verification with sap flow and soil water content measurements. Agric. Water Manage. 2006, 83, 119–129. (39) Carranza, C.; Lanchero, O.; Miranda, D.; Chaves, B. Analisis del crecimiento de lechuga (Lactuca sativa L.) ’Batavia’ cultivada en un suelo salino de la Sabana de Bogota. Agron. Colomb. 2009, 27, 41–48. (40) Eremeev, V.; Joudu, J.; L€a€aniste, P.; M€aeorg, E.; Makke, A.; Talgre, L.; Lauringson, E.; Raave, H.; Noormets, M. Consequences of pre-planting treatments of potato seed tubers on leaf area index formation. Acta Agric. Scand., Sect. B 2008, 58, 236–244. (41) Wolters, A.; Linnemann, V.; Zande, J. C. v. d.; Vereecken, H. Field experiment on spray drift: Deposition and airborne drift during application to a winter wheat crop. Sci. Total Environ. 2008, 405, 269–277. (42) Kaushik, G.; Satya, S.; Naik, S. N. Food processing a tool to pesticide residue dissipation - a review. Food Res. Int. 2009, 42, 26–40. (43) Hamburg, M.; Young, P. Statistical Analysis for Decision Making, 6th ed.; Dryden Press: Fort Worth, London, 1994.
8849
dx.doi.org/10.1021/es201989d |Environ. Sci. Technol. 2011, 45, 8842–8849
ARTICLE pubs.acs.org/est
On the Construction, Comparison, and Variability of Airsheds for Interpreting Semivolatile Organic Compounds in Passively Sampled Air John N. Westgate and Frank Wania* Department of Physical and Environmental Sciences, University of Toronto Scarborough, 1265 Military Trail, Toronto, Ontario, Canada M1C 1A4
bS Supporting Information ABSTRACT: Air mass origin as determined by back trajectories often aids in explaining some of the short-term variability in the atmospheric concentrations of semivolatile organic contaminants. Airsheds, constructed by amalgamating large numbers of back trajectories, capture average air mass origins over longer time periods and thus have found use in interpreting air concentrations obtained by passive air samplers. To explore some of their key characteristics, airsheds for 54 locations on Earth were constructed and compared for roundness, seasonality, and interannual variability. To avoid the so-called “pole problem” and to simplify the calculation of roundness, a “geodesic grid” was used to bin the back-trajectory end points. Departures from roundness were seen to occur at all latitudes and to correlate significantly with local slope but no strong relationship between latitude and roundness was revealed. Seasonality and interannual variability vary widely enough to imply that static models of transport are not sufficient to describe the proximity of an area to potential sources of contaminants. For interpreting an air measurement an airshed should be generated specifically for the deployment time of the sampler, especially when investigating long-term trends. Samples taken in a single season may not represent the average annual atmosphere, and samples taken in linear, as opposed to round, airsheds may not represent the average atmosphere in the area. Simple methods are proposed to ascertain the significance of an airshed or individual cell. It is recommended that when establishing potential contaminant source regions only end points with departure heights of less than ∼700 m be considered.
’ INTRODUCTION Striving to understand both the current levels of semivolatile organic contaminants (SVOCs) in the atmosphere, as well as the effectiveness of efforts to curtail the levels of some that have been identified for restriction or elimination under the Stockholm Convention1 or the LRTAP Protocol,2 several investigators have deployed passive air samplers (PASs).38 The relatively low atmospheric concentrations of SVOCs combined with the low sampling rates of these samplers dictate that PASs must be deployed for at least 2 months to over 1 year to sequester detectable levels of contaminants.9 To better understand the relationship between the levels of SVOCs measured in air to their potential sources it has become popular to compare them to calculated air-mass back trajectories (BTs).1016 Oehme generated BTs specific to episodes of high organochlorine compound concentrations in air at several Norwegian sampling stations,10 while Haugen et al. divided BTs coincident with all of their measurements by origin to more clearly define potential sources of hexachlorocyclohexanes.11 More robust statistical methods of analyzing BTs, such as flow climatologies,17 cluster analysis,18 potential source contribution functions,12 and concentration weighted trajectories,13 require r 2011 American Chemical Society
many air measurements at the same location, preferably with a high temporal resolution, and are not well-suited to the longer sampling times of PAS campaigns, which sometimes only have one sample per location. However, Hageman and et al. adapted cluster analysis in a manner which when combined with a usage estimate gave a metric that correlated with SVOCs measured in mountain snow, a type of natural semipassive air sampler.14 Munn et al. combined BTs using what they called a crossing trajectory analysis;19 the approach was identical to the construction of an airshed, except that the trajectories were hand-selected to coincide with pollution events. The time integrated influence function described by Uliasz is theoretically nearly identical to what is called an airshed in this work.20 Gouin et al. amassed large numbers of BTs into airsheds by binning the end points of the individual segments of the BTs into cells,15 allowing for a visual and semiquantitative analysis of the history of the air that was sampled around the Great Lakes (e.g., refs 15, 22); in such an Received: July 19, 2011 Accepted: September 2, 2011 Revised: August 29, 2011 Published: September 02, 2011 8850
dx.doi.org/10.1021/es202490b | Environ. Sci. Technol. 2011, 45, 8850–8857
Environmental Science & Technology airshed, the fraction of end points that lie in a given cell represents the relative contribution of that cell to the sample. Stohl and co-workers22,23 and Seibert and Frank24 used a stochastic Lagrangian particle dispersion model to define a similar quantity they later called potential emission sensitivity (PES) based on the residence time of modeled particles from a “retroplume”.23 The retroplume has chiefly been applied to relatively short-term, active air samples, but it has recently been used to predict measurements in PASs.25 All of these methods are similar in that they do not require any estimate of emissions per se but can be combined with such information without the need for recalculation. Westgate et al.26 compared airsheds against a simple dispersion model of air movement for quantifying proximity to sources of contaminants. It was found that on a site-by-site basis the two agreed only for some locations and that in some circumstances the more sophisticated but data intensive airsheds were necessary to capture the spatial history of sampled air.26 This work explores more closely the reasons for those differences by defining a “roundness” value for airsheds, allowing a quantitative evaluation of departures from the idealized diffusive model. Attempts are made to address the following questions: How can the significance of an airshed, or an individual cell of an airshed, be ascertained? Are large-scale circulations and terrain effects evident in airsheds, and can an airshed be approximated from them? Do airsheds have long-term stability, or do they vary from seasonto-season or from year-to-year? What is a suitable constraint on the departure heights of trajectories when assigning potential contaminant source regions? To these ends, the airsheds of multiple areas are compared for several attributes, including roundness and seasonal and interannual differences. Suggestions for refining the interpretation of airsheds are proposed, including a minimum number of BTs to ensure the significance of any observed trends, or lack thereof, a limited vertical component when assigning potential source regions, and the employment of equally spaced grid cells rather than a simple grid of 1° of latitude by 1° of longitude quadrilaterals.
’ METHODS Air-Mass Back Trajectories. All of the BTs in this work were generated by the Canadian Meteorological Centre using their trajectory model.27 The nature of these trajectories was described in detail previously.26 Briefly, a 5-day BT for a given arrival point would consist of 20 points in space representing the location of the sampled air every 6 h backward in time, with the intrinsic assumption that the air-mass maintains its integrity. The height associated with each datum is provided in meters above Earth’s surface, but it should be noted that the topographic information in the model is at the same coarse resolution as the meteorological inputs: 100 km resolution for BTs prior to November 30, 2006, and 33 km thereafter. In addition to those analyzed by Westgate al.,26 BTs from two sites in Sichuan Province, China,7 four semirural sites in Ontario, Canada,28 a location on the Devon Ice Sheet in the Canadian Arctic,29 and a site from Yukon Territory, Canada,30 were also examined [see Table S1 in the Supporting Information (SI)]. The Generation of Airsheds from Back Trajectories. An airshed is compiled first by collecting a set of BTs for air-masses arriving at a point on Earth’s surface for the period of interest, hereafter referred to as the “arrival point” or simply the “site”. As
ARTICLE
an example, to generate a 2-month airshed for January and February for location A using 5-day BTs with a 6 h resolution, a BT for air arriving at A at midnight, Greenwich Mean Time (00Z) giving its position at 1800Z, 1200Z, 0600Z, and 00Z, December 31; 1800Z, 1200Z, 0600Z, and 00Z, December 30; etc.; to 0600Z, December 27, is calculated. Each of these positions is the end point of one segment of the BT. A second BT for air arriving at A at 0600Z on January 1 would give that airmass position every 6 h back in time for 20 segments to 1200Z, December 27. A BT would be calculated for air arriving every 6 h until 1800Z, February 28 (or 29). These BT end points are then binned: in effect, a grid of cells is overlaid and the number of end points that fall within each cell is counted. As in a previous work, BTs were calculated for air-masses arriving at three different heights: 50, 100, and 200 m above the Earth’s surface. These are the heights input at the start of the calculation and are sometimes referred to as “starting heights”. Here they are called “arrival heights” to indicate the temporal direction. The three sets of BTs were binned and considered both as individual airsheds and as combined into a single airshed. At the end of November 2006, the resolution of the meteorological data used to calculate BTs increased from 100 to 33 km. BTs that include end points for November 30, 2006, are nonsensical and were excluded from airsheds in this work. End points from before and after the resolution change were treated identically. Constructing the Geodesic Grid. To avoid what is sometimes called the pole problem (see ref 26), a set of equally spaced cells on Earth’s surface was defined. Rather than a simple 1° of latitude by 1° of longitude set, the cell centers were generated by first defining the vertices of an icosahedron. The edges of that polyhedron were bisected and those points, along with the initial 12, were projected onto a unit sphere to define the vertices of an octacontagon. This procedure was repeated for six generations, yielding a set of 40 962 vertices, which were then translated into geographic coordinates to act as nearly equidistant cell centers, with approximately equal areas of ∼12 000 km2 (Figure S6, SI). The term “geodesic” here is used colloquially and is not intended to imply that the grid is based on an exacting measure of Earth’s shape. A short discussion about the geodesic grid appears in the SI. Characterizing Airsheds. Airsheds for different places or different times can be compared by a number of metrics. Here airsheds are compared by area, the areal extent in two dimensions, ignoring height; roundness, how closely the two-dimensional airshed resembles a circle, or a set of concentric circles, with the arrival site as the center; and seasonal or interannual differences for individual sites, the cell-by-cell change in contribution from one season or year to another. Area. As the cells of the geodesic grid are of roughly equal area, the total area covered by an airshed can be approximated simply as the product of the count of cells for that airshed and the area of an individual cell. Cell count then suffices to compare airsheds against one another. Airshed data for this work are stored in comma separated value files for which a strong linear relationship exists between file size and number of data (Figure S7, SI), provided each airshed file stores values with the same data type. File size was thus used as a surrogate for area in the comparisons below. Roundness. The roundness of each airshed was determined by first calculating a departure from roundness. Taking the cell in which the arrival point lies as the center, the first unroundness term is the sum of the absolute differences in the number of end points, p, that fall within each of the cells adjacent to the center, 8851
dx.doi.org/10.1021/es202490b |Environ. Sci. Technol. 2011, 45, 8850–8857
Environmental Science & Technology
ARTICLE
divided by the number of cells in that term, i.e., six for the first term. The second unroundness term considers the 12 cells immediately around the six cells adjacent to the center. Similarly, the kth unroundness term, Uk, would be x
Uk ¼
x
∑fn∑¼ 2 jp1 pnj, m∑¼ 3 jp2 pm j:::jpx1 px jg x
ð1Þ
where x = 6k. Then the roundness, R, is the inverse of the total unroundness: R ¼
1 Uk
∑k
ð2Þ
Values of R thus calculated are on the order of 0.01, so roundness values are presented as 100R. Commonly, when airsheds are combined with other spatially resolved information or when airsheds of different temporal lengths are compared, the number of end points in each cell is divided by the total number of end points in the airshed to give a fractional “density”. For the purposes of comparing “roundness”, the raw number of BT segment end points was used to distinguish between airsheds representing periods of different lengths. Seasonal and Interannual Differences. For several locations two and sometimes more years of BTs were available. Seasonal and interannual differences in airsheds were determined as the sum of the absolute changes in the number of BT end points in each cell from one season or year to another. Ideally, the number of end points composing the airshed is identical for each year, leap years excepted. For ease of interpretation, these differences are expressed as the ratio of changes to the total number of end points in two 365-day three arrival-height airsheds and presented as a percentage; thus, an interannual difference of 100% means that every cell with a nonzero number of BTs in the first year would have no BT end points in the second year, and vice versa. Note that, due to the change in meteorological data resolution and the removal of 18 BTs with erroneous data, there is an overestimation of the interannual difference of ∼1% for some airsheds.
’ DISCUSSION Ascertaining Significance. In an airshed constructed from BTs, cells can be compared to one another using the fraction of end points that fall within them, when normalized by dividing by the total number of cells in the airshed, and this is often termed “density”; that fraction can also be considered to represent the contribution of that cell to the airshed. For instance, from 2004 to 2007 the cell immediately north of the cell in which Izmir, Turkey, lies provides 5% of the airshed of Izmir (Figure S8, SI). An assumption of air mass parcel integrity is built into such an interpretation. When the extreme case of an airshed composed from a single 5-day BT is considered, each of the 20 end points is assigned a 5% contribution to the airshed, and the concept becomes absurd. This can be mitigated by weighting each end point by a function of the inverse of its temporal distance from the arrival point, but a justifiable weighting system must then be devised. However, when a sufficient number of BTs are used to construct an airshed, the effect of the decrease in area with proximity to the arrival cell plays a role. In the case where a geodesic grid is used, this can be viewed as a decreasing number of possible paths as an air-mass approaches the arrival cell. An air-mass
traveling directly to the arrival cell at ∼5 m/s over 5 days would start in one of the 120 cells that are 20 cells distant. After 6 h that air-mass would have passed into one of 114 cells that are 19 cells distant from the arrival cell. By the last step prior to arrival, that air-mass must pass through one of the only six cells directly adjacent to the arrival cell. The result is that cells closer to the arrival cell contain more end points, and weighting by distance becomes somewhat redundant. The center cell—the cell in which the location lies for which BTs are calculated—is a special case. Although all BTs end in that cell, only end points backward in time are employed in creating the airshed. Thus, the center cell only has end points from airmasses that are sufficiently slow to stay within the cell over the 6 h of the first time step or from air-masses that double back on their own paths. The aggregation of many BTs to form an airshed partially alleviates the difficulty of dealing with the uncertainty in the position of the air-mass, which increases with the temporal distance to arrival; Stohl suggested that in general the uncertainty in the position of an air-mass BT was 20% of its distance from the arrival point (or the departure point in the case of forward trajectories).31 Assuming that these errors are random, any true patterns in the airsheds will still be evident against the noise, provided a sufficient number of trajectories are used in the construction of the airshed. In the special case of an airshed in which the air-masses followed the same path constantly, simple statistics could dictate that number. For any realistic airshed, the standard statistical methods become unwieldy, but the geodesic grid and roundness can be used to suggest an appropriate number of trajectories. If the purpose of generating an airshed is to elucidate potential source regions of contaminants, then only those cells with numbers of end points that are significantly greater than the other cells hold interest, that is, unless there is some directionality to the airshed, then air arriving from all directions plays an equal role in the airshed and no individual source regions can be identified. If directionality is defined as a departure from roundness, the number of BTs used to construct the airshed must be sufficient that any departure from roundness can be attributed to actual directionality and not a dearth of BTs. For 5 day BTs with a 6 h time step, the greatest distances from the arrival point, and thus the greatest errors, ought to belong to the 20th time step. The 95th percentile of all 20th time step distances for the BTs available is ∼4.3 Mm (Figure S9, SI), which corresponds to 42 cells distant from the arrival cell. This is used to define the outer limit of a theoretical perfectly round airshed that is composed of 5420 cells. Then 5420 is the minimum number of end points required to construct that perfectly round airshed, and any departure from roundness evident in an airshed constructed from 5420 end points can be considered significant. Thus, using 5-day BTs with a 6 h time step, no less than 10 weeks of BTs are required for generating an airshed. This is on par with the minimum deployment time for passive air samplers.9 When assigning potential contaminant source regions to air samples, it is also desirable to have a test for the significance of the airshed density for individual cells. Again, assuming the interest is in a departure from the theoretical perfectly round airshed, a threshold t can be calculated using ! n ð3Þ t ¼m q 8852
dx.doi.org/10.1021/es202490b |Environ. Sci. Technol. 2011, 45, 8850–8857
Environmental Science & Technology
ARTICLE
where n is the total number of end points in the airshed, q is the number of end points required to generate the theoretical perfectly round airshed with the same number of time steps (see Table 1), and m is a factor determining how large the departure from perfectly round must be. One can scale m so that as the number of end points becomes sufficiently greater than the minimum it goes to unity ! 1 n ð4Þ m ¼ erfc 2π þ 1 2 q or if the interest is only in determining if a cell has a number of end points that is much greater than predicted by perfect roundness, one can simplify and use m¼2
ð5Þ
Any cell that has a number of end points greater than or equal to t can be considered significant, with the constraint that n > q. Area. The file sizes representing the relative area of each complete airshed as well as the individual airsheds from each arrival height can be found in Table S1 of the SI. The area covered by each airshed is a function of two things: the speeds of the airmasses composing the airshed and the roundness. Air-masses traveling at greater rates cover more distance in the 5 days the BTs are allowed to propagate and thus have end points in relatively more cells. A very linear airshed, one for which the air-masses arrive from the same direction much of the time, will have higher end point counts in a small number of cells and thus a smaller area than a rounder airshed with comparable air-mass speeds. Table 1. Recommended Values for the Minimum Number of Trajectory End Points (q) To Ensure the Significance of an Airshed Based on the Temporal Length of the Trajectories with a 6-h Resolution end points per back trajectory
q
1
36
4 (1 day)
275
5
370
10
1330
12 (3 days) 15
1520 2880
20 (5 days)
5420
Airsheds for four relatively close sites in Ontario, Canada, that are at least one cell distant from each other were also compared. The spatial extents of these airsheds were very similar to each other, differing only at the edges, and then only for a few cells (e.g., Figure S10, SI). However, large differences were evident in those cells immediately adjacent to the arrival cell, indicating that airsheds for areas in relatively close proximity can be quite different and highlighting that the cells closer to the arrival cell have greater contributions to an airshed without explicitly weighting cells by distance. Roundness. The roundness of each airshed was calculated to probe any patterns in space and time. Roundness tends to increase as a function of the number of BT end points included in constructing an airshed. Thus, only airsheds spanning very similar temporal ranges, having an equal number of vertical starting points, and having equal temporal resolutions ought to be compared for roundness. The BTs available for the Devon Ice Cap site were for 10 days backward in time. Compared to the airsheds composed from 5 day BTs, airsheds from these gave higher-than-average but not extreme values for roundness, interannual, and interseasonal variation after normalization to the total number of end points. Nonetheless, the Devon Ice Cap airsheds are excluded from the following discussion of roundness. BTs were not calculated for all sites for the same times, and the following analyses are limited by the BTs available (Table S1, SI). Roundness was determined both seasonally and annually, and seasonal airsheds were considered both individually and averaged over multiple years, where data were available. Single Year Roundness. When all arrival heights are considered together, Pallas, Finland, has the roundest single year airshed, with a roundness of 10.1 (Figure 1a). This site is located well inland in moderately flat terrain, and far enough north that it sits within the polar circulation cell nearly all the time. At the opposite extreme is Arica, Chile: the roundness of its airshed is 1.53. This low-lying port city is at the foot of the Andes Mountains, and the barrier-jet winds it experiences are evident in its airshed—nearly all of the BT end points making up this airshed lie to the south and southeast of the arrival point in a tight bundle, while the westerlies are also evident in the large number of low end point count cells to the west (Figure 1b). Although an extreme case, this strongly linear airshed demonstrates that a sampler’s location is insufficient to determine what air it represents: Arica is 10 km from Peru and 100 km from Bolivia, but air measured there is almost exclusively of Chilean origin.
Figure 1. The year-long airshed for Pallas, Finland, is the most round airshed in this work (a). The airshed for Arica, Chile, is the least round (b). The values displayed are the absolute counts of back-trajectory end points for each cell. The hexagons representing the airshed cells are made slightly transparent to allow the landmasses to be seen. 8853
dx.doi.org/10.1021/es202490b |Environ. Sci. Technol. 2011, 45, 8850–8857
Environmental Science & Technology
ARTICLE
Figure 2. Airshed difference maps for Darwin, Australia, for 2005 and 2006 (a) and the Devon Ice Cap site for 1998 and 2003 (b). These represent the greatest interannual differences in this work. Negative values indicate a greater number of end points in a cell in the earlier year and positive values a greater number in the later year. In both instances, the greatest differences occur in the cells nearest the arrival point.
Roundness, Latitude, and Terrain. If roundness were indeed dictated in part by the large-scale winds—the belts of westerlies, the trade winds, etc.—roundness ought to vary as a function of latitude. Indeed, when fit with a fourth-order polynomial, a rough pattern is evident, with low roundness at the equator increasing to maxima in both directions and decreasing again (Figure S11,SI, r2 = 0.51). When the roundness of each airshed is regressed against the local slope calculated from a 5 deg-min elevation model (Figure S12, SI), a weak but significant negative correlation exists (r2 = 0.13, p < 0.05). On the basis of this, those sites whose slopes were above 0.2 were removed from the set yielding a moderately improved fit of roundness and latitude (r2 = 0.65). Nonetheless, latitude on its own is not a good predictor of roundness, even with slopes taken into account. A second order polynomial also fit the revised set, but two sites, Coyhaique, near the southern tip of Chile, and Barrow, Alaska, stand out against the pattern. When roundness as a function of latitude is compared with the multiyear average surface winds for the globe, Coyhaique sits within a band of strong winds circulating around Antarctica, which may explain the low roundness of its airshed (Figure S13, SI). The low roundness of Barrow’s airshed is not as simply explained. Largescale winds cannot be estimated by latitude without accounting for the surfaces over which the air is passing: the continents and oceans are easily discernible on the basis of windspeed alone in a wind field image (Figure S13, SI). Area and Roundness. As mentioned above, the area of an airshed is partly a function of its roundness. For the airsheds considered here, the two are weakly but significantly correlated (r2 = 0.42, p < 0.01, Figure S14, SI). A similar pattern relating roundness to latitude is observed when relating area to latitude. However, the similarity ends when area is compared to slope, indicating that area alone does not contain the same amount of information as the calculated roundness. Interannual and Seasonal Differences in Airsheds. The interannual differences between airsheds were normally distributed with a mean value of 16% (Figure S15, SI). The airshed for Darwin on the north coast of Australia has the largest calculated interannual difference with a change of 21% from 2005 to 2006 (Figure 2a). The bulk of the difference occurs close by the sampling site, where the number of end points is highest. It
appears that many more air masses approached Darwin from the north in 2005 and that wind speeds were low, as suggested by very high end point counts in cells adjacent to the site. In 2006, the winds were more often from the south and were faster, covering a much larger area. The airshed for Kuwait City, Kuwait, had the lowest interannual change at 9% (Figure S16, SI). As with Darwin, the largest differences are seen close to the location of interest, but the differences in end point counts on a cell-by-cell basis are much smaller. The Darwin airsheds are less round than the Kuwait City airsheds, but interannual differences and roundness are not correlated (slope = 2 105, r2 = 2 106). Seasonal differences, as a fraction of total end points, are larger than interannual differences. The average difference between all four seasons for each site ranged from 22% for Bahia Blanca, Argentina, to 38% for Arica, Chile. Individual seasonal differences ranged from 17% for Kuwait City from winter to spring to 53% between winter and summer for the Danum Valley site in Malaysia. Interannual differences are not correlated to seasonal differences: neither is a good predictor of the other. However, with only 2 years of data it is unlikely that these differences represent the overall variability in the airsheds. This seasonal variability and the problem that it cannot easily be predicted pose a difficulty for interpreting PAS data collected over only one or two seasons (e.g., refs 3234). If the airshed over the whole year differs significantly from the single season in which a measurement is taken, that measurement may not wellrepresent the average annual air concentrations in the area of deployment, complicating any comparison between sites or studies. Fourteen consecutive years of BTs were available for the Devon Ice Cap site, allowing a closer investigation of variability for this one location. For the following comparisons only the first 20 time steps of these 40 time step airsheds were used. The differences between consecutive years ranged from 14% to 24%, the largest being between 1999 and 2000 (Figure S17a,b, SI). The greatest difference between any two annual airsheds was between 1998 and 2003, which were 29% different (Figure S17c, d, SI), suggesting that airsheds that are closer temporally may be more similar than those that are further apart. In fact, differences between yearly airsheds tended to increase with time for Devon (Figure S18a, SI, r2 = 0.69, p < 0.01) when compared to 1995, but 8854
dx.doi.org/10.1021/es202490b |Environ. Sci. Technol. 2011, 45, 8850–8857
Environmental Science & Technology
ARTICLE
Figure 3. Airshed maps for New Delhi, India, with all departure heights (a), with the departure heights above 700 m above ground level removed (b), and with only those cells passing the significance criterion t (c). The values displayed are the absolute counts of back-trajectory end points for each cell.
the trend was less strong when compared to 2008’s airshed (Figure S18b, SI, r2 = 0.41, p < 0.05). With the limited data it is not possible to determine if this represents a truly changing airshed, part of a cyclic airshed change, or simply a coincidence. Figure 2b highlights the differences between the 1998 and 2003 airsheds, suggesting a much greater contribution to the sample from the (relatively) populous west coast of Greenland in 1998 than in 2003. It appears that interpretations of long-term trends in atmospheric concentrations of SVOCs that are inherently based on an assumption of long-term stability of air mass origin (e.g., refs 3, 35) ought to include an analysis of the airsheds from the actual deployment time of the samplers to ensure the validity of that assumption. A Constraint on Departure Height. Figure S19 (SI) shows box-and-whisker plots of the heights for each of 20 6-h time steps prior to arrival for the 14 years of trajectories available for the Devon Ice Cap. Because BTs are constrained from passing below 0 m, the 5th percentile, 25th percentile, and the minima of elevations are quite similar, and the median values are consistently lower than the means. An increasing spread of heights with temporal distance from arrival at the center point is seen in all unconstrained values, illustrating the increasing number of possible routes, and reflecting the increasing uncertainty in the estimated position of the air-mass with temporal distance from arrival (Figure S19, SI). This increasing vertical diffuseness is an important consideration when assigning potential contaminant source regions to a site where air sampling has been conducted. It has been common to ignore the vertical component when employing airsheds for this: as stated above, the fraction of end points falling within a cell is considered as the fractional contribution of that cell to the air being sampled. Arguably, however, air masses that are well above the surface have themselves already undergone long-range atmospheric transport, and any contaminants in them are not emissions from within that cell. Thus, when determining potential source regions for a sample, BT end points above some maximum height ought to be excluded from an analysis of contaminant sources. Forward air-mass trajectories that were calculated for other work28 were employed to determine such an appropriate maximum height. When binned together forward trajectories represent not an airshed but rather a footprint area, that is, the area affected by emissions in the starting cell. These trajectory end points were first divided by time step then binned into footprint areas using the geodesic grid cells. As few trajectory end points fall within the cell containing the departure site, a “departure area” was defined as the area covered by the departure cell and the six cells adjacent to it. The fraction of end points that fell
within the departure area for each time step was then calculated (Figure S20a, SI). For the two locations for which forward trajectories were available, Toronto and Ottawa, Canada, only ∼8% of end points remained within the departure area by the fourth time step. This includes any air-masses that “double back” after leaving the area. The source condition cutoff height was then estimated as the 95th percentile of end point heights for the fourth time step, which is 737 and 627 m for the two sites, respectively (Figure S20b, SI), corresponding well with the 600750 m normally associated with the daytime boundary layer thickness. As a rough approximation, 700 m is a reasonable cutoff when using airsheds to assign contaminant source regions. That is not to suggest that BTs should be terminated if they exceed this height but that after combining into an airshed, the end points can be filtered by height. The airsheds for New Delhi, India, with and without a 700 m maximum height are noticeably different (Figure 3). The difference between the two for individual cells is no more than 0.8% of the contribution to the airshed, and the total difference between the two airsheds is only 19%, which is near the mean seasonal or interannual difference for all the airsheds considered here. However, the areal extents are quite different and a large area to the northwest has no air departing from below 700 m, but it appears in the untrimmed airshed. This suggests that any contaminants arriving with the air from this direction had already undergone long-range transport from another location. Implications for the Design and Interpretation of Passive Air Sampling Campaigns. The methods described herein may be employed to aid in the selection of appropriate locations for PASs. Multiple years of BTs ought to be generated and integrated into both yearly and seasonal airsheds to determine the medium term stability of the airshed being studied and to maximize the chances of sampling air from the region of interest. If the goal of a campaign is to link sources and receptors, these airsheds should have end points that are greater than 700 m above the Earth’s surface removed. As airsheds are most different near the point of arrival, campaigns that use transects to explore the effect of distance from a potential source should generate airsheds for each sampler location, provided the samplers are more distant than the grid size of the trajectory model used. These are also superior to a single forward airshed generated for the source. For the purpose of interpreting measurements, airsheds should be constructed for the actual deployment time of the sampler. Determining area and roundness can serve to classify samples: samplers in small linear airsheds represent somewhat more local air, while samplers in large round airsheds represent regional concentrations. If assigning potential sources of contaminants trajectory end points greater than 700 m above ground 8855
dx.doi.org/10.1021/es202490b |Environ. Sci. Technol. 2011, 45, 8850–8857
Environmental Science & Technology should be discarded, as should those failing to meet a criterion for significance. Figure 3c shows the result of applying both the height and significance filters to the airshed of New Delhi, India: in addition to the areas eliminated by the height restriction, contributions to the airshed from Somalia, Sri Lanka, and the southern part of India are found to be insignificant.
’ ASSOCIATED CONTENT
bS
Supporting Information. Figures and tables mentioned above, as well as brief discussions of the errors introduced by the geodesic grid, the roundness of airsheds built from single arrival heights and seasons, the effect of choosing greater arrival heights for BTs, and the choice of temporal lengths of BTs. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Telephone: +1-416-287-7225; e-mail:
[email protected].
’ ACKNOWLEDGMENT The authors wish to thank the Air Quality Models and Applications Section of the Weather and Environmental Prediction Services Branch of Environment Canada for generating the thousands of BTs and acknowledge funding from the Ontario Graduate Scholarship program and the Natural Science and Engineering Research Council of Canada. ’ REFERENCES (1) United Nations Environment Programme. UNEP Preparation of an International Legally Binding Instrument Implementing International Action on Certain Persistent Organic Pollutants; 1998. (2) United Nations Economic Commission for Europe. UN ECE Protocol on Persistent Organic Pollutants under the 1979 Convention on Long-Range Transboundary Air Pollution; 1998. (3) Shunthirasingham, C.; Oyiliagu, C. E.; Cao, X. S.; Gouin, T.; Wania, F.; Lee, S. C.; Pozo, K.; Harner, T.; Muir, D. C. G. Spatial and temporal pattern of pesticides in the global atmosphere. J. Environ. Monit. 2010, 12, 1650–1657. (4) Roots, O.; Roose, A.; Kull, A.; Holoubek, I.; Cupr, P.; Klanova, J. Distribution pattern of PCBs, HCB and PeCB using passive air and soil sampling in Estonia. Environ. Sci. Pollut. Res. 2010, 17, 740–749. (5) Shunthirasingham, C.; Barra, R.; Mendoza, G.; Montory, M.; Oyiliagu, C. E.; Lei, Y. D.; Wania, F. Spatial variability of atmospheric semivolatile organic compounds in Chile. Atmos. Environ. 2011, 45, 303–309. (6) Baek, S. Y.; Choi, S. D.; Park, H.; Kang, J. H.; Chang, Y. S. Spatial and seasonal distribution of polychlorinated biphenyls (PCBs) in the vicinity of an iron and steel making plant. Environ. Sci. Technol. 2010, 44, 3035–3040. (7) Liu, W. J.; Chen, D. Z.; Liu, X. D.; Zheng, X. Y.; Yang, W.; Westgate, J. N.; Wania, F. Transport of semivolatile organic compounds to the Tibetan Plateau: Spatial and temporal variation in air concentrations in mountainous Western Sichuan, China. Environ. Sci. Technol. 2010, 44, 1559–1565. (8) Shunthirasingham, C.; Mmereki, B. T.; Masamba, W.; Oyiliagu, C. E.; Lei, Y. D.; Wania, F. Fate of pesticides in the arid subtropics, Botswana, Southern Africa. Environ. Sci. Technol. 2010, 44, 8082–8088. (9) Hayward, S. J.; Gouin, T.; Wania, F. Comparison of four active and passive sampling techniques for pesticides in air. Environ. Sci. Technol. 2010, 44, 3410–3416.
ARTICLE
(10) Oehme, M. Further evidence for long-range air transport of polychlorinated aromates and pesticides—North-America and Eurasia to the Arctic. Ambio 1991, 20, 293–297. (11) Haugen, J. E.; Wania, F.; Ritter, N.; Schlabach, M. Hexachlorocyclohexanes in air in southern Norway. Temporal variation, source allocation, and temperature dependence. Environ. Sci. Technol. 1998, 32, 217–224. (12) Hoh, E.; Hites, R. A. Sources of toxaphene and other organochlorine pesticides in North America as determined by air measurements and potential source contribution function analyses. Environ. Sci. Technol. 2004, 38, 4187–4194. (13) Hsu, Y. K.; Holsen, T. M.; Hopke, P. K. Comparison of hybrid receptor models to locate PCB sources in Chicago. Atmos. Environ. 2003, 37, 545–562. (14) Hageman, K. J.; Hafner, W. D.; Campbell, D. H.; Jaffe, D. A.; Landers, D. H.; Simonich, S. L. M. Variability in pesticide deposition and source contributions to snowpack in western US National Parks. Environ. Sci. Technol. 2010, 44, 4452–4458. (15) Gouin, T.; Harner, T.; Blanchard, P.; Mackay, D. Passive and active air samplers as complementary methods for investigating persistent organic pollutants in the Great Lakes basin. Environ. Sci. Technol. 2005, 39, 9115–9122. (16) Daly, G. L.; Lei, Y. D.; Teixeira, C.; Muir, D. C. G.; Castillo, L. E.; Jantunen, L. M. M.; Wania, F. Organochlorine pesticides in the soils and atmosphere of Costa Rica. Environ. Sci. Technol. 2007, 41, 1124–1130. (17) Miller, J. M. A 5-year climatology of 5-day back trajectories from Barrow, Alaska. Atmos. Environ. 1981, 15, 1401–1405. (18) Hafner, W. D.; Solorzano, N. N.; Jaffe, D. A. Analysis of rainfall and fine aerosol data using clustered trajectory analysis for National Park sites in the western US. Atmos. Environ. 2007, 41, 3071–3081. (19) Munn, R. E.; Likens, G. E.; Weisman, B.; Hornbeck, J. W.; Martin, C. W.; Bormann, F. H. A meteorological analysis of the precipitation chemistry event samples at Hubbard-Brook (NH). Atmos. Environ. 1984, 18, 2775–2779. (20) Uliasz, M. Lagrangian particle dispersion modeling in mesoscale applications. In Environmental Modeling; Zannetti, P., Ed.; Computational Mechanics Inc., Billerica, MA, 1994; pp 71102. (21) Gouin, T.; Shoeib, M.; Harner, T. Atmospheric concentrations of current-use pesticides across south-central Ontario using monthlyresolved passive air samplers. Atmos. Environ. 2008, 42, 8096–8104. (22) Stohl, A.; Forster, C.; Eckhardt, S.; Spichtinger, N.; Huntrieser, H.; Heland, J.; Schlager, H.; Wilhelm, S.; Arnold, F.; Cooper, O. A backward modeling study of intercontinental pollution transport using aircraft measurements. J. Geophys. Res.-Atmos. 2003, 108, 18. (23) Stohl, A. Characteristics of atmospheric transport into the Arctic troposphere. J. Geophys. Res.-Atmos. 2006, 111, 17. (24) Seibert, P.; Frank, A. Sourcereceptor matrix calculation with a Lagrangian particle dispersion model in backward mode. Atmos. Chem. Phys. 2004, 4, 51–63. (25) Halse, A. K.; Schlabach, M.; Eckhardt, S.; Sweetman, A.; Jones, K. C.; Breivik, K. Spatial variability of POPs in European background air. Atmos. Chem. Phys. 2011, 11, 1549–1564. (26) Westgate, J. N.; Shunthirasingham, C.; Oyiliagu, C. E.; von Waldow, H.; Wania, F. Three methods for quantifying proximity of air sampling sites to spatially resolved emissions of semi-volatile organic contaminants. Atmos. Environ. 2010, 44, 4380–4387. (27) Atmospheric transport models for environmental emergencies; Government of Canada Canadian Meteorological Centre: Dorval, QC, 2001; http://collaboration.cmc.ec.gc.ca/cmc/CMOI/product_ guide/docs/lib/model-eco_urgences_e.pdf. (28) Barthel, P. Einfluss der emission von straßenverkehr auf die deposition von organischen schadstoffen und metallen in kanadischen hochmooren. Masters Dissertation, Lehrstuhl f€ur Hydrologie, Universit€at Bayreuth, 2011. (29) Meyer, T., Muir, D. C. G., Teixeira, C., Wang, X., Young, T., Wania, F. Deposition of brominated flame retardants to the Devon Ice Cap, Nunavut, Canada. Environ. Sci. Technol. Submitted. 8856
dx.doi.org/10.1021/es202490b |Environ. Sci. Technol. 2011, 45, 8850–8857
Environmental Science & Technology
ARTICLE
(30) Sofowote, U. M.; Hung, H. L.; Rastogi, A. K.; Westgate, J. N.; Su, Y. S.; Sverko, E.; D’Sa, I.; Roach, P.; Fellin, P.; McCarry, B. E. The gas/particle partitioning of polycyclic aromatic hydrocarbons collected at a sub-Arctic site in Canada. Atmos. Environ. 2010, 44, 4919–4926. (31) Stohl, A. Computation, accuracy and applications of trajectories— A review and bibliography. Atmos. Environ. 1998, 32, 947–966. (32) Jaward, T. M.; Zhang, G.; Nam, J. J.; Sweetman, A. J.; Obbard, J. P.; Kobara, Y.; Jones, K. C. Passive air sampling of polychlorinated biphenyls, organochlorine compounds, and polybrominated diphenyl ethers across Asia. Environ. Sci. Technol. 2005, 39, 8638–8645. (33) Zhang, G.; Chakraborty, P.; Li, J.; Sampathkumar, P.; Balasubramanian, T.; Kathiresan, K.; Takahashi, S.; Subramanian, A.; Tanabe, S.; Jones, K. C. Passive atmospheric sampling of organochlorine pesticides, polychlorinated biphenyls, and polybrominated diphenyl ethers in urban, rural, and wetland sites along the coastal length of India. Environ. Sci. Technol. 2008, 42, 8218–8223. (34) Zhang, Z.; Liu, L. Y.; Li, Y. F.; Wang, D. G.; Jia, H. L.; Harner, T.; Sverko, E.; Wan, X. N.; Xu, D. D.; Ren, N. Q.; Ma, J. M.; Pozo, K. Analysis of polychlorinated biphenyls in concurrently sampled Chinese air and surface soil. Environ. Sci. Technol. 2008, 42, 6514–6518. (35) Lammel, G.; Dobrovolny, P.; Dvorska, A.; Chroma, K.; Brazdil, R.; Holoubek, I.; Hosek, J. Monitoring of persistent organic pollutants in Africa. Part 2: Design of a network to monitor the continental and intercontinental background. J. Environ. Monit. 2009, 11, 1964–1972.
8857
dx.doi.org/10.1021/es202490b |Environ. Sci. Technol. 2011, 45, 8850–8857
ARTICLE pubs.acs.org/est
Universally Applicable Model for the Quantitative Determination of Lake Sediment Composition Using Fourier Transform Infrared Spectroscopy Peter Rosen,*,† Hendrik Vogel,†,‡ Laura Cunningham,§ Annette Hahn,|| Sonja Hausmann,^ Reinhard Pienitz,# Bernd Zolitschka,|| Bernd Wagner,‡ and Per Persson$ †
)
Climate Impacts Research Centre (CIRC), Umea University, SE-981 07 Abisko, Sweden ‡ Institute of Geology and Mineralogy, University of Cologne, Cologne, Germany § School of Geography and Earth Sciences, University of St. Andrews, Scotland Institute of Geography, University of Bremen, Bremen, Germany ^ Department of Geosciences, University of Arkansas, Fayetteville, Arkansas, United States # tudes Nordiques (CEN), Universite Laval Quebec, Quebec, Canada, Aquatic Paleoecology Laboratory, Centre d’E $ Department of Chemistry, Umea University, Umea, Sweden ABSTRACT: Fourier transform infrared spectroscopy (FTIRS) can provide detailed information on organic and minerogenic constituents of sediment records. Based on a large number of sediment samples of varying age (0340 000 yrs) and from very diverse lake settings in Antarctica, Argentina, Canada, Macedonia/Albania, Siberia, and Sweden, we have developed universally applicable calibration models for the quantitative determination of biogenic silica (BSi; n = 816), total inorganic carbon (TIC; n = 879), and total organic carbon (TOC; n = 3164) using FTIRS. These models are based on the differential absorbance of infrared radiation at specific wavelengths with varying concentrations of individual parameters, due to molecular vibrations associated with each parameter. The calibration models have low prediction errors and the predicted values are highly correlated with conventionally measured values (R = 0.940.99). Robustness tests indicate the accuracy of the newly developed FTIRS calibration models is similar to that of conventional geochemical analyses. Consequently FTIRS offers a useful and rapid alternative to conventional analyses for the quantitative determination of BSi, TIC, and TOC. The rapidity, cost-effectiveness, and small sample size required enables FTIRS determination of geochemical properties to be undertaken at higher resolutions than would otherwise be possible with the same resource allocation, thus providing crucial sedimentological information for climatic and environmental reconstructions.
’ INTRODUCTION Multiproxy approaches are often required for comprehensive paleolimnological reconstructions as information from an individual proxy indicator can be supported by a suite of other indicators thus reducing potential errors or ambiguities. Analyzing a suite of indicators is both time-consuming and expensive, particularly when assessing long sediment records.1 In high-resolution studies, the amount of sample material available from each horizon can also be a limiting factor. Conventional measurements of biogeochemical constituents (including biogenic silica (BSi)) are laborious, time-consuming, and imprecise,2 thus alternative approaches are required. X-ray fluorescence scanners and multisensor core loggers can provide highly resolved, qualitative and semiquantitative information on the inorganic geochemistry, mineralogy, and magnetic properties. Fourier transform infrared spectroscopy (FTIRS) analysis also offers a promising alternative to conventional techniques due to the wealth of information on minerogenic r 2011 American Chemical Society
and organic sediment constituents contained in FTIR spectra, the small sample size required, and the relative speed of analysis.38 The basic principles of FTIRS are that infrared radiation can excite molecular vibrations and, as a consequence of the quantum mechanical behavior, the radiation will be absorbed at specific energies, depending on the composition of the material examined. As most compounds display characteristic infrared spectra, changes in the organic and inorganic composition of sediments can be determined, including changes in carbohydrate, fatty acid, humic material, silicate, and carbonate concentrations.35 This information is of particular interest for paleolimnological studies since sediment is commonly composed of a mixture of various Received: January 18, 2011 Accepted: September 1, 2011 Revised: August 22, 2011 Published: September 01, 2011 8858
dx.doi.org/10.1021/es200203z | Environ. Sci. Technol. 2011, 45, 8858–8865
ARTICLE
5 0.25
0300
Table 2. Number of Surface Sediment Samples and DownCore Samples Analyzed for Each Parameter, Including the Specific Sedimentary Sequences from Which the Down-Core Samples Were Collected
19030 E 351 2 permafrostmire/ subarctic dolomite/sandstone
Sweden 68210 N
Inre
Harrsj€on
Environmental Science & Technology
sample
010 000
15 1.2
granite
010 000
13.7 3.2
granite/syenite
879
3064
35
122
122
193
152
91 725
Ohrid
200
El’gygytgyn Inre Harrsj€on
287 21
010 000 010 000 0130 000 050 000
162 222
504
504
99
1793 99
51 50 62
organic and minerogenic compounds originating from aquatic organisms, chemical in-lake processes, and the erosion of soils in the catchment area of a lake. Absorbance peaks in FTIR spectra relate directly to the concentrations of specific components within the sediment, thus many organic and minerogenic constituents can be quantified using FTIRS. Two of the great advantages of FTIRS, compared to several other methods, are the small sample sizes (10 mg of dried sediment) and negligible pretreatment required. The potential, and possible applications, of FTIRS in paleolimnological work have previously been discussed.68 These studies have shown that FTIRS quantifies selected constituents using either site-specific or regional calibration models.7,8 It remains uncertain, however, if more universally applicable FTIRS calibration models can be developed and ubiquitously applied to lakes from distinctly different settings. This study combines a large number of samples from lakes of varying sizes and settings distributed worldwide (Antarctica, Argentina, Canada, Macedonia/Albania, Siberia, Sweden) to assess if a universally applicable FTIRS calibration model can be developed for the rapid and simultaneous quantification of total organic carbon (TOC), total inorganic carbon (TIC), and BSi. Such calibration models would significantly reduce the amount of work required for paleoclimatic and paleoenvironmental reconstructions, as calibration to specific localities8 will be less important.
020
280 10.8 until 101 106
175 16.5 ultramafics sediments/
100
36
Makkasj€on
0340 000
6.1 1.8 2.3 3.5 <16 surface
mainly granite, gneiss
035 000 core length (m) approx.
012 000
246 8.5
34 2.3
’ MATERIAL AND METHODS
max. depth (m) sediment
volcanic rocks/ molasse-type granites/ metasediments
TOC
30
Vuolep Njakajaure
ages (years)
20350 E 415 3 boreal forest
18450 E 408 13 subalpine birch forest hard shales/ dolomite 17310 E 670 11 alpine
18450 E 400 0.4 subalpine birch forest hard shales/ dolomite 23520 17480 E 1701180 <20 boreal to alpine
20430 E 693 35 800 mediterranean, anthropogenic limestone/ metasediments/ 70 220 W 100 774 cold-semi desert 162530 E 73 18 polar desert
73410 W 494 668 arctic tundra/ permafrost precambrian granitic-gneissic longitude altitude (m a.s.l.) lake area (ha) catchment vegetation bedrock
Laguna Potrok Aike
Seukokjaure
172050 E 492 29 300 arctic tundra/ permafrost volcanic rocks
Sweden 66430 N, Sweden 68200 N Sweden 67460 N Sweden 68200 N Sweden 67070 68480 N Siberia 67300 N
Lake Hoare
TIC
Badsj€on
Macedonia 41010 N
Vuolep
total number of downcore samples
Pingualuit
Argentina 51 580 S Antarctic 77380 S Canada 61170 N country/region latitude
100-lake cal. set El’gygytgyn Ohrid Potrok Aike Hoare Pingualuit variable
Table 1. Location and Environmental Characteristics of the Lakes Included in the FTIRS Models
Badsj€on
Seukok-jaure
Njakajaure
Makkasj€on
number of Swedish lakes with surface samples
BSi
Site-Related Information. The major characteristics of the study lakes are summarized in Table 1. Lake sizes range from 1 to 35 800 ha, with maximum water depth of 1.5280 m, and sediment ages of 0340 000 yrs BP. Further site specific information has previously been published for Lake Hoare,9 Laguna Potrok Aike,10,11 Pingualuit,1 Lake Ohrid,12 El’gygytgyn,13 Seukokjaure,6 Inre Harrsj€on,14 and the surface samples from northern Sweden.6 Analytical Methods. The number of samples included from each sampling location varied by parameter (Table 2). Furthermore, the exact method employed for the conventional geochemical analyses of TIC and TOC also varied among the sampling programs, although all were recognized “Standard Methods”. For example, a CNS elemental analyzer (EuroEA, Eurovector) was used for TOC measurements of samples from Pingualuit. In contrast, the TOC content of samples from Lake Hoare was determined with a 8859
dx.doi.org/10.1021/es200203z |Environ. Sci. Technol. 2011, 45, 8858–8865
Environmental Science & Technology Metalyt CS 1000S (ELTRA Corp.) analyzer, after pretreatment with 10% HCl at 80 C to remove carbonates, while the total carbon (TC) content was measured with an Elementar III (VARIO Corp.) analyzer. The TIC content was calculated by the difference between TC and TOC. Previous publications provided more information about analytical methods used (see above). At all locations, the leaching technique was used for the BSi measurements.15 Prior to FTIRS analysis, samples were freeze-dried and ground. Samples from Sweden were ground by hand using an agate mortar and pestle. All the remaining samples were ground to <63 μm using either a planetary mill or a swing mill. Samples were then diluted and mixed with oven-dried (80 C) potassium bromide (KBr; spectrograde, Fisher Scientific). The mass ratio between sediment and KBr (2%) was chosen to avoid very high absorbance (> 2), which results in low intensities of IR light reaching the detector thus producing noisy data and spectral distortions.16 Samples were analyzed by diffuse reflectance FTIRS under vacuum (4 mbar) conditions, using an FTIR spectrometer (Bruker IFS 66v/S) equipped with a diffuse reflectance accessory (Harrick Inc.). Each sample was scanned 64 times and data were collected at every 2 cm1 at a resolution of 4 cm1 for wavenumbers between 3750 and 400 cm1 (reciprocal centimeters) which is equal to wavelengths from 2666 to 25 000 nm, thus yielding 1735 data points per sample. To avoid variations caused by temperature, all samples were placed in the same temperature-controlled laboratory (25 ( 0.2 C) as the FTIRS device for at least 5 h prior to analysis. Numerical Analyses. Multiple scatter correction (MSC) and baseline correction were used to linearize spectra and remove variation in spectra caused by noise.6,17 Baseline correction performs a linear correction of the spectra so that two points (3750 and 22102200 cm1) equal zero. MSC removes spectral variation arising from different effective path lengths and particle sizes.17 Partial least-squares regression (PLS) was used to develop quantitative calibration models between FTIR spectra of sediment and conventionally measured sediment properties.18 Principal component analysis was used to assess variation in spectra intensity for specific wavenumbers. The predictive performance of the PLS calibration model was assessed by 10% cross-validation (CV). This means that the calibration model was developed using 90% of data of the calibration lakes with the remaining 10% of the lakes used to test the predictions. This process was repeated a total of 10 times as each group, in turn, was set aside. Root mean squared error of cross validation (RMSECV) was used as an estimate of prediction error. SIMCA-P 11.5 (Umetrics AB, SE-907 Sweden) was used for all multivariate data analyses. All 19 Umea, primary sediment properties were square root transformed prior to analysis. For more detailed information see Rosen et al.8 FTIRS calibrations for BSi and TOC were tested on sediments from Lake Pingualuit and the TIC calibration model was tested on Lake Hoare. The model results were compared to conventionally measured values to evaluate model performance.
’ RESULTS AND DISCUSSION Spectral Information. The loading plots for BSi, TIC, and TOC show similar patterns to those reported in Rosen et al.8 The FTIR spectra from sediment with different concentrations of BSi (Figure 1a) confirm the findings from the loading plot, namely that wavenumbers around 450, 800, and between 1050 and 1280 cm1 are particularly important for the FTIRS-BSi model (Figure 2). Previous studies have shown two main bands at 1100 and
ARTICLE
Figure 1. Examples of FTIR spectra from sediment samples with different concentrations of (a) biogenic silica (BSi) (Lake El’gygytgyn), (b) total inorganic carbon (TIC) (Lake Ohrid, Lake Hoare), and (c) total organic carbon (TOC) (Lake Makkasj€on, Lake El’gygytgyn, Swedish surface sediment). For better visualization, spectra have been shifted vertically from their common baseline by adding a constant value to each spectra.
471 cm1 are attributed to stretching and bending vibration modes of the SiO4 tetrahedron, respectively.20 The band at 800 cm1 corresponds to an intertetrahedral SiOSi bending vibration mode. The band near 945 cm1 corresponds to a SiOH mode.19 This band is not pronounced in our FTIRS-BSi model, possibly due to internal condensation reactions during silica maturation in surface sediments and formation of SiOSi linkages19 which can cause changes in the intensity ratio of SiOSi/SiOH. The region around 3600 should not be used for BSi estimates as samples with confirmed low BSi concentration show higher absorbance than samples high in BSi (Figure 1). A previous study6 showed good model performance when only wavenumbers from 1050 to 1250 were included. However, as the internal structure of BSi can vary with age during crystallization and form (e.g., opal-CT), it is probably advisible to also include bands around 450 and 800 cm1 in the model. This could be particularly important when FTIRS calibrations are used to assess BSi in long sediment cores where diagenetic modification of BSi is likely to increase with depth. The loading plot indicates that wavenumbers in the ranges 700725, 860890, 13001560, 17801810, and 24602640 cm1 8860
dx.doi.org/10.1021/es200203z |Environ. Sci. Technol. 2011, 45, 8858–8865
Environmental Science & Technology
ARTICLE
are particularly important for the FTIRS-TIC model (Figure 2). These wavenumbers correspond well to infrared bands of calcite and other carbonate minerals centered around 710, 875, 1425, 1460, 1800, and 2500 cm1.5,21 Depending on the crystal chemistry and structure of the carbonate mineral, these wavenumbers can show slight deviations. The FTIR spectra from sediment with different concentrations of TIC confirm the findings from the loading plot (Figure 1). The loading plots demonstrate that wavenumbers between 1050 and 1750 and 2800 and 3000 cm1 are especially important for the FTIRS-TOC model (Figure 2). These spectral regions include the CO vibrations of carbohydrates (10001200 cm1)5 as well as the characteristic amide I and amide II vibrations of proteins and CO stretching vibrations of carboxyl groups (14001700 cm1) which are commonly derived from humic substances.5 Finally, absorbance in the region 28003000 cm1 is associated with the CH vibrations of CH3, CH2, and CH groups of aliphatic molecules.5 Statistical Performance of the FTIRS-BSi Model. A 4-component PLS model using wavenumbers 4003750 cm1 has a Rcv = 0.93 and a RMSECV of 3.4% (6.2% of the gradient, gradient 055%, n = 816) (Table 3, Figure 3). Based on the loading plot (Figure 2) and known absorption bands for BSi, a model including the wavenumbers 435480, 790830, and 10501280 cm1 was developed.8,19,20 The 6-component PLS model has a Rcv = 0.93
Figure 2. Loading plots showing the relative contribution of different wavenumbers to the PLS regression models for total organic carbon (TOC), total inorganic carbon (TIC), and biogenic silica (BSi). High values indicate wavenumbers are positively correlated to the Y variable and low values indicate wavenumbers are negatively correlated to the Y variable. Loadings refer to the weight vectors (w*c) in the PLS model (component 1) (y-axis) and the corresponding spectral region (x-axis).
Table 3. Statistical Performance of the Calibration Models for Biogenic Silica (% BSi, Opal), Total Inorganic Carbon (% TIC), and Total Organic Carbon (% TOC) Using All Samples and Wavenumbers in the Calibration Set and Selected Samples and Wavenumbers Covering a Smaller Gradient statistics
BSi
PLS components
4
TIC (%)
6
6
816
706a
0
0
55
55
10.6
10.7
10.4
10.5
3.3
1.6
6.0
2.9
0.93
0.94
wavenumbers
435480;
included (cm1)
194
0
3
3.0
9.7
0.4
8.2
0.6
1.0
0.29
0.66
9.6
9.8
0.89
0.92
435480;
700725;
700725;
790830;
790830;
860890;
860890;
10501280
10501280
13001560;
13001560;
17801810; 24602640
17801810; 24602640
0
55 max
9.7
9.0 mean
1.9
9.6 SD
3.0
3.4 RMSECV
0.68
6.2 RMSECV (% gradient)
6.9
0.93 Rcv
3
685
0 min
6
879
816 samples (n)
7
TOC (%)
0.97
4003750
4003750
7
9
6
3164
3164
210
0
0
10
41
41
41
2.7
2.7
24.5
6.8
6.8
7.4
1.2
1.8
1.9
2.9
4.3
6.1
0.98
0.97
0.96
4003750
28003000
4003750
a Based on BSi estimates including all samples, an optimized BSi model was developed including only samples with a small difference between observed versus predicted values (<4%) in the full data set.
8861
dx.doi.org/10.1021/es200203z |Environ. Sci. Technol. 2011, 45, 8858–8865
Environmental Science & Technology
ARTICLE
Figure 3. Scatter plots showing the relationships between conventionally measured and FTIRS inferred biogenic silica (BSi), total inorganic carbon (TIC), and total organic carbon (TOC) using sediment from c. 100 Swedish lakes, Lake El’gygytgyn (Siberia), Lake Ohrid (Albania/Macedonia), Laguna Potrok Aike (Argentina), Lake Hoare (Antarctica), and Lake Pingualuit (Canada). BSi models shown include (a) all wavenumbers (4003750 cm1), (b) a model restricted to known absorption bands for BSi (435480, 790830, 10501280 cm1) and (c) an optimized model based on samples with a prediction error <4% from the model including all wavenumbers. TIC models shown include (d) all wavenumbers (4003750 cm1), (e) a model restricted to known absorption bands for carbonates (700725, 860890, 13001560, 17801810, 24602640 cm1), and (f) an optimized model based on high TIC concentrations using all wavenumbers (310%). TOC models shown include (g) all wavenumbers (4003750 cm1), (h) a model restricted to known absorption bands for TOC (28003000 cm1), and (i) an optimized model based on high TOC concentration (1041%) using all wavenumbers. For statistical information see Table 3.
and a RMSECV of 3.3%. The conventional leaching technique can contain large errors; for example, Conley2 demonstrated that when 30 different laboratories used the leaching method to analyze a sample with high BSi concentration the estimates varied by 1970% (x = 44.3 ( 9.38 ((1 SD)). Because traditional measurements of BSi can contain large errors2 we also developed a model only including samples which showed a good correlation between inferred and measured values. The selected samples have an error less than 4% between conventional analyses versus FTIR inferred BSi values. Out of 814 samples, 706 pass this criterion. This optimized 6-component PLS model has a Rcv = 0.94 and RMSECV of 1.6% (2.9% of the gradient). Statistical Performance of the FTIRS-TIC Model. A 7-component PLS model using wavenumbers 4003750 cm1 has a Rcv = 0.97 and a RMSECV of 0.68% (7% of the gradient, n = 879,
gradient 010% TIC) (Table 3, Figure 3). Based on the positive loadings (Figure 2) and known CO molecular vibrations present in carbonates, a specific model for TIC was developed, incorporating the following wavenumbers: 700725, 860890, 13001560, 17801810, 24602640 cm1.5,8,21 The statistical performance remained high with Rcv = 0.97 and a RMSECV of 0.72%. The FTIRS model overestimates TIC values for higher values. This could potentially result from spectral saturation causing nonlinear relationships between peak absorption and TIC at high concentrations, which can cause problems for models covering a large gradient. This issue can be avoided by developing specific models for low and high concentrations. By using all wavenumbers and TIC range of 03% the prediction error is reduced to 0.29% (Rcv = 0.89; n = 685). When the gradient is set to 310%, RMSECV is reduced to 0.66% (Rcv = 0.92; n = 194). 8862
dx.doi.org/10.1021/es200203z |Environ. Sci. Technol. 2011, 45, 8858–8865
Environmental Science & Technology
ARTICLE
Figure 4. Comparisons between conventionally measured and FTIRS inferred biogenic silica (BSi) and total organic carbon (TOC) in the 8.4-m-long sediment core from Lake Pingualuit, Canada, covering the last c. 35 kyrs. The comparison for total inorganic carbon (TIC) is based on a Holocene sediment sequence from Lake Hoare, Antarctica. Selected wavenumbers refer to known absorption bands for BSi (435480, 790830, 10501280 cm1), carbonates (700725, 860890, 13001560, 17801810, 24602640 cm1), and organic carbon (28003000 cm1).
Statistical Performance of the FTIRS-TOC Model. A 6-component PLS model using wavenumbers 4003750 cm1 has a Rcv = 0.98 and a cross-validated root mean squared error of prediction (RMSECV) of 1.2% (2.9% of the gradient) (Table 3, Figure 3) (041%, n = 3164). Based on the positive loadings (Figure 2) and known CH vibrations of CH3, CH2, and CH groups of organic compounds, a specific model was developed based on the region 28003000 cm1. The statistical performance remained high with Rcv = 0.97 and RMSECV of 1.8%. Model performance was reduced at high TOC concentrations (1041%, Rcv = 0.74). As with the TIC model, this is probably related to spectral saturation (see above). A specific model for high TOC values (1041%) was therefore developed using all wavenumbers; the resulting model showed improved performance (Rcv = 0.96, RMSECV = 1.9% (6.1% of the gradient), n = 210). A model developed for TOC values between 0 and 5%, a common range for TOC in larger oligotrophic lakes, also had a reduced prediction error (RMSECV) of 0.28% (5.6% of the gradient, n = 2807). Loss-on-ignition at 550 C (LOI550 C) is commonly used to estimate the organic carbon content in sediment.22 One hundred samples were analyzed with both LOI 550 C and TOC% using an elemental analyzer.23 The correlation between LOI 550 C and TOC% is lower (R = 0.89, n = 100, 041%) than the correlation between FTIRSTOC and TOC% reported above. Consequently FTIRS may be a better method than LOI for rapid quantification of TOC. Down-Core Application of FTIRS Models. The output from PLS models can be dependent on the gradient for the response variable (Y) as shown for the TIC and TOC models.18 For optimal performance, a subset of samples covering the gradient of interest should be used.24 One way to achieve this is by applying a flexible two-step approach to assess sediment properties with FTIRS. The first step uses the entire calibration set to cover the range of expected values. The second step is to use these initial results to refine the model to a narrower range. This should improve the model performance and reduce the prediction error. For example, if the initial results show a range of 510% in TOC, an optimized model for this range might include
samples from 0 to 15% rather than the 041% range that is used in the global model. The correlation coefficient for the model may decrease but, more importantly, the error of prediction will also decrease. If the model is applied to very different settings, however, there is still a risk that compounds in the sediment with very different molar absorption properties could cause interference thus resulting in a poor fit (large distance to model, DmodX). This can be avoided by developing a specific model, which only includes wavenumbers specific to the compound of interest, which is demonstrated for BSi, TIC, and TOC below. The FTIRS calibration models for TOC and BSi were applied to an 8.4-m-long sediment core from the Arctic crater lake Pingualuit (Figure 4), which spans the last c. 35 kyrs. To test the robustness of our general model, samples from Lake Pingualuit were excluded from the calibration models. The correlation between conventionally measured BSi using the full gradient (055%) and the FTIRS model based on known absorption bands for BSi (Table 3) was R = 0.66 (n = 30, p < 0.005). Two outliers were identified; their exclusion increased the correlation to R = 0.88, similar to the FTIRS-BSi model performance reported above. If the gradient is reduced to 020%, the correlation remains the same (R = 0.88). The two outliers occurred 0.8 cm apart within the core and were situated at a sand layer with large difference in sediment composition (lower TOC, higher density, larger mean grain size, H. Guyard, personal communication). Because subsampling of FTIRS samples and the conventionally analyzed samples was undertaken on different occasions, it is likely that the measurements have not been performed using sediment with comparable sediment properties. This conclusion is supported by a similar mismatch for the FTIRS-TOC% and TOC% measured with the conventional technique within these two outliers. Using the full gradient for TOC and known absorption bands for organic carbon (Table 3), the correlation is R = 0.85 (n = 218, p < 0.005). When the gradient is reduced to 05%, the R value remains almost the same (R = 0.84). Lake Pingualuit only contains very low concentrations of TIC and consequently detailed analyses using conventional methods 8863
dx.doi.org/10.1021/es200203z |Environ. Sci. Technol. 2011, 45, 8858–8865
Environmental Science & Technology were not undertaken. The FTIRS-TIC model was therefore applied to the 1.2-m-long Holocene sediment core from Antarctic Lake Hoare, which contains low (<1%) but variable TIC concentrations. When the whole gradient (010%) is included, the model performs well with a correlation between conventionally measured and FTIR-inferred TIC of 0.86 (Figure 4, n = 122; p < 0.005). Similar results (R = 0.85, p < 0.005) are achieved when the calibration model is restricted to a smaller gradient (03%). Implications and Future Directions. The results demonstrate that FTIRS can be used to simultaneously assess BSi, TIC, and TOC from lake sediments, including long sediment cores. Previous studies have relied on internal calibration to achieve optimal model performance.7 By using samples from very diverse settings and of varying ages, this study significantly advances FTIRS calibration methods by developing universally applicable models for BSi, TOC, and TIC. Furthermore, the greater number of samples included within these newly developed calibration model has also resulted in improved statistical performance. The loading plots indicate that similar wavenumbers are important for the models as shown in previous studies,7,8 demonstrating that the FTIRS calibration models are based on a robust relationship between compound-specific molecular vibrations and compound concentrations. There are, however, some indications of nonlinear relationships between compound-specific absorption and high concentrations of TIC and TOC. Future research will assess if nonlinear PLS methods can overcome these issues. Our results demonstrate that FTIRS is a useful and rapid analytical alternative for quantitative inferences of BSi, TIC, and TOC in lake sediments using small sample quantities (10 mg). This technique could be beneficial for other lake deep drilling projects where long sediment successions need to be analyzed. The calibration models presented here are broadly applicable to many settings, thus greatly reducing the amount of work required to apply these method at new study sites. FTIRS also represents a valuable analytical tool for studies where only small sample quantities are available, such as high-resolution studies of varved sediments. Future research should also test if calibration models based on lacustrine sediments can be applied to marine sediment or whether specific models need to be developed for the marine realm. Considering that the technique is used in soil science for studies on molecular-level processes at mineral, organic, and bacterial surfaces that influence the biogeochemical cycles of elements,25 there are potentially many more applications for FTIRS sediment studies which still need to be explored. In combination with X-ray fluorescence scanners and multisensor core loggers, FTIRS can provide highly resolved, qualitative and quantitative data on the geochemistry, mineralogy, magnetic, and biogeochemical properties of sediment samples. The models presented here can be used to provide crucial sedimentological information and consequently help to improve reconstructions of environmental variability using lake sediments.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected]; tel: +46 980-40043.
’ ACKNOWLEDGMENT We are grateful to the Pingualuit Crater Lake team, the Lake Ohrid team, the PASADO team; Martin Melles and the Lake
ARTICLE
El’gygytgyn team, Christian Bigler, Dan Conley, Dan Hammarlund, Ulla Kokfelt, Sabrina Ortlepp, Nina Reuss, and co-workers for providing sediment and sediment property data. We thank Annika Holmgren, Carin Olofsson, Sabine Stahl, and Thomas Westin for technical help. Financial support was kindly provided by the Swedish Research Council (VR), FORMAS, the Kempe Foundation, the Climate Impacts Research Centre (CIRC), Umea University, The Canadian Foundation for Climate and Atmospheric Sciences (CFCAS), The International Continental Scientific Drilling Program (ICDP), The German Research Foundation (DFG), and the National Science Foundation of the United States (NSF). We would also like to thank 2 anonymous reviewers and the editor for their valuable comments.
’ REFERENCES (1) Pienitz, R.; Melles, M.; Zolitschka, B. Result of recent sediment drilling activities in deep crater lakes. PAGES Newsl. 2009, 17 (3), 117–118. (2) Conley, D. An interlaboratory comparison for the measurement of biogenic silica in sediments. Mar. Chem. 1998, 63, 39–48. (3) Calace, N.; Capolei, M.; Lucchese, M.; Petronio, B. M. The structural composition of humic compounds as indicator of organic carbon sources. Talanta 1999, 49, 277–284. (4) Stehfest, K.; Toepel, J.; Wilhelm, C. The application of microFTIR spectroscopy to analyze nutrient stress-related changes in biomass composition of phytoplankton algae. Plant Physiol. Biochem. 2005, 43, 717–726. (5) Mecozzi, M.; Pietrantonio, E. Carbohydrates proteins and lipids in fulvic and humic acids of sediments and its relationships with mucilaginous aggregates in the Italian seas. Mar. Chem. 2006, 101, 27–39. (6) Rosen, P.; Vogel, H.; Cunningham, L.; Reuss, N.; Conley, D.; Persson, P. Fourier transform infrared spectroscopy, a new method for rapid determination of total organic and inorganic carbon and opal concentration in lake sediments. J. Paleolimnol. 2010, 43, 247–259. (7) Rosen, P.; Persson, P. Fourier-transform infrared spectroscopy (FTIRS), a new method to infer past changes in tree-line position and TOC using lake sediment. J. Paleolimnol. 2006, 35, 913–923. (8) Vogel, H.; Rosen, P.; Wagner, B.; Melles, M.; Persson, P. Fourier transform infrared spectroscopy, a new cost-effective tool for quantitative analysis of biogeochemical properties in long sediment records. J. Paleolimnol. 2008, 40, 689–702. (9) Doran, P. T.; Berger, G. W.; Lyons, W. B.; Wharton, R. A., Jr.; Davisson, M. L.; Southon, J.; Dibb., J. E. Dating Quaternary lacustrine sediments in the McMurdo Dry Valleys, Antarctic. Palaeogeogr., Palaeoclimatol. Palaeoecol. 1999, 147, 223–239. (10) Haberzettl, T.; Fey, M.; L€ucke, A.; Maidana, N.; Mayr, C.; Ohlendorf, C.; Sch€abitz, F.; Schleser, G. H.; Wille, M.; Zolitschka, B. Climatically induced lake level changes during the last two millennia as reflected in sediments of Laguna Potrok Aike, southern Patagonia (Santa Cruz, Argentina). J. Paleolimnol. 2005, 33, 283–302. (11) Zolitschka, B.; Sch€abitz, F.; L€ucke, A.; Clifton, G.; Corbella, H.; Ercolano, B.; Haberzettl, T.; Maidana, N.; Mayr, C.; Ohlendorf, C.; Oliva, G.; Paez, M. M.; Schleser, G. H.; Soto, J.; Tiberi, P.; Wille, M. Crater lakes of the Pali Aike Volcanic Field as key sites of paleoclimatic and paleoecological reconstructions in southern Patagonia, Argentina. J. South Am. Earth Sci. 2006, 21, 294–309. (12) Vogel, H.; Wagner, B.; Zanchetta, G.; Sulpizio, R.; Rosen, P. A paleoclimate record with tephrochronological age control for the last glacial-interglacial cycle from Lake Ohrid, Albania and Macedonia. J. Paleolimnol. 2010, 44, 295–310. (13) Melles, M.; Brigham-Grette, J.; Glushkova, O. Y.; Minyuk, P. S.; Nowaczyk, N. R.; Hubberten, H. W. Sedimentary geochemistry of core PG1351 from Lake Elgygytgyn a sensitive record of climate variability in the East Siberian Arctic during the past three glacial-interglacial cycles. J. Paleolimnol. 2007, 37, 89–104. (14) Kokfelt, U.; Rosen, P.; Schoning, K.; Christensen, T. R.; F€orster, J.; Karlsson, J.; Reuss, N.; Rundgren, M.; Callaghan, T. V.; Jonasson, C.; 8864
dx.doi.org/10.1021/es200203z |Environ. Sci. Technol. 2011, 45, 8858–8865
Environmental Science & Technology
ARTICLE
Hammarlund, D. Ecosystem responses to increased precipitation and permafrost decay in subarctic Sweden inferred from peat and lake sediments. Global Change Biol. 2009, 15, 1652–1663. (15) M€uller, P. J.; Schneider, R. An automated leaching method for the determination of opal in sediments and particulate matter. Deep-Sea Res. 1993, 40, 425–444. (16) Griffiths, P. R.; De Haseth, J. A. Fourier Transform Infrared Spectrometry; Wiley: New York, 1986. (17) Geladi, P.; MacDougall, D.; Martens, H. Linearization and Scatter-Correction for Near-Infrared Reflectance Spectra of Meat. Appl. Spectrosc. 1985, 39, 491–500. (18) Martens, H.; Naes, T. Multivariate Calibration; John Wiley & Sons: New York, 1989. (19) Schmidt, M.; Botz, R.; Rickert, D.; Bohrmann, G.; Hall, S. R.; Mann, S. Oxygen isotopes of marine diatoms and relations to opal-A maturation. Geochim. Cosmochim. Acta 2001, 65 (2), 201–211. (20) Gendron-Badou, A.; Coradin, T.; Maquet, J.; Fr€ohlich, F.; Livage, J. Spectroscopic characterization of biogenic silica. J. Non-Cryst. Solids 2003, 316, 331–337. (21) White, W. B. The carbonate minerals. In The Infrared Spectra of Minerals; Farmer, V. C., Ed.; Mineralogical Society Monograph 4, 1974; pp 227242. (22) Heiri, O.; Lotter, A. F.; Lemcke, G. Loss on ignition as a method for estimating organic and carbonate content in sediments: Reproducibility and comparability of results. J. Paleolimnol. 2001, 25, 101–110. (23) Ohlsson, K. E. A.; Wallmark, P. H. Novel calibration with correction for drift and non-linear response for continuous flow isotope ratio mass spectrometry applied to the determination of d15N, total nitrogen, d13C and total carbon in biological material. The Analyst 1999, 124, 571–577. (24) Kettaneh-Wold, N.; MacGregor, J. F.; Wold, S. Multivariate design of process experiments (mDOPE). Chemom. Intell. Lab. Syst. 1994, 23, 39–50. (25) Schn€urer, Y.; Persson, P.; Nilsson, M.; Nordgren, A.; Giesler, R. Effects of surface sorption on microbial degradation of glyphosate. Environ. Sci. Technol. 2006, 40, 4145–4150.
8865
dx.doi.org/10.1021/es200203z |Environ. Sci. Technol. 2011, 45, 8858–8865
ARTICLE pubs.acs.org/est
Paramagnetic Relaxation Enhancement (PRE) as a Tool for Probing Diffusion in Environmentally Relevant Porous Media Filipe Furtado,†,‡ Petrik Galvosas,‡,§ Frank Stallmach,‡ Ulf Roland,†,* J€org K€arger,‡ and Frank-Dieter Kopinke† †
Department of Environmental Engineering, Helmholtz Centre for Environmental ResearchUFZ, Permoserstrasse 15, 04318 Leipzig, Germany ‡ Department of Interface Physics, University of Leipzig, Linnestrasse 5, 04103 Leipzig, Germany § MacDiarmid Institute for Advanced Materials and Nanotechnology, School of Chemical and Physical Sciences, Victoria University of Wellington, P.O. Box 600, Wellington 6140, New Zealand
bS Supporting Information ABSTRACT: The transport diffusivity of the paramagnetic molecule 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPO) was measured by monitoring its influence on the NMR transverse relaxation time (T2) on surrounding water protons - also known as paramagnetic relaxation enhancement (PRE). Due to the nature of the PRE effect, few paramagnetic molecules are able to simultaneously reduce the T2 of many NMR active nuclei, which represents a significant gain in sensitivity. In an aqueous solution, the minimal detectable TEMPO concentration was around 70 ppm. The value of the diffusivity was estimated by fitting the relaxation data, collected as a function of time, with the appropriate solutions of the second Fick’s law in respect to the corresponding sample geometry and dimensions. Considering the experimentally determined TEMPO relaxivity in water (“TEMPO-water relaxivity”; RTEMPO = (1.05 ( 0.12) 103 ppm1 s1), the obtained diffusion coefficients (D) of TEMPO in homogeneous solution and in a water saturated sand column (Dbulk = (6.7 ( 0.4) 1010 m2 s1 and Dsand = (1.4 ( 0.5) 1010 m2 s1, respectively) are in good agreement with the expected values (literature values: Dbulk = 6.6 1010 m2 s1, 1.3 1010 m2 s1 < Dsand < 2.3 1010 m2 s1). This new approach enables one to determine the diffusivity of paramagnetic molecules in homogeneous (aqueous solution) and porous media with basic NMR equipment, at low concentrations and in a noninvasive manner.
’ INTRODUCTION Mass transfer processes in porous media are of fundamental importance in describing and understanding phenomena such as pollutant bioavailability and biodegradation in soils. Likewise, they are decisive for an efficient removal of organic compounds in the context of soil remediation. Molecular diffusion (thermal Brownian movement and transport diffusion), plays a major role in these processes, but is a challenging quantity to access experimentally. In this context, nuclear magnetic resonance (NMR) spectroscopy and in particular pulsed field gradient (PFG) NMR spectroscopy are appropriate, noninvasive and nuclei selective methods which allow the assessment of the diffusivity of a particular substance in a heterogeneous sample.15 PFG NMR has been extensively used for investigating the self-diffusivity in a variety of systems such as polymers,6,7 zeolites 810 and soils.11,12 Despite the value of the information provided, there are a number of experimental difficulties in using this approach. PFG NMR has been improved to overcome difficulties in measuring slowly diffusing species13 and the implications arising from internal magnetic field gradients.5,1416 However, this is achieved at the cost of signal intensity loss due to enhanced relaxation time weighting. Indeed, short transverse relaxation times (T2) of the molecules of interest may completely hinder PFG NMR experiments and these may result from strong adsorption or entrapment of the target compounds in the matrix 17,18 and from the presence of paramagnetic species in the sample.19 Magic angle spinning (MAS) PFG NMR 20,21 has been applied for reducing the dipolar r 2011 American Chemical Society
interaction between the matrix and the probe molecule thus opening the option to study samples which were difficult to measure using the conventional PFG NMR. However, for studying environmental samples and transport processes especially in the context of soil pollution and remediation, the typical concentration range (often only some mg kg1) of hazardous compounds is an additional challenge when applying NMR spectroscopy. The corresponding spin densities are far below the values occurring in the NMR applications described before. Nevertheless, both longitudinal (T1) and transverse (T2) relaxation times are strongly influenced by the presence of paramagnetic molecules, even in such low concentrations. This effect is known as paramagnetic relaxation enhancement (PRE) 22,23 and has been used in different studies by deliberately introducing 2428 or generating29 paramagnetic species in the sample and exploiting their influence on the relaxation of the NMR-active nuclei. Contrary to conventional PFG NMR and to MRI, as used on references,27,28 the relaxation measurements needed to probe for the PRE do not require a complex and expensive apparatus for data acquisition, since the spatial labeling of spins is not necessary, thus implying that these experiments can be performed in particularly simple NMR devices. Received: April 1, 2011 Accepted: August 29, 2011 Revised: July 13, 2011 Published: August 29, 2011 8866
dx.doi.org/10.1021/es2010946 | Environ. Sci. Technol. 2011, 45, 8866–8872
Environmental Science & Technology
ARTICLE
As an alternative to PFG NMR experiments, a new methodology based on established NMR methods is proposed in the present study which enables monitoring the transport of organic compounds in homogeneous and porous media. It is applicable for low concentrations which until now were hardly accessible to NMR methods and for systems leading to relatively short relaxation times thus excluding conventional PFG NMR methods. Collecting T2 relaxation data as a function of time and making use of PRE in combination with a suitable model for data analysis, the migration of probe molecules can be detected. Exemplarily, the procedure is demonstrated for the diffusion of TEMPO in aqueous solution. TEMPO and related compounds have been widely studied and used30,31 in PRE studies due to their relatively high stability at room temperature32 and their chemical versatility. In principle, a range of physical properties of the tracer compound can be covered by appropriate derivatives of TEMPO, such that the behavior of environmentally relevant pollutants is represented. For a list of appropriate paramagnetic nitroxyl compounds see ref 33. The NMR samples are prepared with a concentration gradient of the paramagnetic species. Afterward equilibration is followed by monitoring the relaxation properties of the system, namely the relaxation of the water molecules being influenced by TEMPO in their vicinity. By considering adequate models of molecular transport the diffusivity of small amounts of TEMPO becomes experimentally accessible. Influence of Porosity on Relaxation. Due to the fast molecular motion in the liquid state,34,35 pure isotropic liquids usually show only one characteristic transverse relaxation time T2 which we will denote as T2,liq. However, liquids confined in porous media have relaxation times shorter than its corresponding bulk phase, mainly due to the influence of the surface relaxation in the pore system.3638 The measured T2 is given by 1 1 FS ¼ þ T2 T2, liq V
ð1Þ
where F is the surface relaxivity (m s1) and S/V is the surface-tovolume ratio (m1) characteristic for the pore system filled with the liquid. Different pore sizes and geometries will have different S/V ratios and therefore, assuming the same surface relaxivity, lead to different relaxation times. Consequently, a pore size distribution will be accompanied by a relaxation time distribution.39,40 Considering that T2 relaxation is an exponential process with respect to the time (see Supporting Information, SI-1) it follows that for a given pore size distribution the decay of the initial magnetization (M0) will have a multiexponential character: " # MðtÞ t ¼ pi exp ð2Þ M0 T2, i i
∑
where pi denotes the fraction of spins which interacts with a given pore i having a characteristic surface-to-volume ratio. Equation 2 is equivalent to a discrete approximation of a Laplace integral, and the relaxation time distribution can, therefore, be obtained by inverse Laplace transformation (ILT) of the relaxation data.41 NMR Paramagnetic Relaxation Enhancement. As known from the literature,23,4245 the presence of paramagnetic species strongly reduces both relaxation time constants T1 and T2 through dipoledipole interactions. This is because the magnetic moment of an electron spin is almost 3 orders of magnitude larger than
that of a nuclear spin. Hence, in isotropic liquids, the concentration of the paramagnetic species (e.g., TEMPO molecules) is related to T21 of the surrounding molecules described by a constant R referred to as relaxivity in literature,23 as shown by eq 3: 1 1 ¼ þ cTEMPO 3 RTEMPO T2, c T2, c0
ð3Þ
where T2,c (c0) is the transverse relaxation time at a given concentration (at zero concentration). Therefore, it is possible to indirectly estimate the concentration of a given paramagnetic species by measuring T2 of the surrounding nuclei provided that the relaxivity is known and all other parameters are kept constant. The relaxation data can then be described by replacing T2,liq in eq 1 by T2,c, in eq 3 leading to 1 1 FS þ cTEMPO 3 RTEMPO ¼ þ T2, c T2, c0 V
ð4Þ
’ EXPERIMENTAL SECTION Experimental Setup and Data Acquisition. Materials. TEMPO from Sigma-Aldrich was used as a paramagnetic probe molecule. Deionized water was ultrafiltered through a 0.22 μm Millipore membrane prior to use. Quartz sand (Fluka, mesh 40/150; BET surface area below 0.1 m2 g1 as determined by N2 adsorption using a Belsorp-mini device from BEL, Japan) was used after washing with deionized water. The porosity ε was estimated by the ratio of the water volume needed to saturate the interparticle volume of the sand and its total volume. A value of about 0.45 was obtained. Samples for Determining the TEMPOWater Relaxivity, Experiment (a). A saturated stock solution46 was prepared by dissolving 9.7 mg of TEMPO in 1 mL of deionized water in a 2 mL vial. 0.2 mL of this solution were transferred to an NMR sample tube for relaxation measurements. Dilutions of the stock solution were prepared with concentrations ranging from saturation until 0.02 ppm by diluting each individual sample, adding 0.5 mL of water to 0.5 mL of the previously prepared solution. Sample for Determining Diffusion in Homogeneous Media, Experiment (b). We carefully placed 0.1 mg of TEMPO on the bottom of a 1 mm inner diameter quartz tube avoiding contact with the tube walls, then covered with 12 μL of water and subsequently sealed. A sample tube of such reduced inner diameter was chosen to prevent the onset of convection, which would interfere with the outcome of the experiments (see refs 47 and 48 and explanation in Supporting Information, SI-2). A scheme of the prepared sample is shown in the Supporting Information, Figure SI-1. A sample without TEMPO (B1) was prepared by analogously filling 12 μL of deionized water into an identical quartz tube. Sample for Determining Diffusion in Heterogeneous Media, Experiment (c). Sand was filled in an NMR tube with an inner diameter of 5.6 mm up to a filling height of 1 cm corresponding to a sand mass of 507.5 mg. Water was then added and the sample was subsequently centrifuged (2000 rpm) for 3 min, ensuring complete water saturation of the pore space. After removing the free water phase above the sand bed, 20 μL of the TEMPOsaturated stock solution was carefully added on the top of the sand layer (see Figure SI-2 in Supporting Information). Subsequently, the sample was sealed and placed in the NMR 8867
dx.doi.org/10.1021/es2010946 |Environ. Sci. Technol. 2011, 45, 8866–8872
Environmental Science & Technology
ARTICLE
spectrometer. A sample without TEMPO (B2) was prepared in an analogous way, namely by adding 20 μL pure water instead of a TEMPO solution. NMR Data Acquisition. All measurements were performed at 298 K on the home-built 400 MHz proton frequency FT spectrometer FEGRIS.13 PRE can be observed in both T1 and T2 relaxation processes, as described in the Supporting Information (SI-1). However, the commonly used inversionrecovery pulse sequence for studying T1 relaxation relies on multiple excitations (and corresponding waiting times between pulses) for data acquisition, which would necessarily correspond to long time intervals compared to the multiecho CPMG pulse sequence49 (one excitation), used for T2 relaxation. For this reason, the latter relaxation mechanism (T2) was in the focus of this study. The data in experiment (a) were collected using an echo spacing (τ) varying from 0.15 to 1.2 ms in order to achieve comparable attenuation of the magnetization for each sample. The data of both time-dependent experiments (b) and (c) were measured with τ = 0.24 ms. In these experiments, data were collected every 12 min after sealing the samples. Analysis of the Relaxation Data. Time-Independent Relaxation Data. The data collected from Experiment (a) for homogeneous solutions were fitted to a monoexponential function. The samples B1 and B2 without TEMPO in experiments b) and c) respectively, were analyzed by a biexponential function and by performing ILT using the commercially available software Prospa v.2.1 from Magritek Limited. Time-Dependent Relaxation Data. During experiments (b) and (c), a series of CPMG measurements was taken at different times. The TEMPO concentration at a given distance from the source will be a function of diffusion time and of the TEMPO diffusivity. The acquired relaxation data of the water molecules reflect the TEMPO concentration along the sample height. Therefore, for a known TEMPO relaxivity, it is possible to model these concentration profiles as function of the diffusion coefficient. In order to be able to use known solutions of the diffusion equation for this process it is assumed that during the experimental time (i) the relaxivity of TEMPO does not change and (ii) TEMPO molecular displacement does not cover the full length of the sample for both systems under study. Furthermore, concerning experiment (b) and regarding the TEMPO crystal at the bottom of the NMR sample tube to act as a constant and infinite TEMPO source, eq 5 represents the solution of Fick’s law,50 for the corresponding initial conditions: r cr ¼ cs erfc pffiffiffiffiffiffiffiffiffiffiffi ð5Þ 2 Dbulk t
Integration over the whole length (r) of the sample yields for the magnetization decay due to T2 relaxation: MðtCPMG Þ ¼ M0
ð6Þ
r
"
exp tCPMG :
0
1 T2, c0
!# r0 þ cs erfc pffiffiffiffiffiffiffiffiffiffiffi 3 RTEMPO dr 0 2 Dbulk t
ð7Þ In this study, eq 7 was the starting equation used to fit the measured relaxation data, having Dbulk as a fitted parameter. For the experiment (c), where a droplet of saturated TEMPO solution was added on top of a water-saturated sand sample, the solution of Fick’s law is50 " ! !# cs h þ r hr ð8Þ erf pffiffiffiffiffiffiffiffiffiffiffi þ erf pffiffiffiffiffiffiffiffiffiffiffi cðh, rÞ ¼ 2 2 Dpore t 2 Dpore t where, in addition to the already defined parameters, h is the thickness of the TEMPO-saturated water layer (see Figure SI-2 in Supporting Information). It is important to note that eq 8 takes into account the decrease in the concentration at the source because the limited amount of TEMPO initially present in the range of length h is distributed over the sample. In contrast, in experiment (b) the mass of the TEMPO crystal is rather irrelevant, as long as it is not dissolved completely, which is the case during the considered experimental time. Similarly to the eq 6, the relaxation rate in the sand sample for a given initial TEMPO concentration cs, at a given distance r to the source, is obtained by replacing cTEMPO in eq 3, with eq 8
" ! !# 1 1 cs h þ r hr ¼ i þ erf pffiffiffiffiffiffiffiffiffiffiffi þ erf pffiffiffiffiffiffiffiffiffiffiffi 3 RTEMPO 2 T2,i c T2, c0 2 Dpore t 2 Dpore t
ð9Þ where i indicates the corresponding phase, contained in the region 0 to h (homogeneous bulk phase, i = bulk) or in the region h to r (heterogeneous sand phase, i = sand). As already mentioned, the values of Ti2,c0 are different, and need to be determined experimentally (see Results and Discussion section). Combining this result with the general relaxation equation (see Supporting Information SI-1) and integrating over the respective lengths r and h lead to: " # " # Z h Z r MðtCPMG Þ tCPMG tCPMG 0 0 ¼ exp bulk dh þ exp sand dr M0 T2, c T2, c 0 0 ð10Þ Tbulk 2,c0
and Z h
Tsand 2,c0
from eq 9 gives the following result: 0 MðtCPMG Þ 1 cs h þ r pffiffiffiffiffi erf ¼ exp tCPMG : bulk þ M0 2 T2, c0 2 Dt 0 0 h r 0 þ erf pffiffiffiffiffi 3 RTEMPO dh 2 Dt " Z r 1 cs h þ r0 pffiffiffiffiffi erf þ exp tCPMG : sand þ 2 T2, c0 2 Dt h h r0 0 þ erf pffiffiffiffiffi ð11Þ R 3 TEMPO dr 2 Dt
Replacing
cs denotes the concentration at the source, cr is the concentration at a distance r from the source, t is the time of diffusion and Dbulk is the TEMPO diffusion coefficient. The relaxation data can, therefore, be described by a sum of exponential decays with varying relaxation times depending on the local TEMPO concentration. The summation has to take into account the contributions of the differently relaxing water species present in the full length of the sample. The relaxation rate for a given TEMPO concentration at the source cs, at a given distance to the source r, is obtained replacing cTEMPO in eq 3, with cr in eq 5, giving: 1 1 r ¼ þ cs erfc pffiffiffiffiffiffiffiffiffiffiffi 3 RTEMPO T2, c T2, c0 2 Dbulk t
Z
"
which is the general equation used to fit the experimental data from experiment (c). 8868
dx.doi.org/10.1021/es2010946 |Environ. Sci. Technol. 2011, 45, 8866–8872
Environmental Science & Technology
Figure 1. Transverse relaxation time (T2) of water in dependence on the TEMPO concentration. The detection limit for the relaxation effect of TEMPO is indicated by the gray dashed line.
ARTICLE
Figure 2. Transverse relaxation rate (1/T2) of water in dependence on the TEMPO concentration.
’ RESULTS AND DISCUSSION Influence of the TEMPO concentration on relaxation of bulk water, experiment (a). These data sets invariably reveal a
monoexponential behavior (see Supporting Information, Figure SI-3). This leads to the conclusion that on the time scale of these measurements and independently of the concentration, the TEMPO molecules are homogeneously distributed in the sample, leading to a uniform T2 reduction of all water molecules. In Figure 1, the transverse relaxation time T2 obtained by a monoexponential fit is plotted as a function of the TEMPO concentration. It can be seen that the minimal detectable concentration is approximately 70 ppm, corresponding to an average distance between TEMPO molecules of about 15 nm. The inverse of the measured T2 as a function of the concentration (Figure 2) yields a linear dependence in which the slope corresponds to the TEMPO relaxivity and the ordinate to the relaxation rate (1/T2) of the bulk liquid, as described by eq 3. The obtained values are RTEMPO = (1.05 ( 0.12) 103 ppm1 s1 and T2,H2O = (2.11 ( 0.01) s. Determination of the TEMPO Bulk Diffusivity, Experiment (b). First, the accurate value of T2,c0 in this system was determined by measuring the blank sample (B1). The data deviate slightly from a monoexponential decay, which reflects a significant S/V ratio of the 1 mm tube (see Supporting Information SI-4). The calculated parameters from a biexponential fit are pa = 0.838 ( 0.001, pb = 0.162 ( 0.001, Ta2,c0 = (1.562 ( 0.002) s, Tb2,c0 = (0.219 ( 0.001) s. Therefore, the previously presented solution of the diffusion equation in this system (eq 5) has to be combined with a distribution of T2 at cTEMPO = 0 (eq 2) and integrated over the whole sample length which gives " Z r MðtCPMG Þ 1 ¼ pi exp tCPMG : i M0 T2, c0 0 i r0 p ffiffiffiffiffiffiffiffiffiffiffi þ cs erfc ð12Þ RTEMPO dr 0 2 Dbulk t 3
∑
with i representing, in this case, the indexes a and b, from the biexponential fit. The semilogarithmic plots of the time dependent relaxation data (Experiment b) and the result of the fit using eq 12 using the T2 values from the blank sample are shown in Figure 3. It is observed that for longer periods after starting the diffusion experiments the CPMG data deviate from a monoexponential (linear) character toward a multiexponential (nonlinear) one. This fact is attributed to the continuous dissolution of the TEMPO crystal, which increases the dissolved concentration
Figure 3. Measured CPMG data of experiment (b) as a function of diffusion time (red circle) and fitted data according to eq 12 (semitransparent surface, blue cross-hatching).
close to the source (r = 0), and spreading over the whole cross section of the tube. Using the length coordinate along the tube axis, the estimated parameters from the blank sample (pa, pb, Ta2,c0, Tb2,c0), the TEMPO concentration at saturation of the water phase and the previously estimated value for the TEMPO-water relaxivity allows to fit the experimental data by the equations above, having the diffusion coefficient Dbulk as a fitted parameter. The result of the fit is represented by the semitransparent surface from which a value for the diffusion coefficient of Dbulk = (6.7 ( 0.4) 1010 m2 s1 is obtained. It is in good agreement with the value Dbulk = 6.6 1010 m2 s1 obtained by Ciszkowska et al. for a temperature of 298 K.51 Using the determined diffusion coefficient and the diffusion equations, the concentration profiles being established throughout the sample in the course of the experiment can be calculated. The concentrations as a function of distance from the source (as shown in Supporting Information, Figure SI-6) reveal that the concentration at the largest distance from the source (at the end of the tube) is still negligible for the longest observation times of the experiment, that is, after 7 h. This result is a confirmation of assumption (ii) and thus validates the used model (eq 5). Determination of TEMPO the Diffusivity through a Sand Bed, Experiment (c). Determination of the Influence of the Sand on Water Relaxation. The accurate distribution of Ti2,c0 values was again determined by measuring the corresponding sample without TEMPO (B2). Figure 4 shows the T2 distribution obtained from fitting the CPMG data (see Supporting Information SI-5) by ILT. Despite the model-free approach in fitting the data by ILT, a direct correlation from the spatial distribution of water in the sample 8869
dx.doi.org/10.1021/es2010946 |Environ. Sci. Technol. 2011, 45, 8866–8872
Environmental Science & Technology
ARTICLE
Figure 4. Transverse relaxation time (T2) distribution of water in a sand bed (sample B2) obtained by ILT of the relaxation data.
with the T2 distribution observed in Figure 4 can be made, The peak centered at T2 = 1.8 s can be attributed to the bulk water layer (i = bulk) over the sand column, whereas the broad peak at T2 < 0.6 s corresponds to the water phase between the sand grains (i = pore). Equation 10 was derived considering that the phases i have only one discrete value of T2 (pore or bulk). However, Figure 4 indicates that we have to take into account a relaxation time distribution for each given concentration of TEMPO instead. Thus, we have to generalize eq 10 and account for T2 distributions (based on the discrete values obtained from ILT leading to the sum terms): 2 3 Z h MðtCPMG , rÞ t CPMG ¼ pj exp4 bulk, j 5dh0 M0 T2, c 0 j
∑
Z þ h
r
∑k pk exp
"
tCPMG k T2,sand, c
# dr 0
ð13Þ
where j and k represent the T2 distribution in the bulk and in the and Tsand,k from pore water phase, respectively. Replacing Tbulk,j 2,c 2,c eq 9 we obtain MðtCPMG Þ ¼ M0
2
Z
h 0
∑j
0
Figure 6. Calculated TEMPO concentrations from NMR measurements at different distances from the source for diffusion times between 0.1 and 15 h after sample preparation.
times confirms assumption (ii) and thus validates the used model (eq 8). The resulting value for the diffusion coefficient can be related to the porosity of the system (ε)48 through the empirical Archie relation with the empirical constant m following the relation
0 1 cs h þ r erf pffiffiffiffiffiffiffiffiffiffiffi pj exp4tCPMG :@ bulk, j þ 2 2 Dsand t T2, c0
h0 r pffiffiffiffiffiffiffiffiffiffiffi R dh0 TEMPO 3 2 Dsand t " Z r 1 cs h þ r0 erf pffiffiffiffiffiffiffiffiffiffiffi þ pk exp tCPMG : sand, k þ 2 2 Dsand t T2, c0 h k 0 hr 0 þ erf pffiffiffiffiffiffiffiffiffiffiffi 3 RTEMPO dr 2 Dsand t
Figure 5. Measured CPMG data of experiment c) (red circle) and fitted data according to eq 14 (semitransparent surface, blue cross-hatching).
þ erf
∑
ð14Þ
from which the diffusivity Dsand can be extracted. The obtained value of the TEMPO diffusion coefficient in the pore space is Dsand = (1.4 ( 0.5) 1010 m2 s1. The semilogarithmic plots of the experimental data and the fit are shown in Figure 5. From the diffusion coefficient and eq 8, again the concentration profiles in the sample can be estimated. As it can be seen in Figure 6, at time zero (gray dashed line) all TEMPO molecules are contained at the top of the sand layer, in the region between 0 and h (integration range i). At longer diffusion times, molecules migrate from the bulk phase into the pore space of the sand bed, lowering the concentration in the bulk phase and increasing it in the pore water. The fact that the concentration remains approximately 0 close to the maximum depth of the sand layer at all experimental
DTEMPO, pore ¼ εm DTEMPO, bulk
ð15Þ
This calculation yields m = 1.9, which is in the range of literature data for unconsolidated sand, from 1.3 to 2.0, as reviewed by Grathwohl et al.52,53 Our study presents self-contained sets of data, which can be fitted and interpreted by the introduced models. The low specific surface area of the quartz sand used justifies neglecting the adsorption of TEMPO on the sand surface on data fitting and in the relaxivity determination. In other cases, where the porous matrices under study have a high specific surface area or a higher affinity for the used probe molecule, different models may need to be used, taking into account the adsorption in the pores.52,53 Adsorption may not only cause an additional barrier for transport, hence reducing the observed effective diffusion coefficient, and a deviation from the idealized system where only one diffusion process occurs, but can as well have an effect on the relaxivity (R) of the water molecules.32,54 It is known that the immobilization of paramagnetic molecules on the surface of the solid matrix may change the surface relaxivity (F) of the matrix itself.55 Therefore these two parameters (R and F in eq 4) need to be determined and considered as a function of the system under study, and included in the subsequent model for data fitting. Tentatively, we propose the measurement of the relaxation times of water at 8870
dx.doi.org/10.1021/es2010946 |Environ. Sci. Technol. 2011, 45, 8866–8872
Environmental Science & Technology different concentrations of the spin-probe and corresponding degrees of surface coverage, thereby experimentally determining ). Furthermore, the presented the second term of eq 13 (Tsand,k 2,c models can be adapted to describe the case where a distribution of diffusion coefficients is present in the sample, by expanding eq 8 to the appropriate distribution. The presented methodology allows the estimation of the diffusion of a given probe molecule in a porous medium. Relying on the strong effect of the diffusive species present in low concentrations on the bulk phase spin relaxation, in this case that of water protons, and using a relatively simple CPMG pulse sequence for relaxation measurement, opens a new opportunity to study transport processes under environmentally relevant conditions. The proposed method is noninvasive. Data analysis and fitting is achieved using the appropriate solutions for the diffusion equations to estimate the concentration profiles and the resulting influence on the relaxation rate of the surrounding water molecules. The method was demonstrated to be successfully applicable for TEMPO as paramagnetic relaxation enhancer in two model systems (point source in a homogeneous water phase and sand column with TEMPO-containing phase at the boundary). This method can be applied to a variety of paramagnetic compounds with varying properties such as polarity, size, chemical nature, sorption affinity, water solubility etc. and being present in low concentrations. Thus, specific information on transport of a variety of chemicals in complex matrices can be obtained, which is of great relevance in the prognosis of the fate of chemical compounds in the environment.
’ ASSOCIATED CONTENT
bS
Supporting Information. On the basic theory of NMR relaxation, experimental conditions as well as with figures showing relaxation data and a sketch of the prepared samples. This material is available free of charge via the Internet at http://pubs. acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone ++49 (341) 235 1762; fax ++49 (341) 235 45 1762; e-mail:
[email protected].
’ ACKNOWLEDGMENT Financial support by the Marie Curie-Early Stage Training project RAISEBIO (Risk Assessment and Environmental Safety Affected by Compound Bioavailability in Multiphase Environments) and by the Deutsche Forschungsgemeinschaft (International Research Training Group “Diffusion in Porous Materials”) is gratefully acknowledged. ’ REFERENCES (1) Stejskal, E. O.; Tanner, J. E. Spin diffusion measurements: Spin echoes in the presence of a time-dependent field gradient. J. Chem. Phys. 1965, 42, 288–292. (2) Callaghan, P. Principles of Nuclear Magnetic Resonance Microscopy; Clarendon Press: Oxford, 1993. (3) Stallmach, F.; K€arger, J. The potentials of pulsed field gradient NMR for investigation of porous media. Adsorption 1999, 5, 117–133. (4) Stallmach, F.; Galvosas, P. Spin echo NMR diffusion studies. Annu. Rep. NMR Spectrosc. 2007, 61, 51–131.
ARTICLE
(5) Price, W. S. NMR Studies of Translational Motion; Cambridge University Press: Cambridge, 2009. (6) Walderhaug, H. PFG NMR study of polymer and solubilizate dynamics in aqueous isotropic mesophases of some poloxamers. J. Phys. Chem. B 1999, 103, 3352–3357. (7) Matsukawa, S.; Ando, I. A study of self-diffusion of molecules in polymer gel by pulsed-gradient spin-echo 1H NMR. Macromolecules 1996, 29, 7136–7140. (8) Gupta, V. Evidence for single file diffusion of ethane in the molecular sieve AlPO4-5. Chem. Phys. Lett. 1995, 247, 596–600. (9) K€arger, J. Diffusion measurement by NMR techniques. In Adsorption and Diffusion; Springer: Berlin, Heidelberg, 2008, 7, 85133. (10) Menjoge, A.; Bradley, S. A.; Galloway, D. B.; Low, J. J.; Prabhakar, S.; Vasenkov, S. Observation of intraparticle transport barriers in FAU/EMT intergrowth by pulsed field gradient NMR. Microporous Mesoporous Mater. 2010, 135, 30–36. (11) Schaumann, G. E.; LeBoeuf, E. J. Glass transitions in peat: Their relevance and the impact of water. Environ. Sci. Technol. 2005, 39, 800–806. (12) Simpson, A. J. Determining the molecular weight, aggregation, structures and interactions of natural organic matter using diffusion ordered spectroscopy. Magn. Reson. Chem. 2002, 40, S72–S82. (13) Galvosas, P.; Stallmach, F.; Seiffert, G.; K€arger, J.; Kaess, U.; Majer, G. Generation and application of ultra-high-intensity magnetic field gradient pulses for NMR spectroscopy. J. Magn. Reson. 2001, 151, 260–268. (14) Cotts, R.; Hoch, M.; Sun, T.; Markert, J. Pulsed field gradient stimulated echo methods for improved NMR diffusion measurements in heterogeneous systems. J. Magn. Reson. 1989, 83, 252–266. (15) Gibbs, S. J.; Johnson, C. S. A PFG NMR experiment for accurate diffusion and flow studies in the presence of eddy currents. J. Magn. Reson. 1991, 93, 395–402. (16) Galvosas, P.; Stallmach, F.; K€arger, J. Background gradient suppression in stimulated echo NMR diffusion studies using magic pulsed field gradient ratios. J. Magn. Reson. 2004, 166, 164–173. (17) Daughney, C. J.; Bryar, T. R.; Knight, R. J. Detecting sorbed hydrocarbons in a porous medium using proton nuclear magnetic resonance. Environ. Sci. Technol. 2000, 34, 332–337. (18) Simpson, M. J.; Simpson, A. J.; Hatcher, P. G. Noncovalent interactions between aromatic compounds and dissolved humic acid examined by nuclear magnetic resonance spectroscopy. Environ. Toxicol. Chem. 2004, 23, 355–362. (19) Kleinberg, R. L. Pore size distributions, pore coupling, and transverse relaxation spectra of porous rocks. Magn. Reson. Imaging 1994, 12, 271–274. (20) Fomba, K. W.; Galvosas, P.; Roland, U.; K€arger, J.; Kopinke, F.-D. New option for characterizing the mobility of organic compounds in humic acids. Environ. Sci. Technol. 2009, 43, 8264–8269. (21) Gratz, M.; Hertel, S.; Wehring, M.; Stallmach, F.; Galvosas, P. Mixture diffusion of adsorbed organic compounds in metal-organic frameworks as studied by MAS PFG NMR. New J. Phys. 2011, accepted for vol. 13. (22) Clore, G. M.; Iwahara, J. Theory, practice, and applications of paramagnetic relaxation enhancement for the characterization of transient low-population states of biological macromolecules and their complexes. Chem. Rev. 2009, 109, 4108–4139. (23) Helm, L. Relaxivity in paramagnetic systems: Theory and mechanisms. Prog. Nucl. Magn. Reson. Spectrosc. 2006, 49, 45–64. (24) Weckhuysen, B. Chemical imaging of spatial heterogeneities in catalytic solids at different length and time scales. Angew. Chem., Int. Ed. 2009, 48, 4910–4943. (25) Deczky, K.; Langford, C. H. Application of water nuclear magnetic resonance relaxation times to study of metal complexes of the soluble soil organic fraction fulvic acid. Can. J. Chem. 1978, 56, 1945–1951. (26) Melton, J. R.; Kantzas, A.; Langford, C. H. Nuclear magnetic resonance relaxometry as a spectroscopic probe of the coordination sphere of a paramagnetic metal bound to a humic acid mixture. Anal. Chim. Acta 2007, 605, 46–52. 8871
dx.doi.org/10.1021/es2010946 |Environ. Sci. Technol. 2011, 45, 8866–8872
Environmental Science & Technology (27) Moradi, A. B.; Oswald, S. E.; Massner, J. A.; Pruessmann, K. P.; Robinson, B. H.; Schulin, R. Magnetic resonance imaging methods to reveal the real-time distribution of nickel in porous media. Eur. J. Soil Sci. 2008, 59, 476–485. (28) Fischer, A. E.; Hall, L. D. Visualization of the diffusion of metal ions and organic molecules by magnetic resonance imaging of water. Magn. Reson. Imaging 1996, 14, 779–783. (29) Mitreiter, I.; Oswald, S. E.; Stallmach, F. Investigation of iron(III) release in the pore water of natural sands by NMR relaxometry. Open Magn. Reson. J. 2010, 3, 46–51. (30) Vallet, P.; van Haverbeke, Y.; Bonnet, P. A.; Subra, G.; Chapat, J. P.; Muller, R. N. Relaxivity enhancement of low molecular weight nitroxide stable free radicals: Importance of structure and medium. Magn. Reson. Med. 1994, 32, 11–15. (31) Bennett, H. F.; Brown, R. D.; Koenig, S. H.; Swartz, H. M. Effects of nitroxides on the magnetic field and temperature dependence of 1/T1 of solvent water protons. Magn. Reson. Med. 1987, 4, 93–111. (32) Ciriano, M. V.; Korth, H. G.; van Scheppingen, W. B.; Mulder, P. Thermal stability of 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPO) and related N-alkoxyamines. J. Am. Chem. Soc. 1999, 121, 6375–6381. (33) March, D. Spin-label EPR for determining polarity and proticity in biomolecular assemblies: Transmembrane profiles. Appl. Magn. Reson. 2010, 37, 435–454. (34) Bergman, D. J.; Dunn, K. J. NMR of diffusing atoms in a periodic porous medium in the presence of a nonuniform magnetic field. Phys. Rev. E 1995, 52, 6516–6535. (35) Levitt, M. Spin Dynamics: Basics of Nuclear Magnetic Resonance; John Wiley & Sons: Chichester New York, 2001. (36) Brownstein, K.; Tarr, C. Importance of classical diffusion in NMR studies of water in biological cells. Phys. Rev. A 1979, 19, 2446–2453. (37) Dunn, K. J.; Bergman, D. J.; Latorraca, G. A. Nuclear Magnetic Resonance; Pergamon: Amsterdam London, 2002. (38) Vogt, C.; Galvosas, P.; Klitzsch, N.; Stallmach, F. Self-diffusion studies of pore fluids in unconsolidated sediments by PFG NMR. J. Appl. Geophys. 2002, 50, 455–467. (39) Latour, L. Pore-size distributions and tortuosity in heterogeneous porous media. J. Magn. Reson., Ser. A 1995, 112, 83–91. (40) Jaeger, F.; Bowe, S.; van As, H.; Schaumann, G. E. Evaluation of 1 H NMR relaxometry for the assessment of pore-size distribution in soil samples. Eur. J. Soil Sci. 2009, 60, 1052–1064. (41) Venkataramanan, L.; Song, Y. Q.; Hurlimann, M. D. Solving Fredholm integrals of the first kind with tensor product structure in 2 and 2.5 dimensions. IEEE Trans. Signal Process. 2002, 50, 1017–1026. (42) Bakhmutov, V. Practical NMR Relaxation for Chemists; John Wiley & Sons: Chichester Hoboken NJ, 2004. (43) Banci, L. Nuclear and Electron Relaxation: The Magnetic NucleusUnpaired Electron Coupling in Solution; VCH: Weinheim New York, 1991. (44) Bertini, I. Solution NMR of Paramagnetic Molecules: Applications to Metallobiomolecules and Models; Elsevier Science Ltd.: Amsterdam New York, 2001. (45) Yazyev, O. V.; Helm, L. Nuclear spin relaxation parameters of MRI contrast agents—Insight from quantum mechanical calculations. Eur. J. Inorg. Chem. 2008, 2, 201–211. (46) Merck Chemicals. 2,2,6,6-Tetramethylpiperidinyloxyl Security Data Sheet. Merck Chemicals, 2008. (47) Hedin, N. Temperature imaging by 1H NMR and suppression of convection in NMR probes. J. Magn. Reson. 1998, 131, 126–130. (48) Goux, W. The impact of Rayleigh-Benard convection on NMR pulsed-field-gradient diffusion measurements. J. Magn. Reson. 1990, 88, 609–614. (49) Meiboom, S.; Gill, D. Modified spin-echo method for measuring nuclear relaxation times. Rev. Sci. Instrum. 1958, 29, 688–691. (50) Crank, J. The Mathematics of Diffusion; Clarendon Press: Oxford, 1979. (51) Zhang, W.; Ma, C.; Ciszkowska, M. Mass transport in thermoresponsive poly(N-isopropylacrylamide-co-acrylic acid) hydrogels studied
ARTICLE
by electroanalytical techniques: Swollen gels. J. Phys. Chem. B 2001, 105, 3435–3440. (52) F€orstner, U.; Grathwohl, P. Ingenieurgeochemie; Springer: Berlin Heidelberg, 2007. (53) Boving, T. Tracer diffusion coefficients in sedimentary rocks: Correlation to porosity and hydraulic conductivity. J. Contam. Hydrol. 2001, 53, 85–100. (54) Gussoni, M.; Greco, F.; Ferruti, P.; Ranucci, E.; Ponti, A.; Zetta, L. Poly(amidoamine)s carrying TEMPO residues for NMR imaging applications. New J. Chem. 2008, 32, 323–332. (55) Bryar, T. Paramagnetic effects of iron(III) species on nuclear magnetic relaxation of fluid protons in porous media. J. Magn. Reson. 2000, 142, 74–85.
8872
dx.doi.org/10.1021/es2010946 |Environ. Sci. Technol. 2011, 45, 8866–8872
ARTICLE pubs.acs.org/est
Gene Quantification by the NanoGene Assay is Resistant to Inhibition by Humic Acids Gha-Young Kim,† Xiaofang Wang,‡ Hosang Ahn,§ and Ahjeong Son*,‡ †
Department of Nuclear Fuel Cycle Technology Development, Korea Atomic Energy Research Institute, Yuseonggu, Daejeon 305-701, Republic of Korea ‡ Department of Civil Engineering, Auburn University, Auburn, Alabama 36849, United States § Materials Research and Education Center, 275 Wilmore Labs, Auburn, Alabama 36849, United States
bS Supporting Information ABSTRACT: NanoGene assay is a magnetic bead and quantum dot nanoparticles based gene quantification assay. It relies on a set of probe and signaling probe DNAs to capture the target DNA via hybridization. We have demonstrated the inhibition resistance of the NanoGene assay using humic acids laden genomic DNA (gDNA). At 1 μg of humic acid per mL, quantitiative PCR (qPCR) was inhibited to 0% of its quantification capability whereas NanoGene assay was able to maintain more than 60% of its quantification capability. To further increase the inhibition resistance of NanoGene assay at high concentration of humic acids, we have identified the specific mechanisms that are responsible for the inhibition. We examined five potential mechanisms with which the humic acids can partially inhibit our NanoGene assay. The mechanisms examined were (1) adsorption of humic acids on the particle surface; (2) particle aggregation induced by humic acids; (3) fluorescence quenching of quantum dots by humic acids during hybridization; (4) humic acids mimicking of target DNA; and (5) nonspecific binding between humic acids and target gDNA. The investigation showed that no adsorption of humic acids onto the particles’ surface was observed for the humic acids’ concentration. Particle aggregation and fluorescence quenching were also negligible. Humic acids also did not mimic the target gDNA except 1000 μg of humic acids per mL and hence should not contribute to the partial inhibition. Four of the above mechanisms were not related to the inhibition effect of humic acids particularly at the environmentally relevant concentrations (<100 μg/mL). However, a substantial amount of nonspecific binding was observed between the humic acids and target gDNA. This possibly results in lesser amount of target gDNA being captured by the probe and signaling DNA.
’ INTRODUCTION Gene quantification techniques such as fluorescent in situ hybridization or quantitative polymerase chain reaction (qPCR) are widely used in environmental science and engineering to detect and quantify specific bacterial genes in environmental samples. Unfortunately these environmental samples also often contain humic compounds that can inhibit the quantification capability of these techniques. This significantly limits the use of these gene quantification techniques to field studies whose samples are not laden with humic compounds. Humic compounds such as humic acid, fulvic acid, or humin are naturally occurring organic compounds that contain anionic functional groups (i.e., phenolic and carboxylic groups) as well as hydrophobic components (i.e., aromatic and aliphatic moieties).1 Humic compounds are known as the most commonly reported group of inhibitors in environmental samples. For the techniques based on DNA hybridization such as fluorescent in situ hybridization, humic compounds appear to have deleterious effects on several reaction components and their interactions.2,3 They lower the efficiency in DNADNA hybridization 4,5 and reduce r 2011 American Chemical Society
the amount of DNA/RNA binding to membrane by occupying some of the nucleic acid binding sites on the membrane.6,7 The effect of humic compounds on the performance of PCR is well documented.811 It was reported that 1 μL of humic acidlike extract was sufficient to inhibit a 100 μL reaction mix.12 Several studies also showed that the humic impurities (<0.1 μg/mL) interfere with the interaction between target DNA and Taq polymerase, which is a key enzyme in PCR amplification.1315 The inhibitory effect of humic acids on qPCR assay may be due to the inhibition of Taq polymerase by humic acids 6,7 and/or the complex formation of Mg2+ ions, vital cofactor for Taq polymerase, with humic acids.12,14 Young et al. suggested that soil humic compounds possessed phenolic groups and they can either denature biological molecules via bonding with amides or oxidize to form a quinone which covalently binds to DNA or Received: April 19, 2011 Accepted: September 7, 2011 Revised: September 2, 2011 Published: September 07, 2011 8873
dx.doi.org/10.1021/es2013402 | Environ. Sci. Technol. 2011, 45, 8873–8880
Environmental Science & Technology
ARTICLE
Figure 1. Five hypotheses that describe the possible mechanisms of humic acids-resistance of the NanoGene assay: 1. adsorption of humic acids on the particle surface, 2. particle aggregation induced by humic acids, 3. fluorescence quenching of QDs by humic acids during hybridization, 4. humic acids mimicking target DNA, and 5. random nonspecific binding between humic acids and target gDNA. MB stands for magnetic beads and QD stands for quantum dot nanoparticles.
proteins.16 Zipper et al. proposed several molecular mechanisms underlying the impact of humic acids on the fluorimetric assay using SYBR Green dye. They include inner filter effect, collisional quenching, and competitive binding between the dye and humic acids. The mechanisms are potentially responsible for the fluorescence quenching of the dyeDNA complex by humic acids.17 Humic impurities are often coextracted with nucleic acids from soil, sediment, and water samples.18 Because extensive DNA purification does not ensure complete removal of humic compounds,9 it is necessary to enable and improve the inhibition resistance of gene quantification assays. We have previously developed a nanoparticle-based gene quantification (NanoGene) assay which was shown to be potentially resistant to a number of environmental inhibitors.19 It uses DNA hybridization with dual nanosize quantum dot (QD) labels and magnetic bead carrier.20 The NanoGene assay captures the target genomic DNA (gDNA) via hybridization with two DNAs which are bound to a magnetic bead (MB) and fluorescent QDs. Magnetic separation is used to consolidate the captured target gene, and fluorescence measurement is used for quantification. In this paper, we specifically addressed the need for an inhibitor resistance gene quantification technique via in-depth characterization of the inhibition resistance of the NanoGene assay to the presence of humic acids in test samples containing target gDNA. Using qPCR as a comparison, the quantification capability of the NanoGene assay was measured for various concentrations of humic acids (7 orders of magnitude) in the test samples. In addition, potential interactions between humic acids and the components of the assay were also investigated for their role in determining the performance of the assay. We hypothesized that the following mechanisms are contributing to the observed inhibition: (1) adsorption of humic acids on the particle surface; (2) particle aggregation induced by humic acids; (3) fluorescence quenching of QDs by humic acids during hybridization; (4) humic acids mimic target DNA; and (5) random nonspecific binding between humic acids and target gDNA. The schematic of the five hypotheses is presented in Figure 1.
’ MATERIALS AND METHODS NanoGene assay. The recent study by Kim and Son 20
describes the details of the NanoGene assay including its
particleparticle and particleDNA conjugates. The procedure of particleparticle and particleDNA conjugates used for the recent study is summarized in Supporting Information. The denatured form of target E. coli O157:H7 gDNA was hybridized with MB-QD655-probe DNA and QD565-signaling probe DNA in 400 μL of DIG easy hybridization buffer (Roche Diagnostic, Basel, Switzerland) for 12 h at 37 °C using a gentle tilt rotation. After washing three times with phosphate buffer and its separation by a magnet (MPC-96S, Invitrogen), the fluorescence of both QDs was measured by a SpectraMax M2 microplate reader (MDS, Sunnyvale, CA). The fluorescence output was normalized (QD565/QD655). The quantification of target gene, eaeA gene of E. coli O157: H7, was performed in the absence and the presence of humic acids (Aldrich, St. Louis, MO). Gene copy numbers of gDNA were calculated based on the concentration of gDNA (ng/μL), gene sequences (151 bp), and molecular weight of each base (660 g/mol). Thus, gDNA of 0.22.0 ng/μL is equivalent to 8.0 105 to 8.0 106 eaeA gene copies. Varying amounts of humic acids (0.0011000 μg/mL of reaction) were added. The environmentally relevant concentration range of humic acids is 0.0230 μg/mL.21 The inhibition effect of humic acids was presented as the quantification capability (%), which was the output (i.e., fluorescence) of the assay performed with various amounts of humic acids divided by the output of assay without humic acids (no inhibition). Quantitative PCR assay. The qPCR assay was used as a comparison in this study. The detailed procedure for qPCR20 is described in Supporting Information. In parallel to NanoGene assay, various amounts of humic acids were injected to the reaction with the gDNA in SYBR Green buffer (total volume of 25 μL). The qPCR assay was performed using the StepOne Real-Time PCR system (Applied Biosystems, Foster City, CA).22 The quantification capability (%) of qPCR was obtained in the same manner as that of the NanoGene assay. Field Emission Scanning Electron Microscopy (FE-SEM). The pattern of adsorbed humic acids on the particle conjugate of NanoGene assay was observed by FE-SEM (JSM-7000F, JEOL, Japan) after incubation. To simulate exposure to humic acids in test samples during hybridization, the MB-QD655 particle conjugate was incubated with humic acids in two concentrations (100 and 1000 μg/mL) in the same manner as the hybridization of the NanoGene assay. The particle conjugate was subsequently 8874
dx.doi.org/10.1021/es2013402 |Environ. Sci. Technol. 2011, 45, 8873–8880
Environmental Science & Technology dispersed in 200 μL of phosphate buffer (pH = 7.4). Five μL of the samples were dispensed on carbon conductive tape (Electron Microscopy Sciences, Hatfield, PA) attached to and dried on the specimen holder (JEOL). The particle conjugate of NanoGene assay and humic molecules were observed by FE-SEM at an accelerating voltage of 20 kV. Adsorption of Humic Acids on the Particles during Hybridization. To quantify the adsorption of humic acids on the particle conjugates of NanoGene assay during hybridization, MB, MB-QD655, and MB-QD655-probe DNA (these particle conjugates will be referred to as MB-QD and MB-QD-DNA, respectively, hereafter) were incubated in 400 μL of hybridization buffer containing humic acids for 12 h. The concentrations of humic acids used were 0, 0.1, 1, 10, 100, and 1000 μg/mL. The particle reagents were subsequently isolated from the solution via magnetic separation. Supernatants were transferred to clear polystyrene 96-well plates (Nunc, Roskilde, Denmark) for absorbance measurement via SpectraMax M2 spectroscopy (MDS, Sunnyvale, CA) at λ = 332 nm. The adsorbed amounts of humic acids on the particles were calculated by subtracting the absorbance of supernatant from the initial absorbance. The amounts (μg HA/mL/particle) of humic acids adsorbed on the particles were normalized by the absorbance of MB (measured at λ = 742 nm). The remaining amounts of humic acids were statistically compared with initial amounts using a paired t test (n = 3, two-tailed). Particle Aggregation, Zeta Potential, and Hydrodynamic Diameter. The particles may aggregate in the presence of humic acids due to the influence from the various functional groups of humic acids. To examine the possibility of particle aggregation in the presence of humic acids, surface charge and hydrodynamic radii of the particles were observed in varying amounts of humic acids (0.0011000 μg/mL). The pH change in solution was also monitored. The surface charges (i.e., zeta potential) of the particles were measured by Zetasizer nano ZS (Malven, UK) with laser Doppler anemometry with phase analysis light scattering in a capillary cell (750 μL, Malven). The dynamic viscosity of the hybridization buffer was determined as 2.6325 cP by the viscosity meter (GV 2200, Gilmont, Barrington, IL). The refractive index of the particles used is 1.5 for the measurements. To estimate the particle aggregation, the size distribution of particles in solution was determined based on the hydrodynamic diameter of the particles. The hydrodynamic diameters were also measured by Zetasizer nano ZS with dynamic light scattering application using the same capillary cell. The measured particle size is the hydrodynamic radius of hydrated/solvated particle, including shape and surface roughness of particle and solvent molecules surrounding particle. All experiments were performed in triplicate and in each instance size of particle were measured ten times. Quenching of Quantum Dots Fluorescence by Humic Acids. The quenching effect of QD fluorescence by the presence of humic acids (0.0011000 μg/mL) was investigated. The MBQD particle conjugate was dispersed in 200 μL of phosphate buffer and subsequently various concentrations of humic acids were added. The mixture of MB-QD particle conjugate and humic acids was transferred to a black 96-well plate for the fluorescence measurement (λem = 360 nm, λex = 660 nm) to observe for potential collisional quenching effect. As compared to the control that contains no humic acids, the result would indicate the level of quenched fluorescence of samples that contained humic acids. The mixture of particles and humic acids
ARTICLE
were incubated for 12 h and the UV-absorption analysis was performed to examine the static quenching effect. Mimicking Target Genomic DNA. The possibility of humic acids mimicking the target gDNA was examined by using humic acids instead of target gDNA in the hybridization process. Various amounts of humic acids (0.0011000 μg/mL) including negative control (no humic acids) were added to the hybridization buffer containing MB-QD655-probe DNA and QD565-signaling probe DNA. The hybridization of humic acids with the probe and signaling probe DNA labeled with MB-QD655 and QD565, respectively, was determined by measuring the normalized fluorescence (QD565/QD655). Nonspecific Binding between Humic Acids and Genomic DNA (Passive Adsorption). To examine the nonspecific binding between humic acids and gDNA, humic acids encapsulated MBs were incubated with various amounts of gDNA. Humic acids encapsulated MBs were prepared by adding 5 μg of humic acids to MBs (2 107 beads) and incubated with coupling agents for covalent bonding (i.e., EDC and NHS) at ambient temperature. After 2 h, the humic acids coated MBs were separated by a magnet and washed three times with phosphate buffer. The humic acids coated MBs were subsequently dispersed in buffer and transferred to a clear polystyrene 96-well plate. The amount of humic acids used for the encapsulation of MBs was determined to be 2.5 μg per 2 107 MBs based on the absorbance at 430 nm. The coated MBs were incubated with various amounts of gDNA (0.012.0 ng per μL of reaction) in hybridization buffer at 37 °C with a slow tilt rotation. After 12 h, the particle conjugates were separated by magnetic field. The amount of gDNA that was not adsorbed on the particles was isolated by centrifugation. The supernatants were collected to measure the residual amount of gDNA by absorbance at 260 nm. The amount of nonspecifically bound gDNA, resulting from the passive adsorption, was obtained by subtracting the gDNA in the supernatant from the prior gDNA applied.
’ RESULTS AND DISCUSSION Resistance of NanoGene Assay to Humic Acids. The quantification capability (%) of the NanoGene assay was not completely inhibited by the presence of humic acids (Figure 2a). But the quantification capability of qPCR reduced to 0% over 1 μg/mL humic acids, although it maintained 90% at 0.1 μg/mL humic acids (Figure 2b). The output (fluorescence) of NanoGene assay slightly decreased at the high concentration of humic acids, however it maintained the gene quantification capability more than 50% at all concentration ranges of humic acids (0.0011000 μg/mL) and target gDNA. Note the range of environmental relevant concentration of humic acids is 0.02 30 μg/mL. Therefore quantification of raw environmental samples using qPCR assay may not be plausible because the inhibition is very significant and the result will show no signals over 1 μg/mL humic acids. To further increase the inhibition resistance of NanoGene assay at high humic acid concentrations, we identified the specific mechanisms that are responsible for the inhibition as described in Figure 1. Adsorption of Humic Acids on the Particles. FE-SEM analysis was performed to visualize the interaction of humic acids with the components (MB-QD) of NanoGene assay. As shown in Figure 3, the humic molecules were observed as a coagulated pattern and resided on the surface of MB-QD at higher concentrations (100 and 1000 μg/mL). It was reported 8875
dx.doi.org/10.1021/es2013402 |Environ. Sci. Technol. 2011, 45, 8873–8880
Environmental Science & Technology that the humic molecules can be held together in supramolecular conformations by weak hydrophobic bond at neutral and alkaline pH.23 Because humic substances have both hydrophobic and hydrophilic functional groups (i.e., amphiphilic), they can form aggregates which have the hydrophobic interior with the highly charged exterior24 and reside on the particle surface. However the amount of information that can be derived from the image
Figure 2. Comparison of the quantification of the eaeA gene of E. coli O157:H7 by (a) NanoGene and (b) qPCR assays in the presence of various concentrations of humic acids. Four tested target gDNAs for both NanoGene and qPCR assays were identical to 8.0 105 through 8.0 106 eaeA gene copy numbers of E. coli O157:H7. The signal and error bar represent the mean and the standard deviation, respectively, based on five measurements. Note that this caption also applies for the following Figures 47.
ARTICLE
pertaining to the interactions (e.g., quantitative data of adsorption) between humic acids and MB-QD is limited. Therefore we performed the following experiment in order to measure the adsorption of humic acids on the surface of particles. During DNA hybridization, humic acids may be adsorbed on the components of NanoGene assay. Therefore it causes the interruption of DNA hybridization and further gene quantification of assay. Various amounts of humic acids were incubated with the particles (i.e., MB, MB-QD, and MB-QD-DNA) under hybridization condition to examine the adsorption of humic acids. Table 1 shows the amounts of adsorbed humic acids on the particles. No humic acids were adsorbed on the particles at lower concentrations (<10 μg/mL of reaction), and a slight adsorption occurred at higher concentrations (101000 μg/mL). To determine the degree of humic acids adsorption at higher concentrations, the initial amount (μg/mL) and remaining (the initial amount the adsorbed amount) portion of humic acids were compared using the paired t test. All tests were nonsignificant (P > 0.01, n = 3). In other words, there was no significant adsorption of humic acids throughout all the concentrations applied (0.11000 μg/mL) on all particles tested: MB, MB-QD, and MB-QD-DNA particle complex. Particle Aggregation: Surface Charge and Particle Size Distribution. We hypothesized the particle complexes may aggregate in the presence of humic acids, as humic acids have various functional groups. To investigate particle aggregation, particle size distribution was determined using dynamic light scattering spectroscopy. Figure 4a shows the hydrodynamic diameter measurement of the particles (i.e., MB, MB-QD, and MB-QD-DNA). The diameter of particles (Figure 4a) was approximately 4 μm regardless of various amounts of humic acids and the particle combinations. The particle size was uniform and less than 5.6 μm (depicted by the line in Figure 4a), which is twice MB’s diameter. The result showed that no significant particle aggregation was induced by humic acids. The stability of particle dispersion in the presence of humic acids was also investigated. The stability of particles in solution decreases as the particle aggregation occurs due to charge neutralization. Surface charge values > ( 30 mV indicate well dispersed particles with no aggregation.25 In the absence of humic acids (negative control), the surface charges of particle conjugates were around 60 mV. As shown in Figure 4b, the surface charges of all three particle conjugates were approximately 60 mV over the range (0.0011000 μg/mL) of humic acids used in the experiment. This meant no aggregation of
Figure 3. FE-SEM image of MB-QDs (a) without humic acids, (b) with 100 μg/mL humic acids, and (c) with 1000 μg/mL humic acids. The locations of putative humic molecules are indicated by the arrows. The scale bar represents 100 nm. 8876
dx.doi.org/10.1021/es2013402 |Environ. Sci. Technol. 2011, 45, 8873–8880
Environmental Science & Technology
ARTICLE
Table 1. Adsorption of humic acids on the particle complex of NanoGene assay: (a) MB, (b) MB-QD, and (c) MB-QD-DNA MB humic acids (μg/mL)
adsorbed (μg HA/mL/particle)
MB-QD P-value
MB-QD-DNA
adsorbed (μg HA/mL/particle)
P-value
adsorbed (μg HA/mL/particle)
P-value
0
0.0
0.0
0.0
0.1
0.0
0.0
0.0
1
0.0
0.0
10
0.34 ( 0.31
0.2244
0.05 ( 0.08
0.4226
0.0
100
4.24 ( 3.26
0.1538
2.31 ( 2.07
0.2018
3.34 ( 4.20
0.2811
1000
21.98 ( 7.67
0.0205
22.66 ( 21.77
0.2138
31.49 ( 5.57
0.0192
Figure 4. Particle aggregation effect by humic acids. (a) Hydrodynamic diameters and (b) zeta potential (surface charge) distribution of the particle complex of the NanoGene assay (i.e., MB, MB-QD, and MBQD-DNA) in the presence of varying humic acids.
particles occurred in the presence of humic acids. In other words, the humic acids do not function as the bridge between particles for further coagulation. In addition, the pH was maintained at 7.4 without being affected by the amount of humic acids. As the concentration of humic acids increases, humic acids may coagulate by clumping with each other instead of interacting with particles. In this regard, we have also determined the size distribution with humic acids only without the particle conjugates. The result showed that increasing humic acids caused the increased coagulation of humic molecules. Figure S1 shows the shift of size distribution of humic acids molecules at different concentrations of humic acids. The size of the most abundant humic molecules increased from 150 to 250 nm as humic acids increased from 100 to 1000 μg/mL. Increasing concentration of humic acids induces the formation of humic coagulants instead of
0.0
particle aggregation. This possibly suggests that the binding affinity between humic acids is stronger (at least identical) than the interaction between humic acids and particles. Fluorescence Quenching of Quantum Dots by Humic Acids. To examine the fluorescence quenching effect of humic acids, the fluorescence of MB-QD in the presence of various humic acids was measured. The relative fluorescence intensities of samples were calculated based on the assumption that the initial fluorescence intensity of MB-QD in the absence of humic acids was 100%. As shown in Figure 5a, the fluorescence intensity of MB-QD was sustained at 80% of its initial value when exposed to humic acids concentration range from 0.001 to 10 μg/mL. However, at the higher concentration of humic acids (100 and 1000 μg/mL), the fluorescence decreased to 40%. One-way ANOVA test indicated that the relative fluorescence (%) at higher concentrations of humic acids was significantly different (P < 0.05, n = 16). This observation is consistent with the earlier finding that humic acids at higher concentrations tend to coagulate each other as shown in both Figures 3 and S1. The coagulation of humic acids may result in erroneous fluorescence measurement of QD particle. In addition, the probability of collisional encounters between humic acids and particles increases with higher concentration of humic acids. The collision between humic acids and MB-QD may generate the loss of excitation energy in the form of heat instead of photon emission (fluorescence), resulting in the decrease of the MB-QD fluorescence. This loss in fluorescence is known as collisional (or dynamic) quenching. The collisional quenching is the interaction of transient excited state and does not affect the absorption spectrum.17,26 On the other hand, the interaction of the fluorophore with quencher can form a nonfluorescent complex, resulting in static quenching. Since the complex may have a different spectrum from the fluorophore, the change of absorption spectra would indicate static quenching.26 As a result of UVvis absorbance scanning, no peak shift was observed in the absorption spectrum of MB-QD incubated with various concentrations of humic acids relative to the absence of humic acids (Figure 5b). The spectra of negative control (no humic acids), which is depicted as a thick line in Figure 5b, showed a similar pattern for all tested samples over the entire concentration range of humic acids. One way ANOVA was used to test the significance of the absorbance peak shift. The ANOVA result showed no significant difference (P > 0.05, n = 7) between the wavelengths for the absorbance peak for all the samples tested. In other words, the presence of humic acids did not affect the optical characteristic of MB-QD by creating other complexes which would be represented by peaks in absorption spectra. This indicated that no static quenching occurred between humic acids and particle complex. 8877
dx.doi.org/10.1021/es2013402 |Environ. Sci. Technol. 2011, 45, 8873–8880
Environmental Science & Technology
ARTICLE
Figure 6. Mimicking effect of humic acids.
gDNA. Figure 7a shows the adsorption isotherm of target gDNA on the surface of humic acids encapsulated MB. Cs in y-axis indicates the mass of target gDNA over the mass of humic acids (ng/μg) and Ce in x-axis indicates the concentration of target gDNA added (ng/μL), where the volume is the reaction volume of the assay. The adsorbed amount of target gDNA per unit amount of humic acids on MBs increased initially and reached a plateau value (5.5 ng/μg). The adsorption isotherm of target gDNA onto the humic acids coated on MBs was fitted using the Langmuir equation (eq 1): q¼
Figure 5. Fluorescence quenching effects by humic acids. (a) Collisional quenching: fluorescence intensity of MB-QD655 with various concentrations of humic acids. (b) Static quenching: UVvis absorbance spectra of MB-QD655 incubated with varying humic acids (0.0011000 μg/mL).
Humic Acids Mimicking Target Genomic DNA. The possibility of humic acids mimicking target gDNA was examined by incubating the particles with humic acids instead of target gDNA. Figure 6 shows the normalized fluorescence obtained by the hybridization of humic acids with the probe DNAs. The DNA hybridization in the NanoGene assay was carried out using humic acids in place of target gDNA. In the event that the humic acids mimic target gDNA, fluorescence (QD565/QD655) will be detected as a result of the hybridization between humic acids and probe/signaling probe DNAs. However, all the signals (Figure 6) were lower than the background level (1.17 RFU) and it is the limit of detection of the NanoGene assay.20 All the signals, except 1000 μg/mL (P > 0.05, n = 7), were significantly different from the limit of detection (P < 0.05, n = 7, Student’s t-test). Thus, the humic acids except 1000 μg/mL did not mimic the target gDNA and no hybridization would occur between humic acids and probe/signaling DNAs. Nonspecific Binding between Humic Acids and Target Genomic DNA. Adsorption test was used to investigate the degree of nonspecific binding between humic acids and target
qmax KA C 1 þ KA C
ð1Þ
where q = the adsorbed target gDNA concentration (ng gDNA/ μg humic acids on MBs), qmax = the maximum concentration of adsorbed target gDNA, KA = constant, and C = the residual concentration of target gDNA in solution. The constants KA and qmax were evaluated from the linearized form represented by eq 2: 1 1 1 1 ¼ þ q qmax KA C qmax
ð2Þ
A plot of 1/q against 1/C gives a straight line with a slope of 1/(qmax KA) and an intercept of 1/qmax. The correlation coefficient (R2) describing the goodness of fit to the linearized Langmuir model was 0.99. The qmax and KA were 5.71 (ng/μg) and 1.50 102, respectively. This means a maximum of 5.71 ng of gDNA can be adsorbed by 1 μg of humic acids (bounded on MBs). The binding between target gDNA and humic acids is likely due to passive adsorption. Passive adsorption can occur via the combination of both electrostatic and hydrophobic interactions. Humic acids have various hydrophilic functional groups as well as hydrophobic center. Similarly gDNA has hydrophilic groups such as amine and phosphate as well as hydrophobic impurities such as proteins. Because humic acids contain organic groups with variable aromaticity, the hydrophobic part of humic acids may attract the hydrophobic impurities of gDNA. In addition, the phenolic groups of humic substances can also denature biological molecules by forming amide bonds or oxidizing to form a quinone which covalently bonds to DNA or proteins.6,16 These various functional sites may induce the nonspecific binding between gDNA and humic acids. 8878
dx.doi.org/10.1021/es2013402 |Environ. Sci. Technol. 2011, 45, 8873–8880
Environmental Science & Technology
ARTICLE
QDs. Another mitigation strategy can involve the use of the additional interceptor molecules (e.g., polymers), which can preferably bind with humic acids, prior to the quantification. The interceptor molecule will be chosen such that it will not bind to the MB-QDs. In this way, the humic acids in the sample will be bound to the interceptor molecule and therefore unable to further bind to the target gDNA. These methods are reasonable propositions because the binding efficiencies between gDNA and the competing humic acids and MB-QDs are in the same order of magnitude. They will be investigated in our follow up studies. Note that qPCR remained inhibited (0% quantification) regardless of the amount of gDNA in the presence of the same or less amount of humic acids (Figure 1b). In summary, we have demonstrated the inhibition resistance of NanoGene assay to the presence of humic acids in test samples for concentration as high as 10 μg of humic acid per mL. At this concentration, the NanoGene assay was able to maintain more than 60% of its quantification capability. The four following mechanisms: (1) adsorption of humic acids on the particles, (2) particle aggregation by humic acids, (3) fluorescence quenching by humic acids, and (4) binding between probe DNA and humic acids, did not show significant contribution to the inhibition effect of humic acids, particularly at the environmentally relevant concentrations (<100 μg/mL). However, substantial nonspecific binding (adsorption) between target gDNA and humic acids was observed. The adsorption may be attributed to the amphiphilic structures of both humic acids and gDNA. This study provides evidence of nonspecific binding as the mechanism by which humic acids can inhibit DNA hybridization in the NanoGene assay and will direct the optimization of the NanoGene assay in future studies.
’ ASSOCIATED CONTENT
bS Figure 7. Nonspecific binding of gDNA and humic acids via passive adsorption. (a) The adsorption isotherm of gDNA on the humic acidsencapsulated MB. (b) Relative gene quantity that are obtained as a function of gDNA (0.4, 1, 2 ng/μL).
To further examine the nonspecific binding between gDNA and humic acids, additional quantification experiment (NanoGene assay) was performed with test samples containing varying amounts of gDNA and the same amount of humic acids (1 μg/μL: the highest humic acids concentration (1000 μg/ mL) in the previous graphs). As shown in Figure 7b, as the amount of gDNA increases, the quantification capability (percentage) of the NanoGene assay also increases. This observation is consistent with the hypothesis that humic acids compete with the NanoGene assay components to bind with the gDNA via nonspecific binding. In addition, the results showed that the humic acids were able to bind a larger absolute amount, but lower overall percentage, of the gDNA at a higher gDNA concentration. Therefore it is indicative that the binding efficiency between humic acids and gDNA is on the same order of magnitude as that between components of NanoGene assay (MB-QDs) and gDNA. To mitigate inhibition via nonspecific binding between humic acids and gDNA, the binding efficiency between MB-QDs and gDNA can be increased via increasing the concentration of MB-
Supporting Information. Preparation of particle components of NanoGene assay, detailed information of qPCR, and hydrodynamic diameter distribution of humic acids. This information is available free of charge via the Internet at http://pubs. acs.org/.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (+1) 334 844 6260; fax: (+1) 334 844 6290; e-mail:
[email protected]; address: 238 Harbert Engineering Center, Auburn University, Auburn, AL 36849.
’ ACKNOWLEDGMENT This work was supported by National Science Foundation (CAREER award program) and USGS-Alabama Water Resources Research Institute. ’ REFERENCES (1) Stumm, W.; Morgan, J. J. Aquatic Chemistry, Chemical Equilibria and Rates in Natural Waters; John Wiley & Sons, Inc.: New York, 1996. (2) Jacobsen, C. S.; Rasmussen, O. F. Development and application of a new method to extract bacterial DNA from soil based on separation of bacteria from soil with cation-exchange resin. Appl. Environ. Microbiol. 1992, 58, 2458–2462. (3) Wilson, I. G. Inhibition and facilitation of nucleic acid amplification. Appl. Environ. Microbiol. 1997, 63 (10), 3741–3751. 8879
dx.doi.org/10.1021/es2013402 |Environ. Sci. Technol. 2011, 45, 8873–8880
Environmental Science & Technology (4) Steffan, R. J.; Goksoyr, J.; Bej, A. K.; Atlas, R. M. Recovery of DNA from soils and sediments. Appl. Environ. Microbiol. 1988, 54 (12), 2908–2915. (5) Tijssen, P. Hybridization with Nucleic Acid Probes, Part I: Theory and Nucleic Acid Precipitation; Elsevier Science Publishers B.V.: Amsterdam, The Netherlands, 1993. (6) Alm, E. W.; Zheng, D.; Raskin, L. The presence of humic substances and DNA in RNA extracts affects hybridization results. Appl. Environ. Microbiol. 2000, 66, 4547–4554. (7) Bachoon, D. S.; Otero, E.; Hodson, R. E. Effects of humic substances on fluorometric DNA quantification and DNA hybridization. J. Microbiol. Methods 2001, 47, 73–82. (8) Chandler, D. P.; Schreckhise, R. W.; Smith, J. L.; Bolton, H. Electroelution to remove humic compounds from soil DNA and RNA extracts. J. Microbiol. Methods 1997, 28, 11–19. (9) Zhou, J.; Bruns, M. A.; Tiedje, J. M. DNA recovery from soils of diverse composition. Appl. Environ. Microbiol. 1996, 62, 316–322. (10) Ogram, A.; Sayler, G. S.; Barkay, T. The extraction and purification of microbial DNA from sediments. J. Microbiol. Methods 1987, 7. (11) Porteus, L. A.; Armstrong, J. L. Recovery of bulk DNA from soil by a rapid, small-scale extraction method. Curr. Microbiol. 1991, 22, 345–348. (12) Tsai, Y.-L.; Olson, B. H. Detection of low numbers of bacterial cells in soils and sediments by polymerase chain reaction. Appl. Environ. Microbiol. 1992, 58 (2), 754–757. (13) Tebbe, C. C.; Vahjen, W. Interference of humic acids and DNA extracted directly from soil in detection and transformation of recombinent DNA from bacteria and a yeast. Appl. Environ. Microbiol. 1993, 59, 2657–2665. (14) Tsai, Y.-L.; Olson, B. H. Rapid method for separation of bacterial DNA from humic substances in sediments for polymerase chain reaction. Appl. Environ. Microbiol. 1992, 58, 2292–2295. (15) McGregor, D. P.; Forster, S.; Steven, J.; Adair, J.; Leary, S. E. C.; Leslie, D. L.; Harris, W. J.; Titball, R. W. Simultaneous detection of microorganisms in soil suspension based on PCR amplification of bacterial 16S rRNA fragments. Biotechniques 1996, 21, 463–471. (16) Young, C.; Burghoff, R. L.; Keim, L. G.; Minak-Bernero, V.; Lute, J. R.; Hinton., S. M. Polyvinylpyrrolidone-agarose gel electrophoresis purification of polymerase chain reaction-amplifiable DNA from soils. Appl. Environ. Microbiol. 1993, 59, 1972–1974. (17) Zipper, H.; Buta, C.; Lammle, K.; Brunner, H.; Bernhagen, J.; Vitzthum, F. Mechanisms underlying the impact of humic acids on DNA quantification by SYBR Green I and consequences for the analysis of soils and aquatic sediments. Nucleic Acids Res. 2003, 31 (7), e39. (18) Jackson, C. R.; Harper, J. P.; Willoughby, D.; Roden, E. E.; Churchill, P. F. A simple, efficient method for the separation of humic substances and DNA from environmental samples. Appl. Environ. Microbiol. 1997, 63, 4993–4995. (19) Kim, G.-Y.; Wang, X.; Son, A. Inhibitor resistance and in-situ capability of nanoparticle based gene quantification. J. Environ. Monit. 2011, 13, 1344–1350. (20) Kim, G.-Y.; Son, A. Development and characterization of a magnetic bead-quantum dot nanoparticles based assay capable of Escherichia coli O157:H7 quantification. Anal. Chim. Acta 2010, 677, 90–96. (21) Brum, M. C.; Oliveira, J. F. Removal of humic acid from water by precipitate flotation using cationic surfactants. Miner. Eng. 2007, 20, 945–949. (22) Carey, C. M.; Kostrzynska, M.; Thompson, S. Escherichia coli O157:H7 stress and virulence gene expression on Romaine lettuce using comparative real-time PCR. J. Microbiol. Methods 2009, 77 (2), 235–242. € (23) Osterberg, R.; Lindovist, I.; Mortensen, K. Particle size of humic acid. Soil Sci. Soc. Am. J. 1993, 57, 283–285. (24) Wershaw, R. L. Model for humus in soils and sediments. Environ. Sci. Technol. 1993, 27, 814–816. (25) Zeta Potential of Colloids in Water and Waste Water; ASTM Standard D 4187-82; American Society for Testing and Materials: Conshohocken, PA, 1985.
ARTICLE
(26) Lakowicz, J. R. Fluorescence Quenching: Theory and Applications, 2nd ed.; Kluwer Academic/Plenum: New York, 1991.
8880
dx.doi.org/10.1021/es2013402 |Environ. Sci. Technol. 2011, 45, 8873–8880
ARTICLE pubs.acs.org/est
Ambient Pressure Proton Transfer Mass Spectrometry: Detection of Amines and Ammonia D. R. Hanson,*,† P. H. McMurry,‡ J. Jiang,‡,§ D. Tanner,|| and L. G. Huey|| †
Chemistry Department, Augsburg College, 2211 Riverside Avenue, Minneapolis, Minnesota 55454, United States Mechanical Engineering, University of Minnesota, Minneapolis, Minnesota, United States § School of Environment, Tsinghua University, Beijing, China School of Earth and Atmospheric Sciences, Georgia Institute of Technology, Atlanta, Georgia, United States
)
‡
bS Supporting Information ABSTRACT: An instrument to detect gaseous amines and ammonia is described, and representative data from an urban site and a laboratory setting are presented. The instrument, an Ambient pressure Proton transfer Mass Spectrometer (AmPMS), consists of a chemical ionization and drift region at atmospheric pressure coupled to a standard quadrupole mass spectrometer. Calibrations show that AmPMS sensitivity is good for amines, and AmPMS backgrounds were suitably determined by diverting sampled air through a catalytic converter. In urban air at a site in Atlanta, amines were detected at subpptv levels for methyl and dimethyl amine which were generally at a low abundance of <1 and ∼3 pptv, respectively. Trimethyl amine (or isomers) was on average about 4 pptv in the morning and increased to 15 pptv in the afternoon, while triethyl amine (or isomers or amides) increased to 25 pptv on average in the late afternoon. The background levels for the 4 and 5 carbon amines and ammonia were high, and data are very limited for these species. Improvements in detecting amines and ammonia from a smog chamber were evident due to improvements in AmPMS background determination; notably dimethyl amine and its OH oxidation products were followed along with impurity ammonia and other species. Future work will focus on accurate calibration standards and on improving the sample gas inlet.
1. INTRODUCTION Ammonia (NH3) and amines (e.g., a primary amine, RNH2, R = CH3, etc.) are the most common basic compounds found in the atmosphere in both the gas14 and particulate phase.511 Laboratory12 and theoretical work13 indicates that these compounds have a potentially large affinity for the particulate phase if strong acids are present. The coaccumulation of amines and or ammonia with acids on atmospheric particles can be important for growth of atmospheric particles.710 Also, nitrogen bases are believed to enhance particle nucleation rates.1417 Currently, there are a variety of methods, e.g. derivatization and chromatography,18,19 infrared absorption, etc.,2022 chemical ionization mass spectrometry (CIMS)23,24 to measure gas phase ammonia. There is also a relatively large set of observations of atmospheric ammonia.3,21 In contrast, the observations of atmospheric amines are rare because of their generally low abundance and the constraints of the measurement techniques. Consequently, the distribution, sinks, and sources of atmospheric amines are not well-established.4 Amines have been detected in the atmosphere using primarily chromatographic methods that are specific for individual compounds. However, these methods often lack high time resolution due to preconcentration, derivatization, and elution steps that may require hours. In recent years mass spectrometry has been used to r 2011 American Chemical Society
measure atmospheric amines. Selegri et al.25 used a modified proton transfer-reaction-mass spectrometer (PTR-MS) to detect amines (e.g., trimethyl amine) and other compounds in rural air in Finland. Eisele26 deployed an atmospheric pressure ionization CIMS and reported detection of amines species such as trimethyl amine and picoline/aniline in addition to ammonia. Amines have been detected in atmospheric particulate matter711 which indicates that these compounds can be important for particle growth and perhaps their formation. In this work a new method for the measurement of atmospheric amines is characterized, Ambient pressure Proton transfer Mass Spectrometry (AmPMS). The AmPMS combines the specificity of chemical ionization with the high sensitivity of atmospheric pressure ionization techniques. Species with very high proton affinities (e.g., amines) are selectively ionized by reaction with large water clusters of the hydronium ion, H3O+(H2O)n. The design, calibration, and basic operating procedure of the AmPMS are presented here. In addition, initial observations of amines by the AmPMS in an urban environment and in a Teflon smog chamber are presented. Received: May 27, 2011 Accepted: September 6, 2011 Revised: September 5, 2011 Published: September 06, 2011 8881
dx.doi.org/10.1021/es201819a | Environ. Sci. Technol. 2011, 45, 8881–8888
Environmental Science & Technology
ARTICLE
2. EXPERIMENTAL SECTION Measurement Sites. Amine measurements are reported from two sites: ambient air in Atlanta from late July to mid-August 2009 and from a smog chamber at the University of Minnesota in 2010. The urban site in Atlanta has hosted a number of field campaigns in the past dozen years and has been described previously in a number of publications.23,2729 The 2009 campaign (NCCN: Nucleation and Cloud Condensation Nuclei) was focused on particle formation, and a suite of particle instruments along with sulfuric acid and particle mass spectrometers were deployed along with AmPMS. AmPMS was also deployed at a smog chamber consisting of a ∼1 m3 Teflon bag in a temperature regulated steel container housed at the University of Minnesota (UM). Small amounts of amines could be introduced into the chamber with a microsyringe as either vapor or aqueous mixtures. The chamber was flushed repeatedly with air from a Pure Air Generator 737 (Advanced Analytical Device Company, Cleves, Ohio, USA.) The chamber was used for particle formation experiments using photolytic generation of sulfuric acid. The chamber and experiments will be described in detail in a future publication (Titcombe, McMurry et al. manuscript in preparation.) Also, results of periodic testing of AmPMS sampling capabilities and calibrations at Augsburg College (AC) are presented. Apparatus Description. AmPMS was developed from a previous instrument focused on laboratory studies of ammonia30 and consists of a chemical ionization and drift region at atmospheric pressure coupled to a standard quadrupole mass spectrometer. Reagent ions, H3O+ 3 (H2O)n, are produced in a radioactive source and then drawn across a flow of sampled air with an electric field where they can react with trace gases. Reagent and product ions enter a vacuum system where they are mass selected with a quadrupole mass spectrometer and detected with a channel electron multiplier. Shown in Figure 1 are the main parts of AmPMS: ion source, drift region, curtain region, ion sampling inlet, and collisional dissociation chamber (CDC). The sample gas inlet with three way solenoid valve and catalytic converter are also shown. Details of these parts are discussed in sequence below. Ion production was initiated with radioactive sources, either 241 Am (2 μCi) or 210Po (∼40 μCi), and the presence of water vapor31 leads rapidly to the formation of protonated water clusters, H3O+ 3 (H2O)n. A flow of ∼0.0200.050 sLpm (std. conditions L min1, 1 atm and 273 K) of UHP nitrogen (99.999%) or clean air, catalytically purified,32 was maintained through the source for cleanliness. This flow was humidified by exposure to a liquid water reservoir containing a few cm3 of deionized water acidified with a few drops of sulfuric acid. Sulfuric acid was added to the reservoir to suppress the vapor pressure of any basic impurities such as ammonia. Ions exit the source through a hole (diameter ∼0.7 cm) in a 0.07 cm thick, 2 cm diameter plate and enter the drift region. They are drawn from the source due to fringing fields near the exit aperture of the ion source. The drift region is enclosed in a small glass manifold and consists of a gap of 1-to-1.5 cm between the exit plate of the ion source and the entrance plate to the curtain region. A voltage of typically 1.4 kV is applied across it resulting in an ionmolecule reaction time of ∼0.7 ms. Sample flow of ∼1 sLpm enters the drift region via a port in the manifold and exits via an opposing port. A third port on the manifold connects to a pressure sensor. The E/N, E is electric field and N is number density, is ∼4 Td (1 Td = 1017 V cm2) in the ambient pressure drift region and ions are essentially at the ambient temperature, and the distribution of protonated water clusters is dominated by clusters n = 5 to 8: H3O+(H2O)58.33 After the
Figure 1. AmPMS ion source, gas entrance and gas exit ports, mass spectrometer inlet, and CDC with octopole ion guide. Schematic depiction of the instrument as deployed for NCCN showing the sampling arrangement, the three way valve, a small flow to the pump, and the catalytic converter. Inset is blowup showing source, drift region, and curtain region.
reagent ions react with amines and ammonia, they and the product ions are drawn into the curtain region. The curtain region is ∼0.45 cm long and is purged with dry nitrogen (∼0.1-to-0.2 sLpm) via a small Teflon manifold (i.d. 0.5 cm) to minimize the amount of water vapor entering the system. An electric field draws ions across the curtain region (E/N = 2-to-4 Td) to the ion sampling orifice, where a portion enters the CDC and vacuum system via a 125 μm hole in a ∼130 μm thick stainless steel disk. Ions pass through the CDC, a region about 6 cm long at a pressure of ∼0.5 hPa maintained with a drag pump (MDP5011, Alcatel), where they are stripped of water ligands. The CDC has an octopole ion guide to center ions onto a 3-to-4 mm orifice, the entrance to the first chamber of the MS system. The average E/N in the CDC is ∼40 Td, and the ionized (protonated) trace species that were monitored had very few water ligands with the exception of the reagent ions: the water-proton clusters did not fully detach water ligands with the most abundant water proton cluster at H3O+•H2O. Note that the octopole ion guide likely causes the ions to experience a range of E/N, both higher and lower than the average value. 8882
dx.doi.org/10.1021/es201819a |Environ. Sci. Technol. 2011, 45, 8881–8888
Environmental Science & Technology Next the ions enter a differentially pumped quadrupole mass spectrometer vacuum system. For deployment in the 2009 Atlanta campaign this system was a Georgia Tech mass spectrometer (300 L/s turbos, 16 mm OD rods) which was equipped with an octopole ion guide in the first chamber at ∼3 103 hPa.34 For AC testing and calibrations and the UM Teflon chamber measurements the vacuum system was equipped with larger pumps, two 1000 L/s turbo pumps (Varian Inc.), and the first chamber was at ∼0.5-to-1 103 hPa and had a series of ion lenses rather than an octopole ion guide. The second vacuum chambers held the quadrupole mass analyzers at low pressures (∼5 105 and ∼105 hPa, for Atlanta and AC and UM, respectively.) Note that the ion throughputs of the two systems were not evaluated, and mass dependent effects were assumed to be negligible. See the SI (Supporting Information) for more discussion of mass dependent effects. Gas Sampling. The gas sample flow rate was ∼0.8 (UM chamber) and 1.5 (ambient air, Atlanta 2009) sLpm drawn through a port in the glass manifold. Flow exiting the manifold was metered downstream with a mass flow meter positioned before a pumping line. For the Atlanta work, the sample gas lines were 1/4” OD Teflon tubing in two sections: a ∼40 cm length between a Teflon three-way valve on the catalytic converter (see below) and a ∼2 m length that ran up and out the trailer and was about 2 m above the ground. A small polyethylene cone to divert rain from the tip of the inlet was attached over the end of this tube for the field instrument. For the UM chamber and AC experiments, the three-way valve was mounted directly on the glass manifold, and the sample line was ∼1 m of 6.2 mm Teflon tubing. For some AC tests of ambient air, this was extended to 8 m. Catalytic Converter. The instrument background was measured by passing the sample gas through a catalytic converter, a ∼30 cm length of 1/4” OD stainless steel tube containing Pt on alumina beads held at 350 °C. The sampling lines and converter are schematically depicted in the lower part of Figure 1. For the field experiments, a flow of ∼0.030 sLpm continually flushed the catalytic converter. For the UM and AC work, the catalytic converter was rebuilt with inlet and outlet tubes crossed so that the hot gas exiting the converter could transfer heat to the gas entering it. Calibrations. The sensitivity of the instrument was evaluated (post-Atlanta mission) for a few amines and some VOCs by evaporation of aqueous solutions of known composition. Small aliquots (110 μL) of dilute aqueous solutions containing one to three compounds were injected through a port in a glass vessel or a large Teflon tee in series with the sample flow. With known amounts of compounds entrained in a flow of clean gas, sensitivities can be obtained from integrated count rates.32 The gas sample flow of 0.31.0 sLpm of catalytically cleaned room air was at ambient relative humidity. Solutions were prepared gravimetrically, and by serial dilution their concentrations were known to an accuracy of 10%. One solution containing methanol, acetone, and 3-pentanone (at concentrations of ∼0.1 to 0.01 M, reagent grade, VWR) and a few solutions with amines (∼ 104 M, methyl and dimethyl amine from ∼40 wt % solutions, Sigma Aldrich; 4-picoline, VWR, 97%) were used. The concentrations of the reagent MA and DMA solutions were verified by titration with known HCl solutions. The amine calibrations were quite variable, presumably due to interactions with the room temperature glass and Teflon surfaces; several trials (3 to 5) were needed before results were reproducible. Even then, they exhibited variabilities at ∼30%, much greater than the typical ∼10% variabilities for methanol
ARTICLE
Table 1. Calibrations of the Laboratory Instrument Using the Evaporation Methodb P.A. (kcal mol1)
Sx/Smax
methanol
180
4 104
acetone 3-pentanone
194 200
3 103 2 102
methyl amine
215
∼0.2a
dimethyl amine
222
∼0.15a
4-picoline
226
∼0.2a
compound
a
As discussed in the text and SI, the calibrations for these compounds were biased low; the Sx/Smax for the laboratory and ambient data for all the amines including methyl amine were assumed to be unity, consistent with the Sunner et al.37 data. b tion was 0.66 to 1.2 ms. Smax according to eq 1 with k = 2 109 cm3 s1.
and acetone.32 Furthermore, the signals ∼30 min after the injection were observed to increase slightly with each trial, indicating significant sticking. Significant signal for the amines existed when clean air was flowing through the calibration cell several days after their calibrations. The methanol-acetonepentanone calibrations did not require several trials for reproducible behavior, and the system was flushed in ∼3 min.
3. RESULTS AND DISCUSSION Calibrations. Shown in Table 1 are the compounds, their proton affinities,35,36 (also see the SI), and their sensitivities expressed as a fraction of the maximum sensitivity, in Hz pptv1
Smax ¼ S0 kt ð2:4 107 cm-3 pptv -1 Þ
ð1Þ
where S0 is the reagent ion signal, k is the ion molecule rate coefficient = 2 109 cm3 s1, and t is the ion drift time (∼1 ms, see the SI.) The factor in parentheses is the number density for a mixing ratio of 1 pptv for 1 atm and 300 K, typical operating conditions for AmPMS. Equation 1 is an approximation of one presented below and is accurate for reactant concentrations up to about 2000 pptv. The observations presented in the preceding section indicate that the calibration procedure did not capture the full allotment of amine and thus results were biased low; the Sx/Smax for the atmospheric data for all the amines including methyl amine were assumed to be unity, consistent with the Sunner et al.37 data. Note that a rough calibration for DMA with the UM chamber discussed below and a calibration for ammonia during NCCN (SI) bolster the assumption that Sx = Smax for the amines. The dependence of sensitivity on PA shown in Table 1 is consistent with previous work. Sunner et al.37 showed that the reactivity of large H3O+(H2O)n clusters with VOCs has a relatively sharp dropoff for compounds with PA below about 204 kcal/mol, the PA for ammonia. This steep dependence simplifies mass spectra of ambient air because highly abundant species such as methanol, acetaldehyde, and acetone are not readily ionized because of their low PA. Calibrations lead to sensitivities that incorporate both ionization efficiency and relative ion transmission with respect to 37 u, the main mass where the reagent ion was observed. The relative ion transmissions of the field MS and the MS instrument in the laboratory are not known, and for simplicity they are assumed to be unity. Future calibration work with an instrument of known 8883
dx.doi.org/10.1021/es201819a |Environ. Sci. Technol. 2011, 45, 8881–8888
Environmental Science & Technology
ARTICLE
relative ion transmission should allow for a better determination of the dependence of sensitivity on PA. Specificity and Detectable Compounds. Because compounds with low proton affinities (<200 kcal/mol) do not react fast with the reagent ions, there is a limited class of compounds that are detected efficiently with AmPMS: amines, amides, imides, etc. The alkyl amines: methyl-, C2-, C3-, C4-, C5-, and C6- amines were detected at m/z = 32 u, 46 u, 60 u, 74 u, 88 u, and 102 u. Note that structure information is not meant to be conveyed by these assignments as many isomers are possible. For example, 46 u could be either protonated ethyl or dimethyl amine but is not likely to be influenced by formamide which has a proton affinity of 196 kcal/mol.35,36 Previous work4 shows that dimethyl amine, DMA, dominates. Also, amides with more than 1 carbon may have high enough proton affinities35,36 that they could be detected at the amine masses for 60 u and greater: for example, 74 u could be a C-3 amide or a C-2 secondary amide. Imines (CdN bond) and imides (cyclic) would not interfere with amine detection, as they generally differ from amines by 2 u. For standard conditions of the AmPMS, these compounds give ions that the CDC dehydrates to primarily the protonated parent compound. For example, ammonia gave 9095% NH4+, the remainder at NH4+•H2O. The amines were likewise dehydrated. The reagent ion, heavily hydrated H3O+, was dehydrated to mostly H3O+•H2O at 37 u (∼10% each at 19 and 55 u) for the Atlanta campaign, while dehydration was more pronounced at times for UM and AC experiments; H3O+ at 19 u could be up to 50%. The amount of hydration of H3O+ can be greatly affected by small changes in the E/N within the CDC. There was no evidence for fragmentation of low molecular weight amines. Yet, ionization of high molecular weight amine/ amide compounds could yield fragments that would interfere with detection of low molecular weight amines, as suggested by a reviewer. Whether high molecular weight base compounds are abundant enough to interfere in this way is not known; however, a recent review suggests that the low molecular weight alkyl amines are the dominant amine species.4 Concentrations from Ion Signals. The most abundant species that reacts readily with the hydrated clusters is ammonia which can be present at several ppbv and thus NH4+ can become a prominent ion. Viggiano et al.38 showed that amines react readily with NH4+(H2O)n clusters so that the sensitivity of AmPMS for amines is not significantly impacted when the reactant ion becomes ammonium. However, S0 must include the NH4+ ion count rate in addition to that for the water clusters H3O+(H2O)02, abbreviated as SH3O+. With SY+ , S0 and S0 = SH3O+ + SNH4+, amine concentrations [Y] were extracted from the net signals for the amine, SY+, using eq 2 and assuming a maximum sensitivity, i.e., k = 2 109 cm3 s1 ½Y ¼ ðSYþ =S0 Þ=kt in molecule cm3
ð2Þ
½Y ¼ SYþ =Smax in pptv, using ð1Þ Ammonia concentrations were extracted from the signals using the following equation ½NH3 ¼ ln½ðSH3 Oþ þ SNH4 þ Þ=SH3 Oþ =kt
ð3Þ
Note that [NH3] from eq 2 agrees well with (3) when ammonia is less than ∼2000 pptv. For a large portion of the NCCN campaign, the SNH4+ count rate including background was 15% or less of S0. However, background count rates for ammonia for NCCN were difficult to separate from the ambient signal because
Figure 2. AmPMS sampling chamber air without added amines. Levels of amines at 100 s of pptv seemed unavoidable even with repeated flushfill heat cycles of the chamber and Teflon bag. Instrument background determination was initiated at 15:52 and during this zero AmPMS sampling was changed to room air and then to dry N2.
of the sampling line arrangement. See the SI for more discussion of the detection of ammonia in Atlanta for a seven day period, the ion chemistry, and instrument sensitivity. The SI also presents the results of the application of (2) and (3) for ammonia and amine concentrations when ammonia was relatively high in AC measurements of urban air in winter. Background. The measurement of the amine background signals of the AmPMS had a relatively fast response time of one minute or less: a set of data is shown in Figure 2. During this zero, the system was taken off the UM chamber and sampled room air whereupon changes in the backgrounds for some signals are apparent. Room air contained much higher levels of amines and other compounds and more water vapor than the chamber air. Whether the converter fully eliminates amines is a concern, but it is clear that composition of the sample air most likely water vapor influences the backgrounds. This is illustrated when (@ 16:12 in the figure) dry UHP nitrogen was sent through the catalytic converter, and a marked change in the background levels for several compounds is apparent. The instrument background for many compounds depended on the relative humidity of purified air (catalytically cleaned) sent in as the sample gas. This effect was not explored further, but frequent zeroing is recommended especially when the RH of sampled gas undergoes a large change. Instrument background signals for NCCN were determined by diverting the flow through the catalytic converter for ten minutes every two hours. It was observed that the background levels for the amines were highly temperature dependent. Increases were on the order of a factor of 5 as outside temperature rose from 20 to 37 °C. The periodic ten minute zeroes were used and were linearly interpolated during the two hour intervening times. For the C-2, C-3, and C-6 amines, uncertainties in background levels were low enough such that single digit pptv ambient levels were readily detected on the minute time scale. The background levels for many compounds were well determined in the ten minute zero: signals reach a plateau in ∼1 min after the zero was initiated. Methyl amine was apparently very low for the entire campaign, perhaps 1 pptv on a background that had only a modest temperature dependent variation, from 2-to-5 pptv. For the C-4 and C-5 amines, the background levels were quite high, 50100 pptv at 4 a.m. when temperature was generally <22 °C increasing to as much as ten times this at high temperatures. The ambient 8884
dx.doi.org/10.1021/es201819a |Environ. Sci. Technol. 2011, 45, 8881–8888
Environmental Science & Technology
ARTICLE
Figure 3. Chamber air sampled during UV photolysis of ozone (10 ppbv) with ∼20 ppbv DMA, 10 ppbv SO2, 20% RH with the balance N2. DMA vapor was injected concomitant with initiation of photolysis and DMA levels are seen to rise quickly and then decay away with the apparent DMA oxidation products growing. Particle count rates from a PHA CPC43 are shown on the right axis; they were binned into two size ranges, roughly 23 nm and ∼4 nm.
signals at 74 and 88 u were dominated by background levels, and only several hour averages are presented for NCCN. For most nights a long time zero was established from 9 p.m. to 4 a.m. local standard time. This was done because the background level for ammonia was not being obtained in the ten minute zeroes. This procedure resulted in the ammonia background level for 4 a.m.; however, its dependency on temperature was not known. Therefore, ambient ammonia levels are not reported for the NCCN campaign, except for a short period discussed in the SI. Chamber Data. Air was sampled at ∼0.8 sLpm from a ∼1 m3 capacity Teflon bag filled with purified air, humidified to ∼20% relative humidity, with trace amounts of SO2 (10 ppbv), O3 (10 ppbv), and an amine (up to 30 ppbv.) This mixture was then irradiated with 254 nm light. Figure 2, discussed above, shows a typical ‘zeroing’ of the AmPMS when sampling chamber air containing several hundred pptv of dimethyl amine (46 u) and an amine/amide that gives 60 u along with other compounds: ∼100 pptv ammonia and methylamine; ∼200 pptv on 74 u; and a few hundred pptv equivalence that gives 44 u, possibly an amide breakup ion at 44 u, CONH2+ or cyanuric acid.39 The zeroing apparatus worked well for all these compounds monitored in the chamber experiments, and its superior performance over the Atlanta deployment could be due to a number of things: more uniform catalysis conditions; the N2 for the source flow and curtain gas was taken from a liquid nitrogen gas pack; a lower and a more uniform in time instrument temperature for the UM work; elimination of tubing between three-way valve and ion drift region; and low exposure to ambient air and particulate matter. Figure 3 shows concentration data as a function of time during an experiment where dimethyl amine vapor was injected into the chamber and signal at 46 u rose rapidly; it was estimated that 20 ppbv was present in the chamber, in good agreement with eq 2 and the assumed maximum efficiency. The amount of amine added was estimated from the partial pressure of DMA over a 40 wt % (14 m) DMA solution at 0 °C using a Henry’s law coefficient of 150 m/atm.41,40 The signals for 60 and 74 u rise as DMA decreases upon OH-initiated oxidation. These signals are likely oxidation products of DMA, possibly N-methylformamide and diformyl amide, respectively.42 A flush and fill cycle began at about 17:25, and large variations in signals are evident. Ammonia levels were low (∼100 pptv) and were detected along with the amines. Particle data using an ultrafine condensation nucleus
Figure 4. The mixing ratios of two alkyl amines (a) 60 u, trimethyl amine and (b) 102 u, triethyl amine (or isomers or isobaric amides, etc.) plotted versus temperature (5 min averages from NCCN.) The three parameter curves shown are for illustration purposes where a constant mixing ratio has been added to an exponential: A+Bexp(-C(1/T-1/To)). With To = 308.15 K, the values are as follows: (a) A = 3 pptv, B = 20 pptv, C = 15,000 K and (b) A = 2.5 pptv, B = 45 pptv, C = 30,000 K.
counter43 is shown on the left axis, and a large number of particles were formed upon UV irradiation. An exponential decay with a time constant of 3 104 s1 is shown as the dotted line. With this first order loss rate, a rate coefficient for OH + dimethylamine of 6.5 1011 cm3 s1,42 an OH concentration of 5 106 cm3 is derived. The growth of probable amide species at 60 and 74 u are prominent during the ∼1 h photolysis. The signals exhibit nonexponential behavior near the end of experiment when the bag was close to empty. NCCN Field Data. Amines (and/or the isobaric amides) and other compounds were monitored with AmPMS during NCCN in Atlanta at Jefferson street station in 2009. The amine concentrations had large variations during the campaign with many days exhibiting a pronounced diurnal cycle: peaking in the late afternoon, correlated with temperature. This is shown for 60 and 102 u in Figure 4 as a plot of five minute averages versus ambient temperature where an equation with an exponential dependence on temperature is shown. The amine species that is detected at 102 u is especially sensitive to temperature, and the exponential portion is quite pronounced for this data. The exponential dependence suggests a very high activation energy (∼60 kcal/mol for 102 u) that may be consistent with aminium salt evaporation.10 MA and DMA (32 and 46 u) had low correlations to temperature: linear R2 values of 0.0 and 0.06, respectively. The ambient levels of methyl and dimethyl amine were found to be less than 5 and 10 pptv, respectively, consistent with previous measurements that indicate they are generally <1% of ammonia 8885
dx.doi.org/10.1021/es201819a |Environ. Sci. Technol. 2011, 45, 8881–8888
Environmental Science & Technology
ARTICLE
Figure 5. Five minute average amine data over a ∼15 day period in Atlanta 2009 for methyl-, dimethyl-, trimethyl-, and triethyl amine or isomers or isobars (amides, etc.) Relative humidity (right axis) is shown in the top plot, and temperature is shown in the bottom plot.
levels.4 The dominant amines (or isobars) observed were trimethyl amine and a six carbon amine. These compounds had late afternoon averages of 15 and 25 pptv, respectively, showed a diurnal cycle, and were temperature dependent. The trimethyl amine observations are somewhat lower than the observations by Selegri et al.25 who report median trimethyl amine levels that ranged from 50 to 80 pptv with peak values in the late afternoon. Representative 5 min averages for ∼15 days (23Jul09 to 07Aug 09) are shown in Figure 5 with nominal concentration in pptv. The instrument background was measured for an extended period on most nights, thus, the data are primarily for 0400 to 2100 local time. A prominent diurnal cycle is displayed for many days. Days with precipitation and overcast conditions showed low levels of all amines (see the 100% humidity time periods.) A few overnight measurements are shown, and generally low levels are exhibited except for around 10 p.m. on 24Jul09 where the amines were elevated. The diurnal hourly averages of the amine data from 23Jul09 to 25Aug09 is shown in Figure 6. The diurnal hourly averages for methyl amine were very low for this time period and essentially zero (<0.2 pptv). Averages for dimethyl amine ranged from 0.5 to 2 pptv peaking in the early afternoon, while trimethyl amine ranged from ∼4 pptv in the early morning to a peak of around 15 pptv in the afternoon. The 4 and 5 carbon amines had overall average levels of about 4 pptv (4 a.m. to 8 p.m.). The background count rates were equivalent to many 100s of pptv so these levels are highly uncertain with upper limits (set as ∼5% of the average background) of 10 pptv for 4 a.m. to noon and 30 pptv for noon to 8 p.m. The most abundant amine detected in NCCN was a 6 carbon amine possibly triethyl amine, and it also exhibited a strong diurnal cycle with average values of ∼3 pptv in the morning peaking at 25 pptv in the late afternoon. Note that signal at 102 u could also arise from a number of amides. Finally, the 10 p.m. hourly average has a small number of samples which was dominated by the ∼10 p.m. 24Jul09 event.
Figure 6. NCCN hourly diurnal averages (LST) for ambient air in Atlanta late July and August, 2009. The 4 and 5 carbon amines 8 h averages are shown. Uncertainty for these two species are on the order of 10 pptv and 30 pptv for the two time periods 4 a.m. to noon and noon to 8 p.m., respectively. Isomers and isobaric species are possible.
The C6 amine was one of the most abundant amines that AmPMS observed during NCCN. A previous observation of particulate C6 amine at the NCCN site has been reported7 where mass spectra of particles with diameters >100 nm were determined to have the signature of triethyl amine. Furthermore, the frequency of the observation of the triethyl amine marker was anticorrelated with relative humidity which qualitatively agrees with the temperature dependency of gas phase C6 amine depicted in Figure 4 assuming gas-particle partitioning is temperature (and water content) dependent. It appears that the C6 amine plays a role in particle growth. The diurnal cycle/RH/temperature variation of the amine signal attributed to C6 amine at another site showed behavior that is consistent with partitioning between gas and particles.8 Particle nucleation at the site has been tied to SO2 and its oxidation product H2SO4,44,11 and recently a weak sensitivity to 8886
dx.doi.org/10.1021/es201819a |Environ. Sci. Technol. 2011, 45, 8881–8888
Environmental Science & Technology ammonia45 for nucleation was observed. Preliminarily for NCCN, and in general previously observed at this site, nanoparticle concentrations indicative of new particle formation happen during sunlit conditions and with ppbv levels of SO2 present. Amines were generally present at the 10 pptv level or higher during the new particle formation events, which may be a level high enough to significantly affect new particle formation, as is predicted for ammonia using ternary classical theory14 and for DMA using ab initio calculations.17 A full correlation study of the 2009 particle observations to the amine data will be presented in the future. Detailed measurements of ambient, nonfeedlot related, gasphase amines and other basic nitrogen species is rare so little guidance is afforded from previous work.4 The ambient data from Atlanta suggest temperature dependent sources such as soil or vegetation; they may be strongly temperature dependent. The amines may also be semivolatile with respect to aerosol and gasaerosol partitioning might be an explanation for the observed diurnal cycle/temperature dependence. A local NH3 source for this site that is temperature or sunlight dependent has been proposed.23 Future work will focus on the following: 1) determining the source of the temperature dependent background signals in particular for ammonia so that measurements in the field will be more reliable, 2) calibration standards for amines, amides, etc., 3) exploring the use of high resolution (Δm < 0.02 u) mass spectrometry to ascertain the relative importance of amides and amines, and 4) more measurements to assess the importance of amines in other locations.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional information on the ionization process, the effects of high ammonia on the instrument’s sensitivity to amines, and an ammonia intercomparison for NCCN. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 612-330-1620. Fax: 612-330-1649. E-mail: hansondr@ augsburg.edu.
’ ACKNOWLEDGMENT This research was supported by the National Science Foundation (grant nos. AGS-0943721 and ATM-0506674). We also appreciate the support of John Jansen (The Southern Company) in providing us with access to the site and with the sampling laboratories and Eric Edgerton (Atmospheric Research and Analysis, Inc.) for his assistance with preparing the site and for the meteorological and NH3 data. R. Stickel’s assistance with the computer control during the Atlanta campaign was essential. J. A. Neuman is gratefully acknowledged for the loan of an ammonia calibration source during the Atlanta field campaign. Conversations with J. Walker are gratefully acknowledged. ’ REFERENCES (1) Cadle, S. H.; Mulawa, P. A. Low molecular weight aliphatic amines in exhaust from catalyst-equipped cars. Environ. Sci. Technol. 1980, 14, 718–723.
ARTICLE
(2) Smith, M. S.; Francis, A. J.; Duxvury, J. M. Collection and analysis of organic gases from natural ecosystems: application to poultry manure. Environ. Sci. Technol. 1977, 1, 51–55. (3) Schlesinger, W. H.; Hartley, A. E. A global budget for atmospheric NH3. Biogeochemistry 1992, 15, 191–211. (4) Ge, X.; Wexler, A. S.; Clegg, S. L. Amines in the Atmosphere I. Review. Atmos. Environ. 2011, 45, 524–546. (5) Kanakidou, M.; Seinfeld, J. H.; Pandis, S. N.; Barnes, I.; Dentener, F. J.; Facchini, M. C.; Van Dingenen, R.; Ervens, B.; Nenes, A.; Nielsen, C. J.; Swietlicki, E.; Putaud, J. P.; Balkanski, Y.; Fuzzi, S.; Horth, J.; Moortgat, G. K.; Winterhalter, R.; Myhre, C. E. L.; Tsigaridis, K.; Vignati, E.; Stephanou, E. G.; Wilson, J. Organic aerosol and global climate modelling: a review. Atmos. Chem. Phys. 2005, 5, 1053–1123. (6) M€akel€a, J. M.; Yli-Koivisto, S.; Hiltunen, V.; Seidl, W.; Swietlicki, E.; Teinil€a, K.; Sillanp€a€a, M.; Koponen, I. K.; Paatero, J.; Rosman, K.; H€ameri, K. Chemical composition of aerosol during particle formation events in boreal forest. Tellus B 2001, 53, 380–393. (7) Angelino, S.; Suess, D. T.; Prather, K. A. Formation of aerosol particles from reactions of secondary and tertiary alkylamines: characterization by aerosol time-of-flight mass spectrometry. Environ. Sci. Technol. 2001, 35, 3130–3138. (8) Pratt, K. A.; Hatch, L. E.; Prather, K. A. Seasonal volatility dependence of ambient particle phase amines. Environ. Sci. Technol. 2009, 43, 5276–5281. (9) Silva, P. J.; Erupe, M. E.; Price, D.; Elias, J.; Malloy, Q.; Li, Q.; Warren, B.; Cocker, D. R. Trimethylamine as precursor to secondary organic aerosol formation via nitrate radical reaction in the atmosphere. Environ. Sci. Technol. 2008, 42, 4689–4695. (10) Smith, J. N.; Barsanti, K. C.; Friedli, H. R.; Ehn, M.; Kulmala, M.; Collins, D. R.; Scheckman, J. H.; Williams, B. J.; McMurry, P. H. Observations of aminium salts in atmospheric nanoparticles and possible climatic implications. Proc. Natl. Acad. Sci. 2010, 15, 6634-6639, doi:10.1073/pnas.0912127107. (11) Smith, J. N.; Moore, K. F.; Eisele, F. L.; Voisin, D.; Ghimire, A. K.; Sakurai, H.; McMurry, P. H. Chemical composition of atmospheric nanoparticles during nucleation events in Atlanta. J. Geophys. Res. 2005, 110, D22S03. (12) Bzdek, B. R.; Ridge, D. P.; Johnston, M. V. Amine exchange into ammonium bisulfate and ammonium nitrate nuclei. Atmos. Chem. Phys. 2010, 10, 3495–3503. (13) Barsanti, K. C.; McMurry, P. H.; Smith, J. N. The potential contribution of organic salts to new particle growth. Atmos. Chem. Phys. 2009, 9, 2949–2957. (14) Coffman, D. J.; Hegg, D. A A preliminary study of the effect of ammonia on particle nucleation in the marine boundary layer. J. Geophys. Res. 1995, 100, 7147–71601995. (15) Weber, R. J.; Marti, J.; McMurry, P. H.; Eisele, F. L.; Tanner, D. J.; Jefferson, A. Measured atmospheric new particle formation rates: Implications for nucleation mechanisms. Chem. Eng. Commun. 1996, 151, 53–64. (16) Ball, S. M.; Hanson, D. R.; Eisele, F. L.; McMurry, P. H. Laboratory studies of particle nucleation: Initial results for H2SO4, H2O, and NH3 vapors. J. Geophys. Res. D 1999, 104, 23709–23718. (17) Kurten, T.; Loukonen, V.; Vehkamaki, H.; Kulmala, M. Amines are likely to enhance neutral and ion-induced sulfuric acid-water nucleation in the atmosphere more effectively than ammonia. Atmos. Chem. Phys. 2008, 8, 7455–7476. (18) Aky€uz, M. Simultaneous determination of aliphatic and aromatic amines in ambient air and airborne particulate matters by gas chromatography-mass spectrometry. Atmos. Environ. 2008, 42, 3809–3819. (19) Huang, G.; Hou, J.; Zhou, X. L. A measurement method for atmospheric ammonia and primary amines based on aqueous sampling, OPA derivatization and HPLC analysis. Environ. Sci. Technol. 2009, 43, 5851–5856. (20) von Bobrutzki; et al. Field inter-comparison of eleven atmospheric ammonia measurement techniques. Atmos. Meas. Tech. 2010, 3, 91–1122010. (21) Norman, M.; Spirig, C.; Wolff, V.; Trebs, I.; Flechard, C.; Wisthaler, A.; Schnitzhofer, R.; Hansel, A.; Neftel, A. Intercomparison of 8887
dx.doi.org/10.1021/es201819a |Environ. Sci. Technol. 2011, 45, 8881–8888
Environmental Science & Technology ammonia measurement techniques at an intensively managed grassland site. Atmos. Chem. Phys. 2009, 9, 2635–2645. (22) Ellis, R. A.; Murphy, J. G.; Pattey, E.; van Haarlem, R.; O’Brien, J. M.; Herndon, S. C. Characterizing a Quantum Cascade Tunable Infrared Laser Differential Absorption Spectrometer (QC-TILDAS) for measurements of atmospheric ammonia. Atmos. Meas. Tech. 2010, 3, 397–406. (23) Nowak, J. B.; Huey, L. G.; Russell, A. G.; Tian, D.; Neuman, J. A.; Orsini, D.; Sjostedt, S. J.; Sullivan, A. P. ; Tanner, D. J.; Weber, R. J.; Nenes, A.; Edgerton, E.; Fehsenfeld, F. C. Analysis of urban gas phase ammonia measurements from the 2002 Atlanta Aerosol Nucleation and Real-Time Characterization Experiment (ANARChE). J. Geophys. Res. 2006, 111, D17308, doi:10.1029/2006JD007113. (24) Norman, M.; Hansel, A.; Wisthaler, A. O2+ as reagent ion in the PTR-MS instrument: detection of gas-phase ammonia. Int. J. Mass. Spectrom. 2007, 265, 382–287. (25) Sellegri, K.; Umann, B.; Hanke, M.; Arnold, F. Deployment of a ground-based CIMS apparatus for the detection of organic gases in the boreal forest during the QUEST campaign. Atmos. Chem. Phys. 2005, 5, 357–372. (26) Eisele, F. First tandem mass spectrometric measurement of tropospheric ions. J. Geophys. Res. 1988, 93, 716–724. (27) Solomon, P. A., et al. , Overview of the 1999 Atlanta Supersite Project, J. Geophys. Res. D 2003, 107, doi:10.1029/2001JD001458, 108. (28) Baumann K.; Ift, F.; Zhao, J.; Chameides, W. L. Discrete measurements of reactive gases and fine particle mass and composition during the 1999 Atlanta Supersite Experiment. J. Geophys. Res. D 2003, 107, doi:10.1029/2001JD001210, 108. (29) Zhang J.; Chameides, W. L. ; Weber, R.; Cass, G.; Orsini, D.; Edgerton, E.; Jongejan, P.; Slanina, J. An evaluation of the thermodynamic equilibrium assumption for fine particulate composition: Nitrate and ammonium during the 1999 Atlanta Supersite Experiment. J. Geophys. Res. D 2003, 108, 8414, doi:10.1029/2001JD001592. (30) Hanson, D. R.; Kosciuch, E. The NH3 mass accommodation coefficient for uptake onto sulfuric acid solutions. J. Phys. Chem. A 2003, 107, 2199–2208. (31) Howard, C. J.; Bierbaum, V. M.; Rundle, H. W.; Kaufman, F. Kinetics and mechanism of the formation of water cluster ions from O2+ and H2O. J. Chem. Phys. 1972, 57, 3491. (32) Hanson, D. R.; Koppes, M.; Stoffers, A.; Harsdorf, R.; Edelen, K. Proton transfer mass spectrometry at 11 hPa with a circular glow discharge ion source: Sensitivities and applications. Int. J. Mass Spectrom. 2009, 282, 28–37. (33) Lau, Y. K.; Ikuta, S.; Kebarle, P. Thermodynamics and kinetics of the gas-phase reactions: H3O+(H2O)n-1 + H2O = H3O+(H2O)n. J. Am. Chem. Soc. 1982, 104, 1462–1469. (34) Slusher, D. L.; Pitteri, S. J.; Haman, B. J.; Tanner, D. J.; Huey, L. G. A chemical ionization technique for measurement of pernitric acid in the upper troposphere and the Polar boundary layer. Geophys. Res. Lett. 2001, 28 (20), 3875–3878. (35) Hunter, E. P.; Lias, S. G. Evaluated gas phase basicities and proton affinities of molecules: An Update. J. Phys. Chem. Ref. Data 1998, 27, 413–652. (36) Proton Affinity Evaluation. In NIST Chemistry WebBook, NIST Standard Reference Database Number 69, Linstrom, P. J., Mallard, W. G., Eds.; National Institute of Standards and Technology, Gaithersburg MD, 20899. http://webbook.nist.gov (accessed May, 2011). (37) Sunner, J.; Nicol, G.; Kebarle, P. Factors determining relative sensitivity of analytes in positive mode Atmospheric Pressure Ionization Mass Spectrometry. Anal. Chem. 1988, 60, 1300–1307. (38) Viggiano, A. A.; Robert, A.; Morris, A.; Dale, F.; Paulson, J. F. Tropospheric reactions of H+(NH3)m(H20)n with pyridine and picoline. J. Geophys. Res. 1988, 93, 9534–9538. (39) Barnes, I.; Solignac, G.; Mellouki, A.; Becker, K. H. Aspects of the atmospheric chemistry of amides. Chem. Phys. Chem. 2010, 11, 3844– 3857.
ARTICLE
(40) Christie, A. O.; Crisp, D. J. Activity coefficients of the n-primary, secondary and tertiary aliphatic amines in aqueous solution. J. Appl. Chem. 1967, 17, 11–14. (41) Sander, R. Henry’s Law coefficients. In NIST Chemistry WebBook, NIST Standard Reference Database Number 69, Linstrom, , P. J., Mallard, , W. G., Eds.; National Institute of Standards and Technology, Gaithersburg MD, 20899. http://webbook.nist.gov (accessed May, 2011). (42) Atkinson, R.; Perry, R. A.; Pitts, J. N. Rate constants for the reactions of the OH radical with (CH3)2NH, (CH3)3N, and C2H5NH2 over the temperature range 298426 K. J. Chem. Phys. 1978, 68, 1850–1853. (43) Stolzenburg, M. R.; McMurry, P. H. An ultrafine aerosol condensation nucleus counter. Aerosol Sci. Technol. 1991, 14, 48–651991. (44) Woo, K. S.; Chen, D. R.; Pui, D. Y. H.; McMurry, P. H. Measurement of Atlanta aerosol size distributions: Observations of ultrafine particle events. Aerosol Sci. Technol. 2001, 34 (1), 75–87. (45) McMurry, P. H., et al. A criterion for new particle formation in the sulfur-rich Atlanta atmosphere, J. Geophys. Res. 2005, 110, D22S02, doi:10.1029/2005JD005901.
8888
dx.doi.org/10.1021/es201819a |Environ. Sci. Technol. 2011, 45, 8881–8888
ARTICLE pubs.acs.org/est
Observation of ClNO2 in a Mid-Continental Urban Environment Levi H. Mielke,† Amanda Furgeson, and Hans D. Osthoff* Department of Chemistry, University of Calgary, 2500 University Drive NW, Calgary, Alberta T2N 1N4, Canada
bS Supporting Information ABSTRACT: In the troposphere, nitryl chloride (ClNO2), produced from uptake of dinitrogen pentoxide (N2O5) on chloride containing aerosol, can be an important nocturnal reservoir of NOx (= NO + NO2) and a source of atomic Cl, particularly in polluted coastal environments. Here, we present measurements of ClNO2 mixing ratios by chemical ionization mass spectrometry (CIMS) in Calgary, Alberta, Canada over a 3-day period. The observed ClNO2 mixing ratios exhibited a strong diurnal profile, with nocturnal maxima in the range of 80 to 250 parts-per-trillion by volume (pptv) and minima below the detection limit of 5 pptv in the early afternoon. At night, ClNO2 constituted up to 2% of odd nitrogen, or NOy. The peak mixing ratios of ClNO2 observed were considerably below the modeled integrated heterogeneous losses of N2O5, indicating that ClNO2 production was a minor pathway compared to heterogeneous hydrolysis of N2O5. Back trajectory analysis using HYSPLIT showed that the study region was not influenced by fresh injections of marine aerosol, implying that aerosol chloride was derived from anthropogenic sources. Molecular chlorine (Cl2), derived from local anthropogenic sources, was observed at mixing ratios of up to 65 pptv and possibly contributed to formation of aerosol chloride and hence the formation of ClNO2.
’ INTRODUCTION The oxides of nitrogen are important trace gas constituents of the troposphere. In urban areas, NOx (= NO + NO2) is emitted mainly in the form of NO from combustion engines used in automobiles and often concentrates near the surface due to inversions, especially in wintertime. High mixing ratios of NOx are of concern because there are links between NOx exposure and adverse health effects (e.g., ref 1), and because NOx promotes formation of secondary pollutants such as ozone (O3) and the peroxycarboxylic nitric anhydrides (PANs, general formula RC(O)O2NO2). Nitrogen oxides are removed from the atmosphere mainly by conversion to nitric acid (HNO3), which deposits onto surfaces and aerosol particulate matter. During daytime, the main HNO3 production (and NOx removal) pathway is the reaction of NO2 with the hydroxyl radical (OH). Nocturnal nitrogen oxide chemistry is dominated by reactions of the photolabile nitrate radical (NO3), produced from reaction of NO2 with O3, and of dinitrogen pentoxide (N2O5), which is in a temperature dependent equilibrium with NO3 and NO2.2,3 NO2 ðgÞ þ O3 ðgÞ f NO3 ðgÞ þ O2 ðgÞ, k1 ¼ 1:2 1013 expð 2450=TÞcm3 molecule1 s1
ð1Þ
NO2 ðgÞ þ NO3 ðgÞ h N2 O5 ðgÞ, k2 ¼ 2:7 1027 expð11000=TÞcm3 molecule1
ð2Þ
Here, temperature (T) is in units of Kelvin. In urban areas, the chemistry of NO3 and N2O5 is often suppressed because NO3 is r 2011 American Chemical Society
rapidly titrated by (freshly emitted) NO, but may still be significant aloft in the residual layer4 or during times of the night when traffic and hence NO emissions are at a minimum. NO3 ðgÞ þ NOðgÞ f 2NO2 ðgÞ, k3 ¼ 1:5 1011 expð170=TÞcm3 molecule1 s1
ð3Þ
At low temperatures and high NOx levels, equilibrium (2) is shifted toward N2O5, and nocturnal HNO3 production (and NOx loss) proceeds mainly by heterogeneous reactions of N2O5.5 het
N2 O5 ðgÞ þ H2 OðaqÞ sf 2HNO3 ðaqÞ
ð4aÞ
In the presence of chloride containing aerosol, reaction 4b competes with NOx loss by reaction 4a: het
N2 O5 ðgÞ þ Cl ðaqÞ sf ClNO2 ðgÞ þ NO3 ðaqÞ
ð4bÞ
The product of reaction 4b, nitryl chloride (ClNO2), was first observed in laboratory experiments by Finlayson-Pitts et al.6 and was initially believed to be of significance only in areas with marine influence, i.e., in the presence of sea salt aerosol.7,8 Formation of ClNO2 can affect air quality in polluted coastal areas,9 because of the lower rate of nocturnal NOx loss relative to reaction 4a and because ClNO2 is sufficiently long-lived at night to photodissociate in the morning to yield NO2 and highly Received: June 8, 2011 Accepted: August 30, 2011 Revised: August 12, 2011 Published: August 30, 2011 8889
dx.doi.org/10.1021/es201955u | Environ. Sci. Technol. 2011, 45, 8889–8896
Environmental Science & Technology
ARTICLE
reactive Cl atoms,10,11 which accelerate photochemical ozone production.12 hv
ClNO2 ðgÞ sf ClðgÞ þ NO2 ðgÞ, jnoon ≈ð40 minÞ1
ð5Þ
In 2006, ClNO2 mixing ratios in ambient air were quantified for the first time along the Texas coast using chemical ionization mass spectrometry (CIMS) with I reagent ion.12 Mixing ratios of ClNO2 were found to exceed 1 part-per-billion by volume (ppbv) in plumes, which suggested that ClNO2 is long-lived at night and that the yield of ClNO2 from reaction 4b was considerably larger than had been predicted. These high yields have since been rationalized by efficient ClNO2 production on aerosol with only partial chloride content8,13,14 and by surfacecatalyzed reaction of N2O5 with HCl.15 The efficiency of reaction 4b observed in coastal areas also suggested that the production of ClNO2 should extend inland and occur wherever nitrogen oxides and sources of aerosol chloride coexist, such as urban areas.12 Recently, Thornton et al.16 observed ClNO2 mixing ratios in excess of 400 parts-per-trillion by volume (pptv) in Boulder, CO. The observed production of ClNO2 1400 km from the nearest coastline implied that a considerable amount of the aerosol chloride had been derived from nonmarine sources.16 Many such sources are known, e.g., biomass burning,17 emissions from cooling towers,18 coal combustion, and waste incineration,19 and the use of road salt in winter.20 In this work, we report measurements of ClNO2 mixing ratios in Calgary, Alberta, Canada, in April 2010. At this location and time of year, production of ClNO2 was favored by low nocturnal temperatures, which shifted equilibrium (2) toward N2O5 and probably increased the N2O5 uptake probability,21 absence of biogenic VOCs that could have consumed NO3, reduced NO3 photolysis rates due to long nights and high daytime solar zenith angles, large NOx emission rates by cars and furnaces, and because the city had applied salt to its roads over the course of the winter. Mixing ratios of ClNO2 were quantified in parallel with those of peroxyacetic and peroxypropanoic nitric anhydride (PAN and PPN) and molecular chlorine (Cl2) by I-CIMS. The observations of ClNO2 reported here are only the second reported at a nonmarine site and the first in Canada. Potential sources of aerosol chloride in the region are discussed.
’ EXPERIMENTAL SECTION Overview of Measurement Locations and Instruments. Calgary is a city of more than 1 million people and is located approximately 70 km east of the Rocky Mountains, 680 km east of Vancouver, and 800 km east of the open ocean (Figure 1) at an elevation of 1100 m above sea level. Measurements were conducted at the University of Calgary (UC), which is located in the northwest quadrant of the city and is surrounded by suburban communities and residential roadways. Auxiliary data were collected at a routine monitoring station (operated by Alberta Environment) located 900 m west of the main measurement site. We sampled from the afternoon of April 16 to the morning of April 17 and during a 2-day consecutive period from the afternoon of April 19 to the early afternoon of April 21, 2010. Two instruments were deployed at the University, the I-CIMS (described below) and a NO/O3 chemiluminescence (CL) “NOx” monitor (Thermo 42TL). The latter was equipped with an internal heated Mo converter to reduce nitrogen oxides to NO and was found in laboratory experiments to not only convert NO2 but also
Figure 1. Map of the sampling region showing the location of the City of Calgary with respect to the ocean. The insert shows the main sample location on the University of Calgary campus. The AB Environment routine monitoring site is located 900 m to the west. The images are from Google Earth.
components of NOz (= NOy NOx = PANs + HONO + HNO3 + NO3 + 2N2O5 + ClNO2 + ...) to NO with high efficiency. The two instruments were housed in the “Penthouse” laboratory located on the roof of the Science B building (lat. 51.07933°N, long. 114.12950°W) and sampled from a common inlet, constructed from Kynar with dimensions of 1/2” outer diameter (o.d.), 3/8” inner diameter (i.d.), and ∼10 m length, which was passed through an open window. ClNO2 has relatively small uptake coefficients22,23 and thus transmits through inlets efficiently. We also did not observe production within the inlet, as we occasionally sampled plumes of NO (in which N2O5 would have been titrated) without observing any attenuation of the ClNO2 signal. The passing efficiency of Cl2 was not assessed during this work and was assumed to be unity; since Cl2 can be scrubbed in inlets,24 the Cl2 data presented in this manuscript are likely lower limits. The tip of the inlet was secured, facing downward, at a height of ∼3 m above the flat rooftop. The total sample flow rate through the common inlet was 4.0 standard liters per minute (slpm), of which 1.0 slpm were sampled by the CL instrument. Both instruments were equipped with polycarbonate filter holders and inline Teflon membrane filters with 2 μm pore size (Pall). Hourly data of meteorology parameters, i.e., wind speed and direction, temperature, relative humidity (RH), and of O3, CO, NO, and NOy mixing ratios collected at the Northwest station during the time of this study were downloaded from the “Clean Air Strategic Alliance” (CASA) Web site.25 Comparison of NO and NOy mixing ratios measured at both locations (Figure S-1 of the Supporting Information, SI) showed that both stations usually sampled the same air mass. Chemical Ionization Mass Spectrometry. The I-CIMS (Figure 2) was purchased from THS instruments and is a compact (width depth height: 21” 40” 52”; weight: 200 lbs) version of, and operated similarly to, the instrument described by Slusher et al.26 and has been partially described elsewhere.27,28 Following sampling of ambient air drawn through the Kynar inlet tubing, the sample flow was split using a PFA Teflon Tee and sampled in part by the CL analyzer and in part by the I-CIMS. A small flow of approximately 38 standard cubic centimeters per minute (sccm) containing photochemically generated 13C-labeled PAN in ultrapure, or zero, air was continuously added to the CIMS sample flow to track PAN matrix 8890
dx.doi.org/10.1021/es201955u |Environ. Sci. Technol. 2011, 45, 8889–8896
Environmental Science & Technology
ARTICLE
Figure 2. Schematic of the University of Calgary I-CIMS and sampling setup. For a detailed description, see text.
effects.26,27,29 The stability of the photochemical source was monitored by periodic flooding of the inlet with N2. The output of the source stabilized over the course of several hours and remained stable (variability 5%) for the duration of the experiments. The sample air was then passed through a section of 1/2” o.d. Teflon tubing heated to 190 °C. This temperature is sufficient to quantitatively dissociate PANs to create peroxyacyl radicals (RC(O)O2) and NO2,30 but is too low to dissociate ClNO2.28 A fraction of the sample air (1.32 slpm) entered the ion molecule reaction region (IMR) through a pinhole, with the remainder being pumped away by a diaphragm pump whose flow was set to 1.68 slpm using a mass flow controller (MFC). In the IMR, the sample air was diluted with N2 passed through a bubbler containing deionized water, which was set to a flow rate of ∼480 sccm. I reagent ions were formed from dissociative electron transfer of CH3I in a 210Po ion source (NRD P-20311000). CH3I (0.5% in N2, ScottMarrin) was delivered at a flow rate of 0.68 sccm and diluted with 1.2 slpm of N2. The pressure in the reaction chamber during ambient measurements presented in this manuscript was 21 Torr. The main reactions in the IMR leading to quantifiable ions were as follows:31,26,32 H2 O
ClNO2 þ I sfðClNO2 ÞI ðm=z 208 and m=z 210Þ
ð6Þ
H2 O
RCðOÞO2 þ I sf IO þ RCO2 ðm=z 59, m=z 61, m=z 73, etc:Þ
ð7Þ
H2 O
Cl2 þ I sfðCl2 ÞI ðm=z 197, m=z 199, and m=z 201Þ
ð8Þ The anions generated in the IMR were transferred (via a pinhole) to the collisional dissociation chamber (CDC), which was operated at 0.5 Torr. The function of the CDC was to break up water cluster ions.26 The anions were then passed through a second octopole ion guide region, operated at 103 Torr, to improve the signal-to-noise ratio, mass-selected by a quadrupole mass filter (range m/z 30 - m/z 300) operated in selected ion mode, and detected using a channeltron detector. The ions
monitored and dwell times are given in Table S-1 of the SI. Each ion was monitored once in a 25 s period. Mass counts were normalized to 1 106 counts of I. Calibration. Response factors of the I-CIMS to PAN, PPN, ClNO2, and Cl2 were determined off-line in a similar fashion as described earlier.27,28 The PAN and PPN response factors were assumed to be equal during this campaign.26,27,29 PPN was eluted from a passive diffusion source and sampled simultaneously by CIMS and thermal dissociation cavity ring-down spectroscopy (TD-CRDS)27 to determine a response factor of 18 counts/pptv. ClNO2 was generated by passing a dilute flow of Cl2 in N2 past an aqueous bed containing sodium nitrite and quantified simultaneously by CIMS and TD-CRDS as described in ref 28. The Cl2 response factor was determined by successive dilutions of the output from a standard cylinder (9.85 ( 5% ppmv Cl2 in N2, Praxair, certified). The ClNO2 response factor at m/z 208, determined using humidified calibration gas (for reasons that are discussed below), was 0.31 counts/pptv, whereas the Cl2 response factor was 0.33 counts/pptv. The accuracy of the PAN, ClNO2 and Cl2 mixing ratios reported in this manuscript was approximately (15%, ( 30%, and (30%, respectively. For 25 s data, the limits of detection (LOD) were 3, 5, and 22 pptv for PAN, ClNO2, and Cl2, respectively; the difference between ClNO2 and Cl2 was caused because we selected a higher dwell time at m/z 208 compared to m/z 197 (Table S-1 of the SI). The Cl2 data presented in this manuscript were averaged to 10 min to improve the LOD to approximately 6 pptv. Humidity Dependence. Reactions 68 are promoted by water vapor within the IMR.26,29,33 The water dependence of the CIMS response was determined in off-line experiments in which the CIMS bubbler (Figure 2) was bypassed, and the sample flow was humidified using two 30 slpm MFCs delivering zero air. The effluent of the first MFC was connected in series to a second bubbler and combined with the effluent of the second MFC which delivered dry zero air. Relative humidity and air temperature of the combined effluent was monitored with a NIST-traceable digital hygrometer (VWR 35519041) and converted to partial pressure of water vapor using the Goff and Gratch equation.34 To the humidified sampled flow were added the output of the 13C-labeled PAN photochemical source27 and either the output of the ClNO2 source 8891
dx.doi.org/10.1021/es201955u |Environ. Sci. Technol. 2011, 45, 8889–8896
Environmental Science & Technology
Figure 3. Relative signal intensities of 13CH313CO2 (m/z 61, shown as green triangles), generated from 13C-labeled PAN, (ClNO2)I (m/z 208, shown as blue circles), and (Cl2)I (m/z 197, shown as downwardfacing triangles) as a function of humidity in parts-per-thousand by volume (ppthv) in the ionmolecule reaction (IMR) region (after dilution with reagent ion and “dry” bubbler flows). The trend lines are hyperbolic expressions 29 fitted to the data. Superimposed are data from Zheng et al. 29 at m/z 61 (shown as diamonds, )) and from Kercher et al. 33 at m/z 208 (shown as red squares).
or the Cl2 gas cylinder. Mixing ratios of ClNO2 were monitored in parallel by TD-CRDS to verify the source stability (data not shown). Figure 3 shows the relative response of the CIMS to 13C-labeled PAN (monitored at m/z 61), ClNO2 (monitored at m/z 208), and Cl2 (monitored at m/z 197) plotted as a function of water vapor mixing ratio in the IMR. The trend lines are hyperbolic expressions29 fitted to the data and are shown here merely as visual guidelines. The response of the CIMS is humidity-dependent at low humidity and is constant above a threshold; this threshold is higher for ClNO2 and Cl2 than for 13C-labeled PAN, consistent with previous reports for PAN and ClNO2 in similar instruments.29,33 With the flow settings as described in the previous section and assuming that the CIMS bubbler achieves saturation with respect to water vapor, we calculate a water vapor mixing ratio of 4 parts-per-thousand by volume (ppthv) in the IMR if dry air is sampled. This amount of water suffices to ensure measurements are RH-independent for the PANs, but not for ClNO2 and Cl2. We found that ClNO2 and Cl2 calibrations conducted with dry zero air (and with the CIMS bubbler operated normally) yielded a response factor 27% lower than those conducted with humidified calibration gases. This implies that the actual humidity achieved with the CIMS bubbler is lower than the calculated value. The ambient air sampled in this campaign contained a sufficiently high amount of moisture to ensure that the CIMS response was independent of RH.
’ RESULTS AND DISCUSSION General Observations. During the three days of measurements, the meteorological conditions (Figure S-2 of the SI) were typical of early spring, i.e., sunny with daytime temperature highs just below +20 °C and the nights frost-free with nocturnal lows of +5 °C. There were no leaves on the deciduous trees, which suggests that biogenic emissions of volatile organic compounds (VOCs) were relatively small and limited to those from grasses and conifers. The dominant wind directions were from the W to NW and from the SE, where downtown and most of city’s people and roadways including a large industrial park and railroad yards are located. An overview of NO, NOy, O3, and CO mixing ratios collected at the routine monitoring site and of PAN
ARTICLE
Figure 4. Time series of ClNO2 and Cl2 mixing ratios observed on the University of Calgary campus.
and PPN mixing ratios collected at UC is provided in the SI section (Figure S-3). ClNO2. Figure 4 presents a time series of ClNO2 mixing ratios derived from m/z 208. ClNO2 exhibited a clear diurnal profile, similar to previous measurements of this compound at other locations.12,16,33 ClNO2 was observed every night during which measurements were made with maxima between 80 and 250 pptv, corresponding to a maximum of 2% of NOy (Figure S-4 of the SI). The observed range of mixing ratios is lower than those observed in coastal areas12,33 and similar to what was observed in Boulder, Colorado.16 Measurable quantities of ClNO2 were observed until noon, consistent with the expected slow photolysis of ClNO2. Mixing ratios of ClNO2 were at a minimum and below the instrumental detection limit of 5 pptv in the early afternoon. Figure 5 shows a pair of partial mass scans during the campaign (the full spectra are shown as Figure S-5 of the SI). The first spectrum was taken on April 16 at 2 p.m., the other on April 17 at 3 a.m. local time. The nocturnal spectrum contains peaks at all masses expected for ClNO2 with the expected 35Cl:37Cl isotope ratio, i.e., m/z 208 and m/z 210 (ClNO2)I, m/z 162 and m/z 164 (Cl)I, and m/z 35 and m/z 37 (Cl). The insert of Figure 5 shows a scatter plot of m/z 210 to m/z 208 for the entire campaign with the expected chloride isotope ratio. Cl2. We observed statistically relevant counts at m/z 197, indicating the presence of Cl2 (Figure 4). Only m/z 197 (35Cl2)I was continuously monitored. The other Cl isotopomers were observed in mass spectra recorded when m/z 197 counts were elevated (see Figure S-6 of the SI for two examples). Concentrations of Cl2 were highest at night, with maxima in the range of 30 to 65 pptv and were sometimes correlated with those of ClNO2, and sometimes anticorrelated (Figure 4). It is not likely that Cl2 was produced in the atmosphere, e.g., by heterogeneous conversion of ClNO2,32 as the aerosol would have to contain high levels of chloride and be very acidic. Further, elevated Cl2 levels were observed during daytime in spite of Cl2 loss due to photolysis. Hence, the observed Cl2 most likely originated from one or more local anthropogenic sources, perhaps from evaporation of chlorinated municipal tap water or fugitive emissions from industrial sources18 located in SE Calgary. Other Chlorine Containing Species. The daytime spectrum, shown in Figures 5 and S-5 of the SI, contains elevated counts at m/z 35 and m/z 37 but at no other Cl containing anions. The absence of counts at m/z 208, 210, 197, 199, and 201 suggests that fragmentation of iodide ion cluster ions was not contributing to the enhanced ion counts at m/z 35 and 37 and suggests the 8892
dx.doi.org/10.1021/es201955u |Environ. Sci. Technol. 2011, 45, 8889–8896
Environmental Science & Technology
Figure 5. Mass spectra of ambient air sampled on April 16, at 13:47 MDT (orange trace) and on April 17, 2010, at 02:47 MDT (blue trace). The mass spectrum acquired at night contains ion counts at m/z 208 and m/z 210 ((35ClNO2)I) and (37ClNO2)I, m/z 162 and m/z 164 ((35Cl)I and (37Cl)I), and m/z 35 and m/z 37 (35Cl and 37Cl). The mass spectrum acquired during the day only contains m/z 35 and m/z 37, suggesting that species other than ClNO2 contribute to counts at these masses. The insert shows ion counts observed at m/z 210 plotted against counts m/z 208 for 25 s and 5 min averaged data, respectively. The observed ratio of 1:(3.17 ( 0.03) is very close to the range in natural variation of the Cl isotopes (1:(3.13 ( 0.01)).42
presence of a Cl containing trace gas other than ClNO2 or Cl2 during daytime in Calgary air. For HCl, we would expect to see adduct ions at m/z 163 (H35Cl)I and m/z 165 (H37Cl)I. However, m/z 165 was not present in any of the mass spectra. We speculate that the unidentified daytime species is hypochlorous acid (HOCl).24 Directional (In)dependence of ClNO2 and Cl2 Production. The night of April 1617 (Figure S-7 of the SI) was an interesting case as we experienced three distinct wind flow regimes. At the beginning of the night (period I), the winds were from the SE. We observed moderate production of ClNO2 and some Cl2 in this air mass. The PPN to PAN ratio, which can be used to gain insight into the chemical history of the air mass sampled,35 was approximately 0.113 ( 0.003 (where the error indicates the precision of the measurement). Halfway through the night, the wind-direction switched to NW (period II). At that time, a sharp increase in ClNO2 mixing ratios and a decrease in Cl2 abundance were observed. The PPN to PAN ratio increased to a maximum of 0.128 ( 0.003), slightly higher than earlier. Further, NOy mixing ratios reached 40 ppbv, the largest of that night. The elevated levels of ClNO2, high NOy abundance, and greater PPN/PAN ratio are indicative of a more polluted and processed air mass of anthropogenic origin. One plausible explanation is that the same air mass was observed twice, once leaving the city and once again coming back. An alternative explanation is subsidance of an “aged” air mass from aloft, coinciding with the change of wind direction. The greater ClNO2 abundance in an “older” and more NOx-polluted air mass is consistent with its production from heterogeneous uptake of N2O5 and implies that this particular air mass was not depleted of aerosol chloride to sustain production of ClNO2. After the initial peak in period II between 1:002:00 a.m., ClNO2, PPN/PAN, and NOy decrease with mixing ratios of NOy reaching 9 ppbv by 3:00 a.m. At the end of
ARTICLE
the night (period III in Figure S-7 of the SI), the winds were still from the NW and the PPN/PAN ratio dropped to about 0.088 ( 0.003. A ratio of <0.1 is indicative of a relatively clean air mass with relatively little anthropogenic VOCs that were present the previous day (when PAN and PPN would have been produced) and little cloud processing.35 Interestingly enough, virtually no Cl2 but up to 150 pptv of ClNO2 were present in this air mass. The latter suggests that ClNO2 production was efficient over a large geographical area, whereas Cl2 originated from one or more nearby point sources located SE of the main sampling site. These conclusions are supported by wind rose plots (Figures S-8 of the SI). On average, the ClNO2 mixing ratios observed in Calgary did not exhibit a significant dependence on wind direction suggesting that ClNO2 production is associated with an area source as opposed to discernible point sources. This lack of directional dependence implies that the ClNO2 precursors, in particular the aerosol chloride, are relatively evenly distributed within the study area. In contrast, high Cl2 mixing ratios were observed only when the local winds were from SE, where downtown and the industrial park with several petrochemical facilities are located, consistent with an anthropogenic Cl2 source. Estimation of ClNO2 Yield from Uptake of N2O5. We used a time-dependent model to put constraints on the branching ratio of reaction (4) into 4b, i.e., the yield of ClNO2 for every molecule of N2O5 lost heterogeneously, ϕ(ClNO2). The main assumptions of this approach are that air masses with similar histories were sampled throughout the night that there are no vertical gradients in NO, NO2, and O3. Since the first assumption is invalid for the first night of the study (as discussed above), the model was only applied to the second and third night of the data set. Mixing ratios of NO2 were estimated by subtracting the mixing ratios of NO, PAN, PPN, and ClNO2 from NOy. Hourly mixing ratios of O3 were interpolated to 1 min data. The first-order loss rate coefficients of NO3 and N2O5, kNO3, and kN2O5, were calculated using the following: kNO3 ≈k3 ½NO
ð9Þ
and 1 kN2 O5 ≈ c̅ γðN2 O5 ÞSA 4
ð10Þ
Expression 9 does not include potential losses of NO3 to VOCs but is nevertheless a reasonable estimate since NO3 primarily reacts with unsaturated biogenic VOCs, whose abundances were likely low during this work, and NO was often very abundant. In eq 10, c is the mean molecular speed of N2O5, γ(N2O5) is the uptake probability of N2O5, and SA is the aerosol surface area density. Since the latter are not constrained by measurements, we used four model cases with assumed values of aerosol surface densities and uptake parameters as shown in Table 1. The range of uptake probabilities chosen, 0.005 to 0.03, was guided by values reported by Thornton et al.16 Eq 10 neglects losses of N2O5 due to dry deposition, which could be non-negligible in an urban area. To estimate the amount of ClNO2 that could have been produced nocturnally, we integrated equations (1-4) for solar zenith angles >85° for the four cases. The results are presented in Figure 6. Depending on input parameters, the total amount of N2O5 lost via reaction (4) was in the range of 0.1 ppbv to 1.2 ppbv, which is small compared to the amount of NOx present and suggests that most of the chemical processing of NOx in this air mass occurs 8893
dx.doi.org/10.1021/es201955u |Environ. Sci. Technol. 2011, 45, 8889–8896
Environmental Science & Technology
ARTICLE
Table 1. Overview of Model Input Parameters case 1
case 2
(high N2O5 reactivity)
(low N2O5 reactivity)
γ(N2O5)
0.03
0.005
0.02
0.02
ϕ(ClNO2)
1
1
1
15%
500
200
250
250
parameter
SA (μm2 cm3)
case 3
case 4
(intermediate (fit to case) data)
Figure 6. Comparison of estimated integrated N2O5 losses with observed ClNO2 mixing ratios for the four scenarios outlined in Table 1. The amount of ClNO2 is less than the total integrated loss of N2O5.
downwind from the city, aloft, and/or during daytime. We obtain reasonable agreement of observed and modeled ClNO2 if a ϕ(ClNO2) value of 0.15 is used in the calculation (case 4). Assuming that the aerosol was internally mixed and deliquesced, a ϕ(ClNO2) value of 0.15 would imply an aerosol chloride concentration in the range of 10 to 50 mM.13 The yield of 15% calculated above is likely too high if there is significant production of ClNO2 aloft, where NO mixing ratios are typically lower,4 and if the long-lived ClNO2 mixes down to the surface. We repeated the model runs assuming NO mixing ratio of zero ppbv, i.e., with kNO3 = 0. Figure S-8 of the SI shows the results. In this scenario, the amount of N2O5 consumed is in the range of 3.3 to 4.7 ppbv, and ϕ(ClNO2) values of 3% (2nd night) and 1.5% (3rd night) reproduce the observed ClNO2 mixing ratios. This yield would imply an aerosol chloride concentration in the range of 1 to 10 mM.13 However, given that the aerosol chloride is most likely derived from a surface source, the first scenario is probably more realistic. In either case, the amount of N2O5 consumed by heterogeneous reactions is considerably larger than the amount of ClNO2 observed. This suggests that the observed ClNO2 can be rationalized by production via reaction 4b alone, and additional ClNO2 source chemistry is not needed to explain the observations.
Sources of Aerosol Chloride. The observed production of ClNO2 in Calgary requires the presence of a substantial amount of aerosol chloride dispersed over a large area. High levels of reactive chlorine, in particular HCl and aerosol chloride, are not uncommon in urban areas.36 Further, chloride is ubiquitous in the environment, as evident from observed chloride levels in rainwater samples across the continent, which have been interpreted as being mainly sea-salt derived.37 However, we do not think sea salt played a major role in the observed formation of ClNO2 because its lifetime is short,7 in particular since sea salt concentration drops rapidly with distance from the coast38 and the western slopes of the coastal and Rocky Mountain ranges favor deposition. We used the online HYSPLIT model39,40 to calculate 2-day back-trajectories (see Figure S-10 of the SI), which indicated that the air originated from the south to east, i.e., the continental U.S. Thus, transport of marine aerosol to the study area was unlikely. Anthropogenic activities within the city of Calgary were the most likely the main source of aerosol chloride and of ClNO2 during the time of these measurements. Coal-fired power plants and cooling towers are insignificant sources of Cl species in the Calgary area, since there are no coal-fired power plants in the vicinity of the city and there are few cooling towers. Some of the aerosol chloride was in all likelihood derived from photolysis of the anthropogenically released Cl2 and subsequent reaction of Cl with hydrocarbons to form HCl, which partitions to the aerosol phase. Another, likely major, source of aerosol chloride in the area was suspension of road dust, which in springtime contains residual road salt that accumulated over the course of the winter. During the time of this work, the city conducted its annual streetcleaning campaign, which generated a considerable amount of suspended dust particles. Ultimately, though, the injection of new anthropogenic Cl does not have to be large to sustain a considerable level of aerosol chloride and hence ClNO2 production because of cycling of Cl between various gas-phase species and the aerosol phase in the atmosphere.41 Implications. The high mixing ratios of ClNO2 and Cl2 observed in this work show that there can be active halogen chemistry attributable to anthropogenic sources in a midcontinental urban environment. The primary fate of ClNO2 and Cl2 is photo dissociation at sunrise to yield NO2 and Cl atoms, which oxidize hydrocarbons and initiate or accelerate photochemical O3 production.18 However, the amounts of NO2 generated in this data set are relatively small compared to the amounts of NOx emitted locally by cars and thus are unlikely to have significant impact on air quality within the city. It would be interesting to investigate how far the observed halogen chemistry extends downwind of the Calgary area, where the recycling of NOx could potentially have greater impact, or what impacts there are during stagnation periods when air quality rapidly deteriorates. The Cl atom source from ClNO2 photolysis observed here could make a substantial difference as it will generate organic peroxy radicals from otherwise fairly unreactive alkanes and thus should be included in atmospheric chemistry and air quality models of landlocked urban areas.
’ ASSOCIATED CONTENT
bS
Supporting Information. 11 Figures and 1 Table. This material is available free of charge via the Internet at http:// pubs.acs.org.
8894
dx.doi.org/10.1021/es201955u |Environ. Sci. Technol. 2011, 45, 8889–8896
Environmental Science & Technology
’ AUTHOR INFORMATION Corresponding Author
*Phone: 403-220-8689; Fax: 403-289-9488; E-mail: hosthoff@ ucalgary.ca. Present Address †
School of Public and Environmental Affairs, Indiana University, Bloomington, IN 47405.
’ ACKNOWLEDGMENT This work was financially supported by an Alberta Ingenuity New Faculty Award, the National Science and Engineering Research Council of Canada (NSERC) in the form of Discovery and Research Tools and Instruments (RTI) grants, and the University of Calgary via startup funds and laboratory space. Amanda Furgeson acknowledges financial support by an NSERC Alexander Graham Bell Scholarship. ’ REFERENCES (1) Gauderman, W. J.; Avol, E.; Lurmann, F.; Kuenzli, N.; Gilliland, F.; Peters, J.; McConnell, R. Childhood asthma and exposure to traffic and nitrogen dioxide. Epidemiology 2005, 16 (6), 737–743. (2) Finlayson-Pitts, B. J.; Pitts, J. N. Chemistry of the Upper and Lower Atmosphere: Theory, Experiments, And Applications; Academic Press: San Diego, Calif., 2000. (3) Sander, S. P.; Friedl, R. R.; Ravishankara, A. R.; Golden, D. M.; Kolb, C. E.; Kurylo, M. J.; Molina, M. J.; Moortgat, G. K.; Keller-Rudek, H.; Finlayson-Pitts, B. J.; Wine, P. H.; Huie, R. E.; Orkin, V. L. Chemical Kinetics and Photochemical Data for Use in Atmospheric Studies, Evaluation Number 15. NASA/JPL: Pasadena, CA, 2006; Vol. 062. (4) Stutz, J.; Alicke, B.; Ackermann, R.; Geyer, A.; White, A.; Williams, E., Vertical profiles of NO3, N2O5, O3, and NOx in the nocturnal boundary layer: 1. Observations during the Texas Air Quality Study 2000. J. Geophys. Res.-Atmos. 2004, 109, D12306; DOI 10.1029/ 2003JD004209. (5) Chang, W. L.; Bhave, P. V.; Brown, S. S.; Riemer, N.; Stutz, J.; Dabdub, D. Heterogeneous atmospheric chemistry, ambient measurements, and model calculations of N2O5: A review. Aerosol Sci. Technol. 2011, 45 (6), 655685; DOI 10.1080/02786826.2010.551672. (6) Finlayson-Pitts, B. J.; Ezell, M. J.; Pitts, J. N. Formation of chemically active chlorine compounds by reactions of atmospheric nacl particles with gaseous N2O5 and ClONO2. Nature 1989, 337 (6204), 241–244. (7) Erickson, D. J.; Seuzaret, C.; Keene, W. C.; Gong, S. L. A general circulation model based calculation of HCl and ClNO2 production from sea salt dechlorination: Reactive chlorine emissions inventory. J. Geophys. Res.-Atmos. 1999, 104 (D7), 8347–8372. (8) Behnke, W.; George, C.; Scheer, V.; Zetzsch, C. Production and decay of ClNO2, from the reaction of gaseous N2O5 with NaCl solution: Bulk and aerosol experiments. J. Geophys. Res.-Atmos. 1997, 102 (D3), 3795–3804. (9) Simon, H.; Kimura, Y.; McGaughey, G.; Allen, D. T.; Brown, S. S.; Osthoff, H. D.; Roberts, J. M.; Byun, D.; Lee, D., Modeling the impact of ClNO2 on ozone formation in the Houston area. J. Geophys. Res.-Atmos. 2009, 114, D00F03; DOI 10.1029/2008jd010732. (10) Ganske, J. A.; Berko, H. N.; Finlayson-Pitts, B. J. Absorption cross-sections for gaseous ClNO2 and Cl2 at 298 K—Potential organic oxidant source in the marine troposphere. J. Geophys. Res.-Atmos. 1992, 97 (D7), 7651–7656. (11) Behnke, W.; Kruger, H. U.; Scheer, V.; Zetzsch, C. Formation of atomic Cl from sea spray via photolysis of nitryl chloride— Determination of the sticking coefficient of N2O5 on NaCl aerosol. J. Aerosol Sci. 1991, 22, S609–S612.
ARTICLE
(12) Osthoff, H. D.; Roberts, J. M.; Ravishankara, A. R.; Williams, E. J.; Lerner, B. M.; Sommariva, R.; Bates, T. S.; Coffman, D.; Quinn, P. K.; Stark, H.; Burkholder, J. B.; Talukdar, R. K.; Meagher, J.; Fehsenfeld, F. C.; Brown, S. S. High levels of nitryl chloride in the polluted subtropical marine boundary layer. Nat. Geosci. 2008, 1 (5), 324–328. (13) Roberts, J. M.; Osthoff, H. D.; Brown, S. S.; Ravishankara, A. R.; Coffman, D.; Quinn, P. K.; Bates, T. S., Laboratory studies of products of N2O5 uptake on Cl containing substrates. Geophys. Res. Lett. 2009, 36, L20808; DOI 10.1029/2009GL040448. (14) Bertram, T. H.; Thornton, J. A. Toward a general parameterization of N2O5 reactivity on aqueous particles: the competing effects of particle liquid water, nitrate and chloride. Atmos. Chem. Phys. 2009, 9 (21), 8351–8363. (15) Raff, J. D.; Njegic, B.; Chang, W. L.; Gordon, M. S.; Dabdub, D.; Gerber, R. B.; Finlayson-Pitts, B. J. Chlorine activation indoors and outdoors via surface-mediated reactions of nitrogen oxides with hydrogen chloride. Proc. Natl. Acad. Sci. U.S.A. 2009, 106 (33), 13647–13654. (16) Thornton, J. A.; Kercher, J. P.; Riedel, T. P.; Wagner, N. L.; Cozic, J.; Holloway, J. S.; Dube, W. P.; Wolfe, G. M.; Quinn, P. K.; Middlebrook, A. M.; Alexander, B.; Brown, S. S. A large atomic chlorine source inferred from mid-continental reactive nitrogen chemistry. Nature 2010, 464 (7286), 271–274. (17) Lobert, J. M.; Keene, W. C.; Logan, J. A.; Yevich, R. Global chlorine emissions from biomass burning: Reactive chlorine emissions inventory. J. Geophys. Res.-Atmos. 1999, 104 (D7), 8373–8389. (18) Tanaka, P. L.; Oldfield, S.; Neece, J. D.; Mullins, C. B.; Allen, D. T. Anthropogenic sources of chlorine and ozone formation in urban atmospheres. Environ. Sci. Technol. 2000, 34 (21), 4470–4473. (19) McCulloch, A.; Aucott, M. L.; Benkovitz, C. M.; Graedel, T. E.; Kleiman, G.; Midgley, P. M.; Li, Y. F. Global emissions of hydrogen chloride and chloromethane from coal combustion, incineration and industrial activities: Reactive Chlorine Emissions Inventory. J. Geophys. Res.-Atmos. 1999, 104 (D7), 8391–8403. (20) Lee, P. K. H.; Brook, J. R.; Dabek-Zlotorzynska, E.; Mabury, S. A. Identification of the major sources contributing to PM2.5 observed in Toronto. Environ. Sci. Technol. 2003, 37 (21), 4831–4840. (21) Griffiths, P. T.; Cox, R. A. Temperature dependence of heterogeneous uptake of N2O5 by ammonium sulfate aerosol. Atmos. Sci. Lett. 2009, 10 (3), 159–163. (22) Schweitzer, F.; Mirabel, P.; George, C. Multiphase chemistry of N2O5, ClNO2, and BrNO2. J. Phys. Chem. A 1998, 102 (22), 3942–3952. (23) Frenzel, A.; Scheer, V.; Sikorski, R.; George, C.; Behnke, W.; Zetzsch, C. Heterogeneous interconversion reactions of BrNO2, ClNO2, Br2, and Cl2. J. Phys. Chem. A 1998, 102 (8), 1329–1337. (24) Lawler, M. J.; Sander, R.; Carpenter, L. J.; Lee, J. D.; von Glasow, R.; Sommariva, R.; Saltzman, E. S. HOCl and Cl2 observations in marine air. Atmos. Chem. Phys. 2011, 11 (15), 7617–7628. (25) Clean Air Strategic Alliance Website; http://www.casadata.org. (26) Slusher, D. L.; Huey, L. G.; Tanner, D. J.; Flocke, F. M.; Roberts, J. M., A thermal dissociation-chemical ionization mass spectrometry (TD-CIMS) technique for the simultaneous measurement of peroxyacyl nitrates and dinitrogen pentoxide. J. Geophys. Res.-Atmos. 2004, 109, D19315; DOI 10.1029/2004JD004670. (27) Furgeson, A.; Mielke, L. H.; Paul, D.; Osthoff, H. D., A photochemical source of peroxypropionic and peroxyisobutanoic nitric anhydride. Atmos. Environ. 2011, 45 (28), 50255032; DOI 10.1016/ j.atmosenv.2011.03.072. (28) Thaler, R. D.; Mielke, L. H.; Osthoff, H. D., Quantification of nitryl chloride at part per trillion mixing ratios by thermal dissociation cavity ring-down spectroscopy. Anal. Chem. 2011, 83 (7), 27612766; DOI 10.1021/ac200055z. (29) Zheng, W.; Flocke, F. M.; Tyndall, G. S.; Swanson, A.; Orlando, J. J.; Roberts, J. M.; Huey, L. G.; Tanner, D. J., Characterization of a thermal decomposition chemical ionization mass spectrometer for the measurement of peroxy acyl nitrates (PANs) in the atmosphere. Atmos. Chem. Phys. 2011, 11 (13), 65296547; DOI 10.5194/acp-11-65292011. 8895
dx.doi.org/10.1021/es201955u |Environ. Sci. Technol. 2011, 45, 8889–8896
Environmental Science & Technology
ARTICLE
(30) Paul, D.; Furgeson, A.; Osthoff, H. D., Measurement of total alkyl and peroxy nitrates by thermal decomposition cavity ring-down spectroscopy. Rev. Sci. Instrum. 2009, 80, 114101; DOI 10.1063/ 1.3258204. (31) McNeill, V. F.; Patterson, J.; Wolfe, G. M.; Thornton, J. A., The effect of varying levels of surfactant on the reactive uptake of N2O5 to aqueous aerosol. Atmos. Chem. Phys. 2006, 6 (6), 16351644; DOI 10.5194/acp-6-1635-2006. (32) Roberts, J. M.; Osthoff, H. D.; Brown, S. S.; Ravishankara, A. R. N2O5 oxidizes chloride to Cl2 in acidic atmospheric aerosol. Science 2008, 321 (5892), 1059. (33) Kercher, J. P.; Riedel, T. P.; Thornton, J. A. Chlorine activation by N2O5: simultaneous, in situ detection of ClNO2 and N2O5 by chemical ionization mass spectrometry. Atmos. Meas. Tech. 2009, 2 (1), 193–204. (34) Goff, J. A.; Gratch, S. Low-pressure properties of water from 160 to 212 °F. Trans. Amer. Soc. Heat. Vent. Eng. 1946, 52, 95–121. (35) Roberts, J. M. PAN and related compounds. In Volatile Organic Compounds in the Atmosphere; Koppmann, R., Ed.; Blackwell Publishing: Oxford, UK, 2007. (36) Graedel, T. E.; Keene, W. C. Tropospheric budget of reactive chlorine. Global Biogeochem. Cycles 1995, 9 (1), 47–77. (37) Junge, C. E.; Gustafson, P. E. On the distribution of sea salt over the United States and its removal by precipitation. Tellus 1957, 9 (2), 164–173. (38) Gustafsson, M. E. R.; Franzen, L. G. Inland transport of marine aerosols in southern Sweden. Atmos. Environ. 2000, 34 (2), 313–325. (39) Draxler, R. R.; Hess, G. D. An overview of the HYSPLIT_4 modelling system for trajectories, dispersion and deposition. Aust. Met. Mag. 1998, 47 (4), 295–308. (40) Draxler, R. R.; Rolph, G. D. HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory) model access via NOAA ARL READY website http://ready.arl.noaa.gov/HYSPLIT.php. (41) Simon, H.; Kimura, Y.; McGaughey, G.; Allen, D. T.; Brown, S. S.; Coffman, D.; Dibb, J.; Osthoff, H. D.; Quinn, P.; Roberts, J. M.; Yarwood, G.; Kemball-Cook, S.; Byun, D.; Lee, D. Modeling heterogeneous ClNO2 formation, chloride availability, and chlorine cycling in Southeast Texas. Atmos. Environ. 2010, 44 (40), 5476–5488. (42) De Laeter, J. R.; Bohlke, J. K.; De Bievre, P.; Hidaka, H.; Peiser, H. S.; Rosman, K. J. R.; Taylor, P. D. P. Atomic weights of the elements: Review 2000—(IUPAC technical report). Pure Appl. Chem. 2003, 75 (6), 683–800.
8896
dx.doi.org/10.1021/es201955u |Environ. Sci. Technol. 2011, 45, 8889–8896
ARTICLE pubs.acs.org/est
PAH Molecular Diagnostic Ratios Applied to Atmospheric Sources: A Critical Evaluation Using Two Decades of Source Inventory and Air Concentration Data from the UK Athanasios Katsoyiannis,* Andrew J. Sweetman, and Kevin C. Jones Lancaster Environment Centre, Lancaster University, Lancaster, LA1 4YQ, U.K.
bS Supporting Information ABSTRACT: Molecular diagnostic ratios (MDRs)—the ratios of defined pairs of individual compounds—have been widely used as markers of different source categories of polycyclic aromatic hydrocarbons (PAHs). However, it is well-known that variations in combustion conditions and environmental degradation processes can cause substantial variability in the emission and degradation of individual compounds, potentially undermining the application of MDRs as reliable source apportionment tools. The United Kingdom produces a national inventory of atmospheric emissions of PAHs, and has an ambient air monitoring program at a range of rural, semirural, urban, and industrial sites. The inventory and the monitoring data are available over the past 20 years (1990 2010), a time frame that has seen known changes in combustion type and source. Here we assess 5 MDRs that have been used in the literature as source markers. We examine the spatial and temporal variability in the ratios and consider whether they are responsive to known differences in source strength and types between sites (on rural urban gradients) and to underlying changes in national emissions since 1990. We conclude that the use of these 5 MDRs produces contradictory results and that they do not respond to known differences (in time and space) in atmospheric emission sources. For example, at a site near a motorway and far from other evident emission sources, the use of MDRs suggests “non-traffic” emissions. The ANT/(ANT + PHE) ratio is strongly seasonal at some sites; it is the most susceptible MDR to atmospheric processes, so these results illustrate how weathering in the environment will undermine the effectiveness of MDRs as markers of source(s). We conclude that PAH MDRs can exhibit spatial and temporal differences, but they are not valid markers of known differences in source categories and type. Atmospheric sources of PAHs in the UK are probably not dominated by any single clear and strong source type, so the mixture of PAHs in air is quickly “blended” away from the influence of the few major point sources which exist and further weathered in the environment by atmospheric reactions and selective loss processes.
’ INTRODUCTION Polycyclic aromatic hydrocarbons (PAHs) are a class of ubiquitous organic chemicals which enter the environment primarily from incomplete combustion or pyrolysis of organic material. Some of these combustion processes are natural (e.g., volcanic activity; forest fires), while others are anthropogenic. Combustion conditions affect the efficiency of burning; “inefficient” low-temperature processes generally generate more PAHs. Major anthropogenic sources include low-temperature burning of coal and wood for domestic space heating, hightemperature burning for power generation, vehicle emissions, cigarette smoking, and the uncontrolled burning of wastes and accidental fires.1 4 The aforementioned sources are believed to emit hundreds of gigagrams of PAHs per year (520 Gg y 1 in 20045). Due to their persistence in the environment and semivolatile character, past PAH emissions that have deposited to soils, vegetation, and water bodies6 can be re-emitted to atmosphere. There is also growing evidence that some low molecular weight PAHs can be formed naturally in soils and volatilize to the atmosphere.7,8 Numerous point and diffuse r 2011 American Chemical Society
sources of PAHs can therefore impact on ambient levels. Many PAHs are mutagenic and some are known carcinogens.9 11 Consequently, there are human health concerns about elevated concentrations of PAHs in the environment. Indeed, air quality standards have been proposed for PAHs in the European Union12 and international actions aim to reduce atmospheric emissions. For all these reasons, there is a need to understand the relative importance of different sources—locally, nationally, regionally, and globally13—so that source reduction measures can be efficient and appropriately targeted. A number of techniques have been developed to improve source apportionment. One of the most common is the use of the molecular diagnostic ratios (MDR). It is argued that some PAHs are emitted in reasonably constant proportions from some of the most important sources of PAHs and that their Received: July 2, 2011 Accepted: August 22, 2011 Revised: August 18, 2011 Published: August 22, 2011 8897
dx.doi.org/10.1021/es202277u | Environ. Sci. Technol. 2011, 45, 8897–8906
Environmental Science & Technology
ARTICLE
Table 1. Characteristic PAH Molecular Diagnostic Ratios14,15 petrogenic
pyrogenic
ANT/(ANT + PHE)
<0.1
>0.1
BaA/(BaA + CHR) FLT/(FLT + PYR)
<0.2 <0.4
>0.35 >0.4
IPY/(IPY + BPE)
<0.2
>0.2
a
fuel combustion
grass/coal/wood combustion
FLT/(FLT + PYR)
0.4 0.5
>0.5
IPY/(IPY + BPE)
0.2 0.5
>0.5
BaP/BPE
nontraffic
traffic
<0.6
>0.6
a
ANT: Anthracene; PHE: Phenanthrene; BaA; Benzo[a]anthracene; CHR: Chrysene; FLT: Fluoranthene; PYR: Pyrene; IPY: Indeno[1,2,3-cd]pyrene; BPE: Benzo[g,h,i]perylene; BaP: Benzo[a]pyrene. concentrations (thus the ratios) remain constant between the source and the receptor. An example of a commonly used ratio is “ANT/(ANT + PHE)” (ANT: anthracene; PHE: phenanthrene). This is said to be indicative of unburned fossil fuel when the ratio is <0.1, but indicative of combustion sources when >0.1.14 Other ratios have been used (see Table 1), often unquestioningly, to identify pyrogenic (combustion sources) or petrogenic (petroleum input) origin, fuel or wood combustion, or traffic-related sources.14 22 However, other studies have highlighted the wide range in emission factors and compound ratios from given source categories, which raises concerns about the use of MDRs for source identification. Additional complications with source apportionment based on MDRs are that atmospheric sources can be mixed and complex in real world settings, and that individual PAH compounds have different atmospheric residence times and reactivities.23 25 These factors will inevitably result in varied and weathered mixtures and ratios in ambient atmospheres. Here we wanted to use a well-established national emissions inventory and data from a long-term ambient monitoring program to test the idea that MDRs respond to spatial and temporal changes in sources. MDRs were derived from a range of rural, semirural, urban, and industrial sites (Figure 1). Some of these sites have been in operation for around 20 years.26 Five MDRs were used, namely the following: 1. ANT (anthracene)/ ANT + PHE (phenanthere); 2. [BaA (benzo[a]anthracene)/ BaA + CHR (chrysene)]; 3. [FLT (fluoranthene)/FLT + PYR (pyrene)]; 4. [IPY (indeno[1,2,3-cd] pyrene)/IPY + BPE (benzo[g,h,i]perylene)]; and 5. BaP (benzo[a]pyrene)/BPE. It has been proposed that MDRs 1 4 can give information about petrogenic or pyrogenic origin, and that 3 and 4 may highlight whether the combustion material is fuel, grass, coal, or wood.14 The fifth MDR has been proposed as an indicator of traffic.14 Their use as diagnostic tools is addressed here using information from the UK inventory and local site knowledge about potential sources. This study is the first to our knowledge to use large atmospheric time-series to examine MDRs together with a 20-year emission inventory. Other studies in which big data sets were used were from Yunker et al.,14 Brandli et al.,15 and Dvorska et al.16
Figure 1. UK map featuring the sampling sites and the population densities.
’ METHODOLOGY In the present study, the data collected by two long monitoring programs (TOMPS and PAHs) and the emission inventory for PAHs compiled in detail for the last two decades are used. Details and information about these programs are presented below. Air Monitoring Data. The UK Department for Environment, Food and Rural Affairs (DEFRA) has supported a longterm program to monitor Toxic Organic Micro Pollutants (TOMPS), including PAHs at a number of rural and urban locations. Some of these sites have been running since 1990 and provide the longest data sets available in Europe. Details of the TOMPS program have been published previously.26 28 Of relevance here is that data are available quarterly (Q1 4, January March, April June, July September, October December). In addition, a wide range of other sites were established later on the PAH Monitoring Network (1999 and onward). Currently about 30 sites are operated. Details of this program are also available.29 Until 2007 the networks were using an Andersen GPS-1 ambient air sampler, although since then the PAHs network samplers were changed to Digitel DHA-80 high volume PM10 aerosol samplers to comply with the requirements of the European Commission’s Fourth Air Quality daughter directive.12 For direct comparability purposes, in the present study, only samples collected with Andersen GPS-1 ambient air samplers are used and discussed. Table S1 in the Supporting Information presents the PAH compounds that are monitored through the PAH monitoring network. 8898
dx.doi.org/10.1021/es202277u |Environ. Sci. Technol. 2011, 45, 8897–8906
Environmental Science & Technology
ARTICLE
Table 2. Descriptive Statistics of the PAH Concentrations (ng m 3) during the TOMPS and the PAHs Monitoring Networksa HAZ
SF
MID
BEL
BIR
average
155
19.5
72.0
41.4
38.1
max
777
134
340
87.8
62.5
min
6.40
6.10
8.30
5.61
12.3
median
124
15.0
57.3
35.4
34.8
EDI
GLA
MAN
LIV
HM
LON
HOL
14.4
27.3
63.0
30.2
12.5
55.4
27.7
25.8
64.7
237
64.9
50.3
267
47.4
9.10
3.51
4.6
18.9
3.0
4.5
12.4
12.5
23.3
45.7
29.3
9.6
39.8
27.6
a
HAZ: Hazelrigg; SF: Stoke Ferry; MID: Middlesborough; BEL: Belfast; BIR; Birmingham; EDI: Edinburgh; GLA: Glasgow; MAN: Manchester; LIV: Liverpool; HM: High Muffles; LON: London; HOL: Holyhead.
Figure 2. (a) Emission inventory estimates per category of emission sources for the years 1990 2009. (b) ΣPAH concentrations at the 6 TOMPS sites versus the emission inventory estimates.
In the present paper, data were compiled from the commencement of monitoring until 2007 2008 and used to derive the MDRs. A range of sites were selected to give long-term trend data [TOMPS sites; see Meijer et al.26 for analysis] and a rural semirural urban industrial gradient. The following were used: Urban [London (LON), Manchester (MAN), Middlesborough (MID), Liverpool (LIV), Birmingham (BIR), Edinburgh (EDI), Glasgow (GLA), and Belfast (BEL)]; rural/semirural [Holyhead (HOL), High Muffles (HM), Stoke Ferry (SF), and Hazelrigg (HAZ)]. The definitions of rural urban are subjective, based on knowledge of the location and surrounding areas. Figure 1 shows the site locations overlaid on a map showing the UK population
density. Using the average total PAH concentration as a guide, the gradient of sites is: HAZ > MID > MAN > LON > BEL > BIR > LIV > HOL > GLA > SF > EDI > HM (Table 2). The evident anomalous site here is HAZ, close to Lancaster. This is semirural, but has yielded concentrations that are routinely higher than urban/industrial locations. The National Emissions Inventory. The UK’s National Atmospheric Emissions Inventory, 1 also funded by DEFRA, compiles data for a range of air pollutants, including PAHs. PAH estimates are available from 1990. The inventory divides sources into a range of categories as highlighted in Figure 2a. Apparently, certain source types (e.g., metal 8899
dx.doi.org/10.1021/es202277u |Environ. Sci. Technol. 2011, 45, 8897–8906
Environmental Science & Technology
ARTICLE
Figure 3. Descriptive statistics for the five MDRs at all sampling sites: (a) ANT/(ANT + PHE), (b) (BaA/(BaA + CHR), (c) FLT/(FLT + PYR), (d) IPY/(IPY + BPE), (e) B[a]P/BPE].
smelting, iron and steel, and aluminum production) are dominated by few local point sources, but the data are compiled as a national estimate of total emissions. Other source categories, such as traffic or space heating, may be broadly linked to population density. In UK urban areas, space heating may typically rely on gas or electricity, while air concentrations in rural areas may be influenced by domestic burning of coal or wood. Previous studies have highlighted that the latter have much higher PAH emission factors than the former and can impact local air quality.30 32 This signal is strongly seasonal, reflecting more domestic burning for space heating in the winter.26 Attempts have been made in recent years to combine the UK emissions inventory with the measurements programs, often in combination with modeling.26 28 This is to check whether the inventory and models are broadly capturing the real trends apparent in ambient measurements. This can give confidence/ support for the inventory, or identify apparent discrepancies and
inform on the need for further emissions estimates, source identification, or model reparameterization.
’ RESULTS AND DISCUSSION Figure 2b overlays the general trends in UK PAH air concentrations for the TOMPS sites on the inventory and Table S2 presents the correlation coefficients of the total PAH concentrations of the various sites with the emission inventory estimates. The latter correlate strongly (p < 0.0005) with the ΣPAHs in LON and MID, but with all the other sites there is a poor correlation. The comparison between inventory estimates and measurements can highlight interesting differences. For example, in 2004 and 2006, an increase in measured PAH concentrations was seen at various sites (HAZ, MID, and MAN being the most characteristic), but this was not predicted in the inventory. 8900
dx.doi.org/10.1021/es202277u |Environ. Sci. Technol. 2011, 45, 8897–8906
Environmental Science & Technology
ARTICLE
Table 3. Between-Sites Correlation Analysis of the ANT/(ANT + PHE) MDR MAN
SF
LON
LIV
MID
BIR
BEL
EDI
GLA
HM
HH
MAN
1.00
SF
0.23
1.00
LON
0.45
0.21
LIV
0.14
0.31
0.02
1.00
MID
0.10
0.03
0.22
0.17
1.00
BIR
0.23
0.08
-0.40
0.40
0.50
1.00
BEL
0.30
0.19
0.02
0.37
0.32
0.50
1.00
EDI GLA
0.45 0.10
0.40 0.08
0.05 0.05
0.83 0.02
0.44 0.30
0.30 0.24
0.86 0.27
1.00 0.30
HM
0.20
0.01
0.05
0.02
0.32
0.02
0.29
0.10
0.00
1.00
HH
0.12
0.27
0.18
0.36
0.08
0.42
0.33
1.00
0.07
0.00
1.00
HAZ
0.18
0.02
0.25
0.00
0.05
0.36
0.17
0.00
0.19
0.16
0.04
HAZ
1.00
1.00
1.00
a
Values in bold and underlined are statistically significant at the 99.95% level of confidence. Values in bold are statistically significant at the 99% level of confidence. Underlined values are statistically significant at the 95% level of confidence.
MDR Results. The detailed presentation of the ratios for all sites and sampling periods is given as Supporting Information (SI), and the descriptive statistics are given in Figure 3a e. In Figure 3a, the ANT/(ANT + PHE) ratio is quite uniform, consistently <0.1. According to the interpretation of the MDRs (Table 1) this value suggests petrogenic origin of the PAHs in all sites. An exception to this behavior is MID; in 1996 1998 and 2007 2008 this value was much >0.1 (Figure S1), showing a very different behavior from the other sites. Although the impression given by Figure 3a is that there is consistency between sites, correlation analysis (Table 3) indicates that there is a strong correlation (p < 0.0005) only between the pairs LON MAN, EDI LIV, and EDI BEL. The rural/semirural sites, HAZ, HOL, HM, and SF do not exhibit any correlations with other sites. The BaA/(BaA + CHR) ratios are usually in the range of 0.2 0.4 (Figure 3b). According to Table 1, this is the “neutral” range that does not show clear dominance of petrogenic or pyrogenic sources. The between-site results are similar, although again MID is an outlier, for which the average/ highest values are affected by the unusually high values of the period 1995 1998. HAZ also exhibits a wide variation in this MDR. For the FLT/(FLT + PYR) MDR, the value is consistently >0.5. The interpretation from Table 1 again suggests pyrogenic origins. Again, MID is unusual, due to its behavior in the period 1995 1998 and 2007 2008. This MDR appears quite consistent between sites, being almost always >0.5; according to Table 1 this indicates vegetation, coal, and wood combustion. The IPY/(IPY + BPE) MDR is said to give the same indications as FLT/(FLT + PYR). As seen in Table 1, values >0.2 suggest pyrogenic origins, when 0.2 < MDR < 0.5, this suggests fuel combustion, and when MDR > 0.5, the pyrogenic origin is attributed to grass/coal/wood combustion. As seen from Figure 3d, the mean and average values of all sites (except MID) fall in the “pyrogenic origin” due to fuel combustion, something that is in disagreement with the aforementioned results of the FLT/(FLT + PYR) ratio. The BaP/BPE MDR (Figure 3e) is the ratio that reportedly shows the effect of traffic. For most sites, the conclusion is mainly “non-traffic” emissions; sites MID, HM, and SF are, according to this ratio, the most influenced by traffic. However, from
local knowledge this is inaccurate. In summary, the MDRs give mixed and contradictory messages about site/source differences. As mentioned, the sites selected are urban, rural, or semirural, but feature also other characteristics, such as being close to industrial zones, close to motorways, ports, etc. The specific emissions that each sampling site is receiving suggest that PAHs would exhibit spatial differences, in concentrations and in profiles. For example the HAZ sampling station is 6 km from the city of Lancaster, within 400 m of a busy motorway and far from residences. Previous studies indicate that emissions close to HAZ are due to vehicles and domestic burning of coal and wood. Major conurbations such as LON or MAN are expected to have contributions from domestic combustion/heating, traffic, and industrial/point source emissions. The background sites of HM and SF are expected to reflect the regional background signal, including biogenic emissions;7 neither is influenced by strong local sources. The MID site is influenced by urban and industrial emissions, while HOL is a semirural site, but also an important port, potentially impacted also by the emissions related to maritime traffic. The aforementioned site-related information about known sources is often not confirmed by the MDRs. For example, the B[a]P/BPE ratio at HAZ has only occasionally exceeded the value of 0.6, suggesting “non-traffic PAHs”. Indeed, its values are generally among the lowest observed (see SI), even though the site has a nearby traffic source. The B[a]P/BPE ratio is also unexpected at MID, because the ratio implies a strong influence of traffic, even though industrial sources are known to be nearby (Figure 3e). This could be attributed to the traffic of heavy vehicles, tracks, and machineries associated with industrial activities. In general, results in Figure 3 do not show major spatial differences, even though known source differences exist. If the ratios are truly reflective of source type (Table 1), the results of the ratios BaA/(BaA + CHR), FLT/(FLT + PYR), and IPY/(IPY + BPE) are all demonstrating pyrolytic or mixed sources and there are only a few cases where one of these suggested a petrogenic source of PAHs. Such examples were the BaA/(BaA + CHR) ratio in Manchester from 1994 (Q1) to 1995 (Q3) and the FLT/(FLT + PYR) in MID from 1996 (Q2) to 1997 (Q3) (SI). The ANT/(ANT + PHE) ratio is in most 8901
dx.doi.org/10.1021/es202277u |Environ. Sci. Technol. 2011, 45, 8897–8906
89 3920
36 1570
4 °C) (h)i
0.010 570 31
0.20 29 6.90
6.35 7
7.47 10
0.01
276 193-39-5
276 191-24-2
252 50-32-8
IPY
BPE
BaP
495
202 129-00-00 PYR
550
375 202 206-44-0 FLT
530
448 CHR 228 218-01-9
0.01
e
0.05 (0.046 )
8
1.37 10
11.77c
46 6.51
5.30 8.8
3
1.33 10
11.56c
4
3.33 10
400 228 56-55-3 BaA
404
1.16 (1.21e)
10.8
>1000
>1000
5.08
5.08
2.5
0.14
0.012
0.11
0.014
11 0.015 0.17
0.013 >1000
690 38
44 5.20
5.61
4
6.67 10 0.66 (0.89 )
8.9
5
8.40 10
340 178 85-01-8 PHE
0.11
340 ANT 178 120-12-7b
0.10
e
6
2.93 10
10.4
5.91
38
650
21.3
2.5
0.11
27 0.011 0.19
0.011 0.17
0.15 1.5
3.4 18.9
6.33 310
>1000 49
48 4.54
4.46 7.6 2
9.07 10 2.59 (4.29e)
7.3 3
2.27 10 1.79 (5.63e)
9.5
190
s 1)h 12
( 10 (s 1)g particles) (s 1)g with OH aire (h) (h)d (h)d KOAb KOWb (pa)a (pa-/m3 mol)a (°c)a registry PAH MW
Values taken from http://www.toronto.ca/health/pdf/cr_appendix_b_pah.pdf.35 b Values taken from: Huckins et al.36 c Values taken from Odabasi et al.37 d Behymer and Hites38 e Kwamena et al.39 f Bunce and Dryfhout.40 g Esteve et al..41 h Brubaker and Hites.42 i Halsall et al.43 a
7 23
6 19
19 °C) (h)i
ARTICLE
10
(gas phase
reactions at 37
OH (gas phase OH (gas phase
on diesel particles) reactions at 298 K) reactions at
OH (heterogeneous (heterogeneous
on graphite reaction black carbon) T1/2 in fly ash) pressure constant point
boiling
Law
vapor
log
log
T1/2 (on T1/2 (on
T1/2 in air f (h) photolysis photolysis Henry’s
Table 4. Physical Chemical Properties of MDR-Related PAHs
CAS
T1/2 in air; T1/2 in air;
reaction with OH
reaction with
reaction with
reaction with
reaction with OH
Environmental Science & Technology
cases <0.1 (Figure 3a) which apparently is indicative of petrogenic origins. However, this is not consistent with the other MDRs, and previous studies have also questioned the efficiency of the ANT/(ANT + PHE) ratio.18 Galarneau23 suggested that the gas phase reactivity of ANT with OH 3 radicals is a lot faster than for PHE and so can quickly weather emitted PAH mixtures. There are studies where small, yet important, differences in the gas/particle phase partitioning of these two PAHs have been reported which will also affect their fate and also the MDR trend.33,34 All ratios are expected to be impacted by the different physicochemical properties of the individual PAHs (Table 4) which will control reactivity and susceptibility to deposition processes. Different vapor pressures, octanol air partition coefficients and half-lives against degradation processes of every compound will definitely change the specific ratio during its way from source to receptor. The ANT/(ANT + PHE) ratio in the present study is strongly seasonal. Figure 4 shows that generally Q4 and Q1 values (winter season) are higher than Q2 and Q3 values (summer). One interpretation is seasonal differences in sources, e.g., domestic burning of coal and wood.23 However, the other explanation is that the ratio is strongly influenced by seasonal differences (higher reaction rates with OH radicals) in the relative rates of loss of these compounds, undermining any attempts to link MDR to sources. Cabrerizo et al.7 studied the air soil exchange of various PAHs and suggested that, especially for PHE, the soil fugacity is substantially higher than air fugacity, which suggests net volatilization. This indicates that volatilization can also contribute substantially to PHE concentrations in the air, especially during the summer, lowering the ratio and resulting in a seasonal signal for this MDR. According to the same study, emissions from soils might also be important for atmospheric FLT and PYR concentrations. Presumably the contribution of biogenic PAHs to the atmospheric burden (and MDRs) will vary seasonally and locally depending on soil properties, environmental conditions, and other (anthropogenic) sources. Dvorska et al.16 report seasonality of MDRs, however, among the examined MDRs in their study, the ANT/(ANT + PHE) was the least seasonal. Temporal Trends. Figure 5 presents MDRs derived from the national emissions estimates (NAEI) over time (Table S3 and Figure S2 (SI) give detailed MDRs for specific emissions for 1995). This shows that most of the ratios are rather insensitive to the underlying changes in source type over time. Comparisons with measured MDRs in ambient air at different sites over time show little agreement (see Table 5). This powerful message is interpreted as evidence that the ambient air MDRs are insensitive and poor indicators of underlying changes in atmospheric source types and strength. Considerations about the Use of MDRs. The use of PAH MDRs has been criticized in the past, especially when applied to source apportionment in soils, sludges, and sediments, etc.23,24 In the present study on the use of MDRs in air, the 5 MDRs used did not give consistent and reproducible results and they do not reflect known differences in sources in space and time. We conclude, following examination of this longterm and spatially comprehensive data set, that PAH MDRs can be different between locations and with time, but they are very poor markers of known differences in source categories and type. We believe this is because atmospheric sources of PAHs in the UK are not dominated by any clear and strong source type, that the mixture of PAHs in air is quickly 8902
dx.doi.org/10.1021/es202277u |Environ. Sci. Technol. 2011, 45, 8897–8906
Environmental Science & Technology
ARTICLE
Figure 4. Examples of ANT/(ANT + PHE) time trends, where the seasonal variation can be observed: (a) MAN, (b) LON, (c) BIR, (d) GLA, (e) HAZ.
“blended” away from the influence of the few major point sources which exist, and further weathered in the environ-
ment by atmospheric reactions, selective loss processes, and biogenic emissions. We do not recommend the use of MDRs 8903
dx.doi.org/10.1021/es202277u |Environ. Sci. Technol. 2011, 45, 8897–8906
Environmental Science & Technology
ARTICLE
Figure 5. MDRs calculated by the estimated emissions of individual PAHs.1
Table 5. Correlation Analysis between the MDRs Calculated Based on the DEFRA Emission Inventory1 and the Average Annual MDRs emission inventory
BEL
BIR
EDI
GLA
HOL
LIV
HM
HAZ
LON
MAN
MID
SF
ANT/(ANT + PHE)
0.28
0.33
0.76
0.33
0.17
0.19
0.20
0.25
0.20
0.55
0.16
0.36
BaA/(BaA + CHR)
0.77
0.65
0.14
0.43
0.50
0.60
0.09
0.29
0.39
0.35
0.32
0.14
FLT/(FLT + PYR)
0.19
0.56
0.40
0.18
0.76
0.36
-0.73
-0.75
0.02
0.29
0.32
0.30
IPY/(IPY + BPE)
0.75
0.12
0.95
0.70
0.73
0.63
0.13
0.04
-0.68
0.08
0.21
0.17
BaP/BPE
0.47
-0.81
0.62
0.47
0.75
0.56
0.04
0.18
0.25
0.11
0.21
0.30
a
Values in bold and underlined are statistically significant at the 99.95% level of confidence. Values in bold are statistically significant at the 99% level of confidence. Underlined values are statistically significant at the 95% level of confidence
in source apportionment work, unless the source is strong, clearly characterized for PAH composition and stability, and
the ambient measurements are made close to the known source. 8904
dx.doi.org/10.1021/es202277u |Environ. Sci. Technol. 2011, 45, 8897–8906
Environmental Science & Technology
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional data and figures. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT The UK Department of Environment, Food and Rural Affairs (Welsh Assembly Government (WAG), the Northern Ireland Executive, represented by the Department of the Environment in Northern Ireland (DOE), and the Scottish Government), provide financial support for the Toxic Organic Micro-Pollutants Program. Further information is available at http://www.airquality.co.uk and http://www.lec.lancs.ac.uk/research/chemicals_management/tomps. php. The authors also thank AEA Energy & Environment, Dr. Robert Lee, Vicky Burnett, Danielle Lock, Susan Hodson, and David Hughes for help with the collection and analysis of the air samples. ’ REFERENCES (1) NAEI: National Atmospheric Emission Inventory. http://www. naei.org.uk (last visited in May 2011). (2) Nikolaou, K.; Masclet, P.; Mouvier, G. Sources and chemical reactivity of polynuclear aromatic hydrocarbons in the atmosphere A critical review. Sci. Total Environ. 1984, 32, 103–132. (3) Ratola, N.; Alves, A.; Lacorte, S.; Barcelo, D. Distribution and sources of PAHs using three pine species along the Ebro River. Environ. Monit. Assess. 2011No. 10.1007/s10661-011-2014-x. (4) Terzi, E.; Samara, C. Dry deposition of polycyclic aromatic hydrocarbons in urban and rural sites of Western Greece. Atmos. Environ. 2005, 39, 6261–6270. (5) Zhang, Y.; Tao, S. Global atmospheric emission inventory of polycyclic aromatic hydrocarbons (PAHs) for 2004. Atmos. Environ. 2009, 43, 812–819. (6) Gigliotti, C. L.; Brunciak, P. A.; Dachs, J.; Glenn, T. R.; Nelson, E. D.; Totten, L. A.; Eisenreich, S. J. Air-water exchange of polycyclic aromatic hydrocarbons in the New York-New Jersey, USA, Harbor Estuary. Environ. Toxicol. Chem. 2002, 21 (2), 235–244. (7) Cabrerizo, A.; Dachs, J.; Moeckel, C.; Ojeda, M.-J.; Caballero, G.; Barcelo, D.; Jones, K. C. Ubiquitous net volatilization of polycyclic aromatic hydrocarbons from soil and parameters influencing their soil air partitioning. Environ. Sci. Technol. 2011, 45 (11), 4740–4747. (8) Nizzetto, L.; Lohmann, R.; Gioia, R.; Jahnke, A.; Temme, C.; Dachs, J.; Herckes, P.; Di Guardo, A.; Jones, K. C. PAHs in air and seawater along a North South Atlantic transect, processes and possible sources. Environ. Sci. Technol. 2008, 42, 1580–1585. (9) IARC, International Agency for Research on Cancer. Monographs on the Evaluation of Carcinogenic Risks to Humans; IARC: Lyon, France, 1991; pp 43 53. (10) Luch, A. The Carcinogenic Effects of Polycyclic Aromatic Hydrocarbons; Imperial College Press, 2004; ISBN 1-86094-417-5. (11) Cincinelli, A.; del Bubba, M.; Martellini, T.; Gambaro, A.; Lepri, L. Gas-particle concentration and distribution of n-alkanes and polycyclic aromatic hydrocarbons in the atmosphere of Prato (Italy). Chemosphere 2007, 68 (3), 472–478. (12) European Commission. Ambient Air Pollution by Polycyclic Aromatic Hydrocarbons (PAH); Position paper; 2001. (13) Shen, H.; Tao, S.; Wang, R.; Wang, B.; Shen, G.; Li, W.; et al. Global time trends in PAH emissions from motor vehicles. Atmos. Environ. 2011, 45, 2067–2073.
ARTICLE
(14) Yunker, M. B.; Macdonald, R. W.; Vingarzan, R.; Mitchell, R. H.; Goyette, D.; Sylvestre, S. PAHs in the Fraser river basin: A critical appraisal of PAH ratios as indicators of PAH sources and composition. Org. Geochem. 2002, 33, 489–515. (15) Brandli, R. C.; Bucheli, T. D.; Ammann, S.; Desaules, A.; Keller, A.; Blum, F.; Stahel, W. A. Critical evaluation of PAH source apportionment tools using data from the swiss soil monitoring network. J. Environ. Monit. 2008, 10, 1278–1286. (16) Dvorska, A.; Lammel, G.; Klanova, J. Use of diagnostic ratios for studying source apportionment and reactivity of ambient polycyclic aromatic hydrocarbons over Central Europe. Atmos. Environ. 2011, 42 (2), 420–427. (17) Tsapakis, M.; Stephanou, E. G. Occurrence of gaseous and particulate polycyclic aromatic hydrocarbons in the urban atmosphere: Study of sources and ambient temperature effect on the gas/particle concentration and distribution. Environ. Pollut. 2005, 133, 147–156. (18) Brandli, R.; Bucheli, T. D.; Kupper, T.; Mayer, J.; Stadelmann, F. X.; Taradellas, J. Fate of PCBs, PAHs and their source characteristic ratios during composting and digestion of source-separated organic waste in full-scale plants. Environ. Pollut. 2007, 148 (2), 520–528. (19) Cai, Q. Y.; Mo, C. H.; Li, Y. H.; Zeng, Q. Y.; Katsoyiannis, A.; Wu, Q. T.; Ferard., J. F. Occurrence and assessment of polycyclic aromatic hydrocarbons in soils from vegetable fields of the Pearl River Delta, South China. Chemosphere 2007, 68, 159–168. (20) Vasilakos, Ch.; Levi, N.; Maggos, Th.; Hatzianestis, J.; Michopoulos, J.; Helmis, C. Gas-particle concentration and characterization of sources of PAHs in the atmosphere of a suburban area in Athens, Greece. J. Hazard. Mater. 2007, 140, 45–51. (21) Wang, W.; Massey Simonich, S. L.; Xue, M.; Zhao, J.; Zhang, N.; Wang, R.; Cao, J.; Tao, S. Concentrations, sources and spatial distribution of polycyclic aromatic hydrocarbons in soils from Beijing, Tianjin and surrounding areas, North China. Environ. Pollut. 2010, 158 (5), 1245–1251. (22) Usenko, S.; Massey Simonich, S. L.; Haggeman, K. J.; Schrlau, J. E.; Geiser, L.; Campbell, D. H.; et al. Sources and deposition of polycyclic aromatic hydrocarbons to western U.S. National Parks. Environ. Sci. Technol. 2010, 44 (12), 4512–4518. (23) Galarneau, E. Source specificity and atmospheric processing of airborne PAHs: Implications for source apportionment. Atmos. Environ. 2008, 42, 8139–8149. (24) Katsoyiannis, A.; Terzi, H.; Cai, Q. Y. On the use of PAH molecular diagnostic ratios in sewage sludge for the understanding of the PAH sources. Is this use appropriate? Chemosphere 2007, 67, 1337–1339 (25) Zhang, X. L.; Tao, S.; Liu, W. X.; Yang, Y.; Zuo, Q.; Liu, S. Z. Source diagnostics of polycyclic aromatic hydrocarbons based on species ratios. A multimedia approach. Environ. Sci. Technol. 2005, 39, 9109–9114. (26) Meijer, S. N.; Sweetman, A. J.; Halsall, C. J.; Jones, K. C. Temporal trends of polycyclic aromatic hydrocarbons in the U.K. atmosphere: 1991 2005. Environ. Sci. Technol. 2008, 42, 3213–3218. (27) Katsoyiannis, A.; Gioia, R.; Sweetman, A. J.; Jones, K. C. Continuous monitoring of PCDD/Fs in the UK atmosphere: 1991 2008 Environ. Sci. Technol. 2010, 44 (15), 5735–5740. (28) Schuster, J. K.; Gioia, R.; Sweetman, A. J.; Jones, K. C. Temporal trends and controlling factors for polychlorinated biphenyls in the UK atmosphere (1991 2008). Environ. Sci. Technol. 2010, 44 (21), 8068–8074. (29) UK-AIR: Monitoring Netwroks. http://uk-air.defra.gov.uk/ networks (last visited in May 2011). (30) Lee, R. G. M.; Jones, K. C. The influence of meteorology and air masses on daily atmospheric PCB and PAH concentrations at a UK location. Environ. Sci. Technol. 1999, 33 (5), 705–712. (31) Lee, R. G. M.; Coleman, P.; Jones, J. L.; Jones, K. C.; Lohmann, R. Emission factors and importance of PCDD/Fs, PCBs, PCNs, PAHs and PM 10 from the domestic burning of coal and wood in the U.K. Environ. Sci. Technol. 2005, 39 (6), 1436–1447. (32) Lohmann, R.; Northcott, G. L.; Jones, K. C. Assessing the contribution of diffuse domestic burning as a source of PCDD/Fs, PCBs, and PAHs to the U.K. Atmosphere. Environ. Sci. Technol. 2000, 34 (14), 2892–2899. 8905
dx.doi.org/10.1021/es202277u |Environ. Sci. Technol. 2011, 45, 8897–8906
Environmental Science & Technology
ARTICLE
(33) Galarneau, E.; Bidleman, T. F.; Blanchard, P. Seasonality and interspecies differences in particle/gas partitioning of PAHs observed by the Integrated Atmospheric Deposition Network (IADN). Atmos. Environ. 2006, 40 (1), 182–197. (34) Lee, F. S. C., Prater, T. J., Ferris, P. PAH emissions from a stratified charge vehicle with and without oxidation catalyst. In Polynuclear Aromatic Hydrocarbons; Jones, P.W., Leber, P., Eds.; Ann Arbor Science: Ann Arbor, Mich, 1979. (35) Toronto Public Health. http://www.toronto.ca/health/pdf/ cr_appendix_b_pah.pdf (last visited in May 2011). (36) Huckins, J. N.; Petty, J. D.; Booij, K. Monitors of Organic Chemicals in the Environment: Semipermeable Membrane Devices; Springer: New York, 2006. (37) Odabasi, M.; Cetin, B.; Sofuoglu, A. Determination of octanolair partition coefficients and supercooled liquid vapor pressures of PAHs as a function of temperature: Application to gas-particle partitioning in an urban atmosphere. Atmos. Environ. 2006, 40 (34), 6615–6625. (38) Behymer, T. D.; Hites, R. A. Photolysis of polycyclic aromatic hydrocarbons adsorbed on simulated atmospheric particulates. Environ. Sci. Technol. 1985, 19 (10), 1004–1006. (39) Kwamena, N. O. A.; Clarke, J. P.; Kahan, T. F.; Diamond, M. L.; Donaldson, D. J. Assessing the importance of heterogeneous reactions of polycyclic aromatic hydrocarbons in the urban atmosphere using the Multimedia Urban Model. Atmos. Environ. 2007, 41 (1), 37–50. (40) Bunce, N. J.; Dryfhout, H. G. Diurnal and seasonal modeling of the tropospheric half-lives of polycyclic aromatic hydrocarbons. Can. J. Chem. 1992, 70, 1966–1970. (41) Esteve, W.; Budzinski, H.; Villenave, E. Relative rate constants for the heterogeneous reactions of NO2 and OH radicals with polycyclic aromatic hydrocarbons adsorbed on carbonaceous particles. Part 2: PAHs adsorbed on diesel particulate exhaust SRM 1650a. Atmos. Environ. 2006, 40, 201–211. (42) Brunbaker, W. W.; Hites, R. A. OH reaction kinetics of polycyclic aromatic hydrocarbons and polychlorinated debenzo-p-dioxins and dibenzofurans. J. Phys. Chem. 1998, 102, 915–921. (43) Halsall, C. J.; Sweetman, A. J.; Barrie, L. A.; Jones, K. C. Modelling the behaviour of PAHs during atmospheric transport from the UK to the Arctic. Atmos. Environ. 2001, 35 (2), 255–267.
8906
dx.doi.org/10.1021/es202277u |Environ. Sci. Technol. 2011, 45, 8897–8906
ARTICLE pubs.acs.org/est
Enantiomer Fractions of Chiral Perfluorooctanesulfonate (PFOS) in Human Sera Yuan Wang,† Sanjay Beesoon,† Jonathan P. Benskin,† Amila O. De Silva,‡ Stephen J. Genuis,§ and Jonathan W. Martin*,† †
Division of Analytical and Environmental Toxicology, Department of Laboratory Medicine and Pathology, University of Alberta, Edmonton, Alberta, Canada, T6G 2G3 ‡ Aquatic Ecosystem Protection Research Division, Water Science and Technology Directorate, Environment Canada, 867 Lakeshore Road, Burlington, Ontario, Canada, L7R 4A6 § Department of Obstetrics and Gynecology, University of Alberta, Edmonton, Alberta, Canada, T6G 2R7
bS Supporting Information ABSTRACT: Perfluorooctane sulfonate (PFOS) is the most prominent perfluoroalkyl contaminant in humans and wildlife, but there is great uncertainty in exposure pathways, particularly with respect to the importance of PFOS-precursors (PreFOS). We explored the hypothesis that nonracemic proportions of chiral PFOS in serum are qualitative and semiquantitative biomarkers of human PreFOS exposure. A new chiral HPLC-MS/MS method was developed for alpha-perfluoromethyl branched PFOS (1m-PFOS, typically 23% of total PFOS) and applied to enantiomer fraction (EF) analysis in biological samples. In blood and tissues of rodents exposed subchronically to electrochemical PFOS, 1m-PFOS was racemic (EF = 0.4850.511) and no evidence for enantioselective excretion was found in this model mammal. 1m-PFOS in serum of pregnant women, from Edmonton, was significantly nonracemic, with a mean EF ((standard deviation) of 0.432 ( 0.009, similar to pooled North American serum. In a highly exposed Edmonton family (mother, father, and 5 children) living in a house where ScotchGard had been applied repeatedly to carpet and upholstery, EFs ranged from 0.35 to 0.43, significantly more nonracemic than in pregnant women. Semiquantitative estimates of % serum 1mPFOS coming from 1m-PreFOS biotransformation in both subpopulations were in reasonable agreement with model predictions of human exposure to PFOS from PreFOS. The data were overall suggestive that the measured nonracemic EFs were influenced by the relative extent of exposure to PreFOS. The possibility of using 1m-PFOS EFs for assessing the relative contribution of 1m-PreFOS (or PreFOS in general) in biological samples requires further application before being fully validated, but could be a powerful tool for probing general sources of PFOS in environments where the importance of PreFOS is unknown.
’ INTRODUCTION Perfluoroalkyl substances are widely used in industrial, commercial, and consumer products, and are globally distributed in wildlife, humans, and the environment. A major perfluoroalkyl substance in biological samples is perfluorooctane sulfonate (PFOS, C8F17SO3). It is generally the most prominent perfluorinated contaminant in biological samples from around the world14 and is regarded as among the most concentrated xenobiotics in human serum in some geographical areas.5 Like other perfluorinated acids, PFOS is not metabolized in the human body, it has a long elimination half-life in human serum (arithmetic mean of 5.4 years),6 it bioaccumulates in foodwebs,79 and it is extremely persistent in the environment. PFOS is a developmental toxicant in animal models10,11 and was found to disrupt the endocrine system.12 Due to its potential for adverse effects on biota and the environment, PFOS was listed as a “Persistent Organic Pollutant” under the Stockholm Convention.13 Nonetheless, its classification under Annex B of this agreement permits continued r 2011 American Chemical Society
production and widescale use of PFOS and its precursors. For example, in China, total production was greater than 200 t in 2006, and by 2009 there were 66 distinct PFOS-precursors registered with the Inventory of Existing Chemical Substances.14 PFOS-precursors (referred to hereafter as PreFOS) are a vast array of manufactured substances that have the potential to degrade to PFOS. These have diverse structures but are generally substituted perfluorooctanesulfonamides (C8F17SO2NRR0 ). These PreFOS compounds have been used as coatings on textiles, paper, carpeting, and food packaging to achieve oil, stain, and water repellence. It is estimated that the historic environmental emissions of PreFOS were higher than direct emission of PFOS,15,16 but most PreFOS molecules that were known to have been manufactured have never been analyzed in any sample.14 Received: July 7, 2011 Accepted: September 1, 2011 Revised: August 30, 2011 Published: September 01, 2011 8907
dx.doi.org/10.1021/es2023434 | Environ. Sci. Technol. 2011, 45, 8907–8914
Environmental Science & Technology It is notable that in human blood samples from China, 3070% of total organic fluorine remained unaccounted for by the perfluorinated analytes routinely monitored.17 Nonetheless, various forms of PreFOS are indeed detectable in wildlife, human samples, house dust, indoor and outdoor air,1820 and even in the oceans and ambient atmosphere, including at remote locations.21,22 Some evidence exists to show that PreFOS may degrade to PFOS through abiotic hydrolysis23 and atmospheric oxidation,24,25 but the current weight of evidence points to biodegradation as the dominant environmental degradation pathway that converts PreFOS to PFOS.2628 It is proposed that direct exposure to PFOS may be the dominant route of PFOS uptake under some scenarios, but the relative importance of PreFOS might increase and surpass direct exposure under others. This may help to explain the sometimes divergent temporal trends of PFOS observed in biota. For example, after the voluntary phase out of PFOS and PreFOS by the major manufacturer in North America between 2000 and 2002,29 PFOS concentrations declined in wildlife from the Western Arctic and Canadian Archipelago.30,31 On the contrary, in Greenland the concentrations in ringed seals32 and polar bears33 have been rising since 2000. Armitage et al.16 proposed that marine foodwebs around Greenland are being increasingly exposed directly to PFOS through ocean water that has been slowly transported north from legacy source regions in North America and Europe, while further west, the dominant sources to the Arctic marine foodweb must still be from atmospherically transported PreFOS in equilibrium with seawater. The PFOS anion is thought to be persistent in the marine water column, thus it is difficult to explain how PFOS can rapidly decline in foodwebs from the Western Arctic and Canadian Archipelago unless exposure to PFOS included semivolatile forms of PreFOS that rapidly partitioned out of seawater as global atmospheric concentrations of these declined. For humans, although serum concentrations of PFOS and PreFOS have declined in the U.S. since the phase out,34 PFOS concentrations in some Chinese populations have been increasing,35,36 and the relative role of PFOS and PreFOS is unknown. Knowing the human exposure sources in all areas of the world would more effectively facilitate exposure mitigation steps. Potential pathways of human exposure to PFOS and PreFOS are many and include food, food packaging, indoor air and dust, outdoor air, water, and contact with commercial materials. In two human exposure modeling studies, the relative contributions of PFOS and PreFOS to human body burdens from various pathways were estimated.37,38 Both models predicted that PreFOS contributes significantly to the PFOS body burden of the general population, but the accuracy of these models is questionable because the yields of PFOS from various forms of PreFOS are not known in humans. Furthermore, many commercially relevant PreFOS molecules have never been examined in the human environment, thus the total human dose of PFOS from PreFOS metabolism cannot be calculated with certainty, and all existing predictions may be underestimations.14 To date, there is only limited evidence to prove that human, wildlife, or environmental burdens of PFOS are affected by PreFOS, despite that PreFOS emissions are estimated to have been greater than PFOS emissions.14 One major obstacle has been the availability of analytical tools that can differentiate directly absorbed PFOS from PFOS formed through the metabolism of PreFOS. Historical emissions of PFOS and PreFOS
ARTICLE
have always contained a mixture of linear and branched isomers, owing to the electrochemical fluorination (ECF) manufacturing method. We demonstrated a general trend whereby branched PreFOS was metabolized more rapidly than linear PreFOS in vitro,39 thus highly branched PFOS isomer profiles in human blood40,41 may be a biomarker of PreFOS exposure. However, this biomarker has not yet been validated in vivo, and the wide applicability of the principle may be limited because fish were shown to rapidly excrete branched PFOS isomers in one study.42 A more recent consideration for PFOS source tracking is enantiomer fraction (EF) analysis. As noted by Arsenault et al.,43 and later by Rayne et al.,44 some of the major branched PFOS and PreFOS isomers contain a chiral carbon center, and thus each may exist as two nonsuperimposable mirror-image molecules, termed enantiomers. Manufactured or abiotically generated PFOS (i.e., from oxidized, photolyzed, or hydrolyzed PreFOS) should contain a racemic (i.e., 50:50) mixture of enantiomers. Conversely, if PreFOS is degraded to PFOS by enzymes, which are always chiral, it is likely there will be preferential metabolism of one PreFOS enantiomer, leading to a nonracemic mixture of the PFOS metabolite enantiomers. Thus, if nonracemic signatures of any PFOS isomer are observed in biological samples, this may be a useful biomarker of exposure to PreFOS. In a proof-ofprinciple in vitro study,45 an enantioselective chromatographic method was developed for a model 1m-PreFOS molecule (C6F13CF(CF3)SO2N(H)CH2(C6H4)OCH3), and its in vitro biotransformation was indeed enantioselective in human liver microsomes, based on disappearance of the parent compound. However, to date there has been no analytical method capable of resolving the enantiomers of the ultimate metabolite, alphaperfluoromethyl branched PFOS (1m-PFOS). Here we developed a novel chiral HPLC-MS/MS method for enantioseparation of 1m-PFOS which normally composes 23% of total PFOS in real samples. This was applied to measure EFs of 1m-PFOS in individual human serum samples for the first time. To aid with the interpretation the human EFs, the EFs of 1mPFOS were also examined in rodents exposed directly to 3M ECF PFOS.46,47
’ MATERIALS AND METHODS Chemicals. Optima-grade methanol, water, and tetrahydrofuran were purchased from Fisher Scientific (Ottawa, ON, Canada). HPLC-grade formic acid (50%) and triethylamine (g99.5%) were from Fluka (Oakville, ON, Canada). An ECF PFOS standard was provided by 3M Co. (St. Paul, MN). The composition of 1m-PFOS in ECF PFOS is 1.6%,48 and the purity of the ECF PFOS was 86.9%. The racemic 1m-PFOS standard in this study was synthesized at Wellington Laboratories Inc. (Guelph, ON, Canada) and was supplied in methanol at a concentration 4.59 mg/mL. Dilutions of the 1m-PFOS standard were prepared in methanol, or 50% methanol in water. Rat Disposition of 1m-PFOS Enantiomers. Archived samples from two previous PFOS isomer pharmacokinetic studies were analyzed here enantioselectively to confirm if direct exposure to 1m-PFOS would result in racemic EFs in exposed animals. Direct exposure to 1m-PFOS that results in racemic EFs in tissues or biological fluids would support the hypothesis that disposition (uptake, distribution, or excretion) of the 1m-PFOS anion is not enantioselective. Such preliminary evidence is essential if we are to claim that nonracemic 1m-PFOS signatures 8908
dx.doi.org/10.1021/es2023434 |Environ. Sci. Technol. 2011, 45, 8907–8914
Environmental Science & Technology in other organisms are a reflection of enantioselective PreFOS metabolism. From a single-dose oral gavage study (ECF PFOS, 400 μg/kg),46 samples from three individual male SpragueDawley rats, following 38 days of depuration, were randomly chosen for analysis. Available samples included urine, blood, liver, and kidney. From a subchronic (84 d) ad libitum dietary exposure study (425 ng/g ECF PFOS spiked in food),47 samples from 2 female and 1 male SpragueDawley rats were randomly selected for analysis. One of the female rats was sacrificed on day 84, while the other 2 rats were sacrificed after a further 60 days of depuration while consuming a PFOS-free diet. Serum, liver, and kidney samples (collected at the time of sacrifice) were analyzed. PFOS was extracted from all samples with an ion-paring method.2 For details of animal husbandry, PFOS administration, sample collection and treatment, please see refs 46 and 47. Sera of Individual Pregnant Women. Eight samples were randomly selected from a cohort of 271 pregnant women’s serum samples, collected during the 15th16th weeks of pregnancy between December 15, 2005 and June 22, 2006 in Edmonton, AB, Canada.49 All women were g18 years of age and delivered at g22 weeks gestation to live singletons. The samples were extracted by a quantitative solid-phase extraction technique.50 Ethics approval was obtained from the University of Alberta Health Research Ethics Board. Human Sera Collected from “High-Exposure” Family. The high-exposure family consisted of 7 members: father (52 yrs), mother (48 yrs), 4 sons (1523 yrs), and 1 daughter (18 yrs). All family members were healthy and were included in the analysis because they had previously been identified to have high exposure to PFOS. Their house carpet had been treated with ScotchGard at least 8 times over a 17-year period between 1991 and 2007. Serum samples were collected in November 2008. The same solid phase extraction method as mentioned above was used for extraction.50 Ethics approval was obtained from the University of Alberta Health Research Ethics Board. Enantioseparation of 1m-PFOS by HPLC-MS/MS. Two Chiralpak QN-AX HPLC columns (2.1 mm I.D. 150 mm each, 5 μm particles, Chiral technologies Inc. PA) in tandem with a C18 guard column (4.0 mm I.D. 3 mm, 5 μm particles, Phenomenex, Torrance, CA) were used for enantioseparation of 1m-PFOS. Chromatographic conditions were optimized in reversed-phase mode on an HPLC system consisting of an Agilent 1100 binary gradient pump, autosampler, and column oven (Agilent Technologies, Palo Alto, CA). Isocratic elution was applied and the mobile phase consisted of tetrahydrofuran, 0.2 M formic acid, triethylamine, and water in the ratio 70:20:0.05:10, by volume. Flow rate was 0.12 mL/min and column temperature was kept at 15 °C. Injection volume was 2 to 20 μL, depending on the concentration of 1m-PFOS in the samples. Detection of 1m-PFOS enantiomers was performed on an API 5000 triple quadrupole mass spectrometer equipped with a Turbo V ion spray source (Applied Biosystems/MDS SCIEX, Concord, ON, Canada) operating in negative ion mode using multiple reaction monitoring (MRM) for data collection. The precursor ion ([M H], m/z 499) and the unique product ion for 1m-PFOS 51 ([C8F17], m/z 419) were monitored (dwell time 150 ms). This MS/MS transition is free from interference by other PFOS isomers.51 The declustering potential was 64 V, the collision energy was 38 eV, the collision cell exit potential was 9 V, the entrance potential was 11 V, the ion spray voltage was 4000 V, and the source temperature was 500 °C.
ARTICLE
Nitrogen was used as nebulizer, auxiliary, curtain, and collision gas. Mass spectral data were collected and processed by Analyst 1.5 software. Peak Fitting and Calculation of EF. To accurately determine the EF, the partially resolved chromatogram of 1m-PFOS enantiomers was deconvoluted and integrated with Peakfit software (v4.06, Aspire Software International, Ashburn, VA), as previously described.52 As the elution order of the 1m-PFOS enantiomers was unknown, EF was defined here as the quotient of the first-eluting enantiomer (E1) peak area divided by the sum peak area of both enantiomers (E1 + E2). Statistical Approach and Definitions of Racemic and Nonracemic. Statistical differences between EFs in various groups of samples were tested by the Student’s t test (α = 0.05). Characterization of any group of samples as “non-racemic” was based on satisfying two criteria. First, the EF for any group of samples had to be significantly different from the racemic ECF standard. Second, the EFs in any group of samples should not overlap with the range of subtle matrix effect(s) measured for racemic 1m-PFOS spiked into various biological matrices. If these two criteria were not met, the group of samples was defined as racemic.
’ RESULTS AND DISCUSSION Enantioseparation of 1m-PFOS. As previously discussed, 1m-PFOS is the only PFOS isomer that produces a detectable m/z 419 product ion over the calibration range on our system.53 Thus, other PFOS isomers should not interfere in the enantioselective analysis of 1m-PFOS, and no special sample preparation or in-line isomer separation was needed. Unfortunately, PFOS lacks common moieties known to interact with enantioselectors of most chiral stationary phases, thus the enantioseparation of any PFOS isomer was predicted to be difficult. It has been acknowledged that chiral recognition requires a minimum of three simultaneous interactions between the stationary phase and one of the enantiomers, with at least one of these interactions being stereochemically dependent.54 Normally these interactions can include hydrogen bonding, dipoledipole interaction, ππ interaction, inclusion, or steric interaction.55 Because PFOS is anionic, a Chiralpak QN-AX column was chosen for the separation because it contains a quinine-based weak anion exchange site situated in a chiral environment.56,57 The column was previously shown capable of enantioseparating chiral acids in reversed-phase with polar organic solvents,58,59 thus also making it appropriate for mass spectrometry detection. For the selected analyte, 1m-PFOS, the chiral carbon is adjacent to the sulfonate moiety, thus theoretically increasing the chances of accomplishing an enantioseparation (i.e., of getting 3 simultaneous interactions on the stationary phase) on the chosen chiral column. Chiralpak QN-AX columns may be used in polar organic mode or reversed-phase mode. In this study, no separation in polar organic mode was achieved, but enantioseparation could be achieved under reversed-phase conditions using tetrahydrofuran as the organic modifier (methanol, acetronitrile, or dioxane were not effective). In practice, however, insufficient separation was accomplished with a single column, thus two QN-AX columns were connected in series (300 mm total length). Organic modifier types and concentrations, acid and alkaline additives, and separation temperature were further optimized. The acid additive was essential (20% of 0.2 M of formic acid), while small 8909
dx.doi.org/10.1021/es2023434 |Environ. Sci. Technol. 2011, 45, 8907–8914
Environmental Science & Technology
ARTICLE
Figure 2. Mean enantiomer fraction (EF ( 1 SD) of 1m-PFOS over a range of ECF PFOS standard injections (n = 9), in blood and tissues of rats subchronically exposed to ECF PFOS (n = 5 for blood, n = 3 for liver and kidney), and in human sera collected in Edmonton, Alberta, from pregnant women (n = 8) and a family (n = 7) with high exposure to PFOS, likely because of repeated carpet treatment applications of ScotchGard in their home. Dashed gray line shows the racemic reference point, EF = 0.5, while the dashed black lines show the boundaries of the subtle matrix effect (Table S1).
Figure 1. Deconvoluted and smoothed chromatograms (m/z 499/419) of 1m-PFOS on Chiralpak QN-AX showing (a) a racemic 1m-PFOS standard (calculated EF = 0.501), (b) an ECF PFOS standard (calculated EF = 0.500), and (c) human serum from a “high-exposure” individual living in a house with repeated ScotchGard applications (calculated EF = 0.350). The inset in (a) shows the original chromatogram, before deconvolution.
amounts of alkaline additive improved enantioseparation (0.05% triethylamine). Although baseline separation was not achieved, peak shapes were symmetric and resolution of the enantiomers was up to 0.85 (Figure 1). Although such separation was not optimal, it was nonetheless adequate for quantitative applications using peak deconvolution and integration software. When using this peak deconvolution method it was shown that the analytical bias in the EF is less than 1%, even when resolution approaches 0.5, and even when the EF is very nonracemic.52 It is germane to note that the other chiral isomers of PFOS did not enantioseparate under our conditions, and were therefore not monitored in real samples. It is notable that column stability was limited by this method. After approximately 100 injections, the retention time of 1mPFOS decreased from 160 to 115 min, and resolution of the enantiomers decreased from 0.85 to 0.55 for all samples presented in the current study. To assess the possible analytical bias resulting from variation in the performance of the method, two human serum extracts were repeatedly analyzed on the column over the usage period. Calculated EFs for each sample only shifted by 0.001 (maximum), thus the effect of any analytical bias was deemed negligible.
Method Sensitivity and Quality Control. The main objectives in this study were to evaluate the EF of 1m-PFOS in biofluids and tissues (humans and experimental rats); concentrations were not calculated. Nonetheless, the limit of detection (LOD) for the racemic 1m-PFOS standard was 0.45 pg oncolumn, and the limit of quantitation (LOQ) was 1.6 pg oncolumn, based on signal-to-noise ratios (S/N = 3 for LOD and S/N = 10 for LOQ). No carry-over was observed immediately after a high concentration 1m-PFOS standard was injected (i.e., up to 3.7 ng on column). There was also no shift in the baseline within the retention window of the enantiomers when blank matrix (liver extract in which 1m-PFOS concentration was nondetectable by this method) was injected, indicating an absence of nonspecific biological interferences. To avoid false positive detection of nonracemic EFs, possible matrix effects that may affect the electrospray ionization efficiency of the enantiomers were investigated. For example, if a matrix component coeluted with only one enantiomer of a racemic mixture, the signal for that one enantiomer may be artificially suppressed or enhanced, and the resulting EF would seem nonracemic. In this study, matrix effects were determined by spiking standard racemic 1m-PFOS into real sample extracts, including rat liver, kidney, urine, serum, and human sera; unfortunately there are no isotope-labeled internal standards for branched PFOS isomers. In the standard addition experiments, the additional peak area for each enantiomer was calculated by subtracting the prespike peak area from the postspike peak area, and in all types of samples the EF of the added spike was in the range of 0.48 to 0.52, indicating only subtle matrix effects (Table S1 in the Supporting Information). Overall, the developed HPLC-MS/MS analytical and peak integration method were regarded to be reliable and sensitive enough for EF determination of 1m-PFOS in real biological samples. The method was previously applied to two samples of pooled human 8910
dx.doi.org/10.1021/es2023434 |Environ. Sci. Technol. 2011, 45, 8907–8914
Environmental Science & Technology serum,14 but the analytical method is presented for the first time here. Absorption, Distribution, and Excretion of 1m-PFOS in Rats. We previously showed that a model chiral 1m-PreFOS molecule could be metabolized enantioselectively,45 thus it was hypothesized that nonracemic PFOS signatures may be a biomarker of PreFOS exposure. Nonetheless, it could not be ruled out that absorption, distribution, or excretion of PFOS are also enantioselective processes, as has been shown for some chiral drugs.60 Therefore, to aid in the interpretation of 1m-PFOS enantiomer profiles in humans, we examined the disposition of 1m-PFOS enantiomers in rats exposed to an ECF PFOS mixture.46,47 If 1m-PFOS disposition in rats is not enantioselective (i.e., EF = 0.5 in all tissues and biofluid), then finding EF 6¼ 0.5 in human samples would support the possibility of PreFOS exposure. The EF ((standard deviation) of 1m-PFOS in the ECF PFOS mixture was 0.498 ( 0.004 based on repeated measurements of the standard diluted in methanol (see example chromatogram in Figure 1b, and measure of the variation in Figure 2). This EF was not significantly different from 0.5, thus 1m-PFOS in ECF PFOS was confirmed to be racemic, as anticipated. The EFs of 1mPFOS in tissues of SpragueDawley rat exposed subchronically to ECF PFOS are shown in Figure 2. Mean EFs in blood, liver, and kidney were 0.485, 0.504, and 0.511, respectively, for all sample time points combined. Although the effect was subtle, blood (p = 0.04) and liver (p = 0.01), but not kidney (p = 0.14), were statistically different from the ECF standard, however these mean values were within the boundaries of the subtle matrix effect (Figure 2, Table S1). Furthermore, in kidney and urine of SpragueDawley rats exposed to a single dose of PFOS by gavage, the 1m-PFOS EFs in kidney (0.497 ( 0.015) and urine samples (0.503 ( 0.023) were racemic. Urine is the primary excretion route of 1m-PFOS in these animals,46 thus there is no evidence that rats can enantioselectively excrete 1m-PFOS. In subsequent sections, these combined results for a model mammal were used to assume that nonracemic EFs are only attributable to biotransformation of PreFOS; but it is acknowledged that these data are not proof of such. Other potential confounders of the EF are also discussed in the final section. Enantiomer Fraction of 1m-PFOS in Human Sera. The average EF of pregnant women (n = 8) in the current work was 0.432 ( 0.009 (Figure 2), ranging from 0.42 to 0.45. The EF for these samples was nonracemic based on statistical difference from the ECF standard (p < 0.001), and no overlap with the matrix effect range. The EF in these samples was also significantly lower than EFs measured in blood of the rats exposed subchronically to ECF PFOS (p < 0.001). Furthermore, these women had similar EFs compared to what we previously reported for human standard reference serum,61 NIST SRM 1589a (EF = 0.41), collected in 1996 from pooled blood samples of donors who consumed fish caught around the Great Lakes, and NIST SRM 1957 (EF = 0.44), collected in 2004 and pooled from serum of Americans. It has been estimated that PreFOS contributes 10% of PFOS to the body burden of most individuals,37 based on a presumed 20% metabolic conversion rate of PreFOS to PFOS. If this is taken to be true for 1m-PFOS, and using a very conservative assumption that metabolism of 1m-PreFOS is entirely enantiospecific (i.e., one enantiomer totally biodegraded while the other was entirely recalcitrant), an EF as low as 0.45, or as high as 0.55, is the expected outcome from a simple modeling exercise
ARTICLE
(see Table S2, and Figure S1 for model calculations and full set of assumptions). Using the same set of assumptions, the current result for pregnant women (i.e., EF = 0.432) implies that the biotransformation of 1m-PreFOS contributes at least 13% to the body burden of 1m-PFOS in this background population. This is comparable to the 10% figure estimated by Fromme et al. for total PFOS.37 Based on exposure modeling,38,62 it was furthermore predicted that a subgroup of humans in a “high-exposure” scenario (i.e., individuals with very high serum PFOS) might receive up to 6080% of PFOS from biodegradation of PreFOS. To test this hypothesis, serum from a “high-exposure” family (n = 7) was analyzed by the new method. Total PFOS concentrations in serum samples of these family members were 215 times higher than those in background Alberta populations.5 The family’s home had high coverage by carpeting which had been treated with ScotchGard at least 8 times between 1991 and 2007. The household furniture, mainly the family room couch, was also treated at least 5 times. Many ScotchGard formulations were known to contain various PreFOS compounds as residual material or active ingredient.14 The EFs in individual serum samples from this family ranged from 0.350 to 0.430 (mean = 0.392 ( 0.027) (Figure 2), strongly nonracemic, and the EF was significantly lower (i.e., more nonracemic) compared to the pregnant woman population (p = 0.003). The lowest EF (EF = 0.35) measured in the high-exposure family suggests that at least 30% of serum 1m-PFOS comes from 1m-PreFOS (Table S2, Figure S1), however the true contribution from 1m-PreFOS could be higher because of the conservative assumption for the model, discussed above. Overall, the current results for the background population and high exposure family corroborate model predictions38 that the relative importance of PreFOS (in this case 1m-PreFOS) exposure may be greater in highexposure circumstances. For all these human EF measurements, it is important to keep in mind that because exposure to PreFOS can come from the diet,37 it is possible that some biotransformation of PreFOS (to PFOS, or to some intermediate) occurs in the foodweb from which dietary items originate (i.e., from microbes through to vertebrates). Barring any enantiospecific absorption or elimination processes (none were observed in the rat study here for 1m-PFOS), the racemic or nonracemic EF signature of 1m-PFOS in a dietary item should therefore be transferred directly to humans following ingestion. Thus, 1m-PFOS EFs measured in humans likely represent an integration of both direct and indirect PreFOS exposure. Direct PreFOS exposure includes the in vivo human metabolism of PreFOS (i.e., absorbed through food, air, dust, or contact with commercial products), whereas indirect PreFOS exposure would include human exposure to PFOS that is resultant from the ex situ metabolism of PreFOS (e.g., PFOS in food that is a metabolite from PreFOS metabolism in the food chain). However, because all organisms in a foodweb are likely to have different relative enantiospecific biotransformation rates of 1m-PreFOS, it is never advisible to use the human serum 1mPFOS EF as a highly quantitative measure of PreFOS exposure. Rather, the strength of 1m-PFOS EF monitoring is as a qualitative biomarker of PreFOS exposure, and calculation of the minimum contribution of 1m-PreFOS to 1m-PFOS is, therefore, the only quantitative information that might be extrapolated from this technique. Nonetheless, it is important to consider the current limitations and potential confounders for both quantitative and qualitative interpretation of the 1m-PFOS EF. 8911
dx.doi.org/10.1021/es2023434 |Environ. Sci. Technol. 2011, 45, 8907–8914
Environmental Science & Technology Future Application, Limitations, and Potential EF Confounders. Significant PFOS and PreFOS manufacturing con-
tinues today, and further applications of enantiospecific PFOS analysis are warranted to examine hypotheses on the pathways of exposure to other human populations and wildlife. However, limitations of the method must be acknowledged in the current data set, and also in future applications. First, PFOS is very persistent in human blood,6 thus it should be noted that the measured 1m-PFOS EFs do not necessarily reflect a present day exposure scenario. Rather, it is most likely that the EFs represent an integrated historical exposure scenario, and long-term temporal trend studies of 1m-PFOS EFs would therefore be most valuable for assessing how the human exposure scenario has changed over the preceding decades. Second, the current estimates of % 1m-PreFOS contribution to body-burdens are highly conservative minima, but they are also specific to 1m-PFOS, which is a relatively minor PFOS isomer in human serum (i.e., 23% of total PFOS41). Exposure to PreFOS will occur as a mixture of linear and branched isomers (likely ∼70% linear PreFOS),14 but the relative metabolic yield of each PFOS isomer from its corresponding PreFOS isomer is unknown for humans, as are the elimination pharmacokinetics of each PFOS isomer once they are formed as metabolites. Therefore, we cannot quantitatively extrapolate the current result for 1m-PFOS to body burdens of total PFOS (i.e., branched and linear). However, given the highly conservative nature of the model used here to estimate % 1m-PFOS coming from 1mPreFOS, we argue that the same % values might be meaningful as a reasonable lower bound estimate for the body burden of total PFOS. Nonetheless, this will require further validation through continued application of the technique. Third, although it is assumed that the extent of biotransformation of PreFOS to PFOS is the major influence on the measured EFs, it remains a possibility that other biological phenomena might confound the interpretation of measured EFs. As examined here, evidence for the enantiospecific elimination of PFOS was not found in SpragueDawley rats, but it is possible that in humans or other animals that this could occur. Age, sex, and physiological status (including pregnancy) may also be important confounders that would make it difficult to interpret 1mPFOS EF data. Many of the above unknowns will only be understood or addressed through further application of this 1m-PFOS EF monitoring method to environmental samples, foodwebs, human studies, and temporal trend analysis in various matrices. Such studies are underway in our lab. Although still underutilized, years of cumulative EF data for other chiral persistent organic pollutants, such as certain PCBs, have proven to be powerful for understanding the processes controlling their environmental fate.63 Further method development aimed at extending the applicability of the current method to other chiral PFOS isomers is advisible to validate the current approach for PFOS. Improvements of the current analytical method, to increase sample throughput or improve resolution and sensitivity, would also be beneficial to future studies on 1m-PFOS.
’ ASSOCIATED CONTENT
bS
Supporting Information. Matrix effect calculations and a quantitative model of minimum relative contribution of 1mPreFOS to total 1m-PFOS in serum. This material is available free of charge via the Internet at http://pubs.acs.org.
ARTICLE
’ AUTHOR INFORMATION Corresponding Author
*Phone: 1-780-492-1190; fax: 1-780-492-7800; e-mail: jon.martin@ ualberta.ca; mail: 10-102 Clinical Sciences Bldg., University of Alberta Edmonton, Alberta, Canada T6G 2G3.
’ ACKNOWLEDGMENT J.W.M. acknowledges Alberta Ingenuity (New Faculty Grant) and NSERC (Discovery Grant) for supporting the salary of Y.W. S.B. and J.P.B. acknowledge support from Alberta Heritage Foundation for Medical Research and Alberta Ingenuity, respectively. Dr. Scott Mabury (University of Toronto) is acknowledged for his significant shared contributions (study conception and design) to the single-dose and subchronic rodent ECF PFOS exposure studies. Ms. Emily Chan (University of Alberta) is acknowledged for extraction of pregnant women’s serum samples. Alberta Health and Wellness is thanked for support of daily laboratory operations. ’ REFERENCES (1) Giesy, J. P.; Kannan, K. Global distribution of perfluorooctane sulfonate in wildlife. Environ. Sci. Technol. 2001, 35, 1339–1342. (2) Hansen, K. J.; Clemen, L. A.; Ellefson, M. E.; Johnson, H. O. Compound-specific, quantitative characterization of organic fluorochemicals in biological matrices. Environ. Sci. Technol. 2001, 35 (4), 766–770. (3) Martin, J. W.; Smithwick, M. M.; Braune, B. M.; Hoekstra, P. F.; Muir, D. C. G.; Mabury, S. A. Identification of long-chain perfluorinated acids in biota from the Canadian Arctic. Environ. Sci. Technol. 2004, 38 (2), 373–380. (4) Smithwick, M.; Muir, D. C. G.; Mabury, S. A.; Solomon, K. R.; Martin, J. W.; Sonne, C.; Born, E. W.; Letcher, R. J.; Dietz, R. Perfluoroalkyl contaminants in liver tissue from east greenland polar bears (Ursus maritimus). Environ. Toxicol. Chem. 2005, 24 (4), 981–986. (5) Gabos, S.; Zemanek, M.; Cheperdak, L.; Kinniburgh, D.; Lee, B.; Hrudey, S.; Le, C.; Li, X. F.; Mandal, R.; Martin, J. W.; Schopflocher, D. Chemical biomonitoring in serum of pregnant women in Alberta (2005). The Alberta Biomonitoring Program 2008, 71–77. (6) Olsen, G. W.; Burris, J. M.; Ehresman, D. J.; Froelich, J. W.; Seacat, A. M.; Butenhoff, J. L.; Zobel, L. R. Half-life of serum elimination of perfluorooctanesulfonate, perfluorohexanesulfonate, and perfluorooctanoate in retired fluorochemical production workers. Environ. Health Perspect. 2007, 115 (9), 1298–1305. (7) Martin, J. W.; Whittle, D. M.; Muir, D. C. G.; Mabury, S. A. Perfluoroalkyl contaminants in a food web from Lake Ontario. Environ. Sci. Technol. 2004, 38 (20), 5379–5385. (8) Houde, M.; Czub, G.; Small, J. M.; Backus, S.; Wang, X.; Alaee, M.; Muir, D. C. G. Fractionation and bioaccumulation of perfluorooctane sulfonate (PFOS) isomers in a Lake Ontario food web. Environ. Sci. Technol. 2008, 42, 9397–9403. (9) Powley, C. R.; George, S. W.; Russell, M. H.; Hoke, R. A.; Buck, R. C. Polyfluorinated chemicals in a spatially and temporally integrated food web in the western Arctic. Chemosphere 2008, 70 (4), 664–672. (10) Lau, C.; Butenhoff, J. L.; Rogers, J. M. The developmental toxicity of perfluoroalkyl acids and their derivatives. Toxicol. Appl. Pharmacol. 2004, 198 (2), 231–241. (11) Shi, X. J.; Du, Y. B.; Lam, P. K. S.; Wu, R. S. S.; Zhou, B. S. Developmental toxicity and alteration of gene expression in zebrafish embryos exposed to PFOS. Toxicol. Appl. Pharmacol. 2008, 230 (1), 23–32. (12) Jensen, A. A.; Leffers, H. Emerging endocrine disrupters: Perfluoroalkylated substances. Int. J. Androl. 2008, 31 (2), 161–169. (13) Wang, T.; Wang, Y. W.; Liao, C. Y.; Cai, Y. Q.; Jiang, G. B. Perspectives on the inclusion of perfluorooctane sulfonate into the 8912
dx.doi.org/10.1021/es2023434 |Environ. Sci. Technol. 2011, 45, 8907–8914
Environmental Science & Technology Stockholm Convention on persistent organic pollutants. Environ. Sci. Technol. 2009, 43 (14), 5171–5175. (14) Martin, J. W.; Asher, B. J.; Beesoon, S.; Benskin, J. P.; Ross, M. S. PFOS or PreFOS? Are perfluorooctane sulfonate precursors (PreFOS) important determinants of human and environmental perfluorooctane sulfonate (PFOS) exposure? J. Environ. Monit. 2010, 12 (11), 1979–2004. (15) Paul, A. G.; Jones, K. C.; Sweetman, A. J. A first global production, emission, and environmental inventory for perfluorooctane sulfonate. Environ. Sci. Technol. 2009, 43 (2), 386–92. (16) Armitage, J. M.; Schenker, U.; Scheringer, M.; Martin, J. W.; MacLeod, M.; Cousins, I. T. Modeling the global fate and transport of perfluorooctane sulfonate (PFOS) and precursor compounds in relation to temporal trends in wildlife exposure. Environ. Sci. Technol. 2009, 43 (24), 9274–9280. (17) Yeung, L. W. Y.; Miyake, Y.; Taniyasu, S.; Wang, Y.; Yu, H. X.; So, M. K.; Jiang, G. B.; Wu, Y. N.; Li, J. G.; Giesy, J. P.; Yamashita, N.; Lam, P. K. S. Perfluorinated compounds and total and extractable organic fluorine in human blood samples from china. Environ. Sci. Technol. 2008, 42 (21), 8140–8145. (18) Martin, J. W.; Muir, D. C. G.; Moody, C. A.; Ellis, D. A.; Kwan, W. C.; Solomon, K. R.; Mabury, S. A. Collection of airborne fluorinated organics and analysis by gas chromatography/chemical ionization mass spectrometry. Anal. Chem. 2002, 74 (3), 584–590. (19) Shoeib, M.; Harner, T.; Ikonomou, M.; Kannan, K. Indoor and outdoor air concentrations and phase partitioning of perfluoroalkyl sulfonamides and polybrominated diphenyl ethers. Environ. Sci. Technol. 2004, 38 (5), 1313–1320. (20) Olsen, G. W.; Church, T. R.; Larson, E. B.; van Belle, G.; Lundberg, J. K.; Hansen, K. J.; Burris, J. M.; Mandel, J. H.; Zobel, L. R. Serum concentrations of perfluorooctanesulfonate and other fluorochemicals in an elderly population from Seattle, Washington. Chemosphere 2004, 54 (11), 1599–1611. (21) Shoeib, M.; Harner, T.; Vlahos, P. Perfluorinated chemicals in the Arctic atmosphere. Environ. Sci. Technol. 2006, 40 (24), 7577– 7583. (22) Ahrens, L.; Barber, J. L.; Xie, Z.; Ebinghaus, R. Longitudinal and latitudinal distribution of perfluoroalkyl compounds in the surface water of the Atlantic Ocean. Environ. Sci. Technol. 2009, 43, 3122–3127. (23) Mendel, A. 3M technical report: “Analytical methodology on fm 3422”; 1977; AR226-0364. (24) Martin, J. W.; Ellis, D. A.; Mabury, S. A.; Hurley, M. D.; Wallington, T. J. Atmospheric chemistry of perfluoroalkanesulfonamides: Kinetic and product studies of the OH radical and Cl atom initiated oxidation of n-ethyl perfluorobutanesulfonamide. Environ. Sci. Technol. 2006, 40 (3), 864–872. (25) D’Eon, J. C.; Hurley, M. D.; Wallington, T. J.; Mabury, S. A. Atmospheric chemistry of N-methyl perfluorobutane sulfonamidoethanol, C4F9SO2N(CH3)CH2CH2OH: Kinetics and mechanism of reaction with OH. Environ. Sci. Technol. 2006, 40 (6), 1862–1868. (26) Rhoads, K. R.; Janssen, E. M.-L.; Luthy, R. G.; Criddle, C. S. Aerobic biotransformation and fate of N-ethyl perfluorooctane sulfonamidoethanol (N-EtFOSE) in activated sludge. Environ. Sci. Technol. 2008, 42, 2873–2878. (27) Xu, L.; Krenitsky, D. M.; Seacat, A. M.; Butenhoff, J. L.; Anders, M. W. Biotransformation of n-ethyl-n-(2-hydroxyethyl)perfluorooctanesulfonamide by rat liver microsomes, cytosol, and slices and by expressed rat and human cytochromes p450. Chem. Res. Toxicol. 2004, 17, 767–775. (28) Xie, W.; Wu, Q.; Kania-Korwel, I.; Tharappel, J. C.; Telu, S.; Coleman, M. C.; Glauert, H. P.; Kannan, K.; Mariappan, S. V. S.; Spitz, D. R.; Weydert, J.; Lehmler, H. J. Subacute exposure to n-ethyl perfluorooctanesulfonamidoethanol results in the formation of perfluorooctanesulfonate and alters superoxide dismutase activity in female rats. Arch. Toxicol. 2009, 83 (10), 909–924. (29) 3M Company. Health and Environmental Assessment of Perfluorooctane Sulfonic Acid and Its Salts; U.S. EPA docket AR-226-1486; U.S. Environmental Protection Agency: Washington, DC, 2003.
ARTICLE
(30) Butt, C. M.; Muir, D. C. G.; Stirling, I.; Kwan, M.; Mabury, S. A. Rapid response of arctic ringed seals to changes in perfluoroalkyl production. Environ. Sci. Technol. 2007, 41 (1), 42–49. (31) Hart, K.; Gill, V. A.; Kannan, K. Temporal trends (19922007) of perfluorinated chemicals in northern sea otters (Enhydra lutris kenyoni) from south-central Alaska. Arch. Environ. Contam. Toxicol. 2009, 56 (3), 607–614. (32) Bossi, R.; Riget, F. F.; Dietz, R. Temporal and spatial trends of perfluorinated compounds in ringed seal (Phoca hispida) from Greenland. Environ. Sci. Technol. 2005, 39 (19), 7416–7422. (33) Dietz, R.; Bossi, R.; Riget, F. F.; Sonne, C.; Born, E. W. Increasing perfluoroalkyl contaminants in East Greenland polar bears (Ursus maritimus): A new toxic threat to the arctic bears. Environ. Sci. Technol. 2008, 42 (7), 2701–2707. (34) Olsen, G. W.; Mair, D. C.; Reagen, W. K.; Ellefson, M. E.; Ehresman, D. J.; Butenhoff, J. L.; Zobel, L. R. Preliminary evidence of a decline in perfluorooctanesulfonate (PFOS) and perfluorooctanoate (PFOA) concentrations in American Red Cross blood donors. Chemosphere 2007, 68 (1), 105–111. (35) Chen, C. L.; Lu, Y. L.; Zhang, X.; Geng, J.; Wang, T. Y.; Shi, Y. J.; Hu, W. Y.; Li, J. A review of spatial and temporal assessment of PFOS and PFOA contamination in China. Chem. Ecol. 2009, 25 (3), 163–177. (36) Wang, Y.; Jiang, G. The research of human exposure to polybrominated diphenyl ethers and perfluoroocatane sulfonate. Chin. Sci. Bull. 2008, 53 (4), 481–492. (37) Fromme, H.; Tittlemier, S. A.; Volkel, W.; Wilhelm, M.; Twardella, D. Perfluorinated compounds - exposure assessment for the general population in western countries. Int. J. Hyg. Environ. Health 2009, 212 (3), 239–270. (38) Vestergren, R.; Cousins, I. T.; Trudel, D.; Wormuth, M.; Scheringer, M. Estimating the contribution of precursor compounds in consumer exposure to PFOS and PFOA. Chemosphere 2008, 73 (10), 1617–1624. (39) Benskin, J. P.; Holt, A.; Martin, J. W. Isomer-specific biotransformation rates of a perfluorooctane sulfonate (PFOS)-precursor by cytochrome p450 isozymes and human liver microsomes. Environ. Sci. Technol. 2009, 43 (22), 8566–8572. (40) K€arrman, A.; Langlois, I.; van Bavel, B.; Lindstr€om, G.; Oehme, M. Identification and pattern of perfluorooctane sulfonate (PFOS) isomers in human serum and plasma. Environ. Int. 2007, 33, 782–788. (41) Riddell, N.; Arsenault, G.; Benskin, J. P.; Chittim, B.; Martin, J. W.; McAlees, A.; McCrindle, R. Branched perfluorooctane sulfonate isomer quantification and characterization in blood serum samples by HPLC/ESI-MS(/MS). Environ. Sci. Technol. 2009, 43 (20), 7902–7908. (42) Sharpe, R. L.; Benskin, J. P.; Laarman, A. H.; Macleod, S. L.; Martin, J. W.; Wong, C. S.; Goss, G. G. Perfluorooctane sulfonate toxicity, isomer-specific accumulation, and maternal transfer in zebrafish (Danio rerio) and rainbow trout (Oncorhynchus mykiss). Environ. Toxicol. Chem. 2010, 29 (9), 1957–66. (43) Arsenault, G.; Chittim, B.; Gu, J.; McAlees, A.; McCrindle, R.; Robertson, V. Separation and fluorine nuclear magnetic resonance spectroscopic (F-19 NMR) analysis of individual branched isomers present in technical perfluorooctanesulfonic acid (PFOS). Chemosphere 2008, 73 (1), S53–S59. (44) Rayne, S.; Forest, K.; Friesen, K. J. Congener-specific numbering systems for the environmentally relevant C4 through C8 perfluorinated homologue groups of alkyl sulfonates, carboxylates, telomer alcohols, olefins, and acids, and their derivatives. J. Environ. Sci. Health, Part A 2008, 43 (12), 1391–1401. (45) Wang, Y.; Arsenault, G.; Riddell, N.; McCrindle, R.; McAlees, A.; Martin, J. W. Perfluorooctane sulfonate (PFOS) precursors can be metabolized enantioselectively: Principle for a new PFOS source tracking tool. Environ. Sci. Technol. 2009, 43 (21), 8283–8289. (46) Benskin, J. P.; De Silva, A. O.; Martin, L. J.; Arsenault, G.; McCrindle, R.; Riddell, N.; Mabury, S. A.; Martin, J. W. Disposition of perfluorinated acid isomers in Sprague Dawley rats; Part 1: Single dose. Environ. Toxicol. Chem. 2009, 28 (3), 542–554. 8913
dx.doi.org/10.1021/es2023434 |Environ. Sci. Technol. 2011, 45, 8907–8914
Environmental Science & Technology
ARTICLE
(47) De Silva, A. O.; Benskin, J. P.; Martin, L. J.; Arsenault, G.; McCrindle, R.; Riddell, N.; Martin, J. W.; Mabury, S. A. Disposition of perfluorinated acid isomers in Sprague-Dawley rats; Part 2: Subchronic dose. Environ. Toxicol. Chem. 2009, 28 (3), 555–567. (48) Kestner, T. Fluorochemical Isomer Distribution by 19F-NMR Spectroscopy; U.S. Environmental Protection Agency Public Docket AR226-0564; Washington, DC, 1997. (49) Chan, E.; Sandhu, M.; Benskin, J. P.; Ralitsch, M.; Thibault, N.; Birkholz, D.; Martin, J. W. Endogenous high-performance liquid chromatography/tandem mass spectrometry interferences and the case of perfluorohexane sulfonate (PFHXS) in human serum; Are we overestimating exposure? Rapid Commun. Mass Spectrom. 2009, 23, 1405– 1410. (50) Kuklenyik, Z.; Reich, J. A.; Tully, J. S.; Needham, L. L.; Calafat, A. M. Automated solid-phase extraction and measurement of perfluorinated organic acids and amides in human serum and milk. Environ. Sci. Technol. 2004, 38, 3698–3704. (51) Benskin, J. P.; Bataineh, M.; Martin, J. W. Simultaneous characterization of perfluoroalkyl carboxylate, sulfonate, and sulfonamide isomers by liquid chromatography-tandem mass spectrometry. Anal. Chem. 2007, 79 (17), 6455–6464. (52) Asher, B. J.; D’Agostino, L. A.; Way, J. D.; Wong, C. S.; Harynuk, J. J. Comparison of peak integration methods for the determination of enantiomeric fraction in environmental samples. Chemosphere 2009, 75 (8), 1042–1048. (53) Benskin, J. P.; De Silva, A. O.; Martin, J. W. Isomer profiling of perfluorinated substances as a tool for source tracking: A review of early findings and future applications. Rev. Environ. Contam. Toxicol. 2010, 208, 111–160. (54) Pirkle, W. H.; Pochapsky, T. C. Considerations of chiral recognition relevant to the liquid chromatographic separation of enantiomers. Chem. Rev. 1989, 89 (2), 347–362. (55) Okamoto, Y.; Yashima, E. Polysaccharide derivatives for chromatographic separation of enantiomers. Angew. Chem., Int. Ed. 1998, 37, 1020–1043. (56) Lindner, W.; Laemmerhofer, M.; Maier, N. Cinchonan based chiral selectors for separation of stereoisomers. U.S. Patent 6313247, 2001. (57) Hoffmann, C. V.; Pell, R.; L€ammerhofer, M.; Lindner, W. Synergistic effects on enantioselectivity of zwitterionic chiral stationary phases for separation of chiral acids, bases, and amino acids by HPLC. Anal. Chem. 2008, 80 (22), 8780–8789. (58) Krawinkler, K. H.; Maier, N. M.; Sajovic, E.; Lindner, W. Novel urea-linked cinchona-calixarene hybrid-type receptors for efficient chromatographic enantiomer separation of carbamate-protected cyclic amino acids. J Chromatogr., A 2004, 1053 (12), 119–131. (59) Gyimesi-Forras, K.; Maier, N. M.; Kokosi, J.; Gergely, A.; Lindner, W. Enantiomer separation of imidazo-quinazoline-dione derivatives on quinine carbamate-based chiral stationary phase in normal phase mode. Chirality 2009, 21 (1), 199–207. (60) Brocks, D. R. Drug disposition in three dimensions: An update on stereoselectivity in pharmacokinetics. Biopharm Drug Dispos. 2006, 27 (8), 387–406. (61) Martin, J. W.; Asher, B. J.; Beesoon, S.; Benskin, J. P.; Ross, M. S. PFOS or PreFOS? Are perfluorooctane sulfonate precursors (PreFOS) important determinants of human and environmental perfluorooctane sulfonate (PFOS) exposure? J. Environ. Monit. 2010, 12 (11), 1979–2004. (62) Vestergren, R.; Cousins, I. T. Tracking the pathways of human exposure to perfluorocarboxylates. Environ. Sci. Technol. 2009, 43 (15), 5565–5575. (63) Lehmler, H.-J.; Harrad, S. J.; H€uhnerfuss, H.; Kania-Korwel, I.; Lee, C. M.; Lu, Z.; Wong, C. S. Chiral polychlorinated biphenyl transport, metabolism, and distribution: A review. Environ. Sci. Technol. 2010, 44, 2757–2766.
8914
dx.doi.org/10.1021/es2023434 |Environ. Sci. Technol. 2011, 45, 8907–8914
ARTICLE pubs.acs.org/est
A Novel Method Using a Silicone Diffusion Membrane for Continuous 222 Rn Measurements for the Quantification of Groundwater Discharge to Streams and Rivers Harald Hofmann,*,†,‡ Benjamin S. Gilfedder,†,‡ and Ian Cartwright†,‡ † ‡
School of Geosciences, Monash University, Clayton, Victoria, Australia National Centre for Groundwater Research and Training, Australia ABSTRACT: 222Rn is a natural radionuclide that is commonly used as tracer to quantify groundwater discharge to streams, rivers, lakes, and coastal environments. The use of sporadic point measurements provides little information about short- to medium-term processes (hours to weeks) at the groundwatersurface water interface. Here we present a novel method for high-resolution autonomous, and continuous, measurement of 222Rn in rivers and streams using a silicone diffusion membrane system coupled to a solid-state radon-in-air detector (RAD7). In this system water is pumped through a silicone diffusion tube placed inside an outer air circuit tube that is connected to the detector. 222Rn diffuses from the water into the air loop, and the 222Rn activity in the air is measured. By optimizing the membrane tube length, wall thickness, and water flow rates through the membrane, it was possible to quantify radon variations over times scales of about 3 h. The detection limit for the entire system with 20 min counting was 18 Bq m3 at the 3σ level. Deployment of the system on a small urban stream showed that groundwater discharge is dynamic, with changes in 222 Rn activity doubling on the scale of hours in response to increased stream flow.
’ INTRODUCTION Documenting the exchange between surface water and groundwater is critical for understanding the hydrology of surface water bodies, such as rivers, lakes, and oceans. Simplification or even neglecting the connectivity between groundwater and surface water has led to many water allocation and water exploitation problems, particularly “double accounting” of water resources.1 Limited knowledge of the interconnectivity between these reservoirs, which may vary on hourly to yearly time scales in response to external forcing, such as precipitation, floods, drought, and human exploitation, is likely to have caused mismanagement. Indeed, the limited data available, most of which comes from coastal and estuarine environments, suggests that groundwater surface water exchange is highly dynamic and can vary on the scale of hours to days (e.g., refs 27). Sporadic geochemical sampling campaigns (e.g., monthly to half-yearly point samples) are unlikely to capture the full complexity of the groundwatersurface water system. To gain a better understanding of short- to mediumterm groundwatersurface water interactions, high-resolution physical and chemical data is needed that can be measured continuously over an extended period of time. 222Rn is an excellent tool for tracing short-term processes at the interface between groundwater and surface water. 222Rn is a radioactive gas produced in rocks and minerals by the decay of 226Ra, within the 238U decay chain. It accumulates in pore spaces after emanating from the surface of mineral grains through α-recoil.8 222Rn is sparingly soluble in water (Henry constant kH = 1.08 102 mol L1 atm1 9), has a short half-life (∼3.8 days), and, as a noble gas, is chemically and r 2011 American Chemical Society
biologically inert. It degasses, however, to the atmosphere from surface water bodies, and high 222Rn activities occur only in the vicinity of zones of groundwater input. This makes it an ideal natural tracer for identifying groundwater input to streams and rivers. 222Rn activities have been used extensively to constrain groundwater discharge to rivers,1015 lakes, 1618 coastal environments,6,1923 and glacial environments.24 It has also been used to a limited extent to trace recharge of aquifers from rivers via river bank infiltration25,26 and to study hyporheic exchange.12,27 Measurement of Radon in Water. 222Rn activities in water are traditionally measured by taking discrete point samples, followed by liquid scintillation counting (LSC) after liquidliquid extraction with a solvent, or by stripping the radon from the sample and measuring the evolved gas by a radon-in-air detector (e.g., Lucas scintillation cell, ionization chamber, or solid-state silicon detector). The Durridge RAD7 has been widely used in environmental sciences for the last 510 years. LSC is an expensive and time-consuming technique that is generally confined to the laboratory. Furthermore, the short half-life of 222Rn and low surface water activities require that samples are rapidly measured, which essentially precludes using 222Rn as a tracer in remote locations, where transport time to the laboratory is more than a Received: August 2, 2011 Accepted: September 2, 2011 Revised: September 1, 2011 Published: September 02, 2011 8915
dx.doi.org/10.1021/es202683z | Environ. Sci. Technol. 2011, 45, 8915–8921
Environmental Science & Technology
ARTICLE
diffusion membrane for 222Rn, despite its excellent properties for long-term deployment in the field (e.g., very robust, permeable to most gases, and inexpensive;40 Kies, personal communication, 2008).
Figure 1. Silicone radon diffusion “tube in a tube”. The PVC mantles the silicone tube and is connected in a closed loop to a desiccant unit and the RAD7 radon-in-air detector.
few days. Although portable LSC devices are available, they are complicated to operate in the field. Moreover, as natural processes can vary significantly in space and time on a variety of scales, continuous field-based monitoring has many advantages. For example, due to the current lack of robust long-term measurement techniques, there is very little known about how rapidly groundwater discharge to rivers varies. Continuous radon measurements in air and water have been undertaken since the 1950s,28 and continuous measurements in water have been increasingly reported over the last 10 years. A method to measure 222Rn continuously in water using an “air water exchanger” was developed by Friedmann and Hernegger,29 Surbeck,30 and Burnett et al.31 The airwater exchanger developed by Burnett et al. 31 was later modified by Dulaiova et al. 32 and Dimova et al.33 to increase the temporal resolution of the extraction device to about 1520 min. It has become a standard method for continuous measurements and has been applied in the field by Santos et al.,4 Santos et al.,6 Peterson et al.,34 and Stieglitz et al.,23 among others. Membranes, such as microporous polymer tubes, e.g., the Membrana Accurel (Membrana Inc.), have also been used to transfer dissolved 222Rn from water to an air circuit for measurement with a radon-in-air detector.22,3537 Although very efficient for the waterair exchange of 222Rn, these membranes clog easily with sediments and biofilms36,37 and are therefore less suitable for long-term autonomous field deployment. This study was driven by our need for a system that can measure 222Rn activities in streams and rivers continuously to better understand the variability of groundwatersurface water interaction. The system is required to work autonomously in the field for at least 34 weeks and thus must be simple, rugged, and reliable. Another requirement is that the system is responsive enough to capture changes in groundwatersurface water interactions over a few hours to days, for example, to quantify bank storage and return flow following high water events. The method described in this paper is based on 222Rn gas exchange through a silicone membrane installed within an airtight poly(vinyl chloride) (PVC) tube (Figure 1). Although Jonsson et al.38 described a system where silicone tubing was used in combination with a liquid scintillation counting device for 222Rn measurements in Icelandic bores, and Surbeck39 used a silicone tube to measure 222 Rn in karst springs, silicone tubing has rarely been applied as a
’ MATERIALS AND METHODS For the evaluation of the efficiency, response time, and accuracy of the new diffusion method, we have simultaneously compared the silicone diffusion membrane with an airwater exchanger based on the design of Burnett et al.31 The airwater exchanger sprays water at approximately 2.5 L min1 into the exchanging chamber, where radon in the water equilibrates with the chamber headspace. The 222Rn activity in the headspace is then measured in closed circuit using a RAD7 radon-in-air detector. Under ideal conditions, the gas airwater equilibrium for 222Rn should be reached after ∼4(V/F), where V is the air volume in the headspace plus detector and the tubing and F is the water flow rate through the airwater exchanger chamber. In our setup, this should take only 2 min; however, it is unlikely that the exchange is ideal and the results presented in the Results and Discussion section suggest that it takes about 25 min. The semipermeable silicone membrane diffusion tube (abbreviated TINT—tube in a tube) developed here for radon sampling, consists of a standard laboratory grade silicone tube (ROTH GmbH, Germany) installed inside a commercially available 13 mm inner diameter (i.d.) reinforced PVC pressure hose (Figure 1). Silicones are polymerized siloxanes, which are a mixture of organic and inorganic polymers of the generic chemical formula (R2SiO)n, where R represents an organic group, such as a methyl group. The silicone tube outer diameters (o.d.) range from 6 to 7 mm. The PVC hose is sealed at both ends with permanently installed tube connectors where the silicone tube enters and exits. During operation, sample water passes through the inner silicone tube and radon diffuses from the water through the membrane into the sheath air of the PVC tube. Two 8 mm tube connectors are installed in the PVC hose at each end to connect the sheath air with the solid-state radon detector in a closed circuit. The sheath air is circulated (∼1 L min1) counter flow to the water using the internal pump of the detector. The water temperature was measured using a T-type thermocouple (TC direct) with a Lascar data logger to compensate for the temperature dependence of the airwater partitioning coefficient (Ostwald solubility coefficient) for 222Rn. According to Meyer and Schweidler (1916, as cited in Weigel41), the Ostwald solubility coefficient α is related to the water temperature by the empirical relationship: α ¼ ð0:105Þð0:405e0:0502T Þ
ð1Þ
where T is the temperature in °C. At equilibrium the radon activity in water can then be calculated from the measured radon activities in the closed air loop by 222
Rnwater ¼ 222
222
Rnair α
ð2Þ
222
where Rnwater and Rnair are the activities of radon in water and air, respectively. Radon Detection System. 222Rn activities were measured using a Durridge RAD7 radon-in-air monitor, which uses a solid-state passivated ion-implanted silicon crystal held at a ground potential of ∼2200 V to count the decay of 222Rn progeny 218Po and 214 Po. The Si chip electrostatically attracts the positively charged 222 Rn daughter isotope 218Po+ (t1/2 = 3.05 min; E = 6.00 MeV) 8916
dx.doi.org/10.1021/es202683z |Environ. Sci. Technol. 2011, 45, 8915–8921
Environmental Science & Technology
ARTICLE
Figure 2. Setup of TINT optimization and TINTairwater exchanger comparison experiments.
and discriminates each α decay pulse from 218Po and its progeny into a pulse-height spectrum consisting of eight windows. This has the advantage that α counts from 218 Po and 214 Po decay can be separated. By using the 218Po channel, the RAD7 can quantify rapid changes in 222Rn activity (secular equilibrium between 222Rn and 218Po ∼15 min, 218Po and 214Po ∼3 h). During this study, the α counts from the A window (218Po) were used to achieve the maximum temporal resolution. Experimental Setup of the TINT. The most crucial variable for high-resolution continuous measurements is the time required for the TINT to attain airwater equilibrium in response to a change in aqueous 222Rn activities. To optimize the diffusion system for maximum response times, we varied the silicone tube length, the wall thickness, and water flow rate through the tube. Three tube lengths (2, 5, and 10 m) were compared simultaneously at a flow rate of 1 L min1 to determine the effect of the membrane surface area on the time required to reach equilibrium. Three 5 m silicone tubes, with wall thicknesses of 0.5, 1.0, and 1.5 mm, were compared simultaneously to determine if the more robust 1.5 mm tubes reduced the exchange rate significantly compared to the 0.5 mm tubing. Finally, three water flow rates were tested on a 10 m TINT with 1 mm wall thickness by adjusting the flow to 0.66, 1.3, and 2.5 L min1. All experiments were conducted under field conditions on the bank of the Avon river, southeast Victoria, Australia. Groundwater was constantly pumped into a 100 L water drum with a submersible impeller pump at a flow rate of 6 L min1 from 5 m bore depth. At this flow rate, the drum was constantly overflowing to ensure a continual renewal of groundwater. A second barrel was filled with river water that had low 222Rn activities (∼1 Bq L1). To simulate rapidly changing 222Rn activities in a river, and to observe the rate of the systems response to this change, we installed a three-way valve between the two barrels and the TINT/exchanger pumps (Figure 2). This allowed us to rapidly switch between water with high and low 222Rn activities, as may occur in a river during a storm event or in a bore if pumping induces surface water inflows. Water was pumped though the airwater exchanger at about 2.5 L min1, whereas the TINTs were generally run at a flow rate of 1 L min1. The TINTs were operated in closed circuit with both drums. All experiments were run using a 5 min counting time (TINT and exchanger), which resulted in a precision of ∼5% relative standard deviation (RSD) when the airwater exchange was at equilibrium.
Figure 3. RAD7 and TINT detection limit experiment.
Point Measurements. Discrete water samples were taken from the water drums throughout the experiment, and 222Rn activities were measured using the RAD7 combined with the commercial RAD H2O kit (Durridge Inc.) (250 mL sample bottles, 10 min integration time). The bottles were filled to the top and then closed using the RAD H2O kit bottle head. 222Rn was stripped from the water by purging the sample with air, which was cycled through the RAD7 in a closed loop. Five replicate measurements of 222Rn activities were made on each sample. A similar procedure was applied to river water samples, using a 500 mL sample bottle with 30 min integration time. The 500 mL sampling bottles had been previously cross-calibrated against the RAD H2O kit. Detection Limit. The background count rate of the RAD7 instrument was measured by running the system in closed circuit with an inline activated carbon filter, which quantitatively adsorbs all radon from the air stream. With this setup, the instrument background (detection limit) for 222Rn at 3σ was 8 Bq m3 using a 20 min counting time (Figure 3). The effect of the TINT on background noise was determined by passing water with low 222 Rn activities (14 Bq m3) thorough a 10 m 1 mm TINT 8917
dx.doi.org/10.1021/es202683z |Environ. Sci. Technol. 2011, 45, 8915–8921
Environmental Science & Technology
ARTICLE
Figure 4. Comparison between a 1 mm 10 m TINT diffusion membrane (black) and the airwater exchanger.
and using a variety of counting times. The detection limit was then estimated by the 3σ value of repeated measurements at this low level.42 Surprisingly, there was very little difference in the detection limit when using 20, 40, 60, or 120 min counting time, e.g., 3σ value of 13 Bq m3 for 2 h versus 18 Bq m3 for 20 min (Figure 3). This suggests that the best option for continuous measurements is to use a 20 min integration time to increase the temporal resolution of the data with little loss in detection probability.
’ RESULTS AND DISCUSSION Comparison with the AirWater Exchanger. The performance of the TINT was compared to the well-established air water exchanger and grab samples. For this experiment, we used a 10 m TINT with 1 mm wall thickness silicone tube. The flow rates for both the airwater exchanger and the TINT were 2.5 L min1. To simulate varying 222Rn activities, we changed the water source from river water to the bore water using the threeway valve (Figure 2). Once equilibrium was reached, we again changed from a high to low 222Rn source by changing the valve back to the river water reservoir. Figure 4 shows that both extraction systems react rapidly to a change in aqueous 222Rn activities. While the airwater exchanger reached equilibrium after 25 min, it took approximately 180 min for the TINT. Nonetheless, the equilibrium 222Rn activities for both the airwater exchanger and the TINT did not differ within the error of the measurements. The 222Rn activities of the point samples taken during the experiment also agreed well with both systems. The 222Rn activities of both point and airwater exchanger measurements increased slightly (from 18 000 to 20 000 Bq m3) over the course of the experiment, probably due to variations in 222Rn activities of the groundwater that is drawn from different parts of the aquifer. After 400 min, the source of water was changed back to river water (222Rn activity ∼1000 Bq m3) to produce rapidly decreasing 222Rn activities. 222Rn activities decreased to river water levels in ∼35 min using the airwater exchanger and ∼200 min using the TINT. The delayed response of the TINT is due to 222 Rn diffusion through the membrane, which, as expected, is slower than the direct airwater exchange of the exchanger.33 The response time of the TINT during periods of increasing and decreasing 222Rn concentrations is similar, between 200 and 300 min. TINT Optimization. The diffusion 222Rn across the silicone tube membrane may be influenced by a number of factors, such
Figure 5. Dependence of equilibrium time on silicone tube length.
as the surface area available for exchange, thickness of the tube wall, and the water flow rate through the tube. The surface area available for exchange is directly proportional to the tube length. The ratio between the surface area A of the silicone tube available for 222Rn exchange and the volume V in the PVC hose (TINT dead volume) does not change with increasing tube length, as A/V scales as 2/r. However, the A/V ratio increases with silicone tube length when the total dead volumes (RAD7750 mL, the desiccant tube ∼350 mL, connection tubing ∼50 mL) are included in the calculation of V. In our system the A/V increases by ∼300% when using a 10 m TINT rather than a 2 m TINT, including the dead volume of 1150 mL. To test the effect this has on the instrument response time, TINTs (wall thickness 1 mm) of 2, 5, and 10 m length were simultaneously compared. As can be seen in Figure 5, the time taken to reach equilibrium varied significantly between the three lengths. The 10 m tube had reached equilibrium after 200 min, the 5 m tube reached a similar equilibrium activity after 350 min, while the 2 m tube had not reached equilibrium when the test was stopped after 500 min. According to Fick’s first law of diffusion, the flux across a membrane is inversely proportional to its thickness. However, a thicker wall will add extra stability and robustness to the TINT design. Thus, if a reduction in the response time with increasing wall thickness is minimal, it may be beneficial to use a thicker silicone tube. To test this, we compared silicone tubing with 0.5, 1, and 1.5 mm wall thickness on a 5 m TINT at a flow rate of 1 L min1. As expected the TINT with 0.5 mm silicone tubing reached equilibrium first (∼250 min), followed by the 1 mm tube (300 min) (Figure 6). The TINT with 1.5 mm wall thickness had not reached equilibrium after 330 min. This tube with 1.5 mm wall thickness had a slightly smaller i.d. (4 mm) than the 0.5 and 1 mm TINTs (5 mm), which may affect the rate of 222Rn diffusion due to the lower surface area available for exchange. When we apply a correction factor proportional to the difference in surface area between the 1.5 mm and the other membranes (20%), the modeled response of the 1.5 mm tube is closer to that of the other tubes, but it is still slower at reaching equilibrium. By differentiating the linear sections of each curve in Figure 6, we find a very strong inverse correlation between the gradient and the wall thickness (r2 = 0.99). Overall, the 1.5 mm silicone tube is 8918
dx.doi.org/10.1021/es202683z |Environ. Sci. Technol. 2011, 45, 8915–8921
Environmental Science & Technology
ARTICLE
Figure 8. Continuous 222Rn, depth, and temperature measurements in an urban creek. Figure 6. Effect of silicone tube wall thickness on diffusion rate and time to equilibrium. The 1.5 mm corrected trend derives from applying a 20% correction factor applied to the 1.5 mm tube due to slightly different surface area available for radon exchange.
container”, i.e., the radon activity in the air circuit being measured. The exhalation rate is calculated from measured data by E¼
Figure 7. Effect of flow rate on radon equilibrium time in a 1 mm 10 m TINT.
prohibitively slow in reaching equilibrium, and so either 0.5 or 1 mm tubing should be used to achieve a satisfactory response. Finally, the effect of flow rate through the silicone tube was determined on a 10 m TINT. The time required to reach airwater equilibrium decreased from 350 to 250 min, as we increased the flow rate from 0.66 to 1.33 and finally to 2.5 L min1 (Figure 7). With the use of those data, it is possible to calculate the diffusion rate constant D for the silicone tube using a slightly modified form of eqs 1 and 2, presented in Jiranek and Svoboda:43 D¼d
E Csc Crc
ð3Þ
where D is the diffusion constant (m2 s1), E is the radon exhalation rate (Bq m2 s1), Csc is the radon activity in the “source” container, which in our case is the water passing through the silicone tube, and Crc is the radon activity in the “receiving
λV ðCrc ðt2 Þ Crc ðt1 Þeλðt2 t1 Þ Þ Að1 eλðt2 t1 Þ Þ
ð4Þ
where λ is the 222Rn decay constant (2.1 106 s1), V is the volume of the air circuit, A is the surface area of the membrane, converted to a sheet by A = 2πrL. t1 is the time at the start of the experiment, and t2 was the time required for the radon activity to reached 1 e1 (∼63%) of the final activity in the closed air circuit. Since we are comparing gas-phase activities with waterphase activities, the equations must be modified to account for the preferential partitioning of 222Rn into the gas phase. Thus, the Crc is converted into water-phase activity using eqs 1 and 2. Using these expressions, the diffusion coefficient varies from 4.6 1010 to 9 1010 m2 s1, increasing with the flow rate. It is thought that increased pressure in the tubes at higher flow rates slightly stretches the molecular lattice (SiO(CH3)2) of the silicone membrane, increasing the effective pore size of the material and, therefore, the diffusion rate. D for the lowest flow (and so the least deformation of the material structure) was similar to the value of 3.2 1010 m2 s1 found by Mamedov et al.44 for a silicone membrane at atmospheric pressure. Field Deployment. To test the suitability of the TINT for continuous field deployment, we installed an optimized system, consisting of a 10 m TINT with 1 mm wall thickness, on the bank of a small stream near Monash University (Scotchman’s creek). A water level logger (pressure sensor) was also placed in the stream to correlate changes in 222Rn with changes in stream depth. The system was run continuously for approximately 4 days. Even over this short time span, 222Rn activities doubled (from 290 to 680 Bq m3) (Figure 8). These variations are considerably larger than can be accounted for by degassing and decay and indicate the dynamic nature of groundwater surface water interactions. The system also has been deployed for 3 months on a research raft in a coastal wetland in Victoria without an external power source and with the replacement of the desiccant on monthly intervals. As with Scotchman’s creek, there are significant changes in 222Rn activity over this time scale (Figure 9). In summary, the results from the optimization experiments show that tube length, membrane wall thickness, and flow rate are all important factors influencing the response of silicone tube 8919
dx.doi.org/10.1021/es202683z |Environ. Sci. Technol. 2011, 45, 8915–8921
Environmental Science & Technology
ARTICLE
’ ACKNOWLEDGMENT This study was supported by the National Centre for Groundwater Research and Training P3 program. The National Centre for Groundwater Research and Training is an Australian Government initiative supported by the Australian Research Council and the National Water Commission. We thank Professor Antoine Kies (University of Luxembourg) for the discussions on the use of membranes for radon and CO2 extraction, in particular the use of silicone tubing. We also thank Chris Pierson from the Geoscience workshop for help in the construction of the TINT setup. Figure 9. Continuous wetlands in Victoria.
222
Rn using the TINT on a raft in coastal
membranes to 222Rn diffusion. When these parameters are optimized, it is possible to construct a continuous 222Rn monitoring system that can capture changes in 222Rn activities, which allow determining variations in groundwater discharge to streams and rivers on a variety of scales from a few hours to a few months. This could be particularly useful for investigating processes, such as bank return flow, the impacts of storm events on groundwater discharge and recharge, and changes in the regional groundwater input to rivers and streams. The TINT system requires minimal power consumption, with a small pump capable of delivering 1 L min1, drawing only a few hundred milliamps, which is essential for long-term deployments in remote areas. The minimum power consumption we were able to achieve included a Rotek Inc. rotary pump drawing significantly under 1000 mA. The system can be supplied with a regular solar panel (e.g., 80 W) and a 712 Ah battery as buffer. Furthermore, the consumption of desiccant is significantly reduced when using a membrane compared to an airwater exchanger, as less humidity enters the air circuit, which results in longer maintenance intervals of at least a few weeks. In contrast to the microporous polymer membranes, which are constructed of porous fibres, clogging by sediments is not a problem, even without prefiltration. In the experiments conducted here, the bore water had high sediment loads (the water was turbid and gray), but this did not have any effect on the response, even after long exposure. The silicone tube is also very flexible, in contrast to the rather brittle microporous polymer tubing, which allows it to be molded into convenient shapes. In our system the TINT is rolled into a loop so that the entire 10 m length only occupies a 40 40 15 cm3 volume. Moreover, there is little chance of water entering and destroying the sensitive detection system as may be the case with the airwater exchanger if an outlet is blocked or if the air return tubing develops a leak. Thus, although the TINT does have a slower response than the other methods, it is more suitable for long-term, autonomous measurement. For systems that vary rapidly, such as tidal areas or radon mapping, the airwater exchanger would be a more suitable system, but regular supervision is required. The first field data presented here demonstrates that groundwater exchange with rivers and wetlands is complex and that it varies on the scale of hours to days. Thus, a continual monitoring system is needed to understand the real complexity of the groundwatersurface water system.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +61 (3) 9905 5786; fax: +61 (3) 9905 4903; e-mail:
[email protected].
’ REFERENCES (1) Sophocleous, M. Interactions between groundwater and surface water: The state of the science. Hydrogeol. J. 2002, 10, 52–67. (2) Taniguchi, M.; Iwakawa, H. Submarine groundwater discharge in Osaka Bay, Japan. Limnology 2004, 5, 25–32. (3) Crusius, J.; Koopmans, D.; Bratton, J.; Charette, M.; Kroeger, K.; Henderson, P.; Ryckman, L.; Halloran, K.; Colman, J. Submarine groundwater discharge to a small estuary estimated from radon salinity measurements and a box model. Biogeosciences 2005, 2, 141–157. (4) Santos, I. R.; Dimova, N.; Peterson, R. N.; Mwashote, B.; Chanton, J.; Burnett, W. C. Extended time series measurements of submarine groundwater discharge tracers (222Rn and CH4) at a coastal site in Florida. Mar. Chem. 2009, 113, 137–147. (5) de Weys, J.; Santos, I. R.; Eyre, B. D. Linking groundwater discharge to severe estuarine acidification during a flood in a modified wetland. Environ. Sci. Technol. 2011, 45, 3310–3316. (6) Santos, I. R.; Peterson, R. N.; Eyre, B. D.; Burnett, W. C. Significant lateral inputs of fresh groundwater into a stratified tropical estuary: Evidence from radon and radium isotopes. Mar. Chem. 2010, 121, 37–48. (7) Burnett, W. C.; Peterson, R. N.; Santos, I. R.; Hicks, R. W. Use of automated radon measurements for rapid assessment of groundwater flow into Florida streams. J. Hydrol. 2010, 380, 298–304. (8) Porcelli, D.; Swarzenski, P. The behavior of U- and Th-series nuclides in groundwater. In Uranium-Series Geochemistry; Bourdon, B., Henderson, G. M., Lundstrom, C. C., Turner, S. P., Eds.; Reviews in Mineralolgy and Geochemistry; Mineralogy Society of America, 2003. (9) Sander, R. Compilation of Henry‘s law constant for inorganic and organic species of potential importance in environmental chemistry, 1999. http://www.henrys-law.org. (10) Wanninkhof, R.; Mulholland, P.; Elwood, J. Gas exchange rates for a first order stream determined by deliberate and natural tracers. Water Resour. Res. 1990, 26, 1621–1630. (11) Cook, P. G.; Favreau, G.; Dighton, J.; Tickell, S. Determining natural groundwater influx to a tropical river using radon, chlorofluorocarbons and ionic environmental tracers. J. Hydrol. 2003, 277, 74–88. (12) Cook, P.; Lamontagne, S.; Berhane, D.; Clark, J. Quantifying groundwater discharge to Cockburn River, South-Eastern Australia, using dissolved gas tracers 222Rn and SF6. Water Resour. Res. 2006, 42, 1–12. (13) Mullinger, N.; Binley, A.; Pates, J.; Crook, N. Radon in Chalk streams: Spatial and temporal variation of groundwater sources in the Pang and Lambourn Catchment, UK. J. Hydrol. 2007, 339, 172–182. (14) Gleeson, T.; Novakowsji, K.; Cook, P.; Kyser, T. Constraining groundwater discharge in a large watershed: integrated isotopic, hydraulic, and thermal data from the Canadian Shield. Water Resour. Res. 2009, 45, W08402. (15) Kies, A.; Hofmann, H.; Tosheva, Z.; Hoffmann, L.; Pfister, L. Using 222Rn for hydrograph separation in a micro basin (Luxembourg). Ann. Geophys. 2005, 45, 101–107. (16) Kluge, T.; Ilmberger, J.; von Rohden, C.; Aeschbach-Hertig, W. Tracing and quantifying groundwater inflow into lakes using a simple method for radon-222 analysis. Hydrol. Earth Syst. Sci. 2007, 11, 1621–1631. (17) Schmidt, A.; Gibson, J.; Santos, I.; Schubert, M.; Tattrie, K.; Weiss, H. The contribution of groundwater discharge to the overall 8920
dx.doi.org/10.1021/es202683z |Environ. Sci. Technol. 2011, 45, 8915–8921
Environmental Science & Technology water budget of two typical boreal lakes in Alberta/Canada estimated from a radon mass balance. Hydrol. Earth Syst. Sci. 2010, 14, 79–89. (18) Dimova, N.; Burnett, W. Evaluation of groundwater discharge into small lakes based on the temporal distribution of radon-222. Limnol. Oceanogr. 2011, 56, 486–494. (19) Kim, G.; Hwang, D.-W. Tidal pumping of groundwater into the coastal ocean revealed from submarine 222Rn and CH4 monitoring. Geophys. Res. Lett. 2002, 29, 1678. (20) Lambert, M. J.; Burnett, W. C. Submarine groundwater discharge estimates at a Florida coastal site based on continuous radon measurements. Biogeochemistry 2003, 66, 55–73. (21) Burnett, W.; Dulaiova, H.; Stringer, C.; Peterson, R. Submarine groundwater: Its measurement and influence on the coastal zone. J. Coastal Res. 2004, SI39, 35–38. (22) Dulaiova, H.; Camilli, R.; Henderson, P. B.; Charette, M. A. Coupled radon, methane and nitrate sensors for large-scale assessment of groundwater discharge and non-point source pollution to coastal waters. J.Environ. Radioact. 2010, 101, 553–563. (23) Stieglitz, C.; Cook, P. G.; Burnett, W. C. Inferring coastal processes from regional-scale mapping of 222Radon and salinity: examples from the Great Barrier Reef, Australia. J. Environ. Radioact. 2010, 101, 544–552. (24) Kies, A.; Surbeck, H.; Tosheva, Z. Natural radioactive isotopes in glacier studies. In The Dynamics and Mass Budget of Arctic Glaciers— Extended Abstracts—Workshop and GLACIODYN Meeting, Pontresina (Switzerland); Oerlemans, J., Tijm-Reijmer, C., Eds.; IASC Working group on Arctic Glaciology, Institute for Marine and Atmospheric Research: Utrecht, The Netherlands, 2007; pp 5760. (25) Hoehn, E.; van Gunten, H. Radon in groundwater: A tool to access infiltration from surface water to aquifers. Water Resour. Res. 1989, 25, 1795–1803. (26) Dehnert, J.; Nestler, W.; Freyer, K.; Treutler, H.-C. Messung der infiltrationsgeschwindigkeit von oberfl€achenwasser mit hilfe des nat€urlichen isotops radon-222. Grundwasser 1999, 1, 19–30. (27) Lamontagne, S.; Cook, P. Estimation of hyporheic water residence time in situ using 222Rn disequilibrium. Limnol. Oceanogr.: Methods 2007, 5, 407–416. (28) George, A. C. World history of radon research and measurement from the early 1900’s to today. AIP Conf. Proc. 2008, 1034, 20–33. (29) Friedmann, H.; Hernegger, F. A Method for continuous measurement of radon in water of springs for earthquake prediction. Geophys. Res. Lett. 1978, 5, 565–568. (30) Surbeck, H. Radon monitoring in soils and water. Nucl. Tracks Radiat. Meas. 1993, 22, 463–468. (31) Burnett, W.; Kim, G.; Lane-Smith, D. A continuous monitor for assessment of 222Rn in the coastal ocean. J. Radioanal. Nucl. Chem. 2001, 249, 167–172. (32) Dulaiova, H.; Peterson, R.; Burnett, W.; Lane-Smith, D. A multi-detector continuous monitor for assessment of 222Rn in the coastal ocean. J. Radioanal. Nucl. Chem. 2005, 263, 361–365. (33) Dimova, N.; Burnett, W. C.; Lane-Smith, D. Improved automated analysis of radon (222Rn) and thoron (220Rn) in natural waters. Environ. Sci. Technol. 2009, 43, 8599–8603. (34) Peterson, R. N.; Santos, I. R.; Burnett, W. C. Evaluating groundwater discharge to tidal rivers based on a Rn-222 time-series approach. Estuarine, Coastal Shelf Sci. 2010, 86, 165–178. (35) Surbeck, H. A radon-in-water monitor based on fast gas transfer membranes. Proceedings of the International Conference on Technologically Enhanced Natural Radioactivity (TENR) Caused by NonUranium Mining, Szczyrk, Poland, 1996. (36) Schmidt, A.; Schlueter, M.; Melles, M.; Schubert, M. Continuous and discrete on-site detection of radon-222 in ground-and surface waters by means of an extraction module. Appl. Radiat. Isot. 2008, 66, 1939–1944. (37) Schubert, M.; Schmidt, A.; Paschke, A.; Lopez, A.; Balcazar, M. In situ determination of radon in surface water bodies by means of a hydrophobic membrane tubing. Radiat. Meas. 2008, 43, 111–120. (38) Jonsson, G.; Theodorsson, P.; Sigurdsson, K. Auto-Radon: A new automatic liquid scintillation system for monitoring radon in water
ARTICLE
and air. In LSC 2005, Advances in Liquid Scintillation Spectrometry; Chalupnik, S., Sch€onhofer, F., Noakes, J., Eds.; Arizona Board of Regents on behalf of the University of Arizona, 2006; pp 119123. (39) Surbeck, H. Dissolved gases as natural tracers in karst hydrogeology; radon and beyond. Proceedings of Multidisciplinary Approach to Karstwater Protection Strategy, UNESCO Course, 2005. (40) Surbeck, H.; Deflorin, O.; Kloos, O. Spatial and temporal variations in the Uranium series background in Alpine groundwater. In Uranium in the Environment, Mining Impact and Consequences; Merkel, B. J., Hasche-Berger, A., Eds.; Springer, 2006; pp 831839. (41) Weigel, F. Radon Chemiker Zeitung 1978, 102, 287–299. (42) Sanders, P.; Lippincott, R.; Eaton, A. Determining quantification level for regulatory purposes. J.—Am. Water Works Assoc. 1996, 88, 104–114. (43) Jiranek, M.; Svoboda, Z. Transient radon diffusion through radon-proof membranes: A new technique for more precise determination of the radon diffusion coeffient. Build. Environ. 2009, 44, 1–10. (44) Mamedov, F.; Cermak, P.; Smolek, K.; Stekl, I. Measurement of radon diffusion through shielding foils for the SuperNEMO experiment. J. Instrum. 2011, 6, C01068.
8921
dx.doi.org/10.1021/es202683z |Environ. Sci. Technol. 2011, 45, 8915–8921
ARTICLE pubs.acs.org/est
Evaluation of the Boron Tolerant Grass, Puccinellia distans, as an Initial Vegetative Cover for the Phytorestoration of a Boron-Contaminated Mining Site in Southern California Amanda R. Stiles,† Chunguang Liu,‡ Yuriko Kayama,† Josephine Wong,† Harvey Doner,§ Roger Funston,|| and Norman Terry*,† †
Department of Plant and Microbial Biology, University of California, Berkeley, California 94720-3102, United States Laboratory of Pollution Processes and Environmental Criteria, Nankai University, Tianjin 300071, China § Department of Environmental Science, Policy and Management, University of California, Berkeley, California 94720, United States Rio Tinto Minerals, Tehachapi, California 93516, United States
)
‡
ABSTRACT: Land damaged by boron (B) mining should be restored to its natural state with a zero net impact on biodiversity. In an earlier study (Environ. Sci. Technol. 2010, 44, 7089 7095), we characterized a Turkish ecotype of the grass, Puccinellia distans, which exhibited extreme tolerance to B. Here we evaluated the use of a US ecotype of P. distans as an initial vegetative cover for the phytorestoration of a B mine in southern California. Hydroponic studies revealed that this P. distans ecotype tolerated B concentrations >100 mg B/L and could be germinated and grown in B-contaminated soils taken from the sites to be restored. P. distans grew well in moderately B-contaminated soil (∼88 mg B/L saturated extract) amended with added organic matter (peat moss); other soil treatments such as gypsum addition or pH correction were not needed. P. distans also grew in severely B-contaminated soil (∼1506 mg B/L) provided that toxic levels of soil B were diluted by the addition of sand and/or organic matter. Our results provide evidence in support of the concept of using the US ecotype of P. distans as an initial vegetative cover for the phytorestoration of B-contaminated soil.
’ INTRODUCTION Boron (B) mining can result in significant damage to the local ecosystem. In mining operations, the inevitable removal of the topsoil with its original vegetation leaves a land surface that typically consists of nutrient-poor, skeletal subsoil.1 After the mining activities have been terminated, it is important to return the mine-damaged land back to its original state insofar as this is possible. In the present work, we are focusing on the phytorestoration of a specific mining site located in the Mojave Desert, a goal that is especially problematic because of the extremely low amounts of precipitation.2 The Mojave Desert has an average precipitation of 95 mm/year in the cool season and 35 mm/year in the warm season (data averaged from 1893 to 20013). Furthermore, large areas of the mining site will suffer from moderate to high levels of B contamination of the soil due to the release of airborne particles, water-borne transport, and the transport of soil from one horizon to another (Roger Funston, personal communication). Specific performance goals for the restoration of this site have been developed with respect to plant density, species diversity, and the extent of plant cover with the overall goal of returning the mined land to its natural state and to have a zero net impact on biodiversity. How may phytorestoration of B-contaminated mining sites be achieved? Soil B includes forms that are water-soluble, adsorbed to solid surfaces, and incorporated in soil solid phases.4 The r 2011 American Chemical Society
water-soluble forms can be leached,5,6 but leaching suffers from the disadvantage that it requires significant amounts of water.5,7,8 Furthermore, leaching is often incomplete; it may initially release only the most soluble B from the soil, and the more strongly sorbed B may eventually be released.9,10 A second option is to replace the original topsoil with a suitable, uncontaminated soil from another source and then plant it with native plant species.1 This may be successful over the short-term, but it may not be sustainable over the long-term because B from lower levels of the soil profile may eventually be solubilized and carried to the soil surface during evapotranspiration (Ajwa and Ba~ nuelos, unpublished). Therefore, covering the soil may only delay the effects of B toxicity.11 A further disadvantage is cost, which would be considerable for a site with a large acreage. A third option is to revegetate B-contaminated soils by growing an initial vegetative cover of a B-tolerant plant, followed by successive plantings of appropriate native plant species. The idea is that, as this initial vegetation cover grows and dies, significant amounts of organic matter will be added to the soil, thereby improving soil tilth (especially if there is significant root penetration), as well as enhancing aeration, water, and nutrient retention.12,13 Furthermore, Received: March 16, 2011 Accepted: September 1, 2011 Revised: June 8, 2011 Published: September 01, 2011 8922
dx.doi.org/10.1021/es200879a | Environ. Sci. Technol. 2011, 45, 8922–8927
Environmental Science & Technology
ARTICLE
Table 1. Major Chemical Characteristics of Three Different Rio Tinto Soilsa soil chemical characteristics
L
M
S
B (mg/L saturation extract)
18.4
88.4
1506
pH (saturation extract) soil organic matter based on carbon (%)
8.2 0.85
9.23 0.1
9.48 0.08
sodium (mg/L saturation extract)
292
398
1862
SAR (sodium adsorption ratio) (mmol L 1)1/2
7.9
34.3
85.2
Est. gypsum requirement (kg/ha)
1070
5470
10600
ECe of saturation extract (dS m 1)
2.25
1.69
7.8
a
Nomenclature is as follows: L: local unmined soil; M: moderately B-contaminated soil; S: severely B-contaminated soil.
the presence of plant roots in the soil ecosystem will support the growth of microbes including arbuscular mycorrhizal fungi (AMF), which form symbiotic associations with many native plants14 and are helpful in facilitating plant establishment in a harsh desert environment.15 The choice of the initial plant species is critical. The plant species selected should not only be able to tolerate high B concentrations but should also be able to survive the harsh, arid conditions of the Mojave Desert. Furthermore, an introduced plant species should not be considered as exotic or invasive. We recently identified a grass species from a semiarid region of Turkey, Puccinellia distans (L.) Parl., that is extremely B-tolerant,16 but because it is a foreign ecotype, we chose instead to use an ecotype of P. distans that is commonly found in California and other regions of the USA. The US ecotype of P. distans is a perennial halophyte that grows to a height of 20 30 cm and is widespread in North America in both saline and arid environments.17 19 The goal of the present work was to evaluate the US ecotype of P. distans in terms of its potential suitability as an initial cover crop for the phytorestoration of B-contaminated soils. The specific objectives were to determine hydroponically its tolerance to high levels of B and its ability to accumulate B as well as its ability to germinate and grow in moderately or severely B-contaminated soils taken from the mining site.
’ MATERIALS AND METHODS Soil Analysis. Soils were sampled at three different sites in and around the Rio Tinto Borax mine, Boron, CA. The soils are designated L (local soil), M (moderately B-contaminated soil), and S (severely B-contaminated soil). The L soil was taken from an unmined, vegetated area close to the mine itself. M and S soils were taken from two different locations in an area known as the East Dump. The three soils were analyzed by Wallace Laboratories (El Segundo, CA). An aqueous extract was used to measure the concentrations of B, sodium and soluble salts, pH, and salinity. The elements were measured using inductively coupled plasma-atomic emission spectrometry (ICP-AES) in accordance with the guidelines of EPA 6010. The results of the soil analyses (see Table 1) were used to determine the soil treatments to be applied. Hydroponic Study. The US ecotype of Puccinellia distans (Comstock Seed, NV) was germinated and cultivated in sand for two weeks. The plants were then individually transplanted to half-Hoagland’s solution contained in Mason jars, one plant per jar. The plants were acclimated in the half-Hoagland’s solution for several weeks prior to the addition of high levels of B supplied
as boric acid. The plants were exposed to 11 boron concentrations (in half-Hoagland’s solution): 0.5 (control), 50, 100, 150, 200, 250, 300, 350, 400, 450, and 500 mg B/L (the control treatment contained sufficient B to prevent B deficiency). Each treatment was replicated three times, with one plant per jar per replicate. The plants were aerated and maintained in a temperature-controlled greenhouse (25 ( 2 °C, 60% RH and 16 h photoperiod) for three weeks. At harvest, the shoots were removed and weighed to obtain their fresh weight (FW) and then dried at 70 °C for 48 h and reweighed to obtain their dry weight (DW). The B concentration of the dried shoot material was analyzed by the UC Davis Analytical Laboratory (DANR) (Davis, CA) using a nitric acid digestion followed by quantification using inductively coupled plasma-optical emission spectrometry (ICP-OES). Germination Trial. For the germination study, we tested two other soils (in addition to L, M, and S); these were an S soil amended with sulfuric acid and gypsum, and horticultural potting compost (as a control). Where necessary, each soil type was sifted to remove large stones. Next, the soil was placed into twelve separate 4 cm 6 cm plastic pots to create twelve replicates. Approximately 100 P. distans seeds were mixed into the top layer of soil in each replicate. The soil was watered with half-Hoagland’s nutrient solution and the extent of germination monitored over the course of six weeks. Preparation of the Moderately B-Contaminated Soil. Moderately-B contaminated soil (88.4 mg B/L saturation extract) was collected and prepared by removing large rocks by hand and then passing the soil through a 2 mm sieve. The soil was amended with gypsum (3 levels), organic matter (3 levels of peat moss), and sulfuric acid (to create two pH levels) supplied in a factorial combination. The factorial consisted of 18 treatments (3 gypsum treatments 3 organic matter treatments 2 pH treatments = 18), with 6 replicates per treatment (for a total of 108 plants). Gypsum was supplied at the rate of 0, 0.5, and 1.0 g gypsum/100 g of contaminated soil (0.5 and 1.0 g gypsum/100 g were well in excess of amounts estimated from the gypsum requirement, Table 1). Peat moss was added to yield 0, 25%, and 50% by volume. The soil was adjusted to pH 7.5 or 9.0 using sulfuric acid. P. distans seeds were germinated in sand for several weeks until they reached ∼10 cm in height at which point they were transferred to the treated soil. Plants were watered daily with pH-adjusted (7.5 and 9.0) distilled water. Plants were harvested 23 days after their transfer, and the shoot DW and shoot B-concentration were determined as described above. The soil was weighed, dried at 105 °C for 24 h, and then reweighed to determine the oven-dry soil water content. Preparation of the Severely B-Contaminated Soil. Severely B-contaminated soil (1506 mg B/L saturation extract) was collected, the rocks were removed, and the soil was sieved through a 2 mm sieve. Dry gypsum powder was added to the soil as a precautionary treatment, 1.2 g gypsum/100 g soil, a value in excess of the estimated gypsum requirement given in Table 1. The soil was amended with sand (to give 4 sand treatments) and organic matter (3 peat moss treatments) supplied in factorial combination with 9 replicates per treatment (for a total of 108 plants). The soil was mixed with sand in the following proportions: 0, 30%, 60%, and 90% (by weight). Organic matter (peat moss) was then added to the four sand-soil combinations in the following proportions: 0%, 25%, and 50% (by volume). In order to ensure that the B was not leached from the soil, a two-pot design was developed to prevent water flow-through, while simultaneously protecting the plants from having their 8923
dx.doi.org/10.1021/es200879a |Environ. Sci. Technol. 2011, 45, 8922–8927
Environmental Science & Technology
ARTICLE
Figure 1. Dry weight and B concentrations of P. distans shoots and roots grown in different B supply concentrations for 21 days: (A) Average DW (g) of P. distans (US ecotype) roots (light bars; n = 3) and shoots (dark bars; n = 3). (B) B concentration (mg B/g) of P. distans (US ecotype) roots (light bars) and shoots (dark bars) (3 replicates were pooled for B analysis). (C) B concentrations (mg B/g) of P. distans roots; US ecotype (solid squares) and Turkish ecotype (open circles). (D) B concentrations (mg B/g) of P. distans shoots; US ecotype (solid squares) and Turkish ecotype (open circles). Data from P. distans (Turkish ecotype) are included from a previous publication16 for comparison.
roots submerged in water. The first step of the two-pot design was to individually place the pots into clear plastic bags that would collect any water leaching from pots. Then, the pot and plastic bag were placed into a slightly larger pot and rotated 45 degrees. This system provides sufficient space to collect any water (and B) that might leach from the soil, while preventing the plants from becoming overwatered. The water (including any leached B) that was collected in the plastic bags was then returned to the plants during the watering schedule. To initiate the study, the P. distans seedlings were cultivated in potting soil mix in a temperature-controlled greenhouse and transplanted to the treated soil. Throughout the study, the plants were watered twice weekly with 50 mL of distilled water. The plants were harvested at 21 days, and the shoot DW and shoot B-concentration were determined as described above. Statistics. All data were analyzed using the program PASW Statistics 18. Multifactor ANOVA tests were conducted to determine the influence of the various factors, and all tests were conducted at a 95% confidence interval (α = 0.05).
’ RESULTS AND DISCUSSION A. Chemical and Physical Analysis of Soil Types from Rio Tinto Borax. Soil B Concentrations. The soluble B concentra-
tions (expressed as mg B/L saturation extract) of the three soils varied widely: soluble B ranged from 18.4 mg B/L (L, unmined soil) to 1506 mg B/L (S, severely B-contaminated soil) (Table 1). Bearing in mind that 1 5 mg soluble B/L is toxic for many plant species,20 both the moderately contaminated M soil (88.4 mg B/L saturation extract) and the severely contaminated S soil (1506 mg B/L) should be considered as seriously B contaminated (Table 1).
Alkalinity, Sodicity, and Salinity. All the soils exhibited high levels of alkalinity, sodicity, and salinity, characteristics that constitute potentially serious problems with respect to plant establishment in a phytorestoration program. The soils from the mined areas, M and S, were highly alkaline with pH values around 9.2 to 9.5, while the L soil was less alkaline with a pH value of about 8.2 (Table 1). Soils with sodium adsorption ratios (SAR) in excess of 13 (mmol L 1)1/2 are considered to be sodic. By this criterion, M and S soils were strongly sodic with SAR values of 34.3 (mmol L 1)1/2 and 85.2 (mmol L 1)1/2, respectively, while Soil L (SAR of 7.9 (mmol L 1)1/2) was nonsodic (Table 1). Salinity levels ranged from moderate (ECe = 1.7 dS m 1 for M) to very high (ECe = 7.8 dS m 1 for S) (Table 1). Soil Organic Matter Content and Infiltration Rate. The extremely low organic matter content of 0.1% or less (as with the B-contaminated soils, M and S, Table 1) is of serious concern for many reasons (see Introduction) and almost certainly contributed to low infiltration and poor drainage. The “Est. gypsum requirement” (Table 1), which provides an estimate of the amount of calcium needed to provide good infiltration, indicates that M and S soils required much greater amounts of calcium than the L soil. B. Boron Tolerance of the US Ecotype of P. distans. The US ecotype of P. distans was moderately B tolerant. Root and shoot biomass was not significantly affected by B supply concentrations up to 100 mg B/L (Figure 1A). Above that value, plant biomass of the US ecotype decreased progressively with each increase in B supply concentration up to 500 mg B/L. Boron concentrations of root and shoot tissue increased over the entire range of B supply: root B concentrations attained levels of 3 3.5 mg B/g DW, while shoot B concentrations reached levels of 12 15 mg B/g DW (Figure 1B). The US ecotype was substantially less B tolerant 8924
dx.doi.org/10.1021/es200879a |Environ. Sci. Technol. 2011, 45, 8922–8927
Environmental Science & Technology than the Turkish ecotype (described in a previous study, 16), which was able to survive at B supply concentrations in excess of 1250 mg B/L (compared to only 250 mg B/L for the US ecotype). The Turkish ecotype of P. distans had the ability to restrict the accumulation of B within its tissues, possibly by a B efflux mechanism in its root cells.16 Throughout the range of B concentrations, the Turkish ecotype of P. distans showed relatively low B concentrations in the root tissue, with a maximum B concentration of approximately 1 mg B/g DW at the highest B concentration supplied (Figure 1C). This is in sharp contrast to the US ecotype of P. distans, in which the B concentration in the roots increased linearly with an increase in the B supply concentration (Figure 1C). The apparent inability of the US ecotype of P. distans to control B accumulation in the roots resulted in a rapid increase in B shoot concentration compared to the Turkish ecotype (Figure 1D). C. Growth of the US Ecotype of P. distans on Moderately (M) and Severely (S) B-Contaminated Soils. In order to use P. distans (ecotype US) as an initial vegetative cover to facilitate reclamation, its seeds must be able to germinate and grow in the contaminated soil. Following growth, the plant material could then be either harvested to remove B from the upper layers of the soil or plowed (disked) into the soil to increase the organic matter content. We first tested the ability of P. distans to germinate in various soil types including L, M, S, and S-amended as well as horticultural potting soil as a control for comparison. Second, we tested the ability of P. distans to grow in the moderately B-contaminated soil (M) amended with organic matter and gypsum at two soil pH levels, 7.5 and 9.0. Third, we tested the ability of P. distans to grow in severely B-contaminated soil (S) amended with various combinations of organic matter and sand. Germination and Seedling Growth. The P. distans seeds were able to germinate successfully in all B-contaminated soils tested, only one of which was amended (the amended-S soil was treated with sulfuric acid and gypsum to counteract high pH and salinity, respectively). Once germinated, the seedlings were able to survive in the control soil, L, and M, but they were not able to survive in the severely B-contaminated soils (S and amended-S). The growth of the seedlings was measured using two different parameters, the average seedling shoot FW and average maximum height of the seedling shoots. As expected, the seedlings grew best in the potting soil with an average FW/pot of 0.27 g ( 0.07 g. The seedlings growing in the L soil attained more than 50% of the FW of seedlings growing in potting compost with a FW of 0.15 g ( 0.05 g, and the seedlings growing in the M soil almost 20% of the growth of the control seedlings with a FW of 0.06 g ( 0.02 g. The average height of the seedlings in the control potting soil was 5.2 cm ( 0.9 cm, while the seedlings in the L and M soil attained approximately 60% of this height with 3.2 cm ( 0.6 and 2.9 cm ( 0.7 cm, respectively. Growth and B Accumulation on Moderately B-Contaminated Soil. The moderately B-contaminated soil was treated with three different amendments: organic matter (peat moss), gypsum, and pH adjustment using sulfuric acid. The addition of 25% organic matter (by volume) to the soil substantially increased shoot growth: shoot DW increased 2.2-fold at pH 7.5 and by 3-fold at pH 9 (Figure 2A). An increase in the organic matter to 50% significantly increased the shoot DW at pH 7.5 but not at pH 9 (Figure 2A). The addition of gypsum had no significant effect on shoot DW, and the effect of pH, although significant, was relatively small. Thus, the greatest improvement in the growth of P. distans was obtained by increasing the organic matter content of the soil regardless of pH or gypsum addition.
ARTICLE
Figure 2. Effect of organic matter and pH treatments in moderately B-contaminated soil on (A) P. distans shoot DW (g), (B) soil B (mg/L), and (C) P. distans shoot B concentrations (mg B/g DW).
How did the addition of organic matter result in such a substantial improvement in growth in the moderately B-contaminated soil? The results show that the progressive increase in shoot DW with an increase in organic matter content (Figure 2A) was correlated to a large extent with the progressive dilution of soil B concentration by the peat moss (Figure 2B). Increasing the organic matter content to 25% lowered soil B concentration by about 40%, and by 70% at 50% organic matter content, regardless of pH or gypsum level. The dilution of soil B by organic matter was accompanied by a substantial reduction in the B concentration of shoot tissue: shoot B concentration was decreased by 70 to 80% with an increase in organic matter to 25% and by 90% with an increase to 50% (Figure 2C). These results suggest that shoot growth was considerably enhanced by the dilution of soil B (and ultimately shoot B) with an increase in organic matter content. It is unlikely that these results are due to the decrease in the concentration of sodium due to dilution because the addition of gypsum did not have a significant effect. 8925
dx.doi.org/10.1021/es200879a |Environ. Sci. Technol. 2011, 45, 8922–8927
Environmental Science & Technology
ARTICLE
Figure 3. Effect of the addition of 0, 30, 60, and 90% sand (by weight) and 0, 25, and 50% organic matter (by volume) in severely B-contaminated soil on (A) soil B (mg/L; average percent error 9%), (B) shoot B (mg B/g DW; average percent error 15%), and (C) shoot DW (g; average percent error 16%).
Growth and B Accumulation on Severely B-Contaminated Soil. The results obtained for moderately B-contaminated soil suggest that the prime cause of growth limitation was due to toxicity associated with high soil B concentrations. This hypothesis was further tested by using severely B-contaminated (S) soil and progressively diluting the soil B concentration with sand and organic matter in a factorial combination. The data showed that the addition of sand or organic matter significantly reduced the B concentration of the severely B-contaminated soil (Figure 3A). The effects of sand addition and organic matter addition were additive: the greater the amount of sand or organic matter the greater the dilution of soil B (Figure 3A). Unlike the situation with the moderately B-contaminated M soil (Figure 2), the dilution of soil B with an increase in sand or organic matter did not result in a progressive reduction in shoot tissue B (Figure 3B). The addition of organic matter did not alter shoot B concentrations at 0 and 30% sand treatments; at 60% and 90% sand, however, organic matter addition significantly reduced shoot B (Figure 3B). The reductions in shoot B were accompanied by corresponding increases in shoot DW (Figure 3C). Similarly, the dilution of soil B with the increase in sand from 60 to 90% resulted in substantial reductions in shoot B (Figure 3B) and corresponding increases in shoot DW (Figure 3C). The Shoot B-Growth Relationship. When shoot DW (regardless of treatment) is plotted against shoot B concentration, there is a high
Figure 4. Effect of shoot B on the shoot DW of P. distans (US ecotype) grown in (A) moderately B-contaminated soil and (B) severely B-contaminated soil. (C) The data for the two soils have been normalized to 100 (100 is the maximum shoot DW for each experiment) and combined.
degree of correlation between the two parameters in the M soil (r2 = 0.87, Figure 4A) and in the S soil (r2 = 0.91, Figure 4B). When the data for the two soils are normalized to 100 (100 is the maximum shoot DW for each experiment) and combined, the correlation between shoot DW and shoot B concentration yields an r2 value of 0.86 (Figure 4C). Since amending the soil with gypsum or sulfuric acid (to lower the pH) had relatively minor effects on plant growth in the M soil, and since plant growth appeared to be strongly dependent on tissue B concentration in plant tissues in both the M and S soils (Figure 4), these results suggest that the overriding influence on plant growth may have been due simply to B toxicity (with other soil factors having relatively minor effects). D. Evaluation of the US ecotype of P. distans as an Initial Vegetative Cover. The data obtained in the present study provide 8926
dx.doi.org/10.1021/es200879a |Environ. Sci. Technol. 2011, 45, 8922–8927
Environmental Science & Technology evidence in support of the concept of using the US ecotype of B-tolerant grass, P. distans, as an initial vegetative cover for the phytorestoration of B-contaminated soil. First, the data from the hydroponic studies show that P. distans is able to tolerate and grow in soluble B concentrations in excess of 100 mg B/L, a concentration significantly higher than those of the moderately B-contaminated soils typical of the sites to be restored. Second, P. distans seeds were able to germinate and produce seedlings that can survive in unamended moderately B-contaminated soil. Third, P. distans was able to grow in M or S soils taken from the mining site provided certain conditions were met: for the M soil, organic matter addition was required, while for S soils dilution of soil B with sand and/or organic matter was necessary. It is clear that the B tolerance of P. distans will allow the initial establishment of vegetation on high B soils, a process that would otherwise be very difficult to achieve given the B sensitivity of most plants. In addition, we hypothesize that the penetration of the contaminated soil by the living roots of P. distans should be beneficial in that it will facilitate the colonization of the rhizosphere and the adjacent soil by other organisms, including bacteria and mycorrhizal fungi. Furthermore, the continuous production and decay of fibrous roots will increase the organic matter content of the soil, thereby reducing B toxicity by diluting soil B. The possibility that P. distans could be used for the phytoextraction of B from the soil profile is considered to be less feasible because B accumulation in P. distans shoots on the M and S soils ranged from only 0.03 0.91% of shoot dry weight; given the high B concentrations of these soils, phytoextraction would require multiple planting and harvesting cycles to significantly reduce levels of soil B. Despite the apparently positive aspects of the findings enumerated above, it should be stressed that these data were obtained for plants well-supplied with water and nutrients and grown under moderate conditions in a greenhouse conditions that are far removed from the challenging and complex environment of the highly arid Mojave Desert. Our next step will be to carry out pilot field studies to test whether P. distans can be grown successfully under the environmental conditions of the actual mining site. If so, we will then move to the next stage in phytorestoration, in which we replace P. distans with native Mojave Desert species chosen initially for their tolerance to B. Over time, these B tolerant native species should alter the soil ecosystem, eventually allowing natural succession to take place. The early successional plant species might then be replaced by later colonizers such as the creosote bush. By this means, it is hoped to create a diverse and completely sustainable ecosystem.
’ AUTHOR INFORMATION
ARTICLE
Restoration Symposium, Roundy, B. A., McArthur, E. D., Haley, J. S., Mann, D. K., Eds.; USDA Forest Service Intermountain Research Station General Technical Report INT-315, 1995; pp 7 15. (3) Hereford, R.; Webb, R. H.; Longpre, C. I. Precipitation history and ecosystem response to multidecadal precipitation variability in the Mojave Desert region, 1893 2001. J. Arid Environ. 2006, 67, 13–34. (4) Keren, R.; Bingham, F. T.; Rhoades, J. D. Plant uptake of boron as affected by boron distribution between liquid and solid phases in soil. Soil Sci. Soc. Am. J. 1985, 49, 297–302. (5) Banuelos, G. S.; Cardon, G. E.; Phene, C. J.; Wu, L.; Akohoue, S.; Zambrzuski, S. Soil boron and selenium removal by three plant species. Plant Soil 1993, 148 (2), 253–263. (6) Tanji, K. A computer analysis on the leaching of B from stratified soil columns. Soil Sci. 1970, 110, 44–51. (7) Bingham, F. T.; Marsh, A. W.; Branson, R.; Mahler, R.; Ferry, G. Reclamation of salt-affected soils in Western Kern County. Hilgardia 1972, 41, 195–211. (8) Reeve, R. C.; Pillsbury, A. J.; Wilcox, L. V. Reclamation of a saline and high B soil in the Coachella Valley of California. Hilgardia 1955, 24, 69–81. (9) Peryea, F. J.; Bingham, F. T.; Rhoades, J. D. Mechanisms for B regeneration. Soil Sci. Am. J. 1985, 49, 840–843. (10) Rhoades, J. D.; Ingvalson, R. D.; Hatcher, J. T. Laboratory determinations of leachable soil B. Soil Sci. Am. J. 1970, 34, 871–875. (11) Nable, R. O.; Banuelos, G. S.; Paull, J. G. Boron toxicity. Plant Soil 1997, 193, 181–198. (12) Karlan, D. L. Soil quality as an indicator of sustainable tillage practices. Soil Tillage Res. 2004, 78 (2), 129–130. (13) Mapa, R. B. Effect of intercropping coconut lands on soil water retention. Biol. Agric. Hortic. 1995, 12 (2), 173–183. (14) Chaudhry, M. S.; Nasim, F. H.; Khan, A. G. Mycorrhizas in the perennial grasses of Cholistan desert, Pakistan. Intl. J. Bot. 2006, 2 (2), 210–218. (15) Camargo-Ricarde, S. L.; Esperon-Rodríguez, M. Effect of the spatial and seasonal soil heterogeneity over arbuscular mycorrhizal fungal spore abundance in the semi-arid valley of Tehuacan-Cuicatlan, Mexico. Rev. Biol. Trop. 2005, 53 (3 4), 339–352. (16) Stiles, A. R.; Bautista, D.; Atalay, E.; Babaoglu, M.; Terry, N. Mechanisms of boron tolerance and accumulation in plants: a physiological comparison of the extremely boron-tolerant plant species, Puccinellia distans, with the moderately boron-tolerant Gypsophila arrostil. Environ. Sci. Technol. 2010, 44 (18), 7089–7095. (17) Binh, D. Q.; Heszky, L. E.; Gyulai, G.; Kiss, E.; Csillag, A. Plant regeneration from callus of Puccinellia distans (L.) Parl. Plant Cell Tiss. Org. 1989, 18 (2), 195–200. (18) Abrams, L. Illustrated flora of the Pacific states: Washington, Oregon, and California, Stanford University Press: Stanford, 1923; Vol 1. (19) Tarasoff, C. S.; Mallory-Smith, C. A.; Ball, D. A. Comparative plant responses of Puccinellia distans and Puccinellia nuttalliana to sodic versus normal soil types. J. Arid Environ. 2007, 70 (3), 403–417. (20) Bagheri, A.; Paull, J. G.; Rathjen, A. J. Genetics of tolerance to high concentrations of soil B in peas (Pisum sativum L.). Euphytica 1996, 87, 69–75.
Corresponding Author
*Phone: (510) 642-3510. E-mail:
[email protected].
’ ACKNOWLEDGMENT We thank the Rio Tinto Borax Corporation for their generous support of this research. ’ REFERENCES (1) Bradshaw, A. The use of natural processes in reclamation advantages and difficulties. Landscape Urban Plan. 2000, 51, 89–100. (2) Allen, E. B. Restoration ecology: limits and possibilities in arid and semiarid lands. In Proceedings of the Wildland Shrub and Arid Land 8927
dx.doi.org/10.1021/es200879a |Environ. Sci. Technol. 2011, 45, 8922–8927
ARTICLE pubs.acs.org/est
Effect of Dissolved Oxygen Changes on Activated Sludge Fungal Bulking during Lab-Scale Treatment of Acidic Industrial Wastewater Shaokui Zheng,* Jingyan Sun, and Hui Han School of Environment, MOE Key Laboratory of Water and Sediment Sciences/State Key Lab of Water Environment Simulation, Beijing Normal University, Beijing 100875, People’s Republic of China
bS Supporting Information ABSTRACT: The cloning and sequencing of fungal 18S rRNA genes followed by the identification of filamentous fungal species by fluorescent in situ hybridization (FISH) and the enumeration of filamentous fungal cells by flow cytometryFISH (FC-FISH) were used to investigate the effect of dissolved oxygen (DO) changes on activated sludge (AS) fungal bulking during a lab-scale treatment of acidic industrial wastewater. By increasing DO levels from <0.5 to >2 mg L1, bulking started to occur due to the outbreak of fungal filaments, whereas the chemical oxygen demand (COD) removals sharply increased from <40 to >70%. Clone library analyses revealed that all clonal fungal sequences were of yeast origin, and that only one and four yeast species were individually detected in AS at two DO levels. Subsequent FISH identification of filamentous yeast species within bulking sludge using self-designed oligonucleotide probes suggested that all probe-reactive cells of Trichosporon asahii had a filamentous morphology and were the dominating filamentous microorganism in the AS. The FC-FISH analyses of bacteria and two main yeast species showed that the DO shift resulted in a sharp increase of T. asahii, by a factor of 4860, which caused filamentous yeast bulking. Subsequently, the restoration of DO levels to <0.5 mg L1 effectively restored the sludge settlement and yeast community, as well as unacceptable COD removals.
1. INTRODUCTION The activated sludge (AS) process is one of the most commonly used technologies in municipal and industrial wastewater treatment because it is cost-effective and straightforward.1,2 Good AS separation (settling) and compaction (thickening) is always a necessary condition to guarantee good effluent quality from the secondary clarifier.1,3 However, the excessive growth of filamentous microorganisms, an event termed AS bulking, frequently leads to poor sludge settling ability, as indicated by a sludge volume index (SVI) greater than 150.3,4 As one of the main approaches to eliminate bulking sludge, specific methods develop targeted control strategies to selectively suppress filamentous microbial proliferation by finding relationships between the dominant filamentous microbe and the operating conditions.1,5 For example, Candidatus “Microthrix parvicella” might be controlled by removing its preferred lipid substrates from the wastewater or by adding chemicals (polyaluminium chloride) that probably inhibit the activities of their surfaceassociated lipases.6 Operating at higher dissolved oxygen (DO) levels may also control C. “Microthrix parvicella” because they are known to be microaerophilic.6,7 The major challenges for specific methods are to properly identify and characterize the filamentous microbes in a bulking sludge, and to investigate the reasons why different filamentous microorganisms may outgrow others in wastewater treatment plants (WWTPs). Previous research has described approximately r 2011 American Chemical Society
30 morphotypes of filamentous bacterial species that appear in domestic WWTPs; however, only some have been found to be responsible for bulking problems.8,9 Furthermore, although conventional culture-dependent methods are widely used to obtain specific microbial sources,10 many filamentous microbes still remain poorly characterized due to problems associated with the cultivation and maintenance of pure cultures.1,11 During the 1990s, culture-independent molecular fingerprinting technologies based on DNA and RNA analyses were introduced to biological wastewater treatment. These methods, e.g., fluorescent in situ hybridization and flow cytometry (FC-FISH),12 allow accurate identification and quantification of the microbial population, as well as the analysis of microbial community structure using clone libraries.2,13 With the development of these molecular technologies, the identification of filaments in AS has increased dramatically in recent years.6 In addition to filamentous bacterial species, some fungal species (including yeasts and molds) are commonly encountered during aerobic biological treatment and cause bulking under acidic pH conditions.1416 For example, the mold Geotrichum candidum was isolated from bulking AS on agar plates and Received: May 27, 2011 Accepted: September 8, 2011 Revised: September 2, 2011 Published: September 08, 2011 8928
dx.doi.org/10.1021/es2018169 | Environ. Sci. Technol. 2011, 45, 8928–8934
Environmental Science & Technology identified by morphological and biochemical characteristics; it was suspected be responsible for bulking during the continuous treatment of salad oil manufacturing wastewater.14 Hu and Strom observed AS fungal bulking at pH 4 and 5 in laboratory fill-anddraw AS systems, during which fungal filament abundance in AS was monitored by semiquantitative microscopic observations.16 However, there has been relatively little investigation of the behavior of fungal species responsible for bulking using cultureindependent technologies. To control filamentous fungal bulking in a WWTP treating acidic industrial wastewaters, the objective of this lab-scale AS research was to determine whether changing the DO levels from microaerobic (<0.5 mg DO L1) to aerobic (>2 mg DO L1) would lead to fungal bulking, and whether the subsequent restoration to a microaerobic level would help to minimize the filamentous fungal growth and restore good AS settlement. Four clone libraries were constructed to document the evolution of the sludge fungal community in response to varied DO levels. Next, the identification of filamentous fungal species within the bulking sludge was conducted by FISH analysis using self-designed 18S rRNA-targeted oligonucleotide probes. Finally, the content of the filamentous fungal species in AS during different phases was elucidated using FC-FISH analyses, and these data were used as a theoretical basis to develop an operational strategy for bulking preventive control. This is the first report on the use of DO regulation to control AS fungal bulking.
2. MATERIALS AND METHODS 2.1. Wastewater and Seed Sludge. Sulfate-rich high-organicstrength industrial wastewater (pH 3.94.7) containing approximately 2000 mg sulfate L1, 27 00030 000 mg chemical oxygen demand (COD) L1, and 450750 mg total nitrogen L1, 80150 mg total phosphorus L1, with a biodegradable index of 0.25 was collected every 2 weeks from a pharmaceutical fermentation factory in Hebei Province, P. R. China, and stored at 4 °C. AS collected from the WWTP was centrifuged at 5000 rpm and room temperature, and used as seed sludge in the subsequent experiments. 2.2. Lab-Scale Treatment System and Its Operation. A 16 L lucite aeration column equipped with a 10 L cylindrical sedimentation column constituted the lab-scale treatment system without sludge return (see Supporting Information). A metering pump transferred wastewater from the feed tank to the bottom of the aeration column to give HRT levels of 60 h. The effluent flowed out from the top of the aeration column via gravity to the top of the sedimentation column and was subsequently discharged. An air pump was used in conjunction with a gas-flow meter to introduce air into diffusers in the bottom of the aeration column to provide the designated DO level. The aeration column containing 10 L wastewater was inoculated with 10 g (wet weight) seed sludge, and the mixture was then cultivated for five days. Next, raw wastewater was pumped into the aeration column continuously, during which time the water sampling and on-site measurements were conducted daily. The DO level of the aeration column was controlled manually at <0.5 mg L1 (Days 017), >2 mg L1 (Days 1843), and <0.5 mg L1 (Days 4360). Additionally, the sludge was collected from the aeration column on Days 17, 30, 43, and 60, and evaluated by scanning electron microscopy (SEM) according to the method described by Vyrides and Stuckey,17 cloning/ sequencing, FISH and FC-FISH analyses.
ARTICLE
2.3. Construction of Fungal 18S rDNA Clone Libraries from Four Sludge Samples. The total fungal community
DNA was extracted from 1.0 g (wet weight) of sludge using a FastDNA Spin Kit for Soil (Bio 101, Q-Biogene, USA) according to the manufacturer’s protocols, and was then precipitated and washed prior to PCR amplification. The 18S rRNA genes were amplified using universal fungal primers: forward primer, nu-SSU0817 (50 -TTAGCATGGAATAATRRAATAGGA-30 ), and reverse primer, nu-SSU-1536 (50 -ATTGCAATGCYCTATCCCCA-30 ), which amplify an rDNA fragment of approximately 700800 bp from all four major phyla of fungi (Ascomycota, Basidiomycota, Chytridomycota, and Zygomycota).13 All reactions were conducted in a 50-μL PCR mixture that contained 1 μL of extracted DNA, 5 μL of 10 PCR buffer, 2.5 mM of each dNTP, 25 pmol of each primer, and 1 U of Taq polymerase. PCR was conducted using a 9600 thermal cycler (Perkin-Elmer, U.S.) under the following conditions: denaturation at 95 °C for 5 min, followed by 30 cycles of 94 °C for 30 s, 54 °C for 30 s, and 72 °C for 50 s, with a final extension at 72 °C for 10 min. Amplification products were electrophoresed in 1% (w/v) agarose gels, stained with ethidium bromide, and visualized under UV light. PCR amplification products were purified using a UNIQ-10 DNA purification kit (Shanghai Sangon, P. R. China) according to the manufacturer’s instructions. The purified PCR products were cloned into the pMD18-T vector (Promega, USA) and transformed into Escherichia coli DH5a High Efficiency Competent Cells (TaKaRa, P. R. China), according to the manufacturers’ instructions. Recombinant cells were plated onto LuriaBertani medium containing ampicillin (150 mg mL1) to identify white-colored recombinant colonies. To confirm the presence of DNA inserts, approximately 100 white colonies were picked randomly for each sample, and subjected to colony PCR using vector primers M13F (50 -CGCCAGGGTTTTCCCAGTCACGAC-30 ) and M13R (50 -AGCGGATAACAATTTCACACAGGA-30 ) (Promega), as described elsewhere.18 Depending on the PCR result, the amplification products from each positive clone, totaling 60, 64, 73, and 70 recombinant clones for the four clone libraries, were further digested individually with two DNA restriction endonuclease treatments (RsaI and HaeIII). The 20-μL reactions contained 2 μL of 10 buffer, 1 μL RasI, 1 μL HaeIII, 8 μL of DNA, and 8 μL of sterile MillQ water, and were incubated at 37 °C for 4 h. The resulting restriction fragment length polymorphism (RFLP) products were separated by electrophoresis on an agarose gel. Finally, the sequences of representative 18S rDNA clones were determined by a commercial laboratory (Beijing Sunbiotech, China, http:// www.sunbiotech.com.cn/). All of the 18S rDNA sequences obtained were compared with 18S rRNA gene sequences available in the GenBank database using the NCBI Blast program (http://www. ncbi.nlm.nih.gov/BLAST/) to find closely related fungal sequences, and checked for chimeric artifacts using the CHIMERA CHECK program (version 2.7) from the Ribosomal Database Project (http://rdp.cme.msu.edu). Sequences suspected of being chimeric were not included in further analyses. Sequences of the selected clones and closely related fungal species obtained from the NCBI databases were aligned using the CLUSTAL W package and corrected by manual inspection. All 18S rRNA gene sequences were deposited in the GenBank database under accession numbers JN638383JN638404. The clone library coverage (C) for each clone library was calculated according the following equation: C = 1 (n/N), where n is the number of unique clones and N is the total number of clones examined. 8929
dx.doi.org/10.1021/es2018169 |Environ. Sci. Technol. 2011, 45, 8928–8934
Environmental Science & Technology 2.4. Probe Design. The 18S rDNA sequences of four yeast species detected in the AS (Issatchenkia occidentalis strain NRRL Y-7552, Trichosporon asahii isolate ZZB-1, Trichosporon lactis partial strain HB865, and Pichia sp. SA8S03) were obtained from GenBank and aligned using the MEGA version 4.1 package to identify the conserved sequences within this group. The probes identified through this process were: EF550374.1 (50 -AGC CCA CGA GCA GAT GCT CGC AGC CCT-30 ) for I. occidentalis, GU323377.1 (50 -ATC CAG TCA AAA AAT AAC GTG GAC-30 ) for T. asahii, AJ319755.1 (50 -AGT GGG CAA AAT AGC GCC ACC-30 ) for T. lactis partial, and FJ153116.1 (50 -ATC ACG ACG ATC GAA ACC GAC CAT ATC AGT-30 ) for Pichia sp. All 18S rRNA-targeted oligonucleotide probes were labeled simultaneously with FITC tags at the 30 - and 50 -ends to achieve high fluorescence intensity during hybridization. 2.5. FISH. All hybridizations were performed with 8 μL of sludge sample dried on a precleaned Teflon coated slide and dehydrated for 2 min each in 50%, 80% and 90% ethanol series. The slides were hybridized with each oligonucleotide probe (100 ng μL1 in Tris-EDTA buffer [100 mM Tris/HCl, pH 8.0, 50 mM EDTA]) at 46 °C for 2 h in hybridization buffer (0.9 M NaCl, 20 mM Tris-HCl, 0.01% [w/v] sodium dodecyl sulfate (SDS), 20% [v/v] formamide, pH 7.2). Microscopic examination of the FISH samples was conducted using an epifluorescence microscope at 600 magnification (Olympus BX61, Japan). Photographs were taken of diluted samples using an Olympus Colorview digital camera under epifluorescence. 2.6. FC-FISH Analyses. The sludge samples were centrifuged, washed, and then sonicated for 90 s at 100 W prior to fixation with fresh 4% (w/v) paraformaldehyde at 4 °C overnight. These samples were then individually subjected to hybridization using a universal eubacteria 16S rRNA-targeted oligonucleotide probe with FITC tags, Eub338 (50 -GCT GCC TCC CGT AGG AGT-30 ), or self-designed 18S rRNA-targeted oligonucleotide probes specific to two yeast species, I. occidentalis and T. asahii. For hybridization, 5.0 μL of each probe and 10 μL of fixed cells were added to 185 μL of prewarmed hybridization buffer, followed by incubation at 46 °C for 3 h in the dark. Additionally, negative controls consisting of hybridization buffer with no probes were prepared. All samples were thoroughly vortexed and sonicated after yellow-green fluorescent beads (TruCount tube, BD, USA) were added according to the manufacturer’s instructions. Finally, the labeled cells were enumerated using a FACSCalibur flow cytometer (BD), and further analyzed using CellQuest software (BD). The concentration of group-specific cells in the sample was calculated relative to the concentration of reference microspheres and was proportional to the number of events identified by the flow cytometer, as follows:
cellconcentration ¼ numberofcellevents=numberofbeadevents beadconcentration
Subsequently, the contents of specific microbial cells were calculated as the ratio of group-specific labeled cells to the total microbial cell number of bacteria and two yeast species. 2.7. Wastewater Analysis. DO and pH measurements were conducted using a Hi9143 DO-meter (Hanna, Italy) and a Hanna8424 pH-meter (Hanna), respectively. The viscosity of the sludge in aeration column was determined by a NDJ-1 rotary viscometer (Shanghai Ande, China). SVI was calculated as the volume occupied by a settled biomass sample after 30 min of settling (in a 100 mL graduated cylinder), divided by the biomass
ARTICLE
Figure 1. Performance of the lab-scale activated sludge system at different dissolved oxygen (DO) levels. The study period was divided into four experimental phases.
dry weight. Water samples were centrifuged at 5000 rpm and room temperature for 10 min, after which time the COD of the resulting supernatant was analyzed using a COD rapid analyzer (Hach, U.S.).
3. RESULTS 3.1. Lab-Scale AS Treatment Performance at Various DO Levels. Figure 1 shows the time course of DO, aeration rate,
effluent COD, COD removal (CODRe), SVI, and viscosity levels at three different DO levels during 60 days of continuous operation of the lab-scale system. As shown in Figure 1, as the DO levels were shifted stepwise from <0.5 mg L1 to >2 mg L1 and then back to to <0.5 mg L1, there was a significant increase and subsequent decrease in the aeration rate (from 125150 to 600750 L h1 and then to 125150 L h1), CODRe (from 3239% to 7179% and then to 3439%), SVI (from 30 to >252 mL g1 and then to 38 mL g1), and viscosity (from 3.8 to 40 mPa S1 and then to 3.5 mPa S1). Under aerobic conditions, the sludge bulking started until Day 35, whereas the COD removal performance remained excellent during this period. To better follow the evolution of the sludge fungal community structure, the sludge was sampled at Day 30, which divided the aerobic conditions into two experimental phases. Therefore, the continuous operation was finally divided into four experimental phases: phases A (Days 017), B (Days 1830), C (Days 3043), and D (Days 4360). At the end of phase C, the bulking resulted in no solids separation in the secondary clarifier. 8930
dx.doi.org/10.1021/es2018169 |Environ. Sci. Technol. 2011, 45, 8928–8934
Environmental Science & Technology
ARTICLE
Figure 2. Scanning electron microscopy (SEM) photos ( 5000) of sludge samples illustrating normal (A) and bulking (B) sludge samples collected individually at Day 17 during phase A (A) and at Day 43 during phase C (B).
Table 1. Fungal 18S rRNA Clone Libraries Constructed with Four Sludge Samples clones of each OUT library
clones
RFLP patterns
closest relative (%similarity; Accession)
(%abundance)
coverage (%)
A
60
10
Issatchenkia occidentalis strain NRRL Y-7552 (9899%; EF550374.1)
60 (100)
100
B
64
16
Issatchenkia occidentalis strain NRRL Y-7552 (9499%; EF550374.1)
23 (34)
98
Trichosporon asahii isolate ZZB-1 (99100%; GU323377.1)
40 (60)
Trichosporon lactis partial strain HB865 (95%; AJ319755.1) C
73
22
39 (53)
Trichosporon asahii isolate ZZB-1 (99100%; GU323377.1)
27 (37)
Trichosporon lactis partial strain HB865 (95%; AJ319755.1) D
70
8
1 (1)
Issatchenkia occidentalis strain NRRL Y-7552 (9499%; EF550374.1)
Pichia sp. SA8S03 (99%; FJ153116.1) Issatchenkia occidentalis strain NRRL Y-7552 (9899%; EF550374.1)
The microaerobic conditions during phase D obviously controlled sludge filamentous fungal bulking, but restored the unacceptable CODRe to that observed during phase A. 3.2. Evolution of Sludge Microbial Community Structure. 3.2.1. SEM Analysis. Figure 2 presents SEM images illustrating the differences between the sludge samples obtained at two DO levels (those obtained during phases A and D had identical microscopic images). On the basis of size and morphology, a variety of morphotypes including bacteria and fungi were observed simultaneously within the sludge. As DO increased, the dominant fungal cells in sludge evolved gradually from relatively short rod-shape cells to relatively long hyphae. 3.2.2. Fungal Community Structure As Revealed by Four Clone Libraries. A total of 267 clones were obtained and finally clustered into 22 RFLP patterns (10, 16, 22, and 8, respectively, RFLP patterns during 4 experimental phases). A representative clone of each pattern was sequenced. Of 267 clones the majority of the clones from the four clone libraries were most closely related to fungal 18S rDNA sequences described previously in the GenBank databases. Approximately 27% of sequences exhibited <98% similarity to known sequences in the GenBank database, which was the species definition level.19 This low level of similarity indicated that the clones may be from newly detected organisms. Additionally, three clones were similar to Pichia spp. (99%) but with low affiliation to any relative (the best match was only 90%). The retrieved sequences, their RFLP patterns, closest relatives (including similarity and accession
100
4 (5) 3 (4) 70 (100)
100
number), abundances in the library, and C values are listed in Table 1. The C values for four clone libraries remained as high as 98100%, suggesting that samples of 267 clones from these clone libraries provided nearly full coverage of the fungal diversity. Of the 267 clones, 192 and 67 were found to have 9499% and 99100% sequence similarity to I. occidentalis strain NRRL Y-7552 (EF550374.1) (72% of the total clones) and T. asahii isolate ZZB-1 (GU323377.1) (25% of the total clones), respectively, representing the majority of the libraries. Only 5 and 3 clones were most closely related to T. lactis partial strain HB865 (AJ319755.1) (<2% of the total clones) and Pichia sp. SA8S03 (FJ153116.1) (<1% of the total clones), with 95% and 99% sequence similarity, respectively. After grouping sequences that differed by e1% into the same operational taxonomic units (OTUs), these clones were finally divided into four OTUs, indicating that the fungal diversity was very low in the AS. Of the four OTUs, I. occidentalis occurred in all four sludge samples, T. asahii and T. lactis occurred only in clone libraries B and C, and Pichia sp. were restricted to clone library C. This suggested that the diversity of the fungal community during phases A and D was much less than that during phases B and C. Additionally, clone library analyses revealed that all clonal fungal sequences were of yeast origin, and that the yeast community shifted markedly in response to the change in DO levels. 3.2.3. Identification of Filamentous Fungal Species by FISH. FISH analyses of bulking sludge were performed using four selfdesigned fluorescently labeled oligonucleotide probes (Figure 3). 8931
dx.doi.org/10.1021/es2018169 |Environ. Sci. Technol. 2011, 45, 8928–8934
Environmental Science & Technology
ARTICLE
Figure 3. FISH performance of four self-designed FITC-labeled oligonucleotide probes toward target yeast species in filamentous fungal bulking sludge samples collected at Day 43 during phase C. A: Probe EF550374.1 for I. occidentalis; B: Probe GU323377.1 for T. asahii; C: Probe AJ319755.1 for T. lactis partial; D: Probe FJ153116.1 for Pichia sp.
All positive cells were morphologically similar and no bacterial rDNA sequences were identified. The EF550374.1-positive cells accounted for the majority of relatively short rod-shape yeast cells with obvious buds. All cells positive with the GU323377.1 probe had thin filamentous morphotypes. Additionally, probes AJ319755.1 and FJ153116.1 only reacted selectively with a minor fraction of rod-shape cells. No further investigations of T. lactis and Pichia sp. were conducted, because these species comprised only a minor fraction of the fungal community and are not filamentous fungal species. 3.2.4. Microbial Cell Concentration by FC-FISH. Bivariate dotplots of side scatter (SSC) versus the fluorescence emission of fluorescein for the four hybridized sludge samples are depicted in Figure 4. The nonhybridized region was gated as R1 using a control sample in which the cells were treated the same as other experimental samples, but without the addition of the probes. Some events detected in region R1 also represent lysed cells and background noise. The hybridized cells were concentrated in region R2, whereas microspheres were located in region R3. The presence of reference beads in region R3 allowed the enumeration of labeled cells. When the labeled cells were gated, they appeared as very small clusters in the FC dot plots, indicating that the cells exhibited limited morphological variability. FC-FISH assays (Figure 5) revealed that the cell concentrations of bacteria and I. occidentalis in sludge samples were relatively stable in narrow ranges of (3.87.1) 109 and (1.82.4) 109 cells mL1, respectively, over time during all four experimental phases. However, the cell concentration of T. asahii increased sharply from 1.0 108 (phase A) to 4.8 109 (phase B) and 6.0 109 cells mL1
(phase C), by a factor of 4860, and then decreased sharply to 3.0 108 cells mL1 (phase D), accounting for 1, 44, 53, and 3% of the total microbial cells in the sludge samples in phases AD, respectively.
4. DISCUSSION Although numerous studies have been conducted to find strategies to control the bulking, the root causes of filamentous overgrowth remain poorly understood, and even contradictory viewpoints persist. For example, polyaluminium chloride, which selectively eliminates filamentous bulking in some cases, does not work in other cases.6 Selectors are the most applied engineering tool worldwide for the prevention of bulking sludge phenomena; however, there are still regular reports citing the failure of selectors in the control of bulking sludge.1 Many studies have demonstrated the necessity of fundamental DO levels for good AS performance since low DO levels frequently leads to bulking.3,4,20 However, the results of the current study demonstrate that a higher DO level has a negative effect on sludge settlement, whereas microaerobic conditions clearly suppress filamentous fungal bulking. Many high-strength industrial wastewaters are often highly acidic, e.g., wastewater from the concentrated rubber latex industry (3.64.7),21 molasses production (4.6),22 and from breweries (∼3).23 When these wastewaters are treated directly by AS without pH adjustment, the extreme or moderate acidophiles that grow optimally at pH <3 or 35, respectively,24 including fungi16,25,26 and acidophilic bacteria,24 will be abundant 8932
dx.doi.org/10.1021/es2018169 |Environ. Sci. Technol. 2011, 45, 8928–8934
Environmental Science & Technology
ARTICLE
Figure 4. Gated FL1 (FITC) versus SSC dot plots of bacteria-, I. occidentalis-, and T. asahii-beads for bulking sludge samples collected during phase C (R1: negative control, R2: hybridized cells, R3: beads).
Figure 5. Evolution of cell concentrations of bacteria and two yeast species (I. occidentalis and T. asahii) in four sludge samples obtained during different experimental phases.
in AS. In this study, although a good coverage of diversity was achieved for all four clone libraries, the phylogenetic analysis of the four clone libraries showed that all fungal species in the AS were solely of yeast origin belonging to various phyla. In many previous studies, yeast species also dominated the microbial community in many AS for industrial wastewater treatment under acidic conditions.25,26 The cloning/sequencing analyses confirmed that the internal relationships between the DO level and sludge yeast community structure are of considerable importance to the treatment performance. Exposing sludge to higher DO levels resulted in a wider variety of yeast species in the bioreactor. Conversely, a microaerobic level had a negative effect on both yeast structural diversity and CODRe level. In this study, FISH analyses were performed using four selfdesigned oligonucleotide probes to distinguish those four yeast species and elucidate their morphology within bulking sludges. Our results suggested that the sole filamentous yeast species in bulking sludge was T. asahii. Different strains of T. asahii have extensive abilities in mashita fermentation27 or lipase production.28 Other species of T. spp. have also been detected in industrial wastewater treatment.14,29,30 Additionally, T. asahii is a clinically important yeast species that can cause life-threatening infections in immunocompromised patients.31 T. asahii has
also been reported to have two different cell morphologies of conidia and hyphae.31 Subsequently, a quantitative analysis of bacteria and two major yeast species, I. occidentalis and T. asahii was developed with the goal of clarifying the ecological factors responsible for controlling filament densities within AS and why yeast filaments can reach excessive abundances. It appears that under nonbulking conditions during phases A and D, filamentous yeast species were actually present inside the sludge, but accounting for only 13% of the total microbial cells. Thus, the filamentous yeast species were not always predominant within sludge. In fact, it is known that filamentous microbes do not regularly represent the dominant metabolic microbial group; e.g., a volume fraction of 120% or even too low to be detected by denaturing gradient gel electrophoresis.1 This result also demonstrates that some yeast sequences present at lower abundance in AS might not be extracted in cloning/sequencing analyses due to the relatively small number of clones evaluated, even if a good coverage of diversity was achieved for each clone library in this study. However, the FC-FISH analysis seemed to overcome the limitations of PCR-based community analytical approaches, which cannot accurately enumerate specific microorganisms, and provides microbial community information on the basis of bacteria and fungi. Generally, bulking sludge is regarded as a microbiological problem (i.e., the occurrence of a filamentous microorganism) or as a morphology problem (growth of a microorganism with a filamentous morphology).1 The results presented here reveal that fungal bulking was caused by both excessive propagation of T. ashaii and a change in its cell morphology in response to DO levels, i.e., both a microbiological and an morphology problem. Previously, it has been reported that T. asahii can spontaneously switch its cell morphology from conidia (blastoconidia or arthroconidia) to hyphae after repeated subculture or in vivo passages.31 Several researchers have pointed out that the morphology of fungal hyphae aids in substrate uptake due to a higher surface-to-volume ratio, which would lead to a relatively higher specific growth rate compared to fungal conidia.31 The ultimate goal of many investigations of bulking is to control populations of the bulking species in AS at a level that prevents bulking problems.6 The results in this report revealed that the DO level was a critical factor affecting the growth of T. asahii, and that higher DO levels may have triggered the outbreak 8933
dx.doi.org/10.1021/es2018169 |Environ. Sci. Technol. 2011, 45, 8928–8934
Environmental Science & Technology of filaments. At the same time, microaerobic conditions helped to suppress the growth of T. asahii and restabilize the AS. In this case, a species-specific method is suggested to control filamentous yeast bulking by including a microaerobic removal stage (selectors) in front of aerobic removal stage to encourage the floc-formers to assimilate the majority of soluble substrate under the microaerobic conditions and to avoid an outbreak of fungal filaments in the presence of plentiful substrate under the aerobic conditions.
’ ASSOCIATED CONTENT
bS
Supporting Information. Figures S1. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +86 10 58809266; e-mail:
[email protected].
’ ACKNOWLEDGMENT This study was financially supported by the Natural Science Foundation of China (21077011) and the Program for Changjiang Scholars and Innovative Research Team in University (IRT0809). ’ REFERENCES (1) Martins, A. M. P.; Pagilla, K.; Heijnen, J. J.; van Loosdrecht, M. C. M. Filamentous bulking sludge--a critical review. Water Res. 2004, 38, 793–817. (2) Chen, M.; Chen, F.; Xing, P.; Li, H.; Wu, Q. L. Microbial eukaryotic community in response to Microcystis spp. bloom, as assessed by an enclosure experiment in Lake Taihu, China. FEMS Microbiol. Ecol. 2010, 74, 19–31. (3) van den Akker, B.; Beard, H.; Kaeding, U.; Giglio, S.; Short, M. D. Exploring the relationship between viscous bulking and ammoniaoxidiser abundance in activated sludge: A comparison of conventional and IFAS systems. Water Res. 2010, 44, 2919–2929. (4) Xie, B.; Dai, X. C.; Xu, Y. T. Cause and pre-alarm control of bulking and foaming by Microthrix parvicella—A case study in triple oxidation ditch at a wastewater treatment plant. J. Hazard. Mater. 2007, 143, 184–191. (5) Tsang, Y. F.; Chua, H.; Sin, S. N.; Tam, C. Y. A novel technology for bulking control in biological wastewater treatment plant for pulp and paper making industry. Biochem. Eng. J. 2006, 32, 127–134. (6) Nielsen, P. H.; Kragelund, C.; Seviour, R. J.; Nielsen, J. L. Identity and ecophysiology of filamentous bacteria in activated sludge. FEMS Microbiol. Rev. 2009, 33, 969–998. (7) Rossetti, S.; Tomei, M. C.; Nielsen, P. H.; Tandoi, V. “Microthirx parvicella”, a filamentous bacterium causing bulking and foaming in activated sludge systems: A review of current knowledge. FEMS Microbiol. Rev. 2005, 29, 49–64. (8) Eikelboom, D. H. Filamentous organisms observed in activated sludge. Water Res. 1975, 9, 365–388. (9) Jenkins, D.; Richard, M. G.; Daigger, G. T. Manual on the Causes and Control of Activated Sludge Bulking, Foaming, And Other Solids Separation Problems; Lewis Publishers: Boca Raton, FL, 2004. (10) Xia, X.; Yang, Z.; Zhang, X. Effect of suspended-sediment concentration on nitrification in river water: importance of suspended sediment-water interface. Environ. Sci. Technol. 2009, 43, 3681–3687. (11) Tang, Y. Z.; Gin, K. Y. H.; Lim, T. H. High-temperature fluorescent in situ hybridization for detecting Escherichia coli in seawater samples, using rRNA-targeted oligonucleotide probes and flow cytometry. Appl. Environ. Microbiol. 2005, 71, 8157–8164.
ARTICLE
(12) Porter, J.; Pickup, R. Nucleic acid-based fluorescent probes in microbial ecology: Application of flow cytometry. J. Microbiol. Methods 2000, 42, 75–79. (13) Borneman, J.; Hartin, R. J. PCR primers that amplify fungal rRNA genes from environmental samples. Appl. Environ. Microbiol. 2000, 66, 4356–4360. (14) Zheng, S.; Yang, M.; Lv, W.; Liu, F. Study on sludge expansion during treatment of salad oil manufacturing wastewater by yeast. Environ. Technol. 2001, 22, 533–542. (15) Li, A. J.; Zhang, T.; Li, X. Y. Fate of aerobic bacterial granules with fungal contamination under different organic loading conditions. Chemosphere 2010, 78, 500–509. (16) Hu, P.; Strom, P. F. Fungal bulking of activated sludge at low pH. Res. J. Water Pollut. Control Fed. 1991, 63, 276–277. (17) Vyrides, I.; Stuckey, D. C. Saline sewage treatment using a submerged anaerobic membrane reactor (SAMBR): Effects of activated carbon addition and biogas-sparging time. Water Res. 2009, 43, 933–942. (18) Anderson, I. C.; Campbell, C. D.; Prosser, J. I. Potential bias of fungal 18S rDNA and internal transcribed spacer polymerase chain reaction primers for estimating fungal biodiversity in soil. Environ. Microbiol. 2003, 5, 36–47. (19) Chen, M.; Chen, F.; Yu, Y.; Ji, J.; Kong, F. Genetic diversity of eukaryotic microorganisms in Lake Taihu, a large shallow subtropical lake in China. Microb. Ecol. 2008, 56, 572–583. (20) Thompson, G.; Forster, C. Bulking in activated sludge plants treating paper mill wastewater. Water Res. 2003, 37, 2636–2644. (21) Saritpongteeraka, K.; Chaiprapat, S. Effects of pH adjustment by parawood ash and effluent recycle ratio on the performance of anaerobic baffled reactors treating high sulfate wastewater. Bioresour. Technol. 2008, 99, 8987–8994. (22) Tondee, T.; Sirianuntapiboon, S.; Ohmomo, S. Decolorization of molasses wastewater by yeast strain, Issatchenkia orientalis No. SF9246. Bioresour. Technol. 2008, 99, 5511–5519. (23) Rao, A. G.; Reddy, T. S. K.; Prakash, S. S.; Vanajakshi, J.; Joseph, J.; Sarma, P. N. pH regulation of alkaline wastewater with carbon dioxide: A case study of treatment of brewery wastewater in UASB reactor coupled with absorber. Bioresour. Technol. 2007, 98, 2131–2136. (24) Johnson, D. B.; Hallberg, K. B. Carbon, iron and sulfur metabolism in acidophilic microorganisms. Adv. Microb. Physiol. 2008, 54, 201–255. (25) Chigusa, K.; Hasegawa, T.; Yamamoto, N.; Watanabe, Y. Treatment of wastewater from oil manufacturing plant by yeasts. Wat. Sci. Technol. 1996, 34, 51–58. (26) Han, H.; Zhang, Y.; Cui, C.; Zheng, S. Effect of COD level and HRT on microbial community in a yeast-predominant activated sludge system. Bioresour. Technol. 2010, 101, 3463–3465. (27) Ongol, M. P.; Asano, K. Main microorganisms involved in the fermentation of Ugandan ghee. Int. J. Food Microbiol. 2009, 133, 286–291. (28) Kumar, S. S.; Gupta, R. An extracellular lipase from Trichosporon asahii MSR 54: Medium optimization and enantioselective deacetylation of phenyl ethyl acetate. Proc. Biochem. 2008, 43, 1054–1060. (29) Monteiro, A. S.; Bonfim, M. R. Q.; Domingues, V. S.; Correa, A.; Siqueira, E. P.; Zani, C. L.; Santos, V. L. Identification and characterization of bioemulsifier-producing yeasts isolated from effluent of a diary industry. Bioresour. Technol. 2010, 101, 5186–5193. (30) Weber, S. D.; Hofmann, A.; Pilhofer, M.; Gerhard, W.; Agerer, R.; Ludwig, W.; K.H., S.; Fried, J. The diversity of fungi in aerobic sewage granules assessed by 18S rRNA gene and ITS sequence analyses. FEMS Microbiol. Ecol. 2009, 68, 246–254. (31) Karashima, R.; Yamakami, Y.; Yamagata, E.; Tokimatsu, I.; Hiramatsu, K.; Nasu, M. Increased release of glucuronoxylomannan antigen and induced phenotypic changes in Trichosporon asahii by repeated passage in mice. J. Med. Microbiol. 2002, 51, 423–432.
8934
dx.doi.org/10.1021/es2018169 |Environ. Sci. Technol. 2011, 45, 8928–8934
ARTICLE pubs.acs.org/est
NOM Fractionation and Fouling of Low-Pressure Membranes in Microgranular Adsorptive Filtration Zhenxiao Cai* and Mark M. Benjamin Dept. of Civil and Environmental Engineering, Box 352700, University of Washington, Seattle Washington 98195-2700, United States
bS Supporting Information ABSTRACT: Membrane fouling by natural organic matter (NOM) was investigated in microgranular adsorptive filtration (μGAF) systems, in which a thin layer of adsorbent is predeposited on low-pressure membranes. The adsorbents tested included heated aluminum oxide particles (HAOPs), ion exchange (IX) resin, and powdered activated carbon (PAC). Size exclusion chromatography (SEC) separated the NOM into four apparent MW fractions with significant UV254. HAOPs and the IX resin performed almost identically with respect to removal of these fractions, and differently from PAC. However, while HAOPs and PAC reduced fouling substantially, IX resin did not, indicating that fouling could not be attributed to the NOM fractions detected by SEC. Rather, the key foulants appear to comprise a very small fraction of the NOM with almost no UV254 absorbance. Alginate, a strongly fouling surrogate for natural polysaccharides, is adsorbed effectively by HAOPs, but not by IX resin or PAC, suggesting that polysaccharides sometimes play a key role in membrane fouling by NOM.
’ INTRODUCTION The primary foulants of low-pressure membranes are thought to be particles 1 3 and natural organic matter (NOM), but researchers have not reached consensus regarding the relative importance of these foulants or the subfractions that are most problematic. Howe and co-workers 2 have implicated colloids in the 100 kDa to 1 μm size range, while others have attributed fouling primarily to the large-MW fraction 4 6 or the neutral hydrophilic fraction of the NOM, especially polysaccharides and proteins.7 11 Our prior studies have demonstrated that deposition of a thin layer of micrometer-sized heated aluminum oxide particles (HAOPs) on membranes prior to delivery of the feed can enhance NOM removal while mitigating fouling; we refer to this process as microgranular adsorptive filtration (μGAF).12,13 The current study was undertaken to compare the performance of various adsorbents in μGAF and to link that behavior to the removal of specific NOM fractions. ’ MATERIALS AND METHODS Source Water, Adsorbents, and Membrane. Water used in the study was collected from Lake Washington (LW) at Portage Bay. Samples were stored at 4 °C and were brought to room temperature (around 20 °C) immediately before the tests. The dissolved organic carbon (DOC) and total organic carbon (TOC) concentrations were in the range of 2.6 3.4 mg/L and were always almost identical, and the UV absorbance at 254 nm r 2011 American Chemical Society
(UV254) was 0.053 0.066 cm 1. In some tests, the water was amended with sodium alginate (Sigma Aldrich). Three adsorbents were investigated: HAOPs, powdered activated carbon (PAC) (PICAPURE L, PICA USA Inc.), and an anion exchange (IX) resin (Amberlite IRA 958, Sigma-Aldrich). The HAOPs were synthesized by heating freshly precipitated Al(OH)3, following a procedure described previously.12,14 The PAC and IX resin particles were too large to be used in μGAF, so they were crushed in an agitator bead mill (Netzsch LabStar). Some characteristics of the pulverized particles and the HAOPs are summarized in Table 1 of the Supporting Information (SI). The membranes used in the tests were 47-mm-diameter, polyethersulfone disks with a nominal pore size of 0.05 μm (Microdyn-Nadir MP005). Each membrane was preconditioned by soaking and rinsing with DI water. Membrane Filtration. All tests used dead-end filtration at a constant flux of 100 L/m2-h (LMH). Adsorbents were predeposited by injecting a concentrated slurry of the particles into the system, feeding deionized (DI) water (Millipore Milli-Q) for 30 min, and then switching the feed to LW water. The surface loading of the adsorbents was fixed at 40 g/m2. Permeate samples were collected at preselected intervals for analysis of UV254 and SEC fractionation, and transmembrane pressure (TMP) was Received: June 28, 2011 Accepted: September 9, 2011 Revised: September 8, 2011 Published: September 11, 2011 8935
dx.doi.org/10.1021/es202219e | Environ. Sci. Technol. 2011, 45, 8935–8940
Environmental Science & Technology
ARTICLE
Figure 2. SEC elution profiles for permeate from a μGAF HAOPs system at different filtration times. (mAU is milli Absorbance Units).
Figure 1. (a) Breakthrough of UV254 and (b) TMP in μGAF systems with various adsorbents. Flux = 100 LMH.
recorded online with a pressure transducer. All μGAF tests were conducted in duplicate and yielded consistent results. A control test with a bare membrane was conducted in parallel with each group of μGAF tests to account for seasonal variability in water quality. Filtration with Column Pretreatment. In some experiments, the source water was pretreated by passage through a column packed with the original-size IX resin or PAC at an empty bed contact time (EBCT) of 5 min. The column pretreatment was started ∼10 h before membrane filtration, and the pretreated water was collected in a continuously mixed reservoir before it was fed to the downstream membrane system. At the end of some runs, the membrane was removed from the cartridge and soaked for two hours in DI water adjusted to pH 12 with NaOH, in an effort to extract the accumulated NOM. The resulting solution was titrated to pH 6.8, using weakly acidic cation exchange resin (Amberlite IRC86, Sigma-Aldrich) to avoid increasing the salt concentration and potentially coagulating the NOM. The neutralized samples were concentrated 10- to 50-fold by freeze-drying (Freezone 4.5, Labconco) prior to analysis by SEC. The SEC chromatogram of redissolved freeze-dried LW samples was almost identical to that of the raw water, indicating that freeze-drying did not significantly alter the NOM characteristics. Alginate Adsorption. Adsorption of alginate was studied by adding various doses of HAOPs, crushed PAC, or original-size IX resin to a solution containing 20 mg/L Na-alginate (6.25 mg/L as DOC). The suspension was mixed for 2 h and was then passed through a 0.45-μm syringe filter prior to DOC analysis. Chemical Analyses. UV254 was measured using a dual-beam spectrophotometer (Perkin-Elmer Lambda-18), and DOC was
determined on filtered samples with a TOC analyzer (Shimadzu TOC-VCSH). Strong correlations were obtained between UV254 and the DOC concentration (SI Figure S1), so UV254 was used as the primary indicator of NOM concentration. The SEC analysis used a DIONEX Ultimate 3000 HPLC system and Ultimate 3000 diode array detector to acquire absorbance data at 200 500 nm. Isocratic flow of 0.01 M NH4HCO3 was delivered through an Agilent PL aquagel OH 30 column (300*7.5, 8 μm) at 0.5 mL/min. SEC chromatograms of LW water at three wavelengths are shown in SI Figure S2. A highly linear calibration curve for apparent molecular weight (AMW) was generated using polyethylene glycol (PEG) standards (Agilent EasiVial PEG) (SI Figure S3), but it suggested that the AMW of the NOM was much larger than that reported by other researchers.15,16 Therefore, although the SEC profiles are useful for comparisons among samples, the absolute AMWs should not be interpreted literally.
’ RESULTS AND DISCUSSION Removal of NOM and Fouling in Filtration Tests. UV254 in the permeate and TMP buildup for several filtration runs are shown in Figure 1. The abscissa in these figures is the specific volume filtered (Vsp), defined as the cumulative permeate volume per unit membrane area. For all tests reported here, a Vsp of 100 L/m2 corresponds to one hour of run time. Consistent with the trends observed in prior research, the bare membrane removed only a few percent of the UV254 in the feed, and the TMP increased throughout the run at a steadily increasing rate. The initial TMP in the system with predeposited HAOPs was essentially the same as in the control system, indicating that the HAOPs contributed negligible resistance to permeation, but the HAOPs system ran for more than three times as long as the control system before the TMP increased to 90 kPa. UV254 removal was much greater in the system with HAOPs, starting at ∼80% and decreasing steadily to ∼50% at the end of the run. As in prior research,13 when the adsorbent was rinsed off at the end of the tests, the membrane permeability was almost completely recovered, indicating that the fouling had occurred in the adsorbent layer or on the membrane surface, not inside the pores. When analogous tests were conducted with PAC, UV254 removal was somewhat less early in the run, but comparable later, and the TMP profile was almost identical. When IX resin was predeposited, UV254 removal was essentially identical to that 8936
dx.doi.org/10.1021/es202219e |Environ. Sci. Technol. 2011, 45, 8935–8940
Environmental Science & Technology
ARTICLE
Figure 4. SEC elution profiles of effluent from PAC and IX resin pretreatment columns. Samples were concentrated 10 by freezedrying.
Figure 3. Normalized breakthrough of different NOM AMW fractions of μGAF systems with predeposited (a) HAOPs; (b) PAC; or (c) IX resin.
with HAOPs, achieving an average UV removal of 70%. Surprisingly though, the IX resin system fouled much faster—in fact, as rapidly as the control system with no adsorbent at all. The dramatic difference in membrane fouling between the systems with IX resin and HAOPs despite their virtually identical UV254 removals, and the identical fouling behavior in the systems with IX resin and with no adsorbent despite their very different UV254 removals, demonstrate clearly that UV254 is not a useful probe for the membrane foulants in NOM. Furthermore, the fact that the bare membrane fouled quickly despite removing very little NOM reinforces the inference that the foulants comprise only a small fraction of the bulk NOM.13 Fouling Potential of Different NOM Fractions. Figure 2 compares the SEC chromatograms of LW water and permeate samples at different times in runs with predeposited HAOPs.
Figure 5. (a) Breakthrough of UV254 and (b) TMP in μGAF systems with IX resin column pretreatment.
Four NOM fractions can be identified in the raw water, with the two highest-AMW peaks partially overlapping. HAOPs selectively removed the high-AMW fractions, especially at the beginning of the run. At a Vsp of 50 L/m2, molecules with AMW > 15 000 (Peaks 1 and 2 in Figure 2) were almost completely removed, along with a significant fraction of those with AMW between 10 000 and 15 000 (Peak 3); by contrast, NOM with AMW <7000 (Peak 4) was not removed. As the run proceeded, the fractions associated with Peaks 2 and 3 gradually broke through, but removal of the fraction associated with Peak 1 remained substantial. 8937
dx.doi.org/10.1021/es202219e |Environ. Sci. Technol. 2011, 45, 8935–8940
Environmental Science & Technology Figure 3 shows breakthrough profiles of the different AMW fractions in systems with each adsorbent, based on SEC peak height. Breakthrough of the AMW fractions in the systems with HAOPs and PAC increased in the same order (greater fractional breakthrough with decreasing AMW), but the shapes of the curves with PAC were all similar to that for Peaks 3 and 4 in the HAOPs system, and different from those for HAOPs Peaks 1 and 2. Both the HAOPs and PAC systems selectively removed the high-MW NOM fractions, and both reduced fouling significantly. However, the IX system fouled rapidly even though it removed the high-AMW fraction as efficiently as the HAOPs system did (Figure 3c). Thus, fouling cannot be attributed to this fraction of the NOM. Rather, the majority of the fouling is apparently contributed by a small portion of the NOM that is partially removed by PAC or HAOPs, but is not removed by IX resin and does not have a strong UV signal. Tests with Column Pretreatment. In an effort to remove much of the NOM with low fouling potential and thereby facilitate the identification of the key foulants, LW water was passed through a column packed with IX resin upstream of the membrane systems; pretreatment with PAC was also explored for comparison. The IX resin column removed 85% of the UV254 from the raw water, leaving predominantly low-AMW NOM in the column effluent (Figure 4); a small amount of additional removal was achieved when the column effluent was passed through a bare membrane (Figure 5a). Although the μGAF system with predeposited IX resin did not mitigate fouling, IX resin pretreatment of the water did: the test ran more than twice as long before the TMP reached 70 kPa as when no pretreatment was included. The average UV254 removal by the IX resin was greater in the pretreatment column than in the μGAF system (85% vs 70%). The fact that membrane fouling was mitigated in the former system but not the latter suggests that the incremental removal included some NOM that contributed to fouling, but had relatively low affinity for the resin. To compare the fouling potential of the NOM removed by the resin with that of the average NOM in the feed, LW water was diluted with deionized water to match the UV254 of the column effluent and was applied to a bare membrane. This run lasted twice as long as the run with IX-pretreated water (Figure 5b). Thus, lowering the NOM concentration by ∼7-dilution was more effective at mitigating fouling then lowering it the same amount by IX treatment, confirming that the IX column selectively removed NOM with lower-than-average fouling potential. When the IX-pretreated water was fed to a μGAF system with predeposited HAOPs, removal of UV254 was essentially the same as by the bare membrane (Figure 5a), but the operation time before the TMP reached 70 kPa doubled. The identical UV254 removal but very different fouling behavior in the systems without and with HAOPs suggests that some strong foulants were reaching the membrane in the former system but not the latter, and that these foulants had negligible UV254. When the same tests were repeated using PAC instead of IX resin to pretreat the water, slightly less UV254 was removed in the pretreatment step, and the NOM in the column effluent was predominantly the high-AMW fraction (Figure 4). When these pretreated waters were membrane-filtered, ∼15% of the original UV254 was removed in the μGAF system with HAOPs, but less than 5% was removed by the bare membrane (Figure 5a). The extra rejection of UV254 by the μGAF system with HAOPs as
ARTICLE
Figure 6. SEC elution profiles of NOM accumulated on membranes receiving different feeds. Profiles are normalized to the peak near AMW 100 000.
compared to the control system after the PAC column, and the absence of any such difference after the IX resin column, are consistent with the prior observation that HAOPs and the IX resin remove the same fractions of UV254-absorbing NOM, but the PAC removes overlapping, but different, fractions. The UV254 rejection by the bare membrane after the PAC pretreatment was less than after the IX resin pretreatment (Figure 5a). If the additional UV254 removal after IX pretreatment were attributable to key foulants, the membrane would have fouled more rapidly in that system. However, the bare membrane downstream of the PAC column fouled faster than that after the IX column (Figure 5b). In an effort to characterize the foulants directly and perhaps detect foulants that are present at very low concentrations in the raw water but accumulate on the membrane during a run, the NOM on the membranes with and without pretreatment was extracted and concentrated. The intensity of the peaks in the SEC profiles of these samples (SI Figure S4) varied substantially, due to the different amounts of foulant deposited and different concentration factors during processing. To account for the concentration factor and to eliminate potential artifacts associated with release of UV-absorbing materials from the titrant, the SEC profiles were normalized by division of the chromatogram by the intensity of the peak at 100 000 AMW that appeared in the blank control but not in the SEC profiles of LW water. The normalized profile from the blank (from processing of a membrane used to treat DI water) was then subtracted from each sample profile. Three UV254-absorbing fractions, with peaks near AMWs of 10 000, 40 000, and 200 000, were detected in the extracted NOM (Figure 6). The 10 000-AMW peak corresponded to a major fraction that had been identified previously in the raw water, but the other two peaks did not. Although these fractions were selectively retained by the membrane, they apparently were not major membrane foulants, since the membrane downstream of the PAC column fouled more rapidly than that downstream of the IX column, even though the latter membrane accumulated more of these fractions than the former did. Collectively, the results suggest that the key foulants in these systems do not absorb UV light. Velten et al. 17 reported SEC/ UV chromatograms of Lake Zurich water that closely match those obtained here for LW water. However, LC/OCD chromatograms of the same samples had an additional peak at high 8938
dx.doi.org/10.1021/es202219e |Environ. Sci. Technol. 2011, 45, 8935–8940
Environmental Science & Technology
ARTICLE
fraction of the NOM that absorbs negligible UV254. Some of the key foulants can be removed by various adsorbents (HAOPs > PAC . IX resin) and appear to include, but not be limited to, hydrophilic NOM such as polysaccharides and proteins. μGAF, especially with HAOPs, can be an effective approach for removing both bulk NOM and the key foulants.
’ ASSOCIATED CONTENT
bS Figure 7. Alginate sorption onto various adsorbents.
Supporting Information. Characteristics of adsorbents; SEC elution profiles for LW and SEC calibration curve; UV254DOC correlation; SEC profiles of NOM rejected on membrane surface. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 206-543-0785; fax: 206-685-9185; e-mail: zhenxiao@ u.washington.edu.
’ ACKNOWLEDGMENT This work is based on work supported by the National Science Foundation under Grant CBET-0931739. ’ REFERENCES
Figure 8. TMP profiles in systems fed alginate-amended solutions.
AMW, which the authors attributed to polysaccharides or other biopolymers.11,17 Therefore, on the basis of the suggestion by other researchers 9,18,19 that alginate can be used as a surrogate for polysaccharides, we next investigated alginate adsorption in batch systems and the contribution of alginate to membrane fouling. In batch tests, sorption of alginate by HAOPs increased steadily with increasing HAOPs dose, reaching ∼90% removal at a dose of 200 mg/L HAOPs (Figure 7). By contrast, the IX resin and PAC adsorbed virtually no alginate. Similarly, Velten et al. 17 reported negligible sorption by activated carbon of the high-AMW “biopolymer” NOM fraction. The fouling rate of a bare membrane increased dramatically when alginate was added to LW water at doses of 0.2 or 1.0 mg/L as DOC (Figure 8). Interestingly, a bare membrane fed 1 mg/L alginate DOC in DI water fouled even faster than one fed the same alginate concentration in LW water, suggesting that interactions of alginate with the background NOM reduced the adverse effects of the alginate. When DI water with 1 mg/L alginate DOC was fed to μGAF systems, fouling was substantially delayed in the system with predeposited HAOPs, but was slightly accelerated in the system with PAC. Since the PAC μGAF system reduced fouling by LW water but not by alginate, fouling in these systems cannot be attributed solely to alginate (or, by extension, hydrophilic NOM). Nevertheless, the results from the HAOPs μGAF support the hypothesis that hydrophilic NOM is often a major contributor to fouling. Implications. The majority of the fouling of low-pressure membranes treating natural water is contributed by a small
(1) Huang, H.; Schwab, K.; Jacangelo, J. G. Pretreatment for low pressure membranes in water treatment: A review. Environ. Sci. Technol. 2009, 43 (9), 3011–3019. (2) Howe, K. J.; Marwah, A.; Chiu, K. P.; Adham, S. S. Effect of coagulation on the size of MF and UF membrane foulants. Environ. Sci. Technol. 2006, 40 (24), 7908–7913. (3) Schafer, A. I.; Schwicker, U.; Fischer, M. M.; Fane, A. G.; Waite, T. D. Microfiltration of colloids and natural organic matter. J. Membr. Sci. 2000, 171 (2), 151–172. (4) Lin, C. F.; Huang, Y. J.; Hao, I. J. Ultrafiltration processes for removing humic substances: Effect of molecular weight fractions and PAC treatment. Water Res. 1999, 33 (5), 1252–1264. (5) Yuan, W.; Zydney, A. L. Effects of solution environment on humic acid fouling during microfiltration. Desalination 1999, 122 (1), 63–76. (6) Yuan, W.; Zydney, A. L. Humic acid fouling during ultrafiltration. Environ. Sci. Technol. 2000, 34 (23), 5043–5050. (7) Carroll, T.; King, S.; Gray, S. R.; Bolto, B. A.; Booker, N. A. The fouling of microfiltration membranes by NOM after coagulation treatment. Water Res. 2000, 34 (11), 2861–2868. (8) Cho, J.; Amy, G.; Pellegrino, J. Membrane filtration of natural organic matter: factors and mechanisms affecting rejection and flux decline with charged ultrafiltration (UF) membrane. J. Membr. Sci. 2000, 164 (1 2), 89–110. (9) Jermann, D.; Pronk, W.; Meylan, S.; Boller, M. Interplay of different NOM fouling mechanisms during ultrafiltration for drinking water production. Water Res. 2007, 41 (8), 1713–1722. (10) Lee, N. H.; Amy, G.; Croue, J. P.; Buisson, H. Identification and understanding of fouling in low-pressure membrane (MF/UF) filtration by natural organic matter (NOM). Water Res. 2004, 38 (20), 4511–4523. (11) Meylan, S.; Hammes, F.; Traber, J.; Salhi, E.; von Gunten, U.; Pronk, W. Permeability of low molecular weight organics through nanofiltration membranes. Water Res. 2007, 41 (17), 3968–3976. (12) Cai, Z. X.; Kim, J. S.; Benjamin, M. M. NOM removal by adsorption and membrane filtration using heated aluminum oxide particles. Environ. Sci. Technol. 2008, 42 (2), 619–623. 8939
dx.doi.org/10.1021/es202219e |Environ. Sci. Technol. 2011, 45, 8935–8940
Environmental Science & Technology
ARTICLE
(13) Kim, J.; Cai, Z. X.; Benjamin, M. M. NOM fouling mechanisms in a hybrid adsorption/membrane system. J. Membr. Sci. 2010, 349 (1 2), 35–43. (14) Kim, J. S.; Cai, Z. X.; Benjamin, M. M. Effects of adsorbents on membrane fouling by natural organic matter. J. Membr. Sci. 2008, 310 (1 2), 356–364. (15) Korshin, G.; Chow, C. W. X.; Fabris, R.; Drikas, M. Absorbance spectroscopy-based examination of effects of coagulation on the reactivity of fractions of natural organic matter with varying apparent molecular weights. Water Res. 2009, 43 (6), 1541–1548. (16) Chin, Y. P.; Aiken, G.; Oloughlin, E. Molecular-weight, polydispersity, and spectroscopic properties of aquatic humic substances. Environ. Sci. Technol. 1994, 28 (11), 1853–1858. (17) Velten, S.; Knappe, D. R. U.; Traber, J.; Kaiser, H.-P.; von Gunten, U.; Boller, M.; Meylan, S. Characterization of natural organic matter adsorption in granular activated carbon adsorbers. Water Res. 2011, 45 (13), 3951–3959. (18) van de Ven, W.; Punt, I.; Kemperman, A.; Wessling, M. Unraveling ultrafiltration of polysaccharides with flow field flow fractionation. J. Membr. Sci. 2009, 338 (1 2), 67–74. (19) Katsoufidou, I. S.; Sioutopoulos, D. C.; Yiantsios, S. G.; Karabelas, A. J. UF membrane fouling by mixtures of humic acids and sodium alginate Fouling mechanisms and reversibility. Desalination 2010, 264 (3), 220–227.
8940
dx.doi.org/10.1021/es202219e |Environ. Sci. Technol. 2011, 45, 8935–8940
ARTICLE pubs.acs.org/est
Nanofiltration Membrane Fouling by Oppositely Charged Macromolecules: Investigation on Flux Behavior, Foulant Mass Deposition, and Solute Rejection Yi-Ning Wang and Chuyang Y. Tang* School of Civil and Environmental Engineering, and Singapore Membrane Technology Centre, Nanyang Technological University, Singapore 639798
bS Supporting Information ABSTRACT: Nanofiltration membrane fouling by oppositely charged polysaccharide (alginate) and protein (lysozyme) was systematically studied. It was found that membrane flux decline in the presence of both lysozyme and alginate was much more severe compared to that when there was only lysozyme or alginate in the feed solution. The flux performance for the mixed foulants was only weakly affected by solution pH and calcium concentration. These effects were likely due to the strong electrostatic attraction between the two oppositely charged foulants. Higher initial flux caused increased foulant deposition, more compact foulant layer, and more severe flux decline. The deposited foulant cake layer had a strong tendency to maintain a constant foulant composition that was independent of the membrane initial flux and only weakly dependent on the relative foulant concentration in feed solution. In contrast, solution chemistry (pH and [Ca2+]) had marked effect on the foulant layer composition, likely due to the resulting changes in the foulant foulant interaction. The mixed alginate lysozyme fouling could result in an initial enhancement in salt rejection. However, such initial enhancement was not observed when there was 1 mM calcium present in the feedwater, which may be attributed to the charge neutralization of the foulant layer.
1. INTRODUCTION Fouling of reverse osmosis (RO) and nanofiltration (NF) membranes by colloids and macromolecules has received considerable attention in the past few decades. Prior fouling studies with various feeds have revealed many important factors that affect fouling, including solution chemistry, hydrodynamic conditions, and membrane properties.1 A recent review of the mechanisms and factors controlling fouling of both RO and NF membranes1 pointed out the importance of mass transfer near the membrane surface and colloid membrane (and colloid colloid) interactions that are strongly affected by the feed solution chemistry. Past studies on single macromolecular feeds showed that electrostatic interactions between molecules play an important role, which was inferred from the effect of solution pH and ionic strength on fouling.2 8 Fouling was generally more severe at pH values close to the isoelectric point (IEP) of the macromolecules and at higher ionic strength due to reduced electrostatic repulsion.6 9 Calcium complexation with macromolecules has also been observed to increase the fouling rate markedly.7 11 The feedwater for NF or RO often contains more than one foulant. The handful of existing studies on mixed foulant feeds have shown different dominant effects and fouling mechanisms, including (1) the prefiltering effect, i.e., one foulant acts as a prefilter for another to reduce fouling in porous microfiltration (MF) and ultrafiltration (UF) membranes;12,13 (2) the r 2011 American Chemical Society
synergistic effect that explains the more severe flux decline by the mixed foulants compared to that by any single foulant;6,14,15 (3) the averaging effect where the rate and the extent of flux reduction for the mixture falls between that of the two single foulants;10,16 and (4) adsorption effect where macromolecules coat inorganic particles that alters the surface properties on the latter.17,18 Nevertheless, the effect of solution chemistry and hydrodynamic conditions on RO and NF fouling by mixed macromolecules is yet to be systematically studied.15 Moreover, an understanding of cake formation and ion rejection due to fouling by mixed macromolecules is still lacking. Little work has been performed to understand of the factors controlling the composition of the foulant cake layer. Therefore, the current study had three main objectives: (1) to systematically investigate solution chemistry and hydrodynamic effects on NF membrane fouling by a feed solution containing oppositely charged macromolecules; (2) to correlate mass deposition with flux decline at various conditions; and (3) to compare the change in salt rejection and correlation with cake deposition for different solution chemistries. Received: August 4, 2011 Accepted: September 19, 2011 Revised: September 14, 2011 Published: September 19, 2011 8941
dx.doi.org/10.1021/es202709r | Environ. Sci. Technol. 2011, 45, 8941–8947
Environmental Science & Technology
ARTICLE
2. MATERIALS AND METHODS 2.2. Chemicals. Unless specified otherwise, all reagents and chemicals were of analytical grade with purity over 99%. Ultrapure water with a resistivity of 18.2 MΩ.cm (Millipore Integral 10 Water Purification System) was used to prepare all working solutions. The pH and ionic compositions of the feed solution were adjusted by the addition of sodium chloride, calcium chloride, hydrochloric acid, and sodium hydroxide. Sodium alginate (ALG, Sigma A2158) and lysozyme (LYS, Fluka 62971) were used to represent polysaccharide and protein foulants of opposite charges under the testing conditions. These model foulants were chosen for their ready availability and their wide use in membrane fouling studies.3,6,10 Alginate is negatively charged at neutral pH due to its high content of carboxylic groups, and LYS is positively charged below pH 10.4. The molecular weights of alginate and LYS are ∼12 80 kDa and 14.3 kDa, respectively.3,6,13 Both foulants were received in powder form with purity above 98% and were stored at 4 °C in the dark. Working solutions were freshly prepared prior to each fouling experiment. 2.3. Nanofiltration Membrane. A commercial nanofiltration membrane NF270 (Dow FilmTec) was used in the current study. According to our prior studies,19,20 this NF membrane is a poly(piperazine)-based thin film composite polyamide membrane. Its surface is smooth (root-mean-square roughness ∼9 nm), negatively charged (zeta potential of ∼ 35 mV at pH 7), and highly hydrophilic (contact angle below 30°).19 21 The permeability and the NaCl rejection of the membrane are ∼0.87 L/m2 3 h 3 psi and 50 60%, respectively, when tested at 90 psi using a 10 mM NaCl feedwater.9 2.4. Fouling Experiment. Membrane fouling experiments were performed using a laboratory-scale crossflow filtration setup under constant pressure conditions. A detailed description of the setup can be found elsewhere.9 The fouling test procedure was adapted from Tang and co-workers.9,15,22 Briefly, membrane coupons (active membrane area ∼42 cm2) were first compacted and equilibrated with the background electrolyte solution (of composition identical to that used for the subsequent fouling test except that foulants were not added) for 2 days under pressure. This was to ensure that any subsequent flux decline after foulant addition was not influenced by the mechanical compaction of membrane. At the end of the 2-day equilibration stage, predissolved foulant solutions were added to achieve a desired total foulant concentration of 20 mg/L. The membrane flux immediately before the addition of foulants was recorded as the initial flux. The fouling experiments continued for 4 days. Pressure and crossflow velocity were maintained constant throughout each fouling experiment. The effects of relative foulant concentration, pH, calcium concentration, and initial flux were evaluated by varying one variable at a time while maintaining the rest at constant conditions. Unless specified otherwise, the following reference conditions were applied: • total foulant concentration of 20 mg/L (LYS + ALG for the mixed system); • total ionic strength of 10 mM (adjusted by the addition of NaCl and CaCl2); • crossflow velocity was 9.5 cm/s with diamond patterned spacer used in the feedwater channel; and • feed tank temperature controlled at 20 ( 1 °C. 2.5. Foulant Extraction and Mass Deposition Analysis. Fouled membrane coupons were gently rinsed with Milli-Q
Figure 1. Zeta potential of alginate and lysozyme in 10 mM NaCl as a function of pH. At least two replicate measurements were performed for each solution condition.
water to remove any labile foulants. Membrane samples (area of 1.267 2.534 cm2) were cut from the fouled coupons and used for foulant extraction. Sodium dodecyl sulfate (SDS, 5% 23 and NaOH (0.1%)24 were used for LYS and alginate extraction, respectively. Mild sonication (below 30 °C, 20 min) was employed after one day soaking of the samples in the extraction solution. The extractant of LYS was analyzed using a protein assay kit (Sigma, QuantiPro BCA Assay Kit, 0.5 30 μg/mL) at UV of 562 nm,25 and that of alginate was examined using phenol sulfuric acid method at UV of 485 nm (UV spectrophotometer, UV-1700 Shimadzu).26 2.6. Zeta Potential Measurement of Macromolecules. Zeta potential of LYS and alginate was measured using a Malvern ZetaSizer Nano ZS according to our prior publication.9 Freshly prepared LYS or alginate solutions were used in the measurements. Measurements were conducted in a background electrolyte of 10 mM NaCl, and the solution pH was adjusted by the addition of HCl or NaOH.
3. RESULTS AND DISCUSSION 3.1. Zeta Potential of Foulants. Figure 1 shows the zeta potential results of LYS and alginate. LYS was positively charged at circumneutral pH (∼ 7 mV at pH 7) with an isoelectric point (pHIEP) ∼ 10.4. It was more positively charged at lower pH. Alginate was negatively charged over the entire pH range evaluated (pH 2 9). Its zeta potential value decreased from ∼ 20 mV at pH 2 to ∼ 70 mV at pH 6. The value was nearly constant above pH 6, likely due to the complete deprotonation of carboxylic groups.6 3.2. Flux Behavior and Foulant Mass Deposition. 3.2.1. Effect of Feed Foulant Composition. The effect of feed foulant composition on membrane fouling was investigated by varying the LYS content (0, 5, 30, 50, 70, and 100 wt % of total feed foulant) while keeping the total feed foulant concentration at 20 mg/L. Figure 2a shows the membrane flux performances under an initial flux of 75 L/m2 3 h. Least flux decline was observed over the 4-day fouling tests when the feed contained either alginate alone (0% LYS, ∼ 30% flux decline) or lysozyme 8942
dx.doi.org/10.1021/es202709r |Environ. Sci. Technol. 2011, 45, 8941–8947
Environmental Science & Technology
ARTICLE
Figure 2. Effect of feed foulant composition on flux behavior and foulant mass accumulation: (a) flux performance; (b) foulant mass accumulation after 96-h fouling. Other test conditions: total foulant concentration of 20 mg/L, initial flux of 75 L/m2 3 h, pH 7, ionic strength of 10 mM, crossflow velocity of 9.5 cm/s, and temperature at 20 ( 1 °C.
Figure 3. Effect of solution chemistry on flux behavior and foulant mass accumulation: (a) flux performance; (b) mass accumulation after 96-h fouling. Other test conditions: initial flux of 75 L/m2 3 h, feedwater containing 10 mg/L lysozyme and 10 mg/L alginate, total ionic strength of 10 mM, crossflow velocity of 9.5 cm/s, and temperature at 20 ( 1 °C.
alone (100% LYS, ∼ 20% flux decline). With a small amount of LYS (5%) in the feed, a much more severe flux reduction (∼ 50%) occurred. This was likely due to the increased foulant mass deposition (Figure 2b) as a result of the electrostatic attraction between the oppositely charged LYS and alginate molecules. The most severe fouling occurred at the LYS relative concentration of 30, 50, and 70% with ∼80% flux loss. The fastest rate of initial flux decline was observed for the feed with 70% LYS (inset of Figure 2a). The reason for explaining the effect of foulant relative concentration is discussed in the following text based on the mass deposition analysis. Figure 2b presents the foulant mass and composition of the extracted cake layer. Compared to the single foulant system (0 or 100% LYS), the mixed foulant system had a much greater amount of foulant mass deposition. A small amount of LYS (e.g., 5% LYS) in the mixed foulants increased the total foulant mass deposition to ∼300 μg/cm2, compared to ∼100 μg/cm2 when there was only alginate in the feed. Even more foulant deposition (400 500 μg/cm2) was observed at LYS contents of
30 70%, which correlated well with the more severe flux reduction under those conditions. Interestingly, the deposited mass ratio of LYS to alginate in the foulant layer was nearly constant (mLYS/malg = ∼4) for feeds with 30 70% LYS content. There seemed to be a strong tendency that the relative quantity of LYS on fouled membranes approached a quite constant value (∼ 80% of the total deposited mass) when sufficient concentrations of LYS and alginate were available in the feed solution. Conceptually, there exists a thermodynamically most stable arrangement of foulant cake deposition (corresponding to a minimum free energy state27). To achieve such a minimum energy arrangement, it is reasonable to assume that the oppositely charged LYS and alginate may deposit in a certain configuration to minimize the electrostatic potential energy (e.g., a mosaic arrangement where each molecule is surrounded by adjacent oppositely charged molecules). Therefore, the mass ratio mLYS/malg is presumably determined by the relative charge densities of two foulants, their geometry, and the presence of other additional interactions (e.g., hydrophobic interaction or specific 8943
dx.doi.org/10.1021/es202709r |Environ. Sci. Technol. 2011, 45, 8941–8947
Environmental Science & Technology interactions). This thermodynamic tendency (i.e., constant mLYS/malg) will be preserved regardless of the foulant mass ratio in the feedwater, as long as both LYS and alginate are present in sufficient quantity (LYS 30 70% in this study). The constant mLYS/malg ratio is somewhat analogous to the precipitation (scaling) of inorganic minerals such as CaSO4 (gypsum) where the calcium to sulfate ratio in the precipitation is fixed (at 1:1 in this particular case) irrespective of their ratio in the bulk solution.28 The fouling behavior at lower LYS content in feed is also interesting. Even when there was only 5% LYS in the feedwater, the relative LYS content in the foulant layer was as high as 53%. The corresponding mLYS/malg ratio was ∼1. Although it was still significantly lower than the “thermodynamic constant ratio” of ∼4, this ratio in the foulant layer was an order of magnitude higher than the corresponding ratio in the feed (∼0.053). This observation may suggest that there might be a tendency for the foulant layer mass ratio to approach the thermodynamic constant ratio even at 5% LYS in feed, although this ratio was also limited by the slow deposition of LYS due to its low feed concentration (i.e., due to the kinetic constraint). The same kinetic consideration could also explain why the feeds with 30, 50, and 70% LYS content fouled membrane more severely. It could predict that the fastest initial flux drop shall occur for the feed containing the “desired” foulant proportion for cake formation (i.e., LYS to alginate concentration ratio of ∼4). In this way, the feed containing 70% LYS initiated faster flux decline compared to 50 and 30% LYS content, likely due to the foulants composed of 70% LYS and 30% alginate having the most proximal composition to the desired proportion of cake formation. 3.2.2. Effect of Solution Chemistry. The effect of pH and calcium concentration on flux behavior is shown in Figure 3a for a feed containing 10 mg/L LYS and 10 mg/L alginate. A nearly identical flux behavior was observed at both pH 5 and pH 7, presumably resulting from the similar strong electrostatic attraction between the oppositely charged macromolecules. Unlike the important role of solution pH in single protein fouling,9 it became less crucial during mixture fouling at least for the conditions explored in the current study. This observation is consistent with our prior study that the flux decline for a feed containing a binary protein mixture was only weakly dependent on pH when it was between the IEPs of the two proteins.15 Notwithstanding the identical flux behavior at different solution pHs, the results of foulant mass deposition showed some remarkable differences (Figure 3b). There were more alginate and less LYS deposition at pH 5 compared to pH 7 at the end of fouling tests, and the relative mass of LYS on the fouled membrane reduced from ∼80% at pH 7 to 66% at pH 5. At the reduced pH, LYS was more positively charged whereas alginate was less negatively charged (Figure 1), which may require a lesser amount of LYS to balance the negatively charged alginate. This is consistent with our hypothesis that the mLYS/malg ratio is dependent on the relative charge density of the foulants—an increased LYS to alginate charge ratio leads to a reduced mLYS/malg. Calcium concentration effect was studied by comparing the feeds containing 1 and 0 mM Ca2+ with a total ionic strength of 10 mM at pH 7 (Figure 3a). The presence of 1 mM Ca2+ only led to slightly more severe flux reduction (86% reduction, compared to 80% without Ca2+). It has been well recognized that alginate can form a cross-linked gel layer in the presence of calcium ions as a result of calcium complexation with alginate carboxylic groups.10 When alginate is the sole foulant in the feed, increasing
ARTICLE
Figure 4. Effect of initial flux on flux behavior and foulant mass accumulation: (a) flux performance; (b) mass accumulation after 96-h fouling. Other test conditions: feedwater containing 10 mg/L lysozyme and 10 mg/L alginate, pH 7, Ca2+ concentration of 1 mM, total ionic strength of 10 mM, crossflow velocity of 9.5 cm/s, and temperature at 20 ( 1 °C.
calcium concentration tends to drastically reduce membrane permeate flux.10,29 In contrast, this marked effect of Ca2+ was not observed in the fouling by the LYS alginate mixture. The reduced Ca2+ effect was probably due to the presence of LYS in feed solution, which can cause a strong LYS alginate attraction and form a dense cake layer even if Ca2+ was not present. Alternatively, the positively charged LYS may compete with Ca2+ for the negatively charged sites ( COO groups) in the alginate, leading to a reduced degree of Ca2+ alginate complexation. Introducing Ca2+ into the feed solution only marginally accelerated flux decline at the beginning (within 10 h) upon foulant addition. During such a fouling process, a compound effect of LYS alginate and Ca2+ alginate complexation may take place (noting that Ca2+ had minimal effect on the fouling by LYS alone, see Supporting Information S1). Mass deposition results (Figure 3b) show a significant increase (more than doubled) in the amount of alginate accumulation on the fouled membrane due to the addition of Ca2+, despite the insignificant variation in flux performance. The increased alginate deposition may be attributed to the Ca2+ alginate 8944
dx.doi.org/10.1021/es202709r |Environ. Sci. Technol. 2011, 45, 8941–8947
Environmental Science & Technology
Figure 5. Foulant layer hydraulic resistance and specific cake layer resistance as a function of total foulant mass deposition. Other test conditions: feedwater containing 10 mg/L lysozyme and 10 mg/L alginate, pH 7, Ca2+ concentration of 1 mM, total ionic strength of 10 mM, crossflow velocity of 9.5 cm/s, and temperature at 20 ( 1 °C.
complexation. On the other hand, LYS deposition was reduced, and the relative mass of LYS on membrane was reduced to ∼52%. This further suggested the existence of the compound effect of LYS alginate and Ca2+ alginate complexation, where both LYS and Ca2+ compete for negatively charged COO sites in alginate. The specific Ca2+ alginate interaction and subsequent reduction in the charge density of alginate may also explain the drastically reduced mLYS/malg ratio (∼4 when there was no calcium and ∼1 in the presence of 1 mM Ca2+). Once again, the reduced LYS to alginate mass ratio corresponds to an increased ratio of charge density. 3.2.3. Effect of Initial Flux. The effect of initial flux was evaluated over an initial flux range of 15 120 L/m2 3 h. The feedwater contained 1 mM Ca2+ at pH 7. Much more severe flux reduction was observed at higher initial flux (Figure 4a), which was consistent with the results from our previous fouling studies with single protein9 and binary protein mixture,15 and as well as from many other existing publications.5,22,30 The fluxes started to converge to an identical limiting value although they began with different initial values; this was consistent with our observation in protein fouling and humic acid fouling.7 9,22 The surface-interaction controlled limiting flux model, which states that the mutual action of hydrodynamic drag and foulant depositedfoulant (foulant membrane) interaction determines foulant deposition and the extent of fouling, is appropriate to explain the limiting flux phenomenon in the current study.1,7 A similar effect of initial flux was also observed for the feed without Ca2+ (S2(a), Supporting Information). Foulant mass deposition results (Figure 4b and S2(b)) show an expected trend that more LYS and alginate deposited on membranes fouled under higher initial flux. It is worthwhile to note that the deposited mass of the two foulants varied synchronously such that their mass ratio remained nearly constant (mLYS/malg = ∼1 for the feed containing 1 mM Ca2+ and ∼4 for the feed without Ca2+). This suggests that increasing initial flux (and/or applied pressure) may have negligible effect on the foulant layer composition and that the mLYS/malg ratio may be primarily determined by the foulant foulant interactions but not the hydrodynamic forces.
ARTICLE
Figure 6. Foulant layer composition as a function of feed composition. The feedwater had a total foulant concentration of 20 mg/L, ionic strength of 10 mM, crossflow velocity of 9.5 cm/s, and temperature at 20 ( 1 °C.
As most hydraulic cake resistance was built up at the initial period where the flux dropped significantly (Figure 4a and S2(a)), the large difference in the deposited cake mass under different initial fluxes may develop within the initial fouling period in the current study. Operating at a higher initial flux (or higher applied pressure) had no noticeable benefit on the longer term membrane flux (say, at 96 h), likely due to the densely formed cake layer. Figure 5 shows that the hydraulic resistance of foulant cake layer Rf increased with the total deposited foulant mass mf. In addition, the specific cake layer resistance (i.e., Rf/mf) also increased at higher mf. The increased specific cake layer resistance may be attributed to the cake layer compaction at the higher initial flux (and applied pressure).9 In other words, the higher initial flux resulted in more foulant deposition, a more compact cake layer, and thus greater flux reduction, which agrees well with a previous characterization study on humic acid fouling with RO and NF membranes.31 The various effects (solution chemistry, feed composition, and initial flux) on the foulant layer composition are presented in Figure 6. These can be summarized into the following three aspects: (1) the foulant composition in the feed solution only had weak influence on the relative mass of deposited foulants as long as both foulants were present in the feed in sufficient concentrations (such that the mLYS/malg ratio was not kinetically limited); (2) the initial flux had negligible effect on the foulant composition in the deposited cake layer; and (3) the solution chemistry such as pH and Ca2+ concentration affected the mLYS/malg ratio as a result of the changes in foulant foulant interactions (e.g., the LYS alginate electrostatic attraction and the calcium alginate specific interaction). 3.3. Effect of Fouling on Salt Rejection. Figure 7 presents the normalized salt rejection performances (normalized against the corresponding virgin membrane rejection) as a function of fouling time for different solution compositions. The effect of feed foulant composition is shown in Figure 7a. It is interesting to observe an initial increase in salt rejection upon fouling for all cases. Tang et al.22 attributed this beneficial effect to the preferential deposition of foulants over membrane defects where the localized flux was at maximum. In this way, even a small amount of foulants deposition may significantly improve salt 8945
dx.doi.org/10.1021/es202709r |Environ. Sci. Technol. 2011, 45, 8941–8947
Environmental Science & Technology
ARTICLE
was clearly observable when there was no Ca2+ in the feed solution (at both pH 5 and pH 7). In contrast, such initial enhancement was not observed for the feed containing 1 mM Ca2+, even though (1) the flux decline in the presence of calcium was only slightly more severe than that without calcium, and (2) the foulant cake layer was likely more compact with the presence of calcium (Supporting Information S3). A possible explanation may be that the foulant charge neutralization associated with the calcium alginate complex formation may result in a cake layer with reduced ability of solute retention. Prior studies have demonstrated that rejection of charged solutes can be enhanced by Donnan exclusion.33 Thus, the loss of rejection in the presence of 1 mM Ca2+ might be attributed to the weakening of Donnan exclusion mechanism in addition to the enhanced solute polarization inside the foulant cake layer.
4. IMPLICATIONS The current study investigated the effect of feed foulant composition, solution chemistry, and initial flux on membrane fouling by oppositely charged macromolecular foulants. It was found, for the first time, that the foulant layer composition was primarily governed by the feed solution chemistry and that this composition had little or weak dependence on initial membrane flux or foulant composition in the feed solution. Meanwhile, the flux behavior was regulated by the interplay of membrane flux and foulant foulant (and foulant membrane) interaction. Although the current study was performed for a nanofiltration membrane, the mechanisms revealed in the current study may also apply in other membrane systems, such as reverse osmosis, ultrafiltration, and microfiltration. This study may have significant implications in membrane fouling control, particularly for membrane based wastewater reclamation plants and membrane bioreactors, where fouling by protein/polysaccharide mixture may likely occur. ’ ASSOCIATED CONTENT Figure 7. Salt rejection as a function of fouling duration: (a) effect of feed foulant composition at pH 7 without Ca2+; (b) effect of solution chemistry (i.e, pH and Ca2+ concentration) for a feed containing 10 mg/L lysozyme and 10 mg/L alginate. Other test conditions: total ionic strength of 10 mM, crossflow velocity of 9.5 cm/s, and temperature at 20 ( 1 °C.
rejection by defects sealing.22 Nevertheless, the rejection started to decrease at later stages of fouling probably due to a combination of several factors: (1) the decreased membrane permeate flux which resulted in a reduced dilution factor to the permeate concentration; (2) the cake-enhanced concentration polarization of solutes when the foulant cake layer became more extensive;32 and (3) the significant change of membrane surface charge properties upon fouling. Consistent with the above explanations, the decrease in salt rejection in the later stage of fouling seemed to be more obvious for more severe fouling conditions that were accompanied by greater permeate flux reduction and more foulant mass deposition (e.g., for the feed containing 30 70% LYS). In contrast, when the feed only contained LYS or alginate, relatively stable rejection was maintained, consistent with the milder membrane fouling for these cases. Figure 7b presents the effect of solution chemistry on salt rejection during fouling. An initial salt rejection enhancement
bS
Supporting Information. S1. Effect of ionic composition on flux behavior during LYS fouling; S2. Effect of initial flux on flux behavior and foulant mass accumulation in absence of calcium; S3. Foulant mass deposition and specific cake layer resistance. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Tel: +65 6790 5267; fax: +65 6791 0676; e-mail: cytang@ntu. edu.sg; mail: 50 Nanyang Avenue N1-1b-35, School of Civil & Environmental Engineering, Nanyang Technological University, Singapore 639798.
’ ACKNOWLEDGMENT We thank the Environment and Water Industry Programme Office (EWI) under the National Research Foundation of Singapore (Project MEWR C651/06/176) for funding the research work carried out in this manuscript. The comments from Professor William B. Krantz have significantly strengthened this paper. We also thank Dow FilmTec for providing membrane samples. 8946
dx.doi.org/10.1021/es202709r |Environ. Sci. Technol. 2011, 45, 8941–8947
Environmental Science & Technology
’ REFERENCES (1) Tang, C. Y.; Chong, T. H.; Fane, A. G. Colloidal interactions and fouling of NF and RO membranes: A review. Adv. Colloid Interface Sci. 2011, 164, 126–143. (2) Palecek, S. P.; Mochizuki, S.; Zydney, A. L. Effect of ionic environment on BSA filtration and the properties of BSA deposits. Desalination 1993, 90 (1 3), 147–159. (3) Palecek, S. P.; Zydney, A. L. Intermolecular electrostatic interactions and their effect on flux and protein deposition during protein filtration. Biotechnol. Prog. 1994, 10 (2), 207–213. (4) Yuan, W.; Zydney, A. L. Humic acid fouling during ultrafiltration. Environ. Sci. Technol. 2000, 34 (23), 5043–5050. (5) She, Q.; Tang, C. Y.; Wang, Y. N.; Zhang, Z. The role of hydrodynamic conditions and solution chemistry on protein fouling during ultrafiltration. Desalination 2009, 249 (3), 1079–1087. (6) Ang, W. S.; Elimelech, M. Protein (BSA) fouling of reverse osmosis membranes: Implications for wastewater reclamation. J. Membr. Sci. 2007, 296 (1 2), 83–92. (7) Tang, C. Y.; Leckie, J. O. Membrane independent limiting flux for RO and NF membranes fouled by humic acid. Environ. Sci. Technol. 2007, 41 (13), 4767–4773. (8) Tang, C. Y.; Kwon, Y. N.; Leckie, J. O. The role of foulant foulant electrostatic interaction on limiting flux for RO and NF membranes during humic acid fouling - Theoretical basis, experimental evidence, and AFM interaction force measurement. J. Membr. Sci. 2009, 326 (2), 526–532. (9) Wang, Y.; Tang, C. Y. Protein fouling of nanofiltration, reverse osmosis, and ultrafiltration membranes - The role of hydrodynamic conditions, solution chemistry, and membrane properties. J. Membr. Sci. 2011, 376 (1 2), 275–282. (10) Lee, S.; Elimelech, M. Relating organic fouling of reverse osmosis membranes to intermolecular adhesion forces. Environ. Sci. Technol. 2006, 40 (3), 980–987. (11) Li, Q.; Xu, Z.; Pinnau, I. Fouling of reverse osmosis membranes by biopolymers in wastewater secondary effluent: Role of membrane surface properties and initial permeate flux. J. Membr. Sci. 2007, 290 (1 2), 173–181. (12) Arora, N.; Davis, R. H. Yeast cake layers as secondary membranes in dead-end microfiltration of bovine serum albumin. J. Membr. Sci. 1994, 92 (3), 247–256. (13) Palacio, L.; Ho, C. C.; Pradanos, P.; Hernandez, A.; Zydney, A. L. Fouling with protein mixtures in microfiltration: BSA-lysozyme and BSA-pepsin. J. Membr. Sci. 2003, 222 (1 2), 41–51. (14) Iritani, E.; Mukai, Y.; Murase, T. Separation of binary protein mixtures by ultrafiltration. Filtr. Sep. 1997, 34 (9), 967–973. (15) Wang, Y.; Tang, C. Y. Fouling of nanofiltration, reverse osmosis and ultrafiltration membranes by protein mixtures: The role of interfoulant-species interaction. Environ. Sci. Technol. 2011, 45 (15), 6373–6379. (16) Zazouli, M. A.; Nasseri, S.; Ulbricht, M. Fouling effects of humic and alginic acids in nanofiltration and influence of solution composition. Desalination 2010, 250 (2), 688–692. (17) Lee, S.; Cho, J.; Elimelech, M. Combined influence of natural organic matter (NOM) and colloidal particles on nanofiltration membrane fouling. J. Membr. Sci. 2005, 262 (1 2), 27–41. (18) Contreras, A. E.; Kim, A.; Li, Q. Combined fouling of nanofiltration membranes: Mechanisms and effect of organic matter. J. Membr. Sci. 2009, 327 (1 2), 87–95. (19) Tang, C. Y.; Kwon, Y.-N.; Leckie, J. O. Effect of membrane chemistry and coating layer on physiochemical properties of thin film composite polyamide RO and NF membranes. I. FTIR and XPS characterization of polyamide and coating layer chemistry. Desalination 2009, 242 (1 3), 149–167. (20) Tang, C. Y.; Kwon, Y.-N.; Leckie, J. O. Effect of membrane chemistry and coating layer on physiochemical properties of thin film composite polyamide RO and NF membranes II. Membrane physiochemical properties and their dependence on polyamide and coating layers. Desalination 2009, 242 (1 3), 168–182.
ARTICLE
(21) Tang, C. Y.; Fu, Q. S.; Criddle, C. S.; Leckie, J. O. Effect of flux (transmembrane pressure) and membrane properties on fouling and rejection of reverse osmosis and nanofiltration membranes treating perfluorooctane sulfonate containing wastewater. Environ. Sci. Technol. 2007, 41 (6), 2008–2014. (22) Tang, C. Y.; Kwon, Y.-N.; Leckie, J. O. Fouling of reverse osmosis and nanofiltration membranes by humic acid -- Effects of solution composition and hydrodynamic conditions. J. Membr. Sci. 2007, 290 (1 2), 86–94. (23) Jones, K. L.; O’Melia, C. R. Protein and humic acid adsorption onto hydrophilic membrane surfaces: Effects of pH and ionic strength. J. Membr. Sci. 2000, 165 (1), 31–46. (24) Guo, X.; Chen, X.; Hu, W. Studies on cleaning the polyvinylchloride ultrafiltration membrane fouled by sodium alginate. Environ. Technol. 2009, 30 (5), 431–435. (25) Brown, R. E.; Jarvis, K. L.; Hyland, K. J. Protein measurement using bicinchoninic acid: Elimination of interfering substances. Anal. Biochem. 1989, 180 (1), 136–139. (26) Dubois, M.; Gilles, K. A.; Hamilton, J. K.; Rebers, P. A.; Smith, F. Colorimetric method for determination of sugars and related substances. Anal. Chem. 1956, 28 (3), 350–356. (27) Greiner, W.; Neise, L.; St€ocker, H.; Rischke, D.Thermodynamics and Statistical Mechanics; Springer: New York, 1995 (28) Stumm, W.; Morgan, J., Aquatic Chemistry: Chemical Equilibria and Rates in Natural Waters, 3rd ed.; Wiley-Interscience: New York, 1996; p 1040. (29) Ang, W. S. Optimization of Chemical Cleaning of Organicfouled Reverse Osmosis Membranes. Yale University, 2008. (30) Wu, D.; Howell, J. A.; Field, R. W. Critical flux measurement for model colloids. J. Membr. Sci. 1999, 152 (1), 89–98. (31) Tang, C. Y.; Kwon, Y.-N.; Leckie, J. O. Characterization of humic acid fouled reverse osmosis and nanofiltration membranes by transmission electron microscopy and streaming potential measurements. Environ. Sci. Technol. 2007, 41 (3), 942–949. (32) Hoek, E. M. V.; Elimelech, M. Cake-Enhanced Concentration Polarization: A New Fouling Mechanism for Salt-Rejecting Membranes. Environ. Sci. Technol. 2003, 37 (24), 5581–5588. (33) Schaep, J.; Vandecasteele, C.; Mohammad, A. W.; Bowen, W. R. Analysis of the salt retention of nanofiltration membranes using the Donnan-steric partitioning pore model. Sep. Sci. Technol. 1999, 34 (15), 3009–3030.
8947
dx.doi.org/10.1021/es202709r |Environ. Sci. Technol. 2011, 45, 8941–8947
ARTICLE pubs.acs.org/est
Regional Characterization of Freshwater Use in LCA: Modeling Direct Impacts on Human Health Anne-Marie Boulay,*,† Cecile Bulle,† Jean-Baptiste Bayart,‡ Louise Desch^enes,† and Manuele Margni† † ‡
cole Polytechnique de Montreal (QC), P.O. Box 6079, Canada H3C 3A7 CIRAIG, Department of Chemical Engineering, E Veolia Environnement, Paris
bS Supporting Information ABSTRACT: Life cycle assessment (LCA) is a methodology that quantifies potential environmental impacts for comparative purposes in a decision-making context. While potential environmental impacts from pollutant emissions into water are characterized in LCA, impacts from water unavailability are not yet fully quantified. Water use can make the resource unavailable to other users by displacement or quality degradation. A reduction in water availability to human users can potentially affect human health. If financial resources are available, there can be adaptations that may, in turn, shift the environmental burdens to other life cycle stages and impact categories. This paper proposes a model to evaluate these potential impacts in an LCA context. It considers the water that is withdrawn and released, its quality and scarcity in order to evaluate the loss of functionality associated with water uses. Regionalized results are presented for impacts on human health for two modeling approaches regarding affected users, including or not domestic uses, and expressed in disability-adjusted life years (DALY). A consumption and quality based scarcity indicator is also proposed as a midpoint. An illustrative example is presented for the production of corrugated board with different effluents, demonstrating the importance of considering quality, process effluents and the difference between the modeling approaches.
’ INTRODUCTION Vital to life, water is a unique natural resource. While it cannot disappear, it can be made unavailable to specific users (ecosystems, human users, and future generations1) either by displacement or quality degradation. This change in availability can lead to environmental impacts. Life cycle assessment (LCA) is a methodology that quantifies potential environmental impacts for comparative purposes in a decision-making context.2 It is used by governments and industries alike to support their impact reduction strategies. In LCA, the resources consumed and emissions generated by a product or service over its entire life cycle are compiled, characterized, and grouped into different impact categories using formal models. While potential environmental impacts from pollutant emissions are characterized in LCA, impacts from water unavailability are not yet fully quantified. Based on a review of existing methods to characterize water use impacts in LCA, Bayart et al.3 suggested a general framework that considers three main impact pathways leading to water deficits for human uses, ecosystems, and future generations (freshwater depletion). This paper focuses solely on human uses and proposes a method that assesses the consequences of insufficient access to water for human needs. Bayart and colleagues3 distinguish the following human users: domestic, agriculture, industry, r 2011 American Chemical Society
fisheries, hydropower, transport, and recreation. A decrease in water availability for human uses can lead to impacts on human health. If there is sufficient economic wealth in the area, users will adapt to the lack of water by compensating with a backup technology (e.g., desalination, water or goods import, etc.). The impacts of these processes can be assessed in existing impact categories through traditional LCA and included in the results by expanding the product system under assessment (i.e., the system for which water use is being studied). In Bayart et al.’s3 review, only one method addresses the impact pathway leading to impacts on human health. Pfister et al.4 proposed a breakthrough by quantifying in DALY (disability-adjusted life years) the impacts of malnutrition stemming from a lack of water for agriculture and addressing spatial and temporal variations for over 10 000 watersheds. These regionalized impacts include a scarcity parameter that accounts for seasonal variations. More recently, Motoshita et al.5 also proposed a methodology to assess the human health impacts of Received: September 10, 2010 Accepted: September 9, 2011 Revised: September 7, 2011 Published: September 09, 2011 8948
dx.doi.org/10.1021/es1030883 | Environ. Sci. Technol. 2011, 45, 8948–8957
Environmental Science & Technology
ARTICLE
Figure 1. Water use impact pathways for human users leading to compensation or human health impacts.
water scarcity on domestic users. Outside the LCA field, Fry et al.6 assessed the avoided health impacts from increased water availability for domestic uses. While these methods can assess potential impacts for consumptive water use, they do not consider the fact that the users’ adaptation capacity may lead to compensation instead of direct human health impacts. In addition, they do not address the consequences of change in water availability due to a degradative use limiting the potential uses (i.e., functionality) of the resource. Water is considered degraded when it is returned to the body of water with a lower quality, while a consumptive use refers to evaporation, integration within a product or the return of the water to a different watershed or the sea.7 As most industrial and domestic water uses can be considered degradative, it is important to account for the fact that returned polluted water will not provide the same function as clean water, but may still provide some in comparison to consumed water.8 Also, none of the existing methodologies consider in-stream users, such as fisheries, nor do they distinguish between surface water and groundwater use. The objective of this work is to develop a characterization model that assesses the potential impacts generated by a loss of water availability or functionality for human uses caused by consumptive or/and degradative use. These potential environmental impacts are modeled using two distinct and complementary impact pathways: one leading to direct human health impacts (in DALY) caused by malnutrition and disease and the other modeling compensation scenarios to overcome water shortage. This
paper focuses on modeling the first impact pathways and will only discuss the second aspect within a comprehensive LCA perspective.
’ MATERIALS AND METHODS General Framework Description. Figure 1 illustrates the impact pathway from a water use inventory to direct and indirect impacts generated by a deficit for human uses. Boulay et al.8 determined two important concepts needed to identify water deficits for human uses: human users and water categories, the latter defining water of sufficient quality and adequate source to be functional for the former. Human users are identified according to domestic use, agricultural use, fisheries (here referring to catch and aquaculture), industry, cooling, transport, hydropower, and recreational use. Three domestic user categories were created in an effort to account for the different qualities of water used for domestic purposes based on local availability, each quality requiring different levels of treatment. Each user is further categorized as in-stream and off-stream according to Bayart et al.3 and as shown in Figure 1. In this paper, off-stream hydropower was not considered. The water categories represent the 17 possible elementary flows described by source (surface, ground or rain) and water quality (see Supporting Information (SI)). Elementary flows describe the exchanges between the assessed product system and the environment. Water quality is evaluated based on a series of parameters and their thresholds, which determine the users for which a specific water category is functional (sufficient 8949
dx.doi.org/10.1021/es1030883 |Environ. Sci. Technol. 2011, 45, 8948–8957
Environmental Science & Technology
ARTICLE
quality and adequate source). For example, groundwater is generally not functional for transport, hydropower or recreational use, and poor quality water is not functional for clean domestic water users (domestic 1) who do not treat their water before consumption.8 Water quality ranges from 1, excellent (functional for all users) to 5, unusable (only functional for transport or hydropower). When facing water scarcity, users can adapt and compensate the loss of functionality previously provided by the water resource (water treatment, import of water or goods). Alternatively, if the socioeconomic situation is not favorable enough, users will directly suffer from a reduction in available water. This could lead to disease caused by limited access to domestic water (lack of hygiene and access to safe water) or agricultural productivity losses leading to malnutrition.9 Model Description. The model characterizes the impacts associated with the amount of water entering and leaving the product system for a given category. Potential impacts on human health are calculated based on the difference between resource extraction and emission into the environment, as per eq 1. HHimpact ¼
17
17
∑ ðCFi Vi, inÞ i∑¼ 1ðCFi Vi, out Þ i¼1
ð1Þ
Where, HHimpact expresses the human health impacts in DALY, CFi is the characterization factor of water category i for the human health impact category (in DALY/m3 of water category (i) and Vi (in and out) is the volume of water category i entering and leaving the process or product system, i.e. the elementary flows (in m3). Characterization factor CFi includes three main components that can be compared to the three factors traditionally used to define emissions-related impact categories: (1) fate, (2) exposure, and (3) effect. As described in eq 2, they respectively represent (1) local water stress, (2) the extent to which user(s) will be affected by a change in water availability and their ability to adapt to this change, and (3) the human health impacts of a water deficit for user j.
Where αi expresses the water stress index of category i (dimensionless), Ui,j the user(s) j that will be affected by the change in water category i availability (dimensionless), AC the adaptation capacity (dimensionless) and E j the effect factor for user j (DALY/ m3). The following section describes these three components. Fate: Water Stress (αi). In eq 2, the stress index represents the level of competition among users due to the physical stress of the resource, addressing quality and seasonal variations and distinguishing surface and groundwater since these two types of resources often do not present the same scarcity in a same region and may not serve the same users. In this paper, the scarcity parameter α*i for surface water was first calculated based on the CU/Q90 ratio proposed by D€oll.10 A discussion on the underlying choice of this scarcity parameter is reported in the SI. The consumed water (CU) in the numerator was calculated using data from the WaterGap model (obtained from the developers).11 While no seasonal effects were taken into account for renewable groundwater resource availability (GWR), they were considered in the denominator for surface water by the Q90 parameter. The latter, the statistical low flow, represents the flow that is exceeded
9 months out of 10. It is therefore a lower value than the average or median flow and allows for the exclusion of the effect of very high flows (e.g., during monsoon season), since this water is rarely fully available unless extensive storage facilities are also available.12 However, the monthly discharge used for the Q90 assessment accounts for the presence of reservoirs. The scarcity parameter α*i for surface and groundwater is described in eqs 3 and 4. αsurf ace, i ¼ αGW, i ¼
CU ð1 fg Þ 1 Pi Q 90
CU fg 1 Pi GWR
ð3Þ
ð4Þ
Where CU represents the consumptive use in km3/yr, Q90 the statistical low flow in km3/yr, fg the fraction of usage dependent on groundwater (obtained from WaterGap), GWR the renewable groundwater resource available in km3/yr and Pi the proportion of available water that is of category i. It should be noted that the less functional a water category is, the more abundant it will be, since all higher quality categories will also meet the category’s functionality requirements. This is a consequence of water categories being defined by upper thresholds and not by ranges. They are functionality-based, and category 3 would therefore include 2 and 1, and so on. The availability of water for a given category is considered through parameter Pi in eqs 3 and 4, which is the fraction of freshwater of category i or better available in a region. The proportion of each water category per watershed is evaluated based on data describing surface and groundwater quality from GEMStat (13). Since seawater may be considered very poor or unusable (as per water categories 4 or 5, see the SI for a detailed description), scarcity for these categories was considered to be null in regions with access to seawater, consequently considering infinite availability. When no water of a certain quality was available, no scarcity was calculated. The consumptive use was not available for each water category, which would ideally be needed to calculate the α*i specific to each water type. Only total water consumption CU was available. Therefore it was assumed that the best quality water is consumed first, before lower quality water, thus resulting in higher scarcity for better quality categories. Alternatively, one could have assumed that water of different quality was consumed in proportion to its availability, which would lead to discarding the parameter Pi. The first proxy was chosen and this, in some cases, may lead to an overestimation of the physical scarcity of good quality water if lower quality water is consumed first. It would be best to use the consumptive use per water category, if it ever becomes available, instead of either of the two proxies. For the assessment of rainwater use from harvesting or as green water, please refer to the SI. The stress index (αi) is then modeled in order to obtain an indicator ranging from 0 to 1 based on accepted water stress thresholds. Assessments of low, moderate, high and very high water stress are associated with water withdrawals of 10, 20, 40, and 80% of available water, respectively.4,12,13 Correlations between these withdrawals-to-availability ratios and consumption-to-availability ratios were generated (see SI) and used to establish corresponding thresholds of 10, 12, 18, and 40%, respectively. Because the stress index (αi) is meant to reflect the competition between users, the αi is set to result in 0 for low water stress (meaning that the consumption of 1 m3 of water will 8950
dx.doi.org/10.1021/es1030883 |Environ. Sci. Technol. 2011, 45, 8948–8957
Environmental Science & Technology
ARTICLE
Figure 2. Adaptation capacity based on World Bank country classification:16 No adaptation for low-income countries, complete (100%) adaptation for high-income countries and partial adaptation for middle-income countries.
not affect other users when water is abundant) and up to 1 for very high stress (meaning that each consumed 1 m3 will deprive other competing users of 1 m3). The in-between data were fit in an S-curve passing through 50% scarcity when the high-stress threshold is reached, as also proposed by Pfister et al.4 and shown in SI equation S1. MidpointWater Stress Indicator (WSI). The proposed midpoint uses the availability of water and differentiates source and quality by weighing the stress of each water type. The results of each flow are aggregated in eq 5 in the same way as in eq 1 for the end point, here using the water stress αi as a midpoint CF. WSI ¼
∑i ðαi Vi, in Þ ∑i ðαi Vi, out Þ
ð5Þ
Where WSI expresses the midpoint result in m3 equivalent of water, αi the stress index of water category i (in m3 of water equivalent per m3 of water of category i withdrawn/released) and Vi (in and out) the volumes of water category i entering and leaving the process or product system (i.e., elementary flows (in m3)). It represents the equivalent amount of water of which other competing users are deprived as a consequence of water use. Affected User(s) (Ui,j). A change in water i availability will not affect all users to the same extent. The impacts depend on (1) the water’s functionality for a specific user j, Fi,j (based on its quality and type of water resource), and (2) the identification of the user(s) most likely to be affected by a change in water availability in the area of interest (Uj). Together, these parameters make it possible to assess the extent to which user j will be affected by a change in availability of water category i, as shown in eqs 6 and 7. Ui, j ¼
Uj Fi, j where j is an offstream user ðUj Fi, j Þ
∑ j, of f stream
Ui, j ¼ Uj Fi, j where j is an instream user
ð6Þ
ð7Þ
Where Ui,j represents the proportion in which user j is affected by a change in water availability for category i (dimensionless), Uj the proportion in which user j is affected by a change in water availability (dimensionless) and Fi,j the functionality of water category i for user j (dimensionless). Functionality, Fi,j. Water categories presented in Boulay et al.8 are related to users through a binary functionality parameter (1 or 0) that reflects whether or not the water category is functional for a given user. The functionality Fi,j, is based on the potential use of water i by user j without any additional treatment. User’s Identification Uj. For offstream users, this parameter represents the user(s) that will be affected by a change in water availability (i.e., the one from which this additional water will be taken). Such users are referred to as marginal users. It is still debated which is the marginal user(s); different approaches have been suggested4,5 and mainly differ on the inclusion or exclusion of domestic users as users potentially affected by a decrease in water availability. A UNESCO report states that “unbridled competition from richer farmers and industrial concerns for water, productive land and fisheries, often put the poor at a serious disadvantage. It is also often very difficult for the poor to assert their rights and needs so as to receive a fair entitlement to public goods and services”.9 However, there is a debate4 as to whether an additional water use as typically assessed in LCA would indeed deprive domestic users or whether the change in water availability would be absorbed by the agricultural sector, which would be considered the sole marginal user due to its lower willingness to pay. In reality, there is not enough information to provide robust evidence of such a choice. To satisfy both ends of the debate, two approaches were chosen. The first, the distribution hypothesis, implies that all users are likely to be affected proportionally to their use. A regional distribution of water withdrawals per user was used to represent the probability of being the affected user. In the second approach, the marginal user hypothesis, agriculture is considered to be the only off-stream user that is affected. Therefore 100% of the water use will affect 8951
dx.doi.org/10.1021/es1030883 |Environ. Sci. Technol. 2011, 45, 8948–8957
Environmental Science & Technology
ARTICLE
Figure 3. Regional water stress index αS2a, based on the ratio of consumptive use over renewable available resource, including local water quality data and modeled based on accepted stress thresholds.
agriculture and no human health impacts will be generated from water deprivation for domestic uses. Results for both scenarios are given, leaving the final choice up to the LCA practitioner depending on the information available at the local level. For in-stream users, a change in availability of 1 m3 of water in a country will deprive each of the in-stream users proportionally to the intensity at which they use the surface water bodies. For human health impacts, this only concerns fisheries. The details of the calculation and hypotheses for both off-stream and in-stream users can be found in SI. Adaptation Capacity (AC). The adaptation capacity defines whether the change in water availability will create deficit or compensation scenarios. The World Bank gross national income (GNI) classification14 was chosen as the socioeconomic parameter to indicate a country’s adaptation capacity (AC) (see Figure 2). Its correlation with access to an improved water source or improved sanitation was reported by the United Nations and reproduced in the SI.15 It is proposed that low-income countries (GNI< $936/cap) will not be able to adapt to a change in water availability and will therefore suffer water deficits, whereas high-income countries (GNI $11 455/cap) will have the means to fully compensate for this type of change. Middle-income countries ($936/cap < GNI < 11 455/cap) are attributed an adaptation capacity proportional to their incomes, meaning that, in these countries, both compensation and deficit occur. This relation is shown in Figure 2 and described in eq 8. AC ¼ ð9:5105 GNIÞ 8:9102 f or 936
$ $ < GNI < 11455 cap cap
ð8Þ Effect FactoR (EJ). Effect factor Ej assesses the importance of human health impacts caused by a water deficit for user j for 5 of the 10 users: domestic (1, 2, and 3), agriculture and fisheries. If a water deficit occurs for the remaining users (transport, hydro, industry, cooling, and recreation), impacts will only be generated through a compensation process when occurring in countries able to compensate. This is reflected by the Ej zero value for these users.
For agriculture and fisheries, it is generally accepted that a lack of water would result in malnutrition health impacts due to lower food availability.4,17 The effect factors (DALY/m3) were determined by first assessing the health impacts generated by malnutrition in DALY/kcal and dividing this value by the amount of water needed to produce one kcal, either from agriculture or fisheries. For domestic use, the effect factor (DALY/m3) relates the human health impacts associated with a lack of hygiene and sanitation when water is scarce to the water deficit for domestic use. It is calculated by dividing the ratio of health burdens from water-related hygiene and sanitation issues by the actual volume of water in deficit for domestic uses (based on a value of 50 L/ cap/day to ensure low health concerns and cover most basic needs).18 The resulting effect factors are 6.53 105, 2.02 105, and 3.11 103 DALY/m3 for agriculture, fisheries, and domestic, respectively. A domestic use deficit is therefore critical, since it shows health impacts that are 2 orders of magnitude greater than those for agriculture or fisheries. The details on how these parameters were obtained are presented in the SI.
’ RESULTS Water Stress. Along with the elementary flows, the water stress indexes (αi) are proposed to calculate the water stress indicator (WSI) at the midpoint level. Figure 3 shows the water stress index for good quality surface water (α2a). Whereas high scarcity is found where expected, in addition, some major watersheds in North America do not have good quality water19 and these quality data are also used as default for the northern watersheds when primary data is lacking. Assuming water quality data are representative of the entire region, no water of this quality would be present in that particular region, hence no assessment of its use would be needed. See the SI for all water category indexes. Human Health Impacts. Figure 4 shows the results of the direct potential impacts on human health (in DALY/m3) from the use of 1 m3 of good-quality surface water (S2a). The CFs in Figure 4a are based on the hypothesis that several users are affected proportionally to their use and therefore include impacts 8952
dx.doi.org/10.1021/es1030883 |Environ. Sci. Technol. 2011, 45, 8948–8957
Environmental Science & Technology
ARTICLE
Figure 4. Human health characterization factors for good quality surface water (category S2a) in DALY/m3 (a) considering that agriculture is the only marginal user affected (along with fisheries), leading therefore to malnutrition and (b) considering all users are affected proportionally to their use (i.e., also including domestic users and therefore considering health burdens due to lack of hygiene and sanitation in addition to malnutrition).
brought about by a lack of hygiene and sanitation related to a water deficit for domestic use and impacts generated by malnutrition from both agriculture and fisheries. The CFs in Figure 4b refer to health impacts generated only by malnutrition from both agriculture and fisheries. As expected, high-income areas such as North America, Europe, and Australia show no direct impacts on human health because they have maximum adaptation capacities. They would, however, generate potential impacts from compensation. CFs for other water types are detailed in the SI. Illustrative Example. A paper mill producing cardboard from recycled fibres was studied as an illustrative example using the generic ecoinvent data set Corrugated board, recycling fiber, single wall, at plant characterized with the Impact 2002+ methodology.20 The generic data set was adapted with available primary industry data from Cascades. For each ton of corrugated board produced, 17.4 m3 of water are withdrawn from a nearby river and 16.4 m3 are released into the same river (average pulp and paper plant data). The effluent is categorized as average quality
water (category 2b, see Boulay et al.8 and SI), while the influent is considered to be good quality surface water (S2a). This scenario B (average effluent S2b) is shown along with two other hypothetical variations: A. the effluent is better treated, up to the good quality surface water of the influent (well-treated effluent, S2a) and C. the effluent is considered null (i.e., 100% of the water is consumed). The purpose of this sensitivity analysis is to assess the variability of the impacts associated with the water that is consumed or used but released at different quality levels. The magnitude of the difference between both hypotheses on the marginal user affected is also illustrated by this example, as discussed below. Impacts on human health are shown for the hypothetical production of cardboard in the Ganges Basin in India. This region was chosen because it shows a low adaptation capacity and high water scarcity. In this respect, adaptation will still occur (AC 6¼ 0), implying that both the direct impacts and impacts generated by compensation scenarios should be addressed. Only the direct 8953
dx.doi.org/10.1021/es1030883 |Environ. Sci. Technol. 2011, 45, 8948–8957
Environmental Science & Technology
ARTICLE
Figure 5. Human health impacts of water use for board production in comparison with other human health contributors from the production process for three different effluent scenarios and for both hypotheses on marginal use: (1) considering all users to be affected proportionally to their us (i.e., including domestic users) and (2) considering that agriculture is the only marginal user affected (along with fisheries), leading therefore to malnutrition (not including domestic use).
impacts are shown in Figure 5. The 10 main contributors (elementary flows) to human health impacts are identified, including water use, which is split into impacts generated by malnutrition and by diseases related to water deficit for domestic use. All three scenarios (A, B, and C) are presented for both hypotheses: (1) Distribution: all users are affected proportionally to their use (i.e., considering health burdens due to the lack of hygiene and sanitation and malnutrition) or (2) Marginal: reporting that only malnutrition impacts occurs. The reduced quality of the effluent in B may significantly contribute to the overall human health damages. The additional impacts from a degradative use (in 1B) are, in this case, even higher than all other human health contributors. However, considering that all water must be consumed (1C) instead of considering the polluted released flow (1B) would more than double the impacts on human health from water use. In addition to these results, compensation scenarios for each user should be modeled with a traditional LCA method to yield impacts scores that would then allow for the integration of indirect impacts into other impact categories.
’ DISCUSSION This methodology modeled the impacts generated by a change in water availability for human uses in an LCA perspective. The water stress parameter α was proposed to calculate a midpoint indicator (WSI) that could be used to characterize physical inventory flows into a common metric, as suggested by
Frischknecht et al.,Pfister et al. and Mila-i-Canals et al.21,4,1 In this respect, α can be considered representative for three areas of protection, including human health, ecosystem quality and resource depletion. However, it is important to note that no ideal midpoint has been found for the impact pathway to allow for a partial characterization that could then be used directly as an input for end point modeling in all categories, as presented in Bayart et al.3 This is due to the fact that each impact pathway involves different mechanisms. Here, the water stress parameter α corresponds to a fate factor (see eq 2). The human health impact pathway was further modeled up to the end point level by developing CFs expressed in DALY/m3. Additionally, indirect impacts must be considered through the compensation scenario by identifying the marginal technology that is appropriate for each water type, as proposed by Weidema et al.22 These indirect impacts could be consistently compared with the direct impacts by further modeling a new LCI (by system expansion) and assessing the related impacts on human health and other categories generated by the compensation scenario, allowing for a coherent comparison with other sources of impacts. The application of this method to a straightforward example illustrates the potential significance of considering water use impacts in an LCA carried out for a region with low adaptation capacity. However, further case studies, including higher income countries, are needed to further evaluate this approach. Moreover, regionalization was shown to be a critical issue since impacts can be dominant or nonexistent from one region to another. 8954
dx.doi.org/10.1021/es1030883 |Environ. Sci. Technol. 2011, 45, 8948–8957
Environmental Science & Technology While other methods have been advanced to assess water use impacts in LCA, the methodology herein differs in the following aspects: Quality. The model does not only consider the consequences of consumptive use leading to reduced access to water. It also assesses the consequences of a loss of functionality for downstream users due to the degradative use of water. Seeing as the inventory procedure integrates the functionality of withdrawn and released water based on quality, the released water and its corresponding functionalities are considered to be returned to the environment, avoiding an overestimation of the potential impacts by considering that the water was consumed. This is especially important for heavy water users (e.g., thermal plants for cooling). Assessing the quality of the returned water would also serve as an incentive for effluent treatment. In the aforementioned example, if the plant were to treat its effluents to attain the same water quality as the influent (2a), water use impacts would be reduced by a factor of 6 as compared to the water released as 2b. Finally, including quality in the methodology also curbs the impacts of using low-quality water vs highquality water since not all users will be affected the same way. Stress. Unlike other widely used scarcity indicators, the one proposed here considers the ratio of consumptive use over a statistical low-flow parameter. The numerator allows for a better representation of the physical scarcity of water without considering the withdrawn water that is released in the same watershed (e.g., for cooling). The denominator takes seasonal variations into account, which is very important in countries that face monsoons and droughts. Moreover, the indicator distinguishes surface water, groundwater, rainwater, and quality, which is significant since the water categories are not available in the same quantities and not functional for all users. A more specific scarcity parameter per water type could be obtained by adjusting the CU term to reflect the actual consumption of each water quality category. However, the regional breakdown of the volume of water consumed for each specific water category is not available. The parameter therefore currently assesses the scarcity of a water category based on its availability and assumes that water of good quality will be consumed before a lower water quality. Users. All users are considered, and some are even differentiated based on the quality of water they require. While not all suffer human health impacts, it would be especially important to consider them in a compensation perspective, since industry and the energy business in developed countries are unlikely to stop their activities due to water shortage without first considering compensation strategies. Human Health. Although a first attempt to model human health impacts for malnutrition has already been proposed by Pfister and colleagues4 with a statistical correlation of R2 = 0.26, in this paper, a probabilistic approach was proposed instead to evaluate the health burdens associated with malnutrition and with water shortages for domestic use. Both show acceptability in terms of the distributionlog-normal, which allowed for the use of geometric averages. The p-values describing the distribution of the correlation are of 0.0839 and 0.15, respectively. A p-value higher than 0.05 reflects an acceptable distribution. Malnutrition was modeled for a water shortage affecting agriculture, fisheries and livestock (through agricultural feed). With regards to domestic use, while infrastructure plays an important role in the health burden caused by water-related diseases, this role did not need to be considered since the impacts of a change in water availability in current socioeconomic conditions were modeled. The health burden for 1 m3 of water that is unavailable for domestic use (whether from lack of access or scarcity) was assessed
ARTICLE
from the correlation between the actual water used for domestic purposes, the water required to avoid health issues and the actual health burden of water-related issues. This rationale is based on strong epidemiological evidence that access to water is correlated with diarrheal diseases, worm infections, and malnutrition.15 However, whether or not access to water for domestic users will be reduced from typical inventory data in LCA, and, consequently, the inclusion or exclusion of the associated impacts, are still open for debate. Pfister and colleagues4 chose to exclude them, but Motoshita and colleagues5 developed a model that suggests they should be included. When addressing the impact pathway, it is important to understand whether a water use in a low-income, water-stressed region would indeed lower water availability for domestic users or rather only affect other users (e.g., agricultural or industries). At this stage, the information to support the choice and determine the marginal affected user is insufficient. For this reason, we chose to present CF for two main scenarios: (1) all users are affected proportionally to their water use in a given region, implying that the burden caused by a lack of hygiene and sanitation is considered in addition to malnutrition, and (2) only agriculture is affected among the off-stream users as it would present the lowest willingness to pay. As demonstrated in the example, human health impacts from water use may be of the same order of magnitude as the human health impacts generated by toxic emissions, thus supporting the conclusion that including or excluding this impact pathway has a relevant influence in a LCA. Compensation. Compensation accounts for the capacity of human users to adapt to a freshwater deficit through the use of technology. Model results indicate that direct human health impacts are expected to be null in developed countries, but indirect impacts may be generated by compensation scenarios. In this respect, the use of backup technologies is not only meant to alleviate depletion, as proposed by Pfister,4 but also to compensate when local resources are not necessarily being depleted but are still used to the extent that they reduce availability for users due to competition. Marginal technology use should therefore be included as having potential impacts within LCA and not solely as a depletion metrics. The volume of water needed would be obtained through the marginal technology in place to compensate for a specific water category. Impacts from this technology should be included in the water use impact assessment by system boundary expansion. However, the assessment should be done consistently with the method proposed here. This implies that only a fraction of the volume, based on the adaptation capacity, is produced from the marginal technology for the withdrawn water category. The identification of the regional marginal technology for each water type should be performed and its impacts included in the scope of the assessment. Limitations. Though it demonstrates spatially differentiated capabilities, the model does not predict absolute or real impacts. It addresses the potential environmental impacts used to characterize environmental interventions within an LCA with an underlying hypothesis of linearity between a change in water availability and the resulting impacts. The default characterization factors in this model may be well-suited to exploring the potential impacts of water use but the robustness of the results and required level of detail can only be evaluated once the model is used and the results are compared with those from other models. Obtaining better data on water quality per watershed or region is necessary to increase the accuracy of the methodology. Current data is not adequately distributed in time or space, and the parameters and uncertainty of the model have yet to be evaluated. Several specific limitations apply to this model. Interactions between 8955
dx.doi.org/10.1021/es1030883 |Environ. Sci. Technol. 2011, 45, 8948–8957
Environmental Science & Technology surface and groundwater or between watersheds are not considered and the consecutive use of the water resource is not taken into account, a fact that could lead to an underestimation of potential impacts. These limitations also apply to (downstream) transboundary watersheds. However, the results could be corrected on a case-by-case basis by identifying the country in which the impacts are most likely to occur for a specific water use or by adding the potential impacts generated by water use in downstream watersheds. Also, user and water availability distribution within a watershed is considered to be homogeneous, implying that there is no difference if the water is withdrawn upstream or downstream of other users. However, it is unlikely that sufficient data would be available on the exact location of the water use to allow for such distinctions or modeling. Moreover, as a modeling choice, the quality of the effluent was taken as is and therefore does not account for dilution or degradation time. While this option is believed to be the best choice since dilution would require the availability of more water, thus creating a loop effect, it may become a limitation in cases in which the natural buffering capacity of the environment would assimilate some of the pollutants. Lastly, limitations associated with water categories, as described in Boulay et al.8 may lead to either an overestimation or underestimation of the impacts by grouping different user functionality requirements into a limited set of water categories. Boulay et al. propose a more detailed functionality-based approach to overcome this. To ensure the integration of the method within daily LCA practices, life cycle inventory databases must be expanded to account for released water volume and therefore support the calculation of water categories. The inclusion of impacts from compensation scenarios may be facilitated by adapting a life cycle inventory database to a consequential approach by including the use of the local marginal source of water, which may come from desalination, reuse, etc. As for other regionalized impact categories, data sets should allow for region selection and appropriate CFs. To facilitate the use of these CFs, a generic data set of effluent water quality by industry type could be generated. Moreover, a water mix similar to a grid mix could be set out based on the local surface/ groundwater consumption data and water quality data that could be used when actual inventory input data is not known.
’ ASSOCIATED CONTENT
bS
Supporting Information. The SI is made up of PDF and Excel documents. The PDF describes the water categories and provides information on water scarcity, user distribution and effect factors, maps and a detailed example of a tailored CF. The Excel document contains all intermediate data, as well as CF for human health at the local (cell) and national scales. One of the tabs is designed to be used as a tool to calculate a tailored CF (Spec CF). A Google Earth layer and other materials are available for download at: www.ciraig.org/wateruseimpacts. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT We thank Ms. Genevieve Martineau (CIRAIG) for her contribution to data collection, Prof. Miguel Chagnon (Universite de
ARTICLE
Montreal) for his support during the statistical analysis, Prof. Petra Doll for her input and WaterGap data, Ms. Kelly Hodgson of GEMStat for providing quality data, and Prof. Ralph K. Rosenbaum for his support and advice during the redaction. The authors also wish to thank the reviewer for the valuable input that led to the significant improvement of this article. We acknowledge the financial support of the industrial partners in the International Chair in Life Cycle Assessment (a research unit of CIRAIG): Alcan, ArcelorMittal, Bell Canada, Cascades, Eco BEC, Groupe EDF/GDFEntreprises Quebec, RECYC-QUE SUEZ, Hydro-Quebec, Johnson & Johnson, Mouvement des caisses Desjardins, RONA, Total, and Veolia Environnement.
’ REFERENCES (1) Mila i Canals, L.; Chenoweth, J.; Chapagain, A. K.; Orr, S.; Anton, A.; Clift, R. Assessing freshwater use impacts in LCA: Part I— inventory modelling and characterisation factors for the main impact pathways. Int. J. Life Cycle Assess. 2009, 14, 28–42. (2) Hauschild, M. Z. Assessing environmental impacts in a life-cycle perspective. Environ. Sci. Technol. 2005, 39, 81A–88A. (3) Bayart, J.-B.; Margni, M.; Bulle, C.; Desch^enes, L.; Pfister, S.; Koehler, A.; Vince, F. Framework for assessment of off-stream freshwater use within LCA. Int. J. Life Cycle Assess. 2010, 15, 439. (4) Pfister, S.; Koehler, A.; Hellweg, S. Assessing the environmental impacts of freshwater consumption in LCA. Environ. Sci. Technol. 2009, 43, 4098–4104. (5) Motoshita, M.; Itsubo, N.; Inaba, A. Development of impact factors on damage to health by infectious diseases caused by domestic water scarcity. Int. J. Life Cycle Assess. 2010, 16, 65–73. (6) Fry, L. M.; Cowden, J. R.; Watkins, D. W.; Clasen, T.; Mihelcic, J. R. Quantifying health improvements from water quantity enhancement: An engineering perspective applied to rainwater harvesting in West Africa. Environ. Sci. Technol. 2010, 44, 9535–9541. (7) Falkenmark, M. Competing freshwater and ecological services in the river basin perspective: An expanded conceptual framework. Water International 2000, 25, 6. (8) Boulay, A.-M.; Bouchard, C.; Bulle, C.; Desch^enes, L.; Margni, M. Categorizing water for LCA inventory. Int. J. Life Cycle Assess. 2010submitted. (9) World Water Assessment Program “The United Nations World Water Development Report 2: Water, a shared responsibility,” 2006. (10) D€oll, P. Vulnerability to the impact of climate change on renewable groundwater resources: a global-scale assessment. Environ. Res. Lett. 2009, 4, 12. (11) Alcamo, J.; Doll, P.; Henrichs, T.; Kaspar, F.; Lehner, B.; Rosch, T.; Siebert, S. Development and testing of the WaterGAP 2 global model of water use and availability. Hydrol. Sci. J. 2003, 48, 317–337. (12) Alcamo, J.; Henrichs, T.; Rosch, T. World Water in 2025— Global Modeling and Scenario Analysis for the World Commission on Water for the 21st Century; Center in Environnemental Systems Research University of Kassel, 2000. (13) Vorosmarty, C. J.; Green, P.; Salisbury, J.; Lammers, R. B. Global water resources: Vulnerability from climate change and population growth. Science 2000, 289, 284–288. (14) Data and Statistics—Country Classification; World Bank, 2008 (15) The United Nations World Water Development Report 3: Water in a Changing World; World Water Assessment Programme, 2009. (16) UNEP. Country income groups (World Bank classification). 2009.http://maps.grida.no/go/graphic/country-income-groups-worldbank-classification (accessed March 1, 2001). (17) The United Nation Water Development Program “Water for People, Water for Life; UNESCO, 2003. (18) Howard, G.; Bartram, J. Domestic Water Quantity, Service Level and Health; World Health Organization, 2003. (19) United Nations Environment Programme Global Environment Monitoring System (GEMS) Water Programme. GEMStat. 2009 8956
dx.doi.org/10.1021/es1030883 |Environ. Sci. Technol. 2011, 45, 8948–8957
Environmental Science & Technology
ARTICLE
(20) Jolliet, O.; Margni, M.; Charles, R.; Humbert, S.; Payet, J.; Rebitzer, G.; Rosenbaum, R. IMPACT 2002+: A new life cycle impact assessment methodology. Int. J. Life Cycle Assess. 2003, 8, 324–330. (21) Frischknecht, R.; Steiner, R.; Braunschweig, A.; Egli, N.; Hildesheimer, G. Swiss Ecological Scarcity Method, The New Version 2006, 2008. (22) Weidema, B.; Frees, N.; Nielsen, A.-M. Marginal production technologies for life cycle inventories. Int. J. Life Cycle Assess. 1999, 4, 48–56.
8957
dx.doi.org/10.1021/es1030883 |Environ. Sci. Technol. 2011, 45, 8948–8957
ARTICLE pubs.acs.org/est
Toxicity and Reductions in Intracellular Calcium Levels Following Uptake of a Tetracycline Antibiotic in Arabidopsis Shaun M. Bowman,† Kathryn E. Drzewiecki,† Elmer-Rico E. Mojica,‡ Amy M. Zielinski,† Alan Siegel,† Diana S. Aga,‡ and James O. Berry*,† † ‡
Department of Biological Sciences, University at Buffalo, Buffalo, New York 14260, United States Chemistry Department, University at Buffalo, Buffalo, New York 14260, United States
bS Supporting Information ABSTRACT: Plant responses to natural stresses have been the focus of numerous studies; however less is known about plant responses to artificial (i.e., man-made) stress. Chlortetracycline (CTC) is widely used in agriculture and becomes an environmental contaminant when introduced into soil from manure used as fertilizer. We show here that in the model plant Arabidopsis (Arabidopsis thaliana), root uptake of CTC leads to toxicity, with growth reductions and other effects. Analysis of protein accumulation and in vivo synthesis revealed numerous changes in soluble and membrane-associated proteins in leaves and roots. Many representative proteins associated with different cellular processes and compartments showed little or no change in response to CTC. However, differences in accumulation and synthesis of NAD-malic enzyme in leaves versus roots suggest potential CTC-associated effects on metabolic respiration may vary in different tissues. Fluorescence resonance energy transfer (FRET) analysis indicated reduced levels of intracellular calcium are associated with CTC uptake and toxicity. These findings support a model in which CTC uptake through roots leads to reductions in levels of intracellular calcium due to chelation. In turn, changes in overall patterns and levels of protein synthesis and accumulation due to reduced calcium ultimately lead to growth reductions and other toxicity effects.
’ INTRODUCTION Veterinary antibiotics are introduced into the environment via cropland application of animal wastes.1 5 These compounds accumulate near animal feeding operations in soil,6,7 groundwater,8 and aquatic environments.8,9 Tetracycline (TC) antibiotics are the most heavily used in the United States, amounting to about 38% of the approximately 27.85 million pounds of antibiotic sold between 2005 and 2007.10 These are highly stable, and accumulation rates in environments receiving yearly applications can exceed the degradation rate,11 increasing their concentration over time. The widespread use of tetracycline antibiotics in agriculture has led to serious concerns about their environmental impact.2,3,5,12 For manure-treated soil in agricultural fields, chlortetracycline (CTC) has been detected at levels as high as 20 mg/kg.7 Concentrations within agricultural lagoons have been shown to be as high as 1000 ug/kg. An example of the distribution potential from agricultural sites was demonstrated in a recent study9 which measured TC concentrations along the length of a Colorado river. While no TCs were detectable in a mountain location upstream of urban or agricultural sites, significant levels of TCs were found in water from sampling locations near or downstream of agricultural areas. The promotion of resistance to TCs and other antibiotics in microorganisms is of major concern and a primary focus of study.13 16 However, few studies have examined the effects of antibiotics on agricultural plants maintained within these r 2011 American Chemical Society
environments. Several plants have been shown to take up CTC from contaminated soil.17 21 Some plants such as pinto beans show significant toxicity effects when grown in the presence of CTC, while others, such as Zea mays (maize) do not.18 22 CTCinsensitive maize plants induce Glutathione s-transferase (GST) when grown in the presence of CTC, and show increased GST activity after CTC uptake. In contrast, CTC-sensitive pinto beans show no increase in GST activity following CTC uptake. CTC is an efficient calcium antagonist,23 25 and our recent findings provide evidence that intracellular calcium chelation by CTC taken up through the roots of pinto bean plants may contribute to toxicity.21 Taken together, these studies support the hypothesis that GST enzymes are induced in response to CTC exposure to detoxify the antibiotic in resistant plants, preventing accumulation of the antibiotic, thereby not enabling calcium chelation. In contrast, the lack of GST detoxification in sensitive plants allows the antibiotic to build up within cells, allowing chelation-related toxicity effects. We show here that Arabidopsis is highly sensitive to CTC exposure. CTC uptake by Arabidopsis roots causes changes throughout the entire plant, including reduced growth rates, Received: March 15, 2011 Accepted: September 1, 2011 Revised: August 29, 2011 Published: September 01, 2011 8958
dx.doi.org/10.1021/es200863j | Environ. Sci. Technol. 2011, 45, 8958–8964
Environmental Science & Technology changes in protein synthesis/accumulation, and changes in intracellular calcium.
’ EXPERIMENTAL SECTION Plant Material, Growth Conditions, Harvesting, and Treatments. Arabidopsis Col-0 seed was obtained from Lehle Seeds.
Arabidopsis expressing the yellow cameleon YC3 calcium sensor protein was obtained from Simon Gilroy (Dept. of Botany, University of Wisconsin, Madison, WI). Plants were germinated and grown in artificial soil in a growth chamber at 24 °C, with 14 h/d illumination at approximately 170 μmol photons m 2 s 2. The pH of treated and untreated soils was approximately 5.5. For soil treatment, plants were grown in untreated soil for seven days, and then watered daily with 20 mg/L CTC (chlortetracycline hydrochloride, Sigma), or water alone (control), and harvested at times indicated in the figures. CTC solutions were prepared fresh immediately before each application. Plants were carefully removed from soil, washed under slow running water to remove soil from roots, and blotted on paper towels prior to weighing. The entire plant was weighed first, and then cut with a razor blade at a position corresponding to the soil line to obtain separate weights for shoots and roots. To compare toxicity during hydroponic treatment, 3-week-old Arabidopsis or 2-week-old maize plants were excised from soil and placed into foil-wrapped containers with water containing freshly prepared 20 mg/L CTC, or water alone, for 24 h. Analysis of Protein Accumulation and Synthesis. Soluble or membrane-bound protein extracts were prepared as described.27,28 Equal amounts of protein were loaded into each lane of an SDS-PAGE gel, electrophoresed, and either silver-stained or transferred to nitrocellulose for immunoblotting. Antisera for Rubisco LSU and SSU, PEPCase, and NAD-Me have been described.28,29 Antisera against plant CoxII was obtained from Agrisera (V€ann€as, Sweden). Antibody reactions were detected using Amersham ABC luminol reagent system, and visualized using a Storm phosphorimager and ImageQuant software (GE Healthcare). In vivo protein synthesis was determined by radioactively labeling Arabidopsis shoots or roots after 12 days growth in CTCtreated or control soil.27,28 For leaves, stems were cut below the lowest leaf under water and placed into a solution containing 100 μCi of [35S]Met/Cys express labeling mix (PerkinElmer NEN Radiochemicals) in 400 μL of water. Roots were washed in water and cut just below the soil line. The cut end was placed into the [35S]-labeling solution; the labeling container was wrapped in foil to achieve darkness. After two hours, proteins were extracted from equal wet weight of material.27,28 Tricarboxylic acid (TCA) precipitation of proteins from each extract was performed as described.27 Protein extracts were used directly for analysis of total protein synthesis by SDS-PAGE, or immunoprecipitated from equal amounts of labeled protein extracts.27 Fluorescence Measurements and LC/MS Analysis. Uptake of TC and CTC was analyzed as described.21 Briefly, Arabidopsis grown in untreated soil for 18 21 days were removed from soil, washed, and roots were submerged in 0.25 mM TC or CTC solution, or water, for 24 h. Leaves were harvested and prepared as described.21 Fluorescence emission spectra of CTC and TC in Arabidopsis extracts and standards were recorded using an SLM model 8100 spectrofluorimeter equipped with a 450 W Xe arc lamp source. Three readings were obtained for each sample and
ARTICLE
standard. Extracts were also analyzed by LC/MS using an LCQ Advantage ion trap mass spectrometer connected to a Surveyor LC system (Thermo Finnigan, San Jose, CA, USA), using electrospray ionization in positive mode. A reversed-phase Thermo Scientific (Fullerton, CA, USA) BetaBasic C18-column (2.1 100 mm, 3 μm particle size) was used for this separation. Details of mobile phase conditions and instrument parameters were described previously.21 Confocal Imaging and Fluorescence Resonance Energy Transfer (FRET) Analysis. FRET analysis was performed using Arabidopsis expressing the YC3.6 cameleon protein.30,31 To maximize uptake of chelation reagents and minimize variation due to soil binding, evaporation, etc., plants grown for 9 or 11 days in untreated soil were transferred to hydroponic growth in water containing freshly prepared 20 mg/L CTC, 5 mM EGTA, or water alone (control), for 24 h, as shown in Supporting Information 6 and 7. Intact plants were then placed onto slides to which had been attached adhesive silicone isolators (19 mm 32 mm 0.5 mm depth, Grace Biolabs, Bend, OR). Plants were first rinsed in H2O, and then placed within the isolator’s chamber, submerged in water, and then covered with a coverslip. Plants were imaged within 30 min. Images were captured from the midsections of the young primary roots; tip and base regions were not imaged. Fluorescence resonance energy transfer (FRET) analysis of the YC3.6-expressing roots was performed according to methods described in ref 30, using an LSM710-InTune Confocal (inverted) Microscope System (Carl Zeiss MicroImaging), with a 20 objective. CFP excitation was done using a 458-nm argon laser line. CFP fluorescence was collected between 472 and 505 nm, and YFP FRET fluorescence was collected between 526 and 536 nm. Combined YFP/CFP images were analyzed for relative amounts of YFP and CFP fluorescence using ImageJ software (http://rsb.info.nih.gov/ij/). Six independent plants were used for each of the three treatments shown in Figure 3, panel D (for a total of 18 independent plants). Three separate images were obtained and analyzed for each of the independently treated plants (thus, a total of 54 independent images were analyzed for YFP/CFP ratios).
’ RESULTS AND DISCUSSION CTC Toxicity in Arabidopsis. For this study plants were exposed to the highest levels detected in contaminated field soils, 20 mg/kg.7 Treated plants began to show toxicity effects after 3 4 days of CTC exposure, including stunted growth, reduced leaf expansion, and yellowing. These effects were not observed in control (untreated) plants. Figure 1 (top two panels) shows a comparison of CTC-treated versus untreated plants after 9 days of CTC watering. The CTC-treated Arabidopsis clearly showed toxicity effects (left panel), with reduced overall weights (right panel) when compared to control plants. For these young plants, separation of shoots and roots was difficult, and we were unable to obtain separate data for these tissues. Figure 1 (bottom panel), and Supporting Information 1, show CTC effects in older Arabidopsis after 13 14 days of CTC watering. For these older plants, separation of shoots and roots was amenable, and accurate wet weight determinations for each tissue were obtained. It is evident that CTC exposure reduced the growth of both shoot and roots. Growth was reduced more than 2.5 fold for Arabidopsis treated with CTC for 9 14 days. Differences between CTC-treated and 8959
dx.doi.org/10.1021/es200863j |Environ. Sci. Technol. 2011, 45, 8958–8964
Environmental Science & Technology
Figure 1. Arabidopsis growth in response to CTC exposure. Col-0 Arabidopsis were germinated and grown in untreated soil for 7 days, and then watered with 20 mg/L CTC, or water (control). Wet weight in grams was recorded; for both graphs standard error is shown. Top left panel: representative soil-grown plants treated with CTC (left row), and control plants (right row) 9 days after initiating CTC watering. Top right panel: graph of total wet weight in grams for 9 day CTC-treated and control plants. Values were based on 10 treated and 11 control plants. Bottom panel: wet weights determined for entire plant, shoots, and roots of treated and untreated plants. Weight wet data were recorded after 13 days of CTC watering, and combined with data for another population after 14 days of CTC watering. Weight determinations were first made for each entire plant, and then for separated shoots and roots. Weights were recorded for 19 treated and 23 control Arabidopsis.
control plants were stronger than those previously reported for CTC-sensitive pinto beans.22,32 With regard to concentrations used in those studies, it was noted that there was considerable adsorption of the antibiotic to some soils, so that effective amounts of CTC taken up by those plants may have been much lower. Taken together, the findings presented here indicate that CTC exposure affects the growth of the entire Arabidopsis plants, and that toxicity effects do not lessen as exposure time to CTC-treated soil increases. In addition, GST activity studies performed using protocols and reagents described in ref 18 with protein extracts from treated and untreated Arabidopsis plants showed no increase in GST activity in response to CTC exposure (not shown). This indicates that like pinto beans, CTC-sensitive Arabidopsis does not increase overall GST activity in response to CTC exposure. Additional experiments, using antisera and probes specific for individual GST isoforms, will be used in future experiments to investigate in more detail the role of GST induction in response to antibiotic exposure in sensitive and resistant plants. Uptake of Tetracycline Antibiotics by Arabidopsis Plants. Uptake of TC and CTC from roots into leaves was investigated by analyzing leaf extracts from plants treated with these
ARTICLE
antibiotics using liquid chromatography/mass spectrometry (LC/ MS). Supporting Information 3 shows the extracted ion chromatograms (m/z 445 for TC and m/z 479 for CTC) of the TC- and CTC-treated Arabidopsis. These chromatograms showed a distinct peak with retention time at 9.10 min, which matched the retention time of the TC standard (not shown). The MS/MS fragmentation of m/z 445 in the TC-treated extract matched that of the TC standard, forming the characteristic fragment ions m/z 427 (loss of NH3) and m/z 410 (loss of NH3 + H2O).21 Similarly, the retention time at 9.63 min observed in CTC-treated plant extract matched that of the CTC standard (not shown). Further fragmentation of m/z 479 produced m/z 462 ([M NH3]+) and m/z 444 ([M (NH3 + H2O)]+) in the CTC-treated extract. This fragmentation pattern is characteristic of the CTC MS/MS spectra.21 These results provide direct evidence of CTC uptake through the roots and translocation into the leaves of Arabidopsis plants, establishing an association between uptake, reduced growth rates, and other observable toxicity effects in this antibioticsensitive plant. CTC Effects on Protein Accumulation and Synthesis. The toxic effects related to the uptake of CTC by Arabidopsis plants could be caused by direct or indirect effects on the accumulation or synthesis of proteins required for normal growth and development. Immunoblot analyses for a representative group of nuclear- and organelle-encoded proteins indicated only minor reductions, or no reductions, in accumulation following 9 days of CTC exposure (Figure 2A, B, C). Among these representative proteins in leaves were the chloroplast-encoded Rubisco large subunit (LSU, plastid localized) (Figure 2A, L, top panel), the nuclear-encoded Rubisco small subunit (SSU, transported into plastids) (Figure 2A, S, middle panel), and the mitochondrialencoded cytochrome oxidase subunit II (CoxII) (Figure 2A, Co, bottom panel). Similarly, phosphoenolpyruvate carboxylase (PEPCase, cytoplasmic) did not show any significant differences in accumulation in CTC-treated leaves or roots (Figure 2B, P, top panel and 2C, P, top panel, respectively). As expected, the Rubisco proteins were not detectable in the roots (not shown). These immunoblots indicate that CTC toxicity was not associated with significant overall reductions in protein accumulation within the cellular cytoplasm or organelles. While substantial changes in protein composition can occur within different cellular compartments under abiotic stress conditions,33 36 the abundance of these specific organelle or cytoplasmic proteins did not occur in response to CTC stress. It is worth noting that we found only slight changes in the Rubisco LSU or SSU proteins in plants displaying CTC toxicity, since significant Rubisco degradation can be associated with many types of plant stress, including heavy metal contaminants.35 Whereas only small differences between treated and untreated Arabidopsis were observed for the representative proteins described above, there were detectable differences in accumulation of the NAD-dependent malic enzyme (NAD-ME). The accumulation of this protein was reduced approximately 2 3 fold (determined using phosphorimager quantification) in the leaves of 9-day treated plants, relative to control plants (Figure 2B, N, middle panel). At the same time, NAD-ME protein levels in roots of CTC-treated plants increased approximately 2 3 fold over control plants (Figure 2C, N, middle panel). In contrast, another mitochondrial enzyme, the organelle-encoded CoxII, did not shown any changes in abundance (Figure 2A, Co, bottom panel). To confirm that changes in NAD-ME production occurred reciprocally in leaves and roots of CTC-treated Arabidopsis 8960
dx.doi.org/10.1021/es200863j |Environ. Sci. Technol. 2011, 45, 8958–8964
Environmental Science & Technology
Figure 2. Accumulation and synthesis of representative proteins in CTC-treated and untreated plants: Soluble protein extracts were prepared from leaves (Columns A, B) or roots (Column C) of CTC-treated (T) and control (C) plants. Extracts were fractionated by SDS-PAGE, blotted to nitrocellulose paper, and reacted with antisera against the Rubisco large and small subunits, and CoxII (Column A; L, S, and Co in top, middle, and bottom panels, respectively), as well as PEPCase and NAD-ME (Columns B and C; P and N in top and middle panels, respectively). Antibody reactions were visualized with a luminol detection system and quantified using a phosphorimager. For the bottom panels of Columns B and C, in vivo synthesis of NAD-ME (N-ip) was determined in leaves and roots of treated and control plants labeled with [35S]Met/Cys as described in the text. NAD-ME was immunoprecipitated from equal amounts of incorporated label, separated by SDSPAGE, visualized, and quantitated using a phosphorimager.
plants, the in vivo synthesis of this protein was examined in leaves and roots of 9-day treated and control plants. NAD-ME was immunoprecipitated from equalized amounts of 35S-labeled leaf or root extracts (Supporting Information Table 1). Relative levels of NAD-ME synthesis were reduced in the leaves of CTC-treated plants (Figure 2B, N-ip, bottom panel), whereas in roots of treated plants NAD-ME synthesis increased (Figure 2C, N-ip, bottom panel). Synthesis in both tissues correlated with changes in the accumulation of this protein, indicating that production/accumulation of this protein was affected differently by CTC exposure in leaves and roots. The mitochondrial NAD-ME has a variety of specialized functions in plants,37,38 including regulation of alternative oxidase (Aox) activity during oxidative stress.33 Hairy root cultures of sunflowers treated with oxytetracycline induced a stress response, where reactive oxygen species (ROS) were produced by enzymes such as NADPH oxidase and peroxidases.39 ROS exuded from the roots into the media caused oxidative root damage. If CTC induced a similar response in Arabidopsis, then associated ROS root damage would induce oxidative stress, leading to the activation of the Aox response pathway (including NAD-ME). In agreement with our findings, such a partial repartitioning of respiration would require the induction of Aoxrelated enzymes and would not likely affect levels of CoxII itself (for example, see ref 40). The reason for lowering of NAD-ME, but not CoxII, in leaves in response to CTC is not clear. Metabolic changes affected by CTC exposure are likely to be complex, and will likely vary in different plant tissues. Because at least one representative protein showed differences in accumulation/synthesis in response to CTC, it was of interest to examine overall levels of protein synthesis. Total synthesis levels (determined from TCA-precipitated 35S-labeled proteins, Supporting Information Table 1) were similar for soluble leaf and root proteins in treated versus control plants. Synthesis levels were slightly lower (2 3 fold) for membrane-bound leaf
ARTICLE
proteins, and slightly higher (2 4 fold) for membrane-bound root proteins. Within these overall levels, there were easily observable differences for several protein bands in treated and control plants (Supporting Information 4). For example, in leaves there were very minor CTC-associated reductions (less than 2) in Rubisco LSU and SSU synthesis rates. In the soluble and membrane-bound root extracts, some protein bands decreased while others increased (beyond overall changes in synthesis), in response to treatment. Alterations in patterns and levels of protein synthesis continued through 12 days post-treatment, in agreement with the continuing of observable toxicity effects through this period (Supporting Information 5). Taken together, these studies of protein synthesis indicate that, while only one identified protein has been specifically shown to change in terms of accumulation/synthesis levels in response to CTC treatment (NAD-ME), there are many changes in protein synthesis occurring within the leaves and roots in response to abiotic stress brought about by CTC uptake in Arabidopsis. Chelation of Intracellular Calcium by CTC. Calcium chelation in plants by TC antibiotics is well-known23 25 and calcium signaling affects many aspects of plant development and gene expression, especially those relating to abiotic stress.41 46 For this study we made use of transgenic Arabidopsis that constitutively expresses a YC3.6 cameleon protein47 from a CaMV 35-S promoter, to measure FRET in response to CTC exposure. To ensure standardization of dosage/uptake and minimize variation of FRET determinations between samples, plants were treated hydroponically with 20 mg/L CTC. As controls, plants were treated with EGTA (to lower intracellular calcium due to chelation) and with H2O alone. Plants were removed from soil after 9 11 days growth, and placed within hole-punched parafilm floating on the treatment media for 24 h (Supporting Information 6). For these studies we used only roots from the intact plants. Arabidopsis leaves (transgenic and wild type) emitted significant fluorescence within the excitation/emission range used here, making it difficult to differentiate true YFP FRET emission. Roots did not show this background. Although CTC itself is capable of fluorescence,48,49 this was not a factor in the FRET determinations shown here. CTC excitation (390 nm) is much lower than that used here, and CTC emission is reduced 50% at room temperature.50 In fact, using wild type (i.e., nontransformed plants that do not express YC3.6) Col0 Arabidopsis, we detected no difference between the background fluorescence in the roots of plants treated with CTC, relative to EGTA- or nontreated plants (not shown). Figure 3A C shows representative images from a control (H2O treated) root region, which is typical of the images used for this FRET study. Figure 3A shows a root region with YFP emission, and Figure 3B shows the same region in cyan emission. Figure 3C shows a color combined image used for analysis of YFP and CFP emission, with the region outlined in red indicating the entire area that was used for ImageJ fluorescence emission analysis. Figure 3D shows a graphical representation of FRET ratios obtained from images of roots from plants treated with 5 mM EGTA, 20 mg/L CTC, and control plants (H2O alone). Analysis of these images revealed significant reductions in YFP/CFP ratios within roots of CTC-treated plants, relative to controls (approximately 25% reduction in the CTC treated). Treatment of plants with the calcium chelator EGTA reduced the YFP/CFP ratios even more (50% reduction), confirming the validity of this FRET assay for determining changes in relative 8961
dx.doi.org/10.1021/es200863j |Environ. Sci. Technol. 2011, 45, 8958–8964
Environmental Science & Technology
ARTICLE
Figure 3. Confocal imaging and FRET analysis of Arabidopsis roots expressing YC3 cameleon protein, in response to EGTA, CTC, and water treatments. (A, B) YFP and CFP emission, respectively, both from the 405 nm wavelength for CFP excitation. (C) Combined image used for quantitation of YFP and CFP emission. The drawn red line shows the region analyzed for emission ratios in this root image. The representative images shown were from an H2O-treated control plant. Images were captured using a 20 objective. (D) YFP/CFP ratios for YC3-expressing roots from plants that were exposed to the treatments indicated. Experimental replicates are described in Experimental Section; standard error is shown.
calcium levels within the roots of these YC3.6-expressing Arabidopsis plants. These FRET data clearly indicate a positive correlation between CTC treatment and calcium reductions within roots. CTC and EGTA appear to cause similar toxicity effects in Arabidopsis during overnight hydroponic treatment (yellowing, leaf curling/wilting, stunted leaf growth); however, at the levels used here the EGTA effects appeared to be more severe (see Supporting Information 7 for a visual comparison of the effects of both treatments). At the resolutions used here, we did not observe any significant effects on overall root morphology for either treatment. Both treatments resulted in reductions in intracellular calcium levels relative to control plants, as determined by confocal FRET analysis, but again, the EGTA treatment caused stronger reductions. Based on these findings, we hypothesize that a comparative analysis of proteins and genes affected by CTC treatment and chelation of calcium by EGTA would likely reveal commonalities in these two toxicity processes. As a central regulator in plants, calcium is known to affect many aspects of growth and development.51 Reductions in calcium are known to decrease plant growth and development, including cell wall and membrane development, photosynthesis, and gene regulation, usually through interactions with calmodulin.41 46,51 Modification of multiple regulatory levels such as cellular signaling, transcription, translation, and protein modification can occur in response to changes in calcium fluctuation. It is therefore not surprising that widespread changes in proteomic patterns (synthesis and accumulation) were observed in response to
calcium chelation caused by CTC uptake. Additional studies will be required to identify specific mRNAs and proteins that fluctuate in direct response to CTC-induced chelation, as well as others that might fluctuate as an indirect response due to reduced growth or general stress. Environmental Relevance of Using Model Species for Phytotoxicity Studies. The experiments presented in this study provide supportive evidence for our model of plant toxicity in CTC sensitive plants. This model proposes that uptake through roots leads to reductions in levels of intracellular calcium due to chelation, leading to changes in overall patterns and levels of protein synthesis and accumulation, ultimately leading to growth reductions and other toxicity effects. This model also suggests that resistant plants, such as maize, produce proteins (GSTs and possibly other proteins)18 20 that modify the antibiotic before it can negatively affect the plant cell’s calcium levels. Although Arabidopsis is not a crop plant, its demonstrated sensitivity to the antibiotic as shown here provides a valuable model system for understanding phytotoxic effects resulting from uptake of antibiotic contaminants that may be introduced into the environment through land-application of manure and biosolids. This study has characterized the toxic effects associated with direct CTC exposure/uptake in Arabidopsis, using upper-end contaminant levels that might be encountered within a highly contaminated environment. It should be noted that these experiments do not necessarily replicate actual field conditions; factors other than CTC exposure that have the potential to reduce or enhance the effects described here (such as retention/absorption by different soils, modification by rhizosphere organisms) were 8962
dx.doi.org/10.1021/es200863j |Environ. Sci. Technol. 2011, 45, 8958–8964
Environmental Science & Technology not analyzed in this laboratory study. Using well-characterized plant systems such as Arabidopsis and maize under controlled experimental conditions will help us understand how pharmaceutical contaminants released into the environment may affect the growth and development of plant species that are vulnerable to antibiotic exposure. Since reuse of treated wastewater for irrigating crops is expected to increase in arid regions, and the use of manure and biosolids as fertilizer will remain commonplace in agriculture, it is important to understand the mechanisms of phytotoxicity that pharmaceuticals may cause to susceptible plants.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional images and analyses of CTC-treated and control plants (soil grown and hydroponically treated), MS/MS spectra demonstrating CTC and TC uptake, and SDS-PAGE showing in vivo labeling of total proteins in 9-day-old and 12-day-old plants; a table showing incorporation of [35S]Met/Cys into proteins of CTC-treated and control plants. This information is available free of charge via the Internet at http://pubs.acs.org/.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 1-716-645-4997; fax: 1-716-645-3369; e-mail: camjob@ buffalo.edu.
’ ACKNOWLEDGMENT We are grateful to Dennis Pietras for excellent plant care and maintenance, and Jim Stamos for preparing the illustrations. This work was supported by NSF Grants CHE 0750321 and MCB 0544234, USDA/NRI Grant 2008-01070, and by an Interdisciplinary Research Development Fund (IRDF) from the University at Buffalo. The Zeiss LSM 710 “In Tune” Confocal Microscope used for FRET imaging was purchased through NSF Major Research Instrumentation grant DBI 0923133. K.E. D. was supported in part by an NSF research education for undergraduate (REU) supplement to MCB 0544234, and by an undergraduate research grant from the Rochester Academy of Science. ’ REFERENCES (1) Kolpin, D. W.; Furlong, E. T.; Meyer, M. T.; Thurman, E. M.; Zaugg, S. D.; Barber, L. B.; Buxton, H. T. Pharmaceuticals, hormones, and other organic wastewater contaminants in US streams, 1999 2000: A national reconnaissance. Environ. Sci. Technol. 2002, 36, 1202–1211. (2) Boxall, A. B. A.; Kolpin, D. W.; Halling-Sorensen, B.; Tolls, J. Are veterinary medicines causing environmental risks? Environ. Sci. Technol. 2003, 37, 286A–294A. (3) Boxall, A. B. A.; Fogg, L. A.; Blackwell, P. A.; Kay, P.; Pemberton, E. J.; Croxford, A. Veterinary medicines in the environment. Rev. Environ. Contam. Toxicol. 2004, 180, 1–91. (4) Hamscher, G.; Pawelzick, H. T.; Hoeper, H.; Nau, H. Antibiotics in soil: Routes of entry, environmental concentrations, fate and possible effects. In Pharmaceuticals in the Environment, 2nd ed.; K€ummerer, K., Ed.; Springer: Berlin, 2004; pp 139 147. (5) Davis, J. G.; Truman, C. C.; Kim, S. C.; Ascough, J. C.; Carlson, K. Antibiotic transport via runoff and soil loss. J. Environ. Qual. 2006, 35, 2250–2260.
ARTICLE
(6) Aga, D.; Goldfish, S. R.; Kulshrestha, P. Application of ELISA in determining the fate of tetracyclines in land-applied livestock wastes. Analyst 2003, 128, 658–662. (7) Aga, D.; O’Connor, S. K.; Ensley, S.; Payero, J. O.; Snow, D.; Tarkalson, D. Determination of the persistence of tetracycline antibiotics and their degradates in manure-amended soil using enzyme-linked immunosorbent assay and liquid chromatography-mass spectrometry. J. Agric. Food Chem. 2005, 53, 7165–7171. (8) Batt, A. L.; Snow, D.; Aga, D. S. Occurrence of sulfonamide antimicrobials in private water wells in Washington County, Idaho, USA. Chemosphere 2006, 64, 1963–1971. (9) Yang, S. W.; Carlson, K. Evolution of antibiotic occurrence in a river through pristine, urban, and agricultural landscapes. Water Res. 2003, 37, 4645–4656. (10) Animal Health Institute. Sales of disease-fighting animal medicines rise; 2008; http://www.ahi.org/files/Media%20Center/Antibiotic% 20Use%202007.pdf. (11) Hamscher, G.; Sczesny, S.; Hoper, H.; Nau, H. Determination of persistent tetracycline residues in soil fertilized with liquid manure by high-performance liquid chromatography with electrospray ionization tandem mass spectrometry. Anal. Chem. 2002, 74, 1509–1518. (12) Park, S.; Choi, K. Hazard assessment of commonly used agricultural antibiotics on aquatic ecosystems. Ecotoxicology 2008, 17, 526–538. (13) National Academy of Sciences Committee on Drug Use in Food Animals. The Use of Drugs in Food Animals: Benefits and Risks; National Academy Press, 1999. (14) Jones, A. H.; Voulvoulis, N.; Lester, J. N. Potential ecological and human health risks associated with the presence of pharmaceutically active compounds in the aquatic environment. Crit. Rev. Toxicol. 2004, 34, 335–350. (15) Schmitt, H.; Stoob, K.; Hamscher, G.; Smit, E.; Seinen, W. Tetracyclines and tetracycline resistance in agricultural soils: Microcosm and field studies. Micro. Ecol. 2006, 51, 267–276. (16) Kim, S.; Aga, D. S. Potential ecological and human health impacts of antibiotics and antibiotic-resistant bacteria from wastewater treatment plants. J. Toxicol. Environ. Health, Part B 2007, 10, 559-573. (17) Kumar, K.; Gupta, S. C.; Baidoo, S. K.; Chander, Y.; Rosen, C. J. Antibiotic uptake by plants from soil fertilized with animal manure. J. Environ. Qual. 2005, 34, 2082–2085. (18) Farkas, M. H.; Berry, J. O.; Aga, D. S. Chlortetracycline detoxification in maize via induction of glutathione s-transferases after antibiotic exposure. Environ. Sci. Technol. 2007, 41, 1450–1456. (19) Farkas, M. H.; Berry, J. O.; Aga, D. S. Determination of enzyme kinetics and glutathione conjugates of chlortetracycline and chloroacetanilides using liquid chromatography/mass spectrometry. Analyst 2007, 132, 664–671. (20) Farkas, M. H.; Berry, J. O.; Aga, D. S. Antibiotic transformation in plants via glutathione conjugation. In Fate of Pharmaceuticals in the Environment and in Water Treatment Systems; Aga, D. S., Ed.; CRC Press: New York, 2008; pp 199 216. (21) Farkas, M. H.; Mojica, E.-R.; Patel, M.; Aga, D. S.; Berry, J. O. Development of a rapid biolistic assay to determine changes in relative levels of intracellular calcium in leaves following tetracycline uptake by pinto bean plants. Analyst 2009, 134, 1594–1600. (22) Batchelder, A. R. Chlortetracycline and oxytetracycline effects on plant growth and development in soil systems. J. Environ. Qual. 1982, 11, 675–678. (23) Jackson, C. K.; Hall, J. L. A fine structure analysis of auxininduced elongation of cucumber hypocotyls, and the effects of calcium antagonists and ionophores. Ann. Bot. 1993, 72, 93–204. (24) Moller, I. M.; Kay, C. J.; Palmer, J. M. Chlortetracycline and the transmembrane potential of the inner membrane of plant mitochondria. Biochem. J. 1986, 237, 765–771. (25) Ohyam, T.; Cowan, J. A. Calorimetric studies of metal binding to tetracycline: Role of solvent structure in defining the selectivity of metal ion-drug interference. Inorgan. Chem. 1995, 34, 3083–3086. (26) Rhee, S. Y.; Beavis, W.; Berardini, T. Z.; et al. The Arabidopsis Information Resource (TAIR): a model organism database providing a 8963
dx.doi.org/10.1021/es200863j |Environ. Sci. Technol. 2011, 45, 8958–8964
Environmental Science & Technology centralized, curated gateway to Arabidopsis biology, research materials and community. Nucleic Acids Res. 2003, 31, 224. (27) Berry, J. O.; Nikolau, B. J.; Carr, J. P.; Klessig, D. F. Transcriptional and posttranscriptional regulation of ribulose 1,5-bisphosphate carboxylase gene expression in light- and dark-grown amaranth cotyledons. Mol. Cell. Biol. 1985, 5, 2238–2246. (28) McCormac, D. J.; Boinski, J. J.; Ramsperger, V. C.; Berry, J. O. C4 Gene expression in photosynthetic and non-photosynthetic leaf regions of Amaranthus tricolor. Plant Physiol. 1997, 114, 801–815. (29) Long, J. J.; Wang, J. -L.; Berry, J. O. Cloning and analysis of the C4 photosynthetic NAD-dependent malic enzyme of amaranth mitochondria. J. Biol. Chem. 1994, 269, 2827–2833. (30) Monshausen, G. B.; Messerli, M. A.; Gilroy, S. Imaging of the yellow cameleon 3.6 indicator reveals that elevations in cytosolic Ca2+ follow oscillating increases in growth in root hairs of Arabidopsis. Plant Physiol. 2008, 147, 1690–1698. (31) Monshausen, G. B.; Bibikova, T. N.; Weisenseel, M. H; Gilroy, S. Ca2+ regulates reactive oxygen species production and pH during mechanosensing in Arabidopsis roots. Plant Cell 2009, 21, 2341–2356. (32) Batchelder, A. R. Chlortetracycline and oxytetracycline effects on plant growth and development in liquid cultures. J. Environ. Qual. 1981, 10, 515–518. (33) Sweetlove, L. J.; Heazlewood, J. L.; Herald, V.; Holtzapffel, R.; Day, D. A.; Leaver, C. J.; Millar, A. H. The impact of oxidative stress on Arabidopsis mitochondria. Plant J. 2002, 32, 891–904. (34) Ndimba, B. K.; Chivasa, S.; Simon, W. J.; Slabas, A. R. Identification of Arabidopsis salt and osmotic stress responsive proteins using two-dimensional difference gel electrophoresis and mass spectrometry. Proteomics 2005, 5, 4185–4196. (35) Feller, U.; Anders, I.; Demirevska, K. Degradation of Rubisco and other chloroplast proteins under abiotic stress. Gen. Appl. Plant Physiol. 2008, 34, 5–18. (36) Heo, J. M.; Livnat-Levanon, N.; Taylor, E. B.; Jones, K. T.; Dephoure, N.; Ring, J.; Xie, J.; Brodsky, J. L.; Madeo, F.; Gygi, S. P.; Ashrafi, K.; Glickman, M. H; Rutter, J. A stress-responsive system for mitochondrial protein degradation. Mol. Cell 2010, 40, 465–80. (37) Berry, J. O.; Zielinski, A. M.; Patel, M. Gene expression in mesophyll and bundle sheath cells of C4 plants. In Advances in Photosynthesis and Respiration Vol. 32. C4 Photosynthesis and Related CO2 Concentrating Mechanisms; Raghavendra, A. S., Sage, R. F., Eds.; Springer: Dordrecht, The Netherlands, 2010; pp 221 256. (38) Chivasa, S.; Berry, J. O.; ap Rees, T.; Carr, J. P. Changes in gene expression during development and thermogenesis in Arum. Aust. J. Plant Physiol. 1999, 26, 391–399. (39) Gujarathi, N. P.; Linden, J. C. Oxytetracycline inactivation by putative reactive oxygen species released to nutrient medium of Helianthus annuus hairy root cultures. Biotechnol. Bioeng. 2005, 92, 393–402. (40) Ederli, L.; Morettini, R.; Borgogni, A.; Wasternack, C.; Miersch, O.; Reale, L.; Ferranti, F.; Tosti, N.; Pasqualini, S. Interaction between nitric oxide and ethylene in the induction of alternative oxidase in ozonetreated tobacco plants. Plant Physiol. 2006, 142, 595–608. (41) Snowden, K. C.; Richards, K. D.; Gardner, R. C. Aluminuminduced genes 1. Induction by toxic metals, low calcium, and wounding and pattern of expression in root tips. Plant Physiol. 1995, 107, 341–348. (42) Hirschi, K. D. Expression of Arabidopsis CAX1 in Tobacco: Altered calcium homeostasis and increased stress sensitivity. Plant Cell 1999, 11, 2113–2122. (43) McCormack, E.; Tsaia, Y.-C.; Braam, J. Handling calcium signaling: Arabidopsis CaMs and CMLs. Trends Plant Sci. 2005, 10, 383–389. (44) Galona, Y.; Nave, R.; Boyce, J. M.; Nachmias, D.; Knight, M. R.; Fromm, H. Calmodulin binding transcription activator (CAMTA) 3 mediates biotic defense responses in Arabidopsis. FEBS Lett. 2008, 582, 943–948. (45) Kim, M. C.; Chung, W. S.; Yun, D. -J.; Cho, M. J. Calcium and calmodulin-mediated regulation of gene expression in plants. Mol. Plant 2009, 2, 13–21. (46) Kudla, J.; Batisti, O.; Hashimoto, K. Calcium signals: The lead currency of plant information processing. The Plant Cell 2010, 22, 541–563.
ARTICLE
(47) Nagai, T.; Yamada, S.; Tominaga, T.; Ichikawa, M.; Miyawaki, A. Expanded dynamic range of fluorescent indicators for Ca2+ by circularly permuted yellow fluorescent proteins. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 10554–10559. (48) Mathew, M. K.; Balaram, P. A reinvestigation of chlortetracycline fluorescence: Effect of pH, metal ions, and environment. J. Inorg. Biochem. 1980, 13, 339–346. (49) Cerella, C.; Mearelli, C.; De Nicola, M.; D’Alessio, M.; Magrini, A.; Bergamaschi, A.; Ghibelli, L. Analysis of calcium changes in endoplasmic reticulum during apoptosis by the fluorescent indicator chlortetracycline. Ann. N.Y. Acad. Sci. 2007, 1099, 490–493. (50) Oliver, A. E.; Baker, G. A.; Fugate, R. D.; Tablin, F.; Crowe, J. H. Effects of temperature on calcium-sensitive fluorescent probes. Biophys. J. 2000, 78, 2116–2126. (51) Hepler, P. K. Calcium: A central regulator of plant growth and development. Plant Cell 2005, 17, 2142–2155.
8964
dx.doi.org/10.1021/es200863j |Environ. Sci. Technol. 2011, 45, 8958–8964
ARTICLE pubs.acs.org/est
Tailored Microarray Platform for the Detection of Marine Toxins T. F. H. Bovee,*,† P. J. M. Hendriksen,† L. Portier,† S. Wangz,†,‡ C. T. Elliott,§ H. P. van Egmond,† M. W. F. Nielen,†,|| A. A. C. M. Peijnenburg,† and L. A. P. Hoogenboom† †
)
RIKILT-Institute of Food Safety, Wageningen UR, Business Unit Bioanalysis & Toxicology, Akkermaalsbos 2, 6708 WB Wageningen, The Netherlands ‡ Division of Toxicology, Wageningen University and Research Centre, Tuinlaan 5, 6703 HE Wageningen, The Netherlands § Institute of Agri-Food and Land Use, School of Biological Sciences, Queen’s University, Belfast BT9 5AY, Northern Ireland Wageningen University, Laboratory of Organic Chemistry, Dreijenplein 8, 6703 HB Wageningen, The Netherlands ABSTRACT: Currently, there are no fast in vitro broad spectrum screening bioassays for the detection of marine toxins. The aim of this study was to develop such an assay. In gene expression profiling experiments 17 marker genes were provisionally selected that were differentially regulated in human intestinal Caco-2 cells upon exposure to the lipophilic shellfish poisons azaspiracid-1 (AZA1) or dinophysis toxin-1 (DTX1). These 17 genes together with two control genes were the basis for the design of a tailored microarray platform for the detection of these marine toxins and potentially others. Five out of the 17 selected marker genes on this dedicated DNA microarray gave clear signals, whereby the resulting fingerprints could be used to detect these toxins. CEACAM1, DDIT4, and TUBB3 were up-regulated by both AZA1 and DTX1, TRIB3 was up-regulated by AZA1 only, and OSR2 by DTX1 only. Analysis by singleplex qRT-PCR revealed the up- and downregulation of the selected RGS16 and NPPB marker genes by DTX1, that were not envisioned by the new developed dedicated array. The qRT-PCR targeting the DDIT4, RSG16 and NPPB genes thus already resulted in a specific pattern for AZA1 and DTX1 indicating that for this specific case qRTPCR might a be more suitable approach than a dedicated array.
’ INTRODUCTION Among the fast growing microscopic algae at the base of the marine food chain, there are a few dozen that produce potent toxins and these algae are found worldwide.14 Public health impact from harmful algal blooms occurs when toxic phytoplankton are filtered from the water by fish, crabs, and especially, shellfish that accumulate the algal toxins to levels that are potentially lethal to humans upon consumption.515 Because of the consequences of global and regional climate changes for algae, it is expected that the occurrence, patterns, and chemistries of marine toxins will change. As a result marine toxins present a factor of growing concern.15,16 On the basis of their chemical properties shellfish toxins can be divided into two broad classes: hydrophilic and lipophilic toxins. One the basis of their adverse effects four classes are recognized, amnesic, paralytic, neurotoxic and diarrhetic poisons. The Amnesic and Paralytic Shellfish Poisons, ASPs and PSPs, respectively, are hydrophilic and have a molecular weight (MW) below 500 Da, while the Diarrhetic Shellfish Poisons, DSPs, have strong lipophilic properties and a MW above 600 Da (up to 2000 Da).5,14,17 Their modes of action are different, complex, and not fully understood, but the DSPs okadaic acid (OA) and dinophysistoxins (DTXs) for instance inhibit protein phosphatases, including the serine/threonine phosphatases 1 and 2A (PP1 and PP2A),2,14 while some PSPs affect the nervous transmission by blocking sodium channels.14,1820 PSPs provoke nerve dysfunction and lead to paralysis and death due to r 2011 American Chemical Society
respiratory failure.21 For other sea fish poisons, including ASPs, the modes of action are still unclear.22 Presently the detection of marine toxins is still largely based on in vivo tests in which mice are injected with an extract of the test sample and lethality is the selected end point for a positive test result.5,23 DSPs can also be detected by feeding rats the test sample potentially resulting in diarrhea. Although EU legislation prescribes these in vivo tests for the determination of OA, DTXs, yessotoxins (YTXs), pectenotoxins (PTXs), and AZAs in shellfish, they are considered highly unethical and will be forbidden from 2015 onward.24 Chemical analytical methods are being developed for the known marine toxins 2527 and most recently an LC-MS/MS method has been developed for the detection of the most prominent lipophilic marine toxins.28 However, these methods are expensive and not high-throughput. Even more important is the fact that they are not able to detect presently unknown toxins and cannot correlate concentrations detected to toxicity because of the absence of toxic equivalence factors for individual toxins expressing their relative potency.29 Moreover, the fact that most classes of marine toxins consist of many analogues and the lack of certified standards and reference materials, makes use of Received: April 5, 2011 Accepted: August 19, 2011 Revised: August 19, 2011 Published: August 19, 2011 8965
dx.doi.org/10.1021/es2011385 | Environ. Sci. Technol. 2011, 45, 8965–8973
Environmental Science & Technology chemical analytical methods for detecting all marine toxins very difficult if not impossible. There are at least 24 PSP type saxitoxins,30 13 DSP type okadaic acid-ester derivatives 2 and 90 DSP type yessotoxin analogues,31 15 different NSP type neurotoxic brevetoxins,32 and around 30 different AZP type azaspiracid analogues.6 The development of suitable alternative test methods to detect marine toxin contaminations, both those that are currently known and those which might emerge, is therefore urgently needed. In vitro bioassays offer the possibility to screen for the presence of whole classes of contaminants and toxins based on their biological effects. As a result, such bioassays are able to detect both known and yet unknown compounds. Well known examples are the use of bacteria to detect antibiotics and the use of genetically modified yeast or mammalian cells for the detection of hormones and dioxin-like compounds.33,34 Toxins of the OA group can be determined by a PP2A enzyme inhibition assay, PTXs and OAs by a cytotoxicity assay based on the BE(2)-M17 neuroblastoma cell line, OAs and YTXs by accumulation of E-cadherin in MCF-7 and Caco-2 cells and SPXs by an inhibition assay based on nicotinic acetylcholine receptor-enriched membranes.3538 However, none of these effect-based in vitro bioassays for marine toxins are suitable enough for routine applications or has been validated successfully. The main objective of the present study was to develop effect-based in vitro bioassays capable of detecting marine toxins by identifying marker genes specifically up- or down regulated in human Caco-2 cells. Human intestinal Caco-2 cells were exposed to pure standards of the lipophilic diarrhetic shellfish poisons AZA1 and DTX1, blank mussel extracts, and to mussel extracts contaminated with AZA1 or DTX1. Seventeen genes were selected that were differentially regulated by examining the gene expression using whole genome Agilent microarrays. Subsequently, specific primers and probes were constructed for the selected 17 marker genes and 2 control genes and used to develop a dedicated low-density microarray which functioning was subsequently tested by exposing Caco-2 cells to pure standards of AZA1, DTX1, and okadaic acid (OA).
’ EXPERIMENTAL SECTION Chemicals. Okadaic acid (OA) and azaspiracid-1 (AZA1) were purchased from the National Research Council, Institute for Marine Biosciences (NRC-CNRC), Halifax, Canada. Dinophysis toxin-1 (DTX1) reference material was kindly donated by Dr P. Hess from the Marine Institute, Oranmore, Ireland. Stock solutions of OA, AZA1 and DTX1 in DMSO were prepared in the concentration range of 150 μM down to 15 nM. DMSO and ethanol were obtained from Merck (Darmstadt, Germany), and β-mercaptoethanol was obtained from Sigma (St. Louis, MO, U.S.A.). Cell Culture and Exposure. The human colorectal adenocarcinoma Caco-2 cell line was obtained from the American Type Culture Collection (ATCC, Manassas, VA, U.S.A.) and used at passages 2535 at a seeding density of 60 000 cells/cm2. The BioWhittaker’s Dulbecco modified Eagle’s minimal essential medium with L-glutamine (DMEM, 4.5 g/L glucose) was obtained from Lonza (Verviers, Belgium). Caco-2 cells were grown in DMEM supplemented with nonessential amino acids (NEAA) from MP Biomedicals (Illkirch, France), penicillinstreptomycin from Sigma (St. Louis, MO, U.S.A.), and heat
ARTICLE
inactivated fetal bovine serum (FBS; 10% v/v) from Gibco BRL (Life Technologies Ltd., Paisly, Scotland). For exposure, cells were seeded in 6-wells plates. When cells reached a confluence of approximately 50%, the cells were exposed to one of the three marine toxins o/n. Concentrations tested were 1, 3, 10, 30, and 100 nM for DTX1, 0.03, 0.1, 0.3, 1, 3, 10, and 20 nM for AZA1 and 0.3, 1, 3, 10, 30, and 100 nM for OA for the tube array experiments and 6.25 and 25 nM of DTX1 and AZA1 for the qRT-PCR analysis. Wells contained 2 mL of medium, and the final DMSO concentration was 0.2%. Exposures were performed in triplicate. RNA Extraction, Complementary Strand Synthesis, Amplification, and Biotin-Labeling. The RNA from the exposed Caco-2 cells was extracted using the QIAshredder and RNeasy Mini kits (Qiagen, Venlo, The Netherlands) according to the manufacturer’s protocols. In short, the medium was removed from the cells and the cells were lysed with 600 μL f RLT buffer with 1% β-mercaptoethanol. After the extraction of the RNA using the QIAshredder and RNeasy columns, the amount and quality of the RNA was evaluated by UV spectrometry (260 and 280 nm wavelength on the Nanodrop spectrophotometer of Nanodrop technologies). Primerset 1 is a mix of reverse primers for all selected genes and the two control genes. The sequence of these rtprimers is given in Table 4. Multiplex cDNA synthesis was performed with 5 μg of purified RNA and 1.5 μL of 0.1 μM primer set 1 in an end volume of 25 μL using the Omniscript RT kit (Qiagen) according to the manufacturer’s instructions. Tubes were placed into a thermocycler for 30 min at 50 °C and subsequently put on ice. Primerset 2 is a mix of biotin-labeled forward primers (lbprimers) for all selected genes and the two control genes and primerset 3 is a mix of primers complementary to all primers of primerset 1. The sequence of the lbprimers (primer set 2) is given in Table 4. Labeling and a linear amplification was then carried out by adding 1.7 μL 3.0 μM Primer set 2 (with a 50 -biotin label) and 1.7 μL 3.0 μM Primer set 3 (a mix of competitor sequences complementary to the primers in primer set 1 in order to avoid an exponential amplification), followed by a multiplex linear amplification reaction of 40 cycles in a thermocycler (melt 15 min at 95 °C, 40 cycles of 45 s at 94 °C, annealing for 45 s at 56 °C and elongation for 45 s at 76 °C). The purity of the cDNA product was determined by UV spectrometry on the Nanodrop spectrophotometer and the A260/ A280 ratio had to be within the range of 1.82. Low-Density Microarray Hybridizations. Table 3 shows the sequence of the manufactured hybridization probes (hbprobes) that were immobilized on the microchip (microchips of 9 mm2, printed with 156 oligonucleotide probes, manufactured by Alere Technologies GmbH, Jena, Germany), where each gene was represented by two oligonucleotide probes and each oligonucleotide probe was printed in duplicate. Biotin-labeled amplification products were heat-denatured (95 °C for 5 min) and hybridized to the DNA microchips in the array tube format The microchips were equilibrated at 30 °C by rinsing them twice for 5 min with 500 μL hybridization buffer (0.9 M NaCl, 60 mM sodium phosphate, 6 mM EDTA, 0.05% triton X-100, pH 7.4). Hybridizations (60 °C for 1 h) were performed with 16 μg of biotinylated cDNA in 100 μL hybridization buffer. Next, the arrays were washed with 500 μL of 2xSSC containing 0.1% (v/v) Triton X-100 (5 min, 30 °C), with 2xSSC (5 min, 20 °C), and finally with 0.2xSSC (5 min, 30 °C) (1xSSC = 0.15 M NaCl, 15 mM sodium citrate, pH 7.0). In addition the microchip contained spot controls for the biotin-marker. 8966
dx.doi.org/10.1021/es2011385 |Environ. Sci. Technol. 2011, 45, 8965–8973
Environmental Science & Technology
ARTICLE
Table 1. Qiagen QuantiTect Primer Assay assay name
gene symbol amplified exons amplicon length
Hs_TMEM179B_1_SG TMEM179B
150 bp
Hs_GAPDH_2_SG
GAPDH
1/2/3
Hs_RGS16_1_SG
RGS16
1/2
Hs_DDIT4_1_SG
DDIT4
Hs_NPPB_1SG
NPPB
119 bp 100 bp
Table 2. 2-log-Fold Up- and Down-regulation of 17 Selected Genes and Two Control Genes in Caco-2 Cells Exposed to Blank Mussel Extracts Spiked with Either 6.25 nM AZA1 or DTX1 (Up-regulation Marked with + and Down-regulation Marked with ) as Analyzed on Whole Genome Agilent Microarrays with 60-mer Probes
97 bp 2/3
upon exposure upon exposure
135 bp
Probe Detection and Data Analysis. After blocking with 2% (w/v) milk powder (Bio-Rad, Hercules, CA, U.S.A.) in 500 μL hybridization buffer containing 0.05% (w/v) Triton X-100 (15 min, 30 °C), the arrays were incubated for 15 min at 30 °C with 100 μL of a 1:2500 dilution of horseradish peroxidase-streptavidin conjugates (1 mg/mL; Thermoscientific, Rockford, Canada) in hybridization buffer containing 0.05% (w/v) Triton X-100. Thereafter, the arrays were washed with 500 μL of 2xSSC containing 0.1% (w/v) Triton X-100 (5 min, 30 °C), with 2xSSC (5 min, 20 °C) and 0.2xSSC (5 min, 20 °C). The reaction was initiated by adding 100 μL of TrueBlue Peroxidase substrate (KPL, Gaithersburg, MD, U.S.A.) and the blue precipitates were recorded by light absorption in a dedicated array tube reader ATR01 (Alere Technologies). Data analysis was conducted with the IconoClust and Partisan software version 3.5r (Alere Technologies), and all transcript levels were compared or normalized to the housekeeping control glyceraldehyde-3-phosphate dehydrogenase (GAPDH). qRT-PCR Analysis. qRT-PCR controls were performed for DDIT4, RGS16, and NPPB. The RNA from the exposed Caco-2 cells was extracted using the QIAshredder and RNeasy Mini kits as described above. Each sample was processed in triplicate. Synthesis of cDNA was performed on 1 μg RNA isolated from cells exposed to DMSO, DTX1, and AZA1 (both at 6.25 and 25 nM) and from a RNA pool (mix) of these 5 treatments using the BioRad iScript cDNA Synthese Kit with iScript and Reverse Transcript (BioRad 170-8891) in the BioRad iCycler. This cDNA synthesis was performed with the following program: 5 min at 25 °C, 30 min at 42 °C, 5 min at 85 °C, and put on ice. After cDNA synthesis the samples were diluted 100 times and the pool was diluted 10, 31.6, 100, 316, 1000, and 3160 times and used to make a calibration line. Next, qRT-PCR was performed with specific and certified QuantiTect primers (Qiagen, Venlo, The Netherlands) and was carried out in a 15 μL reaction mixture containing 8,5 μL SYBR green (BioRad 170-8880), 2.5 μL of the QuantiTect forward/reverse primer mix, 2 μL RNase free water and 2 μL of diluted cDNA. Reactions were carried out in a BioRad HSP9645 PCR plate. Water and a pool without the addition of the reverse transcript were used as negative controls. The plate was covered with a microseal and centrifuged for 1 min. Thermal cycling was performed in a CFX96 Real-Time System (BioRad) starting with a denaturation step at 95 °C for 3 min, followed by 45 cycles at 65 °C for 35 s for annealing, 95 °C for 10 s for denaturation and extension at 65 °C for 1 min. PCR products were checked by melting curve analysis applying an increment of 0.5 °C per 5 s from 65 to 95 °C. Data were analyzed using BioRad software and GAPDH and TMEM179B as controls for normalization. Expression ratios of the genes under investigation were calculated versus the DMSO control.
gene
spot
intensity rank
intensitya 1 to 100b
to AZA1
to DTX1
AK091132
1.51
0.18
3.4
C21orf129
1.76
1.55
17.8
79
C3orf57 CDKN1C
1.75 0.42
0.12 +2.70
4.7 8.4
56 67
CEACAM1
+2.14
+1.51
22.0
82
DDIT4
+2.40
+1.44
18.7
80
GAPDH control
+0.13
0.14
318.9
99
50
LOC387763
+0.60
+3.52
1.0
22
MAFB
+0.09
+3.50
1.7
34
NPPB
0.07
1.82
52.3
90
OSR2 RGS16
0.85 +0.30
+0.81 +2.79
2.8 1.7
46 34
TFEC
1.55
0.95
2.5
43
TMCC1
+1.16
+0.14
1.9
37
TMEM179B control
0.04
+0.04
3.5
50
TNS4
+2.66
+1.80
1.3
28
TRIB3
+1.11
+0.06
4.7
56
TUBB3
+0.72
+2.25
23.4
83
VASN
0.06
+3.10
4.6
55
a
Spot intensity compared to the median intensity (set on 1) of all spots on the microarray. b The relative intensity of the spot when the spot with highest intensity on the array is set at 100 and the spot with lowest intensity at 1.
’ RESULTS Gene Selection. Caco-2 cells were exposed o/n to the lipophilic diarrhetic shellfish poisons AZA1 and DTX1, both at 6.25 and 25 nM, to blank mussel extracts and to mussel extracts naturally contaminated with AZA1 or DTX1. RNA was extracted, reverse transcribed, labeled and analyzed on whole genome Agilent microarrays containing 44 000 probes of 60 bp representing approximately 25 000 human genes. Seventeen genes were selected based on the response to AZA1 or DTX1 exposure by examining the gene expression using these Agilent microarrays (this microarray study will be described elsewhere). Table 2 shows the 17 selected genes and two control genes (GAPDH and TMEM179B) and their 2 log fold up- or down regulation by either AZA1 or DTX1. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and transmembrane protein 179B (TMEM179B) were selected as reference genes since these genes were not affected by the treatments according to the Agilent transcriptomics study. The GAPDH is a well-known housekeeping gene that is often used as a control for microarray analysis.31 Table 2 also provides an indication of the expression level of the genes. No matrix effects were observed on the gene expression profile when Caco-2 cells were exposed to blank mussel extracts (data not shown). Indicating that proper marker genes for AZA1 and DTX1 exposure were chosen.As shown in Table 2, genes were selected that were up-regulated by both AZA1 and DTX1, for example, CEACAM1, DDIT4, TNS4, and 8967
dx.doi.org/10.1021/es2011385 |Environ. Sci. Technol. 2011, 45, 8965–8973
Environmental Science & Technology
ARTICLE
Table 3. Sequence, Length (bp), %GC, and Melt Temperature (Tm) of the Designed Non-Overlapping Hybridization Probes (hbprobes) for the 17 Selected Genes and the Two Control Genes gene
sequence hbprobe
length
% GC
Tm
AK091132-01
CACTAACAGTTCCTTCTGCATCTGTCC
27
48
60.8
AK091132-02
ACACAGTTCCTTTATTTCAGATGTGTTTGTCT
32
34
60.9
C21orf129-01
TCCTCTAGGTGGTATTTCAAAGGGTACG
28
46
60.9
C21orf129-02 C3orf57-01
TCATAAACTGGAGCACATGATCGGGT TCTCCGGTCTAAAGCACAATGTAGCA
26 26
46 46
61.4 61.3
C3orf57-02
ACAGAATATGATTTGTGAGGTAAAGTCACTGT
32
34
60.2
CDKN1C-01
AGTGTACCTTCTCGTGCAGAATACATTTAGA
31
39
61.2
CDKN1C-02
TTTGCACTGAGTTTCAGCAGAGATTAAACA
30
37
61.1
CEACAM1-01
CTGAAAGAGGTACCTGAGTATAGAGAACTCC
31
45
60.6
CEACAM1-02
AGCACAGCATGGATGAGGGAAAG
23
52
60.3
DDIT4-01
AGTAAGATACACAAACCACCTCCACGAC
28
46
61.7
DDIT4-02 GAPDH-01
ACAAACAACAAACACACTTGGTCCCTTC TCTAGACGGCAGGTCAGGTCCA
28 22
43 59
61.9 61.7
GAPDH-02
TCACCACCTTCTTGATGTCATCATATTTGG
30
40
60.7
LOC387763-01
TCAATCCTCCAGACGCAGTAGCAG
24
54
61.5
LOC387763-02
CGTTAAATAAATGCCTTCAGCCATCGC
27
44
61
MAFB-01
CGCCAGCAATTTCAAATGGGAACTT
25
44
60.5
MAFB-02
GCACCATGCGGTTCATACAATCTTGT
26
46
61.7
NPPB-01
CAGCCAGGACTTCCTCTTAATGCC
24
54
60.5
NPPB-02 OSR2-01
CCTTGTGGAATCAGAAGCAGGTGTC ACAAGAAACTAGCATATACTCTTTGTGAAGGG
25 32
52 38
61 60.5
OSR2-02
CAAAGACCCACAACATTATGTACAGAGCT
29
41
60.7
RGS16-01
AGGTTCCCAGCACATTCAGAGGT
23
52
61
RGS16-02
ACAGTGAGGAGTTCCTGACAGGC
23
57
61.2
TFEC-01
TGCTGTTAGATAAGATTCTTGAATTTCGTTTGC
33
33
60.6
TFEC-02
AGCCTTGTATGCCATAAGCAACTGC
25
48
61.4
TMCC1-01
GGTGTGCTCCAGATTGTTGGTTCA
24
50
60.8
TMCC1-02 TMEM179B-01
TGCAGTCCTTCAAGCCACAATTTAGA TGATGGAGTTGCAGAGAGACCTGG
26 24
42 54
60.3 61.2
TMEM179B-02
CTGGGCTTCAGAACAGCTAATTGTAGT
27
44
60.3
TNS4-01
CTGGTCTCAGCCTCTCCTGCAG
22
64
61.6
TNS4-02
CCTGTGACCTTGAGAACCTCATCTCA
26
50
61.1
TRIB3-01
CAGGAGTCCTCCAGGTTCTCCAG
23
61
60.8
TRIB3-02
CAGGGAATCATCTGGCCCAGTCA
23
57
61.4
TUBB3-01
CCTGGAGCTGCAATAAGACAGAGACA
26
50
61.6
TUBB3-02 VASN-01
AGTATCCCCGAAAATATAAACACAAACCAGT TCTTCCCAGAATAAATGGGAAAGGATCTCT
31 30
35 40
60.3 60.7
VASN-02
AAACCAAAGTCCTTGTCTCTGAGTTTGAA
29
38
60.5
TUBB3, up-regulated by AZA1 only, for example, TRIB3, genes that were up-regulated by DTX1 only, for example, CDKN1, MAFB, RGS16, and VASN, down-regulated by both AZA1 and DTX1, for example, C21orf129 and TFEC, down-regulated by AZA1 only, for example, AK091132 and C3orf57, and genes that were down-regulated by DTX1 only, for example, NPPB. DNA Microchip Design. Two nonoverlapping hybridization probes of 2132 bp (hbprobes) were designed for each of the 17 selected genes and the two control genes and these probes were immobilized on the glass chip in the array tube format. Each oligonucleotide hbprobe was printed in duplicates on the microchip. The sequence, %GC and melting temperature (Tm) of the hbprobes are given in Table 3. The advantage of this particular microchip format is that it fits into the bottom of an ordinary reaction tube and, hence, requires no specialized laboratory equipment.
For each target mRNA three types of oligonucleotides were made: (1) a rtprimer, (2) a lbprimer, and (3) a competitor primer complementary to the rtprimer. The rtprimer is used for reverse transcribing the selected mRNA. Subsequently, the lbprimer and competitor primer are used to obtain biotin-labeled sequences. The lbprimer is a forward primer that is 50 -labeled with biotin. This lbprimer is used for the cDNA amplification. The competitor primers are complementary to the rtprimers and are added during the cDNA amplification in order to avoid a PCR-like exponential amplification. The 30 -OH ends of these competitor primers are substituted by an amino group to avoid unwanted extension by DNA polymerase activity. Table 4 shows the sequence, %GC and melt temperature (Tm) of the designed oligonucleotides for the rtprimers and lbprimers. Both the primers and probes were designed as previously described for 8968
dx.doi.org/10.1021/es2011385 |Environ. Sci. Technol. 2011, 45, 8965–8973
Environmental Science & Technology
ARTICLE
Table 4. Sequence, Length (bp), % GC, and Melt Temperature (Tm) of the Designed Reverse Transcription Primers (rtprimer)a and the 50 -Biotin-Labeled Forward Primers (lbprimer) gene
sequence rtprimer
length % GC 0a
Tm
sequence lbprimer
length % GC
Tm
AK091132
50 -AGTGCAGTTATGAATTGTTAACATGT-3
C21orf129 C3orf57
TTCCGAGTGACGTGCAACT TCTGGTAGGTCTGTTAGGAGTCT
26
31
55.4 50 -CTAAAGCTGACCTAAATGGATCAGT-30
25
40
55.9
19 23
53 48
56.9 CCCTAGGAAAGTAAAACAAGGAGT 56.6 CCTGGAGTTTGAAATACATGTCTCT
24 25
42 40
55.4 56.1
CDKN1C
TCCACGCCTGGACATAAATAGA
22
45
56.2 TGGGACCGTTCATGTAGCAG
20
55
57.1
CEACAM1
TTCAGACATGATCTTAGCCAGGA
23
43
56.2 CCTAAGAGGCTTTCTCCAGGA
21
52
55.8
DDIT4
TCAGTAGTGATGCTCCGATCAG
22
50
56.7 GGTGAAGGAAGAGGCACGT
19
58
56.9
GAPDH
TCAGTGTAGCCCAGGATGC
19
58
56.4 GGAAGCTCACTGGCATGG
18
61
55.4
LOC387763 AGGAGCCTCACAGACATATCC
21
52
55.9 CACACCCATTCACACTCACG
20
55
56.5
MAFB
CTGATGCAGGACAAATATCCACA
23
43
56.1 AGTAAAAGGATTTAAGTTGCACTGAC
26
35
55.6
NPPB OSR2
TGTTGACTTTATTTCACCGTGGA CTAAGTGCATCCAACAAGATCCC
23 23
39 48
55.8 GGATCAGCTCCTCCAGTGG 56.7 ACTGTAGAAAGCCACACACTACT
19 23
63 43
56.5 56.6
RGS16
TCATATCTAGGAAACCAGCTCCA
23
43
55.7 GGTTTCAGCCTGACTGTCTC
20
55
55.5
TFEC
TGTTTTCTTGATCTAACCAATGTCAG
26
35
55.4 AGAAGAGACTGATTTTGCTGAAAGA
25
36
55.9
TMCC1
CCTCATCAGGTTTTCGCATCT
21
48
55.7 TGTAGCATGATTTTCCTTGGGATG
24
42
56.6
TMEM179B TGGAGTAAAACTGCAGAGCAGT
22
45
57.2 GTGTCTGCCTGTATCCTTCGA
21
52
56.8
TNS4
TCATGATAGGAGCTGTAGCAGA
22
45
55.4 GTTGCATTATCTTTGGCCAAGG
22
45
55.8
TRIB3
GAGTATCTCAGGTCCCACGT
20
55
55.5 TCAAGCTGTGTCGCTTTGTC
20
50
56.4
TUBB3 VASN
TTTTAAGGGTATCTGACAGCAATAGA GCCTTCATATCATCGTTTGTCTTACA
26 26
35 38
55.4 CTGCAGTATTTATGGCCTCGT 57 AACTGGAAAGGAAGATGCTTTAGG
21 24
48 42
55.5 56.2
a
Sequences for the competitor primers with the aminolink, in order to prevent an exponential amplification, are complementary to the rtprimer. For example, for the AK091132 marker gene this is as follows: 50 -ACATGTTAACAATTCATAACTGCACT-C7-Aminolink-30 .
the developed low-density tube microarray for the detection of mycotoxins.39 DNA Microchip Results. Caco-2 cells were exposed to different concentrations of AZA1, DTX1, and okadaic acid (OA). After a 24 h exposure, the cells were visually inspected under a microscope and subsequently RNA was extracted from the cells. All cells were viable, except those exposed to 30 and 100 nM DTX1, as these concentrations of DTX1 appeared cytotoxic for the Caco-2 cells. Selected marker genes and control genes in purified RNA (5 μg) were converted to biotin-labeled cDNA (cDNA) using the three sets of oligonucleotide primers targeting the 17 selected mRNA transcripts and 2 control transcripts. The labeled PCR products, 136 to 318 bp in length, were then used for hybridization on the developed low density microarray on the tube format. Staining was performed with a streptavidin coupled with a peroxidase and the addition of tetramethylbenzidine. Figure 1 shows the results on the Array Tube obtained with Caco-2 cells exposed to a DMSO control and standards of 10 nM AZA1, 30 nM DTX1, and 100 nM OA. The square spots represent the biotin controls. The two control genes gave clear spots and five of the 17 selected marker genes gave spot intensities above the spot intensity as obtained with the DMSO control. These 5 marker genes were CEACAM1, DDIT4, TRIB3, TUBB3, and OSR2 (Figure 1 and all indicated in bold in Table 2 for the complementary result on the Agilent array). As shown in Figure 2, the GAPDH control was highly expressed and not affected by any treatment, even not by DTX1 concentrations of 30 and 100 nM that were cytotoxic for the Caco-2 cells (Figure 2a). Similar results were obtained with the second control gene TMEM179B (data not shown). Figure 2 also shows the expression profiles of CEACAM1, DDIT4, TRIB3, TUBB3, and OSR2 in Caco-2 cells exposed to different concentrations of the marine toxins. The CEACAM1 marker gene was up-regulated by AZA1 and DTX1 but less clear
Figure 1. Array tube images obtained on the newly developed array tube for the detection of marine toxins with Caco-2 cells exposed to a DMSO control and standards of DTX1, AZA1, and OA. Spots for GAPDH, CEACAM1, DDIT4, TUBB3, OSR2, and VASN are indicated, while the square spots represent the biotin marker controls.
by OA (figure 2b). The up-regulation of this gene by both AZA1 and DTX1 is in agreement with the outcome of the wholegenome microarray (Table 2). The DDIT4 marker gene was upregulated by AZA1, DTX1, and OA from the lowest exposure 8969
dx.doi.org/10.1021/es2011385 |Environ. Sci. Technol. 2011, 45, 8965–8973
Environmental Science & Technology
ARTICLE
Figure 2. Expression profiles obtained on the dedicated low-density DNA tube microarray for the GAPDH, CEACAM1, DDIT4, TRIB3, TUBB3, OSR2, and VASN marker genes in Caco-2 cells exposed to different concentrations of marine toxins. Each sample concentration was performed in triplicate and the bars show the mean intensity of the spot as measured by the tube reader (blue precipitation product) and its standard deviation. 8970
dx.doi.org/10.1021/es2011385 |Environ. Sci. Technol. 2011, 45, 8965–8973
Environmental Science & Technology
ARTICLE
Figure 3. Normalized fold expression of the DDIT4, RGS16, and NPPB marker genes as determined by qRT-PCR of Caco-2 cells exposed to 6.25 and 25 nM of DTX1 or AZA1. Each sample concentration was performed in triplicate and the bars show the fold expression and its standard deviation.
dose onward (Figure 2c). The TRIB3 marker gene was mainly up-regulated by AZA1 and to a much lesser extend by DTX1 (Figure 2d), while TUBB3 was clearly up-regulated by DTX1 and OA but to a much lesser extent by AZA1 (figure 2e). The OSR2 marker gene was not affected by AZA1 but was clearly upregulated by DTX (figure 2f). The array tube results for DDIT4, TRIB3, TUBB3, and OSR2 are also in good agreement with the outcome of the whole-genome microarray (these 5 marker genes are all shown in bold in Table 2). Moreover, Figure 2 shows that the slightly cytotoxic concentration of 30 nM DTX1 inhibited the induction of TRIB3 but not the other genes, while the clear cytotoxic concentration of 100 nM DTX1 also had an inhibiting effect on the expression of CEACAM1, TUBB3, and OSR2. To determine whether the spot intensities on the array tube could be improved for the twelve marker genes that did not reveal spot intensities above background, 2 rounds of linear cDNA amplification were performed instead of one. This gave a slight improvement but only for VASN and TNS4 (data not shown). VASN is a gene up-regulated by DTX1 and OA only (Figure 2g) and TNS4 is a gene that is up-regulated by both AZA1 and DTX1 (Table 2). qRT-PCR Results. Three genes were selected for qRT-PCR analysis, that is, DDIT4, RGS16, and NPPB (underlined in Table 2). DDIT4 was selected as it gave clear spots on the dedicated array tube and RGS16 and NPPB were selected as no spot intensities above the background were obtained for these markers. Figure 3 shows the DDIT4, RGS16, and NPPB qRTPCR results for cells exposed to AZA1 or DTX1. As observed in the microarrays, DDIT4 was clearly up-regulated by AZA1 and DTX1 and both already at the low dose of 6.25 nM, although this was more pronounced for AZA1 than for DTX1. The RGS16 gene was only up-regulated by DTX1 and not by AZA1, whereas NPPB was down-regulated by DTX1 and not by AZA1. These results are completely in line with the whole genome microarray data (underlined in Table 2).
’ DISCUSSION Bioassays can be used as early warning systems for new emerging risks and have already shown their ability to support selection, purification, and identification of unknown compounds in so-called bioassay-directed identification approaches.33,34 Moreover, the combination of receptor- and omics-based assays can also be
helpful to characterize modes of action of newly detected contaminants, for example, as in the case of 2-isopropylthioxanthone. 40 These kind of studies show that chemicalanalytical, bioassay, and -omics-based analyses are complementary and offer the possibility to detect and identify as yet unknown toxic compounds and contaminants. Marine toxins are a problem area, where such an approach is urgently needed. In this study, a tailored microarray platform was designed for the detection of marine toxins based on the differential expression of a set of genes. For this assay an array tube test format was chosen that was previously used for detection of T2/HT2 mycotoxins and estrogens. Exposing Caco-2 cells to different marine toxins resulted in a clear fingerprint on the newly developed dedicated low-density microarray, despite the fact that only 5 out of the 17 selected marker genes gave clear signals. The two control genes were highly expressed, gave clear signals and were not affected by any toxin treatment. The fingerprint of the 5 markers agreed with the whole genome microarray data that were used for the selection of these marker genes: CEACAM1, DDIT4, and TUBB3 were up-regulated by AZA1, DTX1, and OA, TRIB3 was up-regulated by AZA1 only, and OSR2 by DTX1 only. Twelve marker genes gave no detectable signals on the microarray. For the AK091132 gene, it transpired that the forward and backward primers were not properly designed (no PCR product possible). For the other eleven selected markers, the primers and hbprobes or the linear amplification procedure to obtain the labeled PCR product might need further optimization, for example, by using longer hbprobes. The expression level of the genes is very likely another factor that determines the success of detection with the low-density microarray. None of the spots with an intensity below 2.8 times the median intensity of all spots on the whole genome microarray were detected by the dedicated tube array, that is, RGS16, MAFB, LOC387763, TFEC, TMCC1, and TNS4. Moreover, only the up-regulated genes gave clear signals. All genes that were down-regulated, that is, C21orf129 and TFEC by both AZA1 and DTX1, C3orf57 and CDKN1C by AZA1 and NPPB by DTX1, could not be detected by the dedicated array. This left VASN, because it displayed an intensity of 55 on the whole genome microarray and was clearly up-regulated by DTX1 (Table2) but gave no signal on the dedicated array. Figure 2g shows the expression profile of this marker gene and reveals that it was only slightly responsive to 8971
dx.doi.org/10.1021/es2011385 |Environ. Sci. Technol. 2011, 45, 8965–8973
Environmental Science & Technology DTX1 and less so to OA. Performing 2 rounds of linear cDNA amplification instead of one, only gave a slight improvement for VASN and TNS4. An alternative detection method was therefore setup, qRTPCR analysis. This analysis confirmed that DDIT4 was clearly up-regulated by both AZA1 and DTX1, that RGS16 was upregulated by DTX1 only and that NPPB was down-regulated by DTX1 only. These results were completely in line with the whole genome microarray data, but the up-regulation of RGS16 and the down-regulation of NPPB are not detected by the newly developed tailored microarray. Again, this suggests that either the sensitivity of the tube array format is limited or that the primers and hbprobes in combination with the linear amplification procedure to obtain the labeled PCR product need further optimization. Regarding OA, this toxin is known to elicit effects similar to that of its structure analogue DTX1, and in combination, they act additively.5,14 This is in line with the overall data generated, as these show that OA elicits a pattern on the dedicated array that was more similar to that of DTX1 than that of AZA1. For this reason AZA-like toxicity is often classified as a separate class, although they do show diarrhetic effects in animals and humans. Presently, there are no mode of action based in vitro bioassays for marine toxins that offer the advantage over analytical methods for being capable to detect all, both known and unknown, toxins with a similar mode of action simultaneously. The new dedicated low-density tube microarray for marine toxins requires further improvement. The multiplex PCR targeting 17 marker genes and 2 controls is probably very complicated and might need further optimization while hybridization might be improved by using longer hbprobes. However, the results also showed that the outcomes with these 5 marker genes were comparable with the whole genome microarray data and resulted in clear fingerprints of AZA1, DTX1, and OA. This tube array has the advantage that it can be performed at each lab relatively cheaply: the tube array reader is the only equipment that has to be bought. The tube array, however, also has some disadvantages: the procedure is still labor intensive, and it takes about three days from exposing the cells to shellfish extracts until the evaluation of the tube array results. For testing sea food, results should be provided preferably within 24 h, comparable to the presently used in vivo animal tests, which take 16 to 24 h.5,23 In addition the variation on the results seems quite large (Figure 2). However, the current microarray data enable future developments of other fast in vitro bioassays. One option is the development of a multiplex qRT-PCR. This seems very promising, as the singleplex qRT-PCR analysis targeting the DDIT4, RGS16, and NPPB genes already resulted in a specific pattern for both DTX1 and AZA1 (Figure 3). Other possibilities are the development of molecular beacons targeting one of the marker mRNAs in intact living Caco-2 cells or the development of reporter assays. The promoter region of the DDIT4 marker gene for instance, might be used for the development of such a reporter assay, as DDIT4 was already upregulated by DTX1, AZA1, and OA at the lowest concentrations tested. The current data will thus be helpful to develop methods that can be used for high-throughput quality control of seafood and replacement of the current unethical in vivo tests. The benefits are clear: detection might no longer be dependent on a priori knowledge of the type and chemistry of marine toxins involved, and the output will have a relevance on the bioeffect level.
ARTICLE
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected]. Tel: +31(0)317480391. Fax: +31(0)317417717.
’ ACKNOWLEDGMENT This work was supported by the Dutch Ministry of Economic Affairs, Agriculture and Innovation (project number WOT-02001-035) and the European Commission, within the sixth Framework “BioCop” project (contract FOOD-CT-2004-06988; www. biocop.org). The authors would like to thank K. Lancova from the University of Z€urich for visiting RIKILT and her help with the software and Arjen Gerssen from RIKILT for the marine toxin standards. ’ REFERENCES (1) Daranas, A. H.; Norte, M.; Fernandez, J. J. Toxic marine microalgae. Toxicon 2001, 39, 1101–1132. (2) Dominguez, H. J.; Paz, B.; Daranas, A. H.; Norte, M.; Franco, J. M.; Fernandez, J. J. Dinoflagellate polyether within the yessotoxin, pectenotoxin, and okadaic acid toxin groups: Characterization, analysis, and human health implications. Toxicon 2010, 56, 191–217. (3) Furey, A.; O’Doherty, S.; O’Callaghan, K.; Lehane, M.; James, K. J. Azaspiracid poisoning (AZP) toxins in shellfish: Toxicological and health considerations. Toxicon 2006, 56, 173–190. (4) Toyofuku, H. Joint FAO/WHO/IOC activities to provide scientific advice on marine biotoxins (research report). Mar. Pollut. Bull. 2006, 52, 1735–1745. (5) EFSA Panel on Contaminants in the Food Chain (CONTAM). Opinion of the Scientific Panel on Contaminants in the Food chain on a request from the European Commission on marine biotoxins in shellfish—Okadaic acid and analogues. EFSA J. 2008, 589, 162. (6) EFSA Panel on Contaminants in the Food Chain (CONTAM). Opinion of the Scientific Panel on Contaminants in the Food chain on a request from the European Commission on marine biotoxins in shellfish— Azaspiracids. EFSA J. 2008, 723, 152. (7) EFSA Panel on Contaminants in the Food Chain (CONTAM). Opinion of the Scientific Panel on Contaminants in the Food chain on a request from the European Commission on marine biotoxins in shellfish—Yessotoxin group. EFSA J. 2008, 907, 162. (8) EFSA Panel on Contaminants in the Food Chain (CONTAM). Scientific Opinion of the Panel on Contaminants in the Food Chain on a request from the European Commission on Marine Biotoxins in Shellfish—Summary on regulated marine biotoxins. EFSA J. 2009, 1306, 123. (9) EFSA Panel on Contaminants in the Food Chain (CONTAM). Scientific Opinion of the Panel on Contaminants in the Food Chain on a request from the European Commission on marine biotoxins in shellfish—Domoic acid. . 2009, 1181, 161. (10) EFSA Panel on Contaminants in the Food Chain (CONTAM). Scientific Opinion on marine biotoxins in shellfish—Palytoxin group. EFSA J. 2009, 1393, 138. (11) EFSA Panel on Contaminants in the Food Chain (CONTAM). Scientific Opinion on marine biotoxins in shellfish—Emerging toxins: Ciguatoxin group. EFSA J. 2010, 1627, 138. (12) EFSA Panel on Contaminants in the Food Chain (CONTAM). Scientific Opinion on marine biotoxins in shellfish—Cyclic imines (spirolides, gymnodimines, pinnatoxins, and pteriatoxins). EFSA J. 2010, 1628, 139. (13) EFSA Panel on Contaminants in the Food Chain (CONTAM). Scientific Opinion on marine biotoxins in shellfish Emerging toxins: Brevetoxin group. EFSA J. 2010, 1677, 129. (14) Food and Agriculture Organization. Marine Biotoxins. FAO Food Nutr. Pap. 2004, 80. 8972
dx.doi.org/10.1021/es2011385 |Environ. Sci. Technol. 2011, 45, 8965–8973
Environmental Science & Technology (15) Moore, S. K.; Trainer, V. L.; Mantua, N. J.; Parker, M. S.; Laws, E. A.; Backer, L. C.; Fleming, L. E. Impacts of climate variability and future climate change on harmful algal blooms and human health. Environ Health 2008, 7 (Suppl. 2), S4. (16) Peperzak, L. Future increase in harmful algal blooms in the North Sea due to climate change. Water Sci. Technol. 2005, 51, 31–6. (17) Gerssen, A.; Pol-Hofstad, I. E.; Poelman, M.; Mulder, P. P. J.; Van den Top, H. J.; De Boer, J. Marine Toxins: Chemistry, toxicity, occurrence and detection, with special reference to the Dutch Situation. Toxins 2010, 2, 878–904. (18) Briand, J. F.; Jacquet, S.; Bernard, C.; Humbert, J. F. Health hazards for terrestrial vertebrates from toxic cyanobacteria in surface water ecosystems. Vet. Res. 2003, 34, 361–77. (19) Bricelj, V. M.; Connell, L.; Konoki, K.; MacQuarrie, S. P.; Scheuer, T.; Catterall, W. A.; Trainer, V. L. Sodium channel mutation leading to saxitoxin resistance in clams increases risk of PSP. Nature 2005, 434, 763–767. (20) Catterall, W. A.; Cestele, S.; Yarov-Yarovoy, V.; Yu, F. H.; Konoki, K.; Scheuer, T. Voltage-gated ion channels and gating modifier toxins. Toxicon 2007, 49, 124–141. (21) Van Apeldoorn, M. E.; Van Egmond, H. P.; Speijers, G. J.; Bakker, G. J. Toxins of cyanobacteria. Mol. Nutr. Food Res. 2007, 51, 7–60. (22) Wang, D. Z. Neurotoxins from marine dinoflagellates: a brief review. Mar. Drugs 2008, 6, 349–371. (23) Community Reference laboratory for marine biotoxins. EU harmonised standard operating procedure for detection of lipophilic toxins by mouse bioassay, version 5 June 2009. http://www.aesan.msc.es/CRLMB/ docs/docs/metodos_analiticos_de_desarrollo/EU-Harmonised-SOPMBA-Lipophilic-Version5-June2009.pdf, 2009. (24) EC Council Directive 15/2011 of January 10th 2011, replacement of EC/2074/2005/ regarding test methods for the detection of marine toxins. Off. J. Eur. Commun. (2011) L6/3L6/6. (25) Suzuki, T.; Horie, Y.; Koike, K.; Satake, M.; Oshima, Y.; Iwataki, M.; Yoshimatsu, S. Yessotoxin analogues in several strains of Protoceratium reticulatum in Japan determined by liquid chromatography-hybrid triple quadrupole/linear ion trap mass spectrometry. J. Chromatogr. A 2007, 1142, 172–177. (26) Torgersen, T.; Wilkins, A. L.; Rundberget, T.; Miles, C. O. Characterization of fatty acid esters of okadaic acid and related toxins in blue mussels (Mytilus edulis) from Norway. Rapid Commun. Mass Spectrom. 2008, 22, 1127–1136. (27) Gerssen, A.; McElhinney, M. A.; Mulder, P. P. J.; Bire, R.; Hess, P.; De Boer, J. Solid phase extraction for removal of matrix effects in lipophilic marine toxin analysis by liquid chromatography-tandem mass spectrometry. Anal. Bioanal. Chem. 2009, 394, 1213–1226. (28) Gerssen, A.; Van Olst, E. H. W.; Mulder, P. P. J.; De Boer, J. Inhouse validation of a liquid chromatography tandem mass spectrometry method for the analysis of lipophilic marine toxins in shellfish using matrix-matched calibration. Anal. Bioanal. Chem. 2010, 397, 3079–3088. (29) Botana, L. M.; Vilarino, N.; Alfonso, A.; Vale, C.; Louzao, C.; Elliott, C. T.; Campbell, K.; Botana, A. M. The problem of toxicity equivalent factors in developing alternative methods to animal bioassays for marine-toxin detection. Trends Anal. Chem. 2010, 29, 1316–1325. (30) Etheridge, S. M. Paralytic shellfish poisoning: seafood safety and human health perspectives. Toxicon 2010, 56, 108–22. (31) Paz, B.; Daranas, A. H.; Norte, M.; Riobo, P.; Franco, J. M.; Fernandez, J. J. Yessotoxins, A group of marine polyether toxins: an overview. Mar. Drugs 2008, 6, 73–102. (32) Plakas, S. M.; Dickey, R. W. Advances in monitoring and toxicity assessment of brevetoxins in molluscan shellfish. Toxicon 2010, 56, 137–149. (33) Bovee, T. F. H.; Pikkemaat, M. G. Bioactivity-based screening of antibiotics and hormones. J. Chromatogr., A 2009, 1216, 8035–8050. (34) Ede van, K.; Li, A.; Antunes-Fernandez, E.; Mulder, P.; Peijnenburg, A.; Hoogenboom, L. A. P. Bioassay directed identification of natural aryl hydrocarbon agonists in marmalade. Anal. Chim. Acta 2008, 617, 238–245.
ARTICLE
(35) Vieytes, M. R.; Fontal, Ol.; Leira, F.; deSousa, J. M. V. B.; Botana, L. M. A fluorescent microplate assay for diarrheic shellfish toxins. Anal. Biochem. 1997, 248, 258–264. (36) Leira, F.; Alvarez, C.; Cabado, A. G.; Vieitis, J. M.; Vieytes, M. R.; Botana, L. M. Development of a F actin-based live-cell fluorimetric microplate assay for diarrhetic sellfish toxins. Anal. Biochem. 2003, 317, 129–135. (37) Pierotti, S.; Albano, C.; Milandri, A.; Callegari, F.; Poletti, R.; Rossini, G. P. A slot blot procedure dor the measurement of yessotoxins by a functional assay. Toxicon 2007, 49, 36–45. (38) Vilarino, N.; Fonfria, E. S.; Molgo, J.; Araoz, R.; Botana, L. M. Detection of gymnodimine-A and 13-desmethyl C spirolide phycotoxins by fluorescence polarization. Anal. Chem. 2009, 81, 2708–2714. (39) Lancova, K.; Bowens, P.; Stroka, J.; Gmuender, H.; Ellinger, T.; Naegeli, H. Transcriptomic-based bioassays for the detection of type A trichothecenes. World Mycotoxin J. 2009, 2, 247–257. (40) Peijnenburg, A.; Riethof-Poortman, J.; Baykus, H.; Portier, L.; Bovee, T.; Hoogenboom, R. AhR-agonistic, anti-androgenic, and antiestrogenic potencies of 2-isopropylthioxanthone (ITX) as determined by in vitro bioassays and gene expression profiling. Toxicol. in Vitro 2010, 24, 1619–1628.
8973
dx.doi.org/10.1021/es2011385 |Environ. Sci. Technol. 2011, 45, 8965–8973
ARTICLE pubs.acs.org/est
Tissue-Specific Expression of p53 and ras Genes in Response to the Environmental Genotoxicant Benzo(α)pyrene in Marine Mussels Yanan Di,† Declan C. Schroeder,‡ Andrea Highfield,‡ James W. Readman,§ and Awadhesh N. Jha*,† †
School of Biomedical and Biological Sciences, University of Plymouth, Drake Circus, Plymouth, Devon PL4 8AA, U.K. Marine Biological Association of the United Kingdom (MBA), Citadel Hill, Plymouth PL1 2PB, U.K. § Plymouth Marine Laboratory (PML), Prospect Place, The Hoe, Plymouth PL1 3DH, U.K. ‡
bS Supporting Information ABSTRACT: Marine mussels can develop hemeic and gonadal neoplasia in the natural environment. Associated with these diseases are the tumor suppressor (TS) p53 and the proto-oncogene ras coded proteins, both of which are highly conserved among molluscs and vertebrates. We report, for the first time, tissuespecific expression analysis of p53 and ras genes in Mytilus edulis by means of quantitative RT-PCR. A tissue-specific response was observed after 6 and 12 days exposure to a sublethal concentration of a model Polycyclic Aromatic Hydrocarbon (PAH), benzo(α)pyrene (B(α)P). This sublethal concentration (56 μg/L) was selected based on an integrated biomarker analysis carried out prior to gene expression analysis, which included a ‘clearance rate’ assay, histopathological analysis, and DNA strand break measurements. The results indicated that the selected concentration of B(α)P can lead to the induction of DNA strand breaks, tissue damage, and expression of tumor-regulating genes. Both p53 and ras are expressed in a tissue-specific manner, which collaborate with tissue-specific function in response to genotoxic stress. The integrated biological responses in Mytilus edulis strengthen the use of this organism to investigate the fundamental mechanism of development of malignancy in invertebrate which could be translated to other organisms including humans.
’ INTRODUCTION Mutational inactivation of tumor suppressor (TS) genes and activation of oncogenes are two of the most frequently observed and thoroughly studied molecular pathways in human cancer research.1 p53, as a representative TS gene, serves as the major cellular barrier against cancer development. DNA damage or cellular stress can activate the p53 protein via the orchestrated action of multiple post-translational modifications which either increase the stability of p53 or directly enhance its DNA-binding affinity, leading to the arrest of cell cycle progression to allow DNA repair.2 The ras proto-oncogene encodes low molecular weight guanine nucleotide-binding proteins that cycle between inactive GDP-bound and active GTP-bound forms.3 Activated ras oncogenes are frequently detected in human and other animal tumors with point mutations occurring within exons 1 and 2: codons 12, 13, and 61.4 When this gene suffers a mutation in one of the above-mentioned codons, the encoded protein inhibits GTPase activity, leading to altered cell proliferation, differentiation, and cell cycle checkpoint control.5 Recent studies concerning the development of carcinogenesis in mammals and fish reported mutant forms of ras in a large proportion and wide variety of tumors.6 Due to the worldwide distribution and ecological habitat, mussels are extensively used for environmental studies7 and could serve as a model organism for elucidation of neoplastic disorders as leukemia (hemeic neoplasia) and gonadal neoplasia r 2011 American Chemical Society
have been reported in the blue mussel Mytilus sp. around the world.8 In this context, occurrence of malignancies in bivalve molluscs has also been correlated with higher incidence of tumors in human populations who were exposed to increased use of herbicides.9 Defining common mechanisms of tumor formation in mussels, therefore, could provide valuable information for human cancer research. Furthermore, both the gene sequences for p53 and ras have recently become available for Mytilus edulis and show regions of highly conserved sequences across different groups of organisms, including humans.10 Therefore, a study to elucidate expression of tumor-regulating genes in mussels is both timely and warranted, especially when seeking to develop a nonhuman model to explore the underlying mechanisms of neoplasia. Benzo(α)pyrene (B(α)P), a representative polycyclic aromatic hydrocarbon (PAH), is the most extensively studied genotoxicant and carcinogen in aquatic organisms.11 Various groups of aquatic organisms collected from PAH contaminated sites, including the European flounder Platichthys flesu, the oyster Crassostrea virginica, and the marine mussel Mytilus edulis, have been reported to have a greater incidence of neoplastic disease Received: May 6, 2011 Accepted: September 7, 2011 Revised: September 2, 2011 Published: September 07, 2011 8974
dx.doi.org/10.1021/es201547x | Environ. Sci. Technol. 2011, 45, 8974–8981
Environmental Science & Technology
ARTICLE
Figure 1. A: Clearance rates of Isochrysis galbana in Mytilus edulis following 6 and 12 days B(α)P exposure under in vivo condition. Error bars present SEM. B: Induction of DNA strand break (represented as % Tail DNA) in Mytilus edulis. hemolymph following 6 and 12 days in vivo B(α)P exposure. * indicates significant increase of % Tail DNA in treated groups compared with control group (p < 0.05). # indicates significant differences of % Tail DNA between day 6 and day 12 within each treated groups (p < 0.05).
compared to samples collected from uncontaminated sites.1214 This study, therefore, aims to elucidate the gene expression profiles of p53 and ras in M. edulis following exposure to B(α)P. An integrated approach with biomarkers at different levels of biological organization were analyzed prior to gene expression analysis. Gene expression was analyzed in different tissues based on previous studies where neoplasia has been found with different susceptibilities among humans15 and other marine bivalve tissues.16 To our knowledge, this is the first study to evaluate tissue-specific expression profile of tumor-regulating genes (p53 and ras) in the blue mussel M. edulis. Attempts have also been made to link biomarker responses at different levels of biological organization following B(α)P exposure.
’ MATERIAL AND METHODS Mussels Sampling and B(α)P Exposure Setups. Mussels were collected from Trebarwith Strand (North Cornwall, UK,
50°380 N; 4°450 W) and were immediately transported to the laboratory where they were maintained in the aerated tanks with filtered (<10 μm) seawater at 15 °C. Mussels were fed daily with the microalga Isochrysis galbana (Liquifry, Interpet, Dorking, UK), and water was changed daily for at least 14 days for acclimatization prior to the experiment. In no cases were spawning animals used in the experiments. B(α)P was first dissolved in acetone which had previously been confirmed as suitable carrier solvent (data not shown). This was then added to seawater to achieve a final nominal B(α)P concentration of 5.6, 56, and 100 μg/L and a final acetone concentration of 0.01% (v/v). Mussels were placed in the exposure vessels (in triplicate) which were individually aerated using air stones and covered with glass to avoid evaporation of B(α)P. Control treatments were carried out under similar conditions except for the introduction of 0.01% (v/v) acetone only. Triplicate mussels were sampled from each tank after 6 and 12 days exposure for analysis. 8975
dx.doi.org/10.1021/es201547x |Environ. Sci. Technol. 2011, 45, 8974–8981
Environmental Science & Technology
ARTICLE
Determination of B(α)P Concentration in Seawater Using Gas ChromatographyMass Spectrometry (GC-MS) Analysis.
Water samples (9 cm3) were collected into glass vials and dichloromethane (1 cm3, HPLC grade, Rathburn Chemicals Ltd., UK) was added. Phenanthrene d10 (1.1 μg in 10 μL of dichloromethane) was then added as an internal standard. Following thorough shaking, the mixtures were stored in the dark at 4 °C. Immediately prior to analysis, dichloromethane layers were removed into glass microvials. Aliquots of sample extracts (2 μL) were analyzed using an Agilent Technologies 6890N Network GC system interfaced with an Agilent 5973 series Mass Selective detector. B(α)P concentration was calculated based on the internal standard. Integrated Biomarker Analysis. The clearance rate was determined prior to hemolymph collection as described in detail by Canty et al. (2009).17 Individual mussels (9 mussels per treatment) were placed in separate glass beakers containing filtered seawater and allowed to acclimatize for 15 min prior to the addition of Isochrysis galbana algal suspension (20,000 algal cells/mL) in each beaker. Water samples were collected immediately, and 30 min after the introduction of algal suspension, samples were analyzed on a Beckham Coulter Particle Size and Count Analyzer (Z2). The clearance rate of the mussels was then calculated following standard equation.17 The alkaline comet assay was applied for the determination of DNA strand breaks using hemocytes of mussels as routinely used in our laboratory conditions.17,18 Cells were randomly selected for the measurement of DNA breaks after ethidium bromide staining using a Komet 5.0 software (Kinetic Imaging Liverpool, UK). A total of 100 cells were scored for each sample. The tail DNA% was applied to assess the degree of DNA strand break. Tissue samples of adductor muscle, gill, digestive gland, and mantle were dissected for histopathological analysis. Tissues were preserved in 10% neutral, buffered formalin to fix for at least 48 h. The preserved tissues were then processed by a routine histological method.19 Paraffin embedded tissues were sectioned to a thickness of 5 μm and placed on glass slides to be stained with hematoxylin and eosin (H and E). Slides were examined by an Olympus Vanox-T light microscope and photographed using a digital camera (Olympus camedia C-2020 Z) at magnification of 200 and 400. Isolation of Total RNA, Synthesis of cDNA, and Quantitative Real-Time PCR. Partial digestive gland, adductor muscle, gill and mantle tissues were dissected from the individual mussels and preserved in RNAlater before further processing (Qiagen, Sussex, UK). Total RNA was isolated from the tissues using the RNeasy Mini Kit (Qiagen Sussex, UK) and 10 ng of RNA was used for reverse transcription using SuperScript II reverse transcriptase kit and Oligo (dT) primers (Invitrogen, Paisley, UK) following manufacture’s instruction except where stated otherwise. After cDNA synthesis, quantitative real-time PCR using SensiMix NoRef Kit (Quantace) and Roter-Gene 6000 real-time qPCR system (Qiagen) was applied for the analysis of p53 and ras genes expression in different tissues.
’ RESULTS Chemical Analysis Using GC-MS. Known amounts of B(α)P (dissolved in acetone) were injected into the seawater systems to provide nominal concentrations of 5.6, 56, and 100 μg/L. Samples of seawater were collected immediately after the addition and after 6 and 12 days. They were extracted into dichloromethane
Figure 2. Light micrographs of sections through tissues of Mytilus edulis showing histological structure of control and treated mussels stained with H and E. A-E are untreated controls; a-e are exposed to B(α)P. A and a: digestive gland; B and b: gill; C and c: adductor muscle; D and d: mantle (male); E and e: mantle (female), where A-C are 400 magnification and D-E are 200. dt = digestive tubule; ct = connective tissues; at = atrophy; fc = frontal cilia; lc = lateral cilia; gf = gill filaments; amb = adductor muscle block; mgt = male gonad tubule; fgt = female gonad tubule; vct = vesicular connective tissue. Scale bar = 25 μm.
and analyzed by GC-MS. Immediately after addition, concentrations of 49 ( 15 and 92 ( 3 μg/L were recorded for the 56 and 100 μg/L nominal concentrations. For all other samples, levels were below limits of detection of the selected analytical technique (<0.3 μg/L). Integrated Biomarker Responses to B(α)P at Different Levels. Clearance rates as an indicator of physiological response to indicate feeding behavior of mussels remained unaffected by any of the exposure regimes and none of the exposed mussels fed differently to the untreated control mussels. The lowest clearance rates were obtained for the 56 μg/L B(α)P exposure (1.23 L/h and 0.86 L/h after 6 and 12 days exposure, respectively), suggesting individual mussels reacted more promptly with the midrange B(α)P exposure concentration. Following 12 days exposure, 30% decline in clearance rate was detected compared to day 6 for all exposure conditions, including the untreated control (Figure 1A). For DNA strand breaks in hemocytes, after 6 days exposure, maximum effect was seen at the highest exposure concentration (i.e., 100 μg/L) with 67.86 ( 1.26% tail DNA (Figure 1B). A significant concentration-dependent increase for DNA strand 8976
dx.doi.org/10.1021/es201547x |Environ. Sci. Technol. 2011, 45, 8974–8981
Environmental Science & Technology
ARTICLE
Figure 3. Constitutive p53 (A) and ras (B) genes expression pattern in various tissues under untreated conditions and relative quantitative p53 (C) and ras (D) gene expression pattern in various tissues exposed to B(α)P at 56 μg/L for 6 and 12 days. The gene transcript levels were semiquantified in tissues using real-time PCR and expressed relative to actin level. Each histogram represents the means of 6 replicates examination (n = 6), and SEM were indicated by T-bars. Means with the same letter indicate no significant difference based on ANOVA followed by Tukey’s test at p = 0.05. The relative expressions were presented as 2ΔΔCt. * presents significantly induced gene expression.
breaks was observed compared to the untreated control (p < 0.05). As with 6 days exposure, a concentration-dependent increase of DNA strand break was obtained after 12 days exposure. There was a significant decrease in DNA strand breaks in the treated groups after 12 days exposure compared to 6 days at all three concentrations. On the contrary, no significant difference was observed in the untreated control group for both exposure periods. Histological observations indicated no pathological signs of disease in untreated specimens (e.g., hemocyte infiltration, necrosis, or other injuries). In contrast, pathological changes in B(α)P exposed mussels were identified (Figure 2). The digestive gland showed changes in digestive tubule structure, as featured by diffused nuclei and lack of distinction in some epithelial cells. Exposed gills had swollen filaments filled with hemocytes, with the frontal cilia of the epithelial border absent (i.e., filament necrosis). The adductor muscle showed histological abnormalities with loss of muscle bundle structure, increased intracellular spaces, and decreased extracellular spaces in connective tissue. Damage was also identified in treated mantle tissue which showed decreased adipogranular (ADG) and vesicular connective tissue (VCT) cells in the connective tissue as well as loss of gonadal duct structure. Interestingly, histopathological changes were not observed in all B(α)P treated samples (n = 6), and no doseresponse relationship could be identified from the prevalence of histopathological changes. Only selected representative images of control and damaged tissues have been presented
in Figure 2. The prevalence of abnormal tissues after B(α)P exposure was scored across the total number of collected samples (details in the Supporting Information). p53 and ras Expression in Different Tissues. p53 and ras expression abundance for untreated samples were presented as ΔCt (difference of Ct value between target gene, p53/ras, and reference gene, actin) (Figure 3A-B). A higher ΔCt value represents later amplification of the target gene and therefore lower expression. p53 transcription showed a tissue-specific expression pattern. The highest expression was seen in the gill tissue with the lowest ΔCt value (2.02 ( 0.48 and 1.50 ( 0.56 after 6 and 12 days incubation, respectively), followed by the digestive gland, mantle, and adductor muscle in all untreated samples for both the incubation periods. A significantly lower p53 expression level was seen in the adductor muscle compared to other three tissues (p < 0.05). Mantle tissue also showed significantly different expression with higher levels compared to adductor muscle and lower compared to the digestive gland and gill tissues. Gill tissue showed the highest expression of p53 which was significantly different from other tissues apart from the digestive gland (Figure 3A). p53 expression did not change as a function of incubation periods in gill, digestive gland, and mantle. In contrast, significantly higher p53 expression was detected in the adductor muscle following 12 days incubation (ΔCt = 6.15 ( 1.69) compared to 6 days (ΔCt = 10.48 ( 0.75). A similar tissue-specific expression abundance was obtained for ras expression with the trend of gill > digestive gland > mantle > adductor muscle (Figure 3B). 8977
dx.doi.org/10.1021/es201547x |Environ. Sci. Technol. 2011, 45, 8974–8981
Environmental Science & Technology Significantly lower ras expression was detected in the adductor muscle and mantle. The expression level of ras showed no significant difference in all the tissues for both incubation periods in untreated control conditions except for adductor muscle which showed significantly increased ras expression abundance after 12 days incubation. Relative Quantification of p53 and ras Expression in Tissues Induced by B(α)P. Tissues from mussels exposed to 56 μg/L B(α)P were selected for gene expression analysis as this concentration induced the highest response for feeding behavior (i.e., clearance rate) and significant DNA strand breaks at cellular level. The relative quantification of p53 and ras expressions were normalized in different tissues by the 2ΔΔCt method20 (Figure 3C-D). Wide interindividual variation of p53 expression was detected in all tissues. After 6 days exposure, a significant increase in p53 expression was detected in the adductor muscle, approximately 2.34 ( 0.89-fold higher than the control. A small but insignificant decrease in p53 expression was found in the other three tissues. After 12 days exposure, significantly increased p53 expression was detected in both the mantle and adductor muscle with 5.93 ( 0.83- and 3.28 ( 1.23-fold higher expression levels, respectively. Lower induction of p53 expression in gill and digestive gland was observed, but it was not significantly different compared to the control. p53 expression also increased with longer B(α)P exposure period. Approximately 6-fold higher p53 expression was detected in the mantle after 12 days exposure compared to 6 days. The other three tissues also showed a slightly increased p53 expression after a longer exposure time, but the differences were not statistically significant (Figure 3C). There were no significant differences in ras expression in gill, digestive gland, and mantle after 6 days exposure (Figure 3D). A significant 2.29 ( 1.01-fold higher ras expression was only seen in the adductor muscle. After 12 days exposure, the mantle showed significantly (2.18 ( 1.14-fold) higher ras expression, but the adductor muscle and digestive gland showed no significant increase in expression. The gill tissue showed a slight decrease in ras expression but was not statistically significant. Increased ras expression was found in all tissues apart from the adductor muscle after longer B(α)P exposure. Significantly increased ras expression, approximately 2.22-fold, was found in the mantle after 12 compared to 6 days exposure. On the contrary, no significant difference in ras expression was seen in the digestive gland and gill after 12 days compared to 6 days exposure. Slightly decreased expression was seen in the adductor muscle, again indicating tissue specific expression.
’ DISCUSSION The concentration ranges selected in this study have previously shown to induce cellular responses in mussels.21,22 They do, however, exceed the documented solubility of the compound (2.3 μg/L).23 The presence of dissolved organic matter has been shown to enhance solubility 24 and owing to the high Kow (6.04),25 the B(α)P rapidly partitions onto particulates and the mussels themselves. This is consistent with the comparatively low GC-MS measured concentrations, their variability, and the rapid depletion of the B(α)P from the aqueous phase. In the higher B(α)P exposure concentrations (56 and 100 μg/L), there were no observed differences in clearance rates between untreated control and exposed mussels. This suggests that mussels are relatively tolerant to these types and concentration range of organic chemicals and can metabolize them without causing
ARTICLE
marked physiological changes at individual level.17 It is reported that blue mussels exposed to 1 μg/L B(α)P for 6 h, followed by 6 h in clean seawater, had tissue concentrations 7080% of the initial B(α)P concentration in the water,26 suggesting the ability of mussels to accumulate B(α)P in the tissues. However, the clearance rates of mussels after 12 days exposure were significantly lower than 6 days, suggesting other conditions, for example starving stress, rather than direct B(α)P-induced stress may have affected the physiological condition of the mussels. Only a small decrease in clearance rates were observed at 56 μg/L B(α)P exposure. This effect may have been caused by the direct toxic effect of B(α)P as it has been shown to induce blood cell lysosomal membrane damage in mussels.27 Histopathological results carried out on 6 mussels showed changes in the mussel tissues; however, the prevalence of these changes was not concentration-dependent. This sample size was perhaps not large enough to confirm concentration-dependent relationship.28 No evidence of neoplasia was found in any of tissues sampled. Only necrosis was observed in some of the exposed tissues, which could be considered as very early stage of tumor development.29 Neoplasia in invertebrate species is relatively rare compared to incidences reported for vertebrates.30 Attempts to induce neoplasia in vivo in mussels has as of yet been unsuccessful suggesting that its etiology in invertebrates under laboratory conditions potentially involves factors other than exposure to the selected carcinogenic chemicals alone.16 Nonetheless, the histopathological results provide fundamental evidence of the physiological effects of B(α)P that could lead to disease development, such as neoplasis and immune suppression.31 Significant increases in DNA strand breaks were also observed with increasing B(α)P concentrations, demonstrating the capability of B(α)P to induce a genotoxic response. Interestingly, the level of DNA damage showed a significant decrease in DNA strand breaks after 12 days exposure. This suggests the potential involvement of DNA repair, recombination, and cell cycle checking mechanisms32 during the process of mussels responding to the B(α)P exposure. Excessive damaged cells could be lost through apoptotic pathway leading to apparent reduction in the level of DNA damage.33 Toxicant concentrations via dosage or exposure period is required to reach a threshold level before DNA repair systems are initiated.34 Since the in vivo control values were relatively high (30% of Tail DNA) to begin with compared to an in vitro study using H2O2,35 it is likely that the mussels used in our study were already stressed and, therefore, were more sensitive to B(α)P lowering their toxicity threshold. Although both DNA strand breaks and histopathological effects suggest that B(α)P can indeed cause stress in mussels, gene expression studies on key neoplasia associated genes could initiate the interrelated process of mutagenesis and carcinogenesis through common pathways seen in other organisms.36 In normal conditions (i.e., without any induced stress), p53 and ras expressions showed a tissue-specific expression pattern. The highest expression abundance was obtained in gill followed by digestive gland, mantle and adductor muscle tissues. The gill and digestive gland of bivalve molluscs play an important role in food collection, absorption, and digestion,28 and cells in these two tissues are likely to be more susceptible with stresses. Elevated expression of p53 has been shown in mammalian studies where brain was treated by adrenalectomy and excitotoxic treatment.24,37 Increased expression has also been shown with apoptosis of antral ovarian follicles in rats.38 Moreover, a significant 8978
dx.doi.org/10.1021/es201547x |Environ. Sci. Technol. 2011, 45, 8974–8981
Environmental Science & Technology up-regulation of p53 expression in neoplastic hemocytes compared to normal hemocytes has also been reported in Mytilus trossulus.39 All these results confirm that p53 expression changes can be an indicator of cells responding to stress. Interestingly, in our study not all the tissues showed increased p53 and ras expression after exposure to B(α)P. Only the adductor muscle for both exposure periods and the mantle after 12 days exposure showed obvious increased expression for both the genes (i.e., p53 and ras). The adductor muscle is the main tissue to preserve hemolymph and consists of less cell types compared to other tissues. Mantle, as the main tissue to produce sperm, has been found to show gonadal neoplasia (solid tumor) in mussels collected from polluted sites,40 but no evidence of neoplasia occurred in other tissues in the same individuals. These results support that mantle cells or sperm produced during spermatogenesis could be more sensitive to chemicals. A physiologically based pharmacokinetic research on eastern oyster (Crassostrea virginica) found tissue-specific bioaccumulation of dioxin after 28-days post exposure to 2,3,7,8-tetrachlorodibenzo-F-dioxin (TCDD), suggesting the toxico-kinetics of chemicals are variant in tissues.41 Using fish as models for environmental carcinogenesis, it has been reported that different tissues respond differently to carcinogens generating different types of tumor.42 In common with other organisms, it is important to elucidate the metabolic kinetics of B(α)P in mussel tissues to link the tissue-specific expressions of tumor related genes. The high interindividual variation within our samples is likely to be caused by intrinsic and extrinsic factors.7,43,44 Other physiological factors such as feeding and reproduction could also affect the pollutant uptake by the organisms. As a consequence, the body burden of organic pollutants in individuals inhabiting the same site as well as the resultant metabolic and antioxidative responses may vary considerably.45 Therefore, B(α)P uptake and accumulation in different tissues will also show the interindividual variability. The result of longer exposures increased p53 expression which complements the finding of decreased DNA damage after 12 days exposure, indicating potential involvement of DNA repair mechanisms. This is supported by the fact that both p53 and ras function in association with DNA repair and apoptosis processes.46,47 Increased p53 expression suggests that the concentration of 56 μg/L of B(α)P can induce the p53 expression to halt the cell for repair. ras, as pro-oncogene, will be activated too when the cells undergo this stress. Absence of neoplastic cells in the experimental animals suggests that no up-regulation of ras expression caused the response to the ras related pathways which could lead to uncontrolled cell growth over the exposure period. Combining DNA strand break and gene expression results, B(α)P significantly induce DNA damage in hemocytes and increases expression of p53, but this was not significant for all four different tissues. It is consistent with previously documented report that B(α)P induced quantitatively different levels of DNA adducts in different tissues of M. edulis.48 Significant increases in DNA damage are not necessarily related to increased p53/ras expression in our study and is similar to research in human cells, where increased micronuclei formation had no obvious relationship with p53 expression following exposure to a range of genotoxic chemicals under in vitro condition.49 In addition, Banni et al. (2009)50 also found up-regulation of p53 in digestive gland tissue, a large down regulation in hemocytes and no modulation for p53 expression in other tissues following exposure of mussels to B(α)P and crude oil. However, a significant increase in DNA
ARTICLE
strand breaks was observed only in hemocytes. These results further suggest that gene expression in single cell types may give significantly different results in different tissues, bearing in mind that genotoxicity is a cell specific process. In addition, since different tissues have different turnover rates and the fact that B(α)P could cause cell cycle stage specific changes in gene expression,51 it is also likely that tissue-specific gene expression pattern in our study could be related to cell cycle durations. Apart from alteration of p53/ras expression reflecting their functional change in cells, the mutation and post-transcriptional modification of these two genes have also been documented in response to DNA damage.52,53 The localization of p53 in cells also plays an important role in relation to its function in apoptosis process and leukemia formation. The maintenance of tumor phenotype has been reported to require nuclear absence of p53, resulting from its localization in the cytoplasm of leukemic clam hemocytes.54 Vassilenko et al. (2010) also reported an increased number of allelic variants in the leukemic mussels Mytilus trossulus which may arise from separate somatic mutation events in hemocyte precursors or from additional p53-like gene copies in polyploidy.55 In conclusion, while B(α)P induced tissue and DNA damage confirms its action as genotoxicant, it also induces expression of tumor-regulating genes (i.e., p53 and ras) following 12 days exposure but with a high level of interindividual variation. Normal expression of p53 and ras occurred in a tissue-specific manner, reflecting normal physiological function of tissues. Concurrent induction of DNA damage and tissue-specific gene expression has been found in our study. However, the multiple sites targeted by an integrated network of signaling pathways which are highly sensitive to genotoxic stresses must be modified to yield a functional p53 and/or ras.52 The work at translational levels and related pathways of p53 and ras are required to further elucidate the mechanism. Our study goes some way toward achieving these goals and sheds light on tissue-specific expression of tumorregulating genes at whole organism level which should also complement research into human health area.
’ ASSOCIATED CONTENT
bS
Supporting Information. Detailed methods and materials, detailed prevalence of histopathological change result, detailed analysis of p53 and ras expression efficiency, Tables S1 and S2, and Figure S1. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +44 (0) 1752584633. Fax: +44 (0) 1752584605. E-mail:
[email protected].
’ ACKNOWLEDGMENT This research was supported by grants from Overseas Research Students Awards Scheme (ORSAS) and European Regional Development Fund, INTERREG IVA (Grant No. 4059). ’ REFERENCES (1) Solomon, H.; Brosh, R.; Buganim, Y.; Rotter, V. Inactivation of the p53 tumor suppressor gene and activation of the Ras oncogene: cooperative events in tumorigenesis. Disc. Med. 2010, 9 (48), 448–454. 8979
dx.doi.org/10.1021/es201547x |Environ. Sci. Technol. 2011, 45, 8974–8981
Environmental Science & Technology (2) Warshawsky, D. Envrionmental sources, carcinogenicity, mutagenicity, metabolism and DNA binding of nitrogen and sulfur heterocyclic aromatics. Environ. Carcin. Ecotoxicol. 1992, 10, 1–71. (3) Buday, L.; Downward, J. Many faces of Ras activation. Biochim. Biophys. Acta 2008, 1786 (2), 178–187. (4) Lima, I.; Peck, M. R.; Soares, M. V. M.; Guilhermino, L.; Rotchell, J. M. Ras gene in marine mussels: a molecular level response to petrochemical exposure. Mar. Pollut. bull. 2008, 56 (4), 633–40. (5) Rotchell, J. M.; Lee, J.-s.; Chipman, J. K.; Ostrander, G. K. Structure, expression and activation of fish ras genes. Aquat. Toxicol. 2001, 55, 1–21. (6) Nogueira, P. R.; Lourenc-o, J.; Mendo, S.; Rotchell, J. M. Mutation analysis of ras gene in the liver of European eel (Anguilla anguilla L.) exposed to benzo[a]pyrene. Mar. Pollut. Bull. 2006, 52 (12), 1611–1616. (7) Jha, A. N. Genotoxicological studies in aquatic organisms: an overview. Mutat. Res. 2004, 552 (12), 1–17. (8) Ciocan, C. M.; Rotchell, J. M. Conservation of Cancer Genes in the Marine Invertebrate Mytilus edulis. Environ. Sci. Technol. 2005, 39 (9), 3029–3033. (9) Van Beneden, R. J. Molecular analysis of bivalve tumors: models for environmental/genetic interactions. Environ. Health Perspect. 1994, 102 Suppl 12 (13), 81–83. (10) Rotchell, J. M.; du Corbier, F. a.; Stentiford, G. D.; Lyons, B. P.; Liddle, A. R.; Ostrander, G. K. A novel population health approach: Using fish retinoblastoma gene profiles as a surrogate for humans. CBP Part C 2009, 149 (2), 134–40. (11) Venier, P.; Canova, S. Formation of DNA adducts in the gill tissue of Mytilus galloprovincialis treated with benzo [a] pyrene. Aquat. Toxicol. 1996, 34, 119–133. (12) Lyons, B. P.; Stentiford, G. D.; Green, M.; Bignell, J.; Bateman, K.; Feist, S. W.; Goodsir, F.; Reynolds, W. J.; Thain, J. E. DNA adduct analysis and histopathological biomarkers in European flounder (Platichthys flesus) sampled from UK estuaries. Mutat. Res. 2004, 552 (12), 177–186. (13) Gardner, G. R.; Pruell, R. J.; Malcolm, A. R. Chemical induction of tumors in oysters by a mixture of aromatic and chlorinated hydrocarbons amines and metals. Mar. Environ. Res. 1992, 34, 59–63. (14) Krishnakumar, P. K.; Casillas, E.; Snider, R. G.; Kagley, A. N.; Varanasi, U. Environmental contaminants and the prevalence of hemic neoplasia (leukemia) in the common mussel(Mytilus edulis Complex) from Puget Sound washington USA. J. Invert. Pathol. 1999, 73, 135–146. (15) Greenblatt, M. S.; Bennett, W. P.; Hollstein, M.; Harris, C. C. Mutations in the p53 Tumor Suppressor Gene: Clues to Cancer Etiology and Molecular Pathogenesis. Cancer Res. 1994, 54 (15), 4855–4878. (16) Barber, B. J. Neoplastic diseases of commercially important marine bivalves. Aquat. Living Resour. 2004, 17, 449–466. (17) Canty, M. N.; Hutchinson, T. H.; Brown, R. J.; Jones, M. B.; Jha, A. N. Linking genotoxic responses with cytotoxic and behavioural or physiological consequences: differential sensitivity of echinoderms (Asterias rubens) and marine molluscs (Mytilus edulis). Aquat. Toxicol. 2009, 94 (1), 68–76. (18) Tran, D.; Moody, A. J.; Fisher, A. S.; Foulkes, M. E.; Jha, A. N. Protective effects of selenium on mercury-induced DNA damage in mussel haemocytes. Aquat. Toxicol. 2007, 84 (1), 11–18. (19) Al-Subiai, S. N.; Moody, A. J.; Mustafa, S. A.; Jha, A. N., A multiple biomarker approach to investigate the effects of copper on the marine bivalve mollusc, Mytilus edulis. Ecotoxicol Environ. Saf. 2011, In press, DOI: 10.1016/j.ecoenv.2011.07.012. (20) Livak, K. J.; Schmittgen, T. D. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 2001, 25 (4), 402–8. (21) Halldorsson, H. P.; Maurizio; Romano, C.; Svavarsson, J.; Sara, G. Immediate biomarker responses to benzo[a]pyrene in polluted and unpolluted populations of the blue mussel (Mytilus edulis L.) at highlatitudes. Environ. Int. 2008, 34 (4), 483–9. (22) Skarphethinsdottir, H.; Ericson, G.; Zuanna, L. D.; Gilek, M. Tissue differences, dose- response relationship and persistence of DNA
ARTICLE
adducts in blue mussels (Mytilus edulis L .) exposed to benzo [ a ] pyrene. Aquat. Toxicol. 2003, 62, 165–177. (23) Inc., T. Benzo [ a ] pyrene and other polycyclic aromatic hydrocarbons Definition of Polycyclic Aromatic Hydrocarbons (PAH). Public Health 1997, 78–95. (24) Schreiber, S. S.; Sakhi, S.; Dugich-Djordjevic, M. M.; Nichols, N. R. Tumor Suppressor p53 Induction and DNA Damage in Hippocampal Granule Cells after Adrenalectomy. Exp. Neurol. 1994, 130 (2), 368–376. (25) Readman, J. W.; Mantoura, R. F. C.; Rhead, M. M.; Brown, L. Aquatic distribution and heterotrophic degradation of Polycyclic Aromatic Hydrocarbons (PAH) in the Tamar Estuary. Estuarine Coastal Shelf Sci. 1982, 14 (4), 369–389. (26) Durand, F.; Peters, L. D.; Livingstone, D. R. Effect of intertidal compared to subtidal exposure on the uptake, loss and oxidative toxicity of water-born benzo [ a ] pyrene in the mantle and whole tissues of the mussel, Mytilus edulis L. Mar. Environ. Res. 2002, 54, 271–274. (27) Okay, O. S.; Donkin, P.; Peters, L. D.; Livingstone, D. R. The role of algae (Isochrysis galbana) enrichment on the bioaccumulation of benzo [ a ] pyrene and its effects on the blue mussel Mytilus edulis. Environ. Pollut. 2000, 110, 103–113. (28) Aarab, N.; Pampanin, D. M.; Naevdal, A.; Oysaed, K. B.; Gastaldi, L.; Bechmann, R. K. Histopathology alterations and histochemistry measurements in mussel, Mytilus edulis collected offshore from an aluminium smelter industry (Norway). Mar. Pollut. Bull. 2008, 57 (612), 569–74. (29) Vakkila, J.; Lotze, M. T. Inflammation and necrosis promote tumour growth. Nat. Rev. Immunol. 2004, 4 (8), 641–648. (30) Couch, J. A.; Harshbarger, J. C. Effects of carcinogenic agents on aquatic animals: an environmental and experimental overview. Environ. Carcinog. Rev. 1985, 3, 63–105. (31) Depledge, M. H. The ecotoxicological significance of genotoxicity in marine invertebrates. Mutat. Res. 1998, 399 (1), 109–22. (32) Villela, I. V.; de Oliveira, I. M.; da Silva, J.; Henriques, J. A. P. DNA damage and repair in haemolymph cells of golden mussel (Limnoperna fortunei) exposed to environmental contaminants. Mutat. Res. 2006, 605 (12), 78–86. (33) Hook, S. E.; Lee, R. F. Genotoxicant induced DNA damage and repair in early and late developmental stages of the grass shrimp Paleomonetes pugio embryo as measured by the comet assay. Aquat. Toxicol. 2004, 66 (1), 1–14. (34) Ching, E. W. K.; Siu, W. H. L.; Lam, P. K. S.; Xua, L.; Zhang, Y.; Richardson, B. J.; Wu, R. S. S. DNA Adduct Formation and DNA Strand Breaks in Green-lipped Mussels (Perna viridis) Exposed to Benzo [ a ] pyrene: Dose- and Time-Dependent Relationships. Mar. Pollut. Bull. 2001, 42 (7), 603–610. (35) Cheung, V. V.; Depledge, M. H.; Jha, A. N. An evaluation of the relative sensitivity of two marine bivalve mollusc species using the Comet assay. Mar. Environ. Res. 2006, 62, 301–305. (36) Mohn, G.; de Raat, W. Ecological significance of mutagens. Sci. Total Environ. 1993, 134, 1771–1778. (37) Sakhi, S.; Brucet, A.; Sun, N.; Toccot, G.; Baudryt, M. p53 induction is associated with neuronal damage in the central nervous system. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 7525–7529. (38) Qin, G.; Meng, Z. Sulfur Dioxide and Benzo (a) pyrene Modulates CYP1A and Tumor-Related Gene Expression in Rat Liver. Environ. Toxicol. 2009, 25, 169–179. (39) Muttray, A. F.; Schulte, P. M.; Baldwin, S. A. Invertebrate p53like mRNA isoforms are differentially expressed in mussel haemic neoplasia. Mar. Environ. Res. 2008, 66 (4), 412–21. (40) Galimany, E.; Sunila, I. Several Cases of Disseminated Neoplasia in Mussels Mytilus edulis (L.) in Western Long Island Sound. J. Shellfish Res. 2008, 27 (5), 1201–1201. (41) Wintermyer, M.; Skaidas, A.; Roy, A.; Yang, Y.-c.; Georgapoulos, P.; Burger, J.; Cooper, K. The development of a physiologically-based pharmacokinetic model using the distribution of 2,3,7,8-tetrachlorodibenzo-p-dioxin in the tissues of the eastern oyster (Crassostrea virginica). Mar. Environ. Res. 2005, 60 (2), 133–152. 8980
dx.doi.org/10.1021/es201547x |Environ. Sci. Technol. 2011, 45, 8974–8981
Environmental Science & Technology
ARTICLE
(42) Bailey, G. S.; Williams, D. E.; Hendricks, J. D. Fish models for environmental carcinogenesis: the rainbow trout. Environ. Health Perspect. 1996, 104 Suppl (March), 5–21. (43) Klaus, S.; Keijer, J. Gene Expression Profiling of Adipose Tissue: Individual, Depot-Dependent, and Sex-Dependent Variabilities. Nutrition 2004, 20, 115–120. (44) Mitchelmore, C. L.; Chipman, J. K. DNA strand breakage in aquatic organisms and the potential value of the comet assay in environmental monitoring. Mutat. Res. 1998, 399 (2), 135–147. (45) Cheung, C. C. C.; Zheng, G. J.; Li, A. M. Y.; Richardson, B. J.; Lam, P. K. S. Relationships between tissue concentrations of polycyclic aromatic hydrocarbons and antioxidative responses of marine mussels, Perna viridis. Aquat. Toxicol. 2001, 52, 189–203. (46) Shu, K.-X.; Li, B.; Wu, L.-X. The p53 network: p53 and its downstream genes. Colloids Surf. 2007, 55 (1), 10–8. (47) Rajalingam, K.; Schreck, R.; Rapp, U. R.; Albert, S. Ras oncogenes and their downstream targets. Biochim. Biophysic. Acta 2007, 1773 (8), 1177–95. (48) Large, A. T.; Shaw, J. P.; Peters, L. D.; McIntosh, A. D.; Webster, L.; Mally, A.; Chipman, J. K. Different levels of mussel (Mytilus edulis) DNA strand breaks following chronic field and acute laboratory exposure to polycyclic aromatic hydrocarbons. Mar. Environ. Res. 2002, 54 (35), 493–497. (49) Salazar, A. M.; Sordo, M.; Ostrosky-Wegman, P. Relationship between micronuclei formation and p53 induction. Mutat. Res. 2009, 672 (2), 124–128. (50) Banni, M.; Negri, A.; Rebelo, M.; Rapallo, F.; Boussetta, H.; Viarengo, A.; Dondero, F. Expression analysis of the molluscan p53 protein family mRNA in mussels (Mytilus spp.) exposed to organic contaminants. CBP Part C 2009, 149 (3), 414–8. (51) Hamouchene, H.; Arlt, V. M.; Giddings, I.; Phillips, D. H. Influence of cell cycle on responses of MCF-7 cells to benzo[a]pyrene. BMC Genomics 2011, 12 (1), 333. (52) Appella, E.; Anderson, C. W. Post-translational modifications and activation of p53 by genotoxic stresses. Eur. J. Biochem. 2001, 268, 2764–2722. (53) Patra, S. K. Ras regulation of DNA-methylation and cancer. Exp. Cell Res. 2008, 314 (6), 1193–201. (54) Bottger, S.; Jerszyk, E.; Low, B.; Walker, C. Genotoxic StressInduced Expression of p53 and Apoptosis in Leukemic Clam Hemocytes with Cytoplasmically Sequestered p53. Cancer Res. 2008, 68 (3), 777–782. (55) Vassilenko, E. I.; Muttray, A. F.; Schulte, P. M.; Baldwin, S. A. Variations in p53-like cDNA sequence are correlated with mussel haemic neoplasia: A potential molecular-level tool for biomonitoring. Mutat. Res. 2010, 701 (2), 145–152.
8981
dx.doi.org/10.1021/es201547x |Environ. Sci. Technol. 2011, 45, 8974–8981
ARTICLE pubs.acs.org/est
Laser Ablation ICP-MS Co-Localization of Mercury and Immune Response in Fish Benjamin D. Barst,† Amanda K. Gevertz,† Matthew M. Chumchal,‡ James D. Smith,† Thomas R. Rainwater,§ Paul E. Drevnick,|| Karista E. Hudelson,† Aaron Hart,^ Guido F. Verbeck,^ and Aaron P. Roberts†,* †
)
Department of Biological Sciences & Institute of Applied Sciences, University of North Texas, 1155 Union Circle No. 310559 Denton, Texas 76203, United States ‡ Department of Biology, Texas Christian University, 2800 South University Drive, Fort Worth, Texas 76129, United States § Department of Obstetrics and Gynecology, Medical University of South Carolina, Hollings Marine Laboratory, 331 Fort Johnson Road, Charleston, South Carolina 29412, United States INRS-ETE, Universite du Quebec, 490 de la Couronne, Quebec, QC G1K 9A9, Canada ^ Department of Chemistry & Laboratory of Imaging Mass Spectrometry, University of North Texas, 1155 Union Circle No. 305070 Denton, Texas 76203, United States
bS Supporting Information ABSTRACT: Mercury (Hg) contamination is a global issue with implications for both ecosystem and human health. In this study, we use a new approach to link Hg exposure to health effects in spotted gar (Lepisosteus oculatus) from Caddo Lake (TX/LA). Previous field studies have reported elevated incidences of macrophage centers in liver, kidney, and spleen of fish with high concentrations of Hg. Macrophage centers are aggregates of specialized white blood cells that form as an immune response to tissue damage, and are considered a general biomarker of contaminant toxicity. We found elevated incidences of macrophage centers in liver of spotted gar and used a new technology for ecotoxicology studies, laser ablation-inductively coupled plasma-mass spectrometry (LA-ICPMS), to colocalize aggregates and Hg deposits within the tissue architecture. We conclude that Hg compromises the health of spotted gar in our study and, perhaps, other fish exposed to elevated concentrations of Hg.
’ INTRODUCTION Since the industrial revolution, anthropogenic releases of Hg have contaminated ecosystems worldwide. Because of its long residence time in the atmosphere, Hg can be transported far from sources and deposited in even the most remote ecosystems.1 In aquatic ecosystems, the methylation of Hg and its subsequent biomagnification in food webs results in high concentrations of Hg in fish. Hg poses health risks to piscivorous wildlife and humans as well as fish themselves.2,3 In locations solely impacted by atmospheric deposition—constituting the great majority of waterbodies—concentrations of Hg (in axial muscle) in fishes are often less than 2 μg/g wet wt.4 Laboratory studies indicate that whole-body Hg concentrations in fish exceeding 0.3 μg/g wet wt (equivalent to 0.5 μg/g wet wt in axial muscle) may result in effects that are sublethal.5,6 Thus, many wild fish appear at risk of toxic effects from Hg exposure. Current assessments of effects of Hg on wild fish rely on statistical approaches. The simplest approach is to calculate a hazard quotient, by dividing the concentration of Hg in a fish population of interest (generally a summary statistic, e.g., geometric mean) by a threshold concentration above which effects r 2011 American Chemical Society
are predicted to occur. A value >1 indicates the population is at risk of toxic effects. Sandheinrich et al.,7 for example, used this approach to estimate that 44% of walleye (Sander vitreus) populations in the Great Lakes region of North America are at risk of Hg toxicity. Hazard quotients are used for screening-level assessments, however, and do not indicate actual effects. Another approach is to test for changes in anatomy or physiology, especially of “biomarkers”, in fish exposed to elevated concentrations of Hg. A biomarker is a contaminant-induced change in anatomy or physiology of an organism and, ideally, should be validated by a laboratory experiment.8 For example, sex steroid hormones are reproductive biomarkers of Hg in fish, as illustrated in a lab experiment in which fathead minnows (Pimephales promelas) exposed to elevated concentrations of dietary methylmercury (MeHg) had suppressed levels of plasma testosterone and 17-ß estradiol, and the result on reproduction was reduced spawning Received: May 14, 2011 Accepted: September 6, 2011 Revised: August 18, 2011 Published: September 06, 2011 8982
dx.doi.org/10.1021/es201641x | Environ. Sci. Technol. 2011, 45, 8982–8988
Environmental Science & Technology success.9 Field studies have subsequently shown significant negative correlations between Hg concentrations in axial muscle and sex steroid hormones levels in plasma of fish,10,11 but environmental conditions that vary among sites (and may influence fish physiology), as well as the presence of other contaminants, precludes the conclusion that Hg is the cause of suppressed hormone levels. Even in the extreme case of Clay Lake (Ontario, Canada), where a chloralkali plant resulted in northern pike (Esox lucius) with Hg concentrations of 6 16 μg/g wet wt (in axial muscle), Lockhart et al.12 could not attribute changes in physiology of northern pike, compared to a reference site, to Hg. With the exception of Minamata Bay, no study has conclusively linked environmental exposure and toxic effects of Hg in fish. In this study, we use a new approach to link Hg exposure to health effects in spotted gar (Lepisosteus oculatus) from Caddo Lake (TX/LA). Previous field studies have reported elevated incidences of macrophage centers in liver, kidney, and spleen of fish with high concentrations of Hg.13 16 Macrophage centers are aggregates of specialized white blood cells that form in fish as an immune response to tissue damage, and are considered a general biomarker of contaminant toxicity.17 Hg is known to alter cellular calcium homeostasis and this effect leads to the well documented mercury mediated apoptosis and necrosis.18 Increased intracellular calcium has been linked to macrophage activation.19,20 We hypothesize that Hg-mediated areas of cell death resulted in increased macrophage activity, thus explaining why MA and Hg were colocalized (as shown by LA-ICP-MS) in gar liver.
’ EXPERIMENTAL SECTION Study Sites and Sampling. Spotted gar (n = 10) were captured from wetland and open-water habitats of Caddo Lake (TX/LA). Sampling sites and the limnology of the lake have been described elsewhere.21 23 Briefly, Caddo Lake and its associated wetlands cover approximately 10,850 ha and are composed of cypress swamps, marshes, bottomland hardwood forests, grasslands, and pine forests, much of which remain in a relatively undisturbed condition.24,25 The western portion of the lake is shallow (many areas <1 m) and characterized by wetland habitat dominated by bald cypress (Taxodium distichum) and water elm (Planara aquatica), and other aquatic vegetation including fanwort (Cabomba caroliniana), common waterweed (Elodea sp.), yellow pond-lily (Nuphar lutea), and invasive water hyacinth (Eichhornia crassipes).26 In contrast, the eastern portion of the lake is comprised primarily of openwater habitat with an average depth of 1.4 m.27 Fish from Caddo Lake contain some of the highest Hg concentrations recorded in Texas28 and mercury contamination in Caddo Lake is of particular concern because the lake supports high biodiversity, including rare and threatened species24,29 which may be negatively impacted by Hg exposure. The most probable source of Hg loading into Caddo Lake is atmospheric deposition.30 The primary anthropogenic sources of Hg in the region are coal-burning power plants;31 however, a substantial proportion of the Hg deposited in the region may originate from sources outside of North America.32 Fish and grass shrimp (Palaeomonetes kadiakensis) in the western, wetland portion of the lake contain higher concentrations of Hg than those in the eastern, open-water portion of the lake.21,22 The observed differences in Hg accumulation is most likely due to
ARTICLE
differences in the MeHg availability between the two habitats, with the wetland being more conducive to Hg methylation or more efficient at the incorporation of MeHg into the base of the food chain.21,22 Although fish from all locations in Caddo Lake contain elevated levels of Hg, in this study the wetland site was considered “contaminated” and the open-water site was considered to be a “reference” site. Spotted gar were targeted for this study because they were abundant at the reference and contaminated site and known from previous studies to have elevated concentrations of Hg in their tissues.22 Gar were euthanized immediately after capture, total length and wet weight were recorded, and samples of axial muscle and liver tissue were collected. Subsamples of tissues for Hg determination (muscle and liver) and color analysis (liver only) were placed in individual plastic bags and frozen at 20 C. Subsamples of liver tissue for histopathology and LA-ICP-MS were fixed in neutral buffered formalin and later transferred to 70% ethanol for storage. Hg Analysis. Samples were thawed, dried in a 60 C oven, and homogenized to a flour-like consistency with a ball-mill grinder. Total Hg was analyzed in the homogenized samples with a direct Hg analyzer (DMA-80, Milestone Inc. Monroe, CT). Quality assurance included reference and duplicate samples, as previously described.21,22 Reference materials were within certified ranges, and the mean relative percent difference of duplicate samples was 3.63%. Liver Color. Liver color measurements were carried out as previously described by Drevnick et al.13 Briefly, a 100 mg piece of liver was homogenized in 1 mL deionized water, and 0.2 mL chloroform was added. The mixture was centrifuged for 15 min at 12 000 rpm. An aliquot of the supernatant was transferred to a microplate well and measured for absorbance at 400 nm on a BioTek Synergy 2 spectrophotometer (Winooski, VT, USA). Each liver was analyzed in duplicate. A mean percent difference <20% between replicates was deemed acceptable. Macrophage Aggregates. Livers from individual gar were viewed for analysis of MA as previously described by Drevnick et al.13 Livers were randomly assigned a unique number to facilitate blind study, embedded in paraffin, sectioned, and mounted on glass slides.33 For each individual gar, three unstained sections were viewed with fluorescence microscopy (excitation filter 355 425 nm, suppression at 460 nm)13 at 200X magnification. A representative field from each section was photographed, and a digital grid (squares 25 μm 25 μm) was overlaid on the image with Adobe Photoshop (Adobe Systems Inc., San Jose, California). Fifty grid squares (out of a total of 130 per field) were selected with a random number generator. Presence or absence of MA was noted and recorded in each of the selected grid squares. The proportion of grid squares containing MA out of the total 50 assessed was used as a quantitative measure of the extent of liver damage. The average score of the three sections was used in the data analysis.13 LA-ICP-MS. Microscope slides created for the pathological portion of the study were used to investigate the distribution of mercury in spotted gar liver sections by LA-ICP-MS. A microscope slide, with unstained, paraffin-embedded liver tissue, was placed into the chamber of a 213 nm Nd:YAG laser ablation source (New Wave Research, Fremont, CA). A CCD camera enabled zooming and scanning of the slide to locate liver tissue. Five areas of normal liver tissue and five MA were chosen at random and ablated from liver sections from seven individual gar (the three individuals not analyzed had MA scores of zero). 8983
dx.doi.org/10.1021/es201641x |Environ. Sci. Technol. 2011, 45, 8982–8988
Environmental Science & Technology A 15 μm beam diameter and 100 μm raster spacing enabled adequate sampling of MA and decreased the risk of interferences and resampling of previously ablated material. A Varian 820 ICPMS coupled to the laser source was used to monitor 202Hg and 78 Se isotopes. Paraffin near the edge of the liver sections was also ablated, and confirmed a negligible contribution of 78Se and 202 Hg. The mass detector collected data for 100 s each run, the first 30 s of which was laser warm-up. Mean isotope counts for the first 30 s were used as an estimate of background noise and subtracted from the mean counts for the remainder of the run to calculate a signal. After ablation, slides were viewed again under fluorescence microscopy to ensure lipofuscin containing MA were targeted. Table S1 in the Supporting Information describes the instrument parameters used for analysis. Statistical Analyses. Statistics were performed with JMP 4 Statistical Analysis Software (SAS Institute, Cary, NCA). Data were transformed, if necessary, to meet the assumptions of tests. One-tailed student’s t tests were used to compare differences in variables between habitat types. Least-squares regression models were used to describe relationships between variables. A paired t test was used to compare the mean isotope signals (202Hg or 78Se) between normal tissue and MA. Significance of statistical tests was determined with a type I error (α) of 0.05. In addition, Psi-Plot (Poly Software International, Pearl River, NY) was used to create a 3D surface contour describing the relative Hg concentration in a gar liver section.
’ RESULTS AND DISCUSSION Hg Concentrations. A complete presentation of Hg concentrations throughout the Caddo Lake food web is presented in Chumchal et al.23 In spotted gar sampled for this study, Hg concentrations (mean ( SE) in axial muscle were higher from wetland habitats (0.657 ( 0.091 μg/g wet wt) than from openwater habitats (0.239 ( 0.069 μg/g wet wt) of Caddo Lake (p = 0.0032). Chumchal and Hambright22 reported for Caddo Lake that Hg concentrations are higher in food webs from wetland than from open-water habitats. Hg inputs likely do not differ between the two habitat types, as the principal source of Hg to the system is atmospheric deposition.30 Rather, Hg methylation may be enhanced in wetland habitats34 and, if so, would explain why fish in general (several species) are more contaminated in wetlands than in open-waters of Caddo Lake.22 For spotted gar, Hg concentrations are also a function of size (total length) and age,22 but neither of these factors can explain differences in Hg accumulation between habitat types. Gar sampled were from a narrow size range (total length: range = 45 60 cm; mean ( SE = 50.3 ( 2.57 cm from wetland, 55.6 ( 1.85 cm open-water). Gar ages were not measured for this study, but data from Chumchal and Hambright22 indicate there are no differences in age at size between habitat types. Spotted gar from wetland habitats have Hg concentrations known to be toxic to fish. Recent analyses of the available data for Hg toxicity in fish indicate that effects are likely to occur at concentrations in axial muscle exceeding 0.5 μg/g wet wt.5,6 Concentrations in five of the six gar from wetland habitats exceeded this value (hazard quotient of mean concentration >1). In contrast, all four of the gar from open-water habitats had concentrations lower than the threshold value (hazard quotient <1). Based on this simple hazard evaluation, we would expect to find evidence of Hg toxicity in gar from wetland habitats (treatment), but not from open-water habitats (reference).
ARTICLE
Figure 1. Total mercury concentration (μg/g wet wt) in spotted gar muscle and liver from open-water and wetland habitats.
Hg concentrations (mean ( SE) in liver of spotted gar were higher from wetland habitats (9.01 ( 1.90 μg/g wet wt) than from open-water habitats (0.728 ( 0.376 μg/g wet wt) of Caddo Lake (p = 0.0034 ; Figure 1). Hg concentrations in liver were greater than in muscle tissue of gar from both habitats. Chumchal et al.23 also reported higher Hg concentrations in spotted gar liver than in muscle. Though only total Hg was measured in this study, Chumchal et al.,23 analyzed MeHg and total Hg in spotted gar muscle and liver from Caddo Lake. Hg found in the muscle tissue was primarily in the organic form, which is consistent with previous report.35,36 However, in the liver tissue, the majority of Hg was present in the inorganic form.23 Elevated inorganic Hg concentrations in the livers of gar could be due to (1) a dietary source or (2) the way Hg is metabolized, that is, demethylation, which has not been well described in fish. Drevnick et al.13 reported similarly low MeHg and high inorganic Hg concentrations in the livers of northern pike. In that study, the authors speculate that the high inorganic Hg content in the livers was not due to demethylation, but rather to the inorganic Hg content of larval odonates, a common prey item of northern pike in that ecosystem. MeHg concentrations were similar in pike muscle and liver suggesting MeHg was not being converted to inorganic Hg. Conversely, from Chumchal et al.,23 MeHg comprised the majority of total Hg in gar muscle, but only a small percentage of the total Hg in gar livers, suggesting that demethylation could be occurring. At the cellular level, MeHg is transported into the liver where it can be conjugated to glutathione and excreted in the bile.37,38 Inorganic Hg, however, is complexed in the liver with selenium into large polyatomic ions.39,40 These ions are then associated with binding proteins such as selenoprotein P resulting in an accumulation of inorganic Hg in the liver relative to other tissues.41 43 Demethylation has been speculated to occur in birds39 and fish44 but little direct cellular evidence or delineated mechanism has been shown. Mercury speciation data from these two studies is mentioned here because of the possible link to liver damage observed in spotted gar from this study. Hepatic demethylation in fish should be the focus of future studies along with inorganic Hg’s role in liver toxicity. Biomarkers in Liver Tissue. Liver color (as measured by absorbance; λ=400 nm) was significantly different between habitat types. Gar from the wetland habitat had significantly darker livers than those from open-water (p < 0.0001). Liver coloration in spotted gar was positively correlated with total Hg in the muscle (p = 0.01; r2 = 0.58) and total Hg in the liver (p = 0.013; r2=0.56) (Figure 2). 8984
dx.doi.org/10.1021/es201641x |Environ. Sci. Technol. 2011, 45, 8982–8988
Environmental Science & Technology
Figure 2. Relationship between liver color (λ = 400 nm) and total mercury concentration (μg/g wet wt) in the muscle (A) and liver (B), of spotted gar from open-water and wetland habitats.
During histologic examination of gar livers, we identified darkly pigmented deposits as macrophage aggregates, containing melanin and lipofuscin. Melanin is a pigment derived from tyrosine which can scavenge free radicals, bind metal ions, and is produced by hematopoietic cells in fish. Lipofuscin pigment is composed of oxidized lipids and proteins and is a sign of cell membrane or mitochondrial damage due to chronic oxidation.45 Drevnick et al.13 identified lipofuscin as the pigment responsible for the majority of variation in pike liver color. In spotted gar from Caddo Lake liver color and the percent of MA were positively correlated in gar liver (p = 0.013; r2 = 0.56). The presence of lipofusicn in MA could account for some of the variation in gar liver color. Both recent field- and laboratory- based studies have shown positive correlations between Hg exposure and MA in wild fish.15,16 Other studies have found evidence of increased MA in fish from other types of contaminated environments.46,47 Changes in both MA and lipofusicn have been associated with aging,45,48 and this was not controlled for in the present study. However, Drevnick et al.13 measured age using scales and determined it not to be a significant factor in the occurrence of liver damage in Isle Royale northern pike. In that study, age was used as a covariate to statistically show that Hg concentration was the best predictor of liver damage in pike. Also, from Chumchal et al.,23 there does not appear to be any differences in age at size of spotted gar between habitat types of Caddo Lake. Thus, sampling gar of equal size, as was done in this study, from both sides of the lake would likely result in fish of the same age. Habitat-specific differences in hepatic MA were observed for spotted gar. Spotted gar sampled from wetland habitats had approximately five times higher occurrences of MA than those sampled from open-water areas (p = 0.004). Spotted gar contained hepatic MA with a mean occurrence of 9% in wetland areas, 0.31% in open-water, and a maximum of 22% of the
ARTICLE
Figure 3. Relationship between mean hepatic macrophage score and total mercury concentration (μg/g wet wt) in muscle (A) and liver (B), of spotted gar from open-water and wetland habitat types.
histological grids in a single individual. The occurrence of MA was positively correlated with total Hg in the muscle (p = 0.004; r2 = 0.67) and total Hg in the liver (p = 0.009; r2 = 0.59) (Figure 3). LA-ICP-MS. We determined, using laser ablation ICP-MS, that 202 Hg counts were significantly higher in MA than in visibly normal gar liver tissue (p = 0.003). In general, MA had twice as many 202Hg counts as parenchyma (Figures 4 and 5). MA also had significantly higher 78Se than parenchyma (p = 0.031), where counts rarely rose above the estimated background signal. Former studies have employed autometallography (AMG) or electron microscopy (EM) coupled with energy dispersive X-ray analysis (EDAX) in an attempt to show the colocalization of Hg and pathologies, with differing results. Rawson et al.49 used EDAX and EM to suggest Hg localization in the hepatic lysosomes of bottlenose dolphins (Tursiops truncatus), speculating that Hg was the cause of increased lipofuscin deposits. Woshner et al.50 used AMG to show that in the livers of beluga whales (Delphinapterus leucas), Hg was colocalized with stellate macrophages. However, Hg and lipofuscin were rarely colocalized and the authors suggest that diet rather than Hg was the cause of lipofuscin accumulation. Furthermore, Hg was likely found as HgSe in beluga livers,50 which is widely considered to be the endproduct of Hg detoxification in marine mammals.50 52 Higher counts of 78Se in gar macrophages warrants further investigation of the identity of Hg compounds, as it is possible that Hg is bound to Se in these immune cells. However, our study, unlike Woshner et al.50 consistently demonstrates colocalization of Hg and lipofuscin deposits in the liver tissue, suggesting a causative relationship. MA are important in the immune function of fish, and serve as stores for the endproducts of cellular breakdown.48,53 Hg is 8985
dx.doi.org/10.1021/es201641x |Environ. Sci. Technol. 2011, 45, 8982–8988
Environmental Science & Technology
ARTICLE
Figure 4. (A) Image of a spotted gar liver section after laser ablation of normal parenchyma and a macrophage aggregate. Dark patches are macrophage aggregates. Lighter patches are normal parenchyma. (B) Graph showing 202Hg counts in the ablated normal parenchyma and (C) macrophage aggregate.
Figure 5. (A) A preablation image of a spotted gar liver section with a superimposed raster pattern. Dark patches are macrophage aggregates. Lighter patches are normal parenchyma. (B) 3D contour image showing the relative distribution of 202Hg after laser ablation of the section shown in (A).
known to increase intracellular levels of calcium leading to apoptosis and necrosis, thus increasing oxidized lipids and proteins.18 Increases in free calcium in the cell have also been implicated in the activation of macrophage respiratory burst,19,20 which can produce reactive oxygen species capable of breaking the MeHg bond.54 We found elevated incidences of macrophage centers in liver of spotted gar and used a new technology for ecotoxicology studies, LA-ICP-MS, to colocalize lesioned areas and Hg deposits within the tissue. We conclude that Hg compromises the health of spotted gar in our study and, perhaps, other fish exposed to elevated concentrations of Hg. Future studies should focus on the importance of these cells in fish Hg metabolism.
The strong correlations between two measures of pathology (MA and color) and Hg content in both the liver and muscle of spotted gar further implicate Hg as the cause of liver damage. The relationship between liver damage and habitat suggests that Hg is the cause, rather than another contaminant. Liver discoloration and MA were elevated in spotted gar from wetland habitats. Previous reports by Chumchal et al.21 and Chumchal and Hambright22 have documented higher concentrations of Hg in wetland fish in Caddo Lake than fish sampled from open-water. It is unlikely that other contaminants present in Caddo Lake which might biomagnify would display the same habitat-specific differences observed here; Hg relies on wetlands for methylation to achieve the biomagnifying species whereas contaminants such as 8986
dx.doi.org/10.1021/es201641x |Environ. Sci. Technol. 2011, 45, 8982–8988
Environmental Science & Technology organochlorines do not. We did not measure other inorganic and organic compounds in gar tissues for this study. However, the Texas Commission of Environmental Quality in cooperation with U.S. EPA surveyed contaminants in muscle and whole bodies of a variety of fish species from several sites within Caddo Lake. They found that mercury was present in muscle from all sites and at elevated levels. Zinc, magnesium, iron and manganese were present in fillets and whole fish at normal and expected levels while other metals were not detected. Pesticides, PCBs, and perchlorate were not detected in muscle or whole fish samples.28 Implications. Both the hazard quotient and biomarker based approaches suggested that Hg impairs the health of spotted gar from wetland habitats of Caddo Lake. This was supported further, by the novel use of LA-ICP-MS to colocalize Hg and liver damage in these fish. Past studies involving Hg exposure in wild fish, have noted negative correlations between Hg and measures of fish health. Larose et al.55 described negative relationships between Hg and both, liver antioxidant activity and the hepatosomatic index of wild fish. Drevnick et al.13 noted decreased lipid reserves in pike livers and decreased condition factor of fish with higher Hg concentrations. Similarly, Lockhart et al.12 documented decreased fat storage in the liver and emaciation of northern pike from the Hg contaminated Clay Lake. Concentrations of Hg in Caddo Lake spotted gar are within the range of concentrations concluded to impair fish health. Future studies are needed to determine how elevated Hg concentrations are impacting fish health and fitness.
’ ASSOCIATED CONTENT
bS
Supporting Information. One table of LA-ICP-MS parameters. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 940-891-6957; fax 940-565-4297; e-mail: author:aproberts@ unt.edu.
’ ACKNOWLEDGMENT We thank Dr. Ray Drenner (Texas Christian University) for his input on this study and assistance in the field. We would also like to thank Dr. Barney Venables (University of North Texas) for his assistance. Funding for this project was provided by the Caddo Lake Institute, Coypu Foundation, Texas Christian University Research and Creative Activities Fund grant, AFOSR-DURIP FA9550-09-1-0496, and the University of North Texas. ’ REFERENCES (1) Selin, N. E. Global biogeochemical cycling of mercury: A review. Annu. Rev. Environ. Resour. 2009, 34, 43–63. (2) Scheuhammer, A. M.; Meyer, M. W.; Sandheinrich, M. B.; Murray, M. W. Effects of environmental methylmercury on the health of wild birds, mammals, and fish. Ambio 2007, 36, 12–18. (3) Mergler, D.; Anderson, H. A.; Chan, L. H. M.; Mahaffey, K. R.; Murray, M.; Sakamoto, M.; Stern, A. H. Methylmercury exposure and health effects in humans: A worldwide concern. Ambio 2007, 36 (1), 3–11.
ARTICLE
(4) Wiener, J. G.; Spry, D. J. Toxicological significance of mercury in freshwater fish. In Environmental Contaminants in Wildlife: Interpreting Tissue Concentrations; Beyer, W. N., Heinz, G. H., Redmon-Norwood, A. W., Eds.; Lewis Publishers; CRC Press: Boca Raton, FL, 1996; pp 297 340. (5) Dillon, T.; Beckvar, N.; Kern, J. Residue based mercury dose response in fish: An analysis using lethality equivalent test endpoints. Environ. Toxicol. Chem. 2010, 29 (11), 2559-2565; DOI: 10.1002/etc.314. (6) Sandheinrich, M. B.; Weiner, J. G. Mercury in freshwater fish: Recent advances in assessing toxicity of environmentally relevant exposures. In Environmental Contaminants in Biota: Interpreting Tissue Concentrations; Beyer, W. N., Meador, J., Eds.;. Taylor and Francis; CRC Press: Boca Raton, FL, 2011; pp 169 190. (7) Sandheinrich, M. B.; Bhavsar, S. P.; Bodaly, R. A.; Drevnick, P. E.; Paul, E. A. Ecological risk of methylmercury to piscivorous fish populations of the great lakes region. Ecotoxicology 2011, DOI: 10.1007/ s10646-011-0712-3. (8) Biomarkers of Environmental Contamination; McCarthy, J. F., Shugart, L. R., Eds.; Lewis Publishers: Boca Raton, 1990. (9) Drevnick, P. E.; Sandheinrich, M. B. Effects of dietary methylmercury on reproductive endocrinology of fathead minnows. Environ. Sci. Technol. 2003, 37, 4390–4396. (10) Adams, S. M.; Bevelhimer, M. S.; Greeley, M. S., Jr; Levine, D. A.; Teh, S. J. Ecological risk assessment in a large river-reservoir: 6. Bioindicators of fish population health. Environ. Toxicol. Chem. 1999, 18, 628–640. (11) Webb, M. A. H.; Feist, G.; Fitzpatrick, M.; Foster, E.; Schreck, C.; Plumlee, M.; Wong, C.; Gundersen, D. Mercury concentrations in gonad, liver, and muscle of white sturgeon Acipenser transmontanus in the lower columbia river. Arch. Environ. Contam. Toxicol. 2006, 50, 443–451. (12) Lockhart, W.; Uthe, J.; Kenney, A.; Mehrle, P. Methylmercury in northern pike (Esox lucius): Distribution, elimination, and some biochemical characteristics of contaminated fish. Fish. Res. Board Can. 1972, 29, 1519–1523. (13) Drevnick, P.; Roberts, A.; Otter, R.; Hammerschmidt, C.; Klaper, R.; Oris, J. Mercury toxicity in livers of northern pike (Esox lucius) from Isle Royale, USA. Comp. Biochem. Phys.C 2008, 147, 331–338. (14) Meinelt, T.; Kr€uger, R.; Pietrock, M.; Osten, R.; Steinberg, C. Mercury pollution and macrophage centres in pike (Esox lucius) tissues. Environ. Sci. Pollut. Res. 1997, 4, 32–36. (15) Raldua, D.; Bayona, J.; Barcelo, D. Mercury levels and liver pathology in feral fish living in the vicinity of a mercury cell chlor-alkali factory. Chemosphere 2007, 66, 1217–1225. (16) Schwindt, A. R.; Fournie, J. W.; Landers, D. H.; Schreck, C. B.; Kent, M. L. Mercury concentrations in salmonids from western us national parks and relationships with age and macrophage aggregates. Environ. Sci. Technol. 2008, 42, 1365–1370. (17) Hinton, D. E.; Lauren, D. J. Liver structural alterations accompanying chronic toxicity in fishes: Potential biomarkers of exposure. In Biomarkers of Environmental Contamination; Lewis Publishers: Boca Raton, 1990; p 17. (18) Kim, S. H.; Sharma, R. P. Mercury-induced apoptosis and necrosis in murine macrophages: Role of calcium-induced reactive oxygen species and p38 mitogen-activated protein kinase signaling. Toxicol. Appl. Pharmacol. 2004, 196, 47–57. (19) Hotchkiss, R. S.; Bowling, W. M.; Karl, I. E.; Osborne, D. F.; Flye, M. W. Calcium antagonists inhibit oxidative burst and nitrite formation in lipopolysaccharide-stimulated rat peritoneal macrophages. Shock. 1997, 8, 170. (20) Iles, K. E.; Forman, H. J. Macrophage signaling and respiratory burst. Immunol. Res. 2002, 26, 95–105. (21) Chumchal, M.; Drenner, R.; Fry, B.; Hambright, K.; Newland, L. Habitat-specific differences in mercury concentration in a top predator from a shallow lake. Trans. Am. Fish. Soc. 2008, 137, 195–208. (22) Chumchal, M.; Hambright, K. Ecological factors regulating mercury contamination of fish from Caddo Lake, Texas, USA. Environ. Toxicol. Chem. 2009, 28, 962–972. 8987
dx.doi.org/10.1021/es201641x |Environ. Sci. Technol. 2011, 45, 8982–8988
Environmental Science & Technology (23) Chumchal, M. M.; Rainwater, T. R.; Osborn, S. C.; Roberts, A. P.; Abel, M. T.; Cobb, G. P.; Smith, P. N.; Bailey, F. C. Mercury speciation and biomagnification in the food web of Caddo Lake (TX/ LA, USA), a subtropical freshwater ecosystem. Environ. Toxicol. Chem. 2011, 30, 1153–1162. (24) Ingold, J. L.; Hardy, L. M. Web of life: The ecology and biodiversity of Caddo Lake; Louisiana State University: Shreveport, LA, 1996. (25) Consumption of Deer Tissue Collected at Caddo Lake National Wildlife Refuge, Texas Department of State Health Services: Austin, TX, 2006; www.dshs.state.tx.us/epitox/consults/caddo_deer. pdf (accessed October 16, 2010). (26) Van Kley, J. E.; Hine, D. N. The wetland vegetation of Caddo Lake. Tex. J. Sci. 1998, 50, 267–290. (27) Ensminger, P. A. Bathymetric Survey and Physical and ChemicalRelated Properties of Caddo Lake, Louisiana and Texas, August and September 1998; Water Resources Investigations Report 99-4217; U.S. Geological Survey: Denver, CO, 1999. (28) Strategy for Addressing Mercury Contamination in Fish Tissue in Caddo Lake; Texas Commission on Environmental Quality: Austin, TX, 2004; http://caddolakedata.us/media/266/tceq2004.pdf. (29) Caddo Lake Institute Website. http://www.caddolakeinstitute. us/ramsar.html (accessed October 16, 2010). (30) Twidwell, S. Bioaccumulation of mercury in selected east Texas water bodies, Report AS-180; Texas Commision on Environmental Quality: Austin, TX, 2000. (31) United States Environmental Protection Agency Toxics Release Inventory Website. http://www.epa.gov/tri/ (accessed October 16, 2010). (32) Selin, N. E.; Jacob, D. J. Seasonal and spatial patterns of mercury wet deposition in the United States: Constraints on the contribution from North American anthropogenic sources. Atmos. Environ. 2008, 42, 5193–5204. (33) Presnell, J. K.; Schreibman, M. P. Humason’s Animal Tissue Techniques; Johns Hopkins University Press: Baltimore, 1997. (34) St. Louis, V. L.; Rudd, J. W. M.; Kelly, C. A.; Beaty, K. G.; Bloom, N. S.; Flett, R. J. Importance of wetlands as sources of methyl mercury to boreal forest ecosystems. Can. J. Fish. Aquat. Sci. 1994, 51, 1065–1076. (35) Bloom, N. On the chemical form of mercury in edible fish and marine invertebrate tissue. Can. J. Fish. Aquat. Sci. 1992, 49, 1010–1017. (36) Grieb, T. M.; Bowie, G. L.; Driscoll, C. T.; Gloss, S. P.; Schofield, C. L.; Porcella, D. B. Factors affecting mercury accumulation in fish in the upper Michigan peninsula. Environ. Toxicol. Chem. 1990, 9, 919–930. (37) Dutczak, W.; Ballatori, N. Transport of the glutathione-methylmercury complex across liver canalicular membranes on reduced glutathione carriers. J. Biol. Chem. 1994, 269, 9746. (38) Mugesh, G.; Singh, H. B. Synthetic organoselenium compounds as antioxidants: Glutathione peroxidase activity. Chem. Soc. Rev. 2000, 29, 347–357. (39) Eagles-Smith, C.; Ackerman, J.; Yee, J.; Adelsbach, T. Mercury demethylation in waterbird livers: Dose-response thresholds and differences among species. Environ. Toxicol. Chem. 2009, 28, 568–577. (40) Palmisano, F.; Cardellicchio, N.; Zambonin, P. Speciation of mercury in dolphin liver: A two-stage mechanism for the demethylation accumulation process and role of selenium. Mar. Environ. Res. 1995, 40, 109–121. (41) Khan, M. A. K.; Wang, F. Mercury selenium compounds and their toxicological significance: Toward a molecular understanding of the mercury selenium antagonism. Environ. Toxicol. Chem. 2009, 28, 1567–1577. (42) Reeves, M.; Hoffmann, P. The human selenoproteome: Recent insights into functions and regulation. Cell. Mol. Life Sci. 2009, 66, 2457–2478. (43) Schomburg, L.; Schweizer, U.; K€ohrle, J. Selenium and selenoproteins in mammals: Extraordinary, essential, enigmatic. Cell. Mol. Life Sci. 2004, 61, 1988–1995. (44) Gonzalez, P.; Dominique, Y.; Massabuau, J.; Boudou, A.; Bourdineaud, J. Comparative effects of dietary methylmercury on gene
ARTICLE
expression in liver, skeletal muscle, and brain of the zebrafish (Danio rerio). Environ. Sci. Technol. 2005, 39, 3972–3980. (45) Terman, A.; Brunk, U. T. Lipofuscin. Intern. J. Biochem. Cell Bio. 2004, 36, 1400–1404. (46) Capps, T.; Mukhi, S.; Rinchard, J. J.; Theodorakis, C. W.; Blazer, V. S.; Pati~ no, R. Exposure to perchlorate induces the formation of macrophage aggregates in the trunk kidney of zebrafish and mosquitofish. J. Aquat. Anim. Health 2004, 16, 145–151. (47) Khan, R. Health of flatfish from localities in Placentia Bay, Newfoundland, contaminated with petroleum and pcbs. Arch. Environ. Contam. Toxicol. 2003, 44, 485–492. (48) Agius, C.; Roberts, R. Melano macrophage centres and their role in fish pathology. J. Fish Dis. 2003, 26, 499–509. (49) Rawson, A.; Patton, G.; Hofmann, S.; Pietra, G.; Johns, L. Liver abnormalities associated with chronic mercury accumulation in stranded atlantic bottlenosed dolphins. Ecotoxicol. Environ. Saf. 1993, 25, 41–47. (50) Woshner, V.; O’Hara, T.; Eurell, J.; Wallig, M.; Bratton, G.; Suydam, R.; Beasley, V. Distribution of inorganic mercury in liver and kidney of beluga and bowhead whales through autometallographic development of light microscopic tissue sections. Toxicol. Pathol. 2002, 30, 209. (51) Martoja, R.; Berry, J. P. Identification of Tiemannite as a probable product of demethylation of mercury by selenium in cetaceans. A complement to the scheme of the biological cycle of mercury. Vie Millieu 1980, 30, 7–10. (52) Nigro, M.; Leonizio, C. Intracellular storage of mercury and selenium in different marine vertebrates. Mar. Ecol.: Prog. Ser. 1996, 135, 137–143. (53) Wolf, J. C.; Wolfe, M. J. A brief overview of nonneoplastic hepatic toxicity in fish. Toxicol. Pathol. 2005, 33, 75. (54) Suda, I.; Totoki, S.; Uchida, T.; Takahashi, H. Degradation of methyl and ethyl mercury into inorganic mercury by various phagocytic cells. Arch. Toxicol. 1992, 66, 40–44. (55) Larose, C.; Canuel, R.; Lucotte, M.; Di Giulio, R. T. Toxicological effects of methylmercury on walleye (Sander vitreus) and perch (Perca flavescens) from lakes of the boreal forest. Comp. Biochem. Phys. C 2008, 147, 139–149.
8988
dx.doi.org/10.1021/es201641x |Environ. Sci. Technol. 2011, 45, 8982–8988
ARTICLE pubs.acs.org/est
Synergistic Bactericidal Activity of Ag-TiO2 Nanoparticles in Both Light and Dark Conditions Minghua Li,† Maria Eugenia Noriega-Trevino,|| Nereyda Nino-Martinez,|| Catalina Marambio-Jones,† Jinwen Wang,† Robert Damoiseaux,‡,§ Facundo Ruiz,|| and Eric M. V. Hoek*,†,‡ Department of Civil & Environmental Engineering, ‡California NanoSystems Institute, and §Molecular Screening Shared Resource, University of California, Los Angeles, California 90095, United States lvaro Obregon 64, Col. Centro, San Luis Potosí, S. L. P., Mexico Facultad de Ciencias, UASLP, A
)
†
bS Supporting Information ABSTRACT: High-throughput screening was employed to evaluate bactericidal activities of hybrid Ag-TiO2 nanoparticles comprising variations in TiO2 crystalline phase, Ag content, and synthesis method. Hybrid Ag-TiO2 nanoparticles were prepared by either wet-impregnation or UV photo deposition onto both Degussa P25 and DuPont R902 TiO2 nanoparticles. The presence of Ag was confirmed by ICP, TEM, and XRD analysis. The size of Ag nanoparticles formed on anatase/rutile P25 TiO2 nanoparticles was smaller than those formed on pure rutile R902. When activated by UV light, all hybrid Ag-TiO2 nanoparticles exhibited stronger bactericidal activity than UV alone, Ag/UV, or UV/TiO2. For experiments conducted in the dark, bactericidal activity of Ag-TiO2nanoparticles was greater than either bare TiO2 (inert) or pure Ag nanoparticles, suggesting that the hybrid materials produced a synergistic antibacterial effect unrelated to photoactivity. Moreover, less Ag+ dissolved from Ag-TiO2 nanoparticles than from Ag nanoparticles, indicating the antibacterial activities of Ag-TiO2 was not only caused by releasing of toxic metal ions. It is clear that nanotechnology can produce more effective bactericides; however, the challenge remains to identify practical ways to take advantage of these exciting new material properties.
1. INTRODUCTION Nanotechnology is expected to improve our lives in many ways and is expected to give rise to a billion dollar industry over the next decade. One important application is to fight and prevent disease. Photoactive metal oxides and bactericidal metallic nanomaterials have attracted great attention due to their broad antimicrobial capabilities.1,2 Photoactive metal oxides like TiO2 are well-known for their disinfection capability when activated by UV and solar light.1,2 Irradiation of TiO2 with UV greater than its band gap energy, promotes an electron from the valence band to the conduction band. As a result, energy-rich electron hole pairs are formed, which can react with water or oxygen molecules forming various reactive oxygen species (ROS) that are strong oxidants.3 The ROS generation on or near a nano-TiO2 surface can cause oxidative damage to cell membranes and inactivate microorganisms.2 Metallic nanoparticles such as Ag, on the other hand, are another popular category of nanodisinfectants already used in a number of consumer products.4,5 Although the disinfection capability of metallic nanoparticles is widely appreciated, the underlying mechanisms are only partially understood.5 Possible modes of action include release of toxic metal ions, generation of ROS, and cell membrane damage due to direct contact by nanoparticles.4,6 However, the use of either type of nanoparticle as a disinfectant has certain limitations. Photoactive metal oxides have r 2011 American Chemical Society
little or no bacteria inhibition in the absence of UV light, which limits their application. On the other hand, it is concerning that the release of heavy metal ions may cause adverse environmental and human health effects.4 Thus, many studies seek to hybridize photoactive metal oxides with bactericidal metallic nanoparticles leveraging the advantages of both materials, while minimizing the release of potentially toxic ions.7 11 Several types of metal and metal oxides composite materials have been developed and tested for their bactericidal capability. For instance, Ag-TiO2 nanoparticles prepared by wetimpregnation, sol gel, and UV irradiation methods exhibited almost complete destruction of bacterial pathogens under UV or solar illumination.2,7 The metallic nanoparticles act as a conductor and as charge separators because of their high electron conductivity. Consequently, Ag nanoparticles enhanced the photoactivity of TiO2 nanoparticles by removing electrons from TiO2 particles and blocking electron hole recombination.3,7,12 However, the bactericidal effects and mechanisms of Ag-TiO2 nanoparticles in the absence of light are not adequately documented.3,13 Received: May 16, 2011 Accepted: August 25, 2011 Revised: August 12, 2011 Published: August 25, 2011 8989
dx.doi.org/10.1021/es201675m | Environ. Sci. Technol. 2011, 45, 8989–8995
Environmental Science & Technology
ARTICLE
Table 1. Characteristics of Commercial TiO2 and Synthesized Ag-TiO2 Nanoparticles Ag-P25 1:10 reference ID
TiO2 P25
NaBH4
Ag-R902 1:10
Ag-P25 1:50 Ag-P25 1:50 NaBH4
UV
TiO2 R902
Ag-R902 1:50 Ag-R902 1:50
NaBH4
NaBH4
UV
TiO2 size, nm (TEM)
25 35
25 35
25 35
25 35
200 450
200 450
200 450
200 450
TiO2 phase
A/R
A/R
A/R
A/R
R
R
R
R
TiO2 base material
-
P25
P25
P25
-
R-902
R-902
R-902
Ag:TiO2 molar ratio
-
1:10
1:50
1:50
-
1:10
1:50
1:50
reduction method
-
NaBH4
NaBH4
UV
-
NaBH4
NaBH4
UV
theoretical Ag content, w/w%
-
11.89
2.66
2.66
-
11.89
2.66
2.66
actual Ag content, w/w% Ag size, nm (TEM)
-
0.98 3 10
0.36 5 20
0.30 5 20
-
2.15 30 80
0.40 10 40
0.38 10 40
zeta potential, in freshwater, mV
8.4 ( 1.2
32.3 ( 1.9
7.7 ( 3.0
14.4 ( 2.0
28.1 ( 1.6
14.7 ( 2.7
17.3 ( 3.9
16.3 ( 1.7
zeta potential, in deionized
5.8 ( 5.5
36.9 ( 7.6
12.2 ( 2.3
6.8 ( 5.1
0.8 ( 3.0
33.9 ( 3.7
9.0 ( 1.4
7.1 ( 5.7
water, mV aggregate sizea, in freshwater, nm 880 ( 242
157 ( 3
836 ( 264
724 ( 257
335 ( 21
341 ( 18
429 ( 34
443 ( 36
aggregate sizea, in deionized
158 ( 3
177 ( 4
132 ( 2
366 ( 18
296 ( 4
470 ( 46
462 ( 29
157 ( 5
water, nm a
The mean aggregate size is reported between 0 and 15 min measured by dynamic light scattering (DLS).
This study aimed to assess the bactericidal activity of hybrid Ag-TiO2 nanoparticles in both light and dark conditions. Six types of Ag-TiO2 nanoparticles were prepared with different Ag reduction methods, TiO2 base materials, and Ag contents. Details are provided in Table 1. For comparison purposes, the bare TiO2 nanoparticles and a commercial Ag nanoparticle were also examined. High-throughput live/dead assays were used to evaluate viability of Gram-positive and Gram-negative bacteria.
2. EXPERIMENTAL SECTION 2.1. Nanoparticle Synthesis. In total, six types of Ag-TiO2 nanoparticles were prepared as detailed in Table 1. Two commercial TiO2 nanoparticles, DuPont Ti-Pure R-902 (200 450 nm) and Degussa P25 (25 35 nm), were employed as base materials to evaluate the impacts of TiO2 crystalline phase and particle size. On both of these commercial TiO2 nanoparticles, Ag nanoparticles were deposited onto the TiO2 surface by either wet-impregnation or UV photoreduction.11,14 A silver salt (ACS Reagent grade AgNO3, Sigma Aldrich, St. Louis, Missouri, USA) was used as the precursor for deposition of Ag nanoparticles onto TiO2 base nanoparticles. Two different molar ratios of AgNO3 and TiO2 were employed: 1:10 and 1:50. In a typical experiment, 100 mg of commercial TiO2 nanoparticles was dispersed in 100 mL of deionized water and sonicated for approximately five minutes. A predetermined amount of AgNO3 was added immediately to obtain a proper Ag:TiO2 ratio. The solution was allowed to mix for about 30 min at pH = 7 by magnetic stirring. For wet-impregnation reduction, NaBH4 (SigmaAldrich, ACS Reagent), at equivalent molar concentration of AgNO3, was added as reducing agent. For UV photoreduction experiments, the solution was magnetic stirred 30 min inside a UV reactor with lamps UV of wavelength of 354 nm. For both synthetic routes, the pH of the reaction media was adjusted to 10 by adding NH4OH (30% w/w aqueous solution, Sigma Aldrich, ACS Reagent) after the reaction started. The final solution was stirred vigorously for 30 min to allow the reaction to complete. The final products obtained were filtered using nitrocellulose Millipore filters with a pore diameter of 0.1 μm, washed with deionized water, and dried for further
characterization.11 All Ag-TiO2 nanoparticles were stored in sealed containers in the dark at 4 °C for less than 8 months prior to use. Experiments conducted on fresh and 8 month old nanoparticles stored in this manner showed negligible differences in bacterial toxicity. 2.2. Nanoparticle Stock Solutions. Nanoparticle stock solutions (1000 mg/L) used for characterization and bacterial viability assay were prepared by stirring Ag-TiO2 nanoparticles vigorously in ultrapure deionized water followed by 30-min sonication (FS30H, Fisher Scietific, 100W, 42 kHz). The suspensions were sonicated again for at least 5 min before use. All the nanoparticles suspensions were kept in 4 °C and in the dark after use. A new suspension was prepared every 72 h. 2.3. Nanoparticle Physical-Chemical Characterization. Transmission electron microscopy (TEM) analysis was performed on a JEOL JEM-1230 at an accelerating voltage of 100 kV. This allowed an estimation of primary size of both Ag and TiO2 nanoparticles. The crystalline phase of Ag-TiO2 nanoparticles was examined by X-ray diffraction on a GBC-Difftech MMA, with Cu Kα irradiation at λ = 1.54 Å. The wet state properties of Ag-TiO2 nanoparticles acquired in the media were further characterized in both deionized water and freshwater media. A model fresh water electrolyte was prepared using water quality parameters previously reported for a fresh river water in Wheeling, West Virginia.15 This water matrix has total ionic strength of 5.6 mM, pH of 8.5, total dissolved solid of 219 mg/L, and conductivity of 318 μs/cm (hereafter referred to as “freshwater” or “FW”).16 The average hydrodynamic radii and zeta potential of Ag-TiO2 aggregates were determined by DLS and electrophoresis measurements (BI-90 ZetaPALS, Brookhaven Instruments, NY). The UV vis spectra of Ag-TiO2 suspension were acquired using a Perkin-Elmer Lambda 20 UV vis spectrophotometer. The total Ag content and dissolved Ag+ ion concentrations released from Ag-TiO2 nanoparticles were determined by ICP-OES measurement (TJA RADIAL IRIS 1000). The total Ag contents in Ag-TiO2 samples were measured following US-EPA method 3005. To determine the dissolved Ag+ ion, 25 mg/L of Ag-TiO2 nanoparticles were mixed in assay media for 20 h, the same as the high-throughput assay exposure time. Then, the suspension was 8990
dx.doi.org/10.1021/es201675m |Environ. Sci. Technol. 2011, 45, 8989–8995
Environmental Science & Technology
ARTICLE
Figure 1. TEM images of Ag-TiO2 hybrid nanoparticles and size distribution of deposited Ag nanoparticles. The positions of Ag nanoparticles were indicated by white arrows.
filtered through 0.1 μm filter, and the filtrates were collected for ICP analysis.16 2.4. High-Throughput Bacteria Viability Assay. Gram-positive bacteria B. subtilis and Gram-negative bacteria P. putida were used as model bacteria to examine the bactericidal activities of AgTiO2 nanoparticles. P. putida was cultured in trypticase soy broth (TSB). B. subtilis were cultured in Luria Bertani broth (LB). Bacterial cell culture was grown at 25 °C in an incubator with shaking at 150 rpm and harvested at mid exponential growth phase. Cells were washed twice with phosphate buffered saline (PBS) and resuspended in specific solution to achieve an initial concentration of 108 cells/mL before exposure to Ag-TiO2 nanoparticles. The antibacterial activity of Ag-TiO2 nanoparticles was assessed by a previously developed high throughput bacterial viability assay.16 Silver nitrate salt, one commercial Ag nanoparticle (provided by Quantum Sphere Inc., referred to as Ag-NP), P25 TiO2 and R902 TiO2 nanoparticles were tested for comparison purposes as well. In brief, the viable cell percentage was determined by addition of SYTO 9 and propidium iodide Live/ Dead Baclight bacterial viability kit (Invitrogen, Oregon, USA). To initiate the assay, exactly 25 μL of nanoparticles/salt suspension and 25 μL of bacterial suspensions were dispensed into a 384-well clear bottom polystyrene microplate (Greiner Bio-One; Monroe, North Carolina, USA). After exposing bacteria to nanoparticles, Invitrogen Baclight Live/Dead Kit was added to differentiate live and dead cells accordingly. Simultaneous application of both dyes results in green fluorescence of viable cells with an intact membrane and red fluorescence of dead cells with compromised membranes. The green-to-red fluorescence ratio detected by fluorescence microreader (Flexstation, Molecular Devices; Sunnyvale, California, USA) provided the live bacterial percentage in each well (excitation at 485 nm and emission at 630 and 530 nm). Control replicates (media and bacteria, no nanoparticles) and blank replicates (media and nanoparticles, no bacteria) were tested in the same plate as well. 2.5. Inactivation Kinetic Assay. The Ag-TiO2 bactericidal activities were first evaluated by live/dead inactivation kinetic assays in both light and dark conditions using B. subtilis. A 50:50 mixture of 100 mg/L nanoparticles and 108 cell/mL B. subtilis cells in freshwater was created, and 50 μL was dispersed into each well of 384-well microplates (Greiner Bio-One; Monroe, North
Carolina, USA). Five identical microplates were prepared and placed in UV chamber (UV Stratalinker 2400, 254 nm, 4 mW/cm2), and another set of 5 microplates were placed in the dark. After a given exposure time (i.e., 30, 60, 90, 120, and 180 min) one plate each from the dark and UV-light conditions was examined for cell viability. In control experiments, Degussa TiO2 P25 nanoparticles emitted green fluorescence at 530 nm in the presence of SYTO 9/PI stains, creating false positive signal for viable cell counts. Therefore, bacterial inactivation assays using TiO2 P25 and Ag-P25 nanoparticles were corrected by subtracting the corresponding control experiment signal strength. 2.6. Dose Response Assay. In order to further quantify the antibacterial activity of Ag-TiO2 nanoparticles, a set of dose response assays were performed to determine the half maximal inhibitory concentration (IC50) in the dark. UV irradiation caused nearly complete inactivation within 3 h, so nanoparticle dose response assays were not conducted in the presence of UV light. Specifically, 19 different concentrations (from 2 μg/L to 500 mg/L) of nanoparticles were automatically dispensed into specific wells of a 384-well microplate. After 20-h incubation at 35 °C in the dark, live/dead cell viability molecular probes were added. The live bacteria cell percentage was calculated from the absolute live bacterial percentage in the presence of nanoparticles normalized to the live bacterial percentage for nanoparticle-free control wells. The nanoparticle concentration at which 50% of bacteria cells were dead, i.e., the IC50 value, were determined by sigmoidal dose response curve fitting (Prism 4, GraphPad Software, San Diego, California, USA). Each combination of nanoparticles, concentration, water chemistry, and bacteria was repeated at least twice using independent bacteria culture, and nanoparticle suspensions in separated plates, containing a minimum of 4 replicate wells in each plate.
3. RESULTS 3.1. Nanoparticle Characteristics. TEM images (Figure 1) confirmed that Degussa P25 TiO2 primary particle diameter was around 30 nm, while DuPont R902 TiO2 nanoparticles were 200 450 nm in diameter. The diameter of Ag nanoparticles formed on TiO2 P25 surface was 5 to 20 nm regardless of the deposition method or the Ag content. In contrast, the size of Ag 8991
dx.doi.org/10.1021/es201675m |Environ. Sci. Technol. 2011, 45, 8989–8995
Environmental Science & Technology
Figure 2. Aggregation profile TiO2 nanoparticles and Ag-TiO2 nanoparticles in synthetic freshwater as function of time as determined by dynamic light scattering.
nanoparticles formed on DuPont R902 TiO2 nanoparticles was 10 to 80 nm depending on Ag loading (Figure 1). X-ray diffraction patterns (XRD) (Supporting Information Figure S1) revealed that all Degussa P25 based Ag-TiO2 nanoparticles had a hybrid structure of anatase (Joint Committee on Powder Diffraction Standards, JCPDS 21-1272) and rutile (JCPDS 03-1122) phase. All DuPont R902 based nanoparticles showed a pure rutile phase. In samples with high Ag loading, 1:10 Ag:TiO2 mol/mol, metallic Ag peak (JCPDS 04-0783) was observed in XRD spectra. However, no metallic Ag peak was identified in low Ag loading samples (1:50 Ag:TiO2, mol/mol) due to the low Ag contents (<1 w%).12 Both zeta potential and dispersed size were characterized in a synthetic freshwater media and in deionized water (Table 1). Zeta potential of bare TiO2 nanoparticles, P25 and R902, were 5.8 mV and 0.8 mV, respectively, in deionized water. Zeta potential of high Ag content Ag-P25 and Ag-R902 nanoparticles was more negative at 32 mV and 28 mV, respectively, in deionized water, which was close to the zeta potential of Ag nanoparticles at neutral pH range.16,17 The zeta potential of low Ag content Ag-TiO2 nanoparticles were close to the value of the based TiO2 particles. Most nanoparticles appeared more negatively charged in freshwater (pH 8.5) than in deionized water (pH 5.8) probably due to differences in pH.
ARTICLE
Figure 3. Viable cell percentage as function of exposure time under UV light irradiate (a) Degussa P25 based nanoparticles and (b) DuPont R902 based nanoparticles. (Ag-NP refers to the pure Ag nanopariticles provided by Quantum Sphere Inc.).
In general, all the nanoparticles were reasonably stable in deionized water (Figure 2). However, once exposed to freshwater containing divalent ions, P25 and Ag-P25 nanoparticles aggregated rapidly to micrometer sizes, which is consistent with previous studies.18 The aggregate size of these nanoparticles increased from 200 nm to about 1200 nm in only 15 min (Figure 2a). One exception was the high Ag loading due to the strong electrostatic repulsive forces imparted by high content of Ag nanoparticles on the surface. In contrast, TiO2 R902 based Ag-R902 nanoparticles were relatively stable in both solutions (Figure 2b). The UV vis spectra suggest that Degussa P25 based Ag-P25 nanoparticles exhibited increased reactivity in UV regime (Supporting Information Figure S2), indicating UV photoactivity.7 Moreover, for high Ag content samples a peak near 450 nm appeared, which corresponds to Ag nanoparticles.19,20 However, there was no obvious change of UV vis profile for DuPont R902 with deposition of Ag nanoparticles. 3.2. Kinetics of Bacterial Inactivation. Two sets of inactivation kinetics assays were conducted to evaluate the B. subtilis inactivation rate in both UV-light and in the dark. For the assay performed with UV light, both UV irradiation and Ag/UV exerted strong bacterial inactivation (Figure 3); however, the presence of TiO2 P25 nanoparticles enhanced the bacteria 8992
dx.doi.org/10.1021/es201675m |Environ. Sci. Technol. 2011, 45, 8989–8995
Environmental Science & Technology
ARTICLE
Figure 5. IC50 values (expressed as total Ag added at the initiation of assays) of various Ag-TiO2 nanoparticles and controls in dark condition. (*IC50 > 500 mg/L, no IC50 value could be determined up to the maximum dose of 500 mg/L.)
Figure 4. Viable cell percentage as function of exposure time in dark condition (a) Degussa P25 based nanoparticles and (b) DuPont R902 based nanoparticles.
inactivation rate and extent significantly. The enhanced bacterial inactivation by TiO2 P25 nanoparticles was attributed to ROS generation. Irradiation of TiO2 with UV causes formation of various reactive oxygen species (ROS).2,3 Although the bacteria can fight off low levels of ROS via antioxidant defenses such as glutathione/glutathione disulfide (GSH/GSSG) ratio, excess ROS may produce oxidative stress and attack membrane lipids, ultimately, leading to membrane or DNA damage.4,21 In the same assays, P25 based Ag-P25 nanoparticles intensified bactericidal effects comparing to pure TiO2 P25/UV and Ag/UV systems, possibly due to the enhanced ROS generation and release of Ag+ ion. In contrast, only high Ag content R902 based nanoparticles significantly increased inactivation rates; this was most likely due to the release of Ag+ ions (Figure 3b). For inactivation kinetics assays conducted in the dark, both pure TiO2 nanoparticles slightly enhanced cell inactivation (Figure 4). This was possibly due to the adsorption of bacteria cell onto TiO2 surfaces; the stress caused by the adsorption may increase cell mortality. Pure silver nanoparticles reduced the
viable cell count to 28% within 15 min, which was in agreement with previous results by others.4 All Ag-TiO2 hybrid nanoparticles inactivated the bacterial cells as fast or faster than pure Ag nanoparticles. In particular, Ag-P25 nanoparticles exerted greater bactericidal effects than pure Ag nanoparticles in the dark (Figure 4a). 3.3. Dose Response Assay in the Dark. Dose response experiments were conducted in dark conditions to establish quantification the intrinsic bactericidal activity of all nanoparticles. Dose response curves were determined for B. subtilis and P. putida through a 20-h live/dead assay. The IC50 values were reported as total mass of silver actually added in each assay (Figure 5). Both commercial TiO2 nanoparticles had very limited bactericidal activity even at the highest concentration of 500 mg/L in the absence of light. In fact, an IC50 value could not be determined. In contrast, Ag nanoparticles and AgNO3 salt inactivated both types of bacteria effectively in the dark condition. Particularly, the IC50 values of AgNO3 salt were 0.69 and 1.02 mg/L for B. subtilis and P. putida, respectively; more than 1 order of magnitude lower than Ag nanoparticles. This indicated that AgNO3 salt was a more effective bactericide than Ag-NPs. All Ag-TiO2 nanoparticles exerted stronger bactericidal effects than Ag nanoparticles and, in most cases, stronger than AgNO3 salt. Also, Degussa P25 based Ag-P25 nanoparticles appeared more effective than DuPont R902 based Ag-TiO2 nanoparticles. However, the Ag deposition methods and Ag content did not significantly influence Ag-TiO2 bactericidal activity. In general, P. putida was more resistant than B. subtilis to AgNO3, Ag-NP, and Ag-TiO2 bactericidal activity (Figure 5). This difference in antibacterial resistance between Gram-negative and Gram-positive bacteria is in agreement with previously published results.16,22 24 The tolerance of Gram-negative bacteria to various nanoparticles might be attributed to the lipopolysaccharide (LPS) layer, providing bacteria with a protective outer cell membrane. The LPS layer not only contributes to cell integrity but also complexes Ag+ ions and hinders uptake of Ag+ and/or Ag-NPs.23,24 It is not clear if this trend is universal for all Gram-negative and Gram-positive species. 8993
dx.doi.org/10.1021/es201675m |Environ. Sci. Technol. 2011, 45, 8989–8995
Environmental Science & Technology Free Ag+ ion concentrations were determined for all nanoparticles and AgNO3 salts in the water chemistry of the live/dead assays (Figure 5). Silver ions were sparingly soluble in synthetic freshwater due to the presence of inorganic ligands, such as chloride, sulfate, and hydroxides.16 Free Ag ions released by AgP25 hybrid nanoparticles were slightly higher than those determined from Ag-R902 hybrid nanoparticles but were significantly lower than pure Ag nanoparticles and AgNO3 systems.
4. DISCUSSION Both wet-impregnation and photoreduction proved simple and efficient methods to synthesize highly bactericidal Ag-TiO2 nanoparticles. Silver nanoparticles formed on Degussa P25 were smaller than those on DuPont R902 TiO2 nanoparticles. Previous Ag deposition studies indicate that the size and density of Ag particles deposited on TiO2 surface are sensitive to titania crystal structure (rutile vs anatase), primary particle size, and crystallographic orientation, which fundamentally controls the surface energy, the rate of Ag(0) seed formation, and Ag particle growth.25,26 Chan and Barteau demonstrate that the size of Ag nanoparticles formed was proportional to the size of base TiO2 materials, which consists with the observation in this study.9 As a highly reactive photocatalyst, anatase/rutile hybrid TiO2 P25 nanoparticles are very efficient in generating free electrons which reduce surface-bound Ag+ ions and produce Ag(0) seeds. Therefore, the generation rate and density of Ag(0) seeds on P25 surface are expected to be higher than those on the inert rutile R902 TiO2 nanoparticles. Consequently, for P25 nanoparticles, growth of Ag nanoparticles is restricted by the high nucleation rate and smaller, more uniformly distributed Ag nanoparticles formed regardless of the Ag loading. On the other hand, the Ag(0) seeds on R902 particles surface are limited, which favors growth of Ag particles leading to larger Ag particles with increase of Ag loading.9 High-throughput cell viability assays demonstrated that hybrid Ag-TiO2 materials based on Degussa P25 TiO2 produced synergistic bacterial inactivation rates and extents in both light and dark conditions. The mixed anatase/rutile crystal structure of Degussa P25 nanoparticles makes them highly photoactive and capable of creating electron hole pairs that generate ROS in water. The intensified bactericidal activity of Ag-P25 under UV light suggests that Ag nanoparticles deposited on TiO2 P25 act as electron traps preventing electron hole recombination.7,12,19 Photoexited electrons generated are transferred to metallic nanoparticles, instead of losing their activities by recombination. This extends the life of electron hole pairs and promotes photocatalytic activity of Ag-P25 hybrid materials.19 Thus, ROS generation by Ag-P25 and subsequent damage to the cells is enhanced.19 Because R902 TiO2 nanoparticles are not photoactive, R902 TiO2 nanoparticles did not enhance inactivation rates over UV alone. Hence, Ag-R902 hybrid nanoparticles exhibited similar bacteria inactivation rate as pure Ag nanoparticles under UV irradiation. For the experiments carried out in the dark, both kinetic and IC50 data indicated Ag-TiO2 nanoparticles were more effective at inactivating bacteria than either pure Ag nanoparticles or TiO2 nanoparticles alone. Since TiO2 nanoparticles showed almost no toxicity, the substantially lower IC50 of Ag-TiO2 nanoparticles suggests a synergistic effect. The Ag-enhanced bactericidal activity may have arisen from less dissolution of Ag+ ions and reprecipitated into forms like Ag(0) and AgCl(s). In Ag+ dissolution experiments, the free Ag+ ion concentration was
ARTICLE
only ∼40 μg/L. Previous simulations show that the maximum solubility of Ag+ in water chemistry of the live/dead assay is ∼75 μg/L, which is close to the observed free Ag+ concentration for both Ag nanoparticles and AgNO3 salts.16 Given the IC50 for AgNO3 is 690 μg/L for B. subtilis and 1020 μg/L for P. putida, respectively (expressed as total Ag added), the contribution of free Ag+ seems negligible. This suggested that the Ag+ ions dissociated from AgNO3 salts was not the dominant mechanism in antibacterial activities. Rather, the dissociated Ag+ ions undergo a series of precipitation and reduction reactions with anions and other ligands in water. The presence of both Ag(0) and AgCl in such freshwater media was reported previously.16 Moreover, AgCl and AgCl2 toxicity to bacteria and other aquatic invertebrates can be as significant as free Ag+ ion and Ag(0) nanoparticles.27,28 Thus, we conclude that the toxicity of AgNO3 salt control resulted from a mixture of dissolved Ag+, AgCl, and AgCl2 as well as AgCl precipitates and Ag(0) reprecipitates.16,27 As for the Ag nanoparticles and Ag-TiO2 nanoparticles, similar dissolution and reprecipitation reactions are possible, but the kinetics and equilibrium might be different. The Ag nanoparticles dissolve rapidly and then, like AgNO3 salts, precipitate as AgCl settling out from the aqueous phase, whereas Ag-TiO2 nanoparticles produced much lower concentration of dissolved Ag+ ions (Figure 5), suggesting the releasing of Ag+ ion was not the dominant inactivation mechanism for Ag-TiO2 hybrid particles. Thus, the observed synergistic effects in dark condition were most likely attributed to the variation in the dissolution and reprecipitation kinetic and equilibrium between pure Ag nanoparticles and Ag-TiO2 nanoparticles. The Ag-P25 particles exhibited different mechanisms of bacterial inactivation in light and dark conditions. Given the highly photoactive nature of Degussa P25 nanoparticles, the inactivation mechanism under UV-light was largely due to the generation of reactive oxygen species (ROS). In the absence of light, the inactivation mechanism was mainly due to direct contact with Ag nanoparticles and formation of toxic Ag species such as Ag+, Ag(0), AgCl, and AgCl2 .
’ ASSOCIATED CONTENT
bS
Supporting Information. X-ray diffraction (XRD) and UV vis spectra of Ag-TiO2nanparticles. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (310) 206-3735. Fax: (310) 206-2222. E-mail: emvhoek@ ucla.edu. Corresponding author address: University of California, Los Angeles, 5732 Boelter Hall, P.O. Box 951593, Los Angeles, CA 90095-1593.
’ ACKNOWLEDGMENT This research was supported in part by the National Science Foundation and the Environmental Protection Agency through the UC Center for Environmental Implications of Nanotechnology (Cooperative Agreement Number EF 0830117) as well as UC Discovery Industry-University Cooperative Research Program (#GCP07-10239A) with industrial matching funds provided by NanoH2O Inc. 8994
dx.doi.org/10.1021/es201675m |Environ. Sci. Technol. 2011, 45, 8989–8995
Environmental Science & Technology
’ REFERENCES (1) Gumy, D.; Morais, C.; Bowen, P.; Pulgarin, C.; Giraldo, S.; Hajdu, R.; Kiwi, J. Catalytic activity of commercial of TiO2 powders for the abatement of the bacteria (E-coli) under solar simulated light: Influence of the isoelectric point. Appl. Catal., B 2006, 63 (1 2), 76–84. (2) Cho, M.; Chung, H.; Choi, W.; Yoon, J. Linear correlation between inactivation of E-coli and OH radical concentration in TiO2 photocatalytic disinfection. Water Res. 2004, 38 (4), 1069–1077. (3) Yao, Y.; Ohko, Y.; Sekiguchi, Y.; Fujishima, A.; Kubota, Y. Selfsterilization using silicone catheters coated with Ag and TiO2 nanocomposite thin film. J. Biomed. Mater. Res. B 2008, 85B (2), 453–460. (4) Marambio-Jones, C.; Hoek, E. M. V. A review of the antibacterial effects of silver nanomaterials and potential implications for human health and the environment. J. Nanopart. Res. 2010, 12 (5), 1531–1551. (5) Choi, O.; Hu, Z. Q. Size dependent and reactive oxygen species related nanosilver toxicity to nitrifying bacteria. Environ. Sci. Technol. 2008, 42 (12), 4583–4588. (6) Morones, J. R.; Elechiguerra, J. L.; Camacho, A.; Holt, K.; Kouri, J. B.; Ramirez, J. T.; Yacaman, M. J. The bactericidal effect of silver nanoparticles. Nanotechnology 2005, 16 (10), 2346–2353. (7) Page, K.; Palgrave, R. G.; Parkin, I. P.; Wilson, M.; Savin, S. L. P.; Chadwick, A. V. Titania and silver-titania composite films on glasspotent antimicrobial coatings. J. Mater. Chem. 2007, 17 (1), 95–104. (8) Ma, N.; Fan, X. F.; Quan, X.; Zhang, Y. B. Ag-TiO2/HAP/Al2O3 bioceramic composite membrane: Fabrication, characterization and bactericidal activity. J. Membr. Sci. 2009, 336 (1 2), 109–117. (9) Chan, S. C.; Barteau, M. A. Preparation of highly uniform Ag/TiO2 and Au/TiO2 supported nanoparticle catalysts by photodeposition. Langmuir 2005, 21 (12), 5588–5595. (10) Chen, M. X.; Yan, L. Z.; He, H.; Chang, Q. Y.; Yu, Y. B.; Qu, J. H. Catalytic sterilization of Escherichia coli K 12 on Ag/Al2O3 surface. J. Inorg. Biochem. 2007, 101 (5), 817–823. (11) Nino-Martinez, N.; Martinez-Castanon, G. A.; Aragon-Pina, A.; Martinez-Gutierrez, F.; Martinez-Mendoza, J. R.; Ruiz, F., Characterization of silver nanoparticles synthesized on titanium dioxide fine particles. Nanotechnology 2008,19, (6), -. (12) Reddy, M. P.; Venugopal, A.; Subrahmanyam, M. Hydroxyapatite-supported Ag-TiO2 as Escherichia coli disinfection photocatalyst. Water Res. 2007, 41 (2), 379–386. (13) Akhavan, O.; Ghaderi, E. Capping antibacterial Ag nanorods aligned on Ti interlayer by mesoporous TiO2 layer. Surf. Coat. Technol. 2009, 203 (20 21), 3123–3128. (14) Martinez-Castanon, G. A.; Nino-Martinez, N.; Loyola-Rodriguez, J. P.; Patino-Marin, N.; Martinez-Mendoza, J. R.; Ruiz, F. Synthesis of silver particles with different sizes and morphologies. Mater. Lett. 2009, 63 (15), 1266–1268. (15) Tchobanoglous, G.; Schroeder, D. Water quality: Characteristics, modeling, modification; Addison-Wesley Pub. Co.: Reading, MA, 1987. (16) Jin, X.; Li, M.; Wang, J.; Marambio Jones, C.; Peng, F.; Huang, X.; Damoiseaux, R.; Hoek, E. High-throughput Screening of Silver Nanoparticle Stability and Bacterial Inactivation in Aquatic Media: Influence of Specific Ions. Environ. Sci. Technol. 2010, 44 (19), 7321–7328. (17) Sondi, I.; Goia, D. V.; Matijevic, E. Preparation of highly concentrated stable dispersions of uniform silver nanoparticles. J. Colloid Interface Sci. 2003, 260 (1), 75–81. (18) Ji, Z.; Jin, X.; George, S.; Xia, T.; Meng, H.; Wang, X.; Suarez, E.; Zhang, H.; Hoek, E. M. V.; Godwin, H.; Nel, A. E.; Zink, J. I. Dispersion and Stability Optimization of TiO2 Nanoparticles in Cell Culture Media. Environ. Sci. Technol. 2010, 44 (19), 7309–7314. (19) Yu, J. G.; Xiong, J. F.; Cheng, B.; Liu, S. W. Fabrication and characterization of Ag-TiO2 multiphase nanocomposite thin films with enhanced photocatalytic activity. Appl. Catal., 2005, 60 (3 4), 211–221. (20) Akhavan, O. Lasting antibacterial activities of Ag-TiO2/Ag/aTiO2 nanocomposite thin film photocatalysts under solar light irradiation. J. Colloid Interface Sci. 2009, 336 (1), 117–124.
ARTICLE
(21) Nel, A. E.; Madler, L.; Velegol, D.; Xia, T.; Hoek, E. M. V.; Somasundaran, P.; Klaessig, F.; Castranova, V.; Thompson, M. Understanding biophysicochemical interactions at the nano-bio interface. Nat. Mater. 2009, 8 (7), 543–557. (22) Li, M.; Pokhrel, S.; Jin, X.; Madler, L.; Damoiseaux, R.; Hoek, E. M. Stability, bioavailability, and bacterial toxicity of ZnO and irondoped ZnO nanoparticles in aquatic media. Environ. Sci. Technol. 2011, 45 (2), 755–61. (23) Kucerka, N.; Papp-Szabo, E.; Nieh, M. P.; Harroun, T. A.; Schooling, S. R.; Pencer, J.; Nicholson, E. A.; Beveridge, T. J.; Katsaras, J. Effect of cations on the structure of bilayers formed by lipopolysaccharides isolated from Pseudomonas aeruginosa PAO1. J. Phys. Chem. B 2008, 112 (27), 8057–8062. (24) Guine, V.; Spadini, L.; Sarret, G.; Muris, M.; Delolme, C.; Gaudet, J. P.; Martins, J. M. F. Zinc sorption to three gram-negative bacteria: Combined titration, modeling, and EXAFS study. Environ. Sci. Technol. 2006, 40 (6), 1806–1813. (25) Farneth, W. E.; McLean, R. S.; Bolt, J. D.; Dokou, E.; Barteau, M. A. Tapping mode atomic force microscopy studies of the photoreduction of Ag+ on individual submicrometer TiO2 particles. Langmuir 1999, 15 (25), 8569–8573. (26) Hotsenpiller, P. A. M.; Bolt, J. D.; Farneth, W. E.; Lowekamp, J. B.; Rohrer, G. S. Orientation dependence of photochemical reactions on TiO2 surfaces. J. Phys. Chem. B 1998, 102 (17), 3216–3226. (27) Choi, O.; Deng, K. K.; Kim, N. J.; Ross, L.; Surampalli, R. Y.; Hu, Z. Q. The inhibitory effects of silver nanoparticles, silver ions, and silver chloride colloids on microbial growth. Water Res. 2008, 42 (12), 3066–3074. (28) Ward, T. J.; Kramer, J. R. Silver speciation during chronic toxicity tests with the mysid, Americamysis bahia. Comp. Biochem. Physiol., Part C: Toxicol. Pharmacol. 2002, 133 (1 2), 75–86.
8995
dx.doi.org/10.1021/es201675m |Environ. Sci. Technol. 2011, 45, 8989–8995
ARTICLE pubs.acs.org/est
Identification of Environmental Reservoirs of Nontyphoidal Salmonellosis: Aptamer-Assisted Bioconcentration and Subsequent Detection of Salmonella Typhimurium by Quantitative Polymerase Chain Reaction Anurag Jyoti,† Poornima Vajpayee,† Gulshan Singh, Chandra Bali Patel, Kailash Chand Gupta, and Rishi Shanker* Environmental Microbiology, CSIR- Indian Institute of Toxicology Research, Post Office Box 80, Mahatma Gandhi Marg, Lucknow-226001, Uttar Pradesh, India
bS Supporting Information ABSTRACT: In this study, identification of environmental reservoirs of Salmonella enterica subsp. enterica serovar Typhimurium (abbreviated as Salmonella Typhimurium) in sediments, water, and aquatic flora collected from the Ganges River (Ganges riverine material) was carried out by adopting a two-step strategy. Step 1 comprised a selective serovar-specific capture of Salmonella Typhimurium from potential reservoirs. Step 2 involved culture-free detection of selectively captured Salmonella Typhimurium by ttr gene-specific molecular beacon (MB) based quantitative polymerase chain reaction (q-PCR). The ttr gene-specific MB designed in this study could detect 1 colony-forming unit (cfu)/PCR captured by serovar-specific DNA aptamer. Sediments, water, and aquatic flora collected from the Ganges River were highly contaminated with Salmonella Typhimurium. The preanalytical step in the form of serovar-specific DNA aptamer-based biocapture of bacterial cells was found to enhance the sensitivity of the fluorescent probe in the presence of nonspecific DNA . Information about the presence of environmental reservoirs of Salmonella Typhimurium in the Ganges River region may pave the way for forecasting and management of nontyphoidal salmonellosis in south Asia.
’ INTRODUCTION Environmental reservoirs are locations outside the human body within the niche favoring bacterial persistence and replication in the environment and pathogen transmission to susceptible hosts.1 In Asia and other parts of the developing world, natural surface water is being used by a large population for drinking and other domestic purposes.2,3 Furthermore, the prevalence of environmental reservoirs in riverine materials and contamination of surface water due to the addition of untreated sewage from nearby residential areas sets them at high risk for water-borne diseases caused by bacterial pathogens.1 Salmonella frequently occurs in sewage-impacted fresh and marine water environments; it causes gastroenteritis and typhoid in humans worldwide, resulting in a high economic burden due to a substantial increase in medical treatment expenditure.4,5 Salmonella enterica, which causes salmonellosis in humans and animals worldwide, is classified into more than 2400 Salmonella serotypes.6 Salmonella enterica subsp. enterica serovar Typhimurium (abbreviated as Salmonella Typhimurium), an important food and freshwater contaminant encountered frequently in most geographical regions, dominates the global epidemiology of Salmonella.7 The infectious dose of Salmonella Typhimurium lies between 1 and 10 colony-forming units (cfu).8 r 2011 American Chemical Society
Out of the 1.3 billion incidents of human nontyphoidal salmonellosis reported annually worldwide, approximately 3 million cases end with the patient’s death from the disease.9 However, the mortality rate in developed countries (2%) is less than in resource-constrained countries (1824%). Agricultural runoff exhibiting poultry waste, bovine manure, and feces of other asymptomatic animals is an important carrier of Salmonella Typhimurium to riverine materials. In addition, agricultural produces from crops irrigated by contaminated water have also been associated with a large number of outbreaks.10 Hence, the presence of Salmonella Typhimurium in environmental reservoirs, even in low numbers, may cause a public health risk of nontyphoidal salmonellosis. Conventional culture-based protocols for detection and quantification of Salmonella Typhimurium in environmental samples require up to 5 days to obtain a confirmed result and are executable only in samples where the pathogen registers its presence in very high numbers, that is, 102103 cfu/g or 102103 cfu/mL.11,12 Quantitative polymerase Received: June 3, 2011 Accepted: August 29, 2011 Revised: August 24, 2011 Published: August 29, 2011 8996
dx.doi.org/10.1021/es2018994 | Environ. Sci. Technol. 2011, 45, 8996–9002
Environmental Science & Technology
ARTICLE
Table 1. Selectivity of DNA Aptamer for Salmonella Typhimuriuma detection of Salmonella Typhimurium: threshold cycles (CT)b Ac
Salmonella serovars
Bd
29.77 ( 0.988 30.14 ( 1.245
NDe NDe
Salmonella Infantis MTCC 1167c
30.77 ( 1.132
NDe
Salmonella Typhimurium
29.89 ( 1.027
29.69 ( 0.875
29.97 ( 0.978
29.30 ( 0.998
NDe
NDe
Salmonella Typhi MTCC 733c Salmonella Paratyphi A MTCC 735c
ATCC 14028
d
Salmonella Typhimurium ATCC 13311d sterile Milli-Q waterf
106 cells/mL or 104 cells/PCR. b Each assay was conducted with five independent replicates on the same day, and values are mean CT ( standard deviation (SD). c DNA was prepared directly from spiked water samples without bioconcentration of Salmonella Typhimurium cells by Sal aptamer capture of Salmonella Typhimurium cells. d Salmonella Typhimurium cells were bioconcentrated from spiked water samples by serovar-specific DNA aptamer (Sal aptamer) prior to DNA extraction. e ND = CT not detected. f Sterile Milli-Q (Millipore, Billerica, MA) served as negative control. a
chain reaction (q-PCR) assays targeting the invA gene (present in a wide range of Salmonella serotypes, including all subspecies, and absent in other bacterial species and genera) offers several advantages compared to classical bacteriology in terms of speed, detection limit, potential for automation, and cost.7,1214 However, it has been documented that some Salmonella strains exhibit natural deletions within Salmonella pathogenicity island 1, where the inv, spa, and hil loci occur.15 Therefore, q-PCR assays targeting the invA gene could lead to false negative results for such strains.16 Besides, the sensitivity of q-PCR reported for detection of Salmonella from water declines 10fold in the presence of nonspecific DNA.7 Further, these culturefree assays provide information about total Salmonella concentration and fail to detect serovar-specific contamination in environmental samples. Therefore, there is a need for selection of a target that is highly genetically stable in all Salmonella strains. Furthermore, the inclusion of preanalytical sample cleanup steps, consisting of selective serovar-specific capture of target pathogen to improve the sensitivity of q-PCR-based assays for culture-free detection of Salmonella serovars from sewage-impacted surface water and other complex environmental matrices, is required. Aptamers can be selected for a wide range of targets, including bacterial pathogens, through Systematic Evolution of Ligands by EXponential enrichment (SELEX), which involves iterative cycles of absorption, recovery of bound DNA/RNA, and amplification from a combinatorial DNA/RNA library containing a central randomized sequence and flanking fixed sequences (primers) on both sides for amplification.17,18 Recently, it has been reported that aptamers can selectively capture and concentrate Salmonella Typhimurium from spiked chicken feces.19 However, aptamers have not been used yet for capture and selective concentration of any Salmonella serovar from samples retrieved from a natural environment contaminated by untreated sewage. The ttr locus, comprising five genes (ttrA, ttrB, and ttrC, which encode the tetrathionate reductase structural proteins, and ttrS and ttrR, which
encode the sensor and response regulator components of a twocomponent regulatory system) responsible for tetrathionate respiration, is genetically stable in all Salmonella strains.16 Hence, in this study, identification of environmental reservoirs of Salmonella Typhimurium in Ganges riverine samples was carried out by adopting a two-step strategy. Step1 comprised a selective serovar-specific capture of Salmonella Typhimurium from potential environmental reservoirs. Step 2 involved culture-free detection of selectively captured Salmonella Typhimurium by a ttr genespecific molecular beacon (MB) based q-PCR, designed in a highly conserved region of the Salmonella-specific ttr locus.
’ MATERIALS AND METHODS Bacterial Strains. Salmonella enterica subsp. enterica serovar Typhimurium ATCC 14028 (abbreviated as Salmonella Typhimurium) and Escherichia coli DH5α ATCC 35218 were procured from the American Type Culture Collection (ATCC). However, Escherichia coli MTCC 723, Salmonella Paratyphi A MTCC 735, Salmonella Typhi MTCC 733, and Salmonella Infantis MTCC 1167 were procured from the Microbial Type Culture Collection at the Institute of Microbial Technology, Chandigarh, India. Aptamer, Primers, and Molecular Beacon. A DNA aptamer (Sal aptamer: biotin 50 -TATGGCGGCGTCACCCGACGGGGACTTGACATTATGACAG-30 ) reported to capture Salmonella Typhimurium from spiked chiken feces19 was biotinylated at the 50 end to bioconcentrate Salmonella Typhimurium from sewage-impacted surface water. Primers (forward primer ttr6, 50 -CTCACCAGGAGATTACAACATGG-30 , position 42874309 bp; reverse primer ttr4, 50 -AGCTCAGACCAAAAGTGACCATC-30 , position 43594381 bp; product length 95 bp) specific for the ttr gene were adopted from Malorny et al.16 These primers are located in a highly conserved region of the Salmonella-specific ttr locus and were designed on the basis of a multiple alignment of the ttrBCA sequences.16 After multiple alignment of the ttrBCA sequences, a molecular beacon (50 -CCAGGCGACCGACTTTTAGCCACTGACGAGCCTGG-30 ) targeting the ttr gene corresponding to adopted primers was designed in the ttr locus at the ttrA gene (position 43234345 bp) for culture-free quantitative enumeration of Salmonella Typhimurium to identify environmental reservoirs present in Ganges riverine materials impacted by sewage discharge (Supporting Information). The primers, probe, and DNA aptamer were purchased from MetaBion GmbH, Germany. Selectivity of DNA Aptamer for Salmonella Typhimurium. To test the selectivity of the Sal aptamer for bioconcentration of Salmonella Typhimurium, several representative serovars of Salmonella were selected (Table 1). Cultures of the reference strains of these serovars grown in LuriaBertani broth to 0.8 optical density, corresponding to 109 cells/mL. Two sets of 10 mL of sterile water (Milli-Q, Millipore, Billerica, MA) were spiked with 106 cfu/mL of each reference strain to get 104 cfu/PCR in a 25 μL reaction. Cultures were centrifuged at 8000g for 4 min (4 °C). Cells were washed three times with 1 phosphate-buffered saline (PBS) and finally, resuspended in 5 mL of 1 PBST (1 PBS contains 1.47 mM KH2PO4, 2.7 mM KCl, 137 mM NaCl, and 4.3 mM Na2HPO4, pH 7.4; 1 PBST additionally contains 0.5% Tween-20, v/v). Aptamer-conjugated Dynabeads (300 μL) were mixed in resuspended cells and incubated for 30 min at room temperature. Dynabeads were concentrated and supernatant was discarded. Dynabeads were washed 10 times with 1 PBST. Finally, beads were resuspeneded in 250 μL of PBS and stored at 4 °C for further processing. DNA template was 8997
dx.doi.org/10.1021/es2018994 |Environ. Sci. Technol. 2011, 45, 8996–9002
Environmental Science & Technology prepared by boiling the Dynabeads suspended in 250 μL of PBS. Supernatant was collected through aspiration by pipet after concentration of Dynabeads in a concentrator. The cellular debris was removed by centrifugation at 16000g for 5 min at 4 °C. DNA was precipitated from supernatant by use of sodium acetate (0.3 M, pH 5.2) and ice-cold ethanol.7 The precipitated DNA was pelleted by centrifugation at 12000g for 5 min. DNA pellet was washed three times with 70% ethanol and finally dissolved in 50 μL of TE buffer [10 mM tris(hydroxymethyl) aminomethane (Tris) and 1 mM ethylenediaminetetraacetic acid (EDTA), pH 8.0]. DNA was prepared directly from another set of 10-fold diluted culture without Sal aptamer bioconcentration. The quantity and quality of extracted DNA was measured at 260/280 nm with a NanoDrop spectrophotometer (NanoDrop Technologies, Wilmington, DE). DNA (5 μL) was used as template in a real-time q-PCR assay for detection and quantitative enumeration of Salmonella Typhimurium. Each assay was conducted with five independent replicates on the same day, and values are mean CT ( standard deviation (SD). Bioconcentration of Salmonella Typhimurium from Potential Environmental Reservoirs for Culture-Free Detection. An Indian perennial river (Ganges River, Kanpur city, latitude 26.28 N; longitude 80.24 E; altitude 126 m; flowing through northern India) was selected to test the applicability of Sal aptamer for bioconcentration of Salmonella Typhimurium from potential environmental reservoirs and subsequent detection through culture-free MB based q-PCR assay. The Ganges River receives untreated sewage generated by nearby residential setup of urban population in the boundaries of the Kanpur and Unnao cities.20 Replicate (n = 5) grab surface water samples (1 L) were collected in sterilized bottles from six sites (Figure S2, Supporting Information) in the Ganges River covering a 30 km stretch (site 1, Bithur upstream; site 2, Ganges barrage; site 3, Bhairoghat (cremation ghat); site 4, Shuklaganj; site 5, Nanaraoghat; site 6, Jajmau). Surface sediments (05 cm, ∼250 g) were collected at each sampling location (n = 5) from river bed (∼1 m distance) close to the banks with a stainless steel scoop and placed in plastic bags. Aquatic plants/algae growing any at the selected sites were also collected and placed in polythene bags. The samples were collected on the same day, transported on ice, and analyzed immediately after arrival in the laboratory. Water/sediment and aquatic flora (n = 5 each) were also collected from a pristine site negative for Salmonella contamination. Two sets (each comprising five replicates) of 500 mL aliquots of each water sample from each site were concentrated to 5 mL by repeated centrifugation at 18000g for 10 min (4 °C). Cells were pelleted from concentrated samples by centrifugation (8000g for 4 min at 4 °C) and finally resuspended in 1 PBS (5 mL) after three successive washings by same buffer. Similarly, two sets (each comprising five replicates) of 10 g sediment sample collected from each site were dissolved in 100 mL of phosphate-buffered saline. Bacteria were released by vigorous shaking (10 min at 220 rev 3 min1) on a refrigerated rotary shaker (Innova 4230, New Brunswick, NJ). Supernatant from each set of samples was collected separately and concentrated to 5 mL by repeated centrifugation at 18000g for 10 min (4 °C). Cells were pelleted from concentrated samples by centrifugation (8000g for 4 min at 4 °C) and finally resuspended in 1 PBST (5 mL) after three successive washings by the same buffer. Plant samples (roots/leaf/algal mat) collected from each site including the pristine site were processed for releasing the bacteria adhered to the plants as described by Ram et al.21 In brief, bacteria adhering to each plant sample (10 g) were released in
ARTICLE
100 mL of phosphate-buffered saline through repeated washing by vigorous shaking (2 min at 220 rev 3 min1) on a refrigerated rotary shaker (Innova 4230, New Brunswick, NJ) and 45 s of centrifugation at 2000 rpm (653g). Two sets of each sample aliquot (100 mL) of phosphate-buffered saline containing bacteria released from plants and 100 mL of each surface water sample collected from each site in the Ganges River were concentrated to 5 mL by repeated centrifugation at 18000g for 10 min (4 °C). Cells were pelleted from concentrated samples by centrifugation (8000g for 4 min at 4 °C) and finally resuspended in 1 PBST (5 mL) after three successive washings in 1 PBS. One set of environmental samples (plant/water/sediment) was subjected to bioconcentration of Salmonella Typhimurium by use of the Sal aptamer from water samples containing various pathogenic and nonpathogenic bacteria. Briefly, 5 mL concentrated samples were exposed to 0.4 nM aptamer conjugated to Dynabeads (suspended in 300 μL of 1 PBST). Samples were incubated for 30 min at room temperature. Dynabeads were concentrated on a magnetic particle concentrator (Dyna MagTM 2, Invitrogen), and DNA was prepared as described in Supporting Information. The other set of samples (plant/water/sediment), comprising cells suspended in 5 mL of 1 PBS, was processed for isolation of multigenomic DNA through boiling prep as described by Jyoti et al.7 Briefly, DNA template was prepared by boiling 500 μL of concentrated water and removing the debris by centrifugation at 16000g for 5 min at 4 °C. DNA was precipitated from supernatant by use of sodium acetate (0.3 M, pH 5.2) and icecold ethanol.7 The precipitated DNA was pelleted by centrifugation at 12000g for 5 min. The DNA pellet was washed three times with 70% ethanol and finally dissolved in 250 μL of TE buffer. DNA (5 μL) was used as template in 25 μL q-PCR assays (Supporting Information). Quantitative enumeration of Salmonella Typhimurium at each sampling location in potential environmental reservoirs was carried out by use of a standard curve prepared from 10-fold diluted genomic DNA of Salmonella Typhimurium ATCC 14028 (from 1 106 down to 1 100 genomic equivalent (GE)/PCR). Statistical Analysis. For comparison of PCR amplification efficiencies and detection sensitivities among different experiments, slopes of the standard curves were calculated by performing a correlation and regression analysis through iCycle iQ realtime detection system software, version 3.0A. Amplification efficiency (E) was estimated by using the slope of the standard curve and the formula E = [10(1/slope)]1. A reaction with theoretical 100% efficiency will generate a slope of 3.322. The comparison of pollution level of six sites in the Ganges River, in terms of Salmonella Typhimurium load, was performed by oneway analysis of variance (ANOVA).22 Duncan’s multiple range test was used to compare the means.23
’ RESULTS Sensitivity and Specificity of the ttr Gene-Targeting Molecular Beacon. q-PCR in the molecular beacon format de-
scribed here is highly sensitive and specific to Salmonella species. The five reference strains and 15 isolates of Salmonella retrieved from potable water collected from a major city of northern India were positive for the ttr gene (Table S1, Supporting Information). However, no amplification of the ttr gene was observed in E. coli and Enterococcus reference and environmental isolates, lacking the target gene with the assay developed in this study (Table S1, Supporting Information). The detection probability 8998
dx.doi.org/10.1021/es2018994 |Environ. Sci. Technol. 2011, 45, 8996–9002
Environmental Science & Technology
ARTICLE
Table 2. Contamination of Ganges River by Salmonella spp. and Salmonella Typhimuriuma quantitative enumeration of Salmonella spp. and Salmonella Typhimurium in Ganges Riverb water sitesc
Salmonella spp. per 100 mLd
site 1
(2.62 103) ( (8.8 101) f
site 2
sediment Salmonella spp. per 100 gd
Salmonella Typhimurium per 100 ge
(2.37 102) ( (1.1 101) f
(4.37 104) ( (2.05 102) f
(3.37 104( ( (1.50 102) f
(3.46 10 ) ( (1.09 10 ) e
(1.96 10 ) ( (8.8 10 ) e
(5.62 10 ) ( (1.40 10 ) e
(1.96 104) ( (1.02 102) e
site 3
(3.99 10 ) ( (1.60 10 ) a
(2.08 10 ) ( (9.7 10 ) a
(8.51 10 ) ( (1.57 10 ) a
(3.08 109) ( (9.7 106) a
site 4
(4.68 10 ) ( (1.02 10 ) d
(2.26 10 ) ( (8.0 10 ) d
(7.68 10 ) ( (2.69 10 ) d
(3.26 104) ( (1.00 102) d
site 5
(4.51 10 ) ( (1.57 10 ) b
(6.64 10 ) ( (6.52 10 ) b
(4.67 10 ) ( (6.64 10 ) b
(7.64 107) ( (6.98 104) b
site 6
(2.67 10 ) ( (6.64 10 ) c
(3.35 10 ) ( (8.67 10 ) c
(3.99 10 ) ( (1.60 10 ) c
(5.35 106) ( (8.67 104) c
4 9 5 8 8
2 7 4 7 6
Salmonella Typhimurium per 100 mLe 3 8 3
1
5
6
9
1
6 6
6
4
9
4
9
4 6 5 6 7
Salmonella Typhimurium is a nontyphoidal salmonellosis-causing serovar. Values are mean (n = 5) ( SD. Sterile Milli-Q water (Millipore, Billerica, MA) was used as negative control. Column-wise one-way ANOVA, performed separately, was significant at 5% level for each column. Lowercase roman letters show significant difference between means in a column according to Duncan’s multiple range test. c Site 1, Bithoor upstream; site 2, Ganges barrage; site 3, Bhairoghat; site 4, Shuklaganj; site 5, Nanaraoghat; site 6, Jajmau. d DNA was prepared directly from water and sediment samples without Sal aptamer bioconcentration for Salmonella Typhimurium cells. e Salmonella Typhimurium cells were bioconcentrated from water and sediment samples by serovar-specific DNA aptamer (Sal aptamer) prior to DNA extraction. a
of the developed assay was 100% for 1 cfu/PCR (Table S2, Supporting Information). However, the presence of background DNA from E. coli DH5α and enterotoxigenic E. coli (Escherichia coli MTCC 723;106 cfu/PCR each) dropped the sensitivity of the assay 1000-fold (Table S4, Supporting Information). Conjugation of Aptamer to Dynabeads and Binding of Salmonella Typhimurium to Aptamer-Conjugated Dynabeads. The observations indicate that approximately 96.75% of the Sal aptamer was conjugated to the streptavidin-coated Dynabeads (Table S3, Supporting Information). Selectivity of DNA Aptamer for Salmonella Typhimurium. The DNA aptamer is highly selective for the Salmonella Typhimurium (Table 1) . No amplification was recorded when Salmonella cells of other serovars were bioconcentrated by the Sal aptamer. However, the MB designed in this study could detect all the Salmonella reference strains tested in this study. The selected DNA aptamer could concentrate only Salmonella Typhimurium and was found to be highly specific to the serovar. Bioconcentration of Salmonella Typhimurium from Water Spiked with Reference Strain in the Presence of Background Bacterial Flora. Aptamer-conjugated Dynabeads could bioconcentrate Salmonella Typhimurium from water samples spiked with 10-fold diluted culture of Salmonella Typhimurium ATCC 14028 as Dynabeads plated on HiCrome improved salmonella agar showed characteristic pink colonies (Figure S1, Supporting Information). The lowest cell concentration captured by aptamer-conjugated Dynabeads and detected by ttr gene-targeted MB-based q-PCR was 1 cfu/PCR from water samples spiked with Salmonella Typhimurium (Table S4, Supporting Information). Further, Sal aptamer was able to bioconcentrate Salmonella Typhimurium cells from the spiked water in the presence of a high background of pathogenic and nonpathogenic bacteria without any drop in detection limit due to presence of nonspecific DNA (Table S4, Supporting Information). Our observations indicate that the detection limit of the ttr gene-targeted Salmonella specific q-PCR drops 1000-fold in the presence of mixed background of E. coli DH5α (108 cfu/mL or 106 cfu/PCR) and enterotoxigenic E. coli (ETEC; 108 cfu/mL or 106 cfu/PCR). Bioconcentration of Salmonella Typhimurium from Potential Environmental Reservoirs and Subsequent CultureFree Detection. All the water and sediment samples collected from the Ganges River were found to be contaminated by Salmonella
b
(Table 2). However, the level of the Salmonella contamination varied significantly (Duncan's multiple range test, p < 0.05) among the sampling locations (Table 2). Water and sediment samples collected from site 3 exhibit a maximum level (water, 3.99 109 cfu/100 mL; sediment, 8.51 109 cfu/100 g) of Salmonella. However, a minimum level (water, 2.62 103 cfu/100 mL; sediment, 4.37 104 cfu/100 g) of Salmonella contamination in both water and sediment of the Ganges River observed at site 1 (Table 2). Furthermore, it has been observed that Salmonella Typhimurium contamination in water samples collected from selected sites ranged between 2.37 102 and 2.08 108 cfu/100 mL. Sediments from the selected sampling locations also exhibit Salmonella Typhimurium in the range of 3.37 104 to 3.08 109 cfu/100 g (Table 2). The level of the Salmonella Typhimurium was highest at site 3. The inclusion of a preanalytical step in the form of cell capture by the serovar- (Salmonella Typhimurium) specific aptamer allowed enumeration of the Salmonella Typhimurium contamination. Sediments were more contaminated than water (Table 2). Aquatic flora collected from the Ganges River were also contaminated by Salmonella Typhimurium (Table 3). Potamogeton crispus exhibited maximum Salmonella and Salmonella Typhimurium contamination, followed by Potamogeton pectinatus and Spirogyra spp. (Table 3).
’ DISCUSSION The detection of low levels of pathogenic bacteria in complex sample matrices such as environmental samples containing numerous polymerase chain reaction inhibitors is challenging due to the drop in sensitivity of rapid molecular fluorescent probe-based techniques.7,16,18 The sensitivity of the invA gene-targeted MBbased q-PCR reported by Jyoti et al.7 dropped 10-fold in the presence if 106 cfu/PCR E. coli DH5α ATCC 35218. In this study, we observed that addition of enterotoxigenic E. coli (106 cfu/ PCR) along with E. coli DH5α ATCC 35218 (106 cfu/PCR) decreased the sensitivity of the ttr gene-specific MB-based q-PCR assay 1000-fold. Therefore, it could be extrapolated that the assay may perform poorly in environmental samples where multigenomic DNA is present. Also, the assay reported by Jyoti et al. 7 and other fluorescent probe-based assays could quantify only levels of the genus Salmonella.7,1214 Further, these assays fail to provide 8999
dx.doi.org/10.1021/es2018994 |Environ. Sci. Technol. 2011, 45, 8996–9002
Environmental Science & Technology
ARTICLE
Table 3. Total Salmonella and Salmonella Typhimurium Contamination in Potential Environmental Reservoirs of the Ganges River at Kanpur quantitative enumeration of Salmonella (cfu/100 g)a aquatic flora
Salmonella spp.b
Salmonella Typhimuriumc
Potamogeton pectinatusd sample 1e
55 360 ( 190 c
sample 2f
68 990 ( 207 b
1384 ( 50 c 3725 ( 130 b
sample 3g
487 999 ( 19201 a
8958 ( 376 a
Potamogeton crispusd sample1 sample 2g
85 365 ( 341 c 785 289 ( 23 558 a
sample 3h
98 259 ( 365 b
e
3172 ( 142 c 10 856 ( 379 a 5436 ( 190 b
Spirogyra spp.d sample 1e
67 251 ( 326 d
4529 ( 262 d
sample 2f
82 821 ( 4126 c
7523 ( 300 c
sample 3i
282 652 ( 11 306 a
46 689 ( 2120 a
sample 4j
125 689 ( 6081 b
25 826 ( 1033 b
a Values are mean (n = 5) ( SD. Sterile Milli-Q water (Millipore, Billerica, MA) was used as negative control in all experiments. Columnwise one-way ANOVA, performed separately for each aquatic macrophyte and for Spirogyra spp., was significant at 5% level. Lowercase Roman letters in each column for each aquatic macrophyte and for Spirogyra spp. show significant difference between means according to Duncan’s multiple range test. b DNA was prepared directly from water and sediment samples without Sal aptamer bioconcentration for Salmonella Typhimurium cells. c Salmonella Typhimurium cells were bioconcentrated from water and sediment samples by serovar-specific DNA aptamer (Sal aptamer) prior to DNA extraction. d Collected from a pristine area that exhibited no Salmonella contamination. e Site 1, Bithur upstream. f Site 2, Ganges barrage. g Site 3, Bhairoghat (cremation ghat). h Site 4, Shuklaganj. i Site 5, Nanaraoghat. j Site 6, Jajmau.
serovar-specific information on Salmonella contamination in the environmental samples. Therefore, sensitive and robust preanalytical sample processing tools that facilitate separation and concentration of the target organism for subsequent detection and quantification by methods such as florescence probe-based q-PCR are needed. In this study, the application of a DNA aptamer to bioconcentrate the selected serovar of Salmonella Typhimurium from environmental samples to identify potential environmental reservoirs of Salmonella Typhimurium prior to culture-free detection of Salmonella Typhimurium through MB-based q-PCR was explored for the first time. The assay reported in this study was able to bioconcentrate Salmonella Typhimurium from potential environmental reservoirs exhibiting low concentrations of Salmonella Typhimurium, which was subsequently detected by the Salmonella-specific highly conserved region of the ttr genetargeted MB-based q-PCR (lowest detection limit was 1 cfu/ PCR). It has been observed that preanalytical bioconcentration of Salmonella Typhimurium increased the sensitivity of the MBbased q-PCR assay by selective capture of the target cells from samples spiked with reference strain (lowest detection limit was 1 cfu/PCR in the presence of high background microflora). We have tested our assay by analyzing environmental samples collected from a sewage-impacted perennial river, the Ganges River flowing through northern India, that provides water for drinking and other domestic requirements to a large population
in South Asia and exhibits high endemicity for Salmonella contamination.20,24 In this study, we have observed high levels of Salmonella and Salmonella Typhimurium contamination in potential environmental reservoirs (the green alga Spirogyra, the aquatic macrophytes P. crispus and P. pectinatus, and river bed sediments) in riverine materials from the Ganges River. Earlier, it was observed that Cladophora in Lake Michigan serves as an environmental reservoir for Salmonella and potentially other pathogens, such as shiga toxin-producing E. coli, Shigella, and Campylobacter.25 However, that study accounts for Salmonella contamination and lacks serovar-specific information. Recently, the presence of Enterococcus and enterotoxigenic E. coli has also been reported in submerged macrophytes and sediments collected from stream and river banks.26,27 Furthermore, it has been reported that elevated concentrations of fecal indicator bacteria in aquatic sediments and vegetation have prompted concern that environmental reservoirs of fecal indicator bacteria disrupt the correlation between indicator organisms, pathogens, and human health risks.27 Environmental reservoirs of a pathogenic bacterium could disseminate pathogen across large geographical boundaries and act as vectors of infection.1,26 Therefore, environmental reservoirs of Salmonella Typhimurium in riverine materials from the Ganges River can initiate epidemic and may be critical to the disease’s endemicity. It has been suggested earlier that, in a country where cholera is endemic, ingestion of water from ponds or rivers during plankton blooms provide the requisite infectious dose for clinical cholera.28 Therefore, identification of environmental reservoirs of Salmonella Typhimurium is important for defining the potential role of these reservoirs in the transmission pathways of nontyphoidal salmonellosis leading to human illness and is critical to further reduce disease burden, including management of the disease in the endemic areas. Further, quick detection and quantification of Salmonella Typhimurium in potential environmental reservoirs may limit the morbidity and mortality of nontyphoidal salmonellosis in high-endemicity zones like India in south Asia. Salmonella Typhimurium is the most frequently reported serotype of clinical importance and predominates over other serotypes in environmental studies due to its high survival rate outside the host.5,29 Hence, high concentration of Salmonella Typhimurium in river water and potential environmental reservoirs widespread in the riverine system of the Ganges River is possibly attributed to addition of untreated sewage, hospital wastes of medical college and several other clinical setups, poultry wastes, rain storm waters containing fecal pollutants from nearby unauthorized human settlements, agricultural runoff, and the ability of Salmonella Typhimurium to survive in the environment outside the host.20 Increased detection limits in the presence of high background flora by bioconcentration of Salmonella Typhimurium enhanced the applicability of the developed MB-based q-PCR assay in complex matrices containing PCR inhibitors. The present study for the first time provides quantitative information about Salmonella Typhimurium in the high-endemicity zone of Salmonella in south Asia by use of sensitive culture-free protocols involving preanalytical bioconcentration of Salmonella Typhimurium. India, Indonesia, Bangladesh, and Pakistan have been identified as highrisk sites for infections caused by Salmonella species.24,30,31 In conclusion, it could be inferred from the present study that the inclusion of a preanalytical step, in the form of pathogenspecific DNA aptamer-based bioconcentration of the bacterial cell, enhanced the sensitivity of fluorescent probe-based q-PCR 9000
dx.doi.org/10.1021/es2018994 |Environ. Sci. Technol. 2011, 45, 8996–9002
Environmental Science & Technology assays. Furthermore, information about the contamination of surface water by Salmonella Typhimurium in Ganges riverine materials may pave the way for forecasting and management of nontyphoidal salmonellosis in south Asia.
’ ASSOCIATED CONTENT
bS
Supporting Information. Additional text, four tables, and two figures showing design of ttr gene-specific MB, standard curve generation, conjugation of Salmonella Typhimurium-specific DNA aptamer to Dynabeads, binding of Salmonella Typhimurium cells to Sal aptamer-conjugated Dynabeads, enumeration of Salmonella Typhimurium in water samples spiked with reference strain in the presence or absence of high background of nonspecific DNA, and a map depicting sampling sites. This material is available free of charge via the Internet at http://pubs.acs.org/.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 91+ 0522 2613786/2614118/2627586, ext 237; fax: 91+ 0522-2611547; e-mail: rishishanker1@rediffmail.com or
[email protected]. Author Contributions †
A.J. and P.V. contributed equally to this work.
’ ACKNOWLEDGMENT This work was supported by CSIR Network Project NWP-17. Financial assistance to A.J. (SRF), G.S. (SRF), C.B.P. (SRF), and P.V. (scientist: WOS-A) from CSIR and DST, Government of India, New Delhi, respectively, is acknowledged. The paper bears IITR number 2949. ’ REFERENCES (1) Vezzulli, L.; Pruzzo, C.; Huq, A.; Colwell, R. R. Environmental reservoirs of Vibrio cholerae and their role in cholera. Environ. Microbiol. Rep. 2010, 2, 23–33. (2) Moganedi, K. L. M.; Goyvaerts, E. M. A.; Venter, S. N.; Sibara, M. M. Optimisation of the PCR-invA primers for the detection of Salmonella in drinking and surface waters following a pre-cultivation step. Water SA 2007, 33, 195–202. (3) Ram, S.; Vajpayee, P.; Dwivedi, P. D.; Shanker, R. Culture-free detection and enumeration of STEC in water. Ecotoxicol. Environ. Saf. 2011, 74, 551–557. (4) Voetsch, A. C.; Gilder, T. J. V.; Angulo, F. J.; Farley, M. M.; Shallow, S.; Marcus, R. Food Net estimate of the burden of illness caused by non-typhoidal Salmonella infections in the United States. Clin. Infect. Dis. 2004, 38, 27–34. (5) Martinez-Urtaza, J.; Liebana, E.; Garcia-Migura, L.; Perez-Pi~ neiro, P.; Saco, M. Characterization of Salmonella enterica serovar Typhimurium from marine environments in coastal waters of Galicia (Spain). Appl. Environ. Microbiol. 2004, 70, 4030–4034. (6) Foley, S. L.; White, D. G.; McDermott, P. F.; Walker, R. D.; Rhodes, B.; Fedorka-Cray, P, J.; Simjee, S.; Zhao, S. Comparison of subtyping methods for differentiating Salmonella enterica serovar Typhimurium Isolates obtained from food animal sources. J. Clin. Microbiol. 2006, 44, 3569–3577. (7) Jyoti, A.; Ram, S.; Vajpayee, P.; Singh, G.; Dwivedi, P. D.; Jain, S. K.; Shanker, R. Contamination of surface and potable water in South Asia by Salmonellae: Culture-independent quantification with molecular beacon real-time PCR. Sci. Total Environ. 2010, 408, 1256–1263. (8) Humphrey, T. Salmonella, stress responses and food safety. Nat. Rev. Microbiol. 2004, 2, 504–509.
ARTICLE
(9) Chimalizeni, Y.; et al. The Epidemiology and Management of Non Typhoidal Salmonella Infections. In Hot topics in infection and immunity in children IV; Advances in Experimental Medicine and Biology, Vol. 659; Finn, A., Curtis, N., Pollard, A. J., Eds.; Springer: New York, 2010; pp 3360. (10) Duffy, E. A.; Lucia, L. M.; Kells, J. M.; Castillo, A.; Pillai, S. D.; Acuff, G. R. Concentration of Escherichia coli and genetic diversity and antibiotic resistance profiling of Salmonella isolated from irrigation water, packing shed equipment, and fresh produce in Texas. J. Food Prot. 2005, 68, 70–79. (11) Josefsen, M. H.; Krause, M.; Hansen, F.; Hoorfar, J. Optimization of a 12-h TaqMan PCR based method for detection of Salmonella bacteria in meat. Appl. Environ. Microbiol. 2007, 73, 3040–3048. (12) Malorny, B.; L€ofstr€om, C.; Wagner, M.; Kr€amer, N.; Hoorfar, J. Enumeration of Salmonella bacteria in food and feed samples by realtime PCR for quantitative microbial risk assessment. Appl. Environ. Microbiol. 2008, 74, 1299–1304. (13) Malorny, B.; Bunge, C.; Helmuth, R. A real-time PCR for the detection of Salmonella Enteritidis in poultry meat and consumption eggs. J. Microbiol. Methods 2007, 70, 245–251. (14) Hadjinicolaou, A. V.; Demetriou, V. L.; Emmanuel, M. A.; Kakoyiannis, C. K.; Kostrikis, L. G. Molecular beacon-based real-time PCR detection of primary isolates of Salmonella Typhimurium and Salmonella Enteritidis in environmental and clinical samples. BMC Microbiol. 2009, 9, 97. (15) Hensel, M.; Hinsley, A. P.; Nikolaus, T.; Sawers, G.; Berks, B. C. The genetic basis of tetrathionate respiration in Salmonella typhimurium. Mol. Microbiol. 1999, 32, 275–287. (16) Malorny, B.; Paccassoni, E.; Fach, P.; Bunge, C.; Martin, A.; Helmuth, R. Diagnostic real-time PCR for detection of Salmonella in food. Appl. Environ. Microbiol. 2004, 70, 7046–7052. (17) Vikesland, P. J.; Wigginton, K. R. Nanomaterial enabled biosensors for pathogen monitoring - a review. Environ. Sci. Technol. 2010, 44, 3656–3669. € (18) Mairal, T.; Ozalp, V. C.; Sanchez, P. L.; Mir, M.; Katakis, I.; O’Sullivan, C. K. Aptamers: molecular tools for analytical applications. Anal. Bioanal. Chem. 2008, 390, 989–1007. (19) Joshi, R.; Janagama, H.; Dwivedi, H. P.; Senthil Kumar, T. M. A.; Jaykus, L.-A.; Schefers, J.; Sreevatsan, S. Selection, characterization, and application of DNA aptamers for the capture and detection of Salmonella enterica serovars. Mol. Cell. Probes 2009, 23, 20–28. (20) Ram, S.; Vajpayee, P.; Shanker, R. Prevalence of multi antimicrobial agent resistant, shigatoxin and enterotoxin producing Escherichia coli in surface waters of river Ganga. Environ. Sci. Technol. 2007, 41, 7383–7388. (21) Ram, S.; Vajpayee, P.; Shanker, R. Enterotoxigenic Escherichia coli in sewage-impacted waters and aquatic weeds: quantitative PCR for culture-independent enumeration. J. Appl. Microbiol. 2010, 108, 1007–1014. (22) Schefler,W. C. Statistics for Biological Sciences; Addison-Wesley Publishing Company: Menlo Park, CA, 1969. (23) Gomez, K. A.; Gomez, A. A. Statistical Procedures for Agricultural Research; Wiley: New York, 1984. (24) Ochiai, R. L.; Acosta, C. J.; Danovaro-Holliday, M. C.; Baiqing, D.; Bhattacharya, S. K.; Agtini, M. D.; et al. Domi Typhoid Study Group. A study of typhoid fever in five Asian countries: disease burden and implications for controls. Bull. World Health Org. 2008, 86, 260–268. (25) Byappanahalli, M. N.; Sawdey, R.; Ishii, S.; Shively, D. A.; Ferguson, J. A.; Whitman, R. L.; Sadowsky, M. J. Seasonal stability of Cladophora-associated Salmonella in Lake Michigan watersheds. Water Res. 2009, 43, 806–814. (26) Singh, G.; Vajpayee, P.; Ram, S.; Shanker, R. Environmental reservoirs for enterotoxigenic Escherichia coli in south Asian Gangetic riverine system. Environ. Sci. Technol. 2010, 44, 6475–6480. (27) Badgley, B. D.; Thomas, F. I. M.; Harwood, V. J. Quantifying environmental reservoirs of fecal indicator bacteria associated with sediment and submerged aquatic vegetation. Environ Microbiol. 2011, 13, 932–942. (28) Colwell, R. R. Global climate and infectious disease: the cholera paradigm. Science 1996, 274, 2031–2025. 9001
dx.doi.org/10.1021/es2018994 |Environ. Sci. Technol. 2011, 45, 8996–9002
Environmental Science & Technology
ARTICLE
(29) Setti, I.; Rodriguez-Castro, A.; Pata, M. P.; Cadarso-Suarez, C.; Yacoubi, B.; Bensmael, L.; Moukrim, A.; Martinez-Urtaza, J. Characteristics and Dynamics of Salmonella Contamination along the Coast of Agadir, Morocco. Appl. Environ. Microbiol. 2009, 75, 7700–7709. (30) Bahl, R.; Sinha, A.; Poulos, C.; Whittington, D.; Sazawal, S.; Kumar, R.; et al. Costs of illness due to typhoid fever in an Indian urban slum community: implications for vaccination policy. J. Health Popul. Nutr. 2004, 22, 304–310. (31) Brooks, W. A.; Hossain, A.; Goswami, D.; Nahar, K.; Alam, K.; Ahmed, N. Bacteremic typhoid fever in children in an urban slum, Bangladesh. Emerg. Infect. Dis. 2005, 11, 326–329.
9002
dx.doi.org/10.1021/es2018994 |Environ. Sci. Technol. 2011, 45, 8996–9002
ARTICLE pubs.acs.org/est
Differential Effect of Common Ligands and Molecular Oxygen on Antimicrobial Activity of Silver Nanoparticles versus Silver Ions Zong-Ming Xiu, Jie Ma, and Pedro J. J. Alvarez* Department of Civil & Environmental Engineering, Rice University, Houston, Texas 77005, United States
bS Supporting Information ABSTRACT: The antibacterial activity of silver nanoparticles (AgNPs) is partially due to the release of Ag+, although discerning the contribution of AgNPs vs Ag+ is challenging due to their common co-occurrence. We discerned the toxicity of Ag+ versus a commercially available AgNP (35.4 ( 5.1 nm, coated with amorphous carbon) by conducting antibacterial assays under anaerobic conditions that preclude Ag(0) oxidation, which is a prerequisite for Ag+ release. These AgNPs were 20 less toxic to E. coli than Ag+ (EC50: 2.04 ( 0.07 vs 0.10 ( 0.01 mg/L), and their toxicity increased 2.3-fold after exposure to air for 0.5 h (EC50: 0.87 ( 0.03 mg/L) which promoted Ag+ release. No significant difference in Ag+ toxicity was observed between anaerobic and aerobic conditions, which rules out oxidative stress by ROS as an important antibacterial mechanism for Ag+. The toxicity of Ag+ (2.94 μmol/L) was eliminated by equivalent cysteine or sulfide; the latter exceeded the solubility product equilibrium constant (Ksp), which is conducive to silver precipitation. Equivalent chloride and phosphate concentrations also reduced Ag+ toxicity without exceeding Ksp. Thus, some common ligands can hinder the bioavailability and mitigate the toxicity of Ag+ at relatively low concentrations that do not induce silver precipitation. Furthermore, low concentrations of chloride (0.1 mg/L) mitigated the toxicity of Ag+ but not that of AgNPs, suggesting that previous reports of higher AgNPs toxicity than their equivalent Ag+ concentration might be due to the presence of common ligands that preferentially decrease the bioavailability and toxicity of Ag+. Overall, these results show that the presence of O2 or common ligands can differentially affect the toxicity of AgNPs vs Ag+, and underscore the importance of water chemistry in the mode of action of AgNPs.
’ INTRODUCTION Silver nanoparticles (AgNPs) are widely used in consumer products such as textiles, personal care, and food storage containers for their potent antibacterial capacity.112 AgNP-containing products compose the largest group (55.4%) of all of the nanobased consumer products available on the market as of March, 2011.13 Thus, the incidental or accidental release of AgNPs to the environment represents a potential risk to a wide variety of organisms, including indigenous microbial communities that provide critical ecosystem services (e.g., primary productivity, nutrient cycling, waste degradation, and climate regulation).14 However, the mechanisms by which AgNPs exert toxicity are not fully understood, and the role of water chemistry in the mode of action of AgNPs versus released Ag+ has received limited attention. Zero-valent silver metal is insoluble in water,15 but Ag+ can be released from AgNPs (eq 2) following oxidation of Ag(0) on the nanoparticle surface (eq 1):16 4Agð0Þ þ O2 f 2Ag2 O
ð1Þ
2Ag2 O þ 4Hþ f 4Agþ þ 2H2 O
ð2Þ
The released Ag+ is toxic to bacteria due to various mechanisms r 2011 American Chemical Society
that include binding to thiol groups in proteins and disrupting their function, compromising membrane permeability leading to cell lysis and death,17,18 and oxidative stress due to generation of reactive oxygen species (ROS).9,19,20 However, discerning the contribution of Ag+ vs the AgNPs themselves is challenging due to their common co-occurrence during the exposure period, because most antibacterial assays are conducted under aerobic conditions that promote continuous Ag+ release (Figure 1). Some studies suggest that dissolved Ag+ accounts for most if not all of AgNPs’ toxicity, and AgNPs serve mostly as a source of Ag+ . 6,21 These studies used ligands such as cysteine21 to neutralize Ag+ and isolate the effect of AgNPs, but this approach may confound the reactivity of the AgNPs themselves due to their potential association with the added ligand. Other studies showed that both AgNPs and Ag+ contribute to the antibacterial activity7 and toxicity to eukaryotes,2226 although their apparent relative importance varies considerably. For example, the toxicity of AgNPs to bacteria (E. coli, B. subtilis, and S. oneidensis)26 and to Received: June 5, 2011 Accepted: September 9, 2011 Revised: September 9, 2011 Published: September 27, 2011 9003
dx.doi.org/10.1021/es201918f | Environ. Sci. Technol. 2011, 45, 9003–9008
Environmental Science & Technology
Figure 1. Role of oxygen and common ligands on the antibacterial activity of AgNPs.
the ryegrass Lolium multiflorum23 can be higher than that exerted by an equivalent Ag+ concentration, although the mechanism responsible for higher toxicity was not discerned. The antibacterial activity of AgNPs can depend on particle size,5,2729 shape,10 and surface charge.12 However, most commercially available AgNPs are stabilized by various types of organic coatings, and the effect of these coatings on Ag+ release and antibacterial activity has not been systematically addressed in the literature. In fact, the concentration of Ag+ in most toxicity studies with AgNPs is either missing or not explicitly mentioned.12 Another potential confounding factor is the presence of common ligands in water (e.g., Cl, PO43-, S2-, and SO42-) which can associate with Ag+ and induce its precipitation, thus reducing Ag+ bioavailability and toxicity.17,18,30 Furthermore, Ag+ and some potential ligands are likely to occur in relatively low concentrations that form soluble complexes rather than precipitates. Thus, it is important to consider how such common ligands affect the ecotoxicity of Ag+ (or its disinfection efficacy) when present at concentrations that are below their solubility product equilibrium constant (Ksp). This paper compares the toxicity of Ag+ versus a commercially available AgNP suspension by conducting antibacterial assays under both aerobic and anaerobic conditions. The latter eliminated confounding effects associated with oxidative release of Ag+. The differential effect of common ligands on AgNP vs Ag+ toxicity was also considered to address how water chemistry may affect their relative contribution to antimicrobial activity.
’ MATERIALS AND METHODS Chemicals and AgNPs. AgNPs (coated with amorphous carbon), which have been used in previous studies,5,31 were obtained from Novacentrix Corporation (Austin, TX). These particles had a mean size of 35.4 ( 5.1 nm in the exposure medium (2 mM sodium bicarbonate buffer), determined by dynamic light scattering with a Malvern Zetasizer (ZEN 3600, Malvern Instrument, U.K.), and a ζ potential (ζ) of 27.0 ( 1.54 mV when prepared and stored anaerobically vs 30.0 ( 0.42 mV after 10-day equilibration with air, as measured by the same instrument. The nano powder was suspended in Milli-Q water and homogenized by an ultrasonic cleaner (5510, Branson, CT).
ARTICLE
The metal basis purity of the AgNPs is 99.92%, as measured by ICP-OES (Perkin-Elmer, Waltham, MA). AgNO3 and HNO3 (∼69.0%) was obtained from Sigma-Aldrich (St. Louis, MO); Na2S, NaCl, cysteine, LB (LuriaBertani) broth, NaHCO3 and H2O2 (30%) were all obtained from Fisher Scientific (Fair Lawn, NJ). All chemicals used were reagent grade or better unless otherwise specified. Bacteria. E. coli strain K12 (ATCC 25404) was chosen as a model microorganism for inactivation experiments. A single colony of E. coli grown on LB agar plates was inoculated in 10 mL of LB Broth and grown in a shaking incubator at 37 °C overnight. The bacteria were harvested by centrifugation at 10 000 g for 1 min, washed three times with sodium bicarbonate buffer (2 mM, pH 8.1), and resuspended in the same sodium bicarbonate buffer (10 mL) to make the bacteria stock solution. Sodium bicarbonate can buffer the system at relatively low ionic strength (which promotes nanoparticle coagulation) compared to other bacteria media, and was chosen as the exposure medium to avoid ligands that could bind with Ag+/AgNPs and promote precipitation or other confounding effects. Bacteria (1 mL) were added respectively into test tubes to achieve a viable cell concentration of about 107 cells/mL. Preparation of AgNP Stock Solutions. The commercial AgNPs were washed five times with 1% HNO3 and then another five times with deoxygenated (N2-purged) water inside an anaerobic chamber to remove dissolve Ag+ and oxidized silver from the AgNPs, prior to filtration through a cellulose membrane (molecular weight cutoff 10 000) using a Amicon stir cell (Millipore, MA) pressurized with nitrogen. The filtrate was analyzed for total dissolved silver with an ICP-OES/MS (Perkin-Elmer, Waltham, MA) to obtain the concentrations of dissolved silver (mainly Ag+). The AgNPs were washed and filtered continuously until the Ag+ concentration in filtrate was lower than 300 μg/L. This corresponds to a maximum diluted Ag+ concentration of 3 μg/L in the exposure medium, which below the minimum lethal concentration (MLC, 25.4 μg/L) of Ag+ to E. coli. The MLC is defined here as the minimum concentration of Ag+ in the exposure medium (bicarbonate buffer) that causes statistically significant E. coli mortality (p < 0.05) relative to the control set without silver, and was determined by the same procedure described below for the dose response assays (For details please refer to Supporting Information (SI), Figure S1). For reference, the minimum inhibitory concentration (MIC), which is determined by standard procedures in a different medium (LB broth) 32 was measured at 5.25 mg/L for Ag+. After filtration, the retentate was rinsed with Milli-Q water to resuspend the AgNPs, and was then transferred to 50-mL corning tubes (Corning, Lowell, MA) and sonicated at low intensity for 10 s (Sonic Ruptor 250, Omni International, GA) inside the anaerobic chamber to disaggregate AgNPs. This stock was stored and tested inside the anaerobic chamber to avoid confounding effect caused by oxidative Ag+ release during the doseresponse assays. For the aerobic AgNPs doseresponse test, an aliquot of AgNPs were transferred out of the anaerobic chamber and equilibrated with air for 0.5-h or 10-day prior to inoculation. The concentration of the AgNPs stock was determined by nitric acid/hydrogen peroxide digestion (2 mL HNO3, 1 mL H2O2). AgNPs stock (100 μL) was digested in 20-mL disposable scintillation vials for 24 h in triplicate, followed by filtration through 0.22 μm filter (Fisher Scientific, NJ) to get rid of 9004
dx.doi.org/10.1021/es201918f |Environ. Sci. Technol. 2011, 45, 9003–9008
Environmental Science & Technology
ARTICLE
Figure 2. TEM images of AgNPs stored under (a) anaerobic conditions and (b) aerobic conditions after 10-day exposure to air.
impurities. The resulting filtrate volume (∼3 mL) was brought to 10 mL with 1% nitric acid (100-fold dilution) and its concentration was then determined by ICP-OES (detection limit 80 μg/L) or ICP/MS (detection limit 1 μg/L). The Ag+ concentrations in the stock solution (including treatments exposed to air for 0.5 h and 10 days) were determined by filtering 1 mL stock solution through the cellulose membrane and measuring the total dissolved silver concentration in the filtrate by ICP-MS. DoseResponse of AgNPs and Ag+. The AgNPs stock solution was diluted in 10 mL of sodium bicarbonate buffer (2 mM, pH 8.1) to obtain different concentrations (up to 6.2 mg/L), mixed, and equilibrated for 0.5 h before adding E. coli and incubating in the dark for 6 h at 23 °C under anaerobic (inside the chamber) or aerobic conditions (outside the chamber). E. coli mortality in different treatments was then determined by viable plate counts.33 Briefly, all solutions were serially diluted and eight 10-μL droplets from each dilution were placed on LB agar plates. The plates were incubated at 37 °C for 8 h, and the colony forming units (CFU) were counted. The bacteria mortality was calculated as 1 N/N0 100%, where N and N0 are the remaining and initial concentrations of viable bacteria (CFU/mL), respectively. Doseresponse assays were prepared similarly for Ag+. All tests were conducted in triplicate and repeated at least three times to ensure reproducibility. For doseresponse assays to study the effect of common ligands, equivalent concentrations of each ligand (sulfide, cysteine, chloride, and phosphate) were added separately (0.5 h before addition of bacteria) as sodium salts. Ligands were also tested separately without Ag+ or AgNPs to ensure that they did not inhibit the bacteria. Transmission Electron Microscopy (TEM). To compare the different morphology of AgNPs under aerobic versus anaerobic conditions, TEM samples were prepared inside and outside the anaerobic chamber. For the anaerobic sample, a homogeneous diluted suspension (5 μL) of the filtered AgNPs sample was deposited on a 400-mesh copper grid (Ultrathin carbon type-A, Ted Pella Inc., Redding, CA) and dried inside the chamber. The aerobic sample was prepared similarly except that it was air-dried outside the chamber. Imaging was performed by TEM using a JEOL 2010 (JEOL, Tokyo, Japan) operated at 120 kV. Statistical Analyses. Whether differences between treatments were statistically significant was determined using Student’s t test
at the 95% confidence level. All measurements are reported as mean ( one standard deviation with three replicates.
’ RESULTS AND DISCUSSION TEM Characterization of Aerobically and Anaerobically Prepared AgNPs. TEM analysis showed that anaerobically
prepared AgNPs had a spherical shape with a relatively smooth surface (Figure 2a). Following 0.5-h or 10-day exposure to air, the AgNPs had an irregular shape with a rough surface indicative of surface oxidation (Figure 2b). This likely reflects the instability of Ag0 in the presence of oxygen, which can react to form silver oxide (cubic crystal) on the AgNP surface.16,34 The anaerobically prepared AgNPs tended to adhere to the walls of (hydrophobic) polypropylene tubes (SI Figure S2) while the AgNPs that were equilibrated with air for 10 days did not, indicating that the absence of molecular oxygen could affect the hydrophobicity and adsorption characteristics of AgNPs. This corroborates the importance of hydrophobic interactions in the attachment of AgNPs to hydrophobic surfaces 35 and suggests their potential significance in both the transport of AgNPs and their affinity for bacterial surfaces. Ag+ Toxicity under Aerobic versus Anaerobic Conditions. A comparison of AgNP toxicity under aerobic versus anaerobic conditions requires consideration of potential confounding effects associated with Ag+ release, which includes determining whether Ag+ exerts differential toxicity under aerobic versus anaerobic conditions. Figure 3(a) shows that this is not the case; Ag+ exerted similar toxicity under both aerobic and anaerobic conditions. Specifically, E. coli mortality was 97.7 ( 3.5% following 6-h exposure to 320 ug/L Ag+ under anaerobic conditions, compared to 99.8 ( 8.9% under aerobic conditions, and the corresponding EC50 values were statistically undistinguishable (p > 0.05) (93.4 ( 4.4 v. s 95.6 ( 7.5 μg/L). One of the most commonly proposed mechanisms for the antibacterial effect of Ag+ is the generation of ROS,9,20 which requires the presence of O2 as a precursor.20,36 Since ROS could not be produced during anaerobic exposure, the lack of significant difference in toxicity under aerobic versus anaerobic conditions indicates that ROS-induced oxidative stress is not the dominant antibacterial mechanism for Ag+. Other mechanisms 9005
dx.doi.org/10.1021/es201918f |Environ. Sci. Technol. 2011, 45, 9003–9008
Environmental Science & Technology
Figure 3. Toxicity of (a) Ag+ and (b) AgNPs under aerobic vs anaerobic conditions. The AgNPs were prepared under anaerobic conditions and tested inside the chamber or outside after exposure to air for 0.5-h (releasing up to 76 μg/L Ag+) or 10-day (releasing up to 181 μg/L Ag+).
such as the inactivation of thiol-containing proteins 17,18 appear to be more important. Discerning the toxicity of AgNPs versus Ag+. In order to quantify the relative contributions of released Ag+ versus the AgNPs themselves, the nanoparticles were filtered and their doseresponse patterns were compared under both aerobic and anaerobic conditions (Figure 3). The maximum dose of the anaerobically prepared AgNPs (6.2 mg/L), which resulted in 97.7 ( 3.5% mortality, contained a residual Ag+ concentration (3.0 μg/L) that is lower than the MLC for E. coli (25.4 μg/L). Thus, Ag+ released to the bulk solution did not contribute to the antibacterial activity of AgNPs in this assay. Albeit, on the basis of EC50 values, anaerobically prepared filtered AgNPs (Figure 3b) were about 20 less toxic than Ag+ (Figure 3a) (EC50: 2.04 ( 0.07 vs 0.10 ( 0.01 mg/L). The higher bioavailability and uptake potential of Ag+ compared to AgNPs might explain its higher toxicity. The negatively charged bacteria surface (ζ = 26.9 ( 2.6 mV) should have higher affinity for Ag+ than the negatively charged AgNPs (ζ = 30.0 ( 0.4 for aerobic suspension). AgNPs might be more difficult to be assimilated by bacterial cells due to their relatively large size (35.4 ( 5.1 nm) compared to Ag+ (0.26 nm).37 The cell membrane and cytoplasm contain many sulfur-containing proteins that Ag+ can bind to and inactivate.38 Ag+ may also bind to phosphate-containing molecules such as DNA to harm cells. Although small AgNPs (110 nm) can penetrate the cell membrane of E. coli,5 the uptake of larger nanoparticles by bacteria has been shown to be more difficult.39 If AgNPs get into the cytoplasm, then they could theoretically inactivate proteins and DNA, or be oxidized gradually by intracellular ROS, resulting in the release of Ag+ to eventually increase the intracellular toxicity of AgNPs.23
ARTICLE
Figure 4. Chloride (at equivalent concentrations to the highest silver dose tested) significantly mitigated the toxicity of Ag+ (a), but not AgNPs (b). Exposure to Ag+ was under aerobic conditions with 2.94 μmol/L chloride, whereas exposure to AgNPs was under anaerobic conditions with 57.0 μmol/L chloride.
To assess how the oxidation of AgNPs influences their toxicity, doseresponse assays were also conducted with the same anaerobically filtered AgNPs, except that exposure to bacteria was conducted outside the anaerobic chamber after 0.5-h or 10-day equilibration with air. Antibacterial activity increased with AgNPs exposure to air. On the basis of EC50 values, a 2.3 higher toxicity was observed after 0.5-h exposure to air (EC50: 0.87 ( 0.03 mg/L) and 5.1 higher toxicity after 10-day aeration (EC50: 0.87 ( 0.03 mg/L) relative to anaerobic exposure (EC50: 2.04 ( 0.07 mg/L) (Figure 3b), even though aerobic conditions increased the negative charge of AgNPs (ζ increased from 27.0 ( 1.5 mV when stored anaerobically to 30.0 ( 0.4 mV after 10day exposure to air), which is conducive to higher electrostatic repulsion with the negatively charged E. coli cells (ζ =-26.9 ( 2.6 mV). Apparently, the presence of molecular oxygen promoted the release of Ag+16,34 (the Ag+ concentration for the maximum AgNPs dose of 6.2 mg/L was 76 μg/L for the treatment exposed to air for 0.5-h, and 181 μg/L for the treatment exposed to air for 10-days) and increased AgNP toxicity. This underscores the importance to discern the rate and extent of oxidative Ag+ release in studies of the potential eco-toxicity or disinfection capacity of AgNPs. Common Ligands Mitigate Toxicity of AgNPs and Ag+ Differentially. Common ligands that are prevalent in aquatic systems (e.g., Cl, S2-, cysteine, phosphate) may mitigate the toxicity of Ag+ and AgNPs differentially, thus affecting their relative contribution to antibacterial activity. Figure 4 shows that low equimolar concentrations of chloride (Ag+ and Cl at 2.94 μmol/L (0.1 mg/L)) reduced the toxicity of Ag+ by about 55%, while the toxicity of AgNPs decreased by only 9% in the presence 9006
dx.doi.org/10.1021/es201918f |Environ. Sci. Technol. 2011, 45, 9003–9008
Environmental Science & Technology
ARTICLE
AgNPs to polypropylene tubes, and a comparison of Ag+ and ligands concentration with the corresponding solubility product equilibrium constant (Ksp). This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (713) 348-5903, e-mail:
[email protected].
Figure 5. Equivalent ligand concentrations mitigated the toxicity of Ag+ (2.94 μmol/L) by promoting silver precipitation (with sulfide: 1.47 μmol/L or cysteine: 2.94 μmol/L) or complexation (with chloride: 2.94 μmol/L or phosphate: 0.98 μmol/L) under aerobic conditions.
of higher chloride concentration (AgNPs and Cl at 57.4 μmol/L (2.0 mg/L)). These chloride concentrations are much lower than those commonly found in fresh water.40 The preferential mitigation of Ag+ over AgNPs toxicity by a common ligand may partially explain previous reports of higher AgNPs toxicity than their equivalent Ag+ concentrations.23,26 In theory, AgNPs could become more effective means to deliver Ag+ to cells when the bioavailability of the ions is hindered by complexation and/or precipitation with common ligands. Thus, the lower susceptibility of AgNPs may be due to their ability to more effectively deliver Ag+ to the bacteria after attachment to the cell surface or intracellular uptake.5 Several other ligands (e.g., S2-, cysteine, phosphate) were also added separately to compare their potential to reduce the toxicity of Ag+ (Figure 5). The toxicity of Ag+ was completely removed by the addition of cysteine and sulfide (Ksp-Ag2S = 1 1050.1), which have low solubility products that are conductive to silver precipitation (SI Table S1). Significant toxicity reduction was also observed by chloride (55% at the lethal dose) and phosphate (50%). Unlike sulfide and cysteine, chloride and phosphate have relatively high solubility product equilibrium constants (Ksp-AgCl =1 109.7 and Ksp-Ag3PO4 =1 1017.6) that were not exceeded by our tested concentrations (SI Table S1). Therefore, some natural ligands can mitigate the toxicity of Ag+ by forming aqueous complexes below the precipitation potential. Overall, the tested commercial AgNPs (35.4 ( 5.1 nm) were significantly less toxic to E. coli than Ag+, although the presence of common ligands such as chloride, phosphate, and sulfide could alter their relative contribution to antibacterial activity by complexing with Ag+ and preferentially decreasing its toxicity. Although the exact toxicity mechanism(s) for AgNPs was not elucidated, the similar toxicity of Ag+ under aerobic and anaerobic conditions rules out ROS generation (which requires the presence of O2) as a major toxicity mechanism. The mitigation of Ag+ toxicity by low concentrations of common ligands (even at levels that do not induce silver precipitation) suggest a potentially significant natural mechanism to attenuate toxicity, and underscores the importance to consider water chemistry in efforts to use silver as a disinfectant or elucidate the antibacterial mechanisms of AgNPs.
’ ASSOCIATED CONTENT
bS
Supporting Information. Details on the determination of the minimum lethal concentration (MLC), adherence of
’ ACKNOWLEDGMENT This research was sponsored by a Joint U.S.U.K. Research Program (U.S.-EPA and U.K.-NERC-ESPRC) (EPA -G2008STAR-R1). We thank Gautam Kini (Rice University) for his assistance with the particle size and zeta-potential measurements. ’ REFERENCES (1) Shahverdi, A. R.; Fakhimi, A.; Shahverdi, H. R.; Minaian, S. Synthesis and effect of silver nanoparticles on the antibacterial activity of different antibiotics against Staphylococcus aureus and Escherichia coli. Nanomed.-Nanotechnol. 2007, 3 (2), 168–171. (2) Shrivastava, S.; Bera, T.; Roy, A.; Singh, G.; Ramachandrarao, P.; Dash, D. Characterization of enhanced antibacterial effects of novel silver nanoparticles. Nanotechnology 2007, 18, (22), 225103. (3) Yoon, K. Y.; Byeon, J. H.; Park, J. H.; Hwang, J. Susceptibility constants of Escherichia coli and Bacillus subtilis to silver and copper nanoparticles. Sci. Total Environ. 2007, 373 (23), 572–575. (4) Sondi, I.; Salopek-Sondi, B. Silver nanoparticles as antimicrobial agent: A case study on E-coli as a model for Gram-negative bacteria. J. Colloid Interface Sci. 2004, 275 (1), 177–182. (5) Morones, J. R.; Elechiguerra, J. L.; Camacho, A.; Holt, K.; Kouri, J. B.; Ramirez, J. T.; Yacaman, M. J. The bactericidal effect of silver nanoparticles. Nanotechnology 2005, 16 (10), 2346–2353. (6) Lok, C. N.; Ho, C. M.; Chen, R.; He, Q. Y.; Yu, W. Y.; Sun, H.; Tam, P. K. H.; Chiu, J. F.; Che, C. M. Silver nanoparticles: Partial oxidation and antibacterial activities. J. Biol. Inorg. Chem. 2007, 12 (4), 527–534. (7) Fabrega, J.; Fawcett, S. R.; Renshaw, J. C.; Lead, J. R. Silver nanoparticle impact on bacterial growth: effect of pH, concentration, and organic matter. Environ. Sci. Technol. 2009, 43 (19), 7285–7290. (8) Choi, O.; Deng, K. K.; Kim, N. J.; Ross, L.; Surampalli, R. Y.; Hu, Z. Q. The inhibitory effects of silver nanoparticles, silver ions, and silver chloride colloids on microbial growth. Water Res. 2008, 42 (12), 3066–3074. (9) Choi, O.; Hu, Z. Q. Size dependent and reactive oxygen species related nanosilver toxicity to nitrifying bacteria. Environ. Sci. Technol. 2008, 42 (12), 4583–4588. (10) Pal, S.; Tak, Y. K.; Song, J. M. Does the antibacterial activity of silver nanoparticles depend on the shape of the nanoparticle? A study of the gram-negative bacterium Escherichia coli. Appl. Environ. Microb. 2007, 73 (6), 1712–1720. (11) Kim, J. S.; Kuk, E.; Yu, K. N.; Kim, J. H.; Park, S. J.; Lee, H. J.; Kim, S. H.; Park, Y. K.; Park, Y. H.; Hwang, C. Y.; Kim, Y. K.; Lee, Y. S.; Jeong, D. H.; Cho, M. H. Antimicrobial effects of silver nanoparticles. Nanomed.-Nanotechnol. 2007, 3 (1), 95–101. (12) El Badawy, A. M.; Silva, R. G.; Morris, B.; Scheckel, K. G.; Suidan, M. T.; Tolaymat, T. M. Surface charge-dependent toxicity of silver nanoparticles. Environ. Sci. Technol. 2011, 45 (1), 283–7. (13) Rejeski, D.; Kulken, T.; Pollschuk, P.; Pauwels, E. The project on emerging nanotechnologies. http://www.nanotechproject.org/inventories/ consumer/analysis_draft/, 2011. (14) Munn, C. Marine Microbiology; Oxford: BIOS Scientific, 2003. (15) Petrucci, R. H.; Harwood, W. S.; Herring, G. E.; Madura, J. General Chemistry: Principles and Modern Applications, 9th ed.; Prentice Hall: New York, 1997. 9007
dx.doi.org/10.1021/es201918f |Environ. Sci. Technol. 2011, 45, 9003–9008
Environmental Science & Technology (16) Liu, J. Y.; Hurt, R. H. Ion release kinetics and particle persistence in aqueous nano-silver colloids. Environ. Sci. Technol. 2010, 44 (6), 2169–2175. (17) Ratte, H. T. Bioaccumulation and toxicity of silver compounds: A review. Environ. Toxicol. Chem. 1999, 18 (1), 89–108. (18) Wang, J. M.; Huang, C. P.; Pirestani, D. Interactions of silver with wastewater constituents. Water Res. 2003, 37 (18), 4444–4452. (19) Kim, S.; Choi, J. E.; Choi, J.; Chung, K. H.; Park, K.; Yi, J.; Ryu, D. Y. Oxidative stress-dependent toxicity of silver nanoparticles in human hepatoma cells. Toxicol. in Vitro 2009, 23 (6), 1076–1084. (20) Park, H. J.; Kim, J. Y.; Kim, J.; Lee, J. H.; Hahn, J. S.; Gu, M. B.; Yoon, J. Silver-ion-mediated reactive oxygen species generation affecting bactericidal activity. Water Res. 2009, 43 (4), 1027–1032. (21) Navarro, E.; Piccapietra, F.; Wagner, B.; Marconi, F.; Kaegi, R.; Odzak, N.; Sigg, L.; Behra, R. Toxicity of Silver Nanoparticles to Chlamydomonas reinhardtii. Environ. Sci. Technol. 2008, 42 (23), 8959–8964. (22) Meyer, J. N.; Lord, C. A.; Yang, X. Y. Y.; Turner, E. A.; Badireddy, A. R.; Marinakos, S. M.; Chilkoti, A.; Wiesner, M. R.; Auffan, M. Intracellular uptake and associated toxicity of silver nanoparticles in Caenorhabditis elegans. Aquat. Toxicol. 2010, 100 (2), 140–150. (23) Yin, L. Y.; Cheng, Y. W.; Espinasse, B.; Colman, B. P.; Auffan, M.; Wiesner, M.; Rose, J.; Liu, J.; Bernhardt, E. S. More than the ions: the effects of silver nanoparticles on Lolium multiflorum. Environ. Sci. Technol. 2011, 45 (6), 2360–2367. (24) Kawata, K.; Osawa, M.; Okabe, S. In vitro toxicity of silver nanoparticles at noncytotoxic doses to HepG2 human hepatoma cells. Environ. Sci. Technol. 2009, 43 (15), 6046–6051. (25) Laban, G.; Nies, L. F.; Turco, R. F.; Bickham, J. W.; Sepulveda, M. S. The effects of silver nanoparticles on fathead minnow (Pimephales promelas) embryos. Ecotoxicology 2010, 19 (1), 185–195. (26) Suresh, A. K.; Pelletier, D. A.; Wang, W.; Moon, J. W.; Gu, B. H.; Mortensen, N. P.; Allison, D. P.; Joy, D. C.; Phelps, T. J.; Doktycz, M. J. Silver nanocrystallites: biofabrication using Shewanella oneidensis, and an evaluation of their comparative toxicity on gram-negative and gram-positive bacteria. Environ. Sci. Technol. 2010, 44 (13), 5210–5215. (27) Sotiriou, G. A.; Pratsinis, S. E. Antibacterial activity of nanosilver ions and particles. Environ. Sci. Technol. 2010, 44 (14), 5649–5654. (28) Panacek, A.; Kvitek, L.; Prucek, R.; Kolar, M.; Vecerova, R.; Pizurova, N.; Sharma, V. K.; Nevecna, T.; Zboril, R. Silver colloid nanoparticles: synthesis, characterization, and their antibacterial activity. J. Phys. Chem. B 2006, 110 (33), 16248–16253. (29) Carlson, C.; Hussain, S. M.; Schrand, A. M.; Braydich-Stolle, L. K.; Hess, K. L.; Jones, R. L.; Schlager, J. J. Unique cellular interaction of silver nanoparticles: size-dependent generation of reactive oxygen species. J. Phys. Chem. B 2008, 112 (43), 13608–13619. (30) Choi, O.; Cleuenger, T. E.; Deng, B. L.; Surampalli, R. Y.; Ross, L.; Hu, Z. Q. Role of sulfide and ligand strength in controlling nanosilver toxicity. Water Res. 2009, 43 (7), 1879–1886. (31) Zodrow, K.; Brunet, L.; Mahendra, S.; Li, D.; Zhang, A.; Li, Q. L.; Alvarez, P. J. J. Polysulfone ultrafiltration membranes impregnated with silver nanoparticles show improved biofouling resistance and virus removal. Water Res. 2009, 43 (3), 715–723. (32) Andrews, J. M. Determination of minimum inhibitory concentrations. J. Antimicrob. Chemoth. 2001, 48, 5–16. (33) Adams, L. K.; Lyon, D. Y.; Alvarez, P. J. J. Comparative ecotoxicity of nanoscale TiO2, SiO2, and ZnO water suspensions. Water Res. 2006, 40 (19), 3527–3532. (34) Liu, J. Y.; Sonshine, D. A.; Shervani, S.; Hurt, R. H. Controlled release of biologically active silver from nanosilver surfaces. ACS Nano 2010, 4 (11), 6903–6913. (35) Song, J. E.; Phenrat, T.; Marinakos, S.; Xiao, Y.; Liu, J.; Wiesner, M. R.; Tilton, R. D.; Lowry, G. V. Hydrophobic interactions increase attachment of gum arabic- and PVP-coated Ag nanoparticles to hydrophobic surfaces. Environ. Sci. Technol. 2011, 45 (14), 5988–5995. (36) Lyon, D. Y.; Brunet, L.; Hinkal, G. W.; Wiesner, M. R.; Alvarez, P. J. J. Antibacterial activity of fullerene water suspensions (nC(60)) is not due to ROS-mediated damage. Nano Lett. 2008, 8 (5), 1539–1543.
ARTICLE
(37) Kulinowski, K., Environmental impacts of nanosilver: an ICON backgrounder. In ICON: International Council on Nanotechnology: 2008. (38) Feng, Q. L.; Wu, J.; Chen, G. Q.; Cui, F. Z.; Kim, T. N.; Kim, J. O. A mechanistic study of the antibacterial effect of silver ions on Escherichia coli and Staphylococcus aureus. J. Biomed. Mater. Res. 2000, 52 (4), 662–668. (39) Kloepfer, J. A.; Mielke, R. E.; Nadeau, J. L. Uptake of CdSe and CdSe/ZnS quantum dots into bacteria via purine-dependent mechanisms. Appl. Environ. Microb. 2005, 71 (5), 2548–2557. (40) Goldman, C. R.; Horne, A. J., Limnology; McGraw-Hill Inc.: New York, 1983.
9008
dx.doi.org/10.1021/es201918f |Environ. Sci. Technol. 2011, 45, 9003–9008
ARTICLE pubs.acs.org/est
DNA-Damaging Disinfection Byproducts: Alkylation Mechanism of Mutagenic Mucohalic Acids Rafael Gomez-Bombarelli, Marina Gonzalez-Perez, Jorge Arenas-Valga~non, Isaac Fabian Cespedes-Camacho, Emilio Calle, and Julio Casado* Departamento de Química física, Facultad de Ciencias Químicas Universidad de Salamanca, Plaza de los Caídos, 1-5 E-37008 Salamanca, Spain
bS Supporting Information ABSTRACT: Hydroxyhalofuranones form a group of genotoxic disinfection byproduct (DBP) of increasing interest. Among them, mucohalic acids (3,4-dihalo-5-hydroxyfuran2(5H)-one, MXA) are known mutagens that react with nucleotides, affording etheno, oxaloetheno, and halopropenal derivatives. Mucohalic acids have also found use in organic synthesis due to their high functionalization. In this work, the alkylation kinetics of mucochloric and mucobromic acids with model nucleophiles aniline and NBP has been studied experimentally. Also, the alkylation mechanism of nucleosides by MXA has been studied in silico. The results described allow us to reach the following conclusions: (i) based on the kinetic and computational evidence obtained, a reaction mechanism was proposed, in which MXA react directly with amino groups in nucleotides, preferentially attacking the exocyclic amino groups over the endocyclic aromatic nitrogen atoms; (ii) the suggested mechanism is in agreement with both the product distribution observed experimentally and the mutational pattern of MXA; (iii) the limiting step in the alkylation reaction is addition to the carbonyl group, subsequent steps occurring rapidly; and (iv) mucoxyhalic acids, the hydrolysis products of MXA, play no role in the alkylation reaction by MXA.
’ INTRODUCTION Water disinfectants such as chlorine, ozone, chlorine dioxide, or chloramines can react with naturally occurring organic matter, pollutants, or anions such as bromide and iodide during the production of drinking water.1 Disinfection byproducts (DBPs) formed in these reactions, such as halomethanes or haloacetic acids, are responsible for most of the observed mutagenicity of chlorinated tap water25 and have been linked to the incidence of bladder6 and colorectal7 cancer. A group of DBPs of rising interest and yet to be regulated are the halofuranones, which are formed in the chlorination of organic matter, such as humic substances. Their genotoxic and carcinogenic properties have been studied thoroughly and are wellknown.8,9 Among these halofuranones formed in chlorination are mucochloric (MCA, 3,4-dichloro-5-hydroxyfuran-2(5H)-one), and mucobromic acids (MBA, 3,4-dibromo-5-hydroxyfuran-2(5H)-one), both of which are direct genotoxins and potential carcinogens.1016 Aside from their role as anthropogenic pollutants, mucohalic acids are gaining use in organic synthesis, due to their high functionalization and availability.1725 Both mucohalic acids (MXA) are known to alkylate the DNA bases guanosine, adenosine, and cytidine equally in the form of monomers and forming part of DNA, giving rise to etheno, oxalo etheno, and halopropenal derivatives.2633 The reaction of mucohalic acids with hydroxide ions to form mucoxyhalic acids (MOXA) has been proposed in some works as r 2011 American Chemical Society
the initial step in the reaction of MXA with DNA bases.26,2931 However, our previous work suggests that the rate of formation of MOXA in the cellular conditions is negligible.34 This is in agreement with the alkylation mechanism proposed in more recent works,31,33 shown in Scheme 1. Identical reaction paths are expected for adenosine, guanosine, and cytidine: alkylation of the exocyclic secondary amine followed by elimination, hydrolysis, cyclization, and decarboxylation. Additional minor pathways exist, yielding α-hydroxy chlorohydrins from guanosine31 and also dialkylated products in an excess of adenosine.28 In this work, a kinetic study of the alkylation reactivity of MXA using monofunctional model nucleophiles was undertaken. Since mucohalic acids react with the exocyclic primary amino groups of guanosine, cytidine, and adenosine, aniline (AN) was chosen as a model for those groups. 4-(p-nitrobenzyl)pyridine (NBP) is a known trap for alkylating agents and gives its name to a simple colorimetric assay and was chosen to model the reactivity of endocyclic nitrogen atoms. Its use as a nucleophilicity model for the aromatic nitrogen atoms of nucleotides has allowed us to gain insight into the alkylation mechanisms of both strong3540 and weak electrophiles.4144 Received: June 30, 2011 Accepted: September 12, 2011 Revised: September 6, 2011 Published: September 12, 2011 9009
dx.doi.org/10.1021/es202251b | Environ. Sci. Technol. 2011, 45, 9009–9016
Environmental Science & Technology
ARTICLE
Scheme 1. Proposed Reaction Mechanism for the Alkylation of Nucleotides by MXA
In addition, the alkylation reaction of adenosine by mucohalic acids has been modeled in silico to gain insight into the alkylation mechanism and the order in which the successive reaction steps take place.
’ EXPERIMENTAL SECTION Reactions were monitored by UVvis spectroscopy in a Shimadzu UV2401 PC with a thermoelectric six-cell holder temperature control system ((0.1 °C). The reactions were carried out in an excess of nucleophile, and thus the pseudo-first-order approximation was applied. pH was kept constant by using buffer solutions: 0.020 M CH3COOH/CH3COO for 4 < pH < 6 and 0.020 M H2PO4/ HPO42- for 6 < pH < 8. Electrospray ionization mass spectra were recorded on a Waters ZQ4000 spectrometer by direct injection. Reaction with Aniline (AN). Reactions were carried out in water, and in 7:3 (vol) water/dioxane (w/d) mixtures. The kinetic species was monitored at λ = 320 nm for both MBA and MCA, where aniline shows almost null absorption. The concentration of MXA was in the 750 μM range and that of aniline was in the 120 mM range. Reaction with 4-(p-nitrobenzyl)pyridine (NBP). Because the adduct formed in the alkylation of NBP by MXA shows significant absorption in the visible region of the spectrum, a variation of the NBP test was used in which basification of the reaction media to obtain color was not necessary. Thus, the alkylation reaction was monitored continuously in the reaction medium at λ = 485 nm for MBA and λ = 510 nm for MCA. The MXA concentration was in the 30200 μM and that of NBP in the 520 mM range. Because NBP is insoluble in water, reactions were carried out in 7:3 (vol) w/d mixtures. Computational Calculations. Geometries were optimized at the B3LYP/6-31G(d) level, and were characterized as either minima or transition states by harmonic analysis. Solvent effects were included using the IEF-PCM method with default parameters. Calculations were performed using the Gaussian 03 suite of programs. Safety. Mucohalic acids have tested positive in a number of genotoxicity assays, including the Salmonella assay. They should be handled with care and disposed properly. ’ RESULTS AND DISCUSSION Reaction with Aniline. Figure 1 in Supporting Information (SI) depicts the variation in the UVvis spectrum of the reaction mixture along time during the reaction of MXA with excess aniline. The variation in absorbance (A) with time at the wavelength of maximum absorption of the adduct, λ = 320 nm, (Figure 2, SI),
Figure 1. Typical kinetic profile of the alkylation reaction of NBP by MXA. Dashed line ( ): fit to a two-reaction mechanism; solid line (— — —): fit to a three-reaction mechanism. Symbols: experimental, (O) MCA, (Δ) MBA. [NBP] = 0.015 M, T = 25.0 °C. (a) [MCA] = 2.0 105 M, (b) [MBA] = 2.0 105 M, pH = 7.00, 7:3 water/dioxane (vol).
complies with a pseudo-first-order rate equation (eq 1) A ¼ εAN-MXA l½MXAo ð1 ek1alk t Þ AN
ð1Þ
where εAN-MXAis the molar absorption coefficient of the adduct; l is the optic path, and kAN 1alk is the alkylation pseudoconstant. kAN 1alk is proportional to the concentration of aniline and hence AN kAN 1alk ¼ kalk ½AN
ð2Þ
From these results the following experimental rate equation (eq 3) can be derived: d½ADAN-MXA ¼ kAN ð3Þ alk ½MXA½AN dt The fit to eq 1 affords the kAN 1alk and εAN-MXAvalues (εAN-MCA= (30.3 ( 0.2) 103 M1 cm1; εAN-MBA= (27.3 ( 0.3) 103 M1 cm1 at λ = 320 nm). The reaction rate was observed to be pH-independent in the 5.5 < pH < 8.0 range. Table 1 shows the alkylation rate constants obtained in water and aquo-organic reaction media at different temperatures. The two mucohalic acids react at very similar rates, with the reaction with MBA being slightly faster. The decrease in the dielectric constant when switching to 7:3 (vol) water/dioxane (w/d) medium elicits approximately a 6-fold decrease in the rate constant. The reaction rate for the alkylation of aniline is very fast compared to the rate of formation of mucoxyhalic acids (MOXA) from mucohalic acids.34 The hydrolysis half-life of MXA at 25.0 °C and pH = 8.0 is of the order of years,34 whereas the alkylation half-life in the same conditions ranges from minutes to hours. This confirms that the role of MOXA in the alkylation by MXA is very small or inexistent, and thus the presence of hydroxyl groups in the earlier position of halogen atoms must be due to a reaction with water or hydroxide ions succeeding, and not preceding, the alkylation. The fit of the kAN alk (Table 1) values to the EyringWynneJones equation (eq 4), gives the activation parameters for the alkylation reaction (Table 2). k k Δ‡ S° Δ‡ H ° 1 ð4Þ ¼ ln þ T h R R T The low values for the activation enthalpies are consistent with the high reactivity of the aldehyde group of mucohalic acids, ln
9010
dx.doi.org/10.1021/es202251b |Environ. Sci. Technol. 2011, 45, 9009–9016
Environmental Science & Technology
ARTICLE
Table 1. Rate Constants of Aniline Alkylation by Mucohalic Acids as a Function of Temperature and the Reaction Medium
Scheme 2. Proposed Reaction Mechanism for the Alkylation of Aniline by MXA
1 1 a kAN s ) alk (M
MCA
a
MBA
T (°C)
water
7:3 w/d (vol)
water
7:3 w/d (vol)
25.0
0.162
0.0304
0.242
0.0454
27.5
0.190
0.0329
0.273
0.0490
30.0 32.5
0.210 0.235
0.0368 0.0392
0.302 0.335
0.0529 0.0559
35.0
0.275
0.0430
0.370
0.0610
37.5
0.300
0.0475
0.404
0.0640
Values are reproducible to within 5%.
Table 2. Activation Parameters for kAN alk as a Function of the Reaction Medium MCA activation parameters for kAN alk ‡
1
Δ H° (kJ mol ) Δ‡S° (J K1 mol1) Δ‡G° (kJ mol1) 35 °C
Scheme 3. Proposed Reaction Mechanism for the Alkylation of NBP by MXA
MBA
7:3 w/d
7:3 w/d
water
(vol)
water
(vol)
35 ( 1
25 ( 1
29 ( 1
19 ( 1
141 ( 191 191 ( 2 159 ( 2 207 ( 2 78 ( 2
84 ( 2
78 ( 2
83 ( 2
whereas the negative activation entropies are consistent with the idea of alkylation being an addition reaction. The reaction products (ADAN-MXA) were identified as the 2:1 aniline-MXA adducts by mass spectrometry, with m/z = 301.1 for MCA and m/z = 345.1 for MBA (calcd 301.1 and 345.1). This is consistent with the observation that mucohalic acids react with a stoichiometric amount of aniline to form the Schiff base, whereas in an excess of aniline the diadduct is formed by substitution of the halogen atom α to the carboxylic group.24,45 Minor amounts of a secondary product showing monoalkylation and substitution of the halogen in α by a hydroxyl group were also detected (m/z = 226.1 for MCA and m/z = 270.0 for MBA). This allows us to propose the following reaction mechanism (Scheme 2): mucohalic acids exist as an equilibrium between the closed-chain acetal lactone and the open-chain aldehyde.34,46 The attack of aniline on the aldehyde group is the limiting step, whereas the subsequent elimination to form the Schiff base and further reaction with aniline or water take place rapidly. It is known that no substitution of the halogen atom by aniline is observed prior to formation of the Schiff base24,45 and that substitution of α-halogens by water to form mucoxyhalic acids is very slow.34 This, together with our results, suggests that the reactivity of the halogen α to the carboxyl group in the Schiff base is increased with respect to that in mucohalic acids. This enhanced electrophilicity is also consistent with the formation of products with hydroxyl substituents α to the carboxylate group in adducts of MXA with adenosine, guanosine, and cytidine. Cyclic adducts are the main product in the alkylation of cytidine, adenosine, and guanosine in water, as shown in Scheme 1. They are formed by the attack of an aromatic nitrogen in the nucleobases on the carbon β to the carboxylate group after alkylation of the exocyclic amine group.
The fact that the intramolecular reaction takes place at the β instead of the more reactive α position is possibly due to a combination of various factors: (a) the potential six-ring products obtained by intramolecular attack on the carbon α are disfavored; (b) the endocyclic nitrogen atoms are not nucleophilic enough to react with the sp2 α-carbon, while stronger nucleophilic hydroxide ions do react at that site; and (c) substitution in α by water or hydroxide is relatively fast and takes place before cyclization in α can occur. Because the position β to the carboxylate shows almost no reactivity prior to alkylationMXA undergo substitution by hydroxide only in the α positionit may be concluded that the electrophilicity of this carbon must also be enhanced before cyclization. Substitution of chlorine in α by a hydroxyl group yields an enol, whose corresponding keto form would show a strongly increased reactivity on the β-carbon, since it would acquire a more electrophilic sp3 hybridization and yield a favored five-membered ring. Reaction with NBP. The variation in absorbance along time is shown in Figure 3 of the Supporting Information. The adducts show maximum absorption at λ = 510 nm in the case of MCA and λ = 485 nm for MBA. The kinetic profiles of typical reactions are shown in Figure 1. The adduct is seen to be unstable, decomposing quite rapidly in the reaction medium, which suggests a biexponential mechanism. However, closer scrutiny reveals the presence of an induction period, which imposes the need to take into account an extra reaction. When this third step is considered (Scheme 3), the rate NBP eqs 57 are obtained. If kNBP alk is the alkylation rate constant, k1alk 9011
dx.doi.org/10.1021/es202251b |Environ. Sci. Technol. 2011, 45, 9009–9016
Environmental Science & Technology
ARTICLE
Table 3. Rate Constants of NBP Alkylation by Mucohalic Acids as a Function of Temperature at pH = 7.0
Table 4. Activation Parameters for kNBP alk 7:3 w/d (vol)
1 1 s ) 7:3 w/d (vol) 103kNBP alk (M
activation parameters for
kNBP alk
MCA
MBA
T (°C)
MCA
MBA
32 ( 1
43 ( 1
25.0
8.0 ( 0.2
4.7 ( 0.1
Δ‡S° (J K1 mol1)
177 ( 4
144 ( 3
27.5
9.1 ( 0.2
5.6 ( 0.1
Δ‡G° 35 °C (kJ mol1)
86 ( 2
88 ( 2
30.0 32.5
10.3 ( 0.5 11.6 ( 0.5
6.5 ( 0.2 7.5 ( 0.2
35.0
12.6 ( 0.4
8.6 ( 0.1
37.5
14.0 ( 0.3
10.0 ( 0.2
‡
Δ H° (kJ mol )
is the alkylation pseudoconstant, kconv is the rate constant for the conversion of ADNBP1 into ADNBP2, and kdec is the rate constant for the decomposition of ADNBP2, we have d½MXA NBP ¼ kNBP alk ½NBP½MXA ¼ k1alk ½MXA dt
1
ð5Þ
d½ADNBP 1 ¼ kNBP 1alk ½MXA kconv ½ADNBP 1 dt
ð6Þ
d½ADNBP 2 ¼ kconv ½ADNBP 1 kdec ½ADNBP 2 dt
ð7Þ
Integration of eqs 57 affords eq 8:
NBP kdec t kconv t kNBP þ ðkconv kdec Þek1alk t ðkNBP Þ 1alk kconv ððk1alk kconv Þe 1alk kdec Þe NBP NBP ðk1alk þ kconv Þðk1alk þ kdec Þðkdec kconv Þ NBP
½ADNBP 2 ¼ ½MXAo
The quality of the fit of the experimental data to eq 8 is shown in Figure 1. To confirm the need for the inclusion of the three steps, the fit of the kinetic profile to a simplified biexponential mechanism with only two successive reactions is also shown in Figure 1. It can be seen that the quality of the fit is insufficient, and that there is significant deviation from the curve described by the experimental points. Fitting to eq 8 affords values for the three rate constants and the molar absorption coefficients of the adducts. ADNBP2 shows εNBP-MCA-2= (25.6 ( 0.4) 103 M1 cm1 at λ = 510 nm and εNBP-MBA-2= (30.1 ( 1.4) 103 M1 cm1 at λ = 485 nm. ADNBP1 does not show significant absorption at the wavelengths of measurement. Alkylation Reaction. kNBP 1alk was observed to be linear in the concentration of NBP and pH-independent in the working conditions. The values for the second-order alkylation rate constants (kNBP alk ) are shown in Table 3. The alkylation rate constants are higher for MCA, which is coherent with the electronegativity of chlorine and also with the greater steric repulsion of bromine atoms. The values are between 5- and 10-fold lower than those observed for the alkylation of aniline, which is in keeping with the observation that mucohalic acids react primarily at the exocyclic amino groups of DNA bases rather than at the sp2 nitrogen atoms in the ring. As a matter of fact, no nucleobase adducts of MXA with thymine—which lacks exocyclic nitrogen atoms—have been detected. However, the reactivity of MXA toward endocyclic nitrogen atoms is not completely negligible; a minor alkylation of aromatic nitrogen has also been observed in the reaction of MXA and guanosine,31 for whose endocyclic nitrogen atoms NBP is a nucleophilicity model. Table 4 shows the activation parameters—from the fit to eq 4— for the alkylation of NBP. The activation parameters parallel those obtained with aniline: low activation enthalpies, consistent with the electrophilicity of the aldehyde, and high negative activation entropies, consistent with an addition mechanism.
ð8Þ
The values of the activation free energies are slightly higher than the values obtained for the alkylation of aniline in water/ dioxane medium. This is consistent with the lower reactivity of aromatic nitrogen atoms with mucohalic acids. Comparison of the activation parameters with those obtained for the alkylation of NBP by other alkylating agents such as diketene36 (Δ‡G° = 71 ( 2 kJ mol1), β-propiolactone35 (Δ‡G° = 87 ( 2 kJ mol1), potassium sorbate42 (Δ‡G° = 99 ( 6 kJ mol1), or acrylamide43 (Δ‡G° = 104 ( 2 kJ mol1) in the same reaction conditions suggests that mucohalic acids are strong alkylating agents when attacking endocyclic nitrogen atoms, and even more so when reacting with exocyclic amino groups, as is suggested by the higher rate constants obtained in the reaction with aniline. Elimination Reaction. NBP has been shown to be a reasonable nucleophilicity model of guanosine N-7 and other nitrogen atoms in the cycle of nucleobases. However, its structure is very different from that of nucleotides and hence the subsequent transformations undergone by the MXA-NBP adduct after alkylation are hardly extrapolable to DNA. Therefore, they will only be discussed briefly. The values for kconv are shown in Table 1 of the Supporting Information. kconv was seen to be pH-independent in the 5.5 < pH < 8.0 range. Because water elimination reactions occur through a base-catalyzed pathway, the possibility of general base catalysis by the phosphate buffer cannot be ruled out. It can be seen that the values for MBA are almost 4-fold higher than those of MCA. The UVvis spectrum of ADNBP2 (SI, Figure 3) deserves some attention: it shows absorption in the visible part of the spectrum, without the need for basification; the wavelengths of maximum absorption are significantly blue-shifted when compared to NBP adducts with other alkylating agents and the molar absorption coefficients are high as compared to those of other NBP adducts, which are about 5 103 M1 cm1. This suggests that AD2 is the highly conjugated elimination product shown in Scheme 3. 9012
dx.doi.org/10.1021/es202251b |Environ. Sci. Technol. 2011, 45, 9009–9016
Environmental Science & Technology Computational Modeling of Nucleobase Alkylation. To bring together the kinetic evidence obtained with the model substrates, and the products identified by Kronberg et al. in the reaction of MXA and nucleotides, we modeled the proposed reaction mechanism of nucleobase alkylation by MXA in silico (Scheme 4). As shown in Scheme 4, transition states were computed for the following reactions: the alkylation reaction (TS1); the putative ring closure reactions of AD1, affording both five- or six-membered rings (TS2 and TS3); substitution of the α and β halogen atoms in AD1 by hydroxide (TS4 and TS5); the ring-closure reaction of AD2 in both keto and enol forms (TS6 and TS8); the hydrolysis of AD2 in TS7 and TS9. The energy barriers for the reactions and energy differences between different possible isomers are reported in Table 5. Guanosine adducts can exist both as keto and enol tautomers. The results reported correspond to the lower in energy. The energy barriers for the alkylation reaction (TS1) suggest that this is the rate-limiting step for the complete alkylation pathway. Of the three nucleobases studied, guanosine shows the lowest alkylation barrier, which is quite similar to that of aniline (Table 2), with the calculated activation free energies for adenosine being about 1015 kJ mol1 higher. The initial barrier for the alkylation of cytidine is higher than those of the purine bases (MCA Δ‡H° = 95.6 kJ mol1, Δ‡G° = 99.8 kJ mol1), and
ARTICLE
thus no further steps in the alkylation path were computed for this compound. Table 5. Relative Energies of Transition States and Isomeric Forms (kJ mol1) dA
dG
MCA
MBA
MCA
MBA
Δ‡H° Δ‡G° Δ‡H° Δ‡G° Δ‡H° Δ‡G° Δ‡H° Δ‡G° TS1a
74.4
89.3
71.3
84.5
64.3
78.7
62.8
77.3
TS2 b
158.8
168.3
142.4
149.1
272.5
285.3
269.2
284.1
TS3 b
119.9
130.9
112.8
122.1
193.1
200.9
185.2
191.8
TS4 b
33.4
42.4
30.7
38.5
57.3
68.1
49.8
68.7
TS5 b
95.2
105.4
93.5
106.3
102.4
126.1
100.2
120.7
AD2-en c
0
0
0
0
0
0
0
0
187.6
189.5
182.7
185.7
194.3
192.6
189.7
191.4
172.8 TS9 c AD2-ket c 20.1
177.7 12.5
164.7 14.0
170.4 8.2
182.5 14.2
186.9 11.0
175.5 6.8
179.4 4.5
TS8 c
a
TS6 c
75.1
77.5
67.2
69.7
77.0
83.0
75.5
79.5
TS7 c
29.2
34.8
20.9
29.3
28.5
33.7
26.5
31.5
Referred to the reactants. b Referred to AD1. c Referred to AD2.
Scheme 4. Computed Alkylation Pathway for Nucleobases
9013
dx.doi.org/10.1021/es202251b |Environ. Sci. Technol. 2011, 45, 9009–9016
Environmental Science & Technology Computational calculation of the energy barrier for the elimination step is complex, since the reaction is susceptible to general base catalysis by a number of species that are present in both experimental conditions and in vivo, such as phosphate. Experimental evidence from the reaction of aniline suggests that this elimination reaction is very fast, and thus that it has little influence on the global reaction rate. For these reasons, the elimination step in the reaction pathway was not computed. AD1 can undergo a variety of reactions that cleave the halogencarbon bonds: substitution by hydroxide or by endocyclic nitrogen atoms, which yields cyclic products. Of these reactions, the one corresponding to the hydrolysis of the halogen atom α to the carboxylate group (TS4) is by far the most favored. This is consistent with the final reaction products and the proposed reaction pathway (Scheme 1). It also confirms that hydrolysis precludes cyclization, as was the case with aniline. We also computed the barriers for the noncatalyzed hydrolysis reaction, with water as a nucleophile. The barrier is comparatively higher than that for hydroxide ions (Δ‡H° = 82.1 kJ mol1, Δ‡G° = 89.7 kJ mol1 for adenosine-MCA), but the reaction is still faster than the intramolecular alternatives or the hydrolysis at the β-position. Thus, the alkylation pathway proceeds through TS4, or its neutral counterpart, depending on the reaction conditions. Since the amino group in aniline is a stronger nucleophile than the N-1 atom in adenosine or guanosine, double alkylation is observed instead of hydrolysis in the reaction of aniline and MXA. The enol form of AD2 is slightly more stable than the keto form. This is possibly due to the formation of a hydrogen bond between the enol hydrogen and the carboxylate oxygen, and to the extended conjugation that connects the aromatic system and the carboxylate unit. Reactivity, however, is favored in the keto form. The sp2 carbon in the enol form is less reactive in electrophilic reactions, such as hydrolysis or displacement by cyclic nitrogen atoms, as suggested by the high activation free energies for TS8 and TS9, as compared to TS6 and TS7. The activation free energy for the ring-closure reaction (TS6) is quite low—the activation entropy being close to null, as expected for unimolecular reactions—and hence the reaction is expected to proceed rapidly. The activation free energies for the competing hydrolysis are also very low, which suggests that hydrolysis by hydroxide ions is very fast. However, the concentration of hydroxide ions in neutral medium is very low (107 M). Because the effect of reducing the concentration of hydroxide from the reference standard state of 1 M to 107 M is approximately equivalent to an increase of 40 kJ mol1 in the activation free energy, both ring-closure and base-catalyzed hydrolysis could be expected to occur similar at similar rates. The hydrolysis of AD2 by the more abundant water molecules has a higher energy barrier (adenosine: Δ‡H° = 59.7 kJ mol1, Δ‡G° = 81.3 kJ mol1 for MCA; Δ‡H° = 52.6 kJ mol1, Δ‡G° = 79.8 kJ mol1 for MBA) and is somewhat higher than that observed for the formation of the cycle. Thus, it may be concluded that hydrolysis by both hydroxide and water might pose some competition to intramolecular cyclization. However, the formation of nucleotide-adducts showing substitution of the second halogen atom by a hydroxyl group has not been observed experimentally. Analysis of the reported free energy barriers shows that the nucleophilic attack of the primary amine at the carbonyl carbon is the rate-limiting step of the alkylation reaction, and that the rest
ARTICLE
of the steps in the reaction path take place rapidly. However, since the free energy barriers are very similar for all the steps, especially in the case of guanosine, the indetermination in the calculated energies is compatible with more than one step being rate-limiting, rather than a single specific reaction being a bottleneck. When the experimental alkylation reactions of nucleotides with MXA are carried out at high temperatures (∼90 °C), decarboxylated and even deoxalo products are encountered. However, these are hardly detectable when alkylation reactions are carried out at temperatures closer to biological conditions. Because they lack a keto group to assist the reaction, no transition states were found for the decarboxylation of MXA or AD1, which is consistent with the fact that no decarboxylated products have been observed in the alkylation reaction, prior to the formation of AD2. The decarboxylation reactions of the AO2 (and analogue protonated MOXA) somewhat unfavored with free energies of activation about 150160 kJ mol1 (MOCA: Δ‡G° = 161 kJ mol1; MOBA: Δ‡G° = 156 kJ mol1; adenosine-AD2: Δ‡H° = 161 kJ mol1, Δ‡G° = 156 kJ mol1 for MCA; Δ‡H° = 162 kJ mol1, Δ‡G° = 158 kJ mol1 for MBA). The decarboxylation of AD3 is more favored, consistently with the increased amount of decarboxylated AD3 adduct observed in the experiments (adenosine-AD3: Δ‡H° = 113 kJ mol1, Δ‡G° = 116 kJ mol1). These reactions show high activation enthalpies and positive activation entropies (consistent with a unimolecular cleavage that affords two product molecules) which imply a large increase in the decarboxylation rate at higher temperatures; this is in keeping with the observation of decarboxylated products in these conditions. The mutational pattern of MXA—which consists mainly of GCfAT transitions—has been explained by the formation of guanosine and, possibly to a lesser extent, cytidine-etheno derivatives.14 Our results are in good agreement with these conclusions: guanosine shows the lower theoretical free energy barrier for alkylation and thus it is expected to be the main target of mucohalic acids. The formed adducts have shown no tendency to lead to abasic sites and it is difficult to envision such bulky tricyclic adducts mispairing with large purine residues, be they A or G. Of the two pyrimidine nucleobases, only pairing with thymidine leads to a net mutation, and thus, GCfG*CfG*TfAT (G* being the modified guanosine) would be the expected sequence of base pairs, which is in keeping with the results reported in the literature. The combined experimental and theoretical results presented here allow us to conclude the following: (i) Mucohalic acids (MXA) are strong alkylating agents that react directly with amino groups in nucleotides, as suggested by the high reaction rates and low activation free energies measured in the alkylation of aniline and NBP. (ii) MXA preferentially attack the exocyclic amino groups over the endocyclic aromatic nitrogen atoms. (iii) Mucoxyhalic acids, the hydrolysis products of MXA, play no role in the alkylation reaction by MXA. (iv) The limiting step in the alkylation reaction by mucohalic acids is the addition to the carbonyl group. Subsequent elimination, hydrolysis, and cyclization occur rapidly. (v) Kinetic and computational evidence suggests an alkylation mechanism in agreement with both the experimental product distribution and the mutagenic spectra of mucohalic acids. 9014
dx.doi.org/10.1021/es202251b |Environ. Sci. Technol. 2011, 45, 9009–9016
Environmental Science & Technology
’ ASSOCIATED CONTENT
bS
Supporting Information. Figures showing the variation in the UVvis spectra of the AN-MXA and NBP-MXA alkylation mixtures and also fit of the experimental data to eq 1. This information is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +34 923 294486; fax: +34 923 294574; e-mail: jucali@ usal.es.
’ ACKNOWLEDGMENT We thank the Spanish Ministerio de Ciencia e Innovacion and European Regional Development Fund (Project CTQ201018999) for supporting the research reported in this article. R. G.B. thanks the Ministerio de Ciencia e Innovacion, J.A.V. and M.G.P. thank the Junta de Castilla y Leon and I.F.C.C. also thanks the Spanish Ministerio de Asuntos Exteriores y de Cooperacion for PhD grants. ’ REFERENCES (1) Richardson, S. D.; Plewa, M. J.; Wagner, E. D.; Schoeny, R.; DeMarini, D. M. Occurrence, genotoxicity, and carcinogenicity of regulated and emerging disinfection by-products in drinking water: A review and roadmap for research. Mutat. Res. 2007, 636 (13), 178–242. (2) Kronberg, L.; Holmbom, B.; Reunanen, M.; Tikkanen, L. Identification and quantification of the Ames mutagenic compound 3-chloro-4-(dichloromethyl)-5-hydroxy-2(5H)-furanone and its geometric isomer E-2-chloro-4-(dichloromethyl)-4-oxobutenoic acid in chlorine-trated humic water and drinking water extracts. Environ. Sci. Technol. 1988, 22, 1097–1103. (3) Meier, J. R.; Knohl, R. B.; Coleman, W. E.; Ringhand, H. P.; Munch, J. W.; Kaylor, W. H. Ames mutagenicity and concentration of the strong mutagen 3-chloro-4-(dichloromethyl)-5-hydroxy-2(5H)-furanone and of its geometric isomer E-2-chloro-3-(dichloromethyl)4-oxo-butenoic acid in chlorine-treated tap waters. Mutat. Res. 1988, 206, 177–182. (4) Kronberg, L.; Christman, R. F.; Singh, R.; Ball, L. Identification of oxidized and reduced forms of the strong bacterial mutagen (Z)-2chloro-3-(dichloromethyl)-4-oxobutenoic acid (MX) in extracts of chlorine-treated water. Environ. Sci. Technol. 1991, 25, 99–104. (5) Kanniganti, R.; Johnson, J. D.; Ball, L. M.; Charles, M. J. Identification of Compounds in Mutagenic Extracts of Aqueous Monochloraminated Fulvic-Acid. Environ. Sci. Technol. 1992, 26 (10), 1998–2004. (6) Villanueva, C. M.; Cantor, K. P.; Cordier, S.; Jaakkola, J. J.; King, W. D.; Lynch, C. F.; Porru, S.; Kogevinas, M. Disinfection byproducts and bladder cancer: A pooled analysis. Epidemiology 2004, 15 (3), 357–367. (7) Rahman, M.; Driscoll, T.; Cowie, C.; Armstrong, B. Disinfection by-products in drinking water and colorectal cancer: A meta-analysis. Int. J. Epidemiol. 2010, 39 (3), 733–745. (8) McDonald, T. A.; Komulainen, H. Carcinogenicity of the cholorination disinfection by-product MX. Environ. Sci. Health 2005, 23, 163. (9) IARC. Some Drinking-water Disinfectants and Contaminants, Including Arsenic; Lyon, France, 2004. (10) Fekadu, K.; Parzefall, W.; Kronberg, L.; Franzen, R.; SchulteHermann, R.; Knasmuller, S. Induction of genotoxic effects by chlorohydroxifuranones, byproducts of water disinfection, in E. coli K-12 cells
ARTICLE
recovered from various organs of mice. Environ. Mol. Mutagen 1994, 24, 317–324. (11) Hyttinen, J.; Niittykoski, M.; Jansson, K. Lack of uniformity in the mutational spectra of chlorohydroxyfuranones in Salmonella typhimurium strain TA100. Mutagenesis 1995, 10, 321–323. (12) Jansson, K.; Hyttinen, J.; Niittykoski, M.; M€aki-Paakkanen, J. Mutagenicity in vitro of 3,4-dichloro-5-hydroxy-2(5H)-furanone (mucochloric acid), a chlorine disinfection by-product in drinking water. Environ. Mol. Mutagen 1995, 25, 284–287. (13) LaLonde, R. T.; Ramdayal, F. Mucochloric acid action on Phi X174 DNA: A comparison to other chlorine-substituted 2(5H)-furanones. Chem. Res. Toxicol. 1997, 10 (2), 205–210. (14) Knasmuller, S.; Zohrer, E.; Kronberg, L.; Kundi, M.; Franzen, R.; SchulteHermann, R. Mutational spectra of Salmonella typhimurium revertants induced by chlorohydroxyfuranones, byproducts of chlorine disinfection of drinking water. Chem. Res. Toxicol. 1996, 9 (2), 374–381. (15) Isomaa, B.; Holmstrom, T. H.; Lilius, H.; Franzen, R.; Kronberg, L. The Toxicity of the Mutagen Mx and Its Analog, Mucochloric Acid, to Rainbow-Trout Hepatocytes and Gill Epithelial-Cells and to Daphnia magna. Toxicology 1995, 100 (13), 69–77. (16) Lalonde, R. T.; Leo, H. R. Interactive Chlorine-by-Bromine and Hydrogen-by-Hydroxyl Group Replacement Effects in 2(5h)-Furanone Mutagenicity. Chem. Res. Toxicol. 1994, 7 (6), 779–783. (17) Zhang, J.; Blazecka, P. G.; Belmont, D.; Davidson, J. G. Reinvestigation of mucohalic acids, versatile and useful building blocks for highly functionalized alpha,beta-unsaturated gamma-butyrolactones. Org. Lett. 2002, 4 (25), 4559–4561. (18) Bellina, F.; Anselmi, C.; Martina, F.; Rossi, R. Mucochloric acid: A useful synthon for the selective synthesis of 4-aryl-3chloro-2(5H)furanones, (Z)-4-aryl-5-[1-(aryl)methylidenel-3-chloro-2(5H)furanones and 3,4-diaryl-2(5H)-furanones. Eur. J. Org. Chem. 2003, 12, 2290–2302. (19) Blazecka, P. G.; Belmont, D.; Curran, T.; Pflum, D.; Zhang, J. Further utilization of mucohalic acids: Palladium-free, regioselective etherification and amination of alpha,beta-dihalo gamma-methoxycarbonyloxy and gamma-acetoxy butenolides. Org. Lett. 2003, 5 (26), 5015–5017. (20) Zhang, J.; Blazecka, P. G.; Davidson, J. G. First direct reductive amination of mucochloric acid: A simple and efficient method for preparing highly functionalized alpha,beta-unsaturated gamma-butyrolactams. Org. Lett. 2003, 5 (4), 553–556. (21) Zhang, J.; Das Sarma, K.; Curran, T. T.; Belmont, D. T.; Davidson, J. G. Efficient synthesis of novel gamma-substituted gamma-butenolides by Lewis acid catalyzed addition of metal enolates of active methylene compounds to mucohalic acids. J. Org. Chem. 2005, 70 (15), 5890–5895. (22) Das Sarma, K.; Zhang, J.; Huang, Y.; Davidson, J. G. Amino acid esters and amides for reductive amination of mucochloric acid: Synthesis of novel gamma-lactams, short peptides and antiseizure agent levetiracetam (Keppra (R)). Eur. J. Org. Chem. 2006, 3730–3737. (23) Das Sarma, K.; Zhang, J.; Curran, T. T. Novel synthons from mucochloric acid: The first use of alpha,beta-dichloro-gamma-butenolides and gamma-butyrolactams for direct vinylogous aldol addition. J. Org. Chem. 2007, 72 (9), 3311–3318. (24) Bellina, F.; Rossi, R. Mucochloric and mucobromic acids: Inexpensive, highly functionalised starting materials for the selective synthesis of variously substituted 2(5H)-furanone derivatives, sulfur- or nitrogen-containing heterocycles and stereodefined acyclic unsaturated dihalogenated compounds. Curr. Org. Chem. 2004, 8 (12), 1089–1103. (25) Ube, H.; Shimada, N.; Terada, M. Asymmetric Direct Vinylogous Aldol Reaction of Furanone Derivatives Catalyzed by an Axially Chiral Guanidine Base. Angew. Chem., Int. Ed. 2010, 49 (10), 1858–1861. (26) Kronberg, L.; Sj€oholm, R.; Karlsson, S. Formation of 3,N(4)Ethenocytidine, 1,N(6)-Ethenoadenosine, and 1,N(2)-Ethenoguanosine in Reactions of Mucochloric Acid with Nucleosides. Chem. Res. Toxicol. 1992, 5 (6), 852–855. (27) Kronberg, L.; Karlsson, S.; Sj€oholm, R. Formation of Ethenocarbaldehyde Derivatives of Adenosine and Cytidine in Reactions with Mucochloric Acid. Chem. Res. Toxicol. 1993, 6 (4), 495–499. 9015
dx.doi.org/10.1021/es202251b |Environ. Sci. Technol. 2011, 45, 9009–9016
Environmental Science & Technology
ARTICLE
(28) Asplund, D.; Kronberg, L.; Sj€oholm, R.; Munter, T. Reaction of Mucochloric Acid with Adenosine - Formation of 8-(N-6-Adenosinyl) Ethenoadenosine Derivatives. Chem. Res. Toxicol. 1995, 8 (6), 841–846. (29) Kronberg, L.; Asplund, D.; M€aki, J.; Sj€oholm, R. Reaction of mucochloric and mucobromic acids with adenosine and cytidine: Formation of chloro- and bromopropenal derivatives. Chem. Res. Toxicol. 1996, 9 (8), 1257–1263. (30) M€aki, J.; Sj€oholm, R.; Kronberg, L. Formation of oxalo-substituted etheno derivatives in reactions of mucochloric acid with adenosine, guanosine and cytidine. J. Chem. Soc., Perkin Trans. 1 1999, No. 20, 2923–2928. (31) Klika, K. D.; M€aki, J.; Sj€oholm, R.; Kronberg, L. Regioisomeric alpha-hydroxy chlorohydrins from the reaction of mucochloric acid and guanosine. Arkivoc 2006, 65–74. (32) LeCurieux, F.; Munter, T.; Kronberg, L. Identification of adenine adducts formed in reaction of calf thymus DNA with mutagenic chlorohydroxyfuranones found in drinking water. Chem. Res. Toxicol. 1997, 10 (10), 1180–1185. (33) M€aki, J.; Sj€oholm, R.; Kronberg, L. Bi-imidazole nucleosides obtained by ring opening of etheno and substituted etheno derivatives of adenosine. J. Chem. Soc., Perkin Trans. 1 2000, 24, 4445–4450. (34) Gomez-Bombarelli, R.; Gonzalez-Perez, M.; Calle, E.; Casado, J. Reactivity of mucohalic acids in water. Water Res. 2011, 45 (2), 714–720. (35) Manso, J. A.; Perez-Prior, M. T.; García-Santos, M. P.; Calle, E.; Casado, J. A kinetic approach to the alkylating potential of carcinogenic lactones. Chem. Res. Toxicol. 2005, 18 (7), 1161–1166. (36) Gomez-Bombarelli, R.; Gonzalez-Perez, M.; Perez-Prior, M. T.; Manso, J. A.; Calle, E.; Casado, J. Chemical Reactivity and Biological Activity of Diketene. Chem. Res. Toxicol. 2008, 21 (10), 1964–1969. (37) Fernandez-Rodríguez, E.; Manso, J. A.; Perez-Prior, M. T.; García-Santos, M. D. P.; Calle, E.; Casado, J. The unusual ability of alpha-angelicalactone to form adducts: A kinetic approach. Int. J. Chem. Kinet. 2007, 39 (10), 591–594. (38) Perez-Prior, M. T.; Gomez-Bombarelli, R.; Gonzalez-Perez, M.; Manso, J. A.; Calle, E.; Casado, J. Sorbate-Nitrite Interactions: Acetonitrile Oxide as an Alkylating Agent. Chem. Res. Toxicol. 2009, 22 (7), 1320–1324. (39) Manso, J. A.; Perez-Prior, M. T.; García-Santos, M. P.; Calle, E.; Casado, J. Steric effect in alkylation reactions by N-alkyl-N-nitrosoureas: A kinetic approach. J. Phys. Org. Chem. 2008, 21 (11), 932–938. (40) Manso, J. A.; Perez-Prior, M. T.; Gomez-Bombarelli, R.; Gonzalez-Perez, M.; Cespedes, I. F.; García-Santos, M. P.; Calle, E.; Casado, J. Alkylating potential of N-phenyl-N-nitrosourea. J. Phys. Org. Chem. 2009, 22, 386–389. (41) Perez-Prior, M. T.; Manso, J. A.; García-Santos, M. P.; Calle, E.; Casado, J. Sorbic acid as an alkylating agent. J. Solution Chem. 2008, 37 (4), 459–466. (42) Perez-Prior, M. T.; Manso, J. A.; García-Santos, M. P.; Calle, E.; Casado, J. Alkylating potential of potassium sorbate. J. Agric. Food Chem. 2005, 53 (26), 10244–10247. (43) Cespedes-Camacho, I. F.; Manso, J. A.; Perez Prior, M. T.; Gomez Bombarelli, R.; Gonzalez Perez, M.; Calle, E.; Casado, J. Reactivity of acrylamide as an alkylating agent: a kinetic approach. J. Phys. Org. Chem. 2010, 23, 171–175. (44) Gomez-Bombarelli, R.; Palma, B. B.; Martins, C.; Kranendonk, M.; Rodrigues, A. S.; Calle, E.; Rueff, J.; Casado, J. Alkylating potential of oxetanes. Chem. Res. Toxicol. 2010, 23 (7), 1275–81. (45) Simonis, H. The influence of the primary amines on the mucobromine and mucochlorine acids and their esters. Ber. Dtsch. Chem. Ges. 1901, 34 (1), 509–519. (46) Gomez-Bombarelli, R.; Gonzalez-Perez, M.; Perez-Prior, M. T.; Calle, E.; Casado, J., Genotoxic halofuranones in water: Isomerization and Acidity of Mucohalic Acids. J. Phys. Org. Chem. 2011, DOI 10.1002/ poc.1857.
9016
dx.doi.org/10.1021/es202251b |Environ. Sci. Technol. 2011, 45, 9009–9016
ARTICLE pubs.acs.org/est
Sensitivity of Polar and Temperate Marine Organisms to Oil Components Lisette de Hoop,*,† Aafke M. Schipper,† Rob S. E. W. Leuven,† Mark A. J. Huijbregts,† Gro H. Olsen,‡ Mathijs G. D. Smit,§ and A. Jan Hendriks† †
Radboud University Nijmegen, Institute for Water and Wetland Research, Department of Environmental Science, P.O. Box 9010, NL-6500 GL, Nijmegen, The Netherlands ‡ Akvaplan-niva, High North Research Centre for Climate and the Environment, 9296, Tromsø, Norway § Statoil ASA, N-7005, Trondheim, Norway
bS Supporting Information ABSTRACT: Potential contamination of polar regions due to increasing oil exploitation and transportation poses risks to marine species. Risk assessments for polar marine species or ecosystems are mostly based on toxicity data obtained for temperate species. Yet, it is unclear whether toxicity data of temperate organisms are representative for polar species and ecosystems. The present study compared sensitivities of polar and temperate marine species to crude oil, 2-methylnaphthalene, and naphthalene. Species sensitivity distributions (SSDs) were constructed for polar and temperate species based on acute toxicity data from scientific literature, reports, and databases. Overall, there was a maximum factor of 3 difference in sensitivity to oil and oil components, based on the means of the toxicity data and the hazardous concentrations for 5 and 50% of the species (HC5 and HC50) as derived from the SSDs. Except for chordates and naphthalene, polar and temperate species sensitivities did not differ significantly. The results are interpreted in the light of physiological characteristics, such as metabolism, lipid fraction, lipid composition, antioxidant levels, and resistance to freezing, that have been suggested to influence the susceptibility of marine species to oil. As a consequence, acute toxicity data obtained for temperate organisms may serve to obtain a first indication of risks in polar regions.
’ INTRODUCTION The polar regions are subject to contamination supplied by air and water currents emitted in other areas as well as contaminants arising from local activities like tourism, shipping, and infrastructure support for scientific research.1,2 Local petroleumindustry activities might also introduce contaminants to the polar environment, notably through produced water discharges and accidental spills.3 5 Effluents, produced daily during oil extraction, discharge alkylphenols, metals, organic acids, and oil components, such as benzene, toluene, xylene, and naphthalene, into the marine environment.5,6 In addition, oil spills from shipping and drilling activities occur regularly in marine waters.7 In the future, oil exploitation and transportation in polar regions is likely to increase due to depleting resources elsewhere and increased exploration opportunities due to melting of sea ice.8 11 Eventually, more petroleum-industry activities will potentially contribute to increased contamination of the polar marine ecosystems.9,11 Oil may pose a risk to marine species in polar regions due to its persistence in the environment and its tendency to accumulate in biota.5,12 These risks need to be quantified to support environmental decisions to protect polar ecosystems against impacts of pollution.13 So far, however, regulatory risk procedures and threshold values specific to the r 2011 American Chemical Society
polar region are lacking. Even more, it is unclear whether such specific values are needed.3,14 Polar risk assessments are mostly based on toxicity data obtained for temperate species. Yet, the question rises whether toxicity data of temperate organisms are representative for polar species and ecosystems.13,15,16 Adaptations of polar organisms to the harsh conditions of the polar environment have been suggested to influence their sensitivity to contaminants.13,14,16 Speculations about differences in sensitivity between polar and temperate marine species are often related to possible differences in physiological characteristics, including their metabolism, lipid fraction and composition, antioxidant defense system, and antifreeze capacity.13,14,17 Moreover, characteristics of the polar environment, such as low temperatures and marked seasonality, have been suggested to influence the way contaminants behave.15 To date, little is known about potential differences in sensitivity to toxicants between polar and temperate species.18 Moreover, the few studies available do not allow univocal conclusions. Received: July 12, 2011 Accepted: September 8, 2011 Revised: September 6, 2011 Published: September 08, 2011 9017
dx.doi.org/10.1021/es202296a | Environ. Sci. Technol. 2011, 45, 9017–9023
Environmental Science & Technology
ARTICLE
Table 1. Number of Temperate and Polar Marine Species, Subdivided by Phylum, Used to Derive Species Sensitivity Distributions for Different Toxicants toxicant
region
Annelida
Arthropoda
Chordata
Echinodermata
Mollusca
total
naphthalene
temperate
0
11
4
1
3
19
2-methyl-naphthalene
polar temperate
0 0
5 6
4 3
1 0
0 3
10 12
polar
0
5
3
1
4
13
temperate
0
4
2
1
3
10
polar
0
12
7
0
1
20
temperate
4
13
5
1
5
28
polar
0
13
7
0
2
22
crude oil 3 types crude oil 13 types
For example, the Antarctic sea urchin Sterechinus neumayeri was less sensitive to zinc but more sensitive to copper and cadmium than eleven urchin species from temperate regions.19 Contrastingly, some other polar marine invertebrates, primarily amphipods, were on average equally sensitive to copper and less sensitive to cadmium than numerous temperate marine invertebrates.20 Five polar marine amphipods were on average equally or less sensitive to zinc and lead than four nonpolar marine amphipods.21,22 Until now, a comparison of polar and temperate marine species regarding their sensitivity to toxicants other than metals is lacking. The goal of the current study was to compare the sensitivities of polar and temperate marine species to crude oil and individual oil components. To that end, toxicity data available in scientific literature, reports, and databases were collected and species sensitivity distributions (SSDs; cumulative distribution curves) were constructed for each species group (polar and temperate). Possible differences and similarities in sensitivity to oil and oil components between polar and temperate marine species are discussed in relation to their physiological characteristics.
’ METHODS Data Collection. A literature search provided 14 articles and reports with toxicity experiments on polar marine species14,23 31 and temperate marine species.32 35 Additional toxicity data on polar and temperate species were obtained from the following databases: RIVM e-toxBase, U.S. EPA ECOTOX, and PAN Pesticide Database.36 38 Marine species were considered polar when meeting one or more of the following criteria: (1) mentioned as such in peer-reviewed scientific literature; (2) included in the Arctic Register of Marine Species;39 and (3) a minimum of 75% of the distribution records of the species were located within the Arctic or Antarctic marine region.40 The temperate species group included marine organisms from the temperate and subtropical climate zones. Toxicity data comprised acute LC50, EC50, and TLm (median tolerance limit) end point values, with mortality or reduced survival effects for 50% of the test organisms. Only toxicity data with short-term test durations (1 8 days) and salt water test conditions were included. To retrieve sufficient toxicity values, data collection included test results from organisms of different developmental stages, both static and flow-through experiments, and both nominal and measured concentrations (respectively 25% and 75% of the 85% available data information). LC50 and TLm values for crude oil were derived from experiments with water-soluble fractions (WSFs) of several oil types. The WSF constituents are dispersed particulate oil, dissolved hydrocarbon,
and soluble contaminants such as metallic ions.41 WSF is a relatively stable oil-in-water mixture and is, therefore, very useful to assess the toxicity of crude oils for marine organisms.42 Data Treatment. Toxicity data to derive SSDs were available for crude oil and two oil components: naphthalene and 2-methylnaphthalene (Supporting Information (SI)). Two subsets of the crude oil data were used in the SSD construction. One subset consisted of three types of crude oil with similar compositions, i.e., Alaskan North Slope crude oil, Cook Inlet crude oil, and Prudhoe Bay crude oil.43 The other subset consisted of these three types of crude oil combined with 10 other types, i.e., Artem Island, Kuwait, Neftyany Kamni, Norman Wells, SangachalyMore, Shengli, South Louisiana, and Venezuelan Tia Juana crude oil, and Bunker C and No. 2 fuel oil. Toxicity data were first log10-transformed. In case of multiple toxicity values for a single substance and a single species, the geometric mean was determined prior to the transformation. For each toxicant and species group (polar and temperate), a sample mean (μ) and standard deviation (σ) were calculated based on the log10-transformed toxicity data. The parameters μ and σ were used in the integral of the normal distribution (the cumulative distribution function) to derive SSDs,44,45 i.e., the potentially affected fraction (PAF) of species plotted against the environmental concentration of the toxicant. Toxicity is also reported as the hazardous concentration for 5% and 50% of the species (HC5 and HC50) with a 50% confidence limit.46 Potential differences in sensitivity between the polar and temperate marine species groups were investigated by comparing the means (μ) and variances (σ) of the log10-transformed toxicity data. If log10-transformed toxicity values from both species groups were normally distributed according to the Kolmogorov Smirnov and Shapiro Wilk tests, the means were compared with the Independent t test. Alternatively, the Mann Whitney U test was used. The Levene’s test was used to compare the variances. All tests were executed with SPSS 15.0 for Windows.
’ RESULTS Toxicity values were found for 41 polar species and 49 temperate species. The full data set is given in the SI (Tables S1 and S2). A total of 10 28 species were available to derive SSDs for naphthalene, 2-methyl-naphthalene, and crude oil (Table 1). Species were categorized according to five phyla: Annelida, Arthropoda, Chordata, Echinodermata, and Mollusca. The number of values per taxon differed between the temperate and polar species groups and between the toxicants (Table 1). Most data were available for Arthropoda, Chordata, and Mollusca, 9018
dx.doi.org/10.1021/es202296a |Environ. Sci. Technol. 2011, 45, 9017–9023
Environmental Science & Technology
ARTICLE
Figure 1. Species sensitivity distributions for temperate (dotted curves) and polar (solid curves) marine species for (a) 2-methyl-naphthalene, (b) naphthalene, (c) 3 types of crude oil, and (d) 13 types of crude oil.
respectively 54%, 32%, and 20% for polar species and 49%, 20%, and 20% for temperate species. The SSDs of 2-methyl-naphthalene and crude oil showed little difference between polar and temperate species groups (Figure 1). The HC5 values for 2-methyl-naphthalene and crude oil differed a maximum factor of 1.4 between the two groups, whereas the average toxicities (HC50) were within a factor of 2 (Table 2). Only for naphthalene there was a significant difference in both means (p = 0.002) and variances (p = 0.02) of the SSDs, with a factor of 3 difference between the HC50 values and a factor 1.2 between the HC5 values of the temperate and polar species groups (Table 2). A comparison of only arthropods, chordates, and echinoderms led to a factor of 2 significant difference (p = 0.006) between the polar and temperate naphthalene HC50 values (Table 2). No significant difference in sensitivity was found between polar and temperate arthropods (Mann Whitney U test; p = 0.22). The difference between the four polar and four temperate chordates was significant (p = 0.03). The chordates were all fish (Actinopterygii) with exception of the temperate urochordate Ciona intestinalis. Comparing only fish resulted in a p-value of 0.06.
’ DISCUSSION Uncertainties Due to Differences in Test Species. Overall, the results indicate that median hazardous concentrations of polar and temperate marine species for oil and oil components differed less than a factor of 3. The significant difference between the mean sensitivities of polar and temperate species to naphthalene could not be explained by differences in taxonomic groups included in the comparison, as a comparison based on the same taxonomic groups (arthropods, chordates, and echinoderms) yielded a significant difference as well. The SSDs for naphthalene suggested that chordates were responsible for the difference between temperate and polar species (Figure 1b). Uncertainties Due to Test Characteristics. Generally, a higher water temperature in ecotoxicological experiments leads to a lower effect concentration (e.g., LC50), thus to a higher sensitivity of an organism to a toxicant.22,47,48 Although water temperatures were reported in only 60% of the literature sources, we found no distinct indication of a bias in the remaining toxicity data. For instance, test conditions of the most sensitive species in 9019
dx.doi.org/10.1021/es202296a |Environ. Sci. Technol. 2011, 45, 9017–9023
Environmental Science & Technology
ARTICLE
Table 2. Means (μ) and Standard Deviations (σ) of Polar and Temperate Toxicity Data, Probability on Equality of Means and Variances between Polar and Temperate Marine Species Groups (p-value), HC50 (μg/L), and HC5 (μg/L) Values (50% Confidence Limit) μa naphthalene
σa
HC50 (μg/L)
HC5 (μg/L)
0.53
5.0 10
3
0.67 103
0.37
3.4 10
3
0.85 103
3.20
0.18
1.6 10
3
0.79 103
p-value
0.002
0.02
p-valuec temperate
0.006 3.36
0.10 0.37
2.3 103
0.57 103
1.9 10
3
0.55 103
b
temperate
c
temperate polar b
2-methyl-naphthalene
crude oil 3 types
crude oil 13 types
d
3.70 3.53
polar
3.27
0.32
p-value
0.49
0.54
temperate
3.34
0.36
2.2 103
0.55 103
4.3 10
3
0.40 103
d
polar
3.63
0.62
p-value
0.47
0.08
temperate
3.61
0.54
4.1 103
0.52 103
0.61 0.59
3.8 10
0.37 103
polar p-value
d
3.58 0.51
3
a Of the log10-transformed LC50, EC50, and TLm values. b Including all taxonomic groups. c Without temperate molluscs. d Toxicity values were not normally distributed according to the Kolmogorov Smirnov and Shapiro Wilk tests.
the data set included both high and low water temperatures (see Tables S1 and S2, SI). In addition, the 10 species with multiple toxicity and temperature values for a single substance showed no distinct increase or decrease in effect concentrations with an increasing temperature. For example, LC50 values for naphthalene of the pink salmon Oncorhynchus gorbuscha both increased and decreased within a temperature rise of 8 °C.49 Sensitivity to toxicants may vary with life history stages, of which larvae are generally expected to be the most sensitive to oil.50 Furthermore, static tests may give higher estimates of species sensitivities to oil than flow-through tests.34 These uncertainties should be assessed when more data become available. The type of concentration probably generated little uncertainty because means and standard deviations of measured concentrations did not differ significantly from data including nominal concentrations. Comparison with Other Studies. Crude oil, naphthalene, and 2-methyl-naphthalene are expected to exhibit a nonpolar narcotic toxicity mode of action.18,51 53 Nonpolar narcotics penetrate the lipid bilayer region of membranes and thereby alter lipid properties, such as fluidity, thickness, and surface tension, as well as fatty acid composition. Ultimately, the disturbance of the membrane function leads to death of the organism.54 56 Based on a generic critical body burden of 0.10 mol/kg wet weight, we derived LC50 values for the compounds as described in the Supporting Information (text section 1). All measured HC50 values were in the same order of magnitude with the estimated LC50 values for a narcotic toxicity mode of action of naphthalene and 2-methyl-naphthalene (5.7 103 and 1.4 103 μg/L, respectively). Except for chordates and naphthalene, our comparison of SSDs revealed equal sensitivity to oil for polar and temperate species. This is consistent with a qualitative comparison by McFarlin et al. which showed that three polar species were equally sensitive to one type of crude oil type as seven temperate species.57 A quantitative comparison of metal toxicity revealed also equal sensitivities to copper and even higher sensitivities of temperate species to cadmium, zinc, and lead than polar
species.20 By comparing SSDs for two metals, three pesticides, and a narcotic chemical, Kwok et al. found a higher sensitivity of tropical freshwater species than temperate freshwater species.58 Yet, for other metals the reverse was shown, indicating that there is no consistent latitudinal pattern.58 Physiological Characteristics. Differences in sensitivity to oil between polar and temperate marine species may be influenced by differences in physiological characteristics, including metabolism, lipid fraction, lipid composition, antioxidant levels, and resistance to freezing.13,14,17 Generally, the rate of standard metabolism in aquatic organisms decreases 2- to 3-fold with a 10 °C decrease in water temperature.47,59,60 Whereas this decrease is likely to reduce uptake,13 elimination and growth rates are expected to be decreased as well. In addition, the tendency of polar organisms toward gigantism, which causes a lower surface-area-to-volume ratio, has been associated with a reduced contaminant uptake.13 Reduced uptake because of lower temperatures or larger size may explain why polar organisms have been perceived as more tolerant to short-term toxicant exposure than temperate organisms. Yet, sublethal end points such as growth and reproduction may not be indicative of mortality.13,20,50 In long-term toxicity studies on survival, the reduction in uptake may be compensated by a decrease of elimination, yielding similar sensitivities in both types of species. Unfortunately, chronic toxicity data could be obtained for a few polar species only. Furthermore, polar species generally have longer life spans than species from lower latitudes.13,15,61 So, reduced uptake over short periods due to lower temperatures and larger size might be compensated by a more or less proportional decrease of elimination and an increase of life span. Hence, differences in metabolism provide no a priori reason for expecting polar species to be more or less sensitive than temperate species. Marine polar organisms generally have higher lipid contents than temperate organisms.13,14,61 This has been thought to result in a higher uptake of lipophilic contaminants by polar species, possibly leading to increased sensitivity.13 Yet, the binding of lipophilic toxicants may leave a smaller amount of oil components to interfere with cell membranes.61 The larger amount of storage lipids in the Arctic copepod Calanus glacialis, for example, was used to explain 9020
dx.doi.org/10.1021/es202296a |Environ. Sci. Technol. 2011, 45, 9017–9023
Environmental Science & Technology its lower sensitivity to oil components (PAHs; polycyclic aromatic hydrocarbons) in comparison to the copepod Calanus finmarchicus from a lower latitude.61 Following equilibrium fugacity theory, however, one would expect concentrations in fat tissues to be independent of the fraction of lipids.62 Therefore, a higher lipid content is expected to have little influence on the susceptibility of marine organisms to oil. Differences in lipid composition have also been expected to cause a changed sensitivity of polar marine species to oil as compared to temperate species.17 Lipid composition is thought to influence the distribution of a toxicant in an organism at equilibrium, and thereby the concentration at the target site for nonpolar narcotics, i.e., membranes.17 Elevated levels of polyunsaturated fatty acids (PUFA) in cell membranes, which appear to be a special adaptation of polar fish and invertebrates to low temperatures,63 65 may indirectly contribute to a higher sensitivity of polar species to oil. PUFA are primary targets for reactive oxygen species (ROS). Additional production of ROS stimulated by the biotransformation of oil constituents taken up by organisms may lead to an imbalance and thus oxidative damage.64 Yet, levels of antioxidant enzymes, such as vitamins A, C, and E, were found to be higher in polar species than temperate species.63,66 This may lead to a higher tolerance of polar species to the oxidative effects of oil. With few exceptions,63 this may counteract the negative effects of PUFA in polar cell membranes, possibly leading to equal sensitivities of polar and temperate species. Finally, the production of antifreeze peptides (AFPs) and antifreeze glycopeptides (AFGPs) is another adaptation to the cold environment noted in several polar species.67,68 These proteins prevent fish such as Polar cod (Boreogadus saida) and Atlantic cod (Gadus morhua) from freezing by interacting with the cell membranes to protect these against cold damage.69,70 It is unknown whether and to what extent the effect of antifreeze proteins on membranes may influence the susceptibility of organisms to oil. The low LC50 values noted for B. saida for all toxicants (Table S1) can be understood from its specialized excretion process to prevent the loss of antifreeze proteins.71 Due to the lack of glomeruli in the kidneys, toxicants may be excreted via the bile rather than the urine.23,71 Excretion via the bile may be disadvantageous, possibly due to a longer retention time of toxicants due to intestinal microflora and reabsorption into the duodenum.71 Summarizing, our SSDs showed that the sensitivity of polar and temperate marine species to oil and oil components differed on average less than a factor of 3. In addition, most of the differences were not statistically significant and there was no taxonomic group that was consistently more sensitive than the other groups. Apparently, physiological mechanisms suggested to cause differences between polar and temperate species may have little impact on sensitivity to oil. As a consequence, toxicity data obtained for temperate organisms may serve to obtain a first indication of risks in polar regions. Exceptions due to specific mechanisms can be present, however, as noted for example in B. saida. In addition, toxicity data on polar species are limited in terms of quantity and quality. Basic conditions, such as temperature have often not been reported. Chronic toxicity data are largely absent. So, more empirical confirmation is definitely needed.
’ ASSOCIATED CONTENT
bS
Supporting Information. A description of the comparison between measured HC50 and calculated LC50 concentrations in relation to a narcotic toxicity mode of action in text
ARTICLE
section 1; Polar and temperate marine toxicity data in Tables S1 and S2. This information is available free of charge via the Internet at http://pubs.acs.org/.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +3124-3653281; e-mail:
[email protected].
’ ACKNOWLEDGMENT We thank Dick de Zwart and Leo Posthuma for sharing their knowledge on SSDs. We also acknowledge the researchers who answered our questions and provided data. The research was financially supported by Statoil ASA. ’ REFERENCES (1) Chapman, P. M.; Riddle, M. J. Missing and needed: Polar marine ecotoxicology. Mar. Pollut. Bull. 2003, 46, 927–928. (2) VanderZwaag, D.; Huebert, R.; Ferrara, S. The Arctic environmental protection strategy, Arctic council and multilateral environmental initiatives: Tinkering while the Arctic marine environment totters. Denver J. Int. Law Policy 2001, 30 (1), 131–171. (3) Olsen, G. H.; Jaeger, I.; Camus, L.; Honkanen, J.; Carroll, J. Marine hazard assessment of petroleum activities in Arctic environments. Explor. Prod. Oil Gas Rev. 2008, 6 (2), 52. (4) Olsen, G. H.; Smit, M. G. D.; Carroll, J.; Jæger, I.; Smith, T.; Camus, L. Arctic versus temperate comparison of risk assessment metrics for 2-methyl-naphthalene. Mar. Environ. Res., in press (DOI 10.1016/ j.marenvres.2011.08.003). (5) Risk Assessment of Bioaccumulative Substances. Part I: A Literature Review; C107a/09; IMARES: Den Helder, 2009. (6) Smit, M. G. D.; Frost, T. K.; Johnsen, S. Achievements of riskbased produced water management on the Norwegian continental shelf (2002 2008). Integr. Environ. Assess. Manage. 2011, 1–35. (7) Wright, D. A.; Welbourn, P. Environmental Toxicology; Cambridge University Press: Cambridge, U.K., 2002. (8) Borgerson, S. G. Arctic meltdown. The economic and security implications of global warming. Foreign Aff. 2008, 87 (2), 63–77. (9) Casper, K. N. Oil and gas development in the Arctic: Softening of ice demands hardening of international law. Nat. Res. J. 2009, 49, 825–882. (10) Gautier, D. L.; Bird, K. J.; Charpentier, R. R.; Grantz, A.; Houseknecht, D. W.; Klett, T. R.; Moore, T. E.; Pitman, J. K.; Schenk, C. J.; Scheunemeyer, J. H.; Sørensen, K.; Tennyson, M. E.; Valin, Z. C.; Wandrey, C. J. Assessment of undiscovered oil and gas in the Arctic. Science 2009, 324, 1175–1179. (11) Hjorth, M.; Nielsen, T. G. Oil exposure in a warmer Arctic: Potential impacts on key zooplankton species. Mar. Biol. 2011, 1–9. (12) Joyner, C. C. Protection of the Antarctic environment: Rethinking the problems and prospects. Cornell Int. Law J. 1986, 19, 259–273. (13) Chapman, P. M.; Riddle, M. J. Toxic effects of contaminants in polar marine environments. Environ. Sci. Technol. 2005, 200A–207A. (14) Arctic Ecotoxicological Studies Review; Akvaplan-niva: Tromsø, Norway, 2008. (15) Snape, I.; Riddle, M. J.; Filler, D. M.; Williams, P. J. Contaminants in freezing ground and associated ecosystems: Key issues at the beginning of the new millennium. Polar Rec. 2003, 39 (211), 291–300. (16) Olsen, G. H.; Carroll, M. L.; Renaud, P. E.; Ambrose, W. G., Jr.; Olsson, R.; Carroll, J. Benthic community response to petroleumassociated components in Arctic versus temperate marine sediments. Mar. Biol. 2007, 151, 2167–2176. (17) Van Wezel, A. P.; Sijm, D. T. H. M.; Seinen, W.; Opperhuizen, A. Use of lethal body burdens to indicate species differences in susceptibility to narcotic toxicants. Chemosphere 1995, 31 (5), 3201–3209. 9021
dx.doi.org/10.1021/es202296a |Environ. Sci. Technol. 2011, 45, 9017–9023
Environmental Science & Technology (18) de Laender, F.; Olsen, G. H.; Frost, T.; Grøsvik, B. E.; Grung, M.; Hansen, B. H.; Hendriks, A. J.; Hjorth, M.; Janssen, C. R.; Klok, C.; Nordtug, T.; Smit, M.; Carroll, J.; Camus, L. Ecotoxicological mechanisms and models in an impact analysis tool for oil spills. J. Toxicol. Environ. Health, A 2011, 74, 605–619. (19) King, C. K.; Riddle, M. J. Effects of metal contaminants on the development of the common Antarctic sea urchin Sterechinus neumayeri and comparisons of sensitivity with tropical and temperate echinoids. Mar. Ecol.: Prog. Ser. 2001, 215, 143–154. (20) Chapman, P. M.; McDonald, B. G.; Kickham, P. E.; McKinnon, S. Global geographic differences in marine metals toxicity. Mar. Pollut. Bull. 2006, 52, 1081–1084. (21) Chapman, P. M.; McPherson, C. Comparative zinc and lead toxicity tests with Arctic marine invertebrates and implications for toxicant discharges. Polar Rec. 1993, 29 (168), 45–54. (22) Duquesne, S.; Riddle, M. J.; Schulz, R.; Liess, M. Effects of contaminants in the Antarctic environment - potential of the gammarid amphipod crustacean Paramorea walkeri as a biological indicator for Antarctic ecosystems based on toxicity and bioaccumulation of copper and cadmium. Aquat. Toxicol. 2000, 49, 131–143. (23) Xylene LC50 on Arctic Marine Organisms; 3485; Akvaplan-niva: Trømso, Norway, 2008. (24) Gewurtz, S. B.; Laposa, R.; Gandhi, N.; Christensen, G. N.; Evenset, A.; Gregor, D.; Diamond, M. L. A comparison of contaminant dynamics in arctic and temperate fish: A modelling approach. Chemosphere 2006, 63, 1328–1341. (25) Hansen, B. H.; Altin, D.; Vang, S.-H.; Nordtug, T.; Olsen, A. J. Effects of naphthalene on gene transcription in Calanus finmarchicus (Crustacea:Copepoda). Aquat. Toxicol. 2008, 86, 157–165. (26) Hjorth, M.; Dahll€of, I. A harpacticoid copepod Microsetella spp. from sub-Arctic coastal waters and its sensitivity towards the polyaromatic hydrocarbon pyrene. Polar Biol. 2008, 31, 1437–1443. (27) Holth, T. F.; Beylich, B. A.; Skarphedinsdottir, H.; Liewenborg, B.; Grung, M.; Hylland, K. Genotoxicity of environmentally relevant concentrations of water-soluble oil components in cod (Gadus morhua). Environ. Sci. Technol. 2009, 43 (9), 3329–3334. (28) Jensen, M. H.; Nielsen, T. G.; Dahll€of, I. Effects of pyrene on grazing and reproduction of Calanus finmarchicus and Calanus glacialis from Disko Bay, West Greenland. Aquat. Toxicol. 2008, 87 (2), 99–107. (29) Olsen, G. H.; Sva, E.; Carroll, J.; Camus, L.; de Coenc, W.; Smolders, R.; Overaas, H.; Hylland, K. Alterations in the energy budget of Arctic benthic species exposed to oil-related compounds. Aquat. Toxicol. 2007, 83, 85–92. (30) Rice, S. D.; Moles, D. A.; Taylor, T. L.; Karinen, J. F. Sensitivity of 39 Alaskan marine species to Cook Inlet crude oil and No. 2 fuel oil. In International Oil Spill Conference; Washington DC, 1979. (31) Riebel, P. N.; Percy, J. A. Acute toxicity of petroleum hydrocarbons to the Arctic shallow-water mysid, Mysis oculata (Fabricius). Sarsia 1990, 75 (3), 223–232. (32) Calbet, A.; Saiz, E.; Barata, C. Lethal and sublethal effects of napthalene and 1,2-dimethylnaphthalene on the marine copepod Paracartia grani. Mar. Biol. 2007, 151, 195–204. (33) Jiang, Z.; Huang, Y.; Xu, S.; Liao, Y.; Shou, L.; Liu, J.; Chen, Q.; Zeng, J. Advance in the toxic effects of petroleum water accommodated fraction on marine plankton. Acta Ecol. Sin. 2010, 30, 8–15. (34) Moles, A. Sensitivity of ten aquatic species to long-term crude oil exposure. Bull. Environ. Toxicol. 1998, 61, 102–107. (35) Sabourin, T. D. Respiratory and circulatory responses of the blue crab to naphthalene and the effect of acclimation salinity. Aquat. Toxicol. 1982, 2 (5 6), 301–318. (36) RIVM e-toxBase Website. http://www.e-toxbase.eu/. (37) U.S. EPA ECOTOX Website. http://cfpub.epa.gov/ecotox/. (38) Pesticide Action Network Pesticide Database Website. http:// www.pesticideinfo.org/. (39) Arctic Register of Marine Species Website. http://www.marinespecies.org/arms/. (40) Ocean Biogeographic Information System Website. http:// www.iobis.org/maps/distribution.
ARTICLE
(41) Kauss, P. H.; Hutchinson, T. C. The effects of water soluble oil components on the growth of Chlorella vulgaris Beijerinck. Environ. Pollut. 1975, 9, 157–174. (42) Oil in the Sea: Inputs, Fates and Effects; Steering Committee for the Petroleum in the Marine Environment Update, Board on Ocean Science and Policy, Ocean Sciences Board, Commission on Physical Sciences, Mathematics, and Resources, National Research Council: Washington, DC, 1985; http://www.nap.edu/catalog.php?record_id=314#toc. (43) Birtwell, I. K.; Fink, R.; Brand, D.; Alexander, R.; McAllister, C. D. Survival of pink salmon (Oncorhynchus gorbuscha) fry to adulthood following a 10-day exposure to the aromatic hydrocarbon water-soluble fraction of crude oil and release to the Pacific Ocean. Can. J. Fish. Aquat. Sci. 1999, 56, 2087–2098. (44) Aldenberg, T.; Jaworska, J. S. Uncertainty of the hazardous concentration and fraction affected for normal species sensitivity distributions. Ecotoxicol. Environ. Saf. 2000, 46, 1–18. (45) Posthuma, L.; Suter II, G. W.; Traas, T. P. Species Sensitivity Distributions in Ecotoxicology; CRC Press LLC: Boca Raton, FL, 2002. (46) Smith, E. P.; Cairns, J. J. Extrapolation methods for setting ecological standards for water quality: Statistical and ecological concerns. Ecotoxicology 1993, 2, 203–219. (47) Cairns, J. J.; Heath, A. G.; Parker, B. C. The effects of temperature upon the toxicity of chemicals to aquatic organisms. Hydrobiology 1975, 47 (1), 135–171. (48) Hsiao, S. I. C. Effects of crude oils on the growth of arctic marine phytoplankton. Environ. Pollut. 1978, 17, 93–107. (49) Korn, S.; Moles, D. A.; Rice, S. D. Effects of temperature on the median tolerance limit of pink salmon and shrimp exposed to toluene, naphthalene, and cook inlet crude oil. Bull. Environ. Contam. Toxicol. 1979, 21 (4 5), 521–525. (50) Craddock, D. R. Acute toxic effects of petroleum on Arctic and Subarctic marine organisms. In Effects of Petroleum on Arctic and Subarctic Marine Environments and Organisms; Malins, D. C., Ed.; Academic Press, Inc.: New York, 1977; pp 1 93. (51) Di Toro, D. M.; McGrath, J. A.; Stubblefield, W. A. Predicting the toxicity of neat and weathered crude oil: Toxic potential and the toxicity of saturated mixtures. Environ. Toxicol. Chem. 2007, 26 (1), 24–36. (52) Verhaar, H. J. M.; van Leeuwen, C. J.; Hermens, J. L. M. Classifying environmental pollutants, 1: Structure-activity relationships for prediction of aquatic toxicity. Chemosphere 1992, 25 (4), 471–491. (53) Verbruggen, E. M. J.; Beek, M.; Pijnenburg, J.; Traas, T. P. Ecotoxicological environmental risk limits for total petroleum hydrocarbons on the basis of internal lipid concentrations. Environ. Toxicol. Chem. 2008, 27 (12), 2436–2448. (54) Dimitrov, S. D.; Mekenyan, O. G.; Sinks, G. D.; Schultz, T. W. Global modelling of narcotic chemicals: Ciliate and fish toxicity. J. Mol. Struct.: THEOCHEM 2003, 622 (1 2), 63–70. (55) Hendriks, A. J.; Traas, T. P.; Huijbregts, M. A. J. Critical body residues linked to octanol-water partitioning organism composition, and LC50 QSARs: Meta-analysis and model. Environ. Sci. Technol. 2005, 39, 3226–3236. (56) Van Wezel, A. P.; Opperhuizen, A. Narcosis due to environmental pollutants in aquatic organisms: Residue-based toxicity, mechanisms, and membrane burdens. Crit. Rev. Toxicol. 1995, 25 (3), 255–279. (57) McFarlin, K. M.; Perkins, R. A.; Gardiner, W. W.; Word, J. D.; Word, J. Q. Toxicity of physically and chemically dispersed oil to selected Arctic species. In International Oil Spill Conference; Portland, OR, 2011. (58) Kwok, K. W. H.; Leung, K. W. Y.; Lui, G. S. G.; Chu, V. K. H.; Lam, P. K. S.; Morritt, D.; Maltby, L.; Brock, T. C. M.; Van den Brink, P. J.; Warne, M. S. J.; Crane, M. Comparison of tropical and temperate freshwater animal species’ acute sensitivities to chemicals: Implications for deriving safe extrapolation factors. Integr. Environ. Assess. Manage. 2007, 3 (1), 49–67. (59) Karamushko, L. I. Metabolic adaptation of fish at high latitudes. Dokl. Akad. Nauk SSSR 2001, 379, 359–361. (60) Hendriks, A. J. The power of size: A meta-analysis reveals consistency of allometric regressions. Ecol. Modell. 2007, 205 (1 2), 196–208. 9022
dx.doi.org/10.1021/es202296a |Environ. Sci. Technol. 2011, 45, 9017–9023
Environmental Science & Technology
ARTICLE
(61) Jensen, L. K.; Carroll, J. Experimental studies of reproduction and feeding for two Arctic-dwelling Calanus species exposed to crude oil. Aquat. Biol. 2010, 10, 261–271. (62) Hendriks, A. J.; Van der Linde, A.; Cornelissen, G.; Sijm, D. T. H. M. The power of size. 1. Rate constants and equilibrium ratios for accumulation of organic substances related to octanol-water partition ratio and species weight. Environ. Toxicol. Chem. 2001, 20 (7), 1399–1420. (63) Abele, D.; Puntarulo, S. Formation of reactive species and induction of antioxidant defence systems in polar and temperate marine invertebrates and fish. Comp. Biochem. Physiol., Part A: Mol. Integr. Physiol. 2004, 138, 405–415. (64) Hannam, M. L.; Bamber, S. D.; Moody, A. J.; Galloway, T. S.; Jones, M. B. Immunotoxicity and oxidative stress in the Arctic scallop Chlamys islandica: Effects of acute oil exposure. Ecotoxicol. Environ. Saf. 2010, 73, 1440–1448. (65) Gillis, T. E.; Ballantyne, J. S. Mitochondrial membrane composition of two Arctic marine bivalve mollusks Serripes groenlandicus and Mya truncata. Lipids 1999, 34 (1), 53–57. (66) Camus, L.; Gulliksen, B.; Depledge, M. H.; Jones, M. B. Polar bivalves are characterized by high antioxidant defences. Polar Res. 2005, 24 (1 2), 111–118. (67) De Vries, A. L. The role of antifreeze glycopeptides and peptides in the freezing avoidance of Antarctic species. Comp. Biochem. Physiol. 1988, 90B (3), 611–621. (68) De Vries, A. L.; Cheng, C.-H. C. Antifreeze proteins and organismal freezing avoidance in polar fishes. In The Physiology of Polar Fishes; Farrell, A. P., Steffensen, J. F., Eds.; Elsevier Academic Press: San Diego, CA, 2005; pp 155 202. (69) Fletcher, G. L.; Goddard, S. V.; Davies, P. L.; Gong, Z.; Ewart, K. V.; Hew, C. L. New insights into fish antifreeze proteins: Physiological significance and molecular regulation. In Cold Ocean Physiology; P€ortner, H. O., Playle, R. C., Eds.; Cambridge University Press: Cambridge, 1998; pp 239 265. (70) Rubinsky, B.; Arav, A.; Fletcher, G. L. Hypothermic protection a fundamental property of 'antifreeze' proteins. Biochem. Biophys. Res. Commun. 1991, 180, 566–571. (71) Christiansen, J. S.; Dalmo, R. A.; Ingebrigtsen, K. Xenobiotic excretion in fish with aglomerular kidneys. Mar. Ecol.: Prog. Ser. 1996, 136, 303–304.
9023
dx.doi.org/10.1021/es202296a |Environ. Sci. Technol. 2011, 45, 9017–9023
ARTICLE pubs.acs.org/est
Combining Monitoring Data and Modeling Identifies PAHs as Emerging Contaminants in the Arctic F. De Laender,†,* J. Hammer,‡ A.J. Hendriks,§ K. Soetaert,|| and C.R. Janssen† †
Laboratory of Environmental Toxicology and Applied Ecology, Ghent University, Plateaustraat 22, 9000 Ghent, Belgium Aquatic Ecology and Ecotoxicology Department, University of Amsterdam, PO box 94248, 1090 GE Amsterdam, The Netherlands § Department of Environmental Science, Institute for Water and Wetland Research, Radboud University, Heijendaalseweg 135, 6525 AJ Nijmegen, The Netherlands Centre for Estuarine and Marine Ecology, Netherlands Institute of Ecology (NIOO-KNAW), Korringaweg 7, 4401 NT Yerseke, The Netherlands
)
‡
bS Supporting Information ABSTRACT: Protecting Arctic ecosystems against potential adverse effects from anthropogenic activities is recognized as a top priority. In particular, understanding the accumulation and effects of persistent organic pollutants (POPs) in these otherwise pristine ecosystems remains a scientific challenge. Here, we combine more than 20 000 tissue concentrations, a food web bioaccumulation model, and time trend analyses to demonstrate that the concentrations of legacy-POPs in the Barents/ Norwegian Sea fauna decreased 10-fold between 1985 and 2010, which reflects regulatory efforts to restrict these substances. In contrast, concentrations of fossil fuel derived PAHs in lower trophic levels (invertebrates and fish) increased 10 to 30 fold over the past 25 years and now dominate the summed POP burden (25 POPs, including 11 PAHs) in these biota. Before 2000, PCBs dominated the summed POP burden in top predators. Our findings indicate that the debate on the environmental impacts of fossil fuel burning should move beyond the expected seawater temperature increase and examine the possible environmental impact of fossil fuel derived PAHs.
’ INTRODUCTION Protecting the health of Arctic ecosystems is a global concern,1 as direct 2 and indirect changes 3 due to global warming are expected to be most pronounced in these regions. An often overlooked potential source of stress for these ecosystems is the exposure to persistent organic pollutants (POPs), i.e., chemical compounds that are resistant to environmental biodegradation, bioaccumulate in lipid tissues and negatively affect human and environmental health.4,5 In recognition of these issues, legislative actions in the 1970s and 1980s banned the production and/or use of the most toxic and bioaccumulative POPs such as PCBs and DDTs.6,7 Despite these efforts, multiple POPs are still present in many ecosystems today, including remote regions such as the Arctic where they have been found in biota 8 as well as in humans.9 Northward air and sea currents have been identified as transport vectors of (semi-) volatile POPs from temperate regions to the colder Arctic environment where they condense.10 Field studies relating measurements of POPs in biota to observed adverse effects highlight that environmental concentrations of these pollutants have the potential to negatively affect humans 11 as well as the local fauna.12,13 The Arctic Monitoring and Assessment Programme (AMAP) has been documenting tissue concentrations of POPs in Arctic biota over the past 25 years; the data produced by these monitoring r 2011 American Chemical Society
campaigns are freely accessible (http://ecosystemdata.ices.dk/). However, a typical feature of these data is that they are fragmentary, i.e., not all POPs have been measured in all species. This limitation in data availability is problematic if the status of Arctic ecosystems as a whole is to be assessed as information on certain species may be lacking. Bioaccumulation models have demonstrated their capacity to predict tissue concentrations based on water concentrations and easily inferable characteristics of the chemical and the species 1416 and therefore at first seem suitable to fill in crucial data gaps. However, measured water concentrations of POPs —the input of bioaccumulation models—are scarce and are only available for a few points in time which makes it difficult to study temporal trends in Arctic ecosystem health.17 Also, the chemical biotransformation rate in other species than fish remains a poorly known parameter in these models,1820 thus increasing the uncertainty on internal concentrations predicted by bioaccumulation models for these species. In this paper, we combine monitored tissue concentrations, a bioaccumulation model and time trend analyses to reconstruct concentrations of the 4 legacy POP classes (PCBs, DDTs, Received: July 14, 2011 Accepted: September 2, 2011 Revised: August 29, 2011 Published: September 02, 2011 9024
dx.doi.org/10.1021/es202423f | Environ. Sci. Technol. 2011, 45, 9024–9029
Environmental Science & Technology
ARTICLE
HCHs, and HCB) and of PAHs in 21 species occupying various trophic levels (from mussels to Polar bears) in the Barents/ Norwegian Sea food web between 1985 and 2010. The strategy we introduce uses a bioaccumulation model to infer seawater concentrations from concentrations in fish, i.e., species for which biotransformation rates are relatively well-known and for which decadal-scale tissue concentration data are available. Next, statistical modeling is performed to quantify temporal trends in these inferred seawater concentrations. These trends are then used as an input for a bioaccumulation model to estimate concentrations of all 5 POP classes in all 21 species between 1985 and 2010. In addition, we investigate the capacity of the model to accurately predict an independent set of monitoring data under various assumptions on the biotransformation rates in other species than fish.
’ EXPERIMENTAL SECTION Measured Tissue Concentration and Sediment Data. Concentrations of 92 POPs from 5 classes (PCBs, PAHs, HCHs, HCB, and DDTs, see Table S1, Supporting Information, SI, for a list) measured in various tissues (Table S2 of the SI) from 21 different species (Table S3 of the SI) were downloaded from the database of the International Council for the Exploration of the Sea (ICES: http://ecosystemdata.ices.dk/). Only data from the Barents and Norwegian Sea were considered. The tissue concentrations were standardized to lipid content using appropriate lipid contents of the tissue matrix (Table S2 of the SI). Bioaccumulation Modeling. A bioaccumulation model 15 was used to estimate bioaccumulation factors (BAFs) for all 21 species for which tissue concentrations were available. The model estimates a BAF (L kg1 wet weight) of a species at a given trophic level (BAFTL), i.e., the ratio of the tissue concentration (Ctissue μg kg1 wet weight) and the water concentration (Cwater μg L1), using the pollutant’s octanolwater partition coefficient (Kow), the species’ weight (kg) and lipid content and the bioaccumulation factor of a representative prey species (BAFTL-1):
BAFTL ¼ ¼
Ctissue Cwater k0, x, in þ ðk1, x, in 3 BAFTL-1 Þ k0, x, out þ k1, x, out þ k2, x, out þ k3, x, out
ð1Þ
with k0,x,in the pollutant absorption rate from water (L kg1 d1), k1,x,in the pollutant assimilation rate from ingested food (d1), k0,x,out the pollutant excretion rate to the water (d1), k1,x,out the pollutant loss rate via faeces (d1), k2,x,out the dilution rate of the pollutant in the body caused by growth or reproduction (d1) and k3,x,out the rate of biotransformation (d1). Equations of all rate constants (eqs S1S5 of the SI) and the numerical solution procedure are described in the SI. Inferring Water Concentrations from Tissue and Sediment Concentrations. We used BAFs for fish, as predicted by the bioaccumulation model, to infer seawater concentrations Cwater from tissue concentrations Ctissue in fish. Fish were chosen because biotransformation rates k3,x,out, and therefore the modeled BAFs that rely on it, are most certain for fish. For PCBs, DDTs, HCHs, and HCB, 22284 tissue concentrations were available for fish. For PAHs, insufficient concentrations in fish tissue were available, forcing us to use tissue concentrations in mussels (2877 data points) and our modeled BAFs for mussels to
obtain Cwater. The water concentrations we inferred from (fish and mussel) tissue concentrations were compared to seawater concentrations that we inferred from an independent set of sediment concentrations using the following formula: Cwater ¼ Csediment 3 ðKoc 3 OCÞ1
ð2Þ
with Koc the organic carbonwater partitioning coefficient (eq S6 of the SI) and OC the organic carbon content of the sediment, which is about 2%.21 Concentrations of POPs in sediments Csediment (μg kg wet sediment1) from the Norwegian and Barents Sea region were also downloaded from the ICES Web site. We only considered data from the top layer (<10 mm), containing sediment deposited about 10 years before the listed sampling date, based on Barents Sea sedimentation rates 22 of 1 mm y1. Reported concentrations in the sediment (μg kg dry sediment1) were converted to wet weight basis (μg kg wet sediment1) assuming a 90% water content of sediment.21 Time Trend Analysis of Mussel and Fish-Inferred Water Concentrations. For 67 pollutants, either the inferred seawater concentrations were available for less than 10 different years, the measurements spanned a period shorter than 15 years, or the Kow was higher than the maximum Kow for which the bioaccumulation model was calibrated, i.e., 108.5.15 For these pollutants, no time trend analysis was pursued. For the other 25 pollutants, additive models (ams) were fitted to the log-transformed water concentrations (10 base logarithm) as follows: E½log10ðCwater Þ ¼ intercept þ f ðtimeÞ þ ε
ð3Þ
where E[log10(Cwater)] is the expected value of log10(Cwater), which was assumed to be normally distributed, as was the error ε, with mean 0 and σ as a standard deviation, i.e., ε ∼ N(0,σ). The smoothing function f is a non parametric relationship between the predictor (time) and the effect this predictor has on the response variable (log10(Cwater)). Additive models replace multivariate regression by so-called “additive approximation”, which has a number of statistical advantages over the former, as discussed in detail elsewhere.23 All dates in the database were converted to real numbers between 1985 and 2010. By fitting the model (eq 3) to the data (log10(Cwater) and time) estimates of f were obtained, together with the corresponding p-value for these estimates. A cutoff p-value of 0.05 was chosen. Additive models were chosen because they make no assumptions about the shape of the resulting time trend f. Absence of autocorrelation was confirmed (p was always >0.05) by likelihood ratio testing of the ams against corresponding additive mixed models (amms) that extended the residual term of eq 3 with an autocorrelation structure Δ, making ε ∼ N(0,Δσ). Because sampling intervals were irregular, a corCAR1 structure was chosen.23 Model fits were performed using the R-functions gam and gamm from the mgcv package.23 The resulting time trend models for water concentrations of POPs were checked for normality of the residuals (Figure S1 of the SI). Tissue Concentrations of All POPs in All Species and Validation. The time trends for water concentrations of the 25 POPs and the associated residuals were multiplied by the modeled bioaccumulation factors (BAF) of the 21 species that were representative for the aquatic Arctic food web. In this way, tissue concentrations were obtained for all 25 POPs in 21 species of the Arctic food web. There were 6866 observed tissue concentrations in the database that were not used to infer seawater concentrations, i.e., concentrations of PCBs, DDTs, HCHs, and 9025
dx.doi.org/10.1021/es202423f |Environ. Sci. Technol. 2011, 45, 9024–9029
Environmental Science & Technology HCB in other species than fish, and PAH concentrations in other species than mussels. These were used to validate the bioaccumulation model predictions. Predicted tissue concentrations were compared to the corresponding observations and the sum of squared errors (SSE) was calculated. Biotransformation (BT) Rates in Other Species than Fish. Measured BT rates of all 25 POPs in all 21 species are unavailable. Our default assumption on BT rates was therefore that all species had BT rates equal to BT rates for fish estimated from.19 However, we repeated our analysis (from the BAF modeling to the tissue concentration modeling and validation against 6866 observations) in 5 additional scenarios, each time making an alternative assumption on the BT rates, including a scenario in which absence of BT in all species was assumed (scenario 1). Other scenarios were as follows: 2: BT only in fish; 3: BT only in fish and benthos at rates for fish; 4: BT only in fish and mammals at rates for fish; 5: BT in all species at rates for fish (the default assumption). In a sixth scenario, we assumed that BT rates in mammals were faster than in fish, acknowledging that the BT capacity of mammalian cells has been found to be higher than in cells from in poikilotherms.24,25 In absence of compound- and species-specific BT rates, and following earlier modeling exercises,26 we assumed BT rates in mammals to be 5 times those for fish in this sixth scenario. All computational details are provided in the SI, as are the calculated BT rates for fish (Figure S2A&B, SI).
’ RESULTS AND DISCUSSION Inferred Seawater Concentrations. By combining tissue concentrations of PCBs, DDTs, HCHs, and HCB in fish and of PAHs in mussels with a bioaccumulation model, we inferred time trends of dissolved seawater concentrations for these 5 pollutant classes in the Arctic region between 1985 and 2010. The seawater concentrations we inferred corresponded well with the seawater concentrations we independently inferred from sediment concentrations using chemical equilibrium modeling (Figure 1). Note that the uncertainty of the dating of these sediment-inferred concentrations is high because POP concentrations were only available for the whole top layer of the sediment (10 mm), containing sediment deposited 0 to 10 years before the listed sampling date.22 The time trend analyses showed that the most inferred water concentrations significantly (p < 0.05) decreased following bans on production and/or use and our results comply with the decreasing trends found in seabirds and marine mammals from the Canadian Arctic.27 This is shown in Figure 1 for selected POPs; results for other POPs are shown in Figure S3 of the SI. Decreases were apparent for most PCBs (Figure 1a), for the two analyzed DDT breakdown products (Figure 1b, Figure S3i of the SI) and HCHs (Figure 1d,e) and for HCB (Figure 1f). For these POPs, water concentrations in 1985 were 10 times higher than current-day concentrations. For the water concentrations of the PCB congeners with higher molecular weights (e.g., PCB 156, Figure S3g of the SI), a decrease was only apparent after 1995. A possible cause for this delay between the legislative action and the observed decrease might be the slower movement toward the poles and the higher environmental persistence of heavier congeners than of the lighter congeners.28 In contrast, HCHs and HCB are transported relatively quickly to higher latitudes due to their volatility and are also less persistent than heavy congener PCBs,29 which explains why the time trend of these POPs follow legislative actions more closely (Figure 1df).
ARTICLE
Figure 1. Water concentrations as inferred from measured concentrations in fish and Blue mussel, the estimated time trend (solid line) and its point-wise 95% confidence interval (dotted lines, calculated as +1.96 • se). Triangles represent water concentrations inferred from measured sediment concentrations. Horizontal lines indicate the time window over which these values average (minmax); vertical lines are the mean +1 standard deviation. Dashed vertical lines indicate the start of pollutant regulation. DDT was banned before 1985.
For gamma HCH, a clear decrease was only noted after 2000 (Figure 1e), i.e., following the ban by the U.S. in 2002 and by the EU in 2004. A decrease of α-HCH (Figure 1d) and HCB (Figure 1f) from 1985 on reflects European and U.S. bans on the production and use of these pollutants in the 1980s already. For 9 of the 11 PAHs for which a time trend analysis was possible, a 10- to 30-fold increase of the mussel-inferred water concentration over time was found between 1990 and today (Figure 1c and Figure S3js of the SI). Note that modeled PAH concentrations prior to 1990 are uncertain because they exceed the period for which inferred water concentrations were available. In general, PAH concentrations were at least 100 higher than the concentrations we obtained for the legacy POPs. We only found one other published data set containing both water concentrations of PAHs and of legacy POPs in the Norwegian/ Barents Sea region.30 This data set confirmed the higher concentrations of PAHs than of legacy POPs, although the dissolved PAH concentrations it contained were an order of magnitude lower than those found by us. The observed increase in water concentrations of most PAHs was not an artifact from the approach we followed to infer water concentrations from tissue concentrations. The same trend can be observed from the original tissue concentration data for PAHs in mussels (symbols in Figure 2c), from which the water concentrations were inferred. Eight of the 9 PAHs that increased over time have been shown to originate at least partly from fuel combustion:31 benzo[ghi]perylene, benzo[a]pyrene, benzo[a]anthrancene, antracene, pyrene, phenanthrene, fluoranthene, and benzo[e]pyrene. The two PAHs for which the inferred water concentrations did not increase over time (fluorene, acenaphtylene) are PAHs that are generally not produced by fossil fuel burning but are used in chemical synthesis, e.g., as a precursor of pharmaceuticals. Tissue Concentrations. Our approach yielded predictions of 25 POPs in 21 species, i.e., 525 POP-species combinations. Here, we only show predictions for three species that are representative for various tropic levels (TL) of the Norwegian/Barents Sea food web: the Blue mussel (TL = 3), North East Arctic cod (TL = 4), and Polar bear (TL = 6), as well as the original tissue concentrations 9026
dx.doi.org/10.1021/es202423f |Environ. Sci. Technol. 2011, 45, 9024–9029
Environmental Science & Technology
Figure 2. Observed (dots) and modeled tissue concentrations (median: dotted lines; 95% uncertainty interval: solid lines) of 5 pollutant classes in North East Arctic cod (Gadus morhua, blue), Blue mussel (Mytilus edulis, brown) and Polar bear (Ursus maritimus, red). Sum PCBs includes congeners 28, 52, 101, 105, 118, 138, 153, 156, and 180. Sum DDT includes p,p’-DDE and p,p’-TDE. Sum PAHs includes benzo[ghi]perylene, benzo[a]pyrene, benzo[a]anthrancene, antracene, pyrene, phenanthrene, fluoranthene, benzo[e]pyrene, fluorene, acenaphtylene, and indeno[1,2,3cd]pyrene. Sum HCHs is α-HCH and γ-HCH.
in these three species (Figure 2). Predicted lipid normalized bioaccumulation factors for benthos, fish, and mammals as a function of trophic level are summarized in Figure 3 for the 5 POP classes. Predicted and observed lipid normalized concentrations that are not shown in Figure 2 are shown in Figures S4S6 of the SI. Because the inferred water concentrations served as an input for the predictions of tissue concentrations, the temporal trends of the latter are the same as those of the former. The strategy we propose here—converting available tissue concentrations to time trends of seawater concentrations— considerably expanded the current information on POP concentrations in the Arctic food web. For the 21 species and 25 POPs we considered here, the original database contained data on 2 to 25 POPs per species (median = 14 POPs). Our approach yielded tissue data on all 25 POPs for all 21 species and this for the whole period 19852010. Interestingly, also predictions were obtained for species-chemical combinations for which data were already available in the original database. The data that were not used to infer seawater concentrations (PAHs in other species than mussels, other POPs in other species than fish) were independent from our predictions and could thus be used for validation. This validation exercise showed that our predicted tissue concentrations differed less than a factor 5 from the observed concentrations in 87% of the species- pollutant combinations (Figure S7 of the SI), which is comparable to earlier validation efforts of bioaccumulation models 16 indicating that we can confidently use the predicted tissue concentrations of POPs in species for which no data are available. Our scenario analysis in which we made alternative assumptions on the biotransformation (BT) rates always had a higher sum of squared errors, i.e., a lower accuracy, than the default scenario in which all species had BT rates equal to those of fish (scenario 5) (Figure S8 of the SI). Scenario 4, in which it was hypothesized that BT occurs only in fish and mammals—and not in benthos such as mussels— yielded the same SSE as our default assumption. Interestingly, scenario 6, which was identical to scenario 5 but with
ARTICLE
Figure 3. Log 10, transformed bioaccumulation factors (BAFs) of 5 chemical classes as a function of the trophic level of the 21 included species, including fish (blue), benthos (brown), and mammals (red).
mammals having BT rates that were five times higher than fish, was less accurate than our default scenario 5. The rest of this paper will therefore further discuss the results of the default scenario 5. The bioaccumulation model suggested that concentrations of PCBs and DDTs were 101000 times higher at the top (Polar bear) than at the base of the food web (Blue mussel), indicating biomagnification of these pollutants in the Arctic food web (Figure 2a,b). This can also be seen from the increase of the modeled bioaccumulation factor (BAF) with increasing trophic level (Figure 3a,b), which agrees with earlier findings for these pollutants.3234 Modeled concentrations of HCHs and HCB did not increase with the trophic level (Figure 2d,e), again also illustrated by the modeled BAFs (Figure 3d,e). For PAHs, modeled concentrations decreased with increasing trophic level, a process that is referred to as “biodilution” (Figures 2c and 3c) and has been reported in other marine food webs as well, e.g., the Baltic sea,35 Bohai bay China 36 and Tokio Bay.37 Interestingly, these studies often assume that biodilution of PAHs is caused by an alleged increase in BT rate with trophic level. Although inferred BT capacities have been found to be higher in mammals than in cold-blooded animals,24,25 it is uncertain to what extent this results in higher BT rates for Arctic mammals than for lower trophic levels.15 In our default scenario 5, BT in all species was set to rates for fish and simulation results still displayed biodilution for PAHs (Figure 3c). This implies that no trophic increase of the BT rate is needed for biodilution to occur. Our simulations show that the smaller chemical loss rates in higher trophic levels—a typical allometric effect 38 —combined with quasi invariant BT rates makes the proportion of the chemical that is lost by BT higher at higher trophic levels, as is illustrated for pyrene in Blue mussel and Polar bear in the SI (Figure S9 of the SI). Ecological Risks. Tissue concentrations measured in the field have been compared with laboratory-derived toxicity thresholds to assess the health status of Arctic residents.39 Often, the heterogeneity among the species and end points (e.g., biomarker vs individual-level) for which the available toxicity thresholds are valid has made it difficult to interpret the results of such comparisons, let alone to estimate much needed measures of mixture toxicity.17,39 One approach would be to compare the 9027
dx.doi.org/10.1021/es202423f |Environ. Sci. Technol. 2011, 45, 9024–9029
Environmental Science & Technology sum of all POPs in one species with a threshold for a well-defined end point with a direct relation to the population-level. MacLeod et al.40 have compared hydrocarbon burdens to a whole-body “critical body residue” (CBR) for lethal effects of nonspecific acting narcotics, to assess ecological risks. However, because of the heterogeneity of substances and species included here, and thus possibly various modes of toxic action at play, we refrained from such an exercise. In addition, there are a number of additional reasons why such an approach applied to the data at hand may underestimate toxicity. First, from the initially considered 92 POPs, only 25 were amenable to time trend analysis and 67 had to be excluded. Possible effects of these 67 chemicals would therefore not be accounted for. Second, a range of new POPs have been found in Arctic biota,12,41 many of which have an unknown mode of action. The legislative actions to ban the production and use of PCBs, that were triggered by measured concentrations in mammals, appear to have been successful as concentrations at the top of the Arctic food web have dropped substantially over time. In contrast, concentrations at the base of the Arctic food web after 2000 seem to be dominated by PAHs. However, these concentrations seem to be stabilizing during the last 5 years (Figure 2c), which may be explained by the search for and use of alternative energy sources that are independent of fossil fuels,42 the continuous efforts to increase the efficiency of engines,43 or depleting oil reserves.44 Typically, the main consequences for the Arctic environment of fossil fuel burning are often stated to be an increase in atmospheric CO2 concentrations, an increase in seawater temperatures and sea ice melting.45 Our results, which highlight a substantial increase of PAHs in lower trophic levels during the past 25 years, show that an additional dimension should be added to the debate on the consequences of human reliance on fossil fuels.
’ ASSOCIATED CONTENT
bS
Supporting Information. Tables S1S4 and Figures S1 S10. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Tel: +32 (0) 9 264 3779; Fax: +32 (0) 9 264 4199; E-mail:
[email protected].
’ ACKNOWLEDGMENT The authors thank the Institute of Marine Research, (Bergen, Norway), the National Marine Fisheries Service, Northwest Fisheries Science Centre, Seattle, U.S., the Museum of Natural History (Stockholm, Sweden), the Norwegian Institute for Water Research (Oslo, Norway), and the Norwegian Polar Institute (Oslo, Norway) for the tissue concentration data. Hans Mose Jensen from ICES is thanked for supplying the metadata for these concentration data. Jack J. Middelburg, Michiel Kraak, and Rob Leuven are thanked for their comments on earlier versions of the manuscript. F.D.L is a postdoctoral research fellow from the Fund for Scientific Research (FWO, Flanders). Three anonymous reviewers are thanked for their useful suggestions.
ARTICLE
’ REFERENCES (1) Scientists urge greater protection for arctic as ice recedes. Science 2011, 332, (6030), 646-646. (2) IPCC Climate Change 2007: Impacts, Adaptation, and Vulnerability; University Press: Cambridge, 2007. (3) Rosenzweig, C.; Karoly, D.; Vicarelli, M.; Neofotis, P.; Wu, Q. G.; Casassa, G.; Menzel, A.; Root, T. L.; Estrella, N.; Seguin, B.; Tryjanowski, P.; Liu, C. Z.; Rawlins, S.; Imeson, A. Attributing physical and biological impacts to anthropogenic climate change. Nature 2008, 453 (7193), 353–U20. (4) Chen, Y. C. J.; Guo, Y. L.; Hsu, C. C.; Rogan, W. J. Cognitive development of Yu-Cheng (oil disease) children prenatally exposed to heat-degraded PCBs. JAMA-J. Am. Med. Assoc. 1992, 268 (22), 3213– 3218. (5) Kramer, V. J.; Etterson, M. A.; Hecker, M.; Murphy, C. A.; Roesijadi, G.; Spade, D. J.; Spromberg, J. A.; Wang, M.; Ankley, G. T. Adverse outcome pathways and ecological risk assessment bridging to population-level effects. Environ. Toxicol. Chem. 2011, 30 (1), 64–76. (6) UNEP History of POPs Discovery, Use and Ban. http://www. chem.unep.ch/gpa_trial/04histo.htm (7) USEPA, DDT, a review of scientific and economic aspects of the decision to ban its use as a pesticide. 540/175022. WA, 1975. (8) Muir, D.; Braune, B.; DeMarch, B.; Norstrom, R.; Wagemann, R.; Lockhart, L.; Hargrave, B.; Bright, D.; Addison, R.; Payne, J.; Reimer, K. Spatial and temporal trends and effects of contaminants in the Canadian Arctic marine ecosystem: a review. Sci. Total Environ. 1999, 230 (13), 83–144. (9) Dewailly, E.; Ayotte, P.; Bruneau, S.; Laliberte, C.; Muir, D. C. G.; Norstrom, R. J. Inuit exposure to organochlorines through the aquatic food-chain in Arctic Quebec. Environ. Health Perspect. 1993, 101 (7), 618–620. (10) Simonich, S. L.; Hites, R. A. Global distribution of persistent organichlorine compounds. Science 1995, 269 (5232), 1851–1854. (11) Dewailly, E.; Ayotte, P.; Bruneau, S.; Gingras, S.; Belles-Isles, M.; Roy, R. Susceptibility to infections and immune status in Inuit infants exposed to organochlorines. Environ. Health Perspect. 2000, 108 (3), 205–211. (12) Kelly, B. C.; Ikonomou, M. G.; Blair, J. D.; Surridge, B.; Hoover, D.; Grace, R.; Gobas, F. Perfluoroalkyl contaminants in an arctic marine food web: Trophic magnification and wildlife exposure. Environ. Sci. Technol. 2009, 43 (11), 4037–4043. (13) Lockhart, W. L.; Stewart, R. E. A. Biochemical stress indicators in marine mammals. In Synopsis of Research Conducted Under the 1991/ 92 Northern Contaminants Program, Murray, J. L.; Shearer, R. G., Eds. Indian and Northern Affairs: Ottawa, Canada, 1992; pp 158164. (14) Campfens, J.; MacKay, D. Fugacity-based model of PCB bioaccumulation in complex aquatic food webs. Environ. Sci. Technol. 1997, 31 (2), 577–583. (15) Hendriks, A. J.; van der Linde, A.; Cornelissen, G.; Sijm, D. The power of size. 1. Rate constants and equilibrium ratios for accumulation of organic substances related to octanol-water partition ratio and species weight. Environ. Toxicol. Chem. 2001, 20 (7), 1399–1420. (16) Kelly, B. C.; Ikonomou, M. G.; Blair, J. D.; Morin, A. E.; Gobas, F. Food web-specific biomagnification of persistent organic pollutants. Science 2007, 317 (5835), 236–239. (17) AMAP AMAP Assessment 2002: Persistent Organic Pollutants in the Arctic; Arctic Monitoring and Assessment Programme (AMAP): Oslo, Norway, 2004; p xvi +310. (18) Arnot, J. A.; Gobas, F. A review of bioconcentration factor (BCF) and bioaccumulation factor (BAF) assessments for organic chemicals in aquatic organisms. Environ. Rev. 2006, 14 (4), 257–297. (19) Arnot, J. A.; Mackay, D.; Bonnell, M. Estimating metabolic biotransformation rates in fish from laboratory data. Environ. Toxicol. Chem. 2008, 27 (2), 341–351. (20) Buckman, A. H.; Brown, S. B.; Small, J.; Muir, D. C. G.; Parrott, J.; Solomon, K. R.; Fisk, A. T. Role of temperature and enzyme induction in the biotransformation of polychlorinated biphenyls and bioformation 9028
dx.doi.org/10.1021/es202423f |Environ. Sci. Technol. 2011, 45, 9024–9029
Environmental Science & Technology of hydroxylated polychlorinated biphenyls by rainbow trout (Oncorhynchus mykiss). Environ. Sci. Technol. 2007, 41 (11), 3856–3863. (21) Zaborska, A.; Carroll, J.; Papucci, C.; Torricelli, L.; Carroll, M. L.; Walkusz-Miotk, J.; Pempkowiak, J. Recent sediment accumulation rates for the Western margin of the Barents Sea. Deep-Sea Res. Part II-Top. Stud. Oceanogr. 2008, 55 (2021), 2352–2360. (22) Wassmann, P.; Reigstad, M.; Haug, T.; Rudels, B.; Carroll, M. L.; Hop, H.; Gabrielsen, G. W.; Falk-Petersen, S.; Denisenko, S. G.; Arashkevich, E.; Slagstad, D.; Pavlova, O. Food webs and carbon flux in the Barents Sea. Prog. Oceanogr. 2006, 71 (24), 232–287. (23) Wood, S. N. Stable and efficient multiple smoothing parameter estimation for generalized additive models. J. Am. Stat. Assoc. 2004, 99 (467), 673–686. (24) Han, X.; Nabb, D. L.; Yang, C.-H.; Snajdr, S. I.; Mingoia, R. T. Liver microsomes and s9 from rainbow trout (Oncorhynchus mykiss): comparison of basal-level enzyme activities with rat and determination of xenobiotic intrinsic clearance in support of bioaccumulation assessment. Environ. Toxicol. Chem. 2009, 28 (3), 481–488. (25) Mazur, C. S.; Kenneke, J. F. Cross-species comparison of conazole fungicide metabolites using rat and rainbow trout (Onchorhynchus mykiss) hepatic microsomes and purified human CYP 3A4. Environ. Sci. Technol. 2008, 42 (3), 947–954. (26) Arnot, J. A.; Mackay, D.; Parkerton, T. F.; Zaleski, R. T.; Warren, C. S. Multimedia modeling of human exposure to chemical substances: The roles of food web biomagnification and biotransformation. Environ. Toxicol. Chem. 2010, 29 (1), 45–55. (27) Braune, B. M.; Outridge, P. M.; Fisk, A. T.; Muir, D. C. G.; Helm, P. A.; Hobbs, K.; Hoekstra, P. F.; Kuzyk, Z. A.; Kwan, M.; Letcher, R. J.; Lockhart, W. L.; Norstrom, R. J.; Stern, G. A.; Stirling, I. Persistent organic pollutants and mercury in marine biota of the Canadian Arctic: An overview of spatial and temporal trends. Sci. Total Environ. 2005, 351, 4–56. (28) Scheringer, M.; Salzmann, M.; Stroebe, M.; Wegmann, F.; Fenner, K.; Hungerbuhler, K. Long-range transport and global fractionation of POPs: Insights from multimedia modeling studies. Environ. Pollut. 2004, 128 (12), 177–188. (29) Wania, F.; Mackay, D. Tracking the distribution of persistent organic pollutants. Environ. Sci. Technol. 1996, 30 (9), A390–A396. (30) Lohmann, R.; Gioia, R.; Jones, K. C.; Nizzetto, L.; Temme, C.; Xie, Z.; Schulz-Bull, D.; Hand, I.; Morgan, E.; Jantunen, L. Organochlorine pesticides and PAHs in the surface water and atmosphere of the North Atlantic and Arctic Ocean. Environ. Sci. Technol. 2009, 43 (15), 5633–5639. (31) Ravindra, K.; Sokhi, R.; Van Grieken, R. Atmospheric polycyclic aromatic hydrocarbons: Source attribution, emission factors and regulation. Atmos. Environ. 2008, 42 (13), 2895–2921. (32) Borga, K.; Fisk, A. T.; Hoekstra, P. F.; Muir, D. C. G. Biological and chemical factors of importance in the bioaccumulation and trophic transfer of persistent organochlorine contaminants in arctic marine food webs. Environ. Toxicol. Chem. 2004, 23 (10), 2367–2385. (33) Hop, H.; Borga, K.; Gabrielsen, G. W.; Kleivane, L.; Skaare, J. U. Food web magnification of persistent organic pollutants in poikilotherms and homeotherms from the Barents Sea. Environ. Sci. Technol. 2002, 36 (12), 2589–2597. (34) Wiberg, K.; Andersson, P. L.; Berg, H.; Olsson, P. E.; Haglund, P. The fate of chiral organochlorine compounds and selected metabolites in intraperitoneally exposed Arctic char (Salvelinus alpinus). Environ. Toxicol. Chem. 2006, 25 (6), 1465–1473. (35) Nfon, E.; Cousins, I. T.; Broman, D. Biomagnification of organic pollutants in benthic and pelagic marine food chains from the Baltic Sea. Sci. Total Environ. 2008, 397 (13), 190–204. (36) Wan, Y.; Jin, X. H.; Hu, J. Y.; Jin, F. Trophic dilution of polycyclic aromatic hydrocarbons (PAHs) in a marine food web from Bohai Bay, North China. Environ. Sci. Technol. 2007, 41 (9), 3109–3114. (37) Takeuchi, I.; Miyoshi, N.; Mizukawa, K.; Takada, H.; Ikemoto, T.; Omori, K.; Tsuchiya, K. Biomagnification profiles of polycyclic aromatic hydrocarbons, alkylphenols and polychlorinated biphenyls in
ARTICLE
Tokyo Bay elucidated by delta C-13 and delta N-15 isotope ratios as guides to trophic web structure. Mar. Pollut. Bull. 2009, 58 (5), 663–671. (38) Hendriks, A. J. Allometric scaling of rate, age and density parameters in ecological models. Oikos 1999, 86 (2), 293–310. (39) Fisk, A. T.; de Wit, C. A.; Wayland, M.; Kuzyk, Z. Z.; Burgess, N.; Robert, R.; Braune, B.; Norstrom, R.; Blum, S. P.; Sandau, C.; Lie, E.; Larsen, H. J. S.; Skaare, J. U.; Muir, D. C. G. An assessment of the toxicological significance of anthropogenic contaminants in Canadian arctic wildlife. Sci. Total Environ. 2005, 351, 57–93. (40) MacLeod, M.; McKone, T. E.; Foster, K. L.; Maddalena, R. L.; Parkerton, T. F.; Mackay, D. Applications of contaminant fate and bioaccumulation models in assessing ecological risks of chemicals: A case study for gasoline hydrocarbons. Environ. Sci. Technol. 2004, 38, 6225–6233. (41) Riget, F.; Bignert, A.; Braune, B.; Stow, J.; Wilson, S. Temporal trends of legacy POPs in Arctic biota, an update. Sci. Total Environ. 2010, 408 (15), 2874–2884. (42) Verhelst, S.; Wallner, T. Hydrogen-fueled internal combustion engines. Prog. Energy Combust. Sci. 2009, 35 (6), 490–527. (43) Johnson, T. V. Review of diesel emissions and control. Int. J. Engine Res. 2009, 10 (5), 275–285. (44) Ehrenfeld, D. Recalculating future oil reserves. Science 2005, 309 (5731), 54–56. (45) Raper, S. C. B.; Braithwaite, R. J. Low sea level rise projections from mountain glaciers and icecaps under global warming. Nature 2006, 439 (7074), 311–313.
9029
dx.doi.org/10.1021/es202423f |Environ. Sci. Technol. 2011, 45, 9024–9029
ARTICLE pubs.acs.org/est
Effect of Temperature on PV Potential in the World Kotaro Kawajiri* Research Institute of Science for Safety and Sustainability, National Institute of Advanced Industrial Science and Technology 16-1, Onozawa, Tsukuba, Ibaraki, 305-8569, Japan
Takashi Oozeki* Research Center for Photovoltaics, National Institute of Advanced Industrial Science and Technology, AIST Tsukuba Central 2, 1-1-1 Umezono, Tsukuba, Ibaraki, 305-8568, Japan
Yutaka Genchi* Research Institute of Science for Safety and Sustainability, National Institute of Advanced Industrial Science and Technology 16-1, Onozawa, Tsukuba, Ibaraki, 305-8569, Japan ABSTRACT: This work aims to identify the geographic distribution of photovoltaic (PV) energy potential considering the effect of temperature on PV system performance. A simple framework is developed that uses the JIS C 8907 Japanese industrial standard to evaluate the effects of irradiation and temperature on PV potential. The global distributions of PV potential and yearly performance ratio are obtained by this framework. Generally, the performance ratio decreases with latitude because of temperature. However, regions with high altitude have higher performance ratios due to low temperature. The southern Andes, the Himalaya region, and Antarctica have the largest PV potentials. Although PV modules with less sensitivity to temperature are preferable for the high temperature regions, PV modules that are more responsive to temperature may be more effective in the low temperature regions. The correlation between the estimates obtained by our framework and results from a more data-intensive method increases when the temperature effects are considered.
’ INTRODUCTION Recently, as concern about climate change and nuclear power increases, interest in renewable energy is high. However, the geological dispersion of renewable energy sources around the world makes assessment of their potential difficult. A global map of a renewable energy source would help clarify regions with net energy supply or demand, e.g., for connecting the demand and supply of renewable energy sources across international borders.1 Photovoltaic (PV) electric power generation is a promising technology for generating renewable energy from solar irradiation. However, the output of PV is sensitive to its operating conditions, so estimating PV potential accurately is a complex problem. Furthermore, given the limited availability of data for the entire world, a method that achieves accurate estimates with available data is necessary. Most estimates of PV potential use either the power rating method or the energy rating method.2 The power rating method integrates the instantaneous PV power generation over time, thereby accounting for the time-dependency of PV output.3 Huld et al. estimated the PV potential in Europe, using the framework of King et al. to account for ambient temperature and PV module type.4,5 The Photovoltaic Geographical Information System (PVGIS) used the method of Huld et al. to estimate the PV r 2011 American Chemical Society
potential in Europe and Africa.6 The main problem of this method is its complexity and data requirements. Complete instantaneous weather data is not available globally, so no work has estimated the global PV potential by the power rating method. The energy rating method estimates PV potential by multiplying the total solar irradiation during a specific period of time by a performance ratio. The simplicity of the energy rating method and the availability of global weather data 7 has enabled researchers to estimate the PV potential for the world,8 and numerous countries.911 These studies use a constant performance ratio. However, the performance ratio actually changes under different operating conditions, especially ambient temperature,12,13 which limits the accuracy of these studies. Our objective is to identify the areas of the world with the highest PV potential considering the spatial and temporal variation of temperature. JIS C 8907 is the Japanese industrial standard for estimating the performance ratio for PV energy generation as a function of a number of factors including ambient Received: February 23, 2011 Accepted: August 18, 2011 Revised: August 12, 2011 Published: August 18, 2011 9030
dx.doi.org/10.1021/es200635x | Environ. Sci. Technol. 2011, 45, 9030–9035
Environmental Science & Technology
ARTICLE
temperature.14 Because ambient temperature generally has the largest effect on the performance ratio, we develop a modified energy rating method based on the JIS method that estimates the effect of ambient temperature on global PV potential. We use the method to generate a global map of PV potential and annual performance ratio. The results are verified using estimates from PVGIS.
’ METHOD Estimation of PV Potential. We estimate the PV potential for PV systems mounted on a platform above ground and operated under direct connection to the grid without any kind of storage such as batteries. We assume that the PV systems consist of crystalline silicon PV (c-Si PV) modules: c-Si PV includes both multicrystalline silicon PV and single-crystalline silicon PV, which are the most popular modules in the present market.15 The annual PV energy generation EPy (kWh) for each region is obtained by summing the monthly PV energy generation EPm (kWh) in that region. Regions are obtained by gridding the entire globe along 360 ° longitude and 180 °latitude into 64 800 regions.
EPy ¼
∑EPm PAS KHAm GS
KPT ¼ 1 þ αPmax ðTAm þ ΔT 25Þ
ð2Þ
ð6Þ
K 0 ¼ KH KPD KPA KPM KB KC
ð7Þ
In this work, we set the design factor K to 0.75 based on the recommended values shown above. The annual PV potential per nominal power is obtained using the results from eqs 1, 2, and 6: Ypy ¼ ¼
ð4Þ
Here, KPD, KPA, and KPM are coefficients showing losses by time-dependent module function, array circuit, and array load, respectively. Staining and dust on the PV array surface would affect KPD, but we ignore those effects here. We use the value recommended for c-Si PV modules of 0.95 for KPD, the generally recommended value of 0.97 for KPA, and the recommended value
Epy PAS K 0 12 f1 þ αPmax ðTAm þ ΔT 25ÞgHAm GS m ¼ 1
∑
ð8Þ
The yearly performance ratio is obtained from eq 9: Ky ¼
YPy GS HAy 12
ð3Þ
Here, KH, KP, KB, and KC are parameters for irradiation/ energy losses on the PV array surface, in the PV module, in the battery circuit, and in the power conditioner circuit, respectively. We use the JIS recommended value of 0.97 for KH. Although shading on the PV array surface would affect this value, we ignore that effect here. We set the value of KB to be 1 because we have assumed grid-connected operation. We use the recommended value of power conditioners, 0.90, for KC. The effect of ambient temperature on the PV module is included in KP, by a correction term KPT . KP ¼ KPD KPT KPA KPM
K ¼ K 0 KPT ¼ K 0 f1 þ αPmax ðTAm þ ΔT 25Þg
0
Here, K is the performance ratio, PAS is the nominal power of the PV array at standard test conditions in kW, HAm is the monthly total solar irradiation on the PV array in kWh/m2, and GS is the solar irradiance at standard test conditions, which is 1 kW/m2.16 HAm is obtained by multiplying the number of days in the month by the average daily irradiance over the course of the month on equator-pointed surfaces tilted at the latitude angle, which are obtained from the NASA Surface meteorology and Solar Energy database.7 K is obtained from eq 3. K ¼ KH KP KB KC
ð5Þ
Here, αPmax is the maximum power temperature coefficient, which is 0.0041 °C1 for c-Si PV modules; TAm is the 24 h ambient temperature profile averaged over month m; and ΔT is the weighted average of module temperature annual increase. We obtained a value of 18.4 °C for ΔT of PV systems mounted on platforms.17 This value was confirmed to be valid between 20 and 40 °C for 64 Japanese residential PV systems.17 Although regional differences in factors such as wind and humidity might also affect the value, we believe that it should be valid within (0.1 °C. The calculated value of KPT was verified to be within (3% of the measured value for the PV systems. The value of TAm for each region is obtained from the NASA database.7 All parameters except for KPT are assumed to be constant in order to focus on the effect of temperature on the PV potential. Then, K is given as follows.
ð1Þ
The monthly energy generation is obtained from eq 2. EPm ¼
for grid-connected PV modules of 0.94 for KPM. KPT is given by eq 5.
¼K
∑ f1 þ αPmax ðTAm þ ΔT 25ÞgHAm 0m ¼ 1 12
∑
m¼1
ð9Þ
HAm
Weather Conditions in the World. Figure 1 shows the global distribution of annual total irradiation on equator-pointed tilted surfaces obtained by summing the monthly total solar irradiation values in the NASA database, which are averages of 22 years of data from 1983 to 2005.7 These irradiation values include the effects of cloud cover and contain both direct and indirect solar irradiance, both of which can be used by PV systems. Generally, irradiation is largest near the equator and at high altitudes. Regions with irradiation values greater than 2000 kWh/m2 include the southwest region in North America, the Southern Andes region in South America, central and south Africa, midwest Asia, the Himalaya region in Asia, the northwest region of Australia, and Antarctica. Irradiation values in countries having large amounts of installed PV systems, such as Germany and Japan, are relatively small. 9031
dx.doi.org/10.1021/es200635x |Environ. Sci. Technol. 2011, 45, 9030–9035
Environmental Science & Technology
ARTICLE
Figure 1. Global map of annual total irradiation on equator-pointed surfaces tilted at the latitude angle.
Figure 2. Global map of average ambient temperature.
Figure 2 shows the global distribution of annual average ambient temperature.7 Generally, irradiation and temperature are correlated except in Antarctica. However, the annual average ambient temperatures in the Himalaya and Southern Andes regions are much lower than that in other regions with the same latitude due to high altitudes. Temperature decreases with increasing altitude at a rate of about 9.8 °C/km to 4.0 °C/km.18 The altitudes in the Himalaya and Southern Andes regions reach over 5000 m. Therefore, the temperatures in those regions could be up to 50 °C lower than regions at sea level with the same latitude. Irradiation also increases with increasing altitude 19,20 as shown in Figure 1.
’ RESULTS AND DISCUSSION Figure 3 shows the global map of annual energy generation potential by c-Si PV systems. The regions with the largest irradiation values have large PV potentials. In particular, the Himalaya and Southern Andes regions have energy potentials of more than 1800 kWh/kW PV, due to the combination of large irradiation values and low temperatures. The Himalayan region is
especially attractive because it is near regions with large future energy demands, such as China and India. Installation of 1.2 104 km2 of c-Si PV arrays rated at 1800 kWh/kW PV and 0.14 kW/m2 21 in this region could meet the total electricity consumption in China of about 3.1 trillion kWh in 2007.22 This area is less than 4% of the area of the Himalaya region having a PV potential greater than 1800 kWh/kW. Moreover, because CO2 emissions per unit electricity in China and India are larger than those in the developed countries, using PV energy in these regions could have a large mitigation effect on climate change.23 Of course, many problems must be addressed when installing PV systems in high altitude regions, such as transporting the PV system and increased need for maintenance due to the severe environmental conditions. Several high-altitude PV plants are currently in operation.24 Antarctica also has large PV potential because of its high irradiation and low temperature. However, due to the large seasonal variance of PV energy generation, with essentially no energy generation from April to September, together with the 9032
dx.doi.org/10.1021/es200635x |Environ. Sci. Technol. 2011, 45, 9030–9035
Environmental Science & Technology
ARTICLE
Figure 3. Global potential map of PV energy generation by c-Si PV module.
Figure 4. Global potential map of PV energy generation without considering temperature effect.
lack of significant energy demand in this region and the difficulty of maintaining the PV system, the feasibility for utilizing the PV potential in this region is low. If such disadvantages can be overcome, or if some way can be developed to store the generated energy, e.g. in the form of hydrogen or refined metals, then it may be possible to utilize the large potential in this region in the future. Figure 4 shows the global potential map of PV energy generation without considering the temperature effect. Comparing this figure with Figure 3, the effect of temperature on the PV potential is clear. At an ambient temperature of 40 °C, the loss of energy due to temperature is almost 13% for a conventional c-Si PV module. The loss can be reduced by using PV modules having an αPmax greater than 0.0041 °C1, such as amorphous Si or CdTe PV modules.25 This would increase the PV potential in regions with large temperature, such as Africa, Mideast Asia, and Australia. Conversely, the PV potential decreases for larger values of αPmax in regions where the monthly average temperature is less than 6.6 °C, such as Antarctica, the Himalayans, and the southern Andes, because TAm + ΔT 25 becomes negative
forΔT equal to 18.4 °C in eq 5. Those regions should use PV modules, whose performance increases at less than 6.6 °C, such as c-Si or CIS PV modules.25 Figure 5 shows the yearly performance ratio Ky obtained by eq 9. Temperature differences cause the performance ratio to vary from 0.65 to 0.90. We examine the accuracy of the JIS method by comparing the estimates with and without considering temperature of the PV potentials with estimates for capital cities in Europe and Africa obtained from the PVGIS Web site 6 for tilt angle equal to latitude, c-Si PV modules, and system loss of 0.25 (see Figure 6). The regression line for the JIS results without considering temperature overestimates the PV potential by nearly 15% in locations with large PV potential, irradiation, and temperature are often correlated, so the PV potentials in locations having large irradiations could be overestimated if the temperature effect is neglected. The differences between the results from the JIS and PVGIS methods are large in locations such as Reykjavik, Oslo, Douglas in Europe and Asmara, Nouakchott, and Harare in Africa even 9033
dx.doi.org/10.1021/es200635x |Environ. Sci. Technol. 2011, 45, 9030–9035
Environmental Science & Technology
ARTICLE
Figure 5. Global map of PV performance ratio for c-Si PV system.
Figure 6. Comparison of the PV potential estimation results of PVGIS and JIS.
when considering temperature. All of these cities are surrounded by mountains, ocean, or desert. The resultant steep gradients in geographical features cause sharp weather data variations that cannot be captured accurately by low resolution weather data, especially in Africa. The resolution of the NASA database is 1 degree, and that of PVGIS is 1.5 arcminutes.6 Therefore, our results are likely to be less accurate for those cities. Table 1 shows the p-values for linear regressions of HAy, Ky, and YPy obtained by the JIS method with consideration of temperature (T) and without consideration of temperature (NT) against the values obtained by the PVGIS method. The p-values are all below 0.01 except for Ky calculated without considering temperature, especially in Africa. Neglecting the effect of temperature could result in overestimation of the
performance ratio in high temperature regions. In addition, Ky is also affected by both differences of weather data sources and differences between the JIS and PVGIS estimation methods. The NASA and PVGIS weather data differ by geological resolution. However, previous research indicates that the accuracy of PV energy generation increases when the hourly variation of the weather data is considered,4 which the PVGIS method does but the JIS method does not. These differences reduce the correlation of the performance ratio, but not HAy and YPy, when temperature is not considered. The framework we have presented can be used to link PV specifications with the variation of PV potential, e.g., to show what type of PV module is suitable for different regions. These findings could help stakeholders to evaluate the potential for PV energy generation considering temperature in addition to 9034
dx.doi.org/10.1021/es200635x |Environ. Sci. Technol. 2011, 45, 9030–9035
Environmental Science & Technology
ARTICLE
Table 1. p-Values for the correlation between the estimates given by PVGIS and JIS Ky []
YPy [kWh/kW]
HAy [kWh/m2]
T
NT
T
NT
Total
<0.01
<0.01
0.99
<0.01
<0.01
Europe
<0.01
<0.01
0.80
<0.01
<0.01
Africa
<0.01
<0.01
1.00
<0.01
<0.01
irradiation. For example, we identified a large PV potential in the Himalayan region that is close to large future energy consumers such as China and India. Although we have focused on temperature, other factors could be included in JIS framework. For example, one could consider the snowfall effect in KH, the cooling effect by wind by modifying eq 5, and the staining of the PV modules by wind-borne particles in KPD. We hope that this work will provide a first step to investigate the most suitable locations for PV energy generation in the world.
’ AUTHOR INFORMATION Corresponding Author
*Tel: +81-29-861-8114 Fax: +81-29-861-8118 E-mail:
[email protected] (K.K) Tel: +81-29-861-5449 Fax: +81-29861-5829 E-mail:
[email protected] (T.O.). Tel: +81- 29861-8036 Fax: +81-29-861-8118 E-mail:
[email protected] (Y.G.).
’ ACKNOWLEDGMENT We thank Steven Kraines for editing this manuscript. This research was partially supported by the grant of The Research Institute of Science for Safety and Sustainability in The National Institute of Advanced Industrial Science and Technology and also by Grant-in-Aid for Young Scientists (A) (23681008) by Japan Society for the Promotion of Science.
(11) Country reports on PV system performance; IEA-PVPS T205, IEA, 2004. (12) Wysocki J. J.; Rappaport P. Effect of temperature on photovoltaic solar energy conversion. J. Appl. Phys., 1960, 31 (3), 571578; DOI 10.1063/1.1735630. (13) Yukawa, M.; Asaoka, M.; Takahara, K.; Ohshiro, T.; Kurokawa, K. Estimation of photovoltaic module temperature rise. Trans. Inst. Elect. Engnr. Jpn. 1996, 116-B (9), 1101–1110. (14) Estimation method of generating electric energy by PV power system. Japanese Industrial Standard, JIS C 8907 (2005). (15) Yole Development, PV Fab Database 2010, 2010. (16) IEC, Photovoltaic system performance monitoringGuidelines for measurement, data exchange and analysis, IEC 61724 (1998). (17) Development of Evaluation Methodology for PV System; NEDO, 2000. (18) Barry, R. G; Chorley, R. J. Atmosphere, Weather and Climate, 5th ed.; Routledge: London, pp 56. (19) Blumthaler., M; Ambach, W; Ellinger, R. Increase in solar UV radiation with altitude. J. Photochemistry and Photobiology B: Biology 1997, 39, 130–134. (20) Hottel, H. C. A simple model for estimating the transmittance of direct solar radiation through clear atmospheres. Solar Energy 1976, 18, 129–134. (21) NEDO. Investigation of life cycle assessment of photovoltaic system, 2008. (22) World Bank Website; http://data.worldbank.org/data-catalog/world-development-indicators. (23) AIST-LCA ver.5, National Institute of Advanced Industrial Science and Technology. (24) PVRESOURCES Website; http://www.pvresources.com/en/ highest.php. (25) Del Cueto, J. A. Comparison of energy production and performance from flat-plate photovoltaic module technologies deployed at fixed tilt. Proc. 29th IEEE PV Specialist Conference 2002, 20–24.
’ REFERENCES (1) Feresin E. Europe looks to draw power from Africa. Nature, 2007, 450, pp 595; DOI 10.1038/twas08.9a. (2) Kroposki B.; Emery K.; Myers D.; Mrig L. A. comparison of photovoltaic module performance evaluation methodologies for energy ratings. 1995; NREL/TP-4117426. (3) Ransome, S; Funtan, P. Why hourly averaged measurement data is insufficient to model PV system performance accuracy. Proc. 20th PVSEC 2005. (4) Huld T; Gottschalg R; Beyer H. G; Topic M. Mapping the performance of PV modules, effects of module type and data averaging, Solar Energy, 2010, 84, pp 324-338; DOI 10.1016/j.solener.2009.12.002. (5) King D. L; Boyson W. E; Kratochvil J. A. Photovoltaic Array Performance, SAND20043535, 2004. (6) PVGIS Website; http://re.jrc.ec.europa.eu/pvgis/. (7) NASA Website; http://eosweb.larc.nasa.gov/sse/. (8) de Vries B. J. M.; van Vuuren D. P.; Hoogwijk M. M. Renewable energy sources: Their global potential for the first-half of the 21st century at a global level: An integrated approach. Energy Policy, 2007, 35, 25902610; DOI 10.1016/j.enpol.2006.09.002. (9) Suri M.; A.Huld T.; Dunlop E. D.; Ossenbrink H. A. Potential of solar electricity generation in the European Union member states and candidate countries. Solar Energy, 2007, 81, 12951305; DOI 10.1016/ j.solener.2006.12.007. (10) Compared assessment of selected environmental indicators of photovoltaic electricity in OECD cities; IEA-PVPS-T1001, IEA, 2006. 9035
dx.doi.org/10.1021/es200635x |Environ. Sci. Technol. 2011, 45, 9030–9035
ARTICLE pubs.acs.org/est
Effects of a Catalytic Volatile Particle Remover (VPR) on the Particulate Matter Emissions from a Direct Injection Spark Ignition Engine Fan Xu,* Longfei Chen, and Richard Stone Combustion Engines Group, Department of Engineering Science, University of Oxford, Oxford, United Kingdom
bS Supporting Information ABSTRACT: Emissions of fine particles have been shown to have a large impact on the atmospheric environment and human health. Researchers have shown that gasoline engines, especially direct injection spark ignition (DISI) engines, tend to emit large amounts of small size particles compared to diesel engines fitted with diesel particulate filters (DPFs). As a result, the particle number emissions of DISI engines will be restricted by the forthcoming EU6 legislation. The particulate emission level of DISI engines means that they could face some challenges in meeting the EU6 requirement. This paper is an experimental study on the sizeresolved particle number emissions from a spray guided DISI engine and the performance of a catalytic volatile particle remover (VPR), as the EU legislation seeks to exclude volatile particles. The performance of the catalytic VPR was evaluated by varying its temperature and the exhaust residence time. The effect of the catalytic VPR acting as an oxidation catalyst on particle emissions was also tested. The results show that the catalytic VPR led to a marked reduction in the number of particles, especially the smaller size (nucleation mode) particles. The catalytic VPR is essentially an oxidation catalyst, and when post three-way catalyst (TWC) exhaust was introduced to the catalytic VPR, the performance of the catalytic VPR was not affected much by the use of additional air, i.e., no significant oxidation of the PM was observed.
1. INTRODUCTION With the tightening of emissions legislation, more research concerns the particle emissions from gasoline engines, especially those with direct injection. Recently, much attention has been given to the influence of fine particulate material (PM) in the atmosphere on human health. Concerns about possible adverse health effects of particles from engines led to a re-examination of PM standards and measurement protocols by Kittelson.1 The total particle size distribution can be divided into three modes, namely a nucleation mode, an accumulation mode, and a coarse mode. Most of the particle mass exists in the so-called accumulation mode in the 0.10.3 μm diameter range. This is where the carbonaceous agglomerates and associated adsorbed materials reside. The nucleation mode typically consists of particles in the 0.0050.05 μm diameter range. This mode usually consists of volatile organic and sulfur compounds that form during exhaust dilution and cooling, and may also contain solid carbon and metal compounds. The nucleation mode typically contains 120% of the particle mass but more than 90% of the particle number. Various tests have already been carried out on vehicles with different types of engine.26 Generally, it is found that particles are generated during acceleration and cold start in a drive cycle. For port fuel injection gasoline vehicles, after catalyst light-off, particle emissions diminished to near negligible levels at moderate cruise speeds and during deceleration. It has been indicated by various researchers that the PFI and DPF diesel vehicles r 2011 American Chemical Society
showed the lowest particulate emissions, while particle emissions from diesel vehicles without DPFs were the highest.79 The level of particle emissions from a DISI vehicle was between that of the PFI and diesel vehicles.5,6,1014 DISI engines tend to emit higher levels of particulate emissions during stratified operation.5,15 It was shown by Ericsson et al.5 that the difference in particulate number between the homogeneous and the stratified modes was a factor of approximately 40. Although most research has found that particle number (Pn) emissions from DISI engines were higher than those of DPFdiesel or PFI gasoline engines by an order of magnitude,5,6,1014 the research from CONCAWE16 showed that the Pn emissions from DISI vehicles were of the same order as those from the DPF diesel vehicles over the NEDC cycle. So it is of interest to further investigate the particle emissions from the most recent DISI engines. It is well-known that when a DISI engine operates in stratified lean conditions, it will emit a high number concentration of particles. It can be argued that, during stratified combustion, soot was formed as a partially premixed flame propagated through locally rich zones and from pool fires caused by thin films of liquid fuel on the piston surface.1719 However, during stoichiometric operation when using a piezoelectric injector with Received: March 11, 2011 Accepted: September 8, 2011 Revised: June 14, 2011 Published: September 08, 2011 9036
dx.doi.org/10.1021/es2008209 | Environ. Sci. Technol. 2011, 45, 9036–9043
Environmental Science & Technology
ARTICLE
vaporized hydrocarbons so that there is then no possibility of their recondensation. So, it is of interest to investigate the performance of the catalytic stripper based volatile particle remover system on the PM emissions from a spray guided direct injection (SGDI) engine; previous work has utilized a wall guided direct injection engine.26 In the work reported here, engine experiments were carried out to study a catalytic volatile particle remover (VPR), and to find out if it can achieve the transmission efficiencies required by the PMP procedures.
Figure 1. Experimental setup and construction of the catalytic VPR system.
multiple injections in a DISI engine (to further increase homogenization and reduce spray penetration), then the Pn emissions may be lowered. By using turbocharging for downsizing, stoichiometric DISI engines show advantages on fuel economy over turbocharged MPI engines20 and lowered Pn emission compared with lean DISI engines. To investigate the characteristics of particle emissions from engines, a catalytic stripper (CS, an oxidation catalyst) has been tested by Kittelson and co-workers.21,22 It was reported that complete removal of volatile particles could be achieved but that it also led to the removal of 1525% of the solid particles. This loss has been attributed to thermophoretic deposition caused by cooling after the catalytic stripper. A volatile particle remover (VPR) system is required by the European PMP program.10,23 The particle loss in the catalytic stripper led to the development of a solid particle sampling system (SPSS)24 that used dilution immediately after an oxidation catalyst. Khalek24 presents a comprehensive discussion and analysis of the operation of the SPSS, and characterizes its performance using salt (NaCl), ammonium sulfate, and oil aerosols. The ammonium sulfate decomposed and (as with the oil) led to particle penetration through the SPSS that was always below 10%. In contrast, the penetration of the salt aerosol was, within experimental tolerance, 100%. In a subsequent paper Khalek and Bougher25 compared the performance of the SPSS using a catalytic stripper (CS, an oxidation catalyst) with an evaporation tube (ET) using particles of tetracontane (as specified in the European PMP). They conclude that the catalytic stripper outperforms the evaporation tube, since the catalytic stripper will oxidize the
2. EXPERIMENTAL APPARATUS The test engine was a Jaguar naturally aspirated, V8 direct injection spark ignition engine.27 The main characteristics of the engine are listed in Table S1. During these tests, the spark timing was fixed at 30 oCA (crank angle degree) before compression top dead center. A single injection during the intake stroke at 280 oCA before compression top dead center to generate a nominally homogeneous mixture. A schematic of the particulate emissions measurement system and the construction of the catalytic VPR system tested in this work are shown in Figure 1. The geometric data for the catalytic VPR system tested in this work are shown in Figure 1, and its performance as an oxidation catalyst is discussed later in the context of Figure S5 (see Supporting Information). The catalytic VPR system consists of a temperature-controlled heated tube with an oxidation catalyst inside. The effects of the VPR temperature and residence time on the catalytic VPR performance were examined. It can be seen from Figure 1 that in Test 1, the exhaust preTWC was drawn into the catalytic VPR. At each test point, the PM emissions pre-VPR and post-VPR were measured using a Cambustion DMS500 particle sizer.28 The sampling flow rate of the DMS (MFMDMS) was set to 0.02 g/s. A pneumatic switching valve was used to select between pre-VPR and post-VPR. The switching valve was heated by hot air to keep the sampling temperature at around 90 °C, thereby preventing particle formation by condensation. The temperature of the sampling flow pre- and post-VPR, as well as the flow temperature entering DMS, were measured using thermocouples. Downstream of the catalytic VPR, the exhaust flow then entered a dilution/extraction system. As a result, the residence time of the exhaust gas in the catalytic VPR was varied, and the effects of exhaust residence time on PM emissions were studied by means of varying the flow rate of the dilution gas (MFMair) added upstream of the VPR. Also, in this test, the VPR temperature was changed to investigate its effect on the VPR performance. The test conditions are listed in Table 1. The residence time (tr) is calculated assuming a linear variation in the temperature within the catalytic VPR along its length. m_ Rs ðT1 þ T2 Þ=2 ð1Þ V_ ¼ P tr ¼ V =V_ ð2Þ where V_ (m3/s) represents the volume flow rate through the VPR, m_ (g/s) represents the mass flow rate through VPR, Rs (J g1 K1) represents the specific gas constant of exhaust, T1 (K) represents the pre-VPR temperature, T2 (K) represents the post-VPR temperature, P (Pa) represents the exhaust pressure, tr (s) represents the residence time, and V (m3) represents the internal volume of the catalytic VPR. In Test 2, the exhaust was sampled downstream of the TWC (Figure 1). The test procedure was as follows. At each engine 9037
dx.doi.org/10.1021/es2008209 |Environ. Sci. Technol. 2011, 45, 9036–9043
Environmental Science & Technology
ARTICLE
Table 1. Tests Conditions
particle mass was obtained previously from back-to-back tests with the DMS500 and a centrifugal couette flow particle mass analyzer on the same engine.30 massdp ¼ 1:72 1024 dp 2:65 ndp ð4Þ
residence times (s) for test 1 MFMexhaust/(g/s) VPR temperature/°C
0.02
0.09
0.11
0.14
0.22
150
15.9
3.36
2.68
2.05
1.24
250
14.8
3.01
2.35
1.76
1.08
350
13.8
2.98
2.29
1.71
1.02
mass flow of additional gas (g/s) for test 2 VPR-DF engine Lambda 1 0.9
additional gas
1
1.1
1.2
1.3
air
0
0.02
0.04
0.06
N2
0
0.02
0.04
0.06
air
0
0.02
0.04
0.06
N2
0
0.02
0.04
0.06
Lambda (actual/stoichiometric airfuel ratio), supplementary air was first used to study the oxidation effect of the catalytic VPR. By adjusting the needle valve in the additional gas supply route, the mass flow rate of the additional air was controlled to give the Lambda meter reading that was 0.1, 0.2, or 0.3 above the Lambda at which the engine was operating. At each Lambda meter reading, the mass flow rate of the additional gas (MFMadd gas) was measured by MFM 3 and was recorded. Then, the air supply was swapped with nitrogen (N2) to investigate the dilution effects of additional gas on the PM emissions. At each engine operating point, the needle valve was adjusted to let MFM 3 give the same readings as those recorded in the experiments with additional air. By comparing the two sets of results (additional air against N2), the oxidation effect and dilution effect of the additional gas can be examined separately. The VPR dilution factor (VPR-DF) is defined as VPR-DF ¼ 1 þ ðMFMaddgas =MFMexhaust Þ
ð3Þ
where MFMexhaust represents the total mass flow rate of exhaust. When testing with additional air at stoichiometric conditions, the VPR-DF is the same as the Lambda meter reading. The test point conditions are listed in Table 1. During the experiments, both the raw data and log-normal fits of DMS signal were logged into the computer for a number of samples. The DMS sampling line was maintained at 80 °C, and internally controlled dilution air was preheated before being mixed with the sample at entry to the sampling line. Before entry to the DMS sampling line, the pipework and pneumatic switching valve were maintained at above 80 °C. A typical sample flow into the DMS was 2 LPM so the residence time in the sampling system is very short (∼4 s) so diffusion and thermophoretic losses can be ignored. The DMS log-normal output is generated by a Bayesian statistical algorithm,29 which separates the nucleation mode from the accumulation mode particles. The DMS output is a matrix containing number concentration at a given sample number (i.e., time) and particle diameter. The number matrix is then converted to a mass matrix by using an empirical equation. The number matrix and mass matrix are averaged to find a mean concentration plot over particle diameter, which can then be integrated to obtain the total number and total mass of particles emitted by the engine. The formula used to calculate
where dp represents the diameter of the particles (nm), massdp (kg/cm3) represents particle mass concentration of diameter dp, and ndp (no./cm3) represents particle number concentration of diameter dp. From this equation, it can be seen that the larger particles contribute much more to the particle mass than smaller particles. The log-normal fitting is particularly useful as it eliminates the noise at both the high and low ends of the measured raw spectrum. As a result, it is more appropriate to use log-normal fitted data in calculating particle mass distributions from the particle size distribution than to use raw data. European legislation requires the particle number emissions measurement equipment to have counting efficiencies of 50% ((12%) at particle sizes of 23 nm ((1 nm) and of >90% at 41 nm ((1 nm) electrical mobility diameters.23 During data postprocessing, a digital filter using a Wiebe function was applied to the raw measurement data to simulate this legislation requirement for the counting efficiency. The formula for the filter is as follows. " # dp 14 1:09 ð5Þ f ¼ 1 exp 3:54 40 where dp (nm) represents the diameter of the particles. A plot of the filter is shown in the SI (Figure S1), in which the legislation requirements are plotted with cross points. It can be seen that the counting efficiencies represented by the filter fulfill the legislative requirement.
3. RESULTS AND ANALYSIS The experimental results are divided into two parts corresponding to the tests described above. The particle measurement results presented here are an average of data measured for 90 s.
Effects of the VPR Temperature and Residence Time on the Catalytic VPR Performance. Figure S2 shows the tempera-
tures of the pre-VPR and post-VPR sample points under different test conditions for engine-out emissions. It can be seen that the temperatures pre- and post-VPR increase as the mass flow rate of the exhaust becomes higher, showing that the mass flow rate of the exhaust extraction affects the temperature at inlet and outlet of the VPR. This is because as the MFMexhaust increases, the velocity of the exhaust becomes higher, which leads to less cooling per unit mass along the copper pipe connecting the engine exhaust pipe to the catalytic VPR. Figure S3 shows a representative raw size distribution from the DMS data and its log-normal fitting, which was taken pre-VPR with a 250 °C VPR temperature and 0.14 g/s mass flow of exhaust. Symonds et al.29 gave a very comprehensive discussion of how the bilognormal fit is obtained using Bayesian statistics applied to the original electrometer current measurements. It can be seen that the bilognormal result fits the trend of the raw data very well. As the log-normal fitting is very useful in eliminating the noise in the raw data at both the high and low ends of the particle size range, the log-normal fitting results are used in the following analysis. The number transmission efficiency across the catalytic VPR is defined as ηdp ¼ ndp ðdownstreamÞ=ndp ðupstreamÞ 9038
ð6Þ
dx.doi.org/10.1021/es2008209 |Environ. Sci. Technol. 2011, 45, 9036–9043
Environmental Science & Technology
ARTICLE
Figure 2. Number transmission efficiencies versus dp with different sample flow rates taken from the engine exhaust (MFMexhaust) and different VPR temperatures (refer to Table 1) and compared to eq 7.
Figure 3. Number transmission efficiencies for different modes (total, nucleation, and accumulation) for different sample flow rates (MFMexhaust) and different VPR temperatures (refer to Table 1).
where ηdp represents the transmission efficiency of diameter dp, ndp(downstream) (no./cm3) represents particle number concentration of diameter dp downstream the catalytic VPR, and ndp(upstream) (no./cm3) represents particle number concentration of diameter dp upstream the catalytic VPR. The number transmission efficiencies are shown in Figure 2. It can be seen that the transmission efficiencies increase as the particle size increases. The curves for different VPR temperatures and MFMexhaust are similar, except that the curves for MFMexhaust = 0.02 g/s show relatively lower transmission efficiencies at the low end of the particle size range (<50 nm). Also these results show that the catalytic VPR has more effect on the nucleation mode particles than the accumulation mode particles. In some cases, for example at 250 °C VPR temperature and 0.11 g/s MFMexhaust, a sudden rise of the transmission efficiency can be seen at the high end of the particle size range. This is due to noise in the raw DMS data at the high end of the particle size range. Also, it can be seen for particles with dp > 40 nm the loss is about 30% as the exhaust flows through the catalytic VPR (e.g., Figure 2b). So as to account for the solid particle loss, the catalytic VPR transmission efficiencies can be compared with the 70% scaled PMP counting efficiency (eq 7) as shown in Figure 2b. "
f ¼
dp 14 1 exp 3:54 40
1:09 #! 0:7
ð7Þ
where dp (nm) represents the diameter of the particles. The PMP specification is for a 50% efficiency at 23 nm, and >90% at 41 nm. Once the MFMexhaust was above 0.09 g/s, the catalytic VPR transmission efficiencies for different size particles showed similar trends with the scaled PMP counting efficiency. Figure 3 shows the number ratio (post-VPR/pre-VPR) for the total PM, nucleation mode, and accumulation mode. The number reduction ratios of the nucleation mode particles are higher than those of the accumulation mode particles, as their sizes are much smaller than those of the accumulation mode particles. When these results are analyzed in terms of mass (see SI Figure S4), the reduction of total Pm is less than the reduction of total Pn, and the mass ratio of the total particulate matter mass is higher than the number ratio. This indicates that the reduction of PM number in the nucleation mode leads to a more noticeable reduction of total PM number than that of total PM mass. Effects of Dilution Factor on the Catalytic VPR when sampling PM Emissions after a TWC. As explained in Section 2, the purpose of these tests is to see if the catalytic VPR could be used as an oxidation catalyst to reduce the particle number concentration. Only log-normal fitting data are used in the following analysis. The tests with nitrogen as a diluent (as opposed to air) enable the physical and chemical effects to be separated, and in order to separate out the dilution effect of the additional gas, a normalized specific PM emission has been 9039
dx.doi.org/10.1021/es2008209 |Environ. Sci. Technol. 2011, 45, 9036–9043
Environmental Science & Technology
ARTICLE
Figure 4. Normalized specific particle number and mass concentrations with either air or nitrogen addition and Lambdas of 1.0 and 0.9 (refer to Table 1 for the dilution factors) using eqs 8 and 9.
calculated. The normalized specific PM emission is defined as the post-VPR PM concentration times the corresponding VPR dilution factor (VPR-DF) then divided by the pre-VPR PM concentration derived from the log-normal fitting data when no dilution gas is added: Pn, norm ¼ Pn VPR-DF=Pn, VPR-DF ¼ 1
ð8Þ
Pm, norm ¼ Pm VPR-DF=Pm, VPR-DF ¼ 1
ð9Þ
where Pn,norm represents normalized specific particle number emission, Pm,norm represents normalized specific particle mass emission, Pn represents particle number emission (no./cm3), Pm represents particle mass emission (kg/cm3), Pn,VPR-DF=1 represents the particle number emission without dilution gas (no./cm3), and Pm,VPR-DF=1 represents the particle mass emission without dilution gas (kg/cm3). Figure S5 shows how increasing the dilution factor reduces the HC concentration by dilution, and it is for this reason that the PM emissions have been normalized by multiplying with the 9040
dx.doi.org/10.1021/es2008209 |Environ. Sci. Technol. 2011, 45, 9036–9043
Environmental Science & Technology
ARTICLE
Figure 5. Particle numbers in different modes with a VPR dilution factor of 1.2 (refer to Table 1) obtained using log-normal fits to the pre- and postVPR data and showing the effect of the PMP compliant digital filter.
Figure 6. Number transmission efficiencies for different modes (total, nucleation, and accumulation) with either air or nitrogen addition and Lambdas of 1.0 and 0.9 (refer to Table 1 for the dilution factors).
dilution factor. However, when excess air is added it can be seen that hydrocarbon oxidation is very effective, so that any hydrocarbons desorbed from the PM will be oxidized. Using eqs 8 and 9, Figure 4 shows the normalized specific total particle number and mass concentrations calculated from lognormal fitting to the data under various test conditions. Comparing the post-VPR with the pre-VPR particle emissions, there is a consistent reduction in the total number which is independent of the dilution factor, and this is more clearly seen with the rich mixture (Lambda = 0.9). The rich mixture gives a pre-VPR number concentration of typically 2.4 106 no./cm3, and this is about 4 times the concentration at stoichiometric. The trend in the variation of the total mass pre- and post-VPR is less clear. For stoichiometric mixtures there is no change in mass, while for the rich mixture there is a reduction in mass. Since the reduction in mass occurs with dilution by either air or nitrogen, then this is a physical effect, not a chemical effect. This is further analyzed in Figure 5. Data after normal fitting are used in Figure 5, where the data from Lambda = 0.9 with a dilution factor of 1.2 has been selected as a representative data set. The fitted log-normal data have been used, since this eliminates the effects of noise in the measurements. Because the European legislation prescribes the use of a VPR, this has been achieved numerically by using eq 5. Digitally processed measurements from the DMS500 have been compared with measurements made in conformance with the European Method for a range of different engine types (including spray guided gasoline), Braisher et al.,6 and very good agreement was found. The nucleation mode particle number concentration is lower than the accumulation mode particle number concentration because the exhaust sample is after the TWC. As expected from eq 5, the reduction in the nucleation mode particle number
concentration is greater than for the accumulation mode particles. Also, as would be expected from Figure 4 there is no difference between air and nitrogen as diluents. The effect of the numerical VPR (eq 5) is of course greatest for the nucleation mode particles and the difference between pre- and post-catalytic VPR is small after applying the numerical VPR. Further analyzing the total number ratio (post-VPR/pre-VPR) for the total PM, nucleation mode, and accumulation mode (Figure 6), it can be seen that at each measurement point for stoichiometric engine operation, the number ratios are in the following order: nucleation mode < total PM < accumulation mode; this is not always seen for the rich operation cases. Although the oxidation effect of the catalytic VPR has been verified by the HC results (Figure S5), it seems that when the engine was running rich of stoichiometric, the oxidation effects did not play much role in reducing the particle emissions. Apart from chemical reactions, some physical processes were taking place in the catalytic VPR as well. The number transmission efficiencies are shown in the SI (Figure S6). It can be seen that the transmission efficiencies increase as the particle size increases, and the curves for the different engine Lambdas and additional gases are very similar above 10 nm. The engine IMEP was well controlled and this is shown in the Supporting Information (Figure S7).
4. DISCUSSION The role of a catalytic VPR on engine particle emissions has been tested under different operating regimes with a DISI engine. The results can be summarized as follows. (1) With the pre-TWC engine exhaust, nucleation mode particles constitute a large portion of the total particle number but only constitute a small portion of the total 9041
dx.doi.org/10.1021/es2008209 |Environ. Sci. Technol. 2011, 45, 9036–9043
Environmental Science & Technology particle mass. The catalytic VPR led to a significant reduction in the particle number, especially the smaller size (nucleation mode) particles. (2) The VPR temperature and exhaust residence time (MFMexhaust) did not show much effect on the catalytic VPR performance once the MFMexhaust was above 0.09 g/s. The catalytic VPR transmission efficiencies for different size particles showed similar trends for the various VPR temperatures and MFMexhaust tested. Generally, the transmission efficiencies of the VPR follow the trends of the scaled PMP counting efficiency. (3) When post-TWC exhaust was introduced to the catalytic VPR, it showed a moderate reduction effect on the number of particles. With a stoichiometric airfuel ratio, the performance of the catalytic VPR was not affected by the use of additional air. Even with rich mixture combustion, there was still not a noticeable effect with increasing additional air on the catalytic VPR performance. (4) For each airfuel ratio, the transmission efficiency of the catalytic VPR did not change much with different types and varying amounts of additional gas. A comparison between adding air or nitrogen as a diluent showed that the reduction in particle number through the catalytic VPR is more likely to be due to physical than chemical processes.
’ ASSOCIATED CONTENT
bS
Supporting Information. Table S1, Characteristics of the Test Engine; Figure S1, data filter (Wiebe function) described by eq 5; Figure S2, pre- and post-VPR temperatures with different mass flow rate of exhaust (MFMexhaust) and different VPR temperatures; Figure S3, representative size distribution of raw data and its log-normal fitting; Figure S4, mass ratio (postVPR/pre-VPR) for different modes (total, nucleation, and accumulation) for different sample flow rates (MFMexhaust) and different VPR temperatures; Figure S5, HC emissions with either air or nitrogen addition under rich operation; Figure S6, number transmission efficiencies versus dp with various gas additions and relative airfuel ratios; Figure S7, pre- and postVPR temperatures and the IMEP with air or nitrogen addition and relative airfuel ratios of 1.0 and 0.9. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: +44 1865 283467; e-mail:
[email protected].
’ ACKNOWLEDGMENT Jaguar Cars are thanked for the provision of the V8 engine and technical support. ’ REFERENCES (1) Kittelson, D. B. Engines and Nanoparticles: A Review. J. Aerosol Sci. 1998, 29 (5/6), 575–588. (2) Maricq, M. M.; Posdiadlik, D. H.; Chase, R. E. Gasoline Vehicle Particle Size Distribution: Comparison of Steady State, FTP and US06 Measurements. Environ. Sci. Technol. 1999, 33, 2007–2015. (3) Kasper, A.; Burtscher, H.; Johnson, J. P.; Kittelson, D. B.; Watts, W. F.; Baltensperger, U.; Weingartner, E. Particle Emissions from SIEngines During Steady State and Transient Operation Conditions; SAE paper 2005-01-3136.
ARTICLE
(4) Kittelson, D. B.; Watts, W. F.; Johnson, J. P.; Schauer, J. J.; Lawson, D. R. On-road and laboratory evaluation of combustion aerosols. Part 2: Summary of spark ignition engine results. Aerosol Sci. 2006, 37, 931–949. (5) Ericsson, P.; Holmstr€om, M.; Amberntsson-Carlsson, A.; Ohlson, C.; Skoglundh, M.; Andersson, B.; Carlsson, P. A. Characterization of Particulate Emissions and Methodology for Oxidation of Particulates from Non-Diesel Combustion Systems; SAE paper 2008-01-1746. (6) Braisher, M.; Stone, R.; Price, P. Particle Number Emissions from a Range of European Vehicles; SAE paper 2010-01-0786. (7) Geller, M. D.; Ntziachristos, L.; Mamakos, A.; Samaras, Z.; Schmitz, D. A.; Froines, J. R.; Sioutas, C. Physicochemical and redox characteristics of particulate matter (PM) emitted from gasoline and diesel passenger cars. Atmos. Environ. 2006, 40, 6988–7004. (8) Zervas, E.; Dorlhene, P.; Forti, L.; Perrin, C.; Mornique, J. C.; Monier, R.; Ing, H.; Lopez., B. Interlaboratory Study of the Exhaust Gas Particle Number Measurement Using the Condensation Particle Counter (CPC). Energy Fuels 2006, 20, 2426–2431. (9) Lee, H.; Kim, J.; Myung, C. L.; Park, S. Experimental investigation of nanoparticle formation characteristics from advanced gasoline and diesel fueled light duty vehicles under different certification driving modes. J. Mech. Sci. Technol 2009, 23, 1591–1601. (10) Andersson, J.; Giechaskiel, B.; Mu~ noz-Bueno, R.; Sandbach, E.; Dilara, P. Particle Measurement Programme (PMP) Light-duty Interlaboratory Correlation Exercise (ILCE_LD) Final Report; 54th GRPE, 48 June 2007, agenda item 3. GRPE-54-08-Rev.1. (11) Giechaskiel, B.; Dilara, P.; Sandbach, E.; Andersson, J. Particle measurement programme (PMP) light-duty inter-laboratory exercise: Comparison of different particle number measurement systems. Meas. Sci. Technol. 2008, 19, 095401. (12) Mathis, U.; Mohr, M.; Forss, A. M. Comprehensive particle characterization of modern gasoline and diesel passenger cars at low ambient temperatures. Atmos. Environ. 2005, 39, 107–117. (13) Ristim€aki, J.; Keskinen, J.; Virtanen, A.; Maricq, M.; Aakko, P. Cold Temperature PM Emissions Measurement: Method Evaluation and Application to Light Duty Vehicles. Environ. Sci. Technol. 2005, 39, 9424–9430. (14) Ntziachristos, L.; Mamakos, A.; Samaras, Z.; Mathis, U.; Mohr, M.; Thompson, N.; Stradling, R.; Forti, L.; Serves, C. Overview of the European “Particulates” Project on the Characterization of Exhaust Particulate Emissions From Road Vehicles: Results for Light-Duty Vehicles; SAE paper 2004-01-1985. (15) Graskow, B. R.; Kittelson, D. B.; Ahmadi, M. R.; , Morris, J. E. Exhaust Particulate Emissions from a Direct Injection Spark Ignition Engine; SAE paper 1999-01-1145. (16) CONCAWE. Comparison of Particle Emissions from Advanced Vehicles Using DG TREN and PMP Measurement Protocols; Report 2/09; Brussels, 2009. (17) Drake, M. C.; Fansler, T. D.; Solomon, A. S.; Szekely, G. A., Jr. Piston Fuel Films as a Source of Smoke and Hydrocarbon Emissions from a Wall-Controlled Spark-Ignited Direct-Injection Engine; SAE paper 200301-0547. (18) Stojkovic, B. D.; Fansler, T. D.; Drake, M. C.; Sick, V. Highspeed imaging of OH* and soot temperature and concentration in a stratified-charge direct-injection gasoline engine. Proc. Combust. Inst. 2005, 30, 2657–2665. (19) Drake, M. C.; Haworth, D. C. Advanced gasoline engine development using optical diagnostics and numerical modeling. Proc. Combust. Inst. 2007, 31, 99–124. (20) Schwarz, Ch.; Sch€unemann, E.; Durst, B.; Fischer, J.; Witt, A. Potential of the Spray-Guided BMW DI Combustion System; SAE paper 2006-01-1265. (21) Abdul-Khalek, I. S.; Kittelson. D. B. Real Time Measurement of Volatile and Solid Exhaust Particles Using a Catalytic Stripper; SAE paper 950236. (22) Kittelson, D. B.; Watts, W. F.; Savstrom, J. C.; Johnson, J. P. Influence of a catalytic stripper on the response of real time aerosol instruments to diesel exhaust aerosol. J. Aerosol Sci. 2005, 36, 1089– 1107. 9042
dx.doi.org/10.1021/es2008209 |Environ. Sci. Technol. 2011, 45, 9036–9043
Environmental Science & Technology
ARTICLE
(23) ECE/TRANS/WP.29/GRPE/2008/page 62 Appendix 5. Particle Number Emissions Measurement Equipment. (24) Khalek, I. Sampling System for Solid and Volatile Exhaust Particle Size, Number, and Mass Emissions; SAE paper 2007-01-0307. (25) Khalek, I.; Bougher, T. Development of a Solid Exhaust Particle Number Measurement System Using a Catalytic Stripper Technology; SAE paper 2011-01-0635. (26) Khalek, I.; Bougher, T.; Jetter, J. Particle Emissions from a 2009 Gasoline Direct Injection Engine Using Different Commercially Available Fuels; SAE paper 2010-01-2117. (27) Sandford, M.; Page, G.; Crawford, P. The All New AJV8; SAE paper 2009-01-1060. (28) Reavell, K.; Hands, T.; Collings, N. A Fast Response Particulate Spectrometer for Combustion Aerosols; SAE paper 2002-01-2714. (29) Symonds, J.; Reavell, K.; Olfert, J.; Campbell, B.; Swift, S. Diesel soot mass calculation in real-time with a differential mobility spectrometer. Aerosol Sci. 2007, 38, 52–68. (30) Symonds, J.; Price, P.; Williams, P.; Stone, R. Density of particles emitted from a gasoline direct injection engine. 12th ETHConference on Combustion Generated Nanoparticles, June 2008.
9043
dx.doi.org/10.1021/es2008209 |Environ. Sci. Technol. 2011, 45, 9036–9043
ARTICLE pubs.acs.org/est
In-Use Measurement of Activity, Energy Use, and Emissions of a Plug-in Hybrid Electric Vehicle Brandon M. Graver,† H. Christopher Frey,*,† and Hyung-Wook Choi† †
Department of Civil, Construction and Environmental Engineering, North Carolina State University, Campus Box 7908, Raleigh, North Carolina 27695-7908, United States
bS Supporting Information ABSTRACT: Plug-in hybrid electric vehicles (PHEVs) could reduce transportation air emissions and energy use. However, a method is needed for estimating on-road emissions of PHEVs. To develop a framework for quantifying microscale energy use and emissions (EU&E), measurements were conducted on a Toyota Prius retrofitted with a plug-in battery system on eight routes. Measurements were made using the following: (1) a data logger for the hybrid control system; (2) a portable emissions measurement system; and (3) a global positioning system with barometric altimeter. Trends in EU&E are estimated based on vehicle specific power. Energy economy is quantified based on gasoline consumed by the engine and grid energy consumed by the plug-in battery. Emissions from electricity consumption are estimated based on the power generation mix. Fuel use is approximately 30% lower during plug-in battery use. Grid emissions were higher for CO2, NOx, SO2, and PM compared to tailpipe emissions but lower for CO and hydrocarbons. EU&E depends on engine and plug-in battery operation. The use of two energy sources must be addressed in characterizing fuel economy; overall energy economy is 11% lower if including grid energy use than accounting only for fuel consumption.
’ INTRODUCTION In 2008, there were 250 million personal vehicles in the U.S., of which 54% were passenger cars and 40% were sport utility vehicles and light trucks. On average, each vehicle traveled 44.6 km daily.1 Transportation accounts for 28% of U.S. energy use and 33% of national carbon dioxide (CO2) emissions. Two-thirds of U.S. transportation fuel consumption is from imports.2 Highway vehicles contribute 40% of national annual nitrogen oxides (NOx), 56% of carbon monoxide (CO), and 28% of volatile organic compounds (VOCs) emissions.3 Plugin hybrid electric vehicles (PHEVs) have the potential for higher energy efficiency and lower emissions compared to conventional vehicles.4,5 Hybrid electric vehicles (HEVs) are powered by an internal combustion engine (ICE) and electric motors. The motors provide propulsion in combination with or instead of the ICE. The motors can operate as generators, converting shaft output from the ICE or powertrain during braking to recharge the battery.6 In a parallel hybrid, both the ICE and the motors are connected mechanically to the transmission. With series hybrids, only the motor is connected to the transmission, and the ICE generates power for the motor. A mixed hybrid can act either in parallel or series. The first production PHEV, the Chevrolet Volt, is now available.7 PHEV retrofits have been available for the Toyota Prius. The retrofit includes a plug-in battery (PIB) that can only be recharged from the electric grid. The original traction battery r 2011 American Chemical Society
(TB) is recharged only during vehicle operation. Charge Depleting (CD) mode occurs until the PIB reaches a state-of-charge (SOC) set point, after which the PIB is no longer used until later recharged. During subsequent Charge Sustaining (CS) mode, the PIB is not used. During both CD and CS modes, the TB is discharged and recharged depending on operating conditions.8 The TB SOC is maintained between lower and upper set points to prolong battery life.6 Many PHEV studies focus on fuel economy, with less attention to emissions. Some evaluations of PHEV energy use and emissions (EU&E) are based on dynamometer tests that do not represent real-world driving.5,9 Real-world driving is a preferred basis for estimating fuel economy.5,10 Global positioning systems (GPS) have been used to obtain real-world PHEV activity data but not EU&E data.11 Selected PHEV designs have been assessed based on hypothetical travel distances and surveys of driver habits based on conventional light-duty gasoline vehicles (LDGVs).4,8 Idaho National Laboratory (INL) is measuring real-world energy consumption of 184 retrofitted PHEVs.12 CO2 emissions are proportional to fuel consumption and can be inferred from fuel economy.8,13 Fuel use, CO2, CO, HC, and NOx emissions were estimated by Silva et al. for two PHEV Received: April 6, 2011 Accepted: September 8, 2011 Revised: August 28, 2011 Published: September 08, 2011 9044
dx.doi.org/10.1021/es201165d | Environ. Sci. Technol. 2011, 45, 9044–9051
Environmental Science & Technology
ARTICLE
Table 1. Characteristics of Test Routes and Field Study Results field study results
route characteristics
route segmenta
a
route distance
route speed
# signalized
(km)
limits (kph)
intersections
# of runs
total travel
total travel distance
average vehicle
range of driven
time (s)
(km)
speed (kph)
speeds (kph)
A out
16
2472
41
6
8,524
97
42
0100
A in
16
2472
41
6
7,950
93
42
0100
B out
18
2497
32
6
8,579
108
45
0116
B in
18
2497
32
6
9,215
105
40
0116
C out
18
2497
29
6
8,190
106
47
0129
C in
18
2497
29
6
7,667
105
48
0129
D E
16 3
24105 2472
18 0
6 6
6,050 2,482
95 23
56 32
0121 085
1 out
26
40105
8
6
7,488
156
74
0130
1 in
26
40105
8
6
7,385
155
76
0130
2 out
32
40105
17
6
9,894
192
69
0130
2 in
34
40105
17
6
10,406
198
69
0130
3 out
27
4081
23
6
10,414
163
56
0109
3 in
27
4081
23
6
11,340
166
53
0109
Outbound or inbound leg of a route.
configurations using Advanced Vehicle Simulator (ADVISOR) software.14 A life cycle analysis of greenhouse gas emissions from PHEVs concluded that potential decreases in emissions depend on lifetime performance of the PIB and the electricity mix.15 Inuse measurement of activity and EU&E has been completed on a PHEV school bus.16 However, PHEV EU&E varies by vehicle size and weight. Currently, there is no fuel economy reporting standard for vehicles that run on two energy sources. The Society of Automotive Engineers recommends “off-vehicle-charge equivalent fuel use” using an electricity-to-gasoline equivalent of 8835 W-h/L.17 However, this value unrealistically assumes 100% power plant and transmission efficiency. Argonne National Laboratory and the National Renewable Energy Laboratory (NREL) recommend reporting fuel use in kilometers per liter (km/L) and electricity usage in Wh/km, separately.9,18 Objectives here are to (1) demonstrate an approach for quantifying PHEV real-world energy use and tailpipe emissions; (2) assess geographic variability in CD mode grid emissions; and (3) propose a method for quantifying PHEV energy economy.
’ METHODS A portable emissions measurement system (PEMS) is used to quantify the activity and EU&E of a PHEV during actual driving. Electricity grid emissions are quantified based on regional energy mix data and emission factors. Field Study Design. The test vehicle was a 2005 Toyota Prius HEV retrofitted with an A123 Hymotion PIB system. The vehicle was operated on regular unleaded gasoline, with a maximum 10% ethanol content.19 The fuel tank was filled, and the PIB was recharged prior to each day of measurements. A watt-hour meter was used to measure grid electricity consumed during recharge. The PHEV was operated on eight selected routes, described in Table 1. Six routes were used in previous data collection with LDGVs.20 Routes are based on two origin and destination (O/D) pairs: (1) North Carolina State University (NCSU) to
North Raleigh (NR) and (2) NR to Research Triangle Park (RTP). Routes A, B, and C from NCSU to NR are primarily signalized minor and major arterials and represent local commuting within Raleigh. Routes 1, 2, and 3 from NR to RTP are primarily major arterials and freeways representing commuting to the RTP business district. Route D, comprised of major arterials and freeways, represents a typical lunchtime trip. Route E is comprised of minor arterials and represents driving within NCSU. Instruments. Instruments used for field data collection include the following: (1) electronic control unit (ECU) data logger; (2) PEMS; and (3) GPS with barometric altimeter. Externally observable variables (EOVs) and internally observable variables (IOVs) were measured. EOVs can be observed from outside of a vehicle, such as vehicle speed, vehicle acceleration, and road grade. IOVs can be observed within a vehicle, such as engine RPM and electrical current to or from a battery. EOVs could be used to develop modeling approaches for use in traffic simulation and emission factor models. The Kvaser Memorator ECU data logger recorded vehicle speed, voltage, and current to or from the PIB and TB, battery SOC, engine RPM, and fuel flow rate. The OEM-2100 Montana system PEMS, manufactured by Clean Air Technologies International, Inc., includes two parallel, five-gas analyzers.21 PEMS precision, accuracy, and calibration are detailed elsewhere.20,22,23 Exhaust concentrations measured by the PEMS and fuel flow rate from the ECU were used to estimate the mass per time emission rate of each pollutant.23 A Garmin GPSmap 76CSx with barometric altimeter recorded second-by-second location and elevation, from which road grade was estimated. Vehicle Specific Power. Vehicle specific power (VSP) accounts for vehicle kinetic energy, acceleration due to gravity, rolling resistance, and aerodynamic drag. For a typical LDGV, VSP is25 VSP ¼ 0:278v½0:305a þ 9:81r þ 0:132 þ 0:0000065v3
ð1Þ 9045
dx.doi.org/10.1021/es201165d |Environ. Sci. Technol. 2011, 45, 9044–9051
Environmental Science & Technology
ARTICLE
Table 2. Route Average Energy Use and Energy Economy Based on Fuel Use for Engine and Grid Electricity Use for Plug-in Hybrid System grid electricity use
energy use
a
total dist
gas and
gas and
engine starts
grid (km/L)
per km
route
mode
(km)
totalb (kWh)
rate (kWh/km)
gas (L)
grid (Leq)
A
CD
79
6.7
0.06
2.20
2.27
4.47
36
18
1.3
B
CS CD
119 60
--5.4
--0.09
6.47 1.74
--1.82
6.47 3.56
18 34
--17
2.8 1.3
CS
175
---
---
9.24
---
9.24
19
---
2.0
CD
24
2.0
0.08
0.87
0.68
1.55
28
15
2.2
CS
211
---
---
11.2
---
11.2
19
---
1.5
CD
19
1.4
0.07
0.87
0.45
1.32
22
14
0.2
CS
81
---
---
4.77
---
4.77
17
---
1.2
E
CD
5
0.4
0.08
0.19
0.11
0.30
26
17
2.8
1
CS CD
19 35
--2.4
--0.07
1.17 1.40
--0.79
1.17 2.19
16 25
--16
3.4 0.3
CS
282
---
---
16.6
---
16.6
17
---
0.4
2
CD
43
2.5
0.06
1.63
0.83
2.46
26
17
1.1
CS
356
---
---
19.6
---
19.6
18
---
0.7
CD
43
2.3
0.05
1.74
0.76
2.50
25
17
0.6
CS
295
---
---
16.1
---
16.1
18
---
1.1
C D
3 a
energy economy
grid (Leq)
gas (km/L)
There are 6 repeated roundtrip runs on each route. b These results include data for more than one day in the case of Routes A, B, and C.
where VSP = vehicle specific power (kW/ton); v = vehicle speed (km/h); a = vehicle acceleration (km/h per s); and r = road grade (%). Here, VSP accounts for the energy demand of a PHEV.16 For LDGVs, fuel use is typically constant for negative VSP and increases linearly with positive VSP.20 For a PHEV, the batterymotor system will sometimes augment engine output and at times consume some engine output for recharging. Direct and Indirect Energy Use. For validation, daily ECU recorded fuel use was compared to the amount of gasoline used to refill the tank. For both batteries, second-by-second electricity usage associated with either discharge or recharge, in Watt-hours, was recorded. PIB electricity use was compared to the amount of grid electricity used to recharge the PIB after each day of testing. The grid electricity needed to recharge the PIB was used to estimate the corresponding gasoline-equivalent amount of fuel consumed by power plants. The national average power plant heat rate is 10.7 106 J of thermal energy input per kWh generated.2 Gasoline has a heating value of 32.2 106 J/L and density of 737 g/L.26 Thus, each kWh of electricity corresponds to 0.33 L of gasoline equivalent. The resulting equivalent factor of 3.01 kWh/L is approximately one-third of the SAE equivalent factor of 8.84 kWh/L, since it takes into account the inefficiency associated with producing electricity. Regional indirect power generation energy use is estimated based on North American Electric Reliability Corporation (NERC) regions. Data Quality Assurance. The key steps to process ECU, PEMS, and GPS data are as follows: (1) convert data to a second-by-second basis, if necessary; (2) synchronize the data; and (3) screen the data for possible errors, correct errors where possible, and remove invalid data that cannot be corrected. The data processing and quality assurance procedures are detailed elsewhere.16,20,22,23
’ RESULTS Results are given for quality assurance, gasoline and grid electricity use, tailpipe and electric grid emission factors, and gasoline and electricity consumption as a function of VSP. Data Collection and Quality Assurance. Two drivers collected 121,908 s of data over seven days in November 2008. For six days, the vehicle initially operated in CD mode. Only 6.4% of the data were excluded because of data quality issues. Energy Use for Gasoline and Grid Electricity. The amount of grid electricity and gasoline used on each route is given in Table 2. On average, 3.8 kWh of total PIB discharge was used to travel 53.1 km each day that CD mode was observed. A majority of vehicles tested in the INL study had a CD range of between 40.3 and 56.4 km.28 PIB electricity usage rates during CD mode varied from 0.05 to 0.09 kWh/km. In general, routes that contained higher speed interstate and major arterial driving had lower PIB electricity usage rates. ECU-reported daily average fuel use agreed with the actual gasoline consumption to within 3%. Gasoline fuel economy was 28 to 95% higher during CD versus CS mode. Gasoline fuel economies were, on average, 33% higher for routes with less interstate driving and more engine starts per kilometer, such as Routes A and B, than those with fewer engine starts per kilometer, such as Routes 1 and 2. The gasoline and grid equivalent energy economy averaged 42% lower than the gasoline fuel economy for CD mode and was similar to CS mode gasoline fuel economy. The 2005 Toyota Prius HEV has a rated combined highway and city fuel economy of 19.6 km/L.29 The observed CS mode fuel economy was 1.3 km/L worse than the rating perhaps in part because of the additional 85 kg weight of the PIB.30 CD mode fuel economy was 3.0 and 3.8 km/L higher than from NREL’s dynamometer testing and INL’s fleet testing, respectively.5,27 Average CS mode fuel economy was the same as both dynamometer and fleet testing at 18.3 km/L. The overall 9046
dx.doi.org/10.1021/es201165d |Environ. Sci. Technol. 2011, 45, 9044–9051
Environmental Science & Technology
ARTICLE
Table 3. Measured Tailpipe and Estimated Southeast Reliability Corporation (SERC) Region Grid-Based Emission Factorsa,b route
mode
A
CD
CS B
C
D
E
1
2
3
CD
emissions source
CO2 (g/km)
CO (mg/km)
HC (mg/km)
NOx (mg/km)
SO2 (mg/km)
tailpipe
83
119
14
13
<1
grid
72
11
1
102
331
total
155
130
15
115
331
tailpipe
151
106
54
27
<1
tailpipe
85
134
20
12
<1
grid total
77 162
12 146
1 21
109 121
354 354
CS
tailpipe
134
68
35
30
<1
CD
tailpipe
103
202
14
21
<1
grid
71
11
1
99
323
total
174
213
15
120
323
CS
tailpipe
126
38
56
25
<1
CD
tailpipe
134
395
48
9
<1
grid
62
10
1
87
282
total
196
405
49
96
282
CS
tailpipe
163
37
12
32
<1
CD
tailpipe
105
37
<1
2
<1
grid
71
11
1
99
323
total
176
48
1
101
323
CS
tailpipe
148
76
142
7
<1
CD
tailpipe
118
<1
1
16
<1
grid total
58 176
9 9
1 2
81 97
264 264
CS
tailpipe
171
21
20
56
<1
CD
tailpipe
107
12
8
25
<1
grid
49
8
1
69
224
total
156
20
9
94
224
CS
tailpipe
160
52
23
47
<1
CD
tailpipe
114
<1
3
36
<1
grid
45
7
1
64
206
total
159
7
4
100
206
tailpipe
156
37
44
45
<1
CS a
Emissions for electricity generation in SERC are based on 2007 EPA eGRID data for CO2, NOx, and SO2 and 2005 NEI data for CO and HC and account for a 7% transmission loss between the power plant and vehicle charging station and a 28% PIB charging loss. The average electric utility emission factors are 668 g CO2/kWh, 939 mg NOx/kWh, 3.05 g SO2/kWh, 103 mg CO/kWh, and 10 mg HC/kWh. Tailpipe SO2 emissions are based on e30 ppm sulfur in gasoline. b SERC electricity generation mix is 57.1% coal, 24.2% nuclear, 11.7% gas, and 7.0% other sources.
CD and CS mode gasoline fuel economy of 20.0 km/L was similar to that reported by NREL and INL at 20.8 and 20.4 km/L,
respectively.5,27 Overall CD and CS mode equivalent fuel economy, including grid-based electricity, was 42 mpg. 9047
dx.doi.org/10.1021/es201165d |Environ. Sci. Technol. 2011, 45, 9044–9051
Environmental Science & Technology
ARTICLE
Figure 1. Vehicle fuel use versus vehicle specific power.
Direct and Indirect Emissions. Average emission factors for each route are shown in Table 3. Emissions associated with electricity usage are estimated from the electricity generation source mix in the Southeast Electricity Reliability Corporation (SERC) region, which includes North Carolina. CO 2 and NOx emissions from electricity generation, in mass per kWh, are calculated based on U.S. Environmental Protection Agency (EPA) data.31 Grid-based CO and HC emission rates for each fossil fuel were calculated based on EPA National Emissions Inventory data and U.S. Energy Information Administration (EIA) Annual Energy Outlook data32,33 to account for regional variability in emission rates, not national averages given by EPA data. Emission factors are based on a national average electricity transmission loss of 7% and a PIB charging loss of 28%.34 The PEMS does not measure SO2 emissions. However, using the maximum allowable gasoline sulfur concentration of 30 ppm,35 an upper bound of the SO2 emission rate is 2.5 mg/L. In certification tests, the 2005 Toyota Prius had emission rates of 0.0625 mg NOx/km, 5.625 mg HC/km, and 62.5 mg CO/km.36 The EPA Tier 2 Bin 3 emission standard, which this Prius must meet, is 18.6 mg NOx/km, 34.2 mg HC/km, and 1.3 g CO/mile.37 Average measured tailpipe emission rates for HC were approximately 2 to 20 times higher on all routes than the certification data but below the EPA emission standard. Average measured tailpipe emission rates for CO were approximately 4 to 73% lower on most routes than the certification data. Average measured tailpipe NOx emission rates were 3 to 137% higher than the emission standard on all routes except for Route E.
Figure 2. Plug-in battery electricity use versus vehicle specific power during Charge Depleting mode.
Tailpipe CO2 emission rates were 18 to 45% lower during CD versus CS mode for all routes because of decreased fuel use. However, total CO2 emission rates during CD mode were higher than tailpipe emission rates during CS mode for all routes except Route 2. Grid CO2 emission rates were 28 to 48% of total CO2 emission rates in CD mode on all routes. CD mode tailpipe NOx emission rates were lower by 16 to 73% compared to CS mode. Total NOx emission rates during CD mode are higher than tailpipe emission rates during CS mode for all routes. NOx emission rates from electricity use were nearly 3 to 43 times higher than CD mode tailpipe emission rates on all routes. Tailpipe CO emission rates were 13 to 980% higher in CD mode for Routes A through D than in CS mode. Tailpipe HC emission rates were lower in CD versus CS mode on the same routes except for Route D. On these routes, lower speed driving with an increased number of signalized intersections leads to a 43% increase in the number of engine starts. Even though the relative difference in emission rates of CO and HC are large, the absolute differences are small. Grid-related emission rates vary with electricity generation resource mix. For example, the Northeast Power Coordinating Council (NPCC) region, which has the lowest average 9048
dx.doi.org/10.1021/es201165d |Environ. Sci. Technol. 2011, 45, 9044–9051
Environmental Science & Technology
ARTICLE
Figure 4. Tailpipe NOx emissions (mg/s) versus VSP (kW/ton) for Engine On.
Figure 3. Traction battery electricity use versus vehicle specific power during Charge Depleting mode.
CO2 and NOx emissions rates for electricity generation, has 36 and 55% lower grid CO2 and NOx emission rates, respectively, compared to SERC.31 If the PHEV was operated in the Northeast, total CO2 emission rates during CD mode would be similar to or lower than tailpipe emission rates during CS mode for three routes (A, 1, and 2). Total NOx emission rates during CD mode would be higher than tailpipe emission rates during CS mode for all routes. SO2 emission rates from the grid are substantially higher in all regions than from gasoline combustion because the sulfur concentration of gasoline is very low. The highest grid-related CO2 emission rates are in the Midwest Reliability Organization (MRO) region, which are 33% higher than in SERC. NOx emission rates are 86% higher in MRO compared to SERC, while SO2 emission rates are 6% lower. Total CO2 emission rates in CD mode would be 18 to 100% higher than tailpipe emission rates in CS mode, while total NOx emission rates in CD mode would be 4 to 36 times greater than tailpipe emission rates in CS mode. In the SERC region, grid CO emission rates were lower than tailpipe emission rates for local driving on Routes A through E, and grid HC emission rates were lower than or similar to tailpipe emission rates on all routes except for Route 2.
Energy Use and Emissions versus Vehicle Specific Power. For positive VSP, fuel use rate increases monotonically as shown in Figure 1 but not linearly, as is the case for LDGVs. Fuel use is sensitive to VSP at negative values, which includes deceleration and travel on a negative road grade. During CD mode, the fuel use rate for VSP greater than 10 kW/ton is typically 0.2 to 0.4 g/s less than during CS mode. Since power from the PIB is used during CD mode, less engine power is needed to meet the total power demand, and, therefore, less fuel is used. Figure 2(a),(b) shows the CD mode PIB electricity consumption versus VSP for engine off and on, respectively. The engine is considered on when vehicle fuel use is greater than 0 L/s. The PIB discharges for both negative and positive VSP. As VSP increases, the discharge rate increases. For VSP values between 8 and 20 kW/ton, the discharge rate levels off, signifying a maximum discharge rate. When the ICE is off, average PIB discharge is approximately 0 to 0.7 Wh/s higher than when the ICE is on for VSP greater than 5 kW/ton. However, for negative VSP, the PIB discharge rate is higher when the ICE is on. This is due to a lower TB discharge rate when the ICE is on. Electricity consumption of the TB versus VSP for engine on and off during CD mode is given in Figure 3. Results for CS mode are qualitatively similar. In all cases, the TB, on average, recharges during negative VSP and discharges during positive VSP. For both CD and CS modes, recharge and discharge of the TB is greater when the ICE is off. When the ICE is off, recharge and discharge rates are similar for CS and CD modes. TB discharge rates level off for high values of VSP at approximately 2 Wh/s, 9049
dx.doi.org/10.1021/es201165d |Environ. Sci. Technol. 2011, 45, 9044–9051
Environmental Science & Technology signifying a maximum discharge rate. When the ICE is on, discharge rates for CD and CS modes are very small and similar for positive VSP. NOx emission rates are plotted versus VSP for CS mode in Figure 4. NOx emission rates are approximately constant for negative VSP and increase as positive VSP increases. Locations that experience high vehicle speeds and acceleration, such as major arterials and interstates, would observe higher NOx emissions. This is consistent with Table 3, with high tailpipe NOx emission factors observed on Routes 1, 2, and 3.
’ DISCUSSION Energy use differs depending on whether the ICE is on or off and whether the PHEV is in CD or CS mode. Fuel use is approximately 30% lower for positive VSP during CD versus CS mode due to shift of some vehicle power from the ICE to the PIB. The TB recharges and discharges at greater rates when the engine is off. On average, 0.12 kWh of grid energy was used per mile driven during CD mode, comparable to the 0.13 kWh observed during by INL.12 Tailpipe emission rates vary depending on whether the PHEV is in CD or CS mode. Tailpipe emission rates of CO2 and NOx are, on average, 25 and 60% lower, respectively, for CD versus CS mode. Comparisons of total emission rates during CD mode are sensitive to generation mix. Current energy mixes in the selected NERC regions lead to CD mode tailpipe and grid emission rates for CO2 and NOx to be higher than tailpipe emission rates in CS mode for all routes, with a few exceptions. Therefore, the percentage of electricity generated from noncarbon sources need to increase from current levels for total emissions in CD mode to be lower than in CS mode. Both more efficient powertrains and the development of battery charging stations powered by alternative energy could also make PHEVs a viable option for reducing CO2, NOx, and SO2 emissions from transport. The potential future use of Carbon Capture and Sequestration (CCS) for coal-fired power plants is estimated to decrease CO2 and SO2 emission rates by as much as 83 and 97%, respectively, but increase NOx emission rates by approximately 20%. The environmental performance of PHEVs in combination with CCS would thus improve for CO2 and SO2 but worsen for NOx.38 On average, the observed CD range of 53.1 km is greater than the typical average daily driving. It could be possible to make an entire daily commute without consuming any gasoline. This work is based on measurements on only one PHEV. More testing should be completed on additional retrofitted PHEVs as well as production PHEVs. PHEVs with different powertrain configurations and PIB retrofits may result in different EU&E rates. Derived PHEV tailpipe emission rates could be integrated with transportation models to estimate emissions on various temporal scales for a PHEV fleet. One example could be to identify emission hotspots on a road network. Future work is needed to incorporate the PHEV emissions data with such transportation models. Frequency and duration of engine use is sensitive to vehicle operations. Thus, PHEV emissions will not occur uniformly over the transportation network. This episodic nature motivates the need for a microscale model that enables prediction of locations or conditions where PHEV emissions will occur. For example,
ARTICLE
locations for microscale variability in tailpipe NOx emissions include signalized intersections and stop signs where the ICE restarts due to deceleration then acceleration. This is particularly evident during CD mode, where ICE demand is more sporadic. Future work is needed to explore whether excess EU&E rates occur immediately after an ICE restart. The use of two energy sources for a vehicle must be addressed in characterizing fuel economy. Energy economy should be estimated, taking into account the gasoline equivalent of the energy required to generate grid electricity consumed by the vehicle.
’ ASSOCIATED CONTENT
bS
Supporting Information. Turn-by-turn directions for test routes; description of PEMS precision, accuracy, and calibration; test schedule; comparison of OBD versus actual fuel use; comparison of PIB electricity use reported by the OBD and amount reported by the Watt-hour meter; average emission factors from electricity generation for selected NERC regions; and graphs of tailpipe emissions versus VSP. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: (919) 515-1155. Fax: (919) 515-7908. E-mail: frey@ ncsu.edu.
’ ACKNOWLEDGMENT This material is based upon work supported by the National Science Foundation under Grant No. CBET-0756263 and CBET-0853766. Additional financial support was provided by the NSF FREEDM Systems Center and the Southeastern Transportation Center. Advanced Energy, Inc. provided access to their retrofitted Toyota Prius PHEV and ECU/OBD data logger. Ewan Pritchard and Siddharth Ballal of the Advanced Transportation Energy Center and Gurdas Sandhu of North Carolina State University assisted in data collection and processing. Dr. Choi participated in this work as a postdoctoral research fellow at NC State. He is now with the Greenhouse Gas Inventory & Research Center of Korea in Seoul. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. ’ REFERENCES (1) Transportation Energy Data Book: ed. 29; U.S. Department of Energy, Oak Ridge National Laboratory: Oak Ridge, TN, 2010. (2) Annual Energy Outlook 2010 With Projections to 2035; DOE/EIA0383(2010); U.S. Department of Energy, Energy Information Administration: Washington, DC, 2010. (3) EPA. 19702008 Average annual emissions, all criteria pollutants, Current Emissions Trends Summaries from the National Emission Inventory; U.S. Environmental Protection Agency: Research Triangle Park, NC, 2008. http://www.epa.gov/ttn/chief/net/2008inventory.html (accessed October 18, 2010). (4) Markel, T.; Simpson, A. Cost-Benefit Analysis of Plug-In Hybrid Electric Vehicle Technology; NREL/JA-540-40969; U.S. Department of Energy, National Renewable Energy Laboratory: Washington, DC, 2006. 9050
dx.doi.org/10.1021/es201165d |Environ. Sci. Technol. 2011, 45, 9044–9051
Environmental Science & Technology (5) Gonder, J.; Brooker, A. Deriving In-Use PHEV Fuel Economy Predictions from Standardized Test Cycle Results; NREL/CP-540-46251; U.S. Department of Energy, National Renewable Energy Laboratory: Washington, DC, 2009. (6) Toyota. Toyota Hybrid System Technical Training; Toyota Motor Engineering and Manufacturing North America, Inc.: Torrance, CA, 2008. (7) Hirsch, J. Want the first Chevrolet Volt? Bid for it. Los Angeles Times, Dec 1, 2010, p B2. (8) Bradley, T. H.; Frank, A. A. Design, demonstrations and sustainability impact assessments for plug-in hybrid electric vehicles. Renewable Sustainable Energy Rev. 2009, 13 (1), 104-117; DOI 10.1016/ j.rser.2007.05.003. (9) Karner, D.; Francfort, J. Hybrid and plug-in hybrid electric vehicle performance testing by the US Department of Energy Advanced Vehicle Testing Activity. J. Power Sources 2007, 174 (1), 69-75; DOI 10.1016/j.jpowsour.2007.06.069. (10) Gonder, J.; Simpson, A. Measuring and Reporting Fuel Economy of Plug-In Hybrid Electric Vehicles; NREL/JA-540-41341; U.S. Department of Energy, National Renewable Energy Laboratory: Washington, DC, 2007. (11) Gonder, J.; Markel, T.; Thornton, M.; Simpson, A. Using global positioning system travel data to assess real-world energy use of plug-in hybrid electric vehicles. Transp. Res. Rec. 2007, 2017, 26-32; DOI 10.3141/2017-04. (12) INL. North American PHEV Demonstration Fleet Summary Report: Hymotion Prius (V2Green data logger); U.S. Deparment of Energy, Idaho National Laboratory: Idaho Falls, ID, 2010. (13) Stephan, C. H.; Sullivan, J. Environmental and Energy Implications of Plug-In Hybrid-Electric Vehicles. Environ. Sci. Technol. 2008, 42 (4), 1185-1190; DOI 10.1021/es062314d. (14) Silva, C.; Ross, M.; Farias, T. Evaluation of energy consumption, emissions and cost of plug-in hybrid vehicles. Energy Conserv. Manage. 2009, 50 (7) 1635-1643; DOI 10.1016/j.econman.2009.03.036. (15) Samaras, C.; Meisterling, K. Life Cycle Assessment of Greenhouse Gas Emissions from Plug-in Hybrid Vehicles: Implications for Policy. Environ. Sci. Technol. 2008, 42(9), 3170-3176; DOI 10.1021/ es702178s. (16) Choi, H. W.; Frey, H. C. Method for In-Use Measurement and Evaluation of the Activity, Fuel Use, Electricity Use, and Emissions of a Plug-In Hybrid Diesel-Electric School Bus. Environ. Sci. Technol. 2010, 44 (9), 3601-3607; DOI 10.1021/es90330k. (17) Recommended Practice for Measuring the Exhaust Emissions and Fuel Economy of Hybrid-Electric Vehicles; SAE J1711; Society of Automotive Engineers, Inc.: Warrendale, PA, 1999. (18) Duoba, M.; Bocci, D.; Bohn, T.; Carlson, R.; Jehlik, F.; Lohse-Busch, H. Argonne Facilitation of PHEV Standard Testing Procedure (SAE J1711); Presented at the 2009 U.S. Department of Energy Hydrogen and Fuel Cells Program and Vehicle Technologies Program Annual Merit Review, Washington, DC, 2009. (19) The Public Health and Welfare. U.S. Code, Section 7545, Title 42, 2008. (20) Frey, H. C.; Zhang, K.; Rouphail, N. M. Fuel Use and Emissions Comparisons for Alternative Routes, Time of Day, Road Grade, and Vehicles Based on In-Use Measurements. Environ. Sci. Technol. 2008, 42 (7), 2483-2489; DOI 10.1021/es702493v (21) CATI. OEM-2100 Montana System Operation Manual; Clean Air Technologies International, Inc.: Buffalo, NY, 2003. (22) Frey, H. C.; Unal, A.; Rouphail, N. M.; Colyar, J. D. On-Road Measurement of Vehicle Tailpipe Emissions Using a Portable Instrument. J. Air Waste Manage. Assoc. 2003, 53 (8), 992–1002. (23) Frey, H. C.; Rasdorf, W. J.; Kim, K.; Pang, S.; Lewis, P.; Ablhasani, S. Real-World Duty Cycles and Utilization for Construction Equipment in North Carolina; HWY-2006-08; Prepared by North Carolina State University for North Carolina Department of Transportation: Raleigh, NC, 2008. (24) Vojtisek-Lom, M.; Allsop, J. E. Development of Heavy-Duty Diesel Portable, On-Board Mass Exhaust Emissions Monitoring System with NOx, CO2, and Qualitative PM Capabilities; 2001-01-3641; Society of Automotive Engineers: Warrenton, PA, 2007.
ARTICLE
(25) Jimenez-Palacios, J. L. Understanding and quantifying motor vehicle emissions with vehicle specific power and TILDAS remote sensing. Ph.D. Thesis, Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA, 1999. (26) Inventory of U.S. Greenhouse Gas Emissions and Sinks: Fast Facts 19902005. Conversion Factors to Energy Units (Heat Equivalents) Heat Contents and Carbon Content Coefficients for Various Fuel Types; EPA4304-07-002; United States Environmental Protection Agency: Washington, DC, 2007. (27) INL. Hymotion Prius PHEV Fleet Testing (Kvaser data logger) Summary Results to date (6/10); U.S. Department of Energy, Idaho National Laboratory: Idaho Falls, ID, 2010. (28) Charging and Driving Behavior Report for Hymotion Prius (Gridpoint data logger); U.S. Department of Energy, Idaho National Laboratory: Idaho Falls, ID, 2010. (29) Certificate Summary Information for TOYOTA 2005 model year test group 5TYXV01.5MC1 evaporative family 5TYXR0030P41. http://iaspub.epa.gov/otaqpub/display_file.jsp?docid=16163&flag=1 (accessed December 11, 2010). (30) A123 Systems. Hymotion L5 Plug-In Conversion Module. http:// www.a123systems.com/hymotion/pdf/L5_SpecSheet.pdf (accessed October 18, 2010). (31) E.H. Pechan & Associates, Inc. The Emissions & Generation Resource Integration Database for 2007 (eGRID2007) Technical Support Document; EP-D-06-001; Prepared by E.H. Pechan & Associates, Inc. for U.S. Environmental Protection Agency: Washington, DC, 2008. (32) National Emissions Inventory (NEI) Air Pollutant Emissions Trends Data.http://www.epa.gov/ttn/chief/trends/index.html (accessed October 18, 2010). (33) Annual Energy Outlook 2007 With Projections to 2030; DOE/ EIA-0383(2007); U.S. Department of Energy, Energy Information Administration: Washington, DC, 2007. (34) State Electricity Profiles 2009; DOE/EIA-0348(01)/2; U.S. Department of Energy, Energy Information Administration: Washington, DC, 2011. (35) U.S. Environmental Protection Agency. Control of Air Pollution From New Motor Vehicles: Tier 2 Motor Vehicle Emissions Standards and Gasoline Sulfur Control Requirements; Final Rule. Fed Regist. 2000, 65, 6698-6822. (36) Compare Old and New EPA MPG Estimates. http://www. fueleconomy.gov/feg/calculatorSelectEngine.jsp?year=2005&make= Toyota&model=Prius (accessed October 18, 2010). (37) Control of Emissions from New and In-Use Highway Vehicles and Engines. Code of Federal Regulations, Part 86, Title 40, 1999. (38) Cost and Performance Baseline for Fossil Energy Plants Vol. 1: Bituminous Coal and Natural Gas to Electricity, Revision 2; DOE/NETL2010/1397; U.S. Department of Energy, National Energy Technology Laboratory: Pittsburgh, PA, 2010.
9051
dx.doi.org/10.1021/es201165d |Environ. Sci. Technol. 2011, 45, 9044–9051
ARTICLE pubs.acs.org/est
Impact of Fuel Quality Regulation and Speed Reductions on Shipping Emissions: Implications for Climate and Air Quality Daniel A. Lack,†,‡,* Christopher D. Cappa,§ Justin Langridge,†,‡ Roya Bahreini,†,‡ Gina Buffaloe,§ Charles Brock,† Kate Cerully,|| Derek Coffman,^ Katherine Hayden,# John Holloway,† Brian Lerner,†,‡ Paola Massoli,3 Shao-Meng Li,# Robert McLaren,O Ann M. Middlebrook,† Richard Moore,|| Athanasios Nenes,||,[ Ibraheem Nuaaman,#,O Timothy B. Onasch,3 Jeff Peischl,†,‡ Anne Perring,†,‡ Patricia K. Quinn,^ Tom Ryerson,† Joshua P. Schwartz,†,‡ Ryan Spackman,†,‡ Steven C. Wofsy,z Doug Worsnop,3 Bin Xiang,z and Eric Williams†,‡ †
NOAA Earth System Research Laboratory, Boulder, Colorado, United States University of Colorado, CIRES, Boulder, Colorado, United States § Department of Civil and Environmental Engineering, University of California, Davis, California, United States School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, Atlanta, Georgia, United States ^ NOAA Pacific Marine Environment Laboratory, Seattle, Washington, United States # Air Quality Research Division, Environment Canada, 4905 Dufferin St., Toronto, Canada 3 Aerodyne Research Inc., Billerica, Massachusetts, United States O Centre for Atmospheric Chemistry, York University, 4700 Keele St., Toronto, Canada [ Earth & Atmospheric Sciences, Georgia Institute of Technology, Atlanta, Georgia, United States z Department of Earth and Planetary Science, Harvard University, Cambridge, Massachusetts, United States
)
‡
bS Supporting Information ABSTRACT: Atmospheric emissions of gas and particulate matter from a large ocean-going container vessel were sampled as it slowed and switched from high-sulfur to low-sulfur fuel as it transited into regulated coastal waters of California. Reduction in emission factors (EFs) of sulfur dioxide (SO2), particulate matter, particulate sulfate and cloud condensation nuclei were substantial (g90%). EFs for particulate organic matter decreased by 70%. Black carbon (BC) EFs were reduced by 41%. When the measured emission reductions, brought about by compliance with the California fuel quality regulation and participation in the vessel speed reduction (VSR) program, are placed in a broader context, warming from reductions in the indirect effect of SO4 would dominate any radiative changes due to the emissions changes. Within regulated waters absolute emission reductions exceed 88% for almost all measured gas and particle phase species. The analysis presented provides direct estimations of the emissions reductions that can be realized by California fuel quality regulation and VSR program, in addition to providing new information relevant to potential health and climate impact of reduced fuel sulfur content, fuel quality and vessel speed reductions.
1. INTRODUCTION Regulations on the atmospheric emissions from the transportation sector are motivated by the desire to reduce emissions of ozone (O3)-forming chemicals, particulate matter (PM), acid rain- and PM-forming sulfur dioxide (SO2), and other emissions harmful to human health and welfare. Regulation of fuel quality (sulfur, ash or aromatic hydrocarbon content) is one of several approaches that can be used to achieve reductions in these harmful emissions.1 Commercial shipping has had limited fuel quality (or emissions) regulation until recently, even though the shipping industry emits (globally) 3 times more SO2 than road traffic.2 Commercial shipping, although fuel-efficient, mostly consumes low-quality residual fuel (or heavy fuel oil, HFO), r 2011 American Chemical Society
which can have fuel sulfur content (SF) exceeding 3 or 4% (by weight),3 contain elevated concentrations of heavy metals4 and emit significantly more PM (SO4, particulate organic matter (POM) and black carbon (BC)) than more refined fuels.5 In recent years, the contribution of commercial shipping to air pollution has been recognized as significant (e.g., ref 6). In 2005 the International Maritime Organization (IMO) introduced a global cap to SF of 4.5% (reducing to 3.5% in 2012 and 0.5% by Received: April 19, 2011 Accepted: August 23, 2011 Revised: August 19, 2011 Published: September 12, 2011 9052
dx.doi.org/10.1021/es2013424 | Environ. Sci. Technol. 2011, 45, 9052–9060
Environmental Science & Technology 2020),7 motivated by PM reductions for air quality improvements that reductions in sulfur emissions are expected to achieve. Emission control areas (ECAs) have been established through the IMO in the North and Baltic seas to improve regional air quality. These ECAs require consumption of fuels with SF < 1.0%.8,9 In 2009 the US state of California introduced regulations that limit SF consumed within 44.5 km of the Californian coast, which require the use of marine gas oil (MGO) or marine diesel oil (MDO) with a maximum SF of 1.5% or 0.5%, respectively (by January 2012 SF must be <0.1% 10). In 2010 the IMO designated waters within 370 km of the United States and Canadian coast lines as an ECA requiring SF <1% by August 2011, reducing to 0.1% in January 2015.11 Expected benefits from the future global IMO regulations amount to ∼41 200 avoided premature deaths annually (for 2012),12 while up to 8000 avoided premature deaths per year are expected as a result of the future North American ECA regulation (for 2020).13 Consideration of the climate impacts of such regulatory changes has begun only recently. SO4 emissions have a cooling effect on climate due to both light scattering by the particles (direct radiative effect) and from the cloud-forming and modifying ability of cloud condensation nuclei (CCN, indirect radiative effects). Eyring et al.14 estimated the combined direct and indirect radiative forcing (RF) from shipping related SO4 emissions to be 0.44 W m2 (for 2005, globally averaged), with 90% of this from indirect effects. Concurrent emissions of other species (CO2, O3 precursors and BC), were estimated to have a net warming effect of +0.07 Wm2. These forcings are global averages of the effect of both short-lived (e.g., PM) and long-lived (e.g., CO2) forcing agents and have different spatial and temporal impacts.15 Currently, there are no expectations that BC emissions will be reduced due to fuel sulfur regulations (CO2 emissions may decrease slightly due to higher energy content of the more refined fuels), so IMO regulations are expected to decrease the net climate cooling from shipping emissions.16 The newly regulated coastal waters of California provide an opportunity to measure the influence of fuel quality regulation and speed reduction incentive programs on the magnitudes of emissions. These measurements will shed light on the potential air quality and climate effects of the impending regional and global fuel quality regulation, and possible vessel speed reduction (VSR) programs. In previous studies5,17 we showed that correlations between some shipping emissions (e.g., SO4, CCN) and SF are observable in real-world operations. The variability around these correlations is largely due to intership variations in operating conditions, making a quantitative assessment of the potential impacts of fuel quality regulations challenging. The analysis of Winnes and Fridel18 supports our assessment of previous data, suggesting that detailed characterization of emission factors from a single engine (or vessel) switching between high and low sulfur fuel is required (ideally on multiple vessels) to more accurately assess the impact of regulations on emissions. Here we provide emission factor comparisons from a container vessel where total exhaust emissions were measured as the vessel slowed and switched from high to low sulfur fuel near and within the California regulated waters during the 2010 CalNEX field campaign (http://www.esrl.noaa.gov/csd/calnex/).
2. FUEL SWITCH EXPERIMENT AND MEASUREMENT OF EMISSION FACTORS Experiment Details. On the 21st of May, 2010, in collaboration with the Maersk Line shipping company, the NOAA WP-3D
ARTICLE
Figure 1. (a) Map showing section of California fuel sulfur regulation zone (dashed gray), course of the sampled MM for both inbound and outbound days (solid and dashed Red), the flight track of the NOAA WP-3D aircraft (black) and the track of the R/V Atlantis (solid gray). Red triangles mark the approximate location of the start and end of the fuel switch on the inbound journey (reported by Maersk). b) Example plume data for SO2 (blue), SO4 (red) and CO2 (black).
research aircraft19 intercepted the Margrethe Maersk (MM) vessel on its way to the Port of Los Angeles, prior to the vessel starting the fuel switching procedure required by California state law (Figure 1a). The MM is a 371 m, 96 500 tonne container vessel running a 12 cylinder, 68.7 megawatt (MW) main diesel engine (3, 3.8 MW auxiliary engines). The MM was consuming HFO containing 3.15% sulfur and 0.05% ash (by weight) before a gradual blending of MGO containing 0.07% sulfur and <0.01% ash occurred over an 60 min period just outside California regulated waters.20 On average, 60% of emissions were from the main engine, 10% from the auxiliary engines and 30% from boilers20 (all engines switched fuels). The MM also participated in the Californian VSR incentive program,23 changing speed across the fuel switch operation (22 knots prior and 12 knots after). These speed changes and differences in the relative fuel consumption between engines complicates the interpretation of results (discussed in more detail below). The emissions reductions reported here are due to both compliance with regulation (3.15% down to 1.5% SF) as well as the choice of the vessel operator to use MGO with lower SF than required by regulation (1.5% down to 0.07% SF). 9053
dx.doi.org/10.1021/es2013424 |Environ. Sci. Technol. 2011, 45, 9052–9060
Environmental Science & Technology
ARTICLE
Table 1. Summary of Emission Factors Measured from the MM Outside and Within Regulated Waters fuel or emission component
before fuel switch (outside regulated waters)
after fuel switch (within regulated waters)
units
% change
fuel sulfur (SF reporteda
3.15
0.07
%
98%
fuel sulfur (SF) calculatedb
2.6 ((0.4)
0.21 ((0.03)
%
92%
sulfur
25.6 ((4)
2.1 ((0.3)
g kg1
92%
SO2
49 ((7.5)
4.3 ((0.6)
g kg1
91%
measured
lack et al. (2009)d
lack et al. (2009)e
measured
SO4
2.94 ((1.0)
1.5 ((1.6)
0.08 ((0.03)
0.06 ((0.05)
g kg1
97%
POM
0.58 ((0.2)
1.5 ((1.0)
0.17 ((0.06)
0.9 ((1.2)
g kg1
71%
BC
0.22 ((0.09)
0.7 ((0.8)
0.13 ((0.05)
1.1 ((0.8)
g kg1
41%
PMc
3.77 ((1.3)
3.0 ((1.7)
0.39 ((0.14)
1.8 ((1.4)
g kg1
90%
NTot
1.0 1016 ((0.2 1016)
1.4 1016 ((1.0 1016)
1.4 1016 ((0.2 1016)
1.0 x1016 ((0.7 1016
# kg1
+40%
CCN (SS = 0.3%)
4.0 1015 ((0.4 1015)
2.4 x1015 ((2.0 x1015)
0.1 x1015 ((0.01 1015)
0.2 1016 ((0.1 x1015)
# kg1
97.5%
CCN/NTot SO4 / sulfur
40 ((10) 4.1 ((0.7)
34 ((27) 3.9 ((1.4)
0.7 ((0.2) 1.2 ((0.2)
7.4 ((6.0) 1.4 ((1.1)
% %
98% 71%
a Provided by Maersk. b Calculated from EFS/10 26 c Sum of SO4, POM and BC. Does not include SO4-bound water or ash. d Average and standard deviation EFs from vessels using >0.5% SF from Lack et al.5 e Average and standard deviation EFs from vessels using <0.5% SF from Lack et al.5
The WP-3D sampled the emissions plume of the MM before and during the fuel switching operation at approximately 100 m above sea level, 13 km downwind of the vessel (25 min). These times downwind are insufficient for significant atmospheric processing of SO2, SO4, BC, or POM.5,17,21,22 Due to aircraft operational issues the flight was aborted before sampling of low SF emissions could occur. Four days later (24th May, 2010) the NOAA-sponsored Woods Hole Oceanographic Institute research vessel R/V Atlantis sampled the MM emissions 2.57.5 min after emission while within the low-sulfur regulated zone (shown in figures as a triangle data point). The R/V Atlantis sample inlet was approximately 15 m ASL. A direct intercomparison between WP-3D and R/V Atlantis instrumentation was not possible during the campaign. The Supporting Information contains details of common calibrations used between instruments on both platforms. Due to these common calibrations we assume that measurements on both platforms are equally accurate to within the stated uncertainties. Calculation of emissions changes before and after the experiment therefore include these uncertainties. Instrumentation. Measurements taken onboard the NOAA WP-3D research aircraft and the R/V Atlantis included concentrations of CO2, SO2, SO4, POM, BC, particle number (NTot), and CCN as well as particle size distributions (note: NOX data was not available for this analysis). Measurement techniques, uncertainties and references are provided in Supporting Information (Table S1). PM1 mass is estimated as the sum of BC, SO4, and POM mass. CCN are reported at a super saturation (SS) of 0.3%, a SS relevant for pristine stratocumulus and trade-wind cumulus clouds (e.g., ref 24). We determined emission factors (EF: amount emitted per kilogram of fuel burnt) by first determining the ratio between the integrated areas of the data of the plume intercepts for the species of interest and CO2. An example plume encounter from the WP-3D is shown in Figure 1b. The average of CO2 integrated areas from two independent measurement methods were used for WP-3D data. Maximum difference between the integrated areas of the two methods was 10% = CO2 plume integration uncertainty. The measured emission ratios are converted to EFs according to Williams et al.22 and Lack et al.5
Instrument and CO2 plume integration (10%) uncertainties are propagated through the calculation of the EF. Background pollutant levels and plume dilution/mixing are inherently accounted for via normalization of the emission to the measured CO2 concentration. EFs are missing for some plume intercepts due to instrument filter or calibration periods. Engine load as a fraction of maximum load (fLoad) was estimated from the vessel speed (as load ∼ speed3 25) recorded from the regular Automated Information System (AIS) radio broadcasts from the MM, where the maximum vessel speed is 25 knots.
3. RESULTS Summary of Emissions. A summary of EFs and a comparison across the experiment is presented in Table 1. Detailed discussion is presented in the sections that follow. As the MM transitioned from high sulfur to low sulfur fuel and slowed, EFs for SO2, SO4, and CCN dropped by 91%, 97%, and 97.5%, respectively. PM, POM and BC EFs dropped by 90%, 71%, and 41% respectively. EFNTot change was variable and possibly increased after the fuel switch was complete. The various PM EFs for the MM prior to the fuel switch fall within the range of values observed in the comprehensive study by Lack et al.,5 although the POM and BC prior to the fuel switch are about 1/3 of the reported averages (Table 1). Measured PM EFs also compare well to other studies utilizing high SF fuels (e.g., refs 4,18,2629). Sulfur Dioxide Emissions. Compliance with the fuel sulfur regulation provides direct and large reductions in EFSO2 of 91% (Figure 2a). Some fuel sulfur is directly emitted as SO3 (and quickly forms SO4)5,28 and so EFSO4 and EFSO2 are combined (accounting for stoichiometry) to determine an EF of total sulfur (EFS). SF as estimated from EFS (SF ≈ EFS/10 26) changed from 2.6% (HFO) to 0.2% (MGO) across the fuel switch. Maersk records indicate that SF of the fuels dropped from 3.15% HFO to 0.07% MGO (98% drop). The source of this discrepancy is unknown, however several groups18,30 have observed discrepancies (of up to 0.5%) between the SF reported in the fuel analysis and that calculated from emission measurements. Nonetheless, it is 9054
dx.doi.org/10.1021/es2013424 |Environ. Sci. Technol. 2011, 45, 9052–9060
Environmental Science & Technology
ARTICLE
Figure 2. (a) EFSO2 and EFS, (b) EFSO4, (c) EFCCN @ 0.3% SS during fuel switching operation, and (d) fraction of fuel sulfur converted to SO4 versus engine load. Gray line is the trend of previous data from Petzold et al.28 Note: Figure 2a uses a 3rd order polynomial fit EFS = 0.1 + 0.16x + 25.6x2.
Figure 3. (a) EFNTot during experiment, (b) average size distributions (and log-normal fits) before and after the experiment, and median diameter (X) evolution (c) EFPOM, and (d) EFBC during the experiment. For the lowest SF EFBC (R/V Atlantis intercept), three data points of almost identical magnitude are plotted (SP2 and two PAS instruments).
clear EFSO2 is strongly correlated to SF and we anticipate an equivalent reduction in secondary SO4 produced from downwind oxidation of SO2. We fit the general trend in EFS vs plume encounter (black line, Figure 2a) and estimate an SF for each plume encounter from this fit, which is used as the x-axes for Figures 2b,c and 3a,c,d. Particulate Sulfate Emissions. EFs of directly emitted SO4 decreased by 97% during the experiment (Figure 2b). The fraction
of total sulfur emitted as SO431 is 3.5% at high SF (fLoad = 0.7) and 1.2% at low SF (fLoad = 0.1) (Figure 2d). The observed variation in the SO4 fraction with fLoad is in excellent agreement with the results of Petzold et al.28 (gray line Figure 2d), although the fLoad effect does not account for the entire change observed. Therefore both SF and fLoad contribute to the 97% reduction in EFSO4. Cloud Condensation Nuclei, Particle Number Emissions and Particle Size. EFCCN are strongly correlated with EFSO4 and 9055
dx.doi.org/10.1021/es2013424 |Environ. Sci. Technol. 2011, 45, 9052–9060
Environmental Science & Technology were reduced by almost 98% across the experiment (Figure 2c). The ratio between EFCCN and EFNTot (fCCN) gives an indicator of the efficacy of an emitted particle toward CCN formation and decreases from fCCN = 0.4 to 0.007 (98% reduction). The ability of a given particle to act as a CCN (at a given %SS) depends on both the particle composition and size. Additionally, the ability of particles within a size distribution to act as CCN depends on the extent of internal vs external mixing. The composition effect on hygroscopicity can be approximately characterized assuming complete internal mixing, through calculation of the effective “Kappa” parameter (keff) from the observations as follows: ! Vi mi Ftot kef f ¼ ki ¼ ki Vtot mtot Fi i i ! EFi Ftot ¼ ð1Þ ki EFtot Fi i
∑
∑
∑
where Vx is volume, mx is mass, Fx is density, and ki is the speciesspecific hygroscopicity of species i (or of the total).32 We use Fi = 1.7, 1.3, and 1.8 g/cm3 and ki = 0.9, 0.1, and 0.0 for SO4 (from H2SO4), POM and BC, respectively.33 Because the EFs for SO4, POM, and BC all decrease with decreasing fuel sulfur, the calculated keff does not change nearly as dramatically as either the observed EFCCN or the fCCN. In fact, keff is stable around 0.680.73 for all encounters, with the exception of the R/V Atlantis encounter, when SF was minimum, where keff drops to 0.2. Thus it appears that the consistent decrease in EFCCN and fCCN with SF is, in general, not being driven by changes to the particle composition despite the fact that the absolute EFSO4 decreases continuously. Measured size distributions (Figure 3b) show that the median particle size decreased concurrent with the decrease in EFSO4 (number-weighted particle diameter decreased from 60 to 36 nm). The calculated critical dry diameter for CCN activation of particles with the observed keff at 0.3% SS is 60 nm,32 which is consistent with the observation of fCCN = 40% for the high SF emissions. For a change in keff to 0.2, the critical dry diameter at 0.3% SS would increase to about 90 nm. The combination of the decrease in particle size and the sudden drop in keff leads to the very low fCCN for the lowest SF intercept. The measured reduction in EFCCN during the experiment therefore results primarily from changes to the particle size distribution (which most likely result from changes in fLoad), but for the lowest SF (and fLoad) both composition and size changes play a role. Similar to our results, for a test engine operating on HFO, Petzold et al.28 observed a slight shift toward smaller particle sizes as fLoad was decreased (most notable at lower fLoad). The EFNTot do not show a strong dependence on SF (Figure 3a). Lack et al.5 showed reductions in EFNTot between vessels burning high and low sulfur fuel, whereas Winnes and Fridell18 report that the number of smaller particles may increase as SF decreases. As shown in Lack et al.5 these small particles quickly condense onto the larger particles, therefore although initial emissions of NTot may increase, the atmospheric lifetime is shorter than the larger particles. Petzold et al.28 found that EFNTot increased by a factor of 1.65 as fLoad decreased from 85 to 50%. The variability across these studies suggest that NTot emissions are dependent on engine operating parameters including fLoad and SF.
ARTICLE
Particulate Organic Matter Emissions. Reductions in EFPOM (up to 71%) were observed across the experiment (Figure 3c). This reduction may be explained through two factors. First, the refining process for HFO concentrates aromatic and longer chain hydrocarbons, which have delayed burn times in some engines.34 Thus, the higher POM emissions from high SF likely result, in part, from the incomplete combustion of the aromatic and long chain hydrocarbons at high SF. Second, there is larger consumption (and emission) of lubricating oils when HFO is used. However, short-term use of distillate fuels does not always require lubrication oil changes35 and the MM did not alter the lube-oil regime for this fuel switch.20 Petzold et al.28 did not show any link between POM and fLoad for a single test-engine operating on HFO while Lack et al.5 observed a clear correlation between POM and SF. This suggests that the POM reductions observed in Figure 3c are likely due to organic composition changes within the fuel, which correlate to SF. Black Carbon Emissions. EFs of BC appeared to decline across the experiment, although measurement uncertainties indicate a range from 30 to 70% (average of 41%) (Figure 3d). Some measurements of BC were below instrument detection limits despite having measurable CO2 enhancements (the reason for which is currently unknown). To our knowledge there are no published data that would suggest reductions in SF should decrease EFBC. However it has been observed that reductions in slow burning aromatic hydrocarbons within jet turbine fuels reduces BC emissions from these engines.36 Ash, aromatic and long chain hydrocarbon compounds, which are concentrated in HFO, are decreased in refined MGO. We suggest that reduction in these components decreases the concentration of flame quenching nuclei, which decreases BC formation. The results of Righi et al.2 suggest that BC emissions are reduced for cleaner fuels (MGO, biodiesel) relative to HFO. However, recent studies by Agrawal et al.37 (in-use vessel running HFO) and Petzold et al.28 (medium speed diesel (MSD) engine running HFO) showed EFBC increased 1.53 times respectively when fLoad changed from 0.7 to 0.1. While there is a net gain to vessel speed reduction (VSR) in terms of increased fuel efficiency (which acts to reduce absolute emissions of CO2, SO2, and PM, given a constant EF), an increase in the emission factors of BC may actually offset some of the fuel efficiency gains. If the results of Petzold et al.28 and Agrawal et al.37 are applicable to this experiment, the observed decrease in EFBC (Figure 3d) is a lower limit in overall BC reductions due to the change in fuel quality. Alternatively, other results for show MSD engines burning low sulfur MGO suggest that EFBC may increase.38,39 Fuel efficiency gains to absolute BC emissions would then be enhanced by concurrent reductions in the EFBC, and thus the influence of the fuel quality regulations alone on EFBC would be smaller than shown in Figure 2d. Given that the observations in this study and those of Petzold et al.28 and Agrawal et al.37 were for engines or vessels burning HFO, it seems reasonable that the BC reductions observed here are linked to SF rather than fLoad. Certainly more detailed investigation is necessary. Nonetheless, the overall effect of the fuel quality regulation and the VSR program appears to be a decrease in both EFBC and absolute BC emissions. Any BC reduction due to improved fuel quality in ships will provide additional benefits for air quality although may have an uncertain impact of climate (see climate discussion below). Use of higher quality fuels by ships in the Arctic may result in less BC deposition to snow and ice (compared to the use of low quality fuels) resulting in positive climate benefits in that region.40 9056
dx.doi.org/10.1021/es2013424 |Environ. Sci. Technol. 2011, 45, 9052–9060
Environmental Science & Technology
ARTICLE
Figure 4. Emissions reductions (per km of travel) from the MM as a result of the State of California fuel sulfur regulation (gray), vessel speed reduction program (white) and combined (black).
4. DISCUSSION Information Relevant to Impacts of Regional Regulation. On a per-kilometer (km) basis, emissions of most gas and particle pollutants from the MM dropped significantly once the MM entered the region where it is required to be in compliance with the California regulations. Figure 4 (and Supporting Information Table S2) summarizes the emissions for a km of travel outside and inside the regulated waters, calculated from the emission factors presented in Table 1. Estimates of fuel consumption by the MM at the speeds traveled inside and outside of the regulated waters were calculated using eq 2 and data obtained from Maersk:20
CFuel ðkghr1 Þ ¼ Fcons 1000PMW fLoad
ð2Þ
where 1
Fcons ðkgkw:hr Þ ¼ 0:0142
1
fLoad
þ 0:195
ð3Þ
The engine manufacturer literature suggests that a new engine of the type installed on the MM has a fuel consumption rate (FCons) at maximum load of 0.17 kg (kw.hr)1 although 0.195 kg (kw.hr)1 is estimated to be an appropriate average value for in-use slow speed diesel engines.41 FCons varies with engine load according to eq 3.42 FCons for MGO is reduced by 6% due to the specific heat of MGO being 6% higher than HFO on this vessel.20 PMW is the maximum engine power in megawatts (68.7 MW). These data were converted to kilograms of fuel consumed perkilometer (km) of travel, which were then converted to per-km emissions by multiplying CFuel with the measured EFs.
For all but CO2, BC, and NTot, pollutant levels drop by 88% or more (58% for CO2, 75% for BC and 41% for NTot) as a result of the vessel observing both the fuel quality regulation and VSR program (Figure 4). Note that most CO2 reductions arise from the change in fLoad,. Importantly, we can differentiate some of the emissions reductions by the effects of the fuel quality regulation or VSR program. To make this assessment, we have assumed that the observed EF reductions for SO2 and POM are due entirely to the SF change. At high fLoad SO4 formation is 2.9 times higher than at low fLoad (Figure 2d and Petzold et al.28) and this load factor is removed from SO4 emissions by multiplying the low-SF, low-load EFSO4 by 2.9. It is apparent that the emissions of BC, NTot, and CCN are complicated by SF and fLoad and we do not separate by regulation for these species. Note that this analysis is specific to the MM, which was in compliance with the fuel quality regulation and was participating in the VSR program. We reiterate that these results are a snapshot for a single vessel with changing fuel type, fuel consumption distributions across main, auxiliary and boiler engines, and changing speed. Although these factors introduce uncertainly for detailed emissions analysis, the trends for the averaged vessel emissions are evident. Information Relevant to Health Impacts. Reductions in the direct emissions of SO4, BC, and POM per-km of travel of 99%, 75%, and 88%, respectively, will likely have influence on the ambient PM levels near the Californian coast where vessel traffic is significant, especially in the port regions. The reductions in EFBC and EFPOM with improved fuel quality are significant variables that have not been considered in most assessments of the impact of shipping emissions on health. Assuming that reductions in PM emissions leads to reduced mortality, this new information would suggest that greater reductions in mortality 9057
dx.doi.org/10.1021/es2013424 |Environ. Sci. Technol. 2011, 45, 9052–9060
Environmental Science & Technology
ARTICLE
would be found than reported in the North American ECA or global IMO regulation mortality assessments 11,12 (that do not include the BC and POM reductions). In addition, the finding that SO4 emissions decrease with both SF and engine load28 shows that primary SO4 emissions will be further decreased if VSR regulation is introduced. Reductions in SO2 will also significantly reduce secondary SO4 formation. Of further interest is the uncertainty surrounding EFNTot associated with reductions in SF and speed changes. Multiple studies (including the current data) show opposing trends in EFNTot as vessel speed and SF change, and should be investigated further. Information Relevant to Climate Impacts. The indirect RF impacts of PM are difficult to assess and remain the least certain RF agent in global models. For shipping, it is estimated that emitted PM leads to a significant negative RF (i.e., cooling) that substantially exceeds the warming from the emitted CO2.2,14,16 The impact of fuel quality (predominantly reducing the SF) would lead to a reduction in this cooling.2,16 Eyring et al.14 estimate (for 2005) that the globally averaged direct and indirect RF by shipping emissions of SO4 and POM is 0.44 Wm2 (net cooling), which is dominated by the indirect RF (0.41 W m2). CO2, O3 (from NOX emissions), decreased CH4 (from NOX) and BC from shipping together have a globally averaged positive RF of +0.03 Wm2 (net warming). Righi et al.2 estimate this indirect RF would decrease from 0.28 to 0.10 Wm2 if low SF fuels are introduced globally. For the data presented here, although absolute BC emissions decrease, the strong concurrent decrease in CCN emission (from both composition and size changes) could completely offset the cooling gained.43 Given the observed, concurrent reductions in emissions of BC, POM, and CCN (75%, 88%, and 99%, respectively), we conclude that uncertainties in the magnitude of the RF balance from shipping are critically dependent on the composition of emitted PM, size distributions, and the ultimate fate of emitted non-CCN active particles in the atmosphere. The direct RF impact of shipping emissions of PM, although small relative to the indirect effect, will also change due to fuel regulation. Over the past 1520 years, fuel regulation in California for on-road vehicles and nonroad machines has focused on a variety of technological approaches, such as engine rebuilding or addition of emissions control systems.44 The goal (and likely net result) of this regulation was (has been) to reduce primary emissions of BC45 which, if it occurs in isolation, will lead to less warming. However, absorbing BC is usually coemitted with scattering (cooling) SO4 and POM, which may also change upon implementation of a control measure.46 The single scattering albedo (SSA) represents the balance between light scattered and absorbed by a particle and is one of the primary influences on whether a particle warms or cools the atmosphere. The SSA for the MM encounter (for high and low SF) was estimated from the measured EFSO4, EFPOM, and EFBC values using 532 nm mass extinction and mass absorption efficiencies (MEE and MAE) for the different species; SSA 532 ∼ 1
MEESO4 EFSO4
MAEBC EFBC þ MEEPOM EFPOM þ MEEBC EFBC
ð4Þ We use values for the MEE for SO4 and POM from Malm et al.47 (3 m2/g and 4 m2/g) and MEE/MAE values for BC from Bond and Bergstrom48 (9 m2/g and 7.5 m2/g). The SSA for directly emitted PM from the MM decreased from 0.86 to 0.57 across the
experiment. The estimated low-SF SSA value compares favorably with the directly measured dry value of 0.64 (0.2% SF, 532 nm). This is generally consistent with the observations of Lack et al.,5 who found that the SSA decreased from 0.6 to 0.3, on average, as the SF changed from 2.5 to 0.2%. Thus, not only will the absolute PM emissions from ships operating on low sulfur (instead of high sulfur) fuel be decreased, the particles that are emitted will be overall “darker” and can then have a stronger relative warming influence. It seems clear that the implementation of global fuel sulfur regulations will lead to a decrease in the cooling by ship PM emissions, both from changes in indirect and direct RF. We emphasize that the emission reductions observed with the MM introduce previously unaccounted emissions phenomena which may alter the specific RF balance from shipping described by recent model studies.2,16 Local, Regional And Global Policy Connections. The efficacy of Californian shipping fuel quality regulation and vessel speed reduction (VSR) program in reducing emission factors and absolute emissions (emissions per-km of travel with and without the regulation) of SO2, SO4, and (somewhat unexpectedly) POM and BC is evident from the results presented here. EFs of NTot (particle number) appear to increase due to the regulations, although it is likely that these are small particles that will quickly condense or coagulate with existing particles. On an absolute scale (per kilometer of travel), mass reductions of SO2, SO4, and PM are in excess of 96%; BC and POM reductions are 75% and 88% respectively. The regulations will significantly alter the direct climate cooling impacts of the emitted PM by reduction of the SO4 formed just after emission and through secondary formation from SO2 oxidation. In areas where low sulfur fuel is used, significant CCN reductions and particle size reductions will reduce the indirect cooling impacts from enhanced cloud formation, particularly in regions sensitive to inputs of CCN from shipping, such as at ∼30 N. This reduced cooling may be partially offset by a concurrent decrease in the climate warming impact of BC. Our observations suggest that air quality benefits from the fuel quality regulation and the VSR program are likely to be substantial, although these air-quality benefits are likely to occur concurrent with a reduction in anthropogenic cooling that results from shipping PM. If it is determined that air pollution (i.e., human health and welfare) goals can be met through nearcoast regulation (i.e., ECAs), then the implementation of a more nuanced location-dependent global fuel quality regulation may be worthy of consideration. Lastly, possible reductions in BC emissions due to fuel quality changes might suggest a consideration of more refined fuels for future Arctic shipping.40
’ ASSOCIATED CONTENT
bS
Supporting Information. Details on instruments uncertainty, literature and calibrations. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT Thanks to the support of the Maersk Line (in particular Lee Kindberg and Wayne Tober) and the crew and support staff of 9058
dx.doi.org/10.1021/es2013424 |Environ. Sci. Technol. 2011, 45, 9052–9060
Environmental Science & Technology the WP-3D research aircraft and the R/V Atlantis. Thanks also to James Corbett for useful discussions. This work was funded in part by NOAAs Climate Program (NA09OAR4310124, NA09AR4310125), California Air Resources Board, US EPA (RD834558), Canadian Federal Government (PERD Project C12.007) and NSERC.
’ REFERENCES (1) Corbett, J. J.; Winebrake, J. J., The role of international policy in mitigating global shipping emissions. Brown J. World Affairs 2010, XVI, (II). (2) Righi, M.; Klinger, C.; Eyring, V.; Hendricks, J.; et al. Climate impact of biofuels in shipping: global model studies of the aerosol indirect effect. Environ. Sci. Technol. 2011, 45 (8), 3519–3525. (3) IMO Amendments to MARPOL Annex VI, Amendments to the NOX Technical Code; International Maritime Organization: London, April, 2008; p 144. (4) Lyyranen, J.; Jokiniemi, J.; Kauppinenen, E. I.; Joutsensaari, J. Aerosol characterisation in medium speed diesel engines operating with heavy fuel oils. J. Aerosol. Sci. 1999, 30 (6), 771–784. (5) Lack, D. A.; Corbett, J. J.; Onasch, T. B.; Lerner, B., et al., Particulate emissions from commercial shipping. chemical, physical and optical proprties. J. Geophys. Res. 2009, 114, (D00F04), DOI: 10.1029/ 2008/JD011300. (6) Dalsoren, S. B.; Eide, M. S.; Endresen, o.; Mjelde, A.; et al. Update on emissions and environmental impacts from the international fleet of ships: the contribution from major ship types and ports. Atmos. Chem. Phys. 2009, 9 (6), 2171–2194. (7) IMO. Air Pollution from Ships Cut, with Entry Into Force of MARPOL Amendments, 2010, http://www.imo.org/newsroom/mainframe.asp?topic_id=1859&doc_id=13309. (8) IMO. New Rules to Reduce Emissions from Ships Enter Into Force, 2005; http://www.imo.org/newsroom/mainframe.asp?topic_ id=1018&doc_id=4884. (9) IMO North Sea SECA in effect from 22 November, 2007, http://www.imo.org/inforesource/mainframe.asp?topic_id=1472& doc_id=8719. (10) CARB. Final Regulation Order. Fuel Sulfur and Other Operational Requirments for Ocean-Going Vessels Within California Waters and 24 Nautical Miles of the California Baseline; California Air Resources Board: Sacramento, CA, 2009; http://www.arb.ca.gov/regact/2008/ fuelogv08/22992.pdf. (11) IMO In Proposal to Designate an Emission Control Area for Nitrogen Oxides, Sulphur Oxides and Particulate Matter. Submitted by the United States and Canada., IMO MPEC 59, 2009; IMO: 2009. (12) Winebrake, J. J.; Corbett, J. J.; Green, E. H.; Lauer, A.; et al. Mitigating the health impacts of pollution from oceangoing shipping: an assessment of low-sulfur fuel mandates. Environ. Sci. Technol. 2009, 43 (13), 4776–4782. (13) U.S. EPA, EPA Proposal for Control of Emissions from New Marine Compression-Ignition Engines at or above 30 Liters Per Cylinder, EPA-420-F-09-029; United States Environmental Protection Agency Office of Transportation and Air Quality: Washington, DC, 2009. (14) Eyring, V.; Isaksen, I. S. A.; Berntsen, T.; Collins, W. J.; et al. Transport impacts on atmosphere and climate: Shipping. Atmos. Environ. 2010, 44 (37), 4735–4771. (15) Fuglestvedt, J. S.; Shine, K. P.; Berntsen, T.; Cook, J.; et al. Transport impacts on atmosphere and climate: Metrics. Atmos. Environ. 2010, 44 (37), 4648–4677. (16) Lauer, A.; Eyring, V.; Corbett, J. J.; Wang, C.; et al. Assessment of near-future policy instruments for oceangoing shipping: impact on atmospheric aerosol burdens and the earth’s radiation budget. Environ. Sci. Technol. 2009, 43 (15), 5592–5598. (17) Lack, D. A.; Lerner, B.; Granier, C.; Baynard, T., et al. , Light absorbing carbon emissions from commercial shipping. Geophys. Res. Lett. 2008, 35, (L13815), DOI: 10.1029/2008GL033906.
ARTICLE
(18) Winnes, H.; Fridell, E. Particle emissions from ships: Dependence on fuel type. J. Air Waste Manage. 2009, 59, 1391–1398. (19) Brock, C. A.; Cozic, J.; Bahreini, R.; Froyd, K. D.; et al. Characteristics, sources, and transport of aerosols measured in spring 2008 during the aerosol, radiation, and cloud processes affecting Arctic climate (ARCPAC) project. Atmos. Chem. Phys. Discuss. 2010, 10 (11), 27361–27434. (20) Maersk-Line, Personal Communication, 2010. (21) Chen, G.; Huey, L. G.; Trainer, M.; Nicks, D., et al., An Investigation of the chemistry of ship emission plumes during ITCT 2002. J. Geophys. Res. 2005, 110, (D10S90), DOI: 10.1029/2004JD005236. (22) Williams, E.; Lerner, B.; Murphy, P.; Herndon, S. C.; et al. Emissions of NOX, SO2, CO, and C2H4 from commercial marine shipping during Texas Air Quality Study (TexAQS) 2006. J. Geophys. Res. 2009, 114 (D21), D21306. (23) CARB. Vessel Speed Reduction for Ocean-going Vessels; California Air Resources Board: Sacramento, CA, 2009; www.arb.ca.gov/ports/ marinevess/vsr/vsr.htm. (24) Seinfeld, J. H.; Pandis, S. N., Atmospheric Chemistry and Physics: From Air Pollution to Climate Change; John Wiley and Sons: New York, 1998. (25) Harvald, S. Prediction of Power of Ships; Department of Ocean Engineering, Technical University of Denmark: Copenhagen, Denmark., 1977. (26) Cooper, D. A. Exhaust emissions from ships at berth. Atmos. Environ. 2003, 37 (27), 3817–3830. (27) Kasper, A.; Aufdenblatten, S.; Forss, A.; Mohr, M.; et al. Particulate emissions from a low-speed marine diesel engine. Aerosol Sci. Technol. 2007, 41, 24–32. (28) Petzold, A.; Weingartner, E.; Hasselbach, J.; Lauer, A.; et al. Physical properties, chemical composition, and cloud forming potential of particulate emissions from a marine diesel engine at various load conditions. Environ. Sci. Technol. 2010, 44, 3800–3805. (29) Sinha, P.; Hobbs, P. V.; Yokelson, R. J.; Christian, T. J.; et al. Emissions of trace gases and particles from two ships in the southern Atlantic Ocean. Atmos. Environ. 2003, 37 (15), 2139–2148. (30) Miller, W., Personal Communication on measured vs. reported fuel sulfur content of ship fuel, 2011. (31) Lovejoy, E. R.; Hanson, D. R.; Huey, L. G. Kinetics and products of the gas-phase reaction of SO3 with water. J. Phys. Chem. 1996, 100 (51), 19911–19916. (32) Petters, M. D.; Kreidenweis, S. M. A single parameter representation of hygroscopic growth and cloud condensation nucleus activity. Atmos. Chem. Phys. 2007, 7 (8), 1961–1971. √ (33) Pringle, K. J.; Tost, H.; Pozzer, A.; P ∂schl, U.; et al. , Global distribution of the effective aerosol hygroscopicity parameter for CCN activation. Atmos. Chem. Phys. 2010, 10 (12), 5241–5255. (34) American Bureau of Shipping. Notes on Heavy Fuel Oil; American Bureau of Shipping: Houston, 2001; pp 168, http://www.eagle. org/eagleExternalPortalWEB/ShowProperty/BEA%20Repository/ Rules&Guides/Current/31_HeavyFuelOil/Pub31_HeavyFuelOil. (35) Operation on Low-Sulphur Fuels. MAN B&W Two-Stroke Engines; MAN: Copenhagen, 2010; pp 124. (36) Corporan, E.; DeWitt, M. J.; Belovich, V.; Pawlik, R.; et al. Emissions characteristics of a turbine engine and research combustor burning a fischer—tropsch jet fuel. Energy Fuels 2007, 21 (5), 2615–2626. (37) Agrawal, H.; Welch, W. A.; Henningsen, S.; Miller, J. W.; et al. Emissions from main propulsion engine on container ship at sea. J. Geophys. Res. 2010, 115 (D23), D23205. (38) Cappa, C. D.; Williams, E.; Buffaloe, G.; Coffman, D. J., et al. , The influence of operating speed on gas and particle-phase emissions from the R/V Miller Freeman. in preparation. (39) Jayaram, V.; Agrawal, H.; Welch, W. A.; Miller, J. W.; et al. Realtime gaseous, PM and ultrafine particle emissions from a modern marine engine operating on biodiesel. Environ. Sci. Technol. 2011, 45 (6), 2286–2292. (40) Corbett, J. J.; Lack, D. A.; Winebrake, J. J.; Harder, S.; et al. Arctic shipping emissions inventories and future scenarios. Atmos. Chem. Phys. 2010, 10 (19), 9689–9704. 9059
dx.doi.org/10.1021/es2013424 |Environ. Sci. Technol. 2011, 45, 9052–9060
Environmental Science & Technology
ARTICLE
(41) Entec Quantification of Ship Emissions. http://ec.europa.eu/ environment/air/pdf/chapter2_ship_emissions.pdf. (42) U.S. EPA Analysis of Commercial Marine Vessels Emissions and Fuel Consumption Data; United States Environmental Protection Agency: Washington, DC. 2000. (43) Chen, W. T.; Lee, Y. H.; Adams, P. J.; Nenes, A.; et al. Will black carbon mitigation dampen aerosol indirect forcing? Geophys. Res. Lett. 2010, 37 (9), L09801. (44) CARB. The California Diesel Fuel Regualtions; California Air Resources Board: Sacramento, CA, 2004; http://www.arb.ca.gov/regs/ regs.htm. (45) Murphy, D. M.; Capps, S. L.; Daniel, J. S.; Frost, G. J.; et al. Weekly patterns of aerosol in the United States. Atmos. Chem. Phys. 2008, 8 (10), 2729–2739. (46) Bond, T. C.; Sun, H. Can reducing black carbon emissions counteract global warming? Environ. Sci. Technol. 2005, 39 (16), 5921–5926. (47) Malm, W. C.; Day, D. E.; Carrico, C.; Kreidenweis, S. M.; et al. Intercomparison and closure calculations using measurements of aerosol species and optical properties during the Yosemite Aerosol Characterization Study. J. Geophys. Res. 2005, 110 (D14), D14302. (48) Bond, T. C.; Bergstrom, R. W. Light absorption by carbonaceous particles: An investigative review. Aerosol Sci. Technol. 2006, 40, 27–67.
’ NOTE ADDED AFTER ASAP PUBLICATION This paper published September 12, 2011 with an author name spelled incorrectly. Ibraheem Nuaaman's name appeared correct in the version of this paper published September 21, 2011.
9060
dx.doi.org/10.1021/es2013424 |Environ. Sci. Technol. 2011, 45, 9052–9060
ARTICLE pubs.acs.org/est
Microbially Mediated Mineral Carbonation: Roles of Phototrophy and Heterotrophy Ian M. Power,*,†,^ Siobhan A. Wilson,‡,||,# Darcy P. Small,§ Gregory M. Dipple,|| Wankei Wan,§ and Gordon Southam† †
Department of Earth Sciences, The University of Western Ontario, London, Ontario N6A 5B7, Canada Department of Geological Sciences, Indiana University, Bloomington, Indiana 47405-1405, United States § Department of Chemical and Biochemical Engineering, The University of Western Ontario, London, Ontario N6A 5B9 Canada Mineral Deposit Research Unit, Department of Earth and Ocean Sciences, The University of British Columbia, Vancouver, British Columbia V6T 1Z4, Canada
)
‡
bS Supporting Information ABSTRACT: Ultramafic mine tailings from the Diavik Diamond Mine, Canada and the Mount Keith Nickel Mine, Western Australia are valuable feedstocks for sequestering CO2 via mineral carbonation. In microcosm experiments, tailings were leached using various dilute acids to produce subsaline solutions at circumneutral pH that were inoculated with a phototrophic consortium that is able to induce carbonate precipitation. Geochemical modeling of the experimental solutions indicates that up to 2.5% and 16.7% of the annual emissions for Diavik and Mount Keith mines, respectively, could be sequestered as carbonate minerals and phototrophic biomass. CO2 sequestration rates are mainly limited by cation availability and the uptake of CO2. Abundant carbonate mineral precipitation occurred when heterotrophic oxidation of acetate acted as an alternative pathway for CO2 delivery. These experiments highlight the importance of heterotrophy in producing sufficient DIC concentrations while phototrophy causes alkalinization of waters and produces biomass (fatty acids = 7.6 wt.%), a potential feedstock for biofuel production. Tailings storage facilities could be redesigned to promote CO2 sequestration by directing leachate waters from tailings piles into specially designed ponds where carbonate precipitation would be mediated by both chemical and biological processes, thereby storing carbon in stable carbonate minerals and potentially valuable biomass.
’ INTRODUCTION Fossil fuels account for 85% of the global energy supply and are expected to remain an important energy source as developing countries’ demand for energy increases.1 In order to reduce emissions and stabilize atmospheric CO2 it will likely be necessary to implement carbon sequestration strategies.2,3 Mineral carbonation involves dissolution of silicate minerals and subsequent precipitation of carbonate minerals in order to fix CO2 in a solid form (eq 1).1,4 ðMg, CaÞx Siy Oxþ2yþz H2z ðsÞ þ xCO2 ðgÞ f xðMg, CaÞCO3 ðsÞ
ð1Þ
þ ySiO2 ðsÞ þ zH2 O
Disposal sites for carbonate minerals require minimal monitoring because of the stability and environmental safety of these minerals.4 An advantageous rock type for mineral carbonation is serpentinite, given its great abundance, wide availability, and high MgO content.1,5 The rate of mineral carbonation in nature is slow, and therefore, there is a need to accelerate this process in order to use it as a viable option for CO2 sequestration. Research on mineral carbonation has mainly focused on industrial r 2011 American Chemical Society
processes involving aqueous carbonation.5 This involves leaching or dissolution of silicate minerals, commonly using acids, and subsequent precipitation of carbonate minerals under high pressure and temperature.1,4 Mineral acids, organic acids and ligands, and caustic soda have been used to leach serpentinite for the purpose of mineral carbonation.1,46 With current technology, these processes are too expensive for CO2 sequestration because of the high costs of mineral processing (e.g., mining and pulverization), chemical additives, and purification of CO2.4 The costs of mineral processing can be circumvented by using industrial waste products such as cement kiln dust, construction waste, iron and steel slag, fuel ash, and mine wastes as feedstock for mineral carbonation.7 Mining of asbestos, nickel, diamonds, talc, and chromium produces large quantities of ultramafic tailings as a waste product.4 Wilson et al. have documented passive mineral carbonation occurring in ultramafic tailings at historic asbestos mines as being facilitated by their high surface area.8 The Received: May 14, 2011 Accepted: August 31, 2011 Revised: August 30, 2011 Published: August 31, 2011 9061
dx.doi.org/10.1021/es201648g | Environ. Sci. Technol. 2011, 45, 9061–9068
Environmental Science & Technology carbonate minerals in these tailings formed under normal mining conditions, that is, no practices were implemented to accelerate CO2 sequestration. In the present study, ultramafic tailings are used as feedstocks from two active mines, the Diavik Diamond Mine in the Northwest Territories, Canada and Mount Keith Nickel Mine in Western Australia, which both host large stockpiles of tailings. In regards to emissions, the Diavik mine directly emitted ∼162 000 tonnes of CO2 equivalent of greenhouse gases in 2006 and in 2004, the Mount Keith mine emitted ∼382 000 tonnes of CO2 equivalent.9,10 These two mines produce approximately 2 Mt and 11 Mt of tailings per year, respectively.11,12 On an annual basis, complete conversion of these tailings to form magnesium carbonate minerals would allow for these mines to become zero emission facilities and carbon sinks for additional sources of CO2. However, the bulk of the tailings are unweathered being composed of noncarbonated, silicate minerals. Therefore, there is the potential to accelerate mineral carbonation at these sites. In nature, a variety of microorganisms can alter water chemistry in a manner that favors carbonate precipitation.1318 In particular, consortia with both phototrophs (e.g., cyanobacteria and microalgae) and heterotrophs have been shown to have the ability to mediate carbonate precipitation in various environments.15,19,20 In addition, microalgae and cyanobacteria have received much attention for biotechnological applications including biofuel and nutraceutical production. These phototrophs are a very attractive feedstock for biodiesel, bio-oil, or bioethanol production because they can be grown on nonarable land such as mine sites. Many species are able to grow in saline and subsaline waters and their photosynthetic efficiencies can be several times greater than terrestrial plants leading to greater biomass production and CO2 fixation per unit area.21 Coupling carbonate and biomass production in an environment that is suitable for both of these processes would be a novel means of sequestering anthropogenic CO2 and generating valuable byproducts. In this study, we investigated combining microbial carbonate precipitation and biomass production for sequestering CO2. The objectives were to (1) evaluate the suitability of a tailings facility that receives leachate waters as a habitat for carbonate-forming microbes, (2) elucidate the roles of phototrophs and heterotrophs in forming carbonate minerals and producing biomass from a variety of leachate waters, (3) identify limitations to this combined process, and (4) provide estimations of the CO2 sequestration rates. In experiments, Diavik and Mount Keith tailings were leached using various dilute acids and deionized water to simulate potential leaching methods used to release cations for precipitation of carbonate minerals. Carbonate precipitation from these leachates was evaluated by microscopy, X-ray diffraction (XRD), and geochemical modeling and biofuel production was evaluated by analyzing the biochemical composition of the phototrophic consortium used in the experiments.
’ EXPERIMENTAL SECTION Field Sampling. Fieldwork and sample collection were conducted at the Diavik Diamond Mine and the Mount Keith Nickel Mine (Supporting Information (SI) Figure S1). At Diavik, the majority of tailings are transported as slurry to the Fine Processed Kimberlite Containment (fine PKC) and stored in a pond of process water that is continually cycled. Grab samples of tailings with no visible evidence of secondary carbonate minerals
ARTICLE
were obtained from the fine PKC. At Mount Keith, tailings are transported as a suspension in hypersaline process water to the Tailings Storage Facility 2 (TSF2), which is a centralized discharge facility. Samples were collected with the aid of a backhoe from a depth of approximately 5 m to minimize the presence of evaporative minerals. Detailed descriptions of the Diavik tailings and the geology of the Mount Keith site are provided by others.11,12,22 The tailings used in this study were fully characterized by determining their mineralogical and geochemical compositions, particle size distributions, neutralization potentials, and surface areas. General descriptions of the field sites and analytical methods are provided in the SI. Microbial Carbonate Precipitation Experiments. Leaching of Diavik and Mount Keith tailings was performed using dilute solutions of acetic acid, hydrochloric acid, sulfuric acid, a mixture of nitric and phosphoric acids, and deionized water (SI Table S1). These acids were selected to represent a wide variety of acid sources such as from industrial processes (HCl, H2SO4),4,5 overlying soils (CH3COOH),23 microbial sulfur oxidation (H2SO4),24 and nutrient addition (HNO3 and H3PO4) that could be used to release cations from silicate minerals. The concentrations of CH3COOH, HCl, and H2SO4 used for reaction with the Diavik and Mount Keith tailings were 0.1 and 0.25 N, respectively. Acid concentrations depended upon the ability of each tailings sample to neutralize acidity during the leaching period (2 days) to a pH ≈ 7. Although stronger acids could be used to more aggressively leach the tailings, this would result in (1) acidic leachates that would not be suitable for growth of phototrophs and (2) result in potentially less cost-effective and less technologically viable approach to CO2 sequestration in mine tailings. For HNO3 and H3PO4 leaches, 0.018 and 0.0002 N, respectively, were used to produce similar concentrations of nitrate and phosphate that are present in growth media. Deionized water is representative of current leaching by natural carbonic acid in meteoric precipitation at the tailings facilities. Aliquots (7.0 g) of tailings were added to 125 mL Erlenmeyer flasks with 70 mL of acid solution or deionized water and allowed to react on a shaker table for 2 days. After this reaction period, the pH of each of the mixtures was measured. The mixtures were centrifuged and the supernatant was filtered (0.45 μm). A consortium dominated by unicellular algae, cf. Dunaliella sp., was cultured from microbial mats collected from the hydromagnesite playas of Atlin, British Columbia, Canada. These are natural magnesium carbonate deposits in ophiolitic terrane and are a biogeochemical model for CO2 sequestration via mineral carbonation.25 Cultures were grown in 75 mL of BG-11 media in 125 mL Erlenmeyer flasks at room temperature (∼20 C) and in the presence of sunlight.26 Filtered leachate solutions were used to prepare modified BG11 media. Given that the tailings would provide dissolved Mg, Ca and Na, the following ingredients were excluded from the media: MgSO4 3 7H2O, CaCl2 3 2H2O, and Na2CO3(H2O). In the leachate produced using nitric and phosphoric acids, NaNO3 and K2HPO4 3 3H2O were also excluded. Aliquots of 15 mL were dispensed into 20 mL sterile (autoclaved) glass vials. Media was inoculated with 100 μL of culture and incubated at room temperature and in the presence of sunlight. Abiotic controls containing only filtered leachate waters (15 mL) were also prepared. Microbial growth and pH were monitored weekly for 14 weeks. Cells and precipitates were routinely examined using light microscopy and after 12 weeks samples were collected for electron 9062
dx.doi.org/10.1021/es201648g |Environ. Sci. Technol. 2011, 45, 9061–9068
Environmental Science & Technology
ARTICLE
Table 1. Major Cation Concentrations of the Leachates Produced from Reaction with Diavik and Mount Keith Tailings cation concentration (ppm) at t = 0 Ca
Fe
K
Mg
Na
Ni
P
S
Si
avg. pH, biotic/abiotic during experiment
Diavik acetic acid
529
0.3
50
235
466
0.2
7.0
10
26
9.82/9.48 (Δ = 0.34)
hydrochloric acid
504
0.3
49
263
463
0.2
7.7
12
29
8.64/7.49 (Δ = 1.15)
sulfuric acid
438
0.3
44
264
431
0.3
7.6
785
37
8.79/7.83 (Δ = 0.95)
nitric and phosphoric acids
151
0.3
29
122
10
<0.01
1.7
11
11
9.14/8.10 (Δ = 1.05)
deionized water
6
0.4
35
19
421
<0.01
5.9
12
3.2
10.2/8.54 (Δ = 1.69)
127 125
0.2 0.2
78 78
2952 2969
1544 1502
2.2 0.8
5.3 5.3
404 412
<0.02 5.0
9.22/8.71 (Δ = 0.51) 9.16/7.79 (Δ = 1.37)
Mount Keith acetic acid hydrochloric acid sulfuric acid
121
0.2
85
3024
1552
1.1
6.4
4180
4.4
8.98/7.80 (Δ = 1.19)
nitric and phosphoric acids
46
0.3
65
557
1120
<0.01
<0.1
371
<0.02
9.01/7.98 (Δ = 1.03)
deionized water
53
0.3
79
370
1578
<0.01
5.2
360
<0.02
9.26/8.04 (Δ = 1.22)
microscopy and mineralogical analyses. The SI contains further details on the analytical methods. Biomass Production and Collection. Cultures were grown in 2 L Erlenmeyer flasks with 200 μmol of photosynthetically active radiation (PAR)/m2/s at room temperature (∼20 C) in BG-11 media that was well mixed by bubbling with a low volume of filter sterilized air (∼0.3 v/v/min). Cells were harvested by centrifugation for 15 min at 15 103 g using a Sorvall RC-5B refrigerated superspeed centrifuge and washed three times with 50 mL of distilled water by vortexing for 3 min at 2500 rpm. Cells were lyophilized using a Thermo Electron ModulyoD-115 with Savant Valupump VLP200 in the dark for approximately 40 h at 80 ( 20 Pa until there was no further change in mass. The SI contains methods for the biochemical analyses.
’ RESULTS AND DISCUSSION Mine Tailings and Leachate Chemistry. The Diavik tailings were mainly composed of serpentine minerals (predominantly lizardite, 48.2%), high-Mg forsterite (25.2%), and vermiculite (9.8%) (SI Figure S2). Mount Keith tailings were composed primarily of antigorite and lizardite (combined 81.0%) with hydrotalcite-group minerals, which tend more toward chlorine endmembers than carbonate endmembers (SI Figure S3). Other notable minerals were calcite (3.7%) in the Diavik tailings and magnesite (3.4%), brucite (2.5%), and dolomite (0.6%) in the Mount Keith tailings. The Diavik and Mount Keith tailings had surface areas of 47.99 m2/g and 9.35 m2/g and average grain sizes of 0.1 mm and 0.07 mm (SI Figure S4), respectively. The surface area of the Diavik tailings was considerably higher given the amount of vermiculite, but both tailings have surface areas that are orders of magnitude greater than natural bedrock. The neutralization potentials of the Diavik and Mount Keith tailings were 370 and 352 kg/tonne of CaCO3 equivalent, respectively. Although these values are nearly equal, Mount Keith tailings were able to neutralize more concentrated acid (0.25 N compared to 0.1 N for leaching of Diavik tailings) in experiments given more time (leaching time = 2 days) than what is required for determining neutralization potentials by the Sobek method. This may be due to slower reactions with dolomite or finer grained magnesium silicate minerals found in the Mount Keith tailings.
The ability of a tailings sample to neutralize acidity indicates its reactivity to leaching processes. The geochemical composition of the Diavik tailings mainly consisted of SiO2 (41.1%), MgO (32.5%), Fe2O3 (7.6%), CaO (3.7%), and Al2O3 (3.6%), whereas the Mount Keith tailings consisted primarily of MgO (40.3%), SiO2 (31.1%), and Fe2O3 (6.8%). Complete data for each of the tailings are shown in SI Table S2. The historic chrysotile tailings studied by Wilson et al. are similar to the Diavik and Mount Keith tailings in that they are ultramafic, fine grained with high surface areas; making them susceptible to carbonation. The chrysotile tailings contained hydrated magnesium carbonate minerals that precipitated in situ to store atmospheric CO2.8,27 Precipitation of carbonate minerals within the pore spaces of these tailings resulted from extensive evaporative concentration of dissolved magnesium and inorganic carbon released from the dissolution of magnesium silicate minerals by very weak carbonic acid, that is, meteoric precipitation. In contrast to pore waters, leachate waters from this tailings site may not precipitate carbonate minerals in the absence of evaporative concentration. Subsaline solutions (Table 1) were produced from leaching each of the tailings and represent potential feedstocks for carbonation. The solutions of diluted acetic acid, hydrochloric acid, and sulfuric acid were equally effective at leaching both of the tailings. After the addition of nutrients (e.g., NaNO3), the major cations in the Diavik solutions were Ca > Na > Mg . K > Si and for the Mount Keith solutions they were Mg . Na . Ca > K. As expected, solutions produced by leaching tailings with less concentrated nitric and phosphoric acids and deionized water had much lower cation concentrations. In experiments, precipitation of aragonite and calcite does not represent a net sequestration of CO2 because the only sources of dissolved calcium are carbonate minerals already present in ore prior to mining. Dissolved magnesium would have come from both the dissolution of magnesium silicate and, in the case of Mount Keith, carbonate and hydroxide minerals such as brucite, dolomite and magnesite. Precipitation of magnesite and dolomite is kinetically inhibited at ambient temperatures owing to the strong hydration of Mg2+ ions in solution.28,29 Hydromagnesite may also be kinetically inhibited at ambient temperatures.30 Therefore, the precipitation of kinetically favored magnesium carbonate minerals, such as dypingite [Mg5(CO3)4(OH)2 3 ∼5H2O] and nesquehonite, 9063
dx.doi.org/10.1021/es201648g |Environ. Sci. Technol. 2011, 45, 9061–9068
Environmental Science & Technology
Figure 1. SEM micrographs of cells and precipitates from the experimental systems. A: Diavik acetic acid precipitate showing algal cells embedded (arrow) in a calcium carbonate aggregate; note the heterotrophic bacteria associated with algal cells. B: Cells and dypingite rosettes (arrows) formed in the Mount Keith acetic acid solution. C: An algal cell completely encrusted by fine silicate rosettes from the Diavik deionized water precipitate. D: High magnification view of silicate precipitates on an algal cell from the Mount Keith sulfuric acid precipitate.
represents the best opportunity for sequestering CO2. Although these minerals are relatively less stable and more soluble than magnesite,30 they are still capable of storing CO2 on geological time scales. For example, nesquehonite and dypingite precipitate from water in a wetland that is part of a carbonate playa near Atlin, British Columbia, Canada.25 These minerals dehydrate to form hydromagnesite, the most abundant carbonate mineral present in the playas, which are underlain by glacial till and presumably began forming soon after the last deglaciation demonstrating their stability over the Holocene epoch. The minor relief of the playas above the surrounding land further demonstrates that hydrated magnesium carbonate minerals are stable when present at near-surface conditions, especially in an alkaline environment such as an ultramafic tailings facility. In the present study, we evaluated the ability of the Atlin consortium to induce carbonate precipitation from leachate solutions that were initially undersaturated with respect to kinetically favorable carbonate minerals. Carbonate Precipitation Experiments. The Altin consortium, dominated by cf. Dunaliella sp., was able to grow in all of the leachate solutions and cause alkalinization of these waters. The presence of relatively high concentrations of chlorine (HCl leach) or organics (CH3COOH leach) did not impede growth. During the experiment, approximately 70% of the water was evaporated from the vials. In the biotic systems, the Atlin consortium was able to raise the pH value significantly in all the systems, typically by one pH unit, when compared to the abiotic controls (SI Figure S5). Precipitation of abundant carbonate minerals occurred in solutions produced from reaction of tailings with dilute acetic acid. When examined using scanning electron microscopy (SEM), cells were seen embedded on the outside of large crystal aggregates and rod-shaped heterotrophic bacteria were associated with algal cells (Figure 1A). The Mount Keith acetic acid precipitates had rosette morphology and magnesium carbonate composition based on energy dispersive X-ray spectroscopy
ARTICLE
(EDS) (Figure 1B). Rosette precipitates of magnesium carbonate minerals, although less common, were also seen in the Diavik acetic acid precipitate. The Diavik precipitate mainly consisted of magnesium-containing calcite and minor amounts of partially dehydrated dypingite, whereas the Mount Keith precipitate mainly consisted of dypingite and halite that formed when the sample was dried (XRD patterns in SI Figure S6). All of the Diavik solutions produced noticeable calcium carbonate over the course of the experiment. The calcium carbonate was seen as translucent crystal aggregates when viewed using light microscopy. Algal cells were entombed in the calcium carbonate along with casts of cells that had been infilled. In addition, precipitates with a silicate composition were seen directly on cells (Figure 1C). Abundant amounts of magnesium carbonate minerals did not precipitate from the Mount Keith solutions with the exception of the acetic acid solution. Similar fine grained silicate precipitates were seen in association with cells, but more commonly directly on cells with numerous cells being completely encrusted (Figure 1D). These precipitates had rosette morphology and were typically less than 0.5 μm in diameter with a magnesium and calcium silicate composition. Aragonite precipitated in some solutions after extensive evaporation. Several example micrographs of the precipitates are given in SI Figures S7 and S8. Biomass Production. The approximate composition of the Atlin consortium showed a protein content of 44 ( 2 wt.%, which is higher than in most species of photosynthetic unicellular organisms and indicates that this consortium is a potential source of single-cell protein,31 which could be hydrolyzed and used in fermentations as a nitrogen source, along with any carbohydrates in the biomass. The value of the proteins would depend on the amino acid composition, which was not determined. Carbohydrate content was relatively low (22 ( 1 wt.%), but at least twice the lipid content (11 ( 1 wt.%), suggesting that carbohydrates may be the preferred energy-storage medium for this culture.31 Ash content was relatively high at 12 ( 2 wt.%. The combination of inorganics inside and outside the cells and any reduced carbon amounts to the total ash measured; however, the composition of this material was not determined. The biomass was relatively high in pigments with a low chlorophyll a (0.81 ( 0.03 wt.%) to chlorophyll b (0.32 ( 0.02 wt.%) ratio of 2.5:1 and total chlorophyll exceeding 1 wt.% of the dry biomass. This indicates acclimation to low light,32 which is consistent with the light level used; 10% of full-strength sunlight at the Earth’s surface. Carotenoid pigments were 0.045 ( 0.002 wt.% of the biomass. Acid hydrolysis of cell biomass and excreted saccharides in the growth media (SI Table S3) showed that the total hydrolyzable carbohydrate content was 21 ( 1 wt.%, which agreed with the carbohydrate content measured by the phenol sulfuric acid method, 22 ( 1 wt.%, within experimental error (n = 3; P < 0.05). The monosaccharides in the hydrolysate of the cell biomass were 53 wt.% hexoses that are easily fermented by yeast and 47 wt.% pentoses, which cannot be fermented by standard yeast but can be fermented by other microorganisms and genetically engineered yeast.33 The growth media contained a relatively high concentration of hydrolyzable carbohydrates, 160 mg/L or about 15 wt.% of the biomass weight. This hydrolysate contained 41 wt.% hexoses and 59 wt.% pentoses. Although the concentration of excreted saccharides is very low compared to the concentration of carbon source in a typical fermentation broth it would be possible to ferment these sugars to produce ethanol or butanol.34 9064
dx.doi.org/10.1021/es201648g |Environ. Sci. Technol. 2011, 45, 9061–9068
Environmental Science & Technology The fatty acid content measured by gas chromatography (SI Figure S9) was significantly lower than the lipid content measured gravimetrically; 7.6 ( 0.3 wt.% versus 11 ( 1 wt.%, respectively (n = 3; P > 0.05). However most of the lipids in microalgae are often polar lipids, especially when the lipid content is low.31 The extra weight of the nonfatty acid portions of phospholipids and other nonfatty acid lipids such as pigments may be the reason for this discrepancy. The most abundant fatty acid in the cells was α-linolenic acid, which accounted for about 37 wt.% of the total fatty acids and over 95 wt.% of the ω-3 fatty acids. Although α-linolenic acid is an essential nutrient it is not very valuable because it is present in high amounts in common foods such as flax, hemp, canola, and soya and does not have the same health benefits as eicosapentaenoic acid or docosahexaenoic acid.35 Approximately 40 wt.% of the fatty acids had three units of unsaturation; ∼13 wt.% had two units of unsaturation; ∼20 wt.% had one unit of unsaturation; and ∼27 wt.% were saturated. Since only the minority of fatty acids had chain lengths over eighteen or more than three double bonds the oil from this species of algae should be more suitable for biodiesel than the oil of other species such as Nannochloropsis sp., which is very rich in eicosapentaenoic acid.31 Although the lipid content was relatively low at about 11 ( 1 wt.% this was determined for late-exponential phase biomass in nutrient-sufficient cultures. It is often possible to increase lipid or carbohydrate content dramatically, often by several times, by allowing the biomass to accumulate energy storage metabolites in a second-phase of nitrate or phosphate limited growth.31 Therefore, it may be possible to increase the oil and/or carbohydrate content of this consortium for biodiesel and/or fermentation carbon source production. If this could be achieved this consortium would be ideal for biofuel applications since it can be grown with minimal monetary and energy cost for growth media, cultivation, and harvesting from existing mine tailing ponds, while simultaneously sequestering CO2 as biomass. Roles of Phototrophy and Heterotrophy in Carbonate Precipitation. Field observations at Diavik suggest that mineral carbonation under some conditions for tailings disposal is limited to abiotically formed evaporative crusts (SI Figure S1C,D). The tailings at both sites are largely unweathered and therefore have tremendous potential for further carbon sequestration. In the present study, a variety of acids were used to simulate possible accelerated dissolution scenarios, but natural weathering (simulated using deionized water), although slowest, may be the most practical. Carbonate precipitation will occur if pore waters in tailings or leachate waters are sufficiently supersaturated with respect to a given carbonate mineral; however, leachate waters may often be undersaturated with respect to kinetically favorable magnesium carbonate minerals. Phototrophic and heterotrophic microbes are able to alter the geochemical conditions to cause leachate waters to become sufficiently supersaturated. The three important abiotic factors in carbonate precipitation are pH and the concentrations of cations (e.g., Ca2+ and Mg2+) and dissolved inorganic carbon (DIC). Experimental leaching of the ultramafic tailings produced subsaline solutions rich in Mg2+ and Ca2+ ions that were then available for carbonate precipitation. Geochemical modeling of the starting solutions (Table 1) using PHREEQC showed that increasing the pH value would generally result in these solutions becoming supersaturated with respect to hydromagnesite.36 Although dypingite has been precipitated in magnesium-rich solutions in both abiotic37 and biotic18 experiments, there is no thermodynamic data available for modeling its
ARTICLE
precipitation. When leachate solutions were inoculated with the Atlin consortium, phototrophs produced OH ions during photosynthesis while forming biomass or producing energy (eq 2).13,38 HCO3 þ H2 O þ hv f CH2 O þ OH þ O2 v
ð2Þ
This release of OH ions outside the cell results in alkalinization, that is, an increase in the pH value, which in turn increases the concentration of CO32- anions. This effect is seen in the elevated pH values of the biotic systems (avg. pH 9.22), which were generally one pH unit above corresponding abiotic controls (avg. pH 8.17) (Table 1 and SI Figure S5). This can greatly increase the saturation of carbonate minerals (e.g., aragonite and dypingite), allowing for their precipitation (eq 3 and 4). Ca2þ þ CO3 2 f CaCO3
ð3Þ
5Mg2þ þ 4CO3 2- þ 2OH þ 5H2 O f Mg5 ðCO3 Þ4 ðOHÞ2 3 5H2 O
ð4Þ
Heterotrophic bacteria are often associated with phototrophs in natural systems and can increase DIC concentrations, such as the acetate oxidizing, heterotrophic bacteria in this study. In natural systems, this may occur by either aerobic or anaerobic (e.g., sulfate reduction) oxidation of organics (eq 5 and 6). 2CHO2 þ O2 f 2HCO3
ð5Þ
4CHO2 þ SO4 2 þ Hþ f 4HCO3 þ HS
ð6Þ
The decomposition of organic matter by heterotrophic bacteria is part of the cycling of carbon that is often first fixed by phototrophs. The acetic acid leachate solutions allowed for the growth of aerobic, acetate oxidizing bacteria that were part of the Atlin consortium. Oxidation of acetate (CH3COO) produces CO2; acting as an alternative pathway for CO2 delivery in the experiments (eq 7). CH3 COO þ 2O2 f 2CO2 þ OH þ H2 O
ð7Þ
Acetate oxidation also increases the pH value through production of OH, whereas the CO2 produced may then dissolve into solution forming CO32- at alkaline pH. Similarly, von Knorre and Krumbein conducted calcium carbonate precipitation experiments by growing heterotrophic bacteria on agar plates in an artificial seawater medium containing 15 mM sodium acetate.20 They found that 126 of the 128 isolates tested were able to induce carbonate precipitation, which they concluded was mainly due to physiological activities that increased carbonate alkalinity. As well, Mg carbonate and Mg-rich carbonate minerals have been precipitated with the aid of aerobic heterotrophs in batch experiments.39 An important observation of these microcosm experiments was the formation of magnesium and calcium silicate minerals on cell surfaces. This limits CO2 sequestration as it removes cations from solution as precipitates that are not carbonate minerals. This observation indicates that although experiments were open to the atmosphere, uptake of CO2 from the atmosphere was ratelimiting in the microcosms and was insufficient to allow for abundant precipitation of magnesium carbonate minerals even though solutions had high concentrations of magnesium and alkaline pH values as a result of alkalinization by phototrophs. The acetic acid systems were an exception because the oxidation of acetate acted as an alternate means of delivering CO2. 9065
dx.doi.org/10.1021/es201648g |Environ. Sci. Technol. 2011, 45, 9061–9068
Environmental Science & Technology
ARTICLE
Table 2. Estimated CO2 Sequestration Rates (Tonnes/Year) Based on Annual Tailings Production values are in
mineralized CO2 dissolved
CO2 mineralized as carbonate
CO2 captured as
total CO2 sequestered
tonnes/year
during leach
minerals (ara/hmg)
biomass
(% in carbonate/% in biomass)
Diavik (2 Mt tailings/year; emissions = 162 000 t CO2 equivalent/year) acetic acid
1160
1290
711
422 from the PKC
4000
21%
79%
hydrochloric acid
1110
1200
0
2740 from north inlet
3250
3%
97%
storage reservoir
3250
3%
97%
3410
7%
93%
3190
1%
99%
63 300
83%
17%
62 000
83%
17%
59 100
82%
18%
19 000
44%
56%
17 100
38%
62%
sulfuric acid
962
1050
0
nitric and phosphoric acids
332
363
214
deionized water
13
14
22
growing season: 12 weeks
Mount Keith (11 Mt tailings/year; emissions = 382 000 t CO2 equivalent/year) acetic acid 1534 1710 52 500 hydrochloric acid
1510
1700
51 200
sulfuric acid
1460
1580
48 400
nitric and phosphoric acids
556
596
8390
deionized water
640
696
6460
Precipitation of dypingite from natural waters, sampled near Atlin, British Columbia, has been achieved using a phototrophic consortium; although these waters had very high alkalinities (e.g., 2661 and 4727 mg HCO3/L).18 Wilson et al. also noted that uptake of atmospheric CO2 into brine solutions (2430 mg Mg/ L) was rate-limiting for the precipitation of dypingite with initial precipitation requiring 15 days.37 It was also noted that carbonate mineral precipitation outpaced uptake of CO2 gas. Additional carbon sources or more proficient means of delivering CO2 are likely required for abundant precipitation of magnesium carbonate minerals. This differs from calcium carbonate precipitation given that the formation of magnesite is kinetically slow because of the strong hydration of Mg2+ ions.28 Ferrini et al. synthesized nesquehonite at ambient temperatures by sparging CO2 through a concentrated magnesium chloride solution at room temperature and adjusting the pH with ammonia.40 They proposed that this process could be developed for sequestering CO2 from point sources using waste brine solutions. Alternatively, increasing the concentration of DIC could be achieved using natural biocatalysts.41 For example, carbonic anhydrase is a metalloenzyme that is produced by both prokaryotes and eukaryotes, which catalyzes the hydration of CO2, that is, the dissolution of CO2 in an aqueous form.40,42,43 Mitchell et al. investigated bacterial ureolysis in synthetic brine solutions, which also increases carbonate alkalinity and pH to induce calcium carbonate precipitation.42 They noted that wastewater urea sources could be utilized and that this would reduce labile carbon, thereby preventing the release of CO2 by forming carbonate minerals. Similarly, a net sequestration of CO2 can be achieved if the CO2 from the oxidization of organic carbon was first fixed (from the atmosphere) by phototrophs before being precipitated as carbonate minerals. Carbon Sequestration Strategies. Mining is an ongoing, profitable industrial activity that produces tailings with an inherent ability to sequester CO2 through a natural weathering process. Precipitation of carbonate minerals from leachate solutions rich in magnesium and calcium is advantageous in that carbonate minerals are both benign and stable. An ideal scenario would be to have specially designed carbonate precipitation ponds with leachate waters being directed into these ponds from tailings piles. A means of accelerated leaching could be implemented post ore processing or at the tailings facility if it were deemed cost-effective. In the pond, a microbial consortium
10 600 from TSF2 growing season: 52 weeks
would be cultivated to induce precipitation of carbonate minerals with microbial species being selected based on their ability to mediate carbonate precipitation and the quality of their biomass. Additional nutrients and evaporative conditions would help create ideal conditions. Carbonate minerals would be deposited in these ponds while biomass would be harvested for potential use as biofuel or other valuable byproducts. In the experiments, CO2 was derived from uptake from the atmosphere, dissolution of tailings carbonate minerals, and microbial oxidation of acetate. Sinks of CO2 included precipitation of secondary carbonate minerals and formation of biomass. CO2 sequestration rates were calculated using the experimental data and have been scaled to the annual tailings production for Diavik (2 Mt/yr) and Mount Keith (11 Mt/yr) (Table 2). Tailings slurries were assumed to be approximately 50% solids by mass, thereby requiring 2 GL and 11 GL of water per year for Diavik and Mount Keith, respectively. This assumes that tailings are leached only during the year that they are produced; whereas leaching of the entire tailings storage facilities would access approximately 1 order of magnitude more material. The calculated values of mineralized CO2 dissolved from the acid leaching were based on the Ca concentrations in the starting solutions assuming partial dissolution of calcite and dolomite in the Diavik and Mount Keith tailings. Magnesite, present in the Mount Keith tailings, was assumed unreactive as magnesite has been shown to be highly resistant to dilute acids.43 PHREEQC modeling of the starting solutions assumed a modest evaporation of 10% and used the average pH value obtained during each of the biotic experiments (Table 1). The pH of the solutions was fixed in order to mimic experimental conditions. These conditions were generally obtained after two weeks and coincided with reaching the stationary growth phase. Aragonite and hydromagnesite (a proxy for dypingite) were allowed to precipitate and solutions were in equilibrium with atmospheric concentrations of CO2. Net CO2 sequestered per year as carbonate minerals for the Diavik site ranged from 23 to 841 tonnes/year (deionized water to acetic acid system), whereas for Mount Keith rates ranged from 6520 to 52 700 tonnes/year (deionized water to acetic acid system) (Table 2). Although the reacted tailings were not processed further in this study, they would still be a valuable feedstock for further leaching. The acidity of the leaching solution would need to be adjusted because primary carbonate minerals are present only in fresh tailings. Use of carbonate-free 9066
dx.doi.org/10.1021/es201648g |Environ. Sci. Technol. 2011, 45, 9061–9068
Environmental Science & Technology tailings or continual leaching of carbonate-bearing tailings is preferred in comparison to the use of fresh tailings that have primary carbonate minerals. Carbonate precipitation would benefit from additional carbon sources, such as waste organics (simulated by acetic acids in experiments) and waste gas streams, or improved means of delivering CO2 to the system as opposed to relying on uptake of CO2 from the atmosphere. Carbonate precipitation at Mount Keith was comparably greater than Diavik given that leachate solutions contained greater concentrations of cations. In cultures, biomass production was 1 g of dry biomass/L over 2 weeks to obtain stationary growth. We have used a conservative growth rate of half this amount (0.5 g/L) assuming nonideal conditions and accounted that 1.83 g of CO2 is required to produce 1 g of dry biomass.21 Biomass production will depend on the site’s climate and the availability of water and nutrients (e.g., wastewater). In addition to the fine Processed Kimberlite Containment (PKC) at Diavik there is the North Inlet Water Treatment Plant, which is a reservoir in that 13 GL of water passes through each year.44 This water is not likely suitable for carbonate precipitation as it is not associated with tailings, but may be suitable for biomass production as algal blooms have been observed at this site. Based on climate data we have estimated that suitable growing conditions at Diavik and Mount Keith are 12 and 52 weeks per year, respectively, and have therefore scaled the water available for biomass production based on the fraction of the year in which suitable growing conditions are present.45,46 On an annual basis, CO2 sequestered as biomass was estimated to be 422 tonnes from the PKC and 2740 tonnes from the North Inlet treatment reservoir at Diavik and 10 600 tonnes from the Tailings Storage Facility 2 (TSF2) at Mount Keith. The Diavik and Mount Keith mines emit approximately 162 000 and 382 000 tonnes of CO2 equivalent in the production of 2 Mt and 11 Mt of mine tailings per year, respectively. Mines are industrialized sites that provide nonarable land, large water reservoirs (e.g., tailings ponds and water treatment facilities), and abundant fine grained mine wastes that have an inherent ability to sequester CO2. These sites could be modified to accelerate leaching of these tailings and better suit the cultivation of phototrophic microbes that are able to mediate carbonate precipitation and produce valuable biomass. From microcosm experiments, estimated sequestration rates indicate that 2.5% and 16.6% of Diavik’s and Mount Keith’s yearly emissions, respectively, could be sequestered. Carbonate precipitation rates will mainly depend on the availability of cations and the rate of CO2 uptake into solutions. Heterotrophic oxidation of acetate demonstrated an alternative, effective means of accelerating CO2 uptake in comparison to uptake of CO2 from the atmosphere; however, it is necessary to account for the source of CO2 in any carbon sequestration strategy. Rather than acetate oxidation, wastes organics or injection of waste gas streams with elevated partial pressures of CO2 could provide a more efficient means of CO2 uptake. Carbonate mineral precipitation coupled with biomass production has the potential to provide a low-energy, lowtemperature, and relatively passive means of sequestering CO2 at mine sites and possibly in other human-made environments.
’ ASSOCIATED CONTENT
bS Supporting Information. Details for the field sites, experimental methods, and additional figures for this study are provided.
ARTICLE
This material is available free of charge via the Internet at http:// pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 1-604-822-1929; fax 1-604-822-6088
[email protected]. Present Addresses ^
Mineral Deposit Research Unit, Department of Earth and Ocean Sciences, The University of British Columbia, Vancouver, British Columbia V6T 1Z4, Canada. # School of Geosciences, Monash University, Clayton, VIC 3800, Australia.
’ ACKNOWLEDGMENT We acknowledge the support of the Natural Sciences and Engineering Research Council of Canada (NSERC), BHP Billiton, and Diavik Diamond Mines Inc. through a Collaborative Research and Development grant and a Carbon Management Canada grant to G.M. Dipple and G. Southam. Additional support to I.M. Power was provided by an NSERC PostGraduate Scholarship and NSERC Alexander Graham Bell Canada Graduate Scholarships supported work by S.A. Wilson and D.P. Small. We thank Colleen English and her staff, Scott Wytrychowski and Dan Guigon of Diavik Diamond Mine Inc. and Josh Levett, John Tomich, and Ben Grguric of BHP Billiton. We thank Ameer Taha of Certo Labs Inc. and Ken Stark and Flaviu Ciobanu of University of Waterloo for fatty acid quantification. We also thank the associate editor and three reviewers for their comments and suggestions that helped to improve this manuscript. This is publication 276 of the Mineral Deposit Research Unit. ’ REFERENCES (1) Alexander, G.; Maroto-Valer, M. M.; Gafarova-Aksoy, P. Evaluation of reaction variables in the dissolution of serpentine for mineral carbonation. Fuel 2007, 86, 273–281. (2) Hoffert, M. I.; Caldeira, K.; Benford, G.; Criswell, D. R.; Green, C.; Herzog, H.; Jain, A. K.; Kheshgi, H. S.; Lackner, K. S.; Lewis, J. S.; Lightfoot, H. D.; Manheimer, W.; Mankins, J. C.; Mauel, M. E.; Perkins, L. J.; Schlesinger, M. E.; Volk, T.; Wigley, T. M. L. Advanced technology paths to global climate stability: Energy for a greenhouse planet. Science 2002, 298 (5595), 981–987. (3) Metz, B.; Davidson, O.; de Coninck, H. C.; Loos, M.; Meyer, L. A. IPCC, 2005: IPCC Special Report on Carbon Dioxide Capture and Storage. Prepared by Working Group III of the Intergovernmental Panel on Climate Change; Cambridge University Press: Cambridge, 2005; p 442. (4) Teir, S.; Revitzer, H.; Eloneva, S.; Fogelholm, C. J.; Zevenhoven, R. Dissolution of natural serpentinite in mineral and organic acids. Int. J. Miner. Process. 2007, 83, 36–46. (5) Lackner, K. S.; Wendt, C. H.; Butt, D. P.; Joyce, E. L., Jr.; Sharp, D. H. Carbon dioxide disposal in carbonate minerals. Energy 1995, 20 (11), 1153–1170. (6) Maroto-Valer, M. M.; Fauth, D. J.; Kuchta, M. E.; Zhang, Y.; Andresen, J. M. Activation of magnesium rich minerals as carbonation feedstock materials for CO2 sequestration. Fuel Process. Technol. 2005, 86, 1627–1645. (7) Renforth, P.; Washbourne, C. L.; Taylder, J.; Manning, D. A. C. Silicate production and availability for mineral carbonation. Environ. Sci. Technol. 2011, 45 (6), 2035–2041. (8) Wilson, S. A.; Raudsepp, M.; Dipple, G. M. Verifying and quantifying carbon fixation in minerals from serpentine-rich mine 9067
dx.doi.org/10.1021/es201648g |Environ. Sci. Technol. 2011, 45, 9061–9068
Environmental Science & Technology tailings using the Rietveld method with X-ray powder diffraction data. Am. Mineral. 2006, 91 (89), 1331–1341. (9) Towards Sustainable Mining Progress Report 2007: GHG Emissions and Energy Management Progress Report; Mining Association of Canada, 2007; http://www.mining.ca/www/Towards_Sustaining_Mining/Technical_Data_and_Bulletins.php (accessed June 15, 2009). (10) Mt Keith Nickel Operations: Environmental Data. BHP Billiton Sustainable Development Reports, BHP Billiton, 2005; http:// hsecreport.bhpbilliton.com/wmc/2004/performance/mko/data/ index.htm (accessed July 10, 2011). (11) Baker, M. J.; Blowes, D. W.; Logsdon, M. J.; Jambor, J. L. Environmental geochemistry of kimberlite materials: Diavik diamonds project, Lac de Gras, Northwest Territories, Canada. Explor. Min. Geol. 2003, 10, 155–163. (12) Grguric, B. A. Minerals of the MKD5 nickel deposit, Mount Keith, Western Australia. Aust. J. Mineral. 2003, 9 (2), 55–71. (13) Thompson, J. B.; Ferris, F. G. Cyanobacterial precipitation of gypsum, calcite, and magnesite from natural alkaline lake water. Geology 1990, 18 (10), 995–998. (14) Warthmann, R.; van Lith, Y.; Vasconcelos, C.; McKenzie, J. A.; Karpoff, A. M. Bacterially induced dolomite precipitation in anoxic culture experiments. Geology 2000, 28, 1091–1094. (15) Riding, R. Microbial carbonates: The geological record of calcified bacterial-algal mats and biofilms. Sedimentology 2000, 47 (Suppl. 1), 179–214. (16) Wright, D. T. The role of sulphate-reducing bacteria and cyanobacteria in dolomite formation in distal ephemeral lakes of the Coorong region, South Australia. Sediment. Geol. 1999, 126, 147–157. (17) Roberts, J. A.; Bennett, P. C.; Gonzalez, L. A.; Macpherson, G. L.; Milliken, K. L. Microbial precipitation of dolomite in methanogenic groundwater. Geology 2004, 32 (4), 277–280. (18) Power, I. M.; Wilson, S. A.; Thom, J. M.; Dipple, G. M.; Southam, G. Biologically induced mineralization of dypingite by cyanobacteria from an alkaline wetland near Atlin, British Columbia, Canada. Geochem. Trans. 2007, 8, 13. (19) Rodriguez-Navarro, C.; Jimenez-Lopez, C.; Rodriguez-Navarro, A.; Gonzalez-Mu~noz, M. T.; Rodriguez-Gallego, M. Bacterially mediated mineralization of vaterite. Geochim. Cosmochim. Acta 2007, 71 (5), 1197–1213. (20) von Knorre, H.; Krumbein, W. E. Bacterial calcification. In Microbial Sediments; Riding, R. E., Awramik, S. M., Eds.; Springer: Berlin, 2000; p 331. (21) Chisti, Y. Biodiesel from microalgae. Biotechnol. Adv. 2007, 25 (3), 294–306. (22) Wilson, S. A.; Raudsepp, M.; Dipple, G. M. Quantifying carbon fixation in trace minerals from processed kimberlite: A comparative study of quantitative methods using X-ray powder diffraction data with applications to the Diavik Diamond Mine, Northwest Territories, Canada. Appl. Geochem. 2009, 24 (2), 2312–2331. (23) Bennett, P. C.; Rogers, J. R.; Choi, W. J. Silicates, silicate weathering, and microbial ecology. Geomicrobiology Journal 2001, 18 (1), 3–19. (24) Power, I. M.; Dipple, G. M.; Southam, G. Bioleaching of ultramafic tailings by Acidithiobacillus spp. for CO2 sequestration. Environ. Sci. Technol. 2010, 44, 456–462. (25) Power, I. M.; Wilson, S. A.; Thom, J. M.; Dipple, G. M.; Gabites, J. E.; Southam, G. The hydromagnesite playas of Atlin, British Columbia, Canada: A biogeochemical model for CO2 sequestration. Chem. Geol. 2009, 206, 302316. (26) Stanier, R. Y.; Kunisawa, R.; Mandel, M.; Cohenbaz., G Purification and properties of unicellular blue-green algae (order Chroococcales). Bacteriol. Rev. 1971, 35 (2), 171–205. (27) Wilson, S. A.; Dipple, G. M.; Power, I. M.; Thom, J. M.; Anderson, R. G.; Raudsepp, M.; Gabites, J. E.; Southam, G. Carbon dioxide fixation within mine wastes of ultramafic-hosted ore deposits: Examples from the Clinton Creek and Cassiar chrysotile deposits, Canada. Econ. Geol. 2009, 104 (1), 95–112.
ARTICLE
(28) H€anchen, M.; Prigiobbe, V.; Baciocchi, R.; Mazzotti, M. Precipitation in the Mg-carbonate system—Effects of temperature and CO2 pressure. Chem. Eng. Sci. 2008, 63 (4), 1012–1028. (29) Land, L. S. Failure to precipitate dolomite at 25C from dilute solution despite 1000-fold oversaturation after 32 years. Aquat. Geochem. 1998, 4, 361–368. (30) K€ onigsberger, E.; K€onigsberger, L.; Gamsjager, H. Low-temperature thermodynamic model for the system Na2CO3-MgCO3CaCO3-H2O. Geochim. Cosmochim. Acta 1999, 63, 3105–3119. (31) Williams, P. J. L.; Laurens, L. M. L. Microalgae as biodiesel & biomass feedstocks: Review & analysis of the biochemistry, energetics & economics. Energy Environ. Sci. 2010, 3 (5), 554–590. (32) Ensminger, I.; Busch, F.; Huner, N. P. A. Photostasis and cold acclimation: Sensing low temperature through photosynthesis. Physiol. Plant. 2006, 126 (1), 28–44. (33) Hasunuma, T.; Sanda, T.; Yamada, R.; Yoshimura, K.; Ishii, J.; Kondo, A., Metabolic pathway engineering based on metabolomics confers acetic and formic acid tolerance to a recombinant xylosefermenting strain of Saccharomyces cerevisiae. Microb. Cell Fact. 2011, 10. (34) Antoni, D.; Zverlov, V. V.; Schwarz, W. H. Biofuels from microbes. Appl. Microbiol. Biotechnol. 2007, 77 (1), 23–35. (35) Williams, C. M.; Burdge, G. Long-chain n-3 PUFA: plant v. marine sources. Proc. Nutr. Soc. 2006, 65 (1), 42–50. (36) Parkhurst, D. L.; Appelo, C. A. J. User’s Guide to PHREEQC (Version 2) a Computer Program for Speciation, Batch Reaction, OneDimensional Transport, And Inverse Geochemical Calculations, WaterResources Investigations Report 99-4259; U.S. Geological Survey: Washington, DC, 1999; p 312. (37) Wilson, S. A.; Barker, S. L. L.; Dipple, G. M.; Atudorei, V. Isotopic disequilibrium during uptake of atmospheric CO2 into mine process waters: Implications for CO2 sequestration. Environ. Sci. Technol. 2010, 44, 9522–9529. (38) Riding, R. Cyanobacterial calcification, carbon dioxide concentrating mechanisms, and ProterozoicCambrian changes in atmospheric composition. Geobiology 2006, 4, 299–316. (39) Sanchez-Roman, M.; Romanek, C. S.; Fernandez-Remolar, D. C.; Sanchez-Navas, A.; McKenzie, J. A.; Pibernat, R. A.; Vasconcelos, C. Aerobic biomineralization of Mg-rich carbonates: Implications for natural environments. Chem. Geol. 2011, 281 (34), 143–150. (40) Ferrini, V.; De Vito, C.; Mignardi, S. Synthesis of nesquehonite by reaction of gaseous CO2 with Mg chloride solution: Its potential role in the sequestration of carbon dioxide. J. Hazard. Mater. 2009, 168 (23), 832–837. (41) Lee, S. W.; Park, S. B.; Jeong, S. K.; Lim, K. S.; Lee, S. H.; Trachtenberg, M. C. On carbon dioxide storage based on biomineralization strategies. Micron 2010, 41 (4), 273–282. (42) Mitchell, A. C.; Dideriksen, K.; Spangler, L. H.; Cunningham, A. B.; Gerlach, R. Microbially enhanced carbon capture and storage by mineral-trapping and solubility-trapping. Environ. Sci. Technol. 2010, 44 (13), 5270–5276. (43) Alaasm, I. S.; Taylor, B. E.; South, B. Stable isotope analysis of multiple carbonate samples using selective acid-extraction. Chem. Geol. 1990, 80 (2), 119–125. (44) Rio Tinto: Diavik Diamond Mine Sustainable Development Report. 2010. (45) Environment Canada. National Climate Archive, climate. weatheroffice.ec.gc.ca (accessed August 17, 2009). (46) Stolberg, D. J. Rehabilitation Studies on Tailings Storage Facilities in an Arid Hypersaline Region; Ph.D. Thesis, The University of Queensland: Brisbane, 2005.
9068
dx.doi.org/10.1021/es201648g |Environ. Sci. Technol. 2011, 45, 9061–9068
ARTICLE pubs.acs.org/est
Impedance Characteristics and Polarization Behavior of a Microbial Fuel Cell in Response to Short-Term Changes in Medium pH Sokhee Jung,†,|| Matthew M. Mench,‡,§ and John M. Regan*,† †
Department of Civil and Environmental Engineering,The Pennsylvania State University, University Park, Pennsylvania 16802, United States ‡ Department of Mechanical, Aerospace, and Biomedical Engineering, University of Tennessee, Knoxville, Tennessee 37966, United States § Energy and Transportation Science Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, United States ABSTRACT: pH oppositely influences anode and cathode performance in microbial fuel cells. The differential electrochemical effects at each electrode and the resultant full-cell performance were analyzed in medium pH from 6.0 to 8.0. Potentials changed 60 mV/pH for the anode and 68 mV/pH for the cathode, coincident with thermodynamic estimations. Open circuit voltage reached a maximum (741 mV) at pH 7, and maximum power density was highest (712 mW/m2) at pH 6.5 as the cathode performance improved at lower pH. Maximum current density increased and apparent half-saturation potential (EKA) decreased with increasing medium pH due to improved anode performance. An equivalent circuit model composed of two time constant processes accurately fit bioanode impedance data. One of these processes was consistently the ratelimiting step for acetate-oxidizing exoelectrogenesis, with its pH-varying charge transfer resistance R2 ranging from 2- to 321-fold higher than the pH-independent charge transfer resistance R1. The associated capacitance C2 was 23 orders of magnitude larger than C1. R2 was lowest near EKA and increased by several orders of magnitude at anode potentials above EKA, while R1 was nearly stable. However, fits deviated slightly at potentials above EKA due to emerging impedance possibly associated with diffusion and excessive potential.
’ INTRODUCTION Microbial fuel cells (MFCs) differ from electrochemical fuel cells in that they use microbes to catalyze reactions on the anode (and on the cathode in biocathode designs) instead of an inorganic catalyst. This requirement for maintaining a viable biocatalyst necessitates physiologically permissive medium conditions. An improved understanding of the catalytic activity of the bioanode as a function of solution conditions would contribute to enhanced MFC operation and performance. However, the electrochemistry of the bioanode is complex due to changing microbial communities and interactions within these communities in the anode biofilm.1 Electrochemical impedance spectroscopy (EIS) coupled with equivalent circuit (EC) analysis24 has been used to elucidate anodic electrochemical reactions in MFCs. Development of anode biofilms significantly decreased anodic polarization resistance (interchangeable with charge transfer resistance), indicating their catalytic role in electron transfer to the anode.58 In a ferricyanide-cathode MFC, initial development of anode biofilm during the first five days decreased anodic polarization resistance by ∼40% (from 2.6 to 1.5 kΩ cm2 at 0.27 A/m2) with a simultaneous increase in power density by ca. 120%.7 In an aircathode MFC, anodic charge transfer resistance decreased by ∼75% (from 0.073 to 0.017 kΩ cm2 at 2.63 A/m2) with increasing anodic capacitance during 70 days of long-term operation.8 r 2011 American Chemical Society
To enable precise impedance analysis on extracellular electron transfer (exoelectrogenesis), pure cultures were also studied. Potentiostatic EIS on a Geobacter sulfurreducens-covered anode showed that charge transfer resistance was the lowest at the imposed potential near the midpoint potential in cyclic voltammetry.6 In their impedance analysis, adding a second time constant through an additional circuit element did not improve data fitting.6 In a subsequent study using Shewanella oneidensis DSP10, three RC time constants were used to differentiate three charge transfer reactions.9 One important controlling parameter for MFC performance is the medium pH. It was found that acidification of the anode biofilm affects current generation because microbial activity is inhibited in low pH.1012 For example, G. sulfurreducens grown using fumarate as an electron acceptor had a considerably lower growth rate at pH 6 (0.04 h1) than pH 7 (0.21 h1),11 and riboflavin production by S. oneidensis DSP10 was reduced significantly at pH 5 with appreciably decreasing power compared with neutral pH.13 It was demonstrated that alkaline medium and high buffer concentration enhances bioanode performance due Received: May 22, 2011 Accepted: September 8, 2011 Revised: September 6, 2011 Published: September 08, 2011 9069
dx.doi.org/10.1021/es201737g | Environ. Sci. Technol. 2011, 45, 9069–9074
Environmental Science & Technology to the increasing flux of proton shuttles out of the anode biofilm.10,14 Increasing the pH from 7 to 9 resulted in a 42% improvement in maximum power density in a cloth-electrode assembly system.14 Maximum current density at pH 8 was 9.9 A/m2-anode, and only ca. 25% of this current was attained at pH 6 in a single-chamber microbial electrolysis cell (MEC).10 Anode biofilm acidification was experimentally substantiated by fluorescent detection of the pH gradient in a pure G. sulfurreducens biofilm.11 On the other hand, low pH enhances cathodic oxygen reduction, presumably due to increased availability of protons to participate in this reaction. It was observed that a lower pH created better performance in terms of current and power density in a granular-carbon cathode,15 and it was also observed in an upflow MFC that lower catholyte pH benefitted cathodic oxygen reduction.16 A 2.5-fold increase of maximum power density was observed with a catholyte of pH 1 compared to that of pH 7.5 in a two-chamber MFC.17 Even in a biocathode, neutralizing the catholyte using strong acid was beneficial for current generation.18 We focused on characterizing the trade-off associated with differential performance effects of pH on each electrode, with acidic medium enhancing cathode performance but inhibiting bioanode performance. In varying pH conditions, impedance elements of the bioanode in a single-chamber MFC were characterized using a new equivalent circuit model. Furthermore, we characterized performance of each electrode and the whole cell using polarization behavior.
’ MATERIALS AND METHODS MFC Construction and Operation. A single-chamber bottle MFC (350 mL) with a carbon paper anode (3.14 cm2 projected area, E-TEK) and an air cathode (7 cm2 of Pt-coated area, 0.5 mg Pt cm2 Pt, and four PTFE diffusion layers) was constructed as previously described.19,20 A Nylon membrane filter with 0.22 μm pores (Magna nylon, N02SP04700) was used to cover the solution-facing side of the cathode to prevent the formation of biofilm directly on the cathode. The small anode size and large medium volume allowed monitoring over long time periods without appreciable changes in the medium condition. A Ag/AgCl reference electrode was located 5 mm from the anode electrode. Anode biofilm enrichment was conducted in 50 mM phosphate-buffered GM medium (pH 7).21 The MFC was inoculated with suspension from an electricity-producing MFC and operated with 10 mM sodium acetate (Sigma-Aldrich, MO) as an electron donor under 600 Ω of external resistance (Rext) at 30 °C for 60 days for anode biofilm enrichment. Electrochemical tests were conducted upon observation of reproducible polarization behavior at a medium pH of 7. To characterize pH effects, GM media with pHs ranging from 6.0 to 8.0 at 0.5 pH unit increments were created with 5 M solutions of HCl or NaOH. The conductivities of these media were kept uniform at 8.9 mS/cm by adjusting with a 5 M solution of NaCl as needed. Electrochemical Analysis. Before each electrochemical measurement, the MFC was operated for 4 h in the fresh medium condition, starting with a medium pH of 8.0 and decreasing incrementally to pH 6.0. Polarization tests, linear sweep voltammetry (LSV), and EIS were performed in separate batches for each pH condition. Between each pH change, the MFC was operated at the initial medium pH of 7 for 12 h. For all pH conditions, the final pH never decreased more than 0.2 pH units from the initial pH. After conducting each of these tests over the pH range, the entire operational procedure was repeated, and
ARTICLE
Figure 1. The proposed equivalent circuit, BIOANODE-1, models exoelectrogenic acetate oxidation occurring in the bioanode.
polarization tests were performed a second time at each pH. Averages and standard deviations of electrode potentials at each pH were calculated based on these replicated polarization tests. Polarization curves were created by decreasing Rext from open circuit to 2 Ω and recording voltage after 30 min at each Rext using a multimeter (Keithley Instruments, OH). LSV and EIS were performed using a potentiostat Reference 600 (Gamry Instrument Inc.). For LSV, the MFC circuit was disconnected for 2 h to create open circuit potential (OCP) for the anode. Then, a scan rate of 0.1 mV/s with a 1-mV step size was applied to the anode over the potential range from OCP to 0.2 V. For potentiostatic EIS, the bioanode was first either poised at each test potential until it produced stable current or disconnected for 2 h to attain OCP. Then EIS was performed at each DC potential with the following conditions: AC potential of 10 mV rms, initial frequency 106 Hz, final frequency 10 mHz, 10 points/decade of data acquisition. Equivalent Circuit Model. Our BIOANODE-1 model (Figure 1) was composed of RC time constants representing two successive reactions of exoelectrogenic acetate oxidation. A constant phase element (CPE) was incorporated to model a nonideal capacitor, defined as 1/Z = T(jω)α. T (S 3 sα) is a numerical value of the admittance (1/|Z|) at ω = 1, where S is siemens and s is second. The exponent α is an empirical nonideality constant, j2 = 1, and ω (s1) is the radial frequency (ω = 2πf).22 Capacitance (C) is calculated using C = (TRp)1/α/Rp, where Rp (Ω) is the charge transfer resistance.22 Impedance spectra were fitted with the BIOANODE-1 model by χ2-minimization using Echem Analyst (Gamry Instrument Inc.). Calculations. The theoretical potentials of the anode and cathode in a specific medium condition were calculated using the Nernst equation according to eqs 14 for the tested pH range at 303 K (30 °C) þ þ 8e T CH3 COO þ 4H2 O 2HCO 3 þ 9H
Ean, M ¼
Eoan
! RT ½CH3 COO ln , E° ¼ 0:187 2 þ 9 8F ½HCO 3 ½H
O2 þ 2H þ þ 2e T H2 O2 Ecat, M ¼
Eocat
! RT ½H2 O2 ln , E° ¼ 0:695 2F ½O2 ½H þ 2
ð1Þ ð2Þ ð3Þ ð4Þ
Eo is the standard potential (at 298 K, 1 bar, 1 M, reported vs standard hydrogen electrode (SHE)), R is the ideal gas constant (8.314 J/Kmol), T is the temperature (K), and F is the Faraday constant (96,485C/mol-eq).23 Potential values were reported 9070
dx.doi.org/10.1021/es201737g |Environ. Sci. Technol. 2011, 45, 9069–9074
Environmental Science & Technology
ARTICLE
Figure 2. Averages and standard deviations of anode potential (Ean, circles) and cathode potential (Ecat, triangles) measured at different external resistances in medium pHs ranging from 6.0 to 8.0 (n = 2 based on replicated polarization tests across the entire pH range). Thick solid lines are theoretical potentials of the anode (bottom) and cathode (top) at approximate experimental conditions (cathode: [H2O2] = 0.01 mM and [O2] = 0.2 atm, anode: [CH3COO] = 10 mM and [HCO3] = 5 mM, 303 K (30 °C)), and thick broken lines are standard potentials at 1 M or 1 atm of chemical species except for varying pH.
Figure 3. Linear sweep voltammetry (LSV, scan rate 0.1 mV/s) curves superimposed with anode polarization curves. Vertical lines are the theoretical open circuit potentials of the bioanode at a given pH calculated using the Nernst equation. Yellow circles designate midpoint potential (EKA) values.
Table 1. Electrochemical Data in Each Medium pH from Polarization Curvesa pH
a
6.0
6.5
7.0
7.5
8.0
Anode OCP (mV)
438
482
517
539
555
Cathode OCP (mV)
288
251
224
170
145
OCV (mV) Pmax (mW/m2)
725 495
733 712
741 687
709 654
700 636
jmax (A/m2)
1.85
4.69
5.01
5.05
6.15
EKA (mV)
352
375
416
440
431
EKA and jmax were calculated from LSV curves.
with respect to Ag/AgCl using Ag/AgCl = SHE 197 mV.24 Power density and current density were calculated using the apparent surface area of biofilm-covered anode (3.14 cm2).
’ RESULTS Polarization Behavior. Anode and cathode potentials were affected by the pH over the entire tested range of 6.08.0. As pH was lowered, both the anode potential (Ean) and the cathode potential (Ecat) increased (Figure 2 and Table 1), indicating that anode performance deteriorated and cathode performance was enhanced as bulk pH decreased. The OCP was described using the Nernst equation (Figure 2), assuming concentrations of 0.01 mM H2O2 at the cathode and 5 mM HCO3 and 10 mM acetate in the solution at the anode. These bicarbonate and acetate concentrations were the initial medium conditions, and this low hydrogen peroxide concentration provided the best fit to the cathode OCP data. Potential changes per pH unit increase were 60 mV/pH for the anode and 68 mV/pH for the cathode. Open circuit voltage (OCV) reached a maximum at pH 7, but maximum power density (Pmax) was highest at pH 6.5 (Table 1). Maximum current density increased, and the midpoint potential
Figure 4. Bioanode impedance plots at Ean of 250, 300, 350, 425, and 450 mV in medium of pH 7. Points are experimental data, lines are fitted curves using the BIOANODE-1 equivalent circuit. A - Nyquist plot, B - bode plot.
(apparent half-saturation potential EKA derived from LSV curves) decreased as pH increased (Table 1, Figure 3). At each medium pH, maximum current density was attained when Ean was approximately 100 mV greater than EKA. Anode polarization curves superimposed on LSV curves illustrated that both methods yielded comparable results (Figure 3), except that an Ean range with a maximum current plateau was not observed using Rext alterations (i.e., the polarization curve data) in high pH conditions. The duplicated polarization tests showed very high reproducibility of the reactor performance while subjected to the short-term pH perturbations. Impedance Parameters of the Bioanode. The BIOANODE-1 model generated excellent fits to the impedance data (Figure 4). The χ2 value ranged from approximately 104 to 103.25 Fits deviated slightly from experimental data above EKA 9071
dx.doi.org/10.1021/es201737g |Environ. Sci. Technol. 2011, 45, 9069–9074
Environmental Science & Technology
ARTICLE
α1, 8.0 107 Ω for R2, 6.0 104 F for C2, and 0.99 for α2. R1 and C1 in the abiotic anode were slightly higher than the bioanode, but R2 was 4 orders of magnitude larger and C2 was 2 orders of magnitude larger.
Figure 5. Parameter values from fittings of impedance data with the equivalent circuit BIOANODE-1. The lowest Ean for EIS was the OCP.
due to the emergence of a third arc at the lowest frequencies (Figure 4). The low frequency arc (LFA) and this third arc enlarged as Ean further increased above EKA, and they finally combined into a single arc. However, χ2 was still in the same order. Because the third impedance arc was not considered in the BIOANODE-1 model, impedance spectra at potentials below EKA should be well-characterized using our model. Medium pH did not affect R1, but reducing medium pH increased R2 (Figure 5). At potentials below EKA, R1 averaged 11.3 ( 0.8 Ω, and R2 ranged from 2- to 321-fold higher than R1. C2 (1.20 102 ( 0.47 102 F) was 2 or 3 orders of magnitude larger than C1 (5.89 105 ( 0.77 105 F). The nonideality constants were 0.80 ( 0.03 for α1 and 0.78 ( 0.08 for α2. Below EKA, C2 increased and C1 decreased with increasing Ean. pH and Ean did not affect ohmic resistance (RΩ, 12.4 ( 0.8 Ω) because it is related to electrode resistance, medium conductivity, and proximity of the reference electrode to the working electrode, all of which were constant throughout these experiments. The presence of the third arc significantly affected fitting as indicated by the C2 surge at potentials above EKA (Figure 5). The nonideality constant α2 fluctuated with Ean variations, but α1 did not. Over the entire potential range, α1 averaged 0.82 ( 0.04 and α2 was 0.73 ( 0.12. In the overall potential range, R2 (1201.8 ( 2552.4 Ω) was 2 orders of magnitude greater than R1 (10.8 ( 1.1 Ω) (Figure 5). R2 was lowest near the EKA and increased 2 or 3 orders of magnitude above EKA, while R1 decreased by only a few ohms as Ean approached EKA (Figure 5). Though impedance parameters generated from current-producing potentials had a consistent trend, those at OCP showed neither any consistency with the medium pH changes nor any correlation with the values attained from current-generating potentials (Table 1). An abiotic control anode had an OCP of 510 mV at pH 7, which is close to the thermodynamic estimate for water hydrolysis. Acetate addition did not affect anode polarization, demonstrating that acetate was not oxidized abiotically at the anode. The Nyquist plot of the abiotic anode (not shown) consisted of two arcs. Impedance parameters of the abiotic anode at open circuit and pH 7 were 13.5 Ω for R1, 4.3 105 F for C1, 0.88 for
’ DISCUSSION Sequential reactions of microbial substrate oxidation and electron transfer via riboflavin were modeled in a previous equivalent circuit (EC) with three time constants.9 However, this EC did not generate accurate fitting to our data. An EC for hydrogen oxidation in the anode of a PEM fuel cell 26 was used in this experiment. It consists of two processes, the electrosorption of hydrogen (H2 T 2Ha) and the electron transfer from hydrogen onto the anode interface (Ha T H+ + e).26 While this EC provided an excellent fit to our EIS data, a complete mechanistic interpretation of these data from our MFC is precluded by factors such as the uncharacterized complexity of the microbial community, the associated mechanisms of charge transfer, the spatial heterogeneity of the biofilm, and the potential pH effects on all of these components. Increasing Ean lowered both the high- and low-frequency impedance in this study, showing that increasing Ean facilitated the overall kinetics of exoelectrogenic electron transfer to the solid anode surface at potentials below EKA. However, current did not proportionally increase with increasing Ean above EKA. R1 and R2 started to increase with increasing Ean above EKA. While R1 increased a slight amount, R2 significantly increased due to the emergence of the third arc. Either proton diffusion resistance or excessive potential might increase those resistances. The pH gradient between the anode biofilm and the medium controls the diffusion of protonated buffer species out of the biofilm, so that low medium pH limits current density.10 Our results showed that R2 was greatly affected by pH, while R1 was insensitive to pH change (Figure 5). While microbial processes are generally sensitive to pH, it cannot be determined without additional targeted experiments which charge transfer reaction was captured by R1 and insensitive to pH over the range of this study. Diffusion of protonated species from the anode biofilm may be related to the third arc detected in the low frequency domain (Figure 4). Protons are produced during microbial substrate oxidation, which combine with buffer species such as carbonate or phosphate. Proton diffusion can be modeled by a Warburg element in equivalent circuits, usually shown as a straight line in the low frequency domain of a Nyquist plot. Pure diffusion creates a 45° line, but other factors such as capacitance influence the angle and curvature of the Warburg element. For example, mass transport processes for a porous gas diffusion electrode usually give a semicircle rather than a straight line in a Nyquist arc due to finite diffusion. 15,27 In PEM fuel cell systems, the main contributor to the overall impedance in low overpotential is normally charge transfer resistance (Rp), which is dependent on potential according to the Tafel equation V ∼ logRp1, but diffusion resistance (mass transport resistance) becomes dominant in the overall impedance in high overpotential.15 In our data, the third arc became a significant portion of total impedance at high current density in the saturating Ean region, appearing at potentials above EKA and finally combining with the LFA, creating a semicircle (Figure 4). A two-chamber MFC did not have such an increase of impedance because of its low maximum current density (ca. 0.3 A/m2-anode).7,21 In a previous study, the lowest 9072
dx.doi.org/10.1021/es201737g |Environ. Sci. Technol. 2011, 45, 9069–9074
Environmental Science & Technology
diffusive species, or pH gradient. Acidotolerant exoelectrogens might be another solution to pH inhibition. A strain of Acidiphilium sp. was able to produce up to 3 A/m2-anode at ca.390 mV in aerobic low-pH medium, which leads to a low ohmic resistance configuration due to shorter distance between anode and air-cathode and high cathode performance due to high proton concentration.32 The BIOANODE-1 model created a nice fit to the experimental data as indicated by χ2, even in the presence of a third arc. Addition of a CPE component allows the arc to elongate and match data, even though the real physical process may not be understood.22 In our results, the third arc was not discretely modeled but rather was fitted with the LFA charge transfer resistance. A Finite Warburg element was added to our model, but it did not generate a well-matched fitting (data not shown). For future study, the BIOANODE-1 model can be improved by considering diffusion resistance in a high-flux system or excessive electrical energy in EC modeling.
’ AUTHOR INFORMATION Corresponding Author
*Phone: 814-865-9436. Fax: 814-863-7304. E-mail: jregan@ engr.psu.edu. Present Addresses
)
charge transfer resistance (0.408 kΩ cm2) was measured at an applied potential near the EKA of 355 mV (158 mV vs SHE) for G. sulfurreducens.6 Our results also demonstrate charge transfer resistance was lowest near EKA at each pH condition. We showed that medium pH affected EKA, which has been regarded as an intrinsic constant of an anode biofilm.28 These results imply that proton diffusion affects EKA. Charge transfer resistance R1 (average ∼10 Ω) was minor compared with R2 (average ∼400 Ω). At the maximum current density of our anode biofilm (6.8 A/m2), the processes described by R1 created ∼20 mV of potential loss, while the R2 processes imposed ∼260 mV of potential loss (3.14 cm2 of anode area 6.8 A/m2 impedance Ω at 6.8 A/m2). Future research should focus on discriminating the processes contributing to these discrete EIS signatures and engineering the exoelectrogenic microbial processes to minimize these overpotentials. C2 was several orders of magnitude greater than C1, demonstrating higher charge storage in the LFA process(es) than the HFA process(es). The CPE is attributed to heterogeneous electrode properties or reactions (e.g., surface roughness, nonhomogeneous reaction rates on a surface, or variation of catalyst thickness).29 The LFA had a higher α (0.82) than that of the HFA (0.73), but it is unknown whether this difference in nonideality is significant. C2 increased sharply at anode potentials just above EKA. In these potential ranges, EC fitting was not good due to the emergence of the third arc. So, this increase in the fitted C2 value might be due to overestimation. Anode performance was enhanced as the bulk pH was increased as previously shown.30 This result was not consistent with a previous study using open circuit EIS, which showed the bioanode process preferred a neutral pH and had a slightly higher charge transfer resistance above neutral conditions.13 Perhaps open circuit EIS might not be a representative way to evaluate electrode impedance in MFCs but rather closed-circuit conditions should be used.15 Because anode catalysts are living microorganisms, the unavailability of an electron acceptor at open circuit might create unrepresentative impedance values, and current production (microbial respiration) should be allowed to characterize living microbial catalysts. Even in hydrogen fuel cells, the impedance data at current-generating potentials were suggested to be more representative of overall electrode processes than that at OCV with an exception that EIS of a symmetrical gas-feeding can be measured only at OCV.15 Alkaline electrolyte fuel cells have less cathodic activation overpotential than PEM fuel cells with an acidic electrolyte system.31 In an alkaline electrolyte system, hydroxide (O2 + 4e + 2H2O T 4OH) generated at the cathode is utilized for the anodic hydrogen oxidation (2H2 + 4OH T 4H2O + 4e), where protons are not involved in the electrochemical reaction.31 In a previous MFC study, the charge transfer resistance of the cathode decreased in high pH condition,13 suggesting that their system behaved like an alkaline electrolyte system. In our result, high proton availability in low electrolyte pH led to improved cathode performance, demonstrated by increasing cathodic potential. Our thermodynamic estimations suggested that protons instead of hydroxide are involved in electrochemical reactions in our MFC system. Although an acidic medium is beneficial for cathodic oxygen reduction, it decreases anodic substrate oxidation. High anode surface area, buffer concentration, or medium pH can prevent the acidification of anode biofilm because diffusion of protonated species is proportional to diffusive surface area, concentration of
ARTICLE
Current address: Sustainability Consulting Group, Samsung SDS, Seoul, South Korea, 135-918.
’ ACKNOWLEDGMENT This research was supported by National Science Foundation Grant CBET-0834033. The authors appreciate Dr. Xin Wang (Assistant Professor in Environmental Engineering, Nankai University) for his comments on this experiment, Dr. Kyu Taek Cho (Presently of the Environmental Energy Technologies Division, Lawrence Berkeley National Laboratory) for his explanations on fuel cell systems, and Dr. Judodine Patterson (Alcoa Technical Center, PA) for her manuscript revision. We also thank Dr. Digby Macdonald (Distinguished Professor in Material Science and Engineering, Penn State University) for his comments. ’ REFERENCES (1) Jung, S.; Regan, J. M. Influence of external resistance on electrogenesis, methanogenesis, and anode prokaryotic communities in microbial fuel cells. Appl. Environ. Microbiol. 2011, 77 (2), 564–571. (2) Ciureanu, M.; Wang, H. Electrochemical impedance study of electrode-membrane assemblies in PEM fuel cells: I. Electro-oxidation of H2 and H2/CO mixtures on Pt-based gas-diffusion electrodes. J. Electrochem. Soc. 1999, 146 (11), 4031–4040. (3) Kim, J.-D.; Park, Y.-I.; Kobayashi, K.; Nagai, M.; Kunimatsu, M. Characterization of CO tolerance of PEMFC by AC impedance spectroscopy. Solid State Ionics 2001, 140 (34), 313–325. (4) Wagner, N.; Gulzow, E. Change of electrochemical impedance spectra (EIS) with time during CO-poisoning of the Pt-anode in a membrane fuel cell. J. Power Sources 2004, 127 (12), 341–347. (5) He, Z.; Huang, Y. L.; Manohar, A. K.; Mansfeld, F. Effect of electrolyte pH on the rate of the anodic and cathodic reactions in an aircathode microbial fuel cell. Bioelectrochemistry 2008, 74 (1), 78–82. (6) Marsili, E.; Rollefson, J. B.; Baron, D. B.; Hozalski, R. M.; Bond, D. R. Microbial biofilm voltammetry: Direct electrochemical characterization of catalytic electrode-attached biofilms. Appl. Environ. Microbiol. 2008, 74 (23), 7329–7337. (7) Ramasamy, R. P.; Ren, Z. Y.; Mench, M. M.; Regan, J. M. Impact of initial biofilm growth on the anode impedance of microbial fuel cells. Biotechnol. Bioeng. 2008, 101 (1), 101–108. 9073
dx.doi.org/10.1021/es201737g |Environ. Sci. Technol. 2011, 45, 9069–9074
Environmental Science & Technology (8) Borole, A. P.; Aaron, D.; Hamilton, C. Y.; Tsouris, C. Understanding long-term changes in microbial fuel cell performance using electrochemical impedance spectroscopy. Environ. Sci. Technol. 2010, 44 (7), 2740–2745. (9) Ramasamy, R. P.; Gadhamshetty, V.; Nadeau, L. J.; Johnson, G. R. Impedance spectroscopy as a tool for non-intrusive detection of extracellular mediators in microbial fuel cells. Biotechnol. Bioeng. 2009, 104 (5), 882–891. (10) Torres, C. I.; Kato Marcus, A.; Rittmann, B. E. Proton transport inside the biofilm limits electrical current generation by anode-respiring bacteria. Biotechnol. Bioeng. 2008, 100 (5), 872–81. (11) Franks, A. E.; Nevin, K. P.; Jia, H.; Izallalen, M.; Woodard, T. L.; Lovley, D. R. Novel strategy for three-dimensional real-time imaging of microbial fuel cell communities: Monitoring the inhibitory effects of proton accumulation within the anode biofilm. Energy Environ. Sci. 2009, 2 (1), 113–119. (12) Gil, G. C.; Chang, I. S.; Kim, B. H.; Kim, M.; Jang, J. K.; Park, H. S.; Kim, H. J. Operational parameters affecting the performance of a mediator-less microbial fuel cell. Biosens. Bioelectron. 2003, 18 (4), 327–334. (13) Biffinger, J. C.; Pietron, J.; Bretschger, O.; Nadeau, L. J.; Johnson, G. R.; Williams, C. C.; Nealson, K. H.; Ringeisen, B. R. The influence of acidity on microbial fuel cells containing Shewanella oneidensis. Biosens. Bioelectron. 2008, 24 (4), 900–905. (14) Fan, Y. Z.; Hu, H. Q.; Liu, H. Sustainable power generation in microbial fuel cells using bicarbonate buffer and proton transfer mechanisms. Environ. Sci. Technol. 2007, 41 (23), 8154–8158. (15) Freguia, S.; Rabaey, K.; Yuan, Z.; Keller, J. G. Non-catalyzed cathodic oxygen reduction at graphite granules in microbial fuel cells. Electrochim. Acta 2007, 53 (2), 598–603. (16) Zhang, F.; Jacobson, K. S.; Torres, P.; He, Z. Effects of anolyte recirculation rates and catholytes on electricity generation in a litre-scale upflow microbial fuel cell. Energy Environ. Sci. 2010, 3, 1347–1352. (17) Erable, B.; Etcheverry, L.; Bergel, A. Increased power from a two-chamber microbial fuel cell with a low-pH air-cathode compartment. Electrochem. Commun. 2009, 11 (3), 619–622. (18) Liang, P.; Fan, M.; Cao, X.; Huang, X. Evaluation of applied cathode potential to enhance biocathode in microbial fuel cells. J. Chem. Technol. Biotechnol. 2009, 84 (5), 794–799. (19) Logan, B.; Cheng, S.; Watson, V.; Estadt, G. Graphite fiber brush anodes for increased power production in air-cathode microbial fuel cells. Environ. Sci. Technol. 2007, 41 (9), 3341–3346. (20) Cheng, S.; Liu, H.; Logan, B. E. Increased performance of single-chamber microbial fuel cells using an improved cathode structure. Electrochem. Commun. 2006, 8 (3), 489–494. (21) Jung, S.; Regan, J. M. Comparison of anode bacterial communities and performance in microbial fuel cells with different electron donors. Appl. Microbiol. Biotechnol. 2007, 77 (2), 393–402. (22) Macdonald, J. R. Impedance spectroscopy. Ann Biomed. Eng. 1992, 20 (3), 289–305. (23) Logan, B. E. Microbial fuel cells; Wiley-Interscience: 2008. (24) Bard, A.; Faulkner, L. Electrochemical methods: Fundamentals and applications; John Wiley & Sons: 2001. (25) Taylor, J. An introduction to error analysis: The study of uncertainties in physical measurements; Univ Science Books: 1997. (26) Ciureanu, M.; Mikhailenko, S. D.; Kaliaguine, S. PEM fuel cells as membrane reactors: Kinetic analysis by impedance spectroscopy. Catal. Today 2003, 82 (14), 195–206. (27) Xie, Z.; Holdcroft, S. Polarization-dependent mass transport parameters for ORR in perfluorosulfonic acid ionomer membranes: An EIS study using microelectrodes. J. Electroanal. Chem. 2004, 568, 247–260. (28) Marcus, A. K.; Torres, C. I.; Rittmann, B. E. Conduction-based modeling of the biofilm anode of a microbial fuel cell. Biotechnol. Bioeng. 2007, 98 (6), 1171–1182. (29) Baron, D. B.; Labelle, E.; Coursolle, D.; Gralnick, J. A.; Bond, D. R. Electrochemical measurement of electron transfer kinetics by Shewanella oneidensis MR-1. J. Biol. Chem. 2009, 284 (42), 28865–28873.
ARTICLE
(30) Yuan, Y.; Zhao, B.; Zhou, S.; Zhong, S.; Zhuang, L. Electrocatalytic activity of anodic biofilm responses to pH changes in microbial fuel cells. Bioresour. Technol. 2011, 102 (13), 6887–91. (31) Mench, M. M. Fuel cell engines; John Wiley and Sons, Inc.: New York, NY, 2008. (32) Malki, M.; De Lacey, A. L.; Rodriguez, N.; Amils, R.; Fernandez, V. M. Preferential use of an anode as an electron acceptor by an acidophilic bacterium in the presence of oxygen. Appl. Environ. Microbiol. 2008, 74 (14), 4472–6.
9074
dx.doi.org/10.1021/es201737g |Environ. Sci. Technol. 2011, 45, 9069–9074
ARTICLE pubs.acs.org/est
Volatile Organic Compounds in Small- and Medium-Sized Commercial Buildings in California Xiangmei (May) Wu,† Michael G. Apte,‡ Randy Maddalena,‡ and Deborah H. Bennett†,* † ‡
Department of Public Health Sciences, University of California, Davis, California, United States Lawrence Berkeley National Laboratory, Berkeley, CA, United States
bS Supporting Information ABSTRACT: While small- and medium-sized commercial buildings (SMCBs) make up 96% of the commercial buildings in the U.S., serving a large variety of uses, little information is available on indoor air quality (IAQ) in SMCBs. This study investigated 37 SMCBs distributed across different sizes, ages, uses, and regions of California. We report indoor concentrations and whole building emission rates of a suite of 30 VOCs and aldehydes in these buildings. There was a considerable range in the concentrations for each of the contaminants, especially for formaldehyde, acetaldehyde, decamethylcyclopentasiloxane, d-limonene, 2-butoxyethanol, toluene, 2,2,4-trimethylpentanediol diisobutyrate, and diethylphthalate. The cause of higher concentrations in some building categories generally corresponded to expected sources, for example, chloroform was higher in restaurants and grocery stores, and formaldehyde was higher in retail stores and offices. Factor analysis suggests sources in SMCBs include automobile/traffic, cleaning products, occupant sources, wood products/coating, and plasticizers. The comparison to health guidelines showed that formaldehyde concentrations were above the chronic RELs required by the OEHHA (9 μg/m3) in 86% of the buildings. Data collected in this study begins to fill the knowledge gap for IAQ in SMCBs and helps us understand the indoor sources of VOCs to further improve indoor air quality in SMCBs.
’ INTRODUCTION Indoor air quality (IAQ) in commercial buildings is a public health concern and has been associated with Sick Building Syndrome symptoms and mucous membrane and lower respiratory irritation.1 Volatile organic compounds (VOCs) are a common group of indoor air pollutants, with some being classified as known or probable human carcinogens and some having neurologic, developmental, or reproductive effects at high levels. They are commonly found in building materials and furnishings, such as cabinetry and carpeting, and numerous consumer products, e.g., paints, adhesives, and air refreshers.2,3 The U.S. population spends approximately 87% of their time indoors, and of that, 21% is in nonresidential environments, primarily in commercial buildings.4 Small- and medium-sized commercial buildings (SMCBs), as defined here, are any low-rise building (less than four stories) with roof-top heating, ventilation, and air-conditioning (HVAC) units. These buildings are commonly found in small office complexes, schools, and strip malls, serving as places of work and for commercial, educational, and recreational activities.5 SMCBs make up 96% of the commercial buildings in the U.S. 6 While there have been a number of studies on indoor VOCs in large commercial buildings,712 especially office buildings, very little information is available on SMCBs which have more diverse uses.13 r 2011 American Chemical Society
VOC levels in SMCBs are subject to the impact of chemical emissions from building materials and activities that take place in the buildings, building ventilation rates, and outdoor air pollution. SMCBs, such as dry cleaners, restaurants, and printing establishments, have substantial, often unique, indoor VOC sources. Office spaces can have localized sources, including copiers, intake and recirculation of polluted outside air, and introduction of pollutants brought into the building by occupants. Retail stores contain a wide range of new products that outgas a variety of VOCs, leading to higher levels indoors of these compounds.12 Thus, to the extent that the quality of the commercial indoor environment affects people’s health and well-being, the time spent in SMCBs has a potential to significantly impact public health. However, due to the largely disaggregated, heterogeneous nature of commercial buildings, information on indoor pollutant levels in SMCBs is very limited. To fill this knowledge gap, we collected information on the operation and maintenance of HVAC and air filtration systems, indoor pollutant levels in SMCBs, and the relation between the two. This paper focuses Received: June 22, 2011 Accepted: September 2, 2011 Revised: August 30, 2011 Published: September 02, 2011 9075
dx.doi.org/10.1021/es202132u | Environ. Sci. Technol. 2011, 45, 9075–9083
Environmental Science & Technology on indoor levels of VOCs. Potential sources, differences in concentrations between types of SMCBs, and comparison with relevant health guidelines and existing literature are also explored.
’ METHODS Buildings Sampled. Thirty-seven unique buildings were included in the study, with three buildings sampled on two occasions, resulting in 40 sampling days. The selection of buildings was semi-randomized, providing a minimum coverage of the vast variety of SMCBs across space, age, size, and building use category. Buildings were almost evenly distributed across each of five regions of California: north-coastal, north-inland, southcoastal, south-inland, and central-inland. A total of 25 buildings built before 2000 and 12 buildings built after 2000 were included. We classified buildings into three size categories: small (1000 12 000 ft2), medium (12 00025 000 ft2), and medium/large (25 00050 000 ft2). Thirty buildings were one-story, and seven were two stories. Twenty-four small buildings (two sampled twice), seven medium buildings (one sampled twice), and six medium/large buildings were included. The buildings varied in use, including seven retail establishments, five restaurants, eight offices, two each of gas station convenience stores, hair salons, healthcare facilities, grocery stores, dental offices, and fitness gyms, along with three miscellaneous buildings. Details on the building distribution can be found in the Supporting Information (SI). Sample Collection and Analysis. For each building in the study, the air exchange rate and the concentrations of a suite of air pollutants including hydrocarbons, aldehydes, and chlorinated compounds were determined. These compounds, listed with their abbreviations in Table 1, were selected either because they are of health concern or were anticipated to have higher concentrations in commercial buildings. Compounds were measured at one or two indoor locations, depending on building size, and at one outdoor location for 4 h during normal working hours while the buildings were occupied. Samplers were placed at about 1 m above the ground, located as centrally as possible given logistic constraints. A total of 66 indoor VOC samples and 62 aldehyde samples were collected from 40 sampling days. The majority of the compounds were collected onto multibed sorbent tubes with a primary bed of Tenax-TA sorbent backed with a section of Carbosieve III and analyzed by EPA method TO-17, using thermodesorptionGC/MS (Series 6890Plus/ 5973; Agilent Technologies).14 Formaldehyde, acetaldehyde, and acetone were collected using a commercially available cartridge packed with acidified 2,4-dinitrophenylhydrazine (DNPH) coated silica gel and analyzed by EPA method TO-11A, using highperformance liquid chromatography (HPLC) (1200 Series; Agilent Technologies).15 All samples were analyzed at Lawrence Berkeley National Laboratory. Ozone scrubbers were used on all DNPH cartridges to prevent aldehyde loss due to reaction with ozone on the absorbent. The VOC sample flow rate was 1.25 L/h for the initial five buildings, and for the remainder of the project, a flow rate of 1 L/h indoors and 2.5 L/h outdoors was used. Sampling flow rate for carbonyl was 0.5 L/min. More details on chemical analysis can be found in the SI. The building ventilation rate during operating hours was measured using a tracer (sulfur hexafluoride) decay method and was used in calculating whole building emission rate. The measurement method for ventilation rates can be found in the SI. The quality assurance measures are also detailed in the SI. A total of 30% of samples were processed as travel blanks that were
ARTICLE
carried to the field and returned unopened, and 24% were processed as field blanks that were carried to the field, installed on the sample pump, and then removed without pulling air through the sample and returned to the lab. Only the low molecular weight carbonyls had background concentrations near or slightly above the MDL. Duplicate samples were collected in parallel to the primary samples for 93.5% of the sample locations, resulting in a total of 1972 individual compound duplicate measurements where the compound concentrations were greater than the MDL. The difference between sample pairs was less than 10% for 65% of the measurements, with 92% of the measurements having a difference of less than 50%. Collection efficiency was determined for VOCs under field conditions using sampling tubes mounted in series, and the results ranged from 85% to 100%. Concentrations below MDL were replaced with 1/2 MDL, while for samples between the MDL and the analytical limit of quantification (LOQ), determined as 10 times the standard deviation of low-level spikes, were reported as the value determined in the laboratory. Concentrations that were above the calibration range of the analytical method were replaced with a value slightly higher than the maximum calibration point, an underestimate of the actual value. Data Analysis. Summary statistics were calculated for all the VOCs, with the average indoor concentration calculated and used in analyses for buildings where multiple sites were sampled collected. On the basis of potential expected sources, we classified the buildings into a specified building use category from the following list: nonmedical offices, grocery/restaurant, retail, dentist/health care, gas/auto services, hair salon/gym (both had significant use of personal care products), and miscellaneous. Three buildings fell into the miscellaneous category, including a public assembly building, a building used for religious purposes, and a daycare center. As VOC concentrations were log-normally distributed, the analysis of variance (ANOVA) was conducted on log-transformed concentrations to determine which compounds had significant concentration differences between building use categories. Factor analysis was conducted on log-transformed indoor VOC concentrations to identify sources. Pearson correlation coefficients were first calculated to identify correlated compounds. Concentrations of m/p-xylene were highly correlated with ethylbenzene (R = 0.97) and o-xylene (R = 0.96) concentrations; thus, m/p-xylene was excluded from the analysis. Methylene chloride was also excluded due to a number of missing samples. Then, a trial analysis was conducted with 62 observations and 28 compounds, with multiple indoor samples considered as different observations. Though not statistically valid, seven factors with eigenvalues greater than 1 were extracted. To confirm the findings in the first trial, a second attempt only including the compounds with loadings >0.5 on the first five factors extracted was conducted. Since the initial factor pattern matrix was not unique, an oblique rotation after an orthogonal VARIMAX rotation was performed to achieve more interpretable factor loadings. Factor loadings greater than 0.50 are considered to be significant. Note that, with the limited number of observations from the sample of 37 buildings, the factor analysis does not quite reach statistical validity, but is helpful in identifying major source categories. Whole building emission rates, S (μg/h/m2), the apparent source strength for the combined mix of individual indoor 9076
dx.doi.org/10.1021/es202132u |Environ. Sci. Technol. 2011, 45, 9075–9083
Environmental Science & Technology
ARTICLE
Table 1. Distribution of Indoor VOC Concentrations and Whole Building Emission Rates in SMCBs (N = 40, including three repeat measurements)a indoor concentration (μg/m3) compoundb
percent > MDLc (%)
geometric mean
benzene
100
0.69
0.29
toluene
100
4.47
0.44
ethylbenzene m/p-xylene
100 100
0.62 1.63
o-xylene
100
styrene
100
formaldehyde
whole building emission rate (μg/h/m2) Nd
geometric mean
2.11
29
1.00
0.01
4.49
200f
35
7.90
0.10
301
0.08 0.22
34.7 87.7f
32 30
1.06 2.71
0.09 0.22
45.2 114
0.69
0.13
10.7
32
1.28
0.03
13.7
0.37
0.03
3.6
40
0.71
0.05
13.9
100
16.4
1.41
102
39
25.4
2.23
393
acetaldehyde
100
8.94
0.74
72.5
37
14.5
1.00
443
acetone
100
28.3
1.10
1380
38
64.8
7.82
3405
hexanal
100
3.45
0.53
27.5
40
6.18
1.56
277
benzaldehyde octanal
98 95
2.54 1.46
1.11 nd
5.28 16.1
24 39
3.14 2.65
0.35 0.61
19.8 219
range
range
nonanal
97
4.25
nd
20.9
38
7.80
1.44
557
decanal
92
3.33
nd
104
35
12.3
1.92
271
methylene chloridee
100
0.83
0.25
17.1
30
1.39
0.08
44.1
carbon tetrachloride
98
0.46
0.05
2.87
23
0.72
0.03
8.72
chloroform
100
0.3
0.05
2.62
40
0.73
0.02
36.2
trichloroethylene
70
0.02
nd
1.93
26
0.07
0.01
11.0
tetrachloroethylene 1,4-dichlorobenzene
94 88
0.18 0.05
nd nd
118 2.16
32 35
0.30 0.14
0.03 0.02
178 4.39
α-pinene
100
1.67
0.26
33.8
40
2.85
0.47
122
d-limonene
100
8.18
0.28
1100f
40
21.7
0.68
2313
α-terpineol
86
0.11
nd
15.6
36
0.52
0.02
172
n-hexane
100
1.04
0.17
180f
34
1.51
0.15
534
naphthalene
100
0.17
0.04
1.45
40
0.35
0.02
4.8
2-butoxyethanol
100
4.21
0.02
356f
39
8.91
0.69
1329
d5-siloxane phenol
100 100
24.9 1.97
1.30 0.63
120f 17.0
40 32
76.1 3.23
4.84 0.02
1636 16.5
TXIB
100
1.09
0.16
58.6
40
1.77
0.29
48
diethylphthalate
100
0.39
0.09
2.42
40
0.66
0.10
11.6
a
Note that one sample collected in a retail store in a room that conducted screen printing work had many extremely high concentrations and was excluded from this distribution, as this could be classified as a light industrial source rather than a retail source. The distribution by building use category, including geometric/arithmetic means and standard deviation, median, and max, are presented in Tables S5 and S6 (SI). b d5-siloxane = decamethylcyclopentasiloxane; TXIB = 2,2,4-trimethyl-1,3-pentanediol diisobutyrate. c Detection percents were calculated based on the original samples, not the average for each building. A building could have detectable concentration at one sampling site but not at the other one. d The distribution of whole building emission rates were only calculated for the buildings with indoor/outdoor differences greater than zero, namely those having potential source(s). e Measurements of methylene chloride for six buildings were excluded due to suspected blank contamination. f Maximum values were actually above quantitative range and replaced by a value slightly higher than the maximum value we observed. nd, not detectable.
sources, were calculated for each building assuming a steadystate, well-mixed box model, incorporating outdoor concentrations, building ventilation rates, and building volumes with the following steady-state mass balance equation: S¼
Q ðCi Co Þ A
where Q is the total ventilation rate of the building as measured by the tracer decay method (m3 h1), Ci is the measured concentration in the building for the specified compound (μg/ m3), Co is the measured concentration outside of the building for the specified compound (μg/m3), and A is the floor area of the building space (m2). This model assumes that penetration of VOCs from the outdoors is unity16 and that losses by reaction are less
than losses by air exchange, since the time scale of decay is at most 1 103/h for VOCs indoors compared with an air exchange rate of 0.1/h1/h.17 All statistical analyses were conducted using SAS 9.2 for Windows (SAS Institute Inc., Cary, NC).
’ RESULTS Indoor Concentrations and the Variability by Building Use Category. The geometric means and ranges of measured indoor
VOC concentrations across all buildings are presented in Table 1 (see Table S5 of the SI for the distributions by building use). The majority of the compounds were detected in all of the buildings, with the highest geometric mean levels observed for acetone, decamethylcyclopentasiloxane (d5-siloxane), acetaldehyde, d-limonene, and toluene. 9077
dx.doi.org/10.1021/es202132u |Environ. Sci. Technol. 2011, 45, 9075–9083
Environmental Science & Technology
ARTICLE
Figure 1. Comparison of indoor concentrations for selected VOCs by building use category: (a) d5-siloxane, (b) DEP, (c) o-xylene, (d) benzene, (e) acetaldehyde, (f) chloroform, (g) naphthalene, and (h) formaldehyde. Note that the scale for each individual VOC is different.
We compared indoor VOC concentrations by building use, and statistically significant differences (p < 0.05) were observed for ethylbenzene, o-xylene, chloroform, tetrachloroethylene (PCE), naphthalene, phenol, 2,2,4-trimethylpentanediol diisobutyrate (TXIB), and diethyl phthalate (DEP). Concentrations of acetaldehyde, m/p-xylene, d5-siloxane, octanal, and benzene were
marginally different (p = 0.050.09) by building use category. Indoor concentration distributions of selected compounds by building use are presented in Figure 1. Phenol, DEP, and d5-siloxane are used in personal care products, and higher concentrations of these compounds were found in buildings where occupants use personal care products or 9078
dx.doi.org/10.1021/es202132u |Environ. Sci. Technol. 2011, 45, 9075–9083
Environmental Science & Technology with higher occupant density. The concentrations of d5-siloxane were significantly higher in hair salons/gyms and dentist/healthcare facilities, which might use more products containing d5siloxane than other buildings. The level of d5-siloxane was also sometimes above the quantitative range in offices and groceries/ restaurants, possibly due to high occupant density in the buildings (Figure 1a). Phenol is used in the manufacture of resins and nylon but also used in cosmetics, e.g., sunscreen and hair dyes and as a disinfectant and antiseptic. The concentrations of phenol were significantly higher in offices than other types of buildings. DEP was higher in hair salons/gyms and dentist/healthcare facilities (Figure 1b). High concentrations of ethylbenzene, m/p-xylene and oxylene were observed in a medium-sized healthcare building, and it is possible that these chemicals were used as solvents in the laboratory located in this building. Except for the case of the healthcare building, these three compounds had relatively higher concentrations in gas stations/fleet services and retailers (see oxylene as an example in Figure 1c). Groceries/restaurants and miscellaneous buildings had low levels of these compounds. Concentrations of benzene were also high in gas stations/fleet services (Figure 1d), probably from gasoline evaporation. Concentrations of benzene above 2 μg/m3 were observed in a retail store that included a room where screen-printing was conducted and an office where we suspect smoking occurred due to a spike in ultrafine particle concentrations at lunch time while field staff were not present. Acetaldehyde occurs naturally in some foods such as ripe fruits and is also a byproduct of yeast used in baking. It was significantly higher in groceries/restaurants than other types of buildings, with concentrations up to 72.5 μg/m3 (Figure 1e). Acetaldehyde is also a byproduct of combustion, and is emitted from many building materials and consumer products such as adhesives, coatings, lubricants, inks, and nail polish remover. Concentrations of chloroform were also significantly higher in groceries/ restaurants (Figure 1f), potentially due to volatilization from tap water used for cleaning and cooking (both grocery stores included on-site cooking facilities). While PCE was statistically significantly higher in retail stores, the authors do not consider this result generalizable, as two retail stores had extremely high concentrations of PCE, one being in the retail space that conducted on-site screen printing, which may have used PCE, and the second being in a bookstore with a high outdoor concentration measured at the site. Naphthalene concentrations were high in some retail stores and low in groceries/restaurants and miscellaneous buildings (Figure 1g). Octanal concentrations were also found to be higher in retail stores. An extremely high concentration of TXIB (∼60 μg/m3), used as a plasticizer in vinyl flooring, was observed in a fitness gym that had recently replaced the flooring. Retail stores also had higher TXIB concentrations relative to buildings of other uses. There were other compounds with sources of note that did not reach statistical significance between building use categories. Acetone is used in a variety of medical and cosmetic applications, which corresponds to the high concentrations observed in hair salons/gyms and dentist offices/healthcare buildings. Formaldehyde is emitted from carpet, pressed wood products, and many building materials and consumer products or results from indoor secondary chemical reactions, and higher concentrations were found in offices and retail stores, as well as one high concentration (101.7 μg/m3) observed in a hair salon, likely from use of a
ARTICLE
particular, unknown product (Figure 1h). High concentrations of d-limonene, 2-butoxyethanol, and toluene were also observed, corresponding to sources from cleaning/polishing/waxing agents, paints/adhesives, and solvent-containing materials, respectively.18 Whole building emission rates were also estimated, as shown in Table 1 (see Table S6 of the SI for the distributions by building use). Similar to the indoor concentrations, acetone, d5-siloxane, acetaldehyde, d-limonene, toluene, as well as 2-butoxyethanol had high whole building emission rates in SMCBs. The variations of whole building emission rates were similar to the pattern of the variations of indoor concentrations, that is, compounds with significant variation on indoor concentrations also had varied whole building emission rates by building use category. Significant differences (p < 0.05) by building use category were observed for chloroform, PCE, octanal, naphthalene, phenol, and TXIB, and marginal significant differences (p = 0.050.09) were observed for o-xylene, hexanal, DEP, and benzene. In addition, a gas station convenience store, an office, and a restaurant were measured in both the summer and winter. Higher summer concentrations were observed for a number of VOCs, including many aldehydes, chlorinated compounds, and esters, for example, formaldehyde, acetone, nonanal, trichloroethylene, and phenol. This was probably due to the higher emission rate of these chemicals as temperature goes up. In addition, photochemical reactions with ozone, which generate aldehydes, are expected to be more intense in the summer.19 In contrast, d-limonene and d5-siloxane had lower concentrations in the summer than in the winter. The reason was unclear, since the measured ventilation rates were higher in winter in these buildings due to comfortable winter temperatures in California, which was supposed to reduce indoor VOC concentrations. Comparison to Health Guidelines. The indoor concentrations were compared to the reference concentrations (RfC) for chronic inhalation exposure defined in the U.S. EPA Integrated Risk Information System (IRIS; U.S. EPA 2010), the Permissible Exposure Limits (PELs) defined by the Occupational Safety and Health Administration (OSHA; U.S. Department of Labor, 1993), and the reference exposure levels (RELs) suggested by the California Office of Environmental Health Hazard Assessment (OEHHA; California Environmental Protection Agency, 2008), with the values of all guidelines listed in the SI. We note that the measured concentrations are for a 4-h period; however, as the concentrations were measured during a typical business day, they are likely to be somewhat representative of typical concentrations. The majority of the compounds met all three guidelines in all buildings. However, three compounds, formaldehyde, acetaldehyde, and PCE, had concentrations above the guidelines in some buildings. Elevated air concentrations of formaldehyde and acetaldehyde could lead to respiratory irritation and trigger asthma attacks or damage the reparatory tract.20 PCE is a central nervous system depressant and can cause dermal irritation.21 Concentrations were above the chronic RELs required by the OEHHA for formaldehyde (9 μg/m3) in 86% of the buildings. Concentrations of formaldehyde in three buildings, including a retail store, an office, and a hair salon, were above the acute inhalation REL of 55 μg/m3. In addition, an extremely high concentration of PCE was observed in the retail store with screen printing, with a level that was 3 times higher than the chronic inhalation REL (35 μg/m3). Additionally, the concentration in a bookstore was 34.8 μg/m3, which was very near the REL; however, the high indoor concentration was likely due to high outdoor concentrations. More than 50% of the 9079
dx.doi.org/10.1021/es202132u |Environ. Sci. Technol. 2011, 45, 9075–9083
Environmental Science & Technology
ARTICLE
Table 2. Results of Factor Analysisa principal factor compounds
1
2
3
4
5
communality
benzene
0.87
0.71
toluene
0.70
0.66
ethylbenzene
0.65
0.53
formaldehyde
0.77
acetaldehyde
0.73
0.76 0.56
octanal
0.75
0.58
nonanal
0.81
0.65
decanal d-limonene
0.61 0.55
0.47 0.64
0.69
0.70
0.35
α-terpineol n-hexane d5-siloxane
0.78
0.70 0.78
0.63
TXIB
0.74
0.60
diethylphthalate
0.69
0.66
a
The analysis includes 62 observations and 14 compounds. Factor loadings shown are after an oblique rotation and an orthogonal VARIMAX rotation. Factor loadings greater than 0.30 in absolute value are shown and loadings greater than 0.50 are considered to be significant.
buildings had acetaldehyde concentrations above the US EPA RfC for chronic inhalation exposure (9 μg/m3). A number of the compounds measured are classified as carcinogens. Levels of benzene, carbon tetrachloride, and chloroform in the majority of buildings had levels between the EPA inhalation risk level concentrations of 106 and 105, while acetaldehyde levels were primarily between the 105 and 104 levels, and 88% of the buildings had formaldehyde concentrations above the 104 level (specifics listed in Table S10 of the SI). Source Apportionment. The final factor analysis was conducted with 62 observations and 14 compounds (as shown in Table 2), and five factors were obtained, explaining 63% of the variance in the indoor concentrations of the VOCs in the analysis. Factor 1 represents outdoor sources, with high loadings of benzene, toluene, ethylbenzene and n-hexane, primarily emitted from automobile sources. Factor 2, with high loadings of d-limonene, α-terpineol, and d5-siloxane, are possibly related to cleaning products. Factor 3 represents higher molecular weight aldehydes, while factor 5 represents lower molecular weight aldehydes. Factor 4 had high loadings of TXIB and DEP, probably related to plasticizer sources. Since it was suspected that the plasticizer factor was driven by a gym with recent floor renovation, we repeated the analysis without the gym, and obtained the same factor pattern. We acknowledge the limitations of this analysis due to the small number of observations and multiple observations from the same buildings, which may not allow us to build an unbiased correlation matrix. The analysis may not be statistically valid, though interpretable results were obtained. Formaldehyde and Potential Sources. One hypothesis for the high formaldehyde concentrations is that the building ventilation rates might not be sufficient based on required standards. To test the hypothesis, we calculated the hypothetical indoor formaldehyde concentrations based on the actual outdoor concentrations and actual whole building emission rate, combined with the air exchange rate based on the required California Title 24 ventilation value based on building area. The actual
whole building emission rate was calculated from the measured indoor and outdoor concentrations combined with the actual air exchange rate. However, increasing the ventilation to the required rate in all buildings was not sufficient to reduce formaldehyde concentration to below the OEHHA standard in 76% of the buildings, indicating the need for source reduction of formaldehyde from building materials and consumer products (details on Title 24, the analysis, and results are in the SI). Two common indoor sources of formaldehyde in commercial buildings are composite/compressed wood furniture and carpet, either by direct emissions or through secondary reactions from compounds emitted directly from the carpet or other compounds absorbed onto carpets.3,19,22 Comparisons suggest that indoor formaldehyde concentrations were significantly higher with the presence of carpet (Table 3). An association with new carpet or new wood furniture was not found, potentially due to the small sample size. An association with wood furniture was not found, potentially because wood furniture was present in all buildings and we were unable to obtain specific information on composite/compressed versus solid wood furniture, as this is difficult to identify. The comparison excluded an extremely high concentration observed in a hair salon, which was likely from products used in the salon. Since the presence of carpet shows a significant impact on indoor formaldehyde concentration, we further examined the interaction between carpeting and building use category using a regression model with log-transformed formaldehyde concentrations as the dependent variable. Indoor formaldehyde concentrations were significantly different (p = 0.04) between different types of buildings (we note that for these analyses, one hair salon with a high concentration, likely related to product use, was excluded). By including “presence of any carpet” in the regression model, the carpet variable is significant (p = 0.04), while the difference by building use category becomes marginally significant (p = 0.08), indicating that the presence of carpet more than building use category plays the major role in determining indoor formaldehyde concentration. Evaluating “primary flooring being carpet” resulted in both building use category (p = 0.02) and carpet (p = 0.01) as significant factors in influencing indoor formaldehyde concentrations.
’ DISCUSSION This study reports VOC levels in a variety of SMCBs, distributed across different sizes, ages, uses, and regions of California. Data collected in this study address the knowledge gap for air quality in SMCBs and thus help us understand the indoor sources of VOCs to facilitate further improvements to indoor air quality. The comparison to health guidelines highlights concern on the indoor level of formaldehyde. Elevated air concentrations can lead to eye, throat, and respiratory irritation, as well as cancer. At high concentrations, formaldehyde may trigger asthma attacks. Our analysis indicated that increasing ventilation rates could not fully resolve indoor formaldehyde pollution. To address this, it is recommended to reduce formaldehyde emission rates from building materials, furniture, and other products. There are only a limited number of studies on commercial buildings available in the U.S. The Building Assessment Survey and Evaluation (BASE) study measured concentrations of a number of VOCs in 100 office buildings nationwide in the 1990s, including primarily large buildings.23 The geometric 9080
dx.doi.org/10.1021/es202132u |Environ. Sci. Technol. 2011, 45, 9075–9083
Environmental Science & Technology
ARTICLE
Table 3. Impact of Presence of Carpet and Wood Furniture on Indoor Formaldehyde Concentration (μg/m3) category any carpet present primary flooring is carpet new carpet is primary floor covering new wood furniture present a
N
mean
STD
25th %
median
75th %
95th %
comparisona p = 0.007
Y
26
23.3
13.0
13.9
21.4
27.2
51.2
N
10
14.2
17.3
5.73
9.79
11.6
61.2
Y
22
24.5
13.7
16.5
22
32.5
51.2
N
15
14.8
14.2
5.99
11.4
20.5
61.2
Y
5
16.3
9.69
8.91
17.9
25.2
25.5
N
33
20.9
15.0
10.5
18.6
22.8
54.4
Y
6
20.0
11.8
11.4
20.6
25.2
38.1
N
30
20.9
15.4
10.1
18.3
25.5
54.4
p = 0.01 p = 0.60 p = 0.96
Comparison is based on log transformed indoor concentration of formaldehyde, using analysis of variance.
Table 4. Comparison of Indoor VOC Concentrations (μg/m3) with Other Studies 10 small/medium offices 12 California office bldgs (Daisey et al.8) 100 BASE bldgs (EHE23) A call center office bldg (Hodgson et al.10) compound
geometric mean
range
geometric mean
range
geometric mean
range
benzene
0.70
0.422.11
0.98
<0.12.7
3.45
nd17.3
toluene
4.84
1.72113
2.60
0.5817
9.37
nd365
ethylbenzene
0.69
0.233.97
0.50
0.270.98
1.66
nd29.6
m/p-xylene
1.67
0.4914.0
2.12
0.934.6
5.66
nd96.4
o-xylene
0.77
0.276.20
0.66
0.301.4
2.18
nd38.4
0.40
<0.10.95
styrene
0.51
0.123.60
formaldehyde acetaldehyde
23.0 11.0
10.554.4 4.7330.9
acetone
27.3
8.6885.1
hexanal
3.19
0.9511.6
0.46
<0.21.9
0.47
<0.11.5
0.93
nd8.51
15.0 7.00
N/A N/A
geometric mean
range
4.63
2.8310.9
2.17
1.354.86
14.6 5.40
6.6330.7 2.5212.8
32.5
4.03223
33.3
9.5097.4
4.20
nd19.5
4.71
2.509.01
benzaldehyde
3.08
1.465.28
nonanal
4.20
1.597.05
3.71
1.1623.6
methylene chloride
0.89
0.254.02
1.58
nd360
carbon tet.
0.58
0.311.17
1.00
nd3.86
chloroform TCE
0.15 0.03
0.050.74 nd0.28
0.41 0.40
nd9.63 nd17.5
PCE
0.18
0.031.57
1.78
nd33.0
1,4-DCB
0.10
0.030.95
0.74
nd60.9
α-pinene
2.07
0.2915.8
0.61
nd12.2
2.12
0.504.85
d-limonene
5.72
0.28167
1.20
<0.25.6
6.57
nd137
4.62
1.2327.3
0.55
<0.11.6
18.8 37.9
6.3392.3 16.7112
n-hexane
0.82
0.306.08
naphthalene
0.22
0.070.57
2-butoxyethanol d5-siloxane
4.59 25.5
0.90176 7.40120
phenol
4.38
TXIB
0.89
1.80
0.236.9
2.50
nd20.6
0.65
nd8.80
4.98
nd102
1.9417.0
1.72
nd10.4
0.421.86
0.77
nd8.37
1.60
<0.427
means of the concentrations of formaldehyde, acetaldehyde, α-pinene, and phenol in small/medium office buildings that we observed were 40%300% higher than those in the large office buildings in the BASE study (Table 4). In contrast, aromatic compounds, n-hexane, naphthalene, and several chlorinated compounds showed from 2 to >10 times lower concentrations in SMCB offices. Daisey et al. measured VOCs in 12 California office buildings during the same time period, finding lower concentrations of the aromatic compounds and n-hexane than those measured in BASE, with levels generally comparable to levels measured in the SMCB offices.8 Note that BASE and Daisey et al.’s studies were conducted in the 1990s and might not represent current levels in large office buildings, such that temporal changes in product
composition and use may be responsible for part of the differences. Formaldehyde is emitted from composite/pressed wood furniture, and as the popularity of composite/compressed wood furniture continues to increase, so may have formaldehyde concentrations. Hodgson and Levin have observed decreases of concentrations of benzene, PCE, and other aromatics hydrocarbons in U.S. from the 1980s through the 2000s, which was attributed to the reduction of the production and use of these chemicals.13 A recent study conducted in a large call center office building in Northern California reported similar concentrations for many VOCs to the levels we observed in the small/medium office buildings.10 However, again, the formaldehyde and acetaldehyde concentrations in SMCB were 1.62.0 times of those in the call center. 9081
dx.doi.org/10.1021/es202132u |Environ. Sci. Technol. 2011, 45, 9075–9083
Environmental Science & Technology There are also a few existing studies on VOC levels in retail stores or restaurants. Loh et al. reported VOC concentrations in a number of primarily large retail stores and restaurants in Boston, MA, and found that the geometric mean concentrations for formaldehyde and several aromatic hydrocarbons were higher in stores than in other microenvironments, particularly in certain store types.12 For example, formaldehyde was highest in houseware and furniture stores, and toluene was particularly high in multipurpose stores. We also found high levels of formaldehyde and toluene in retail stores, with geometric means of 28.5 and 9.9 μg/m3, respectively. Except for formaldehyde and PCE, the concentrations of aromatic hydrocarbons and other chlorinated compounds were relatively lower in the small/medium retail stores we measured than those reported by Loh et al. In addition, Loh et al. reported high chloroform levels in restaurants (1.1 μg/m3), which is similar to the levels we observed in groceries/restaurants, with a geometric mean of 1.04 μg/m3. Hotchi et al. observed concentrations >10 μg/m3 for formaldehyde, 2-butoxyethanol, toluene, and d5-siloxane in the sales area of a large mulitipurpose store in the San Francisco Bay Area.11 However, except for 2-butoxyethanol which is likely related to floor cleaning and waxing, the concentrations of the other three compounds observed in small/medium retail stores were 1.51.8 times higher than the concentrations in the mulitpurpose store, and the concentrations of acetone, d-limonene, and PCE were 1355 times higher in small/medium retail stores. We were unable to draw conclusions based on the limited existing data on IAQ in retail stores, since the type of products sold in stores varies. SMCBs serve more varied uses than large commercial buildings, which primarily function as offices, and the type of products and the activities in the building influence the IAQ. Thus, it is challenging to obtain representative samples covering the large variety of retail stores to provide any comparable data. Data on whole building emission rates in SMCBs are extremely scarce. Whole building emission rates can be used in environmental modeling, in order to evaluate the risks and impacts of the changing ventilation rate and to compare against emission rates from products. However, whole building emission rates for commercial buildings were rarely reported in previous studies and such factors were only available for limited chemicals. The whole building emission rates per unit area obtained from small/medium offices in our study were 110-fold lower than those reported by Hodgson et al. for a large call center office building,10 though the indoor VOC concentrations were similar, indicating a greater dilution rate due to the high ventilation rate of a large, densely occupied building. Jia et al. measured VOC concentrations in office and work zone, respectively, in 10 small/medium mixed-use buildings in southeast Michigan.24 The source emission rates for the building they reported for office space are basically the same as what we observed in offices for benzene, toluene, m/p-xylene, α-pinene, phenol, and naphthalene. Among the compounds measured in both studies, the only exception is that the mean source emission rate for d-limonene we observed was 16.5 mg/h, much higher than the mean of 6.9 mg/h reported by Jia et al., though with similar medians and maxima, indicating more intense use of fragrant cleaning products in the office buildings we studied. A rough comparison was also made to a residential study conducted in inner-city homes in Los Angeles, CA, where emission rates were reported in units of μg/h for an average floor area of 900 ft2 (range 5002500 ft2).25 The median whole building emission rate of VOCs in SMCBs were from 2.5 times higher for
ARTICLE
benzene to 28 times higher for trichlorobenzene than residential environments. However, conclusions cannot be made until more careful comparison on a per unit area basis is conducted. Though limited by the sample size, the factor analysis helps us understand the sources in SMCBs. The sources highlighted in our analysis include automobile/traffic sources, cleaning products, occupant sources, wood products/coating, and plasticizers. SMCBs are mostly located next to major roads and have parking lots close to the buildings, which affects low-rise buildings more than large buildings. Moreover, the building envelope of SMCBs are usually not as tight as large commercial buildings. As a result, automobile sources appear to have a larger impact on the IAQ in SMCBs. A factor related to automobiles was also observed in the analysis of BASE data.1 Factor analysis also suggests air fresheners and cleaning sources, which are represented by d-limonene and α-terpineol, and occupant sources, such as personal care products, presented by high loading of d5siloxane. Aldehydes usually appear in a single factor in previous studies,1,10 which is associated with compressed wood products and furniture coating sources. However, in our analysis, the lower molecular weight and higher molecular weight aldehydes were extracted into two separate factors. TXIB and DEP appear in the same factor, indicating plasticized material sources. The indoor air quality in SMCBs is more variable than that in large commercial buildings, with more variability in indoor sources, ventilation rates, and maintenance conditions. A full understanding of indoor air quality in SMCBs requires larger studies to cover the great variability of SMCBs.
’ ASSOCIATED CONTENT
bS
Supporting Information. Details on building selection, chemical analysis and QA/QC, ventilation rates, whole building emission rates, VOC concentrations by building use category, and relevant VOC guidelines. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*Tel: (530) 754-8282; fax: (530) 752-3239; e-mail: dhbennett@ ucdavis.edu.
’ ACKNOWLEDGMENT This research was funded by the California Energy Commission (CEC), Public Interest Energy Research (PIER) Program through contract 500-02-023 with the California Air Resources Board (ARB). The authors would like to thank all of the building operators and owners and our field staff, Amber Trout and Michael Powers. Additionally, all work conducted at LBNL is supported by the Director, Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231. ’ REFERENCES (1) Apte, M. G.; Daisey, J. M. In VOCs and “Sick Building Syndrome”: Application of a New Statistical Approach for SBS Research to U.S. EPA BASE Study Data; presented at Indoor Air 99, Edinburgh, Scotland, 1999. 9082
dx.doi.org/10.1021/es202132u |Environ. Sci. Technol. 2011, 45, 9075–9083
Environmental Science & Technology (2) Brown, S. K. Chamber assessment of formaldehyde and VOC emissions from wood-based panels. Indoor Air—Int. J. Indoor Air Qual. Clim. 1999, 9 (3), 209–215. (3) Kelly, T. J.; Smith, D. L.; Satola, J. Emission rates of formaldehyde from materials and consumer products found in California homes. Environ. Sci. Technol. 1999, 33 (1), 81–88. (4) Klepeis, N. E.; Nelson, W. C.; Ott, W. R.; Robinson, J. P.; Tsang, A. M.; Switzer, P.; Behar, J. V.; Hern, S. C.; Engelmann, W. H. The National Human Activity Pattern Survey (NHAPS): A resource for assessing exposure to environmental pollutants. J. Exposure Anal. Environ. Epidemiol. 2001, 11 (3), 231–252. (5) Lee, K. D.; Norford, L. K., High Performance Commercial Building Systems, Report on metrics appropriate for small commercial customers and report on selection of case study sites. Final Report to CEC Public Interest Energy Research Program, Report No. E2P2.1T3b, Building Technology Program, Department of Architecture, Massachusetts Institute of Technology, October 8th, 2001; 2001. (6) EIA. Commercial Buildings Energy Consumption Survey (preliminary data); Energy Information Administration: Washington, DC, 2003; http://www.eia.doe.gov/emeu/cbecs. (7) Burton, L. E.; Baker, B.; Hanson, D.; Girman, J. G.; Womble, S. E.; McCarthy, J. F., Baseline information on 100 randomly selected office buildings in the United States (BASE): Gross building characteristics. Proceedings of Healthy Buildings 2000 2000, 1, http://www.epa. gov/iaq/base/pdfs/base_a2_396.pdf. (8) Daisey, J. M.; Hodgson, A. T.; Fisk, W. J.; Mendell, M. J.; Brinke, J. T. Volatile organic compounds in twelve California office buildings: Classes, concentrations and sources. Atmos. Environ. 1994, 28 (22), 3557–3562. (9) Shields, H. C.; Fleischer, D. M.; Weschler, C. J. Comparisons among VOCs measured in three types of US commercial buildings with different occupant densities. Indoor Air—Int. J. Indoor Air Qual. Clim. 1996, 6 (1), 2–17. (10) Hodgson, A. T.; Faulkner, D.; Sullivan, D. P.; DiBartolomeo, D. L.; Russell, M. L.; Fisk, W. J. Effect of outside air ventilation rate on volatile organic compound concentrations in a call center. Atmos. Environ. 2003, 37 (3940), 5517–5527. (11) Hotchi, T.; Hodgson, A. T.; Fisk, W. J., Indoor Air Quality Impacts of a Peak Load Shedding Strategy for a Large Retail Building. LBNL-59293 2006. (12) Loh, M. M.; Houseman, E. A.; Gray, G. M.; Levy, J. I.; Spengler, J. D.; Bennett, D. H. Measured concentrations of VOCs in several non-residential microenvironments in the United States. Environ. Sci. Technol. 2006, 40 (22), 6903–6911. (13) Hodgson, M.; Levin, H. Volatile organic compounds in indoor air: A review of concentrations measured in North America since 1990. Report LBNL-51715; 2003. (14) Woolfenden, E. A.; McClenny, W. A. Compendium Method TO17: Determination of volatile organic compounds in ambient air using active sampling onto sorbent tubes; EPA/625/R-96/010b; U.S. EPA: Cincinnati, OH, 1997. (15) Willbury, W. T.; Tejada, S.; Lonneman, W.; Kleinndienst, E. Compendium Method TO-11A: Determination of formaldehyde in ambient air using absorbent cartridge followed by high performance liquid chromatography (HPLC) [active sampling methodology]; EPA/625/R-96/010b; U.S. EPA: Cincinnati, OH, 1999. (16) Lewis, C. W. Sources of air pollutants indoors: VOC and fine particulate species. J. Exposure Anal. Environ. Epidemiol. 1991, 1, 31–44. (17) Finlayson-Pitts, B. J.; Pitts Jr, J. N. Atmospheric Chemistry: Fundamentals and Experimental Techniques; John Wiley & Sons: New York, 1986. (18) Nazaroff, W. W.; Weschler, C. J. Cleaning products and air fresheners: Exposure to primary and secondary air pollutants. Atmos. Environ. 2004, 38 (18), 2841–2865. (19) Morrison, G. C.; Nazaroff, W. W. Ozone interactions with carpet: Secondary emissions of aldehydes. Environ. Sci. Technol. 2002, 36 (10), 2185–2192.
ARTICLE
(20) O’Brien, P. J.; Siraki, A. G.; Shangari, N. Aldehyde sources, metabolism, molecular toxicity mechanisms, and possible effects on human health. CRC Crit. Rev. Toxicol. 2005, 35 (7), 609–662. (21) Hake, C. L.; Stewart, R. D. Human exposure to tetrachloroethylene: Inhalation and skin contact. Environ. Health Perspect. 1977, 21, 231. (22) Morrison, G. Interfacial chemistry in indoor environments. Environ. Sci. Technol. 2008, 42 (10), 3495–3499. (23) EHE. A Screening-Level Ranking of Organic Chemicals and Metals Found in Indoor Air. Report 12278; Environmental Health and Engineering, Inc.: Newton, MA, 2002. (24) Jia, C.; Batterman, S.; Godwin, C.; Charles, S.; Chin, J. Y. Sources and migration of volatile organic compounds in mixed use buildings. Indoor Air 2010, 20 (5), 357–369. (25) Sax, S. N.; Bennett, D. H.; Chillrud, S. N.; Kinney, P. L.; Spengler, J. D. Differences in source emission rates of volatile organic compounds in inner-city residences of New York City and Los Angeles. J. Exposure Sci. Environ. Epidemiol. 2004, 14, S95–S109.
9083
dx.doi.org/10.1021/es202132u |Environ. Sci. Technol. 2011, 45, 9075–9083
ARTICLE pubs.acs.org/est
Greenhouse Gas Emission Evaluation of the GTL Pathway Grant S. Forman,†,* Tristan E. Hahn,‡ and Scott D. Jensen§ †
Sasol Synfuels International, Sasol North America, 900 Threadneedle, Suite 100, Houston, Texas 77079-2990, United States U.S.A Sasol Synfuels International, 33 Baker Street, Rosebank 2196, Johannesburg, South Africa § Baker & O’Brien, Inc., 12221 Merit Drive, Suite 1150, Dallas, Texas 75251, United States ‡
bS Supporting Information ABSTRACT: Gas to liquids (GTL) products have the potential to replace petroleum-derived products, but the efficacy with which any sustainability goals can be achieved is dependent on the lifecycle impacts of the GTL pathway. Life cycle assessment (LCA) is an internationally established tool (with GHG emissions as a subset) to estimate these impacts. Although the International Standard Organization’s ISO 14040 standard advocates the system boundary expansion method (also known as the “displacement method” or the “substitution method”) for life-cycle analyses, application of this method for the GTL pathway has been limited until now because of the difficulty in quantifying potential products to be displaced by GTL coproducts. In this paper, we use LCA methodology to establish the most comprehensive GHG emissions evaluation to date of the GTL pathway. The influence of coproduct credit methods on the GTL GHG emissions results using substitution methodology is estimated to afford the Well-to-Wheels (WTW) greenhouse gas (GHG) intensity of GTL Diesel. These results are compared to results using energy-based allocation methods of reference GTL diesel and petroleum-diesel pathways. When substitution methodology is used, the resulting WTW GHG emissions of the GTL pathway are lower than petroleum diesel references. In terms of net GHGs, an interesting way to further reduce GHG emissions is to blend GTL diesel in refineries with heavy crudes that require severe hydrotreating, such as Venezuelan heavy crude oil or bitumen derived from Canadian oil sands and in jurisdictions with tight aromatic specifications for diesel, such as California. These results highlight the limitation of using the energy allocation approach for situations where coproduct GHG emissions reductions are downstream from the production phase.
’ INTRODUCTION Achieving an equitable balance between energy security1 and climate change issues2 is one of the biggest strategic challenges facing both developed and developing countries today. In the United States, this challenge is made even more acute by the size of the transportation sector and the reliance on imported oil.3 Solving this energy strategy dilemma requires identifying domestic liquid fuel alternatives that can be anticipated to contribute to domestic energy security while at the same time provide benefits to the U.S. economy and reduce existing GHG emissions from current crude oil import sources.4 The U.S. has large recoverable amounts of conventional and unconventional sources of natural gas, which could, in principle, be utilized to produce transportation fuels, such as gas to liquids (GTL) diesel via the Fischer Tropsch process, thus increasing the nation’s energy security by reducing dependence on imported petroleum.3 These energy sources must be critically analyzed to determine their greenhouse gas (GHG) emission impacts while concurrently considering scalability and security of supply issues. Environmental regulations, such as low carbon5 and renewable fuels standards6 are being promoted on a global basis7 because their implementation may potentially reduce transport related r 2011 American Chemical Society
GHG emissions, lowering the transportation sector’s use of, and reliance upon, petroleum.8 Investigation of various conventional9 and future fuel10 types and production pathways with GHG and energy benefits requires close examination of the energy use and emission burdens of the entire life cycle. This can be done using life cycle assessment (LCA), which is an internationally standardized methodology for the systematic and quantitative evaluation of environmental impacts from functionally equivalent products or services, through all stages of the product life cycle,11 which for petroleum-derived transportation fuels includes feedstock extraction, fuel production and fuel combustion in the vehicle engine. An LCA is conducted by following each fuel life-cycle step to derive the total energy use and emission burdens for a given amount of fuel produced and used. The final LCA impacts, which include GHG emissions, for a given fuel pathway are usually compared to those of baseline fuels so that the relative merits of new fuel pathways can be assessed for policy implementation.12 Received: June 23, 2011 Accepted: September 6, 2011 Revised: August 29, 2011 Published: September 22, 2011 9084
dx.doi.org/10.1021/es202101b | Environ. Sci. Technol. 2011, 45, 9084–9092
Environmental Science & Technology For a process that results in two or more products, an LCA must define how environmental impacts will be assigned to those products. The framework for handling coproducts in an LCA is presented in ISO 14040:2006,13,14 and advocates avoiding allocation when possible by including within the boundary of the assessment production processes for materials that are replaced by coproducts. A common approach in LCA and net energy analysis, known as the “system expansion” method (also known as the “substitution” or “displacement” method) credits saved energy and emissions burdens to coproducts associated with the products displaced in the market. This method subtracts the energy and materials required to produce the material, along with any downstream benefits or penalties, that are displaced by the coproduct. This method has been adopted as the default for dealing with biofuel coproducts in transportation LCA models and used in biofuel regulations development,15 18 and the substitution method is used wherever needed in the life cycle of a given fuel pathway. For example, a certain amount of the residual biomass energy content in corn or sugar cane ethanol production can ultimately be converted to excess electricity, which is exported to the local regional grid. This excess electricity reduces more carbon-intensive electricity demand from regional power plants and associated emissions, resulting in a credit being applied back to the original pathway.19 Despite this, the ISO preference for system expansion is not uniformly agreed upon, or adopted by LCA practitioners.20,21 While the substitution method is generally advocated for conducting LCAs, an unavoidable consequence of its implementation is the outcome can vary greatly depending on the assumptions made in the assessment for processes which produce economically useful coproducts in addition to the fuel.22,23 This situation is further exacerbated by the results from LCA being highly sensitive to definitions of system boundaries, life cycle inventories, process efficiencies, and functional units.24,25 Synthetic paraffinic fuels, whether produced from Fischer Tropsch synthesis, hydrogenation of natural oils or direct synthesis by bacteria all have very low levels of sulfur (e1 ppm) and aromatics (<1%) relative to petroleum derived fuels. The GTL process can produce GTL diesel, GTL naphtha, and liquefied petroleum gas (LPG) and potentially many other coproducts, such as GTL lubricant base oils, GTL aviation jet kerosene, GTL waxes, and normal paraffins, which all share the same qualities of very low levels of sulfur and aromatics while at the same time being rich in hydrogen relative to conventional analogues.26 Typically, GTL diesel is the highest yield product produced from the GTL process (40 80%, depending on product slate). The high cetane number (>74) and very low levels of sulfur (e1 ppm) and aromatics (<1%) in GTL diesel ensure a more efficient and cleaner-burning combustion environment, affording a substantial reduction in engine wear and exhaust emissions, including particulate matter (30% reduction in optimized engines), nitrogen oxides (45% reduction in optimized engines), carbon monoxide (85% reduction in optimized engines), and hydrocarbons (60% reduction in optimized engines), contributing to improvements in local air quality relative to conventional diesel.27 29 Studies to assess the emissions reduction from a diesel engine when using GTL diesel, as well as different blends of GTL diesel showed that the impact of GTL diesel blend concentration is often nonlinear, giving greater emissions benefits than expected at low concentrations of GTL diesel.30 In addition, a large potential for further reductions in soot and NOx emissions has been
ARTICLE
identified if the engine is recalibrated for optimum use of the GTL diesel.31,32 As local air quality problems persist, there is continued pressure to further reduce soot and NOx emissions from diesel engines.33 Lower levels of aromatic hydrocarbons in diesel fuel reduce emissions of both particulate matter and NOx.34 Despite this, since aromatics are a substantial component of diesel fuel refined from crude oil, and cannot be converted into other diesel components easily and without considerable use of energy and hydrogen, complete elimination of aromatics from crude-oil derived diesel is not a realistic option in most cases. Similarly, technologies used by refiners to remove sulfur from diesel fuel are least effective in reducing sulfur in so-called sterically hindered dibenzothiophene and similar species, which are typically the only ones remaining in less than 15 ppm diesel fuel, commonly referred to as ultralow sulfur diesel, or ULSD.35,36 Thus, although diesel fuel specifications have increasingly been tightened to meet environmental objectives, and sulfur and aromatics specifications are gradually moving toward those of GTL diesel, commercially available crude-oil derived diesel fuel is unlikely to ever meet these values. Recently, the California Air Resources Board (CARB) has recognized that the superior qualities of GTL diesel can be exploited to reduce or eliminate NOx increases observed in biodiesel-containing blends, mandated by renewable fuels standards.37 These fundamental superior physical properties extend to all GTL products, which can result in a downstream GHG emissions benefit relative to petroleum-derived analogous products, and should require LCA practitioners to thoroughly investigate the GHG emissions benefits of all GTL products using ISOpreferred substitution methodology. LCAs of the GTL pathway exist in the literature, but data on the GHG emissions benefits of GTL coproducts is lacking. LCAs of the GTL diesel pathway that have not employed coproduct substitution methodology because of lack of available data and substitution complexities have afforded lifecycle GHG emissions slightly higher than current petroleum-derived conventional diesel baselines, 38 53 while LCAs of GTL diesel that have used the systems boundary expansion method have generally produced lifecycle GHG emissions equivalent to current conventional diesel baselines.54 58 On the basis of the foregoing, there exists a need to comprehensively quantify and update as many of the GHG coproduct emissions benefits attributable to the GTL pathway as possible and to integrate these to determine the lifecycle GHG emissions of GTL diesel. This study builds on industrial knowledge of the GTL process and GTL products to demonstrate how by using the substitution method, the GHG emissions benefits associated with GTL coproducts affords significant carbon dioxide (CO2) benefits downstream from the production phase, resulting in a different WTW emissions result compared to using the energy allocation method.
’ METHODOLOGY AND ASSUMPTIONS Overview of Systems Boundaries. The modeling boundary for this study is from wellhead to wheels, which includes embodied energy and materials for raw gas extraction, gas processing, the GTL process, final product distribution and combustion (Figure 1). This includes a scenario where GTL diesel is used as a neat fuel (scenario A), and a scenario where GTL diesel is blended with petroleum-derived diesel (scenario B). 9085
dx.doi.org/10.1021/es202101b |Environ. Sci. Technol. 2011, 45, 9084–9092
Environmental Science & Technology
ARTICLE
Figure 1. Gas-to-liquids (GTL) process systems boundary with coproduct displacement.
Figure 1 shows how the model has been expanded to include analysis of GTL coproduct GHG emissions benefits relative to conventional, petroleum derived analogues and an integrated calculation of the net energy and GHG emissions due to impacts from inputs, outputs, processes and coproduct displacement for GTL diesel production. Main Assumptions for Each Lifecycle Stage. The GTL pathway model evaluated in this study is based on global current or imminent GTL production, and therefore is not necessarily indicative of any specific GTL plant in any specific country, but represents an industrially relevant average of the GTL process. Just as no two oil refinery configurations or operations are identical, GTL plant configurations and product slates will differ depending on local market need. As such, there exists a need to evaluate GTL refineries as an industry average, in a similar fashion to the way that crude oil refineries are assessed. Since only a single product slate from raw gas (LPG; 8.4%, condensate; 17.8%, GTL Naphtha; 23.1%, GTL diesel; 40.7%, GTL normal paraffin; 2.4%, GTL lubricant base oils; 7.5%) was considered in the current study, a different mix of GTL products could afford GHG emissions different to those presented here. Additional detailed information about the GTL product slate and material and energy inputs and outputs for the GTL process is contained in the Supporting Information (SI). Raw Gas Extraction and Processing. In this study it is assumed that conventional natural gas is extracted offshore and processed onshore at a gas processing facility. During extraction, it is assumed that 13.6 g/m3 raw gas is lost as vented (0.4% of raw gas) or flared gas (0.2% of raw gas). Sulfur produced in the gas processing plant is recovered as elemental sulfur. GTL Processing. The model GTL facility is located in Qatar. In this study GTL total energy efficiency (67.6%) is defined as the
fraction of internal energy in the raw gas and unstabilized condensate stream that is present in the carbon product streams.59 This figure includes all internal energy present in the stabilized field condensate stream. GTL total carbon efficiency (81.8%) is defined as the fraction of carbon in the raw gas and unstabilized condensate stream that is present in the carbon product streams. This figure includes all carbon present in stabilized field condensate and CO2 in the feed gas stream. GTL Diesel Distribution and Combustion (Scenario A). The major product from the GTL process, GTL diesel, is assumed to be transported to California and used as a neat fuel in a compression-ignition direct ignition (CIDI) diesel engine. GTL Diesel Distribution, Blending to CARB Diesel and Combustion (Scenario B). Scenario B is identical to Scenario A, except GTL diesel from the GTL process is blended in a California refinery to produce CARB diesel, which is subsequently used in a CIDI diesel engine. The treatment of the refinery blending GHG emissions benefit is detailed in the following section. GHG Emissions Reductions of GTL CoProducts. Since many of the GHG emissions reductions of the GTL process occur downstream of the GTL facility itself, correct handling of coproduct GHG emissions benefits are essential to derive an accurate WTW GHG emissions profile of the GTL pathway. In this study avoided impacts are accounted for by using the substitution method (Figure 1). The sections below summarize the data sources, methods and assumptions employed in the analysis of avoided GHG emission impacts. Additional detailed information can be found in the SI. Condensate and LPG. The associated upstream production process that operates to feed the GTL plant, depending on the characteristics of the gas reservoir and the amount of condensate contained in the produced gas, can produce significant volumes 9086
dx.doi.org/10.1021/es202101b |Environ. Sci. Technol. 2011, 45, 9084–9092
Environmental Science & Technology of gas condensate, which can add substantially to the overall liquid product output and value of an integrated upstream/GTL project. Therefore, when assessing the overall WTW GHG emissions impact of GTL (and other natural gas-based fuel pathways), the associated gas condensate produced by the upstream portion of the facilities needs be carefully considered. The gas processing facility separates LPG and condensate from the methane-rich feed gas to the GTL plant at carbon efficiencies (98%) that are significantly higher than those for the GTL process alone (79.1%). The net effect is an increase in the overall efficiency of the entire system relative to the GTL process alone. In this study it is assumed that LPG is burned locally in a commercial boiler, and displaces natural gas, which is the major energy source in Qatar. Condensate is assumed to be shipped from the Middle East to China and India (average of 4000 nautical miles) and displaces petroleum-derived naphtha. GTL Naphtha. Steam cracking of GTL Naphtha results in a higher olefin yield than cracking conventional naphtha alone, and is substantially more selective to the production of ethylene, propylene and butadiene.60 In addition, because GTL Naphtha has an essentially zero aromatics content, cracking of GTL Naphtha results in reduced coking of furnace tubes and catalyst relative to conventional naphtha, allowing for extended run durations. Consequently, the GHG emissions intensity of GTL Naphtha cracking is lower than petroleum-derived naphtha cracking. GTL Normal Paraffin. The GTL C9 C14 normal paraffin stream is an ideal feedstock for linear alkyl benzene (LAB) or linear alkyl benzene sulfonate (LAS) production for detergent manufacture because it is linear and has essentially zero sulfur and aromatics.61 Conventional normal paraffin production requires energy intensive prefractionation, hydrotreating and separation from petroleum-derived kerosene to produce the LAB or LAS feedstock.62 To simplify the case, in this study it is assumed that fully saturated GTL normal paraffin production displaces petroleum derived normal paraffin production.63 It is assumed that GTL normal paraffin and petroleum derived normal paraffin are of equal quality, meaning processes downstream of normal paraffin production are equal in GHG intensity. This assumption is conservative, because an alternative scenario for GTL normal paraffin extraction is separation of an olefin-rich (25% olefin) C9 C14 normal paraffin stream from the Fischer Tropsch unit, which would significantly reduce the need for the energy intensive dehydrogenation step that is required in LAB production. GTL Lubricant Base Oils. GTL lubricant base oils are high in saturates, have excellent oxidation stability, are virtually sulfur free and possess a high viscosity index.64 GTL lubricant base oils can be used to produce 0W-20 and 5W-20 motor oil grades 65,66 and provide significant quality benefits over petroleum derived lubricant base oils, matching in many areas those of the more energy-intensive chemically derived synthetic lubricant base oils (polyalphaolefins or PAO’s). These GTL lubricant base oils have a fuel efficiency advantage67 over petroleum-derived lubricant base oils, and the fuel efficiency benefit is gained from the high viscosity index of these oils, enabling the oils to provide good lubrication (i.e., sufficient viscosity) to the engine over a broad range of operating temperatures. According to the GF-5 standard recently introduced for passenger car applications, high performance fuel efficient SAE 0W and 5W oils will be required to provide fuel efficiency benefits in the region of 0.5 1.2%.68 Currently, the lack of suitable lubricant base oils is restricting widespread implementation; GTL69 and
ARTICLE
hydrogenated vegetable oil (HVO)70 lubricant base oils are able to provide a solution for this lack of supply. In this study it is assumed that GTL lubricant base oils (50% 0W and 50% 5W) displace existing petroleum Group II lubricant base oils in the U.S. market, and generate an average fuel efficiency saving of 0.85% against petroleum derived lubricant base oils.71 Relative to Group II lubricant base oils, GTL lubricant base oils (Group III+) impart less internal engine frictional losses, or less drag on the crankshaft, pistons and valve train, which in turn promotes increased vehicle fuel efficiency. In addition to providing fuel efficiency savings, Fischer Tropsch derived lubricant base oils are extremely stable and could enable greatly extended drain intervals. Typical drain intervals in North America are in the region of 5000 8000 miles.72 A reasonable drain interval for 0W and 5W oils such as those produced by GTL lubricant base oils in normal service with certain manufacturer’s vehicles could be in excess of 15 000 miles.73 Despite the longer drain intervals of high quality lubricant base oils that are being supported by original equipment manufacturer (OEM) warranties, a change in consumer habits is required, thus in the current study it is conservatively assumed that drain intervals between GTL lubricant base oils and conventional base oils are identical. Additional details of the benefits of GTL lubricant base oils can be found in the SI. GHG Emissions Reductions of Refinery Blending of GTL Diesel (Scenario B). Although GTL diesel can be used as a pure, neat diesel fuel (Scenario A), a refiner has the option of utilizing it as a high quality diesel blending component (Scenario B). GTL diesel is an already refined, hydrogen-rich, essentially zero-aromatic fuel,74 which when blended in a refinery could potentially reduce the severity of refining processing required to produce, for example, CARB diesel, thus reducing incremental CO2 intensive hydrogen production. When GTL diesel is blended with diesel fuel derived from crude oil, the sulfur and aromatics content and density of the blend will typically be reduced, and the cetane number increased. All of these changes are directionally favorable in terms of the pressures exerted on refiners by clean fuel specifications in recent years. In contrast, intermediate diesel components, such as fluid catalytic cracked (FCC) light-cycle oil (LCO), are short in hydrogen because they are rich in aromatics. For example, LCO generally contains about 60 85 wt % aromatics.75 In order to send LCO to the diesel pool, it needs to be subjected to simultaneous desulfurization, denitrogenation, hydrogenation, and dearomatization.76 These highly aromatic streams can consume 3 5 times as much hydrogen as virgin distillates when producing diesel products. In addition to being more difficult to desulfurize, these highly aromatic streams can result in a diesel product that does not meet the cetane or density specifications, especially when a refiner processes high-aromatic crudes.77,78 According to the U.S. Department of Energy, U.S. demand for hydrogen is currently about 9 million tons per year.79 Refiners obtain about 95% of this hydrogen through steam reforming of natural gas80 by either purchasing hydrogen (through local infrastructure) or by building a hydrogen plant in the refinery. The steam methane reforming (SMR) process used in centralized plants to produce hydrogen for hydrotreaters emits significantly more CO2 than hydrogen produced on a weight basis. An LCA of hydrogen production via SMR indicates that the overall global warming potential (GWP) of the SMR system is 11.9 kg CO2e per kg of hydrogen produced.81 Thus, because more severe refinery processing is required to lower diesel aromatics content,82 9087
dx.doi.org/10.1021/es202101b |Environ. Sci. Technol. 2011, 45, 9084–9092
Environmental Science & Technology
ARTICLE
Table 1. Net GHG Emissions of the GTL Diesel Pathway
coproducta
stage
scenario
scenario B:
A: neat
GTL diesel
GTL
as blend
dieselb
componentc
gCO2e/ MJ
gCO2e/MJ
NG extractiond
12.1
12.1
NG transporte
0.5
0.5
gas pretreatmentf
7.7 5.7
7.7 5.7
Feedstock Production
condensateg LPGh Total
0.9
0.9
15.5
15.5
35.1
35.1
Fuel Production GTL planti GTL naphthaj
10.1
10.1
LPG GTL lubricant base oilsk
0.4 26.3
0.4 26.3
0.7
0.7
GTL normal paraffinl refinery blendingm distributionn
0.0 2.8
5.4 2.8
Total
1.2
-4.2
Operation
72.0
74.9
Total WTW
88.7
86.2
Emissions a
The substitution (displacement) method was used to estimate energy and emission credits of GTL coproducts. b Neat GTL Diesel is used as a fuel in a compression-ignition direct ignition (CIDI) diesel engine. c GTL diesel is blended in a California refinery to produce CARB diesel, which is used in a CIDI diesel engine. d Extracted offshore and processed onshore at a gas processing facility. Includes provision for venting and flaring (0.4% of the raw gas is vented, and 0.2% is flared). e Raw gas is transported 47 miles by pipeline to Gas Treatment Plant (GTP). f Natural gas recovery efficiency is 98%. g Displaces petroleum derived naphtha. h Displaces natural gas. i GTL total energy efficiency 67.6%, total carbon efficiency 81.8%. j Displaces petroleum derived naphtha. k Displaces existing petroleum Group II lubricant base oils in the U.S. market. l Displaces petroleum derived normal paraffin production. m Blending of GTL diesel in California refineries. n Shipped 7200 nautical miles to California. Road transport 50 miles.
and increased refinery processing produces more GHG emissions, incremental CO2 intensive hydrogen production could potentially be reduced when hydrogen rich GTL diesel is blended in California petroleum refineries to produce CARB diesel. Analysis of the possible ways GTL diesel can be blended in California petroleum refineries reveals a large degree of potential complexity. For example, under a constant CARB diesel output scenario, a refiner can potentially purchase GTL diesel and displace the hardest to treat diesel components, such as LCO and light coker gas oil (LCGO), to sales as residual fuel oil blendstock. Alternatively, a refiner could purchase GTL diesel, lower their crude throughput, and reduce the severity of their operations to achieve the same yield of CARB diesel. This is further exacerbated by the fact that no two petroleum refineries in California have the same configurations.83 Each petroleum
refinery is likely to have different crude feedstocks, cracking and treatment severities of intermediate streams to produce different yields of final products. In short, each refiner will find a different optimum use for GTL diesel to lower their GHG emissions to produce their required product slate. Notwithstanding the above, in the current study the complexity required to model what is described above was balanced with the need to develop a simpler model to assess a realistic GHG emissions credit for blending GTL diesel in a California petroleum refinery. In the current study, three different blending scenarios were developed. In the first scenario, the incremental hydrogen required to produce CARB diesel84 from ULSD was avoided by blending 0.364 volume equivalents of GTL diesel (1% aromatics max) with 0.636 volume equivalents of ULSD (25.7% aromatics) to make 1 volume equivalent of CARB diesel (16.7% aromatics). In the remaining scenarios, LCGO or LCO was removed from the hydrocracker feed in different refinery configurations and displaced to residual fuel oil with an equivalent volume of GTL diesel added as diesel blendstock. This resulted in a reduced feed to diesel hydrodesulfurisation (HDS) or hydrodearomatization (HDA) units, and the energy displacement was calculated. For each of these scenarios, to simplify the analysis, it was assumed that there was no change in the total volume of CARB diesel production. Thus, each barrel of GTL diesel added to the blend pool displaces a barrel of LCO or LCGO into the residual fuel oil pool. This simplification is necessary, since the alternative scenario involves cutting the input crude oil rate, which would require adjustment of many refinery variables, such as adjustment in distillation cutpoints and cracking severities. If this latter approach was taken, an array of possible results could be obtained, depending on the operation adjustments that were made to accommodate the reduction in crude oil feed rate. Lifecycle Assessment Model. Details of refinery simulation models, LCA software, system boundaries and allocation methodology can be found in the SI. Emissions are represented in grams of CO2 equivalent per megajoule (MJ) of fuel content (g CO2e/MJ).
’ RESULTS AND DISCUSSION GHG Emissions Results. Table 1 presents the comparison of the GHG components for the WTW GHG emissions of neat GTL diesel (scenario A) and GTL diesel used as a CARB diesel blend component (scenario B), respectively. The detailed GHG emissions for GTL diesel including combustion are presented in the SI. For the GTL simulations carried out in the current study, the WTW GHG emissions are characterized by net production phase emissions that are close to zero. This can be rationalized by the relative carbon-balancing effect of coproduct credits resulting from the superior physical properties of GTL products relative to petroleum-derived analogues. In this instance, the substitution method has a significant effect on the GTL WTW emissions results. Further examination reveals that GTL lubricant base oils have a large and positive effect on the overall WTW GHG emissions. Although the fuel efficiency benefit of replacing a Type II lubricant base oil with a high quality (Group III+) GTL lubricant base oil is relatively small (+ 0.85%), over a 5000 mile drain interval period, this is equivalent to a saving of 20 100 g of CO2e per drain interval (4.0 g CO2e/mile driven by a gasoline powered motor vehicle using GTL lubricant base oils). A sensitivity analysis (see SI) reveals that additional CO2e benefits are possible with greater fuel efficiency gains and drain intervals. 9088
dx.doi.org/10.1021/es202101b |Environ. Sci. Technol. 2011, 45, 9084–9092
Environmental Science & Technology
ARTICLE
Table 2. Comparison of GHG Emissions of the GTL Diesel and Petroleum Diesel Pathways Using Substitution and Allocation Methodologies GTL diesel: scenario Aa
GTL diesel: scenario Ba
GTL dieselb
ULSDb
CARB dieselc
gCO2e/MJ
gCO2e/MJ
gCO2e/MJ
gCO2e/MJ
gCO2e/MJ
substitution
substitution
allocation
allocation
allocation
feedstock
20.3
20.3
7.9
5.1
8.1
fuel
37.9
37.9
23.2
10.2
11.7
operation
72.0
74.9
72.1
75.8
74.9
0.0
0.0
0.0
103.1
91.1
94.7
methodology
coproducts total a
41.6 88.7
47.0 86.2
Current Study. b Reference Value, GREET 1.8d. c Reference Value, CARB.
Note that this fuel efficiency benefit applies to a large number of vehicles (90%+ of the U.S. vehicle fleet), both old and new, and is totally independent of driver behavior. The potential for fuel efficiency gains would be significantly greater in developing countries where GTL lubricant oils would displace even lowerquality (Group I) lubricant oils. For example, if a GTL lubricant oil were to replace a Group I lubricant base oil (currently used in significant quantity in China), then the fuel efficiency gains could be 3% or higher.85 A sensitivity study of the effect of GTL lubricant base oils on the WTW GHG emissions of GTL diesel can be found in the SI. For the case where GTL diesel is used as a blend component for CARB diesel (scenario B), a slightly lower WTW GHG emissions value is obtained relative to the case where neat GTL diesel is used as a neat fuel in a CIDI vehicle (scenario A). Although the difference in GHG emissions between both scenarios are relatively small, the greater operation phase emissions of CARB diesel relative to GTL diesel is more than offset by GHG emissions savings that occur when hydrogen rich GTL diesel is blended in California refineries to produce CARB diesel. Similar blending GHG emissions reductions would be expected to accrue to other hydrogen-rich fuel sources such as products derived from biomass-to-liquids (BTL) and HVO processes. These results suggest that the optimal reduction in refining intensity could be achieved when GTL diesel is blended with heavy crudes that require severe hydrotreating, such as Venezuelan heavy crude oil or Bitumen derived from Canadian oil sands and in jurisdictions with tight aromatic specifications for diesel, such as California. In the absence of tight aromatic fuels specifications, blending GTL diesel into conventional diesel, such as ULSD, would likely be identical to using neat GTL diesel. Table 2 presents a comparison of the WTW GHG emissions results for the GTL pathway in the current study that uses the substitution method with reference values for GTL diesel (GREET 1.8d), ULSD (GREET 1.8d), and CARB diesel (CARB) that use allocation methodology. A notable component is the significant effect that substitution methodology has on the WTW results for the GTL pathway. Using scenario B results in the current study as an example, the WTW GTL emissions of GTL diesel using the substitution method (86.2 gCO2e/MJ) are 17% lower than when allocation is used (103.1 gCO2e/MJ, GREET 1.8d). Using the substitution method, the WTW GHG emissions of the GTL diesel pathway are lower than petroleum diesel references, such as ULSD and CARB diesel. In the current study, the feedstock and fuel GHG emissions of GTL diesel under a substitution approach are much larger than
under an allocation approach. The reason is that the substitution emissions have to be considered as gross emissions. The net emissions are obtained by subtracting the substitution benefits of the coproducts to the gross emissions. Indeed, under the substitution approach, 3.1 kg of natural gas is needed to produce 1 kg of GTL diesel and associated coproducts. The quantity of natural gas is much lower (1.2 kg of natural gas) under the allocation scenario because feedstock GHG emissions correspond to only a proportion of natural gas which is needed to produce GTL diesel, excluding coproducts. Similarly, the emissions of the fuel production step are much higher under the substitution approach because the GHG emissions correspond to the impact related to the production of GTL diesel as well as the associated coproducts. This is why it is important to consider the total of the gross GHG emissions and the coproducts credits from the substitution approach before making any comparison with the results obtained with the allocation approach. In principle, because the substitution method was used in the current study to capture the full emissions benefits of GTL that are ignored by using allocation, the potential for distorted results exists. Several authors 22,52 have recently noted that the substitution method can generate distorted LCA results if the coproducts are actually main products (for the cases of biodiesel and renewable diesel from soybeans). In the current study GTL diesel comprises 36% (by weight) of the raw gas stream, and 53% of the GTL process, while the most abundant coproduct of the raw gas stream is condensate (27%). However, the emissions from GTL are comparable in magnitude to petroleum fuels, which is why distorted results are not being generated in this study. Several issues suggest future work. Using the substitution method, the composition of the product slate from the GTL pathway appears to have an influence on the final lifecycle GHG emissions of GTL diesel, implying that the environmental advantages of GTL could be optimized with changes to the composition of the GTL product slate. Given the relevancy of the refinery blending results in the current study to other paraffinic fuels, such as HVO, a sensitivity analysis of major parameters used when blending GTL diesel and refinery intermediates could be of value. Technology and system-level improvements suggest further environmental benefits of GTL. For example, recent results suggest that combustion at altitude of a 1:1 blend of synthetic paraffinic kerosene (SPK) aviation fuel, obtainable from the GTL process, and petroleum-derived JetA kerosene results in significantly less ice particles being formed in contrails with particles formed having a larger average diameter (average ice particle diameter = 2 μm) relative to combustion of JetA kerosene alone (average ice 9089
dx.doi.org/10.1021/es202101b |Environ. Sci. Technol. 2011, 45, 9084–9092
Environmental Science & Technology particle diameter = 1 μm). This has the effect of reducing the optical depth and radiative forcing effect of linear contrails from 12 mW/m2 in JetA kerosene to 5 mW/m2 in the 1:1 SPK/JetA blend.86 The consequence of these effects would be a smaller overall radiative forcing (RF) impact from combustion of the 1:1 SPK/JetA blend relative to combustion of JetA alone. Although the level of scientific understanding remains low, linear contrails87 and contrail cirrus88 are generally thought to be significant radiative-forcing components associated with aviation. In summary, to the best of our knowledge this study represents the most comprehensive investigation of the WTW GHG emissions of the GTL pathway to date. These results suggest that the GTL pathway can realize GHG emissions reductions relative to a petroleum diesel baseline. A notable component of the GTL pathway is the displacement of coproducts that occur by (a) avoiding increases in energy and processing intensity associated with carbon rejection, hydrogen addition and supporting energy intensive processes by displacement of conventional product streams with higher quality GTL products and (b) downstream GHG emissions reductions of the GTL production phase that occur when higher quality GTL products displace conventional products. The GTL process incurs the cost of producing high quality hydrogen rich products that are virtually free of aromatics and sulfur at the production phase of the process by having thermal and carbon efficiencies lower than a petroleum diesel pathway. This penalty is more than compensated for by the coproduct displacement credits of GTL products described herein. On this basis, it is important to take into account coproducts in life-cycle evaluation of the energy use and environmental effects of GTL diesel. Until now, most LCA studies of the GTL pathway have chosen not to do so because of lack of available data and substitution complexities. These results help to explain differences in previous estimates of lifecycle GHG emissions of the GTL pathway with those obtained in the current study. When coproduct displacement credits are not considered (energy allocation methodology), the lifecycle GHG emissions of the GTL pathway are typically slightly higher than the petroleum-derived diesel pathway. However, when the full GHG emissions benefits of GTL diesel and coproducts are considered using substitution methodology, the lifecycle GHG emissions of the GTL pathway are lower than the petroleum-based diesel baseline. By producing hydrogen-rich paraffinic products that are free of aromatic and sulfur compounds, the net GHG emissions avoidance that can be realized by GTL coproducts represents a significant difference between the GHG emissions profiles of other fossil fuel pathways. In this context, the GTL process is more akin to a gas refinery which produces premium products that exceed even the tightest product specifications. Given the increasing abundance of unconventional natural gas resources in North America, GTL appears capable of contributing to domestic energy security, while providing benefits to the U.S. economy and reducing existing GHG emissions from current crude oil import sources.
’ ASSOCIATED CONTENT
bS
Supporting Information. Data and details of methods, coproduct allocation assumptions, analyses and results. This material is available free of charge via the Internet at http://pubs. acs.org.
ARTICLE
’ AUTHOR INFORMATION Corresponding Author
*Phone: 1-281-588-3445; e-mail:
[email protected].
’ ACKNOWLEDGMENT We gratefully acknowledge the support of Sasol Synfuels International by providing data and giving permission to publish this manuscript. ’ REFERENCES (1) Leiby, P. N. Estimating the Energy Security Benefits of Reduced U.S. Oil Imports; Prepared for the U.S. Department of Energy, Oak Ridge National Laboratory: Oak Ridge TN, February 2007. (2) America’s climate choices: Panel on advancing the science of climate change. In Advancing the Science of Climate Change, National Research Council; Washington, D.C.: The National Academies Press, 2010. (3) U.S. Energy Information Administration, Annual Energy Outlook 2011, Early Release Overview. http://www.eia.gov/neic/speeches/ newell_12162010.pdf (accessed March 2011). (4) Affordable, Low-Carbon Diesel Fuel from Domestic Coal and Biomass; National Energy Technology Laboratory/U.S. Department of Energy, 2009. (5) Simeroth, D. California’s Low Carbon Fuel Standard, California Air Resources Board; Hart’s World Refining & Fuels Conference, May 2007. (6) Sissine, F. Energy Independence and Security Act of 2007: a Summary of Major Provisions (RL34294). CRS Report for Congress, Washington DC, 2007. http://assets.opencrs.com/rpts/RL34294_ 20071221.pdf (accessed October 2010) (7) Sperling, D.; Sonia, Y. Toward a global low carbon fuel standard. Transp. Policy 2010, 17 (1), 47–49. (8) Andress, D.; Nguyen, T. D.; Das, S. Low-carbon fuel standard— Status and analytic issues. Energy Policy 2010, 38 (1), 580–591. (9) Wang, M.; Lee, H.; Molburg, J. Allocation of energy use and emissions to petroleum refining products: Implications for life-cycle assessment of petroleum transportation fuels. Int. J. Life Cycle Assess. 2004, 9 (1), 34–44. (10) Kaufman, A. S.; Meier, P. J.; Sinistore, J. C.; Reinemann, D. J. Applying life-cycle assessment to low carbon fuel standards—How allocation choices influence carbon intensity for renewable transportation fuels. Energy Policy 2010, 38 (9), 5229–5241. (11) Baumann H. Decision Making and Life Cycle Assessment. thesis. Chalmers University of Technology. Technical Environmental Planning Report 1995. (12) Wang, M. Well-to-wheels energy and greenhouse gas emission results and issues of fuel ethanol. In The Life-Cycle Carbon Footprint of Biofuels; Outlaw, J. L., Ernstes, D. P., Eds.; Farm Foundation: Oak Brook, IL, 2008; pp 19 34. (13) ISO 14040: Environmental Management Life Cycle Assessment Principals and Framework; International Organization for Standardization: Geneva, Switzerland, 2006a. (14) ISO 14044: Environmental Management Life Cycle Assessment Requirements and Guidelines; International Organization for Standardization: Geneva, Switzerland, 2006b. (15) Delucchi, M. A. A Lifecycle Emissions Model (LEM): Lifecycle Emissions from Transportation Fuels, Motor Vehicles, Transportation Modes, Electricity Use, Heating and Cooling Fuels, and Materials; Institute of Transportation Studies, University of California at Davis: Davis, CA, 2008. (16) 2008 GHGenius Update; (S&T)2 Consultants, Inc., prepared for Natural Resources Canada: Ottawa, Ontario, Canada, 2008. (17) E3 Database: An Introduction into the Life-Cycle Analysis Calculation Tool; LBST: 3 Ottobrunn, Germany, 2008. (18) Mendoza, A.; van Ruijven, T.; Vad, K.; Wardenaar, T. The Allocation Problem in Bio-Electricity Chains. M.Sc. Thesis, Industrial Ecology, Leiden, The Netherlands, 2008. 9090
dx.doi.org/10.1021/es202101b |Environ. Sci. Technol. 2011, 45, 9084–9092
Environmental Science & Technology (19) Ekvall, T.; Weidema, B. P. System boundaries and input data in consequential life cycle inventory analysis. Int. J. Life Cycle Assess. 2004, 9 (3), 161–171. (20) Ekvall, T.; Finnveden, G. Allocation in ISO14041—A critical review. J. Clean. Prod. 2001, 9, 197–208. (21) Bjorklund, A. Survey of approaches to improve reliability in LCA. Int. J. Life Cycle Assess. 2002, 7, 64–72. (22) Wang, M.; Huo, H.; Arora, S. Methods of dealing with coproducts of biofuels in life-cycle analysis and consequent results in the U.S. context. Energy Policy 2010, 39 (10), 726 5736. (23) Xie, X.; Wang, M. ; Han, J. Assessment of fuel-cycle energy use and greenhouse gas emissions for fischer tropsch diesel from coal and cellulosic biomass. Env. Sci. Technol. 2011, 45 (7), 3047 3053. (24) Farrell, A. E.; Plevin, R. J.; Turner, B. T.; Jones, A. D.; O’Hare, M.; Kammen, D. M. Ethanol can contribute to energy and environmental goals. Science 2006, 311 (5760), 506–508. (25) Hill, J.; Nelson, E.; Tilman, D.; Polasky, S.; Tiffany, D. Environmental, economic, and energetic costs and benefits of biodiesel and ethanol biofuels. Proc. Natl. Acad. Sci. U. S. A. 2006, 103 (30), 11206–11210. (26) Dancuart, L. P.; Steynberg, A. P. Fischer Tropsch based GTL technology: a new process? Stud. Surf. Sci. Catal. 2007, 163, 379–399. (27) Alleman T. L.; McCormick R. L.; Vertin, K. Assessment of Criteria Pollutant Emissions from Liquid Fuels Derived From Natural Gas; National Renewable Energy Laboratory: Golden, CO, April 2002. (28) Larsson, M. An Experimental Study of Fischer Tropsch Fuels in a Diesel Engine. thesis, Department of Applied Mechanics, Chalmers University of Technology, G€oteborg, Sweden, 2007. (29) Tsujimura, T.; Goto, S.; Matsubara, H. A study on PM emission characteristics of diesel vehicle fueled with GTL. Proc. J. SAE Annu. Congr. 2006, 54 (06), 1–6. (30) Maly, R. Effect of GTL diesel fuels on emissions and engine performance (DaimlerChrysler). In 10th Diesel Engine Emissions Reduction Conference, August 29 September 2, 2004, Coronado, California). http://www1.eere.energy.gov/vehiclesandfuels/pdfs/deer_ 2004/session2/2004_deer_maly.pdf (accessed February 2011). (31) H.O. Herrmann, H. O.; S. Keppeler, S.; Botha, J. J.; Schaberg, P.; Schnell, M. The Potential of Synthetic Fuels to meet Future Emission Regulations. In 27th Motor Symposium, Vienna, 2006. (32) Lilik, G. K.; Boehman, A. L. Advanced diesel combustion of a high cetane number fuel with low hydrocarbon and carbon monoxide emissions. Energy Fuels 2011, 25 (4), 1444–1456. (33) Nakamura, D. Product sulfur specs will determine future refining configurations. Oil Gas J., 2009, 107(13). (34) California Air Resources Board & California Environmental Protection Agency, California Diesel Fuel Fact Sheet, www.epa.gov/ttn/ atw/urban/dieselfs.pdf (accessed March, 2011). (35) Alleman T. L.; McCormick R. L.; Vertin, K. (NREL) Assessment of Criteria Pollutant Emissions from Liquid Fuels Derived From Natural Gas; National Renewable Energy Laboratory: Golden, CO, April 2002. (36) Thijssen, J. The Impact of Future Diesel Fuel Specifications and Engine Emissions Standards on SOFC; U.S. Department of Energy, National Energy Technology Laboratory, June, 2004. (37) Mitchell, M.; Okamoto, B. Biodiesel and Renewable Diesel Rulemaking, 2nd Public Workshop; California Air Resources Board: Sacramento, CA, May 2010. (38) Norton, P.; Vertin K.; Bailey, B. (National Renewable Energy Lab, NREL), Clark, N. N.; Lyons, D. W. (Wed Virginia Univ., WVU), Goguen, S.; Eberhardt, J. (US Department of Energy, DOE). Emissions from Trucks using Fischer Tropsch Diesel Fuel, International Fall Fuels and Lubricants Meeting and Exposition, San Francisco, California, October 19 22, 1998; SAE Technical Paper Series 982526. (39) Greene, D. L. An Assessment of Energy and Environmental Issues Related to the Use of Gas-to-Liquids Fuels in Transportation; Oak Ridge National Laboratory, Center for Transportation Analysis: Oak Ridge TN, November 1999. (40) Bagajewicz, M.; Sujo, D.; Martinez, D.; Savelski, M. Driving Without Petroleum?—A Comparative Guide to Biofuels, Gas-to-Liquids and Coal-to-Liquids as Fuels for Transportation, May 2007.
ARTICLE
(41) Johnson, E. Carbon Footprints of Biofuels & Petrofuels; SRI Consulting, October 2007. (42) Jaramillo, P.; Griffin, M.; Matthews, S. Comparative analysis of the production costs and life-cycle GHG emissions of FT liquid fuels from coal and natural gas. Environ. Sci. Technol. 2008, 42 (20), 7559–7565. (43) An Assessment of Energy and Environmental Issues Related to the Use of Gas-to-Liquid Fuels in Transportation; Oak Ridge National Laboratory for U.S. DOE: Oak Ridge, TN, 1999. (44) Kavalov, B. (JRC) Techno-Economic Analysis of Natural Gas Application As an Energy Source for Road Transport in The EU; Joint Research Center: Brussels, January, 2004. (45) Edwards R. Well-to-Wheels Analysis of Future Automotive Fuels and Powertrains in the European Context, Concawe, Eucar, European Commissions, well-to-wheels report, version 2a, December 2005. (46) Beer, T.; Grant, T.; Morgan G.; Lapszewicz, J.; Anyon, P.; Edwards, P.; Nelson, P.; Watson, H.; Williams D. Comparison of Transport Fuels -LifeCycle Emissions Analysis of Alternative Fuels for Heavy Vehicles; Australian Greenhouse Office, CSIRO, 2006. (47) Marano, J. J.; Ciferno, J. P. Life-Cycle Greenhouse-Gas Emissions Inventory for Fischer Tropsch Fuels; U.S. Department of Energy: Washington, DC, June 2001. (48) General Motors Well-to-Wheel Energy Use and Greenhouse Gas Emissions of Advanced Fuel/Vehicle Systems—North American Analysis, June 2001. (49) Wang, M.; Wu, M.; Huo H. Life-Cycle Energy and Greenhouse Gas Results of Fischer Tropsch Diesel Produced from Natural Gas, Coal, and Biomass; Center for Transportation Research, Argonne National Laboratory: Argonne, IL, May 2007. (50) Brandt, A. R; Farrell, A. E. Scraping the bottom of the barrel: greenhouse gas emission consequences of a transition to low-quality and synthetic petroleum resources. Clim. Change 2007, 84, 241–263. (51) Van Vliet, O. P. R.; Faaij, A. P. C.; Turkenburg, W. C. Fischer Tropsch diesel production in a well-to-wheel perspective: A carbon, energy flow and cost analysis. Energy Convers. Manage. 2009, 50, 855–876. (52) Stratton, R. W.; Wong, H. M.; Hileman, J. I. Quantifying variability in life cycle greenhouse gas inventories of alternative middle distillate transportation fuels. Environ. Sci. Technol., 2011, 10, 4637 4644. (53) Hao, H.; Wang, H.; Song, L.; Li, X.; Ouyang, M. Energy consumption and GHG emissions of GTL fuel by LCA: Results from eight demonstration transit buses in Beijing. Appl. Energy 2010, 87, 3212– 3217. (54) CONCAWE, EUCAR, JRC. Well to Wheels Analysis of Future Automotive Fuels and Powertrains in the European Context, March 2007. (55) An Assesment of Health and Environmental Impacts of GTL Diesel; Institute for Energy and Environmental Research (IFEU): Takoma Park, MD, August 2005. (56) Use of a Life Cycle Approach to Compare the Environmental Implications of Sasol Chevron’s Slurry Phase Distillate Technology with Alternative Processes producing Transportation Fuels—European case update; SasolChevron, March 2006. (57) Shell Middle Distillate Synthesis (SMDS) Update of a Life Cycle Approach to Assess the Environmental Inputs and Outputs, and Associated Environmental Impacts, of Production and Use of Distillates from a Complex Refinery and SMDS Route; Shell: The Hague, The Netherlands, May 2003. (58) Gas to Liquids Life Cycle Assessment Synthesis Report, ConocoPhillips, Sasol Chevron and Shell International Gas; Prepared by Five Winds International, August 2004. (59) Based on Higher Heating Value (HHV). (60) Dancuart, L.; Mayer, J.; Tallman, M. J.; Adams, J. Performance of the Sasol SPD Naphtha as Steam Cracking Feedstock. Pet. Chem. Div. Preprints 2003, 48 (2), 132. (61) Maase, B.; Dirkzwager, H. Gas to Liquids Paraffins and Their Impact on the Detergent Industry, CESIO Conference, 2004. (62) Franke, M.; Schwalbach; Berna, J. L.; Cavalli, L.; Renta, C.; Stalmans, M.; Strombeek-Bever, M.; Thomas, H. A Life cycle inventory for the production of petrochemical intermediates in Europe: Paraffins, olefins, benzene, ethylene and ethylene oxide. Tenside, Surfactants, Deterg. 1995, 32, 384. 9091
dx.doi.org/10.1021/es202101b |Environ. Sci. Technol. 2011, 45, 9084–9092
Environmental Science & Technology (63) Normal Paraffins—World Markets 2000 2010; Colin A. Houston & Associates, Inc. (CAHA): Aiken, SC. (64) Beret, S.; Deskin, S.; Rosenbaum, J.; Van der Mervwe, D. Lubricants from Gas-to-Liquids (GTL) Base Oils; Lubricants (Chevron), December 2006. (65) Kobayashi, M.; Saitoh, M.; Togawa, S.; Ishida, K. Branching structure of diesel and lubricant base oils prepared by isomerization/ hydrocracking of Fischer Tropsch Waxes and α-olefins. Energy Fuels 2009, 23 (1), 513–518. (66) Clark, R. H.; Wedlock D. J.; Cherillo, R. A. Future fuels and lubricant base oils from shell gas to liquids (GTL) technology, SAE Series Technical Paper Series, 2005-01-2191; SAE International: Warrendale, PA. (67) 0W-20 and 5W-20 motor oil grades have a lighter viscosity than lower quality oils, such as 5W-30 and therefore has less internal engine frictional losses, or less drag on the crankshaft, pistons and valvetrain, which in turn promotes increased fuel efficiency. (68) ILSAC GF-5 Standard for Passenger Car Engine Oils. http:// www.gf-5.com/uploads/File/ILSAC_GF-5_Dec-22-09_final.pdf (accessed October, 2010). (69) Dos Santos, A. A. C. B.; Sousa-Aguiar, E. F. Surging GTL promises lube market opportunities, Oil Gas J., 2010, 108(28), 98(6). (70) Lehmus, M. (Neste Oil) Contribution of High Quality Base Oils to Cleaner Traffic. In 15th World Base Oils & Lubricants Conference, London, U.K., February 2011. (71) Tanaka, H.; Nagashima, T.; Sato, T.; Kawauchi, S. (Honda R+D Co., Ltd) The Effect of OW-20 Low Viscosity Engine oil on Fuel Economy, SAE Series Technical Paper Series, 1999-01-3468; SAE International: Warrendal, PA. (72) Extended drain intervals are supported by original equipment manufacturer (OEM) warranties; see Amsoil Product Recommendation and Drain Interval Chart. www.barmacinc.com/amsoilpdfs/.../ AMSOIL%20Product%20Change%20Interval%20Guide.pdf (accessed March, 2011). (73) McFall, D. Drain Intervals: How Long Must We Wait?, March 2003, Lubes’n’creases, www.donsoil.com/drainintervals.pdf (accessed March, 2011). (74) Effect of GTL Fuels on Emissions and Engine Performance, Daimler Chrysler presentation September 2, 2004. (75) Light Cycle Oil (LCO) Material Safety Data Sheet. Houston Refining, 2006. (76) O’Rear, D. J.; Bacha, J. D.; Tiedemann, A. N. Thermally stable blends of Fischer Tropsch and LCO diesel fuel components. Energy Fuels 2004, 18, 682–684. (77) Karras, G. Combustion emissions from refining lower quality oil: What is the global warming potential? Environ. Sci. Technol. 2010, 44 (24), 9584–9589. (78) Anabtawi, J. A.; Ali, S. A.; Ali, M. A. Impact of gasoline and diesel specifications on refining industry. Energy Sources 1993, 18, 203–214. (79) Zhai, Y.; Pierre, D.; Weiling Deng, R. S.; Ferrin, P.; Nilekar, A. U.; Peng, G.; Herron, J. A.; Bell, D. C.; Saltsburg, H.; Mavrikakis, M.; Flytzani-Stephanopoulos, M. Alkali-stabilized Pt-OHx species catalyze low-temperature water-gas shift reactions. Science 2010, 329, 1633–1636. (80) Best Available Techniques for Hydrogen Production by Steam Methane Reforming, IGC Document 155/09/E; European Industrial Gases Association AISBL, 2009. (81) Spath P. L.; Mann, M. R. Life Cycle Assessment of Hydrogen Production via Natural Gas Steam Reforming, NREL/TP-570-27637; National Renewable Energy Laboratory: Golden, CO, February 2001. (82) Wang, M. Estimation of Energy Efficiencies of U.S. Petroleum Refineries; Center for Transportation Research, Argonne National Laboratory: Argonne, IL, March 2008. (83) Worrell, E.; Galitsky, C. Profile of the Petroleum Refining Industry in California; California Industries of the Future Program, Energy Analysis Department, Lawrence Berkeley National Laboratory: Berkeley, CA, July 2003. (84) Diesel Fuel Comparison Study Workshop; California Environmental Protection Agency, California Air Resources Board (CARB): Sacramento, CA, January 8, 2009.
ARTICLE
(85) Total, Fuel Economy Lubricants for Passenger Cars, http://www. total-fuel-economy-lubricants.com/#/home (accessed March, 2011). (86) Sustainable Way for Alternative Fuels and Energy in Aviation; Final Report; April 2011: Submitted to the European Commission under contract Tren/F2/ 408.2008/SI2.518403/SI2.519012. http:// www.swafea.eu/LinkClick.aspx?fileticket=llISmYPFNxY%3d&tabid=77 (accessed August, 2011). (87) Lee, D. S.; Pitari, G.; Grewec, V.; Gierens, K.; Penner, J. E.; Petzold, A.; Prather, M. J.; Schumann, U.; Bais, A.; Berntsen, T.; Iachetti, D.; Lim, L. L.; Sausen, R. Transport impacts on atmosphere and climate: Aviation. Atmos. Environ. 2010, 44, 4678–4734. (88) Burkhardt, U.; K€archer, B. Global radiative forcing from contrail cirrus. Nature Clim. Change 2011, 1, 54–58.
9092
dx.doi.org/10.1021/es202101b |Environ. Sci. Technol. 2011, 45, 9084–9092
ARTICLE pubs.acs.org/est
Novel Online Monitoring and Alert System for Anaerobic Digestion Reactors Fang Dong,† Quan-Bao Zhao,† Wen-Wei Li,*,† Guo-Ping Sheng,† Jin-Bao Zhao,† Yong Tang,† Han-Qing Yu,*,† Kengo Kubota,‡ Yu-You Li,‡ and Hideki Harada‡ † ‡
Department of Chemistry, University of Science and Technology of China, Hefei, 230026 China Department of Civil Engineering, Tohoku University, Sendai 980-8579, Japazn
bS Supporting Information ABSTRACT: Effective monitoring and diagnosis of anaerobic digestion processes is a great challenge for anaerobic digestion reactors, which limits their stable operation. In this work, an online monitoring and alert system for upflow anaerobic sludge blanket (UASB) reactors is developed on the basis of a set of novel evaluating indexes. The two indexes, i.e., stability index S and auxiliary index a, which incorporate both gas- and liquid-phase parameters for UASB, enable a quantitative and comprehensive evaluation of reactor status. A series of shock tests is conducted to evaluate the response of the monitoring and alert system to organic overloading, hydraulic, temperature, and toxicant shocks. The results show that this system enables an accurate and rapid monitoring and diagnosis of the reactor status, and offers reliable early warnings on the potential risks. As the core of this system, the evaluating indexes are demonstrated to be of high accuracy and sensitivity in process evaluation and good adaptability to the artificial intelligence and automated control apparatus. This online monitoring and alert system presents a valuable effort to promote the automated monitoring and control of anaerobic digestion process, and holds a high promise for application.
’ INTRODUCTION Anaerobic digestion (AD) processes are sensitive to environmental fluctuations and frequently suffer from operation instability. Thus, real-time monitoring and robust control of AD processes are essential to avoid possible instability due to disturbance. For this purpose, intensive studies have focused on fast, reliable, and online monitoring of AD processes as well as the development of efficient feedback alert and control strategies.1 A number of indexes have been frequently adopted in the process monitoring, including volatile fatty acids (VFAs), pH, redox potential, biogas production rate, and composition. Compared with the easy-to-measure parameters like pH and redox potential, VFA and biogas production are widely considered as the two most crucial and direct indicators of the system status.2,3 In a normally operating AD system, VFAs produced by acidogens and acetogens can be immediately utilized by the methanogens for methane production, leaving a low VFA concentration.4 Thus, the VFA accumulation is usually interpreted as the result of methanogenesis inhibition or organic overloading, and implies a risk of process upset.5 However, VFA alone is insufficient to reveal the overall reactor status. In this context, the biogas composition and production rate can also provide useful information.6,7 Therefore, a combined detection of both liquidand gas-phase parameters is considered as an efficient strategy to provide comprehensive insight into an AD process.810 r 2011 American Chemical Society
A variety of techniques for online monitoring of VFAs are available, such as fluorescence spectroscopy,11,12 near-infrared (NIR) spectroscopy,13 titration,14 and gas chromatography.15,16 Among them, titration is the most practical option, attributed to its high simplicity, rapidness, and cost-effectiveness.14,1719 On the other hand, NIR spectroscopy technique is gaining a great popularity for online monitoring of gas-phase parameters.20 Thus, incorporating titration and NIR spectroscopy techniques might offer a relatively simple way to achieve a comprehensive and accurate monitoring of AD process. In addition to parameter detection, effective analysis of the obtained data is of equal importance for reactor status evaluation and diagnosis. Because of the high complexity of AD systems and the unclear interrelations of many involved parameters, it is difficult to extract useful “hidden” information from the large data sets of individual parameters and to find out the true situation as well as potential risks. For AD process evaluation or diagnosis, a common practice is to set threshold values for some individual indicators like pH and VFA, and to judge the reactor status on the basis of the detected values.8,2123 However, such an evaluation is usually inaccurate, because information Received: June 30, 2011 Accepted: September 14, 2011 Revised: September 11, 2011 Published: September 14, 2011 9093
dx.doi.org/10.1021/es202245f | Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology
ARTICLE
Table 1. Operation Conditions for Shock Tests test
duration
6-fold organic overload shock
24 h
operation condition influent COD of 60 000 mg/L influent NaHCO3 of 3000 mg/L HRT of 24 h
5 °C temperature shock
96 h
influent COD of 10 000 mg/L influent NaHCO3 of 3000 mg/L HRT of 24 h temperature of 5 °C
6-fold hydraulic shock
24 h
influent COD of 1667 mg/L influent NaHCO3 of 500 mg/L HRT of 4 h
toxicant shock
slug dose influent COD of 10 000 mg/L influent NaHCO3 of 3000 mg/L HRT of 24 h chloroform dosage of 80 mg/L
Figure 1. Schematic of online monitoring and alert system for a UASB reactor (dotted lines: data; solid lines: liquid or gas pipes).
about the relationship between the individual parameters and the reactor status is still lacking and thus such an evaluation relies heavily on the professional knowledge and experiences of operators. Moreover, these parameters can only reveal the current reactor status, but it is actually, in most cases, too late for an effective process control once the threshold values are reached. Indeed, no effective method for AD reactor diagnosis and risk prediction is currently available. As such, it is imperative to propose and apply objective indexes for AD reactor diagnosis. Such indexes should be ideally accurate and sensitive to environmental fluctuations, reveal the change dynamics of reactor status, and be adaptive to online monitoring, autoalert, and control systems. In this work, an effective online monitoring and alert system is developed for an upflow anaerobic sludge blanket (UASB) reactor. Particularly, two new evaluating indexes, stability index S and auxiliary index a, are proposed for a quantitative assessment of reactor status, which for the first time take into account the changing dynamics of both liquid- and gas-phase parameters of UASB reactors. On the basis of a comprehensive investigation into the relationship between the indexes and reactor performance, a database of S and a under various disturbances is established to offer a foundation for online diagnosis and early warning of AD process.
’ MATERIALS AND METHODS UASB Reactor and Online Monitoring and Alert System. In this work a bench-scale Plexiglas-made UASB reactor was used, which consisted of a reaction zone of 2.0 L and a gassolids separator zone. The online monitoring system consisted of a gasphase detection unit and a liquid-phase detection unit that connected to a computer through a data acquisition card (PCI1602, ICP DAS Co., China), as illustrated in Figure 1. The liquidphase detection unit was comprised of an automatic sample loading and ejecting device and an online titration device, controlled by computer through 485 serial ports. A precise meter pump (BT100-1F, Longer Co., China) was used as the sample loading and ejecting device. The online titration device consisted of an inject pump (TJ-3A, Longer Co., China), a pH glass
electrode (LE409, Mettler Toledo Co., USA), and a titration cell. A magnetic stirrer was used for the liquid mixing in the titration cell. The gas-phase detection unit consisted of methane and carbon dioxide sensors (AGM10 and AGM32, Onward Co., China) and a gas flow measurement device for monitoring of biogas production rates. The software of alert system was designed on a LabView virtual instrument platform (National Instruments Co.,USA), which incorporates multifunctions of data acquisition, analysis, recording, and display. Specifically, the indexes S and a profiles were calculated, recorded, and displayed real-time based on the LabView software. Experimental Operations. The seed sludge for the UASB reactor was taken from an anaerobic digester in a local citrateprocessing wastewater treatment plant. The pH and volatile suspended solids (VSS) of seed sludge were 7.2 and 42.2 g/L, respectively. The reactor temperature was maintained at 35 ( 1 °C using a heating jacket, except during the temperature shock testing period. The reactor was operated at a fixed loading rate of 10 g/L chemical oxygen demand (COD) and 24-h hydraulic retention time (HRT) except as otherwise specified. Sucrose-rich synthetic wastewater was used as the feedstock, which contained 3000 mg/L NaHCO3 as the buffer and sufficient amount of inorganic nutrients as follows (in mg/L): NH 4 HCO3 , 405; K2HPO4 3 3H2O, 155; CaCl2, 50; MgCl2 3 6H2O, 100; FeCl2, 25; NaCl, 10; CoCl2 3 6H2O, 5; MnCl2 3 4H2O, 5; AlCl3, 2.5; (NH4)6Mo7O24, 15; H3BO3, 5; NiCl2 3 6H2O, 5; CuCl2 3 5H2O, 5; and ZnCl2, 5. After the reactor reached a steady state in terms of COD removal and methane production, four consecutive shock tests were conducted to investigate the response of the UASB reactor to various operating shocks, including organic overloading, hydraulic loading, temperature, and toxicant shocks (Table 1). The organic overloading shock was imposed by switching the feed COD concentration to six fold while keeping the flow rate unchanged. The hydraulic loading shock was provided by elevating the flow rate while accordingly reducing the feed COD and bicarbonate concentrations to maintain a constant organic loading rate (OLR). The temperature shock was achieved by directly shifting the reactor temperature from 37 to 5 °C. For the toxicant shock, toxicants that need a relatively high concentration to induce significant toxic effect, e.g., formaldehyde, 9094
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology
ARTICLE
Figure 2. Index variations under 6-fold organic overloading shock: (A) S and a; and (B) total VFA concentration and MPR.
and surfactant, which shows chronic toxicity,24 were not chosen here in order to avoid OLR interference.25 In this study, chloroform was selected as the toxicant, because of its significant toxic effect on anaerobic microorganisms even at a very low level.26 The toxicant shock was imposed through slug dose of chloroform at 80 mg/L. An interval of over two months was adopted between every two shock tests to ensure a stable operation before another shocking test. The variations of the gas- and liquid-phase parameters during the entire test period were measured online and the indexes S and a were calculated. Monitoring and Analysis. The VFA concentration and pH were measured online using the liquid-phase detection unit with five-point titration technique.19 The volume and composition of methane and carbon dioxide were detected online by the gasphase detection unit. All detections were conducted once per hour except during the temperature shock test, in which a 4-h measurement frequency was adopted. The pH electrode and titration cell were cleaned or washed regularly to prevent possible fouling. The monitoring data were collected by the data acquisition card and real-time recorded by computer. VSS and COD concentrations were measured according to the Standard Methods.27 Evaluating Indexes. Two evaluating indexes are proposed here for reactor status evaluation. The stability index S is used to describe the VFA accumulation in the liquid phase, while the auxiliary index a is used to evaluate the variation trend of
methane production in the gas phase, as follows: S¼
1 dVFA 100 QCH4 dt
ð1Þ
a¼
dQCH4 dt
ð2Þ
where (dVFA)/(dt) (mmol/L/h) is the change rate of total VFA concentration at time t and QCH4 (mmolCH4/L/h) is methane production rate (MPR). The signs of indexes also serve as useful signals for reactor state evaluation. To benefit an automatic process diagnosis, a diagnosis software based on the statistic analysis and logic judgment programs is designed for data analysis. In this study, if the majority (i.e., 70%) of the index data show the same sign in a time span of at least 5 h, this sign would be automatically assigned to the index. Specifically, if a change of sign occurs, the system will not give an immediate judgment. Instead, it would evaluate whether such a sign change is caused by a real change of reactor status or just a temporary fluctuation according to the statistics of this point and the subsequent points in at least 5 h. On the basis of the signs and values of both indexes, the current state and possible risks of reactor can be determined. In this study, the dangerous state is defined as sustaining VFAs accumulation and corresponding inhibition in methane production. Under this state, the majority of S would be positive and a would be negative according to the eqs 1 and 2, and a higher value 9095
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology
ARTICLE
Table 2. Operation Conditions and Shock Types of Tests in Literature shock type
duration
shock condition
references 25
4-fold organic overload shock
hours 03
influent COD increased from 5000 to 20 000 mg/L
4-fold hydraulic shock
hours 03
HRT decreased from 12 to 3 h
25
toxicant shock
slug addition
10 mg chloroform/L was dosed
25
temperature shock
days 070
temperature increased from 37 to 55 °C
30
temperature shock
days 3038
temperature increased from 55 to 65 °C
31
of S implies a higher degree of risk. To better reflect the degree of danger and stability, several threshold values are set here. Under the premise of the positive S coupled with negative a, the alert system will give alarm if the absolute value of S exceeds 50.0. Notably, this threshold for alarm is only a type of emergency alarm that indicates a serious situation. Actually, even at the beginning of shock, the diagnosis system has already detected the occurrence of shock according to signs of S and a. In a special case, if both indexes are close to zero, i.e., |S| is less than 4.0, and |a| is less than 1.5, this refers to a typical steady state. Or, if both indexes have frequently changed signs (positive and negative signs occur alternately and both occurrences are below 70%), this can be regarded as an unsteady state. In this way, the system status can be diagnosed and a warning signal will be given by the alert system in advance if a dangerous state or trend is identified.
’ RESULTS Reactor Performance under Steady-State Conditions. The gas production rate and compositions as well as the effluent pH, COD, and VFAs during the reactor steady-state operation period were monitored to evaluate the reactor performance. The average gas production rate was in a range of 407.7431.5 mL/h. The methane and carbon dioxide contents were 65.568.0% and 20.127.0%, respectively. The effluent pH was 7.147.21. The effluent COD and VFA concentrations were 41.596.0 and 39.095.5 mg/L, respectively. Organic Overloading Shock. The typical response of the AD process to an organic overloading shock is as follows:28 increase in VFA concentration, sludge volume index, and MPR; decrease in COD removal efficiency and methane content; change of pH depending on the buffer capacity of the feed. According to the variations of operating conditions and S and a values, the organic overloading shock test could be divided into 4 phases (Figure 2). When the organic overloading shock was applied in phase 1 (P1, hours 014), initially no obvious inhibition was observed at this stage. The S value was positive because of a rising VFA concentration. Meanwhile, a also showed positive values as the MPR kept increasing. In phase 2 (P2, hours 1524), the shock condition continued and an obvious inhibition occurred. S remained positive values, but a turned negative as the MPR started to decline. The shock was stopped since phase 3 (P3, hours 2537), and the reactor rapidly restored, with S and a values changed to negative and positive, respectively. In phase 4 (P4, after hour 38), the reactor became restabilized, and both S and a values were close to zero. A lasting warning signal was given by the alert system throughout P2 according to the identification result of diagnosis software, because most S showed positive values while a was mostly negative. The absolute value of S exceeded 1000.0 at hour 24, indicating a high risk of reactor collapse due to severe VFA accumulation. From hour 25, the organic overloading shock was terminated, and the S and a in P3 changed to negative and
positive values, respectively. Such changes suggest that the shock strength was weakened or stopped, resulting in a reduced VFA concentration and recovery of the methane production capacity. Accordingly, the alert stopped with the recovery of reactor status. In P4, both the S and a values approached zero. The MPR and VFA concentration became stabilized, and the reactor was tending toward stable operation. Hydraulic Loading Shock. The variations of S and a during the 6-fold hydraulic loading shock are shown in Figure S1 (Supporting Information). Most of the S showed negative values, and the absolute values were all below 3.0 in P1 (hours 124). Alternative positive and negative values of a were observed almost in half-and-half, with all the absolute values below 1.4. These results suggest that the MPR remained stable in P1. After hour 24 (P2), the hydraulic shock stopped. The value of S mostly became positive and its absolute values were below 2.4 during hours 2536. The total VFA concentration increased slightly. The a values also showed positive and negative fluctuation, and the absolute values fell to below 0.8. The MPR remained stable in this stage. After hour 37, both the absolute values of S and a were close to zero, and the reactor restored stable performance. No significant change in both indexes was observed during the hydraulic loading shock, indicating that the reactor performance was less affected by such a hydraulic loading shock. This was consistent with previous studies, in which the COD removal varied only slightly under the condition of decreased HRT while unchanged OLR.29 Temperature Shock. The changes of S and a during the temperature shock are shown in Figure S2. During the temperature shock (P1, hours 196), most of S exhibited positive values, indicating an increase in VFA concentration. The absolute values of S mounted from 3.0 up to a maximum of 60.0 at hour 28, suggesting an apparently increasing trend of VFA accumulation. Meanwhile, negative values were found for a in P1, indicating a considerable suppression of methane production. Accordingly, warning signals were given by the alert system throughout P1. Noticeably, the signs of a here were apparently different from those in the hydraulic and organic overloading shocks, which mostly showed positive values. With temperature shock, the methane production was almost completely inhibited during hours 6096 and the absolute values of a were below 0.1. With the ending of temperature shock after hour 96 (P2), the S value changed drastically from positive to negative. A rapid drop in the total VFA concentration was observed. At the same time, the a value changed from negative to positive and a gradual restoration of methane production was observed. As a consequence, the alert signal automatically terminated. Although both S and a values fluctuated between positive and negative values within P3 (hours 188264), they became stable and approached zero in P4 (after hour 265), indicating that the UASB reactor restored stable performance. Toxicant Shock. The response of S and a to the toxicant shock is illustrated in Figure S3. Both S and a exhibited violent 9096
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology
ARTICLE
exhibited negative values during this period. This indicates a rapid VFA accumulation and a severe inhibition in methane production. A warning signal was given by the system in P1. After hour 5 (P2), a remained negative and the biogas production dropped to near zero, while S value changed from positive to negative in P2. Because the VFA could not be consumed by the methanogens, as evidenced by the completely ceased gas production, this means that the VFA production process was also inhibited. The slight decreased VFA concentration in the reactor might be attributed to the effluent discharge. The sharp increase in the absolute value of S and the ceased methane production at hour 12 indicate a severe system collapse. After hour 12, a was near zero and no positive value of a was observed, analogue to that under other shock-ceasing conditions. This further confirms the full stop of methane production and collapse of the UASB system. Verification. To further evaluate the practicability of the two indexes S and a, the experimental data of shock tests reported in previous studies were used for verification. The operating conditions and shock types in these tests are summarized in Table 2, whereas the calculated results of S and a are shown in Figure 3. As shown in Figure 3C, the S and a calculated from the literature data show changing profiles highly similar to those in this study, except for the toxicant shock test. For the toxicant shock test by Tay and Zhang,25 the calculated values of a increased for a short period before dropping to zero again, indicating the recovery of methane production and stable performance after the ending of shock. In contrast, in our study an irreversible collapse of reactor occurred due to the much higher dosage of chloroform, which had caused unrecoverable toxic damage. Generally, these two evaluating indexes demonstrate good applicability to various AD systems.
’ DISCUSSION
Figure 3. Profiles of S and a based on data in literature: (A) 4-fold organic overloading shock;25 (B) 4-fold hydraulic loading shock;25 (C) toxicant shock;25 (D) 55 °C temperature shock;30 and (E) 65 °C temperature shock.31
responses to the toxicant shock in the initial 4 h (P1). The S index mostly showed positive values above 50.0, while a mostly
Merits of the Monitoring and Alert System. The constructed monitoring and alert system shows several distinct advantages over the conventional systems. First, comprehensive but reasonably simple online monitoring of AD process is realized, which offers the basis for a rapid and accurate evaluation of AD reactor status. Second, the alert system demonstrates high capacity in process diagnosis and early warning, attributed to the introduction of the two new indexes, S and a. Third, the system, which is characterized by automated execution, modularized design, and easy use, holds great promise for future practical application. All monitoring processes are automatically executed with software program. Simple and commercially available components are adopted for the hardware of this system to facilitate modularization. Most of all, no complex operation and maintenance are required for this system, and customized services for reactor monitoring and online data processing can be achieved by using the virtual instrument software program. The two indexes, S and a, are proposed for diagnosis of AD reactor status. They can reflect reactor status and changing trend accurately, and have a rapid response to the operating condition disturbance. On the basis of the diagnosis results, an early alert can be given by the system to prevent the upcoming risks. Accuracy of Indexes. The S and a indexes, which incorporate the key operating parameters in both liquid- and gas-phase, offer meaningful information on anaerobic reactor status. Specifically, a combined evaluation of values and signs of both indexes reveals the hidden pattern of dynamic change in VFA accumulation and 9097
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology
ARTICLE
Table 3. Correlation between the Sign of Indexes and Reactor Status S
a
positive
positive
organic overload; methanogens are
positive
negative
methanogens are
P2 of Figure 2A, P1 of Figure S2(A),
evidently suppressed
P1 of Figure S3(A), P2 of Figure 3A, P1 of Figure 3C,
reactor status
example P1 of Figure 2A, P1 of Figure 3A
not evidently suppressed
P1 of Figure 3D, P1 of Figure 3E negative
positive
activity of methanogens has restored.
P3 of Figure 2A, P2 of Figure S2(A), P2 of Figure 3C, P2 of Figure 3E
negative
negative
acidogenesis and methanogenesis are both inhibited
P2 of Figure S3(A), P3 of Figure 3A
frequent
frequent
reactor state is unstable; at the
P3 of Figure S2(A), P2 of Figure 3D
change of sign
change of sign
transitional stage
close
close
reactor is operated at
to zero
to zero
steady state or all the biochemical
P4 of Figure 3A, P1P4 of Figure 3B, P3 of Figure 3C,
processes in the reactor cease
P3 of Figure 3D, P3 of Figure 3E
methane production. Based on the experimental results of the shock tests and literature data verification, a database about the correlation between the two indexes and anaerobic reactor status was established (Table 3). Such a combination of different signs and magnitudes of the two indexes corresponds well to different reactor status and shock types. Here, the absolute values of S and a reflect the intensity of changes or disturbance in anaerobic reactors, whereas their signs indicate a specific reactor status and shock type. As shown in Table 3, positive S is found under many different operating conditions and reactor status, implying that the value of S alone is insufficient to clearly identify reactor status. Noticeably, the signs of a also vary significantly and offer valuable supplementary information. On the basis of a combined analysis of S and a at the initial stage of shock, it is possible to preliminarily distinguish the shock type and find out whether anaerobes are significantly inhibited. Generally, a positive S and positive a correspond to an organic overloading shock, while a positive S and negative a indicate a temperature or toxicant shock. At the late stage of an organic overloading shock (Figures 2A and 3A), a positive S and negative a are obtained regardless of the shock intensity, which suggests a methanogenesis-inhibited state. In contrast, a negative S and positive a at the late stage of shock means consumption of VFA by methanogens and a gradual adaptation and reactor recovery after shock. In addition, the changing trends of S and a are also informative. If S value changes from positive to negative, while a value remains negative for a long time, the inhibition of both VFA and methane production is suggested. Apparently, there exist some relationships between the indexes of S and a and the reactor status. Compared with the conventional individual parameters, such as pH, VFA, and MPR, which rely on their preset threshold values for judgment and usually lead to subjective and incorrect conclusions, these two integrated evaluating indexes proposed in this study offer more objective, accurate, and comprehensive assessment on the status and shock types of anaerobic reactors. Sensitivity of Indexes. Both S and a demonstrate high sensitivity, mainly attributed to an integrated analysis of both the absolute value and the sign as well as the use of MPR as a regulatory factor. With this regulatory factor, the changes of S can be properly and significantly magnified or buffered, so that the reactor changing
P4 of Figure 2A, P1, P2 of Figure S1(A), P4 of Figure S2(A),
trend can be more clearly revealed. This ensures a high sensitivity of the indexes and a capability of the system for early warning. For example, in the organic overloading shock test (Figure 2), the positive S and a in P1 corresponded to the VFA accumulation and MPR increase. Such an MPR increase was attributed to the uninhibited methanogenesis at the beginning. As a consequence, the increase in S was weakened by the methane production, which serves as a buffer against the factor dVFA/dt. Thus, S showed no drastic increase. The moderate change of S was consistent with the reactor status. In P2, however, the MPR severed as amplifier of S, because methanogenic inhibition occurred. At this phase, the S remained positive, while a shifted to negative. The increase in dVFA/dt was magnified by the MPR, leading to a maximum S of over 1000.0 at the end of shock. This was consistent with the significantly deteriorated reactor performance. Accordingly, alert was given by the system throughout P2. It is worthwhile to note that, however, the shock had already been detected by the system even at the beginning of shock. For both S and a, their signs are dependent on the transient changing trend of the monitoring parameters. The sign will change and respond immediately once the reactor becomes instable or undergoes changing state, even when the conventional evaluating parameters are still below their threshold values. For example, in the organic overloading shock test, the VFA concentration, methane production, and pH all showed little changes at the initial shock stage (within 1 h): the VFA concentration increased slightly from 2.0 to 2.2 mmol/L, the maximum MPR increased from 6.4 to 7.5 mmol/L/h, and pH and methane component kept at around 7.3 and 66.0%, respectively. However, both S and a values changed substantially. S and a jumped rapidly from near zero to 5.6 and 3.2, respectively, with a positive sign (the sign can change instantaneously as a result of shock). These changes suggest a changing reactor status and offer hints to the type and level of the encountered shocks. The positive sign of S and a clearly suggested an organic overloading shock. As a consequence of diagnosis results, alert was given by the system. The time of earliest alert given was at hour 15, which was 5 h earlier than the occurrence of severe acidification (pH < 6.2) and 8 h before the full stop of methane production. In contrast, at hour 15 the VFA concentration was 20.1 mmol/L; MPR was 36.9 mmol/L/h, which was even much higher than that at hour 0 (6.4 mmol/L/h); pH value was 6.93 and acidification 9098
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology did not occur. Apparently, these conventional parameters revealed no sign of risk at that moment. Therefore, indexes S and a are not only more effective and accurate in assessing the reactor status and operation condition changes, but also more effective in risk prediction, because risks can be detected far in advance. Similar high prediction accuracy of risks was repeatedly demonstrated in the other shock tests. Therefore, compared to the conventional parameters that can reveal the reactor status only after occurrence of deterioration, the two indexes proposed here can give synchronous and real-time information on the reactor status and rapidly reflect any slight change early before the dangers occur. Adaptability of Indexes to Diagnosis Program. With S and a, a simple and reliable diagnosis program can be developed for a fast and automatic evaluation of anaerobic reactors, without the need for human intuition or intervention. Moreover, these two indexes are suitable for a logic working pattern of artificial intelligence programming, because the positive and negative signs resemble 0 and 1 of the computer binary system. Therefore, an integration of the two indexes with the online monitoring and alert system can provide a practical tool for online artificial intelligence diagnosis of anaerobic reactor status. By further employing an additional feedback automatic control system, intelligent operating and automatic control of anaerobic reactors could be realized. In summary, an efficient online monitoring and alert system is developed for monitoring and diagnosing the status of a UASB reactor. The core of this system is a set of integrated indexes, a stability index S and an auxiliary index a, which offer the basis for a quantitative and comprehensive evaluation of anaerobic digestion reactors. The great accuracy and sensitivity of the two indexes enable a reliable early warning of potential risks. Their good adaptability and relatively simplicity also ensure a high possibility for being incorporated into artificial intelligence and automated control systems. Therefore, this online monitoring and alert system holds great promise for practical application in online monitoring, diagnosis, and control of anaerobic digestion reactors for waste treatment. Moreover, with a further development of advanced monitoring and data interpretation techniques as well as a deeper understanding of microbial metabolic dynamics, it may also be expanded to other fields of anaerobic fermentation processes. Nevertheless, despite the good performance of our system, it may face challenges for practical application in large-scale reactors, such as liquid residual in pipeline, pipe jam, electrode fouling, and fluid mixing. In addition, the setting of threshold values for S and a indexes is important and may vary significantly with reactor type and diagnosis objective, because there is inevitably a balance between prediction accuracy and degree of foreseeability. Thus, selection of appropriate threshold setting values is necessary. These all warrant further investigations.
’ ASSOCIATED CONTENT
bS
Supporting Information. Figures S1S3. This information is available free of charge via the Internet at http://pubs.acs. org.
’ AUTHOR INFORMATION Corresponding Author
*W.-W.L. Fax: +86 551-3607453; e-mail:
[email protected]. H.-Q.Y. Fax: +86 551-3607592; e-mail:
[email protected].
ARTICLE
’ ACKNOWLEDGMENT We thank the NSFC-JST Joint Project (21021140001) and Ministry of Science and Technology of China (2008BADC4B18) for the partial support of this study. ’ NOMENCLATURE a auxiliary index COD chemical oxygen demand HRT hydraulic retention time MPR methane production rate OLR organic loading rate S stability index VFA volatile fatty acid VSS volatile suspended solids UASB upflow anaerobic sludge blanket ’ REFERENCES (1) Madsen, M.; Nielsen, J. B. H.; Esbensen, K. H. Monitoring of anaerobic digestion processes: A review perspective. Renewable Sustainable Energy. Rev. 2011, 15, 3141–3155. (2) Bjorklund, R. B.; Christiansson, A.; Anders, E. W.; Ejlertsson, J. Electrode specific information from voltammetric monitoring of biogas production. Talanta 2010, 81, 1578–1584. (3) Ward, A. J.; Hobbs, P. J.; Holliman; Jones, D. L. Optimisation of the anaerobic digestion of agricultural resources. Bioresour. Technol. 2008, 99, 7928–7940. (4) Lahav, O.; Morgan, B. E. Titration methodologies for monitoring of anaerobic digestion in developing countries - A review. J Chem. Technol. Biotechnol. 2004, 79, 1331–1341. (5) Buyukkamaci, N.; Filibeli, A. Volatile fatty acid formation in an anaerobic hybrid reactor. Proc. Biochem. 2004, 39, 1491–1494. (6) Mathiot, S.; Esoffier, Y.; Ehlinger, F.; Couderc, J. P.; Leyris, J. P.; Moletta, R. Control parameter variations in an anaerobic fluidized-bed reactor subjected to organic shockloads. Water Sci. Technol. 1992, 25, 93–101. (7) Steyer, J.-P.; Buffiere, P.; Rolland, D.; Moletta, R. Advanced control of anaerobic digestion processes through disturbances monitoring. Water Res. 1999, 33, 2059–2068. (8) Boe, K.; Batstone, D. J.; Steyer, J. P.; Angelidaki, I. State indicators for monitoring the anaerobic digestion process. Water Res. 2010, 44, 5973–5980. (9) García-Dieguez, C.; Molina, F.; Roca, E. Multi-objective cascade controller for an anaerobic digester. Proc. Biochem. 2011, 46, 900–909. (10) Nielsen, H. B.; Uellendahl, H.; Ahring, B. K. Regulation and optimization of the biogas process: Propionate as a key parameter. Biomass Bioenergy 2007, 31, 820–830. (11) Li, W. H.; Sheng, G. P.; Lu, R.; Yu, H. Q.; Li, Y. Y.; Harada, H. Fluorescence spectral characteristics of the supernatants from an anaerobic hydrogen-producing bioreactor. Appl. Microbiol. Biotechnol. 2010, 89, 217–224. (12) Palacio-Barco, E.; Robert-Peillard, F.; Boudenne, J. L.; Coulomb, B. On-line analysis of volatile fatty acids in anaerobic treatment processes. Anal. Chim. Acta 2010, 668, 74–79. (13) Zhang, M. L.; Sheng, G. P.; Mu, Y.; Li, W. H.; Yu, H. Q.; Harada, H.; Li, Y. Y. Rapid and accurate determination of VFAs and ethanol in the effluent of an anaerobic H2-producing bioreactor using near-infrared spectroscopy. Water Res. 2009, 43, 1823–1830. (14) Feitkenhauer, H.; Sachs, J.v.; Meyer, U. Online titration of volatile fatty acids for the process control of anaerobic digestion plants. Water Res. 2002, 36, 212–218. (15) Lenz, M.; Hullebusch, E. D. V.; Farges, F.; Nikitenko, S.; Corvini, P. F. X.; Lens, P. N. L. Combined speciation analysis by X-ray absorption near-edge structure spectroscopy, ion chromatography, and solid-phase microextraction gas chromatography-mass spectrometry to 9099
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
Environmental Science & Technology
ARTICLE
evaluate biotreatment of concentrated selenium wastewaters. Environ. Sci. Technol. 2011, 45, 1067–1073. (16) Ward, A. J.; Hobbs; Bruni, E.; Lykkegaard, M. K.; Feilberg, A.; Adamsen, A. P. S.; Jensen, A. P.; Poulsen, A. K. Real time monitoring of a biogas digester with gas chromatography, near-infrared spectroscopy, and membrane-inlet mass spectrometry. Bioresour. Technol. 2011, 102, 4098–4103. (17) Moosbrugger, R. E.; Wentzel, M. C.; Ekama, G. A.; Marais, G. v. R. A 5 pH point titration method for determining the carbonate and SCFA weak acid/bases in anaerobic systems. Water Sci. Technol. 1993, 28, 237–45. (18) Anderson, G. K.; Yang, G. Determination of bicarbonate and total volatile acid concentration in anaerobic digesters using a simple titration. Water Environ. Res. 1992, 64, 53–59. (19) Lahav, O.; Morgan, B.; Loewenthal, R. Rapid, simple, and accurate method for measurement of VFA and carbonate alkalinity in anaerobic reactors. Environ. Sci. Technol. 2002, 36, 2736–2741. (20) Pejcic, B.; Eadington, P.; Ross, A. Environmental monitoring of hydrocarbons: A chemical sensor perspective. Environ. Sci. Technol. 2007, 41, 6333–6342. (21) Michaud, S.; Berner, N.; Buffiere, P.; Roustan, M.; Moletta, R. Methane yield as a monitoring parameter for the start-up of anaerobic fixed film reactors. Water Res. 2002, 36, 1385–1391. (22) Nielsen, J. B. H.; Lomborg, C. J.; Popiel, P. O.; Esbensen, K. H. Online near infrared monitoring of glycerol-boosted anaerobic digestion processes: evaluation of process analytical technologies. Biotechnol. Bioeng. 2008, 99, 302–313. (23) Punal, A.; Roca, E.; Lema, J. M. An expert system for monitoring and diagnosis of anaerobic wastewater treatment plants. Water Res. 2002, 36, 2656–2666. (24) Ginkel, S. W. V.; Kortekaas, S. J. M.; Lier, J. B. V. The chronic toxicity of alcohol alkoxylate surfactants on anaerobic granular sludge in the pulp and paper industry. Environ. Sci. Technol. 2007, 41, 4711–4714. (25) Tay, J. H.; Zhang, X. Y. Stability of high-rate anaerobic systems. I: Performance under shocks. J. Environ. Eng. 2000, 126, 713–725. (26) Speece, R. E. Anaerobic Biotechnology for Industrial Wastewaters; Archae Press: Nashville, TN, 1996. (27) APHA-AWWA-WPCF. Standard Methods for the Examination of Water and Wastewater, 20th ed.; Washington, DC, 1999. (28) Nachaiyasit, S.; Stuckey, D. C. The effect of shock loads on the performance of an anaerobic baffled reactor (ARB). 1. Step changes in feed concentration at constant retention time. Water Res. 1997, 31, 2737–2746. (29) Nachaiyasit, S.; Stuckey, D. C. The effect of shock loads on the performance of an anaerobic baffled reactor (ARB). 2. Step and transient hydraulic shocks at constant feed strength. Water Res. 1997, 31, 2747–2754. (30) Bouskova, A.; Dohanyos, M.; Schmidt, J. E.; Angelidaki, I. Strategies for changing temperature from mesophilic to thermophilic conditions in anaerobic CSTR reactors treating sewage sludge. Water Res. 2005, 39, 1481–1488. (31) Lau, I. W. C.; Herbert, H. H. P. Effect of temperature shock to thermophilic granules. Water Res. 1997, 31, 2626–2632.
9100
dx.doi.org/10.1021/es202245f |Environ. Sci. Technol. 2011, 45, 9093–9100
ARTICLE pubs.acs.org/est
Amine-Based Nanofibrillated Cellulose As Adsorbent for CO2 Capture from Air Christoph Gebald,†,‡,§ Jan Andre Wurzbacher,†,§ Philippe Tingaut,‡ Tanja Zimmermann,‡ and Aldo Steinfeld†,||,* †
Department of Mechanical and Process Engineering, ETH Z€urich, 8092 Zurich, Switzerland EMPA Material Science and Technology, 8600 D€ubendorf, Switzerland § Climeworks LLC, 8044 Z€urich, Switzerland Solar Technology Laboratory, Paul Scherrer Institute, 5232 Villigen, Switzerland
)
‡
bS Supporting Information ABSTRACT: A novel amine-based adsorbent for CO2 capture from air was developed, which uses biogenic raw materials and an environmentally benign synthesis route without organic solvents. The adsorbent was synthesized through freeze-drying an aqueous suspension of nanofibrillated cellulose (NFC) and N-(2-aminoethyl)-3-aminopropylmethyldimethoxysilane (AEAPDMS). At a CO2 concentration of 506 ppm in air and a relative humidity of 40% at 25 °C, 1.39 mmol CO2/g was absorbed after 12 h. Stability was examined for over 20 consecutive 2-hadsorption/1-h-desorption cycles, yielding a cyclic capacity of 0.695 mmol CO2/g.
1. INTRODUCTION CO2 capture from atmospheric air—usually referred to as “air capture”—can be coupled with CO2 sequestration in geological formations to reduce the atmosphere’s CO2 content, or with the conversion into CO2-neutral liquid hydrocarbon fuels using renewable energy.13 A recent report of the American Physical Society (APS)4 questioned the economic viability of air capture, but assumes old designs and outdated values. Recent studies, including this, introduced amine-based adsorbents feasible for air capture featuring CO2 capacities comparable to those obtained under flue gas conditions.57 Chemical systems investigated for air capture include those based on calcium oxide and hydroxide,811 sodium hydroxide,1217 potassium hydroxide,18 amines immobilized on silica nanoparticles,19 amine modified mesoporous silicas,6,7,20 amine impregnated natural fibers21 and basic ion exchange resins.2225 Because of the need to process 2600 moles of air per mole of CO2 captured, the feasibility of the chemical system will be strongly dependent on its ability to tolerate moisture without the need to compress, heat, or cool the stream of ambient air. Aminebased adsorbents are especially suitable for this purpose, as amines react selectively with atmospheric CO2 in the presence of moisture at ambient temperature and pressure while pure CO2 can subsequently be released upon heating to about 100 °C.6 In previous studies, the synthesis of amine-based solid adsorbents was carried out either via covalent bonding or physisorption onto the support,5,26 often using mesoporous silicas as support.27 r 2011 American Chemical Society
Covalent bonding further reduces the risk of amine volatilization during the adsorption/desorption cycles.27 The reaction mechanism of CO2 with immobilized amines has been analyzed in detail for silica supports modified with aminosilanes. For dry CO2 adsorption, the formation of silylpropylcarbamate, ammonium carbamate and carbamic acid was observed.2831 Danon et al. did not confirm the presence of carbamic acid after dry CO2 adsortion.30 For humid CO2 adsorption, ammonium carbamate formation was observed as the dominant mechanism (Reaction Scheme 1) and carbamic acid was formed to a lesser extent as compared to dry CO2 adsorption.28 CO2 þ RNH2 þ RNH2 T RNHCOO þ RNH3 þ
ð1Þ
For the enthalpy change of reaction for dry CO2 adsorption, values of 50 to 118 kJ/mol CO2 were reported.29,3139 For humid CO2 adsorption the enthalpy change of reaction was reported to be 56 kJ/mol CO2 and additionally 47 to 53 kJ/mol for H2O adsorption.32,33 Amine-based adsorbents proved to reversibly adsorb 0.9 mmol/g ( 0.09 mmol CO2/g (TRI-PE-MCM-41, a triaminesilane modified pore expanded mesoporous silica),20 1.72 mmol CO2/g (HAS6, a hyperbranched aminosilica),6 2.36 mmol Received: July 1, 2011 Accepted: September 13, 2011 Revised: September 13, 2011 Published: September 14, 2011 9101
dx.doi.org/10.1021/es202223p | Environ. Sci. Technol. 2011, 45, 9101–9108
Environmental Science & Technology CO2/g (PEI/silica, a polyethylenimine modified porous silica), 2.26 mmol CO2/g (A-PEI/silica, a polyethylenimine modified porous silica stabilized with 3-aminopropyltriethoxysilane) and 2.19 mmol CO2/g (T-PEI/silica, a polyethylenimine modified porous silica stabilized with tetrapropyl orthotitanate)7 from air. Previous synthesis of amine-based adsorbent materials employed mainly inorganic materials as support.5 To promote the application of renewable materials, we suggest the use of nanofibrillated cellulose (NFC) as a solid support. NFC is an abundant natural material composed of cellulose fibril aggregates, 10100 nm in diameter and several micrometers length.40,41 Furthermore, NFC is rich in surface hydroxyl groups, which can be used to anchor chemical functionalities covalently to the cellulose backbone. Amine functionality has been introduced on NFC through grafting of aminosilanes, which has been the compound of choice to bond amines covalently to mesoporous silicas,42 cellulose fibers4349 and NFC.50 However, the hydrolysis and self-condensation of aminosilanes with three alkoxy groups is fast and leads to the build-up of three-dimensional siloxane networks.51 This characteristic is unfavorable for air capture, as functional amine sites and pores are likely to be blocked by siloxane networks. This problem may be overcome by the use of anhydrous organic solvents.27 We decided to use aminosilanes having two alkoxy functions instead of three alkoxy functions, as dialkoxy silanes build up only two-dimensional linear structures or ring structures upon self-condensation,52 hence, avoiding the undesired blocking of functional amine sites and enabling the use of water as solvent. Recently, it has been reported that the use of pure water is superior to water/alcohol mixtures for aminosilane grafting on cellulose.53 The overall objective of this paper is to propose a novel aminebased adsorbent suitable for air capture. The new adsorbent was characterized by BET, SEM, FTIR, elemental analysis and its CO2 capacity as well as CO2 uptake rate were evaluated through adsorption of humid air at 40% relative humidity and desorption at 90 °C in Ar.
2. EXPERIMENTAL SECTION Materials. N-(2-aminoethyl)-3-aminopropylmethyldimethoxysilane (AEAPDMS, 97% purity) was obtained from ABCR (Germany) and used as received. As starting material for the isolation of NFC commercially available refined fibrous beech wood pulp suspension was used (Arbocel P 10111, 13.5% w/w aqueous suspension, Rettenmeier & S€ ohne GmbH & Co. KG, Germany). Synthesis. A 0.5% w/w NFC hydrogel was obtained through a two-step mechanical isolation process from the refined fibrous beech wood pulp suspension.54 AEAPDMS was added to this NFC hydrogel until the total silane concentration was 4% w/w. After sittring for 24 h, the suspension was centrifuged (Hettich Rotanta/AP) at 3600 rpm for 20 min and the remaining water was removed by freeze-drying. For this last processing step, the retentate was poured in a copper cylinder that was then immersed in liquid N2 and the frozen sample was dried in a freezedryer (Leybold Lyovac GT2). Once dried, the material was removed from the copper cylinder and heated in an inert Ar atmosphere to 120 °C for 2 h. The resulting material is in the following referred to as AEAPDMS-NFC-FD. For comparison, the synthesis procedure above was repeated without silane addition and the produced sample is referred to as NFC-FD. Material Characterization. The morphology was examined by scanning electron microscopy (SEM, FEI Nova NanoSEM
ARTICLE
230). Specific surface areas were determined by N2 adsorption (Micrometritics TriStar) by BrunauerEmmettTeller (BET) method with samples degassed (Micrometritics FlowPrep 060) at 105 °C for 24 h in dry N2 flow. The density was calculated from weight and volume, which was determined by measuring the diameter and height of the dried cylindrical product. The composition was determined through elemental analysis and Fourier-transform-infrared-spectroscopy (FTIR, Digilab BioRad FTS 6000 spectrometer, ATR mode, 32 scans, 4 cm1 resolution). CO2 Adsorption/Desorption Measurements. CO2 adsorption/ desorption measurements were performed in a fully automated system, schematically shown in Supporting Information Figure S1. It consists of a 40 mm-i.d 30 mm-height cylindrical packed bed filled with 1.2 g of adsorbent material. AEAPDMS-NFC-FD was broken in irregularly shaped pieces of mean size approximately 10 mm. The temperature of the packed bed was measured by a K-type thermocouple. The gas flow was controlled by two electronic mass flow controllers (Bronkhorst EL-FLOW). One of the gas streams was bubbled through a water bath kept at 25 °C. The relative humidity before and after the packed bed arrangement was monitored by two electronic humidity sensors (Vaisala HMP110). After passing the off-gas through a filter and a condenser, the CO2 content was analyzed by an IR analyzer (Siemens Ultramat 23) equipped with two detectors for the CO2 concentration ranges of 01000 ppm and 05%, at 1 Hz sampling rate and 0.2% range detection limit. Pressurized air with a CO2 concentration of 506 ppm was used for CO2 adsorption. Argon was used for CO2 desorption. All CO2 adsorption experiments were carried out at an air flow rate of 1 L/min at 25 °C and 40% relative humidity. All CO2 desorption experiments were performed at an Ar flow rate of 0.8 L/ min containing a level of moisture which corresponds to a relative humidity of 40% at 25 °C. In this study Ar was used as a purge gas during desorption to enable precise measurement of the amount of desorbed CO2 in the off-gas. In an industrial application, concentrated CO2 is obtained by a temperature-vacuum-swing process.55 With the onset of CO2 desorption, the packed bed arrangement was heated externally by a water bath at 95 °C until the packed bed temperature reached 90 °C ( 1 °C. After CO2 desorption, the packed bed was cooled by a water bath at 20 °C until the packed bed temperature reached 25 °C. Prior to all CO2 adsorption measurements, preadsorbed CO2 was removed through CO2 desorption at the defined conditions for 1 h. To determine the equilibrium CO2 capacity of the sample, CO2 adsorption was performed until the CO2 outlet concentration was higher than 97% of the CO2 inlet concentration, which occurred after 12 h of CO2 adsorption in this study. To determine the cyclic CO2 capacity of the sample and to examine the stability of the AEAPDMS grafting, consecutive 2 h-adsorption/1 h-desorption cycles were carried out. For all adsorption and desorption experiments, the CO2 uptake and release, Δq, was calculated by integrating the signal of the IR gas analyzer, Z t ¼ tads=des n_ gas ðc0 c1 Þ dt Δq ¼ ms t¼0 where tads/des is the adsorption and desorption time, respectively, n_ gas the molar flow rate of the respective gas stream, c0 and c1 the CO2 concentrations at the inlet and outlet of the packed bed, respectively, and ms the mass of the adsorbent material contained in the packed bed. FTIR spectra after CO2 adsorption and desorption were recorded from a sample which was removed 9102
dx.doi.org/10.1021/es202223p |Environ. Sci. Technol. 2011, 45, 9101–9108
Environmental Science & Technology
ARTICLE
Table 1. BET Surface Area, Density, And Amine Loading of NFC-FD and AEAPDMS-NFC-FD BET surface
density
amine loading
sample
area [m2/g]
[kg/m3]
[mmol N/g]
NFC-FD AEAPDMS-NFC-FD
26.8 7.1
26 61
4.9
Figure 2. FTIR spectra of NFC-FD and AEAPDMS-NFC-FD.
Figure 1. (a) SEM of NFC hydrogel, (b) NFC-FD, (c) AEAPDMSNFC-FD.
from the packed bed arrangement after 12 h of humid CO2 adsorption and 1 h of CO2 desorption, respectively.
3. RESULTS AND DISCUSSION Material Characterization. SEMs of the NFC before and after freeze-drying are shown in Figure 1a and b, respectively. Before freeze-drying, the NFC hydrogel morphology was composed of entangled cellulose nanofibrils of submicrometer diameter. After freeze-drying, the NFC-FD morphology was composed of cellulose sheet structures with single irregularly distributed cellulose nanofibrils attached to the cellulose sheets. Aggregation of cellulose nanofibrils to form sheets upon freezing and drying was previously observed.5658 The cellulose sheet forming was explained through ice crystal growth during freezing, which squeezed cellulose nanofibrils around the ice crystal in the form of cellulose sheets.58 The presence of AEAPDMS in the NFC suspension further increased the sheet-forming tendency and eliminated single cellulose nanofibrils, as shown in Figure 1c. Ice crystal growth presumably caused cellulose nanofibril aggregation while the hydrolyzed silane molecules in solution are pushed into the spaces between the ice crystals, forming dense cellulose-silane aggregates. Comparing Figure 1a and Figure 1c indicates the potential of AEAPDMS-NFC-FD: by finding ways to keep the nanofibrillar structure intact after amine modification and drying, more amine groups can be accessible for surface reaction with CO2, which will enhance the CO2 uptake rate of the adsorbent.
The BET surface area of NFC-FD was 26.8 m2/g and the density was 26 kg/m3 (Table 1). These values are comparable to those of cellulose aerogels prepared from NFC.56,58,57 The BET surface area of AEAPDMS-NFC-FD was 7.1 m2/g and the density 61 kg/m3, indicating that the surface modification reduced the available surface area but increased the density of the sample. For comparison, the BET surface area was 367 m2/g for TRI-PE-MCM-4120 and 45 m2/g for HAS6.6 Nevertheless, previous studies indicated that, for CO2 adsorption from gas streams with CO2 partial pressures of less than 0.01 atm, the equilibrium CO2 capacity of the adsorbent is mainly influenced by the amine loading and less dependent on the surface area.59,6 The amine loading of AEAPDMS-NFC-FD was obtained through elemental analysis and determined to be 4.9 mmol N/g. For comparison, the amine loading was 7.95 mmol N/g for TRI-PE-MCM-41,42 9.9 mmol N/g for HAS66 and 10.5 mmol N/g for T-PEI/silica.7 The FTIR spectra of NFC-FD and AEAPDMS-NFC-FD are shown in Figure 2. The spectrum of NFC-FD exhibited typical bands for cellulose, such as OH stretching at 3345 cm1, CH stretching at 2900 cm1, CH2 symmetric bending at 1430 cm1, OH and CH bending as well as C—C and C—O stretching at 1380, 1317, and 1256 cm1, COC skeletal vibrations around 1055 cm1, C1H deformation with ring vibration contribution and OH bending at 898 cm1, and OH out of plane bending at 666 cm1.50,60,61 The spectrum of AEAPDMS-NFC-FD after thermal treatment showed successful grafting of AEAPDMS on NFC through NH2 bending at 1600 cm1.29,6265 The band at 1662 cm1, similar to the one for mesoporous silica modified with N-(2-aminoethyl)-3-aminopropyltrimethoxysilane,62 was assigned to aminosilane employed herein and is associated with asymmetric NH3+ deformation, formed through the interaction of primary amines with water or silanol groups.28,62,6466 The 9103
dx.doi.org/10.1021/es202223p |Environ. Sci. Technol. 2011, 45, 9101–9108
Environmental Science & Technology
Figure 3. CO2 outlet concentration during humid CO2 adsorption on AEAPDMS-NFC-FD (solid line) and fractional uptake of the CO2 capacity (dashed line).
bands at 2917 cm1, 2877 cm1, and 2810 cm1 were assigned to CH stretching and the band at 1446 cm1 was assigned to CH2 bending of the silane propyl chain.29,62,65 The grafting conditions employed herein were sufficient for covalent bonding of silane molecules, as corroborated by the appearance of signals associated with vibrations from silicon-based linkages at 1256 cm1 (νSiC), and between 1180 and 700 cm1 (Si—OH, Si—O—C, νSiC, νas SiC, νas SiOSi, νs SiOSi).29,44,50 CO2 Adsorption/Desorption Measurements. CO2 Adsorption. The CO2 outlet concentration as a function of the adsorption time of AEAPDMS-NFC-FD is shown in Figure 3. CO2 adsorption on AEAPDMS-NFC-FD was stopped after 12 h of experiment, when the CO2 uptake rate became small (0.67 μmol/g/min) and the CO2 outlet concentration reached 97% of the inlet concentration. By integration, the CO2 loading of AEAPDMS-NFC-FD was calculated to be 1.39 mmol CO2/g. The amine efficiency of AEAPDMS-NFC-FD was calculated to be 28%. The highest CO2 loading reported to date was 2.36 mmol CO2/g obtained by Choi et al.7 for CO2 adsorption from dry Ar with 400 ppm CO2 concentration on PEI/silica having an amine density of 10.5 mmol N/g, resulting in an amine efficiency of 22%. Note that the amine efficiency of Choi et al.7 is given for CO2 adsorption from a dry gas, which usually increases when a wet gas stream is used. Moreover, Choi et al.6 achieved 1.72 mmol CO2/g for CO2 adsorption from humid air with 400 ppm CO2 concentration on HAS6 having an amine loading of 9.9 mmol N/g, corresponding to an amine efficiency of 17% in the fully loaded state. Belmabkhout et al.20 reported an equilibrium CO2 capacity of 0.98 mmol CO2/g for CO2 adsorption from dry air with 400 ppm CO2 concentration on TRI-PEMCM-41, where the amine efficiency was 12%. Its capacity increased from 1.9 mmol CO2/g to 2.04 mmol CO2/g when moisture was introduced in a gas stream with 5% CO2 concentration. Lackner22 reported that a commercially available strong base ion-exchange resin having a charge density of 1.7 mmol/g can theoretically absorb 0.85 mmol CO2/g. In the patent literature, Olah et al.19 reported a CO2 loading of 0.61 mmol CO2/g sorbent on amine modified fumed silica nanoparticles for dry air containing 380 ppm of CO2, and Gebald et al.21 reported a CO2 loading of 1.87 mmol CO2/g on amine modified air oxidized carbon fibers for humid CO2 adsorption from air having a CO2 concentration of 500 ppm. To fully characterize amine-based adsorbents, both the CO2 capacity and CO2 adsorption rate as a function of time are reported, as shown in Figure 4. Choi et al.6,7 reported the kinetic
ARTICLE
Figure 4. CO2 adsorption rate of AEAPDMS-NFC-FD (solid line) and adsorbed amount CO2 (dashed line).
Figure 5. Comparison of adsorption capacity and adsorption half time of AEAPDMS-NFC-FD (star), HAS6 (sphere), PEI/silica (rhombus), A-PEI/silica (triangle), T-PEI/silica (square).6,7.
data in the form of adsorption half time (time to 50% of final CO2 uptake), but this indicator is only meaningful in combination with the CO2 capacity as a small adsorption half time does not necessarily correspond to a high CO2 adsorption rate. Nevertheless, the adsorption half time is also given here for comparison (Figure 5). Note that particle size, residence time, linear velocity of the incoming gas and length over diameter of the packed bed investigated here differ from those used by Choi et al.6,7 The adsorption half time of AEAPDMS-NFC-FD was 92 min, corresponding to an average CO2 adsorption rate of 7.6 μmol/g/min during the adsorption half time. The adsorption half time and corresponding average CO2 adsorption rate was 309 min and 3.8 μmol/g/min for PEI/silica, 196 min and 5.8 μmol/g/min for A-PEI/silica, 210 min and 5.2 μmol/g/min for T-PEI/silica and 167 min and 5.2 μmol/g/min for HAS6.6,7 The amine-based solid adsorbent TRI-PE-MCM-41 synthesized by Belmabkhout et al.20 needed 15 min to reach 90% of its equilibrium CO2 capacity for CO2 adsorption from a dry gas mixture with 1000 ppm CO2 concentration. Lackner22 reported that 2500 kg of strong base ion-exchange resin are needed to capture 0.25 mol CO2/s from air, corresponding to an uptake rate of 6 μmol/g/min, which is comparable to that of AEAPDMS-NFC-FD during the first 120 min (6.7 μmol/g/min). In spite of the relatively small BET surface area and a particle diameter of 10 mm, the CO2 9104
dx.doi.org/10.1021/es202223p |Environ. Sci. Technol. 2011, 45, 9101–9108
Environmental Science & Technology
Figure 6. Comparison of CO2 capacity as a function of time of AEAPDMS-NFC-FD (dashed line), HAS6 (spheres), PEI/silica (rhombuses), A-PEI/silica (triangles), and T-PEI/silica (squares)6,7.
Figure 7. FTIR spectra of AEAPDMS-NFC-FD after humid CO2 adsorption.
adsorption rate of AEAPDMS-NFC-FD is comparable to that of HAS6, PEI/silica, A-PEI/silica and T-PEI/silica in powder form,6,7 as shown in Figure 6, where the CO2 adsorption rate of the literature adsorbents was assumed as constant. From this result we conclude that the bulk CO2 uptake rate of 10 mm AEAPDMS-NFC-FD particles is feasible for air capture. The favorable CO2 uptake rate can be explained by the highly porous structure of AEAPDMS-NFC-FD, which is composed of cellulose sheets separated by pores in the micrometer range (see Figure 1). For comparison purposes, NFC-FD was also tested for CO2 adsorption. The CO2 outlet concentration reached the value of the CO2 inlet concentration immediately after gas introduction, confirming that no physisorption of CO2 occurs on the solid support without modification. The FTIR spectrum of AEAPDMS-NFC-FD after 12 h of humid CO2 adsorption is shown in Figure 7. FTIR spectra of CO2 adsorption on aminosilane modified NFC have not been reported yet. The FTIR spectrum obtained after CO2 adsorption on AEAPDMS-NFC-FD is tentatively compared to FTIR spectra obtained from CO2 adsorption on amine-modified silicas. The band at 1568 cm1 was ascribed to (N)COO asymmetric stretching,2830,62,6466 the band at 1466 cm1 to symmetric NH3+ deformation and the bands at 1410 cm1, 1369 cm1 and 1310 cm1 to symmetric stretching of (N)COO.28,30 From CO2 absorption in amine solutions it is known that CO2 reacts
ARTICLE
Figure 8. Multicycle adsorption (squares) and desorption (spheres) experiments with AEAPDMS-NFC-FD.
with primary and secondary amines to form carbamic acid, carbamate and bicarbonate species.28,67,68 By comparing the FTIR spectra obtained in this work to the FTIR spectra obtained from CO2 absorption in amine solution, the band at 1369 cm1 could be further assigned to bicarbonate.67 However, this assignment remains questionable. First the formation of carbamate species, through deprotonation of carbamic acid, is kinetically and thermodynamically favored over bicarbonate formation for primary and secondary amines.67,68 Second aminosilane modification of NFC rendered the NFC surface less hydrophilic when compared to unmodified NFC,44 questioning the feasibility of comparing the FTIR spectra obtained in this work to FTIR spectra obtained from CO2 scrubbing in amine solutions. CO2 Desorption. CO2 desorption from AEAPDMS-NFC-FD was fast and completed after 30 min, as shown in Supporting Information Figure S2. By integration, the amount of desorbed CO2 was calculated to be 1.41 mmol CO2/g, which confirmed the amount of adsorbed CO2 within 1.5% error. More than 85% of the CO2 is desorbed within 19 min at a packed bed temperature below 80 °C. The FTIR spectra of AEAPDMS-NFC-FD after CO2 desorption is shown in Supporting Information Figure S3. The peaks that evolved after CO2 adsorption disappeared and the FTIR spectrum of the pristine AEAPDMS-NFC-FD was re-established. Cyclic CO2 Capacity. Figure 8 shows 20 consecutive CO2 adsorption/desorption cycles. No decrease in cyclic capacity was observed, which confirmed the structural stability of AEAPDMSNFC-FD under the conditions used. During 2 h of CO2 adsorption at 25 °C and 40% relative humidity and 1 h of CO2 desorption in Ar at 90 °C and 40% relative humidity, the average cyclic CO2 capacity was 0.695 mmol CO2/g. During CO2 desorption, humidity was added to avoid urea formation and a related loss of amine functionality.69 In an industrial application, the cycle duration might deviate from the one used in this study. Short cycle times maximize the amount of CO2 captured per mass of adsorbent per time and therefore reduce the capital cost. Long cycle times favor high CO2 loading, which minimizes the energy requirement for desorption. The optimal duration results from a trade-off between these two aspects.
’ ASSOCIATED CONTENT
bS
Supporting Information. Figures S1S3. This material is available free of charge via the Internet at http://pubs.acs.org.
9105
dx.doi.org/10.1021/es202223p |Environ. Sci. Technol. 2011, 45, 9101–9108
Environmental Science & Technology
’ AUTHOR INFORMATION Corresponding Author
Phone: +41-44-6327929; fax: +41-44-6321065; e-mail: aldo.steinfeld@ ethz.ch.
’ ACKNOWLEDGMENT € F-STIFTUNG We thank the Swiss foundation GEBERT-RU for financial support. ’ NOMENCLATURE c0 CO2 concentration at the inlet of the packed bed CO2 concentrations at the outlet of the c1 packed bed mass of the adsorbent material contained ms in the packed molar flow rate of gas stream (compressed n_ gas air or Ar) Δq adsorption/desorption CO2 capacity adsorption time tads desorption time tdes Acronyms
ABCR AEAPDMS
Supplier of specialty chemicals N-(2-aminoethyl)-3-aminopropylmethyldimethoxysilane AEAPDMS-NFC-FD aminosilane modified nanofibrillated cellulose aerogel A-PEI/silica aminosilane stabilized polyethylenimine modified porous silica7 BET BrunauerEmmettTeller method FTIR Fourier-Transform-Infrared-Spectroscopy HAS6 hyperbranched aminosilica6 MEA monoethanolamine NFC nanofibrillated cellulose NFC-FD nanofibrillated cellulose aerogel PEI polyethylenimine PEI/silica polyethylenimine modified porous silica7 SEM scanning electron microscopy T-PEI/silica tetrapropyl orthotitanate stabilized polyethylenimine modified porous silica7 TRI-PE-MCM-41 aminosilane modified pore expanded mesoporous silica20
’ REFERENCES (1) Chueh, W. C.; Falter, C.; Abbott, M.; Scipio, D.; Furler, P.; Haile, S. M.; Steinfeld, A. High-flux solar-driven thermochemical dissociation of CO2 and H2O using nonstoichiometric ceria. Science 2010, 330 (6012), 17971801; DOI: 10.1126/science.1197834. (2) Graves, C.; Ebbesen, S. D.; Mogensen, M.; Lackner, K. S. Sustainable hydrocarbon fuels by recycling CO2 and H2O with renewable or nuclear energy. Renewable Sustainable Energy Rev. 2011, 15 (1), 123; DOI: 10.1016/J.Rser.2010.07.014. (3) Keith, D. W. Why capture CO2 from the atmosphere? Science 2009, 325 (5948), 16541655; DOI: 10.1126/Science.1175680. (4) Direct Air Capture of CO2 with Chemicals; The American Physical Society: College Park, MD, 2011; http://www.aps.org/policy/reports/ popa-reports/loader.cfm?csModule=security/getfile&PageID=244407 (accessed August 8th, 2011).
ARTICLE
(5) Choi, S.; Drese, J. H.; Jones, C. W. Adsorbent materials for carbon dioxide capture from large anthropogenic point sources. Chemsuschem 2009, 2 (9), 796854; DOI: 10.1002/cssc.200900036. (6) Choi, S.; Drese, J. H.; Eisenberger, P. M.; Jones, C. W. Application of amine-tethered solid sorbents for direct CO2 capture from the ambient air. Environ. Sci. Technol. 2011, 45 (6), 24202427; DOI: 10.1021/Es102797w. (7) Choi, S.; Gray, M. L.; Jones, C. W. Amine-tethered solid adsorbents coupling high adsorption capacity and regenerability for CO2 Capture from ambient air. Chemsuschem 2011, 4 (5), 628635; DOI: 10.1002/cssc.201000355. (8) Nikulshina, V.; Gebald, C.; Steinfeld, A. CO2 capture from atmospheric air via consecutive CaO-carbonation and CaCO3-calcination cycles in a fluidized-bed solar reactor. Chem. Eng. J. 2009, 146 (2), 244248; DOI: 10.1016/j.cej.2008.06.005. (9) Nikulshina, V.; Galvez, M. E.; Steinfeld, A. Kinetic analysis of the carbonation reactions for the capture of CO2 from air via the Ca(OH)(2)-CaCO3-CaO solar thermochemical cycle. Chem. Eng. J. 2007, 129 (13), 7583; DOI: 10.1016/J.Cej.2006.11.003. (10) Nikulshina, V.; Hirsch, D.; Mazzotti, M.; Steinfeld, A. CO2 capture from air and co-production of H2 via the Ca(OH)(2)-CaCO3 cycle using concentrated solar power - Thermodynamic analysis. Energy 2006, 31 (12), 17151725; DOI: 10.1016/J.Energy.2005.09.014. (11) Nikulshina, V.; Steinfeld, A. CO2 capture from air via CaOcarbonation using a solar-driven fluidized bed reactor—Effect of temperature and water vapor concentration. Chem. Eng. J. 2009, 155 (3), 867873; DOI: 10.1016/J.Cej.2009.10.004. (12) Zeman, F. Energy and material balance of CO2 capture from ambient air. Environ. Sci. Technol. 2007, 41 (21), 75587563; DOI: 10.1021/esO7O874m. (13) Zeman, F. Experimental results for capturing CO2 from the atmosphere (R&D note). AIChE J. 2008, 54 (5), 13961399; DOI: 10.1002/aic.11452. (14) Stolaroff, J. K.; Keith, D. W.; Lowry, G. V. Carbon dioxide capture from atmospheric air using sodium hydroxide spray. Environ. Sci. Technol. 2008, 42 (8), 27282735; DOI: 10.1021/es702607w. (15) Nikulshina, V.; Ayesa, N.; Galvez, M. E.; Steinfeld, A. Feasibility of Na-based thermochemical cycles for the capture of CO2 from air— Thermodynamic and thermogravimetric analyses. Chem. Eng. J. 2008, 140 (13), 6270; DOI: 10.1016/j.cej.2007.09.007. (16) Baciocchi, R.; Storti, G.; Mazzotti, M. Process design and energy requirements for the capture of carbon dioxide from air. Chem. Eng. Process. 2006, 45 (12), 10471058; DOI: 10.1016/j.cep.2006.03.015. (17) Mahmoudkhani, M.; Keith, D. W. Low-energy sodium hydroxide recovery for CO2 capture from atmospheric air-Thermodynamic analysis. Int. J. Greenhouse Gas Control 2009, 3 (4), 376384; DOI: 10.1016/J.Ijggc.2009.02.003. (18) Stucki, S.; Schuler, A.; Constantinescu, M. Coupled CO2 recovery from the atmosphere and water electrolysis—Feasibility of a new process for hydrogen storage. Int. J. Hydrogen Energy 1995, 20 (8), 653663; DOI: 10.1016/0360-3199(95)00007-Z. (19) Olah, G. A.; Goepert, A.; Meth, S.; Prakash, G. K. S. Solid sorbent useful for absorbing carbon dioxide from a gas mixture comprises an amine and nano-structured solid support. WO2008021700-A1; US2008293976-A1, WO2008021700-A1, February 21, 2008; B01J-020/ 30 200835. (20) Belmabkhout, Y.; Serna-Guerrero, R.; Sayari, A. Amine-bearing mesoporous silica for CO2 removal from dry and humid air. Chem. Eng. Sci. 2010, 65 (11), 36953698; DOI: 10.1016/J.Ces.2010.02.044. (21) Gebald, C.; Wurzbacher, J. A.; Steinfeld, A. Structure used for adsorption and desorption of carbon dioxide from gas mixture, comprises fiber filaments, and sorbent with amine groups which captures carbon dioxide from gas mixture by adsorption and desorption cycles. WO2010091831-A1; EP2266680-A1, WO2010091831-A, August 19, 2010; B01D-053/04 201056. (22) Lackner, K. S. Capture of carbon dioxide from ambient air. Eur. Phys. J. Special Topics 2009, 176, 93106; DOI: 10.1140/Epjst/E200901150-3. 9106
dx.doi.org/10.1021/es202223p |Environ. Sci. Technol. 2011, 45, 9101–9108
Environmental Science & Technology (23) Lackner, K. S.; Liu, P. Removal of carbon dioxide from air by contacting air with first ion exchange resin, contacting carbon dioxide sorbent with first ion exchange resin, contacting sorbent with weak base anion exchange resin, and heating ion exchange resin. WO2008061210A2; WO2008061210-A3; WO2008061210-A9, WO2008061210-A2, May 22, 2008; B01D-053/02 200846. (24) Wright, A. B.; Lackner, K. S.; Leon-Guerrero, E.; Ginster, U.; Catterson, T.; Madison, K.; Ishikawa, R.; Grimm, G.; Malcomson, M.; Liu, P. Removing carbon dioxide from gas involves bringing gas in contact with resin in chamber, evacuating gas from chamber, wetting resin, collecting water vapor and carbon dioxide from resin, and separating carbon dioxide from water vapor. WO2008131132-A1; WO2008131132B1, WO2008131132-A1, October 30, 2008; B01D-053/02 200878. (25) Wang, T.; Lackner, K. S.; Wright, A. A Moisture swing sorbent for carbon dioxide capture from ambient air. Environ. Sci. Technol. 2011, 45 (15), 66706675; DOI: 10.1021/es201180v. (26) Hicks, J. C.; Drese, J. H.; Fauth, D. J.; Gray, M. L.; Qi, G. G.; Jones, C. W. Designing adsorbents for CO2 capture from flue gashyperbranched aminosilicas capable,of capturing CO2 reversibly. J. Am. Chem. Soc. 2008, 130 (10), 29022903; DOI: 10.1021/ja077795v. (27) Wang, Q.; Luo, J.; Zhong, Z.; Borgna, A. CO2 capture by solid adsorbents and their applications: current status and new trends. Energy Environ. Sci. 2011, 4 (1), 4255; DOI: 10.1039/C0EE00064G. (28) Bacsik, Z.; Ahlsten, N.; Ziadi, A.; Zhao, G.; Garcia-Bennett, A. E.; Martin-Matute, B.; Hedin, N. Mechanisms and kinetics for sorption of CO2 on bicontinuous mesoporous silica modified with n-propylamine. Langmuir 2011; DOI: 10.1021/la202033p. (29) Knofel, C.; Martin, C.; Hornebecq, V.; Llewellyn, P. L. Study of carbon dioxide adsorption on mesoporous aminopropylsilane-functionalized silica and titania combining microcalorimetry and in situ infrared spectroscopy. J. Phys. Chem. C 2009, 113 (52), 2172621734; DOI: 10.1021/Jp907054h. (30) Danon, D.; Stair, P. C.; Weitz, E. FTIR Study of CO2 adsorption on amine-grafted SBA-15: Elucidation of adsorbed species. J. Phys. Chem. C 2011, 115 (23), 1154011549; DOI: 10.1021/jp200914v. (31) Pinto, M. L.; Mafra, L.; Guil, J. M.; Pires, J.; Rocha, J. Adsorption and Activation of CO2 by amine-modified nanoporous materials studied by solid-state NMR and (13)CO2 adsorption. Chem. Mater. 2011, 23 (6), 13871395; DOI: 10.1021/cm1029563. (32) Satyapal, S.; Filburn, T.; Trela, J.; Strange, J. Performance and properties of a solid amine sorbent for carbon dioxide removal in space life support applications. Energy Fuels 2001, 15 (2), 250255; DOI: 10.1021/ef0002391. (33) Knowles, G. P.; Delaney, S. W.; Chaffee, A. L. Diethylenetriamine[propyl(silyl)]-functionalized (DT) mesoporous silicas as CO2 adsorbents. Ind. Eng. Chem. Res. 2006, 45 (8), 26262633; DOI: 10.1021/Ie050589g. (34) Knowles, G. P.; Graham, J. V.; Delaney, S. W.; Chaffee, A. L. Aminopropyl-functionalized mesoporous silicas as CO2 adsorbents. Fuel Process. Technol. 2005, 86 (1415), 14351448; DOI: 10.1016/j. fuproc.2005.01.014. (35) Hornebecq, V.; Knofel, C.; Descarpentries, J.; Benzaouia, A.; Zelenak, V.; Mornet, S.; Llewellyn, P. L. Functionalised micro-/mesoporous silica for the adsorption of carbon dioxide. Microporous Mesoporous Mater. 2007, 99 (12), 7985; DOI: 10.1016/j.micromeso.2006.09.018. (36) Shi, Y.; Wei, J. W.; Shi, J. J.; Pan, H.; Zhao, W.; Ye, Q. Adsorption of carbon dioxide on organically functionalized SBA-16. Microporous Mesoporous Mater. 2008, 116 (13), 394399; DOI: 10.1016/j.micromeso.2008.04.028. (37) Zukal, A.; Dominguez, I.; Mayerova, J.; Cejka, J. Functionalization of delaminated zeolite ITQ-6 for the adsorption of carbon dioxide. Langmuir 2009, 25 (17), 1031410321; DOI: 10.1021/la901156z. (38) Belmabkhout, Y.; Sayari, A. Effect of pore expansion and amine functionalization of mesoporous silica on CO2 adsorption over a wide range of conditions. Adsorption 2009, 15 (3), 318328; DOI: 10.1007/ s10450-009-9185-6. (39) Serna-Guerrero, R.; Belmabkhout, Y.; Sayari, A. Modeling CO2 adsorption on amine-functionalized mesoporous silica: 1. A
ARTICLE
semi-empirical equilibrium model. Chemical Eng. J. 2010, 161 (12), 173181; DOI: 10.1016/J.Cej.2010.04.024. (40) Eichhorn, S. J.; Dufresne, A.; Aranguren, M.; Marcovich, N. E.; Capadona, J. R.; Rowan, S. J.; Weder, C.; Thielemans, W.; Roman, M.; Renneckar, S.; Gindl, W.; Veigel, S.; Keckes, J.; Yano, H.; Abe, K.; Nogi, M.; Nakagaito, A. N.; Mangalam, A.; Simonsen, J.; Benight, A. S.; Bismarck, A.; Berglund, L. A.; Peijs, T. Review: current international research into cellulose nanofibres and nanocomposites. J. Mater. Sci. 2010, 45 (1), 133; DOI: 10.1007/s10853-009-3874-0. (41) Klemm, D.; Heublein, B.; Fink, H. P.; Bohn, A. Cellulose: Fascinating biopolymer and sustainable raw material. Angew. Chem., Int. Ed. 2005, 44 (22), 33583393; DOI: 10.1002/anie.200460587. (42) Harlick, P. J. E.; Sayari, A. Applications of pore-expanded mesoporous silica. 5. Triamine grafted material with exceptional CO2 dynamic and equilibrium adsorption performance. Ind. Eng. Chem. Res. 2007, 46 (2), 446458; DOI: 10.1021/ie060774. (43) Abdelmouleh, M.; Boufi, S.; ben Salah, A.; Belgacem, M. N.; Gandini, A. Interaction of silane coupling agents with cellulose. Langmuir 2002, 18 (8), 32033208; DOI: 10.1021/La011657g. (44) Abdelmouleh, M.; Boufi, S.; Belgacem, M. N.; Duarte, A. P.; Ben Salah, A.; Gandini, A. Modification of cellulosic fibres with functionalised silanes: development of surface properties. International Journal of Adhesion and Adhesives 2004, 24 (1), 4354; DOI: 10.1016/ S0143-7496(03)00099-X. (45) Abdelmouleh, M.; Boufi, S.; Belgacem, M. N.; Dufresne, A.; Gandini, A. Modification of cellulose fibers with functionalized silanes: Effect of the fiber treatment on the mechanical performances of cellulosethermoset composites. J. Appl. Polym. Sci. 2005, 98 (3), 974984; DOI: 10.1002/App.22133. (46) Bel-Hassen, R.; Boufi, S.; Salon, M. C. B.; Abdelmouleh, M.; Belgacem, M. N. Adsorption of silane onto cellulose fibers. II. The effect of pH on silane hydrolysis, condensation, and adsorption behavior. J. Appl. Polym. Sci. 2008, 108 (3), 19581968; DOI: 10.1002/App.27488. (47) Salon, M. C. B.; Abdelmouleh, M.; Boufi, S.; Belgacem, M. N.; Gandini, A. Silane adsorption onto cellulose fibers: Hydrolysis and condensation reactions. J. Colloid Interface Sci. 2005, 289 (1), 249261; DOI: 10.1016/j.jcis.2005.03.070. (48) Salon, M. C. B.; Gerbaud, G.; Abdelmouleh, M.; Bruzzese, C.; Boufi, S.; Belgacem, M. N. Studies of interactions between silane coupling agents and cellulose fibers with liquid and solid-state NMR. Magn. Reson. Chem. 2007, 45 (6), 473483; DOI: 10.1002/Mrc.1994. (49) Xie, Y. J.; Hill, C. A. S.; Xiao, Z. F.; Militz, H.; Mai, C. Silane coupling agents used for natural fiber/polymer composites: A review. Composites, Part A 2010, 41 (7), 806819; DOI: 10.1016/j.compositesa.2010.03.005. (50) Lu, J.; Askeland, P.; Drzal, L. T. Surface modification of microfibrillated cellulose for epoxy composite applications. Polymer 2008, 49 (5), 12851296; DOI: 10.1016/j.polymer.2008.01.028. (51) Beari, F.; Brand, M.; Jenkner, P.; Lehnert, R.; Metternich, H. J.; Monkiewicz, J.; Siesler, H. W. Organofunctional alkoxysilanes in dilute aqueous solution: new accounts on the dynamic structural mutability. J. Organomet. Chem. 2001, 625 (2), 208216; DOI: 10.1016/S0022328X0100650-7. (52) Bennevault-Celton, V.; Maciejak, O.; Desmazieres, B.; Cheradame, H. Condensation of alkoxysilanes in alcoholic media II. Oligomerization of aminopropylmethyldiethoxysilane and co-oligomerization with dimethyldiethoxysilane. Polym. Int. 2010, 59 (9), 12731281; DOI: 10.1002/ Pi.2863. (53) Salon, M. C. B.; Belgacem, M. N. Competition between hydrolysis and condensation reactions of trialkoxysilanes, as a function of the amount of water and the nature of the organic group. Colloids Surf., A 2010, 366 (13), 147154; DOI: 10.1016/j.colsurfa.2010.06.002. (54) Zimmermann, T.; Bordeanu, N.; Strub, E. Properties of nanofibrillated cellulose from different raw materials and its reinforcement potential. Carbohydr. Polym. 2010, 79 (4), 10861093; DOI: 10.1016/J. Carbpol.2009.10.045. (55) Wurzbacher, J. A.; Gebald, C.; Steinfeld, A. Separation of CO2 from air by temperature-vacuum swing adsorption using diamine9107
dx.doi.org/10.1021/es202223p |Environ. Sci. Technol. 2011, 45, 9101–9108
Environmental Science & Technology
ARTICLE
functionalized silica gel. Energy Environ. Sci. 2011, 4, 35843592; DOI: 10.1039/C1EE01681D. (56) Paakko, M.; Vapaavuori, J.; Silvennoinen, R.; Kosonen, H.; Ankerfors, M.; Lindstrom, T.; Berglund, L. A.; Ikkala, O. Long and entangled native cellulose I nanofibers allow flexible aerogels and hierarchically porous templates for functionalities. Soft Matter 2008, 4 (12), 24922499; DOI: 10.1039/B810371b. (57) Aulin, C.; Netrval, J.; Wagberg, L.; Lindstrom, T. Aerogels from nanofibrillated cellulose with tunable oleophobicity. Soft Matter 2010, 6 (14), 32983305; DOI: 10.1039/C001939a. (58) Sehaqui, H.; Salajkova, M.; Zhou, Q.; Berglund, L. A. Mechanical performance tailoring of tough ultra-high porosity foams prepared from cellulose I nanofiber suspensions. Soft Matter 2010, 6 (8), 18241832; DOI: 10.1039/B927505c. (59) Tang, Y. D.; Landskron, K. CO2-sorption properties of organosilicas with bridging amine functionalities inside the framework. J. Phys. Chem. C 2010, 114 (6), 24942498; DOI: 10.1021/Jp9095612. (60) Liang, C. Y.; Marchessault, R. H. Infrared spectra of crystalline polysaccharides 0.2. Native celluloses in the region from 640 to 1700 cm1. J. Polym. Sci. 1959, 39 (135), 269278; DOI: 10.1002/ pol.1959.1203913521. (61) Sun, R. C.; Liu, C. F.; Ren, J. L.; Xu, F.; Liu, J. J.; Sun, J. X. Isolation and characterization of cellulose obtained from ultrasonic irradiated sugarcane bagasse. J. Agric. Food Chem. 2006, 54 (16), 57425748; DOI: 10.1021/jf060929o. (62) Hiyoshi, N.; Yogo, K.; Yashima, T. Adsorption characteristics of carbon dioxide on organically functionalized SBA-15. Microporous Mesoporous Mater. 2005, 84 (13), 357365; DOI: 10.1016/j.micromeso.2005.06.010. (63) Khatri, R. A.; Chuang, S. S. C.; Soong, Y.; Gray, M. Carbon dioxide capture by diamine-grafted SBA-15: A combined Fourier transform infrared and mass spectrometry study. Ind. Eng. Chem. Res. 2005, 44 (10), 37023708; DOI: 10.1021/Ie048997s. (64) Bacsik, Z.; Atluri, R.; Garcia-Bennett, A. E.; Hedin, N. Temperature-induced uptake of CO2 and formation of carbamates in mesocaged silica modified with n-propylamines. Langmuir 2010, 26 (12), 1001310024; DOI: 10.1021/La1001495. (65) Huang, H. Y.; Yang, R. T.; Chinn, D.; Munson, C. L. Aminegrafted MCM-48 and silica xerogel as superior sorbents for acidic gas removal from natural gas. Ind. Eng. Chem. Res. 2003, 42 (12), 2427 2433; DOI: 10.1021/Ie020440u. (66) Wang, X. X.; Schwartz, V.; Clark, J. C.; Ma, X. L.; Overbury, S. H.; Xu, X. C.; Song, C. S. Infrared study of CO2 sorption over “molecular basket” sorbent consisting of polyethylenimine-modified mesoporous molecular sieve. J. Phys. Chem. C 2009, 113 (17), 72607268; DOI: 10.1021/Jp809946y. (67) Attalla, M. I.; Robinson, K.; McCluskey, A. An FTIR spectroscopic study on the effect of molecular structural variations on the CO2 absorption characteristics of heterocyclic amines. Chemphyschem 2011, 12 (6), 10881099; DOI: 10.1002/cphc.201001056. (68) Versteeg, G. F.; Van Dijck, L. A. J.; Van Swaaij, W. P. M. On the kinetics between CO2 and alkanolamines both in aqueous and nonaqueous solutions. An overview. Chem. Eng. Commun. 1996, 144, 113158; DOI: 10.1080/00986449608936450. (69) Sayari, A.; Belmabkhout, Y. Stabilization of amine-containing CO2 adsorbents: dramatic effect of water vapor. J. Am. Chem. Soc. 2010, 132 (18), 63126314; DOI: 10.1021/Ja1013773.
9108
dx.doi.org/10.1021/es202223p |Environ. Sci. Technol. 2011, 45, 9101–9108
CORRESPONDENCE/REBUTTAL pubs.acs.org/est
Comment on “Hydroxycarboxylic Acid-Derived Organosulfates: Synthesis, Stability and Quantification in Ambient Aerosol” n the recent article by Olson et al.1 a method is presented for the synthesis of analytical standards of glycolic (GAS) and lactic acid sulfate (LAS), which are used for their quantification in ambient particulate matter (PM2.5) using hydrophilic-interaction liquid chromatography coupled to electrospray ionization tandem mass spectrometry (HILIC/ESI-MS/MS). In addition, the stability of these standards under acidic conditions and the influence of solvents for filter extraction on the stability of the organosulfates are discussed. The objectives of this comment are to draw attention to some weaknesses of the analytical method used for quantification of the organosulfates, which are influencing the end results, suggest possible improvements, and request additional information. Use of methanol and acetonitrile as extraction solvents: The authors report that if methanol is used for extraction of filters, artifact compounds (methyl carboxylates) of the targeted organosulfates are formed, which will influence their quantification. Therefore, they propose to use water for filter extraction of the very hydrophilic organosulfates like those studied and acetonitrile for extraction of less hydrophilic ones. The aqueous standards and samples were injected on a bare silica column (for HILIC).2 In Figure 2,1 a sample HILIC-MS/MS chromatogram is shown. The chromatographic peaks for GAS and LAS (observed at ion transitions: m/z 155 f m/z 97 and m/z 169 f m/z 97, respectively) look very broad and are split, despite their relatively low retention times. Such peak shapes are expected when the injection solvent used is strongly eluting under the chromatographic conditions. Since water is the strongest eluting solvent under HILIC it is strongly advised not to use fully aqueous solutions for injection because they result in peak broadening and peak distortion (e.g., peak splitting), especially for early eluting peaks.3 Therefore, we recommend to use an injection solvent with a composition that is at most as strong as that of the initial mobile phase composition. In this way, problems with peak shape and peak splitting will be alleviated, allowing a higher column separation efficiency, higher peak heights and higher signal-to-noise, and ensuring as such lower detection limits. In our opinion, acetonitrile deserves to be further investigated for the extraction of GAS and LAS from filters, and if it turns out to be suitable, to be considered for the extraction procedure and as injection solvent. In addition, a mixture of acetonitrile and 10 mM aqueous ammonium formate (90/10 v/v)—initial mobile phase composition2—would be worthwhile exploring. Quantification method: The authors state in the experimental section that “normalization of target compound response can be best addressed by one of the following methods: standard addition, mass labeled analogs of the compounds of interest as internal standards, or performing experiments at different dilutions”. However, they did not mention which quantification method was applied in their study. Taking into account their statements “for GAS and LAS no interference was observed, so co-elution likely does not affect quantification”, “ion suppression
I
r 2011 American Chemical Society
is also not expected to be significant in aerosols samples”, and “GAS and LAS are not commercially available in a mass labeled form, and the cost of synthesis of such mass labeled forms was too high and beyond the scope of this project”, we presume that an external standardization method was used for quantification of GAS and LAS, without having investigated matrix effects and the variability of the ESI process and having addressed the low separation between GAS (Rt = 6.4 min) and LAS (Rt = 6.6 min) (Figures 2 and S2).1 Without using a proper quantification method, like internal standardization or standard addition in LC/MS, the analytical errors can, however, be higher than anticipated. In this respect, we previously showed that matrix effects can be very significant when HILIC-MS/MS is used for atmospheric aerosol analysis;4 hence, the statement: “ion suppression is also not expected to be significant in aerosol samples as it increases with the concentration of ionizing species and aerosol samples have very low analyte concentrations relative to applied charge” does not stand here for several reasons. First, aerosols are very complex mixtures of different inorganic and organic species with different polarities. Second, the retention factors (k) for GAS and LAS (Figure 21) are around 2.23 and 2.33 (calculated according to ref 6, and taking into account the column characteristics in ref 2), respectively. It is well-known that matrix effects are most pronounced around and near the column void time, owing to elution of less retained species from the sample matrix.4,5 The low retention of GAS and LAS and their very broad peaks (∼ 1 min) increase very much the possibility for aerosol matrix compounds to interfere with the ionization of the polar organosulfates studied, thereby affecting their proper quantification. Thus, a careful examination of the matrix effects and their possible impact on the analyte quantification would be warranted. Detailed information about method validation (i.e., recovery, accuracy, linearity, limits of detection/ quantification, specificity, and the influence of the sample matrix) is missing in the article.7 For potential users who would like to conduct analysis of real aerosol samples, this additional information would be very valuable and appreciated. Zoran Kitanovski and Irena Grgic* Laboratory for Analytical Chemistry, National Institute of Chemistry, Hajdrihova 19, PO Box 660, SI-1001 Ljubljana, Slovenia
’ AUTHOR INFORMATION Corresponding Author
*Phone: + 386 1 4760200; fax: +386 1 4760300; e-mail: irena.
[email protected].
’ REFERENCES (1) Olson, C. N.; Galloway, M. M.; Yu, G.; Hedman, C.; Lockett, M. R.; Yoon, T. P.; Stone, E. A.; Smith, L. M.; Keutsch, F. N. Published: September 15, 2011 9109
dx.doi.org/10.1021/es202517d | Environ. Sci. Technol. 2011, 45, 9109–9110
Environmental Science & Technology
CORRESPONDENCE/REBUTTAL
Hydroxycarboxylic acid-derived organosulfates: synthesis, stability and quantification in ambient aerosol. Environ. Sci. Technol. 2011, 45, 6468 6474, DOI: 10.1021/es201039p. (2) Stone, E. A.; Hedman, C. J.; Sheesley, R. J.; Shafer, M. M.; Schauer, J. J. Investigating the chemical nature of humic-like substances (HULIS) in North American atmospheric aerosols by liquid chromatography tandem mass spectrometry. Atmos. Environ. 2009, 43, (27), 4205 4213; DOI 10.1016/j.atmosenv.2009.05.030. (3) Ruta, J.; Rudaz, S,; McCalley, D. V.; Veuthey, J. L.; Guillarme, D. A systematic investigation of the effect of sample diluent on peak shape in hydrophilic interaction liquid chromatography. J. Chromatogr., A 2010, 1217, (52), 8230 8240; DOI 10.1016/j.chroma.2010.10.106. (4) Kitanovski, Z.; Grgic, I,; Veber, M. Characterization of carboxylic acids in atmospheric aerosols using hydrophilic interaction liquid chromatography tandem mass spectrometry. J. Chromatogr,. A 2011, 1218, (28), 4417 4425; DOI 10.1016/j.chroma.2011.05.020. (5) Antignac, J. P.; de Wasch, K,; Monteau, F.; De Brabander, H.; Andre, F.; Le Bizec, B. The ion suppression phenomenon in liquid chromatography-mass spectrometry and its consequences in the field of residue analysis. Anal. Chim. Acta 2005, 529, (1 2), 129 136; DOI 10.1016/j.aca.2004.08.055. (6) Snyder, L. R.; Kirkland, J. J.; Dolan, J. W. Introduction to Modern Liquid Chromatography, 3rd, ed.; John Wiley & Sons: Hoboken, NJ, 2010. (7) Dejaegher, B.; Mangelings, D.; Vander Heyden, Y. Method development for HILIC assays. J. Sep. Sci. 2008, 31, (9), 1438 1448; DOI 10.1002/jssc.200700680.
9110
dx.doi.org/10.1021/es202517d |Environ. Sci. Technol. 2011, 45, 9109–9110
CORRESPONDENCE/REBUTTAL pubs.acs.org/est
Reply to Comment on “Hydroxycarboxylic Acid-Derived Organosulfates: Synthesis, Stability and Quantification in Ambient Aerosol” §
I
Environmental Chemistry and Technology, University of WisconsinMadison, Madison, Wisconsin 53706, United States
’ ASSOCIATED CONTENT
bS
Supporting Information. Further details about the analysis of glycolic acid sulfate and lactic acid sulfate in ambient air sample filters. This material is available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION Corresponding Author
*E-mail:
[email protected]. Present Addresses
)
n response to the request for additional information on the quantitative HPLC-MS/MS parameters for the analysis of glycolic acid sulfate (GAS) and lactic acid sulfate (LAS) in ambient air sample filters, we uploaded additional Supporting Information available via the Internet at http://pubs.acs.org, which provides details about method recovery, accuracy, linearity, limits of detection, and specificity. We understand that aqueous injection is not ideal for HILIC chromatography; however, this concession was needed to be able to study water-soluble ambient aerosol samples without introducing extraction artifacts. We believe that the methanol effect, in which carboxylic acid groups in organosulfates form methyl esters, as further described in the article, validates this approach. Acetonitrile may ultimately indeed prove to be a better extraction solvent for the study of polar organosulfates, including GAS and LAS in ambient air samples, but further study and validation is needed to show that acetonitrile will not adversely affect the molecules of interest. The authors of the comment are correct in their assumption that external standard quantification was used for this study. For consistency with ambient samples, standard solutions were diluted in aqueous solution to create calibration curves. Ion suppression is an important concern in analysis of environmental samples by LC/MS. As stated in the published article, ion suppression will interfere with measurement and occurs when readily ionized ions at high concentration (e.g., sulfate) compete with analytes of interest for applied charge. Under the chromatographic conditions utilized, inorganic sulfate and other inorganic ions were separated from LAS and GAS analytes, removing such competition for ionization. The field samples were also analyzed at full strength and at 1:20 dilution. Comparison of the two resulting concentrations, after having applied the dilution factor, showed good agreement and indicated that ion suppression had not significantly impacted their quantification. Efforts are currently under way in our laboratories to improve upon this analytical method, including the synthesis of mass labeled GAS and mass labeled LAS for internal standardization. The use of these mass labeled internal standards provides the best approach to account for potential matrix effects in future studies. We are grateful for the insights made in the comment and the commenter’s new work in the field of ambient aerosol research,2 and for the opportunity to provide additional details to clarify our analytical methodology.
Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02138, United States ^ Department of Chemistry, University of Iowa, Iowa City, IA 52242
’ REFERENCES (1) Olson, C. N.; Galloway, M. M. ; Yu, G.; Hedmann, C. J.; Lockett, M. R. ; Yoon, T. P.; Stone, E. A.; Smith, L. M.; Keutsch F. N. Hydroxycarboxylic acid-derived organosulfates: synthesis, stability and quantification in ambient aerosol. Environ. Sci. Technol. 2011, 45, 6468 6474, DOI: 10.1021/es201039p. (2) Kitanovski, Z; Grgic, I; Veber, M. Characterization of carboxylic acids in atmospheric aerosols using hydrophilic interaction liquid chromatography tandem mass spectrometry. J. Chromatogr., A 2011, 1218(28), 4417 4425.; DOI 10.1016/j.chroma.2011.05.020.
Corey N. Olson,† Melissa M. Galloway,† Ge Yu,† Curtis J. Hedman,‡ Matthew R. Lockett,†,|| Tehshik P. Yoon,† Elizabeth A. Stone,§,^ Lloyd M. Smith,† and Frank N. Keutsch†,* †
Department of Chemistry, University of Wisconsin-Madison, Madison, Wisconsin 53706, United States
‡
Wisconsin State Laboratory of Hygiene, Environmental Health Division, Madison, Wisconsin 53718, United States
r 2011 American Chemical Society
Published: September 15, 2011 9111
dx.doi.org/10.1021/es203122z | Environ. Sci. Technol. 2011, 45, 9111–9111
