No title

VIEWPOINT pubs.acs.org/est Rainwater Management and Harvesting Strategies for Human Needs: An Indian Perspective Ravi ...

Author: Jerald Schnoor

5 downloads 460 Views 100MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

VIEWPOINT pubs.acs.org/est

Rainwater Management and Harvesting Strategies for Human Needs: An Indian Perspective Ravi Rangarajan and Prosenjit Ghosh* Centre for Earth Sciences, Indian Institute of Science, Bangalore, India

I

ndia is a growing economy with more than a billion people, and is dependent on monsoonal precipitation for its water needs. Recent data shows that only 1% of the existing water resources on the entire planet can be utilized by human population. The remaining 99% of the existing water resources consist of 97% saltwater and 2% ice caps and are nonexploitable.1 These numerical facts become even more alarming when it is estimated that the world’s population recently exceeded 6 billion people, and ∼25% of them stay in the Indian subcontinent. The annual need of 1800 cubicmeters of water per person is barely met and governmental agencies demand schemes for the better management of fresh water supply through river systems2 and other reservoirs of fresh water. A signiﬁcant proportion of fresh water demand is met through precipitation received during the summer time through the South-West monsoon and partly during winter through the North-East monsoon. The observations on global warming and changing rainfall pattern, although uncertain, predicts a condition where driergets-drier and wetter-gets-wetter.3 The regions experiencing large scarcity of water due to excessive exploitation needs to adopt mechanisms to deal with the present scarcity situation. The idea and plan of rainwater harvesting to meet the water r 2011 American Chemical Society

need of the present population is highly attractive and it requires the tapping of water bodies like lakes and shallow aquifers. Within the Indian subcontinent, the regions covered by thick alluvium exhibit least water scarcity (Figure 1).4 The exploitation of water resources for human demand is largely based on available water through natural agencies like river and groundwater. The water scarcity index map for the region which captures the extent of exploitation of available water resources shows higher depletion for states in South Indian Peninsula. The regions in south and central India with bedrock made up of granite gneiss and Deccan basalt rock types are characterized by shallow aquifer having shorter residence time of groundwater. The rain fed groundwater reservoirs are over exploited due to high demand and shortage of supply through rainfall. A signiﬁcant fraction of the rainfall ﬂows into the ocean without being arrested by any aquifers or water bodies. The potential retrieval of this water to compensate for the scarcity of groundwater in the region can be addressed through rainwater harvesting program either through lakes or shallow aquifer. The economic exploitation of these resources, which carry the signature of monsoon-fed rainwater, has been well demonstrated in the study conducted on a brand of bottled water from the southern Indian city of Bangalore.5 A simultaneous monitoring of the stable isotopic ratios in regional rainwater and bottle water purchased at regular monthly time intervals showed isotopic imprints of rainwater in the bottle water. Based on the observations, it was concluded that commercially available bottled water is manufactured by tapping and treating of the monsoon-fed fresh water available in the region. The study provides a clear demonstration for the scope of harvesting monsoon rainfall to meet the need of portable water in this region. Further, the approach can be implemented in major scale through community based harvesting, augmentation of this scheme in municipality and a systematic distribution of conserved water through proper channelized network. Existing water bodies like lakes, ponds and wells, available in the urban locations can be eﬃciently utilized for this purpose. An approach of a similar kind in other region may allow circumventing the deﬁciency of drinking water at regional and community levels.

Received: September 14, 2011 Accepted: September 26, 2011 Published: October 20, 2011 9469

dx.doi.org/10.1021/es203308p | Environ. Sci. Technol. 2011, 45, 9469–9470

Environmental Science & Technology

VIEWPOINT

Figure 1. (a) Showing the Water Scarcity Index (WSI)4 of Peninsular India along with the average net groundwater availability during the year 2010 (Data Source: Central Ground Water Board, Govt. of India, 2010 report) and the Population statistics of the year 2011(http://www.world-gazetteer.com). Major cities with high population density are marked. (b) Bangalore City which houses large inland water bodies, acts as an ideal ground for harvesting rainwater for human needs.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

(5) Rangarajan, R.; Ghosh, P. Tracing the source of bottled water using stable isotope techniques. Rapid Commun. Mass Spectrom. 2011, 25, 3323 3330.

’ REFERENCES (1) Rooy, T. B.V. Bottling up our natural resources: the ﬁght over bottled water extraction in the United States. J. Land Use and Environ. Law 2003, 18, 267–298. (2) Fairless, D. Muddy waters. Nature 2008, 452, 278–281. (3) Science news. http://www.sciencedaily.com/releases/2010/02/ 100226093238.htm (accessed Aug 28, 2011). (4) Smakhtin, V.; Ravenga, C.; Doll, P. A pilot global assessment of environmental water requirements and scarcity. Water Int. 2004, 29 (3), 307–317. 9470

dx.doi.org/10.1021/es203308p |Environ. Sci. Technol. 2011, 45, 9469–9470

VIEWPOINT pubs.acs.org/est

Anthrax Cleanup Decisions: Statistical Confidence or Confident Response Igor Linkov,*,† John B. Coles,† Paul Welle,† Matthew Bates,† and Jeﬀrey Keisler‡ †

Environmental Laboratory, US Army Engineer Research and Development Center, US Army Corps of Engineers, Vicksburg, Mississippi ‡ University of Massachusetts, Boston, Massachusetts

“C

an you guarantee this building is safe?” This question is diﬃcult to answer in the wake of a biological attack. It is likewise diﬃcult to decide when people may return to a place where dangerous agents were once dispersed. This decision becomes especially challenging when even minute quantities of the suspected contaminating agent could pose a signiﬁcant threat (e.g., weaponized anthrax). Following the 2001 US anthrax attacks, the Government Accountability Oﬃce (GAO) directed the U.S. Department of Homeland Security (DHS) to develop a defensible strategy for making such decisions following biological incidents.1 Statistically based sampling and analysis has historically provided the information to guide remediation of contaminated sites. However, classical statistics approaches could require thousands of tests2 to conclude that the number of anthrax spores is below the level of 0.1 spores per square meter (a concentration which could still pose a signiﬁcant risk). Supplemental experimental data (e.g., dispersion studies, preliminary sampling) may be diﬃcult to obtain due to ﬁnancial constraints and could be limited to speciﬁc experimental conditions. Moreover, classical statistics expresses results as conﬁdence or tolerance intervals and does not actually state a probability of r 2011 American Chemical Society

contamination necessary for risk-based decision making, which poses a challenge of communicating test results in a meaningful way.3 For example, the conﬁdence interval statement, “with 90% conﬁdence the probability of contamination is less than 5%,” means “if the probability of contamination was really 5% or greater, there is only a 10% chance we would have observed no contamination.” Likewise, the tolerance interval statement “with 90% conﬁdence, 95% of the room has no contamination” is not reassuring. Modern Bayesian statistical approaches (developed starting in the 1950s, building from a tradition going back to Bayes and Laplace in the 1700s-1800s) facilitate inferences about the probability of remaining contamination, eﬀectively overcoming much of the aforementioned challenge. Bayesian methods can smoothly incorporate data from laboratory experiments (e.g., decontaminating diﬀerent surfaces such as steel and concrete), which is useful when conditions are dynamic. Additionally, when relevant expert knowledge exists, Bayesian statistics can smoothly incorporate it into decision making, in place of further sampling. Finally, Bayesian methods can support direct statements about: probabilities, e.g., “there is a 95% probability that the room is not contaminated”; probability ranges, e.g., “are 95% conﬁdent that the probability of the room being clean is between 2% and 4%”; or probability distributions, e.g., “the probability distribution over the number of spores follows a beta distribution with these parameters” or even “there is a 1% chance that there are ten or more spores remaining” (which could be useful in less hazardous situations where some risk of exposure may be tolerable). An important caveat is that the probability statements are still fundamentally an assertion of the expert beliefs based on subjective inputs regarding properties of contaminants and cleanup methods under diﬀerent conditions tempered by the logical implications of the observed data. They should not be viewed as a way to “launder” opinions into fact, and if experts do not know enough to provide strong judgments, new data will still be needed to gain adequate certainty about treatment success. The other key limitation of the Bayesian approach is that people— including subject matter experts—are known to have a variety of systematic biases in making subjective probability estimates. Scientiﬁc research is replete with examples of overconﬁdence, where experts were slow to update their beliefs in the face of new Received: September 29, 2011 Accepted: October 4, 2011 Published: October 26, 2011 9471

dx.doi.org/10.1021/es203479t | Environ. Sci. Technol. 2011, 45, 9471–9472

Environmental Science & Technology

VIEWPOINT

assessment and remediation approach could more rapidly assess and respond to future biological threats.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]. US Army Corps of Engineers, 696 Virginia Rd., Concord, MA 01742.

’ ACKNOWLEDGMENT Permission was granted by the Chief of Engineers to publish this information.

Figure 1. Proposed cleanup decision framework.

information. But there is also a rich literature on how to minimize these biases and validate results based on small (and thus, relatively inexpensive) sets of objective data. Correctly applied Bayesian methods ﬁll an important gap when empirical data are limited. Figure 1 proposes a basic framework in which Bayesian methods could combine expert judgments about the threat with lab-data about decontamination eﬃcacy to provide an estimate of remaining threat. Because it is able to integrate onsite judgment from ﬁrst responders and experts in the ﬁeld, such a model could help guide remediation and testing. The likelihood of remaining contamination is one input to a larger decision context for which experts from pertinent ﬁelds (e.g., counter-terrorism, law-enforcement, epidemiology, and policy), can provide other relevant information. Speciﬁcally, onsite observations (room characteristics, contaminated material composition, and wind/draft potential) are represented by the dark blue nodes. Onsite test data results (characterization, remediation test strips) are represented by light blue nodes. The red nodes are informed by historical laboratory data and oﬀsite expert judgments about the probabilistic relations between these nodes and their predecessors, and Bayesian methods are then used to calculate probability distributions on each red node’s values, given those of its predecessors. The probability distribution on the number of surviving spores is derived directly from: (1) the initial number of spores (more generally, colony forming units), (2) the percentage that would be killed by remediation under laboratory conditions, and the (3) eﬃcacy of onsite remediation compared to laboratory conditions. Issues 1 3 above are each informed by oﬀ or onsite data and judgment about their predecessors, as described. The distribution on the number (possibly zero) of remaining spores is an input to the overall threat calculation. With both strengths and limitations, the Bayesian approach is well-suited to guide remedial decisions in problems like anthrax remediation. Consistent procedures and inputs from both objective and subjective sources and procedures are important factors in making traceable and justiﬁable decisions based on estimates of probability given the available data. The better and more empirical the data, the less need there is to rely on judgment (and, in fact, Bayesian and classical statistical approaches converge to the same conclusions as suﬃciently rich data become available). Through the continued development and integration of Bayesian and classical statistical method, a consistent, reliable, and scalable

’ REFERENCES (1) Agencies Need to Validate Sampling Activities in Order to Increase Conﬁdence in Negative Results, GAO-05-251; Government Accountability Oﬃce: Washington, DC, 2005. (2) Price, P. N., Sohn, M. D.; Lacommare, K. S. H.; , McWilliams, J. A.; Framework for evaluating anthrax risk in buildings. Environ. Sci. Technol. 2009, 43, 1783 1787. (3) Benchmark Dose Analysis for Bacillus anthracis Inhalation Exposures in the Nonhuman Primate and Application to Risk-Based Decision Making, EPA 600/R-10/138; Environmental Protection Agency: Washington, DC, 2010.

9472

dx.doi.org/10.1021/es203479t |Environ. Sci. Technol. 2011, 45, 9471–9472

CRITICAL REVIEW pubs.acs.org/est

Technical, Economical, and Climate-Related Aspects of Biochar Production Technologies: A Literature Review Sebastian Meyer,*,† Bruno Glaser,‡ and Peter Quicker§ †

Bioenergy Unit, Ecofys, 81243 M€unchen, Germany Soil Biogeochemistry, Martin-Luther-University Halle-Wittenberg, 06120 Halle, Germany § Unit of Technology of Fuels, RWTH Aachen, 52062 Aachen, Germany ‡

ABSTRACT: For the development of commercial biochar projects, reliable data on biochar production technologies is needed. For this purpose, peer-reviewed scientiﬁc articles on carbonization technologies (pyrolysis, gasiﬁcation, hydrothermal carbonization, and ﬂash carbonization) have been analyzed. Valuable information is provided by papers on pyrolysis processes, less information is available on gasiﬁcation processes, and few papers about hydrothermal and ﬂash carbonization technologies were identiﬁed. A wide range of data on the costs of char production (between 51 US$ per tonne pyrolysis biochar from yard waste and 386 US$ per tonne retort charcoal) and on the GHG balance of biochar systems (between 1054 kg CO2e and +123 kg CO2e per t dry biomass feedstock) have been published. More data from pilot projects are needed to improve the evaluation of biochar production technologies. Additional research on the inﬂuence of biochar application on surface albedo, atmospheric soot concentration, and yield responses is necessary to assess the entire climate impact of biochar systems. Above all, further ﬁeld trials on the ability of diﬀerent technologies to produce chars for agricultural soils and carbon sequestration are essential for future technology evaluation.

1. INTRODUCTION In recent years, biochar application to soil has been put forward as a tool to mitigate global warming and improve soil properties.1 3 In spite of considerable scientiﬁc work on the eﬀects of biochar application to soil with respect to crop yields and stabilization of plant-derived carbon in agricultural soils, the commercial production of biochar for soil improvement and C sequestration is still very limited today. Parties interested in the development of commercial biochar need reliable and comprehensive data on the diﬀerent technologies available for biochar production. For this reason, this paper summarizes the available peer-reviewed scientiﬁc literature (ISI Web of Knowledge) about the technological, economical, and climate-relevant aspects of carbonization technologies. Biochar is deﬁned as “charred organic matter applied to soil in a deliberate manner, with the intent to improve soil properties” in Lehmann et al.4 Although biomass-derived char can be used as energy carrier, as adsorber, and for further applications, this paper focuses on the production of chars for the improvement of soil properties. Carbonized organic matter can have fundamentally diﬀerent physical and chemical properties depending on the technology (e.g., torrefaction (a pyrolysis process at low temperature), slow pyrolysis, intermediate pyrolysis, fast pyrolysis, gasiﬁcation, hydrothermal carbonization (htc), or ﬂash carbonization) used for its production. Research on torreﬁed material as soil amendment has started only recently.5 In contrast to considerable r 2011 American Chemical Society

research which has already been carried out to assess the value of charcoal as soil amendment,6 10 no publication was identiﬁed which examines the use of chars from modern gasiﬁers as soil amendment. Charcoal can be produced both in traditional earthen charcoal kilns where pyrolysis, gasiﬁcation, and combustion processes are carried out in parallel below the earthen kiln layer and in modern charcoal retorts where pyrolysis and combustion processes are physically separated by a metal barrier. Two papers have been published on the suitability of htc-char for the stabilization of organic carbon,11,12 and another on the suitability of htc-char for the improvement of soil properties.13 Only one publication is available today in the ISI Web of Knowledge on the suitability of the use of carbonized material from ﬂash carbonization as a soil amendment.14 It is important to be aware that the results of the indicated publications with carbonized material from torrefaction, hydrothermal carbonization, and ﬂash carbonization did not show an improvement of plant growth after the addition of carbonized material. As phytotoxic components have been found in torreﬁed material5 and torreﬁed material has hydrophobic properties, this technology is treated in less detail in this review. Apart from that, all main technology routes already mentioned have been fully Received: December 22, 2010 Accepted: September 30, 2011 Revised: September 13, 2011 Published: September 30, 2011 9473

dx.doi.org/10.1021/es201792c | Environ. Sci. Technol. 2011, 45, 9473–9483

Environmental Science & Technology

CRITICAL REVIEW

Table 1. Publications Identiﬁed and Reviewed Per Category technological

proﬁtability

climate

maturity

analyses

impacts

pyrolysis

2

7

6

gasiﬁcation

7

2

1

hydrothermal carbonization

2

0

0

ﬂash carbonization

1

1

0

technology type

included in this literature review as today’s knowledge on the suitability of carbonized material from modern gasiﬁcation, hydrothermal carbonization, and ﬂash carbonization for the improvement of soil properties is still very limited. However, it is indispensable to further assess the ability of the diﬀerent technologies to produce carbonized material suitable to increase the fertility of agricultural soils and to store carbon over a long period of time. In this context, special care has to be taken to avoid the use of chars contaminated with polycyclic aromatic compounds or dioxins for agricultural purposes. A detailed discussion of dioxin formation is presented by McKay,12 and limits for dioxin and polycyclic aromatic hydrocarbon levels in compost and sewage sludge in European countries can be found in Libra et al.12 It should be noted that in ﬁeld trials, often mixtures of char and compost are used with the aim to produce a soil amendment similar to the fertile Terra Preta soils in the Amazon region.1 Though char makes up a minor weight component of this soil amendment, it is an essential part of the ﬁnal product.

2. METHODOLOGY To identify the relevant literature for this review, the ISI Web of Knowledge was explored with the following method: By searching for articles containing the keywords “pyrolysis”, “gasiﬁcation”, “hydrothermal carbonization”, and “ﬂash carbonization” in connection with the keywords “reliability”, “availability”, “durability”, “development + hours”, and “scale up”, the technological maturity of carbonization technologies was assessed. To retrieve publications that analyze the economical proﬁtability of carbonization technologies, the keywords “proﬁtability”, “economics”, “production costs + char”, and “return + char” were used. Regarding the climate impact of carbonization technologies, the keywords “GHG balance”, “LCA”, “albedo”, and “atmospheric soot” were selected. The available peer-reviewed scientiﬁc literature about the technological, economical, and climate-relevant aspects of the diﬀerent technologies varies considerably. This can be seen in the overview on the number of publications identiﬁed and reviewed per technology and assessment aspect (Table 1). In addition to that, information on carbonization technologies is often focused on the production of energy carriers only. This will be reﬂected in the following chapters. As this paper concentrates on publications in the ISI, it cannot be excluded that additional publications are available in other scientiﬁc databases. 3. OVERVIEW OF CARBONIZATION TECHNOLOGIES To produce carbonized organic matter, pyrolysis, gasiﬁcation, hydrothermal carbonization, and ﬂash carbonization technologies can be used. Pyrolysis can be diﬀerentiated from gasiﬁcation

by the (nearly) complete absence of oxygen in the conversion process.16 Pyrolysis technologies can be further diﬀerentiated by the reaction time of the pyrolysis material (e.g., slow and fast pyrolysis processes) and the heating method (e.g., pyrolysis processes started by the burning of fuels, by electrical heating, or by microwaves). Bridgwater16 and IEA Bioenergy17 diﬀerentiate pyrolysis technologies according to the temperature and the residence time of the pyrolysis process (see Table 2). In gasiﬁcation processes, the biomass is partially oxidized in the gasiﬁcation chamber18 at a temperature of about 800 °C16 at atmospheric or elevated pressure. As already indicated by its name, the main product of this process is gas, only small amounts of char and liquids are formed. The hydrothermal carbonization of biomass is realized by applying elevated temperature (180 220 °C) to biomass in a suspension with water under elevated pressure for several hours. It yields solid, liquid, and gaseous products.19 Libra et al.12 refer to hydrothermal carbonization as “wet pyrolysis”. Because no oxygen is supplied to the reactor with the biomass water suspension, this classiﬁcation is justiﬁed. For the ﬂash carbonization of biomass, a ﬂash ﬁre is ignited at elevated pressure (at about 1 2 MPa) at the bottom of a packed bed of biomass. The ﬁre moves upward through the carbonization bed against the downward ﬂow of air added to the process. In total about 0.8 1.5 kg of air per kg of biomass are delivered to the process. The reaction time of the process is below 30 min, and the temperature in the reactor is in the range of 300 600 °C. The process results mainly in gaseous and solid products. In addition to that, a limited amount of condensate is formed. While the oxygen input into the carbonization process is a typical feature of gasiﬁcation technologies, both process temperature and the product spectrum (distribution among solid, liquid, and gaseous outputs) of ﬂash carbonization are uncommon for gasiﬁcation processes. It should be noted that typical solid product yields obtained by gasiﬁcation and fast pyrolysis processes are signiﬁcantly lower as compared to the solid product yields of slow pyrolysis, ﬂash carbonization, hydrothermal, carbonization and torrefaction (see Table 2). It is important to take into account that the development history of the diﬀerent technologies reviewed varies considerably: The development of coal gasiﬁcation started already a few centuries ago26 whereas the development of charcoal kilns has taken place over a time span of millennia.21

4. TECHNOLOGICAL MATURITY OF CARBONIZATION TECHNOLOGIES To understand the challenges that need to be solved to ensure a high annual availability of a biochar production system, Table 3 lists technical points that need special attention to ensure a longterm operation of the respective technologies. 4.1. Pyrolysis Technologies. Bridgwater et al.18 assumed an overall annual availability of 85% for an electricity production process based on liquids produced by fast pyrolysis. This assumption is used in a model to calculate the electricity production costs of the process. The assumption itself is based on the precondition that a buffer storage for pyrolysis liquids limits unplanned generation shutdowns. Thus, the pyrolysis process itself can have a considerably lower availability than 85%. 4.2. Gasification. Bridgwater et al.18 assumed a mean annual availability of 80% for an electricity production process based on 9474

dx.doi.org/10.1021/es201792c |Environ. Sci. Technol. 2011, 45, 9473–9483

Environmental Science & Technology

CRITICAL REVIEW

Table 2. Solid Product Yields, Solid Product Carbon Content, and Carbon Yield of Diﬀerent Technologies

process type

typical process temperature

typical residence time

typical solid product yield

typical carbon content

typical carbon

on a dry wood feed-stock basis [in mass %]

of the solid product [in mass % ]

yield: (mass carbon, product/ mass carbon, feedstock)

61 84%

51 55%

0.67 0.85

16, 20

≈ 30%

95%

≈ 0.58

16, 21

12 26%

74%

0.2 0.26

reference

torrefaction

∼ 290 °C

10

slow pyrolysis

∼ 400 °C

minutes to days

fast pyrolysis

∼ 500 °C

∼1s

gasiﬁcation

∼ 800 °C

∼10 to 20 s

≈ 10%

htc

∼ 180 250 °C

1 12 h

<66%b

<70%a

≈ 0.88

12, 24

ﬂash carbonization

∼ 300 600 °C

<30 min

37%

≈ 85%

≈ 0.65

25

60 min

16, 17, 22, 23 15, 16

a

The carbon content of 70% of the product indicated in Tsukashima24 is related to the dry, ash-free product. b Lower solid product yields for htc at both shorter and longer residence times are reported by Libra et al.12

Table 3. Selection of Technology-Speciﬁc Challenges technology type

technological challenges

reference 27

Jensen et al., Bridgwater et al. 18

pyrolysis • achieving and maintaining high, controlled heat rates and a correct reaction temperature; a low gas-vapor residence time at a moderate temperature • a rapid removal of char and eﬀective liquids recovery can be challenging in fast pyrolysis systems • the release of chlorine from feedstock with high Cl content may result in corrosion of the reactor containment and in formation of deposits during pyrolysis gas conversion

Nunes et al.,33 Buchireddy et al. 28

gasiﬁcation • aerosol formation • soot formation due to repolymerization • dehydration of tars in the gas phase and interaction with other contaminants on ﬁne particles • condensation of heavier tar components on cooler surfaces • blockage of particulate ﬁlters and clogging of fuel lines/injectors in internal combustion engine • corrosion caused by tars

Funke and Ziegler,19 Libra et al. 12

hydrothermal carbonization • the elasticity limit of the materials used for the pressure tank may not be exceeded during operation • feeding against pressure in continuous systems is challenging regarding material and safety aspects • a heat recovery from the hot process water and post-treatment installations for the htc-char might be necessary ﬂash carbonization

Wade et al.29 • sudden pressure rise in carbonization container observed at ignition under speciﬁc process conditions with certain feedstock’s • the elasticity limit of the materials used for the pressure tank may not be exceeded during operation

a dual fuel diesel engine fed by the atmospheric gasification of wood chips and diesel as auxiliary fuel. In the model, it is assumed that the ash produced from the atmospheric gasification process contains 33% char on a weight basis. Bridgewater et al.18 further assumed the same annual availability (80%) for an electricity production process based on an integrated gas turbine combined cycle fed by the pressurized gasification of wood chips. It has to be noted that this technology is still in an early development stage. Yin et al.30 described a circulating ﬂuidized bed biomass gasiﬁcation and power generation system based on rice husk

installed in 1998 which has been operating for 10 000 h within two years of operation. A considerable part of the char produced in the gasiﬁer is removed from the product gas and returned to the gasiﬁer. According to Pr€oll et al.,31 the 8 MWth dual ﬂuidized bed steam gasiﬁcation plant for solid biomass in G€ussing, Austria has been operated for 24 000 h between April 2002 and the end of 2006. This translates into an average plant availability of about 58%. In 2005, an average availability of about 69% was reached (6000 operation hours in one year).32 9475

dx.doi.org/10.1021/es201792c |Environ. Sci. Technol. 2011, 45, 9473–9483

Environmental Science & Technology

CRITICAL REVIEW

Table 4. Annual Availability of Electricity Production Systems process name

annual availability

comment

fast pyrolysis

>85%

model assumption

Bridgwater et al.18

atmospheric gasiﬁcation

80%

model assumption

Bridgwater et al.18

pressurized gasiﬁcation

80%

model assumption

Bridgwater et al.18

ﬂuidized bed gasiﬁcation

90%

model assumption

Yassin et al.34

circulating ﬂuidized bed gasiﬁcation

57%

empirical data

Yin et al.30

circulating ﬂuidized bed gasiﬁcation

58%; 69%

empirical data

Pr€oll et al.,31 Kreuzeder et al. 32

Nunes et al.33 described the negative eﬀect of tar formation on the operational availability of gasiﬁcation, but do not indicate quantitative numbers on the annual availability of gasiﬁcation processes. Yassin et al.34 assumed an availability of an electricity production system based on the ﬂuidized bed gasiﬁcation of residual waste of 90% (329 days per year) within a model to evaluate the techno-economic performance of energy from waste ﬂuidized bed gasiﬁcation. 4.3. Summary. Most information is available on gasification processes, less information is available on pyrolysis technologies (Table 4). Papers on the availability and reliability of carbonization technologies are often not based on empirical data and do not cover hydrothermal and flash carbonization technologies at all. The available knowledge is focused on systems for the production of pyrolysis oil, synthesis gas, electricity, or heat. In summary, a comparison of the technological maturity of biochar production technologies based on scientific literature in not possible at the moment. In Table 4, data on the annual availability of production processes as indicated in the reviewed papers are summarized. The difference between the assumed availability rates and the empirical data is evident.

5. PROFITABILITY ANALYSES OF CARBONIZATION TECHNOLOGIES The focus of the literature review in this section was on publications issued not earlier than the year 2000, since proﬁtability analyses are subject to rapidly changing economic framework conditions. Apart from that, they are often only valid for a speciﬁc project in a speciﬁc region at a speciﬁc point in time. Depending on the type of technology used, biogenic energy carriers (biogenic oil, synthesis gas), electricity, or heat are produced together with the char and constitute the main product, byproduct, or residue of the biomass conversion process. A considerable part of the described processes aim primarily at the provision of bioenergy. In these cases, the indicated economic data cannot be directly used to assess the proﬁtability of the described technologies for biochar production. However, the indicated economic data can be used as a benchmark to assess under which conditions the production of biochar would be more (or less) proﬁtable compared to the production of a bioenergy carrier. It is important to be aware that the (bio)energy and the biochar markets compete for the same feedstock, and that biochar-based soil amendments compete with other products (e.g., peat, pure compost) and other methods used to improve soil properties. To enable a direct comparison, the published economic data have been converted to US $ at historical exchange rates and has been inﬂationadjusted to the reference year 2010 (US$2010).

reference

5.1. Pyrolysis. Islam and Ani35 carried out a techno economic

assessment of fluidized bed fast pyrolysis systems with rice husk throughputs of 100 and 1000 kg/h. The systems were assumed to be installed in Malaysia in a study carried out in 2000. Since the study focuses on the production of the pyrolysis oil, production costs of 0.38 US$ (0.5 US$2010) per kg primary pyrolysis oil [at 100 kg feed/hour] and 0.18 US$ (0.23 US$2010) per kg primary oil [at 1000 kg feed/hour] have been calculated. Because solid char is coproduced by the fluidized bed fastpyrolysis system, the sales price of the primary pyrolysis oil could be reduced if the char would be sold. Unfortunately, production costs for the solid char cannot be derived from this study. However, the energy value of char from fast pyrolysis within the process can be calculated. Based on the higher heating value of char from fast pyrolysis (about 28 MJ/kg)22 and the price of wood pellets as an alternative biogenic heating fuel (0.05 h2011/kWh; 0.0202 $2010/MJ), the energy value of the char is as high as 560 US$2010 /tonne. It should be noted here that the value of the char is considerably lower if a cheaper conventional fuel is replaced. For example, Badger et al.36 calculate with residual heating oil as replacement at a price of 0.0109 $2010/MJ. Bridgwater et al.18 calculated electricity production costs of 0.091 h (0.098 US$2010)/kWh at 20 MWel and 0.199 h (0.215 US$2010)/kWh at 1 MWel for a modeled electricity production process in UK based on liquids produced by fast pyrolysis and diesel as auxiliary fuel. Taking into account learning eﬀects and assuming a 50% reduction of capital costs of fast pyrolysis modules after 10 installations, they calculated electricity production costs of 0.073 h (0.078 US$2010)/kWh at 20 MWel and to 0.146 h (0.157 US$2010)/kWh at 1 MWel. In their model it is assumed that char and oﬀ-gas produced during the pyrolysis are burned to cover the internal heat demand of the installation. As calculated already above, the energy value of the char can be as high as 560 US$2010/tonne. Lin and Hwang37 assessed the proﬁtability of charcoal production from discarded Cryptomeria branches and wood tops using a still-operational earthen kiln in Taiwan. This analysis was based on empirical data combined with market research. Charcoal production cost of 3707 US$ (3747 US$2010)/tonne can be derived from the analysis. If the revenues of selling wood vinegar are subtracted from the sum of production cost, the char could be sold at a whole-sale price of 1840 US$ (1860 US$2010) per tonne. Considering charcoal prices in Taiwan amounted to 3030 US$ (3063 US$2010)/tonne at the time of the analysis according to the Lin and Hwang,37 the production process was judged to be economically feasible in their publication. However, the charcoal sales price assumed in this calculation—it is not stated in the mentioned publication if the indicated sales price is a retail price or an end customer price—is extremely high and exceeds even the end customer price for retort barbecue charcoal in Germany in 2011 (2700 US$2010). Thus, it is doubtful whether the 9476

dx.doi.org/10.1021/es201792c |Environ. Sci. Technol. 2011, 45, 9473–9483

Environmental Science & Technology assumed sales price can be realized for the total production volume indicated in the calculation (13 000 kg charcoal/year). Whereas charcoal production might thus not be proﬁtable in the described example, charcoal production for energy applications— both in simple charcoal kilns and in modern retort systems—is clearly proﬁtable for many charcoal producers around the globe.38 Thus, Lin and Hwang37 might overestimate charcoal production cost in earthen-kilns. For comparison, Norgate and Langberg39 calculated charcoal production cost of 373 US$ (386 US$2010)/ tonne based on a continuous charcoal retort. In a very detailed assessment, Roberts et al.40 calculated the economic viability of a modeled continuous drum kiln “slow” pyrolysis plant with a throughput of 10 tonnes dry feedstock mass per hour at a temperature of 450 °C and a drum residence time of several minutes in the United States. Aim of the process is to produce biochar for soil management and synthesis gas for heat provision. Taking into account revenues from selling the biochar (its value is calculated on basis of the potassium and phosphate content of the biochar and an improved nitrogen fertilizer use eﬃciency caused by the application of the char), a tipping fee for the disposal of yard waste, the sales of heat produced from the synthesis gas, avoided composting costs, and the sales of GHG oﬀset certiﬁcates (at a price of 20 US$ per tonne CO2), biochar produced on the basis of yard waste yields a positive return of 16 US$ per tonne dry feedstock. This calculation takes into account the opportunity costs of switching from yard waste based compost production to biochar production. Biochar produced on the basis of corn stover and switch grass cannot be produced in a proﬁtable way under the same assumptions. For these types of feedstock, negative returns of 17 US$ to 30 US$ per tonne dry feedstock have been calculated. It should be noted here that the costs for the transportation of feedstock from dispersed locations to the pyrolysis plant can play a major role in the overall production costs of biochar. In this respect, the opportunity of sourcing biochar feedstock from a single location—e.g., from a composting collection station as assumed in the calculation cited above—can be a clear cost advantage. Under these framework conditions, the direct production cost (not taking into account the opportunity costs for not producing and selling compost) amount to 50 US$ per tonne dry feedstock (or 172 US$ per tonne biochar at 29% wt biochar yield), while the total direct revenues (not taking into account avoided costs for composting) amount to 112 US$ per tonne dry feedstock (or 368 US$ per tonne biochar). Of the direct revenues, 35 US$ per tonne dry feedstock (121 US$ per tonne biochar) are gained by sale of heat produced from the synthesis gas of the pyrolysis process. Only 11 US$ per tonne dry feedstock (38 US$ per tonne biochar) are associated with the agricultural value of biochar. Without the revenues from the sales of GHG oﬀset certiﬁcates, the pyrolysis process would be just at the edge of proﬁtability. Two aspects regarding the proﬁtability analysis of the yard waste analysis need further examination: Is it possible to create GHG oﬀset certiﬁcates from a biochar project? Since this is not possible on the market for legally binding GHG emission reductions, only the voluntary market would oﬀer chances to sell GHG oﬀset certiﬁcates from biochar projects. In addition to that, potential yield increases associated with the application of biochar—yet not connected to its potassium and phosphorus content alone— have not been included yet in the value of biochar and might improve the proﬁtability of the analyzed system substantially. Brown et al.38 reported total annual operating costs including ﬁxed costs, capital depreciation, and coproduct credits of about

CRITICAL REVIEW

71 million US$2010/year for a slow pyrolysis facility with 262.000 tonnes of biochar production per year. From these ﬁgures, the production costs for one tonne biochar can be calculated at 272 US$2010. For a fast pyrolysis facility with an output of 172 million liters of biogenic gasoline from biooil, total annual operating costs of 67,500,000 US$2010 and biogenic gasoline production costs of 0.39 US$2010/liter can be derived from the study. If the indicated production cost could be realized, this facility would be highly proﬁtable. The paper assumes that only 26% of the coproduced char is used within the fast pyrolysis process to provide heat, the remaining char is sold. 5.2. Gasification. Yin et al.30 indicated a payback period of less than two years for a circulating fluidized bed biomass gasification and power generation system installed in 1998 in the Fujian Province of China. As already stated before, a considerable part of the char produced in the gasifier is removed from the product gas and returned to the gasifier. For this system with a throughput of 1500 kg rice husk/hour providing a power output of about 800 kW, total investment costs of 510,000 US$ (625,000 US$2010) are necessary. However, the amount of data provided in this article to underline the claimed payback period of two years is very scarce. In addition to that, the indicated investment costs (about 640 US$ (784 US$2010)/kW) are very low. The energy value of char from gasification within the process is calculated in the last part of this subchapter. Bridgwater et al.18 calculated electricity production costs of 0.1 h (0.11 US$2010)/kWh at 20 MWel and 0.22 h (0.24 US$2010)/ kWh at 1 MWel installed capacity for an electricity production process based on a dual fuel diesel engine fed by the atmospheric gasiﬁcation of wood chips and diesel as auxiliary fuel. They assumed electricity production costs of about 0.09 h (0.097 US $2010)/kWh at 20 MWel and of about 0.26 h (0.28 US$2010)/ kWh at 1 MWel for an electricity production process based on an integrated gas turbine combined cycle fed by the pressurized gasiﬁcation of wood chips. In this model, it is assumed that the ash produced from the atmospheric gasiﬁcation process contains 33% wt char, resulting in an overall carbon conversion eﬃciency of the system of 99.5%. Thus, nearly no char is left after the biomass conversion process. Peer-reviewed proﬁtability analyses of gasiﬁcation systems aiming at the sale of the char produced in the process are not available in the ISI to the knowledge of the authors. This can be partly explained by the development focus of this technology which is clearly set on the provision of energy, and by the technical challenges still connected with the biomass gasiﬁcation technology. It is important to be aware that the operators of gasiﬁers will only sell the coproduced char of the gasiﬁers if a char price is paid which at least covers the cost for an alternative heating fuel for the gasiﬁcation process. Jorapur and Anil41 indicate a higher heating value of 18.9 MJ for gasiﬁer char. This would correspond to an energy value of about 380 US$2010/ tonne gasiﬁer char if wood pellets would be used to replace the char used as fuel in the gasiﬁer. 5.3. Flash Carbonization. Antal et al.42 stated that the actual capital investment incurred in the fabrication and setup of a flash carbonization demonstration reactor (1.73 m D 2.74 m H) at the University of Hawaii were 270,000 US$ (290,000 US$2010). To these costs, US$ 30,000 (32,300 US$2010) have to be added in case a 1.27 MPa (12.7 bar) air compressor is installed to provide pressure to the reactor. In case a 2.17 MPa (21.7 bar) air compressor would be used, about US$ 120,000 (129,000 US $2010) has to be added to the costs of the reactor. According to 9477

dx.doi.org/10.1021/es201792c |Environ. Sci. Technol. 2011, 45, 9473–9483

Environmental Science & Technology

CRITICAL REVIEW

Table 5. Production Costs for Char and Energy Carriers As Indicated in Reviewed Papers process type

production costs

comment

reference

ﬂuidized bed fast pyrolysis

0.23 0.5 US$2010 per kg pyrolysis oil

no sales of coproduced char assumed

Islam and Ani35

ﬂuidized bed fast pyrolysis

0.39 US$2010/L biogenic gasoline

26% of coproduced char used within the

Brown et al.38

electricity production based

0.098 US$2010/kWh el at 20 MWel ;

from pyrolysis oil on fast pyrolysis

process char used to cover the internal heat demand

0.215 US$2010/kWh el at 1 MWel

Bridgwater et al.18

of the process

fast pyrolysis

560 US$2010/tonne chara

energy value of fast pyrolysis char in the process

own calculation

slow pyrolysis (earthen kiln)

3747 (1860) US$2010/tonne charcoal

sales of coproduced wood vinegar excluded

Lin and Hwang37

slow pyrolysis (Lambiotte continuous retort)

373 US$ (386 US$2010)/tonne charcoal

production costs based on wood production and wood processing (charring) cost

Norgate and Terry39

slow pyrolysis (drum kiln)

51 US$2010 per tonne char from yard waste

sales of heat from syngas included in the

Roberts et al.40

slow pyrolysis

272 US$2010 per tonne char from corn stover

corn stover feedstock costs: 83 $/tonne

Brown et al.38

investment: 510,000 US$2010; capacity:

Yin et al.30

(included) in the production cost

production costs

circulating ﬂuidized bed gasiﬁcation

1500 kg feed/h electricity production based on atmospheric gasiﬁcation

0.11 US$2010/kWh el at 20 MWel ;

only 0.5% of the carbon in the feedstock is

0.24 US$2010/kWh el at 1 MWel

Bridgwater et al.18

converted to char

electricity production based on pressurized gasiﬁcation

0.097 US$2010/kWh el at 20 MWel ; 0.128 US$2010/kWh el at 1 MWel

no information on char production and char use available

Bridgwater et al.18

gasiﬁcation

380 US$2010/tonne chara

energy value of gasiﬁcation char in the process

own calculation

ﬂash carbonization

no information available

investment: 419,000 US$2010; Capacity:

Antal et al.42

430 kg char or 1300 kg feedstock/h (at 2.17 MPa) a

Char value has been calculated based on the costs for an alternative heating fuel (wood pellets).

the authors, the 2.17 MPa systems have an output of 8.4 tonnes/ day of fixed-carbon, whereas the 1.27 MPa system has an output of 6.1 tonnes/day of fixed-carbon at 24 h of operation. A rough profitability analysis is indicated in the paper with the aim to compare the two systems from an economical point of view. This analysis—which resulted in very short payback periods of 3.7 and 1.3 years for the 1.27 and the 2.17 MPa systems, respectively— however cannot be used to assess the overall profitability of the two systems due to the limited amount of cost data included in the calculation. It is possible that only little economic information on this process has been published by the authors for confidentiality reasons. 5.4. Summary. Most information is available on pyrolysis (especially slow pyrolysis) processes. Though the information provided on the economics of a slow pyrolysis system aimed at the production of heat and biochar is very detailed, it is only partly based on empirical data of an already installed system. In summary, a thorough comparison of the profitability of biochar production technologies based on scientific literature is not possible at the moment. In Table 5, data on the economic viability of the different production processes as indicated in the reviewed papers are summarized.

6. GREENHOUSE GAS (GHG) BALANCE OF BIOCHAR PRODUCTION AND APPLICATION Before discussing the value of biochar technologies for climate change mitigation, it is important to understand the overall context of mitigation strategies. Suﬃciency (lifestyle changes),

eﬃciency, and renewable fuel switch strategies help to avoid the emission of greenhouse gases before they enter the atmosphere. Biochar systems can help to mitigate global warming also after fossil CO2 has already been released to the atmosphere. Whereas biochar systems thus oﬀer the opportunity to act also if other climate change mitigation strategies should fail, it is important to not weaken the necessary eﬀorts in the ﬁeld of suﬃciency, eﬃciency, and renewable fuel strategies. This risk would become very concrete if the sale of GHG certiﬁcates from biochar projects on the carbon compliance market would enable utilities to oﬀset their fossil fuel emissions in “temporary” emission reduction projects, instead of implementing eﬃciency or fuel switch measures needed in an existing cap-and-trade system. In contrast to that, trading GHG certiﬁcates from biochar projects on the voluntary market would not reduce the mitigation pressure in the carbon compliance market. For a discussion of the same mechanism in the context of forestry oﬀset projects, see Streck et al.43 To fully assess the GHG balance of biochar conversion technologies, information on feedstock provision (including direct and indirect land use change eﬀects), conversion process, byproducts use, biochar application, biochar stability in soil, inﬂuence of biochar application on soil related N2O, CH4, and CO2 emissions and on plant growth—including associated impacts on land use—is needed.12,44,45 To compare the greenhouse gas impact of the production and use of biochar to a reference scenario with an alternative use of the feedstock, it is necessary to provide detailed information on this reference scenario. To comprehensively assess the climate-related eﬀects 9478

dx.doi.org/10.1021/es201792c |Environ. Sci. Technol. 2011, 45, 9473–9483

Environmental Science & Technology

CRITICAL REVIEW

Table 6. Climate Impacts of Biochar Production Technologies

9479

dx.doi.org/10.1021/es201792c |Environ. Sci. Technol. 2011, 45, 9473–9483

Environmental Science & Technology

CRITICAL REVIEW

Figure 1. Impact of biochar production on the natural carbon cycle.

of biochar application, insight into the impacts of biochar application on surface albedo46 and on black carbon concentration in the atmosphere47 is needed in addition to the information summarized above. In the following sections it is indicated to which extent these aspects have been included in the reviewed literature articles. Table 6 gives an overview of the analyzed publications which are described in detail in the following sections. From Table 6 it is evident that biochar sequestration in the soil is one of the main factors positively inﬂuencing the GHG balance of the biochar systems. Taking into account the variety in char yields of diﬀerent technologies (see Table 2) and the diﬀerences in the long-term stability of chars made from diﬀerent technologies (see Nguyen and Lehmann;51 Steinbeiss et al.11), the full eﬀect on biochar systems on the natural carbon cycle has to be understood. Whereas the production of char is in most cases slowing down the carbon cycling of the charred fraction of the feedstock (as compared to a reference situation with the natural decay of, e.g., forest residues or agricultural residues)52 the release of CO2 from the biomass conversion process will speed up the carbon cycle of the uncharred feedstock fraction compared to the reference situation (Figure 1). For a detailed discussion of this eﬀect in the context of bioenergy and biochar use, see Repo et al.53 and the supplementary information to the publication of Woolf et al.39 6.1. Pyrolysis Technologies. Woolf et al.39 calculated the maximum sustainable technical global potential for the contribution of biochar systems to climate change mitigation. In this very comprehensive paper, all GHG related impacts of pyrolysis biochar production and application have been assessed. The potential impacts of biochar systems on the atmospheric soot concentration—via biomass smoke and via black carbon dust becoming airborne—and on the surface albedo are mentioned in

the Supporting Information to the paper, but have not been examined in detail. In the alpha scenario of the publication, 66 Gigatonnes CO2 C equivalent net avoided GHG emissions have been calculated over a time period of 100 years for the production and application of biochar derived from about 1.01 Gigatonnes biogenic carbon per year. Assuming an average carbon content of 50% for dry biomass, 2.02 Gigatonnes of biomass feedstock are used in the scenario for the production of biochar. As yield improvements are assumed to be triggered by biochar application, the increment in biomass production is reinvested into additional production of biochar in this scenario. Assuming a total consumption of 2.3 Gigatonnes of dry biomass feedstock for the biochar production, this scenario results in average net avoided GHG emissions of 1054 kg CO2e/tonne dry feedstock and year. The average avoided GHG emissions indicated above are in line with the results of a recent publication of Hammond et al.:54 in this paper, a carbon abatement of 0.7 1.3 t CO2 equivalent per oven dry tonne of feedstock processed has been calculated. They also ﬁt with the calculations of Roberts et al.40 in those of their scenarios which assume that biochar is produced from unused residues. Roberts et al.40 calculated comprehensive greenhouse gas balances for the production and application of biochar produced from diﬀerent feedstock in a slow pyrolysis process. The authors included most of the climate-relevant factors but did not account for impact of biochar application on soil CH4 and soil CO2 emissions, on the surface albedo and the soot concentration in the atmosphere. The latter two aspects do not impact on the GHG balance of biochar production and application itself, but they inﬂuence the sum of climate relevant eﬀects of biochar application. To calculate the GHG impact of the production and 9480

dx.doi.org/10.1021/es201792c |Environ. Sci. Technol. 2011, 45, 9473–9483

Environmental Science & Technology use of biochar, a reference scenario has to be taken into account to describe changes in emissions when the biochar system is implemented. Under the assumption used by Roberts et al.,40 the following results have been calculated: Choosing a reference scenario in which yard waste is used for composting, 885 kg CO2e per t dry biomass feedstock can be saved when switching from yard waste composting to the production and application of biochar. Choosing a reference scenario in which corn stover is left as residue on the ﬁeld (thus not used to provide bioenergy in the reference scenario), 793 864 kg CO2e per t dry biomass feedstock can be saved when switching to the production and application of biochar. The range of emission reduction depends on the moisture content of the corn stover used for biochar production. However, if corn stover were used to produce electricity (thereby replacing natural gas based electricity generation) in the reference scenario, GHG emissions would increase by 123 kg CO2e per t dry biomass feedstock when switching to use the stover in a biochar system. Switching from a reference scenario with agricultural crop production to the cultivation of switch grass and the subsequent use of this feedstock for biochar production might either reduce GHG emission by 442 kg CO2e per t dry biomass or increase GHG emissions by 32 kg CO2e per t dry biomass, depending on the amount of GHG emissions assumed to be triggered via indirect land use change eﬀects. Indirect land-use change eﬀects are caused when an existing production of agricultural goods is displaced by the cultivation of energy crops on the same plot of land. As an eﬀect of that, other land areas, e.g., primary rain forest might be converted to arable land to compensate for the decrease in the previous production of agricultural goods. In contrast to Roberts et al.,40 Hammond et al.54 do not account for indirect land use change eﬀects when assessing the GHG balance of biochar systems. Thus, the latter calculate substantial carbon abatements for biochar systems even in the case of wood chips from short rotation coppice being used for biochar production. A less comprehensive GHG balance for a slow pyrolysis-based biochar system has been carried out by Gaunt and Lehmann.49 The authors did not account for indirect land-use change impacts triggered by the production of energy crops, the emissions connected to the provision and the conversion of the biomass feedstock, the energy use necessary for biochar application, the impact of biochar application on plant growth and associated land use eﬀects, soil CH4 and soil CO2 emissions, the impacts on the surface albedo, and the soot concentration in the atmosphere. Based on their assumptions, they calculated GHG emission reductions of 10.7 t CO2 ha 1 yr 1 for corn stover. For direct comparison, Roberts et al.40 calculated emissions reductions of 7 t CO2e ha 1 yr 1 in a similar scenario for corn stover if this feedstock would have remained a ﬁeld residue in the reference scenario. If corn stover and switch grass were used for pyrolysis based electricity production in the reference scenario (thereby replacing natural gas-derived electricity), switching to a pyrolysis system optimized for biochar production would reduce GHG emissions by 8.5 t CO2e ha 1 yr 1 in the case of using corn stover as feedstock and by 7.6 t CO2e ha 1 yr1 in the case of using wheat straw as feedstock according to Gaunt and Lehmann.49 Because direct and indirect land-use change eﬀects were not taken into account for using energy crops as feedstock, the calculated emission reductions for energy crop scenarios are not indicated here. Whereas Libra et al.,12 did not calculate a complete GHG assessment of biochar systems, they observed both a reduction in

CRITICAL REVIEW

soil N2O emission and CO2 eﬄux four weeks after mixing sandy loam brown earth mixed with pyrolysis biochar. 6.2. Gasification. Searcy and Flynn50 calculated emission reductions of switching from coal-based electricity production to a straw-based integrated gasification and combined cycle electricity production system at 839 g CO2e per kWh of produced electricity (1680 kg CO2e/tonne dry feedstock). It should be noted that this technology is still in an early development stage. Apart from that, no extraction of biochar out of the integrated gasification and combined cycle was assumed in this study. 6.3. Summary. Peer-reviewed greenhouse gas balances on the production and application of biochar were identified for pyrolysis technologies only. The more recent publications in this field—Woolf et al.39 and Roberts et al.40—are far more comprehensive compared to older papers since they cover more GHG related impacts of the biochar systems. Depending on the types and previous use of the biomass feedstock, both reductions and increases of GHGs have been calculated for biochar systems.39,40,49 GHG reductions are often not achieved if dedicated energy crops are used as feedstock for the production of biochar or if biomass residues are already used for the provision of bioenergy in the reference scenario. From a GHG perspective alone, the most recent studies give a clear indication under which conditions biochar systems can contribute to mitigate climate change. However, as explained above, the calculations do not yet take into account all relevant climate impacts of biochar systems: Insights into the impacts of biochar application on the surface albedo and on the black carbon (soot) concentration in the atmosphere have not been added yet to the GHG balances of the biochar systems in the publications reviewed. In this context, it should be noted that Vaccari et al.53 reported positive soil temperature anomalies up to 2 °C in open field biochar plots during the initial phases of durum wheat production. These aspects should also be taken into account when comparing the full climate benefit of bioenergy vs biochar systems (see also Woolf et al.39). Last but not least—as already indicated—it is important to use biochar-based mitigation options complementary to existing mitigation strategies instead of replacing the latter.

7. OUTLOOK The peer-reviewed papers analyzed provided valuable insights into technical reliability, economic feasibility, and climate impact of diﬀerent carbonization technologies. Most papers focus on pyrolysis technologies, less information is available on gasiﬁcation processes. Publications on gasiﬁcation processes often do not take into account the potential suitability of the char as a product for soil improvement. Very little information about hydrothermal and ﬂash carbonization technologies with relevance for this review has been published yet. Two comprehensive studies on the economic proﬁtability and the GHG balance of a slow pyrolysis process have been identiﬁed which point at economic and environmental chances of biochar systems (Roberts et al.40 Woolf et al.39). Data from pilot projects would be essential to further improve assessments on the technical reliability and economic proﬁtability of biochar production technologies. To complement the assessments on the climate impact of biochar production technologies, additional information on the impact of biochar application on surface albedo, atmospheric soot concentration, and yield responses would be needed. The GHG balance of biochar systems itself has been 9481

dx.doi.org/10.1021/es201792c |Environ. Sci. Technol. 2011, 45, 9473–9483

Environmental Science & Technology already quite well examined in recent papers with promising results regarding climate mitigation opportunities. In most of the studies, only one assessment dimension of a technology is analyzed. Comparisons between diﬀerent technologies—see for example the papers of Bridgewater et al.18 or Brown et al.38—are rare as well. Empirical data on the annual availability of technologies which (co)produce biogenic chars has only been published from gasiﬁcation systems aimed at the production of electricity (in this sector, annual availability rates of 60 70% were measured). A wide range of data on the costs of char production (between 51 $2010 per tonne pyrolysis biochar from yard waste and 386 US$2010 per tonne retort charcoal) and on the GHG balance of biochar systems (between 1054 kg CO2e and +123 kg CO2e per t dry biomass feedstock) can be retrieved from the literature. A comprehensive assessment of the technical, economic, and environmental strength and weakness of biochar production technologies is unfortunately still not possible yet on the basis of the available scientiﬁc peer-reviewed literature. This is at least partly explainable by the fact that the production of biomass-based chars for the improvement of agricultural soils is still a relatively new topic of scientiﬁc interest. Further research of both the public and the private sector on the indicated knowledge gaps and its publication is necessary to support project developers, technology developers, and policy makers with a comprehensive and detailed picture on the diﬀerent options to produce biochar for soil improvement and climate change mitigation.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]; phone: 0049-176-23595765.

’ ACKNOWLEDGMENT S.M. thanks his supervising tutors, his colleagues from Ecofys— special thanks go to Carlo Hamelinck—and three anonymous reviewers for providing valuable comments and suggestions for this work. In addition, S.M. thanks Anne for her interest in carbon stories. Finally, we thank all authors, companies, and institutions who provided the knowledge on biochar production technologies needed to compile this paper. ’ REFERENCES (1) Glaser, B.; Haumaier, L.; Guggenberger, G.; Zech, W. The 'Terra Preta' phenomenon: A model for sustainable agriculture in the humid tropics. Naturwissenschaften 2001, 88 (1), 37–41. (2) Kimetu, J. M.; Lehmann, J.; Ngoze, S. O.; Mugendi, D. N.; Kinyangi, J. M.; Riha, S.; Verchot, L.; Recha, J. W.; Pell, A. N. Reversibility of soil productivity decline with organic matter of diﬀering quality along a degradation gradient. Ecosystems 2008, 11 (5), 726–739. (3) Molina, M.; Zaelke, D.; Sarma, K. M.; Andersen, S. O.; Ramanathan, V.; Kaniaru, D. Reducing abrupt climate change risk using the Montreal Protocol and other regulatory actions to complement cuts in CO2 emissions. Proc. Natl. Acad. Sci., U.S.A. 2009, 106 (49), 20616–20621. (4) Lehmann, J., Joseph, S., Eds. Biochar for Environmental Management. Science and Technology; Earthscan: London, 2009. (5) Trifonova, R.; Postma, J.; Schilder, M. T.; van Elsas, J. D. Microbial Enrichment of a Novel Growing Substrate and its Eﬀect on Plant Growth. Microb. Ecol. 2009, 58 (3), 632–641. (6) Iswaran, V.; Jaukhri, K. S.; Sen, A. Eﬀect of charcoal, coal and peat on the yield of moong, soybean and pea. Soil Biol. Biochem. 1980, 12 (2), 191–192.

CRITICAL REVIEW

(7) Lehmann, J.; Pereira da Silva, J.; Steiner, C.; Nehls, T.; Zech, W.; Glaser, B. Nutrient availability and leaching in an archaeological Anthrosol and a Ferralsol of the Central Amazon basin: Fertilizer, manure and charcoal amendments. Plant Soil 2003, 249, 343–357. (8) Yamato, M.; Okimori, Y.; Wibowo, I. F.; Anshori, S.; Ogawa, M. Eﬀects of the application of charred bark of Acacia mangium on the yield of maize, cowpea and peanut, and soil chemical properties in South Sumatra, Indonesia. Soil Sci. Plant Nutr. 2006, 52 (4), 489–495. (9) Steiner, C.; Teixeira, W.; Lehmann, J.; Nehls, T.; Mac^edo, J.; de; Blum, W.; Zech, W. Long term eﬀects of manure, charcoal and mineral fertilization on crop production and fertility on a highly weathered Central Amazonian upland soil. Plant Soil 2007, 291, 275–290. (10) Rajeev, J.; Anil, K. R. Sugarcane leaf-bagasse gasiﬁers for industrial heating applications. Biomass Bioenergy 1997, 13 (3), 141–146. (11) Steinbeiss, S.; Gleixner, G.; Antonietti, M. Eﬀect of biochar amendment on soil carbon balance and soil microbial activity. Soil Biol. Biochem. 2009, 2009 (41), 1301–1310. (12) Libra, J. A.; Ro, K. S.; Kammann, C.; Funke, A.; Berge, N. D.; Neubauer, Y.; Titirici, M.-M.; Fuhner, C.; Bens, O.; Kern, J.; Emmerich, K.-H. Hydrothermal carbonization of biomass residuals: A comparative review of the chemistry, processes and applications of wet and dry pyrolysis. Biofuels 2011, 2 (1), 71–106. (13) Rillig, M. C.; Wagner, M.; Salem, M.; Antunes, P. M.; George, C.; Ramke, H.-G.; Titirici, M.-M.; Antonietti, M. Material derived from hydrothermal carbonization: Eﬀects on plant growth and arbuscular mycorrhiza. Appl. Soil Ecol. 2010, 45 (3), 238–242. (14) Deenik, J. L.; McClellan, T.; Uehara, G.; Antal, J. M. J.; Campbell, S. Charcoal Volatile Matter Content Inﬂuences Plant Growth and Soil Nitrogen Transformations. Soil Sci. Soc. Am. J. 2010, 74 (4), 1259–1270. (15) Luoga, E. J.; Witkowski, E. T. F.; Balkwill, K. Economics of charcoal production in miombo woodlands of eastern Tanzania: Some hidden costs associated with commercialization of the resources. Ecol. Econ. 2000, 35 (2), 243–257. (16) Bridgwater, A. V. The production of biofuels and renewable chemicals by fast pyrolysis of biomass. Int. J. Global Energy Iss. 2007, 27 (2), 160–203. (17) IEA Bioenergy. IEA Bioenergy Task 34: Pyrolysis. http://www. pyne.co.uk/?_id=76 (accessed May 6, 2010). (18) Bridgwater, A. V.; Toft, A. J.; Brammer, J. G. A techno-economic comparison of power production by biomass fast pyrolysis with gasiﬁcation and combustion. Renewable Sustainable Energy Rev. 2002, 6 (3), 181–248. (19) Funke, A.; Ziegler, F. Hydrothermal carbonization of biomass: A summary and discussion of chemical mechanisms for process engineering. Biofuels Bioprod. Bioreﬁn. 2010, 4 (2), 160–177. (20) Yan, W.; Acharjee, T. C.; Coronella, C. J.; Vasquez, V. R. Thermal Pretreatment of Lignocellulosic Biomass. Environ. Prog. Sustainable Energy 2009, 28 (3), 435–440. (21) Antal, M. J.; Gronli, M. The art, science, and technology of charcoal production. Ind. Eng. Chem. Res. 2003, 42 (8), 1619–1640. (22) DeSisto, W. J.; Hill, I. N.; Beis, S. H.; Mukkamala, S.; Joseph, J.; Baker, C.; Ong, T.-H.; Stemmler, E. A.; Wheeler, M. C.; Frederick, B. G.; van Heiningen, A. Fast Pyrolysis of Pine Sawdust in a Fluidized-Bed Reactor. Energy Fuels 2010, 24, 2642–2651. (23) Repo, A.; Tuomi, M.; Liski, J. Indirect carbon dioxide emissions from producing bioenergy from forest harvest residues. GCB Bioenergy 2011, 3 (2), 107–115. (24) Tsukashi, H. Infrared Spectra of artiﬁcial coal made form submerged wood at Uozu Toyama Prefecture Japan. Bull. Chem. Soc. Japan 1966, 39 (3), 460–&. (25) Antal, M. J.; Mochidzuki, K.; Paredes, L. S. Flash carbonization of biomass. Ind. Eng. Chem. Res. 2003, 42 (16), 3690–3699. (26) Hamper, M. J. Manufactured gas history and processes. Environ. Forensics 2006, 7 (1), 55–64. (27) Jensen, P. A.; Frandsen, F. J.; Dam-Johansen, K.; Sander, B. Experimental investigation of the transformation and release to gas phase of potassium and chlorine during straw pyrolysis. Energy Fuels 2000, 14 (6), 1280–1285. 9482

dx.doi.org/10.1021/es201792c |Environ. Sci. Technol. 2011, 45, 9473–9483

Environmental Science & Technology (28) Buchireddy, P. R.; Bricka, R. M.; Rodriguez, J.; Holmes, W. Biomass Gasiﬁcation: Catalytic Removal of Tars over Zeolites and Nickel Supported Zeolites. Energy Fuels 2010, 24, 2707–2715. (29) Wade, S. R.; Nunoura, T.; Antal, M. J. Studies of the ﬂash carbonization process. 2. Violent ignition behavior of pressurized packed beds of biomass: A factorial study. Ind. Eng. Chem. Res. 2006, 45 (10), 3512–3519. (30) Yin, X. L.; Wu, C. Z.; Zheng, S. P.; Chen, Y. Design and operation of a CFB gasiﬁcation and power generation system for rice husk. Biomass Bioenergy 2002, 23 (3), 181–187. (31) Pr€oll, T.; Rauch, R.; Aichernig, C.; Hofbauer, H. Fluidized Bed Steam Gasiﬁcation of Solid Biomass - Performance Characteristics of an 8 MWth Combined Heat and Power Plant. Int. J. Chem. Reactor Eng. 2007, (Vol. 5, Article A 54). (32) Kreuzeder, A.; Pfeifer, C.; Hofbauer, H. Fluid-Dynamic Investigations in a Scaled Cold Model for a Dual Fluidized Bed Biomass Steam Gasiﬁcation Process: Solid Flux Measurements and Optimization of the Cyclone. Int. J. Chem. Reactor Eng. 2007 (Vol. 5, Article A 31). (33) Nunes, S. M.; Paterson, N.; Herod, A. A.; Dugwell, D. R.; Kandiyoti, R. Tar formation and destruction in a ﬁxed bed reactor simulating downdraft gasiﬁcation: Optimization of conditions. Energy Fuels 2008, 22 (3), 1955–1964. (34) Yassin, L.; Lettieri, P.; Simons, S. J.; Germana, A. Technoeconomic performance of energy-from-waste ﬂuidized bed combustion and gasiﬁcation processes in the UK context. Chem. Eng. J. 2009, 2009 (146), 315–327. (35) Islam, M. N.; Ani, F. N. Techno-economics of rice husk pyrolysis, conversion with catalytic treatment to produce liquid fuel. Bioresour. Technol. 2000, 73 (1), 67–75. (36) Badger, P.; Badger, S.; Puettmann, M.; Steele, P.; Cooper, J. Techno-Economic Analysis: Preliminary assessment of pyrolysis oil production costs and material energy balance associated with a transportable fast pyrolysis system. Bioresources 2011, 6 (1), 34–47. (37) Lin, Y.-J.; Hwang, G.-S. Charcoal from biomass residues of a Cryptomeria plantation and analysis of its carbon ﬁxation beneﬁt in Taiwan. Biomass Bioenergy 2009, 33 (9), 1289–1294. (38) Brown, T. R.; Wright, M. M.; Brown, R. C. Estimating proﬁtability of two biochar production scenarios: slow pyrolysis vs fast pyrolysis. Biofuels, Bioprod. Bioreﬁn. 2011, 5 (1), 54–68. (39) Woolf, D.; Amonette, J. E.; Street-Perrott, F. A.; Lehmann, J.; Joseph, S. Sustainable biochar to mitigate global climate change. Nat. Commun. 2010, 1, 56. (40) Roberts, K. G.; Gloy, B. A.; Joseph, S.; Scott, N. R.; Lehmann, J. Life Cycle Assessment of Biochar Systems: Estimating the Energetic, Economic, and Climate Change Potential. Environ. Sci. Technol. 2010, 44 (2), 827–833. (41) Norgate, T.; Langberg, D. Environmental and Economic Aspects of Charcoal Use in Steelmaking. ISIJ Int. 2009, 49 (4), 587–595. (42) Antal, J. M. J.; Wade, S. R.; Nunoura, T. Biocarbon production from Hungarian sunﬂower shells. J. Anal. Appl. Pyrol. 2007, 79 (1 2), 86–90. (43) Streck, C.; Tuerk, A.; Schlamadinger, B. Forestry oﬀsets in emissions trading systems: A link between systems? Mitigat. Adapt. Strategies Global Change 2009, 14, 455–463. (44) Panichelli, L.; Gnansounou, E. Estimating greenhouse gas emissions from indirect land-use change in biofuels production: concepts and exploratory analysis for soybean-based biodiesel. J. Sci. Ind. Res. 2008, 67 (11), 1017–1030. (45) Gelfand, I.; Zenone, T.; Jasrotia, P.; Chen, J.; Hamilton, S. K.; Robertson, G. P. Carbon debt of Conservation Reserve Program (CRP) grasslands converted to bioenergy production. Proc. Natl. Acad. Sci. U.S. A. 2011, 108 (33), 13864–13869. (46) Verheijen, F.; Jeﬀery, S.; Bastos, A.; van der Velde, M.; Diafas, I. Biochar Application to Soils - A Critical Scientiﬁc Review of Eﬀects on Soil Properties, Processes and Functions; Oﬃce for the Oﬃcial Publications of the European Communities: Luxembourg, 2010. (47) Ramanathan, V.; Chung, C.; Kim, D.; Bettge, T.; Buja, L.; Kiehl, J. T.; Washington, W. M.; Fu, Q.; Sikka; Wild, M. Atmospheric brown

CRITICAL REVIEW

clouds: Impacts on South Asian climate and hydrological cycle. Proc. Natl. Acad. Sci. U.S.A. 2005, 102 (15), 5326. (48) McKay, G. Dioxin characterisation, formation and minimisation during municipal solid waste (MSW) incineration: Review. Chem. Eng. J. 2002, 86 (3), 343–368. (49) Gaunt, J. L.; Lehmann, J. Energy balance and emissions associated with biochar sequestration and pyrolysis bioenergy production. Environ. Sci. Technol. 2008, 42 (11), 4152–4158. (50) Searcy, E.; Flynn, P. C. A criterion for selecting renewable energy processes. Biomass Bioenergy 2010, 34 (5), 798–804. (51) Nguyen, B. T.; Lehmann, J. Black carbon decomposition under varying water regimes. Org. Geochem. 2009, 40 (8), 846–853. (52) Kuzyakov, Y.; Subbotina, I.; Chen, H.; Bogomolova, I.; Xu, X. Black carbon decomposition and incorporation into soil microbial biomass estimated by 14C labeling. Soil Biol. Biochem. 2009, 41 (2), 210–219. (53) Vaccari, F.P.; Lugato, E.; Genesio, L.; Castaldi, S.; Fornasier, F.; Miglietta, F. Biochar as a strategy to sequester carbon and increase yield in durum wheat. Eur. J. Agron. 2011, 34 (4), 231–238. (54) Hammond, J.; Shackley, S.; Sohi, S.; Brownsort, P. Prospective life cycle carbon abatement for pyrolysis biochar systems in the UK. Energy Policy 2011, 39 (5), 2646–2655.

9483

dx.doi.org/10.1021/es201792c |Environ. Sci. Technol. 2011, 45, 9473–9483

CRITICAL REVIEW pubs.acs.org/est

Establishing Policy Relevant Background (PRB) Ozone Concentrations in the United States Elena C. McDonald-Buller,† David T Allen,†,* Nancy Brown,‡ Daniel J. Jacob,§ Daniel Jaﬀe,|| Charles E. Kolb,^ Allen S. Lefohn,# Samuel Oltmans,3 David D. Parrish,3 Greg Yarwood,O and Lin Zhang§ †

University of Texas at Austin, Austin Texas Lawrence Berkeley National Laboratory, Berkeley, California § Harvard University, Cambridge, Massachusetts University of Washington, Bothell, Washington ^ Aerodyne Research, Inc., Billerica, Massachusetts # ASL & Associates, Helena, Montana 3 National Oceanic and Atmospheric Administration, Boulder, Colorado O ENVIRON International, Novato, California

)

‡

ABSTRACT: Policy Relevant Background (PRB) ozone concentrations are deﬁned by the United States (U.S.) Environmental Protection Agency (EPA) as those concentrations that would occur in the U.S. in the absence of anthropogenic emissions in continental North America (i.e., the U.S, Canada, and Mexico). Estimates of PRB ozone have had an important role historically in the EPA’s human health and welfare risk analyses used in establishing National Ambient Air Quality Standards (NAAQS). The margin of safety for the protection of public health in the ozone rulemaking process has been established from human health risks calculated based on PRB ozone estimates. Sensitivity analyses conducted by the EPA have illustrated that changing estimates of PRB ozone concentrations have a progressively greater impact on estimates of mortality risk as more stringent standards are considered. As deﬁned by the EPA, PRB ozone is a model construct, but it is informed by measurements at relatively remote monitoring sites (RRMS). This review examines the current understanding of PRB ozone, based on both model predictions and measurements at RRMS, and provides recommendations for improving the deﬁnition and determination of PRB ozone.

’ INTRODUCTION Among the most ubiquitous air quality problems that aﬀect the U.S. are enhanced concentrations of ground-level ozone, which is a secondary pollutant formed by the photochemical reactions of its precursors that include oxides of nitrogen (NOx) and volatile organic compounds (VOC). Exposure to ozone has been associated with adverse human health eﬀects, including decreased lung function, exacerbation of asthma and respiratory conditions, premature mortality, natural and agricultural ecosystem injury and loss, and deterioration of the built (i.e., materials) environment.1 8 Primary and secondary National Ambient Air Quality Standards (NAAQS) for ozone have been established to protect public health and public welfare, respectively. Review of the criteria and standards, by the U.S. Environmental Protection Agency (EPA), is required at ﬁve-year intervals by Section 109(d)(1) of the Clean Air Act. As a result of this process, the primary and secondary NAAQS for ground-level ozone have become increasingly stringent over the past several decades, r 2011 American Chemical Society

with signiﬁcant changes in averaging time, level, and form. In March 2008, the EPA established primary and secondary NAAQS of 0.075 ppm for ozone concentrations averaged over 8 h.9 During a reconsideration of these standards, in January 2010 the EPA proposed to strengthen the primary NAAQS to an 8 h averaged ozone concentration in the range of 0.060 and 0.070 ppm and to establish a new cumulative, seasonal secondary standard in the range of 7 15 ppm-hours.10 In September 2011, the EPA’s draft ozone NAAQS, which it had submitted for review to the President’s Oﬃce of Management and Budget on July 11, 2011, were withdrawn.11 Future rulemaking is to be based on EPA’s12 new periodic review of the air quality criteria and standards for ozone to be completed in 2013. The components of the periodic review consist of an Integrative Received: July 3, 2011 Accepted: October 10, 2011 Revised: October 3, 2011 Published: October 10, 2011 9484

dx.doi.org/10.1021/es2022818 | Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology Review Plan (IRP; an outline of the review process and key science-policy questions), an Integrative Science Assessment (ISA; an evaluation and integration of the policy-relevant science), the Risk and Exposure Assessments (REAs; quantitative estimates of health and welfare exposures and risks associated with current ambient levels and current and alternative regulatory air quality standards), and Policy Assessment (PA; a presentation of the scientiﬁc basis for the policy options for consideration by the EPA Administrator prior to establishing proposed and ﬁnal rules).12 The determination of background ozone concentrations has important implications for the ozone rulemaking process. Background ozone concentrations used to inform decisions about setting the primary and secondary NAAQS are referred to as Policy Relevant Background (PRB) ozone concentrations.8 For the review completed in March 2008, the EPA8,13,14 deﬁned PRB ozone concentrations as those that would occur in the United States in the absence of anthropogenic emissions in continental North America (i.e., the United States (U.S.), Canada, and Mexico). In this context, PRB concentrations represent levels that are not controllable by regulations either in the U.S. or through agreements with the neighboring North American countries.8,13,14 Sources that contribute to PRB ozone concentrations include natural sources globally and anthropogenic sources from outside of North America. Processes that contribute to PRB ozone concentrations include photochemistry associated with natural emissions of volatile organic compounds (VOCs), nitrogen oxides (NOx), and carbon monoxide (CO) from sources such as biogenic emissions (not including agricultural activities), wildﬁres, lightning, the long-range transport of ozone and its precursors from outside of North America, and stratospheric-tropospheric exchange (STE) of ozone.8,13,14 The approach the EPA has used to establish PRB ozone concentrations has changed over time. For the review completed in 1997, the EPA15,16 estimated an annual average ozone background concentration near sea level in the U.S. to be in the range of 0.020 0.035 ppm, which included a stratospheric contribution of 0.005 0.015 ppm and a 0.01 ppm contribution from the photochemical oxidation of methane and carbon monoxide. In addition, the natural ozone background concentration for a 1 h daily maximum at sea level during the summer in the U.S. was estimated to be 0.03 0.05 ppm.15 These estimates were based on observations at sites in the continental U.S. with low maximum hourly average ozone concentrations that appeared to be relatively isolated from anthropogenic sources.15,16 The EPA17 adopted a constant value of 0.04 ppm for PRB ozone. Relatively remote monitoring sites (RRMS) are sites that are not strongly inﬂuenced by but not necessarily free from the eﬀects of nearby pollution sources.8 The estimate of PRB ozone concentrations at remote locations, such as Trinidad Head (CA), Mt. Bachelor Observatory (OR), Gothic (CO), and Yellowstone National Park (WY), have provided important insights regarding the relative importance of processes that contribute to PRB ozone concentrations.18,19 However, for the review completed in March 2008,8,13,14,17 the EPA determined that PRB ozone concentrations could not be derived solely from measurements of ozone at RRMS because of long-range transport from anthropogenic source regions within North America. Instead, estimates of PRB ozone concentrations were established based on predictions of the global chemical transport model, GEOS-Chem, described in Fiore et al.20 for the 2001 April to September season. At the time, GEOS-Chem was

CRITICAL REVIEW

the only model documented in the literature for estimating PRB ozone. In addition to a standard simulation including anthropogenic and natural sources of emissions, Fiore et al.20 conducted a simulation in which North American anthropogenic emissions were set to zero (North American background or PRB) and a simulation in which global anthropogenic emissions were set to zero and methane was set to its 700 ppbv preindustrial value (natural background). Diﬀerences between the PRB and natural ozone simulations reﬂected the contributions of intercontinental pollution inﬂuences and anthropogenic methane. GEOS-Chem predictions indicated that PRB ozone concentrations varied with season, altitude, and meteorological conditions. The EPA determined that 4 h average PRB ozone concentrations at the surface were generally in the range of 0.015 0.035 (0.025 ( 0.010) ppm in the afternoon, but were actually lower under meteorological conditions conducive to high ozone episodes (i.e., concentrations >0.084 ppm). PRB ozone concentrations were estimated to be highest during spring, due to contributions from hemispheric pollution and stratospheric intrusions, and were predicted to decline into summer.8,13,14,17 The stratospheric contribution to surface ozone was estimated to typically be below 0.020 ppm and was more frequently enhanced at high altitude than low altitude sites.8,13,14,17 In the existing regulatory framework for review of the NAAQS described above,12 PRB ozone estimates are established in the ISA and have a crucial role in the REAs. Risks are only estimated for ambient ozone concentrations that exceed PRB levels, which EPA17 considers most relevant for policy decisions. For the review completed in March 2008, the EPA used estimates of PRB ozone concentrations for 12 urban areas (Atlanta, Boston, Chicago, Cleveland, Washington DC, Detroit, Houston, Los Angeles, New York, Philadelphia, Sacramento, and St. Louis) to calculate human health risks.17 Risk assessments reﬂected two diﬀerent types of human studies (i.e., controlled human exposure and epidemiological).17 For its controlled human exposure risk analyses, the EPA calculated the following to estimate risk associated with hourly ozone concentrations in excess of PRB: (1) the expected risk given the personal exposures associated with ambient ozone concentrations; (2) the expected risk given the personal exposures associated with estimated PRB ambient ozone concentrations; and (3) subtracted the latter from the former. For its epidemiological risk analyses, the EPA used diﬀerent exposure metrics, including the 24-h average and the daily 1 h and 8 h maximum ozone concentrations. As an example, for the concentration response function relating daily mortality to daily 1-h maximum ozone concentrations, the daily changes in 1 h maximum ozone concentrations were calculated. For the epidemiology-based risk assessment associated with levels of ozone above PRB levels, the following steps were implemented: (1) using monitor-speciﬁc input streams of hourly ozone concentrations for a speciﬁc year, the 1 h maximum ozone concentration for each day was calculated; (2) using the stream of hourly PRB ozone concentrations, the 1 h maximum PRB ozone concentration for each day was calculated; and (3) for each day, the latter was subtracted from the former. Sensitivity analyses performed by the EPA17 relating varying PRB ozone concentrations to risk estimates demonstrated its importance in the REA process. Estimates assuming lower PRB ozone levels resulted in increased estimates of nonaccidental mortality incidence per 100 000 that were often 50 100% greater than the base case estimates.17 Similarly, estimates 9485

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology assuming higher PRB ozone levels resulted in decreased estimates of nonaccidental mortality incidence per 100 000 that were g50% less than the base case estimates.17 The EPA’s sensitivity analyses illustrated that changing estimates of PRB ozone concentrations have a progressively greater impact on estimates of mortality risk as more stringent standards are considered.17 In March 2011, the EPA21 released the ﬁrst external review draft of the ISA for the ongoing periodic review. The draft ISA includes relevant emerging studies conducted after the release of the 2006 criteria documents,8,13,14 among these are studies by Cooper et al.,22 Zhang et al,23 Oltmans et al.,24 Parrish et al.,25 Langford et al.,26 Kaynak et al.,27 and Wang et al.28 Future drafts of the ISA will incorporate other research as it becomes available. In April 2011, the EPA issued its plans for the health risk and exposure assessment29 for the ongoing periodic review. Among the changes from the review completed in March 2008, the EPA plans to model population exposures to ambient ozone in three or more of the 12 urban areas modeled, as well as in a highelevation area such as Denver.29 The GEOS-Chem model with 0.5 0.67 (∼50 50 km2) horizontal resolution over North America as described by Zhang et al.30 is expected to be used to derive PRB ozone estimates with multiple scenarios including a base case or current atmosphere scenario, for which a model performance evaluation will be conducted using surface and satellite measurements, and three additional emissions scenarios isolating the contributions of internationally transported air pollutants to U.S. ozone concentrations.29 The primary objectives of the review presented here are to examine the current understanding of PRB ozone, based on both model predictions and measurements at RRMS and to provide recommendations for improving the deﬁnition and determination of PRB ozone. The review is based on both information available in the literature and information presented at a workshop held on March 30 April 1, 2011 at The University of Texas at Austin: (http://www.utexas.edu/ research/ceer/prb/). The workshop, entitled Workshop on Policy Relevant Background Ozone Concentrations in the United States, included 10 invited participants (all of the authors of this paper with the exception of Dr. Lin Zhang) with expertise in global and regional chemical-transport models, remote/rural ambient surface and airborne monitoring, satellite data retrievals, and emissions inventory development and analysis. On March 30, 2011, the ﬁrst day of the workshop, the participants made presentations and responded to questions in their areas of expertise. These presentations were open to the public on The University of Texas at Austin campus and were broadcast live over the web, drawing a national audience. The ﬁndings and recommendations were initially formulated by the participants at closed sessions during March 31 and April 1, 2011. They were reﬁned through a series of analyses performed in response to reviewer comments, following the workshop. All of the authors participated in this process. In addition to the authors, the discussions also involved Dr. Joseph Pinto as an active participant from the U.S. EPA. The workshop was sponsored by the American Petroleum Institute, for which two representatives attended as observers only. Our review is organized into a series of ﬁve topics that address the deﬁnition of PRB ozone, spatial and temporal variations across the U.S., the application and performance of models used to determine PRB, the role of measurements, and the potential implications of sources of PRB ozone for attainment demonstrations and the SIP

CRITICAL REVIEW

development process. Each section includes major ﬁndings followed by policy and/or research recommendations.

’ THE CONTEXT OF PRB OZONE: DEFINING BACKGROUND AND BASELINE OZONE CONCENTRATIONS National and international activities have attempted to distinguish between the terms “background” and “baseline” ozone concentrations in the context of understanding the long-range transport of air pollution. The Task Force on Hemispheric Transport of Air Pollution (TF HTAP), created in December 2004 by the Convention on Long-range Transboundary Air Pollution (LRTAP Convention), acknowledged that the terms global or hemispheric background concentrations and baseline concentrations are often used interchangeably.31 TF HTAP31 deﬁned baseline concentrations of a pollutant as “...an observation made at a site when it is not inﬂuenced by recent, locally emitted or produced pollution.”, and global or hemispheric background concentrations as “...a model construct that estimates the atmospheric concentration of a pollutant due to natural sources only.” In the context of baseline concentrations, the Task Force31 stated that neither a strict deﬁnition of recently produced local sources of anthropogenic pollution, nor a means to eliminate traces of local pollution emitted many days earlier and well-mixed with other air masses, exist. For ozone and other pollutants that have longer lifetimes, models are required to determine global or hemispheric background concentrations at sites where naturally occurring concentrations are well mixed with anthropogenic sources.31 The TF HTAP further recognized the terms “urban background” and “rural or regional background” in the literature, which are based on observations. Urban background concentrations were described as those “...observed in urban areas away from the direct inﬂuence of heavily-traﬃcked roads and chimney stacks.” Rural or regional background concentrations were described as “...those observed at locations where there is little inﬂuence from urban sources of pollution.” 31 The Committee on the Signiﬁcance of International Transport of Air Pollutants of the National Research Council32 acknowledged EPA’s deﬁnition of PRB ozone as a model construct, but noted that it was unclear whether the PRB included tropospheric ozone from North American sources that have already traveled around the globe. The Committee32 described the term “background O3” as ambiguous, hypothetical, and not directly measurable and the term would always have to be qualiﬁed. Baseline ozone was used by the Committee32 to “describe a measurable quantity, the statistically deﬁned lowest abundances of ozone in the air ﬂowing into a country, which is typical of clean-air, remote marine sites at the same latitude.” The Committee32 recognized that baseline ozone varies with location and season in the Northern Hemisphere and can change over time. In addition, regional and continental emissions of ozone precursors “...will also contribute to an increasingly diﬀuse background that is indistinguishable from baseline O3 in most urban airsheds and beyond local control.” 32 In an overview of the role of background ozone on air quality issues, Reid et al.33 noted that “diﬀerent interpretations in the understanding of what constitutes background widen the uncertainty in reported values”. In their context, the ozone background concentration is the lowest level that can be achieved in a jurisdiction. It has implications for both local air quality policy and the contribution to health and ecosystem impacts. 9486

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology It is obvious that the deﬁnitions and use of the terms background and baseline ozone concentrations vary widely and can be subject to ambiguities across the air quality scientiﬁc and regulatory communities. While this review cannot remove this ambiguity, for clarity it is necessary to precisely deﬁne the meanings of these terms that will be used in this review. Here, a baseline ozone concentration is deﬁned as an observation made at a site when it is not inﬂuenced by recent nationally emitted or produced pollution. In practice, the determination of the absence of recent pollution is based upon measurements of short-lived tracer species (e.g., radon or NOx) or transport modeling. Since there is no practical method for determining if national emissions have perhaps been circulated globally, well-aged and well-mixed national inﬂuences are necessarily included in baseline concentrations. This deﬁnition is consistent with that of the TF HTAP, but explicitly recognizes that U.S. generated pollution should be excluded on all spatial scales (local, regional, etc.) to the extent possible in the determination of baseline ozone concentrations. This deﬁnition does not deﬁne baseline ozone concentrations as necessarily the lowest ozone concentration measured at a site, nor does it represent baseline ozone concentrations as a single value. This review recognizes PRB ozone concentrations as a model construct in accordance with EPA’s deﬁnition. Diﬀerences between PRB and baseline ozone concentrations can be quantiﬁed by evaluating the models used to construct PRB ozone concentrations against observations at sites used to determine baseline ozone concentration. These diﬀerences will reﬂect North American pollution contributions to transported midlatitude ozone, unrecognized U.S. pollution contributions, and errors in the models. We note that although EPA has established a relatively speciﬁc regulatory deﬁnition of PRB ozone concentrations,8,13,14 ambiguities in this deﬁnition and its application still exist. These are largely associated with the question of whether emissions sources could be subject to control or not. For example, uncertainties surround the inclusion of ﬁres that are natural versus anthropogenic in origin, of agricultural emissions, of ocean-going vessels and air transport operations, and of North American anthropogenic methane emissions. Findings. Inconsistencies exist in the use of the terms “background” and “baseline” ozone concentrations. The EPA has established a relatively speciﬁc regulatory deﬁnition of PRB ozone, but ambiguities in the deﬁnition and its application still exist. Policy Recommendations. The definitions and applications of the terms “background” and “baseline” ozone concentrations in the scientific, risk and exposure, and policy assessment air quality communities and among air quality managers should be harmonized. Further clarity is required in the regulatory deﬁnition of PRB ozone recognizing the contributions and uncertainties surrounding biomass burning, biogenic emissions related to agriculture, ocean-going vessels and air transport operations, and North American anthropogenic methane emissions.

’ SPATIAL AND TEMPORAL VARIATIONS IN PRB OZONE Global and regional chemical transport models20,30,34 as well as measurements of baseline ozone concentrations35 at RRMS have demonstrated that PRB ozone concentrations vary spatially and temporally across the U.S. Temporal and spatial variations

CRITICAL REVIEW

Figure 1. (a) CASTNet ozone monitoring sites in the continental United States in 2006 from Zhang et al.30 Sites in the intermountain west are indicated in red. Pluses denote sites above 1.5 km altitude. (b) Frequency distributions of maximum daily 8-h averaged (MDA8) ozone concentrations in March August 2006 for the ensemble of lowaltitude (<1.5 km) and high-altitude CASTNet sites in the U.S. from Zhang et al.30 Model results (red) are compared to observations (black). Also shown are frequency distributions for the North American (PRB) background (solid blue) and natural background (dashed green).

can arise from regional diﬀerences in PRB ozone sources, such as ﬁres, long-range transport, synoptic-scale meteorological variability, and STE, or from diﬀerences in ozone removal processes including deposition and chemical destruction. Figure 1 shows the frequency distributions of maximum daily 8 h averaged (MDA8) ozone concentrations observed at the ensemble of Clean Air Status and Trends Network (CASTNet) low altitude (<1.5 km) and elevated (>1.5 km) sites during 2006. These observations are compared with GEOS-Chem predictions, sampled at the appropriate pressure level, by Zhang et al.,30 and natural and North American (i.e., PRB ozone) background ozone concentrations. GEOS-Chem predictions of the natural background were determined by zeroing global anthropogenic NOx, nonmethane volatile organic compounds (NMVOC), and CO emissions and setting methane concentrations to a preindustrial value of 700 ppbv. North American (PRB) ozone background concentrations were determined by 9487

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology

CRITICAL REVIEW

Figure 2. GEOS-Chem mean PRB MDA8 ozone concentrations (ppb) for the (a) spring (March/April/May) and (b) summer (June/July/August) of 2006, and (c) annual 4th highest GEOS-Chem PRB MDA8 ozone concentrations for 2006 2008 from Zhang et al.30

zeroing anthropogenic NOx, NMVOC, and CO emissions in continental North America. Figure 1 indicates that the variation in PRB ozone concentrations is in part attributable to their tendency to increase with altitude. All elevated sites are located in the western U.S.; PRB ozone is generally higher in the mountainous western U.S. than in the eastern U.S. Mean PRB values for the U.S. in spring-summer (6-month average) are 27 ( 8 ppbv at low altitude sites (<1.5 km) and 40 ( 7 ppbv at high-altitude sites. These values were 9 13 ppbv higher than the natural background due to intercontinental pollution including anthropogenic methane and were on average 4 ppbv higher than those reported in earlier studies with GEOS-Chem by Fiore et al.20 and Wang et al.;28 the diﬀerences were attributed to a combination of increasing Asian emissions, higher model lightning, and higher model resolution. In another recent study using the Community Multiscale Air Quality (CMAQ) regional model, Mueller and Mallard.34 examined the relative roles of natural emissions (including biogenic, oceanic, geogenic, and ﬁres) and background sources (i.e., model boundary conditions established from the GEOS-Chem simulations of Fiore et al.36) on ozone concentrations at CASTNet sites for the year 2002. MDA8 ozone concentrations due to boundary and natural sources were higher in the western than the eastern U.S.34

Time scales of systematic temporal variation in PRB ozone range from hourly (i.e., diurnal proﬁles), to seasonal to interannual.30 Figure 2 shows seasonal mean GEOS-Chem predictions of PRB ozone concentrations during 2006 and annual GEOS-Chem predictions of the fourth highest PRB ozone concentrations during 2006 2008 from Zhang et al.30 Figure 2 demonstrates the strong geographic and seasonal variations that exist in PRB ozone concentrations across the U.S. In some regions, PRB concentrations could approach 60 70 ppb, a range previously under consideration for revisions to the ozone NAAQS. Finding. Strong spatial and temporal variability exists in PRB ozone across the U.S. Policy Recommendation. The EPA should consider the spatial and temporal variability of PRB ozone in the periodic review of the air quality criteria and standards for ozone, in particular recognizing differences that may exist between populated high- and low- elevation areas and between years and seasons. Research Recommendation. The scientific communities associated with global and regional modeling and measurements at relatively remote sites that are used to inform estimates of PRB ozone should continue to establish and assess the distributions of 9488

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology

CRITICAL REVIEW

PRB ozone concentrations in the U.S. To the extent possible, coordination between the scientific and policy communities should occur in the definition and use of relevant terminology, in the study of key processes that affect PRB ozone estimates, and in the compilation of measurement and modeling data reflecting the state of the science. Coordination should occur within the time frame needed to support the periodic review process.

’ MODELING OF PRB OZONE The lifetime of ozone in the troposphere can be as long as months,37 comparable to the time scale for hemispheric mixing. The relatively long lifetime of ozone and the EPA’s current regulatory deﬁnition of PRB ozone necessitate a reliance on global-scale chemical transport models. The EPA’s earlier and ongoing periodic reviews of the air quality criteria and standards for ozone have relied solely on successive generations of the GEOS-Chem model.20,30,38 A number of other mature global models for tropospheric ozone, with diﬀerent heritages, have been developed by the scientiﬁc community. The full diversity of these models has not been applied to estimate the distributions of PRB ozone either for the regulatory process or in the peer-reviewed literature. Model intercomparisons including evaluations with ozone observations have been reported by Reidmiller et al.,38 the Intergovernmental Panel on Climate Change,39 Atmospheric Composition Change the European Network (ACCENT),40 and the United Nations Economic Commission for Europe TF HTAP.41 These intercomparisons have revealed large diﬀerences between simulations of ozone for speciﬁc sites and regions, even for models with comparable skills in their overall abilities to reproduce observations. Although community intercomparisons of global ozone models oﬀer insights on model diﬀerences in global budgets and individual processes, these cannot be readily related to PRB estimates because of nonlinear chemistry and because of the detail needed for PRB values in policy applications. Application of multiple global models and model intercomparisons of PRB ozone estimates, including with observations at RRMS and aloft, are needed to provide a better appreciation of model uncertainty and understanding of the processes contributing to PRB ozone. The recent study by Mueller and Mallard34 indicates the potential for a regional model such as CMAQ to also be used in the study of PRB, although a global model is still required to provide boundary conditions. The application of regional models and the nesting of ﬁner-scale regional models within global models for estimating PRB ozone and for understanding how processes, such as ﬁres, drought, and changes in emissions inventories, aﬀect estimates of PRB ozone, should be explored more thoroughly in the future. State-of-science global models of tropospheric ozone, such as GEOS-Chem, can reproduce monthly mean MDA8 ozone concentrations at RRMS typically within 5 ppb, and also provide a good simulation of synoptic-scale variability in MDA8 ozone.28 In an evaluation of 15 global models and one hemispheric chemical transport model for the HTAP project, Reidmiller et al.38found that although wide variation existed between simulated maximum and minimum ozone concentrations, the ensemble mean represented observations at U.S. CASTNet sites in most regions and seasons well, with mean annual biases typically less than 5 ppbv. At RRMS, PRB ozone often accounts for a large fraction of total ozone and ozone variability, both in the model and in observational

Figure 3. Comparison of GEOS-Chem predictions and observed MDA8 ozone concentrations at elevated (>1.5 km) CASTNet sites in the intermountain western U.S. during March-May 2006. Model results are from Zhang et al.30 The vertical bars are model statistics of U.S. background ozone in red (obtained by zeroing U.S. anthropogenic emissions) and North American background (PRB) ozone in blue, as a function of observed ozone in 10 ppb bins. The statistics show minima, 25th percentile, medians, 75th percentiles, and maxima.

analyses that attempt to separate baseline from pollution inﬂuences. This provides some conﬁdence that the models can provide reasonable estimates of monthly mean MDA8 PRB ozone ((10 ppb). However, there are regions in the U.S. where global models show consistent biases that could be relevant to PRB ozone estimates. For example, models are generally unable to simulate the very low ozone concentrations observed at Gulf Coast sites in summer during onshore ﬂow from the Gulf of Mexico.20,30,36,38,41 which could reﬂect marine boundary layer chemistry or stratiﬁcation that is not properly represented. Models generally ﬁnd little ozone production in wildﬁre plumes for short aging times (days) because NOx emissions are low and conversion to peroxyacetylnitrate (PAN) is rapid.42 44 In contrast, observations show large ozone production from at least some regional wildﬁres that may signiﬁcantly elevate ozone at low altitude sites on a monthly basis.43,45 47 and persist over long distances from the burned region.48,49 Yet another diﬃculty for models is the complex topography in some regions of the U.S., which may promote external inﬂuences on surface ozone through ﬁne-scale orographic ﬂow and subsidence.50 GEOS-Chem and other global models also have diﬃculty representing the ﬁne structures of ozone events observed at RRMS in the U.S., including events for which the contribution of PRB ozone is likely important. Stratosphere-troposphere exchange can contribute to PRB ozone at low altitude and, in particular, at elevated sites (Figure 3).18,19,26,45,51 56 Fire plumes transported on intercontinental scales can contain very high ozone concentrations.48,57,58These plumes are generally transported in the free troposphere above the boundary layer, and have a strongly layered structure that is diﬃcult to capture with Eulerian models because of numerical diﬀusion under stretchedﬂow conditions.59 Numerical diﬀusion broadly aﬀects the ability of models to capture observed maxima, particularly at mountain sites. The eﬀect is expected to be less at surface sites due to dilution of the plumes during entrainment into the boundary layer23,60 Findings. Several mature global models are available, but the full diversity of these models has not been applied to estimate the 9489

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology

CRITICAL REVIEW

Figure 4. Measured and modeled MDA8 ozone concentrations (left) for Gothic, Colorado (38.96N, 106.99W, 2926 m asl) during March 1, 2006 August 31, 2006. The black line shows the observed MDA8 ozone concentrations; the red line shows GEOS-Chem simulations from Zhang et al.;30 the blue line shows the PRB variations as calculated by GEOS-Chem. Numerical values in black, red and blue give the corresponding average MDA8 value for the entire period. Maximum MDA8 values for the year occurred on April 19th (83 ppbv) and 20th (88 ppbv), respectively, and were not captured by the GEOS-Chem model (black arrow). The right side shows HYSPLIT back-trajectories for April 20th, which indicate rapid descent from the upper troposphere (350 500 hPa). Trajectories were initialized at 300, 500, and 700 m above ground level. This rapid airmass descent is consistent with the 8 h average water vapor mixing ratio observed during the MDA8 period on April 20th of 1.5 g/kg compared to a mean of 4.6 ( 2.5 g/kg (1 sigma) for all days during this entire period, consistent with a free tropospheric source.

distributions of PRB ozone either in the EPA’s regulatory process or in the peer-reviewed literature. Agreement between global model predictions and measurements of monthly mean MDA8 ozone at RRMS provides some conﬁdence in the ability of global chemical transport models to predict PRB ozone in many parts of the U.S. Global chemical transport models exhibit biases in monthly mean MDA8 ozone in some regions of the U.S., including the Gulf Coast, regions aﬀected by ﬁres, and regions with complex topography, which have implications for model estimates of PRB ozone. They also have diﬃculty representing the ﬁne structures of ozone events at RRMS that include contributions from PRB ozone sources. The application of regional models or the nesting of ﬁner-scale regional models within global models for estimating PRB ozone has been only minimally explored. Research Recommendations. Comparisons of multiple, independent global model predictions of the distribution of PRB ozone in the U.S. that can be performed on a time scale consistent with EPA’s periodic review process for the ozone standards should be a high priority for the EPA and the global modeling community. Models used to provide PRB ozone estimates should be extensively evaluated with surface, sonde, and satellite ozone observations. Consistent protocols for evaluation of global model performance should be developed, and among other requirements, should include an assessment of the strengths and weaknesses in the representation of key chemical and physical processes and thorough documentation and review of emissions inventories. Research should focus on key processes that aﬀect predictions of MDA8 ozone concentrations in the U.S. by global models. Priorities include ﬁres, vertical transport in regions of complex topography, and physical and chemical processes in the marine boundary layer of the Gulf of Mexico. The diﬃculties Eulerian models encounter when simulating the ﬁne structures of ozone events observed at RRMS and their eﬀects on PRB ozone distributions should be evaluated. Speciﬁc questions that should be addressed during model evaluation are: (1) Do problems exist mainly at mountain sites or do they extend

to low elevation sites? (2) Do enhanced ozone events that have their origins in PRB sources signiﬁcantly aﬀect the enhancements in the overall ozone distribution, and if so, can the models identify these patterns? (3) Can model limitations in simulating the ﬁne structure of the ozone event distribution be overcome? and (4) Can model limitations be overcome by blending model and observational approaches? Research should explore the application of regional models and the nesting of ﬁner-scale regional models within global models for estimating PRB ozone and for understanding how processes, such as ﬁres, STE, drought, and changes in emissions inventories, aﬀect estimates of PRB ozone.

’ THE ROLE OF MEASUREMENTS IN THE DETERMINATION OF PRB OZONE PRB ozone is a model construct that must be informed by and evaluated based on observational data. The current regulatory deﬁnition of PRB includes certain sources of ozone and excludes others. Global models can distinguish estimates of PRB ozone from total predicted ozone, but comparable source apportionment of measured ambient ozone generally is not possible because emissions from multiple, disparate sources interact to form ozone in a highly nonlinear manner and unique source tracers are lacking. However, as described in the previous section, caution must be exercised in applying global models to estimate the distributions of PRB ozone. For example, Figure 3 indicates that GEOS-Chem overpredicts lower observed concentrations and does not capture observations greater than 65 ppb of MDA8 ozone concentrations at elevated (>1.5 km) CASTNet sites in the intermountain western U.S. during March-May 2006. Figure 4 shows similar results speciﬁcally for the Gothic Colorado site that experienced enhanced ozone concentrations during April 2006, likely from the descent of air from the upper troposphere, which was not replicated by the GEOS-Chem model. Observations are essential to validate models and improve the conﬁdence in their performance, to better understand the causes of enhanced ozone, to indicate geographic areas of strength and weakness, and to guide model improvements where needed. Measurements of baseline ozone at RRMS approximate PRB 9490

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology

CRITICAL REVIEW

Figure 5. Hourly averaged surface ozone concentrations by month at Trinidad Head, California during the daytime (upper left panel) and for Mt. Bachelor, Oregon for free tropospheric sampling conditions (upper right panel) for 2004 through 2009. The diamond represents the mean, the horizontal bar represents the median, the box represents the 25th and 75th percentiles, and the whiskers represent the 5th and 95th percentiles. Average spring (March/April/May) ozone proﬁles derived from ozone sondes at Trinidad Head are shown in the bottom left panel for 1997 through 2010. These data represent averages over 500 m in altitude with (1 standard deviation. The symbols in blue are the seasonal surface ozone average at Trinidad Head and the symbol in red is the Mt. Bachelor seasonal average plotted at the altitude of the observatory. Proﬁle results from the GEOS-Chem model for 2006 are shown as green pluses. The contribution from Policy Relevant Background in the model proﬁle is shown by orange circles. Representative standard deviations at several altitudes for the model are shown. The bottom right panel shows ozone proﬁles during the summer months (June/July/August) at Trinidad Head and from the model. The model results are plotted slightly oﬀset in altitude from the nominal altitude for clarity. The slight diﬀerence between the Trinidad Head sonde data and the Mt. Bachelor values is due to large scale diﬀerences in subsidence at the two locations.61

ozone, and provide powerful constraints for and evaluations of model-derived estimates of PRB ozone. A surface monitoring network designed speciﬁcally for informing PRB ozone estimates does not exist in the U.S. Ozone measurements at a number of relatively remote surface sites in the U.S. are regularly representative of baseline ozone concentrations. However, care must be taken when interpreting these measurements as approximating PRB ozone. Data from two of these sites, located near the U.S. west coast, the higher elevation Mt. Bachelor, Oregon site61 and marine boundary layer location and ozone sonde launch site at Trinidad Head, California,24,62 are shown in Figure 5, and data from these sites have been used widely to gain an observational perspective on western U.S. PRB ozone levels. At Mt. Bachelor,

multiple constituents are measured that provide markers indicative of possible PRB sources that contribute to measured (enhanced) ozone events at NAAQS thresholds.19,23,61 When interpreting such data, two considerations are important. First, baseline ozone and PRB ozone are not synonymous in that quantiﬁcation of baseline ozone attempts to avoid recent local or regional inﬂuences, while PRB ozone includes some of these inﬂuences such as surface deposition and local and regional ozone production from natural sources such as biogenic emissions and wildﬁres. Second, determination of baseline ozone concentrations by ﬁltering recent local and regional inﬂuences is a diﬃcult and somewhat uncertain process. For example, mountain top site data are often ﬁltered according to water vapor 9491

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology

CRITICAL REVIEW

Figure 6. Global tropospheric ozone distributions at 500 hPa from the TES and OMI satellite instruments for the diﬀerent seasons of 2006. The central two columns show the GEOS-Chem model ozone ﬁelds smoothed by the diﬀerent averaging kernels (sensitivities) of the two instruments, demonstrating that most of the diﬀerences between TES and OMI are due to diﬀerent instrument sensitivities rather than bias. White areas indicate lack of data meeting the retrieval quality criteria. From Zhang et al.74

measurements, which are designed to eliminate all boundary layer inﬂuences and retain only free troposphere data,19,63 and marine boundary layer data can be ﬁltered according to tracers of continental inﬂuences (radon, CO2 depletion, etc.),25,64 wind sector selection,24,25 or transport modeling.64 Nevertheless, when these complications are properly considered, baseline measurements of ozone do provide observational data for testing the ability of models to diagnose PRB ozone contributions. The GEOS-Chem predictions of ozone proﬁles and the North American (PRB) background ozone over Trinidad Head are also shown in Figure 5 averaged over the spring and summer of 2006. The model under-predicts the sonde measurements in the free troposphere by about 5 ppbv, but agrees well with measurements at Mt. Bachelor. The PRB ozone is on average 5 ppbv lower in spring and 7 ppbv lower in summer, reﬂecting inﬂuences from the North American anthropogenic sources in the baseline ozone concentrations. Interior continental surface sites that are generally free from local (<100 km) anthropogenic contributions of ozone can inform the understanding of PRB sources, including ﬁres and STE and provide valuable information for evaluating global models as described above. Some of the existing CASTNet ozone monitoring sites qualify for that purpose. However, additional measurements of carbon monoxide (CO), chemically speciated ﬁne particulate matter (PM), volatile organic compounds (VOCs), oxides of nitrogen (NOx) and total odd nitrogen (NOy) at a time resolution comparable to ozone measurements are required to provide stronger constraints on the model and aid in the diagnosis of model performance. In recent years, ground-based and aircraft campaigns have been carried out along the west coast of North America that have provided information on the characteristics of air reaching the west coast of the U.S, including the baseline ozone distribution, sources that contribute to this baseline, and the impacts on continental interior ozone distributions.65 68 These short-term studies have deployed advanced instrumentation that measure a large number of ozone precursors as well as other intermediate

species and secondary products (carbonyls, free radicals, aerosol composition). The resulting data sets provide not only stringent tests of model calculations, but also direct means to assess emission sources and to characterize atmospheric transformation and transport processes.65 68 Commercial aircraft equipped with appropriate instrumentation can also be a reliable and costeﬀective means of improving the vertical characterization of the troposphere, as demonstrated by the European MOZAIC program (1994 2010)69 and expected by the recently initiated IAGOS (In-Service Aircraft for the Global Observing System; http://www.iagos.org/) program. In its recently issued plans for the health risk and exposure assessment29 for the ongoing periodic review of the air quality criteria and standards for ozone, the EPA noted that the performance evaluation of the GEOS-Chem model will be conducted using both surface and satellite measurements. Satellite-borne sensors have been used in the detection and quantiﬁcation of the long-range transport of tropospheric ozone and its precursor species and aerosol pollution.32,70 Detection of tropospheric ozone from satellites is challenging because of the larger stratospheric signal. In addition, stratospheric/tropospheric exchange events (folds) that transport stratospheric ozone into the free troposphere can be diﬃcult to distinguish from pollution plumes entrained into the free troposphere, particularly if only ozone concentration ﬁelds are utilized.23,32,71 Nonetheless, satellite data sets have provided unique large-scale views with repeat coverage that have allowed plumes from intense pollution episodes to be tracked around the northern hemisphere.32,72,73 Validation and intercomparison of tropospheric ozone observations from the tropospheric emission spectrometer (TES) and ozone monitoring instrument (OMI) satellite sensors, operating in the IR and in the UV respectively, lends conﬁdence in the use of satellite observations for testing models of tropospheric ozone.74 Figure 6 from Zhang et al.74 shows the global distributions of tropospheric ozone at 500 hPa from the TES and OMI sensors, along with comparisons to the GEOS-Chem model. In addition, satellite pollutant concentration ﬁeld data sets coupled 9492

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology

CRITICAL REVIEW

Figure 7. Average NO2 columns over the western U.S. during the summer of 2005 derived from the SCIAMACHY satellite from Kim et al.80 Boxes highlight emissions from some U.S. electrical power plants (labeled with P) and cities (labeled with C). Since NO2 has a short atmospheric lifetime (several hours), emission sources can be identiﬁed from their geographic position without confounding eﬀects from transport.

with inverse model analyses of CO, nitrogen dioxide (NO2) and formaldehyde (HCHO) have been used to constrain emissions of CO, NOx, and VOCs respectively .75 79 For example, Figure 7 shows average NO2 columns due to NOx emissions from electrical power plants and urban areas in the western U.S. during the summer of 2005 derived from the SCIAMACHY satellite.80 However, all U.S. satellite sensors for measuring tropospheric composition are beyond their design lifetimes (including MODIS, MISR, MOPITT, AIRS, OMI, TES) and there is no plan for immediate successor systems to provide continuity. Serious degradation of the observation system from space is to be expected in the coming few years, only partly mitigated by the continuing European program (GOME-x, IASI sensors). Findings. PRB ozone is a model construct that must be informed and validated by data from surface, airborne, and space-based measurements. Surface and airborne measurements of species other than ozone will be required to identify and quantify PRB ozone sources and to constrain models for PRB ozone estimates. Satellite-borne sensors have been used in the detection and quantiﬁcation of the long-range transport of tropospheric ozone and its precursor species, and in the evaluation and constraint of emissions estimates of important ozone precursor species. Many U.S. satellite sensors for measuring tropospheric composition are beyond their design lifetimes. Degradation or loss of these systems will be only partly mitigated by the continuing European programs. Research Recommendations. Joint research teams composed of investigators with expertise in modeling of PRB ozone and measurements at relatively remote sites should develop integrated assessments of the distributions of PRB ozone in the U.S. Existing, geographically dispersed remote monitoring sites throughout the U.S. should be enhanced by adding measurements of CO, NOy, and PM mass.

Deployment of additional measurements including VOCs, halocarbons, mercury (Hg), chemically speciated ﬁne PM, and NOx should be considered at research sites to allow more reﬁned chemical ﬁngerprinting of air masses. Policy Recommendations. The U.S. should consider instrumentation of U.S. commercial aircraft for long-term observations of atmospheric trace gases. Current U.S. satellite sensors capable of characterizing and quantifying emissions and long-range transport of ozone and critical precursor species should be maintained as long as possible. Planning and implementation of new U.S. atmospheric composition satellite missions should continue as vigorously as possible. Engagement in this process could provide attention to sensor speciﬁcations that may facilitate the collection of PRB relevant data. Two U.S. atmospheric composition missions are in their early planning phases, with expected launches in the 2020s, at the National Aeronautics and Space Administration (NASA); a geostationary air quality mission (GEO-CAPE) and a global atmospheric composition mission (GACM). Such a combination of geostationary (North America) and low-elevation orbit (global) perspectives would be of considerable value for better constraining estimates of PRB ozone.

’ INFLUENCE OF SOURCES OF PRB OZONE ON ATTAINMENT DEMONSTRATIONS AND SIP DEVELOPMENT The EPA provides guidance on the use of models and other analyses for ozone attainment demonstrations and State Implementation Plan (SIP) development.81 The initial step in the recommended modeling procedure is the development of a conceptual description of ozone formation in the local area. Based on the results of the conceptual description, potential episodes or seasons to use for photochemical grid modeling are analyzed and eventually a time period for modeling is selected. 9493

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology Conceptual descriptions are conducted based largely on observational data. The ﬁnal steps of the modeling procedure include: (i) running the air quality model with basecase emissions to evaluate the performance and performing diagnostic tests to improve the model, as necessary, and (ii) performing future year modeling (including additional control strategies, if necessary) and applying the attainment test at each monitor.81 The attainment test is based on a relative rather than absolute application of the model estimates, such that the ratio of the model’s future to current (baseline) predictions or relative response factor (RRF) is considered. Future ozone concentrations for the attainment test are estimated at existing monitoring sites by multiplying the modeled relative response factor at locations near each monitor by the observation-based, baseline ozone design value for that site.81 Resulting predicted future concentrations are compared to NAAQS. If these do not meet the attainment test, the future year modeling is typically repeated with the inclusion of one or a suite of emissions control strategies until the attainment test can be passed. Many sources of PRB ozone do not yet have an explicitly acknowledged role in the development and evaluation of attainment demonstrations and SIPs. In some areas of the U.S., PRB ozone concentrations may provide a signiﬁcant contribution on days identiﬁed as having high ozone, especially with a lower threshold that would accompany a more stringent NAAQS. Identifying and understanding the role of sources of PRB ozone relative to local, regional, or continental contributions to observations at monitoring sites during the development of a conceptual description may inﬂuence the selection of historical time periods for air quality modeling. The EPA does have a rule establishing criteria and procedures for use in determining whether air quality monitoring data have been inﬂuenced by exceptional events.82 While the concept of exceptional events and elimination of certain types of events from air quality planning activities has been used for emission sources such as wildﬁres,83 there is no precedent for its use for some of the sources of PRB ozone. Because of the nature of the approach for determining attainment, it is important that states understand the relative eﬀects of sources of PRB ozone on model responsiveness (i.e., stiﬀness) to local and regional emissions control strategies under consideration. Finding. Many sources of PRB ozone do not have an explicitly acknowledged role in the development and evaluation of ozone attainment demonstrations and State Implementation Plans, but may impact components of these processes. Policy Recommendations. EPA should inform states about existing and emerging developments in the quantification of PRB ozone and assess the potential role of PRB ozone in attainment demonstrations.

’ SUMMARY PERSPECTIVES PRB ozone concentrations8,13,14 are deﬁned by the EPA as those concentrations that would occur in the U.S. in the absence of anthropogenic emissions in continental North America (U.S., Canada, and Mexico). This review found that further clarity is required in the regulatory deﬁnition of PRB ozone and its application, recognizing the contributions and uncertainties surrounding biomass burning, biogenic emissions related to agriculture, ocean-going vessels and air transport operations, and North American anthropogenic methane emissions. Estimates of PRB ozone have had an important role historically in the

CRITICAL REVIEW

EPA’s human health and welfare risk analyses used in establishing NAAQS. The margin of safety for the protection of public health in the ozone rulemaking process has been established from human health risks calculated based on PRB ozone estimates. Sensitivity analyses conducted by the EPA17 have illustrated that changing estimates of PRB ozone concentrations have a progressively greater impact on estimates of mortality risk as more stringent standards are considered. Many sources of PRB ozone do not currently have a role in the development and evaluation of attainment demonstrations and SIPs, but may impact components of these processes as well. As deﬁned by the EPA, PRB ozone is a model construct, but it is informed by data from surface, airborne, and space-based measurements. Consequently, collaborative research teams with expertise spanning these areas are needed for integrated assessments of the distributions of PRB ozone. Strong spatial and temporal variability exists in the distributions of PRB ozone concentrations across the U.S. A recent assessment30 with the GEOS-Chem model found 6-month mean PRB values for the U.S. in spring-summer of 27 ( 8 ppbv at low altitude sites (<1.5 km) and 40 ( 7 ppbv at high-altitude CASTNet sites. In some regions, PRB concentrations approach 60 70 ppb, a range previously under consideration for revisions to the ozone NAAQS. Several mature global models are available, but the full diversity of these models has not been applied to estimate the distributions of PRB ozone either in the EPA’s regulatory process or in the peer-reviewed literature. Comparisons of multiple, independent global model predictions of the distribution of PRB ozone should be a high priority for the EPA and the global modeling community. Future research should focus on key processes that aﬀect predictions of MDA8 ozone concentrations in the U.S. by global models, including ﬁres, STE, vertical transport in regions of complex topography, physical and chemical processes in the marine boundary layer of the Gulf of Mexico, as well as replication of the ﬁne structures of ozone events observed at RRMS.

’ AUTHOR INFORMATION Corresponding Author

*Phone: (512) 471-0049; fax: (512) 471-1720; e-mail: allen@ che.utexas.edu.

’ ACKNOWLEDGMENT We express our appreciation to Dr. Harvey Richmond, EPA Senior Risk Analyst, for his review of the manuscript. We also thank Chris Rabideau and Ted Steichen, the American Petroleum Institute observers for their attendance at the workshop, and Dr. Joseph Pinto of the Environmental Protection Agency for his attendance and participation in the workshop. We thank the American Petroleum Institute for providing ﬁnancial support for the workshop, and the reviewers of the manuscript for their interest and insights. ’ REFERENCES (1) Bell, M. L.; McDermott, A.; Zeger, S. L.; Samet, J. M.; Dominici, F. D. Ozone and short-term mortality in 95 U.S. urban communities, 1987 2000. J. Am. Water Works Assoc. 2004, 292 (19), 2372–2378. (2) Bell, M. L.; Peng, R. D.; Dominci, F. The ozone-response curve for ozone and risk of mortality and the adequacy of current ozone regulations. Environ. Health Perspect. 2006, 114, 532–536. 9494

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology (3) Chen, T.; Gokhale, J; Shofer, S.; Kuschner, W. G. Outdoor air pollution: ozone health eﬀects. Am. J. Med. Sci. 2007, 333 (4), 244–248. (4) Sanhueza, P. A.; Reed, G. D.; Davis, W. T.; Miller, T. L. An environmental decision-making tool for evaluating ground-level ozonerelated health eﬀects. J. Air Waste Manage. 2003, 53 (12), 1448–1459. (5) Fuhrer, J. Ozone risk for crops and pastures in present and future climates. Naturwissenschaften 2009, 96 (2), 173–194. (6) Kline, L. J.; Davis, D. D.; Skelly, J. M.; Savage, J. E.; Ferdinand, J. Ozone sensitivity of 28 plant selections exposed to ozone under controlled conditions. Northeast. Nat. 2008, 15 (1), 57–66. (7) Musselman, R. C.; Lefohn, A. S.; Massman, W. J.; Heath, R. L. A critical review and analysis of the use of exposure and ﬂux-based ozone indices for predicting vegetation eﬀects. Atmos. Environ. 2006, 40, 1869–1888. (8) U.S. EPA. Air Quality Criteria for Ozone and Other Photochemical Oxidants, EPA 600/R-05/004aF; U.S. Environmental Protection Agency, Oﬃce of Air Quality Planning and Standards, 2006a; Vol. I. (9) National Ambient Air Quality Standards for Ozone; Final Rule. Fed. Regist. 2008, 73(60), 16436 16514. (10) National Ambient Air Quality Standards for Ozone, Proposed Rule. Fed. Regist. 2010, 75 (11), 2938 3052. (11) The White House, Oﬃce of the Press Secretary, http://www. whitehouse.gov/the-press-oﬃce/2011/09/02/statement-presidentozone-national-ambient-air-quality-standards (accessed September 2011). (12) U.S. EPA. Integrated Review Plan for the Ozone National Ambient Air Quality Standards Review, EPA 452/D-09-001; U.S. Environmental Protection Agency, Oﬃce of Air Quality Planning and Standards, 2009. (13) U.S. EPA. Air Quality Criteria for Ozone and Other Photochemical Oxidants, EPA 600/R-05/004bF; U.S. Environmental Protection Agency, Oﬃce of Air Quality Planning and Standards, 2006b; Vol. II. (14) U.S. EPA. Air Quality Criteria for Ozone and Other Photochemical Oxidants EPA 600/R-05/004cF; U.S. Environmental Protection Agency, Oﬃce of Air Quality Planning and Standards, 2006c; Vol. III. (15) U.S. EPA. Air Quality Criteria for Ozone and Other Photochemical Oxidants, EPA 600/P-93/004aF; U.S. Environmental Protection Agency, Oﬃce of Air Quality Planning and Standards, 1996a; Vol. I. (16) U.S. EPA. Review of the National Ambient Air Quality Standards for Ozone, Assessment of Scientiﬁc and Technical Information, EPA 452/R96-007; U.S. Environmental Protection Agency, Oﬃce of Air Quality Planning and Standards, 1996b. (17) U.S. EPA. Review of the National Ambient Air Quality Standards for Ozone: Policy Assessment of Scientiﬁc and Technical Information, EPA 452/R-07-007, OAQPS Staﬀ Paper; U.S. Environmental Protection Agency, Oﬃce of Air Quality Planning and Standards, 2007a. (18) Lefohn, A. S.; Wernli, H.; Shadwick, D.; Limbach, S.; Oltmans, S. J.; Shapiro, M. The importance of stratospheric-tropospheric transport in aﬀecting surface ozone concentrations in the Western and Northern Tier of the United States. Atmos. Environ. 2011, 45, 4845–4857. (19) Ambrose, J. L., Reidmiller, D. R., Jaﬀe, D. A. Causes of high O3 in the lower free troposphere over the Paciﬁc Northwest as observed at the Mt. Bachelor Observatory. Atmos. Environ. 2011, DOI: 10.1016/j. atmosenv.2011.06.056. (20) Fiore, A.; Jacob, D. J.; Liu, H.; Yantosca, R. M.; Fairlie, T. D.; Li, Q. Variability in surface ozone background over the United States: Implications for air quality policy. J. Geophys. Res. 2003, 108(D24) 4787, DOI: 10.1029/2003JD003855 (21) U.S. EPA. Integrated Science Assessment for Ozone and Related Photochemical Oxidants, EPA 600/R-10/076A; U.S. Environmental Protection Agency, Oﬃce of Air Quality Planning and Standards, 2011. (22) Cooper, O. R.; Parrish, D. D.; Stohl, A.; Trainer, M.; Nedelec, P.; Thouret, V.; Cammas, J. P.; Oltmans, S. J.; Johnson, B. J.; Tarasick, D.; Leblanc, T.; McDermid, I. S.; Jaﬀe, D.; Gao, R.; Stith, J.; Ryerson, T.; Aikin, K.; Campos, T.; Weinheimer, A.; Avery, M. A. Increasing springtime ozone mixing ratios in the free troposphere over western North America. Nature 2010, 463, 344–348http://dx.doi.org/10.1038/ nature08708380093.

CRITICAL REVIEW

(23) Zhang, L.; Jacob, D. J.; Boersma, K. F.; Jaﬀe, D. A.; Olson, J. R.; Bowman, K. W.; Worden, J. R.; Thompson, A. M.; Avery, M. A.; Cohen, R. C.; Dibb, J. E.; Flock, F. M.; Fuelberg, H. E.; Huey, L. G.; McMillan, W. W.; Singh, H. B.; Weinheimer, A. J. Transpaciﬁc transport pollution and the eﬀect of recent Asian emission increases on air quality in North America: an integrated analysis using satellite, aircraft, ozonesonde, and surface observations. Atmos. Chem. Phys. 2008, 8, 6117–6136. (24) Oltmans, S. J.; Lefohn, A. S.; Harris, J. M.; Shadwick, D. S. Background ozone levels of air entering the west coast of the U.S. and assessment of longer-term changes. Atmos. Environ. 2008, 42, DOI: 10.1016/j.atmosenv.2008.03.034, 6020-6038. (25) Parrish, D. D.; Millet, D. B.; Goldstein, A. H. Increasing ozone in marine boundary layer inﬂow at the west coasts of North America and Europe. Atmos. Chem. Phys. 2009, 9, 1303–1323. (26) Langford, A. O.; Aikin, K. C.; Eubank, C. S.; Williams, E. J. Stratospheric contribution to high surface ozone in Colorado during springtime. Geophys. Res. Lett. 2009, 36, L12801, DOI: 10.1029/ 2009GL038367. (27) Kaynak, B.; Hu, Y.; Martin, R. V.; Russell, A. G.; Choi, Y.; Wang, Y. The eﬀect of lightning NOx production on surface ozone in the continental United States. Atmos. Chem. Phys. 2008, 8, 5151–5159. (28) Wang, H.; Jacob, D. J.; Le Sager, P.; Streets, D. G.; Park, R. J.; Gilliland, A. B.; van Donkelaar, A. Surface ozone background in the United States: Canadian and Mexican pollution inﬂuences. Atmos. Environ. 2009, 43, 1310–1319. (29) U.S. EPA. Ozone National Ambient Air Quality Standards: Scope and Methods Plan for Health Risk and Exposure Assessment, EPA 452/P11-001; U.S. Environmental Protection Agency, Oﬃce of Air Quality and Planning and Standards, 2011. (30) Zhang, L; Jacob, D. J.; Downey, N. V.; Wood, D. A.; Blewitt, D.; Carouge, C. C.; van Donkelaar, A.; Jones, D. B. A.; Murray, L. T.; Wang, Y. Improved estimate of the policy-relevant background ozone in the United States using the GEOS-Chem global model with 1/2 2/3 horizontal resolution over North America. Atmos. Environ. 2011in press. (31) Task Force on Hemispheric Transport of Air Pollution. Hemispheric Transport of Air Pollution 2010 Part A: Ozone and Particulate Matter; United Nations Publication, 2010; ISBN: 978-92-1-116977-5. (32) National Research Council (NRC). Global Sources of Local Pollution; National Academy Press: Washington, DC, 2009. (33) Reid, N.; Yap, D.; R. Bloxam, R. The potential role of background ozone on current and emerging air issues: An overview. Air Qual. Atmos. Health 2008, 1, 19–29. (34) (a)Mueller, S. F.; Mallard, J. W. Contributions of natural emissions to ozone and PM(2.5) as simulated by the Community Multiscale Air Quality (CMAQ) Model. Environ. Sci. Technol. 2011, 45(11), 4817 4823, DOI: 10.1021/es103645m. (b)Errata in contributions of natural emissions to ozone and PM2.5 as simulated by the Community Multiscale Air Quality (CMAQ) Model, Environ. Sci. Technol. 2011, 45(18), 7950. (35) Lefohn, A. S.; Oltmans, S. J.; Dann, T.; Singh, H. B. Present-day variability of background ozone in the lower troposphere. J. Geophys. Res. 2001, 106 (D9), 9945–9958. (36) Fiore, A. M.; Jacob, D. J.; Bey, I.; Yantosca, R. M.; Field, B. D.; Fusco, A. C. Background ozone over the United States in summer: origin, trend, and contribution to pollution episodes. J. Geophys. Res. 2002, 107 (D15), 4275, DOI: 10.1029/2001JD000982. (37) Wang, Y. H.; Jacob, D. J.; Logan, J. A. Global simulation of tropospheric O3 NOx hydrocarbon chemistry 3. Origin of tropospheric ozone and eﬀects of nonmethane hydrocarbons. J. Geophys. Res. 1998, 103 (D9), 10757–10767. (38) Reidmiller, D. R.; Fiore, A. M.; Jaﬀe, D. A.; Bergmann, D.; Cuvelier, C.; Dentener, F. J.; Duncan, B. N.; Folberth, G.; Gauss, M.; Gong, S.; Hess, P.; Jonson, J. E.; Keating, T.; Lupu, A.; Marmer, E.; Park, R.; Schultz, M. G.; Shindell, D. T.; Szopa, S.; Vivanco, M. G.; Wild, O.; Zuber, A. The inﬂuence of foreign vs. North American emissions on surface ozone in the US. Atmos. Chem. Phys. 2009, 9, 5027–5042. (39) Climate Change 2001: The Scientiﬁc Basis. Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel 9495

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology on Climate Change; Houghton, J. T., Ding, Y., Griggs, D. J., Noguer, M.; van der Linden, P., Dai, X., Maskell, K., Johnson, C. A., Eds.; IPCC, Cambridge University Press, 2001. (40) Stevenson, D. S.; Dentener, F. J.; Schultz, M. G.; Ellingsen, K.; van Noije, T. P. C.; Wild, O.; Zeng, G.; Amann, M.; Atherton, C. S.; Bell, N.; Bergmann, D. J.; Bey, I.; Butler, T.; Cofala, J.; Collins, W. J.; Derwent, R. G.; Doherty, R. M.; Drevet, J.; Eskes, H. J.; Fiore, A. M.; Gauss, M.; Hauglustaine, D. A.; Horowitz, L. W.; Isaksen, I. S. A.; Krol, M. C.; Lamarque, J.-F.; Lawrence, M. G.; Montanaro, V.; M€uller, J.-F.; Pitari, G.; Prather, M. J.; Pyle, J. A.; Rast, s.: Rodriguez, J. M.; Sanderson, M. G.; Savage, N. H.; Shindell, D. T.; Strahan, S. E.; Sudo, K.; Szopa, S. Multimodel ensemble simulations of present-day and near-future tropospheric ozone. J. Geophys. Res. 2006, 111 (D08301), DOI: 10.1029/ 2005JD006338. (41) Fiore, A. M.; Dentener, F. J.; Wild, O.; Cuvelier, C.; Schultz, M. G.; Hess, P.; Textor, C.; Schulz, M.; Doherty, R. M.; Horowitz, L. W.; MacKenzie, I. A.; Sanderson, M. G.; Shindell, D. T.; Stevenson, D. S.; Szopa, S.; Van Dingenen, R.; Zeng, G.; Atherton, C.; Bergmann, D.; Bey, I.; Carmichael, G.; Collins, W. J.; Duncan, B. N.; Faluvegi, G.; Folberth, G.; Gauss, M.; Gong, S.; Hauglustaine, D.; Holloway, T.; Isaksen, I. S. A.; Jacob, D. J.; Jonson, J. E.; Kaminski, J. W.; Keating, T. J.; Lupu, A.; Marmer, E.; Montanaro, V.; Park, R. J.; Pitari, G.; Pringle, K. J.; Pyle, J. A.; Schroeder, S.; Vivanco, M. G.; Wind, P.; Wojcik, G.; Wu, S.; Zuber, A. Multimodel estimates of inter-continental source-receptor relationships for ozone pollution. J. Geophys. Res. 2009, 114, D04301–D04321. (42) Jacob, D. J.; Wofsy, S. C.; Bakwin, P. S.; Fan, S,M; Harriss, R. C.; Talbot, R. W.; Bradshaw, J. D.; Sandholm, S. T.; Singh, H. B.; Browell, E. V.; Gregory, G. L.; Sachse, G. W.; Shipham, M. C.; Blake, D. R.; Fitzjarrald, D. R. Summertime photochemistry at high northern latitudes. J. Geophys. Res. 1992, 97, 16421–16431. (43) McKeen, S. A.; Wotawa, G.; Parrish, D. D.; Holloway, J. S.; Buhr, M. P.; H€ubler, G.; Fehsenfeld, F. C.; Meagher, J. F. Ozone production from Canadian wildﬁres during June and July of 1995. J. Geophys. Res. 2002, 107(D14), 4192, DOI: 10.1029/2001JD000697. (44) Alvarado, M. J.; Logan, J. A.; Mao, J.; Apel, E.; Riemer., D.; Blake, D.; Cohen, R. C.; Min, K. E.; Perring, A. E.; Browne, E. C.; Wooldridge, P. J.; Diskin, G. S.; Sachse, G. W.; Fuelberg, H.; Sessions, W. R.; Harrigan, D. L.; Huey, G.; Liao, J.; Cae-Hanks, A.; Jimenez, J. L.; Cubison, M. J.; Vay, S. A.; Weinheimer, A. J.; Knapp, D. J.; Montzka, D. D.; Flocke, F. E.; Pollack, I. B.; Wennberg, P. O.; Kurten, A.; Crounse, J.; St. Clair, J. M.; Wisthaler, A.; Mikoviny, T.; Yantosca, R. M.; Carouge, C. C.; Le Sager, P. Nitrogen oxides and PAN in plumes from boreal ﬁres during ARCTAS-B and their impact on ozone: An integrated analysis of aircraft and satellite observations. Atmos. Chem. Phys. 2010, 10, 9739–9760. (45) Jaﬀe, D. A. Let the data speak! What do observations say about PRB in the western U.S.? Workshop on Policy Relevant Background Ozone Concentrations in the United States: Austin, Texas, 2011, http:// www.utexas.edu/research/ceer/prb/ (accessed October 20, 2011). (46) Jaﬀe, D. A.; Chand, D.; Hafner, W.; Westerling, A.; Spracklen, D. Inﬂuence of ﬁres on O3 concentrations in the western U.S. Environ. Sci. Technol. 2008, 42 (16) 5885 5891, DOI: 10.1021/es800084k. (47) Val Martin, M.; Honrath, R. E.; Owen, R. C.; Pﬁster, G.; Fialho, P.; Barata, F. Signiﬁcant enhancements of nitrogen oxides, black carbon, and ozone in the North Atlantic lower free troposphere resulting from North American boreal wildﬁres. J. Geophys. Res. 2006, 111, D23S60, DOI: 10.1029/2006JD007530. (48) Jaﬀe, D. A.; Bertschi, I.; Jaegle, L.; Novelli, P.; Reid, J. S.; Tanimoto, H.; Vingarzan, R.; Westphal, D. L. Long-range transport of Siberian biomass burning emissions and impact on surface ozone in western North America. Geophys. Res. Lett. 2004, 31, 16106, DOI: 10.1029/2004GL020093. (49) Oltmans, S.; Cooper, O.; Lefohn, A. Searching for PRB: Characterizing ozone levels at the west coast of North America from ozonesonde and surface observations. Workshop on Policy Relevant Background Ozone Concentrations in the United States: Austin, Texas, 2011, http://www.utexas.edu/research/ceer/prb/ (accessed October 20, 2011).

CRITICAL REVIEW

(50) Parrish, D. D.; Aikin, K. C.; Oltmans, S. J.; Johnson, B. J., Sweeny, C. Impact of transported background ozone inﬂow on summertime air quality in a California ozone exceedance area. Atmos. Chem. Phys. 2010, 10, 10093 10109, DOI: 10.5194/acp-10-10093-2010. (51) Ludwig, F. L.; Reiter, E.; Shelar, et al. The Relation of Oxidant Levels to Precursor Emissions and Meteorological Features: v. I, Analysis and Findings, EPA 450/3-77-022a; U.S. Environmental Protection Agency, Oﬃce of Air Quality Planning and Standards, 1977. (52) Cooper, O. R.; Stohl, A.; H€ubler, G.; Hsie, E. Y.; Parrish, D. D.; Tuck, A. F.; Kiladis, G. N.; Oltmans, S. J.; Johnson, B. J.; Shapiro, M.; Moody, J. L.; Lefohn, A. S. Direct transport of mid-latitude stratospheric ozone into the lower troposphere and marine boundary layer of the tropical Paciﬁc Ocean. J. Geophys. Res. 2005, 110, D23310, DOI: 10.1029/2005JD005783. (53) Lefohn, A. S. Searching for an adequate estimate of ozone PRB. Workshop on Policy Relevant Background Ozone Concentrations in the United States: Austin, Texas, 2011, http://www.utexas.edu/research/ ceer/prb/ (accessed October 20, 2011). (54) Jaﬀe D. A. Relationship between surface and free tropospheric ozone in the western U.S. Environ. Sci. Technol. 2011b, 45, 432 438 DOI: 10.1021/es1028102. (55) Trickl, T.; Feldmann, H.; Kanter, H. J.; Scheel, H.-E.; Sprenger, M.; Stohl, A.; Wernli, H. Forecasted deep stratospheric intrusions over Central Europe: case studies and climatologies. Atmos. Chem. Phys. 2010, 10, 499–524. (56) Nowak, J. B. et al. Gas-phase chemical characteristics of Asian emission plumes observed during ITCT 2K2 over the eastern North Paciﬁc Ocean. J. Geophys. Res. 2004, 109, D23S19, DOI: 10.1029/ 2003JD004488. (57) Bertschi, I. T.; Jaﬀe, D. A.; Jaegle, L.; Price, H. U.; Dennison, J. B. PHOBEA/ITCT 2002 airborne observations of transpaciﬁc transport of ozone, CO, volatile organic compounds, and aerosols to the northeast Paciﬁc: Impacts of Asian anthropogenic and Siberian boreal ﬁre emissions. J. Geophys. Res. 2004, 109, D23S12, DOI: 10.1029/ 2003JD004328. (58) Oltmans, S. J.; Lefohn, A. S.; Harris, J. M.; Tarasick, D. W.; Thompson, A. M.; Wernli, H.; Johnson, B. J.; Novelli, P. C.; Montzka, S. A.; Ray, J. D.; Patrick, L. C.; Sweeney, C.; Jeﬀerson, A.; Dann, T.; Davies, J.; Shapiro, M.; Holben, B. N. Enhanced ozone over western North America from biomass burning in Eurasia during April 2008 as seen in surface and proﬁle observations. Atmos. Environ. 2010, 44, 4497–4509. (59) Rastigejev, Y.; Park, R.; Brenner, M.; Jacob, D. J. Resolving intercontinental pollution plumes in global models of atmospheric transport. J. Geophys. Res. 2010, 115, D02302, DOI: 10.1029/ 2009JD012568. (60) Hudman, R. C.; Jacob, D. J.; Cooper, O. R.; Evans, M. J.; Heald, C. L.; Park, R. J.; Fehsenfeld, F.; Flocke, F.; Holloway, J.; H€ubler, G.; Kita, K.; Koike, M.; Kondo, Y.; Neuman, A.; Nowak, J.; Oltmans, S.; Parrish, D.; Roberts, J. M.; Ryerson, T. Ozone production in transpaciﬁc Asian pollution plumes and implications for ozone air quality in California. J. Geophys. Res. 2004, 109, D23S10, DOI: 10.1029/2004JD004974. (61) Weiss-Penzias, P.; Jaﬀe, D. A.; Swartzendruber, P.; Dennison, J. B.; Chand, D.; Hafner, W.; Prestbo, E. Observations of Asian air pollution in the free troposphere at Mt. Bachelor Observatory in the spring of 2004. J. Geophys. Res. 2006, 110, D10304, DOI: 10.1029/ 2005JD006522. (62) Newchurch, M. J.; Ayoub, M. A.; Oltmans, S.; Johnson, B.; Schmidlin, F. J. Vertical distribution of ozone at four sites in the United States. J. Geophys. Res. 2003, 108(D1), 4031, DOI: 10.1029/ 2002JD002059. (63) Macdonald, A. M.; Anlauf, K. G.; Leaitch, W. R.; Chan, E. Interannual variability of ozone and carbon monoxide at the Whistler high elevation site: 2002 2006. Atmos. Chem. Phys. Discuss. 2011, 11, 17621–17664. (64) Derwent, R. G.; Simmonds, P. G.; Manning, A. J.; Spain, T. G. Trends over a 20-year period from 1987 to 2007 in surface ozone at the atmospheric research station, Mace Head, Ireland. Atmos. Environ. 2007, 41, 9081–9098. 9496

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

Environmental Science & Technology (65) Singh, H. B.; Brune, W. H.; Crawford, J. H.; Flocke, F.; Jacob, D. J. Chemistry and transport of pollution over the Gulf of Mexico and the Paciﬁc: Spring 2006 INTEX-B campaign overview and ﬁrst results. Atmos. Chem. Phys. 2009, 9, 2301–2318. (66) Jacob, D. J.; Crawford, J. H.; Maring, H.; Clarke, A. D.; Dibb, J. E.; Emmons, L. K.; Ferrare, R. A.; Hostetler, C. A.; Russell, P. B.; Singh, H. B.; Thompson, A. M.; Shaw, G. E.; McCauley, E.; Pederson, J. R.; Fisher, J. A. The Arctic Research of the Composition of the Troposphere from Aircraft and Satellites (ARCTAS) mission: design, execution, and ﬁrst results. Atmos. Chem. Phys. 2010, 10, 5191–5212. (67) Parrish, D. D.; Kondo, Y.; Cooper, O. R.; Brock, C. A.; Jaﬀe, D. A.; Trainer, M.; Ogawa, T.; H€ubler, G.; Fehsenfeld, F. C. Intercontinental Transport and Chemical Transformation 2002 (ITCT 2K2) and Paciﬁc Exploration of Asian Continental Emission (PEACE) experiments: An overview of the 2002 winter and spring intensives. J. Geophys. Res. 2004, 109, D23S01, DOI: 10.1029/2004JD004980. (68) Kotchenruther, R. A.; Jaﬀe, D. A.; Beine, H. J.; Anderson, T. L.; Bottenheim, J. W.; Harris, J. M.; Blake, D. R.; Schmitt, R. Observations of Ozone and Related Species in the Northeast Paciﬁc during the PHOBEA Campaigns: 2. Airborne observations. J. Geophys. Res. 2001, 106, 7463–7483. (69) Marenco, A.; Thouret, V.; Nedelec, P.; Smit, H.; Helten, M.; Kley, D.; Karcher, F.; Simon, P.; Law, K.; Pyle, J.; Poschmann, G.; Von Wrede, R.; Hume, C.; Cook, T. Measurement of ozone and water vapor by Airbus in-service aircraft: The MOZAIC airborne program. J. Geophys. Res. 1998, 103, 25,631–25,642. (70) National Research Council (NRC). Global Air Quality an Imperative for Long-term Observational Strategies; National Academy Press: Washington, DC, 2001. (71) Schoeberl, M. R.; Ziemke, J.R.; ; Bojkov, B.; Livesey, N.; Duncan, B.; Strahan, S.; Froidevaux, L.; Kulawik, S.; Bhartia, P. K.; Chandra, S.; Levelt, P. F.; Witte, J. C.; Thompson, A. M.; Cuevas, E.; Redondas, A.; Tarasick, D. W.; Davies, J.; Bodeker, G.; Hansen, G.; Johnson, B. J.; Oltmans, S.; V€omel, H.; Allaart, M.; Kelder, H.; Newchurch, M.; Godin-Beekmann, S.; Ancellet, G.; Claude, H.; Andersen, S. B.; Kyr€o, E.; Parrondos, M.; Yela, M.; Zablocki, G.; Moore, D.; Dier, H.; von der Gathen, P.; Viatte, P.; St€ubi, R.; Calpini, B.; Skrivankova, P.; Dorokhov,V.; de Backer, H.; Schmidlin, F. J.; Coetzee, G.; Fujiwara, M.; Thouret, V.; Posny, F.; Morris, G.; Merrill, J.; Leong, C. P.; Koenig-Langlo, G.; Joseph, E. A trajectory-based estimate of the tropospheric ozone column using the residual method. J. Geophys. Res. 2007, 112, D24S49, DOI: 10.1029/2007JD008773. (72) Ziemke, J. R.; Chandra, S.; Duncan, B. N.; Froidevaux, L.; Bhartia, P. K.; Levelt, P. F.; Waters, J. W. Tropospheric ozone determined from Aura OMI and MLS: Evaluation of measurements and comparison with the Global Modeling Initiative’s Chemical Transport Model. J. Geophys. Res. 2006, 111, D19303, DOI: 10.1029/2006JD007089. (73) Liu, X.; Bhartia, P. K.; Chance, K.; Spurr, R. J. D.; Kurosu, T. P. Ozone proﬁle retrievals from the Ozone Monitoring Instrument. Atmos. Chem. Phys. 2010, 10, 2521 2537, DOI: 10.5194/acp10-2521-2010. (74) Zhang, L.; Jacob, D. J.; Liu, X.; Logan, J. A.; Chance, K.; Eldering, A.; Bojkov, B. R. Intercomparison methods for satellite measurements of atmospheric composition: application to tropospheric ozone from TES and OMI. Atmos. Chem. Phys. 2010, 10, 4725–4739. (75) Clerbaux, C.; Edwards, D. P.; Deeter, M.; Emmons, L.; Lamarque, J. F.; Tie, X. X.; Massie, S. T.; Gille, J. Carbon monoxide pollution from cities and urban areas observed by the Terra/MOPITT mission. Geophys. Res. Lett. 2008, 35, L03817, DOI: 1029/ 2007GL032300. (76) Kopacz, M.; Jacob, D. J.; Henze, D. K.; Heald, C. L.; Streets, D. G.; Zhang, Q. Comparison of adjoint and analytical Bayesian inversion methods for constraining Asian sources of carbon monoxide using satellite (MOPITT) measurements of CO columns. J. Geophys. Res. 2009, 114, D04305, DOI: 10.1029/2007JD009264. (77) Kurokawa, J.; Yumimoto, K.; Uno, I.; Ohara, T. Adjoint inverse modeling of NOx emissions over China using satellite observations of NO2 vertical column densities. Atmos. Environ. 2009, 43, 1878–1887.

CRITICAL REVIEW

(78) Zhao, C.; Wang, Y. Assimilated inversion of NOx emissions over East Asia using OMI NO2 column measurements. Geophys. Res. Lett. 2009, 36, L06805, DOI: 10.1029/2008GL037123. (79) Millet, D. B.; Jacob, D. J.; Boersma, K. F.; Fu, T.-M.; Kurosu, T. P.; Chance, K.; Heald, C. L.; Guenther, A. Spatial distribution of isoprene emissions from North America derived from formaldehyde column measurements by the OMI satellite sensor. J. Geophys. Res. 2008, D02307, DOI: 10.1029/2007JD008950. (80) Kim, S. W.; Heckel, A.; Frost, G. J.; Richter, A.; Gleason, J. P.; Burrows, J. P.; McKeen, S.; Hsie, E.-Y.; Granier, C.; Trainer, M. NO2 columns in the western United States observed from space and simulated by a regional chemistry model and their implications for NOx emissions. J. Geophys. Res. 2009, 114, D11301, DOI: 10.1029/ 2008JD011343. (81) U.S. EPA. Guidance on the Use of Models and Other Analyses for Demonstrating Attainment of Air Quality Goals for Ozone, PM2.5, and Regional Haze, EPA 454/B-07-002; U.S. Environmental Protection Agency, Oﬃce of Air Quality Planning and Standards, 2007b. (82) Treatment of data inﬂuenced by exceptional events. Fed. Regist. 2007, 72(55) 13559 13581. (83) U.S. EPA. http://www.epa.gov/ttn/analysis/exevents.htm (accessed September 2011)

9497

dx.doi.org/10.1021/es2022818 |Environ. Sci. Technol. 2011, 45, 9484–9497

POLICY ANALYSIS pubs.acs.org/est

A Biorefinery for Mobility? S. Pacca* and J. R. Moreira University of Sao Paulo, Av. Arlindo Bettio, 1000 Sao Paulo, SP 03828-000 Brazil 55-11-3091-8173 ABSTRACT: Biofuels are considered as a carbon neutral alternative to hydrocarbons in the transport sector and this approach has triggered concerns about the impact the production of biofuels might have on land usage. Another option that might also lead to reduced emissions in the transport sector is electricity based on renewable energy sources such as biomass. Below, we assess the beneﬁts and drawbacks of the joint production of ethanol and electricity in a sugar cane based reﬁnery, and the use of both energy forms in privately owned automobiles. In this analysis, we have considered technology for energy production that is currently available and cost competitive. The results show that the amount of land that is required to power our current automobile use needs is less than what is typically stated. According to our results that are based on 2010 values, 2 million ha of land are suﬃcient to power the Brazilian automobile ﬂeet, 25 million ha are enough to satisfy the needs of the U.S. ﬂeet, and 67 million ha are suﬃcient to cover the global autofuel requirements. When minor eﬃciency gains are considered, 19 million ha will be enough to satisfy the fuel needs of the U.S. ﬂeet in 2030, whereas land required to supply the Brazilian and global ﬂeet remain basically unchanged. Our analysis shows that the harvested energy density of sugar cane is 306 GJ/ha/yr, which is 1.7 times the value usually reported in the literature for biofuels. As a result, taking advantage of the primary energy potential of sugar cane, only 4% of the world’s available cropland area would be suﬃcient to produce fuels that would power the global car ﬂeet.

1. INTRODUCTION 1.1. Motivation for This Effort. The recent debate surrounding the expansion of biofuels as one of the possible solution to mitigate greenhouse gas emissions provoked several questions directly associated with the land area that would need to be set aside for energy crops. The conflict between food versus fuel and the direct and indirect land use change effects on the carbon balance are two examples of issue involved in this debate. The results of the study described below, involving a bottom up assessment of mobility related fuel needs, demonstrate that the amount of land required to support an efficient, state-of-the-art bioenergy based global car fleet is relatively modest. 1.2. Existing Literature on Biofuels and Land Use Change (LUC) Effects. The carbon mitigation potential of biofuels has been scrutinized by several authors.1 4 Besides life cycle emissions, researchers are concerned about how changes in direct and indirect land use will affect the carbon balance of biofuels. Two of the important issues relate to the use of vast areas to produce biofuels and its effect on the carbon balance,5,6 and the competition involving fuel versus food production.7,8 Recently, these concerns have sparked considerable debate and different types of analyses of the problem. One metric to judge the performance of biofuels is referred to as the ecosystem carbon payback time (ECPT).9 This metric takes into account the comparison of the amount of carbon released as a consequence of the removal of indigenous r 2011 American Chemical Society

vegetation with emission savings caused by the substitution of biofuels for fossil fuels. Thus, lower ECPTs imply more sustainable biofuels. Therefore, ceteris paribus, the higher the energy yield of a feedstock, the lower the ECPT provided that yield enhancement is not a result of massive life cycle fossil fuel inputs, such as chemical fertilizers. Consequently, the quest to uncover feedstocks that have the highest biofuel yields has been at the heart of the strategy to select future biofuel feedstocks.10 In this regard, given the current stateof-the-art, the advantage of sugar cane in comparison with other biomass feedstocks has already been established. Although Borjesson and Tufvesson have stated that sugar beet has the highest biomass energy yield per ha,11 their analysis ignored the fact that sugar cane can be used as a biofuel production feedstock and it did not consider the contribution of crop residues. Moreover, in addition to its high biofuel yield, sugar cane can be utilized to produce electricity that also can be employed to power passenger vehicles.12,13 In the future, biofuel yields might be enhanced by the conversion of cellulose into ethanol.14,15 Although this challenge has been described in several recent reports, the technology is not yet commercially available despite having been conceptually Received: February 10, 2011 Accepted: October 3, 2011 Revised: September 21, 2011 Published: October 03, 2011 9498

dx.doi.org/10.1021/es2004667 | Environ. Sci. Technol. 2011, 45, 9498–9505

Environmental Science & Technology demonstrated more than 50 years ago to be practical in both the laboratory and trial commercial plants. Increasing biofuel output per land area is only part of the solution for increasing energy eﬃciency per unit area of biomass. This represents a supply side based approach whereas several of those conducting energy analyses have argued that demand side management (DSM) is an alternative strategy for rational energy use.16 For example, Romm & Lovins 1993 have suggested that plans that focus on DSM are important in maintaining a competitive economy.17 DSM is based on the service provided by energy consumption and, in the case of biofuels, the service is personal mobility. Therefore, a DSM approach applied to personal mobility would entail scaling back fuel demand while maintaining the same mobility level.18 Inducing society to adopt automobiles that use renewable energy sources in a very eﬃcient way is another important component of the solution. Although it is recognized that public transportation is an excellent alternative for meeting the mobility needs of people, several analysts believe that automobiles will remain as the predominant method for mobility in the near future.19,20 And yet, in 2007 the global ﬂeet of automobiles was 823 million and it is expected that this ﬁgure will reach 1452 million in 2030.21 In the assessment given below, the assumption is made that the land available for bioenergy production is limited and that the major use of alternative liquid fuel based bioenergy will be for transportation. This proposal also is based on the thought that electricity for other end-use sectors besides transport can be generated by using many other renewable energy sources. 1.3. Mobility Per Hectare (ha) and the Optimal Use of Sugar Cane Based Biomass. The debate surrounding the impact of biofuel production on land use entails the issue of the energy production capacity of land, which we refer to as power density (GJ/ha/yr or W/ha). In contrast, personal transportation needs are measured in terms of distance. Therefore, an efficient personal transportation technology would be aimed at using a minimal amount of energy per km driven by an automobile. In terms of the mobility per ha indicator, expressed in km/ha/yr, the goal is maximization of the power density of the bioenergy crop and the minimization of the energy consumed per km. A variation in this indicator is expressed in terms of vehicles per ha if an average distance driven per year per automobile is adopted. Thus, instead of only looking for biofuels that have greater yields or for new ways to increase ethanol production, an alternative approach would be to develop combined technologies that lead to a greater mobility per ha of land. As a result, our analysis focuses on the mobility potential per unit of land area because, historically, land has been used to meet various competing human needs. The overall goal is to maximize the distance driven in private automobiles in one year based on the biomass energy produced from 1 ha of land on which sugar cane is grown. Importantly, a sugar cane bioreﬁnery produces both ethanol and electricity, energy forms that can be used to power automobiles. The advantage of reaching this goal is that both the end use technology and the bioreﬁnery are already commercially available and cost competitive. Moreover, their large scale deployment is quite realistic. At the current time, commercial hybrid automobiles are powered by an internal combustion engine and an electric motor. These are highly eﬃcient vehicles that might be fueled with ethanol. In comparison, electric vehicles, which store electricity in batteries, are becoming available. At the same time, modern

POLICY ANALYSIS

sugar mills in Brazil are producing 90 L of ethanol (1.98 109J) and exporting 125 kWh of electricity (4.5 108 J) per tonne of processed sugar cane, and in 2007 the average sugar cane yield in southeast of Brazil was 85 tonnes per hectare.22 Thus, a state-ofthe-art sugar mill is a highly eﬃcient energy source because it coproduces from a common feedstock 7650 L of ethanol (1.68 1011 J) and 10.6 MWh (3.8 1010 J) of surplus electricity per ha per year as well as the steam demands of the mill. The adoption of 1ha as a basic unit for the evaluation of diﬀerent bioenergy production technologies and the maximum energy yield of diﬀerent energy forms has been proposed previously. In 2000, Bystricky et al, compared diﬀerent feedstock options in Germany and accounted for their potential for electricity and heat production in the form of biogas and fuel used for transportation.23 Nevertheless, this study did not consider the use of electricity and fuel for transportation but, instead, focused on a ﬁxed energy output in the form of only electricity, heat and fuel, and evaluated the potential of diﬀerent biomass sources that could be employed for these three purposes. In addition, a large-scale corn based biogas production facility was the option the authors believed would have the least GHG emissions but a sugar cane-based facility was ignored. In their assessment, biomass energy needs to be complemented by traditional sources to meet the desired combined annual output of roughly 273 GJ/ha. An alternative assessment of the land area required to fulﬁll the needs of the transportation sector has considered synthetic oil, based on the gas to liquid (GTL) conversion processes.24 While production of 13.8 million barrels per day of biofuel (4.8 1016 J) requires 530 million ha of land, the production of the same volume of synthetic oil, using photovoltaic modules, biomass and gasiﬁcation, requires 107 million ha. However, the technologies required for this purpose are not yet commercially available and the cost associated with this approach is still prohibitive. In contrast, bioreﬁneries are currently producing ethanol and electricity from bagasse22,25,26 and the use of bioelectricity and an assessment of its carbon balance when used as a transportation fuel have been already evaluated.12,13 The potential of energy derived from sugar cane as a carbon mitigation option is based on the fact that it comes in the form of ethanol and bioelectricity. According to the Directive 2009/28 of the European Union, the use of sugar cane ethanol avoids 71% of the CO2 emissions in comparison to typical liquid fuels.27 On average, liquid fuels used in Europe emit 83.9 g of CO2 eq per MJ. In addition, the carbon mitigation potential of sugar cane would be enhanced if bioelectricity production is maximized and its use displaces fossil fuels. The eﬀectiveness of bioelectricity in the transportation sector as a carbon mitigation option depends on the fuel mix that is used to generate electricity.13 However, if a business strategy is implemented that guarantees that electricity would be used to fulﬁll transportation needs, bioelectricity would become an attractive replacement for gasoline. In the approach we have devised, we have evaluated the possibility of displacing gasoline by the two sugar cane products: ethanol and bioelectricity. Based on this approach, two schemes are possible for this biomass based system. First, in the short run, ethanol can be directly consumed in internal combustion engines and bioelectricity can be consumed by electric vehicles (EV) that are leased from bioelectricity suppliers. This is not the ﬁrst time that the latter approach has 9499

dx.doi.org/10.1021/es2004667 |Environ. Sci. Technol. 2011, 45, 9498–9505

Environmental Science & Technology

POLICY ANALYSIS

Figure 1. Energy ﬂows of a sugar mill when processing one tonne of sugar cane.

been used because the EV1 was leased by the General Motors Corporation in the 90s. It should be mentioned that electric vehicles have not become popular in the automobile market owing to diﬃculties associated with battery technology (eg., toxicity, loss of charge, cost, lifetime). A second scheme entails the use of plug-in hybrid vehicles, which would consume both ethanol and bioelectricity derived from sugar cane. However, the electricity storage capacity of these vehicles must be at a level that would enable them to rely on bioelectricity 38% of the time in order for them to be compatible with an optimal use of both sugar cane derived energy forms. If either the ﬁrst or second business strategies were adopted, public policies would help tailor the demand to the optimized supply of biofuel and bioelectricity. Tying bioelectricity supply with the electricity demands of personal automobiles is not important but the crucial issue is that bioelectricity would be delivered to the grid in order to avoid the need to create new electricity supply systems. This proposal represents an alternative to recent ones concerning the coproduct allocation of biomass products,26 which is typically based on their energy content, economic value, mass, and other criteria. The diﬃculty with this classical approach is that all of the allocation methods have limitations because the characteristics and uses of the coproducts make comparisons diﬃcult. However a sugar cane bioreﬁnery is unique because it produces ethanol and electricity as its energy products. Therefore, using sugar cane as a source for ethanol and electricity for the same ﬁnal use, which is personal mobility, facilitates accounting for the full energy content of sugar cane, its carbon mitigation potential, and resource constraints. Another DSM related feature of a sugar cane bioreﬁnery is the rational use of feedstock energy. Energy eﬃciency is enhanced by reducing energy losses in the conversion process. For example, the exclusive production of electricity in thermal power plants always produces signiﬁcant amounts of waste heat, the use of which would improve the eﬃciency of this energy conversion system.28 Globally, the amount of conversion losses is 1.7 times greater than that of gross electricity production. In temperate countries, this heat is partially used for heating purposes in

inclement months by employing cogeneration schemes.29 31 However, the demand for heat energy does not exist in tropical countries, unless it is employed as a power input in industrial processes. This is the case in sugar mills where heat is an important input for the joint production of sugar and ethanol. In fact, ethanol processing that takes place in chemical plants requires heat that is easily obtained by cogeneration schemes. As a gross comparison, most sugar mills consume 450 kg of steam and 20 kWh (7.2 107J) of electricity per tonne of processed sugar cane. The saturated steam needed for this purpose has a pressure in the range of 10 bar and, consequently, an enthalpy of 2.8 MJ/kg. In comparison, electricity produced in a thermal power plant operating at 35% eﬃciency requires 10.3 MJth/kWh. Thus, the total energy consumption of a sugar mill is 1190 MJth as heat and 206 MJth as electricity per tonne of processed sugar cane. This comparison demonstrates that ethanol and sugar processing are signiﬁcant heat sinks and very modest electricity consumers. Thus, the share of harvested primary energy that is used in the conversion of sugar cane to its useful energy forms is signiﬁcant. In Figure 1 is shown a complete energy ﬂow diagram of the operation of a sugar mill. The total useful energy produced based on the primary energy of 1 tonne of sugar cane harvested is 3.62 MJ/tcane or 308 GJ/ha, which gives an eﬃciency of 65%. As the data in Figure 1 show, the quantitative overall eﬃciency for external uses (ethanol + exported electricity) is 45.3%. 1.4. Land Use Demand and Bioenergy for Transport. The analysis presented above is based on arable land area. Because land is a finite resource, its use for cultivation of an energy crop should be sustainable and efficient.23 Thus, the land required to supply our biofuel needs should be considered in a quantitative way and minimized.32 As a matter of fact, the land required to power vehicles and other global energy demands has been assessed in several recent studies and various authors have addressed the issue of land demand for energy, biofuels, and bioethanol production. About 850 million ha are required to meet the present global energy needs of our society given the state-of-the-art technologies that are available for biomass energy conversion.33 It is estimated that 1540 million ha of cropland are present worldwide, 9500

dx.doi.org/10.1021/es2004667 |Environ. Sci. Technol. 2011, 45, 9498–9505

Environmental Science & Technology of which 904 million ha are located in developing countries. A recent long-term forecast by the FAO expects that global agricultural production will rise by 1.5% per year over the next three decades, a rate that is signiﬁcantly larger than that of the projected population growth.34 Thus, competition between food, feed, and fuels could be avoided in the future by increasing the production of biomass for energy in a balanced fashion through improvement of agricultural management.35 Assuming that an average yield of biomass dedicated land is 50 (dry) tonnes per ha and that 75% of the biomass is converted to biofuels, the land area required to displace all oil consumed worldwide in 2008 would be 100 million ha.36 This is the most optimistic scenario, because on the other extreme the author claims that 3000 million ha would be required if a 5 tonne biomass per ha yield is considered and only 25% of the biomass is converted into biofuels. Of course, biomass can be produced on abandoned crop land, which along with pasture land is estimated to be between 385 and 472 million ha worldwide. Thus, it should be possible to produce 32 41 EJ of primary energy on this land area or 7 8% of the global primary energy demand.8 In a previous assessment, it was estimated that the production of biomass on abandoned land would be suﬃcient to meet only 5% of the world’s primary energy needs in 2006.6 In order to produce 35 billion gallons (133 109 liters) of ethanol in 2017, an amount recommended by the U.S. Advanced Energy Initiative, it is expected that 31 million hectares need to be cultivated, if corn grain is used as the source, and 47.2 million hectares if corn stover is used.14 In another assessment, in 2008 Groom et al. estimated that in order to meet 50% of the U.S. transportation fuel demand (550 billion liters of gasoline per year in 2006) between 290 and 485 million ha would be required if corn is the ethanol source, whereas between only 85 to 105 million ha are needed if sugar cane is used.32 A European assessment indicates that the land available for nonfood crops in 2010, 2020, and 2030 is 13.2, 20.5, and 26.2 million ha, respectively.37 The increases during these periods are a consequence of several factors including fallow land, food crop yield increases, changes in the export/import balance, and population decline. Another evaluation showed that between 18.5 and 21.1 million ha are required to meet the 2020 biofuel target of the European Union.38 Thus, the available land area in Europe is suﬃcient to meet the EU biofuel target, which requires that only 10% of fossil fuels dedicated to transportation derive from renewable sources (including biofuels) by the year 2020.39 The transportation fuel requirements for 27 EU countries are expected to be 147 billion liters of gasoline and 265 billion liters of diesel fuel by 2020.40 Actually, EU countries are already replacing approximately 3% of traditional liquid fuels by those coming from renewable sources, mainly in the form of biofuels. The demand for biofuels by 2020 represents an additional 5% fossil fuel replacement.40 Thus, 7.5 and 13.5 billion liters of gasoline and diesel fuel must be replaced by biofuels, which when diﬀerences in heat contents are considered equates to an extra demand of 10 billion liters of ethanol and 25 billion liters of biodiesel. In 2009, Tian et al. projected that 6.3 million ha of land will be available in China for cultivation of nonfood energy crops without compromising food safety.41 In comparison, 15.8 million hectares will be available for bioethanol production in Brazil by 2020, again without compromising increasing food production.42 These values should be considered in light of the fact that the

POLICY ANALYSIS

Table 1. Land Area Required to Meet Global Primary Energy Demand reference

area (ha) 8.5 108

Lal 2010 Henry 2010 Campbell et al. 2008

3.1 108 to 9.4 109 4.8 109 to 5.9 109

Field et al. 2008b

8,6 109

a

a,c

Heaton et al. 2008b

a,c

Groom et al. 2008 € Ozdemir et al. 2009a,d

3.2 109 1 109 to 5.8 109 4.6 108 to 5.2 108

a

Assuming that oil consumption corresponds to 33.2% of total energy consumption. b Assuming that abandoned agriculture land area is similar to value presented by Campbell et al. 2008. c Assuming that demand in the U.S. is 50% of the global gasoline consumption, and gasoline is 50% of the U.S. oil consumption. d Based on EU 2020 biofuels target; biofuels displace 55% of diesel; diesel consumption in EU accounts for 26% of global consumption; diesel consumption accounts for 28% of oil consumption.

world’s total cropland (arable land and permanent crops) area is 1527 million hectares.43 In summary, most studies indicate that vast land areas are required to meet future global energy needs, and yet, the land available for energy production is limited. The land areas normalized by global primary energy needs suggested by several authors are given in Table 1. Note that the land area required to meet the global energy demand is between 300 and 9000 million ha. However, our assessment indicates that these projections of massive land areas needed to meet our global energy needs are too high when optimization of sugar cane based energy production and automobile ﬂeet are taken into account.

2. MATERIALS AND METHODS Several assumptions are made in our assessment. First, the assessment is based on utilizing a sugar cane mill for energy production, a hybrid automobile powered by an internal combustion engine that consumes ethanol with an eﬃciency of 15 km/liter, which corresponds to the eﬃciency of a commercially available automobile assuming 19.55 km per liter using gasoline as the fuel. Second, we assume that the fuel performance (measured as km/L) of the ethanol engine is 75% of that of a gasoline powered engine. Third, the assessment is based on using an electric vehicle with a 6.5 km/kWh eﬃciency.12 Finally, instead of having the internal combustion and electric engines in two diﬀerent automobiles, we propose having one plug-in hybrid automobile that operates utilizing these two technologies. This is the operational principle for hybrid plug-in vehicles, which have begun being commercialized since the beginning of 2011. Furthermore, we assume that the average yield of a sugar cane plantation is 85 tonnes per ha and that 90 L of ethanol are produced per tonne of sugar cane. In addition, we assume that the bagasse byproduct will be used as feedstock in boilers to produce electricity and heat. In order to achieve a surplus of 125 kWh of electricity per tonne of sugar cane, we require that 50% of the sugar cane straw is also harvested and, together with the bagasse byproduct, is used for power production. The electricity conversion data, arising from the analysis presented by Campbell et al. in 2009, is then employed to determine the fuel value of electricity. 9501

dx.doi.org/10.1021/es2004667 |Environ. Sci. Technol. 2011, 45, 9498–9505

Environmental Science & Technology

POLICY ANALYSIS

Figure 3. Number of cars, amount of electricity equivalent required for EVs, amount of ethanol, and electricity necessary for hybrid ﬂeet, number of ICV cars and of EVs in hybrid ﬂeet, and sugar cane area. Figure 2. Number of cars per hectare of harvested sugar cane.

The production of sugar cane, having the performance data listed above and an end-use energy proﬁle that includes ethanol and electricity, is then placed in the context of replacing fossil fuels as the energy source for automobiles. Consequently, the same approach used by Campbell et al in 2009 can be applied to determine the land required to supply a suﬃcient quantity of fuel to power the automobile ﬂeets in the U.S. and Brazil. These two countries are used for this case study because the U.S. has the largest gasoline powered automobile ﬂeet in the world and Brazil is the largest sugar cane derived bioethanol producer globally. The U.S. automobile ﬂeet in 2007 contained 222.2 million vehicles whereas at the same time in Brazil the number was 30.3.21 In 2007, each automobile in the U.S. was driven 19 801 km with an average eﬃciency of 9.6 km per liter.44 We have adjusted the sizes based on OPEC’s projections to determine the 2010 ﬂeet sizes that we project to be 229 million automobiles in the U.S. and 33 million in Brazil. Therefore, 474 (229 million cars 20 000 km/yr/9.6 km/L) billion liters of gasoline were required to power all vehicles in the U.S. in 2010. The annual distance driven by each vehicle in the U.S. is assumed to be 20 000 km where as for the rest of the world the annual distance driven is assumed to be 12 000 km. In our optimized model we also consider a hybrid automobile fueled with ethanol at an eﬃciency of 15 km/liter, which we believe is expected for an ethanol fueled hybrid automobile constructed using the current technology. The eﬃciency measured as km/L of conventional ﬂex fuel Otto engines, which currently power passenger vehicles commercialized in Brazil, is 1.3 times greater when fueled with gasoline than with ethanol, a consequence of the fact that the heat value of ethanol (21.34 GJ/m3) is lower than that of gasoline (32.22 GJ/m3). However, ethanol engines have higher compression ratios (1:12) and, therefore, the comparative factor based on the distance driven on a volume basis is 1.3.

3. RESULTS The result of our assessment based on the factors summarized above shows that, considering 20 000 km per vehicle per year, 1 ha of land cultivated with sugar cane would suﬃcient to power 9.2 vehicles per year if the hybrid car and the EV ﬂeets are designed to perfectly match the bioethanol and the bioelectricity supplied by processing sugar cane in a modern sugar mill and that

Figure 4. Number of cars, amount of gasoline consumed, amount of ethanol equivalent required for a neat ethanol ﬂeet, ethanol and electricity to power hybrid ﬂeet, and sugar cane area for hybrid ﬂeet.

utilizes a high pressure steam boiler such as those that have been installed recently in Brazil (Figure 2). However, if the hybrid automobiles and electric powered vehicles operate 12 000 km per year, it would be possible to power 11.6 vehicles per year (EV cars in the U.S. = 6.5 km 3 kWh 1 125 kWh 3 tonne 1 85 tonne 3 ha 1 1 car 3 year 1 1/20 000 km = 3.4 car 3 year 1 3 ha 1; EV cars in Brazil = 6.5 km 3 kWh 1 125 kWh 3 tonne 1 85 tonne 3 ha 1 1 car 3 year 1 12 000 km 1 = 5.8 car 3 year 1 3 ha 1). If all automobiles are replaced by electric vehicles, 697 TWh/ year would be required to power the entire ﬂeet in the U.S. and 60 TWh/year would be needed to power the Brazilian ﬂeet (Figure 3). These consumption values are modest when compared to the present electricity use in both countries (3741 TWh,44 and 466 TWh, respectively).45 However, if both ethanol and bioelectricity are used in an optimized proportion, 385 billion liters of ethanol and 262 TWh of electricity would be enough to power the U.S. ﬂeet and, in Brazil, 33 billion liters of ethanol and 23 TWh would be suﬃcient to operate the ﬂeet for a one-year period. These energy consumption ﬁgures equate to a U.S. ﬂeet comprised of 143 million hybrid and 86 million electric vehicles. In comparison, the proportional Brazilian ﬂeet would be comprised of 20 million hybrid and 12 million electric vehicles. Alternatively, the ﬂeet in both countries could consist of only 9502

dx.doi.org/10.1021/es2004667 |Environ. Sci. Technol. 2011, 45, 9498–9505

Environmental Science & Technology

Figure 5. Relative global yield increase for selected crops normalized to 1961 values.

hybrid plug-in vehicles that are powered by ethanol and electricity. Assuming that sugar cane grown in Brazil is the feedstock, the land area required to power the ﬂeet in the U.S. equates to 25 million hectares while the land required to produce fuel to power the Brazilian ﬂeet would be 2 million hectares or about one-third of the current area dedicated to ethanol production in Brazil. The same approach can be extended to deriving estimates for fuel production to power the world’s ﬂeet of automobiles, which currently comprises 886 million vehicles.21 For example, fuel in the form of sugar cane bioenergy could be used to power a hybrid ﬂeet comprised of ethanol fueled and plug in electric vehicles. In order to power the world’s 886 million automobile ﬂeet, 1299 billion liters of gasoline are currently required (Figure 4). In contrast, 1688 billion liters are required If pure ethanol is used as a gasoline substitute. However, if both ethanol and electricity are used in a tandem manner in hybrid plug-in vehicles, 1055 billion liters of ethanol and 717 TWh of electricity would be required to power this automobile ﬂeet, which would come from 67 million hectares cultivated with sugar cane. An extension of this analysis to 2030 includes the assumption that eﬃciencies for sugar cane, ethanol, and electricity production will increase by 1% per year. Historically, the global sugar cane yield has increased at a rate of 0.9% over the last 50 years, whereas increases in the annual yield rate increments of corn, rice, and soybeans were 3.4%, 2.6%, and 2.3%, respectively.33 According to this scenario, it would be possible to power 15 automobiles using one hectare of land planted with sugar cane. We assumed that the automobile ﬂeet in the U.S., Brazil, and globally will increase at respective annual rates of 1, 2.5, and 3%.21 Based on these values, it should be possible for 19 million ha to produce a suﬃcient fuel quantity to power the U.S. ﬂeet in 2030 and only 2 million ha would be required to power the Brazilian ﬂeet. Finally, the area required to power the global ﬂeet would be 66 million ha or 4% of the present total cropland area. It should be noted that the yields of other crops have increased at greater rates than that of sugar cane. Therefore, it is expected that sugar cane has a signiﬁcant yield improvement potential.46,47 In Figure 5 are shown the FAO data derived yield growth of various crops relative to that in 1961. If over the past 30 years the annual rate of sugar cane’s yield had increased by greater than 1%, the total increase would still be less than for other crops. Importantly, more eﬃcient production of sugar cane would lead

POLICY ANALYSIS

to reduction in the area required to power the global automobile ﬂeet in 2030 giving a value of 38.5 ha. Even considering the fact that electricity produced in Brazil cannot be transported to the U.S., it is possible to conceive of a compensation system in which bioelectricity is produced and consumed in Brazil for other uses such as residential, industrial, commercial, and public purposes. This bioelectricity could replace natural gas that is consumed in Brazil, which would then be exported to the U.S. and utilized for electricity generation. Taking into consideration the fact that no natural gas shortage exists in the U.S., there remains great interest in extending the lifetime of fossil fuel reserves. Therefore, by employing this socalled swap plan emissions from electricity production in the U.S. would be indirectly mitigated in Brazil, where the bioenergy is available. Another possible mechanism to compensate for ethanol production in Brazil and not the U.S. would be through the use of 100% electric vehicles in Brazil (6.5 km/kWh) and hybrid ethanol fueled vehicles in the U.S. (15 km/L). An interesting conclusion can be formulated by utilizing the scenario that (1) the annual driving distance per vehicle in the U.S. is 20 000 km and that it is 12 000 km in the rest of the world, (2) 2010 state-ofthe-art technology is employed for ethanol and electricity production in sugar cane mills (90 L/tonne; 125 kWh/tonne; 85 tonnes/ha), and (3) 2008 state-of-the-art end use technology is employed for automobiles. Considering these inputs, the projection is that the use of one electric vehicle in Brazil would be equivalent to the use of one ethanol vehicle in the U.S. Thus, it would be possible to power 5.8 vehicles with each one of the technologies (that is 11.6 cars/ha) based on the output of 1 ha (ethanol fueled cars in the U.S. = 15 km 3 liter 1 90 L 3 tonne 1 85 tonnes 3 ha 1 1 car 3 year 1 1/20 000 km = 5.8 car 3 year 1 3 ha 1; EV cars in Brazil = 6.5 km 3 kWh 1 125 kWh 3 tonne 1 85 tonne 3 ha 1 1 car 3 year 1 1/12 000 km =5.8 car 3 year 1 3 ha 1). Reaching the goals we have set requires the use of less than 70 million ha to power either the 2010 or the 2030 global automobile ﬂeet, an easily obtainable target when surplus arable land and pasture is included for the production of sugar cane. Expansion of sugar cane production to pasture land is currently taking place in Brazil with little if any environmental, social, or economic impact.42 Another remarkable conclusion can be made about the ﬁnal useful energy that can be obtained from sugar cane. If optimized production of ethanol and electricity are attained, a total of 1055 billion liters of ethanol and 717 TWh would be derived from 68 million ha or about 306 GJ/ha/yr of secondary energy (ethanol and steam) before transformation into electricity. This is a high yield for biomass. The Blue Map scenario of the World Energy Outlook 2008, forecasts the use of 29.8 EJ of liquid biofuels based on ethanol from sugar cane, corn, lignocelluloses, oil seed biodiesel, and biomass to liquid (BTL) conversion processes, generated using an area of 160 Mha.48 This value corresponds to an energy density of only 186 GJ/ha/yr. The diﬀerence between this projection and our assessment is a consequence of two factors. First, our plan is based on sugar cane derived fuel and that the yield of sugar cane is much higher than that of other ethanol feedstocks (corn, wheat, even some short rotation crops). Second, electricity coproduced from the processing of sugar cane into ethanol is a component of our plan. In arriving at the present assessment, we are concerned with several issues including carbon mitigation, land availability, and commercially available technologies. Regarding carbon 9503

dx.doi.org/10.1021/es2004667 |Environ. Sci. Technol. 2011, 45, 9498–9505

Environmental Science & Technology mitigation a recent study performed by EPA49 calculated the amount of GHGs emissions per unit of energy for the production of ethanol from sugar cane in Brazil. The lifecycle assessment study took into account direct (LUC) and indirect land use change (ILUC) and concludes that ethanol produced in Brazil and transported to the U.S. can oﬀset 60% of the emissions compared with gasoline or even 90% if the sugar mill performs bioelectricity generation eﬃciently. Regarding land availability the recent IPCC SRREN50 recognizes that land availability may not be a serious issue if proper policies and biofuel expanded production is well managed. Although we cannot identify at this point any drawbacks related to the use of sugar cane as a feedstock for bionergy production, we do recognize that in a broader assessment one or more issues might require consideration and that they might have a potentially negative impact on our overall conclusions. We want to emphasize that we are not proposing that all sugar cane required to fulﬁll the current mobility scenario be produced in one country. On the contrary, production needs to be spread over several countries that have suitable conditions for cultivation,51 and in which environmental and social impacts are minimized. Institution of the proposed schemes requires the development of new policies and regulation on both the supply and demand side. Actions that might be required on the supply side include capacity and institutional building in potential sugar cane producing countries, technology transfer covering biomass and automobile manufacturing, removal of ethanol trade barriers, and the operation of independent power producers. The demand side might require actions that include enhancement of public awareness concerning climate change and the beneﬁts of using sugar cane based energy carriers for transport, and the institution of economic policies that support the acquisition of hybrid vehicles and EVs.

4. DISCUSSION In the assessment presented above, we demonstrated that 67 million hectares of land are suﬃcient to power all of the world’s current automobiles. In addition, assuming that the ﬂeet is 60% larger than the current one and that eﬃciency gains will occur, we forecast that in 2030 66 million hectares will be enough land to produce fuel to power all automobiles in the world. Therefore, if the appropriate technology is both developed and applied, biomass based energy could replace fossil fuels using a modest amount of land. That such a small land area would be required for this purpose is remarkable because food versus fuel competition and land use changes are frequently the most troubling issues that serve as potential barriers for the widespread use of ethanol as a replacement for liquid fossil fuels to power automobiles. Thus, our assessment suggests that it is important to consider land availability under realistic biofuel scenarios and, when this is done, it demonstrates that land is not a limitation for biofuel production. ’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

POLICY ANALYSIS

’ ACKNOWLEDGMENT S.P. is grateful to the “Conselho Nacional de Desenvolvimento Cientíﬁco e Tecnologico (CNPq)” for the start up grant n° 555731/ 2010 for the life cycle assessment (LCA) laboratory at the “Instituto de Eletrotecnica e Energia” of the “Universidade de Sao Paulo”, Brazil. ’ REFERENCES (1) Fargione, J.; Hill, J.; Tilman, D.; Polasky, S.; Hawthorne, P. Land clearing and the biofuel carbon debt. Science 2008, 319, 1235–1238. (2) Searchinger, T.; Heimlich, R.; Houghton, R. A.; Dong, F.; Elobeid, A.; Fabiosa, J.; Tokgoz, S.; Hayes, D.; Yu, T.-H. Use of U.S. croplands for biofuels increases greenhouse gases through emissions from land-use change. Science 2008, 319, 1238–1240. (3) Searchinger, T. D.; Hamburg, S. P.; Melillo, J.; Chameides, W.; Havlik, P.; Kammen, D. M.; Likens, G. E.; Lubowski, R. N.; Obersteiner, M.; Oppenheimer, M.; Philip Robertson, G.; Schlesinger, W. H.; David Tilman, G. Fixing a critical climate accounting error. Science 2009, 326, 527–528. (4) Hertel, T. W.; Golub, A. A.; Jones, A. D.; O’Hare, M.; Plevin, R. J.; Kammen, D. M. Eﬀects of US maize ethanol on global land use and greenhouse gas emissions: Estimating market-mediated responses. BioScience 2010, 60, 223–231. (5) Heaton, E. A.; Dohleman, F. G.; Long, S. P. Meeting US biofuel goals with less land: The potential of Miscanthus. Global Change Biol. 2008, 14, 2000–2014. (6) Field, C.; Campbell, J.; Lobell, D. Biomass energy: The scale of the potential resource. Trends in Ecol. Evol. 2008, 23, 65–72. (7) Chakravorty, U.; Hubert, M.-H.; Nøstbakken, L. Fuel Versus Food. Annu. Rev. Resour. Econ. 2009, 1, 645–663. (8) Campbell, J. E.; Lobell, D. B.; Genova, R. C.; Field, C. B. The global potential of bioenergy on abandoned agriculture lands. Environ. Sci. Technol. 2008, 42, 5791–5794. (9) Gibbs, H. K.; Johnston, M.; Foley, J. A.; Holloway, T.; Monfreda, C.; Ramankutty, N.; Zaks, D. Carbon payback times for crop-based biofuel expansion in the tropics: The eﬀects of changing yield and technology. Environ. Res. Lett. 2008, 3, 034001. (10) Worldwatch Institute. Biofuels for transport: Global potential and implications for sustainable energy and agriculture; Earthscan: London; Sterling VA, 2007. (11) B€ orjesson, P.; Tufvesson, L. M. Agricultural crop-based biofuels —Resource eﬃciency and environmental performance including direct land use changes. J. Clean. Prod. 2011, 19, 108–120. (12) Campbell, J. E.; Lobell, D. B.; Field, C. B. Greater transportation energy and GHG oﬀsets from bioelectricity than ethanol. Science 2009, 324, 1055–1057. (13) Lemoine, D. M.; Plevin, R. J.; Cohn, A. S.; Jones, A. D.; Brandt, A. R.; Vergara, S. E.; Kammen, D. M. The climate impacts of bioenergy systems depend on market and regulatory policy contexts. Environ. Sci. Technol. 2010, 44, 7347–7350. (14) Heaton, E. A.; Dohleman, F. G.; Long, S. P. Meeting US biofuel goals with less land: The potential of Miscanthus. Global Change Biol. 2008, 14, 2000–2014. (15) Milliken, J.; Joseck, F.; Wang, M.; Yuzugullu, E. The advanced energy initiative. J. Power Sources 2007, 172, 121–131. (16) Cullen, J. M.; Allwood, J. M. The eﬃcient use of energy: Tracing the global ﬂow of energy from fuel to service. Energy Policy 2010, 38, 75–81. (17) Romm, J. J.; Lovins, A. B. Fueling a competitive economy. Foreign Aﬀairs 1992, 71, 46–62. (18) Sager, J.; Apte, J. S.; Lemoine, D. M.; Kammen, D. M. Reduce growth rate of light-duty vehicle travel to meet 2050 global climate goals. Environ. Res. Lett. 2011, 6, 024018. (19) He, K.; Huo, H.; Zhang, Q.; He, D.; An, F.; Wang, M.; Walsh, M. P. Oil consumption and CO2 emissions in China’s road transport: 9504

dx.doi.org/10.1021/es2004667 |Environ. Sci. Technol. 2011, 45, 9498–9505

Environmental Science & Technology Current status, future trends, and policy implications. Energy Policy 2005, 33, 1499–1507. (20) Han, J.; Hayashi, Y. Assessment of private car stock and its environmental impacts in China from 2000 to 2020. Transp. Res., Part D: Transp. Environ. 2008, 13, 471–478. (21) OPEC World Oil Outlook 2010; Organization of the Petroleum Exporting Countries: Vienna, Austria, 2010. (22) Macedo, I. C.; Seabra, J. E. A.; Silva, J. E. A. R. Green house gases emissions in the production and use of ethanol from sugarcane in Brazil: The 2005/2006 averages and a prediction for 2020. Biomass Bioenergy 2008, 32, 582–595. (23) Bystricky, M.; Kn€odlseder, T.; Weber-Blaschke, G.; Faulstich, M. Comparing environmental impacts of electricity, heat and fuel from energy crops: Evaluating biogas utilization pathways by the basket of beneﬁt methodology. Eng. Life Sci. 2010, 10, 570–576. (24) Agrawal, R.; Singh, N. R.; Ribeiro, F. H.; Delgass, W. N. Sustainable fuel for the transportation sector. Proc. Nat. Acad. Sci. 2007, 104, 4828–4833. (25) Pacca, S.; Moreira, J. R. Historical carbon budget of the brazilian ethanol program. Energy Policy 2009, 37, 4863–4873. (26) Wang, M. Huo, H.; Arora, S. Methods of dealing with coproducts of biofuels in life-cycle analysis and consequent results within the U.S. context. Energy Policy 2010. (27) EU DIRECTIVE 2009/28/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 23 April 2009 on the promotion of the use of energy from renewable sources and amending and subsequently repealing Directives 2001/77/EC and 2003/30/ EC;2009. (28) Combined Heat and Power; Evaluating the Beneﬁts of Greater Global Investment; IEA: Paris, 2008. (29) Streckiene, G.; Martinaitis, V.; Andersen, A. N.; Katz, J. Feasibility of CHP-plants with thermal stores in the German spot market. Appl. Energy 2009, 86, 2308–2316. (30) Torchio, M. F.; Genon, G.; Poggio, A.; Poggio, M. Merging of energy and environmental analyses for district heating systems. Energy 2009, 34, 220–227. (31) Amiri, S.; Trygg, L.; Moshfegh, B. Assessment of the natural gas € potential for heat and power generation in the County of Osterg€ otland in Sweden. Energy Policy 2009, 37, 496–506. (32) Groom, M. J.; Gray, E. M.; Townsend, P. A. Biofuels and biodiversity: Principles for Creating Better Policies for Biofuel Production. Conserv. Biol. 2008, 22, 602–609. (33) Lal, R. Managing soils for a warming earth in a food-insecure and energy-starved world. J. Plant Nutr. Soil Sci. 2010, 173, 4–15. (34) World Bank Global Economic Prospects 2009: Commodities at the Crossroads; World Bank: Washington, DC, 2008; p 196. (35) Dornburg, V.; van Vuuren, D.; van de Ven, G.; Langeveld, H.; Meeusen, M.; Banse, M.; van Oorschot, M.; Ros, J.; Jan van den Born, G.; Aiking, H.; Londo, M.; Mozaﬀarian, H.; Verweij, P.; Lysen, E.; Faaij, A. Bioenergy revisited: Key factors in global potentials of bioenergy. Energy Environ. Sci. 2010, 3, 258. (36) Henry, R. J. Evaluation of plant biomass resources available for replacement of fossil oil. Plant. Biotechnol. J. 2010, 8, 288–293. (37) Krasuska, E.; Cadorniga, C.; Tenorio, J. L.; Testa, G.; Scordia, D. Potential land availability for energy crops production in Europe. Biofuels, Bioprod. Bioreﬁn. 2010, 4, 658–673. € (38) Ozdemir, E. D.; H€ardtlein, M.; Eltrop, L. Land substitution eﬀects of biofuel side products and implications on the land area requirement for EU 2020 biofuel targets. Energy Policy 2009, 37, 2986–2996. (39) Impacts of the EU biofuel target onagricultural markets and land use: A comparative modelling assessment; EUR—Scientiﬁc and Technical Research series; Joint Research Centre—Eurpopean Commission: Luxembourg, 2010. (40) The Impact of a Minimum 10% Obligation for Biofuel Use in the EU-27 in 2020 on Agricultural Markets; Impact Assessment of the Renewable Energy Roadmap—March 2007; European Commission—Directorate General for Agriculture and Rural Development: Brussels, 2007.

POLICY ANALYSIS

(41) Tian, Y.; Zhao, L.; Meng, H.; Sun, L.; Yan, J. Estimation of un-used land potential for biofuels development in (the) People’s Republic of China. Appl. Energy 2009, 86, S77–S85. (42) Gauder, M.; Graeﬀ-H€onninger, S.; Claupein, W. The impact of a growing bioethanol industry on food production in Brazil. Appl. Energy 2011, 88, 672–679. (43) FAOSTAT http://faostat.fao.org/site/405/default.aspx (accessed May 28, 2011). (44) EIA-USDOE. Annual Energy Review 2009; Energy Information Administration—U.S. Department of Energy: Washington, DC, 2010. (45) EPE. Brazilian Energy Balance; Empresa de Pesquisa Energetica: Rio de Janeiro, 2010; p 276. (46) Eaton, J. M.; McGoﬀ, N. M.; Byrne, K. A.; Leahy, P.; Kiely, G. Land cover change and soil organic carbon stocks in the Republic of Ireland 1851 2000. Clim. Change 2008, 91, 317–334. (47) Tobias, C. M.; Sarath, G.; Twigg, P.; Lindquist, E.; Pangilinan, J.; Penning, B. W.; Barry, K.; McCann, M. C.; Carpita, N. C.; Lazo, G. R. Comparative genomics in switchgrass using 61,585 high-quality expressed sequence tags. Plant Genome J. 2008, 1, 111. (48) International Energy Agency. World Energy Outlook 2008; International Energy Agency/Turpin Distribution: Paris/New Milford CT, 2008. (49) EPA Renewable Fuel Standard Program (RFS2) Regulatory Impact Analysis; Environmental Protection Agency: Washington, DC, March 26, 2010; p 1120. (50) Chum, H. Faaij, A. Moreira, J. R. Berndes, G. Dhamija, P. Dong, H. Gabrielle, B. Goss Eng, A. Lucht, W. Mapako, M. Masera Cerutti, O. McIntyre, T. Minowa, T.; Pingound, K. Bioenergy. In IPCC Special Report on Renewable Energy Sources and Climate Change Mitigation; Edenhofer, O. Pichs-Madruga, R. Sokona, Y. Seyboth, K. Matschoss, P. Kadner, S. Zwickel, T. Eickemeier, P. Hansen, G. Schlomer, S.; von Stechow, C., Eds.; Cambridge University Press: Cambridge, UK, 2011. (51) Fischer, G. Global Agro-Ecological Assessment for Agriculture in the 21st Century: Methodology and Results; International Institute for Applied Systems Analysis/Food and Agriculture Organization of the United Nations: Laxenburg Austria/Rome, 2002.

9505

dx.doi.org/10.1021/es2004667 |Environ. Sci. Technol. 2011, 45, 9498–9505

POLICY ANALYSIS pubs.acs.org/est

Economic Incentives and Regulatory Framework for Shale Gas Well Site Reclamation in Pennsylvania Austin L. Mitchell and Elizabeth A. Casman* Department of Engineering and Public Policy, Carnegie Mellon University, 5000 Forbes Avenue, Baker Hall 129, Pittsburgh, Pennsylvania, 15213 ABSTRACT: Improperly abandoned gas wells threaten human health and safety as well as pollute the air and water. In the next 20 years, tens of thousands of new gas wells will be drilled into the Marcellus, Utica, and Upper Devonian shale formations of Pennsylvania. Pennsylvania currently requires production companies to post a bond to ensure environmental reclamation of abandoned well sites, but the size of the bond covers only a small fraction of the site reclamation costs. The economics of shale gas development favor transfer of assets from large entities to smaller ones. With the assets go the liabilities, and without a mechanism to prevent the new owners from assuming reclamation liabilities beyond their means, the economics favor default on well-plugging and site restoration obligations. Policy options and alternatives to bonding are discussed and evaluated.

T

he emergence of technologies for economic recovery of natural gas from tight shale formations across the U.S. is responsible for a resurgence in domestic natural gas production. Even though the national average wellhead price has dropped by more than two-thirds in three years, shale gas production continues to increase. The Marcellus shale formation underlies numerous Appalachian states and is considered to be the largest gas-bearing shale formation in the U.S. Rapid development of this resource, evidenced by thousands of new wells in the region since the ﬁrst well in 2004, is charting a new course for natural gas supply and utilization in the Northeast. In Pennsylvania, where there are more drilled wells than any other Appalachian state, this development already dwarfs past oil and gas booms in areal extent and production

’ ECONOMIC, ENVIRONMENTAL, AND HUMAN HEALTH RISKS OF IMPROPERLY ABANDONED SHALE GAS WELLS Disturbance of the surface environment and subsurface geological strata is a necessary outcome of shale gas development in Appalachia. Surface disturbance is caused by the construction of well pads, impoundments, access roads, and pipelines. Reclamation of the disturbed surface occurs in two stages. Shortly after a well begins production the size of the well pad is reduced and the impoundment is removed. Full reclamation does not occur until after a well is abandoned (permanently taken out of production) because site access is necessary for routine maintenance and removing produced water (brine that comes up with gas). If a well site is not properly reclaimed after abandonment, the well pad and access roads may cause permanent changes to the r 2011 American Chemical Society

natural environment. The deterioration of erosion control features increases siltation, which results in the loss of nutrient-rich topsoil and increased sedimentation of nearby surface waters, impairing natural habitats of aquatic species.13 Compared to natural forest clearing occurrences (e.g., ﬁre), the recruitment, growth, and mortality rate of native plant species at reclaimed oil and gas well sites in boreal forests was found to be signiﬁcantly worse.4 Without restoration of topsoil and proper revegetation, the regeneration of natural habitat will be delayed and the environmental impacts of forest fragmentation, including loss of biodiversity and introduction of invasive species, will be exacerbated. The adverse eﬀects of forest fragmentation on the nesting success of migratory birds have been documented,5 and the impacts extend to other plant and animal species dependent on shade, humidity, and tree canopy protection characteristic of deep forest environments in the region.6,7 The construction of well pads, water impoundments, and access roads is projected to disturb 129 000310 000 acres of forested land in Pennsylvania.6 In northern Pennsylvania forests, where largest blocks of public forests exist, the potential for lasting forest fragmentation and associated environmental impacts could negatively aﬀect economic interests related to timber management, game, and tourism.7 To reach the Devonian Shale formations, wellbores transect a mile or more of geologic strata, including fresh and saline aquifers and shallow gas-bearing formations. Shale gas wells will need to be plugged to prevent environmental damage Received: June 27, 2011 Accepted: October 10, 2011 Revised: October 4, 2011 Published: October 10, 2011 9506

dx.doi.org/10.1021/es2021796 | Environ. Sci. Technol. 2011, 45, 9506–9514

Environmental Science & Technology

POLICY ANALYSIS

still exists after a well has been plugged and increases with time.9,1416 The impacts and remediation costs resulting from gas migration and groundwater contamination due to failures at unplugged and improperly abandoned gas wells is well documented in Pennsylvania and elsewhere.10,12,1921 Property values can be negatively aﬀected if gas wells contaminate groundwater used for drinking.2224 Moreover, the presence of an improperly abandoned gas well may prevent landowners from using their property for other purposes.25 Stray gas, which is mostly methane, is also a potent source of greenhouse gas emissions.26

Figure 1. Simple representation of shale gas well anatomy. Layers of cement and steel casing are used to isolate production zones from freshwater aquifers. To properly close a shale gas well, the wellhead and steel production casing are removed and cement plugs are installed to prevent ﬂuid movement in the wellbore and annulus. This diagram is not drawn to scale.

caused by the disturbance of the subsurface, namely the movement of oil, gas, and brine to the surface and between geologic formations connected by the wellbore. General plugging procedures in most states, including Pennsylvania, begin with the removal of steel production casing, which extends from the surface to producing formations, for scrap value. Next, a series of cement plugs will be installed in the wellbore to isolate freshwater and saline aquifers and gas producing formations.8 (Figure 1) Unplugged wells may provide a direct pathway to the environment for ﬂuids in the wellbore,9 which results in ecological harm, property damage, and surface and groundwater contamination. Additional pathways in the annulus (an industry term for the space between two concentric objects, such as between the wellbore and casing or between casing and tubing) may develop that would allow oil, gas, and brine to move vertically across geologic formations and contaminate groundwater. Substances dissolved in the brine may include those that occur naturally in the shale formations (some radioactive) and others injected during the hydraulic fracturing process (some toxic). Also upwardly migrating gas, known as stray gas, represents an explosion hazard if not properly vented away from buildings and drinking water wells.1012 The risk that annular pathways will develop increases over time as chemical, mechanical, and thermal stresses causes deterioration of well structures and components. Failure modes of improperly abandoned wells (deﬁned here as nonproducing wells not in compliance with Pennsylvania plugging requirements or inactive status rules) include the formation of cracks in the cement casing or packers, corrosion of steel production casing, faulty valves, and leaking temporary plugs or surface caps.9,1317 Properly performed, the plugging process reinforces existing casing and seals and prevents ﬂuid movement in the wellbore, which may retard the deterioration of vital well components and structures. Therefore, prompt plugging once a shale gas well becomes uneconomic may reduce the risk of negative environmental and human health impacts,13,14 while also avoiding additional plugging costs that may be incurred if the mechanical integrity of a casing has been compromised.18 However, the risk of failures leading to ﬂuid migration pathways

’ THE SAUDI ARABIA OF NATURAL GAS AND THE SWISS CHEESE OF APPALACHIA Approximately 350 000 conventional oil and natural gas wells have been drilled in Pennsylvania since the 1859 discovery of oil in Titusville. 11 Many of these legacy wells that are no longer producing oil or gas were never plugged. Some leak gas, oil, and/or brine into freshwater aquifers and the surface environment.27,28 To remedy this situation, Pennsylvania’s Oil and Gas Act of 1984 required all wells from which economic beneﬁts were accrued after 1979 to be plugged according to the latest standards and the well sites reclaimed by their owners. To promote compliance with this statute and cover the cost in the event of owner insolvency, a bonding requirement was established. In 1985, Pennsylvania started plugging oil and gas wells lacking a legally responsible owner, known as orphan wells, and supported these activities with fees on new oil and natural gas well permits ($200 and $50 per well for the Orphan Well Plugging Fund and Abandoned Well Plugging Fund, respectively), monies collected for regulatory violations, and grants distributed by Pennsylvania’s taxpayer-funded Growing Greener program.29 From 2007 to 2008, the most recent years for which data are available, a total of $1,066,000 in Growing Greener grants were awarded to reclaim orphan and abandoned wells. 30,31 Before the current shale gas boom, the Pennsylvania Department of Environmental Protection (PADEP) estimated that at 2004 funding rates it would take around 160 years to plug all the existing orphan wells in the Commonwealth.11 ’ COSTS OF SITE RESTORATION AND SHALE GAS WELL CLOSURE Pennsylvania’s 1984 Oil and Gas Act deﬁnes a natural gas operator’s drinking water, site restoration, and well closure responsibilities. Once a well is abandoned, the owner has 12 months to properly plug it and restore the well pad to its previous condition. Restoration of the production well pad (which typically covers 13 acres32) may involve regrading of land, removing access roads and impoundments, restoring top soil, planting native ﬂora, or other necessary restoration required for compliance with Pennsylvania’s Clean Streams Law of 1937. Operators must also remove all equipment used in the production of gas as part of the well abandonment process. This equipment includes the production casing (innermost steel casing that extends down to the production zone), Christmas tree (a grouping of pipes, valves, and ﬁttings used to control the ﬂow of gas from a well), dehydrator, compressor, and tank battery. The cost to plug a deep shale gas well has not been formally estimated by the PADEP, however, it is understood that the cost to plug a well depends primarily on its measured depth (full length of wellbore including horizontal portions). Plugging costs 9507

dx.doi.org/10.1021/es2021796 |Environ. Sci. Technol. 2011, 45, 9506–9514

Environmental Science & Technology increase when the condition of the wellbore is poor or access to the site is diﬃcult. For orphan oil and gas wells in Southwestern Pennsylvania, the PADEP estimates the total cost to plug and restore the site of a well approximately 914 m (3000 feet) in depth averages $60,000, but per well reclamation costs have also exceeded $100,000.18 Reclamation costs of wells drilled into the Devonian Shale (Marcellus, Utica, and Upper Devonian), which range from 1524 to 2744 m deep, will be greater because costs are strongly correlated with depth. Using reclamation data from 255 orphan wells in Wyoming, Andersen and Coupal (2009) estimated the relationship between reclamation costs and depth.33 They estimated that total reclamation costs (well plugging, site restoration, and equipment removal) were approximately $34.45 per meter ($10.50 per foot). They also noted that economies of scale exist when more than one well is on each well pad, which is the norm for wells in the Marcellus Shale. Summarizing data from approximately 1000 individual well completion reports catalogued by the Pennsylvania Department of Conservation and Natural Resources,34 the average measured depth of hydraulically fractured shale gas wells completed in Pennsylvania during 2010 was approximately 3254 m (10 675 feet). Thus, for a single well, at $34.45 per meter, the average reclamation cost for a well in the Marcellus Shale will be in the vicinity of $100,000. However, in some cases the costs for plugging and abandonment of a shale gas well in Pennsylvania have been substantially higher. For instance, in 2010, Cabot Oil & Gas Corporation estimated that it spent $2,190,000 to properly abandon three vertical Marcellus Shale gas wells in Susquehanna County, Pennsylvania, about $700,000 per well.35

’ PENNSYLVANIA BONDING REQUIREMENTS ON PRIVATE LANDS DO NOT INCENTIVIZE RECLAMATION Issues of operator insolvency due to the boom and bust cycles of oil and gas development complicate eﬀorts to hold liable parties responsible and provide for timely environmental reclamation. In theory, requiring that operators post bonds prior to drilling bolsters traditional liability rules by incentivizing compliance.36 In Pennsylvania, bonded monies are released one year after the PADEP deems regulatory requirements associated with reclamation have been satisﬁed. If the level of bonding is set less than the associated reclamation costs, companies could be tempted to pursue strategies that avoid their liabilities. Oil and gas bonding requirements vary across states and on federal lands, but most have established minimum bonding levels (blanket or for individual wells).25 In general, the dollar amount of state and federal bonds for oil and gas wells often do not reﬂect expected reclamation costs. The full eﬀect of this imbalance has not yet been felt because oil and gas wells may have long life spans (up to 50 years, which can be prolonged further on paper via regulatory allowances), and bonding requirements are relatively new.36 Pennsylvania’s experience with bonding of coal mining sites may be indicative of what to expect. From 1985 to 1999, bonds for surface mining permits covering approximately 10% of total acreage were forfeited.37 Since the cost to reclaim a mine in most cases was higher than the amount bonded, funding to bring abandoned mine lands into compliance has generally been inadequate.3739 In 1986, only 33% of acreage covered by forfeited bonds had been reclaimed, according to a U.S. General Accounting Oﬃce study. The discrepancy was attributed to inadequate funding from forfeited bonds and legal delays in bond forfeiture.39 Following a lawsuit and increased Federal scrutiny thereafter, Pennsylvania modiﬁed its

POLICY ANALYSIS

regulatory framework related to the reclamation of abandoned mine lands.38 Pennsylvania now requires mine operators to perform sitespeciﬁc estimation of reclamation liabilities to ensure posted bonds cover the full cost of reclamation.40 Today, shale gas operators in Pennsylvania must post either a bond of $2500 for each well or a blanket bond of $25,000 to cover all the wells they drill in the state. This is the same dollar amount required in 1984, despite statutory provisions that empower the Environmental Quality Board to adjust the level of bonding to match projected reclamation costs every two years. A bond of $2500 is inadequate to cover the costs to plug a deep shale gas well and restore the land (approximately 100700 thousand dollars). The inadequacy of the blanket bond is even more pronounced, as many operators are expected to drill thousands of wells. For example, Chesapeake Energy, operating in a joint venture with Statoil, plans to drill up to 17 000 shale gas wells in Appalachia over the next 20 years.41 The Oil and Gas Act prohibits private landowners from securing ﬁnancial assurances from the operator independent of Pennsylvania regulations. The situation is diﬀerent on Pennsylvania’s state-owned land. Pennsylvania includes a condition in all of its lease agreements for drilling in state forests that requires operators to submit additional individual well bonds. The dollar amount required scales with the measured depth, so operators in state forests are required to post bonds of $50,000100,000 per well drilled.42 It is important to note that the substantial bonds required in drilling leases in state forests did not preclude a successful lease auction, proceeds of $128 million far exceeded original expectations of $60 million.43 This suggests that bonds in the $100,000 range are not prohibitive for large exploration and production companies, though they may be an obstacle for smaller concerns.

’ TRANSFERRING ASSETS SHIFTS ENVIRONMENTAL LIABILITY Over the next two decades, drilling rates of 1000 or more new shale gas wells per year are projected, as production from Pennsylvania’s Marcellus Shale is expected to reach approximately 110 million cubic meters (4 billion cubic feet) of natural gas per day by 2015.44,45 To sustain such high levels of production, the shale gas industry needs to constantly drill and complete new wells because gas production rapidly declines in the ﬁrst few years of production. Figure 2 shows a type curve published by a Marcellus Shale operator, EQT Production.46 A type curve is a gas production curve modeled from initial and historic production data and reservoir characteristics. The precipitous decline in production rate of gas is typical of deep shale gas wells in Pennsylvania and elsewhere. (Refracking is a process that can be used to increase production in a declining well. Because there are no reliable data published on this practice in Appalachia it is excluded from this analysis.) Industry economics are dominated by high initial gas production rates. For a typical well, assuming a constant price of $176.6 per thousand cubic meters of gas ($5/Mcf) and a $5.3 million cost to drill and complete a new well,46 the internal rate of return (IRR) asymptotes near 79% after the seventh year, after which production revenue dwindles compared to that of the initial years. Assuming a 10% discount rate, 81% of the net present value (NPV) of gross revenue would be realized in 10 years. Compared to the potential revenue from gas sales, the present value of long-term shale gas liabilities, which are discounted 9508

dx.doi.org/10.1021/es2021796 |Environ. Sci. Technol. 2011, 45, 9506–9514

Environmental Science & Technology

Figure 2. Expected gas production rate (solid line) and cumulative production curve (dashed) for EQT Production’s Marcellus Shale operations.46.

4050 years, has negligible impact on near-term accounting. The problem of failing to internalize reclamation liabilities emerges when the liabilities begin to exceed the current asset value. The steep decline in production may drive divestment of shale gas assets by primary exploration and production companies well before the expected closure of a shale gas well. The transfer of marginally producing assets to smaller independent operators or surface owners is common practice in the oil and gas industry.4749 Sometimes surface owners take ownership of a marginally producing well for household use. In such cases, the Oil and Gas Act permits oil and gas asset transfers as long as the prospective owner satisﬁes the applicable bonding requirements. In Pennsylvania, there exists no formal regulatory mechanism to prevent fully bonded owners from assuming shale gas assets with reclamation liabilities substantially above their own ﬁnancial means. Large liabilities covered by limited resources could lead to large-scale insolvency, similar to the situation that spawned Pennsylvania’s pervasive abandoned acid mine drainage and orphan well problems.50 In Pennsylvania and other U.S. states, individual and blanket bonds may be satisﬁed using a number of ﬁnancial instruments and often do not even require monies to be transferred. Requiring only the demonstration of assets is common, especially for large operators. When an operator cannot demonstrate suﬃcient assets to cover liabilities, third party backing, usually in the form of a surety bond, may be obtained for a percentage of the bond’s face value. Since surety companies or banks underwriting the bond are liable if an operator is unable to perform reclamation, bond rates are set according to an individual operator’s risk of insolvency.36 Today’s low bonding levels make it possible for hundreds of independent operators satisfy the Pennsylvania’s blanket bonding requirements.51 These operators are capable of producing marginal amounts of oil and gas economically, which allows them to maximize potential economic beneﬁts by extending the productive lifetime of oil and gas wells.52 The ability to transfer well ownership to independent operators beneﬁts the industry, but a potential consequence of increasing bonding minima could be that smaller operators may face steep risk premiums or not qualify for third party backing and be excluded from participation. Primary exploration and production companies rely on divestment of existing assets to fund new drilling operations. Blocking

POLICY ANALYSIS

independent operators from the market may force these companies to temporarily abandon their uneconomic wells and apply for inactive status instead. In Pennsylvania, nonproducing wells may be granted inactive status for a period of ﬁve years, but to be granted an annual extension the operator only has to declare regulatory compliance and the capacity to produce gas in the future from the inactive well. Inactive status and similar provisions in other states grant operators the ability to temporarily abandon a gas well until technology advances or favorable gas prices improve the economics of production, though in practice the decision to reopen a well is expected to be dominated by reclamation and other liabilities.13 Inactive status could be used to defer the costs of reclamation indeﬁnitely. According to PADEP records, almost 17 000 conventional oil and gas wells did not report or produce oil or gas for three consecutive years (20072009), and were listed as active at the end of 2009.While it may be the case that many of the operators of these wells simply failed to report production, poor compliance with reporting requirements prevents the PADEP from enforcing plugging requirements or administering the inactive status program. In 2009 alone, only 38% of the Commonwealth’s conventional oil and gas wells reported production, which indicates a majority of the wells drilled in Pennsylvania may represent environmental liabilities as opposed to a source of revenue.53 Incentives (ﬁnes) are needed to improve compliance with production reporting requirements, though reporting alone will not close this loophole. The delay between production and reclamation temporally separates revenue generation from the future liabilities. Others have recognized this undesirable trend and instituted remedies. Growth in the number of nonproducing (idle) wells in Alberta and Saskatchewan led these two Canadian provinces to implement a Licensee Liability Rating Program as a measure of insolvency risk and to minimize state ﬁnancial exposure to orphan wells. The program requires individual operators to provide ﬁnancial assurance equivalent to the diﬀerence between the operators’ assets (active wells and assets) and liabilities (inactive wells and abandoned assets).54,55 Some U.S. states oﬀer tax breaks to promote marginal well production, while others require additional bonds or levy annual fees for inactive wells to incentivize new production or plugging, and to fund compliance monitoring. 25,52

’ REGULATORY POLICY AND FINANCIAL ASSURANCE OPTIONS When bonding requirements are smaller than expected liabilities, there is a ﬁnancial incentive to not comply with reclamation requirements. Individual well bonding requirements that match reclamation costs would remedy this situation, especially with the blanket bonds, where misalignments with reclamation costs can be huge. Eliminating the blanket bond would be a common sense ﬁrst step for Pennsylvania. However, simply increasing the bond requirement to match reclamation costs may not be the best alternative because more operators will need to obtain third party backing. In theory, reliance on third party backing favors operators that manage assets and liabilities eﬀectively since the underwriting ﬁrms would assess the risk of insolvency of individual operators. However, the same may not be true for third party backers. Insolvency of these ﬁnancial ﬁrms is a real concern and the eﬀects may be large.36,56 Furthermore, bonds are inherently inﬂexible to changes in the cost of performing reclamation, to the economics of gas extraction 9509

dx.doi.org/10.1021/es2021796 |Environ. Sci. Technol. 2011, 45, 9506–9514

Environmental Science & Technology when wells start to lose pressure, and the way ﬁnancial risk is shared in the industry. This is problematic if reclamation costs deviate dramatically from the average. For instance, following methane migration into the aquifer supplying drinking water to 14 households in Dimock, Pennsylvania, the estimated costs for individual water ﬁltration units and supply replacement via permanent pipeline were approximately $8,000 and $800,000 per household, respectively.57,58 Underwriting ﬁrms will only market surety bonds when the amount and term of liability are strictly deﬁned,36 so bonds are not well suited to cover uncertain liabilities. Bonds would also fail to provide funding for maintenance and monitoring of plugged and abandoned wells and the potential environmental issues that may arise postreclamation. After the release of a bond, recovery of additional environmental costs would require aggrieved citizens or the State to pursue civil action. The State may also block the issuance of new permits to operators with outstanding reclamation liabilities, but for operators without ongoing interests in Pennsylvania, this enforcement mechanism will be limited.

’ ALTERNATIVES TO BONDS To pay for the long-term treatment of acid mine discharges, coal mine operators in Pennsylvania may establish trust accounts under contract with the State. Funding requirements are based on operator estimates of the present value of capital costs and operating expenses of pollution control projects, which depend on the inﬂation rate and the expected growth of the trust account. As irrevocable beneﬁciaries of the trust, the State will reimburse coal mine operators one year after the performance of work, or in the case of nonperformance, the State may use accumulated funds to do the work.59 If reclamation trust accounts were to be used for the shale gas industry, it would be the responsibility of the operator to determine current (time zero) reclamation costs as part of the drilling permit and the responsibility of the state to approve that ﬁgure. If fully funded trust accounts were tied to individual wells rather than pooling them, timely plugging would become independent of the solvency of the last operator. For the mining industry, trusts are designed so that they will be fully funded one-year after production ends. The size the trust is estimated from eq 1, which shows the calculation for the present value of reclamation costs. " # RC ð1 þ VolÞ ð1Þ PV ¼ ð1 þ ½E IÞt Where RC = estimated cost of reclamation in current dollars, E = expected annual return on investments in trust, I = inﬂation rate, Vol = volatility premium, proportional to amount invested in stock market, and t = time in years, duration of production For the shale gas industry, the contract between the State and individual operator would specify the ﬁrm responsible for managing the trust account and investment strategy. An inﬂation rate of 3.1%, bond yield of 5.25%, and market return of 11.2% are recommended by the PADEP for eq 1. At most, 80% of the trust may be invested in stock. A 20% volatility premium is required for the portion of the trust invested in stock.59 It is the responsibility of the PADEP to ensure an operator’s inﬂation, bond yield, and market return assumptions reﬂect current conditions. This contract would also detail the irrevocable rights held by the State to claim monies held in the trust.

POLICY ANALYSIS

Figure 3. Comparison of ﬁnancial assurance mechanisms for funding a reclamation liability costing $100,000 at time zero. Assumptions: gas is produced according to the EQT Production type curve (Figure 2); the inﬂation rate is 3.1%; and monies invested in the trust have an assumed annual return of 5.25%, following PADEP guidance for bond yields.59 The “no risk” cash bond option is not shown as it is equal to the cost of reclamation. The funds collected by a predrilling fee and severance tax collected for ﬁve years are contrasted. Delayed collection options run the risk of collecting insuﬃcient funds for reclamation of the well if the number of productive years is less than the number of years used to determine present value of reclamation costs. At any given year, the funding shortfall is measured as the diﬀerence between the projected reclamation cost line and the respective delayed option line.

We compare three potential mechanisms to fund well reclamation costs estimated using eq 1: cash bond, severance tax on gas production, and a discounted predrilling fee. The properly sized cash bond represents a “no risk” scenario for Pennsylvania because operators would be required to deposit the full cost of reclamation as a precondition for drilling permit approval. Compared to the other forms of bonding allowed by the PADEP, the State Treasurer would manage the bonded monies and the risks associated with operator or third-party default or insolvency would be eliminated. A severance tax on gas production would gradually collect and reinvest monies to reach the future value of reclamation. Pennsylvania’s Governor, Tom Corbett, opposes levying taxes on the natural gas industry, but has supported a onetime, per well fee to pay for local impacts of the natural gas industry. To fund a reclamation trust via a discounted predrilling fee, we assume that the fee would need to be assessed in an amount equal to the present value of expected reclamation costs at the time of well closure. The severance tax and predrilling fee represent delayed funding mechanisms, so the annual growth and security of the trust as well as the productive lifetime of a shale gas well are important variables. The cost to perform reclamation is compared to funds accrued in a reclamation trust by a severance tax (calculated for two diﬀerent anticipated well lifespans) and a predrilling fee in Figure 3. To fully fund a reclamation trust by year 16, a predrilling fee of $65,975 and a severance tax of $0.87/TCM ($0.25/Mcf) collected for ﬁve years would need to be assessed. A severance tax of $0.15/TCM ($0.004/Mcf) on the ﬁrst ﬁve years of production would be assessed if full funding of the trust is not required until year 51. The cash bond option is not graphed because it is equivalent to the inﬂated reclamation cost each year. The options are fully 9510

dx.doi.org/10.1021/es2021796 |Environ. Sci. Technol. 2011, 45, 9506–9514

Environmental Science & Technology

POLICY ANALYSIS

funded when they intersect the reclamation cost line. If the well is abandoned before the reclamation trust is whole, the diﬀerence between the accumulated funds and the inﬂated reclamation costs will be the shortfall. No empirical evidence exists to suggest the economic lifetime of a shale gas well will reach generic industry predictions of 4050 years. Well productivity and the economics of shale gas production have equal weight in an operator’s decision to keep a well open. The use of unrealistic expectations of well economics has implications for the application of delayed funding mechanisms and risks underfunding reclamation trust accounts. Figure 3 shows that even if a 15-year lifetime is assumed (reclamation costs discounted from year 16), the diﬀerence between the reclamation cost and the funding levels in the trust are substantial for wells abandoned sooner. For the purpose of estimating reclamation costs, it would be wise for Pennsylvania to require that reclamation costs by funded within 10 years, regardless of the actual life span of the well. Actual production will deviate from industry type curves. Figure 4 shows the cumulative production from horizontal shale gas wells in Pennsylvania that began producing gas from January 2010 through July 2011 compared to the EQT Production type curve (Figure 2).

Figure 4. Reported cumulative production of 294 individual horizontal Marcellus Shale gas wells that began producing after 1/1/ 2010.51 Three continuous cumulative production curves are modeled: EQT Production’s type curve (Figure 2), a 60% EQT, and 35% EQT. Cumulative production predicted by the 60% EQT and 35% EQT curves is exceeded by 50% and 75% of horizontal Marcellus Shale gas wells, respectively.

While nearly a quarter of the wells exceeded the EQT curve, half of the wells produced less than 60% of the EQT curve and 25% of the wells produced 35% or less of the EQT estimate. The variability in cumulative production indicates that industry type curves should not be used to set the terms of ﬁnancial assurance policy. If a 5-year severance tax is calculated from EQT Production’s type curve and applied to the cumulative production of all the wells in Figure 4, independent of the tax rate, the amount of money collected in a trust would only be 62% of the target funding level, assuming that excess funds are returned to the operator.

’ THE IMPACT OF THESE REGULATORY OPTIONS ON THE INDUSTRY BOTTOM LINE From the point of view of industry ﬁnances, the diﬀerent funding mechanisms have similar impacts on the internal rate of return (IRR) of a producing well, even if total production is low. Table 1 contrasts the IRRs resulting from implementation of (1) the current bond requirement ($2,500), (2) a cash bond equivalent to the reclamation cost, (3) a predrilling fee, and (4) a 5-year severance tax. We assume 50 years of revenue from production, but use a 10-year funding timeline to minimize the risk of underfunding the reclamation trust. Though these are rough calculations based on simple assumptions, Table 1 shows that levying a predrilling fee and small severance tax on the ﬁrst ﬁve years of production would quickly fund a trust account with minimal impact on the project’s IRR. From the industry point of view, paying the full cost of reclamation in an up-front bond is the least attractive alternative. However, actual implantation of any ﬁnancial assurance requires an industry-wide evaluation of ﬁnancial assumptions ’ RISKS TO THE STATE From the State’s point of view, there is a risk that the well will become uneconomic prior to year 10, especially if production is much less than EQT Production’s type curve. If this occurs, the shortfall of the 5-year severance tax would be greatest. The problem of underperforming wells or dry holes, however, is not adequately addressed, and unless the “no risk” cash bond is employed, it is expected that both delayed funding options will result in inadequate funding of the reclamation trust account. In the coal industry, operators are required to make underfunded trust accounts whole either by direct payments into the trust or supplementary bonds. If regulations are strictly enforced to prevent dry holes and uneconomic wells from being granted

Table 1. Gross Revenue IRRs Incorporating the Implementation Cost of Financial Assurance Mechanismsa reclamation cost $100,000

$700,000

gas production curve model IRR with current bond

IRR with “no risk” cash bond

IRR with predrilling fee IRR with 5-year severance tax

EQT

78.7%

76.7%

77.1%

78.1%

60% EQT

34.3%

33.2%

33.5%

33.8%

35% EQT

13.2%

12.7%

12.8%

12.9%

EQT 60% EQT

78.7% 34.3%

65.6% 27.6%

68.4% 29.0%

74.3% 30.7%

35% EQT

13.2%

10.2%

10.8%

11.0%

a

Drilling and completion cost of $5.3 million and $176.6/TCM ($5/Mcf) price of gas is assumed. The pre-drilling fee and 5-year severance tax are calculated to fully-fund the reclamation trust by year 11. Two target reclamation costs are contrasted, $100,000 and $700,000. The pre-drilling fees are $76,000 and $535,000 for targets of $100,000 and $700,000, respectively. A severance tax rate of $1.01/TCM ($0.029/Mcf) is required for reclamation cost of $100,000 and the EQT production curve. The rate increases to $20.01/TCM ($0.57/Mcf) for reclamation cost of $700,000 and the 35% EQT production curve. TCM = thousand cubic meters. Mcf = thousand cubic feet. 9511

dx.doi.org/10.1021/es2021796 |Environ. Sci. Technol. 2011, 45, 9506–9514

Environmental Science & Technology inactive status, the risk of these wells becoming State liabilities decreases. The risk of underfunded reclamation trusts due to dry holes or otherwise underperforming wells could be reduced if individual operators pooled monies in a reclamation trust. In this case, the severance tax would need to be based on the value of the pooled trust, aggregate production data, and total reclamation liability. To prevent operators from shirking environmental responsibility and ensure the State has adequate resources in case of insolvency, adjustments to the severance tax rate may be necessary so that pooled funds cover the sum of expected reclamation costs. PADEP may readjust trust funding levels for the mining industry to reﬂect changes in pollution control costs of plus or minus 10%.59 However, regulatory inertia or poor oversight pose a threat to the achievement of adequate funding levels, as demonstrated by the lack of adjustment in oil and gas well bonding levels for more than a quarter-century. In theory, the potential for a downward adjustment of the required funding level incentivizes operators to invest in new technologies (or enhanced “pollution control”) to lower the cost of reclamation and to have excess funds returned.60

’ DISAGGREGATING ENVIRONMENTAL ACCIDENTS FROM WELL SITE RESTORATION AND CLOSURE While bond forfeiture is commonly associated with operator failure to perform site restoration and plug abandoned wells, the intent of current bonding system for oil and gas wells is much broader. At any point during the productive life of a well, noncompliance with the Oil and Gas Act or an order of the PADEP may be grounds for bond forfeiture. Restoration of water supplies impacted by nearby shale gas operations is an example. The formation of a competitive bond market requires that liabilities be well-deﬁned in amount and time. Therefore, neither bonds nor trust accounts are the appropriate tool for environmental accidents that occur during production. A remedy could be for Pennsylvania to adopt ﬁnancial assurance rules that separate expected liabilities from uncertain events such as casing failure or other environmental accidents. Requiring active operators to obtain liability insurance for uncertain events is a partial solution. Insurance companies would need to quantify potential risks and determine an eﬃcient way to pool risk across multiple wells or operators. However, in the absence of a responsible operator, the State or aﬀected citizen is likely to bear the cost in the event of an environmental issue postreclamation. ’ CONCLUSION The ﬁnancial assurance mechanisms that Pennsylvania uses to ensure compliance with Pennsylvania’s Oil and Gas act of 1984 are outdated and allow ownership transfers to entities less likely to be able to cover the expected costs of reclamation. Without strict enforcement of gas production reporting requirements, the PADEP will be unable to monitor compliance with plugging requirements and prevent abuse of the inactive status program. Timely plugging and abandonment should be the goal of PADEP policy because the long-term environmental and human health risks of shale gas development will increase over time and with the risk of operator insolvency. However, increasing the bonding requirements to fully cover reclamation costs, which is within the PADEP’s mandate, will not address well-known limitations of environmental bonds and may limit participation in shale gas development to larger companies.

POLICY ANALYSIS

Alternative mechanisms to ensure operators pay for future reclamation costs include a cash bond, a predrilling fee, and a severance tax. If operators were to deposit the full cost of reclamation in the form of a cash bond, the risk of underfunding will be lowest. Taxing gas production to fund an individual-well trust account for future reclamation poses no additional barrier to operator entrance. This approach may force the State to assume the risk of reclaiming dry holes unless wells are pooled and a severance tax adjustable to funding levels in the trust, total reclamation liabilities, and aggregate production is developed. Comparing all three mechanisms, we found that generating funds directly from the revenue stream during the most lucrative years of gas production has the lowest impact on an operator’s IRR. Though the industry generically predicts wells to operate for 4050 years, reliance on these assumptions to deﬁne the terms of ﬁnancial assurance increases the risk of underfunding and cannot be justiﬁed. Separate handling of reclamation and accidental environmental liabilities would promote the development of a competitive bond market if the current system is kept in place.

’ AUTHOR INFORMATION Corresponding Author

*Phone: (412) 268 3756; fax (412) 268 3757; e-mail: casman@ andrew.cmu.edu. Author Contributions

Both authors contributed to the content and writing of this manuscript.

’ ACKNOWLEDGMENT We are grateful to W. Michael Griﬃn (CMU), Joel Tarr (CMU), Roma Sidortsov (Vermont Law School), and Susan Ghoweri (PADEP) for their valued advice and perspective. This work was funded by the Climate Decision Making Center (SES3045798) and by the Center for Climate and Energy Decision Making (SES-0949710), both through cooperative agreements between the National Science Foundation and Carnegie Mellon University; and by the Gordon and Betty Moore Foundation (Award Number 1625) in support of the Carnegie Energy Research Initiative. ’ NOMENCLATURE BCF, billion cubic feet Mcf, thousand cubic feet TCM, thousand cubic meters PADEP, Pennsylvania Department of Environmental Protection IRR, internal rate of return NPV, net present value ’ REFERENCES (1) Oil Industry International Exploration and Production Forum (E&P Forum) and United Nations Environmental Programme (UNEP). Environmental Management in Oil and Gas Exploration and Production: An Overview of Issues and Management Approaches; UNEP: London, U.K., 1997. (2) Angradi, T. R. Fine sediment and macroinvertebrate assemblages in Appalachian streams: A ﬁeld experiment with biomonitoring applications. J. North Am. Benthol. Soc. 1999, 18 (1), 49–66. (3) Berkman, H.; Rabeni, C. Eﬀect of siltation on stream ﬁsh communities. Environ. Biol. Fishes 1987, 18 (4), 285–294. 9512

dx.doi.org/10.1021/es2021796 |Environ. Sci. Technol. 2011, 45, 9506–9514

Environmental Science & Technology (4) MacFarlane, A. Revegetation of Wellsites and Seismic Lines in the Boreal Forest. Honour’s Thesis University of Alberta, Edmonton, Alberta, 1999. (5) Robinson, S. K.; Thompson Iii, F. R.; Donovan, T. M.; Whitehead, D. R.; Faaborg, J. Regional forest fragmentation and the nesting success of migratory birds. Science 1995, 267 (1), 1987–1990. (6) Johnson, N. Pennsylvania Energy Impacts Assessment; The Nature Conservancy—Pennsylvania Chapter : Harrisburg, PA, 2010; p 47. http://www.nature.org/media/pa/tnc_energy_analysis.pdf. (7) Pennsylvania Department of Conservation and Natural Resources, Bureau of Forestry. Pennsylvania Statewide Forest Resource Assessment Pennsylvania Department of Conservation and Natural Resources: Harrisburg, PA, 2010; p 210 www.dcnr.state.pa.us/forestry/farmbill/pdfs/assessment.pdf. (8) Railroad Commission of Texas. Well Plugging Primer. In Railroad Commission of Texas Oil and Gas Division Well Plugging Section: 2000; p 20 http://www.rrc.state.tx.us/forms/publications/plugprimer1. pdf. (9) Gurevich, A. E.; Endres, B. L.; Robertson, J. O., Jr; Chilingar, G. V. Gas migration from oil and gas ﬁelds and associated hazards. J. Pet. Sci. Eng. 1993, 9 (3), 223–238. (10) Pennsylvania Department of Environmental Protection. Stray Natural Gas Migration Associated with Oil and Gas Wells; Pennsylvania Department of Environmental Protection: Harrisburg, 2009; http:// www.dep.state.pa.us/dep/subject/advcoun/oil_gas/2009/Stray% 20Gas%20Migration%20Cases.pdf. (11) Pennsylvania Department of Environmental Protection, Orphan Oil and Gas Wells and the Orphan Well Plugging Fund. In Pennsylvania Department of Environmental Protection: 2007. http:// www.elibrary.dep.state.pa.us/dsweb/Get/Document-82185/5500-FSDEP1670.pdf (12) National Energy Technology Laboratory. Methane Emissions Project Borough of Versailles, PA; U.S. Department of Energy, 2007; http://www.netl.doe.gov/newsroom/versailles/Versailles%20Methane %20Emissions%20Project%20-%20Final%20Report.pdf (13) Muehlenbachs, L. Internalizing Production Externalities: A Structural Estimation of Real Options in the Upstream Oil and Gas Industry. Doctor of Philosophy, University of Maryland, College Park, MD, 2009. (14) Nichol, J. R.; Kariyawasam, S. N. Risk Assessment of Temporarily Abandoned or Shut-in Wells; Department of the Interior, Minerals Management Service, C-FER Technologies, 2000; p 72 www.boemre. gov/tarprojects/329/329AA.pdf. (15) Khandka, R. K. Leakage Behind Casing; Norwegian University of Science and Technology, Department of Petroleum Engineering and Applied Geophysics Trondheim, Norway, 2007. (16) Dusseault, M. B.; Gray, M. N. Why Oilwells Leak: Cement Behavior and Long-Term Consequences; Society of Petroleum Engineers Inc.: Beijing, China, 2000; http://www.onepetro.org/mslib/servlet/ onepetropreview?id=00064733&soc=SPE. (17) Gasda, S. E.; Bachu, S.; Celia, M. A. Spatial characterization of the location of potentially leaky wells penetrating a deep saline aquifer in a mature sedimentary basin. Environ. Geol. 2004, 46 (6), 707–720. (18) Ghoweri, S., Personal Communication. In 2011. (19) Bureau of Land Management. Coalbed Methane Development in the Northern San Juan Basin of Colorado; U. S. Department of the Interior: San Juan, CO, 1999; p 129, http://cogcc.state.co.us/Library/ sanjuanbasin/blm_sjb.htm. (20) Chilingar, G. V.; Endres, B. Environmental hazards posed by the Los Angeles Basin urban oilﬁelds: An historical perspective of lessons learned. Environ. Geol. 2005, 47 (2), 302–317. (21) Staﬀord, S. L.; Weaver, T. J.; Hedin, R. S. Geochemistry, hydrogeology, and eﬀects from the plugging of artesian ﬂows of acid mine drainage: Clarion River Watershed, Northwestern Pennsylvania. In National Meeting of the American Society of Mining and Reclamation; The American Society of Mining and Reclamation: Morgantown, WV, 2004; http://www.asmr.us/Publications/Conference%20Proceedings/ 2004/1792-Staﬀord%20PA.pdf.

POLICY ANALYSIS

(22) Rabinowitz, H. Economic Eﬀects of Groundwater Contamination on Real Estate; University of Wisconsin, Water Resources Center, 1995; http://digital.library.wisc.edu/1711.dl/EcoNatRes.WRCGRR9506. (23) Boxall, P.; Chan, W.; McMillan, M. The impact of oil and natural gas facilities on rural residential property values: a spatial hedonic analysis. Resour. Energy Econ. 2005, 27 (3), 248–269. (24) Leggett, C. G.; Bockstael, N. E. Evidence of the eﬀects of water quality on residential land prices. J. Environ. Econ. Manage. 2000, 39 (2), 121–144. (25) Interstate Oil and Gas Compact Commission. Protecting Our Country’s Resources: The states Case; The Interstate Oil and Gas Compact Commission: Oklahoma City, OK, 2008; p 68, http://iogcc.myshopify. com/products/protecting-our-countrys-resources-the-states-case-orphaned-well-plugging-initiative-2008. (26) Intergovernmental Panel on Climate Change. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, 2007. http://ipcc.ch/publications_ and_data/ar4/wg1/en/contents.html (accessed April 14, 2011). (27) Pennsylvania Department of Environmental Protection. Executive Summary The Oil and Gas Act; Pennsylvania Department of Environmental Protection: Harrisburg, PA, 2009; www.dep.state.pa.us/dep/deputate/minres/oilgas/ORPHRPT4.pdf. (28) Pennsylvania Department of Environmental Protection. Bureau of Oil and Gas Management, Abandoned & Orphan Wells Listing; Pennsylvania Department of Environmental Protection: Harrisburg, PA,2011; http://www.dep.state.pa.us/dep/deputate/ minres/oilgas/AbandedOrphanWells.xls. (29) Pennsylvania Department of Environmental Protection. Bureau of Oil and Gas Management, Pennsylvania’s Plan for Addressing Problem Abandoned Wells and Orphaned Wells, Document number 550-800-001; Pennsylvania Department of Environmental Protection: Harrisburg, 2000; p 7; http://www.elibrary.dep.state.pa.us/dsweb/Get/Version48262/550-0800-001.pdf. (30) Pennsylvania Department of Environmental Protection. Second Year Growing Greener II Report; Pennsylvania Department of Environmental Protection: Harrisburg, PA, 2007; http://www.portal.state.pa. us/portal/server.pt?open=18&objID=503089&mode=2. (31) Pennsylvania Department of Environmental Protection. Third Year Growing Greener II Report; Pennsylvania Department of Environmental Protection: Harrisburg, PA, 2008; http://www.portal.state.pa. us/portal/server.pt?open=18&objID=503088&mode=2. (32) New York State Department of Environmental Conservation. Well Permit Issuance for Horizontal Drilling And High-Vol. Hydraulic Fracturing to Develop the Marcellus Shale and Other Low-Permeability Gas Reservoirs ; New York State Department of Environmental Conservation: Albany, NY, 2009; ftp://ftp.dec.state.ny.us/dmn/download/ OGdSGEISFull.pdf. (33) Andersen, M.; Coupal, R., Economic issues and policies aﬀecting reclamation in wyoming’s oil and gas industry. In National Meeting of the American Society of Mining and Reclamation, Billings, MT, 2009. (34) Pennsylvania Department of Conservation and Natural Resources. Bureau of Topographic and Geologic Survey, Well Completion Reports . In Pennsylvania Internet Record Imaging System/Wells Information System (PA*IRIS/WIS); Pennsylvania Department of Environmental Protection: Harrisburg, 2010. (35) Cabot Oil & Gas Corporation. Summary of Cabot’s Good Faith Eﬀorts, 2010. http://cabotog.com/pdfs/ExhibitB.pdf (accessed April 15, 2011). (36) Boyd, J. Financial Responsibility for Environmental Obligations: Are Bonding and Assurance Rules Fulﬁlling Their Promise?; Resources for the Future: Washington, DC, 2001; p 71, www.rﬀ.org/documents/RFFDP-01-42.pdf. (37) Pennsylvania Department of Environmental Protection. Assessment of Pennsylvania’s Bonding Program for Primacy Coal Mining Permits ; Pennsylvania Department of Environmental Protection Oﬃce of Mineral Resources Management Bureau of Mining and Reclamation: Harrisburg, PA, 2000; p 43, http://www.dep.state.pa.us/dep/deputate/ minres/bmr/bonding/. 9513

dx.doi.org/10.1021/es2021796 |Environ. Sci. Technol. 2011, 45, 9506–9514

Environmental Science & Technology (38) Oﬃce of Surface Mining Reclamation and Enforcement. National Priority ReviewAdequacy of Bonding in the Approved Pennsylvania Program; U.S. Oﬃce of Surface Mining Reclamation and Enforcement: Pittsburgh, PA, 2010; p 407, http://www.arcc.osmre. gov/Divisions/PFD/PDFs/2010/2010_Pennsylvania_Bonding_Adequacy_Study.pdf. (39) U.S. General Accounting Oﬃce. Surface Mining: Diﬃculties in Reclaiming Mined Lands in Pennsylvania and West Virginia; General Accounting Oﬃce: Washington, DC, 1986; http://www.gao.gov/ products/RCED-86-221. (40) Pennsylvania Department of Environmental Protection. Conventional Bonding for Land Reclamation—Coal; Department of Environmental Protection Bureau of Mining and Reclamation: Harrisburg, PA, 2006; p 30 http://www.elibrary.dep.state.pa.us/dsweb/Get/Document-72980/Draft%20563-2504-001.pdf. (41) STATOIL, Marcellus shale gas. 2010. www.statoil.com/ en/About/Worldwide/USA/Pages/ShaleGasMarcellus.aspx (accessed April 15, 2011). (42) DCNR Minerals 2007 Standard Oil Gas Lease. http://www. dcnr.state.pa.us/forestry/sfrmp/documents/Minerals_2007_Standard_Oil_Gas_Lease.pdf (accessed May 24, 2011). (43) DCNR. Gov. Rendell says responsible planning generates $128.4 million from state forest land lease, safeguards natural resources. In News and Information Resource; Pennsylvania Department of Conservation and Natural Resources: Harrisburg, PA, 2010; http://www. dcnr.state.pa.us/news/resource/res2010/10-0120-gaslease.aspx (accessed June 21, 2011). (44) Sherman, T. Market Eﬀects of the Marcellus, Pittsburgh, PA, 2010; http://www.dugeast.com/PastConferences/ (accessed November 16, 2010). (45) Considine, T. J. The Economic Impacts of the Marcellus Shale: Implications for New York, Pennsylvania, and West Virginia; Natural Resource Economics, Inc.: Larmie, WY, 2010; p 44, http://www.api. org/policy/exploration/hydraulicfracturing/upload/APIEconomicImpactsMarcellusShale.pdf. (46) E. Q. T. Production, Marcellus Decline Curve. Events & Presentation, 2011. http://ir.eqt.com/events.cfm?AcceptDisclaimer= yes (accessed April 15, 2011). (47) Koplow, D.; Martin, A., Fueling Global Warming: Federal Subsidies to Oil in the United States. In Industrial Economics Inc: Cambridge, MA, 1998; p 140 http://archive.greenpeace.org/climate/ oil/fdsuboil.pdf (48) Hager, A. V.; Shaw, K. L., Idle and Deserted Wells: Who Plugs and Who Pays? In Proceedings of the Annual Institute of the Rocky Mountain Mineral Law Foundation, 1999; Vol. 45 www.mayerbrown. com/energy/article.asp?id=2144&nid=10908 (49) U.S. General Accounting Oﬃce. Alaska’s North Slope Requirements for Restoring Lands After Oil Production Ceases. In U.S. General Accounting Oﬃce: Washington, DC, 2002; p 114, http://www.gao. gov/products/GAO-02-357. (50) Tarr, J. A. Devestation and Renewal: An Environmental History of Pittsburgh and Its Region; University of Pittsburgh Press: Pittsburgh, PA, 2003. (51) Pennsylvania Department of Environmental Protection. Production Reports 20032011. PA DEP Oil & Gas Reporting Website, 2011. https://www.paoilandgasreporting.state.pa.us/publicreports/Modules/Production/ProductionHome.aspx (accessed August 30, 2011). (52) Interstate Oil and Gas Compact Commission. Marginal Wells: Fuel For Economic Growth; The Interstate Oil and Gas Compact Commission: Oklahoma City, OK, 2008; www.energy.psu.edu/swc/ news/2008-Marginal-Well-Report.pdf. (53) Pennsylvania Department of Environmental Protection. Production Reports 20032009. PA DEP Oil & Gas Reporting Website, 2009. https://www.paoilandgasreporting.state.pa.us/publicreports/Modules/Production/ProductionHome.aspx (accessed August 30, 2011). (54) ECRB. Directive 006 Licensee Liability Rating (LLR) Program and License Transfer Process. In The Energy Resources Conservation Board 2009. http://www.ercb.ca/docs/documents/directives/Directive006.pdf (accessed August 15, 2011)

POLICY ANALYSIS

(55) ER. Guideline PD-G01: Licensee Liability Rating (LLR) Program Guideline. In The Ministry of Energy and Resources: 2010. http://www.ir.gov.sk.ca/adx/aspx/adxGetMedia.aspx?DocID= 10419,10418,3680,3384,5460,2936,Documents&MediaID=38399& Filename=LLR+Guideline+January+2011.pdf (accessed May 25, 2011). (56) U.S. General Accounting Oﬃce. Surface Mining Cost and Availability of Reclamation Bonds; U.S. General Accounting Oﬃce: Washington, DC, 1988; http://www.gao.gov/products/PEMD-88-17. (57) Cabot Oil & Gas Corporation. 2010. http://www.cabotog. com/pdfs/ExhibitB.pdf (accessed April 15, 2011). (58) Pennsylvania Department of Environmental Protection, Consent Order and Settlement Agreement. 2010. http://www.cabotog. com/pdfs/FinalA_12-15-10.pdf (accessed April 14, 2011). (59) Pennsylvania Department of Environmental Protection, Financial Assurance and Bond Adjustments for Mine Sites with post-mining Discharges. In Pennsylvania Department of Environmental Protection: Harrisburg, PA, 2007; http://www.elibrary.dep.state.pa.us/dsweb/Get/ Document-64460/563-2504-450.pdf. (60) Shogren, J. F.; Herriges, J. A.; Govindasamy, R. Limits to environmental bonds. Ecol. Econ. 1993, 8, 109–133.

9514

dx.doi.org/10.1021/es2021796 |Environ. Sci. Technol. 2011, 45, 9506–9514

POLICY ANALYSIS pubs.acs.org/est

Unearthing Potentials for Decarbonizing the U.S. Aluminum Cycle Gang Liu, Colton E. Bangs, and Daniel B. M€uller* Industrial Ecology Programme and Department of Hydraulic and Environmental Engineering, Norwegian University of Science and Technology, S.P. Andersens vei 5, 7491 Trondheim, Norway

bS Supporting Information ABSTRACT: Global aluminum demand is anticipated to triple by 2050, by which time global greenhouse gas (GHG) emissions are advised to be cut 50 85% to avoid catastrophic climate impacts. To explore mitigation strategies systematically, a dynamic material ﬂow model was developed to simulate the stocks and ﬂows of the U.S. aluminum cycle and analyze the corresponding GHG emissions. Theoretical and realistic reduction potentials were identiﬁed and quantiﬁed. The total GHG emissions for the U.S. aluminum cycle in 2006 amount to 38 Mt CO2-equivalence. However, the U.S. has increasingly relied on imports of aluminum embodied in various products. The in-use stock is still growing fast in most product categories, which limits current scrap availability for recycling and emissions saving. Nevertheless, there is still large emission mitigation potential through recycling. The potentials from “100% old scrap collection” and “low emission energy” were each calculated to be higher than all process technology potential. Total emissions will decrease dramatically and mitigation priorities will change signiﬁcantly under a stock saturation situation as much more old scrap becomes available for recycling. The nature of in-use stock development over the coming decades will be decisive for the aluminum industry to reach deeper emission cuts.

1. INTRODUCTION Aluminum is the second most used metal worldwide and global aluminum demand is anticipated to triple at least by 2050.1 Primary production of aluminum from bauxite is very energy and greenhouse gas (GHG) emissions intensive. While the IPCC advises cutting global GHG emissions by 50 85% by 2050 to avoid catastrophic climate impacts that would accompany a 2 °C global average temperature increase,2 aiming for such a reduction in the aluminum industry would be extremely challenging only through technology improvements.3 For example, the perﬂuorocarbon (PFC) emissions intensity has limited scope for further improvement after an 86% reduction over the past twenty years,4 so the growing demand for primary aluminum would result in a dramatic increase of total emissions. Recycling of aluminum scrap requires up to 95% less energy than the production of primary aluminum; however, currently scrap availability is limited due to high accumulation of aluminum in products in use. Understanding the dynamics of the entire aluminum cycle would facilitate the exploration of mitigation strategies from a systems approach that goes far beyond the potential of process technology improvements. Several studies on the emissions of the aluminum industry applied a life cycle assessment (LCA) approach5 and concentrated mainly on primary aluminum production6,7 and lightweighting use of aluminum in vehicles.8,9 These studies neglect aggregate eﬀects, interactions within the entire cycle, and the time dimension, and thus cannot provide a sectoral or regional context for discussions on absolute emissions reduction. These shortcomings can be avoided by employing a material/substance r 2011 American Chemical Society

ﬂow analysis (MFA/SFA) approach. The aluminum cycles of Denmark,10 Italy,11 the U.S.,12 and China 13 for a single year or selected years have been characterized. Dynamic models for calculating aluminum scrap generation using historical consumption data and product lifetimes were introduced in the 1970s,14 and recently further reﬁned for Germany,15 the UK,16 and the U.S.17,18 These models primarily focus on mass stocks and ﬂows, and only a few have been integrated with an environmental dimension to further enhance the policy relevance for long-term issues of environmental impact mitigation.19 21 Schwarz et al.22 and Allwood et al.3 built material ﬂow models on the global scale to discuss future aluminum mass ﬂows and emissions. However, their models do not include in-use stocks explicitly, and therefore have limited potential for explaining changes in scrap availability, which is critical for insights into emissions reduction potential through recycling. In this paper, we develop a model which simulates the dynamic anthropogenic aluminum cycle and allows an integrated analysis of material ﬂows and corresponding energy use and GHG emissions. The U.S. was selected as a case country due to its large market and long history for aluminum use. We ﬁrst analyze and quantify all relevant stocks and ﬂows of the U.S. aluminum cycle for the period of 1900 2008, then present a detailed analysis of the emissions for the cycle in 2006. Subsequently, Received: June 28, 2011 Accepted: October 4, 2011 Revised: September 5, 2011 Published: October 04, 2011 9515

dx.doi.org/10.1021/es202211w | Environ. Sci. Technol. 2011, 45, 9515–9522

Environmental Science & Technology

POLICY ANALYSIS

Figure 1. System deﬁnition of the U.S. anthropogenic aluminum cycle and its consequent energy use and GHG emissions. Four types of semis α = a d (rolled, extruded, cast, and other mill products); twelve types of ﬁnal and obsolete products β = a l (BC, TAU, TAE, TOT, PCA, POT, ME, ECA, EOT, CD, OTN, and OTD); two types of scrap γ = a b (can scrap and other general scrap); seven types of energy carriers δ = a d (U.S. smelting contract mix, U.S. grid mix, natural gas, heavy oil, hard coal, propane, and diesel and light fuel oil). P12 and P13 are placed outside the system boundary to portray that only emissions allocated to the U.S. aluminum cycle is considered for these two processes.

Table 1. Product Categories, Codes, and Examples Used in This Study code

product category

product examples

BC TAU

building and construction transportation: automobiles and light trucks

rooﬁng, cladding, window and door frames engine blocks, suspension components, automobile frames and body panels, wheel rims

TAE

transportation: aerospace

aircraft frames and decking

TOT

transportation: others

railway cars, marine vessels, motorcycles and bicycles

PCA

packaging: cans

beverage cans, aerosol cans

POT

packaging: others

foil for ﬂexible packaging, semirigid food containers

ME

machinery and equipment

irrigation pipe, ladders, oﬃce and hospital equipment

ECA

electrical: cables

wire, cables

EOT CD

electrical: others consumer durables

transformers and capacitors, electric lamps air conditioners, refrigerators, dishwashers, cookware

OTN

other uses: non-destructive use

other uses except destructive use

OTD

destructive uses

metallurgical products for steelmaking

theoretical emission reduction potentials are quantiﬁed and several realistic options are discussed.

2. METHODS The historic U.S. aluminum cycle was calculated using a system deﬁnition described in Figure 1. Primary aluminum is produced by electrolytic reduction of alumina (Hall Heroult smelting). Molten aluminum from smelters is alloyed, cleaned, and cast into diﬀerent kinds of ingots. These ingots are further transformed into diﬀerent semiproducts mainly by rolling, extrusion, and casting, and eventually manufactured into ﬁnal products. New scrap from all stages of production and manufacturing and old scrap from products leaving use are recycled in reﬁners and remelters. The processes Manufacturing and Use are divided into 12 subprocesses reﬂecting diﬀerent product categories (Table 1). All ﬂows and stocks were calculated as aluminum metallic equivalent using a dynamic MFA model as detailed in the

Supporting Information. Three approaches were used to calculate the ﬂows: use of industry and government statistical data, calculation by transfer coeﬃcients estimated from literature and expertise, and derivation by the mass balance principle. Aluminum in products leaving use and accumulated in in-use stock was calculated based on historic consumption and lifetime for each product category. To get more accurate consumption data, the statistical shipment data of diﬀerent product categories to manufacturing were adjusted by assumed yield ratios, and historic trade statistics of around 110 products were used to account for the import and export ﬂows of aluminum contained in parts and ﬁnal products (“indirect trade”). Recent years after 2006 were excluded because of incomplete trade data. Due to a lack of historical data regarding the age of products upon disposal, a parameter variation was performed for the lifetimes (normal distribution) of all product categories. Energy use and GHG emissions corresponding with the U.S. aluminum cycle were calculated using coeﬃcients based on the 9516

dx.doi.org/10.1021/es202211w |Environ. Sci. Technol. 2011, 45, 9515–9522

Environmental Science & Technology

POLICY ANALYSIS

Figure 2. Historic aluminum ﬂows in the U.S. assembled by markets, 1900 2008. Mt/a = million metric tons per annum.

output of each process. All processes were considered except Manufacturing and Use, because emissions from these two processes are diﬃcult to allocate to a single material and instead reported by other sectors. Energy use was diﬀerentiated between that occurring on-site at each process (primary), and that occurring oﬀ-site from “well-to-gate” in the energy infrastructure (secondary). Besides the consequent primary and secondary energy related emissions, a third type of GHG emissions is called process emissions, which mainly includes CO2 emissions from anode production and consumption and PFC emissions from smelters (CF4 and C2F6). Emissions from transportation of raw materials and products, which were estimated to have a 5% share of total emissions for the global aluminum cycle,23 were excluded in the model, except for those related to energy carriers. Emission mitigation options were analyzed based on the result for 2006. Theoretical reduction potentials for those identiﬁed process technology and system-wide improvements were each calculated as the diﬀerence between current total emissions and emissions with the entire mitigation potential achieved.

The entire mitigation potentials for the major processes were calculated as the according emissions of theoretical minimum energy requirements reported by the U.S. Department of Energy.24 Several realistic options were identiﬁed and quantiﬁed using a sensitivity analysis approach. A hypothetical saturated in-use stock situation is compared to the current situation to illustrate the importance of stock dynamics on total emissions and mitigation options.

3. RESULTS AND DISCUSSION 3.1. Historic Flows and Stocks. The historic U.S. aluminum flows are assembled by markets (bauxite, alumina, aluminum, semiproducts, final products, and scrap) in Figure 2. All markets show a slow initial development and a rapid penetration after World War II. Thereafter the production curves from upstream to downstream show a progressive decoupling, which demonstrates the substitution of domestic production by increasing net import of aluminum-containing goods (aluminum, semis, and 9517

dx.doi.org/10.1021/es202211w |Environ. Sci. Technol. 2011, 45, 9515–9522

Environmental Science & Technology

POLICY ANALYSIS

Figure 3. Historic U.S. aluminum in-use stock: absolute (left) and per capita (right), 1900 2006. The thick lines indicate the medium lifetime scenario, and the bottom and top dot lines are for the long and short lifetime scenarios, respectively. See lifetime scenarios in detail in Tables S3 and S4 in the Supporting Information.

final products) and the recovery of scrap. While total domestic production growth of all markets slowed down since the 1970s, the total net import has doubled to satisfy the growing demand for final products. In general, the U.S. is a net importer of aluminum in all preconsumer markets and a net exporter of scrap, a phenomenon similar to the U.S. iron cycle.25 The import dependency has gradually shifted along the production chain, i.e., from bauxite before 1970s, to alumina around 1980s, and aluminum and final products after 1990s, which increases not only foreign dependence but also the potential leakage of embodied GHG emissions of the U.S. aluminum industry to other countries. Since 2001, secondary recovery exceeds primary production, and castings production, which predominantly uses recycled aluminum, approaches the level of extrusion. However, a large share of secondary production still comes from new scrap (around 60% in recent years), which reflects inefficiency of the whole aluminum processing and production chain and increases GHG emissions as a result of additional energy use for new scrap remelting. The absolute and per-capita U.S. aluminum in-use stocks in 2006 were estimated to be 146 Mt and 490 kg, respectively, for the medium lifetime scenario (Figure 3). The categories BC, TAU, and TOT constitute the largest components of the in-use stock, representing about two-thirds of the total in 2006. The packaging categories (PCA and POT), though having a high share in annual consumption, form a negligible in-use stock due to short product lifetimes. Our absolute in-use stock result of 2006 generally agrees with the crude estimation of the United States Geological Survey (USGS) for 2002, 142 Mt,26 but is higher than the estimation of Hatayama et al. with 120 Mt for 2003 17 and McMillan et al. with 91.1 97.6 Mt for 2007.18 The diﬀerences most probably come from our explicit consideration of manufacturing scrap and full inclusion of indirect trade. The total net import relative to the total apparent consumption of aluminum ﬁnal products varies from 0 to 15% since 1962 and remains around 20% in recent years in our model (Figure S3 in the Supporting Information). This falls within a similar range as U.S. indirect trade of other metals,27 and is higher than the aforementioned two studies.17,18 Simulation results also show a growing trend of U.S. aluminum in-use stocks in most product categories, with a relatively small impact from lifetime assumptions. Results of the short and long lifetime scenarios diﬀer from that of the medium by only about 18% 25%, indicating the fact that aluminum in-use stocks are still growing rapidly in the U.S. This diﬀers from a

previous observation for the U.S. iron in-use stock, which saturated and has remained stable on a per-capita level since the 1980s.28 The stock increase was initially dominated by the penetration of aluminum applications in the BC category before the late 1980s, and has since gradually shifted to transportation, especially in the TAU category. This reﬂects the increasing use of aluminum in automobiles for weight reduction and consequent energy savings. Although aluminum has historically been employed in automobiles primarily in the form of castings, a wider variety of products including extrusions, stamped sheet parts, and forgings have started to penetrate the markets in recent years, and are expected to be increasingly used in car body and other structural applications.9 Nevertheless, the per-capita in-use stocks in some categories such as ECA, EOT, and PCA witnessed saturation after the 1980s, which may reﬂect the mature state of aluminum use in the electrical engineering and packaging sectors in the U.S. The average plateau levels for ECA and EOT are approximately 40 and 12 kg/cap, respectively. Used beverage cans (UBCs, obsolete PCA) form the largest source of obsolete scrap in the U.S., followed by end-of-life vehicles (obsolete TAU), and retired consumer durables (obsolete CD) (Figure 4). Aluminum in UBCs was estimated to be 1.4 Mt in 2006, making up around 26% of the total in all obsolete products. This can be explained by their high apparent consumption and short lifetimes. The obsolete BC ﬂow, on the contrary, only comes in ﬁfth due to its long lifetime, despite having the largest in-use stock. However, the continuously declining collection ratio of UBCs and the fast growing net export of old scrap in recent years resulted in a substantial decrease of domestic recovery. Additionally, since the obsolete product markets are poorly understood and thus diﬃcult to be quantiﬁed, the simulated aluminum ﬂow leaving use is generally higher than the aluminum ﬂow entering collection (the sum of landﬁlled29 and collected30). The diﬀerence indicates an important gap of the generation and use of obsolete products, e.g., 0.4 Mt in 2006, which could be further explored from two aspects: (i) the stock growth of obsolete products, e.g., uncollected used products in the back yard, which is not yet understood suﬃciently; and (ii) the net export of obsolete products, especially second-hand vehicles, which is not tracked by trade statistics. Hence, the fraction of the two unknown ﬂows could not be determined directly, but it was estimated that the U.S. exported 1.21 million old passenger cars in 2005,31 which is approximately equal to 0.15 Mt aluminum, or 35% of the gap. 9518

dx.doi.org/10.1021/es202211w |Environ. Sci. Technol. 2011, 45, 9515–9522

Environmental Science & Technology

POLICY ANALYSIS

Figure 4. Simulated U.S. historic aluminum ﬂows in obsolete products and comparison with USGS reported recovery:30 diﬀerent categories (left) and the total (right), 1900 2006.

Figure 5. The U.S. anthropogenic aluminum cycle in 2006. Detail may not add to totals due to rounding.

3.2. Energy Use and GHG Emissions of the Contemporary U.S. Aluminum Cycle. The stocks and flows presented above

were assembled to generate the historic U.S. aluminum cycle at a detailed level. A snapshot for 2006 is visualized in Figure 5 and its consequent energy use and GHG emissions were calculated. The results indicate that 464 PJ or 129 TWh of energy was expended in total in 2006, which is 31% lower than the value calculated by the U.S. Department of Energy for 2003.24 This is reasonable considering primary production fell by 16% in the U.S. between 2003 and 2006, and most of the process energy data used in that study were from 1995 while data in our model are mostly from 2005 or more recent. Thus the difference may reflect process improvements in that ten-year span. The total GHG emissions amount to about 38 Mt CO2-equivalence, which is equal to 0.53% of total U.S. GHG emissions in 2006.32 This proportion is lower than the world average, as global aluminum production was estimated to cause 1% of global GHG emissions.1 However, it should be noted that emissions embodied in trade are not considered in our territorial (or production-based) model. Because the U.S. aluminum cycle is highly dependent on imports, the results of a consumption-based approach are expected to be higher. This assertion is supported by the study of primary aluminum ingot in North America.6 When broken down by process and source (Figure 6), the dominating smelting process (71%) and reﬁning and anode

Figure 6. GHG emissions of the U.S. aluminum cycle in 2006, by process and emission source.

production together comprise 83% of total emissions, which is much higher than the semimanufacturing processes (11%) and scrap remelting and reﬁning (5%). Secondary emissions are the most signiﬁcant source at 67%, followed by natural gas and process emissions, indicating that a large share of emissions is induced by electricity use. With the other fossil fuels contributing 9519

dx.doi.org/10.1021/es202211w |Environ. Sci. Technol. 2011, 45, 9515–9522

Environmental Science & Technology

POLICY ANALYSIS

Figure 7. Theoretical reduction potentials and realistic measures on global warming potential (GWP-100) of the U.S. aluminum cycle under the current situation and a hypothetical in-use stock saturation situation.

relatively much less, it appears that fuel switching has limited potential left as a mitigation measure. 3.3. Options for GHG Emissions Mitigation. The emission reduction potentials of several identified mitigation options are presented in Figure 7. They are listed with theoretical potentials serving as umbrella categories for more realistic short-term and longterm measures. Emission reductions are displayed across four aggregated process categories (mining and refining, smelting, semimanufacturing, and waste management and recycling) of the aluminum life cycle to visualize where emission savings are occurring. The theoretical reduction potential for all process technology improvements adds up to a 61% reduction relative to the baseline situation in 2006. The majority of reduction potential lies within smelting due to its high electricity intensity and 58% coal power share in the contract mix. While upgrading U.S. reﬁners to the world best available technology (BAT) is not expected to make

a signiﬁcant diﬀerence, BAT in smelting has the largest potential of all the strategies deemed available in the short-term (13% reductions), though this may be diﬃcult and costly considering the long lifetime and capital investment of smelting potlines. The inert anode cell paired with the wetted cathode is expected to bring emission reductions reaching to 23% of the total, though technical and commercialization feasibility has not yet been proven.1 Continuous strip casting allows molten metal to be directly cast into slab or strip and has demonstrated 25% energy savings for sheet and foil production,24 though when applied to the system the emission savings amount to only a 1.5% reduction. Finally, oxy-fuel combustion, in which natural gas is burned with pure oxygen to increase energy eﬃciency, has been demonstrated in an aluminum remelting furnace with overall energy savings of 50 60%.2 When this technology is applied to all scrap remelting and reﬁning in the system, 2% emission reductions are achieved. 9520

dx.doi.org/10.1021/es202211w |Environ. Sci. Technol. 2011, 45, 9515–9522

Environmental Science & Technology Whereas process technology improvements deal typically with improving process energy eﬃciency and end-of-pipe treatment, system-wide improvements capture synergies between processes and sectors that facilitate more radical material eﬃciencies, smarter energy sourcing, and better integration of technological, social, and policy factors. Low emission energy, which implies that all electricity and fuels for the aluminum cycle can be produced with near-zero emissions, shows the highest theoretical potential (83% reduction) of those analyzed. Speciﬁc options include fuel switching to natural gas in the short term, and decarbonizing the electricity supply and applying carbon capture and storage (CCS) to coal power plants in the long term. The next three potentials address the most signiﬁcant sources of unrecovered metal in the system: landﬁlling, scrap export, and obsolete product stock accumulation and export. Each has a large theoretical reduction potential (81%, 47%, and 15%, respectively), since all additionally recovered old scrap was assumed to replace metal that would have come from primary production. Cans and durable goods are the major sources of aluminum entering landﬁlls, with the 2006 can collection rate at a staggeringly low 43%29 compared to the 70% world average33 and the durable goods collection at a negligible level.29 Improving can collection to 90% and durable goods collection to 50% are two signiﬁcant measures that can reduce system-wide emissions by 19% each. Keeping scrap and obsolete products within the U.S. formal recycling system has moderate reduction potentials, while it is less practical and makes no diﬀerence from a global perspective. Next, improving semimanufacturing and manufacturing yield reduces scrap generation at these processes which reduces both energy required for scrap melting and recasting and metal loss to unrecovered dross. Each can achieve around 10% theoretical reductions with yield maximization. Avoiding scrap remelting in a “nondestructive recycling” (e.g., extrusion by solid bonding of new scrap) system34 is a potential practical measure here, though this strategy does not address the actual generation of scrap. If the stock approaches saturation, total emissions will decrease dramatically and mitigation priorities will change signiﬁcantly as much more old scrap becomes available for recycling. In the case of full saturation (no net addition to stock) on the 2006 level, secondary production has the potential to replace domestic primary production, decrease the current reliance on imported ingot, and reduce total emissions by 81%. This would help eliminate domestic emissions from primary production and shift the focus to semimanufacturing and recycling processes. Therefore the nature of in-use stock development over the coming decades will be critical for industry and government to prioritize relevant innovation investments and policy instruments for deeper emission reductions. Ironically, increasing domestic recycling will actually increase territorial based emissions under this situation (in the range of 1 6%), because the consequent recovery increase further exceeds demand, which will therefore either be exported or replace imports. This demonstrates that a territorial based emissions accounting provides little incentives for global emissions reduction. It is worth mentioning that in many cases emission reduction potentials between diﬀerent measures cannot be summed due to complex system interactions. The short-term and long-term scenarios illustrate how recent and future mitigation potential can be determined by combining measures. It is evident that short-term measures alone (23% reduction) are not signiﬁcant enough to meet the IPCC target with the current system, but

POLICY ANALYSIS

with long-term potentials (87% reduction) the cycle is able to meet the target. Under the stock saturation situation, the longterm scenario achieves reductions of 58%. This is due largely to the application of CCS on coal power for semimanufacturing electricity consumption, a measure that brings relatively modest reductions in the current situation. Challenges are likely to arise, however, to achieve those realistic potentials and approach those theoretical potentials in practice. Besides the massive investment requirement and uncertain feasibility of new process technologies, the lock-in eﬀect would further hamper technology penetration and the consequent emissions reduction. Practical accessibility of old scrap is highly dependent on the socio-economic context, consumer behaviors, and recycling infrastructure. And more importantly, the thermodynamic barriers and accumulation of tramp elements over time through repeated recycling will introduce huge challenges to materials engineering and product design.35 The remanufacturing of products36 and intelligent use of aluminum such as lightweighting for automobiles are acknowledged to reduce GHG emissions as well. However, the sectoral approach applied here is unable to include the energy use and emissions in Manufacturing and Use due to complexities with allocation arising from the intricate web of materials, processes, and sectors these phases contain. Nevertheless, the model we developed is an example of how the intimate connection between material metabolisms and associated energy and emission ﬂows can be explored using a dynamic MFA framework. The analyzed patterns of U.S. aluminum cycle and in-use stock growth can shed light on other countries and materials. The immense importance of in-use stock dynamics for recycling and emission reduction illustrated here provides a new perspective to inform long-term industry and government policy.

’ ASSOCIATED CONTENT

bS

Supporting Information. Model details, parameter estimations and data sources, and details of diﬀerent mitigation strategies. This information is available free of charge via the Internet at http://pubs.acs.org/.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]; tel: + 47 73594754; fax: +47 73591298.

’ ACKNOWLEDGMENT We thank Georg Rombach from Norsk Hydro for fruitful discussions. Special thanks also go to Henry F. Sattlethight and Nicholas A. Adams from the Aluminum Association and E. Lee Bray from United States Geological Survey for providing part of the data and helpful clariﬁcation. ’ REFERENCES (1) IEA. Energy Technology Transitions for Industry: Strategies for the Next Industrial Revolution; The International Energy Agency (IEA): Paris, France, 2009. (2) IPCC. Climate Change 2007: The Fourth Assessment Report of the Intergovernmental Panel on Climate Change; Cambridge University Press: Cambridge, U.K. and New York, 2007. 9521

dx.doi.org/10.1021/es202211w |Environ. Sci. Technol. 2011, 45, 9515–9522

Environmental Science & Technology (3) Allwood, J. M.; Cullen, J. M.; Milford, R. L. Options for Achieving a 50% Cut in Industrial Carbon Emissions by 2050. Environ. Sci. Technol. 2010, 44 (6), 1888–1894. (4) Results of the 2009 Anode Eﬀect Survey; International Aluminium Institute: London, 2010. (5) Environmental Proﬁle Report for the European Aluminium Industry: Life Cycle Inventory Data for Aluminium Production and Transformation Processes in Europe; European Aluminium Association: Brussels, 2008. (6) McMillan, C. A.; Keoleian, G. A. Not All Primary Aluminum Is Created Equal: Life Cycle Greenhouse Gas Emissions from 1990 to 2005. Environ. Sci. Technol. 2009, 43 (5), 1571–1577. (7) Koltun, P.; Tharumarajah, A.; Grandﬁeld, J. Greenhouse Emissions in Primary Aluminium Smelter Cast Houses - A Life Cycle Analysis. Mater. Sci. Forum 2010, 630, 27–34. (8) Kim, H.-J.; McMillan, C.; Keoleian, G. A.; Skerlos, S. J. Greenhouse Gas Emissions Payback for Lightweighted Vehicles Using Aluminum and High-Strength Steel. J. Ind. Ecol. 2010, 14 (6), 929–946. (9) Bertram, M.; Buxmann, K.; Furrer, P. Analysis of greenhouse gas emissions related to aluminium transport applications. Int. J. Life Cycle Assess. 2009, 14 (0), 62–69. (10) Aluminium: Massestrømsnanalyse og vurdering af muligheder for reduktion af tab (Aluminium: Substance ﬂow analysis and loss reduction feasibility study); Danish Environmental Protection Agency (In Danish): Copenhagen, 1999. (11) Amicarelli, V.; Lagioia, G.; de Marco, O. Aluminium industrial metabolism: A commodity science contribution. Forum Ware Int. 2004, 1, 1–11. (12) Aluminum Recycling in the United States in 2000; United States Geological Survey: Washington, DC, 2006. (13) Chen, W.; Shi, L.; Qian, Y. Substance ﬂow analysis of aluminium in mainland China for 2001, 2004 and 2007: Exploring its initial sources, eventual sinks and the pathways linking them. Resour., Conserv. Recycl. 2010, 54 (9), 557–570. (14) Bever, M. B. The recycling of metals II. Nonferrous metals. Conserv. Recycl. 1976, 1 (1), 137–147. (15) Melo, M. T. Statistical analysis of metal scrap generation: The case of aluminium in Germany. Resour., Conserv. Recycl. 1999, 26 (2), 91–113. (16) Iron, Steel and Aluminium in the UK: Material Flows and their Economic Dimensions; Centre for Environmental Strategy, University of Surrey: London, 2004. (17) Hatayama, H.; Daigo, I.; Matsuno, Y.; Adachi, Y. Assessment of the Recycling Potential of Aluminum in Japan, the United States, Europe and China. Mater. Trans. 2009, 50 (3), 650–656. (18) McMillan, C. A.; Moore, M. R.; Keoleian, G. A.; Bulkley, J. W. Quantifying U.S. aluminum in-use stocks and their relationship with economic output. Ecol. Econ. 2010, 69 (12), 2606–2613. (19) Bruggink, P. R.; Martchek, K. J. Worldwide recycled aluminum supply and environmental impact model, Light Metals. In The 133th TMS (The Minerals, Metals, and Materials Society) Annual Meeting, Charlotte, North Carolina, 2004; Tabereaux, A. T., Ed.; Charlotte, North Carolina, 2004; pp 907 911. (20) Cheah, L.; Heywood, J.; Kirchain, R. Aluminum Stock and Flows in U.S. Passenger Vehicles and Implications for Energy Use. J. Ind. Ecol. 2009, 13 (5), 718–734. (21) Dahlstr€om, K.; Ekins, P. Combining economic and environmental dimensions: Value chain analysis of UK aluminium ﬂows. Resour., Conserv. Recycl. 2007, 51 (3), 541–560. (22) Schwarz, H. G.; Briem, S.; Zapp, P. Future carbon dioxide, emissions in the global material ﬂow of primary aluminium. Energy 2001, 26 (8), 775–795. (23) Life Cycle Assessment of Aluminium: Inventory Data for the Primary Aluminium Industry - Year 2005 Update; International Aluminium Institute: London, 2007. (24) U.S. Energy Requirements for Aluminum Production: Historical Perspective, Theoretical Limits and Current Practices; U.S. Department of Energy: Washington, DC, 2007.

POLICY ANALYSIS

(25) M€uller, D. B.; Wang, T.; Duval, B.; Graedel, T. E. Exploring the engine of anthropogenic iron cycles. Proc. Natl. Acad. Sci., U.S.A. 2006, 103 (44), 16111–16116. (26) Metal stocks in use in the United States; Fact Sheet 3090; U.S. Geological Survey: Washington, DC, 2005. (27) Johnson, J.; Graedel, T. E. The “Hidden” Trade of Metals in the United States. J. Ind. Ecol. 2008, 12 (5 6), 739–753. (28) M€uller, D. B.; Wang, T.; Duval, B. Patterns of Iron Use in Societal Evolution. Environ. Sci. Technol. 2011, 45 (1), 182–188. (29) Municipal Solid Waste Generation, Recycling, and Disposal in the United States: Facts and Figures; U.S. Environmental Protection Agency: Washington, DC, 1960 2009. (30) Minerals Yearbook: Aluminum; United States Geological Survey: Washington, DC, 2006. (31) Fuse, M.; Kosaka, H.; Kashima, S. Estimation of world trade for used automobiles. J. Mater. Cycles Waste Manage. 2009, 11 (4), 348–357. (32) Inventory of U.S. Greenhouse Gas Emissions and Sinks: 1990 2009; U.S. Environmental Protection Agency: Washinton, DC, 2011. (33) Global Aluminium Recycling: A Cornerstone of Sustainable Development; The Global Aluminium Recycling Committee, International Aluminium Institute: London, 2009. (34) Allwood, J. M.; Ashby, M. F.; Gutowski, T. G.; Worrell, E. Material eﬃciency: A white paper. Resour., Conserv. Recycl. 2011, 55 (3), 362–381. (35) Gaustad, G.; Olivetti, E.; Kirchain, R. Toward Sustainable Material Usage: Evaluating the Importance of Market Motivated Agency in Modeling Material Flows. Environ. Sci. Technol. 2011, 45 (9), 4110–4117. (36) Gutowski, T. G.; Sahni, S.; Boustani, A.; Graves, S. C. Remanufacturing and Energy Savings. Environ. Sci. Technol. 2011, 45 (10), 4540–4547.

9522

dx.doi.org/10.1021/es202211w |Environ. Sci. Technol. 2011, 45, 9515–9522

ARTICLE pubs.acs.org/est

Novel Methoxylated Polybrominated Diphenoxybenzene Congeners and Possible Sources in Herring Gull Eggs from the Laurentian Great Lakes of North America Da Chen,*,†,‡ Robert J. Letcher,*,†,‡ Lewis T. Gauthier,† Shaogang Chu,† Robert McCrindle,§ and Dave Potter|| †

)

Wildlife and Landscape Directorate, Science and Technology Branch, Environment Canada, National Wildlife Research Centre, Carleton University, Ottawa, ON, K1A 0H3, Canada ‡ Department of Chemistry, Carleton University, Ottawa, ON, K1S 5B6, Canada § Chemistry Department, University of Guelph, Guelph, ON, N1G 2W1, Canada Wellington Laboratories, Research Division, Guelph, ON, N1G 3M5, Canada

bS Supporting Information ABSTRACT: An increasing number of brominated ﬂame retardants and other brominated substances are being reported in herring gull eggs from the Laurentian Great Lakes basin. Yet, in extracts from gulls’ eggs, numerous bromide anion response peaks in electron capture negative ion (ECNI) mass chromatograms remain unidentiﬁed. Using archived herring gull egg homogenates, we characterize the structures of three major and three minor, new and unique brominated substances. After extensive cleanup and separation to isolate these substances from the extracts, high-quality ECNI and electron impact (EI) mass spectra revealed fragmentation patterns consistent with congeners of methoxylated polybrominated diphenoxybenzene (MeO-PBDPB), where four congeners contained ﬁve bromines and the other two contain four and six bromines, respectively. Optimized, semiquantitative analysis revealed sum concentrations of the MeOPBDBP congeners ranged from <0.2 to 36.8 ng/g ww in pooled egg homogenates (collected in 2009) from fourteen herring gull colony sites across the Great Lakes, with the highest concentration being for ChannelShelter Island in Saginaw Bay (Lake Huron). To our knowledge, there are no published reports on the environmental presence and sources of MeO-PBDPBs. We hypothesize that these MeO-PBDPBs are degradation products of the polybrominated diphenoxybenzenes, for example, tetradecabromodiphenoxybenzene (currently marketed as SAYTEX 120) or polybromo 3P2E. MeOPBDPBs in Great Lakes herring gull eggs indicates their bioaccumulation potential, and raises concerns about their origin, environmental behavior and inﬂuences on wildlife and environmental health.

’ INTRODUCTION Brominated ﬂame retardants (BFRs) are used to protect the public from ﬁres by reducing the ﬂammability of combustible materials, and comprise a large and diverse class of additive and reactive substances used in commercial polymeric materials. As recently reviewed in Covaci et al.,1 to date, at least 75 diﬀerent BFRs have been commercially produced. Environmental studies have focused largely on polybrominated diphenyl ethers (PBDEs), polybrominated diphenyls (PBBs), hexabromocyclododecanes (HBCDs) and tetrabromobisphenol A (TBBP-A). As a consequence of substantial, longer-term use, these BFRs have been reported in humans, wildlife, air, water, soil, and/or sediment, even in samples from remote areas such as the Arctic.14 The subsequent restrictions on PBDE applications have resulted in the increased use of alternative ﬂame retardant chemicals, either newly engineered or already in continuous use, to meet ﬂammability standards. Replacement BFRs other than PBDEs, HBCD and TBBP-A and that are new to the market or newly/recently observed r 2011 American Chemical Society

in the environment include decabromodiphenylethane (DBDPE), 1,2-bis(2,4,6-tribromophenoxy)ethane (BTBPE), 2-ethyl-1-hexyl 2,3,4,5-tetrabromobenzoate (EHTBB), 1,2-dibromo-4-(1,2-dibromoethyl)cyclohexane (TBECH) and bis(2-ethyl-1-hexyl)tetrabromophthalate (BEHTBP).58 However, many others may be currently produced and in use by industry and yet remain unknown in the environment. Brominated substances in the environment, including biota, are not necessarily BFRs, but rather can have a variety of sources. Numerous BFRs such as PBDEs have been reported to degrade and/ or metabolize to other forms that themselves are environmentally bioaccumulative.912 For example, BDE-209 and other PBDEs may debrominate in ﬁsh and wildlife.1012 Metabolic processes convert Received: April 25, 2011 Accepted: October 3, 2011 Revised: September 28, 2011 Published: October 03, 2011 9523

dx.doi.org/10.1021/es201325g | Environ. Sci. Technol. 2011, 45, 9523–9530

Environmental Science & Technology

Figure 1. Locations of the 14 herring gull colony sites from the Laurentian Great Lakes: (1) Granite Island; (2) Agawa Rocks; (3) Big Sister Island; (4) Gull Island; (5) Double Island; (6) Channel-Shelter Island; (7) Chantry Island; (8) Middle Island; (9) Port Colborne; (10) Niagara River; (11) Hamilton Harbour; (12) Toronto Harbour; (13) Snake Island; and (14) Strachan Island. The gray, white and black bars represent the concentrations (ng/g wet weight) of three major MeO-PBDPB congeners U1, U2, and U3, respectively. A complete summary of the concentrations of all MeOPBDPB congeners, as well as polybrominated diphenyl ethers (PBDEs) and hexabromocyclododecane, in egg homogenate pools from the fourteen colony sites was provided in the Table S1 (Supporting Information).

certain PBDEs (e.g., BDE-47) to hydroxylated PBDE products, which themselves can be persistent in an organism.12 Furthermore, brominated substances can be of natural origin rather than being anthropogenic. Gribble13 reviewed the vast array of naturally occurring halogenated compounds in marine life including algae, sponges, and other animals. Overall, the extent of possible, yet environmentally unknown ﬂame retardants and other brominated substances raises new challenges and concerns for scientists to investigate their sources, bioavailability and ecosystem behavior. The herring gull (Larus argentatus) has been considered a key bioindicator species in the Laurentian Great Lakes of North America,14 and recently we have characterized several novel FRs in the Great Lakes herring gull eggs.57,15 However, in the BFR-containing fractions isolated from these herring gull eggs, numerous (and some rather intense) bromide anion responsive peaks observed in electron capture negative ionization (ECNI) mass spectra remain to be identiﬁed.5 The current study reports on the identiﬁcation of several novel brominated contaminants, as well as quantiﬁcation and assessment of recent spatial distribution and potential sources, in herring gull eggs from fourteen colony sites spanning the Laurentian Great Lakes.

’ EXPERIMENTAL SECTION Standards and Chemicals. To our knowledge, there are currently no pure standards commercially available for methoxylated

ARTICLE

polybrominated diphenoxybenzene congeners. Diatomaceous earth (DE) was purchased from J.T. Baker (Mallinckrodt Baker, NJ) and treated in a muffle furnace at 600 °C overnight (>12 h) prior to use. Bakerbond SPE silica gel (SiOH) disposable, solid phase extraction (SPE) columns (6 mL, 500 mg, 4760 μm) were purchased from VWR (Mississauga, ON, Canada). Solvents used were HPLC grade (Caledon Laboratories, Georgetown, ON, Canada). The 6-MeOBDE-137 and 40 -MeO-BDE-201 standards were purchased from AccuStandards Inc. (New Haven, CT). Samples. For qualitative identification of target unknown analytes, herring gull eggs were collected annually during the 1999 2001 period from the Channel-Shelter Island (Lake Huron) colony. These eggs (n = 1013 each year) were pooled and archived at 40 °C at Environment Canada’s National Wildlife Specimen Bank. Herring gull eggs collected in late-April to early-May of 2009, from 14 colony sites in the Laurentian Great Lakes (Figure 1), were used to quantify and examine the spatial distribution of target analytes. For each colony site, n = 1013 individual eggs were pooled on an equal wet weight basis. The Great Lakes egg samples were collected as part of Environment Canada’s Great Lakes Herring Gull Monitoring Program.16 Egg Sample Preparation. Approximately 40 g of pooled egg homogenate from Channel-Shelter Island (19992001) were extracted in ten replicates (4 g each). Each replicate was ground with DE and then subjected to accelerated solvent extraction (ASE) (Dionex ASE 200, Sunnyvale, CA) with 50:50 dichloromethane:hexane (DCM:HEX). A method blank (DE only) was also extracted along with the egg samples. The extract was purified by gel permeation chromatography (GPC) (OI Analytical, TX) followed by cleanup on a Bakerbond SPE cartridge. After prewashing with 6 mL HEX, the target analytes were eluted from the cartridge with 8 mL 20:80 DCM:HEX. The generated fraction was further cleaned and separated on a column (250 11 mm i.d.) packed with 7-g silica gel (Grade 62, 60200 mesh, Sigma-Aldrich). The first fraction containing PBDEs was eluted with 100 mL HEX, followed by 30 mL 20:80 DCM:HEX. The second fraction eluted with 16 mL 80:20 DCM:HEX contained the target analytes and were concentrated for various gas chromatographymass spectrometry analyses. Gas ChromatographyLow Resolution Mass Spectrometry (GCLRMS). The target compounds were analyzed using an Agilent 6890 GC (Agilent Tech., Palo Alto, CA) coupled to a low resolution, single quadrupole mass analyzer (Agilent 5973 MS), and in both ECNI and electron impact (EI) ionization modes. The column used was a 15-m DB-5 HT column (0.25 mm i.d., 0.1 μm, J&W Scientific, Agilent Tech.). The injector was operated in pulsed-splitless mode, held at 240 °C. The initial oven temperature was held at 100 °C for 2 min, increased to 250 at 25 °C/min, then to 260 at 1.5 °C/min, and finally to 325 at 25 °C/min (held for 7 min). For EI (ion source 230 °C) and ECNI (ion source 250 °C), full scan (scan range 30800 m/z) mass spectra were generated for the target analytes. The heated transfer line temperature was 280 °C and the quadrupole temperature was 150 °C. Methane was used as reagent gas for ECNI-MS. Figure 2 shows the GC full scan ECNI mass chromatograms of a partial time range of 14 to 18 min for the Channel-Shelter herring gull egg fraction (showing three major (U1, U2, and U3) and three minor (U4, U5, and U6) analytes) and PBDE standard mixture (containing BDE-194, -205, and -206). Gas ChromatographyHigh Resolution Mass Spectrometry (GCHRMS). Accurate mass determination was performed using an Agilent 6890 GC coupled to a Waters Autospec Ultima MS. 9524

dx.doi.org/10.1021/es201325g |Environ. Sci. Technol. 2011, 45, 9523–9530

Environmental Science & Technology

ARTICLE

Figure 2. Gas chromatographylow resolution mass spectrometry (electron-capture negative ionization mode; full scan m/z 30 800 amu) mass chromatograms of methoxylated polybrominated diphenoxybenzene (MeO-PBDPB) congeners (U1U6) in a fraction from a representative egg pool from Channel-Shelter herring gulls (top) and a PBDE standard mixture (bottom). The mass chromatograms are for a partial GC elution time range of 1418 min.

The fraction containing target analytes was injected onto the GCHRMS system and analyzed in full-scan mode (501000 amu). Voltage scan experiments were created to scan narrow mass ranges that encompassed the molecular fragment of interest. These narrow mass ranges also included 1 to 3 reference peaks from the mass-calibrant, perfluorokerosene (PFK). The instrument was tuned to a resolution of at least 10 000. A calibration curve was constructed using the PFK masses in a portion of the TIC where no peaks eluted. A secondary reference correction (Masslynx 4.1) was applied to the acquired data file to autocorrect each mass in the spectrum using the PFK reference peaks. The corrected data file was then opened and the background was subtracted from the region of interest to generate the final mass spectrum. The Masslynx 4.1 elemental composition program was utilized to determine the best fitting formula for each isotope signal in the mass spectrum of interest. All injections

were in the splitless mode at a temperature of 280 °C. The oven temperature program was the same as that for GCLRMS analysis. Quantitative Analysis in Herring Gull Eggs. The target analytes (U1U6) (Figure 2) are identified as methoxylated polybrominated diphenoxybenzene (MeO-PBDPB) congeners, and the characterization is described in detail in the Results and Discussion section. To semiquantify the six MeO-PBDPBs, approximately 1 g egg pool homogenate from each of fourteen gull colonies and two method blanks were spiked with 20 ng of 6-MeO-BDE-137 as an internal standard and subjected to ASE extraction. After gravimetric determination of lipid content using 10% of the ASE extract, the remaining extract was subjected to GPC, followed by SPE cleanup. The fraction eluted with 8 mL 20:80 DCM:HEX was evaporated and reconstituted to 250 μL in isooctane, and analyzed on the Agilent 6890 GC-5973 MS using 9525

dx.doi.org/10.1021/es201325g |Environ. Sci. Technol. 2011, 45, 9523–9530

Environmental Science & Technology

ARTICLE

Figure 3. Electron-capture negative ionization (scan range m/z 30 800 amu) and electron impact (scan range m/z 30 800 amu) mass spectra (GCLRMS) of the methoxylated polybrominated diphenoxybenzene (MeO-PBDPB) congener U1 in the extract of Channel-Shelter Island herring gull egg. The base structure of this congener is also shown (hydrogen atoms are omitted for clarity).

ECNI mode. The analytical column and temperature program, as well as GC parameters, were the same as those described in the previous section. Semiquantification of target analytes U1, U2, U5, and U6 was achieved via selected ion monitoring (SIM) for 79 Br and 81Br, and based on the calibration curve for the 40 MeO-BDE-201. This could be accomplished due to the comparable molecular structures and equitable cleanup and isolation from egg homogenate, as well as their close retention times on GC (i.e., 16.25 min for 40 -MeO-BDE-201 versus 16.17 min for U1). Due to the lack of suitable methoxylated PBDE standards

that elute on GC at similar retention times as U3, BDE-206 was used instead as the quantification standard for U3 (Figure 2). This could be accomplished as the quantification based on the calibration curves for 40 -MeO-BDE-201 and BDE-194 produced comparable results for U1 or U2 (Supporting Information (SI) Table S1). Although the peaks of BDE-206 and U3 were slightly overlapped (Figure 2), the quantification results for U3 via selected ion monitoring (SIM) for 79Br and 81Br were deemed credible, as our previous report indicated that the BDE-206 concentrations in herring gull eggs from the same locations were 9526

dx.doi.org/10.1021/es201325g |Environ. Sci. Technol. 2011, 45, 9523–9530

Environmental Science & Technology

ARTICLE

Figure 4. An electron impact (positive) ionization fragmentation pathway scheme of one representative, methoxylated polybrominated diphenoxybenzene (MeO-PBDPB) congener (Br5). Hydrogen atoms are omitted for clarity.

mostly below 0.1 ng/g ww.5 For the same reason, U4 was semiquantified based on the calibration curve for BDE-170. Quality Control and Assurance. The precision, accuracy and recovery efficiency of the MeO-PBDPB analytical method was assessed using commercial chicken eggs (n = 6) spiked with 20 ng each of BDE-194, BDE-206 and 6-MeO-BDE-137. The mean ((standard deviation) recoveries were 91((8.6)% and 93((7.5)% for BDE-194 and -206, respectively. The recoveries of internal standard 6-MeO-BDE-137 were 95((14)%. Based on 6-MeO-BDE-137, the method limits of quantification (MLOQs) of the MeO-PBDPBs were estimated as an analyte response six times the standard deviation of the noise and they were 0.2 ng/g ww, assuming that the MeO-PBDPBs have the same ECNI responses as 6-MeO-BDE-137. MeO-PBDPBs were not detectable (nd: signal/noise <3) in any of the method blanks.

’ RESULTS AND DISCUSSION Mass Spectral Characterization of MeO-PBDPBs in Herring Gull Eggs. Thorough cleanup and separation of MeO-PBDPBs

from the egg matrix, particularly the application of 7-g silica gel SPE, greatly reduced the interferences (and thus increased the sensitivity of target MeO-PBDPBs relative to noise) in the GCMS analyses by removal of lipids and closely eluting compounds, particularly PBDEs. This is illustrated in the GCLRMS(ECNI) mass chromatograms for the extract from Channel-Shelter Island egg homogenate before and after the application of 7-g silica gel SPE (SI Figure S1). The optimally isolated MeO-PBDPB-containing fraction facilitated the acquisition of high quality ECNI and EI mass spectra for the target analytes. In the absence of pure chemical standards, the LRMS(ECNI and EI) (Figure 3; SI Figures S2S4) and HRMS(EI) (Figures S5 and S6) spectral evidence strongly supports the identity of U1 to U6 as congeners of MeO-PBDPBs. Furthermore, the EI mass spectrum for U1 (Figure 3) is consistent with the fragmentation pathway scheme for a possible U1 congener illustrated in Figure 4. We now closely examine the mass spectral data to substantiate the proposed EI fragmentation scheme and

the assertion of the unknown analytes in herring gull egg extract as being MeO-PBDPB congeners. The LRMS(ECNI) mass spectrum for U1 reveals the dominance of the [Br] fragment ion isotopes at m/z 79 and 81 amu, and to a lesser extent by the [HBr2] fragment isotope cluster at m/z 159, 161, and 163 amu (Figure 3). Ion clusters such as [M-Br], [M-HBr2] and [M-H2Br3] are present in low abundances, whereas [M] is not detectable. This ECNI ionization pattern resembles that of most MeO-PBDE congeners.17 However, the ECNI spectrum of U1 in general provides very limited structural information. Additional LRMS(EI) analysis indicates that the ion cluster centered at m/z 686 dominates the EI mass spectrum of U1 (Figure 3). Essentially no ions are observed above m/z 686 to the maximum m/z 800 amu (and to a maximum of m/z 1000 amu by GCHRMS(EI)) (SI Figure S5). Thus the ion cluster centered at m/z 686 amu is the molecular ion ([M]+), which exhibits the isotopic signature and relative abundances corresponding to ﬁve bromine atoms. The GCHRMS(EI) analysis reveals that the elemental composition of U1 is C19H11O3Br5 since the exact mass of [M]+ (12C191H1116O379Br5) is m/z 681.6005 amu and is 0.002 amu from the calculated mass (SI Figure S6). Similarly for the other major ions in the cluster, [M+2]+, [M+4]+, [M+6]+, [M+8]+ and [M+10]+, the exact masses also have the same level of mass accuracy. The EI mass spectrum for U1 (Figure 3) exhibits an abundant [M-Br2]+ fragment (with the isotopic signature of three bromine atoms), which resembles the spectra of certain PBDE congeners (e.g., BDE-47) that have bromine atoms ortho to the ether linkage.17 This [M-Br2]+ fragment is formed by the elimination of two ortho bromines, upon electron impact, and involves the formation of a dibenzofuran-like ion.1719 In addition to [M-Br2]+, a doubly charged ion of [M-Br2]2+ that is characteristic for the mass spectra of certain PBDE congeners (e.g., m/z 242 amu for BDE-153)17 is also observed in the U1 spectrum (ion cluster centered at m/z 263 amu; Figure 3). Therefore, this similarity in fragmentation behavior to give M+, [M-Br2]+ and [M-Br2]2+ indicates partial structural similarities for U1 and some PBDE congeners. The ions at m/z 356, 432, and 451 amu appear to 9527

dx.doi.org/10.1021/es201325g |Environ. Sci. Technol. 2011, 45, 9523–9530

Environmental Science & Technology be unique in the EI spectrum for U1. The ion at m/z 356 amu is the second most abundant in the U1 mass spectrum, and the HRMS(EI) analysis indicated (data not shown) its exact mass to be m/z 353.8898 amu, which corresponds to an elemental composition of C13H8O2Br2 with an error of 0.0007 amu from the calculated exact mass. The mass diﬀerence between m/z 356 and m/z 526 ([M-Br2]+) may correspond to the loss of a phenoxy group with a bromine substituent (C6H4OBr). Therefore, the transition from M+ to [M-Br2]+ and then to the ion at m/z 356 amu may represent the loss of two ortho bromine substituents to form a dibenzofuran-like structure, followed by loss of a brominated phenoxy group. A similar fragmentation pattern is consistently observed in the EI mass spectra of polychlorinated diphenoxybenzenes,20 which may suggest a diphenoxybenzene-like structure for U1. A loss of CH3Br from [M]+ (minor ion fragment cluster centered at m/z 592 amu) is observed in the LRMS(EI) spectrum of U1. For MeO-PBDE congeners with a methoxy group and a bromine atom ortho to the aromatic ether linkage, for example, 6-MeO-BDE-47, EI has been shown to produce a stable brominated dibenzo-p-dioxin ion due to the loss of CH3Br from the molecular ion.17,21 The [M-CH3Br]+ fragment ion is present in the U1 spectrum at m/z 592 amu in low abundance, whereas loss of CH3Br from [M-Br2]+ produces an abundant ion [M-Br2CH3Br]+ at m/z 432 amu. Thus, a very likely EI fragmentation pathway for U1 is formation of a dibenzofuran ion due to loss of Br2 from [M]+, followed by loss of CH3Br to form a dibenzo-pdioxin structure. Formation of such a fragment ion (m/z 432 amu) that contains both dibenzo-p-dioxin and dibenzofuran moieties appears to require a diphenoxybenzene base structure for the molecular ion, as well as a methoxy group ortho to one of the phenyl ether atoms. This proposed backbone structure, that is, diphenoxybenzene, is further supported by the presence of the ion at m/z 366 amu in the EI spectrum (Figure 3). This fragment ion cluster may be produced by loss of Br2 from the dibenzofuran ion [M-Br2]+ and consequently contains two dibenzofuran moieties. The EI and ECNI spectra of U2 were essentially the same as for U1 (SI Figure S2), and thus according to the mass spectral rationale for U1, U2 appears to have a structure very similar to that for U1, that is, a diphenoxybenzene backbone bearing a methoxy group and ﬁve bromine atoms. U3 has a molecular ion at m/z 766 amu and the HRMS(EI) analysis determines its element composition as C19H18O3Br6 (SI Figures S3 and S6). Its EI fragmentation pattern is very similar to that of U1. The important ions that characterize the U1 spectrum such as [M]+, [M-Br2]+, [M-Br2CH3Br]+ and [M-Br2Br2]+ are also present in the U3 spectrum (SI Figure S3). Similar EI fragmentation patterns were also observed in the mass spectra for the three minor analytes U4 (SI Figure S4), U5, and U6. Considering the evidence of LRMS(EI), HRMS(EI) and GCMS(ECNI) spectra combined, it is clear that the base structures of U1, U2, U5 and U6 are MeO-pentabromoDPBs, U3 a MeO-hexabromoDPB and U4 a MeO-tetrabromoDPB (Figure 2). However, explicit full structures of these congeners are diﬃcult to establish conclusively in the absence of appropriate standards. Using a potential structure for the U1 congener as an example (Figure 4), the EI mass spectra provide hints for the possible substitution patterns of the bromine atoms and the methoxy group in three phenyl rings (labeled as A, B, and C). As discussed above, the ion at m/z 356 amu (C13H8O2Br2) contains a dibenzofuran structure, two bromine atoms and a methoxy group. Assuming that the dibenzofuran structure is formed

ARTICLE

between phenyl rings A and B, these two rings must have four bromines in total before the dibenzofuran is formed (due to loss of Br2). Therefore, only one bromine atom is in ring C. The ion at m/z 453 amu (C13H8O3Br3) is believed to have a quinone structure and contains two phenyl rings, a methoxy group and three bromines (Figure 4). It may be produced by loss of a terminal ring containing two bromines from the [M]+ ion. The cleaved terminal ring, likely corresponding to the ion at m/z 235 amu, should not be ring C as this ring has only one bromine substituent. Therefore, ring A is the cleaved terminal ring and the ion at m/z 453 amu contains rings B and C. Because the ions at m/z 356 amu (containing rings A and B) and at m/z 453 amu (containing rings B and C) both contain a methoxy group, this sole methoxy group should be in ring B, and ortho to one of the phenyl ether atoms. As terminal rings A and C contain two and one bromine atoms, respectively, the remaining two bromines should be in ring B. To form the ion at m/z 432 amu that contains both dioxin and dibenzofuran moieties, the only bromine atom in ring C and at least one bromine atom in ring A should be ortho to the phenyl ether linkage. The existence of the ion containing two dibenzofuran moieties (i.e., m/z 366 amu) may suggest that the two bromines in ring B are ortho to diﬀerent phenyl ethers. However, Hites17 indicates that PBDE congeners with just one ortho bromine substituent (e.g., BDE-28) can also form a dibenzofuran ion, possibly due to a ring rearrangement. Hence some of the bromine atoms that are proposed to have an ortho position may be located on non-ortho ring sites. This raises the possibility of other congener structures. Spatial Distribution and Possible Sources of MeO-PBDPBs in Herring Gulls. The spatial distribution of MeO-PBDPB contamination in the Great Lakes basin was examined by a preliminary assessment of recent egg pool homogenates of herring gulls from fourteen colony sites (Figure 1). Highest concentrations were observed in the pool from the ChannelShelter Island (Lake Huron), where U1, U2, and U3 concentrations were 11.5, 10.9, and 13.6 ng/g ww, respectively, and the ∑MeO-PBDPB (including U1 U6) concentration was 36.8 ng/g ww (SI Table S1). Similar proportions of the three major MeO-PBDPB congeners were also measured in egg homogenates from most of the other thirteen colony sites, although comparatively to Channel-Shelter Island the concentrations were at least 1 order of magnitude lower, that is, ∑MeO-PBDPBs ranging <0.28.5 ng/g ww (SI Table S1). Since the MeOPBDPB accumulation is likely driven by lipid association, the concentrations may be more appropriately reported on a lipid weight (lw) basis, that is, ∑MeO-PBDPB concentration = 460 ng/g lw in the egg pool from Channel-Shelter Island and <2.4113 ng/g lw from the other 13 colony sites. Nonetheless, such a spatial distribution pattern indicates the likely source(s) of contamination are in the gull foraging area in proximity to the Channel-Shelter Island. The Saginaw Bay, where the ChannelShelter Island is located, is on the eastern side of the U.S. state of Michigan. A number of industrial manufacturing plants, producing elastomers, resins, fluids, polymers and many others, have been located nearby the Bay area, which may be potential MeOPBDPB sources to the Bay by direct releases of industrial wastewaters or via sewage treatment plants. The accumulation of MeO-PBDPBs in gull eggs from other colonies in the lower Great Lakes, for example., Middle Island (8.5 ng/g ww) and Toronto Harbor (3.3 ng/g ww), may also reflect the impacts from neighboring industrial and urban sources. Southern Lake Michigan is surrounded by metropolitan cities and industrial 9528

dx.doi.org/10.1021/es201325g |Environ. Sci. Technol. 2011, 45, 9523–9530

Environmental Science & Technology manufacturing plants. Although no eggs from the southern Lake Michigan colonies could be examined in this study, contamination attributed to industrial and urban influences may be expected in that region. Gulls from the northwestern Great Lakes colonies (e.g., Granite and Big Sister Island) may undertake postbreeding migrations to the south of Lake Michigan with the onset of winter.22 Increasing ice cover in the northern Great Lakes limits gulls’ access to aquatic prey, whereas southern Lake Michigan does not freeze completely. Therefore, the presence of MeO-PBDPBs in these two northwestern colonies (3.9 and 5.2 ng/g ww, respectively) may be largely attributed to their overwintering in the southern Lake Michigan area. Similarly, the colonies from northern Lake Huron (e.g., Double Island) may be exposed during their overwintering in the southern Huron or Lake Ontario areas.22 The occurrence and distribution of MeO-PBDPBs in the Great Lakes herring gull eggs diﬀered greatly from those of PBDEs as reported previously.7 The sum concentration of the seven major PBDEs (∑7PBDE: BDE-28, 47, 99, 100, 153, 154, and 183) in the same 2009-collected herring gull egg pool homogenates ranged from 240 to 545 ng/g ww (412 ng/g ww for the Channel-Shelter Island colony) (Letcher unpublished data, Table S1), exhibiting a relatively even distribution across colonies when compared to the ∑MeO-PBDPBs. The ∑PBDE concentrations were generally 13 orders of magnitude higher than the ∑MeO-PBDPBs in the same colonies. However, in some colonies (e.g., Agawa Rocks, ChannelShelter Island, Gull Island, Niagara River and Toronto Harbor), the ∑MeO-PBDPBs were comparable to or somewhat lower than HBCD (3.219.1 ng/g ww) and BDE-209 (4.337.9 ng/g ww) in the 2009-collected egg homogenates (Letcher unpublished data, Table S1). However, the ∑MeO-PBDPB concentrations were higher than several newly emerging BFRs such as BTBPE (<0.060.20 ng/g ww) and TBECH (0.110.54 ng/g ww) that had been reported in the 2006-colected herring gull egg homogenate pools.6 Like PBDEs, the distributions of BDE-209, HBCD, BTBPE, or TBECH in the Great Lakes gull eggs were diﬀerent from that of MeO-PBDPBs, for which a particular site (Channel-Shelter Island) exhibited orders of magnitude greater contamination than other sites. Statistical analyses also indicated that no signiﬁcant correlations were observed between the concentrations of ∑MeO-PBDPBs and those of ∑7PBDE, BDE-209, HBCD, BTBPE, or TBECH in egg pool homogenates from the colonies across the Great Lakes (p > 0.05 in all cases). Nonetheless, to our best knowledge, there are no published literature reports on the environmental presence of MeOPBDPBs prior to the current study. Absolutely nothing is known about their sources and environmental behavior. Enlightened by the relationship between PBDEs and MeO-PBDEs, we hypothesize that the MeO-PBDPB congeners are metabolites/degradation products of PBDPBs such as tetradecabromodiphenoxybenzene (TDBDPB; also commercially known as SAYTEX 120) or polybromo 3P2E ﬂame retardants. SAYTEX 120 is a currently used BFR that ﬁnds primary applications in high performance polyamide and linear polyester engineering resins and alloys.23 Previous studies have suggested that BDE-209 can be degraded into less brominated PBDE congeners via photochemical or metabolizing pathways.912 MeO-PBDEs may be formed in sediments via hydroxylation of PBDE congeners followed by methylation by microorganisms or in organisms via biotransformation,24 although the origin directly from PBDE metabolism was not seen

ARTICLE

in laboratory exposure studies.12,25 We also hypothesize that MeOPBDPBs are produced via transformations of PBDPBs that may be impurities in technical SAYTEX 120 or produced by degradation of SAYTEX 120. In addition to SAYTEX 120, polybromo 3P2E may be another potential source for MeO-PBDPBs. Polybromo 3P2E belongs to a general class of halogenated polyphenyl ethers that are incorporated as ﬂame retardants into a number of polymer and foam rubber products.26 It is a mixture of congeners with a general structure of polybrominated dephenoxybenzene. For example, one of the congeners introduced in the United States Patent 3760003 was 2,5-dibromo-p-bis(2,4-dibromophenoxy)benzene,26 which may be a potential precursor for MeOPBDPB-Br5. It is interesting to note that the halogenated polyphenyl ethers have been produced by a chemical company that has plants located in proximity to the Saginaw Bay,26 where the highest MeO-PBDPB concentrations were observed. Our hypotheses are primarily based on the structural similarities between MeOPBDPBs and potential precursors such as TDBDPB and polybromo 3P2E, although nothing is known about the environmental presence and behavior of these proposed precursor candidates. In our opinion, there is the possibility of exposure and accumulation via terrestrial diet consumption of herring gulls and subsequent in ovo transfer to their eggs. There is also a more remote possibility that the MeO-PBDPBs may be of natural origin, rather than being anthropogenic, and accumulated in the herring gull food web. For example, the octabrominated tetrahydroxylated diphenoxybenzene, structurally resembling the present MeO-PBDPBs, was reported as a natural product by the marine worm Ptychodera ﬂava laysanica.13 This raises the possibility that MeO-PBDPBs or precursors (including hydroxylated PBDPBs) are accumulated low in the herring gull food web, and possibly methylated via some biotransformation processes. Nonetheless, the discovery of MeO-PBDPBs in Great Lakes herring gull eggs raises concerns about their sources, environmental behavior and potential hazards to wildlife and human health. Theoretically MeO-PBDBPs may represent a group of chemicals that include a large number of congeners, which, other than those we have discovered, may also exist in the environment. Therefore, to better elucidate MeO-PBDPB contamination, research is urgently needed to investigate the sources and transformation pathways of MeO-PBDPBs, the possible environmental presence of additional MeO-PBDPBs and potential precursors, the environmental behaviors (e.g., biomagniﬁcation potential) in Great Lakes food webs and potential adverse eﬀects on wildlife.

’ ASSOCIATED CONTENT

bS

Supporting Information. Table S1 and Figures S1 S6. This material is available free of charge via the Internet at http:// pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: 613-998-6696 (R.J.L.), 613-998-0474 (D.C.). Fax: 613998-0458 (R.J.L.). E-mail: [email protected] (R.J.L.); da. [email protected] (D.C.).

’ ACKNOWLEDGMENT We thank Chip Weseloh, Craig Hebert, Kim Williams, Doug Crump, Guy Savard, and other ﬁeld personnel involved in the 9529

dx.doi.org/10.1021/es201325g |Environ. Sci. Technol. 2011, 45, 9523–9530

Environmental Science & Technology herring gull egg collection and archiving. Special thanks go to Brock Chittim and Wellington Laboratories. Funding support was provided by Environment Canada’s Chemicals Management Plan (to R.J.L.). Postdoctoral support for D.C. is in part from the Natural Science and Engineering Research Council (NSERC) of Canada (to R.J.L.).

’ REFERENCES (1) Covaci, A.; Harrad, S.; Abdallah, M. A.; Ali, N.; Law, R. J.; Herzke, D.; de Wit, C. A. Novel brominated ﬂame retardants: A review of their analysis, environmental fate and behaviour. Environ. Int. 2011, 37, 532–556. (2) Hites, R. A. Polybrominated diphenyl ethers in the environment and in people: A meta-analysis of concentrations. Environ. Sci. Technol. 2004, 38, 945–956. (3) de Wit, C. A. An overview of brominated ﬂame retardants in the environment. Chemosphere 2002, 46, 583–624. (4) Letcher, R. J.; Bustnes, J. O.; Dietz, R.; Jenssen, B. M.; Jørgensen, E. H.; Sonne, C.; Verreault, J.; Vijayan, M. M.; Gabrielsen, G. W. Exposure and eﬀects assessment of persistent organohalogen contaminants in arctic wildlife and ﬁsh. Sci. Total Environ. 2010, 408, 2995–3043. (5) 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. (6) 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. (7) Gauthier, L. T.; Hebert, C. E.; Weseloh, D. V. C.; Letcher, R. J. Dramatic changes in the temporal trends of polybrominated diphenyl ethers (PBDEs) in herring gull eggs from the Laurentian Great Lakes: 19822006. Environ. Sci. Technol. 2008, 42, 1524–1530. (8) Tomy, G. T.; Pleskach, K.; Arsenault, G.; Potter, D.; McCrindle, R.; Marvin, C. H.; Sverko, E.; Tittlemier, S. Identiﬁcation of the novel cycloaliphatic brominated ﬂame retardant 1,2-dibromo-4-(1,2-dibromoethyl)cyclohexane in Canadian arctic beluga (Delphinapterus leucas). Environ. Sci. Technol. 2008, 42, 543–549. (9) S€oderstr€om, G.; Sellstr€om, U.; de Wit, C. A.; Tyslkind, M. Photolytic debromination of decabromodiphenyl ether (BDE-209). Environ. Sci. Technol. 2004, 38, 127–132. (10) Stapleton, H. M.; Brazil, B.; Holbrook, R. D.; Mitchelmore, C. L.; Benedict, R.; Konstantinov, A.; Potter, D. In vivo and in vitro debromination of decabromodiphenyl ether (BDE-209) by juvenile rainbow trout and common carp. Environ. Sci. Technol. 2006, 40, 4653–4658. (11) Van den Steen, E.; Covaci, A.; Jaspers, V. L. B.; Dauwe, T.; Voorspoels, S.; Eens, M.; Pinxten, R. Accumulation, tissue-speciﬁc distribution and debromination of decabromodiphenyl ether (BDE 209) in European starlings (Sturnus vulgaris). Environ. Pollut. 2006, 148, 1–6. (12) Hakk, H.; Letcher, R. J. Metabolism in the toxicokinetics and fate of brominated ﬂame retardantsA review. Environ. Int. 2003, 29, 801–826. (13) Gribble, G. W. Naturally occurring organohalogen compounds. Acc. Chem. Res. 1998, 31, 141–152. (14) Hebert, C. E.; Norstrom, R. J.; Weseloh, D. V. C. A quarter century of environmental surveillance: The Canadian Wildlife Service’s Great Lakes Herring Gull Monitoring Program. Environ. Rev. 1999, 7, 147–166. (15) Letcher, R. J.; Chu, S. High-sensitivity method for determination of tetrabromobisphenol-S and tetrabromobisphenol-A derivative ﬂame retardants in Great Lakes herring gull eggs by liquid chromatographyatmospheric pressure photoionizationtandem mass spectrometry. Environ. Sci. Technol. 2010, 44, 8615–8621. (16) Hebert, C. E.; Norstrom, R. J.; Weseloh, D. V. C. A quarter century of environmental surveillance: The Canadian Wildlife Service’s Great Lakes Herring Gull Monitoring Program. Environ. Rev. 1999, 147–166.

ARTICLE

(17) Hites, R. A. Electron impact and electron capture negative ionization mass spectra of polybrominated diphenyl ethers and methoxylated polybrominated diphenyl ethers. Environ. Sci. Technol. 2008, 42, 2243–2252. (18) Cooper, R. H.; Kozloski, R. P.; Gelbin, A.; Taroua, M. Comparison and interpretation of mass spectral data for polybrominated diphenyl ether (PBDEs) congeners and polyhalogenated biphenyl congeners. Organohalogen. Compd. 2003, 61, 223–226. (19) Eljarrat, E.; Lacorte, S.; Barcelo, D. Optimization of congenerspeciﬁc analysis of 40 polybrominated diphenyl ethers by gas chromatography/mass spectrometry. J. Mass Spec. 2002, 37, 76–84. (20) Timmons, L.; Cannon, M.; Grese, D.; Brown, R.; Haile, C.; Murrill, E. Identiﬁcation of chlorinated phenyl and phenoxy substituted dibenzodioxin, dibenzofuran, and diphenyl ether homologs in commercial grade pentachlorophenol. Anal. Lett. 1984, 17, 277–296. (21) Malmvaern, A.; Zebuehr, Y.; Jensen, S.; Kautsky, L.; Greyerz, E.; Nakano, T.; Asplund, L. Identiﬁcation of polybrominated dibenzo-pdioxin in blue mussels (Mytilus edulis) from the Baltic Sea. Environ. Sci. Technol. 2005, 39, 8235–8242. (22) Hebert, C. E. Winter severity aﬀects migration and contaminant accumulation in northern Great Lakes herring gulls. Ecol. Appl. 1998, 8, 669–679. (23) AlbemarleÒ Corporation. SAYTEXÒ 120 ﬂame retardant. 1999. http://www.albemarle.com/TDS/Flame_retardants/bc1003f. pdf (accessed September 2010). (24) Haglund, P. S.; Zook, D. R.; Buser, H.; Hu, J. Identiﬁcation and quantiﬁcation of polybrominated diphenyl ethers and methoxy-polybrominated diphenyl ethers in Baltic Sea. Environ. Sci. Technol. 1997, 31, 3281–3287. (25) Munschy, C.; Heas-Moisan, K.; Tixier, C.; Pacepavicius, G.; Alaee, M. Dietary exposure of juvenile common sole (Solea solea L.) to polybrominated diphenyl ethers (PBDEs): Part 2. Formation, bioaccumulation and elimination of hydroxylated metabolites. Environ. Pollut. 2010, 158, 3527–3533. (26) Asadorian, A. A.; Langer, H. G.; Funsher, J. A. United States Patent US3760003. 1973. http://www.freepatentsonline.com/3760003. pdf (accessed September 2010).

9530

dx.doi.org/10.1021/es201325g |Environ. Sci. Technol. 2011, 45, 9523–9530

ARTICLE pubs.acs.org/est

Aerobic and Anaerobic Nonmicrobial Methane Emissions from Plant Material Zhi-Ping Wang,†,* Zong-Qiang Xie,† Bao-Cai Zhang,‡ Long-Yu Hou,† Yi-Hua Zhou,‡ Ling-Hao Li,† and Xing-Guo Han†,§ †

State Key Laboratory of Vegetation and Environmental Change, Institute of Botany, Chinese Academy of Sciences, Nanxincun 20, Xiangshan, Beijing 100093, China ‡ State Key Laboratory of Plant Genomics and National Center for Plant Gene Research, Institute of Genetics and Developmental Biology, Chinese Academy of Sciences, Beijing 100101, China § Institute of Applied Ecology, Chinese Academy of Sciences, Shenyang 110016, China ABSTRACT: Methane (CH4) may be generated via microbial and nonmicrobial mechanisms. Nonmicrobial CH4 is also ubiquitous in nature, such as in biomass burning, the Earth's crust, plants, and animals. Relative to microbial CH4, nonmicrobial CH4 is less understood. Using fresh (living) and dried (dead) leaves and commercial structural compounds (dead) of plants, a series of laboratory experiments have been conducted to investigate CH4 emissions under aerobic and anaerobic conditions. CH4 emissions from fresh leaves incubated at ambient temperatures were nonmicrobial and enhanced by anaerobic conditions. CH4 emissions from dried leaves incubated at rising temperature ruled out a microbialmediated formation pathway and were plant-species-dependent with three categories of response to oxygen levels: enhanced by aerobic conditions, similar under aerobic and anaerobic conditions, and enhanced by anaerobic conditions. CH4 emissions in plant structural compounds may help to fully understand nonmicrobial CH4 formation in plant leaves. Experiments of reactive oxygen species (ROS) generator and scavengers indicate that ROS had a signiﬁcant role in nonmicrobial CH4 formation in plant material under aerobic and anaerobic conditions. However, the detailed mechanisms of the ROS were uncertain.

1. INTRODUCTION Methane (CH4) is an important trace gas, contributing to global warming and atmospheric redox chemistry. The change in atmospheric CH4 concentrations from 715 nL 3 L1 in 1750 to 1774 nL 3 L1 in 2005 ﬁgures out an average radiative forcing of 0.48 W 3 m2, ranking CH4 as the second most important anthropogenic greenhouse gas after CO2.1 CH4 has been traditionally considered an end product of organic matter degradation by microbes. The microbes are a limited group of obligate prokaryotes called methanogens that thrive under anaerobic conditions.2,3 Microbial CH4 has been widely studied in the past decades and understood profoundly.2 Nonmicrobial CH4 is also widespread in nature, such as in biomass burning,4,5 the Earth's crust,6 plants,7 and animals.8,9 However, it has been less understood. Nonmicrobial CH4 emissions by plants and its global strength still remain controversial.10 Previous studies described plant CH4 emissions as aerobic since plant tissues/compounds were incubated under aerobic conditions.7,1113 However, recent studies indicated that nonmicrobial CH4 was also generated in plant leaves when they were incubated under anaerobic conditions.10,14 Earlier studies demonstrated hypoxia-induced generation of nonmicrobial CH4 in mitochondria and eukaryotic cells of animals.8,9 Thus, we would propose that it might be better to use aerobic and r 2011 American Chemical Society

anaerobic nonmicrobial CH4 that are deﬁned as those from organisms, including plants and animals, when incubated under aerobic and anaerobic conditions, respectively. In this study, we postulated that nonmicrobial CH4 formation in plant material may occur under both aerobic and anaerobic conditions. To test this hypothesis, we concentrated on a comparison of nonmicrobial CH4 emissions from fresh and dried leaves of plants between aerobic and anaerobic conditions. Several structural compounds of plants such as pectin, lignin, and cellulose were examined to aid a better understanding of the eﬀect of oxygen levels on the emissions from plant leaves. Furthermore, experiments of reactive oxygen species (ROS) generators and scavengers were conducted to investigate the potential role of ROS in nonmicrobial CH4 formation in plant material.

2. MATERIALS AND METHODS 2.1. Plant Species Collection. A total of nine plant species were collected from the Xilin River basin in the Inner Mongolia12 Received: June 14, 2011 Accepted: September 30, 2011 Revised: September 27, 2011 Published: September 30, 2011 9531

dx.doi.org/10.1021/es2020132 | Environ. Sci. Technol. 2011, 45, 9531–9537

Environmental Science & Technology and the Beijing Botanical Garden (39°590 2400 N, 116°120 3600 E; 66 m above sea level) throughout the year 2010. The plants were of distinctive morphotypes, including five wood, one shrub, and three herb species. For the purpose of clarity, plant species abbreviations are used in ﬁgures: AF (Artemisia frigida), LG (Larix gmelinii), LS (Lactuca sativa), MS (Medicago sativa), PB (Populus beijingensis), PT (Pinus tabulaeformis), QW (Quercus wutaishanica), RP (Robinia pseudoacacia), and SO (Spinacia oleracea). 2.2. Sample Preparation. Fresh leaves were detached to limit transpiration as a potential microbial CH4 source. The leaves were collected in plastic bags and transported to the laboratory within 15 min. They were immediately washed in deionized water and air-dried for about 0.5 h prior to commencement of measurements. Dried leaves were obtained by oven-drying fresh leaves at 40 °C to constant mass. Plant structural compounds were commercial citrus pectin (CAS no. 9000-69-5), lignin (CAS no. 8068-05-1), and cellulose (CAS no. 9004-34-6). They were obtained from SigmaAldrich Chemical Co., Shanghai, China. 2.3. Chemical Addition. Fenton reagent was used to generate • OH,15 whereas 1,4-diazabicyclo[2.2.2]octane (DABCO, C6H12N2), potassium iodide (KI), and D-mannitol [C6H8(OH)6] were reported to be scavengers of 1O2, H2O2, and •OH, respectively.16,17 In ROS generator experiments, fresh or dried leaves were impregnated consecutively by 2 mL of 20 mM Na2EDTA 3 2H2O (disodium ethylenediaminetetraacetate dihydrate) and 2 mL of 20 mM FeSO4 3 7H2O, sealed in gastight serum bottles, and flushed with CH4-free compressed oxygen or nitrogen. Immediately afterward, 2 mL of deionized water (equivalent to 0% H2O2), 2 mL of 1% H2O2, or 2 mL of 2% H2O2 were added via syringe into the samples. Plant structural compounds were also treated as above but with 1 mL of chemical solutions. In ROS scavenger experiments, dried leaves were soaked in 0, 5, or 50 mM DABCO, KI, or D-mannitol solutions for about 5 h, removed, and then air-dried for a few days. The air-dried leaves containing the chemical were used as samples. 2.4. Laboratory Incubation. CH4 emissions were examined from fresh and dried leaves and structural compounds of plants in closed-bottle laboratory incubations in the dark. For each plant sample, a few grams of prepared plant material was placed in a 120-mL serum bottle. Parallel blanks were employed to determine whether background CH4 concentrations in serum bottles changed in the absence of plant material. If blanks had undetectable change in CH4 concentrations, they were usually omitted in figures for the purpose of clarity. A flushing method was used to establish aerobic and anaerobic conditions.14 In brief, the bottles were immediately sealed with butyl rubber stoppers (diameter 20 mm) and flushed for 15 min with CH4-free compressed oxygen (O2), nitrogen (N2), hydrogen (H2), or helium (He) by use of “inletoutlet” needles inserted through the stoppers at a rate of 400 mL 3 min1, respectively. To avoid plant structural compound being flushed out of the bottle, a piece of glass microfiber filter (Whatman GF/A, diameter 12.5 cm) was used to separate a slow flushing (200 mL 3 min1) from the sample. Before usage, glass microfiber filters were baked for a few hours in an oven at 200 °C to remove possible organic contaminants. Initial CH4 concentrations were measured immediately prior to incubations. Fresh leaves were incubated at ambient temperatures. At the end of each experiment, their dry matter was determined by oven-drying at 40 °C to constant mass. To increase the signal-tonoise ratio, dried leaves and structural compounds of plants were incubated at rising temperature, unless stated otherwise.

ARTICLE

Figure 1. CH4 emissions from (a) fresh leaves at ambient temperatures of 2324 °C for 3 h and (b) dried leaves at rising temperature of 70 °C for 1 h under oxygen, hydrogen, nitrogen, and helium conditions.

2.5. Extraction of Structural Compounds in Plant Cell Wall. Structural compounds were extracted from dried leaves of selected plants. The dried leaves were extracted with 70% ethanol and chloroform/methanol (1:1 v/v) to prepare alcohol-insoluble residues (AIRs) of cell walls.18 For the measurement of pectin content, AIRs were extracted by 1% ammonium oxalate (w/v), and then soluble pectin was precipitated and trapped on a crucible to weigh. To determine the contents of uronic acids, AIRs were destarched and methanolyzed in 1 M methanolic hydrochloric acid. The trimethylsilyl derivatives were generated with Trisil reagent and finally analyzed by Agilent GC 7900/ 5975C MS with a DB-1 column.19 The content of lignin was determined as described previously.19 In brief, AIRs were treated with 72% 1 N sulfuric acid to remove polysaccharides, and then the insoluble lignin was trapped and weighed after being thoroughly dried. To determine the content of crystalline cellulose, AIRs were destarched with amylase and then hydrolyzed in Updegraff reagent (8:1:2 v/v/v in acetic acid/nitric acid/ water) at 100 °C for 30 min. After centrifuge collection and washing, the cellulose was hydrolyzed and used for anthrone assay.18 2.6. CH4 Concentration Measurement. CH4 concentrations in the headspace of serum bottles were analyzed at various time intervals by use of a Hewlett-Packard 5890 series II gas chromatograph. The GC running conditions were described previously.14 A 5-mL gas sample was withdrawn from a 120-mL serum bottle by syringe and immediately replaced by 5 mL of CH4-free compressed oxygen or nitrogen to maintain headspace pressure. 2.7. Statistical Analysis. Emission rate was calculated by CH4 accumulation over time and recorded as nanograms per gram dry weight per hour. Value is mean ( 1 standard deviation (n = 3 in Figures 1 and 46 and n = 4 in Figures 2 and 3). Statistical analysis was performed by use of a Statistical Analysis System program.20 Duncan’s multiple range test was employed to compare 9532

dx.doi.org/10.1021/es2020132 |Environ. Sci. Technol. 2011, 45, 9531–9537

Environmental Science & Technology

Figure 2. CH4 emissions from fresh leaves of plants at ambient temperatures of 2223 °C under (a) a cycle of 01 h aerobic (100% O2), 23 h anaerobic (0% O2), and 45 h aerobic conditions and (b) a cycle of 01 h anaerobic, 23 h aerobic, and 45 h anaerobic conditions. Between incubation periods, the samples were appropriately ﬂushed again.

the variation in CH4 emission rates among treatments at P < 0.05. One-way analysis of variance was used to evaluate statistical difference in CH4 emission rates between aerobic and anaerobic conditions. The different letters indicated significant differences (P < 0.05) in each group of treatments. If statistically significant differences were easily self-explanatory, the different letters were omitted for the purpose of clarity.

3. RESULTS AND DISCUSSION 3.1. Aerobic and Anaerobic Incubation Conditions. CH4 emissions from fresh and dried leaves of plants had significant differences (P < 0.05) between the treatments of O2 and the other gases, with the exception of those from dried leaves of A. frigida. For all plant species investigated, however, the emissions had no significant differences (P > 0.05) in the treatments of H2, N2, and He (Figure 1). H2 is an available growth substrate for a large diversity of anaerobic bacteria, notably obligate methanogens; CH4 generation by anaerobic bacteria using H2 as substrate contributes approximately 1050% to total CH4.2 In this study if H2 would provide a substrate for CH4 generation, the emission in H2 treatment should be higher than those in the treatments of N2 and He. Accordingly, H2 did not serve as a substrate for CH4 generation during incubation periods. On the other hand, temporal kinetic experiments showed that no significant microbial CH4 was generated over a few hours of incubation, since methanogens need adequate time to multiply.10 These indicate that CH4 emitted from the leaves was indeed nonmicrobial. O2 provided

ARTICLE

aerobic conditions while H2, N2, and He provided anaerobic conditions during these short-term incubations. 3.2. Nonmicrobial CH4 Emissions from Fresh Leaves. When fresh leaves of A. frigida, M. sativa, and Q. wutaishanica were alternately incubated under aerobic and anaerobic conditions, their CH4 emissions were significantly higher (P < 0.05) under anaerobic than aerobic conditions. However, CH4 emissions from fresh leaves of L. gmelinii, P. beijingensis, and P. tabulaeformis had no statistically significant difference from zero. Despite this, the emissions were enhanced by anaerobic conditions (Figure 2). Accordingly, it is concluded from results obtained from these experiments and previous studies in plants10,14 and animals8,9 that instant CH4 formation in living organisms is a nonmicrobial process enhanced by anaerobic conditions. Keppler et al.7 suggested a nonenzymatic process for CH4 formation in plants. Nisbet et al.21 did not find necessary biochemical pathways to synthesize CH4 in plants. Thus, enhanced CH4 formation under anaerobic conditions might be due to physiological activities of living organisms, such as a passive consequence of physiological processes. It is well-known that microbial CH4 is generated under anaerobic conditions.2 As a result, in nature both microbial and nonmicrobial CH4 should be simultaneously generated under anaerobic environment. In a number of ﬁeld studies where signiﬁcant CH4 emissions were observed, it has been reported that these were transmitted to the atmosphere by plants.2224 Such ﬁeld emissions might now also include a contribution from nonmicrobial source. However, with current knowledge, it is diﬃcult to distinguish between microbial and nonmicrobial CH4 generated in nature. 3.3. Nonmicrobial CH4 Emissions from Dried Leaves. When dried leaves of R. pseudoacacia and Q. wutaishanica were alternately incubated at ambient and rising temperatures, their CH4 emissions were repeatedly provoked by rising temperature (70 °C) but were undetectable at ambient temperatures (Figure 3a,b). Microbial CH4 emission was usually observed as a parabolic curve with respect to temperature; the emission peak corresponded to the most appropriate temperature of 2530 °C required by enzymatic metabolism of microbes.25 Accordingly, the emissions at rising temperature excluded microbial activity as the source. Nonmicrobial CH4 emissions from the dried leaves incubated at rising temperature had three categories of response to oxygen levels. Specifically, the emissions were enhanced by aerobic conditions in M. sativa, P. beijingensis, R. pseudoacacia, S. oleracea, and L. sativa (left of the left dashed line); similar under aerobic and anaerobic conditions in A. frigida and L. gmelinii (between the two dashed lines), and enhanced by anaerobic conditions in P. tabulaeformis and Q. wutaishanica (right of the right dashed line) (Figure 3c,d). Three categories of response were also reflected in nonmicrobial CH4 emissions between the treatments of O2 and the other gases (Figure 1b). When pectin and lignin were incubated at rising temperatures, their CH4 emissions were enhanced by aerobic conditions (Figure 4). This may be used to interpret that the emissions from dried leaves were enhanced by aerobic conditions. Commercial pectin or lignin incubated in serum bottles did not have an opportunity to react with other compounds like those in dried leaves. This might be one reason why the emissions from these pure structural compounds were much lower than those from dried leaves when based on structural compound equivalents (Figure 3c,d; Table 1). Previous studies used ultraviolet radiation as a trigger to drive nonmicrobial CH4 formation in terrestrial 9533

dx.doi.org/10.1021/es2020132 |Environ. Sci. Technol. 2011, 45, 9531–9537

Environmental Science & Technology

ARTICLE

Figure 3. CH 4 emissions from dried leaves of plants under (a, b) a cycle of ambient and rising temperatures and (c, d) a cycle of aerobic and anaerobic conditions. Dried leaves of (a) R. pseudoacacia and (b) Q. wutaishanica were alternately incubated at ambient temperature of 21 °C in the periods of 13 and 46 h and rising temperature of 70 °C in the periods of 01, 34, and 67 h. Dried leaves of plants were alternately incubated at rising temperature of 70 °C under (c) a cycle of 01 h aerobic, 23 h anaerobic, and 45 h aerobic conditions and (d) a cycle of 01 h anaerobic, 23 h aerobic, and 45 h anaerobic conditions. Between incubation periods, the samples were appropriately ﬂushed again.

Figure 4. CH4 emissions from plant structural compounds at rising temperatures of (a) 70 °C and (b) 90 °C for 3 h under aerobic and anaerobic conditions.

plant tissues and compounds.2628 Rising temperature, such as heat wave in summer and biomass burning, has wide implications for terrestrial plants on the Earth's surface. CH4 emission was lower in pectin than lignin when rising temperature was a driver (Figure 4). This is inconsistent with results by Vigano et al.,27 where there was higher CH4 emission in pectin than lignin under ultraviolet irradiation. Accordingly, diﬀerent structural compounds prefer to accept their distinctive drivers in nonmicrobial CH4 formation. In addition, almost no CH4 emission was observed in cellulose when incubated at rising temperatures (Figure 4). This may suggest that nonmicrobial CH4 formation was not derived from the cellulose of dried leaves at rising temperature.

Figure 5. Eﬀects of ROS generator, Fenton’s reagent, on CH4 emissions in (a) fresh leaves of P. tabulaeformis, (b) dried leaves of P. tabulaeformis and R. pseudoacacia, and (c) plant structural compounds. The samples were incubated at ambient temperatures. Treatments had aerobic and anaerobic conditions; 1% and 2% H2O2; PT and RP species; P (pectin) and L (lignin). Undetectable CH4 emissions in plant material inﬁltrated with deionized water as blanks (0% H2O2) and cellulose treatments were omitted for the purpose of clarity. 9534

dx.doi.org/10.1021/es2020132 |Environ. Sci. Technol. 2011, 45, 9531–9537

Environmental Science & Technology

ARTICLE

Cell wall compounds in dried leaves of three types of plants are listed in Table 1. AIRs were obtained via washing dried leaves with ethanol, in which polysaccharides are insoluble. Pectic polysaccharides mainly consist of sugar residues, methyl esters, and O-acetyl groups,28 while in sugar residues galactouronic acid (GalUA) and glucuronic acid (GlcUA) are major components.29 On average, the ﬁrst type of plants with enhanced CH4 emissions under aerobic conditions had a higher proportion of AIRs to biomass when compared with the other two types. Again, the ﬁrst type had higher pectin and lower lignin and cellulose contents in

Figure 6. Eﬀect of ROS scavengers on CH4 emissions in dried leaves of (a) P. tabulaeformis and R. pseudoacacia and (b) P. tabulaeformis. The leaves were incubated at rising temperature of 70 °C for 1 h.

AIRs and higher GalUA and GlcUA contents in destarched AIRs. The values in pectin are incomparable with GalUA and GlcUA contents, since destarched AIRs are diﬀerent from and much lower than AIRs. For dried leaves in each type of plant, however, contents of their structural compounds had large variabilities, for example with coeﬃcients of variation (CV) of 128.1%, 65.5%, and 109.9% in the ﬁrst, second, and third types of plants, respectively. Together with less plants examined, thus, it is uncertain that the response of nonmicrobial CH4 emissions to oxygen levels is classiﬁed by contents of the structural compounds. To profoundly understand the response, more plant species and structural compounds need to be investigated in the future. Previous studies found that methoxyl groups of plant pectin and/or lignin serve as precursors for nonmicrobial CH4 formation.2628 If precursors would restrictedly come from pectin and/or lignin, the result of enhanced nonmicrobial CH4 emissions by aerobic conditions in the structural compounds (Figure 4) cannot explain the emissions from dried leaves of the other two types of plants (Figure 3c,d). This indicates that more precursors should be responsible for nonmicrobial CH4 formation in dried leaves. The precursors might include oxidative and reductive categories that coexist in the dried leaves. Response of nonmicrobial CH4 formation to oxygen levels presumably depends upon a mixture of various categories of precursors. 3.4. Role of ROS in Nonmicrobial CH4 Formation in Plant Material. ROS are exceedingly reactive and short-lived.15,28 Previous studies did not directly monitor ROS.28,30 Because of difficulty in monitoring ROS, we used ROS generator and scavengers as done previously 28,30 to examine potential role of ROS in mechanisms of nonmicrobial CH4 formation in plant material. More CH4 was irritatingly emitted by ROS generator, H2O2, in all categories of plant material under anaerobic than aerobic

Table 1. Cell Wall Compounds in Dried Leaves of Plantsa pectin/AIRs

GalUA/D-AIRs

GlcUA/D-AIRs

lignin/AIRs

cellulose/AIRs

AIRs/biomass (%)

(μg 3 mg1)

(μg 3 mg1)

(μg 3 mg1)

(μg 3 mg1)

(μg 3 mg1)

M. sativa

72.6

18.9

78.5

9.4

122.5

73.2

P. beijingensis

80.1

57.9

78.8

3.6

330.6

72.4

R. pseudoacacia S. oleracea

66.1 72.9

7.4 5.0

63.5 48.0

6.3 ndb

446.4 58.6

67.1 49.4

L. sativa

68.0

152.2

79.2

nd

139.5

107.7

mean

71.9

48.3

69.6

6.4

219.5

73.9

SD

5.4

61.8

13.8

2.9

162.4

21.2

CV (%)

7.5

128.1

19.8

45.6

74.0

28.6

A. frigida

63.1

17.6

95.4

11.4

309.7

164.6

L. gmelinii mean

64.3 63.7

6.4 12.0

45.7 70.6

3.4 7.4

566.0 437.8

71.8 118.2

species

SD

0.9

7.9

35.2

5.6

181.2

65.6

CV (%)

1.4

65.5

49.8

76.0

41.4

55.5

P. tabulaeformis

62.2

0.6

25.3

1.5

450.3

228.9

Q. wutaishanica

67.1

5.1

82.4

2.9

312.1

125.6

mean

64.7

2.9

53.9

2.2

381.2

177.2

3.4 5.3

3.2 109.9

40.4 75.1

1.0 44.7

97.8 25.6

73.0 41.2

SD CV (%)

a Three groups of plant species showed distinctive responses in nonmicrobial CH4 emissions to aerobic and anaerobic conditions (see Figure 3c,d). AIRs are alcohol-insoluble residues of cell wall, while D-AIRs are destarched alcohol insoluble residues. b Content is under detection threshold.

9535

dx.doi.org/10.1021/es2020132 |Environ. Sci. Technol. 2011, 45, 9531–9537

Environmental Science & Technology conditions (Figure 5). The emissions showed logarithmic kinetics with respect to time from fresh leaves of P. tabulaeformis and linear kinetics from dried leaves of P. tabulaeformis and R. pseudoacacia. The linear kinetics may be interpreted as due to a large precursor reservoir that continuously served CH4 formation. The emissions also had logarithmic kinetics in pectin but were almost undetectable in lignin. CH4 emissions were signiﬁcantly (P < 0.05) constrained by the ROS scavengers immerged in dried leaves when incubated at rising temperature (Figure 6). KI showed increasing inhibitory eﬀect on the emissions from dried leaves of P. tabulaeformis and R. pseudoacacia. DABCO and D-mannitol, even at their low concentration treatments, had signiﬁcant inhibition on the emissions from dried leaves of P. tabulaeformis. The diﬀerences in inhibitory magnitude may be due to KI, DABCO, and Dmannitol acting as distinctive scavengers of H2O2, 1O2, and •OH, respectively.16,17 Similar/identical change trends in nonmicrobial CH4 emissions were observed between aerobic and anaerobic conditions (Figures 5 and 6). This indicates that ROS were involved in nonmicrobial CH4 formation in plant material while other factors could be responsible for the diﬀerences in emission rates between aerobic and anaerobic conditions. However, it is necessary to mention that the detailed mechanisms of the ROS were unclear. Less nonmicrobial CH4 emissions were observed in pectin than lignin at rising temperatures (Figure 4), whereas more emissions were stimulated by ROS generator in pectin than lignin at ambient temperatures (Figure 5c). Previous studies suggested that nonmicrobial CH4 is generated via pyrogenic and thermogenic reactions in biomass burning4,5 and within the Earth's crust.6 Thus, free radicals and pyrogenic and thermogenic reactions might together be responsible for nonmicrobial CH4 formation in dried leaves when incubated at rising temperature. Plant tissues naturally generate certain ROS during growth via the Fenton or HaberWeiss reactions (see ref 28). On the other hand, in nature plants are frequently subjected to various forms of environmental stress such as extreme weather, solar UV radiation, soilwater deﬁcit and ﬂooding, hypoxia and hyperoxia, wounding, herbicides, and pathogens.10 These environmental stress factors stimulate ROS generation in plant cells.31 Thus, the ROS’ role in nonmicrobial CH4 formation simulated in laboratory conditions may be extended to natural situations. Previous studies indicated that microbial CH4 formation in soils may occur under aerobic conditions.32,33 This is inconsistent with the traditionally held view that microbial CH4 is generated under anaerobic conditions. This study conﬁrms that nonmicrobial CH4 formation in plant material occurred under both aerobic and anaerobic conditions. Accordingly, it is clearly shown that CH4 formation, regardless of via microbial or nonmicrobial mechanisms, does not completely depend upon oxygen levels. Conrad2 suggested that aerobic microbial CH4 may be generated via the coincidence of electron donors and electron acceptors. When electron donor availability coincides with electron acceptors in a medium involved by microorganisms, sequential reduction occurs largely to produce microbial CH4 under aerobic conditions. The electron coincidence might provide a basic point for understanding microbial and nonmicrobial CH4 formation under aerobic and anaerobic conditions. Whether nonmicrobial CH4 formation is also realized via the coincidence of electron donors and electron acceptors in a plant medium needs further work to test.

ARTICLE

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected] or [email protected]; phone: 0086-01-6283 6635; fax: 0086-01-6859 7569.

’ ACKNOWLEDGMENT We greatly thank two anonymous referees and editors for their constructive comments that improved the paper. We are also very grateful to Frank Keppler and John T. G. Hamilton for their helpful comments. This research was supported by the general program of the National Natural Science Foundation of China (30970518), the Key Project of National Natural Science Foundation of China (30830026), and funding from the State Key Laboratory of Vegetation and Environmental Change (2011zyts07). ’ REFERENCES (1) Forster, P.; et al. 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; pp 129234. (2) Conrad, R. Soil microorganisms as controllers of atmospheric trace gases (H2, CO, CH4, OCS, N2O, and NO). Microbiol. Rev. 1996, 60, 609–640. (3) Conrad, R. Quantiﬁcation of methanogenic pathways using stable carbon isotopic signatures: a review and a proposal. Org. Geochem. 2005, 36, 739–752. (4) Crutzen, P. J.; Andreae, M. O. Biomass burning in the tropics: impact on atmospheric chemistry and biogeochemical cycles. Science 1990, 250, 1669–1678. (5) Andreae, M. O.; Merlet, P. Emission of trace gases and aerosols from biomass burning. Global Biogeochem. Cycles 2001, 15, 955–966. (6) Etiope, G.; Klusman, R. W. Geologic emissions of methane to the atmosphere. Chemosphere 2002, 49, 777–789. (7) Keppler, F.; Hamilton, J. T. G.; Brass, M.; R€ockmann, T. Methane emissions from terrestrial plants under aerobic conditions. Nature 2006, 439, 187–191. (8) Ghyczy, M.; Torday, C.; Boros, M. Simultaneous generation of methane, carbon dioxide, and carbon monoxide from choline and ascorbic acid: a defensive mechanism against reductive stress? FASEB J. 2003, 17, 1124–1126. (9) Ghyczy, M.; Torday, C.; Kaszaki, J.; Szabo, A.; Cz obel, M.; Boros, M. Hypoxia-induced generation of methane in mitochondria and eukaryotic cells - An alternative approach to methanogenesis. Cell. Physiol. Biochem. 2008, 21, 251–258. (10) Wang, Z. P.; Keppler, F.; Greule, M.; Hamilton, J. T. Nonmicrobial methane emissions from fresh leaves: eﬀects of physical wounding and anoxia. Atmos. Environ. 2011, 45, 4915–4921. (11) Ferretti, D. F.; Miller, J. B.; White, J. W. C.; Lassey, K. R.; Lowe, D. C.; Etheridge, D. M. Stable isotopes provide revised global limits of aerobic methane emissions from plants. Atmos. Chem. Phys. 2007, 7, 237–241. (12) Wang, Z. P.; Han, X. G.; Wang, G. G.; Song, Y.; Gulledge, J. Aerobic methane emission from plants in the Inner Mongolia steppe. Environ. Sci. Technol. 2008, 42, 62–68. (13) Bruhn, D.; Mikkelsen, T. N.; Øbro, J.; Willats, W. G. T.; Ambus, P. Eﬀects of temperature, ultraviolet radiation and pectin methyl esterase on aerobic methane release from plant material. Plant Biol. 2009, 11 (Suppl. 1), 43–48. (14) Wang, Z. P.; Gulledge, J.; Zheng, J. Q.; Liu, W.; Li, L. H.; Han, X. G. Physical injury stimulates aerobic methane emissions from terrestrial plants. Biogeosciences 2009, 6, 615–621. 9536

dx.doi.org/10.1021/es2020132 |Environ. Sci. Technol. 2011, 45, 9531–9537

Environmental Science & Technology

ARTICLE

(15) Fry, S. C.; Dumville, J. C.; Miller, J. G. Fingerprinting of polysaccharides attacked by hydroxyl radicals in vitro and in the cell walls of ripening pear fruit. Biochem. J. 2001, 357, 729–737. (16) Buxton, G. V.; Greenstock, C. L.; Helman, W. P.; Ross, A. B. Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals (•OH/•O) in aqueous solution. J. Phys. Chem. Ref. Data 1988, 17, 513–886. (17) Wang, Y. B.; Feng, H. Y.; Qu, Y.; Cheng, J. Q.; Zhao, Z. G.; Zhang, M. X.; Wang, X. L.; An, L. Z. The relationship between reactive oxygen species and nitric oxide in ultraviolet-B-induced ethylene production in leaves of maize seedlings. Environ. Exp. Bot. 2006, 57, 51–61. (18) Foster, C. E.; Martin, T. M.; Pauly, M. Comprehensive compositional analysis of plant cell walls (lignocellulosic biomass) Part II: Carbohydrates. J. Vis. Exp. 2010, e1837. (19) Li, M.; Xiong, G.; Li, R.; Cui, J.; Tang, D.; Zhang, B.; Pauly, M.; Cheng, Z.; Zhou, Y. Rice cellulose synthase-like D4 is essential for normal cell-wall biosynthesis and plant growth. Plant J. 2009, 60, 1055–1069. (20) SAS/STAT User’s Guide, Release 8.0; SAS Institute: Cary, NC, 1999. (21) Nisbet, R. E. R.; Fisher, R.; Nimmo, R. H.; Bendall, D. S.; Crill, P. M.; Gallego-Sala, A. V.; Hornibrook, E. R. C.; Lopez-Juez, E.; Lowry, D.; Nisbet, P. B. R.; Shuckburgh, E. F.; Sriskantharajah, S.; Howe, C. J.; Nisbet, E. G. Emission of methane from plants. Proc. R. Soc. B: Biol. Sci. 2009, 276, 1347–1354. (22) Megonigal, J. P.; Guenther, A. B. Methane emissions from upland forest soils and vegetation. Tree Physiol. 2008, 28, 491–498. (23) Rice, A. L.; Butenhoﬀ, C. L.; Shearer, M. J.; Teama, D.; Rosenstiel, T. N.; Khalil, M. A. K. Emissions of anaerobically produced methane by trees. Geophys. Res. Lett. 2010, 37, No. L03807. (24) Mukhin, V. A.; Voronin, P. Yu. Methane emission from living tree wood. Russ. J. Plant Physiol. 2011, 58, 344–350. (25) Dunﬁeld, P.; Knowles, R.; Dumont, R.; Moore, T. R. Methane production and consumption in temperature and subarctic peat soils: Response to temperature and pH. Soil Biol. Biochem. 1993, 25, 321–326. (26) Keppler, F.; Hamilton, J. T. G.; McRoberts, W. C.; Vigano, I.; Braß, M.; R€ockmann, T. Methoxyl groups of plant pectin as a precursor of atmospheric methane: evidence from deuterium labelling studies. New Phytol. 2008, 178, 808–814. (27) Vigano, I.; van Weelden, H.; Holzinger, R.; Keppler, F.; McLeod, A.; R€ockmann, T. Eﬀect of UV radiation and temperature on the emission of methane from plant biomass and structural components. Biogeosciences 2008, 5, 937–947. (28) Messenger, D. J.; McLeod, A. R.; Fry, S. C. The role of ultraviolet radiation, photosensitizers, reactive oxygen species and ester groups in mechanisms of methane formation from pectin. Plant Cell Environ. 2009, 32, 1–9. (29) Mohnen, D. Pectin structure and biosynthesis. Curr. Opin. Plant Biol. 2008, 11, 266–277. (30) McLeod, A. R.; Fry, S. C.; Loake, G. J.; Messenger, D. J.; Reay, D. S.; Smith, K. A.; Yun, B. W. Ultraviolet radiation drives methane emissions from terrestrial plant pectins. New Phytol. 2008, 180, 124–132. (31) Apel, K.; Hirt, H. Reactive oxygen species: metabolism, oxidative stress, and signal transduction. Annu. Rev. Plant Biol. 2004, 55, 373–399. (32) DeGroot, C. J.; Vermoesen, A.; VanCleemput, O. Laboratory study of the emission of N2O and CH4 from a calcareous soil. Soil Sci. 1994, 158, 355–364. (33) Yavitt, J. B.; Fahey, T. J.; Simmons, J. A. Methane and carbon dioxide dynamics in a northern hardwood ecosystem. Soil Sci. Soc. Am. J. 1995, 59, 796–804.

9537

dx.doi.org/10.1021/es2020132 |Environ. Sci. Technol. 2011, 45, 9531–9537

ARTICLE pubs.acs.org/est

Long-Term Temporal Variability in Hydrogen Peroxide Concentrations in Wilmington, North Carolina USA Rainwater Katherine M. Mullaugh,*,† Robert J. Kieber,† Joan D. Willey,† and G. Brooks Avery, Jr.† †

Department of Chemistry and Biochemistry, University of North Carolina Wilmington, Wilmington, North Carolina 28403-5932, United States

bS Supporting Information ABSTRACT: Measurements of hydrogen peroxide (H2O2), pH, dissolved organic carbon (DOC), and inorganic anions (chloride, nitrate, and sulfate) in rainwater were conducted on an event basis at a single site in Wilmington, NC for the past decade in a study that included over 600 individual rain events. Annual volume weighted average (VWA) H2O2 concentrations were negatively correlated (p < 0.001) with annual VWA nonseasalt sulfate (NSS) concentrations in low pH (<5) rainwater. Under these conditions H2O2 is the primary aqueousphase oxidant of SO2 in the atmosphere. We attribute the increase of H2O2 to decreasing SO2 emissions which has had the eﬀect of reducing a major tropospheric sink for H2O2. Annual VWA H2O2 concentrations in low pH (<5) rains showed a signiﬁcant increase over the time scale of this study, which represents the only long-term continuous data set of H2O2 concentrations in wet deposition at a single location. This compositional change has important implications because H2O2 is a source of highly reactive free radicals so its increase reﬂects a higher overall oxidation capacity of atmospheric waters. Also, because rainwater is an important mechanism by which H2O2 is transported from the atmosphere to surface waters, greater wet deposition of H2O2 could inﬂuence the redox chemistry of receiving watersheds which typically have concentrations 23 orders of magnitude lower than rainwater.

’ INTRODUCTION Hydrogen peroxide (H2O2) is a chemically labile oxidant which plays a central role in a number of important redox processes occurring within the troposphere including the reactions of a number of highly reactive free radicals and trace metals. The main source of H2O2 in the atmosphere is self-reaction of hydroperoxy radicals (HO2•) which are photochemically produced in the gas phase.1,2 Studies by our group3 and others2,4 have shown aqueous phase production of H2O2 is also possible. Much of the gas phase H2O2 produced in the atmosphere is subsequently scavenged into the aqueous phase. Wet deposition is an important atmospheric removal mechanism with additional sinks including gas phase and aqueous phase reactions.5 Due to the high water solubility of H2O2,6 the aqueous phase chemistry of hydrogen peroxide in the atmosphere is of particular interest. As an important oxidant of S(IV), it plays a central role in H2SO4 generation in the atmosphere. Although other species like O3 and Fe(III) can act as oxidants of S(IV), previous work has shown the rate constant of S(IV) oxidation is highest when H2O2 is the oxidant, particularly under conditions of low pH (<5).7 Additionally, the photolysis of H2O2 is a major source of hydroxyl (•OH) and hydroperoxyl (HO2•) radicals in the atmosphere and hence partially controls the oxidizing capacity of the atmosphere. The highly reactive radicals generated from H2O2 r 2011 American Chemical Society

are responsible for important aqueous phase reactions including the production of organic acids and organic peroxides.5 Any change in the abundance of H2O2 in precipitation will also be important for sensitive terrestrial and aquatic environments. Concentrations of H2O2 in rainwater are typically 23 orders of magnitude higher than surface water concentrations so the contribution of H2O2 from rainwater to surface waters is important albeit episodic. Any change in the wet deposition of H2O2 could be particularly signiﬁcant for the open ocean where it has a relatively long half-life.8,9 Greater wet deposition of H2O2 could inﬂuence the redox chemistry of these systems, such as altering the speciation of trace metals like iron or promoting the oxidation of organic compounds. Rainwater has been collected on an event basis on the campus of the University of North Carolina Wilmington since 1985. This data set has been invaluable in evaluating changes in the chemistry of precipitation in this area, which has experienced large population growth and increased environmental regulations. Our group has previously reported signiﬁcant decreases in Received: July 29, 2011 Accepted: October 4, 2011 Revised: September 29, 2011 Published: October 26, 2011 9538

dx.doi.org/10.1021/es202634s | Environ. Sci. Technol. 2011, 45, 9538–9542

Environmental Science & Technology dissolved organic carbon (DOC) and hydrogen ion (H+) concentrations in precipitation, which have been attributed to better pollution control including enhanced emission regulations since 1995 and the introduction of reformulated gasoline in 1996.10 A signiﬁcant remaining tropospheric chemistry question is how long-term changes in the composition of the atmosphere aﬀects levels of labile oxidants such as H2O2. This point was underscored in a recent review2 where it was suggested that peroxide concentrations in the aqueous phase may be increasing due to desulfurization in North America and Western Europe, although the author pointed out this hypothesis was impossible to explore because of a lack of long-term measurements of H2O2 in precipitation. Here we present the ﬁrst decade-long study of H2O2 in rainwater which represents the only long-term continuous data set known to the authors at the time of publication, though researchers have previously reported long-term trends in H2O2 from ice core data.11 It is essential that the rain was sampled on an event basis and analyzed quickly because of the lability of hydrogen peroxide. This time period is important because it encompasses dramatic decreases in atmospheric SO2 inputs in North America and the introduction of various emission control technologies, both of which could play a central role in controlling H2O2 concentrations in precipitation.

ARTICLE

was standardized monthly by a titration by phenylarsine oxide. The method has an analytical precision of 2% RSD at ambient rainwater concentrations with a detection limit of 2 nM.12 Supporting Analysis (pH, DOC, and Anions). Rainwater pH was determined upon collection using a Ross electrode calibrated with low ionic strength 4.10 and 6.97 buffers. Ionic strength adjuster (pHix Orion Research Incorporated, Boston, Massachusetts) was added to each sample to match the ionic strength of samples to that of buffers. Measurements were (0.01 pH units in the pH 4 range and (0.03 for samples with pH 5 or above. Dissolved organic carbon (DOC) was determined by high temperature combustion (HTC) using a Shimadzu TOC 5000 total organic carbon analyzer equipped with an ASI 5000 autosampler. Standards were prepared from potassium hydrogen phthalate (KHP) in DIW. Each sample was injected 4 times with a relative standard deviation (RSD) e 3%. Anions (Cl, NO3, and SO42‑) were measured using suppressed ion chromatography using a Dionex IC system outﬁtted with an IonPac AS14A column and ASRS 300 4-mm suppressor. Standards were prepared from the sodium salts of each anion in DIW. Anions measurements had a RSD e 5%. Nonseasalt sulfate (NSS) was calculated using the following equation which assumes a constant seawater ratio of Cl to SO4 which works at our coastal location where almost all Cl comes from sea spray13

’ EXPERIMENTAL SECTION Sample Collection and Storage. The rain sampling site used

in this study was an open area of longleaf pine, wire grass, and turkey oak on the campus of the University of North Carolina at Wilmington (34°13.90 N, 77°52.70 W, 8.5 km from the Atlantic Ocean). All rainwater event samples were collected using an Aerochem Metrics (ACM) Model 301 Automatic Sensing Wet/Dry Precipitation Collector which housed a 4 L glass beaker placed within a HDPE plastic bucket. All glassware, including the rain collection beakers, was combusted at 450 °C in a muffle furnace for a minimum of 4 h to remove organics prior to use. Samples were collected on an event basis no more than 12 h after precipitation ceased, but this period of time was generally much shorter. Samples were analyzed immediately after collection for hydrogen peroxide and pH. Stability tests in our laboratory demonstrate that pH and hydrogen peroxide concentrations are stable over 2024 h time periods (Figure S1 in the Supporting Information). Rainwater was ﬁltered (0.2-μm pore size, polysulfone) and 100 μL of 6 M HCl was added to a 40-mL ﬁltered aliquot for DOC analysis. Samples were stored at 4 °C in combusted glass vials until analysis could be carried out, usually within two to three weeks of the rain event. Anion samples were unﬁltered and stored at 4 °C until analysis. A total of 637 individual rain events were sampled over the course of ten years for this study. Hydrogen Peroxide Analysis. Hydrogen peroxide was analyzed at the time of sample collection by a fluorescence decay technique involving the peroxidase-mediated oxidation of the fluorophore scopoletin by H2O2 in rain buffered at a pH of 7 with a phosphate buffer.12 Fluorescence measurements were made on a Turner Designs (Sunnyvale, CA) Model 7200000 fluorometer equipped with a H2O2 module wavelength filter (λex = 365 ( 10 nm, λem = 486 ( 10 nm). Calibration curves were obtained by recording the decrease in fluorescence upon addition of hydrogen peroxide spikes to the sample or by preparing standards in deionized water (DIW) (Milli-Q Plus Ultra Pure Water, resistivity >18 MΩ). H2O2 standard solutions were prepared fresh from a stock solution of 0.1 M H2O2, which

½NSS ¼ ½SO2 4 ð0:0517 3 ½Cl Þ

ð1Þ

Previous work at our site has demonstrated that using either sodium ion concentration or chloride ion concentration gives results that are within 1% of each other,13 less than the analytical precision of the sulfate determination. For the 637 rain events analyzed, NSS contributed an average of 76% of the total sulfate. However, this percentage varied widely and was largely driven by the origin of the air mass at the time of each rain event. The NSS in terrestrial storms typically contributed >90% of the total sulfate and marine storms usually had <10% NSS. Volume Weighted Averages and Standard Deviation Calculations. All rain events included in the study were those in which pH, H2O2, and anion values were available. To minimize the impact of highly variable concentrations of analytes in small volume rain events which would otherwise mask temporal trends, annual volume weighted averages (CVWA) were calculated. This approach is necessary in rainwater research because of the large variation in analyte concentrations between samples depending on sample volume, season, time of day, and air mass origin.10,1417 (Ranges of H2O2 and NSS concentrations are provided in Table S1.) The following equation is for the calculation of CVWA for N number of individual rain events CVWA ¼

N

N

∑ Pi 3 Ct = i∑¼ 1 Pi i¼1

ð2Þ

where Ci represents the concentration of an analyte measured in an individual rain event, and Pi is the amount of precipitation collected for that event. Volume weighted standard deviations (SDVWA) were determined by vﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ u N u 2 u u 1 i ¼ 1 Pi 3 ðCi CVWA Þ ð3Þ SDVWA ¼ u uN 3 N t Pi

∑

∑

i¼1

9539

dx.doi.org/10.1021/es202634s |Environ. Sci. Technol. 2011, 45, 9538–9542

Environmental Science & Technology

ARTICLE

Figure 2. Annual volume weighted averages and standard deviations for nonsealt sulfate (NSS) in Wilmington precipitation for the years indicated. Mann Kendall Trend Analysis shows a negative trend (p = 0.037).

Figure 1. (A) Annual volume weighted average hydrogen peroxide concentration in μM in Wilmington, NC rainwater plotted versus year from 2001 through 2010. Mann Kendall trend analysis shows a weak positive trend (p = 0.045). (B) Annual precipitation data for Wilmington, NC for 20012010 from data provided by NOAA.19

Error bars in all figures represent ( one volume weighted standard deviation unit.

’ RESULTS AND DISCUSSION The annual volume weighted average (VWA) concentration of hydrogen peroxide in Wilmington, NC rainwater over the time period 20012010 is presented in Figure 1A. There was a weak positive trend of increasing peroxide concentrations with time over this time period (Mann Kendall trend analysis18 p = 0.045). Variations in annual concentrations could not be explained by varying precipitation amounts because all years had total rainfall between 125 and 180 cm, with the exception of 2001 and 2007, which were drought years with total rainfalls of 96 and 85 cm respectively (Figure 1B).19 No anomalous behavior with respect to volume weighted average concentrations of H2O2 was observed in 2001 or 2007. Of the other analytes measured (DOC, H+, Cl, NO3, and NSS), only NSS showed a clear trend, decreasing with time between 2001 and 2010 (Mann Kendall trend analysis18 p = 0.037) (Figure 2). The decrease in annual VWA NO3 concentrations in Wilmington rainwater over the same time period was not signiﬁcant (p = 0.076, Figure S2), most likely due to increased traﬃc in this region. Local data reﬂect a 35% increase in traﬃc on the major roadway in front of our campus between 2004 and 2009, the only years in our study for which traﬃc counts were available.20 Other researchers have reported decreases in NO3 and NSS concentrations in precipitation across the continental United States21 and in Europe22,23 where more stringent emission regulations have been put in place. The annual volume weighted average hydrogen ion concentration decreased

between 1985 and 2005, and dissolved organic carbon (DOC) decreased between 1995 and 200510 although these parameters have not changed signiﬁcantly in the past decade in Wilmington rainwater. Most SO2 is emitted from burning oil and coal for electricity generation,24 whereas the primary source of NOx is on road vehicle emissions.25 Both SO2 and NOx emissions have seen a continued decrease in the United States as enhanced regulations have taken eﬀect. The US Environmental Protection Agency reports a 42% decrease in SO2 emissions nationally from 2001 to 2009, the last year data are available. This was larger than the approximate 25% decrease observed locally in Wilmington, NC.24 Similarly, the EPA reports a 32% decrease in the national average of NO2 emissions over the same time period. Because no NO2 emission data were available for Wilmington, NC, we compared this national trend to that observed in Charleston, SC, a similarly sized coastal city in the southeastern United States, and found the percent of NO2 emission decreases to be about the same as the national average.25 Oxidation of NOx to produce nitric acid (HNO3) takes place primarily in the gas phase by hydroxyl radicals or the hydrolysis of N2O5 (in equilibrium with NO3 radical and NO2). Nitric acid is subsequently removed from the gas phase through wet or dry deposition or uptake on particle surfaces.26 Sulfur dioxide, however, is primarily oxidized in the aqueous phase where it ﬁrst reacts with water to form HSO3 with trace quantities of H2SO3 and SO32‑, which is then oxidized by H2O2 or O3 to form H2SO4.7 The chemistry of SO2 oxidation by hydrogen peroxide in rainwater has been studied in great detail because of its importance in the production of acid rain. At pH < 5, the aqueous phase oxidation of SO2 to SO42‑ is controlled by H2O2 while at higher pH values, ozone (O3) is the dominant oxidant.7 When the annual volume weighted averages of H2O2 concentrations are sorted into rain events with a pH < 5 (Figure 3A) and those with a pH > 5 (Figure 3B), the pattern of increasing annual VWA H2O2 concentrations is signiﬁcant only in low pH rain samples (Mann Kendall trend analysis18 p = 0.010). Rain events with a pH < 5 constitute approximately two-thirds of the total rain events sampled in Wilmington, NC over this time period (Table S1). The relationship between annual VWA concentrations of H2O2 and NSS was investigated for all rain samples. The inverse linear correlation was signiﬁcant (p < 0.001) when just rain samples with a pH < 5 were included in the analysis (Table S2, 9540

dx.doi.org/10.1021/es202634s |Environ. Sci. Technol. 2011, 45, 9538–9542

Environmental Science & Technology

Figure 3. Annual volume weighted average hydrogen peroxide concentrations in μM for low pH (<5) and high pH (>5) rain samples. Mann Kendall trend analysis indicates a positive trend in the low pH rains (p = 0.010) and no trend in the pH > 5 rains (p > 0.05).

Figure 4A), which is consistent with the known chemistry of the aqueous-phase oxidation of SO2 by H2O2. At higher pH values, O3 is the main oxidant, and no relationship between NSS and H2O2 was observed under these conditions (Figure 4B and Table S2). Implications. The recent decrease of SO2 emissions has removed a major sink for H2O2 in the atmosphere, which may have resulted in the observed increase in H2O2 concentrations in precipitation. The increase in H2O2 driven by lower SO2 emissions has been suggested by other researchers,2 but this is the first data set to provide a lengthy enough sampling campaign to support this hypothesis. H2O2 is one of the most important oxidants in the atmosphere because it is a main source of aqueous phase •OH radicals.5 Changes in the concentration of H2O2 in precipitation presented in this manuscript suggest that there may be a rise of radically mediated transformations potentially altering the overall speciation of organic compounds and trace metals in atmospheric waters. The annual wet deposition of H2O2 in Wilmington, NC calculated from annual VWA H2O2 concentrations (Figure 1A) and local annual precipitation amounts (Figure 1B) is presented in Figure 5. When 2007, the second driest year in the 78-year climatological record in Wilmington, NC, was omitted the increase in wet deposition of hydrogen peroxide between 2001 and 2010 is signiﬁcant (Mann Kendall trend analysis18 p = 0.024). The average deposition during the preceding 5 years (ca. 27 mmol H2O2 m2 yr1 not including 2007) is more than double the approximately 10 mmol H2O2 m2 yr1 observed in 2001 and 2002. The 20012002 deposition data are near the annual wet deposition of 12 mmol H2O2 m2 yr1 determined at this location between October 1992 and October 1994.27

ARTICLE

Figure 4. (A) Inverse linear correlation between annual VWA concentrations of nonseasalt sulfate (NSS) and H2O2 concentrations for rainwater samples with a pH < 5. The slope of the linear regression is 0.6 NSS/H2O2 (r = 0.871, p < 0.001). The value marked by “X” represents the VWA for data collected at this site from 19921994. These individual events were not sorted according to pH, but approximately 80% of the events over this time period had pH < 5.27 (B) Annual VWA concentrations of nonseasalt sulfate (NSS) and H2O2 concentrations for rainwater samples with a pH > 5 showing no signiﬁcant correlation.

Figure 5. Annual wet deposition of H2O2 in Wilmington, NC determined from annual volume weighted averages and annual local rainfall amounts.

During this earlier study the VWA H2O2 concentration was 9.6 μM and NSS was 13.7 μM, which are both between the VWA’s for 2001 and 2002 (marked by “X” in Figure 4A). Comparison to the earlier data suggests that the concentrations of H2O2 and NSS at this location did not change in the ten years prior to the temporal trends presented in Figures 1A and 2. We estimate an individual rain event at this location typically deposits 350 μmol H2O2 m2 based on a rain amount of 20 mm 9541

dx.doi.org/10.1021/es202634s |Environ. Sci. Technol. 2011, 45, 9538–9542

Environmental Science & Technology (the average rainfall of all the events included in this study) and the VWA H2O2 concentration for 2010 of 17.3 μM. The addition of higher amounts of H2O2 to surface waters via rain is signiﬁcant given the disparity between typical H2O2 concentrations in rainwater (1100 μM) and surface water (10200 nM). The importance of rainfall to surface water concentrations was demonstrated quantitatively at the Bermuda Atlantic Time Series Station (BATS) during August 1999 and March 2000. Rainwater was responsible for a 2-fold increase in H2O2 concentrations throughout the 25-m mixed layer making wet deposition the dominant source of oceanic H2O2 at this location.28 The increase in wet deposition of H2O2 is particularly signiﬁcant in the oligotrophic open ocean such as BATS where it has a relatively long half-life of approximately 100 h.8,9 As a labile oxidant, the infusion of excess H2O2 could greatly inﬂuence the redox chemistry of surface waters such as altering the speciation or bioavailability of trace metals such as iron and the oxidation state of organic compounds.

’ ASSOCIATED CONTENT

bS

Supporting Information. Figures S1 and S2 and Tables S1S5. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT Our rainwater research program at UNCW has been continuously supported by a variety of NSF Atmospheric Chemistry grants since 1994 the most recent of which are AGS 0646153 and AGS 1003078. Event-based sampling and timely analysis of rainwater required by this project would not have been possible without the contributions of undergraduate students, master’s students, and postdocs. ’ REFERENCES (1) Gunz, D.; Hoﬀmann, M. R. Atmospheric chemistry of peroxides: A review. Atmos. Environ. 1990, 24A, 1601–1633. (2) Moller, D. Atmospheric hydrogen peroxide: Evidence for aqueous phase formation from a historic perspective and a one year measurement campaign. Atmos. Environ. 2009, 43, 5293–5936. (3) Kieber, R. J.; Smith, J.; Mullaugh, K. M.; Southwell, M. W.; Avery, G. B.; Willey, J. D. Inﬂuence of dissolved organic carbon on photochemically mediated cycling of hydrogen peroxide in rainwater. J. Atmos. Chem. 2009, 64 (23), 149–158. (4) Anastasio, C.; Faust, B. C.; Allen, J. M. Aqueous phase photochemical formation of hydrogen peroxide in authentic cloud waters. J. Geophys. Res. 1994, 99, 8231–8248. (5) Sakugawa, H.; Kaplan, I. R.; Tsai, W.; Cohen, Y. Atmospheric hydrogen peroxide. Environ. Sci. Technol. 1990, 24, 1452–1461. (6) Sander, R. Henry’s Law Constants. http://webbook.nist.gov (accessed June 7, 2001). (7) Calvert, J. G.; Lazarus, A.; Kok, G. L.; Heikes, B. G.; Walega, J. G.; Lind, J.; Cantrell, C. A. Chemical mechanisms of acid generation in the troposphere. Nature 1985, 317, 27–35. (8) Petasne, R. G.; Zika, R. G. Hydrogen peroxide lifetimes in south Florida coastal and oﬀshore waters. Mar. Chem. 1997, 215–225. (9) Cooper, W. J.; Saltzman, E. S.; Zika, R. G. The contribution of rainwater to variability in surface ocean hydrogen peroxide. J. Geophys. Res. 1987, 92, 2970–2980.

ARTICLE

(10) Willey, J. D.; Kieber, R. J.; Avery, G. B. Changing chemical composition of precipitation in Wilmington, North Carolina, USA: Implications for the continental USA. Environ. Sci. Technol. 2006, 40, 5675–5680. (11) Sigg, A.; Neftel, A. Evidence for a 50-percent increase in H2O2 over the past 200 years from a Greenland Ice Core. Nature 1991, 351 (6327), 557–559. (12) Kieber, R. J.; Helz, R. G. Two method veriﬁcation of hydrogen peroxide determinations in natural waters. Anal. Chem. 1986, 58, 2312– 2315. (13) Willey, J. D.; Kiefer, R. H. A contrast in winter rainwater composition: maritime versus continental rain in eastern North Carolina. Mon. Weather Rev. 1990, 118, 488–494. (14) Jickells, T. D.; Knap, A. H.; Church, T. M. Trace metals in Bermuda rainwater. J. Geophys. Res. 1984, 89, 1423–1428. (15) Khare, P.; Satsangi, G. S.; Kumar, N.; Maharaj, K.; Srivastava, S. S. HCHO, HCOOH, CH3COOH in air and rain water at a rural tropical site in north central India. Atmos. Environ. 1997, 31, 3867–3875. (16) Willey, J. D.; Kieber, R. J.; Eyman, M. S.; Avery, G. B. Rainwater dissolved organic carbon: Concentrations and global ﬂux. Global Biogeochem. Cycles 2000, 14, 139–148. (17) Topol, L. E.; Levon, M.; Flanagan, J.; Schwall, R. J.; Jackson, A. E. Quality assurance management for precipitation systems; EPA/600/ 4-82-042a; Environmental Protection Agency: Research Triangle Park, NC, 1985. (18) Helsel, D. R.; Hirsch, R. M. Statistical Methods in Water Resources, 3rd ed.; Elsevier Science Pub. Co.: 2000. (19) NOAA, Local Climatological Data and Annual Summary with Comparative Data. In National Climatic Data Center: Wilmington, NC, 2010. (20) Wilmington and New Hanover County Planning Department Report; Urban Area Metropolitan Planning Organization: Wilmington, NC, 2009. (21) Lehmann, C. M. B.; Bowersox, V. C.; Larson, S. M. Spatial and temporal trends of precipitation chemistry in the United States, 19852002. Environ. Pollut. 2005, 135 (3), 347–361. (22) van der Swaluw, E.; Asman, W. A. H.; van Jaarsveld, H.; Hoogerbrugge, R. Wet deposition of ammonium, nitrate and sulfate in the Netherlands over the period 19922008. Atmos. Environ. 2011, 45, 3819–3826. (23) Santos, P. S. M.; Otero, M.; Santos, E. B. H.; Duarte, A. C. Chemical composition of rainwater at a coastal town on the southwest of Europe: What changes in 20 years? Sci. Total Environ. 2011, 209 (18), 3548–3553. (24) National Trends in Sulfur Dioxide Levels. http://www.epa. gov/airtrends/sulfur.html (June 7, 2011). (25) National Trends in Nitrogen Dioxide Levels. http://www.epa. gov/airtrends/nitrogen.html (June 8, 2011). (26) Brasseur, G. P.; Orlando, J. J.; Tyndall, G. S. Atmospheric Chemistry and Global Change; Oxford University Press: New York, 1999. (27) Willey, J. D.; Kieber, R. J.; Lancaster, R. Coastal rainwater hydrogen peroxide: Concentration and deposition. J. Atmos. Chem. 1996, 25, 149–165. (28) Kieber, R. J.; Cooper, W. J.; Willey, J. D.; Avery, G. B. Hydrogen peroxide at the Bermuda Atlantic Time Series Station. Part 1: Temporal variability of atmospheric hydrogen peroxide and its inﬂuence on seawater concentrations. J. Atmos. Chem. 2001, 39, 1–13.

9542

dx.doi.org/10.1021/es202634s |Environ. Sci. Technol. 2011, 45, 9538–9542

ARTICLE pubs.acs.org/est

Tertiary-Treated Municipal Wastewater is a Significant Point Source of Antibiotic Resistance Genes into Duluth-Superior Harbor Timothy M. LaPara,*,† Tucker R. Burch,† Patrick J. McNamara,† David T. Tan,† Mi Yan,‡ and Jessica J. Eichmiller§ †

Department of Civil Engineering, University of Minnesota, Minneapolis, Minnesota 55455, United States Department of Bioproducts and Biosystems Engineering, University of Minnesota, St. Paul, Minnesota 55108, United States § Department of Soil, Water, and Climate, University of Minnesota, St. Paul, Minnesota 55108, United States ‡

bS Supporting Information ABSTRACT: In this study, the impact of tertiary-treated municipal wastewater on the quantity of several antibiotic resistance determinants in Duluth-Superior Harbor was investigated by collecting surface water and sediment samples from 13 locations in Duluth-Superior Harbor, the St. Louis River, and Lake Superior. Quantitative PCR (qPCR) was used to target three diﬀerent genes encoding resistance to tetracycline (tet(A), tet(X), and tet(W)), the gene encoding the integrase of class 1 integrons (intI1), and total bacterial abundance (16S rRNA genes) as well as total and human fecal contamination levels (16S rRNA genes speciﬁc to the genus Bacteroides). The quantities of tet(A), tet(X), tet(W), intI1, total Bacteroides, and human-speciﬁc Bacteroides were typically 20-fold higher in the tertiary-treated wastewater than in nearby surface water samples. In contrast, the quantities of these genes in the St. Louis River and Lake Superior were typically below detection. Analysis of sequences of tet(W) gene fragments from four diﬀerent samples collected throughout the study site supported the conclusion that tertiary-treated municipal wastewater is a point source of resistance genes into Duluth-Superior Harbor. This study demonstrates that the discharge of exceptionally treated municipal wastewater can have a statistically signiﬁcant eﬀect on the quantities of antibiotic resistance genes in otherwise pristine surface waters.

’ INTRODUCTION Over the past several decades, antibiotic-resistant bacterial infections have become increasingly prevalent, increasing morbidity and mortality as well as the cost of treatment.13 In response to these clinical concerns, there has been increasing focus on environmental reservoirs of antibiotic resistance over the past several years.48 Antibiotic use in agriculture, for example, has been heavily scrutinized9,10 and recently banned in the European Union. In contrast, the role of treated municipal wastewater has received relatively little attention as a reservoir of resistance, in spite of numerous reports suggesting that bacteria resistant to multiple antibiotics1113 and antibiotic resistance genes1421 are abundant in municipal wastewater. Determining the relative importance of treated municipal wastewater as a reservoir of antibiotic resistance is a potentially diﬃcult task. The ﬁrst challenge is to enumerate “antibiotic resistance” in some meaningful way. Historically, antibiotic resistance would have been quantiﬁed by cultivating bacteria based on their phenotypic resistance to a speciﬁc antibiotic or set of antibiotics. This approach, however, is insuﬃcient because cultivation-based methods are well-known to underestimate the quantities and diversity of bacteria.22,23 The second challenge is to distinguish the impact of treated municipal wastewater from r 2011 American Chemical Society

the background level of resistance because antibiotic resistant bacteria and antibiotic resistance genes are natural phenomena5,24 and because other human activities (i.e., other than the release of municipal wastewater) have presumably perturbed the majority of surface waters to some extent. In this study, we examined the impact of tertiary-treated municipal wastewater on the quantities of three tetracycline resistance genes (tet(A), tet(X), and tet(W)) and the integrase gene of class 1 integrons (intI1) in the St. Louis River, DuluthSuperior Harbor, and Lake Superior. This ecosystem represents an ideal locale for studying the importance of treated municipal wastewater as a reservoir of antibiotic resistance because the St. Louis River and Lake Superior are surprisingly pristine surface waters with very low background levels of bacteria,25 which suggests that the levels of antibiotic resistant bacteria also should be very low. Furthermore, the quality of treatment at the Western Lake Superior Sanitary District (WLSSD), which operates the municipal wastewater treatment facility in Duluth, MN, is exemplary. Received: August 9, 2011 Accepted: October 7, 2011 Revised: September 26, 2011 Published: October 07, 2011 9543

dx.doi.org/10.1021/es202775r | Environ. Sci. Technol. 2011, 45, 9543–9549

Environmental Science & Technology

Figure 1. Map of the St. Louis River, Duluth-Superior Harbor, and Lake Superior, identifying the locations from which samples were collected.

The WLSSD facility treats approximately 40 million gallons of residential, commercial, and industrial wastewater each day via a conventional system consisting of bar screens, grit removal, and a state-of-the-art, high-purity oxygen activated sludge process. The WLSSD wastewater treatment facility, however, is unique in that it further treats the wastewater by passing it through a mixed media ﬁlter (consisting of anthracite coal, silica sand, and garnet) before disinfecting (using sodium hypochlorite) the wastewater and discharging it to Duluth-Superior Harbor.

’ MATERIALS AND METHODS Sample Collection. Surface water (sample volume = 250 mL) and sediment (sample mass = ∼0.75 g wet sediment) samples were collected on October 1, 2010 from the St. Louis River, Duluth-Superior Harbor, and Lake Superior while aboard the R/V Blue Heron (Figure 1). Most of the surface water samples were collected manually at a distance of 0.5 m below the water surface using sterile polystyrene bottles. A small fraction of the samples (those from Lake Superior) were collected using an SBE 32 Carousel Water Sampler (Sea-Bird Electronics, Inc., Bellevue, WA) at a depth of 5 m below the water surface. Sediment samples were collected using either a multicorer (Ocean Instruments, San Sweden). Diego, CA) or a gravity-corer (HTH Teknik; Lulea, Sediment samples represent a composite sample of the top 2.5 cm of sediment. As soon as possible after collection (typically less than 30 min; always less than 2 h), surface water samples were passed through a 47 mm diameter nitrocellulose ﬁlter (pore size = 0.22 μm) to concentrate microbial biomass. Filters were then immersed in 0.5 mL of lysis buﬀer (120 mM phosphate buﬀer, pH = 8.0, 5% sodium dodecyl sulfate) to preserve the sample until genomic DNA could be extracted and puriﬁed. All samples were stored on ice while they were transported to the University of Minnesota (within 12 h), after which they were stored at 20 °C until processed further. Genomic DNA Extraction. Water samples (preserved in lysis buffer) underwent three consecutive freezethaw cycles and an incubation of 90 min at 70 °C to lyse cells. Genomic DNA was then extracted and purified from these samples using the FastDNA Spin Kit (MP Biomedicals, Solon, OH) according to manufacturer’s instructions. Genomic DNA was also extracted from sediment samples (∼500 mg of wet weight per sample)

ARTICLE

using a bead beater to lyse cells. Genomic DNA was then extracted and purified from sediment samples using a FastDNA Spin Kit for soil (MP Biomedicals, Solon, OH). All genomic DNA extractions were performed in triplicate and stored at 20 °C until needed. Community Analysis. The composition of the bacterial communities in the aquatic samples was compared by automated ribosomal intergenic spacer analysis (ARISA). The ribosomal intergenic spacer (ITS) regions of Bacteria were amplified using primers ITSF (50 -GTC GTA ACA AGG TAG CCG TA-30 ) and ITSReub (50 -GCC AAG GCA TCC ACC-30 )26 as described previously.27 Fragment analysis was performed by denaturing capillary electrophoresis at the Biomedical Genomics Center at the University of Minnesota using an ABI 3130xl Genetic Analyzer (Applied Biosystems, Foster City, CA). The length of the fragments was estimated using the Map Marker 1000 size standard. Quantitative PCR. Quantitative real-time PCR (qPCR) was used to quantify the presence of three genes encoding tetracycline resistance (tet(A), tet(W), and tet(X)) and the integrase gene of class 1 integrons (intI1) as described previously.15 These genes were targeted in this study because our prior work demonstrated that these genes were easily detectable in untreated wastewater solids15,16 and because these genes encode proteins that confer tetracycline resistance via each of the three known mechanisms of resistance.28 qPCR was also used to quantify the 16S rRNA genes of all members of the domain Bacteria as well as total and human-specific Bacteroides spp. as described previously.2931 The qPCR analysis was conducted using an Eppendorf Mastercycler ep realplex thermal cycler (Eppendorf, Westbury, NY) or an ABI Prism7000 Sequence Detection System (Applied Biosystems). Each qPCR run consisted of initial denaturation for 10 min at 95 °C, followed by 40 cycles of denaturation at 95 °C for 15 s, and anneal and extension at 60 °C (most targets) or at 56 °C (human-speciﬁc Bacteroides) for 1 min. A 25 μL reaction mixture contained 12.5 μL of iTaq SYBR Green Supermix with ROX (Bio-Rad, Hercules, CA), 25 μg bovine serum albumin (Roche Applied Science, Indianapolis, IN), optimized quantities of forward and reverse primers, and a speciﬁed volume of template DNA (usually 0.5 μL). The precise volume and concentration of template DNA were empirically optimized for each sample to generate the lowest detection limit while minimizing inhibition of PCR. Additional information on the qPCR primers, their quantiﬁcation limits, and their associated products are provided in the Supporting Information. The quantity of target DNA in unknown samples was calculated based on a standard curve generated using known quantities of template DNA. Standards for qPCR were prepared by PCR ampliﬁcation of genes from positive controls, followed by ligation into a cloning vector (either the StrataClone PCR kit (Stratagene, Santa Clara, CA) or pGEM-T Easy (Promega, Madison, WI)), and transformation into E. coli JM109. Plasmids were puriﬁed using the alkaline lysis procedure.32 Plasmid DNA was quantiﬁed by staining with Hoechst 33258 dye and measured on a TD-700 ﬂuorometer (Turner Designs, Sunnyvale, CA) using calf thymus as a DNA standard. Tenfold serial dilutions of plasmid DNA were prepared and run on the thermal cycler to generate standard curves (r2 > 0.99). Following qPCR, melting curves were generated and analyzed to verify that nonspeciﬁc ampliﬁcation did not occur. Clone Libraries. Fragments of tet(W) genes from four different surface water samples (samples SLR5, DH2, WW, and LS2) 9544

dx.doi.org/10.1021/es202775r |Environ. Sci. Technol. 2011, 45, 9543–9549

Environmental Science & Technology were amplified by PCR, purified, ligated into the pGEM-T Easy cloning vector, transformed into Escherichia coli JM109, and plated onto LB agar plates supplemented with 40 μg/mL of ampicillin. This resulted in libraries of tet(W) gene fragments from each of these samples, allowing their nucleotide sequences to be determined. Approximately 30 colonies from each library were randomly picked so that plasmids could be extracted and purified using the alkaline lysis method. Extracted plasmids were then used as template for nucleotide sequence analysis using M13F and M13R as sequencing primers. Bidirectional sequence information was then used to produce a consensus sequence. Approximately 20% of the plasmids contained primerdimer rather than a genuine tet(W) gene fragment; these sequences were excluded from further analysis. Data Analysis. Nonmetric multidimensional scaling (nMDS) was used on triplicate ARISA profiles to evaluate differences in bacterial community composition based on the presence and intensity of peaks in the electropherograms. The relative

ARTICLE

intensity of peaks, obtained by dividing the individual intensities by total intensity of all the peaks, was used in the analysis. Peaks falling below 1% of the total intensity were excluded from the analysis. nMDS was performed using the ade4 package in R, version 2.4.1.33 Prior to statistical analysis, samples with gene concentrations below the method detection limit were assigned a value equal to half the detection limit. All gene concentrations were then logtransformed, and this log-transformed data set was used for all subsequent statistical analysis. One-way analysis of variance (ANOVA) was performed with R version 2.12.0 for all gene targets. An F-test was conducted to determine if results from a speciﬁc surface water sample location exhibited gene concentrations that were signiﬁcantly diﬀerent from results at the other sample locations. Tukey’s honestly signiﬁcant diﬀerence (HSD) test was conducted for each gene target to determine the diﬀerence in mean gene concentrations between each possible pair of surface water samples sites. Pearson correlation coeﬃcients of gene concentrations were also calculated using R version 2.12.0 for all possible pairs of gene targets. The detailed results of all statistical analyses (i.e., P values and/or Pearson correlation coeﬃcients) are provided in the Supporting Information. All nucleotide sequences were initially compared with sequences in the GenBank database34 to verify that the cloned fragments were genuine tet(W) gene fragments. Sequences were then aligned using the ClustalW algorithm35 using DNAMan version 7.0 software (Lynnon Biosoft, Vaudreuil-Dorion, Quebec, Canada). To avoid artifacts stemming from misampliﬁcation during PCR and nucleotide sequencing error,36 all sequences for which there was not a replicate were excluded from further analysis.

’ RESULTS

Figure 2. Results of nonmetric multidimensional scaling (nMDS) analysis of bacterial community composition as determined by automated ribosomal intergenic spacer analysis. Ellipses show the 95% conﬁdence limit of triplicate water samples. Samples were collected from the St. Louis River (identiﬁed as “SLR”), Duluth-Superior Harbor (identiﬁed as “DH”), and Lake Superior (identiﬁed as “LS”); the precise locations from which samples were collected are shown in Figure 1.

Bacterial Community Composition. The composition of the bacterial communities in surface water samples collected along a length of the St. Louis River, Duluth-Superior Harbor, and Lake Superior was assayed by automated ribosomal intergenic spacer analysis (ARISA) (Figure 2). The bacterial community composition gradually transitioned along the length of the St. Louis River, into Duluth-Superior Harbor, and out into Lake Superior. In contrast, the composition of bacteria in the treated municipal wastewater from the Western Lake Superior Sanitary District (WLSSD) was significantly different than all of the surface water samples. Quantitative PCR. The amount of bacterial biomass was quantified in the surface water samples by real-time PCR of 16S rRNA gene fragments (Figure 3). Bacterial biomass in the

Figure 3. Quantities (gene copies per mL) of 16S rRNA genes in water samples collected from the St. Louis River, Duluth-Superior Harbor, the outfall from the Western Lake Superior Sanitary District, and Lake Superior. Values shown are the arithmetic means; error bars show the standard deviation of the mean. The locations from which samples were collected are shown in Figure 1. 9545

dx.doi.org/10.1021/es202775r |Environ. Sci. Technol. 2011, 45, 9543–9549

Environmental Science & Technology

ARTICLE

Figure 4. Quantities (gene copies per mL of water) of tet(A), tet(W), tet(X), and the integrase gene of class 1 integrons (intI1) in samples collected from the St. Louis River, Duluth-Superior Harbor, the outfall from the Western Lake Superior Sanitary District, and Lake Superior. Values shown are the arithmetic means; error bars show the standard deviation of the mean. The locations from which samples were collected are shown in Figure 1.

different surface water samples varied substantially from as high as 3.6 106 gene copies per mL (sample location = SLR2) to as low as 2.1 105 gene copies per mL (sample location = LS2). These quantifications of 16S rRNA gene copies are substantially lower than that previously reported from the Haihe River in China (108109 copies per mL)37 and from a drinking water source in Michigan (3.4 109 copies per mL),38 but are consistent with previously reported direct cell counts from Lake Superior (1 105 cells/mL).25 The quantity of bacterial biomass in the treated WLSSD effluent was 5.4 106 gene copies per mL, which was higher than any surface water sample. The quantities of three diﬀerent genes that encode resistance to tetracycline (tet(A), tet(X), and tet(W)) as well as the quantity of the integrase gene (intI1) of class 1 integrons were also determined along the St. Louis River, in Duluth-Superior Harbor, and in Lake Superior (Figure 4; for the same data normalized to 16S rRNA genes, see Supporting Information). The quantities of tet(A) and tet(X) followed similar patterns in the aquatic samples; both of these genes were at relatively high concentrations in the WLSSD eﬄuent (tet(A): 6.3 102 copies per mL; tet(X): 1.2 103 copies per mL), slightly above the detection limit at several locations within Duluth-Superior Harbor, and below the detection limit in the St. Louis River and in Lake Superior. The pattern of intI1 genes was somewhat similar to that observed with tet(A) and tet(X), except that a more distinct hump-shaped proﬁle, albeit slightly skewed into Duluth-Superior

Harbor, was observed; this hump-shaped proﬁle began in the St. Louis River and encompassed all but one sample collected from Duluth-Superior Harbor. An entirely diﬀerent proﬁle was observed with respect to the quantity of tet(W) genes, which were quantiﬁable in every aquatic sample with only the WLSSD eﬄuent (1.8 104 gene copies per mL) and one sample from Duluth-Superior Harbor (sample DH4: 5.3 103 gene copies per mL) being statistically greater (P < 0.05) than the other samples. Because the quantities of 16S rRNA genes were relatively constant among the diﬀerent water samples (i.e., within an order of magnitude), the quantities of tet(A), tet(X), tet(W), and intI1 normalized to 16S rRNA genes follow similar patterns to those described above (see Supporting Information for more details). The quantities of 16S rRNA genes from all Bacteroides spp. in the aquatic samples followed a trend similar to that observed with the tet(W) quantities (Figure 5A). The highest quantity of Bacteroides spp. was found in the WLSSD eﬄuent (6.8 103 gene copies per mL), but otherwise most of the samples had relatively low concentrations that were similar. In contrast, the quantities of human-speciﬁc Bacteriodes spp. followed a trend like that of tet(A) and tet(X), in which a relatively high concentration was detected in the WLSSD eﬄuent (1.0 102 gene copies per mL); two samples from Duluth-Superior Harbor had quantities slightly higher than the detection limit, but then all other samples were below the detection limit. 9546

dx.doi.org/10.1021/es202775r |Environ. Sci. Technol. 2011, 45, 9543–9549

Environmental Science & Technology

ARTICLE

Figure 5. Quantities (gene copies per mL of water) of 16S rRNA genes from all Bacteriodes spp. and from human-speciﬁc Bacteroides spp. in water samples collected from the St. Louis River, Duluth-Superior Harbor, the outfall from the Western Lake Superior Sanitary District, and Lake Superior. Values shown are the arithmetic means; error bars show the standard deviation of the mean. The locations from which samples were collected are shown in Figure 1.

Table 1. Arithmetic Means (Units = gene copies per wet gram of sediment; n = 3) of Various Genes Detected in Sediment Samples Collected near the WLSSD Outfall, from Duluth-Superior Harbor, and from Lake Superior (See Figure 1 for Actual Locations)a sample location

all Bacteroides spp.

all 16S rRNA

tet(A)

tet(X)

tet(W)

intI1

WW

1.9 10

10

(6.5 10 )

6.9 10 (3.9 10 )

2.1 10 (6.8 10 )

2.9 10 (5.4 10 )

1.9 10 (9.3 10 )

2.4 106 (6.6 105)

DH1

2.1 10

10

(5.1 10 )

2.3 10 (1.4 10 )

7.4 10 (1.9 10 )

1.2 10 (2.3 10 )

6.5 10 (3.7 10 )

2.5 106 (5.5 105)

DH3

1.7 10

10

(5.1 10 )

4.9 10 (2.8 10 )

3.7 10 (6.8 10 )

b.d.

2.6 10 (1.4 10 )

7.7 105 (4.3 104)

LS2

1.9 1010 (5.7 109)

2.3 104 (3.2 103)

1.2 105 (1.2 104)

b.d.

1.1 104 (4.1 103)

4.9 105 (8.7 104)

9 9 8

4 5 4

4 5 4

6 5 5

5 5 4

4 4

b

3 3

4 4 4

3 4 4

a

Human-speciﬁc Bacteroides spp. were also targeted by real-time PCR, but were below the quantiﬁcation limit in all four sample locations. The numbers in parentheses represent the standard deviation of the mean. b b.d., below detection.

Four sediment samples were also collected from DuluthSuperior Harbor and Lake Superior (Table 1). Each of these samples had similar concentrations of total bacteria, as measured by qPCR of 16S rRNA genes. The quantities of the other genetic markers tracked in this study, however, varied signiﬁcantly depending on sample location (except for human-speciﬁc Bacteroides spp., which were not detected in any of the sediment samples). The highest concentrations of these genetic markers were detected in sediment samples collected from near the WLSSD outfall (samples WW and DH1) compared to the samples collected from the DuluthSuperior Harbor channel (sample DH3) and from Lake Superior (sample LS2). PCR Cloning of tet(W) Gene Fragments. In a previous study, tet(W) gene sequences corresponded to the location from which they originated (i.e., from agriculture, from municipal wastewater, etc.).39 Nucleotide sequences, therefore, were determined from four different clone libraries (from samples SLR5, WW, DH2, and LS2) of tet(W) gene fragments to determine whether or not the type of tet(W) genes varied in the St. Louis River, Duluth-Superior Harbor, and Lake Superior. Comparing only nucleotide sequences for which a matching nucleotide sequence was detected (i.e., singletons were excluded from consideration), only two distinct clones were detected.

The first of these clone types (100% sequence identity to GenBank accession no. GU116971) comprised 100% of the clone library from the St. Louis River sample (sample = SLR5; n = 17), slightly less than half of the clone library from the Duluth-Superior Harbor sample (sample = DH2; 8 out of 17 clones), and the majority of the clones from the Lake Superior sample (sample = LS2; 17 out of 20 clones). In contrast, the second clone type (100% sequence identity to GenBank accession no. AP012212) represented 100% of the library from the sample collected from the tertiary-treated wastewater (sample = WW; n = 14), slightly more than half of the Duluth-Superior Harbor clone library (9 out of 17 clones), and a small fraction of the LS2 library (3 out of 20 clones).

’ DISCUSSION The importance of municipal wastewater treatment as a necessary component of modern society is without question.40,41 The primary goal of municipal wastewater treatment is to protect surface water quality from the adverse eﬀects of the relatively high concentration of nutrients (biodegradable carbon, nitrogen, and phosphorus) in the sewage; the secondary goal of municipal wastewater treatment is to protect public health from direct exposure to pathogens (usually via accidental ingestion of surface water).41 An unintended consequence of municipal wastewater 9547

dx.doi.org/10.1021/es202775r |Environ. Sci. Technol. 2011, 45, 9543–9549

Environmental Science & Technology treatment, however, is the creation of a centralized location where bacteria from the microﬂora of healthy and unhealthy humans coalesce. Municipal wastewater and municipal wastewater treatment, therefore, simultaneously represent a pertinent reservoir of resistance and a potential opportunity to ameliorate this reservoir of resistance, respectively. The present study demonstrates that treated municipal wastewater is a statistically signiﬁcant point source of three tetracycline resistance determinants as well as the integrase gene of class 1 integrons into Duluth-Superior Harbor. The tertiary-treated wastewater had approximately 20-fold higher concentrations of various antibiotic resistance determinants than the local background levels in the St. Louis River and Lake Superior. Furthermore, the concentrations of antibiotic resistance genes generally correlated to either all Bacteroides spp. (a measure of total fecal material) or human-speciﬁc Bacteroides spp. (a measure of human-generated fecal material) (see Supporting Information for more details). Finally, the sequence of tet(W) gene fragments in tertiary-treated wastewater was unique compared to that compared to that found in the St. Louis River and Lake Superior, again suggesting that the tertiary-treated municipal wastewater was a signiﬁcant source of antibiotic resistant determinants into Duluth-Superior Harbor where approximately equal amounts of these two gene sequences were detected. The present study is unique and novel because of its ability to clearly identify tertiary-treated municipal wastewater as a point source of antibiotic resistance genes, which have been identiﬁed as an emerging pollutant of concern.42 Previous studies in which treated municipal wastewater was implicated as a source of antibiotic resistance determinants were substantially more convoluted because multiple sources of antibiotic resistance genes existed, such as agricultural activity and industrial wastewater discharges.39,43 In contrast, the current study is considerably more straightforward to interpret because of the general transition from pristine (St. Louis River) to relatively perturbed (Duluth-Superior Harbor) back to pristine (Lake Superior), with virtually no known anthropogenic sources of antibiotic resistance genes other than a large input of tertiary-treated municipal wastewater from WLSSD (ﬂow rate = 40 million gallons per day) and a small input of secondary-treated municipal wastewater from Superior, Wisconsin (ﬂow rate = 5 million gallons per day; near sample location DH4). In conclusion, municipal wastewater treatment operations need to be more carefully considered as an important factor in the global ecology of antibiotic resistance. Municipal wastewater contains numerous types of waste, of which human fecal material is known to have substantial concentrations of both antibiotic resistant bacteria and antibiotic resistance genes.44 Municipal wastewater treatment operations undoubtedly remove a very large fraction of the antibiotic resistance genes in untreated sewage prior to discharging the treated eﬄuent. This study demonstrates that even tertiary-treated municipal wastewater is a statistically signiﬁcant source of antibiotic resistance genes in otherwise pristine surface waters; additional research is needed to determine the importance of treated municipal wastewater in the overall proliferation of antibiotic resistance.

’ ASSOCIATED CONTENT

bS

Supporting Information. Information regarding the qPCR primers and conditions (Table S1) and statistical analyses (Tables S2S10). Additional results are also included regarding

ARTICLE

tetracycline resistance and intI1 genes normalized to 16S rRNA genes (Figure S1). This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Tel.: (612) 624-6028. Fax: (612) 626-7750. E-mail: lapar001@ umn.edu.

’ ACKNOWLEDGMENT This work was supported by the National Science Foundation (CBET-0967176) and by the Minnesota Environment and Natural Resources Trust Fund (for ship time on the R/V Blue Heron). The majority of this work was performed during the laboratory portion of CE 8551: Environmental Microbiology— Molecular Theory and Methods during the Fall 2010 semester at the University of Minnesota. The assistance of the crew of the R/V Blue Heron is gratefully acknowledged. Dan Engstrom, Bill Arnold, Cale Anger, Noah Hensley, and Hao Pang are also gratefully acknowledged for their assistance with the collection and processing of sediment samples. ’ REFERENCES (1) Andersson, D. I.; Hughes, D. Antibiotic resistance and its cost: is it possible to reverse resistance? Nat Rev. Microbiol. 2010, 8, 260–271. (2) Levy, S. B. The antibiotic paradox: How miracle drugs are destroying the miracle; Plenum Press: New York, 1992. (3) Levy, S. B. Antibiotic resistance: Consequences of inaction. Clin. Infect. Dis. 2001, 33, S124–S129. (4) Alonso, A.; Sanchez, P.; Martínez, J. L. Environmental selection of antibiotic resistance genes. Environ. Microbiol. 2001, 3, 1–9. (5) Allen, H. K.; Donato, J.; Wuimi Wang, H.; Cloud-Hansen, K. A.; Davies, J.; Handelsman, J. Call of the wild: Antibiotic resistance genes in natural environments. Nat. Rev. Microbiol. 2010, 8, 251–259. (6) Canton, R. Antibiotic resistance genes from the environment: a perspective through newly identiﬁed antibiotic resistance mechanisms in the clinical setting. Clin. Microbiol. Infect. 2009, 15, 20–25. (7) D’Costa, V. M.; Griﬃths, E.; Wright, G. D. Expanding the soil antibiotic resistome: exploring environmental diversity. Curr. Opin. Microbiol. 2007, 10, 481–489. (8) Martínez, J. Antibiotics and antibiotic resistance genes in natural environments. Science 2008, 321, 365–367. (9) Lipsitch, M.; Singer, R. S.; Levin, B. R. Antibiotics in agriculture: when is it time to close the barn door? Proc Natl. Acad. Sci. U.S.A. 2002, 99, 5752–5754. (10) Smith, D. L.; Dushoﬀ, J.; Morris, J. G., Jr. Agricultural antibiotics and human health. PLoS Med. 2005, 2, 731–735. (11) Merlin, C.; Bonot, S.; Courtois, S.; Block, J.-C. Persistence and dissemination of the multiple-antibiotic-resistance plasmid pB10 in the microbial communities of wastewater sludge microcosms. Water Res. 2011, 45, 2897–2905. (12) Novo, A.; Manaia, C. M. Factors inﬂuencing antibiotic resistance burden in municipal wastewater treatment plants. Appl. Microbiol. Biotechnol. 2010, 87, 1157–1166. (13) Ramsden, S. J.; Ghosh, S.; Bohl, L. J.; LaPara, T. M. Phenotypic and genotypic analysis of bacteria isolated from three municipal wastewater treatment plants on tetracycline-amended and ciproﬂoxacinamended growth media. J. Appl. Microbiol. 2010, 109, 1609–1618. (14) Auerbach, E. A.; Seyfried, E. E.; McMahon, K. D. Tetracycline resistance genes in activated sludge wastewater treatment plants. Water Res. 2007, 41, 1143–1151. (15) Diehl, D. L.; LaPara, T. M. Eﬀect of temperature on the fate of genes encoding tetracycline resistance and the integrase of class 1 9548

dx.doi.org/10.1021/es202775r |Environ. Sci. Technol. 2011, 45, 9543–9549

Environmental Science & Technology integrons within anaerobic and aerobic digesters treating municipal wastewater solids. Environ. Sci. Technol. 2010, 44, 9128–9133. (16) Ghosh, S.; Ramsden, S. J.; LaPara, T. M. The role of anaerobic digestion in controlling the release of tetracycline resistance genes and class 1 integrons from municipal wastewater treatment plants. Appl. Microbiol. Biotechnol. 2009, 84, 791–796. (17) Moura, A.; Henriques, I.; Smalla, K.; Correia, A. Wastewater bacterial communities bring together broad-host range plasmids, integrons and a wide diversity of uncharacterized gene cassettes. Res. Microbiol. 2010, 161, 58–66. (18) Munir, M.; Wong, K.; Xagoraraki, I. Release of antibiotic resistant bacteria and genes in the eﬄuent and biosolids of ﬁve wastewater utilities in Michigan. Water Res. 2011, 45, 681–693. (19) Zhang, T.; Zhang, M.; Zhang, X.; Fang, H. H. P. Tetracycline resistance genes and tetracycline resistant lactose-fermenting Enterobacteriaceae in activated sludge of sewage treatment plants. Environ. Sci. Technol. 2009, 43, 3455–3460. (20) Zhang, X.; Zhang, T.; Zhang, M.; Fang, H. H. P; Cheng, S.-P. Characterization and quantiﬁcation of class 1 integrons and associated gene cassettes in sewage treatment plants. Appl. Microbiol. Biotechnol. 2009, 82, 1169–1177. (21) Zhang, X.-X.; Zhang, T. Occurrence, abundance, and diversity of tetracycline resistance genes in 15 sewage treatment plants across China and other global locations. Environ. Sci. Technol. 2011, 45, 2598–2604. (22) Amann, R. I.; Ludwig, W.; Schleifer, K. H. Phylogenetic identiﬁcation and in situ detection of individual microbial cells without cultivation. Microbiol. Rev. 1995, 59, 143–169. (23) Staley, J. T.; Konopka, A. Measurement of in situ activities of nonphotosynthetic microorganisms in aquatic and terrestrial habitats. Annu. Rev. Microbiol. 1985, 39, 321–346. (24) Davies, J.; Davies, D. Origins and evolution of antibiotic resistance. Microbiol. Mol. Biol. Rev. 2010, 74, 417–433. (25) Cotner, J. B.; Ogdahl, M. L.; Biddanda, B. A. Double-stranded DNA measurement in lakes with the ﬂuorescent stain PicoGreen and the application to bacterial bioassays. Aquat. Microb. Ecol. 2001, 25, 65–74. (26) Cardinale, M.; Brusetti, L.; Quatrinia, P.; Borin, S.; Puglia, A.; Rizzi, A.; Zanardini, E.; Sorlini, C.; Corselli, C.; Daﬀonchio, D. Comparison of diﬀerent primer sets for use in automated ribosomal intergenic spacer analysis of complex bacterial communities. Appl. Environ. Microbiol. 2004, 70, 6147–6156. (27) Nelson, D. K.; LaPara, T. M.; Novak, P. J. Eﬀects of ethanolbased fuel contamination: Microbial community changes, production of regulated compounds, and methane generation. Environ. Sci. Technol. 2010, 44, 4525–4530. (28) Chopra, I.; Roberts, M. 2001. Tetracycline antibiotics: mode of action, applications, molecular biology, and epidemiology of bacterial resistance. Microbiol. Mol. Biol. Rev. 2001, 65, 232-260. (29) Berhardt, A. E.; Field, K. G. 2000. Identiﬁcation of nonpoint sources of fecal pollution in coastal waters by using host-speciﬁc 16S ribosomal DNA genetic markers from fecal anaerobes. Appl. Environ. Microbiol. 2000, 66, 15871594. (30) Berhardt, A. E.; Field, K. G. A PCR assay to discriminate human and ruminant feces on the basis of host diﬀerences in BacteroidesPrevotella genes encoding 16S rRNA. Appl. Environ. Microbiol. 2000, 66, 4571–4574. (31) Seurinck, S.; Deﬁordt, T.; Verstraete, W.; Siciliano, S. D. Detection and quantiﬁcation of the human-speciﬁc HF183 Bacteroides 16S rRNA genetic marker with real-time PCR for assessment of human faecal pollution in freshwater. Environ. Microbiol. 2005, 7, 249–259. (32) Sambrook, J.; Fritsch, E. F.; Maniatis, T. Molecular Cloning: A laboratory manual, 2nd ed.; Cold Spring Harbor Laboratory: Cold Spring Harbor, NY, 1989. (33) Thioulouse, J.; Chessel, D.; Doledec, S.; Olivier, J. M. Ade-4: a multivariate analysis and graphical display software. Stat. Comput. 1997, 7, 75–83. (34) Benson, D. A.; Karsch-Mizrachi, I.; Lipman, D. J.; Ostell, J.; Rapp, B. A.; Wheeler, D. L. GenBank. Nucleic Acids Res. 2002, 30, 17–20.

ARTICLE

(35) Thompson, J. D.; Higgins, D. G.; Gibson, T. J. CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-speciﬁc gap penalties and weight matrix choice. Nucleic Acids Res. 1994, 22, 4673–4680. (36) Ling, L. L.; Keohavong, P.; Dias, C.; Thilly, W. G. Optimization of the polymerase chain reaction with regard to ﬁdelity: modiﬁed T7, Taq, and vent DNA polymerases. PCR Methods Appl. 1991, 1, 63–69. (37) Yi, L.; Daqing, M.; Rysz, M.; Qixing, Z.; Hongjie, Z.; Lin, X.; Alvarez, P. J. J. Trends in antibiotic resistance genes occurrence in the Haihe River, China. Environ. Sci. Technol. 2010, 44, 7220–7225. (38) Xi, C.; Zhang, Y.; Marrs, C. F.; Ye, W.; Simon, C.; Foxman, B.; Nriagu, J. Prevalence of antibiotic resistance in drinking water treatment and distribution systems. Appl. Environ. Microbiol. 2009, 75, 5714–5718. (39) Storteboom, H.; Arabi, M.; Davis, J. G; Crimi, B.; Pruden, A. Tracking antibiotic resistance genes in the South Platte River basin using molecular signatures of urban, agricultural, and pristine sources. Environ. Sci. Technol. 2010, 44, 7397–7404. (40) Arthurson, V. Proper sanitization of sewage sludge: a critical issue for a sustainable society. Appl. Environ. Microbiol. 2008, 74, 5267–5275. (41) Tchobanoglous, G.; Burton, F. L.; Stensel, H. D. Wastewater engineering: Treatment and Reuse, 4th ed.; Metcalf and Eddy, Inc., McGraw-Hill: Boston, MA, 2003. (42) Pruden, A.; Pei, R.; Storteboom, H.; Carlson, K. H. Antibiotic resistance genes as emerging contaminants: studies in northern Colorado. Environ. Sci. Technol. 2006, 40, 7445–7450. (43) Graham, D. W.; Olivares-Rieumont, S.; Knapp, C. W.; Lima, L.; Werner, D.; Bowen, E. Antibiotic resistance gene abundances associated with waste discharges to the Almendares River near Havana, Cuba. Environ. Sci. Technol. 2011, 45, 418–424. (44) Sommer, M. O. A.; Dantas, G.; Church, G. M. Functional characterization of the antibiotic resistance reservoir in the human microﬂora. Science 2009, 325, 1128–1131.

9549

dx.doi.org/10.1021/es202775r |Environ. Sci. Technol. 2011, 45, 9543–9549

ARTICLE pubs.acs.org/est

Spectroscopic Evidence for Ternary Complex Formation between Arsenate and Ferric Iron Complexes of Humic Substances Christian Mikutta*,† and Ruben Kretzschmar† †

Institute of Biogeochemistry and Pollutant Dynamics, Department of Environmental Sciences, ETH Zurich, 8092 Zurich, Switzerland

bS Supporting Information ABSTRACT: Formation of ternary complexes between arsenic (As) oxyanions and ferric iron (Fe) complexes of humic substances (HS) is often hypothesized to represent a major mechanism for AsHS interactions under oxic conditions. However, direct evidence for this potentially important binding mechanism is still lacking. To investigate the molecular-scale interaction between arsenate, As(V), and HS in the presence of Fe(III), we reacted fulvic and humic acids with Fe(III) (1 wt %) and equilibrated the Fe(III)-HS complexes formed with As(V) at pH 7 (molar Fe/As ∼10). The local (<5 Å) coordination environments of As and Fe were subsequently studied by means of X-ray absorption spectroscopy. Our results show that 4.5 12.5 μmol As(V)/g HS (25 70% of total As) was associated with Fe(III). At least 70% of this As pool was bound to Fe(III)-HS complexes via inner-sphere complexation. Results obtained from shell ﬁts of As K-edge extended X-ray absorption ﬁne structure (EXAFS) spectra were consistent with a monodentate binuclear (2C) and monodentate mononuclear (1V) complex stabilized by H-bonds (RAs Fe = 3.30 Å). The analysis of Fe K-edge EXAFS spectra revealed that Fe in Fe(III)-HS complexes was predominantly present as oligomeric Fe(III) clusters at neutral pH. Shell-ﬁt results complied with a structural motif in which three corner-sharing Fe(O,OH)6 octahedra linked by a single μ3-O bridge form a planar Fe trimer. In these complexes, the average Fe C and Fe Fe bond distances were 2.95 Å and 3.47 Å, respectively. Our study provides the ﬁrst spectroscopic evidence for ternary complex formation between As(V) and Fe(III)-HS complexes, suggesting that this binding mechanism is of fundamental importance for the cycling of oxyanions such as As(V) in organic-rich, oxic soils and sediments.

’ INTRODUCTION Arsenic (As) is a toxic trace element jeopardizing water and soil resources worldwide. Depending on the prevailing redox conditions, As is predominantly present as As(III) (arsenite) in anoxic and As(V) (arsenate) in oxic environments. While the interactions of both As species with mineral matter have received much attention in the past,1 4 natural organic matter (NOM) as a ubiquitous sorbent for As has long been neglected by geochemists.5 8 However, a growing body of direct evidence now suggests that both As(V) and As(III) can bind to NOM.9 17 Other studies provided additional indirect evidence for As-NOM interactions. For example, Gonzalez A. et al.18 used sequential extractions and inferred that up to 73% of As in a naturally Asenriched minerotrophic peatland was associated with the organic matter fraction. Correlations between dissolved As and dissolved organic matter have also been taken as indicators for As-NOM interactions.19,20 Although there is overwhelming evidence that inorganic As species can sorb to NOM, our knowledge about the governing binding mechanisms is still surprisingly limited. Yet, this knowledge is essential to accurately predict the fate of As in the environment. Several binding mechanisms have been hypothesized for As(V) and As(III), which include (i) the formation of ternary complexes with a polyvalent metal cation r 2011 American Chemical Society

forming a bridge between negatively charged As oxyanions and organic ligands10,21 (Figure 1a, b), (ii) the formation of outersphere complexes with protonated amino groups of NOM,9 and (iii) the formation of covalent bonds between phenolate/carboxylate groups of NOM and As(III) or As(V).13 Ternary As complex formation is by far the most popular binding mechanism invoked to explain As binding to NOM,10 12,14 17,21 27 suggesting that it is widely accepted and pointing toward its environmental relevance. Perusal of these studies, however, revealed a complete lack of direct evidence for ternary As complexes because neither were spectroscopic methods employed in these studies nor could the presence of particulate Fe be ruled out. The fact that the existence of ternary As complexes is regularly spread in the literature without solid foundation calls for an evaluation of this potentially important binding mechanism. Similar complexes have also been proposed for other oxyanions such as phosphate28 32 or selenite,33 but in all cases no convincing spectroscopic data have been presented to date. The main Received: July 5, 2011 Accepted: October 10, 2011 Revised: October 6, 2011 Published: October 10, 2011 9550

dx.doi.org/10.1021/es202300w | Environ. Sci. Technol. 2011, 45, 9550–9557

Environmental Science & Technology

Figure 1. Schematic representation of binding modes of As(V) in ternary As(V) Fe(III) HS complexes in which As(V) forms one or more polar-covalent bonds. For simplicity the organic ligand is represented by succinate (a, b) and oxalate (c, d). In (a) the As(V) tetrahedron is linked to an Fe(O,OH)6 octahedron via a single μ-oxo bridge (corner-sharing polyhedrons). In (b) two μ-oxo bridges link the As(V) tetrahedron with one octahedron resulting in edge-sharing polyhedrons. Graphs (c) and (d) illustrate possible binding modes of As(V) inferred from this study: Monodentate binuclear 2C complex on a trimeric Fe(III) cluster (c) and monodentate mononuclear 1V complex with H-bonding (arrow) to an adjacent Fe(O,OH)6 octahedron (d).

objective of this study was therefore to test the formation of ternary complexes between oxyanions and Fe(III)-NOM complexes. To this end, we reacted As(V) with Fe(III) complexes of humic substances (HS) at pH 7 and studied the local coordination environments (<5 Å) of As and Fe by means of K-edge X-ray absorption spectroscopy.

’ EXPERIMENTAL SECTION Humic Substances. Pahokee peat humic acid (1S103H) and Pahokee peat fulvic acid (2S103F) were purchased from the International Humic Substances Society. These materials were used without further purification. Details about their composition are given in Table S1 of the Supporting Information (SI). Sample Preparation. Fe(III)-HS complexes were prepared similar to Karlsson and Persson.34 Two hundred milligrams of humic and fulvic acid powders were weighed into 10-mL polypropylene vials sealed with Al foil to exclude potential photoreactions. Then, 2.4 mL of an acidified 14.92 mM Fe(III) nitrate stock solution were added, resulting in a target Fe loading of 10,000 μg/g organic matter. The measured pH values of the humic and fulvic acid suspensions after Fe(III) addition were 1.5 and 1.7, respectively. The suspensions were subsequently sonicated for 5 min and shaken on a reciprocating shaker (1000 rpm) at room temperature (∼24 °C) for one hour. Afterward, the pH of the suspensions was raised to about pH 5 with 2 M KOH, and 358 μL of a 0.01 M Na2HAsO4 stock solution were then spiked to the suspensions, yielding target As concentrations of 1343 μg/g HS and final (Fe/As)spiked molar ratios of 10. The pH was then adjusted with base to between 6

ARTICLE

and 7, and the final volume was adjusted to 5.0 mL with ultrapure water (Milli-Q, Millipore, 18.2 MΩ cm). The suspensions were then reacted in the dark on a rotary shaker with occasional pH adjustments to pH 7.0 using KOH or HNO3 of varying strengths. The final ionic strength was approximately 0.1 M in both treatments. After two days, the pH stabilized at pH 7.0 ( 0.1, and after three days the suspensions were directly injected into liquid N2 and lyophilized at 30 °C (0.37 mbar). The freezedried materials were homogenized in an agate mortar and filled into Al sample holders, which were sealed with Kapton tape. X-ray Absorption Spectroscopy. The samples were analyzed by X-ray absorption spectroscopy at the bending magnet beamline BM23 at the European Synchrotron Radiation Facility (ESRF, Grenoble, France). X-ray absorption near edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) spectra of Fe and As were collected in fluorescence mode using a 13-element Ge detector. The white beam was monochromatized with a Si(111) monochromator (2.0 10 4 ΔE/E), and Si mirrors were used to eliminate higher harmonics in the beam. Prior to mounting the samples to the sample holder stage, the samples were shock-frozen in liquid N2. During the measurements the samples were kept at about 80 K with a cryostream. The monochromator was calibrated with an Fe foil (K-edge: 7112 eV) for the Fe measurements and an Au foil (L3edge: 11,919 eV) for the As measurements. We constantly monitored the spectra of either foil in transmission mode in order to account for small energy shifts (<2 eV) during the measurements. The vertical exit-slit size was fixed at 1 mm, and the horizontal slit size was adjusted to between one and 2.5 mm to ensure a proper counting statistics. The pre-edge region was scanned with 5-eV steps. Along the edge the step size was reduced to 0.3 eV. For both elements the EXAFS was collected with Δk = 0.04 Å 1 up to 14 Å 1. Up to 11 scans were recorded for each individual sample. Data Analysis. Spectral data processing was performed in Athena.35 The Autobk algorithm was applied for background removal using a linear pre-edge line between 200 and 80 eV before the edge step, E0, and a normalization range from 150 to 740 eV (Fe) or 760 eV (As). By default a quadratic polynomial was used as post-edge line. The frequency cutoff parameter, Rbkg, was set to 0.9 for Fe and 0.8 for As. The k-weight in the background function determination was set to three. Fourier transforms of the normalized k3-weighted χ(k) data were calculated over a k-range of 1 12 Å 1 (Fe) or 3 12 Å 1 (As) using a Kaiser-Bessel apodization window with a window parameter of 3 Å 1. Least-squares fitting of k3-weighted EXAFS spectra was performed in R-space (Fe) and q-space (As) using the software Artemis.35 The fit k-weights were set to three. Theoretical phaseshift and amplitude functions were calculated with the planewave formalism using the ab initio FEFF 8.4 code.36 All scattering paths used to model the As K-edge EXAFS spectra were extracted from the structure of mapimite.37 For Fe K-edge EXAFS spectra, we used the structures of goethite (α-FeOOH)38 and an Fe(III)-acetate (‘complex B’)39 as FEFF input in order to calculate all relevant single and multiple scattering paths. Wavelet-transform analysis was used to qualitatively identify contributions of high and low Z backscattering atoms in the Fourier-transform peaks occurring between 2.2 and 4.0 Å R + ΔR (Fe) or 2.0 and 3.4 Å R + ΔR (As). We utilized the Fortran-based HAMA code to calculate the Morlet wavelet transforms of k1weighted EXAFS spectra.40 For appropriate resolution in R- and k-space, we found a wavelet parameter combination of k = 4 and 9551

dx.doi.org/10.1021/es202300w |Environ. Sci. Technol. 2011, 45, 9550–9557

Environmental Science & Technology

ARTICLE

Figure 2. Morlet wavelet transforms of k1-weighted Fe (top) and As (bottom) K-edge EXAFS spectra of reference compounds and Fe(III)-HS complexes reacted with As(V). The wavelet transforms were calculated over a R + ΔR-range of 2.2 4.0 Å (k = 4 and σ = 1.5) for Fe and over a R + ΔRrange of 2.0 3.4 Å (k = 7 and σ = 2) for As. (a) Ferrihydrite, (b) X-ray amorphous FeAsO4, (c) Fe(III)-EDTA, (d) As(V) + Fe(III)-fulvate, (e) As(V) + Fe(III)-humate, (f) As(V) adsorbed to ferrihydrite, (g) X-ray amorphous FeAsO4, (h) Na2HAsO4•7H2O, (i) As(V) + Fe(III)-fulvate, and (j) As(V) + Fe(III)-humate. The ﬁrst modulus in the wavelets of Fe at k ∼3 Å 1 is due to signal leakage from the ﬁrst O shell, multiple scattering (MS) within the Fe(O,OH)6 octahedron, and/or backscattering from low Z atoms of an organic ligand (c-e). For As the ﬁrst modulus at k ∼2 3 Å 1 originates from the ﬁrst O shell and MS within the AsO4 tetrahedron.

σ = 1.5 well suited to discriminate atomic contributions in the Fourier transforms of Fe K-edge EXAFS spectra. In contrast, a combination of k = 7 and σ = 2 was employed for the As K-edge EXAFS spectra. All spectra analyzed by wavelet transformation were truncated at 13 Å 1.

’ RESULTS AND DISCUSSION Wavelet-Transform Analysis. We used the wavelet-transform analysis to qualitatively test for the presence of Fe backscatterers in the second coordination shell of both Fe and As. Therefore, we first calculated the Morlet wavelet transforms of k1-weighted Fe K-edge EXAFS spectra of the HS samples over a R + ΔR-range of 2.2 4.0 Å and compared the resulting wavelet plots with those of Fe(III) reference compounds (ferrihydrite, X-ray amorphous FeAsO4, Fe-EDTA) in Figure 2. The second coordination shell of Fe in the crystal structure of ferrihydrite and FeAsO4 comprises Fe atoms; consequently, these phases showed an Fe backscattering signal in the wavelet plots at k ∼6.5 Å 1 (Figure 2a, b). This signal was absent for Fe(III)-EDTA (Figure 2c) because this ligand forms a strong hexadentate mononuclear complex with Fe(III). In contrast to Fe-EDTA, Fe(III)-HS complexes exhibited a pronounced wavelet modulus at k ∼6.5 Å 1 (Figure 2d, e), signifying the presence of secondshell Fe in these samples. A backscattering signal of As occurring at k ∼9.5 Å 1 was barely visible in the wavelet-transform plots of the HS samples. However, this cannot reasonably be expected owing to their high molar Fe/As ratio of approximately 10. Even the FeAsO4 reference with its molar Fe/As ratio of 1.0 as determined by X-ray fluorescence spectrometry41 yielded a comparably weak As backscattering signal at k ∼9.5 Å 1 (Figure 2b).

The Morlet wavelet transforms of k1-weighted As K-edge EXAFS spectra of the Fe(III)-HS complexes reacted with As(V) are also shown in Figure 2 in addition to those of As model compounds, including As(V) adsorbed to ferrihydrite, X-ray amorphous FeAsO4, and Na2HAsO4•7H2O. The latter compound was included in the analysis because it (i) has no Fe in its crystal structure and (ii) may have formed upon lyophilization of the HS samples. The wavelet transforms were calculated over an R + ΔR-range of 2.0 3.4 Å. Reference compounds in which As has second-shell Fe neighbors (As(V) adsorbed to ferrihydrite and FeAsO4) showed a pronounced backscattering signal of Fe at k ∼ 6.5 Å 1, which was lacking for the Na2HAsO4•7H2O reference (Figure 2f-h). The wavelet-transformed As K-edge EXAFS spectra of the HS samples (Figure 2i, j) possessed Fe backscattering signals of similar shape but weaker intensity compared to FeAsO4 (Figure 2g), thus corroborating the existence of second-shell Fe neighbors of As in As(V)-reacted Fe(III)-HS complexes. The wavelet-transform results thus document that Fe(III) in the HS samples was at least partially hydrolyzed and that As(V) was coordinated to Fe atoms. In the next sections, we discuss the likely coordination environments of Fe and As revealed by EXAFS shell-ﬁt analysis. Shell-Fit Analysis of Fe K-Edge EXAFS Spectra. The Fe K-edge XANES spectra of our HS samples exhibited a maximum in their first derivatives at ∼7127 eV, which is consistent with Fe(III) reference compounds (Figure S1, SI). Thus, no reduction of Fe(III) in the HS suspensions was observed. Figure 3 shows the Fe K-edge EXAFS spectra and corresponding Fourier transforms of As(V) Fe(III) HS mixtures and reference compounds. Ferrihydrite showed a strong Fourier-transform signal at ∼2.6 Å, corresponding to edge-sharing Fe(O,OH)6 linkages. 9552

dx.doi.org/10.1021/es202300w |Environ. Sci. Technol. 2011, 45, 9550–9557

Environmental Science & Technology

ARTICLE

Figure 3. Iron K-edge EXAFS spectra (A) as well as magnitude (B) and real part (C) of the Fourier transform of reference compounds and Fe(III)-HS complexes reacted with As(V). Solid lines represent experimental data and shell ﬁts are shown as dotted lines. Dashed vertical lines indicate the position of Fe in edge- (E) and double corner-sharing (DC) Fe(O,OH)6 octahedra in the structure of ferrihydrite. (FA = fulvate, HA = humate).

This peak was neither observed for the mononuclear Fe(III)EDTA complex nor for the X-ray amorphous FeAsO4 (Figure 3). The absence of the ∼2.6-Å peak for the latter phase accords with its proposed structure consisting of chains of corner-sharing Fe(O,OH)6 octahedra with bridging AsO4 tetrahedra.42 Similar to Fe(III)-EDTA and FeAsO4, no distinct Fourier-transform peaks at ∼2.6 Å were visible for the HS samples, suggesting only minor contributions of edge-sharing Fe(O,OH)6 linkages typical of Fe(III) oxyhydroxides (Figure 3). As a consequence, we tested whether a structural model of oligomeric Fe clusters devoid of edgesharing octahedral linkages conforms to the recorded Fe EXAFS spectra. To this end, we resorted to a structural motif that was identified for a trinuclear Fe(III)-acetate complex, [Fe3O(CH3COO)6(H2O)3][AuCl4]•6H2O.39 This complex contains a planar [Fe3O] core, in which the central μ3-O bridge links three Fe(O,OH)6 octahedra arranged in a regular triangle (Figure 1c, d). Shell-ﬁts were performed on the Fourier-transformed k3-weighted Fe K-edge EXAFS spectra (k-range: 1 12 Å 1, ﬁt k-weight = 3). Our ﬁt strategy involved a ﬁrst ﬁt in R-space over a distance of 0.9 2.0 Å, which accommodates the ﬁrst O coordination shell of Fe. Fit parameters included the energy-shift parameter, ΔE0, the coordination number of O, the mean Fe O half path length as well as the Debye Waller parameter. The passive amplitude reduction factor, S02, was ﬁxed to 0.71. This value was used by Mikutta (2011)43 for mixtures of ferrihydrite and organic Fe(III) complexes and is close to the 0.78 used by Karlsson et al.44 Thus, the total number of ﬁt variables was four, and the number of independent data points, Nind, calculated by

the Nyquist theorem was 7 (Nind = 2 Δk ΔR/π, where Δk is the k-range of the EXAFS spectrum and ΔR is the R-range in the Fourier transform). In a second step, we ﬁxed the determined EXAFS parameters and extended the R-window to 4.0 Å. Peaks of the Fourier transform occurring at ΔR + R > 2.0 Å (Figure 3) were then accounted for by an Fe C single scattering (SS) path, an Fe Fe SS path, and multiple scattering (MS) within the Fe(O,OH)6 octahedron. According to our structural motif, we ﬁxed the coordination number of the Fe Fe path to two and only adjusted its mean half path length and Debye Waller parameter. Tests revealed that only the four-legged collinear Fe O Fe-O MS path was necessary in order to properly ﬁt the data. The coordination number of this path was constrained to its theoretical value, CNMS = CNFe O (Table 1). The second ﬁt thus contained seven ﬁt variables and 21 independent data points. The ﬁnal model ﬁts are illustrated in Figure 3, and the respective EXAFS parameters are summarized in Table 1. Evidently, all features in the Fe K-edge EXAFS and the Fourier transforms are adequately described by the model (Figure 3). Oxygen coordination numbers close to six and average Fe O bond distances of 1.99 Å are in agreement with those determined for Fe(III) complexes of humic acid.34 Likewise, Fe C interactions in the second scattering shell at ∼2.95 Å are in the range of reported values for Fe(III) complexes of humic acid and Fe(III) in organic soils.34,45 The Fe C bond distances, however, were considerably longer (∼0.1 Å) than those determined for ligands forming mononuclear 5-membered chelate ring structures, for example, ferrioxamine B and trisoxalatoiron.46 48 Iron-C bond distances of less than 2.9 Å, corresponding to 5-membered chelate ring 9553

dx.doi.org/10.1021/es202300w |Environ. Sci. Technol. 2011, 45, 9550–9557

Environmental Science & Technology

ARTICLE

Table 1. Iron and As K-Edge EXAFS Shell-Fit Results for Fe(III)-HS Complexes Reacted with As(V)g sample

path

CN (-)a

R (Å)b

σ2 (Å2)c

ΔE (eV)d

R-factor (-)e

3.6 (5)

0.004

4.1 (5)

0.004

4.7 (12)

0.007

4.6 (11)

0.006

Iron FA

HA

Fe O

5.8 (2)

1.99 (0)

0.007 (0)

Fe C

2.7 (9)

2.96 (2)

0.010 (5)

Fe Fe

2

3.47 (1)

0.015 (1)

Fe O Fe O (collinear)

5.8

4.45 (2)

0.007 (3)

Fe O

5.8 (2)

1.99 (0)

0.006 (0)

Fe C

2.5 (10)

2.94 (2)

0.012 (6)

Fe Fe

2

3.48 (1)

0.015 (2)

Fe O Fe O (collinear)

5.8

4.46 (2)

0.005 (3)

As O

4

1.70 (0)

0.003 (0)

As Fe

0.7 (1)

3.29 (1)

0.007f

Arsenic FA

HA

Fe O O (triangular)

12

3.08

0.003

As O As O (collinear)

4

3.40

0.012

As O As O (noncollinear)

12

3.40

0.006

As O

4

1.70 (0)

0.003 (0)

As Fe Fe O O (triangular)

0.5 (1) 12

3.30 (1) 3.09

0.003f 0.003

As O As O (collinear)

4

3.40

0.011

As O As O (noncollinear)

12

3.40

0.006

a Coordination number. b Mean half path length. c Debye Waller parameter. d Energy-shift parameter. e R-factor = ∑i(datai-ﬁti)2/∑i datai2. f Fit uncertainty was estimated at 0.001. g Values in bold were ﬁxed in the ﬁt. Parameters in italics were deﬁned in terms of single scattering paths and subsequently covaried in the ﬁts. Fit uncertainties in parentheses are given for the last signiﬁcant ﬁgure. The passive amplitude reduction factor, S02, was ﬁxed to 0.71 for Fe and 1.00 for As. (FA = fulvate, HA = humate).

structures, have also been reported for humic acid (pH 3.0 7.2) with Fe loadings of less than 10,000 μg/g.34 By contrast, Fe C bond distances of ∼2.95 Å imply predominantly 6-membered chelate ring structures in our HS samples and suggest a coordination change from predominantly 5- to 6-membered chelate ring structures with increasing Fe loading. A similar change in the binding mode from ﬁve- to six-O-ring chelates with increasing metal loading of NOM has recently been described for Cu(II).49 The ﬁtted Fe Fe bond distance of approximately 3.47 Å accords well with corner-sharing Fe(O,OH)6 octahedra and is consistent with the ﬁtted average Fe O bond distances and an Fe O(μ3)Fe bond angle of 120°. Our shell-ﬁt results thus corroborate ﬁndings of Vilge-Ritter et al.,50 who concluded based on X-ray absorption spectroscopy and small-angle X-ray scattering that single-corner sharing Fe(O,OH)6 trimers are the dominant Fe species in freshwater NOM. Our results also agree with those of Gustafsson et al.,45 who proposed a mixture of Fe(III)-NOM complexes with [(O5Fe)2O] and [(O5Fe)3O] inner cores (pH 2.6 3.9). The general compliance of our EXAFS data with the trimeric Fe model thus indicates that the majority of Fe in Fe(III)-HS complexes is hydrolyzed at neutral pH and exists in the form of small oligomeric Fe(III) species. However, a detailed wavelet-transform analysis (Figure S4, SI) also revealed contributions of edge-sharing Fe(O,OH)6 linkages in the Fe K-edge EXAFS of our samples, which could arise from nanometer-sized Fe(III) oxyhydroxides, particularly ferrihydrite. In order to quantify the amount of ferrihydrite, we included a short Fe Fe SS path in our ﬁt scheme. The Debye Waller parameter of this path was ﬁxed to 0.013 Å2, a typical value for ferrihydrite.43,51 The resulting shell ﬁts are detailed in the SI. In brief, we found that the implementation of a short Fe Fe SS path did only

marginally reduce the misﬁt between the model and the data (Figure S5 and S6, Table S2, SI). Yet, ﬁtted CNs of the short Fe Fe SS path were 0.2 on average and may thus suggest about 6 mol % of colloidal Fe in the form of ferrihydrite. Shell-Fit Analysis of As K-Edge EXAFS Spectra. The As Ke-edge XANES spectra of the HS samples exhibited a maximum in their first derivatives at ∼11,873 eV, which is consistent with As(V) reference compounds (Figure S2, SI). Thus, reduction of As(V) to As(III) in the HS suspensions or during sample analysis was not observed. Figure 4 shows the As K-edge EXAFS spectra and Fourier transforms of Fe(III)-HS complexes reacted with As(V) in concert with those of As(V) model compounds. In contrast to the Na2HAsO4•7H2O reference, the HS samples exhibited Fourier-transform peaks at a radial distance R + ΔR of ∼2.8 Å. These peaks coincide with those typical of As(V) references that contain Fe atoms in the second coordination shell of As, for example, As(V) adsorbed to ferrihydrite and FeAsO4 (Figure 4). A qualitative spectrum comparison thus already confirms Fe in the second coordination shell of As, in agreement with our wavelet analysis (Figure 2). Shell-ﬁts were performed on the Fourier-ﬁltered k3-weighted As K-edge EXAFS spectra (k-range: 3 12 Å 1, R + ΔR-range: 0.8 3.4 Å, ﬁt k-weight = 3, 14.5 independent data points). Our ﬁt approach was identical to that used by Mikutta et al.41 Two SS paths were included: As O and As Fe. The coordination number of the As O path was set to four, and the Debye Waller parameter of the As Fe path was ﬁxed to a predetermined value (see the SI). Multiple scattering was considered by a triangular As O O, a collinear As O As O, and a noncollinear As O As O MS path. Implementation of MS did not increase the number of variables in the ﬁt because the MS path 9554

dx.doi.org/10.1021/es202300w |Environ. Sci. Technol. 2011, 45, 9550–9557

Environmental Science & Technology

ARTICLE

Figure 4. Arsenic K-edge EXAFS spectra (A) as well as magnitude (B) and real part (C) of the Fourier transform of reference compounds and Fe(III)HS complexes reacted with As(V). Solid lines represent experimental data and shell ﬁts are shown as dotted lines. The dashed vertical lines mark the position of the As Fe single scattering (SS) path at R + ΔR ∼2.8 Å. (FA = fulvate, HA = humate).

degeneracies were ﬁxed to their theoretical values, and the Debye Waller parameters as well as mean half path lengths were related to the respective value of the As O SS path. The triangular MS path was constrained by a path degeneracy of 12, a mean half path length of 1.8165 (=1 + (2/3)(1/2)) RAs O, and a Debye Waller parameter of σ2As O. The degeneracy of the collinear MS path was set to four, its mean half path length to 2 RAs O, and its Debye Waller parameter was calculated as σ2 = 4 σ2As O. The degeneracy of the noncollinear MS path was 12, its mean half path length 2 RAs O, and its Debye Waller parameter was deﬁned as 2 σ2As O. The model thus contained ﬁve variables: the mean half path lengths and of the two SS paths, the Debye Waller parameter of the As O path, the degeneracy of the As Fe path, and the energy-shift parameter, ΔE0. The passive amplitude reduction parameter, S02, was set to one because it was ﬁtted to 0.96, which was not statistically diﬀerent from unity. All EXAFS parameters determined for As accord with previous studies on As(V) sorption to Fe(III) oxyhydroxides: Four O were found at a mean distance of 1.70 Å with an average Debye Waller parameter of 0.003 Å2, which were followed by 0.5 0.7 Fe atoms at ∼3.30 Å (Table 1). The observed As Fe bond distance is commonly explained by a monodentate binuclear (2C) As(V) complex in which As(V) shares two O with two edge-sharing Fe(O,OH)6 octahedra.1,3,52,53 Since the predominance of such octahedral linkages was ruled out, As(V) is likely bridging two corner-sharing Fe(O,OH)6 octahedra. Classical monodentate mononuclear (1V) and bidentate mononuclear (1E) complexes of As(V) as those depicted in Figure 1a and 1b can be excluded because the As Fe distances of these complexes would be considerably longer (>3.5 Å for 1V) or shorter (<2.9 Å for 1E) than those determined for a 2C complex.1,53 The fact that the coordination number of second-shell Fe is less than unity

may suggest that a portion of As(V) is adsorbed in a mononuclear 1V complex stabilized by a H-bond to a singly coordinated OH group of an adjacent Fe(O,OH)6 octahedron as was proposed for As(V) adsorption to goethite.54 Other reasons for the low Fe coordination numbers could be sodium arsenate precipitation upon lyophilization, outer-sphere complexation of As(V) to oligomeric Fe clusters4 and/or protonated amino groups of HS molecules.9 Depending on the underlying As(V) binding geometry (mono- vs binuclear), the ﬁtted Fe coordination numbers imply that 4.5 12.5 μmol As(V)/g HS (25 70% of total As) was associated with Fe. When the ﬁtted CNs of the short Fe Fe SS path (Table S2, SI) are taken into account, we conservatively estimate that 70 90% of As(V) associated with Fe was bound to oligomeric Fe(III)-HS complexes rather than ferrihydrite. For this assessment we assumed (i) 6 mol % ferrihydrite in the samples, (ii) a ratio of 0.25 between surface and bulk Fe atoms in ferrihydrite,55 and (iii) a complete surface saturation of ferrihydrite with As(V) bound in a binuclear complex. Thus, based on the wavelet-transform analyses and shell-ﬁts of Fe and As K-edge EXAFS spectra presented here, we conclude that at neutral pH As(V) forms an inner-sphere complex with oligomeric Fe(III) clusters, in which octahedral corner linkages predominate. The binding geometries of As(V) that accord with our data are illustrated in Figure 1c and 1d. Environmental Implications. Our results confirm the assumption commonly made in the literature that As(V) is bound to NOM by formation of ternary complexes with Fe(III), in which Fe(III) bridges between organic functional groups and As(V). However, the ‘bridging Fe(III)’ exists as poly- rather than mononuclear Fe(III) species at neutral pH and a moderate Fe loading of organic matter (1 wt %). We believe that this inner-sphere binding of As(V) to Fe(III)-NOM complexes has 9555

dx.doi.org/10.1021/es202300w |Environ. Sci. Technol. 2011, 45, 9550–9557

Environmental Science & Technology wide-ranging environmental implications as both NOM and Fe are ubiquitous in surface-near environments including soils, wetlands, and aquatic systems. In these environments NOM could serve as important sorbent for As(V) under oxic conditions, potentially lowering the mobility and availability of As for uptake by plants and microorganisms. Under favorable geochemical conditions, however, Fe(III)-NOM complexes may also become mobilized and transported in dissolved or colloidal forms.56,57 Dissolved or colloidal Fe(III)-NOM complexes may then become an effective mobile carrier for As(V), thus increasing As mobility as was recently demonstrated in column experiments.26 Ternary complex formation may also increase the availability of As(V) to metal-reducing bacteria. Under anoxic conditions, As(V) in ternary complexes is probably more prone to microbial reduction than As(V) adsorbed to or coprecipitated with metal oxyhydroxides. A preferred utilization of Fe(III) in ternary complexes as electron acceptor for dissimilatory iron-reducing bacteria would also facilitate the rapid release and subsequent reduction of As(V). This may have signiﬁcant implications for Fe and As cycling following soil ﬂooding, e.g., in mining-aﬀected NOM-rich river ﬂoodplains or As-contaminated rice paddy soils in Southeast Asia. However, these eﬀects have not yet been investigated. In the current literature, the role of NOM in controlling As(V) mobility in soils and aquifers is mostly discussed in terms of competitive sorption to metal oxyhydroxides and of NOM oxidation driving the microbial reduction of Fe(III) and As(V) under anoxic conditions.23,58 While these As-mobilizing processes are of undisputable importance, we propose that As binding to NOM is another important process aﬀecting As mobility in NOM-rich environments, which has been underestimated in the past.

’ ASSOCIATED CONTENT

bS

Supporting Information. Properties of the humic substances used, ﬁrst derivatives of normalized Fe and As K-edge XANES spectra, determination of the Debye Waller parameter of the As Fe single scattering paths, test for edge-sharing octahedral Fe linkages, and additional shell ﬁts of Fe K-edge EXAFS spectra accounting for a short Fe Fe single scattering path. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: +41-44-6336024. Fax: +41-44-6331118. E-mail: christian. [email protected].

’ ACKNOWLEDGMENT We would like to thank K. Barmettler for his support in the soil chemistry laboratory and M. Hoﬀmann for his help during the synchrotron measurements. The XAS experiments were performed on the BM23 beamline at the European Synchrotron Radiation Facility (ESRF), Grenoble, France. We are grateful to M. Chorro at ESRF for providing assistance in using beamline BM23. ’ REFERENCES (1) 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.

ARTICLE

(2) 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. (3) Fendorf, S.; Eick, M. J.; Grossl, P.; Sparks, D. L. Arsenate and chromate retention mechanisms on goethite. 1. Surface structure. Environ. Sci. Technol. 1997, 31, 315–320. (4) Catalano, J. G.; Park, C.; Fenter, P.; Zhang, Z. Simultaneous inner- and outer-sphere arsenate adsorption on corundum and hematite. Geochim. Cosmochim. Acta 2008, 72, 1986–2004. (5) Cullen, W. R.; Reimer, K. J. Arsenic speciation in the environment. Chem. Rev. 1989, 89, 713–764. (6) Smith, E.; Naidu, R.; Alston, A. M. Arsenic in the soil environment: A review. Adv. Agron. 1998, 64, 149–195. (7) Smedley, P. L.; Kinniburgh, D. G. A review of the source, behaviour and distribution of arsenic in natural waters. Appl. Geochem. 2002, 17, 517–568. (8) Wang, S. L.; Mulligan, C. N. Eﬀect of natural organic matter on arsenic release from soils and sediments into groundwater. Environ. Geochem. Health 2006, 28, 197–214. (9) Thanabalasingam, P.; Pickering, W. F. Arsenic sorption to humic acids. Environ. Pollut. 1986, 12, 233–246. (10) Redman, A. D.; Macalady, D. L.; Ahmann, D. Natural organic matter aﬀects arsenic speciation and sorption onto hematite. Environ. Sci. Technol. 2002, 36, 2889–2896. (11) Ko, I.; Kim, J.-Y.; Kim, K.-W. Arsenic speciation and sorption kinetics in the As-hematite-humic acid system. Colloids Surf., A 2004, 234, 43–50. (12) Lin, H.-T.; Wang, M. C.; Li, G.-C. Complexation of arsenate with humic substance in water extract of compost. Chemosphere 2004, 56, 1105–1112. (13) Buschmann, J.; Kappeler, A.; Lindauer, U.; Kistler, D.; Berg, M.; Sigg, L. Arsenite and arsenate binding to dissolved humic acids: Inﬂuence of pH, type of humic acid, and aluminum. Environ. Sci. Technol. 2006, 40, 6015–6020. (14) Chen, Z.; Cai, Y.; Solo-Gabriele, H.; Snyder, G. H.; Cisar, J. L. Interactions of arsenic and the dissolved substances derived from turf soils. Environ. Sci. Technol. 2006, 40, 4659–4665. (15) Ritter, K.; Aiken, G. R.; Ranville, J. F.; Bauer, M.; Macalady, D. L. Evidence for the aquatic binding of arsenate by natural organic matter-suspended Fe(III). Environ. Sci. Technol. 2006, 40, 5380– 5387. (16) Bauer, M.; Fulda, B.; Blodau, C. Groundwater derived arsenic in high carbonate wetland soils: Sources, sinks, and mobility. Sci. Total Environ. 2008, 401, 109–120. (17) Liu, G.; Cai, Y. Complexation of arsenite with dissolved organic matter: Conditional distribution coeﬃcients and apparent stability constants. Chemosphere 2010, 81, 890–896. (18) Gonzalez A., Z. I.; Krachler, M.; Cheburkin, A. K.; Shotyk, W. Spatial distribution of natural enrichments of arsenic, selenium, and uranium in a minerotrophic peatland, Gola di Lago, Canton Ticino, Switzerland. Environ. Sci. Technol. 2006, 40, 6568–6574. (19) Dobran, S.; Zagury, G. J. Arsenic speciation and mobilization in CCA-contaminated soils: Inﬂuence of organic matter content. Sci. Total Environ. 2006, 364, 239–250. (20) Rothwell, J. J.; Taylor, K. G.; Ander, E. L.; Evans, M. G.; Daniels, S. M.; Allott, T. E. H. Arsenic retention and release in ombrotrophic peatlands. Sci. Total Environ. 2009, 407, 1405–1417. (21) Warwick, P.; Inam, E.; Evans, N. Arsenic’s interaction with humic acid. Environ. Chem. 2005, 2, 119–124. (22) de Mello, J. W. V.; Talbott, J. L.; Scott, J.; Roy, W. R.; Stucki, J. W. Arsenic speciation in arsenic-rich Brazilian soils from gold mining sites under anaerobic incubation. Environ. Sci. Pollut. Res. 2007, 14, 388–396. (23) Bauer, M.; Blodau, C. Arsenic distribution in the dissolved, colloidal and particulate size fraction of experimental solutions rich in dissolved organic matter and ferric iron. Geochim. Cosmochim. Acta 2009, 73, 529–542. (24) Wang, S.; Mulligan, C. N. Enhanced mobilization of arsenic and heavy metals from mine tailings by humic acid. Chemosphere 2009, 74, 274–279. 9556

dx.doi.org/10.1021/es202300w |Environ. Sci. Technol. 2011, 45, 9550–9557

Environmental Science & Technology (25) Sharma, P.; Ofner, J.; Kappler, A. Formation of binary and rernary colloids and dissolved complexes of organic matter, Fe and As. Environ. Sci. Technol. 2010, 44, 4479–4485. (26) Sharma, P.; Rolle, M.; Kocar, B.; Fendorf, S.; Kappler, A. Inﬂuence of natural organic matter on As transport and retention. Environ. Sci. Technol. 2011, 45, 546–553. (27) Liu, G.; Fernandez, A.; Cai, Y. Complexation of arsenite with humic acid in the presence of ferric iron. Environ. Sci. Technol. 2011, 45, 3210–3216. (28) Bloom, P. R. Phosphorus adsorption by an aluminum-peat complex. Soil Sci. Soc. Am. J. 1981, 45, 267–272. (29) Gerke, J.; Hermann, R. Adsorption of orthophosphate to humic-Fe-complexes and to amorphous Fe-oxide. Z. Pﬂanzenern€ahr. Bodenkd. 1992, 155, 233–236. (30) Hutchison, K. J.; Hesterberg, D. Dissolution of phosphate in a phosphorus-enriched ultisol as aﬀected by microbial reduction. J. Environ. Qual. 2004, 33, 1793–1802. (31) Guardado, I.; Urrutia, O.; Garcia-Mina, J. M. Some structural and electronic features of the interaction of phosphate with metal-humic complexes. J. Agric. Food. Chem. 2008, 56, 1035–1042. (32) Gerke, J. Humic (organic matter)-Al(Fe)-phosphate complexes: An underestimated phosphate form in soils and source of plant-available phosphate. Soil Sci. 2010, 175, 417–425. (33) Gustafsson, J. P.; Johnsson, L. The association between selenium and humic substances in forested ecosystems - laboratory evidence. Appl. Organomet. Chem. 1994, 8, 141–147. (34) Karlsson, T.; Persson, P. Coordination chemistry and hydrolysis of Fe(III) in a peat humic acid studied by X-ray absorption spectroscopy. Geochim. Cosmochim. Acta 2010, 74, 30–40. (35) Ravel, B.; Newville, M. ATHENA, ARTEMIS, HEPHAESTUS: data analysis for X-ray absorption spectroscopy using IFEFFIT. J. Synchrotron Rad. 2005, 12, 537–541. (36) Ankudinov, A. L.; Ravel, B.; Rehr, J. J.; Conradson, S. D. Realspace multiple-scattering calculation and interpretation of x-ray-absorption near-edge structure. Phys. Rev. B 1998, 58, 7565–7576. (37) Ginderow, D.; Cesbron, F. Structure de la mapimite, Zn2Fe3(AsO4)3(OH)4.10H2O. Acta Crystallogr. 1981, 37, 1040–1043. (38) Kaur, N.; Singh, B.; Kennedy, B. J.; Graefe, M. The preparation and characterization of vanadium-substituted goethite: The importance of temperature. Geochim. Cosmochim. Acta 2009, 73, 582–593. (39) Turte, K. I.; Shova, S. G.; Meriacre, V. M.; Gdaniec, M.; Simonov, Y. A.; Lipkowski, J.; Bartolome, J.; Wagner, F.; Filoti, G. Synthesis and structure of trinuclear iron acetate [Fe3O(CH3COO)6(H2O)3] [AuCl4].6H2O. J. Struct. Chem. 2002, 43, 108–117. (40) Funke, H.; Scheinost, A. C.; Chukalina, M. Wavelet analysis of extended x-ray absorption ﬁne structure data. Phys. Rev. B 2005, 71, 94110–1 7. (41) 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. (42) Paktunc, D.; Dutrizac, J.; Gertsman, V. Synthesis and phase transformations involving scorodite, ferric arsenate and arsenical ferrihydrite: Implications for arsenic mobility. Geochim. Cosmochim. Acta 2008, 72, 2649–2672. (43) Mikutta, C. X-ray absorption spectroscopy study on the eﬀect of hydroxybenzoic acids on the formation and structure of ferrihydrite. Geochim. Cosmochim. Acta 2011, 75, 5122–5139. (44) Karlsson, T.; Persson, P.; Skyllberg, U.; M€orth, C.-M.; Giesler, R. Characterization of iron(III) in organic soils using extended X-ray absorption ﬁne structure spectroscopy. Environ. Sci. Technol. 2008, 42, 5449–5454. (45) Gustafsson, J. P.; Persson, I.; Kleja, D. B.; Van Schaik, J. W. J. Binding of iron(III) to organic soils: EXAFS spectroscopy and chemical equilibrium modeling. Environ. Sci. Technol. 2007, 41, 1232–1237. (46) Persson, P.; Axe, K. Adsorption of oxalate and malonate at the water-goethite interface: Molecular surface speciation from IR spectroscopy. Geochim. Cosmochim. Acta 2005, 69 (3), 541–552.

ARTICLE

(47) Edwards, D. C.; Myneni, S. C. B. Hard and soft X-ray absorption spectroscopic investigation of aqueous Fe(III)-hydroxamate siderophore complexes. J. Phys. Chem. A 2005, 109, 10249–10256. (48) Duckworth, O. W.; Bargar, J. R.; Sposito, G. Quantitatives structure-activity relationships for aqueous metal-siderophore complexes. Environ. Sci. Technol. 2009, 43, 343–349. (49) Manceau, A.; Matynia, A. The nature of Cu bonding to natural organic matter. Geochim. Cosmochim. Acta 2010, 74, 2556–2580. (50) Vilge-Ritter, A.; Rose, J.; Masion, A.; Bottero, J.-Y.; Laine, J.-M. Chemistry and structure of aggregates formed with Fe-salts and natural organic matter. Colloids, Surf. A 1999, 147, 297–308. (51) Manceau, A. Critical evaluation of the revised akdalaite model for ferrihydrite. Am. Mineral. 2011, 96, 521–533. (52) Waychunas, G. A.; Rea, B. A.; Fuller, C. C.; Davis, J. A. Surface chemistry o ferrihydrite: Part 1. EXAFS Studies of the geometry of coprecipitated and adsorbed arsenate. Geochim. Cosmochim. Acta 1993, 57, 2251–2269. (53) Manceau, A. The mechanism of anion adsorption on iron oxides: Evidence for the bonding of arsenate tetrahedra on free Fe(O, OH)6 edges. Geochim. Cosmochim. Acta 1995, 59, 3647–3653. (54) Loring, J. S.; Sandstr€om, M. H.; Noren, K.; Persson, P. Rethinking arsenate coordination at the surface of goethite. Chem.— Eur. J. 2009, 15, 5063–5072. (55) 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. (56) Batchelli, S.; Muller, F. L. L.; Chang, K.-C.; Lee, C.-L. Evidence for strong but dynamic iron-humic colloidal associations in humic-rich coastal waters. Environ. Sci. Technol. 2010, 44, 8485–8490. (57) Krachler, R.; Krachler, R. F.; von der Kammer, F.; S€uphandag, A.; Jirsa, F.; Ayromlou, S.; Hofmann, T.; Keppler, B. K. Relevance of peat-draining rivers for the riverine input of dissolved iron into the ocean. Sci. Total Environ. 2010, 408, 2402–2408. (58) McArthur, J. M.; Banerjee, D. M.; Hudson-Edwards, K. A.; Mishra, R.; Purohit, R.; Ravenscroft, P.; Cronin, A.; Howarth, R. J.; Chatterjee, A.; Talukder, T.; Lowry, D.; Houghton, S.; Chadha, D. K. Natural organic matter in sedimentary basins and its relation to arsenic in anoxic ground water: the example of West Bengal and its worldwide implications. Appl. Geochem. 2004, 19, 1255–1293.

’ NOTE ADDED AFTER ASAP PUBLICATION There was an error in the Abstract in the version published ASAP on October 24, 2011. The corrected version was published on November 11, 2011.

9557

dx.doi.org/10.1021/es202300w |Environ. Sci. Technol. 2011, 45, 9550–9557

ARTICLE pubs.acs.org/est

Probabilistic Multicompartmental Model for Interpreting DGT Kinetics in Sediments P. Ciﬀroy,*,† Y. Nia,† and J. M. Garnier‡ †

EDF, Division Recherche et Developpement, Departement Laboratoire National d’Hydraulique et Environnement, 78401 Chatou, France ‡ CEREGE, BP80, Universite Paul Cezanne, 13545 Aix en Provence, France ABSTRACT: Extensive research has been performed on the use of the DIFS (DGT-Induced Fluxes in Soils and Sediments) model to interpret diﬀusive gradients in thin-ﬁlm, or DGT, measurements in soils and sediments. The current report identiﬁes some areas where the DIFS model has been shown to yield poor results and proposes a model to address weaknesses. In particular, two major ﬂaws in the current approaches are considered: (i) many studies of accumulation kinetics in DGT exhibit multiple kinetic stages and (ii) several combinations of the two ﬁtted DIFS parameters can yield identical results, leaving the question of how to select the ‘best’ combination. Previously, problem (i) has been addressed by separating the experimental data sets into distinct time segments. To overcome these problems, a model considering two types of particulate binding sites is proposed, instead of the DIFS model which assumed one single particulate pool. A probabilistic approach is proposed to ﬁt experimental data and to determine the range of possible physical parameters using Probability Distribution Functions (PDFs), as opposed to single values without any indication of their uncertainty. The new probabilistic model, called DGT-PROFS, was tested on three diﬀerent formulated sediments which mainly diﬀer in the presence or absence of iron oxides. It was shown that a good ﬁt can be obtained for the complete set of data (instead of DIFS-2D) and that a range of uncertainty values for each modeling parameter can be obtained. The interpretation of parameter PDFs allows one to distinguish between a variety of geochemical behaviors, providing useful information on metal dynamics in sediments.

’ INTRODUCTION Information on metal exchange between the particulate phase and the solution in sediments and soils can be assessed by using the diﬀusive gradient in thin-ﬁlms (DGT) technique which is a dynamic in situ measuring technique of labile metals in solution.13 The principle behind DGT is that the sampler provides a localized region of low metal concentration, which promotes a diﬀusive ﬂux of the metal into the sampler.4 Interpretation of trace metal ﬂuxes into the DGT sampler indicates the degree of depletion of the solution metal concentration at the device interface (as determined by the thickness of the diﬀusive layer), the kinetics of desorption of the metal, and the size of the pool(s) of labile metal in the particulate phase.5,6 DGT measurements in sediments can be interpreted using dynamic models known as DGT-Induced Fluxes in Soils and Sediments (1D-DIFS6 and 2D-DIFS7), which allow describing the following values quantitatively: (i) the distribution ratio Kdl between labile adsorbed metal and metal dissolved in solution and (ii) response time Tc, which describes metal resupply kinetics from the solid phase. Despite their undisputable performance and wide dissemination among DGT users, the DIFS models have been shown to be ﬂawed in some cases. Two speciﬁc cases will be investigated in this paper. r 2011 American Chemical Society

First, the DIFS model considers a single pool of labile adsorbed metal. Some authors have suggested that the model of the system approaching equilibrium that features a single pair of forward and reverse rate constants is inaccurate and that multiple types of sorption sites should be used, each characterized by diﬀerent aﬃnities and sorption kinetics. For example, Ernstberger et al.8 used a single solid phase metal pool to describe accumulation kinetics, but the short-term desorption of Zn from the soil was not well-ﬁtted using the 1D and 2D-DIFS models.9 Additionally, the authors focused on ﬁtting the kinetics for extended exposure times (50 to approximately =500 h), because they were unable to simultaneously ﬁt short- and long-term experimental values. Other authors911 demonstrated that in some cases, especially those with short exposure times, 1D and 2D-DIFS models yield a poor ﬁt to experimental values. To overcome this limitation, they separated the experimental data into two segments: the ﬁrst segment represents a fast pool of metal exchange which represents a rapid exchange of metal (on the order of minutes to

Received: January 4, 2011 Accepted: June 27, 2011 Revised: June 9, 2011 Published: June 27, 2011 9558

dx.doi.org/10.1021/es104221b | Environ. Sci. Technol. 2011, 45, 9558–9565

Environmental Science & Technology hours); the second segment is generally observed over the course of about 48 h,11 representing a slow metal exchange. The second drawback to 2D-DIFS modeling results is that they clearly demonstrate that a number of combinations of Kdl and Tc can be ﬁtted to an experimental data set with equivalent results. For example, error maps for (Kdl:Tc) parameters by Lehto et al.9 demonstrated that for some soils, the ﬁtting procedure exhibits low sensitivity to the value of Tc parameter over several orders of magnitude and that several combinations of (Kdl:Tc) are equally valid. Because of these limitations, the aim of this study was to develop a new model and associated software for interpreting DGT measurements in soils and sediments, called DGT-PROFS or DGT PRObabilistic Fitting of parameters for Soils and Sediments, which will serve the following purposes: • to explicitly consider two pools of particulate metal, without subdividing the experimental data set into two segments (fast and slow processes) during the ﬁtting procedure. This approach will provide information about the distribution of exchangeable metal between the fast pool and the slow pool, respectively, and their desorption constants; • to use a probabilistic ﬁtting procedure to determine the most probable combination of thermodynamic and kinetic parameters, eﬀectively answering the following question: what is the probability that Kdl = X? The DGT-PROFS model was tested on DGT measurements performed on three formulated sediments to reduce the complexity of the system and measurements were monitored from 4 to 96 h. The results obtained using the DGT-PROFS model were compared to those obtained using the 2D-DIFS model.

’ MATERIAL AND METHODS Experimental Design. The following section describes an experimental procedure aimed at measuring the accumulated mass adsorbed on DGT as a function of time for three different formulated sediments called hereafter A, B, and C. These three sediments were selected from a larger data set of sediments because they exhibited contrasting behaviors see the Results section. The experimental procedure follows different stages: 1 Preparation of generic sediment. The generic sediment was formulated with common minerals, i.e., 70% acid washed Fontainebleau sand, 22% kaolinite clay, and 8% crushed natural calcite. The sediment was then mixed with mineral water (0.75 L 3 kg1 dry weight sediment) and homogenized for 4 h. 2 Addition of iron (oxyhydr)oxides in sediments B and C. Goethite (R-FeOOH) and ferrihydrite (FeOOH) were prepared as described in refs 12 and 13 and suspended in Montcalm water at a concentration of approximately 5 g 3 L1 for 16 h. X-ray diﬀraction analysis conﬁrmed the identity of the synthesized iron (oxyhydr)oxides (results not shown). These suspensions were added separately to two generic sediments to achieve a ﬁnal concentration of 2 g 3 kg1 (dry weight) and mixed for 48 h. The sediments coated with goethite and ferrihydrite were labeled B and C, respectively. 3 Spiking with metals. Sediments A and B were spiked with copper by addition of Cu(NO3)2 at a concentration of 50 mg 3 kg1, and sediment C was spiked with cadmium by addition of Cd(NO3)2 at a concentration of 2 mg 3 kg1. After addition of the metal, all three sediments were agitated over 48 h. The pH of the sediments was regularly checked

ARTICLE

and adjusted if necessary to obtain a nominal pH of 7. Sediments were then distributed in equal quantities to plastic beakers and equilibrated for 8 days at a controlled temperature of 20 ( 0.5 C. 4 Deployment of DGT units in sediments. The DGT units (piston type with 2 cm-diameter window) were obtained from DGT Research (Lancaster, UK; http://www.dgtresearch.com). Units were loaded with a Chelex-100 resin and open pore gel (0,76 mm). Prior to deployment in sediments, the gel was conditioned in an electrolyte solution as described in Zhang et al., 1995. As a control, two DGT devices were used as blanks. DGT devices were inserted into sediments to guarantee complete contact of the collecting pore with the sediment.1,5,8,14 5 DGT sampling and treatment processing. DGT devices were removed from the sediment solutions at diﬀerent times, i.e., 4, 8, 16, 24, 48, and 96 h. To verify the reproducibility of the measurements, ﬁve replicates (performed in diﬀerent beakers) were collected at 16 and 96 h, while three replicates were collected at the other times. DGT probes were retrieved and washed with deionized water to remove sediment particles before they were disassembled. Metals were extracted from the Chelex resins by immersion in 1 M nitric acid for at least 24 h at 4 C. After retrieval of the probes, the pH of the sediment solution was immediately measured, and pore water was extracted by centrifugation at 2500 g for 20 min. The collected pore water was then ﬁltered through a 0.22 μm PVDF membrane ﬁlter (Filter device PURADISC 25, Whatman, USA), acidiﬁed to 1% (v/v) with nitric acid (65% Suprapur), and stored at 4 C for analysis. The Cd and Cu concentrations were quantiﬁed in the extraction solution and in porewater by graphite furnace atomic absorption spectroscopy. Description of the Model. The model developed here assumes that transport in both the diffusion layer (the DGT gel) and the sediment porewater is solely driven by molecular diffusion and that all labile metal species in the porewater have a single self-diffusion coefficient (that is related to those of the free metal). As for the 1D-DIFS model,2,6 ours is a one-dimensional model that operates along the axis perpendicular to the DGT interface. Unlike Harper et al.,2 however, we assert here that the metal associated to particles is distributed among two particulate pools, a labile pool characterized by weak interactions and/or easily accessible metals and an ‘inert’ pool characterized by stronger, less accessible interactions, respectively. A consecutive reaction scheme that assumes that sorption onto weak sites is a prerequisite for sorption onto strong sites is presented here, but a parallel scheme involving simultaneous sorption onto weak and strong sites is also possible. The set of chemical reactions to be considered is as follows k ads, 1

Mzþ þ Sweak T MSweak k des, 1

ð1Þ

kads, 2

MSweak þ Sstrong T MSstrong þ Sweak for the ‘consecutive‘ scheme k des, 2

k ads, 2

Mzþ þ Sstrong T MSstrong for the ‘parallel’ scheme k des, 2

ð2Þ

where Mz+ is the free metal in porewater (mg cm3), Sweak is the concentration of weak sites on the particles (mg g1), Sstrong is the concentration of strong sites on the particles (mg g1), 9559

dx.doi.org/10.1021/es104221b |Environ. Sci. Technol. 2011, 45, 9558–9565

Environmental Science & Technology

ARTICLE

MSweak and MSstrong are the concentration of metal associated to weak and strong particulate sites, respectively (mg cm3), and kads,1, kdes,1, kads,2, and kdes,2 are the sorptiondesorption rate constants at weak and strong sites. The kinetic distribution of metals in the system will thus be partially governed by the initial available weak and strong site concentrations ({Sweak} and {Sstrong}). For the sake of conformity to experimental measurements and simpliﬁcation of the equations, the conditional rate constants k0 ads,1, k0 ads,2, and k0 des,2 (in s1) are deﬁned as follows 0

ð3Þ

0

ð4Þ

0

ð5Þ

k ads, 1 ¼ k ads, 1 3 ðSweak Þ k ads, 2 ¼ k ads, 2 3 ðSstrong Þ k des, 2 ¼ k ads, 2 3 ðSweak Þ

Diﬀusion and chemical reactions in the sediment produce the following set of mass-balance equations based on the consecutive sorption reaction scheme ∂CMzþ ∂2 CMzþ 0 ¼ Dsed -k ads, 1 3 CMzþ þ k des, 1 3 Sed 3 CMSweak ∂t ∂x 2

ð6Þ

0

k ads, 1 3 CMzþ dCMSweak 0 ¼ -ðk des, 1 þ k ads, 2 Þ 3 CMSweak dt Sed 0

þ k des, 2 3 CMSstrong dCMSstrong 0 0 ¼ k ads, 2 3 CMSweak k des, 2 3 CMSstrong dt

ð7Þ ð8Þ

where Dsed is the diﬀusion coeﬃcient of the metal in sediment (cm2 3 s1) (taken from the literature2), Sed is the concentration of particles in the sediments (g 3 cm3) (known from experimental design), and t is the time (s). Only the free metal Mz+ (and inorganic small complexes) is assumed to diﬀuse in the DGT gel, and so a single kinetic equation is proposed to describe the transport in the diﬀusive layer of the DGT device ∂CMzþ ∂2 CMzþ ¼ Dgel ð9Þ ∂t ∂x 2 Equations at the sediment-DGT device interface are designed according to the same principles as those presented in Harper et al.2 The model described above was implemented on the Ecolego platform (see www.facilia.se), where advanced methods for probabilistic and sensitivity analysis are available. This platform allows the use of features such as (i) deﬁning Probability Density Functions (PDF) for each uncertain parameter, (ii) using the Monte Carlo method for the propagation of parametric uncertainties, and (iii) making it possible to account for correlations between parameters in the Monte Carlo process and (iv) regression methods and Fourier-based methods for conducting sensitivity analysis of selected input and output parameters. Some of these functions are essential in the probabilistic ﬁtting procedure described below. Fitting Procedure. According to eqs 6 to 9, six parameters must be fitted a priori, i.e., the sorption and desorption rate constants k0 ads,1, kdes,1, k0 ads,2, and k0 des,2 and the diffusion coefficients Dsed and Dgel.

In porous media such as sediments, molecular diﬀusion coeﬃcients depend on the following: • tortuosity, which can directly be related to porosity by the Millington and Quirk15 relationship; • metal speciation in the sediment. Indeed, compared to free metal and small inorganic species, metals bound to humic substances generally have a lower diﬀusion coeﬃcient.16,17 Thus, diﬀusion of the diﬀerent metal species (i.e., free, inorganic, and humic species) through the sediment and gel layer generates ‘apparent’ diﬀusion coeﬃcients that are lower than those of free metal.18 Thus, the variability of Dsed and Dgel implicitly includes several sources of uncertainty related to metal speciation and physical characteristics of the sediment. In our case, however, these parameters were assumed to be known because the porosity of our formulated sediments was high, being freshly deposited sediment where no compaction had occurred and because no organic acids were added to the sediments. According to the Millington and Quirk relationship, tortuosity, and subsequently the diﬀusion coeﬃcient in sediment, is maximized. The absence of organic acids makes it possible to propose that complexation processes in the pore water are minimized, resulting in a maximized diﬀusion coeﬃcient. Maximum diﬀusion coeﬃcients reported in the literature were then considered. It should be noted that in a companion paper accepted for publication,18 formulated sediments containing humic acids were investigated, and diﬀusion coeﬃcients were considered to be uncertain parameters. In the experiments described above, the following concentrations were measured before the deployment of the DGT device into the sediment: (i) the total concentration on particles, i.e. CMSweak (t = 0) + CMSstrong (t = 0) and (ii) the total concentration of the metal in porewater. We introduced a new parameter, Rweak, which represents the proportion of metal adsorbed to labile sites at time t0, deﬁned as Rweak ¼

CMSweak ðt ¼ 0Þ CMSweak ðt ¼ 0Þ þ CMSstrong ðt ¼ 0Þ

ð10Þ

This new parameter serves to simplify the analysis and represents a simple physical phenomena. Several reasonable assumptions were considered to reduce the number of parameters to be ﬁtted. One important assumption was that before deployment of the DGT device, the whole sediment system was assumed to be at equilibrium making it possible to simplify eq 8 to 0

k ads, 2 CMSstrong ðt ¼ 0Þ 1 Rweak ¼ 0 ¼ Rweak CMSweak ðt ¼ 0Þ k des, 2

ð11Þ

The value of k0 ads,2 can thus be derived from k0 ads,2 and Rweak. Similarly, by assuming that no diﬀusion occurs before DGT deployment, eq 6 can be simpliﬁed to 0

k ads, 1 CMSweak ðt ¼ 0Þ ¼ k des, 1 CMzþ ðt ¼ 0Þ Sed 3 Rweak 3 ºCMSweak ðt ¼ 0Þ þ CMSstrong ðt ¼ 0Þß CMzþ ðt ¼ 0Þ

¼

ð12Þ 0

The value of k ads,1 can thus be derived from kdes,1 and Rweak. Finally, the three remaining parameters (Rweak, kdes,1, kdes,2) must be ﬁt using experimental data. 9560

dx.doi.org/10.1021/es104221b |Environ. Sci. Technol. 2011, 45, 9558–9565

Environmental Science & Technology

ARTICLE

Figure 1. Steps followed for ﬁtting model parameters Step 1: deﬁnition of noninformative PDFs for each parameter Step 2: spaghetti plot showing the simulations conducted with 10000 random parameters combinations (experimental data points are represented by white points) Step 4: ranking of the 100 best simulations Step 5: determination of PDFs for each parameter by a weighted bootstrap procedure Step 6: determination of a possible parameters correlation Step 7: informative PDFs available for probabilistic simulation.

As discussed in the Introduction section, previous research using DIFS models demonstrated that several combinations of parameters can equally be ﬁt to a given set of experimental data. To address this problem, we proposed a probabilistic approach that would allow identiﬁcation of the most probable combinations of parameters. Instead of determining a single value for each of them, the objective is to deﬁne Probability Density Functions (PDF) that will describe a range of the most probable values of the parameters. The procedure that was used for ﬁtting the parameters is demonstrated using an example in the Results section below.

’ RESULTS AND DISCUSSION Reproducibility and Physicochemical Control. All physicochemical measurements involved with controlling the quality of the experimental procedure (i.e., estimation of the reproducibility through replicates, pH, and porewater variation over the duration of the experiment, etc.) and their associated results are detailed in ref 18. Briefly, it was observed that (i) sediment pH varied minimally (<0.07 pH unit) from the initial pH over the course of the tests, (ii) Cd and Cu porewater concentration was 9561

dx.doi.org/10.1021/es104221b |Environ. Sci. Technol. 2011, 45, 9558–9565

Environmental Science & Technology

ARTICLE

Figure 2. Time dependence of mean ﬂuxes to DGT for sediments A, B, and C. The gray line represents the mean curve obtained from 1000 random simulations. The upper and lower dotted lines represent the 5th and 95th percentiles, respectively. The black line represents the 2D-DIFS model.

essentially constant over the exposure period, varying <3% and suggesting that equilibration with the solid phase was complete after the sediment preparation time (8 days), and (iii) replicate variability in the measurements of DGT at each deployment time is limited in the case of the sediments B and C (RSD < 10%) and greater for experiment A (RSD< 18%). Probabilistic Estimation of Model Parameters: Detailed Procedure. In this section, only the consecutive DGT-PROFS model (and not the parallel scheme) was used to investigate the applicability of this new model for interpreting DGT measurements in sediments. This section aims to describe the steps followed for deriving the PDFs for each of the model parameters. Sediment A was chosen as an example to illustrate this procedure: Step 1: Selection of Prior PDFs for Each Parameter. Each of the parameters to be fitted is described by a PDF that characterizes the range of potential values that can be provided from preliminary knowledge. For example, sequential extraction techniques performed on sediment particulate matter19 can yield a preliminary quantification of exchangeable and nonexchangeable adsorbed metals. In the absence of preliminary experiments, uniform or log-uniform PDFs were selected to maximize the entropy of the system (Figure 1, Step 1). Step 2: Random Simulations. Ten thousand values of each parameter are randomly sampled by a Monte Carlo-Latin Hypercube procedure. The calculation described above was performed for each of the 10000 combinations of parameters, providing a ‘spaghetti’ plot (Figure 1 Step 2). Step 3: Error Calculations. For each of the simulations, mean error and maximum error values were then calculated as follows

εmean, p ¼

1 n

∑i

ðjMmeasured ðti Þ Mp, calculated ðti ÞjÞ Mmeasured ðti Þ

εmax , p ¼ max i

ðjMmeasured ðti Þ Mp, calculated ðti ÞjÞ Mmeasured ðti Þ

where M(ti) refers to the accumulated mass on the DGT probe at time ti, the index p refers to the pth combination of parameters, and i refers to the ith time of measurement during experiments. Step 4: Simulations Ranking. The results of the simulations were ranked according to two criteria. First, the 1000 simulations that produced the minimal mean error εmean,p were selected. Among these, 100 simulations that minimized the maximum error εmax,p were selected. Step 5: Determination of Posterior PDFs. Once these top 100 combinations were identified, the 100 values of each parameter were fit to a PDF using a weighted bootstrap procedure.20 For simplicity, the example of Rweak is presented here. A score was affected to each Rweak; in our study, we chose the following value as the score 1 ScoreðRweak, p Þ ¼ pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ Rank Rweak, p where RankRweak,p is the rank of combination p. This procedure allows one to assign more importance in the fitting procedure to one value of the parameters (i.e., those minimizing εmean and εmax) over another. We therefore performed a nonequiprobable sampling with replacement from data selected above, where the probability of drawing each Rweak value corresponded to the score determined for each of the variable pairs. The number of samples generated by the bootstrap procedure was 1000, and the number of data drawn for each one corresponded to that of the initial data set (i.e., 100). These data were then fitted to normal or log-normal PDFs. Figure 1 provides an example of a PDF that can be obtained using this procedure, demonstrating the application of the technique to the value Rweak of sediment A. 9562

dx.doi.org/10.1021/es104221b |Environ. Sci. Technol. 2011, 45, 9558–9565

3.6

3.9

7.9

8.5

Consecutive model. b Parallel model. c LN(a;b;c) means a log-normal distribution, with a: geometric mean value; b: 5th percentile; c: 95th percentile. N(a;b;c) means a normal distribution, with a: mean value; b: 5th percentile; c: 95th percentile. The value in the ‘correlation’ columns correspond to the Pearson’s correlation coeﬃcients between the indicated parameters. Kd,tot and Kd,weak are the distribution coeﬃcients, deﬁned respectively as the ratio between total metal in sediment and metal concentration in pore water, experimentally measured, and the ratio of metal associated to weak sites in sediment and the metal concentration in pore water, determined from eq 12. Mean error is deﬁned in the text in the section Probabilistic estimation of model parameters, Step 3 Correlation between Rweak and kdes,1 is deﬁned in the section Probabilistic estimation of model parameters. a

1.9 103 5.2 101 0.7 N(0.36;0.32;0.40) Cb

N(1.50 10-5; 1.2 105; 1.8 105) LN(8.1 10-9;1.5 1010;4.2 107)

1.4 103 5.2 101 LN(1.67 10-8;7 1010;4 107) N(0.37;0.29;0.45) Ca

N(1.48 10-5;8.5 106;2.1 105)

0.7

7.7 101 4.8 10 LN(1.5 10 ;3.4 10 ;6.4 10 ) LN(2.1 10 ; 4.2 10 ;10 ) LN(0.016;0.012;0.022) B

2.0 102 1.4 103 0.7

3

8 9

LN(1.1 10-8;1.1 109;1.1 107)

-8

3 5

kdes,2 (s1)

LN(2.2 10-6; 106;5 106) LN(0.14;0.07;0.27)

-4

a

Aa

strong sites:

desorption rate from desorption rate from

kdes,1 (s1) experiment metal and total particulate metal: Rweak

ratio between weak particulate

Table 1. Fitted Parameter PDFs for Sediments A, B, and C, Respectivelyc

weak sites:

correlation Rweak-kdes,1 Kd,tot (cm3 3 g1) Kd,weak (cm3 3 g1) mean error (%)

Environmental Science & Technology

ARTICLE

Step 6: Correlations. Additionally, once the 100 ‘best’ combinations were identified in Step 4, correlations between parameters could be modeled shown in Figure 1, Step 6 (see the Discussion section). Step 7: Calibration Verification. Once PDFs were fitted for all the parameters, the DGT-PROFS model was run again replacing the noninformative PDFs used in step 1 (i.e., the uniform or loguniform PDFs) with the informative PDFs obtained from step 5 (e.g., normal or log-normal PDFs). Correlations between parameters at step 6 were introduced in the DGT-PROFS model, and the Iman-Conover technique21 was applied to generate parameter samples respecting these correlations. Step 8: Results Summary. The spaghetti plot calculated with parameters obtained from informative PDFs was compared to the experimental data. To clarify the interpretation, the plot features three curves representing the mean, fifth percentile, and 95th percentile values. Comparison between Experimental Results and Predicted Models. Figure 2 shows the measured and simulated average fluxes into the DGT probes for each experiment. It can be observed that the measured fluxes into the probe are different for all three sediments. For sediment A, the mean flux into DGT probe is essentially constant over time. For sediment C, the flux significantly decreases over time. Sediment B exhibits an intermediate behavior, characterized by a significant decrease of the flux over short deployment times and a stabilization of the flux over longer deployment times. Qualitatively, using the terminology proposed by Harper et al.,6 these sediments can be classified as follows: (i) for sediment A, it was observed that the resupply of metals at the interface is sufficiently rapid to maintain a constant flux into the DGT probe over time. It can thus be considered representative of a ‘sustained’ or ‘partial’ case. (ii) For sediment B and, to a lesser extent sediment C, it was observed that the resupply of metals at the interface decreased with time, and, as a consequence, the flux into the DGT probe decreases during the deployment period. These sediments can be considered representative of ‘diffusive’ or ‘partial-non steady state’ cases. This observation demonstrates that the systems tested in the current study are representative of those observed in natural sediments. For any length of deployment of the DGT, the comparison between model and experimental ﬂuxes (Figure 2) demonstrates that the average curve generally agrees well with experimental data and that the area encompassed by the ﬁfth and the 95th percentiles includes experimental data points. The range of simulated values are generally tight (see sediments B and C), revealing that the uncertainty in the prediction is low. However, the narrowness of the uncertainty range also demonstrates that all the combinations generated from the informative PDFs produce similar kinetic results; obtaining greater precision in the estimation of parameters is not possible, as many combinations of parameters can yield similar results. This result demonstrates why a probabilistic approach that produces PDF estimations rather than single values should be preferred. Estimating single values without any indication of the range of potential solutions that yield similar results can allow one to distinguish between different sediments, while they cannot be diﬀerentiated based on a probabilistic point of view. It can also be observed that the uncertainty range for sediment A is larger; the experimental data point exhibit ﬂuctuations that result in decreased precision in the estimation of the parameters. As an example, the ﬁfth and 95th percentiles for Rweak in sediment A are 0.07 and 0.27, reﬂecting a relatively large uncertainty. High-quality experimental data are essential to decrease the uncertainty in the estimation of the 9563

dx.doi.org/10.1021/es104221b |Environ. Sci. Technol. 2011, 45, 9558–9565

Environmental Science & Technology parameters. In Figure 2, simulations conducted with the 2DDIFS model are also presented, using the ﬁtting procedure included in the software to estimate the labile distribution coefﬁcient Kd,l and the time constant Tc. As previously discussed, DIFS exhibited poor agreement when short- and long-term experimental data are simultaneously used for the ﬁtting process. Interpretation of Parameters Values. An analysis of the physicochemical factors governing geochemical behavior in sediments (e.g., pH, iron oxide type, aging, the presence of organic matter) is beyond the scope of the current study, as such an analysis would be irrelevant with only three sediments. A more detailed discussion is provided in Nia et al.,18 which analyzed 12 different sediments. This section discusses how fitted parameter PDFs obtained using the DGT-PROFS model can be interpreted. Consecutive and parallel schemes were modeled for the chosen sediments. As an example, parameter values for sediment C are presented in Table 1. The parameters values are similar for both models, demonstrating that the results are not sensitive to the selected modeling scheme. Based on this observation, only the consecutive scheme will be discussed here. Analysis of the ﬁtted parameter allows one to clearly discriminate between contrasting geochemical behaviors. For example, sediments A, B, and C clearly exhibit quite diﬀerent Rweak ranges (see Table 1). For sediment B, most of the metal is predicted to be associated with strong binding sites, as evidenced by the low Rweak value (mean = 1.6%). In contrast, a signiﬁcant fraction of the metal in the system is predicted to be associated with weak sites in sediment C (mean = 37%) and to a lesser extent sediment A (mean = 14%). Such information is of great interest for providing a more complete understanding of metal speciation within the particulate structure of sediments, allowing one to distinguish between diﬀerent types of metal binding sites. Additionally, values obtained for Rweak enhance the relevance of taking into account two diﬀerent types of binding sites on particulate matter. To illustrate that, simulations were conducted by assuming that only one type of site is present in the sediment and setting Rweak = 1. In such a simulation, the same ﬂaws are observed as are present under the DIFS model, i.e., that ﬁtting the complete set of experimental data requires one to separate it into two segments for short and long time periods (results not shown). Desorption rate values also provide information regarding the rate of resupply of metal into the sediment that can be supported by the investigated systems. kdes,1 values diﬀer between sediments by approximatively 2 orders of magnitude. In contrast, kdes,2 values are quite similar, demonstrating that the desorption from strong sites is weakly dependent on the identity of the iron-(oxyhydr)oxide constituent of the sediment. It is also useful to compare the results obtained from the DGTPROFS model with those derived from the DIFS models. The labile distribution coeﬃcient Kdl as deﬁned in the DIFS model describes the partition of metal between a ‘labile’ particulate sink and the pore water. This variable can be substituted by our Kd,weak distribution coeﬃcient, calculated by the DGT-PROFS model and deﬁned as the ratio between metal associated with weakly binding sites and metal in the pore water, which is calculated from eq 12. The values of Kd,weak calculated for the selected sediments are presented in Table 1; the large range of Kd,weak values demonstrates that this parameter also allows one to discriminate between the behaviors of diﬀerent sediments. We also compared Kd,weak values with total distribution coeﬃcient Kd,tot, deﬁned as the ratio between the total metal adsorbed to the sediment and metal found in the pore water. The ratio Kd,tot/Kd,weak

ARTICLE

is 7.1, 62.5, and 2.7 for sediments A, B, and C, respectively, indicating that similarly to the Rweak values, the distribution of metal between weak and total binding sites diﬀers highly among sediments. However, it can be noted that Ernstberger et al.22 also compared Kd values among a number of sediments (in their case, Kd,tot and the DIFS Kdl). They obtained results for ﬁve diﬀerent soils that diﬀered by multiple orders of magnitude diﬀerent from the results presented in this paper. Furthermore, the values of Kd,tot/Kdl calculated there were approximatively 1 for most soils containing Cd, Ni, and Zn. Due to the similarity between Kd,tot and Kdl, a future area of investigation could focus on whether the DIFS model is able to discriminate between labile and total metal binding sites in a sediment. In some cases, modeling interactions between parameters were observed, such as between Rweak and kdes,1 for sediment A and C. Such interactions demonstrate that some parameters are linked in the modeling process, which reﬂects the fact that several combinations of parameters, such as Rweak and kdes,1, can yield similar results. For instance, associating a ‘high’ Rweak value with a ‘small’ kdes,1 value is equivalent to associating a ‘small’ Rweak value to a ‘high’ kdes,1 value. This observation supports the probabilistic approach described here, as a single combination of parameters is not suﬃcient to describe the myriad of possible values.

’ AUTHOR INFORMATION Corresponding Author

*Phone: 0033 130877259. Fax: 0033 1307-877336. E-mail: philippe.ciﬀ[email protected].

’ REFERENCES (1) Zhang, H.; et al. In situ high resolution measurements of ﬂuxes of Ni, Cu, Fe, and Mn and concentrations of Zn and Cd in porewaters by DGT. Geochim. Cosmochim. Acta 1995, 59, 4181–4192. (2) Harper, M. P.; et al. Kinetics of metal exchange between solids and solutions in sediments and soils interpreted from DGT measured ﬂuxes. Geochim. Cosmochim. Acta 1998, 62, 2757–2770. (3) Degryse, F.; et al. Predicting availability of mineral elements to plants with the DGT technique: a review of experimental data and interpretation by modelling. Environ. Chem. 2009, 6, 198–218. (4) Zhang, H.; et al. Kinetics of Zinc and Cadmium release in freshly contaminated soils. Environ. Toxicol. Chem. 2006, 25, 664–670. (5) Zhang, H.; et al. In Situ Measurements of Solution Concentrations and Fluxes of Trace Metals in Soils Using DGT. Environ. Sci. Technol. 1998, 32, 704–710. (6) Harper, M. P.; Davison, W.; Tych, W. DIFS-a modelling and simulation tool for DGT induced trace metal remobilisation in sediments and soils. Environ. Modelling Software 2000, 15, 55–66. (7) Sochaczewski, L.; et al. 2D DGT induced ﬂuxes in sediments and soils (2D DIFS). Environ. Modelling Software 2007, 22, 14–23. (8) Ernstberger, H.; et al. Measurement and Dynamic Modeling of Trace Metal Mobilization in Soils Using DGT and DIFS. Environ. Sci. Technol. 2002, 36, 349–354. (9) Lehto, N. J.; et al. Quantitative assessment of soil parameter (KD and TC) estimation using DGT measurements and the 2D DIFS model. Chemosphere 2008, 71, 795–801. (10) Nowack, B.; Koehler, S.; Schulin, R. Use of Diﬀusive Gradients in Thin Films (DGT) in Undisturbed Field Soils. Environ. Sci. Technol. 2004, 38, 1133–1138. (11) Cornu, J. Y.; et al. Temporal evolution of redox processes and free Cd dynamics in a metal-contaminated soil after rewetting. Chemosphere 2007, 70, 306–314. 9564

dx.doi.org/10.1021/es104221b |Environ. Sci. Technol. 2011, 45, 9558–9565

Environmental Science & Technology

ARTICLE

(12) Rusch, B.; Hanna, K.; Humbert, B. Coating of quartz silica with iron oxides: Characterization and surface reactivity of iron coating phases. Colloids Surf., A 2010, 353, 172–180. (13) Iron Oxides in the Laboratory: Preparation and Characterization; Schwertmann, U. Cornell, R. M., Eds.; WileyVCH: 2000; New York. (14) Sangi, M. R.; Halstead, M. J.; Hunter, K. A. Use of the diﬀusion gradient thin ﬁlm method to measure trace metals in fresh waters at low ionic strength. Anal. Chim. Acta 2002, 456, 241–251. (15) Millington, R. J.; Quirk, J. P. Permeability of porous solids. Faraday Soc. Trans. 1961, 57, 1200–1206. (16) Unsworth, E. R.; Zhang, H.; Davison, W. Use of Diﬀusive Gradients in Thin Films To Measure Cadmium Speciation in Solutions with Synthetic and Natural Ligands: Comparison with Model Predictions. Environ. Sci. Technol. 2005, 39, 624–630. (17) VanDerVeeken, P. L. R.; Pinheiro, J. P.; VanLeeuwen, H. P. Metal Speciation by DGT/DET in Colloidal Complex Systems. Environ. Sci. Technol. 2008, 42, 8835–8840. (18) Nia, Y.; et al. Kinetic resupply of metal in artiﬁcial sediments coated with iron oxides using DGT and DGT-PROFS model: Eﬀect of aging and humic acids. Chemosphere 2011, accepted for publication. (19) Tessier, A.; Campbell, P. G. C.; Bisson, M. Sequential extraction procedure for the speciation of particulate trace metals. Anal. Chem. 1979, 51, 844–850. (20) Duboudin, C.; Ciﬀroy, P.; Magaud, H. Eﬀects of data manipulation and statistical methods on species sensitivity distributions. Environ. Toxicol. Chem. 2004, 23, 489–499. (21) Iman, R. L.; Conover, W. J. A Distribution-Free Approach to Inducing Rank Order Correlation Among Input Variables. Commun. Statist.-Simula Computa 1982, 3, 311–334. (22) Ernstberger, H.; et al. Desorption Kinetics of Cd, Zn, and Ni Measured in Soils by DGT. Environ. Sci. Technol. 2005, 39, 1591–1597.

9565

dx.doi.org/10.1021/es104221b |Environ. Sci. Technol. 2011, 45, 9558–9565

ARTICLE pubs.acs.org/est

Effects of Phosphate on the Transport of Escherichia coli O157:H7 in Saturated Quartz Sand Lixia Wang,† Shangping Xu,‡ and Jin Li*,† † ‡

Department of Civil Engineering and Mechanics, University of Wisconsin-Milwaukee, Milwaukee, Wisconsin 53201, United States Department of Geosciences, University of Wisconsin-Milwaukee, Milwaukee, Wisconsin 53201, United States

bS Supporting Information ABSTRACT: Consumption of groundwater contaminated with E. coli O157: H7 has led to several waterborne disease outbreaks over the past decade. A thorough understanding of the transport of E. coli O157:H7 within the soilgroundwater system is critical to the protection of public health. Although phosphate is ubiquitous in the natural environment, the inﬂuence of phosphate on the transport of E. coli O157:H7 in the groundwater system remains unknown. In this research, we performed column transport experiments to evaluate the eﬀect of phosphate on the transport of E. coli O157:H7 cells within saturated sand. The pH of the solutions was maintained at 7.2, the ionic strength varied from 10 to 100 mM, and the phosphate concentration ranged from 0 to 1 mM. Our results show that (1) phosphate could enhance the transport of E. coli O157:H7 cells under both ionic strength conditions; (2) E. coli O157:H7 displayed lower retention in sand under higher ionic strength conditions; (3) increased phosphate in the mobile aqueous phase led to the release of previously immobilized E. coli O157:H7 cells. The response of E. coli O157:H7 cells to variations in phosphate concentrations and ionic strength conditions are explained using the extended DLVO (XDLVO) theory and the steric repulsion caused by extracellular macromolecules. In summary, our results suggest that phosphate could widen the spread of E. coli O157:H7 cells, and potentially other types of bacterial cells, within the soilgroundwater system.

1. INTRODUCTION Escherichia coli O157:H7 is a shiga-toxin producing, pathogenic strain of E. coli that can cause fatal diseases such as hemorrhagic diarrhea and hemolytic uremic syndrome (HUS).1,2 The Centers for Disease Control and Prevention (CDC) estimates that E. coli O157:H7 causes ∼73 000 illnesses, ∼2200 hospitalizations, and ∼61 deaths each year in the United States, and that drinking water is responsible for ∼15% of the reported infections.1,3 Several recent E. coli O157:H7 outbreaks were linked to the consumption of contaminated groundwater, highlighting the importance of understanding the transport of E. coli O157:H7 within the soil-groundwater system.46 Cattle represent the most important sources of E. coli O157: H7.4,79 Survey results showed that E. coli O157:H7 was commonly detected in cattle manure and its density could reach as high as 105 CFU g1.1015 Additionally, it was found that E. coli O157:H7 in cattle manure could survive over an extended period of time.1618 When manure from cattle farms is applied to agricultural ﬁelds as a fertilizer, E. coli O157:H7 cells can leach out, inﬁltrate through the soil zone, and subsequently contaminate the underlying groundwater.6,19 Findings from the latest investigations into the transport and retention of E. coli O157:H7 in saturated porous media have indicated some unique behaviors of this bacterium.2026 Castro and Tufenkji,25 Haznedaroglu et al.,20 Schinner et al.,23 and r 2011 American Chemical Society

Kim et al.24 reported that under low pH (∼5.7) conditions there was minimal variation in the retention of E. coli O157:H7 when ionic strength was varied within the range of 1100 mM. Under high pH conditions (pH > 8.4), however, the deposition of E. coli O157:H7 decreased with increasing ionic strength.24 When the extracellular macromolecules on the surface of E. coli O157:H7 cells were removed through proteinase K digestion, the attachment of E. coli O157:H7 became positively related with ionic strength under low pH conditions, suggesting that cell surface macromolecules had a strong inﬂuence on the transport of E. coli O157:H7.22,24 Phosphate is a ubiquitous chemical species that can be found in minerals, soils, human bodies, and water. Phosphate is also a key ingredient used in numerous domestic and industrial applications, for example, detergents, metal surface coating, fertilizers, and drinking water distribution pipe corrosion control. The addition of phosphate to water mains has been shown to improve the water quality by reducing the occurrence of coliform bacteria and inhibit bioﬁlm growth, despite the fact that phosphate serves as an essential nutrient for microorganisms.27 Several studies Received: April 4, 2011 Accepted: September 28, 2011 Revised: September 6, 2011 Published: September 28, 2011 9566

dx.doi.org/10.1021/es201132s | Environ. Sci. Technol. 2011, 45, 9566–9573

Environmental Science & Technology investigating the role of phosphate in bacterial interaction with iron-coated sand28 and iron oxyhydroxide particles27 have attributed the reduced bacterial adhesion to charge modiﬁcation of the grain surface from positive to negative by the adsorbed phosphate ions. In soil solution and, particularly in agricultural soil that receives manure and/or chemical fertilizers, phosphate concentrations can reach mM levels.2934 As phosphate was reported to enhance the transport of bacterial cells (e.g., Pseudomonas ﬂuorescens) and viruses within natural porous media,35,36 little is known about its eﬀect on the transport of E. coli O157:H7. The main goal of this research is to examine the eﬀect of phosphate on the transport of E. coli O157:H7 in quartz sand packed columns. In addition, we evaluate the release of E. coli O157:H7 cells attached to sand surfaces when phosphate concentration in the mobile aqueous phase is elevated.

2. MATERIALS AND METHODS 2.1. Preparation of Bacteria Suspension. E.coli O157:H7 strain EDL933 stx1 and stx2 double mutant obtained from Dr. C.-H. Yang’s lab in the department of Biological Sciences at the University of Wisconsin-Milwaukee were used in the experiments. Preserved cells stored in 20% glycerol at 80 °C were streaked onto LuriaBertani (LB) agar plates supplemented with 20 μg/mL chlorampfenicol, 50 μg/mL kanamycin, and 150 μg/mL rifampin, and the plates were incubated at 37 °C overnight. One single colony was picked by a sterile needle and transferred into culture tubes that contained 10 mL sterile LB broth amended with antibiotics of similar concentrations. Following 6 h incubation at 37 °C, 0.5 mL of the starter culture was added to a flask that contained 250 mL sterile LB broth amended with antibiotics, and incubated at 37 °C with 200 rpm shaking for 18 h. The bacterial cells were harvested through centrifugation (3000g, 10 min, and 4 °C) and the pellets were rinsed twice using appropriate electrolyte solutions to remove the growth media. Cell concentration was then adjusted to ∼3 107cells/mL and the suspension was ready for column transport experiments. The cell surface macromolecules were left unaltered in the suspension used in the column transport experiments. The motility of the bacteria did not change after the double mutation. 2.2. Electrolyte Solutions. Six different types of electrolyte solutions were used in the experiment. The total ionic strength of the electrolyte solutions was either 10 mM or 100 mM (adjusted using NaCl). Under each ionic strength condition, the phosphate concentrations varied as 0 mM (no phosphate), 0.1 mM (77 μM Na2HPO4, 23 μM NaH2PO4) and 1 mM (770 μM Na2HPO4, 230 μM NaH2PO4), respectively. Because the presence of phosphate changed water pH to ∼7.2, the pH of the solutions that contained no phosphate was raised to 7.2 using 0.1 mM NaHCO3. 2.3. Column Transport Experiments. High-purity quartz sand (US Silica) with a size range of 0.3540.420 mm was used in the experiments. The porosity of the sand was 0.344. The sieved sand was treated alternately with hot, concentrated nitric acid and diluted NaOH solutions to remove surface iron oxide/ hydroxide coatings and organic materials, as well as fine particles attached to sand surfaces. Following each cleaning step, the sand was thoroughly rinsed with deionized water. The clean sand was dried in an oven at 55 °C and then stored in high-density polyethylene (HDPE) containers until use. Duplicate glass chromatography columns (Kontes) measuring 15 cm in length and 2.5 cm in diameter were wet-packed using

ARTICLE

the clean quartz sand. The packed columns were equilibrated with more than 20 pore volumes of background electrolyte solution. Peristaltic pumps (Cole-Parmer MasterFlex) were used to regulate the ﬂow and the speciﬁc discharge was maintained at 0.31 cm/min. Following the equilibrium step, bacterial suspension was injected into the columns and the injection lasted 60 min (∼3 pore volumes). The column eﬄuent was connected to ﬂow-through quartz cells and the concentration of the bacterial cells was monitored every 30 s using a spectrophotometer (Shimadzu UV-1700) by measuring the absorbance at a wavelength of 220 nm.37,38 Upon the termination of cell suspension injection, the columns were ﬂushed with background electrolyte solution until the absorbance of the eﬄuent returned to zero. It was previously shown that, for the relatively uniform sand used in this research, the eﬀect of dispersion on the transport of colloid-sized particles was negligible.39 The kinetics of the deposition of E. coli O157:H7 cells within the saturated sand packs under clean-bed conditions can be estimated by calculating the deposition rate coeﬃcient, kd (min1), from the early cell breakthrough concentrations in the eﬄuent:40,41 U c ln kd ¼ ð1Þ ε3L co Where ε is the porosity of the sand (cm3/cm3), U is the speciﬁc discharge (cm/min), L is the length of the column (cm) and C/C0 is the normalized breakthrough concentration relevant to clean-bed conditions, which was obtained from average bacterial breakthrough concentrations between 1.8 and 2.0 pore volumes.25,40 Experiments were also performed to investigate the potential release of E. coli O157:H7 cells due to increased phosphate concentration in the mobile aqueous phase. For this purpose, column transport experiments were ﬁrst conducted using a solution that had an ionic strength of 100 mM but did not contain any phosphate. Once the columns were ﬂushed, a solution with similar ionic strength and pH (i.e., 100 mM and 7.2) but contained 0.1 mM phosphate was injected into the columns. When the released cells were ﬂushed out from the columns (i.e., the absorbance of the eﬄuent returned to background values), a 100 mM solution that contained 1 mM phosphate was then injected into the columns. The experiments stopped after the pulse release of the bacterial cells was completed. 2.4. XDLVO Interaction between E. coli O157:H7 Cells and Quartz Sand. The transport of bacterial cells within saturated porous media is governed by the energy interactions between bacterial cells and the surface of solid matrix (e.g., sand). According to the extended DerjaguinLandauVerwey Overbeek (XDLVO) theory, the forces include the Lifshitzvan der Waals (LW) interactions, the electrostatic double layer (EDL) repulsion, and the Lewis acidbase (AB) (i.e., hydrophobic) interaction:4244 ΦTotal ¼ ΦLW þ ΦEDL þ ΦAB

ð2Þ

The LW, EDL, and AB interaction energies (ΦLW, ΦEDL, and ΦAB) can be calculated using the following equations:4552 ΦLW ¼ 9567

Aab 6h

ð3Þ

dx.doi.org/10.1021/es201132s |Environ. Sci. Technol. 2011, 45, 9566–9573

Environmental Science & Technology A ¼ 24πho 2 ð

ARTICLE

pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ γb LW γw LW Þð γs LW γW LW Þ

Φ

EDL

1 þ expð khÞ ¼ πε0 εw ab 2ψb ψs ln 1 þ expð khÞ 2 2 þ ðψb þ ψs Þln½1 expð 2khÞ

ð4Þ

ΦAB ¼ 2πab λw ΔGAB ho exp

ð5Þ

h0 h λw

ð6Þ

pﬃﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃ γw þ ð γb þ γs γw Þ pﬃﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃ þ γw ð γb þ þ γs γw Þ pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ γb γs þ γb þ γs

ΔGAB ho ¼ 2½

ð7Þ

where A represents the Hamaker constant; ab is the equivalent radius of the bacterial cells;h is the separation distance between the bacterium and sand surface; h0 represents the minimum equilibrium distance between the cell and sand surface and equals to 0.157 nm; γLW, γ+ and γ are the LW, electron-accepting and electron-donating interfacial tension parameters, respectively; ε0 and εw are the dielectric permittivity of vacuum and water, respectively; k is the inverse of Derby length; ψb and ψs are the surface potentials of the bacterial cells and sand, respectively; λw (= 0.6 nm) is the characteristic decay length of AB interactions in water; and ΔGAB ho represents the hydrophobicity interaction free energies per unit area corresponding to h0. For the interfacial tension parameters (i.e., γLW, γ+, or γ), the subscripts of b, w, and s represent bacteria, water, and sand, respectively. For water, the values of γLW, γ+, or γ are 21.8, 25.5, and 25.5 mJ m2, respectively.53 Values of γLW (39.2 mJ m2), γ+ (1.4 mJ m2), or γ (47.8 mJ m2) for quartz were determined in Morrow et al.52 and were used in this research. Values of γLW, γ+, or γ for E. coli O157:H7 cells were determined by measuring the contact angles (θ) using three different probe liquids with known surface tension parameters:53 pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ γi L ð1 þ cos θÞ ¼ 2 γi LW γLW þ 2 γi þ γ þ 2 γi γþ ð8Þ where the subscript i represents water (γ = 72.8, γ = 21.8 and γ+ = γ = 25.5 mJ m2 ), glycerol (γ L = 64.0, γLW = 34.0, γ + = 3.92 and γ = 57.4 mJ m2 ) or diiodomethane (γL = 50.8, γLW = 50.8 and γ + = γ = 0 mJ m 2 ). 53 The contact angles were acquired with a Rame-Hart goniometer using bacterial lawns produced by filtering cells onto porous membrane.45 2.5. Steric Interaction Between E. coli O157:H7 Cells and Quartz Sand. In biological systems, the classic DLVO model often failed to fully explain the bacterial transport and deposition behavior observed in experiments due to the presence of extracellular macromolecules on bacterial surface.4043 The steric repulsion between two parallel surfaces similarly coated by macromolecules is described by the deGennes equation:54,55 " 3=4 # kT 2L 9=4 D P¼ 3 For D < 2L ð9Þ s D 2L L

LW

Where P is the pressure between the two parallel surfaces, D is the separation distance, L is the thickness of brush layer and s is the average distance between anchoring sites. For E. coli O157:H7, of the value of L is ∼30 nm56 and the values of s is ∼2.2 nm.57 If one plate has the brush and the other plate is bare, 2D should substitute D and the pressure should be divided by 2:58 " 3=4 # 1 kT L 9=4 D P ¼ For D < L ð10Þ 2 s3 D L Integration using Derjaguin’s approximation, we have the steric force expression for a sphere-plate system:54,55 " 3=4 # 1 kT Z L L 9=4 x F ¼ 2πR 3 dx 2 s D x L " # 5=4 4L kT D 7=4 L ¼ πR 3 5 þ 7 12 ð11Þ 35 s L D The integration of F gives the steric interaction energy (ΦSteric) for a sphereplate system ΦSteric ¼

Z L

FðxÞdx " # 3=4 3=4 4 kT D D ¼ þ 308L3 420DL2 þ 132D2 L πR 3 20D3 385D s L L D

ð12Þ 2.6. E. coli O157:H7 Cell Characterization. Zeta potential values of bacterial cells and sand were used to represent surface potentials in eq 5.50 Cell suspensions were prepared in a similar fashion, as in the column transport experiments and the quartz sand was pulverized to colloid-sized particles and then suspended in the electrolyte solutions. The electrophoretic mobility of the bacterial cells and colloidal quartz sand in each solution was then measured using a ZetaPALS analyzer (Brookhaven Instruments Corporation). The Smoluchowski equation was used to convert electrophoretic mobility values into zeta potentials. To measure cell sizes, photos of E. coli O157:H7 cells suspended in various solutions were obtained using a Nikon Eclipse 50i microscope, equipped with a Photometric coolsnap ES digital camera and the MetaMorph software. The length and width of the cells were then determined using the ImageJ q software ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ and the equivalent C radii of the cells were calculated as ( LC W π ), where LC and WC represent the length and width of the cell, respectively.59

3. RESULTS AND DISCUSSION 3.1. Transport of E. coli O157:H7 within Sand Packs. Results from the column transport experiments show that higher percentages of E. coli O157:H7 cells could travel through the sand columns when the concentration of phosphate progressively increased from 0 to 1 mM, indicating that phosphate can promote the transport of E. coli O157:H7 (Figure 1). At a constant ionic strength of 10 mM, 55.3% ((1.6%), 32.2% ((0.3%), and 23.0% ((0.8%) of the E. coli O157:H7 cells were immobilized within the sand columns for phosphate concentrations of 0, 0.1, and 1 mM, respectively. Accordingly, the deposition rate coefficient (kd) decreased from 0.054 ((0.003) min1 to 0.019 ((0.0007) min1 when phosphate concentration increased from 0 to 1 mM (Figure 2). A similar trend was 9568

dx.doi.org/10.1021/es201132s |Environ. Sci. Technol. 2011, 45, 9566–9573

Environmental Science & Technology

Figure 1. Breakthrough concentrations of E. coli O157:H7. The ionic strength of the solution was (A) 10 mM and (B) 100 mM, respectively. Concentrations of phosphate were 0, 0.1, and 1 mM. In (B), the release of immobilized E. coli O157:H7 cells was investigated by injecting solutions with similar ionic strength (100 mM) but progressively higher phosphate concentrations.

Figure 2. Deposition rate coeﬃcient, kd (min1), as a function of ionic strength and phosphate concentration. Error bars represent standard deviation of duplicate experiments.

observed when the ionic strength was maintained at 100 mM. The fraction of immobilized E. coli O157:H7 cells decreased from 33.0% ((1.7%) to 12.1% ((1.4%), and the deposition rate coefficient decreased from 0.025 ((0.0007) min1 to 0.0005 ((0.0001) min1 when phosphate concentration increased from 0 to 1 mM. This clear increase in breakthrough is difficult to dispute, even in the absence of a mass balance. It was reported that under high pH (>8.4) conditions, the retention of E. coli O157:H7 cells within quartz sand decreased

ARTICLE

Figure 3. (A) Equivalent radius of E. coli O157:H7 cells suspended in various electrolyte solutions. (B) Zeta potential of bacterial cells (left axis) and quartz sands (right axis) under diﬀerent ionic strength and phosphate concentration conditions. Note that the zeta potential values of bacterial cells and quartz sands were plotted at diﬀerent scales.

with increasing ionic strength.24 Consistent with ﬁndings reported in previous research, results from our study show that the deposition of E. coli O157:H7 cells decreased with increasing ionic strength under a pH of 7.2, regardless of phosphate concentrations (Figures 1 and 2). For instance, in the absence of phosphate, the deposition rate coeﬃcients were 0.054 ((0.003) min1 and 0.025 ((0.0007) min1 for 10 and 100 mM of ionic strength, respectively. 3.2. XDLVO Interaction Energy Profiles. The measured contact angles of water, glycerol, and diiodomethane on E. coli O157:H7 lawns were 22.1° ((0.1°), 27.0° ((1.8°) and 63.0° ((0.7°), respectively. The values of γLW, γ+or γ for E. coli O157: H7 cells were calculated as 26.9, 47.6, and 4.82 mJ m2, respectively. Using the values previously determined for quartz in Morrow et al.,52 the Hamaker constant in eq 4 for the bacterium-waterquartz system was estimated as 1.522 1021 J. The estimated 2 value of ΔGAB ho was 24.94 mJ/m , suggesting a repulsive AB interaction between the E. coli O157:H7 cells and the quartz sand. The equivalent cell radius was around 0.85 μm and showed little variation with diﬀerent ionic strength and phosphate concentrations (Figure 3A). The zeta potentials of both the E. coli O157:H7 cells and quartz sand were negative (Figure 3B). In general, the zeta potentials of sand were ∼30 mV less negative when ionic strength increased from 10 mM to 100 mM due to the compression of electric double layer. The zeta potentials of the E. coli O157:H7 cells were close to neutral and, in contrast to the trend observed for quartz sand, an increase in ionic strength 9569

dx.doi.org/10.1021/es201132s |Environ. Sci. Technol. 2011, 45, 9566–9573

Environmental Science & Technology

Figure 4. XDLVO energy interaction proﬁles between E. coli O157:H7 cells and surface of quartz sands. The energy interactions were expressed in kT, where k is Boltzmann constant and T is absolute temperature in Kelvin.

led to a slight decrease in the zeta potential of the bacterial cells (Figure 3B). For both quartz sand and bacterial cells, phosphate decreased zeta potential values. This could be related to adsorption of phosphate onto the surface of quartz sand (e.g., through the bonding between phosphorus and oxygen at the surface of quartz) and bacterial cells, which could increase the negative surface charge under the pH conditions employed in this research.60 The calculated XDLVO energy interaction proﬁles are shown in Figure 4. Under an ionic strength of 10 mM, there was no repulsive energy barrier for cell attachment to sand surface when phosphate was absent (Figure 4A). Energy barriers were present when phosphate concentrations were either 0.1 mM or 1 mM. The energy barrier values were 0.86 kT (0.1 mM phosphate) and 1.33 kT (1 mM phosphate), respectively, where k is the Boltzmann constant and T is the absolute temperature in Kelvin. Similarly, when ionic strength was 100 mM, the energy barrier changed from absent for no phosphate to ∼53 kT for both 0.1 mM and 1 mM phosphate (Figure 4B). The energy barriers, although small, could hinder the attachment of E. coli O157:H7 cells to the surface of quartz sand and thus change a system that would otherwise be favorable for deposition, and make it unfavorable for deposition. This trend is consistent with results from the column transport experiments, which suggest that phosphate increased the transport of E. coli O157:H7 cells. Additionally, the magnitude of the energy barriers was generally higher for the 100 mM ionic strength conditions than the 10 mM ionic strength conditions. This is consistent with the observation that the transport of E. coli O157:H7 cells within the sand columns increased with higher ionic strength (Figures 1 and 2).

ARTICLE

The total XDLVO energy interaction proﬁles reﬂect the summation of the LW, EDL, and AB interactions. The LW and AB components of the overall interaction energy are independent of water chemistry parameters and remain the same for all conditions. Ionic strength, however, had a signiﬁcant impact on the zeta potential values of both the bacterial cells and the sands (Figure 3B). As the sand zeta potential became less negative when ionic strength increased from 10 to 100 mM, the zeta potential of E. coli O157:H7 cells became more negative. In response to the changes in the zeta potential values, the calculated EDL interactions between bacterial cells and quartz sand under 100 mM ionic strength conditions were more repulsive than the EDL interactions under 10 mM ionic strength conditions. 3.3. Release of Immobilized E. coli O157:H7 cells. Results from those column transport experiments using a solution that had an ionic strength of 100 mM and contained no phosphate, show that ∼33% of the bacterial cells were immobilized within the sand packs (Figure 1B). Upon the completion of the experiments, solutions with similar ionic strength but progressively higher phosphate concentrations (0.1 mM and then 1 mM) were sequentially injected into the columns to examine the release of the immobilized E. coli O157:H7 cells. The results show that the increase in phosphate concentration in the mobile aqueous phase led to the release of previously retained E. coli O157:H7 cells (Figure 1B). Maximum cell concentrations in the first and second release pulses reached ∼16% and ∼9% of influent concentration. Integration of the release pulses shows that the total quantity of released cells for the first and second pulses was equivalent to 2.5% ((0.5) and 1.4% ((0.5) of the total amount of bacterial cells injected into the columns, respectively. When combined, ∼12% of the immobilized bacterial cells were flushed out during the two-stage phosphate perturbation experiments. According to the XDLVO theory, when the repulsive barrier (and thus secondary energy minimum) is absent in the energy interaction proﬁle of bacterial cells and sands, the primary energy minimum is primarily responsible for cell deposition, which is considered irreversible.50 The observed release of immobilized E. coli O157:H7 was thus contrary to the XDLVO energy interaction proﬁles, which suggests the absence of an energy barrier and secondary minimum (100 mM, no phosphate). The energy interaction proﬁles shown in Figure 4, however, were calculated using the average cell zeta potential values. Considering the variations in the measured cell zeta potential values (Figure 3B), repulsive interaction barrier and secondary energy minimum could be present in the interaction energy proﬁles of a fraction of the bacterial cells and sand surfaces, and cell deposition within the secondary energy minimum could thus occur. For these cells, perturbations in water chemistry (i.e., phosphate concentration) could lead to their release. This is consistent with the observation that only a fraction (∼12%) of the immobilized bacterial cells was released. For the fraction of E. coli O157:H7 cells that deposited into the primarily minimum, the increase in phosphate concentration actually made the release less likely because, as suggested in Hahn et al.,61 the energy barrier for cell release increased. Our results also highlight the vital importance of the energy interaction within 5 nm from the surface in cell deposition and release.61 According to the colloid ﬁltration theory,62,63 the clean-bed deposition rate coeﬃcient can be expressed as a function of the product of the collector eﬃciency (η) and collision eﬃciency (α). As the collector eﬃciency (η) can be estimated from the correlation equation reported in the literature,63 the collision 9570

dx.doi.org/10.1021/es201132s |Environ. Sci. Technol. 2011, 45, 9566–9573

Environmental Science & Technology

Figure 5. Steric interaction energy proﬁle between E. coli O157:H7 cells and surface of quartz sands. The energy interaction was expressed in kT, where k is Boltzmann constant and T is absolute temperature in Kelvin.

eﬃciency (α) can then be estimated from the experimentally determined deposition rate coeﬃcient and collector eﬃciency. The calculated values of α were all less than 1, suggesting that the deposition condition was not completely favorable (Supporting Information (SI) Table S1). This was also consistent with our observation that the retention of E. coli O157:H7 was reversible, even if the XDLVO theory predicted the absence of energy barrier. In addition, the values of α decreased with phosphate concentrations (SI Table S1). 3.4. Steric Interactions. As shown in the previous section, the XDLVO theory and our experimental observation did not always agree with each other (e.g., although XDLVO predicted no energy barrier, the deposition of E. coli O157:H7 was reversible under the 100 mM, no phosphate condition). Steric interactions due to the presence of extracellular macromolecules on bacterial surface were reported to be partly responsible for the discrepancies between model and experimental results.46 The model results (Figure 5) indicate that the steric interaction between E. coli O157:H7 surface and quartz sand was signiﬁcantly higher than the XDLVO forces at comparable distances. This is qualitatively consistent with our observation that retention of E. coli O157:H7 is reversible when the XDLVO theory predicts the absence of energy barrier. Additionally, it has been hypothesized that the conformational changes caused by the deprotonation of bacterial surface lipopolysaccharides (LPS) carboxylic and phosphoric functional groups allowed for greater penetration of the counterions into the polymer layer, which in turn decreased the attachment of E. coli O157:H7 cells onto the surface of quartz sand.24 Elevated phosphate concentrations may have caused conformational changes of the bacterial extracellular polymers and consequently more repulsive steric interactions between the cell and quartz sand. More experimental investigations are needed to further examine the relationship between phosphate and the conformation of E. coli O157:H7 surface macromolecules, as well as the associated eﬀects on the steric interactions. 3.5. Environmental Implications. Cattle manure represents a major source of E. coli O157:H7.4,79 Manure produced in cattle farms is commonly applied to agricultural fields as a fertilizer. The E. coli O157:H7 cells that are introduced into the soil through this process could leach out and contaminate the underlying groundwater. Findings from this research suggest that phosphate, a key ingredient used in numerous domestic and industrial applications, for example, detergents, metal surface coating,

ARTICLE

fertilizers, and drinking water distribution pipe corrosion control, could potentially change a system that would otherwise be favorable for E. coli O157:H7 cells deposition, and make it unfavorable for deposition. Since groundwater is the primary source of drinking water, particularly in rural areas, the enhanced transport of E. coli O157:H7 cells could translate into greater public health risks. Phosphate concentrations in the soil may also change as a result of fertilizer (including manure) application, plant uptake, rainfall, irrigation, and plant evatranspiration. Our results indicate that increase in phosphate concentration can lead to a pulse-type release of previously immobilized E. coli O157:H7 cells. This release represents another mechanism that can cause the wider spread of E. coli O157:H7 cells within the soilgroundwater system. Our results also provide insights into the mechanisms contributing to the reduced occurrence of coliform bacteria and biofilm inhibition in drinking water distribution systems through the addition of phosphate.

’ 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.

’ AUTHOR INFORMATION Corresponding Author

*Phone: (414) 229-6891; fax: (414) 229-6958; e-mail: li@uwm. edu.

’ REFERENCES (1) Rangel, J. M.; et al. Epidemiology of Escherichia coli O157: H7 outbreaks, United States, 19822002. Emerging Infect. Dis. 2005, 11 (4), 603–609. (2) Karch, H.; Tarr, P. I.; Blelaszewska, M. Enterohaemorrhagic Escherichia coli in human medicine. Int. J. Med. Microbiol. 2005, 295 (67), 405–418. (3) Mead, P. S.; et al. Food-related illness and death in the United States. Emerging Infect. Dis. 1999, 5 (5), 607–625. (4) Smith, J. E.; Perdek, J. M. Assessment and management of watershed microbial contaminants. Crit. Rev. Environ. Sci. Technol. 2004, 34 (2), 109–139. (5) Centers for Disease Control and Prevention. Outbreak of Escherichia coli O157:H7 and campylobacter among attendees of the Washington County Fair - New York, 1999. Morbidity and Mortality Weekly Report, 1999. 43: p. 803805. (6) O’Connor, D., Report of the Walkerton Inquiry. 2002. (7) Caprioli, A.; et al. Enterohaemorrhagic Escherichia coli: Emerging issues on virulence and modes of transmission. Vet. Res. 2005, 36 (3), 289–311. (8) Chase-Topping, M. E.; et al. Risk factors for the presence of highlevel shedders of Escherichia coli O157 on Scottish farms. J. Clin. Microbiol. 2007, 45 (5), 1594–1603. (9) Valcour, J. E.; et al. Associations between indicators of livestock farming intensity and incidence of human Shiga toxin-producing Escherichia coli infection. Emerging Infect. Dis. 2002, 8 (3), 252–257. (10) Himathongkham, S.; et al. Survival of Escherichia coli O157: H7 and Salmonella typhimurium in cow manure and cow manure slurry. FEMS Microbiol. Lett. 1999, 178 (2), 251–257. (11) Zhao, T.; et al. Prevalence of enterohemorrhagic Escherichia coli O157-H7 in a survey of dairy herds. Appl. Environ. Microbiol. 1995, 61 (4), 1290–1293. (12) Cho, S.; et al. Prevalence and characterization of Escherichia coli O157 isolates from Minnesota dairy farms and county fairs. Journal of Food Prot. 2006, 69 (2), 252–259. 9571

dx.doi.org/10.1021/es201132s |Environ. Sci. Technol. 2011, 45, 9566–9573

Environmental Science & Technology (13) Faith, N. G.; et al. Prevalence and clonal nature of Escherichia coli O157:H7 on dairy farms in Wisconsin. Appl. Environ. Microbiol. 1996, 62 (5), 1519–1525. (14) Ezawa, A.; et al. High prevalence of enterohemorrhagic Escherichia coli (EHEC) O157 from cattle in selected regions of Japan. J. Vet. Med. Sci. 2004, 66 (5), 585–587. (15) Hancock, D. D.; et al. Multiple sources of Escherichia coli O157 in feedlots and dairy farms in the northwestern USA. Prev. Vet. Med. 1998, 35 (1), 11–19. (16) Kudva, I. T.; Blanch, K.; Hovde, C. J. Analysis of Escherichia coli O157: H7 survival in ovine or bovine manure and manure slurry. Appl. Environ. Microbiol. 1998, 64 (9), 3166–3174. (17) Lung, A. J.; et al. Destruction of Escherichia coli O157: H7 and Salmonella enteritidis in cow manure composting. J. Food Prot. 2001, 64 (9), 1309–1314. (18) Wang, G. D.; Zhao, T.; Doyle, M. P. Fate of enterohemorrhagic Escherichia coli O157:H7 in bovine feces. Appl. Environ. Microbiol. 1996, 62 (7), 2567–2570. (19) Gagliardi, J. V.; Karns, J. S. Leaching of Escherichia coli O157: H7 in diverse soils under various agricultural management practices (vol 66, pg 877, 2000). Appl. Environ. Microbiol. 2000, 66 (9), 4172–4172. (20) Haznedaroglu, B. Z.; et al. Relative transport behavior of Escherichia coli O157:H7 and Salmonella enterica Serovar Pullorum in packed bed column systems: inﬂuence of solution chemistry and cell concentration. Environ. Sci. Technol. 2009, 43 (6), 1838–1844. (21) Kim, H. N.; et al. Surface characteristics and adhesion behavior of Escherichia coli O157:H7: Role of extracellular macromolecules. Biomacromolecules 2009, 10 (9), 2556–2564. (22) Kim, H. N.; Walker, S. L.; Bradford, S. A. Macromolecule mediated transport and retention of Escherichia coli O157:H7 in saturated porous media. Water Res. 2010, 44 (4), 1082–1093. (23) Schinner, T.; et al. Transport of selected bacterial pathogens in agricultural soil and quartz sand. Water Res. 2010, 44 (4), 1182–1192. (24) Kim, H. N.; Bradford, S. A.; Walker, S. L. Escherichia coli O157: H7 transport in saturated porous media: Role of solution chemistry and surface macromolecules. Environ. Sci. Technol. 2009, 43 (12), 4340– 4347. (25) Castro, F. D.; Tufenkji, N. Relevance of nontoxigenic strains as surrogates for Escherichia coli O157: H7 in groundwater contamination potential: Role of temperature and cell acclimation time. Environ. Sci. Technol. 2007, 41 (12), 4332–4338. (26) Bradford, S. A.; Simunek, J.; Walker, S. L. Transport and straining of E coli O157: H7 in saturated porous media. Water Resour. Res. 2006, 42 (12), W12S12. (27) Appenzeller, B. M. R.; et al. Inﬂuence of phosphate on bacterial adhesion onto iron oxyhydroxide in drinking water. Environ. Sci. Technol. 2002, 36 (4), 646–652. (28) Park, S. J.; Lee, C. G.; Kim, S. B. The role of phosphate in bacterial interaction with iron-coated surfaces. Colloids Sur., B 2009, 68 (1), 79–82. (29) Nelson, N. O.; Mikkelsen, R. L. Polyethersulfone membrane ﬁlters for sampling soil water from in situ soils and intact soil columns for phosphate analysis. Commun. Soil Sci. Plant Anal. 2006, 37 (34), 377–388. (30) Wang, Z. Y.; Kelly, J. M.; Kovar, J. L. In situ dynamics of phosphorus in the rhizosphere solution of ﬁve species. J. Environ. Qual. 2004, 33 (4), 1387–1392. (31) Wang, Z. Y.; Wen, S. F.; Li, A. F. Analysis of phosphate in rhizosphere soil solution samples using capillary electrophoresis. Chin. J. Anal. Chem. 2010, 38 (1), 87–90. (32) Rajan, S. S. S. Comparison of phosphate fertilizers for pasture and their eﬀect on soil solution phosphate. Commun. Soil Sci. Plant Anal. 2002, 33 (1314), 2227–2245. (33) Bierman, P. M.; et al. Soil solution chemistry of sewage-sludge incinerator ash and phosphate fertilizer amended soil. J. Environ. Qual. 1995, 24 (2), 279–285. (34) Gilley, J. E.; Eghball, B.; Marx, D. B. Nutrient concentrations of runoﬀ during the year following manure application. Trans. ASABE 2007, 50 (6), 1987–1999.

ARTICLE

(35) Zhuang, J.; Jin, Y. Interactions between viruses and goethite during saturated ﬂow: Eﬀects of solution pH, carbonate, and phosphate. J. Contam. Hydrol. 2008, 98 (12), 15–21. (36) Blanford, W. J.; et al. Inﬂuence of water chemistry and travel distance on bacteriophage PRD-1 transport in a sandy aquifer. Water Res. 2005, 39 (11), 2345–2357. (37) Walczak, J. J.; et al. Inﬂuence of tetracycline resistance on the transport of manure-derived Escherichia coli in saturated porous media. Water Res. 2011, 45 (4), 1681–1690. (38) Brown, D. G.; Jaﬀe, P. R. Eﬀects of nonionic surfactants on bacterial transport through porous media. Environ. Sci. Technol. 2001, 35 (19), 3877–3883. (39) Xu, S. P.; Gao, B.; Saiers, J. E. Straining of colloidal particles in saturated porous media. Water Resour. Res. 2006, 42, (12), W12S16, DOI: 10.1029/2006WR004948 (40) Walker, S. L.; Redman, J. A.; Elimelech, M. Inﬂuence of growth phase on bacterial deposition: Interaction mechanisms in packed-bed column and radial stagnation point ﬂow systems. Environ. Sci. Technol. 2005, 39 (17), 6405–6411. (41) Kretzschmar, R.; et al. Experimental determination of colloid deposition rates and collision eﬃciencies in natural porous media. Water Resour. Res. 1997, 33 (5), 1129–1137. (42) Camesano, T. A.; Logan, B. E. Probing bacterial electrosteric interactions using atomic force microscopy. Environ. Sci. Technol. 2000, 34 (16), 3354–3362. (43) Tong, M. P.; et al. Detachment-inﬂuenced transport of an adhesion-deﬁcient bacterial strain within water-reactive porous media. Environ. Sci. Technol. 2005, 39 (8), 2500–2508. (44) Tong, M. P.; Camesano, T. A.; Johnson, W. P. Spatial variation in deposition rate coeﬃcients of an adhesion-deﬁcient bacterial strain in quartz sand. Environ. Sci. Technol. 2005, 39 (10), 3679–3687. (45) Ong, Y. L.; et al. Adhesion forces between E-coli bacteria and biomaterial surfaces. Langmuir 1999, 15 (8), 2719–2725. (46) Bayoudh, S.; et al. Quantiﬁcation of the adhesion free energy between bacteria and hydrophobic and hydrophilic substrata. Mater. Sci. Eng., C 2006, 26 (23), 300–305. (47) Bayoudh, S.; et al. Assessing bacterial adhesion using DLVO and XDLVO theories and the jet impingement technique. Colloids and Surfaces B-Biointerfaces 2009, 73 (1), 1–9. (48) Farahat, M.; et al. Adhesion of Escherichia coli onto quartz, hematite and corundum: Extended DLVO theory and ﬂotation behavior. Colloids Surf., B 2009, 74 (1), 140–149. (49) Elimelech, M. Particle deposition on ideal collectors from dilute ﬂowing suspensions—Mathematical formulation, numerical-solution, and simulations. Sep. Technol. 1994, 4 (4), 186–212. (50) Redman, J. A.; Walker, S. L.; Elimelech, M. Bacterial adhesion and transport in porous media: Role of the secondary energy minimum. Environ. Sci. Technol. 2004, 38 (6), 1777–1785. (51) Huang, X. F.; Bhattacharjee, S.; Hoek, E. M. V. Is surface roughness a “scapegoat” or a primary factor when deﬁning particlesubstrate interactions? Langmuir 2010, 26 (4), 2528–2537. (52) Morrow, J. B.; et al. Macro- and nanoscale observations of adhesive behavior for several E coli strains (O157: H7 and environmental isolates) on mineral surfaces. Environ. Sci. Technol. 2005, 39 (17), 6395–6404. (53) van Oss, C. J., Acid-base interfacial interactions in aqueous media Colloids Surf., A 1993. 78: p. 1-49. (54) Butt, H. J.; Cappella, B.; Kappl, M. Force measurements with the atomic force microscope: Technique, interpretation and applications. Surf. Sci. Rep. 2005, 59 (16), 1–152. (55) Israelachvili, J. N., Intermolecular and Surface forces, 2nd ed.; Academic Press: San Diego, CA, 1991; Vol. xxi. (56) Strauss, J.; Burnham, N. A.; Camesano, T. A. Atomic force microscopy study of the role of LPS O-antigen on adhesion of E. coli. J. Mol. Recognit. 2009, 22 (5), 347–355. (57) Neidhardt, F. C., Curtiss, R., III, Ingraham, J. L., Lin, E. C. C., Low, K. B., Magasanik, B., Reznikoﬀ, W. S., Riley, M., Schaechter, M., 9572

dx.doi.org/10.1021/es201132s |Environ. Sci. Technol. 2011, 45, 9566–9573

Environmental Science & Technology

ARTICLE

Umbarger, H. E., Eds.; Escherichia coli and Salmonella: cellular and molecular biology. 2nd ed. Washington, D.C.: ASM Press; 1996. (58) Oshea, S. J.; Welland, M. E.; Rayment, T. An atomic-force microscope study of grafted polymers on mica. Langmuir 1993, 9 (7), 1826–1835. (59) Haznedaroglu, B. Z.; Bolster, C. H.; Walker, S. L. The role of starvation on Escherichia coli adhesion and transport in saturated porous media. Water Res. 2008, 42 (67), 1547–1554. (60) Morel, F. Hering, J.G., Principles and Applications of Aquatic Chemistry; New York: Wiley, 1993; Vol. xv. (61) Hahn, M. W.; Abadzic, D.; O’Melia, C. R. Aquasols: On the role of secondary minima. Environ. Sci. Technol. 2004, 38 (22), 5915–5924. (62) Yao, K. M.; Habibian, M. M.; Omelia, C. R. Water and waste water ﬁltration—Concepts and applications. Environ. Sci. Technol. 1971, 5 (11), 1105–1112. (63) Tufenkji, N.; Elimelech, M. Correlation equation for predicting single-collector eﬃciency in physicochemical ﬁltration in saturated porous media. Environ. Sci. Technol. 2004, 38 (2), 529–536.

9573

dx.doi.org/10.1021/es201132s |Environ. Sci. Technol. 2011, 45, 9566–9573

ARTICLE pubs.acs.org/est

Heavy Metal Sorption at the Muscovite (001)Fulvic Acid Interface Sang Soo Lee,*,† Kathryn L. Nagy,† Changyong Park,‡,§ and Paul Fenter‡ †

Department of Earth and Environmental Sciences, 845 West Taylor Street MC-186, University of Illinois at Chicago, Chicago, Illinois 60607, United States ‡ Chemical Sciences and Engineering Division, Argonne National Laboratory, 9700 South Cass Avenue, Argonne, Illinois 60439, United States

bS Supporting Information ABSTRACT: The role of fulvic acid (FA) in modifying the adsorption mode and sorption capacity of divalent metal cations on the muscovite (001) surface was evaluated by measuring the uptake of Cu2+, Zn2+, and Pb2+ from 0.01 m solutions at pH 3.7 with FA using in situ resonant anomalous X-ray reﬂectivity. The molecular-scale distributions of these cations combined with those previously observed for Hg2+, Sr2+, and Ba2+ indicate metal uptake patterns controlled by cationFA binding strength and cation hydration enthalpy. For weakly hydrated cations the presence of FA increased metal uptake by approximately 60140%. Greater uptake corresponded with increasing cationFA aﬃnity (Ba2+ ≈ Sr2+ < Pb2+ < Hg2+). This trend is associated with diﬀerences in the sorption mechanism: Ba2+ and Sr2+ sorbed in the outer portion of the FA ﬁlm whereas Pb2+ and Hg2+ complexed with FA eﬀectively throughout the ﬁlm. The more strongly hydrated Cu2+ and Zn2+ adsorbed as two distinct outer-sphere complexes on the muscovite surface, with minimal change from their distribution without FA, indicating that their strong hydration impedes additional binding to the FA ﬁlm despite their relatively strong aﬃnity for FA.

’ INTRODUCTION Natural organic matter (NOM) in both dissolved and solid forms plays a signiﬁcant role in controlling the disposition of toxic heavy metal elements in the environment. Dissolved organic matter (DOM) binds metal ions in solution, changing their speciation and mobility.13 Various thermodynamic models have been developed to predict metal binding by DOM,46 although they are mostly case-speciﬁc and still need to be reﬁned for general applications.7 In many soils and sediments, NOM, including DOM, binds to mineral surfaces and can signiﬁcantly alter the uptake of metals to both the organic matter and minerals.812 Unlike the case for DOM-metal binding in solution, there are no models that describe the systematics of metal binding at the DOMmineral interface. The development of such models requires observations at the molecular scale using surface-sensitive approaches to distinguish the various modes of ionmineralDOM interactions. Muscovite mica has often been chosen as a mineral substrate in experimental systems because its basal plane, the (001) surface, is similar to the dominant surfaces of many clay minerals, which, along with many micas, are the main constituents of argillaceous soils and sediments. The surface cleaves easily to provide a large atomically ﬂat surface with a permanent negative charge (∼1e per unit cell area, AUC). The morphology of adsorbed DOM on the basal surface of muscovite has been characterized by atomic force microscopy (AFM).1316 The earlier AFM images were interpreted as showing the formation of aggregates of DOM on the r 2011 American Chemical Society

surface after reaction in 10200 mg DOM/L solutions.1315 The DOM aggregates were sparsely distributed leaving large areas of the surface exposed to solution. The coverage, size, and sorption stability of the aggregates increased with increasing cation concentration and decreasing pH, indicating that aggregate formation is related to the degree of cationorganic complexation and the hydrophobicity of the DOM. A more recent study showed that at acidic to near neutral pH DOM aggregates were located on top of a 536 Å thick organic ﬁlm that covered the surface.16 The DOM ﬁlm covered a larger area of the surface than the aggregates and therefore would be expected to aﬀect the overall sorptive capacity of the underlying substrate. X-ray reﬂectivity (XR) is an in situ, nondestructive method suitable for probing the distribution of DOM and simultaneous uptake of metal cations at the mineralwater interface. Particularly, the approach measures surface signals averaged over a relatively large area (∼1 mm2), enhancing the sensitivity to an organic ﬁlm with wide coverage and diminishing the sensitivity to sparsely distributed aggregates. Continuous organic ﬁlms about 612 Å thick were modeled from XR data obtained on the muscovite (001) surface after reaction in 100 mg/kg H2O Elliott Soil Fulvic Acid II (ESFA) solutions at pH 26.17 At pH 3.7, Received: April 18, 2011 Accepted: October 5, 2011 Revised: September 15, 2011 Published: October 05, 2011 9574

dx.doi.org/10.1021/es201323a | Environ. Sci. Technol. 2011, 45, 9574–9581

Environmental Science & Technology

ARTICLE

Table 1. Best-Fit Model Parameters of the RAXR Dataa sampleb

χ2 (R-factor)

inner-sphere peak c cIS

zIS

uIS

outer-sphere peak c cOS uOS

zOS

zbroad

broad peak c cbroad

ubroad

metal only muSr53.5

1.35 (0.006)

1.38(f)

0.014(5)

0.29(f)

4.52(6)

0.19(1)

0.66(7)

8.43(31)

0.06(2)

1.38(43)

muHg12.0 24

1.15 (0.005)

0.62(5)

0.06(1)

0.13(f)

3.58(4)

0.14(1)

0.68(6)

9.57(36)

0.06(2)

2.26(55)

muPb103.7 32

1.54 (0.012)

1.90(3)

0.29(4)

0.27(11)

4.39(42)

0.19(5)

1.60(43)

9.58(61)

0.05(1)

2.00(f)

muCu103.7 32 muZn103.7 32

1.32 (0.007) 1.23 (0.007)

0.14(19) 1.28(72)

0.03(1) 0.01(1)

0.29(f) 0.29(f)

3.97(4) 3.90(4)

0.28(4) 0.20(4)

0.92(11) 0.40(10)

7.69(106) 5.37(75)

0.19(7) 0.14(5)

3.45(88) 1.77(38)

muSr5ESFA3.5 21

1.18 (0.008)

1.38(f)

0.03(1)

0.25(f)

muHg1ESFA2.024

1.22 (0.004)

metal + FA 4.40(2)

0.20(1)

0.17(5)

6.66(9)

0.20(3)

1.28(11)

2.28(7)

0.15(2) e

1.28(11)

6.28(27)

0.47(5)

4.71(25)

inner-sphere peak c

outer-sphere peak c

diﬀuse proﬁled

zIS

cIS

uIS

zOS

cOS

uOS

zdiﬀuse

cdiﬀuse

k1diﬀuse 5.4(15)

muPb10SRFA3.7

1.28 (0.008)

1.91(2)

0.27(3)

0.20(9)

2.91(12)

0.42(4) e

1.46(7)

7.24(33)

0.21(5)

muCu10SRFA3.7

1.08 (0.009)

0.14(f)

0.05(1)

0.29(f)

3.71(10)

0.22(3)

0.67(12)

4.61(54)

0.20(8)

2.9(11)

muZn10SRFA3.7

1.17 (0.013)

1.28(f)

0.00(1)

0.29(f)

4.07(5)

0.15(1)

0.58(7)

4.97(39)

0.28(11)

4.9(20)

a The numbers in parentheses indicate 1σ uncertainties of the last digit(s) of the ﬁtting parameters. f indicates parameter ﬁxed during ﬁtting. b mu: muscovite, SRFA: Suwannee River Fulvic Acid, ESFA: Elliott Soil Fulvic Acid II. The subscripted number after each metal name indicates the metal concentration in units of 103 m. All FA solutions contained 100 mg/kg of a dissolved FA. The number at the end of each sample name indicates the solution pH. c zj, cj, and uj: height (Å), occupancy (atom per unit cell area, AUC), and distribution width (Å) of a Gaussian peak j. d zdiﬀuse, cdiﬀuse, k1diﬀuse: height of the ﬁrst peak (Å), total occupancy (atom/AUC), and debye length (Å) of the broad diﬀuse proﬁle (see Supporting Information). e Modeled as a part of metalFA complexes.

Ba2+ adsorbed to the muscovite surface from a premixed BaCl2ESFA solution mostly as an apparent inner-sphere (IS) complex.18 The electron density of the fulvic acid (FA) ﬁlm was higher with Ba2+ than without Ba2+, implying that some Ba2+ was located within the ﬁlm. However, the amount and distribution of Ba2+ in the ﬁlm could not be quantiﬁed because XR is not element-speciﬁc. Resonant anomalous X-ray reﬂectivity (RAXR)19 was applied to probe the distribution of Sr2+, a cation having aﬃnity for organic matter similar to that of Ba2+,20 at the muscoviteESFA interface.21 The results showed that about 1040% of the adsorbed cation accumulated in the outer part of a FA ﬁlm at pH 3.55.5 while the remaining Sr2+ adsorbed on the muscovite surface as an outer-sphere (OS) complex.21 Compared to these cations, Hg2+, a cation with a greater aﬃnity for organic matter,20,22,23 did not form any discrete IS or OS complex on the muscovite surface reacted in a premixed Hg(NO3)2 and ESFA solution at pH 2. Instead, a large fraction of Hg was incorporated within the FA ﬁlm, resulting in increases both in metal uptake (by ∼140% compared to that without FA) and in ﬁlm thickness (by ∼20% compared to that without Hg).24 The way in which dissolved metals interact with the muscovite (001)solution interface in the presence of DOM must depend on speciﬁc properties of the cations, but those trends are unclear at present. Here we present new XR and RAXR data for Pb, Zn, and Cu uptake onto the muscovite (001) surface from premixed metalFA solutions at acidic pH, a condition typically observed in carbonate-depleted clayey organic-rich soil layers, e.g., in some humults ultisols or spodozols,25,26 or in many metal-contaminated environments, e.g., acid mine drainages27,28 and associated pit lakes.29 The results for total coverage and adsorbed cation distribution are combined with those observed previously for Ba, Sr, and Hg and characterized systematically as a function of the

relative cation aﬃnity for organic matter and the cation hydration strength.

’ EXPERIMENTAL SECTION Sample Preparation. Experimental solutions were prepared by dissolving a high-purity (g99.99%) nitrate of Cu, Zn, Sr, or Pb (Aldrich Chemical Co., Inc.), Suwannee River Fulvic Acid (SRFA) from the International Humic Substances Society (IHSS), or both together in deionized water (DIW; ∼18.2 MΩ). Solutions containing SRFA were prepared with a relatively high FA content (100 mg/kg H2O; i.e., mg of dry FA in 1 kg of DIW) to ensure the formation of a FA film on the muscovite surface over the time of the reaction (>2 h).16,17 High concentrations of metal cation (510 103 m; molality) were used to minimize competitive effects of hydronium and other cations sourced from the muscovite and FA. The high concentrations also controlled the ionic strength of the solutions without addition of other electrolytes, which could increase the complexity of the system. Therefore, the observed structural changes are expected to derive purely from changing cationFA interactions. The pH of the SRFA solution without any adjustment was 3.7, close to the log K1 value (3.81) of the FA.30 The pH of all other solutions was also adjusted to 3.7 using high-purity 0.1 M HNO3 except that for muscovite reacted in a 5 103 m Sr(NO3)2 solution without FA (muSr53.5, Table 1) whose pH was 3.5 for comparison to results from the previous experiment conducted in a premixed 5 103 m Sr(NO3)2 and 100 mg/kg ESFA solution at the same pH (muSr5ESFA3.5, Table 1).21 Prepared solutions were stored in brown polypropylene bottles in a refrigerator until used. For each experiment, a gem-quality single crystal muscovite (Asheville Schoonmaker Mica Company) was cleaved to expose a fresh (001) surface and immersed vertically in a 50-mL centrifuge 9575

dx.doi.org/10.1021/es201323a |Environ. Sci. Technol. 2011, 45, 9574–9581

Environmental Science & Technology

ARTICLE

Figure 2. Total electron-density proﬁle derived from the best-ﬁt model of XR for muscovite (001) in contact with a 100 mg/kg SRFA solution at pH 3.7 (muSRFA3.7). Those for muscovite (001) in a 100 mg/kg ESFA solution at pH 3.7 (muESFA3.7)18 and a 100 mg/kg PPFA solution at pH 3.6 (muPPFA3.6)18 are plotted for comparison. The electron density was normalized to that of bulk water, and plotted with a band indicating the 1σ uncertainty as a function of height from the surface. The proﬁle below 0 Å (corresponding to the muscovite substrate beneath the top oxygen plane) is not shown. Figure 1. Normalized RAXR signal using the resonant amplitude normalization [(|Ftot(q,E)|2 |FNR(q)|2)/(2|FNR(q)|), where Ftot and FNR are total and nonresonant structure factors, respectively] for muscovite in solutions containing heavy metals (Pb, Cu, and Zn in comparison to Hg and Sr)21,24 in the absence (cyan circles) and presence (pink squares) of FA. Spectra are oﬀset by 2 units, and open and ﬁlled symbols are used alternately for easier visual comparison. Each spectrum is labeled with the q value (Å1) where the spectrum was measured. The curves through data points are calculated intensities derived from the best-ﬁt models. Solid (for solid symbol data) and dashed (for open symbol data) gray horizontal lines guide theoretical reﬂectivity when there is no resonant atom at the interface. The deviation of the RAXR signal from the reference lines is proportional to the ion coverage at small q.24

tube containing one of the solutions for at least 2 h,17 after which the wet muscovite was transferred to a thin-film sample cell for XR measurements as described previously.17,18,21,24 Specular X-ray Reflectivity. Measurements were made in situ at beamlines 6-ID-B (MU-CAT) and 11-ID-D (BESSRC), Advanced Photon Source, Argonne National Laboratory (Figure S1 in Supporting Information (SI)). X-ray experiments on samples containing FA were conducted in the dark. The stability of the experimental samples over the measurement time of approximately 1 h was confirmed by periodically measuring reflectivity at two reference points defined by momentum transfer values q = 0.85 and 1.83 Å1. Only experiment muPb10SRFA3.7 (muscovite in 10 103 m Pb and SRFA at pH 3.7; Table 1) showed a small but significant (∼6%) variation in reference point reflectivities. Measurements for muCu10SRFA3.7 and muZn10SRFA3.7 were duplicated with separately prepared samples. The XR data were fit with parameterized models (SI text) to obtain optimized structures for the total electron density profiles at the interfaces. Resonant Anomalous X-ray Reflectivity. Measurements were obtained by scanning the photon energy (E) near the X-ray absorption edge of the target metal at fixed q values (from 0.25 0.35 to 3.34.3 Å1) (Figure 1). One set of data typically included 1020 spectra measured over 36 h. The stability of the experimental system was monitored by periodically repeating measurements at low q values (0.250.57 Å1). The RAXR signals at the reference points varied less than 5% in amplitude, except for muPb10SRFA3.7, which showed a slow but continuous decrease

in signal amplitude (e.g., ∼20% decline after 3 h). This implies that adsorbed Pb was mobile during X-ray exposure. A similar result was observed after 2 h of X-ray measurements of adsorbed Hg on a pre-FA-coated muscovite (001) surface at pH 2.0.24 The RAXR data were ﬁt by a model with two cation positions (represented as inner-sphere (IS) and outer-sphere (OS) positions in Table 1) near the muscovite surface, followed by a broad ion proﬁle. Initial cation distributions were guided by a semiquantitative cation proﬁle derived from model-independent analysis31 (Figures S2 and S3). For data collected in a pure metal-salt solution, the broad proﬁle was modeled using a single Gaussian peak.24,32 With FA, a slightly better quality of ﬁt could be obtained using a broad asymmetric peak simulated by overlapping a series of equally spaced Gaussian peaks whose occupancies decrease exponentially as a function of distance from the surface (SI text).

’ RESULTS AND DISCUSSION Fulvic Acid Sorption on the Muscovite (001) Surface. The total electron-density profile for muSRFA3.7 (Suwannee River FA adsorbed on muscovite at pH 3.7) has a broad peak near 2.5 Å followed by a broader profile with the maximum electron density located near 4.6 Å (Figure 2). This pattern is similar to those determined previously for FA sorbed on muscovite from a 100 mg/kg Elliott Soil FA II (ESFA) solution at pH 3.7 (muESFA3.7) and a 100 mg/kg Pahokee Peat FA (PPFA) solution at pH 3.6 (muPPFA3.6).18 The broad peak near the surface has a lower electron density in muSRFA3.7 than in muESFA3.7 and muPPFA3.6. This suggests that the fraction of SRFA adsorbed directly on the surface either has a lower electron density or covers less of the surface than the similar fraction of ESFA or PPFA. The higher electron density for the PPFA sample may be explained by its higher ash content (4.61 wt %) compared to that for SRFA (0.46 wt %) or ESFA (1.00 wt %).18,30 The next peak in the electron-density proﬁle of muSRFA3.7 extends to about 7 Å from the surface and is narrower and closer to the surface compared to that of muESFA3.7. The muSRFA3.7 proﬁle also has a third electron-dense region above ∼7 Å, indicating that a fraction of sorbed SRFA molecules extends farther from the surface. A similar pattern is observed in muPPFA3.6, but is relatively less prominent in muESFA3.7. 9576

dx.doi.org/10.1021/es201323a |Environ. Sci. Technol. 2011, 45, 9574–9581

Environmental Science & Technology

ARTICLE

Figure 3. Fractional change in heavy metal uptake at the muscovite (001)solution interface in the presence of 100 mg/kg FA in relation to the calculated molar ratio between metalFA complexes and free metal cation in solution and the metal cation hydration enthalpy. The molar ratio [Me2+ FA]/[Me2+] was calculated based on the solution composition (Table S2). Fractional changes in cation coverage (cMeFA cMe)/cMe (%)) (see text) were calculated based on the RAXR results of ﬁve metals (marked in a pink dashed pentagon; except Ba). The color map with contours, generated by the best-ﬁt empirical expression (eq 1), illustrates the trends in how the hydration enthalpy of the cation and aﬃnity of the cation for FA under the experimental conditions control metal uptake at the interface. Total electron-density proﬁles derived from the best-ﬁt models of the interfacial structure with adsorbed metal in the absence (blue dashed line) and presence (red solid line) of FA are also shown. Refer to Table 1 for sample codes, and Figure 2 for descriptions of the axes for proﬁles associated with individual elements. The electron-density proﬁles measured in FA solutions without metals (green dot-dashed line) are plotted for comparison. The element-speciﬁc proﬁles of adsorbed metals in the absence and presence of dissolved FA are shown in sky-blue and pink shaded areas, respectively. Note that the distribution of Ba is estimated based on the diﬀerence between total electron-density proﬁles (e.g., muBa5ESFA3.7 muESFA3.7) and was not determined by RAXR.

Table 2. Characteristics of Fulvic Acid Films Adsorbed on the Muscovite (001) Surface in the Absence and Presence of Heavy Metals average layer density (FWeq) samplea muSRFA3.7 muPb10SRFA3.7

layer thickness (Å) 6.0(5.96.3) 25.8(24.726.5)

with metals

without metals 1.02(0.971.07)

1.29(1.251.32) 1.11(1.071.14)

muCu10SRFA3.7

8.0(7.98.1)

1.09(1.041.14) 0.99(0.941.04)

muZn10SRFA3.7

8.1(8.08.2)

1.12(1.111.16) 1.02(0.961.07)

muESFA2.017

12.1(11.912.1)

1.12(1.071.16)

muHg1ESFA2.024

14.9(14.714.9)

1.34(1.291.39) 1.13(1.081.18)

muESFA3.717,18 muSr5ESFA3.521

7.2(7.17.2) 10.9(10.911.0)

1.02(0.961.07) 1.08(1.051.12) 0.99(0.961.03)

a

Refer to Table 1 for sample codes. The numbers in parentheses are the ranges of the values calculated from the lower (σ) and upper (+σ) limits of electron-density proﬁles derived from the best-ﬁt models.

All three FA samples demonstrate the common feature of an approximately 10 Å thick ﬁlm with a structure composed of a

directly adsorbed fraction of FA having a higher electron density and a remnant tail that has a lower electron density. Effect of Fulvic Acid on Pb Uptake on Muscovite. The amount of adsorbed Pb is enhanced in the presence of FA as shown by the RAXR data for muPb10SRFA3.7 vs muPb103.732 (Table 1). The total coverage of Pb [0.90(7) atom per unit cell area, AUC = 46.72 Å2]33 in the presence of FA is about twice as high as that (∼0.5 atom/AUC) needed to compensate the muscovite surface charge (∼1e/AUC), implying that some Pb is bonded to sorbed FA molecules. Comparing the electrondensity profiles of muPb103.7 and muPb10SRFA3.7 shows that the distribution of the additional Pb matches that of the FA film, indicating a direct association of Pb with the sorbed FA (Figure 3). The total electron density of the FA film in muPb10SRFA3.7 is also higher than that in muSRFA3.7 (Table 2) because of the presence of electron dense Pb in the layer. A large fraction of Pb is adsorbed as an apparent IS complex to the muscovite surface, while it is not possible to distinguish an OS complex because the modeled electron density of this species would be superimposed on that of PbFA complexes (Figure 3). The muPb10SRFA3.7 profile shows a small increase in the Pb distribution located immediately 9577

dx.doi.org/10.1021/es201323a |Environ. Sci. Technol. 2011, 45, 9574–9581

Environmental Science & Technology adjacent to the surface (i.e., <1 Å). The feature derives from the tail of the broad Gaussian peak at ∼2.9 Å, and is likely an artifact of using this simple model to analyze the data. The presence of two distinct forms of Pb (IS Pb and Pb within the FA ﬁlm) agrees in general with observations from total external reﬂection X-ray standing wave (TER-XSW) measurements that showed sorbed Pb both on the surfaces of hematite and corundum and within preformed bioﬁlms (Burkholderia cepacia) at >106 M Pb in solution.34 The TER-XSW approach is better suited for investigating the distribution of an ion within a large-scale matrix (such as bioﬁlms), but it has intrinsically a lower spatial resolution than RAXR (e.g., ∼25 Å vs ∼1 Å, respectively). Therefore, it is unknown if the fraction of Pb reported near the surface34 represents a true surface-adsorbed species. The fractional enhancement of cation uptake by FA (compared to that without FA) is larger for Pb than Sr (Table 1 and Figure 3). The dissimilar atomic-scale distributions of the cations in muPb10SRFA3.7 and muSr5ESFA3.5 indicate that the uptake is controlled by diﬀerent sorption mechanisms. Whereas these two cations have similar hydration enthalpies, Pb has a stronger aﬃnity for FA and forms organic complexes in the solution. Strontium does not complex signiﬁcantly with FA in the solution. The enhanced sorption of Sr is, unlike Pb, localized in the outer part of the FA ﬁlm, suggesting that, although Sr and FA were mixed in solution prior to adsorption, the Sr adsorbed independently after the FA.21 A similar result using only XR data had been observed for Ba, which is slightly less strongly hydrated than Sr but has a similar aﬃnity for FA.18 The total electron-density proﬁle above the muscovite surface for muBa5ESFA3.7 is higher than that of muESFA3.7, and can be attributed partly to an accumulation of Ba in the layer (Figure 3). The distribution of Pb throughout the FA ﬁlm indicates that its enhanced uptake results mostly from sorption of metalorganic complexes, similar to results for Hg at pH 2.0 (muHg1ESFA2.0) (Figure 3).24 The total coverage of Hg [0.62(5) Hg/AUC], however, was smaller than that of Pb, in part because of greater competition by hydronium for sorption sites at pH 2.0 compared to pH 3.7 (Table 1). Also, little Hg occurred as distinct IS and OS species. The HgFA complexes would have been protonated at pH 2.0, resulting in decreased electrostatic repulsion at the surface and increased hydrophobicity of the complexes.2,17,35,36 Effect of Fulvic Acid on Copper and Zinc Uptake on Muscovite. The element-specific electron-density profiles of muCu10SRFA3.7 and muZn10SRFA3.7 show broad peaks at about 4 Å from the surface (Figure 3). The heights and occupancies of these peaks match those of the adsorbed OS (OSads) complex of Cu or Zn from solutions without FA (i.e., muCu103.7 and muZn103.7, respectively),32 indicating that the OSads species is not altered by the presence of FA. For each metal there is a second broader peak extending to ∼10 Å from the surface. This broader distribution is similar to that of an extended OS (OSext) complex observed in muCu103.7 and muZn103.7 (i.e., without FA).32 The position of the OSext species in the absence of FA is stabilized by multiple layers of water molecules, including those in higher-order hydration shells of the cation and the hydration layer at the muscovite surface.32 The lack of a signiﬁcant increase in metal uptake in the presence of FA was unexpected especially for Cu2+ which has been reported to bind strongly to organic matter.4,37 Calculations based on the nonideal competitive adsorption-Donnan model show that the amount of Cu2+ bonded to SRFA in solution at

ARTICLE

pH 3.7 should be larger than that of Sr2+ at pH 3.5 (Table S3). Zn2+ has a slightly smaller aﬃnity for organic matter compared to Cu2+,38 but still has a larger aﬃnity than Ba2+ and Sr2+, especially for binding to phenolic or thiolic groups (Table S2). The cations Cu2+ and Zn2+ may occur mainly in nonsorbing metalorganic complexes that remain in solution during the experiments.39,40 Extended X-ray absorption ﬁne structure (EXAFS) spectroscopy of Zn in organic-rich soils with a relatively high metal content (0.510 mg Zn/g of soil) at near neutral pH (5.67.3) showed that most Zn is bonded to organic matter with ﬁrst-shell coordination ligands that are a mixture of oxygen (or nitrogen) and sulfur.41 Copper(II) complexed with multiple ligands (e.g., malate or malonate groups) and formed ﬁve- and six-membered chelate ring structures in solutions containing DOM (∼300500 g C/kg) and Cu (100 6500 mg/kg) at pH 4.5 and 5.5 as determined by EXAFS spectroscopy.42 If the positive charge of each metal cation is shielded by ligands, then sorption to the muscovite would be controlled by the net charge and hydrophobicity of the cationorganic complex. This phenomenon also helps to explain the RAXR results for Hg, in which the enhanced metal coverage observed at pH 2.0 declined at higher pH24 in part because of deprotonation of functional groups that did not bind Hg and in part because of increased hydrolysis of Hg to form neutral inorganic species. It is also possible that some Zn2+ and Cu2+ formed complexes within sorbed aggregates1315 which occurred at low surface coverage below the detection limits of XR and RAXR. A large fraction of Zn and Cu in the solutions should have been present in simple (hydrated) ionic form according to thermodynamic calculations (Table S2). Therefore, some aqueous Zn2+ and Cu2+ might have interacted with the sorbed FA ﬁlm, similarly to Ba2+ and Sr2+.18,21 Part of the sorbed Zn and Cu identiﬁed as OSext species might be bonded within the outer region of FA; however, the locations of the peaks assigned to OSext species do not match well the location of the outer region of the ﬁlm, supporting the interpretation that the majority of the cations are independently adsorbed species. It is not possible to determine why Zn2+ and Cu2+ sorbed to the FA ﬁlm to a lesser extent than Sr2+ and Ba2+ based on the limited data in the current study. The reason may be related in part to the diﬀerence in hydration strength of the cations. The results show that the total amount of metals, especially within the FA layer, increases as the magnitude of the cation hydration enthalpy decreases (i.e., the hydration strengths become weaker) (Figure 3). This relationship indicates that strongly hydrated cations are less sorptive to the SRFA ﬁlm than to the muscovite (001) surface at pH 3.7. Changes in the Internal Structure of Adsorbed FA by Adsorbed Metal Cations. The presence of metal cations leads to changes in the thickness, layer density, and detailed structure of the FA film on the muscovite (001) surface. In muCu10SRFA3.7 and muZn10SRFA3.7, the total electron density at 68 Å above the surface is greater than in muCu103.7 and muZn103.7 (Figure 3). This height range does not match the heights of any adsorbed Cu or Zn species, indicating that it is a region dominated by adsorbed FA. The position is 12 Å higher above the surface than the outer region of the film in muSRFA3.7, suggesting that the presence of the OS Cu and Zn complexes effectively increases the thickness of the organic film (Table 2). The overall average layer density increases mainly because of the added electron density from the cations. The layer densities corrected for the occupancy of the cations are 0.991.02 FWeq 9578

dx.doi.org/10.1021/es201323a |Environ. Sci. Technol. 2011, 45, 9574–9581

Environmental Science & Technology

ARTICLE

(1 FWeq corresponds to the electron density of bulk water = 0.33 e/Å3)18 similar to that for muSRFA3.7 (1.02 FWeq) (Table 2). The electron density g10 Å above the surface is slightly higher than that when metal cations are absent, suggesting that the presence of small amounts of OSext species can induce some additional sorption of FA at the interface. Interfacial Hg and Pb, considered to be largely in the form of metalorganic complexes, also increased the electron density (g1.3 FWeq) of the FA ﬁlms (Table 2). Similarly to the experiments for Cu and Zn, the densities adjusted by subtracting the contributions from the metals are comparable to those for the pure FA layers, indicating this increase results from the presence of the metals alone (Table 2). However, the greater layer thicknesses (g15 Å) must result from additional FA on the surface. For Sr and Ba, the prominent enhancement in the electron density observed in the outer part of the FA ﬁlm conﬁrmed the incorporation of these cations in this speciﬁc location. The overall ﬁlm thickness increased slightly (Table 2), indicating that the adsorbed metals attracted more FA to the surface. Empirical Model of Metal Uptake at the MineralNOM Interface. At the muscoviteFA solution interface three effective ligands, the waters of hydration, the FA, and the mineral surface, compete to bind metal cations. Effects of this competition are characterized here for the first time for a group of divalent cations in terms of the metal coverage and its molecularscale distribution at the interface. The fractional changes in the amount of cation uptake relative to the system without FA [= (cMeFA cMe)/cMe, where cMeFA and cMe are the metal ion coverages (atom/AUC) with and without FA, respectively] for all metals (except Ba) were characterized by RAXR, and can be expressed empirically in terms of the dependence on cationFA binding strength and cation hydration enthalpy: ðcMeFA cMe Þ=cMe ¼ 3:2ðlog10

’ ASSOCIATED CONTENT

KFA 0:49Þ

1:011ðjΔHhyd j 2130Þ þ 0:62

DOM molecules, which could sorb diﬀerently to mineral surfaces than other fractions of DOM.22,23 For less strongly hydrated cations, the fractional coverage increases in the order Ba2+ ≈ Sr2+ < Pb2+ < Hg2+, in sequence of increasing cation aﬃnity for FA. This trend is related, at the molecular-scale, to a transition from additional uptake of metal in the outer part of the FA ﬁlm (presumably by electrostatic eﬀects) to uptake eﬀectively throughout the entire ﬁlm via sorption of metalFA complexes. At the same time, sorbed FA does not appear to alter the binding of the strongly hydrated Cu and Zn to the mineral surface, in these cases demonstrating the importance of metal binding to the bare mineral surface. Weakly hydrated cations with a smaller aﬃnity for organic matter tend to bind electrostatically to both the negatively charged muscovite surface and the negatively charged functional groups of the FA ﬁlm formed on the muscovite surface. In soils or sediments where the concentrations of these cations are relatively small these cations may be readily displaced by background cations, such as Ca2+ or Na+. A moderately hydrated cation with a larger aﬃnity for organic matter can also sorb as organic complexes, and may be less exchangeable and therefore less mobile, at least at low pH. As pH increases these metalorganic complexes may be released to solution owing to increased electrostatic repulsion between the muscovite surface and the sorbed NOM as a consequence of deprotonation reactions of the NOM. These results conﬁrm that the complex interactions among ions, NOM, and mineral surfaces can be monitored systematically. This and similar types of molecular-scale characterization will be essential in the development of more robust predictive models for assessing the transport of toxic metals in nature.

bS ð1Þ

where KFA = [Me2+FA]/[Me2+] is the molar ratio of cation FA complexes to the free cation species in solution calculated based on the reported metalligand binding constants (Table S2) and ΔHhyd is the cation hydration enthalpy (kJ/mol). We note that KFA and ΔHhyd are not necessarily independent variables. The apparent high precision of some model parameters (SI) is mostly a result of the limited number of data points. This simple empirical equation may not explain fully the complex nature of metalFAmuscovite interactions. However, the equation effectively reproduced the observed data trends [with a good quality of fit (i.e., χ2 < 1 and R-factor <1%)]. Equation 1 summarizes the observed trends in intrinsic metalFAmineral interactions in acidic pH solutions with relatively high concentrations of dissolved metal and DOM, and may be useful for predicting uptake for other divalent cations such as Ni2+ or Cd2+ under similar conditions (for example, acid mine drainage, pit lakes, or water in pores in the top or leached layers of contaminated soils).2527,29 In practice, the empirical nature of this model limits its direct application to other systems. For example, in sediments in contact with fresh waters, such as lakes or rivers, where metal content is relatively low, metalDOM binding can be controlled by minor functional groups whose aﬃnity for the metal cation can be larger than the more abundant carboxylic and phenolic groups. Similarly, the mobility of Hg2+ may be determined by binding to thiol-rich

Supporting Information. Detailed information about materials, XR and RAXR data and models, metal speciation calculation, description of lead and copper(II) adsorption speciation, and development of an empirical metal uptake model. This material is available free of charge via the Internet at http:// pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: (630)252-6679; fax: (630)252-9570; e-mail: sslee@ anl.gov; present address: Chemical Sciences and Engineering Division, Argonne National Laboratory, Argonne, Illinois 60439, United States. Present Address §

HP-CAT, Geophysical Laboratory, Carnegie Institution of Washington, Argonne, IL 60439.

’ ACKNOWLEDGMENT This work was funded by the Geosciences Research Program, Oﬃce of Basic Energy Sciences, United States Department of Energy under Grant DE-FG02-06ER15364 to the University of Illinois at Chicago and under Contract DEAC02-06CH11357 to UChicago Argonne, LLC as operator of Argonne National Laboratory, and the National Science Foundation under grant EAR-0455938 to the University of 9579

dx.doi.org/10.1021/es201323a |Environ. Sci. Technol. 2011, 45, 9574–9581

Environmental Science & Technology Illinois at Chicago. Use of the Advanced Photon Source was supported by the U.S. Department of Energy, Oﬃce of Science, Oﬃce of Basic Energy Sciences, under Contract DE-AC0206CH11357 to UChicago Argonne, LLC as operator of Argonne National Laboratory.

’ REFERENCES (1) Raspor, B.; N€urnberg, H. W.; Valenta, P.; Branica, M. Studies in seawater and lake water on interactions of trace metals with humic substances isolated from marine and estuarine sediments. I. Characterisation of humic substances. Mar. Chem. 1984, 15, 217–230. (2) Sposito, G. The Surface Chemistry of Soils; Oxford University Press: New York, 1984. (3) Ticknor, K. V.; Vilks, P.; Vandergraaf, T. T. The eﬀect of fulvic acid on the sorption of actinides and ﬁssion products on granite and selected minerals. Appl. Geochem. 1996, 11, 555–565. (4) Kinniburgh, D. G.; Milne, C. J.; Benedetti, M. F.; Pinheiro, J. P.; Filius, J.; Koopal, L. K.; van Riemsdijk, W. H. Metal ion binding by humic acid: Application of the NICA-Donnan model. Environ. Sci. Technol. 1996, 30, 1687–1698. (5) Gustafsson, J. P. Modeling the acid-base properties and metal complexation of humic substances with the Stockholm Humic Model. J. Colloid Interface Sci. 2001, 244, 102–112. (6) Filius, J. D.; Meeussen, J. C. L.; Lumsdon, D. G.; Hiemstra, T.; van Riemsdijk, W. H. Modeling the binding of fulvic acid by goethite: The speciation of adsorbed FA molecules. Geochim. Cosmochim. Acta 2003, 67, 1463–1474. (7) Lenoir, T.; Matynia, A.; Manceau, A. Convergence-optimized procedure for applying the NICA-Donnan model to potentiometric titrations of humic substances. Environ. Sci. Technol. 2010, 44, 6221– 6227. (8) D€uker, A.; Ledin, A.; Karlsson, S.; Allard, B. Adsorption of zinc on colloidal (hydr)oxides of Si, Al and Fe in the presence of a fulvic acid. Appl. Geochem. 1995, 10, 197–205. (9) Frimmel, F. H.; Huber, L. Inﬂuence of humic substances on the aquatic adsorption of heavy metals on deﬁned mineral phases. Environ. Int. 1996, 22, 507–517. (10) Schroth, B. K.; Sposito, G. Eﬀect of landﬁll leachate organic acids on trace metal adsorption by kaolinite. Environ. Sci. Technol. 1998, 32, 1404–1408. (11) B€ackstr€om, M.; Dario, M.; Karlsson, S.; Allard, B. Eﬀects of a fulvic acid on the adsorption of mercury and cadmium on goethite. Sci. Total Environ. 2003, 304, 257–268. (12) Strathmann, T. J.; Myneni, S. C. B. Eﬀect of soil fulvic acid on nickel(II) sorption and bonding at the aqueous-boehmite (γ-AlOOH) interface. Environ. Sci. Technol. 2005, 39, 4027–4034. (13) Balnois, E.; Wilkinson, K. J.; Lead, J. R.; Buﬄe, J. Atomic force microscopy of humic substances: Eﬀects of pH and ionic strength. Environ. Sci. Technol. 1999, 33, 3911–3917. (14) Maurice, P. A.; Namjesnik-Dejanovic, K. Aggregate structures of sorbed humic substances observed in aqueous solution. Environ. Sci. Technol. 1999, 33, 1538–1541. (15) Namjesnik-Dejanovic, K.; Maurice, P. A. Conformation and aggregate structures of sorbed natural organic matter on muscovite and hematite. Geochim. Cosmochim. Acta 2000, 65, 1047–1057. (16) Gibson, C. T.; Turner, I. J.; Roberts, C. J.; Lead, J. R. Quantifying the dimensions of nanoscale organic surface layers in natural waters. Environ. Sci. Technol. 2007, 41, 1339–1344. (17) Lee, S. S.; Fenter, P.; Park, C.; Nagy, K. L. Fulvic acid sorption on muscovite mica as a function of pH and time using in-situ X-ray reﬂectivity. Langmuir 2008, 24, 7817–7829. (18) Lee, S. S.; Nagy, K. L.; Fenter, P. Distribution of barium and fulvic acid at the mica-solution interface using in-situ X-ray reﬂectivity. Geochim. Cosmochim. Acta 2007, 71, 5763–5781. (19) Park, C.; Fenter, P. A.; Sturchio, N. C.; Regalbuto, J. R. Probing outer-sphere adsorption of aqueous metal complexes at the oxide-water

ARTICLE

interface with resonant anomalous X-ray reﬂectivity. Phys. Rev. Lett. 2005, 94, 076104–14. (20) Martell, A. F.; Smith, R. M. Critical Stability Constants: Other Organic Ligands; Plenum Press: New York, 1976; Vol. 3, p 495. (21) Lee, S. S.; Park, C.; Fenter, P.; Sturchio, N. C.; Nagy, K. L. Competitive adsorption of strontium and fulvic acid at the muscovitesolution interface observed with resonant anomalous X-ray reﬂectivity. Geochim. Cosmochim. Acta 2010, 74, 1762–1776. (22) Haitzer, M.; Aiken, G. R.; Ryan, J. N. Binding of mercury(II) to dissolved organic matter: The role of the mercury-to-DOM concentration ratio. Environ. Sci. Technol. 2002, 36, 3564–3570. (23) Skyllberg, U. Competition among thiols and inorganic sulﬁdes and polysulﬁdes for Hg and MeHg in wetland soils and sediments under suboxic conditions: Illumination of controversies and implications for MeHg net production. J. Geophys. Res. 2008, 113, G00C031-14. (24) Lee, S. S.; Nagy, K. L.; Park, C.; Fenter, P. Enhanced uptake and modiﬁed distribution of mercury(II) by fulvic acid on the muscovite (001) surface. Environ. Sci. Technol. 2009, 43, 5295–5300. (25) Herbert, R. B. J. Partitioning of heavy metals in podzol soils contaminated by mine drainage waters, Dalarna, Sweden. Water, Air Soil Pollut. 1997, 96, 39–59. (26) Appel, C.; Ma, L. Concentration, pH, and surface charge eﬀects on cadmium and lead sorption in three tropical soils. J. Environ. Qual. 2002, 31, 581–589. (27) Nordstrom, D. K.; Alpers, C. N. Negative pH, eﬄorescent mineralogy, and consequences for environmental restroration at the Ion Mountain Superfund site, California. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 3455–3462. (28) Johnson, D. B.; Hallberg, K. B. Acid mine drainage remediation options: A review. Sci. Total Environ. 2005, 338, 3–14. (29) Castro, J. M.; Moore, J. N. Pit lakes: Their characteristics and the potential for their remediation. Environ. Geol. 2000, 39, 1254– 1260. (30) Ritchie, J. D.; Perdue, E. M. Proton-binding study of standard and reference fulvic acids, humic acids, and natural organic matter. Geochim. Cosmochim. Acta 2003, 67, 85–96. (31) Park, C.; Fenter, P. A. Phasing of resonant anomalous X-ray reﬂectivity spectra and direct Fourier synthesis of element-speciﬁc partial structures at buried interfaces. J. Appl. Crystallogr. 2007, 40, 290–301. (32) Lee, S. S.; Fenter, P.; Park, C.; Sturchio, N. C.; Nagy, K. L. Hydrated cation speciation at the muscovite (001)-water interface. Langmuir 2010, 26, 16647–16651. (33) Schlegel, M. L.; Nagy, K. L.; Fenter, P.; Cheng, L.; Sturchio, N. C.; Jacobsen, S. D. Cation sorption on the muscovite (001) surface in chloride solutions using high-resolution X-ray reﬂectivity. Geochim. Cosmochim. Acta 2006, 70, 3549–3565. (34) Templeton, A. S.; Trainor, T. P.; Traina, S. J.; Spormann, A. M.; Brown, G. E., Jr. Pb(II) distributions at bioﬁlm-metal oxide interfaces. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 11897–11902. (35) Zhou, Z. L.; Rowland, S.; Mantoura, R. F. C.; Braven, J. The formation of humic coatings on mineral particles under simulated estuarine conditions - A mechanistic study. Water Res. 1994, 28, 571– 579. (36) Wang, L. L.; Chin, Y.-P.; Traina, S. J. Adsorption of (poly)maleic acid and an aquatic fulvic acid by goethite. Geochim. Cosmochim. Acta 1997, 61, 5313–5324. (37) Saito, T.; Koopal, L. K.; Nagasaki, S.; Tanaka, S. Analysis of copper binding in the ternary system Cu2+/humic acid/goethite at neutral to acidic pH. Environ. Sci. Technol. 2005, 39, 4886–3893. (38) Cabaniss, S. E. Forward modeling of metal complexation by NOM: I. A priori prediction of conditional constants and speciation. Environ. Sci. Technol. 2009, 43, 2838–2844. (39) Boily, J.-F.; Fein, J. B. Proton binding to humic acids and sorption of Pb(II) and humic acid to the corundum surface. Chem. Geol. 2000, 168, 239–253. 9580

dx.doi.org/10.1021/es201323a |Environ. Sci. Technol. 2011, 45, 9574–9581

Environmental Science & Technology

ARTICLE

(40) Arias, M.; Barral, M. T.; Da Silva-Carvalhal, J.; Mejuto, J. C.; Rubinos, D. Interaction of Hg(II) with kaolin-humic acid complexes. Clay Miner. 2004, 39, 35–45. (41) Karlsson, T.; Skyllberg, U. Complexation of zinc in organic soils - EXAFS evidence for sulfur associations. Environ. Sci. Technol. 2007, 41, 119–124. (42) Manceau, A.; Matynia, A. The nature of Cu bonding to natural organic matter. Geochim. Cosmochim. Acta 2010, 74, 2556–2580.

9581

dx.doi.org/10.1021/es201323a |Environ. Sci. Technol. 2011, 45, 9574–9581

ARTICLE pubs.acs.org/est

Photoreduction of Chlorothalonil Fungicide on Plant Leaf Models S. Monadjemi,†,‡ M. El Roz,†,‡ C. Richard,†,‡ and A. Ter Halle†,‡,* †

Laboratoire de Photochimie Moleculaire et Macromoleculaire (LPMM), Clermont Universite, Universite Blaise Pascal, BP 10448, 63000 Clermont-Ferrand, France ‡ CNRS, UMR 6505, LPMM, F-63171 Aubiere, France

bS Supporting Information ABSTRACT: Photodegradation is seldom considered at the surface of vegetation after crop spraying. Chlorothalonil, a broadspectrum foliar fungicide with a very widespread use worldwide, was considered. To represent the waxy upper layer of leaves, tests were performed within thin paraﬃn wax ﬁlms or in n-heptane. Laser ﬂash photolysis together with steady-state irradiation in n-heptane allowed the determination of the photodegradation mechanisms Chlorothalonil ability to produce singlet oxygen was measured; noteworthy its eﬃciency is close to 100%. Additionally, chlorothalonil photodegradation mainly proceeds through reductive dechlorination. In these hydrophobic media, a radical mechanism was evidenced. Photochemical tests on wax ﬁlms under simulated solar light show that formulated chlorothalonil is more reactive than pure chlorothalonil. The ﬁeld-extrapolated half-life of photolysis on vegetation was estimated to 5.3 days. This value was compared to the half-lives of penetration and volatilization available in the literature. It appears that chlorothalonil dissipation from crops is ruled by both photodegradation and penetration. The relative importance of the two paths probably depends on meteorological factors and on physicochemical characteristics of the crop leaf cuticle.

’ INTRODUCTION Solar-radiation-induced degradation is an important route of pesticide dissipation in the environment.1 It may even be the major process for some pesticides in surface water. However, photodegradation is seldom considered at the surface of vegetation after crop spraying, although it was noted about 30 years ago that “a meaningful proportion of pesticides are not stable to sunlight on crops after ﬁeld application”.2 Pesticide dissipation from crop leave surfaces after spraying includes physical transport processes (volatilization, wash-oﬀ and plant uptake) and transformation processes (photodegradation, thermal degradation). For example, it was demonstrated that herbicide sulcotrione dissipation from maize leaves in the ﬁeld is predominantly dependent on photodegradation compared to the other processes (volatilization, wash oﬀ, penetration).3 Dissipation from crops after spraying directly aﬀects the pesticide eﬃciency. In practice, losses are generally counterbalanced by high application rates. If these losses were taken into consideration and estimated, they could be overcome by formulation adjustments for example. It was shown that an optimal adjuvant combination can reduce the eﬀective pesticide dose by as much as 10-fold,4 which obviously represents an important progress with regard to the environmental impact. Chlorothalonil (2,4,5,6-tetrachloroisophthalonitrile, CT ﬁgure 1) is a broad-spectrum nonsystemic foliar fungicide.5 This is one of the most commonly used fungicides worldwide. CT is used to control r 2011 American Chemical Society

many fungal diseases in a wide range of crops, especially vegetables and fruits. CT is also the most common pesticide used to control fungal diseases in turf. Turf application rates are among the highest of all labeled used patterns. CT has been available since 1964, and its use is still common. In 2001, it ranked 13th in pesticide usage with approximately 4 to 5 million kg active ingredient to crops in the U.S.6 New systemic fungicides, such as sterol biosynthesis inhibitor (tebuconazole and propiconazole) and strobilurin fungicides (azoxystrobin and triﬂoxystrobin), have provided new options for disease management. However, these fungicides have highly speciﬁc modes of action, which have prompted concerns in the development of fungicide resistance. The combination of these new fungicides with chlorothalonil has been recommended as resistance prevention.7 The environmental fate of CT has been well documented since the early 1970s. Degradation in soil is primarily microbial, and the major reported metabolite is 4-hydroxy-2,5,6- trichloroisophthalonitrile (OH-CT).8 In surface water, CT is resistant to hydrolysis, photolysis and volatilization. CT fate in surface water is ruled by microbial metabolization; the main metabolite is OHCT.9 At several estuarial locations, CT was detected with maximum Received: April 27, 2011 Accepted: September 27, 2011 Revised: September 27, 2011 Published: September 27, 2011 9582

dx.doi.org/10.1021/es202400s | Environ. Sci. Technol. 2011, 45, 9582–9589

Environmental Science & Technology

ARTICLE

’ EXPERIMENTAL SECTION Kinetic Equations. The quantum yields and rate constants for CT photodegradation are represented in Scheme 1. The reaction rate constants are specified for CT deactivation (kd, s1), CT reactivity with oxygen to produce singlet oxygen (kO2, M1s1), CT deactivation by oxygen (k0 O2, M1s1) CT reactivity with nheptane (kh, s1) and with a hydrogen donor molecule if added (kRH, M1s1). CT apparent first order constant of photodegradation kdeg is expressed as follows:

kdeg ¼ Ie ΦISC

kRH ½RH þ kh 0 kRH ½RH þ kh þ kd þ KO2 ½O2 þ kO2 ½O2

ΦISC is the quantum yield of intersystem crossing and Ia the photon fluence rate. Quantum yield singlet oxygen production (ΦSO) and of photodegradation (Φdeg) are expressed as follows: Figure 1. Structure of chlorothalonil and absorption spectra in methanol (dash line) and in heptane (solid line) at 104 M (with a 1 cm optical path).

levels of 1.1 ng L1, while OH-CT was detected in river waters at higher levels (up to 14 ng L1)10,11. This metabolite appears to be more toxic than the parent compound.9 Aside from OH-CT, many other degradation products resulting from reductive dechlorination or oxidation/hydrolyzation reactions via biotic13,14 and abiotic routes9,1518 have been identiﬁed under laboratory conditions. CT aqueous photochemistry has been investigated but the studies focused on kinetic aspects;15,18,19 the mechanism of CT photodegradation has never been discussed. In addition, CT photolysis has been reported to be not signiﬁcant in the aquatic media.9 In this paper, we aimed at estimating and rationalizing CT photodegradation on leaves. CT photoreactivity on leaves can be drastically diﬀerent from that in aquatic media, as was demonstrated for other pesticides.20 Photochemical degradation of pesticides on crops occurs mainly within the outermost micrometer of plants. Once the pesticide has entered the leaves the photochemical reactions are decreased due to shielding eﬀect; the plant metabolism also comes into play and prevails. The surface of leaves, the cuticle, is mainly made of cuticular wax. A ﬁrst good approach to estimate the phenomenon in the ﬁeld is a set of laboratory tests on thin wax ﬁlms. At least some precautions should be taken, the pesticide concentration should be within the range encountered in the ﬁeld, because some concentration eﬀects have been demonstrated21 and simulated solar irradiations are recommended. Under these conditions it was demonstrated for example, that herbicide sulcotrione rate of photolysis is of the same order of magnitude with the laboratory set and ﬁeld monitoring.22 Furthermore, tests were also performed in n-heptane in this study. This is not a classical solvent used for environmental studies, but its hydrophobicity makes it a suitable solvent in this particular case. The transition from wax paraﬃn (linear alkane chains containing 2040 carbon atoms) to n-heptane is only a reduction of the aliphatic chain length with no modiﬁcation of the functional groups between the two media. Laser ﬂash photolysis together with steady-state irradiation allowed the determination of the mechanisms of chlorothalonil photodegradation. Finally, the ﬁeldextrapolated rate of photolysis was compared with rates of penetration and volatilization available in the literature and discussed.

Φdeg ¼ ΦISC

kRH ½RH þ kh kRH ½RH þ kh þ kd þ kO2 ½O2

and ΦSO ¼ ΦISC

kO2 ½O2 kRH ½RH þ kh þ kd þ kO2 ½O2

Chemicals and Standards. All solvents and chemicals were used as received. Chlorothalonil (99.3%) was purchased from Fluka (Saint-Quentin Fallavier, France). Commercial formulations of chlorothalonil, Fongil FL (500 g L1 chlorothalonil), were obtained from a regular agricultural shop. 1,3 Diphenylisobenzofuran (97%), perinaphthenone (97%), isoprene (99%), paraffin wax (mp 7080 °C) and spectrophotometric-grade nheptane (99%) were all purchased from Aldrich (Saint-Quentin Fallavier, France). Synperonic 10/6 (Polyoxyethylene C9C11 alcohol) was provided by Uniquema (Paterson, NJ). 4-Nonylphenol (mixture of compounds with branched side chains) was provided by TCI (Tokyo Kasei, Japan). Anthracene (99%) was provided by Alfa Aesar (Alfa Aesar, Schiltingheim, France). Water was purified using a Millipore Milli-Q system (Millipore αQ, resistivity 18 MΩ cm1, DOC < 0.1 mg L1). Pyridine (99%) was supplied by Lancaster (Alfa Aesar, Schiltingheim, France) and 4-nitroanisole (97%) by Aldrich (Saint-Quentin Fallavier, France). Methanol (99%, HPLC grade) was provided by Riedel de Ha€en (Saint-Quentin Fallavier, France). High-Performance Liquid Chromatography (HPLC). Irradiated solutions of CT were analyzed by HPLC. Samples irradiated in acetonitrile were analyzed using a reversed phase column as well as the methanol solutions used to rinse the irradiated wax films. Irradiated samples in n-heptane were monitored by normal phase HPLC. CT detection limit was around 106 M which is below the concentration range of the study. Separations in the reversed-phase mode were performed on a Hewlett-Packard series 1050 HPLC system consisting of an autosampler, a pump, and a diode-array detector. The column used for these separations has the following specifications: C18, Nucleodur supplied by Machery-Nagel, 150 mm long, 4.6 mm internal diameter, and 5 μm particle size. The mobile phase consisted of 70% methanol and 30% water with a flow rate of 1 mL min1 and UV detection at 232 nm. CT retention time in these conditions was 7.5 min. Separations in the normal-phase mode were performed on a MerckHitachi HPLC system consisting of an autosampler (AS-2000), a pump (L-6200A), and a UV/vis detector (L-4250). The column 9583

dx.doi.org/10.1021/es202400s |Environ. Sci. Technol. 2011, 45, 9582–9589

Environmental Science & Technology

ARTICLE

Scheme 1. Proposed Mechanism of CT Homolytic Photodehalogenation in n-Heptanea

Both quantum yields were measured at 313 nm at 298 K in air-saturated n-heptane (oxygen concentration in n-heptane is 2.8 103 M); ΦISC is close to unit and Φdeg equals (2.0 ( 0.1)102. The primary photoproducts are the dechloro-congeners, as further dechlorinated congeners have been identiﬁed by GC-MS; they are probably phototransformed according to the same degradation path as the parent molecule. a

used for these separations has the following specifications: Silica, Sunfire supplied by Waters, 250 mm long, 4.6 mm internal diameter, and 5 μm particle size. The mobile phase consisted of 1% isopropanol and 99% n-heptane with a flow rate of 1 mL min1 and UV detection at 232 nm. CT retention time in normal phase was 3.8 min. Gas Chromatography with Electron Impact Mass Spectrometry (GC/EI-MS). The method was adapted from Penuela et al.15 A Hewlett-Packard 6890 gas chromatograph coupled to a Hewlett-Packard 5973 mass spectrometer was used for GC-MS in electron impact (EI) mode. A HP-5MS capillary column (30 m 0.25 mm inner diameter, 0.25 μm film thickness) was programmed to heat from 60 to 300 at 10 °C min1; the final temperature was held 20 min. The injector and transfer line temperatures were both held at 280 °C. Helium was used as the carrier gas at a flow rate of 1.2 mL min1. The injection volume was 1 mL in the split mode. For the mass spectrometer, the ionization energy was 70 eV, and the temperature of the ion source was 280 °C. CT and its photodegradation products were identified by GC/EI-MS. The retention time were 14.3 min for CT, 12.7 and 12.8 min for the monodechlorinated congeners, 10.8 and 11 min for the didechlorinated congeners and 8.8 min for the tridechlorinated congeners detected. Laser Flash Photolysis (LFP). Laser flash photolysis experiments were carried out using an Applied Photophysics LKS.60

apparatus equipped with a Nd3+:YAG laser Quanta-Ray GCR130. Samples were irradiated using the fourth harmonic (266 nm, 9 ns pulse duration, the beginning of the measure starts 20 ns after the laser flash) in a quartz cuvette. The experimental details are described elsewhere.23 Solutions were deoxygenated by nitrogen bubbling directly in the cell and changed after each shot to avoid excessive exposure. The energy transfer reaction between the triplet excited state of chlorotalonil (104 M) and anthracene (4 104 M) was studied in an argon deoxygenated mixture of n-heptane/isooctane (50:50, v/v) upon at 266 nm. The anthracene triplet excited-state build-up was monitored at 421 nm in the absence and in the presence of chlorothalonil. The quantum yield of the chlorothalonil triplet excited-state formation was estimated assuming a complete energy transfer between CT and anthracene (see the Supporting Information (SI)). Potassium peroxodisulfate was used as a chemical actinometer (ε450ΦISC = 1900 ( 150 M1cm1).23 The quenching of the chlorothalonil triplet excited state by phenol was investigated by varying the phenol concentration within the range 3 1042 103 M in air-saturated n-heptane. Steady-State Irradiation Experiments in Solution. For the irradiation experiments, a stock solution of 2.7 mg of CT in 100 mL of n-heptane (104 M) was prepared. Irradiations were performed at 313 nm in a quartz cuvette (1 cm optical length) 9584

dx.doi.org/10.1021/es202400s |Environ. Sci. Technol. 2011, 45, 9582–9589

Environmental Science & Technology

ARTICLE

To achieve an amount of surfactant that is ﬁve times higher in weight than the active ingredient, surfactants containing solutions were prepared as follows: 33 mg of polyoxyethylene alcohol (spreading agent, synperonic 10/6) was added to 250 mL of CT stock solution (104 M) and 36 mg of 4-nonylphenol was added to 250 mL of CT stock solution (104 M). Quality Control. For quantum yield measurements, all experiments were performed at least in triplicate and analyzed twice by HPLC. For kinetic measurements on wax films, each point was realized in quadruplicate. Results presented are an average of replicates given with standard deviation error.

using a mercury arc lamp (200 W) equipped with a Schoeffel monochromator. To deoxygenate the solutions, argon was bubbled for 10 min prior to irradiation. The cuvette was then closed using a valve system. Due to the volatility of n-heptane, the CT concentration was recalculated after deoxygenation. This recalculation was done using the absorbance of the solution. To measure the quantum yield of singlet oxygen formation (ΦSO), the probe 1,3 diphenylisobenzofuran was used in the concentration range 105 to 104 M in n-heptane,24 details of the method used is described in the SI . ΦSO was calculated with a method described in the SI. To assess singlet oxygen reactivity toward CT, perinaphthenone (105 M) was used as a sensitizer and excited at 365 nm. Actinometry was measured according to the method described by Calvert and Pitts,25 the photon fluence rate was 1.67 1015 photons cm2 s1 at 313 nm and 5.86 1015 photons cm2 s1 at 365 nm. Steady-State Irradiation Experiments on Wax Films. The procedure has been described recently.20 Recommendations for chlorothalonil application are between 500 and 1000 g/ha. The applied rates on paraffin films were set within theses ranges. Chlorothalonil solution in methanol (104 M) was added with a micropipet on top of wax films (1.5 mL for each dish of 3.2 cm diameter). Methanol was allowed to evaporate overnight. Fongil solutions were prepared in water, with a dilution factor of 36 000. The surface concentration of chlorothalonil was ultimately equal to 5 108 g m2. After irradiation at 550 W m2 (irradiance within the range 300800 nm), films were rinsed with methanol, and samples were analyzed by HPLC. The recovery yields were estimated as 95% ( 5% using nonirradiated samples. Control samples were run concomitantly in the dark.

’ RESULTS AND DISCUSSION Phototransformation in Solution. CT Absorption Characteristics. In n-heptane, CT shows two bands of absorption at 311

(ε=1570 M1cm1) and 324 nm (ε=2250 M1cm1) (Figure 1). There is a slight hypsochromic shift (1 nm) and higher absorption coefficients compared to more polar solvents such as methanol. The CT absorption spectrum in aqueous solution is not reported in the literature due to a low solubility26 (<1 mg L1 at pH 7). The CT absorption spectrum in n-heptane overlaps with solar radiation spectrum; CT is therefore susceptible to direct photolysis. Implication of CT Triplet State. Laser flash photolysis at 266 nm of CT in argon-saturated n-heptane (A266 = 0.1) yields the transient absorption spectrum at the end of the pulse shown in Figure 2. Two maxima at 320 and 575 nm are observed. The first-order-decay rate constant is 2.2 105 s1 in the absence of oxygen and 2.4 106 s1 in air-saturated n-heptane (see inset of figure 2). Because the oxygen concentration is 2.8 103 M,27 a bimolecular rate constant of 7.8 108 M1s1 is obtained for the oxygen quenching. This reactivity toward oxygen strongly suggests that the intermediate species is the triplet excited state. To confirm this hypothesis, we carried out energy transfer reactions using anthracene as an energy acceptor. From these experiments, one could estimate a quantum yield of inter system crossing (ΦISC) of at least 0.70 (see SI). Steady-state monochromatic irradiations at 313 nm were also performzed to measure the pseudofirst-order rate constants and the quantum yields of the CT phototransformation. In air-saturated n-heptane, CT shows a low quantum yield of photodegradation (Φdeg = (2.0 ( 0.1) 102, Table 1 entry 1). This value is higher than that reported by Millet et al. in hexane (Φdeg = (2.2 ( 0.3) 103), but the difference may be explained by the imprecision of polychromatic quantum yield measurements in this latter case.26 In deoxygenated n-heptane, the quantum yield of photodegradation is increased by a factor of 4 to 5 (Φdeg = (9 ( 0.1) 102, Table 1 entry 2). As oxygen is a triplet quencher, the lowering of Φdeg in air-saturated medium confirms the involvement of the CT triplet excited state in the reaction. This result also shows that the chemical transformation of CT is not due to reaction with oxygen.

Figure 2. Transient absorption spectrum measured at the end of the pulse upon excitation at 266 nm in a deoxygenated solution of CT (104 M) in heptane. Inset: absorbance decay monitored at 320 nm in deoxygenated (dot line) and in air-saturated media (solid line).

Table 1. Chlorothalonil Rate Constant (kdeg) and Quantum Yield of Photodegradation (Φdeg) in Solution entry 1 2

solvent n-heptane n-heptane

conditions air saturated deoxygenated

(kdeg ( σ) (s1)

(Φdeg ( σ)

(1.1 ( 0.06) 10

4

(2.0 ( 0.1) 102

(6.8 ( 0.1) 10

4

(9.0 ( 0.1) 102

4

3 4

n-heptane +5% isopropanol n-heptane +2% phenol

air saturated air saturated

(3.4 ( 0.2) 10 (4.9 ( 0.2) 104

(3.4 ( 0.1) 102 (5.4 ( 0.2) 102

5

acetonitrile

air saturated

(7.5 ( 0.2) 106

(9.7 ( 0.2) 104

9585

dx.doi.org/10.1021/es202400s |Environ. Sci. Technol. 2011, 45, 9582–9589

Environmental Science & Technology

Figure 3. GC-MS chromatogram of an irradiated chlorothalonil solution.

Because the CT triplet excited state is eﬃciently quenched by oxygen, we tried to measure the ability of CT to generate singlet molecular oxygen using chemical trapping. ΦSO was estimated close to unit (see sSI). Thus, CT very eﬃciently sensitizes the formation of singlet oxygen. The fact that ΦISC is only 0.7 while ΦSO equals 1 is surprising because singlet oxygen is expected to arise from the triplet excited state and thus ΦISC and ΦSO should be close. 1,3-Diphenylisobenzofuran has for long shown its ability as a chemical scavenger for singlet oxygen titration and thus ΦSO is reliable. The experiment of energy transfer to anthracen in contrast probably underestimates ΦISC. First, absorbances are measured 1 μs after the pulse end when the absorbance of triplet anthracen is maximum, thus a part of triplets is lost. Second, the anthracen concentration may be not high enough for full trapping of the chlorothalonil triplets, but it cannot be increased for solubility reasons. Further investigations were initiated to determine whether CT could be oxidized by singlet oxygen. Perinaphtenone, a classical singlet oxygen sensitizer, was selectively excited at 365 nm, where CT absorption is zero. No depletion of CT concentration was observed under these conditions, demonstrating the stability of CT against singlet oxygen. CT Photodehalogenation. In the literature, CT and its degradation products have been primarily analyzed by gas chromatography.28 Here, GC-MS analysis of photoproducts formed in air-saturated or deoxygenated n-heptane mainly yields dechlorinated compounds (Figure 3). The EI spectrum of CT exhibits a parent ion at m/z 266 with the quadruplet in the correct ratio for four Cl atoms. In contrast, the primary photoproducts show a parent ion at m/z 230 in accordance with the replacement of a chlorine atom by a hydrogen atom. This attribution is confirmed by the good agreement with the mass spectra reported in the literature regarding aqueous photolysis of CT.15,16,18 There are three possible isomers of trichloro-1,3dicyanobenzene, but only two were detected in this study. The isomer discrimination is not possible without standards. Further reductive dechlorination also occurs; two isomers of dichloro1,3-dicyanobenzene have also been detected. Their mass spectra also correspond with that reported in the literature. One isomer of monochloro-1,3-dicyanobenzene was also detected. Its identification has already been clearly demonstrated15 because it presents a characteristic fragmentation pattern (ions

ARTICLE

at m/z 162, 134, 127, 100, and 74). The mass spectra of CT and its dechlorinated congeners are given in the SI. In comparison, the literature data concerning the distribution pattern of CT degradation products in aqueous media are more complex, depending on the constitution of the aqueous solution, that is, whether humic matter was added and whether a solvent was used to dissolve CT. HO-CT is proposed to be formed by a photoassisted nucleophilic substitution but was not detected because it is very quickly degraded by light at a rate exceeding that of production.18 In pure water, only trichloro-1,3-dicyanobenzene has been detected and quantiﬁed.18 The presence of natural organic matter allows photoreduction through the formation of dichloro- and chloro-1,3-dicyanobenzene. In conclusion, the presence of hydrogen donors such as organic matter favors CT reductive dechlorination in water. It is important to note that the recovery of CT degradation products is always very low. When the formation of the dehalogenated congeners is quantiﬁed with standards, they are found in very low chemical yields, even if no other degradation products were detected.15 Given the photoproducts identiﬁcation, reductive dehalogenation is the only identiﬁed route for CT phototransformation in nheptane. CT photodegradation is also promoted by good hydrogen donor compounds. The CT quantum yield of photolysis in polar aprotic solvents such as acetonitrile shows a marked decrease compared to n-heptane (Φdeg = (9.7 ( 0.2) 104, Table 1 entry 4). On the contrary, the addition of a good hydrogen donor such as 2-propanol or phenol in aerated nheptane increases the quantum yield of photodegradation by 1.7 to 2.7 (Table 1 entries 3 and 4). Besides the photoreduction of CT, photoaddition could be envisaged. According to Jahn et al., CT can undergo a photoaddition to oleﬁnic compounds of the plant cuticle.29 They have investigated the formation of bound residues to the cuticle by immunoassay. If bound residue formation was evidenced, it never exceeded 2% of the initial amount of CT. To directly investigate this photoaddition reaction, we irradiated n-heptane solutions of CT in the presence of isoprene. No adduct formation was detected by GC-MS. In literature data, CT has also been proposed to form adducts in solution of pure methanol or ethanol.30 In this study, the addition of alcohol to n-heptane in proportions of up to 5% did not lead to the formation of adducts. Mechanism of rReaction. Laser flash photolysis experiments have given evidence of the formation of a CT triplet excited state, and the experiments regarding energy transfer to anthracene have shown that the intersystem crossing quantum yield is high (ΦISC > 0.7). Similar values were reported for polychlorobenzenes in cyclohexane.31 Moreover, we demonstrated that there is also an efficient energy transfer on oxygen with a singlet-oxygen quantum yield close to unity. Thus, most of CT triplet excited state is deactivated in aerated solution, explaining the low CT quantum yield of photodegradation (Φdeg = (2 ( 0.2) 102). Because the CT triplet state is very efficiently populated, one can postulate that the photochemical reactions proceed from it. Photoproduct identiﬁcation shows that the main reaction path is a reductive dechlorination. The photolysis of aryl chlorides has been widely studied because of their environmental signiﬁcance,32,33 and several paths for the reductive photodechlorination have been proposed.30,31,34,35 The homolytic cleavage of the CCl bond may occur (path a, scheme 1). Alternatively, the reaction may be initiated by an electron transfer by a H-donor molecule followed by a proton transfer to form two radicals (path b, scheme 1).34,35 The presence of the cyano substituent, which has 9586

dx.doi.org/10.1021/es202400s |Environ. Sci. Technol. 2011, 45, 9582–9589

Environmental Science & Technology

ARTICLE

Table 2. Chlorothalonil Rate Constant of Photodegradation on Paraﬃn Wax (kdeg) Measured in the Suntest Reactor at 500 W m2 b with Addition of Various Additivesa rate constant pure chlorothalonil

kdeg = 0.065 ( 0.02 d1

chlorothalonil + nonyl phenol

kdeg = 0.65 ( 0.2 d1

chlorothalonil + with a

kdeg = 1.8 ( 0.2 d1

synthetic alcohol ethoxylate Synperionic 10/6f photolysis of formulated chlorothalonyl

kdeg = 0.33 ( 0.17 d1

g

kdeg = 0.14 d1

ext

photolysis of formulated chlorothalonyl extrapolated to outdoor conditions CT volatilization from crops d

Figure 4. Phototransformation of chlorothalonil on paraﬃn wax under simulated sunlight irradiation (Suntest 500 W/m2) as a function of irradiation time. Chlorothalonil surface concentration is set to 500 g/ha, it is either 2 deposited pure, 9 in addition with a spreading agent Synperionic 10/6, b in a commercial formulation, f in addition with nonylphenol. The dashed lines represent the ﬁrst-order ﬁtted decays. The error bars represent the standard deviations based on three replicates.

a strong electro-withdrawing eﬀect, may favor the latter path.36,37 To test this alternative pathway, we investigated the inﬂuence of phenol on the triplet excited-state lifetime by laser ﬂash photolysis. The presence of phenol within the concentration range 3 104 to 2 103 M actually reduced the triplet lifetime in airsaturated n-heptane. The rate constant of the reaction was estimated to be kRH= 1.8 109 M1 s1. These considerations lead us to assess that the path b is possible. The products resulting from the attack of the chlorine radical have not been detected by GC-MS. In the experiment with phenol, the chlorophenol is probably too photoreactive to accumulate in the solution,23 but we could detect phenoxyphenol, which conﬁrms a radical mechanism. The mechanism is proposed in Scheme 1. Photodegradation on Paraffin Wax. Photochemical processes occurring at the surface of crop leaves, vegetables or fruits after spraying were studied using a previously established laboratory procedure on model wax films.20,38 Pure or formulated CT decays follow pseudofirst-order kinetics with correlation values higher than 0.65 (R2, Figure 4). Pure CT presents a photodegradation rate constant (kdeg) of (6.5 ( 2) 102 d1 (R2 = 0.77). This value corresponds to a photolysis half-life of about ten days under continuous irradiation. Formulated CT photoreactivity is higher with a photolysis half-life of two days (kdeg = (0.33 ( 0.17) d1, R2 = 0.7). This photoreactivity increase induced by the additives has already been observed on wax films2022 but is difficult to rationalize because formulations compositions are unknown. The addition of a classically used adjuvant from the family of synthetic alcohol ethoxylate (Synperionic 10/6) has also an accelerating effect on CT photolysis (Table 2). Such additives allow a better spreading of the molecule at the film surface, resulting in a better light absorption and consequently an increased photoreactivity. Additionally, the experiments performed in nheptane have shown that CT photoreduction is enhanced by the addition of H-donor molecules like phenol. The commercial formulations are usually constituted of alcoholic functionalities or other good hydrogen-donor constituents that could have the same effect. To further demonstrate this effect, we prepared a mixture of

CT penetration in apple fruit

e

kvol = 0.007 d1 kpen = 0.11 d1

a

The extrapolated rate constant of photodegradation of formulated chlorothalonyl to outdoor conditions includes the alternation of night and day (extkdeg)c. Chlorothalonyl rate constant of volatilization measured in the ﬁeld (kvol)d and chlorothalonyl rate constant of penetration (kpen)e. b Corresponding to an irradiance of 51 W m2 between 300 and 400 nm. c Average irradiance of 22 W m2 in central Europe in summer time conditions44. d Measured in a wheat ﬁeld42. e Measured into apple fruit under laboratory conditions with radiolabeled chlorothalonyl40. f A synthetic alcohol ethoxylate used as a spreading agent. g Commercial formulation Fongil.

pure CT and nonylphenol for application on wax films. An increase of photoreactivity by a factor of about nine was observed (kdeg = (0.65 ( 0.2) d1, R2 = 0.75). Environmental Implications. CT is described to provide a rather long protective action at the plant surfaces, and the typical recommended spray interval is 1014 days39. CT dislodgeable residues are defined as the part obtained by soaking the leaves in an organic solvent for a few seconds, corresponding to the portion present at the leaf surface or in the upper layer of the cuticle. The dissipation of dislodgeable foliar CT on cranberry in a bog followed first-order kinetics with an estimated half-life of 12.7 days.40 CT dissipation from tomatoes in a greenhouse presents a half-life of dissipation of 8.8 days,41 whereas field dissipation on the surface banana leaves under tropical conditions had half-lives of 3.9 days.42 In the following discussion we are going to compare the different path of dissipation. The CT penetration into the plant rate is estimated to be slow. Experiments with radio-labeled CT measured 46% and 29.4% of dislodgeable radioactive residue in apple fruit after 5 days and 12 days, respectively.43 CT is indeed described as a nonsystemic foliar fungicide with a low penetration rate inside the leaves.44 Based on these results, the penetration rate coeﬃcient is kpen = 0.11 d1 (half-life 6.3 days). CT volatilization was measured in a wheat ﬁeld.45 The cumulated emission from the ﬁeld reached 0.6% and 4.9% of the application dose after 31 h and one week, respectively. Based on these results, the rate coeﬃcient is kvol = 0.007 d1. These measurements are in good agreement with another study in which the CT volatilization from crop canopies has been modeled and estimated to about 5% of the application dose after a week.46 The sunlight irradiance between 300 and 400 nm in central Europe in summertime conditions averages 22 W m2 over the 24 h of a day,47 whereas the tests in the photosimulator are 9587

dx.doi.org/10.1021/es202400s |Environ. Sci. Technol. 2011, 45, 9582–9589

Environmental Science & Technology performed at a constant irradiance of 51 W m2. We can thus extrapolate the formulated CT rate constant of photodegradation to outdoor conditions: extkdeg = 0.14 d1. This value corresponds to an extrapolated photolysis half-life of 5 days. This value is the ﬁrst extrapolated datum regarding the CT photolysis rate at a plant surface. The CT rate of photodegradation and rate of penetration are of about the same order of magnitude, while volatilization is much lower (table 2). CT dissipation from crops is then ruled by both photodegradation and penetration. The relative importance of the two paths probably depends on meteorological factors and on physicochemical characteristics of the cuticle of the crop. It is interesting to note that while CT fate in the aquatic media is reported to be ruled by microbial degradation and photolysis is a minor path,15 this study implies that CT photolysis on crops is one of the predominant processes. This ﬁrst approach invites photochemical investigation on real system; a ﬁrst step would be the use of plants under controlled conditions and ﬁnally the undertaking of a ﬁeld monitoring. Additionally CT has the ability to produce singlet oxygen very eﬃciently (near 100%). The oxidant power of singlet oxygen could have some impact on the constituents of crop leaves. Singlet oxygen produced by chlorothalonil could also take place in its fungicide activity; like some phototoxic phytoalexins implied into plant defense mechanisms.

’ 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

*E-mail: [email protected].

’ REFERENCES (1) Environmental Protection and Toxic Substances; Pesticide Fact Sheet, Name of Chemical: Azoxystrobin; United States Oﬃce of Prevention, Pesticides: Arlington, VA, 1997. (2) Rabson, R.; Plimmer, J. R. Photoalteration of pesticides— Summary of workshop. Science 1973, 180 (4091), 1204–1205. (3) ter Halle, A.; Piquet, A.; Richard, C. An actual scenario that demonstrates sulcotrione photodegradation on maize leaves after spraying. Environ. Chem. 2007, 4 (4), 256–259. (4) Green, J. Adjuvant outlook for pesticides. Pestic. Outlook 2000, 11 (5), 196–199. (5) Sherrarda, R. M.; Bearrb, J. S.; Murray-Guldec, C. L.; Rodgers, J. H.; Shahd, Y. T. Feasibility of constructed wetlands for removing chlorothalonil and chlorpyrifos from aqueous mixtures. Environ. Pollut. 2004, 127, 385–394. (6) Kiely, T.; Donaldson, D.; Grube, A. Pesticides Industry Sales and Usage: 2000 and 2001 Market Estimates; U.S. EPA: Washington, DC, 2004. (7) Keinatha, A. P.; Holmesb, G. J.; Evertsc, K. L.; Egeld, D. S.; Langston, D. B. Evaluation of combinations of chlorothalonil with azoxystrobin, harpin, and disease forecasting for control of downy mildew and gummy stem blight on melon. Crop Prot. 2007, 26, 83–88. (8) Putnam, R. A.; Nelson, J. O.; Clark, J. M. The persistence and degradation of chlorothalonil and chlorpyrifos in a cranberry bog. J. Agric. Food Chem. 2003, 51 (1), 170–176. (9) Mozzachio, A. M.; Rusiecki, J. A.; Hoppin, J. A.; Mahajan, R.; Patel, R.; Beane-Freeman, L.; Alavanja, M. C. Chlorothalonil exposure

ARTICLE

and cancer incidence among pesticide applicator participants in the agricultural health study. Environ. Res. 2008, 108 (3), 400–403. (10) Kern, S.; Fenner, K.; Singer, H. P.; Schwarzenbach, R. P.; Hollender, J. Identiﬁcation of transformation products of organic contaminants in natural waters by computer-aided prediction and high-resolution mass spectrometry. Environ. Sci. Technol. 2009, 43 (18), 7039–7046. (11) Yamamoto, A.; Miyamoto, I.; Kitagawa, M.; Moriwaki, H.; Miyakoda, H.; Kawasaki, H.; Arakawa, R. Analysis of chlorothalonil by liquid chromatography/mass spectrometry using negative-ion atmospheric pressure photoionization. Anal. Sci. 2009, 25 (5), 693–697. (12) The Agrochemical Handbook; Unwin Brothers limited Old Working, Royal Society of chemistry: Surrey, UK. 1991. (13) Regitano, J. B.; Tornisielo, V. L.; Lavorenti, A.; Pacovsky, R. S. Transformation pathways of C-14-chlorothalonil in tropical soils. Arch. Environ. Contam. Toxicol. 2001, 40 (3), 295–302. (14) Carlo-Rojas, Z.; Bello-Mendoza, R.; Figueroa, M. S.; Sokolov, M. Y. Chlorothalonil degradation under anaerobic conditions in an agricultural tropical soil. Water Air Soil Poll. 2004, 151 (14), 397–409. (15) Penuela, G. A.; Barcelo, D. Photodegradation and stability of chlorothalonil in water studied by solid-phase disk extraction, followed by gas chromatographic techniques. J. Chromatogr., A 1998, 823 (12), 81–90. (16) Sakkas, V. A.; Lambropoulou, D. A.; Albanis, T. A. Study of chlorothalonil photodegradation in natural waters and in the presence of humic substances. Chemosphere 2002, 48 (9), 939–945. (17) Szalkowski, M. B.; Stallard, D. E. Eﬀect of Ph on hydrolysis of chlorothalonil. J. Agric. Food Chem. 1977, 25 (1), 208–210. (18) Kwon, J. W.; Armbrust, K. L. Degradation of chlorothalonil in irradiated water/sediment systems. J. Agric. Food Chem. 2006, 54 (10), 3651–3657. (19) Avila, L. A.; Massey, J. H.; Senseman, S. A.; Armbrust, K. L.; Lancaster, S. R.; McCauley, G. N.; Chandler, J. M. Imazethapyr aqueous photolysis, reaction quantum yield, and hydroxyl radical rate constant. J. Agric. Food Chem. 2006, 54 (7), 2635–2639. (20) ter Halle, A.; Drncova, D.; Richard, C. Phototransformation of the herbicide sulcotrione on maize cuticular wax. Environ. Sci. Technol. 2006, 40 (9), 2989–2995. (21) Eyheraguibel, B.; ter Halle, A.; Richard, C. Photodegradation of bentazon, clopyralid, and triclopyr on model leaves: Importance of a systematic evaluation of pesticide photostability on crops. J. Agric. Food Chem. 2009, 57 (5), 1960–1966. (22) Lavieille, D.; ter Halle, A.; Richard, C. Understanding mesotrione photochemistry when applied on leaves. Environ. Chem. 2008 5 (6), 420–425. (23) Bonnichon, F.; Richard, C. Phototransformation of 3-hydroxybenzonitrile in water. J. Photochem. Photobiol., A 1998, 119 (1), 25–32. (24) Ohyashiki, T.; Nunomura, M.; Katoh, T. Detection of superoxide anion radical in phospholipid liposomal memebrane by ﬂuorescence quenching method using 1,3 diphenylisobenzofuran. Biochim. Biophys. Acta 1999, 1421, 131–139. (25) Calvert, J. C. Pitts, J. N. Photochemistry; John Wiley and Sons: London, 1966. (26) Millet, M.; Palm, W. U.; Zetzsch, C. Abiotic degradation of halobenzonitriles: Investigation of the photolysis in solution. Ecotoxicol. Environ. Saf. 1998, 41 (1), 44–50. (27) Battino, R.; Rettich, T. R.; Tominaga, T. The solubility of oxygen and ozone in liquids. J. Phys. Chem. Ref. Data 1983, 12 (2), 163–178. (28) Chaves, A.; Shea, D.; Danehower, D. Analysis of chlorothalonil and degradation products in soil and water by GC/MS and LC/MS. Chemosphere 2008, 71 (4), 629–638. (29) Jahn, C.; Schwack, W. Determination of cutin-bound residues of chlorothalonil by immunoassay. J. Agric. Food Chem. 2001, 49 (3), 1233–1238. (30) Samanta, S.; Kole, R. K.; Ganguly, L. K.; Chowdhury, A. Photochemical transformation of the fungicide chlorothalonil by ultra violet radiation. Bull. Environ. Contam. Toxicol. 1997, 59 (3), 367–374. 9588

dx.doi.org/10.1021/es202400s |Environ. Sci. Technol. 2011, 45, 9582–9589

Environmental Science & Technology

ARTICLE

(31) Kawamura, Y.; Takeda, M.; Uchiyama, M. Photolysis of chlorothalonil in benzene. J. Pestic. Sci 1978, 3 (4), 397–400. (32) Izadifard, M.; Langford, C. H.; Achari, G. Photocatalytic dechlorination of polychlorinated biphenyls using leuco-methylene blue sensitization, broad spectrum visible lamps, or light emitting diodes. Environ. Sci. Technol. 2010, 44 (23), 9075–9079. (33) Bendig, P.; Vetter, W. Photolytical transformation rates of individual polybrominated diphenyl ethers in technical octabromo diphenyl ether (DE-79). Environ. Sci. Technol. 2010, 44 (5), 1650–1655. (34) Bunce, N. J. Photolysis of aryl chlorides with aliphatic-amines. J. Org. Chem. 1982, 47 (10), 1948–1955. (35) Freccero, M.; Fagnoni, M.; Albini, A. Homolytic vs heterolytic paths in the photochemistry of haloanilines. J. Am. Chem. Soc. 2003, 125 (43), 13182–13190. (36) Neunteufel, R. A.; Arnold, D. R. Radical ions in photochemistry. I. The 1,1-diphenylethylene cation radical. J. Am. Chem. Soc. 1973, 95 (12), 4080–4081. (37) Bonnichon, F.; Grabner, G.; Guyot, G.; Richard, C. Photochemistry of substituted 4-halogenophenols: eﬀect of a CN substituent. J. Chem. Soc., Perkin Trans. 2 1999, No. 6, 1203–1210. (38) Lavieille, D.; ter Halle, A.; Bussiere, P. O.; Richard, C. Eﬀect of a spreading adjuvant on mesotrione photolysis on wax ﬁlms. J. Agric. Food Chem. 2009, 57 (20), 9624–9628. (39) Elliott, V. J.; Spurr, H. W. Temporal dynamics of chlorothalonil residues on peanut foliage and the inﬂuence of weather factors and plantgrowth. Plant. Dis 1993, 77 (5), 455–460. (40) Putnam, R. A.; Nelson, J. O.; Clark, J. M. The persistence and degradation of chlorothalonil and chlorpyrifos in a cranberry bog. J. Agric. Food Chem. 2003, 51 (1), 170–176. (41) Kurz, M. H. S.; Goncalves, F. F.; Adaime, M. B.; da Costa, I. E. D.; Primel, E. G.; Zanella, R. A gas chromatographic method for the determination of the fungicide chlorothalonil in tomatoes and cucumbers and its application to dissipation studies in experimental greenhouses. J. Brazil. Chem. Soc. 2008, 19 (6), 1129–1135. (42) Chaves, A.; Shea, D.; Cope, W. G. Environmental fate of chlorothalonil in a Costa Rican banana plantation. Chemosphere 2007, 69, 1166–1174. (43) Gilbert, M. Fate of chlorothalonil in apple foliage and fruit. J. Agric. Food Chem. 1976, 24 (5), 1004–1007. (44) Caux, P. Y.; Kent, R. A.; Fan, G. T.; Stephenson, G. L. Environmental fate and eﬀects of chlorothalonil: A Canadian perspective. Crit. Rev. Environ. Sci. Technol. 1996, 26 (1), 45–93. (45) Carole, B.; Rousseau-Djabri, M. F.; Benjamin, L.; Brigitte, D.; Dominique, F.; Olivier, B.; Barriuso, E. Fungicide volatilization measurements: Inverse modeling, role of vapor pressure, and state of foliar residue. Environ. Sci. Technol. 2010, 44 (7), 2522–2528. (46) Leistra, M.; Van Den Berg, F. Volatilization of parathion and chlorothalonil from a potato crop simulated by the PEARL model. Environ. Sci. Technol. 2007, 41 (7), 2243–2248. (47) Frank, R.; Klopﬀer, W. Spectral solar photon irradiance in central-europe and the adjacent north-sea. Chemosphere 1988, 17 (5), 985–994.

9589

dx.doi.org/10.1021/es202400s |Environ. Sci. Technol. 2011, 45, 9582–9589

ARTICLE pubs.acs.org/est

Gas Chromatographic Analysis with Chiral Cyclodextrin Phases Reveals the Enantioselective Formation of Hydroxylated Polychlorinated Biphenyls by Rat Liver Microsomes Izabela Kania-Korwel,§ Michael W. Duﬀel,‡ and Hans-Joachim Lehmler§,* §

Department of Occupational and Environmental Health, College of Public Health, University of Iowa, Iowa City, Iowa 52242, United States ‡ Department of Pharmaceutical Sciences and Experimental Therapeutics, College of Pharmacy, University of Iowa, Iowa City, Iowa 52317, United States

bS Supporting Information ABSTRACT: Chiral PCB congeners are major components of PCB mixtures and undergo enantioselective biotransformation to hydroxylated (OH-)PCBs by cytochrome P450 enzymes. While it is known that biotransformation results in an enantiomeric enrichment of the parent PCB, it is currently unknown if OH-PCBs are formed enantioselectively. The present study screened seven commercial capillary gas chromatography columns containing modiﬁed β- or γ-cyclodextrins for their potential to separate the atropisomers of methylated derivatives of OH-PCB. The atropisomers of 3-, 4- and 5-methoxy derivatives were at least partially separated on one or more columns. A subsequent biotransformation study was performed with rat liver microsomes to assess if hydroxylated metabolites are formed enantioselectively from PCBs 91, 95, 132, and 149. The OH-PCBs were extracted from the microsomal incubations, derivatized with diazomethane and analyzed as the respective methoxylated (MeO)PCB derivatives using selected columns. The 5-hydroxylated metabolites of PCBs 91, 95, 132, and 149 were the major metabolites, which is consistent with PCB’s biotransformation by cytochrome P450 2B enzymes. All 5-hydroxylated metabolites displayed a clear, congener-speciﬁc enantiomeric enrichment. Overall, this study demonstrates for the ﬁrst time that chiral PCBs, such as PCB 91, 95, 132, and 149, are enantioselectively metabolized to OH-PCBs by cytochrome P450 enzymes.

’ INTRODUCTION Polychlorinated biphenyls (PCBs) remain an important class of environmental contaminants, even forty years after their production was banned in the US. PCB levels did not show a clear decline in the general US population from 1999 to 2004,1 thus suggesting ongoing PCB exposure via the diet 2 and by inhalation of indoor air.3 Human epidemiological data consistently show a negative association between developmental exposure to environmental PCBs and cognitive function in infancy or childhood.4 Speciﬁcally, in utero and lactational exposures to multiple-ortho substituted PCB congeners correlate with decreased intelligence quotients, impaired learning and memory, attentional deﬁcits, and lowered reading comprehension. These PCB congeners and their hydroxylated metabolites cause developmental neurotoxicity in laboratory studies by mechanisms involving altered Ca2+ signaling, interference with thyroid hormone signaling and decreased dopamine content.5 Ryanodine receptor (RyR) sensitization has been shown to be the most sensitive mechanism mediating PCB’s eﬀect on Ca2+ signaling. A recent study demonstrates that PCB 136 enantiospeciﬁcally sensitizes RyRs, with only ()-PCB 136 being active.6 Although PCBs are considered to be persistent organic pollutants, many PCB congeners, especially congeners with r 2011 American Chemical Society

vicinal hydrogen substituents, are biotransformed through a complicated metabolic pathway to hydroxylated (OH-PCB) and methylsulfonylated metabolites (MeSO2PCB).7 OH-PCB can be formed by direct insertion of oxygen into an aromatic CH bond or via an epoxide intermediate. The epoxide may rearrange to OH-PCBs, be conjugated with glutathione in a glutathione transferase catalyzed reaction or react with other cellular nucleophiles, such as proteins and DNA. Several OH-PCBs, but not analogous MeSO2PCBs, have been shown to be neurotoxic and to sensitize RyRs.8 Many of these neurotoxic metabolites display axial chirality and exist as rotational isomers which are nonsuperimposable mirror images of each other.9 Such nonsuperimposable molecules are called enantiomers or, in the case of the rotational isomers of multiple-ortho substituted PCB congeners, atropisomers. Chiral PCB metabolites can be formed by biotransformation of both chiral or prochiral PCBs. Analogous to the parent compounds,9 it is likely that chiral OH- and MeSO2PCBs also undergo enantiomeric enrichment Received: April 29, 2011 Accepted: October 3, 2011 Revised: September 27, 2011 Published: October 03, 2011 9590

dx.doi.org/10.1021/es2014727 | Environ. Sci. Technol. 2011, 45, 9590–9596

Environmental Science & Technology

ARTICLE

Figure 1. Formation of putative metabolites of PCB atropisomers with the OH-group in the 2,3,6-trichlorinated ring by rat liver microsomes (PCB 91: R2 = Cl; R1 = R3 = H; PCB 95: R3 = Cl; R1 = R2 = H; PCB 132: R1 = R2 = Cl; R3 = H; PCB 149: R2 = R3 = Cl; R1 = H).

in vivo and display enantioselective toxicity. Indeed, several studies report enantiomeric enrichment of MeSO2PCB in wild-life,1014 human 15 and laboratory animal studies.16,17 A study by Norstr€om et al. showed that meta- and para-methylsulfonyl metabolites of PCB 132 are formed enantiospeciﬁcally in rats.17 In contrast, the enantiomeric enrichment of chiral OHPCBs is poorly investigated and to date only one study demonstrated the enantiomeric enrichment of two OH-PCBs after intraperitoneal administration of racemic PCB 136 in rats.18 It is currently unclear if the enantiomeric enrichment of chiral PCB metabolites, in particular OH-PCBs, in vivo is due to their enantioselective formation by cytochrome P450 enzymes (P450 enzymes) or other, enantioselective phase II biotransformation processes. A recent study by Warner et al. provides indirect evidence that PCB metabolism by cytochrome P450 enzymes may result in the enantioselective formation of OH-PCBs.19 While the authors demonstrated that PCB 132 atropisomers are enantioselectively metabolized by recombinant rat and human cytochrome P450 enzymes, it is unclear which metabolites were formed and if these metabolites displayed enantiomeric enrichment. In the present study, we use a suite of putative metabolites of neurotoxic PCBs 91, 95, 132, and 149 to ﬁrst develop gas chromatographic methods for the separation of the OH-PCB atropisomers. Subsequently, these separation methods were utilized to investigate the enantioselective formation of OHPCB by hepatic microsomes. These studies demonstrate the enantioselective formation of 5- hydroxylated PCBs by cytochrome P450 enzymes, an observation that has implication for understanding the mechanisms of PCB neurotoxicity and, ultimately, assessing the risk of developmental neurotoxicity in PCB exposed populations.

’ EXPERIMENTAL SECTION Methoxylated Derivatives of PCBs. A series of racemic 4-, 5-methoxy and 4,5-dimethoxy- derivatives of PCBs 91, 95, 132, and 149 and respective NIH-shift products (Figure 1) were synthesized and characterized as described previously 18 (see

Table S1 of the Supporting Information, SI, for a list of compounds). The nomenclature of the PCBs is according to the revised Ballschmiter nomenclature.20 The nomenclature of the metabolites follows the recommendation of Maervoet et al. 21 and is presented in Table S1 of the SI. Enantioselective Gas Chromatography. Atropisomers of methoxylated PCBs were separated using an Agilent 7890A gas chromatograph equipped with an electron capture detector (μ-ECD) using the following enantioselective columns: HPChiral-20B (20B) and Cyclosil-B (CB) from Agilent, ChirasilDex (CD) from Varian, BGB-172 (BGB) for BGB Analytic, and Chiral-Dex B-DM (BDM), Chiral-Dex B-PM (BPM), and Chiral-Dex G-TA (GTA) columns from Supelco Analytical (see Table S3 of the SI for details). The compounds were first analyzed individually using the following temperature program: 50 °C for 1 min, 10 °C/min to 140 °C, hold for 20 min, 1 °C/min to column maximum temperature, hold for 20 min 22 to establish retention times and elution order as well as identify atropisomers of methoxylated PCBs separating on these columns for further optimization. The injector and detector were kept at 250 °C. The flow was set to 1 mL/min. Subsequently, the separation of methoxylated PCBs that partially separated or showed signiﬁcant peak widening on a particular column was further optimized. Because the resolution of PCB atropisomers improves with decreasing analysis temperature (Figure S1 of the SI),18,23 temperature programs with a long isothermal hold at progressively lower temperatures were used: 50 °C for 1 min, 10 °C/min to X, 10 °/min to 225 °C, hold for 10 min, where X is in the range from 140 to 180 °C. The ﬂow in these analyses was 3 mL/min to achieve elution in a reasonable amount of time. The length of the temperature programs varied from 220 min at 180 °C to 500 min at 140 °C. In some cases, an analysis time of 8 h was not suﬃcient to elute the methoxylated metabolite. These long analysis times were considered impractical for enantioselective analysis, even if improvement of separation was observed. Atropisomers of PCBs in the microsomal incubation mixtures (see below) were analyzed on the CD column using the following temperature program: 2 °C/min 9591

dx.doi.org/10.1021/es2014727 |Environ. Sci. Technol. 2011, 45, 9590–9596

Environmental Science & Technology from 100 to 150 °C, 0.2 °C/min to 185 °C, 15 °/min to 200 °C.22,24 The conditions provided acceptable resolution (from 0.63 to 0.89) for all four studied congeners. Microsomal Incubations and Extraction. Microsomal incubations consisted of 0.1 M phosphate buffer (pH 7.4), 3 mM magnesium chloride, 0.5 mM NADPH and 0.77 mg/mL of microsomal protein in a final volume of 16 mL. After 5 min of preincubation, 10 mM solution of PCB in DMSO was added (80 μL, 0.5% total volume) and the samples were incubated at 37 ( 1 °C in a shaking water bath for 30 min. Control samples with DMSO alone were incubated at the same time and no metabolites were detected in these blank samples. The reaction was quenched by adding 2 mL of ice-cold 0.5 M NaOH and subsequent heating of the samples at 90 °C for 10 min. Metabolites and remaining parent compound were extracted from the acidified reaction mixture (1 mL of 6 M HCl) with 2-propanol (2 mL) and hexane-MTBE (5 mL, 1:1 v/v).24,25 Metabolites were separated from the parent compound by partitioning of OH-PCBs into KOH (10 mL, 0.5M) and reextracting them with hexane-MTBE (8 mL, 9:1 v/v) from the acidified solution (3 mL of 6 M HCl). After exchanging the solvent to hexane, OH-PCBs were derivatized with diazomethane 26 and the MeO-PCBs were cleaned-up using sulfuric acid 27,28 before gas chromatographic analysis. Gas Chromatographic Analysis. In addition to the enantioselective analysis, samples from microsomal incubations were analyzed on an Agilent 6890 gas chromatograph (GC) with a 5975 mass selective detector in both total and selective ion monitoring modes. The GC was equipped with a SLB-5 ms column (Supelco, 60 m, 250 μm 0.25 μm). The following temperature program was used: 100 °C hold for 1 min, 5 °C/min to 250 °C, hold for 20 min, and 5 °C/min to 280 °C, hold for 3 min. The injector temperature was 280 °C and the MS temperatures were 280, 230, and 150 °C for transfer line, source, and quadrupole, respectively. In total ion scan mode, a mass range of m/z 50 to 500 was recorded. In selective ion monitoring mode, ions m/z 326, 356, and 386 were used to scan for metabolites of pentachlorobiphenyls, and 360, 390, and 420 for metabolites of hexachlorobiphenyls. The same chromatographic conditions were employed for GC-ECD analyses with a SPB-1 column (Supelco, 60 m, 250 μm 0.25 μm). The detector temperature for these analyses was 300 °C and the flow rate was 1.0 mL/min.

’ RESULTS AND DISCUSSION Enantioselective Separation of Methoxylated PCBs. A major obstacle toward studying the enantioselective metabolism of PCBs is the lack of good separation methods for methoxylated derivatives of relevant chiral OH-PCBs. Although we recently reported the separation of MeO-PCB atropisomers with the CD column, the resolution on this column was unsatisfactory.18 Therefore, a set of seven enantioselective columns was obtained from commercial sources to optimize the separation of methoxylated PCB atropisomers (Table S3 of the SI). This set of columns is similar to the one used previously by Wong and Garrison 22 in a study of the separation of PCB atropisomers. Six columns contained a β-cyclodextrin-based stationary phase, a stationary phase that is frequently used in gas and liquid chromatographic separations of PCB atropisomers 22,29 and methoxylated PCB derivatives.18,30 These six columns differ in the modification of the β-cyclodextrin as well as the polysiloxane backbone. For example, the CD and BPM columns contain a

ARTICLE

permethylated β-cyclodextrin phase that is chemically bound to the polysiloxane backbone in the CD but not the BPM column. Additionally, a single γ-cyclodextrin column was studied (GTA). This chiral selector has also been used for the enantioselective separation of PCBs.22 Atropisomeric separations of a few MeO-PCBs were comparatively straightforward. Speciﬁcally, the atropisomers of 591 (see Table S1 of the SI for the MeO-PCB nomenclature) were separated on all 7 columns used in this study (Figures S2B and S3 of the SI), 50 -132 atropisomers were resolved on 6 columns (except of 20B column) and 595 atropisomers were separated on 5 columns (Table S1 of the SI). These three compounds have a 5-methoxy group and a 2,3,6 chlorine substitution pattern in the same phenyl ring. Interestingly, Haglund reported that a 2,3,6- chlorine substitution pattern in one phenyl ring favors the separation of PCBs on β-cyclodextrin phases.29,31 This suggests that a methoxy substituent in 5-position may have a relatively small eﬀect on the enantioselective separation of MeO-PCBs on the columns investigated. Similarly, all seven NIH shift products have a 2,3,4,6-substitution pattern (i.e., MeO-PCBs with a 2,4,6-chlorine substitution pattern and a 3-methoxy groups in the same phenyl ring) resolved to some extent of the diﬀerent enantioselective columns, with two of the seven NIH shift products separating on four diﬀerent columns (398 and 3154) and all other atropisomers separating on at least one column (Tables S1 and S2 of the SI). The only exception was 350, which did not separate on any column. Similarly, chiral PCB congeners with a 2,3,4,6substitution pattern were resolved easily on various chiral columns.22 The observation that 398, but not 350 atropisomers separated on most stationary phases is consistent with earlier observations on the CD phase.18 Analogous to the parent PCBs,29 a 2,3-dichloro substitution in the less lipophilic, nonmethoxylated phenyl ring is required to achieve a separation of the MeO-PCB atropisomers of the NIH shift products on the CD column. This relationship between chemical structure and resolution appears to extend to the chiral stationary phases investigated in this study. In contrast, the resolution of monomethoxylated compounds with a 4-methoxy group was poorer compared to the structurally analogous 5-methoxylated PCBs. The best separations were observed for 491 (Rs = 0.79; BDM column), 495 (Rs = 0.78; BGB column) and 4136 (Rs = 0.83; CB column), whereas 40 -132 atropisomers were not resolved on any column used in this study. Chiral PCBs with 4,5-dimethoxy groups did not resolve on any of the columns under investigation. The only exception was 4,595, which resolved on the BDM and CB columns. Together, these observations suggest that a 4-methoxy group, but not a 4-chlorine substituent prevents the separation of atropisomers on the chiral columns investigated. In the case of pentachlorinated PCBs with 4-methoxy groups, the more lipophilic methoxylated phenyl ring is likely to partition into the respective cavity of the chiral selector.18,29 Therefore, one plausible explanation for the overall poor separation of PCB atropisomers with a 4-methoxy group is an unfavorable interaction of the 4-methoxy substituent with functional group in the cavity of the chiral selector. However, further studies are needed to conﬁrm this hypothesis. Formation of Hydroxylated Metabolites. After optimizing the enantioselective separations, a series of microsomal metabolism experiments were performed with racemic PCBs 91, 95, 132, and 149 to determine the metabolite profile and, ultimately, 9592

dx.doi.org/10.1021/es2014727 |Environ. Sci. Technol. 2011, 45, 9590–9596

Environmental Science & Technology

ARTICLE

Table 1. Comparison of the Yields and Enantiomeric Fractions of Hydroxylated Metabolites of PCB 91, PCB 95, PCB 132, and PCB149 Determined Using Diﬀerent Columnsa PCB metabolite

yieldb

BDM

CB

CD

20%

0.54e

0.54f

0.54e

PCB 91 591

0.43

PCB 95 595

8.8%

0.36e nr

0.32e

0.33e

0.33

0.33 0.39

0.68 0.71e,d

16%

0.30e

0.31e,d

0.31

2.6%

0.66

0.65

0.65

PCB 149 5149

0.46

0.64

X-95 PCB 132 50 -132

BGBc

nr

0.46 nr

a

The enantiomeric fraction of the parerent PCB is shown for comparison. b Expressed as percent of the PCB used in the respective incubation (0.8 μmol). c The elution order of the atropisomers was reversed on the BGB column. d Analysis conducted at 170 °C. e Racemic standard coelutes with another peak. f EF of racemic standard 0.55. Abbreviations of columns: BGB - BGB-172; BDM - Chiral-Dex B-DM, BPM - ChiralDex B-PM; CB - Cyclosil-B; CD - Chirasil-Dex (CD); GTA - Chiral-Dex G-TA. See Table S3 for additional information regarding the columns.

assess if the atropisomers of the major OH-PCBs are formed enantioselectively. Highly ortho-substituted PCBs with 2,3,6trichloro substituted phenyl rings are readily metabolized in reactions catalyzed by cytochrome P450 enzymes to OH-PCBs with the OH-group in meta or para position (Figure 1). In addition, an NIH shift can result in the formation of a 2,4,6trichloro-3-hydroxy-substitution pattern. The cytochrome P450 isoforms responsible for the metabolism of these PCB congeners are cytochrome P450 2B enzymes, such as 2B1 and 2B4.19,32,33 Since rat cytochrome P450 2B enzymes are induced by phenobarbital in rat livers, the present study used liver microsomes obtained from phenobarbital-treated rats. The cytochrome P450 2B activity, measured as BROD activity, was 13 100 pmol/min/ mg protein, which represents almost a 100-fold increase compared to corn-oil treated animals (data not shown). PCBs 91, 95, 132, and 149, were incubated at concentrations of 50 μM for 30 min with liver microsomes. These PCB congeners are highly neurotoxic 8 and have a 2,3,6-trichloro substitution pattern in one ring. On the basis of the work by Schnellmann et al. with the structurally similar PCB 136,34 a concentration of 50 μM was used to maximize the formation of metabolites and, thus, allow GC-MS detection of trace metabolites. The metabolites were extracted, separated from the excess of the parent PCB, derivatized with diazomethane and analyzed as MeO-PCB derivatives by GC-MS in both total ion and selective ion monitoring modes as well as with GC-ECD to identify the metabolites (Figures S4S7 of the SI). For all four congeners, the major metabolite formed had the OH-group in the 5-position of the 2,3,6-trichloro substituted phenyl ring. The amounts of the metabolites formed in the microsomal incubations displayed the rank order 591 > 50 -132 > 595 > 5149 (Tables 1 and S4 of the SI). In addition to 591, three minor metabolites were formed in the incubation with PCB 91. We were able to identify these metabolites as 491, 3100 (NIH shift product), and 4,591. PCB 95 formed 595 and a second major monohydroxylated pentachlorobiphenyl X-95. The retention time of X-95 did not correspond to any putative PCB 95 metabolites with the

hydroxyl group in the 2,3,6-trichloro substituted ring system, which suggests that the hydroxy group of X-95 is present on the second, less chlorinated ring. This metabolite is most likely 30 -95, a metabolite that was reported in rats in vivo.35 In addition to 50 132, three minor metabolites were detected in incubations with PCB 132. We were able to unambiguously identify these metabolites as 40 -132, 30 -140 (NIH shift product) and 40 ,50 132 using authentic standards. Only a single metabolite, 5149, was found in incubations of PCB 149. Under the experimental conditions employed in this study, no appreciable amounts of NIH-shift products were detected, with only trace amounts of 3-100 and 30 -140 being detected in incubations with PCB 91 and PCB 132, respectively. In contrast, two in vivo studies reported the formation of NIH-shift metabolites of PCB 95 and 136 in rodent animal models.18,35 With exception of PCB 95, no metabolites with a hydroxyl group in the second, lower chlorinated ring were observed. This is consistent with the presence of a 4-chlorine substituent in these congeners, which essentially prevents metabolic attack in this ring system. Finally, only dihydroxylated PCB 91 and PCB 132 metabolites were detected in the microsomal incubations. Since both 4- and 5-136 are readily metabolized to 4,5-136 by recombinant enzymes 33 and human liver microsomes,34 the fact that only small quantities of dihydroxylated metabolites were formed for PCBs 91 and 132 is likely due to the relatively short incubation time (30 min) used in the present studies. Overall, the metabolite proﬁles observed for all four congeners are in agreement with previous in vitro and in vivo studies. Two unidentiﬁed monohydroxylated metabolites were reported by Warner et al. in an in vitro experiment where PCB 95 was metabolized by rat cytochrome P450 2B1 enzyme.19 Also, a single monohydroxylated metabolite was reported for PCB 91 in the same study. A comparative in vivo study by Sundstrom et al. 35 reported 40 -95 as the major metabolite in quail, whereas 5-95 was the major metabolite in the rat and mouse, with 3-103 (NIH shift product) and 40 -95 being only minor metabolites. In the rat, mouse, and guinea pig, 50 -132 was also the major metabolite of PCB 132. 40 - and 40 ,50 -132 were also formed, but in diﬀerent ratios depending on the species.36 There are currently no studies reporting the formation of hydroxylated metabolites of PCB 149. Enantioselective Analysis of Microsomal Incubations of Chiral PCBs. Four chiral columns were used to determine the enantiomeric enrichment of the major OH-PCBs formed in the microsomal incubations discussed above (Tables 1 and S4; Figures 2 and S8S11 of the SI). All enantioselective analyses were performed isothermally at 160 °C unless otherwise stated (Table 1). The BDM column provided baseline separation of all major, 5-substituted metabolites of the PCB congeners investigated (Figure S2A of the SI). The EF values obtained with this column, as well as the CB and CD columns were comparable and revealed a congener-dependent enantiomeric enrichment for all OH-PCBs. The results from the BGB column showed good agreement with BDM, CB, and CD columns; however, the elution order of the atropisomers was reversed. Since the absolute configurations of neither the parent compounds nor the corresponding OH-PCBs shown in Figure 1 has been established, it is currently not possible to determine which of the atropisomers shown in Figure 1 is formed selectively and if there is a congener-specific difference in how R and S atropisomers are metabolized. The parent PCBs showed a congener speciﬁc enantiomeric enrichment (Table 1). In the case of PCBs 91, 132, and 149, the 9593

dx.doi.org/10.1021/es2014727 |Environ. Sci. Technol. 2011, 45, 9590–9596

Environmental Science & Technology

ARTICLE

enrichment of the second eluting atropisomer was also observed for 50 -132 (EF = 0.31). Due to the comparatively low levels of the minor metabolites and coelution problems, it was not possible to determine the EFs of the minor metabolites; however, it is likely that these metabolites also display some enantiomeric enrichment. Overall, this study demonstrates for the ﬁrst time that chiral PCBs, such as PCB 91, 95, 132, and 149, are enantioselectively metabolized to OH-PCBs by cytochrome P450 enzymes. Considering the high cytochrome P450 2B activity in the microsomal preparation used in the incubations, it is likely that the enantioselective formation of 5-hydroxylated PCB metabolites is due to cytochrome P450 2B enzymes, which is consistent with PCB metabolism studies using recombinant enzymes. The enantioselective formation of OH-PCBs therefore in part explains the enantiomeric enrichment of the respective parent compounds in in vitro and in vivo studies.9 Further studies are needed to better understand the role of cytochrome P450 2B enzymes in the enantiomeric enrichment of both PCBs and OH-PCBs, and the toxicity of pure OH-PCB atropisomers. The later question is of particular interest from an environmental health perspective because of the enantiomeric enrichment of OH-PCBs reported in this study and the recently documented eﬀect of OH-PCBs on Ryanodine receptor sensitization.8

’ ASSOCIATED CONTENT

bS

Figure 2. Enantiomeric enrichment of 5-hydroxylated metabolites of PCB 91 (A), PCB 95 (B), PCB 132 (C), and PCB 149 (D) in incubations with rat liver microsomes. Samples were analyzed on the DBM column at 160 °C.

second eluting atropisomers were enriched, whereas the ﬁrst eluting atropisomer was enriched for PCB 95. In the case of PCB 132 and 149, this corresponds to an enrichment of (+)-PCB 132 and (+)-PCB 149.31,37 The most pronounced enantiomeric enrichment was observed for PCB 95 (EF = 0.64) and PCB 132 (EF = 0.39). These observations are in agreement with a previous study by Warner demonstrating an enantiomeric enrichment of PCBs due to metabolism by recombinant rat and human cytochrome P450 enzymes 19 as well as several in vivo studies reporting an enantiomeric enrichment of these congeners in rodents and humans. For example, in a study by Kania-Korwel et al.,38 the EFs of PCB 95 (EF = 0.63) and PCB 149 (EF = 0.45) in the liver of rats treated with Aroclor 1254 were comparable to the results obtained in this study. The most intriguing observation of the present study is that the major OH-PCB metabolites formed in the microsomal incubations displayed a clear, congener-speciﬁc enantiomeric enrichment. There was very little enrichment of the ﬁrst eluting atropisomer observed for 5-91, with an EF of 0.54. In the case of 5-149, the ﬁrst eluting congener was also enriched (EF = 0.65). Both 5-95 and the unknown monomethoxylated PCB 95 (X-95) showed an enrichment of the second eluting atropisomer, with comparable EFs of approximately 0.33 for both metabolites. An

Supporting Information. Description of microsomes preparation and cytochrome P450 enzyme activities; structures, nomenclature and resolution of all studied methoxylated PCBs on all columns investigated, both in temperature programmed and isothermal analysis; description of enantioselective columns; enantiomeric fractions of methoxylated PCBs in microsomal incubations; dependence of resolution on temperature in isothermal analysis for 5-91 and 50 -132; resolution of all methoxylated PCBs on BDM column and programmed temperature; resolution of 5-91 on all columns and programmed temperature; GC-MS and GC-ECD analysis of OH-PCBs formed in microsomal incubations of PCB 91, PCB 95, PCB 132, and PCB 149; comparison of the enantiomeric enrichment of 5-methoxylated derivatives of PCB 91, PCB 95, PCB 132, and PCB 149 isolated from microsomal incubations to the racemic standard. This material is available free of charge via the Internet at http:// pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: (319) 335-4211; fax: (319) 335-4290; e-mail: [email protected].

’ ACKNOWLEDGMENT The authors would like to thank Drs. Stelvio Bandiera and Eugene Hrycay (University of British Columbia) for the characterization of the microsomes, Drs. Yang Song, Sandhya M. Vyas, and Sudhir N. Joshi (University of Iowa) for the synthesis of the methoxylated PCB standards, and Ananya Pramanik (University of Iowa) for help with the microsomal metabolism studies. The OH- and MeO-PCB 136 metabolites were a generous gift from E.A. Mash and S.C. Waller of the Synthetic Chemistry Facility Core of the Southwest Environmental Health 9594

dx.doi.org/10.1021/es2014727 |Environ. Sci. Technol. 2011, 45, 9590–9596

Environmental Science & Technology Sciences Center, funded by NIH Grant ES06694. The project described was supported by Grant Nos. ES05605, ES013661, and ES017425 from the National Institute of Environmental Health Sciences.

’ REFERENCES (1) Fourth National Report on Human Exposure to Environmental Chemicals; Center for Disease Control and Prevention: Atlanta, GA, 2009; www.cdc.gov/exposurereport/pdf/FourthReport_ExecutiveSummary.pdf (2) Schecter, A.; Colacino, J.; Haﬀner, D.; Patel, K.; Opel, M.; Papke, O.; Birnbaum, L. Perﬂuorinated compounds, polychlorinated biphenyl, and organochlorine pesticide contamination in composite food samples from Dallas, Texas. Environ. Health Perspect. 2010, 118 (6), 796–802. (3) Harrad, S.; Ibarra, C.; Robson, M.; Melymuk, L.; Zhang, X.; Diamond, M.; Douwes, J. Polychlorinated biphenyls in domestic dust from Canada, New Zealand, United Kingdom, and United States: Implications for human exposure. Chemosphere 2009, 76 (2), 232–238. (4) Schantz, S. L.; Widholm, J. J.; Rice, D. C. Eﬀects of PCB exposure on neuropsychological function in children. Environ. Health Perspect. 2003, 111 (3), 357–376. (5) Mariussen, E.; Fonnum, F. Neurochemical targets and behavioral eﬀects of organohalogen compounds: An update. Crit. Rev. Toxicol. 2006, 36 (3), 253–289. (6) Pessah, I. N.; Lehmler, H. J.; Robertson, L. W.; Perez, C. F.; Cabrales, E.; Bose, D. D.; Feng, W. Enantiomeric speciﬁcity of ()-2, 20 ,3,30 ,6,60 -hexachlorobiphenyl toward ryanodine receptor types 1 and 2. Chem. Res. Toxicol. 2009, 22 (1), 201–207. (7) James, M. O., Polychlorinated biphenyls: Metabolism and metabolites. In PCBs: Recent Advances in Environmental Toxicology and Health Eﬀects; Robertson, L. W.; Hansen, L. G., Eds.; The University Press of Kentucky: Lexington, KY, 2001; pp 3546. (8) Pessah, I. N.; Hansen, L. G.; Albertson, T. E.; Garner, C. E.; Ta, T. A.; Do, Z.; Kim, K. H.; Wong, P. W. Structure-activity relationship for noncoplanar polychlorinated biphenyl congeners toward the ryanodine receptor-Ca2+ channel complex type 1 (RyR1). Chem. Res. Toxicol. 2006, 19 (1), 92–101. (9) 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. 2009, 44 (8), 2757–2766. (10) Jorundsdottir, H.; Norstr€om, K.; Olsson, M.; Pham-Tuan, H.; H€uhnerfuss, H.; Bignert, A.; Bergman, A. Temporal trends of bis(4chlorophenyl) sulfone, methylsulfonyl-DDE, and -PCBs in Baltic guillemot (Uria aalge) egg 19712001—A comparison to 4,4’-DDE and PCB trends. Environ. Pollut. 2006, 141, 226–237. (11) Larsson, C.; Norstr€om, K.; Athanansidais, I.; Bignert, A.; K€onig, W. A.; Bergman, A. Enantiomeric speciﬁcity of methylsulfonyl-PCBs and distribution of bis(4-chlorophenyl) sulfone, PCB and DDE methyl sulfones in grey seal tissues. Environ. Sci. Technol. 2004, 38, 4950–4955. (12) Wiberg, K.; Letcher, R.; Sandau, C. D.; Duﬀe, J.; Norstrom, R.; Haglund, P.; Bidleman, T. F. Enantioselective gas chromatography/ mass spectrometry of methylsulfonyl PCBs with application of arctic marine mammals. Anal. Chem. 1998, 70, 3845–3852. (13) Chu, S.; Covaci, A.; Haraguchi, K.; Voorspoels, S.; van de Vijver, K.; Das, K.; Bouquegneau, J.-M.; de Coen, W.; Blust, R.; Schepens, P. Levels and enantiomeric signatures of methyl sulfonyl PCB and DDE metabolites in livers of harbor porpoises (Phocoena phocoena) from the Southern North Sea. Environ. Sci. Technol. 2003, 37, 4573–4578. (14) Karasek, L.; Hajslova, J.; Rosmus, J.; H€uhnerfuss, H. Methylsulfonyl PCB and DDE metabolites and their enentioselective gas chromatographuc separation in human adipose tissues, seal blubber and pelican muscle. Chemosphere 2007, 67, S22–227. (15) Ellerichmann, T.; Bergman, A.; Franke, S.; H€uhnerfuss, H.; Jakobsson, E.; K€onig, W. A.; Larsson, C. Gas chromatographic enantiomer separations of chiral PCB methyl sulfons and identiﬁcation of

ARTICLE

selctively retained enantiomers in human liver. Fres. Environ. Bull. 1998, 7, 244–257. (16) Larsson, C.; Ellerichmann, T.; H€uhnerfuss, H.; Bergman, A. Chiral PCB methyl sulfones in rat tissues after exposure to technical PCBs. Environ. Sci. Technol. 2002, 36, 2833–2838. (17) Norstr€ om, K.; Eriksson, J.; Haglund, J.; Silvari, V.; Bergman, A. Enantioselective formation of methyl sulfone metabolites of 2,20 ,3,30 ,4,60 hexachlorobiphenyl in rat. Environ. Sci. Technol. 2006, 40 (24), 7649 –7655. (18) (a) Kania-Korwel, I.; Vyas, S.; Song, Y.; Lehmler, H. J. Gas chromatographic separation of methoxylated polychlorinated biphenyl atropisomer. J. Chromatogr. A 2008, 1207, 146–154. (b) Joshi, S. N.; Vyas, S. M.; Duﬀel, M. W.; Parkin, S.; Lehmler, H. J. Synthesis of Sterically Hindered Polychlorinated Biphenyl Derivatives. Synthesis 2011, 1045–1054. (19) Warner, N. A.; Martin, J. W.; Wong, C. S. Chiral polychlorinated biphenyls are biotransformed enantioselectively by mammalian cytochrome P-450 isozymes to form hydroxylated metabolites. Environ. Sci. Technol. 2009, 43, 114–121. (20) Ballschmiter, K.; Bacher, R.; Mennel, A.; Fischer, R.; Riehle, U.; Swarev, M. The determination of chlorinated biphenyls, chlorinated dibenzodioxins and chlorinated dibenzofurans by GC-MS. J. High Resol. Chromatogr. 1992, 15, 260–270. (21) Maervoet, J.; Covaci, A.; Schepens, P.; Sandau, C. D.; Letcher, R. A reassessment of the nomenclature of polychlorinated biphenyl (PCB) metabolites. Environ. Health Perspect. 2004, 112 (3), 291–294. (22) Wong, C. S.; Garrison, A. W. Enantiomer separation of polychlorinated biphenyl atropisomers and polychlorinated biphenyl retention behavior on modiﬁed cyclodextrin capillary gas chromatography columns. J. Chromatogr. A 2000, 866 (2), 213–220. (23) Vetter, W. Enantioselctive fate of chiral chlorinated hydrocarbons and their metabolites in environmental samples. Food Rev. Int. 2001, 17 (2), 113–182. (24) Milanowski, B.; Lulek, J.; Lehmler, H.-J.; Kania-Korwel, I. Assesment of disposition of chiral polychlorinated biphenyls in female mdr 1a/b knockout versus wild-type mice using multivariate analyses. Environ. Int. 2010, 36 (8), 884–892. (25) Kania-Korwel, I.; El-Komy, M. H. M. E.; Veng-Pedersen, P.; Lehmler, H. J. Clearance of polychlorinated biphenyl atropisomers is enantioselective in female C57Bl/6 mice. Environ. Sci. Technol. 2010, 44 (8), 2828–2835. (26) Kania-Korwel, I.; Zhao, H.; Norstrom, K.; Li, X.; Hornbuckle, K. C.; Lehmler, H. J. Simultaneous extraction and clean-up of polychlorinated biphenyls and their metabolites from small tissue samples using pressurized liquid extraction. J. Chromatogr. A 2008, 1214, 37–46. (27) Kania-Korwel, I.; Hornbuckle, K. C.; Peck, A.; Ludewig, G.; Robertson, L. W.; Sulkowski, W. W.; Espandiari, P.; Gairola, C. G.; Lehmler, H.-J. Congener speciﬁc tissue distribution of Aroclor 1254 and a highly chlorinated environmental PCB mixture in rats. Environ. Sci. Technol. 2005, 39, 3513–3520. (28) Kania-Korwel, I.; Hornbuckle, K. C.; Robertson, L. W.; Lehmler, H.-J. Inﬂuence of dietary fat on the enantioselective disposition of 2,20 ,3,30 ,6,60 -hexachlorobiphenyl (PCB 136) in female mice. Food Chem. Toxicol. 2008, 46 (2), 637–644. (29) Haglund, P. Enantioselective separation of polychlorinated biphenyl atropisomers using chiral high performance liquid chromatography. J. Chromatogr. 1996, 724, 219–228. (30) Pham-Tuan, H.; Larsson, C.; Hoﬀmann, F.; Bergman, A.; Fr€oba, M.; H€uhnerfuss, H. Enantioselective semipreparative HPLC separation of PCB metabolites and their absolute structure elucidation using electronic and vibrational circular dichroism. Chirality 2005, 17, 266–280. (31) Haglund, P.; Wiberg, K. Determination of the gas chromatographic elution sequences of the (+)- and ()-enantiomers of stable enantiomeric PCBs on Chirasil-Dex. J. High Resolut. Chromatogr. 1996, 19, 373–376. (32) Duignan, D.; Sipes, I.; Leonard, T.; Halpert, J. Puriﬁcation and characterization of the dog hepatic cytochrome P-450 isozyme responsible for the metabolism of 2,20 ,4,40 ,5,50 -hexachlorobiphenyl. Arch. Biochem. Biophys. 1987, 255, 290–303. 9595

dx.doi.org/10.1021/es2014727 |Environ. Sci. Technol. 2011, 45, 9590–9596

Environmental Science & Technology

ARTICLE

(33) Waller, S. C.; He, Y. A.; Harlow, G. R.; He, Y. Q.; Mash, E. A.; Halpert, J. R. 2,20 ,3,30 ,6,60 -hexachlorobiphenyl hydroxylation by active site mutants of cytochrome P450 2B1 and 2B11. Chem. Res. Toxicol. 1999, 12 (8), 690–699. (34) Schnellmann, R.; Putnam, C.; Sipes, I. Metabolism of 2,20 ,3,30 ,6,60 -hexachlorobiphenyl and 2,20 ,4,40 ,5,50 -hexachlorobiphenyl by human hepatic microsomes. Biochem. Pharmacol. 1983, 32 (21), 3233–3239. (35) Sundstr€om, G.; Jansson, B. The metabolism of 2,20 ,3,50 ,6pentachlorobiphenyl in rats, mice and quails. Chemosphere 1975, 4 (6), 361–370. (36) Haraguchi, K.; Kato, Y.; Koga, N.; Degawa, M. Species diﬀerences in the tissue distribution of catechol and methylsulphonyl metabolites of 2,4,5,20 ,50 -penta and 2,3,4,20 ,30 ,60 -hexachlorobiphenyls in rats, mice, hamsters and guinea pigs. Xenobiotica 2005, 35 (1), 85–96. (37) Harju, M. T.; Haglund, P. Determination of the rotational energy barriers of atropisomeric polychlorinated biphenyls. Fres. J. Anal. Chem. 1999, 364, 219–223. (38) Kania-Korwel, I.; Garrison, A. W.; Avants, J. K.; Hornbuckle, K. C.; Robertson, L. W.; Sulkowski, W. W.; Lehmler, H.-J. Distribution of chiral PCBs in selected tissues in the laboratory rat. Environ. Sci. Technol. 2006, 40, 3704–3710.

9596

dx.doi.org/10.1021/es2014727 |Environ. Sci. Technol. 2011, 45, 9590–9596

ARTICLE pubs.acs.org/est

Binding of HgII to High-Affinity Sites on Bacteria Inhibits Reduction to Hg0 by Mixed FeII/III Phases Bhoopesh Mishra,* Edward J. O’Loughlin, Maxim I. Boyanov, and Kenneth M. Kemner Biosciences Division, Argonne National Laboratory, Argonne, Illinois 60439, United States

bS Supporting Information ABSTRACT: Magnetite and green rust have been shown to reduce aqueous HgII to Hg0. In this study, we tested the ability of magnetite and green rust to reduce HgII sorbed to 2 g 3 L 1 of biomass (Bacillus subtilis), at high (50 μM) and low (5 μM) Hg loadings and at pH 6.5 and 5.0. At high Hg:biomass loading, where HgII binding to biomass is predominantly through carboxyl functional groups, Hg LIII-edge X-ray absorption spectroscopy showed reduction of HgII to Hg0 by magnetite. Reduction occurred within 2 h and 2 d at pH 6.5 and 5.0, respectively. At low Hg:biomass loading, where HgII binds to biomass via sulfhydryl functional groups, HgII was not reduced by magnetite at pH 6.5 or 5.0 after 2 months of reaction. Green rust, which is generally a stronger reductant than magnetite, reduced about 20% of the total HgII bound to biomass via sulfhydryl groups to Hg0 in 2 d. These results suggest that HgII binding to carboxyl groups does not signiﬁcantly inhibit the reduction of HgII by magnetite. However, the binding of HgII to biomass via sulfhydryl groups severely inhibits the ability of mixed FeII/III phases like magnetite and green rust to reduce HgII to Hg0. The mobility of heavy metal contaminants in aquatic and terrestrial environments is greatly inﬂuenced by their speciation, especially their oxidation state. In the case of Hg, reduction of HgII to Hg0 can increase Hg mobility because of the volatility of Hg0. Since Hg is typically present in aquatic and terrestrial systems at low concentrations, binding of HgII to high-aﬃnity sites on bacteria could have important implications for the potential reduction of HgII to Hg0 and the overall mobility of Hg in biostimulated subsurface environments.

’ INTRODUCTION Mercury (Hg) is a contaminant of global concern, as bioaccumulation of methylmercury poses signiﬁcant risk to aquatic ecosystems and human health.1 Although elemental mercury (Hg0) is far less reactive and toxic than the water-soluble ionic HgII species, the high mobility of Hg0 (due to low vapor pressure) and the relative ease of oxidization of Hg0 to HgII render Hg0 an environmental hazard. Historical records from lake sediments provide compelling evidence that long-range atmospheric transport of Hg0 results in signiﬁcant inputs of Hg to remote areas.2 The reduction of HgII to Hg0 results from both abiotic and microbially mediated processes and is a key component of global Hg biogeochemical cycling.3 In soils and sediments, the reduction of HgII to Hg0 is generally attributed to direct microbial processes.4 However, abiotic reduction pathways,5 9 including photoreduction,10,11 can also contribute signiﬁcantly to HgII reduction to Hg0. The mobility of heavy metal contaminants in aquatic and terrestrial environments is greatly inﬂuenced by their speciation, especially their oxidation state. For example, CrVI, UVI, and TcVII species tend to be more soluble and hence mobile than CrIII, UIV, and TcIV species. Thus, stimulating the in situ activity of native metal-reducing bacteria by the addition of organic substrates (e.g., acetate, ethanol) could potentially immobilize many heavy metals and radionuclides in contaminated environments. However, r 2011 American Chemical Society

the activity of metal-reducing bacteria (FeIII-reducing bacteria in particular) can lead to the reduction of FeIII oxides to FeIIbearing phases such as magnetite, vivianite, siderite, and green rust,12,13 some of which can be eﬀective reductants for HgII to Hg0 reduction.10,11 Thus, promotion of metal-reducing conditions for immobilization of heavy metals and radionuclides can lead to increased mobility of Hg. Understanding the geochemical processes that mediate Hg transformations in aquatic and terrestrial environments is necessary to predict its fate and transport. O’Loughlin et al.10 showed the reduction of HgII to Hg0 by green rust and suggested that other FeII phases may also reduce HgII. Indeed, reduction of HgII by magnetite has recently been reported.11 Aqueous HgII reduction by magnetite occurs within minutes, and reaction rates increase with increasing magnetite surface area and solution pH.11 The same study showed that chloride, an environmentally important inorganic ligand with strong binding aﬃnity for HgII, inhibits the rate and extent of HgII reduction by magnetite. Although the reduction of aqueous HgII to Hg0 by green rust and magnetite establishes the potential for abiotic HgII reduction Received: May 27, 2011 Accepted: September 13, 2011 Revised: September 13, 2011 Published: September 14, 2011 9597

dx.doi.org/10.1021/es201820c | Environ. Sci. Technol. 2011, 45, 9597–9603

Environmental Science & Technology under FeIII-reducing conditions, the redox properties of Hg can be profoundly altered by the presence of organic ligands.14,15 Field observations indicate the eﬀect of organic ligands by showing the coexistence of HgII with high levels of FeII in groundwater containing very low levels of chloride ions.16 It is possible that strong complexation of HgII with organic ligands signiﬁcantly aﬀects its availability for reduction by magnetite and other reactive FeII/III minerals resulting from the activity of FeIIIreducing bacteria. Hence, the eﬀect of organic ligands on HgII reduction by magnetite and other FeII-bearing minerals must be evaluated to improve understanding of the geochemical processes that inﬂuence Hg transformations in the subsurface. Studies on the speciation of HgII indicate that complex organic ligands such as natural organic matter (NOM) form stable Hg complexes through their sulfhydryl, carboxyl, and amine groups.17 23 X-ray absorption spectroscopy (XAS) has shown that HgII interacts strongly with bacterial cell envelope through sulfhydryl and carboxyl functional groups.24 A systematic study of CdII binding to both gram-positive and gram-negative bacteria suggests that CdII (and HgII, which has similar coordination properties) binds to the high-aﬃnity sulfhydryl groups (about 2% of total functional groups), followed by much higher extents of adsorption to the more abundant carboxyl and phosphoryl groups at higher metal:biomass ratios.25 Because typical Hg concentrations in contaminated environments are low and cell density in biostimulated environments may be high, preferential binding of HgII to sulfhydryl groups on bacterial cells could signiﬁcantly impact the availability of HgII for reduction. This bacterial binding can aﬀect the overall redox behavior of HgII, in both natural environments and bioremediation settings where Hg can be a cocontaminant with other metals and radionuclides that are being immobilized. Understanding the interplay between factors inﬂuencing the reduction of HgII by reactive FeII phases in the presence and absence of biomass will improve understanding of abiotic HgII reduction in contaminated environments. We have investigated the eﬀects of Hg binding to bacteria on the reduction of HgII to Hg0 by magnetite and green rust. We hypothesized that sorption of HgII to biomass would inhibit abiotic reduction of HgII by mixed FeII/III phases. To test this hypothesis, after HgII adsorption to Bacillus subtilis (a common soil bacterium that is neither a methylator [i.e., cannot produce methylmercury] nor a dissimilatory metal reducer) at diﬀerent metal:biomass ratios, we introduced a stoichiometric excess of magnetite or green rust, then used synchrotron XAS to determine the speciation and coordination environment of solidphase-associated Hg. Experiments were done as a function of pH (5.0 and 6.5), total Hg concentration (5 and 50 μM), and reaction time (2 h to 2 months).

’ METHODS AND MATERIALS Bacterial Growth Conditions. The procedures for growth and washing of B. subtilis 168 for use in this study were similar to those described earlier.25,26 Briefly, B. subtilis was cultured in tryptic soy broth with 0.5% yeast extract and incubated for 24 h at 32 °C on a shaker. The cells were collected by centrifugation (5800 g for 60 min) and rinsed five times with 0.1 M NaClO4 (the background electrolyte used in the HgII sorption experiments). The resulting cell density, reported on a wet mass basis, corresponds to approximately 8 times the dry mass of the cells. HgII Adsorption to Biomass. Washed bacteria were suspended in Teflon centrifuge tubes in 0.1 M NaClO4 electrolyte

ARTICLE

at 32 °C to form a suspension of 2 g 3 L 1 of bacteria (wet mass). HgII was added from a stock solution created from a commercially available (GFS Chemicals) reagent grade 5 mM HgII standard solution in 5% HNO3, which was titrated to pH 3.0 with 1 M NaOH. The pH of each system (pH 5 or 6.5) was adjusted with 1 M HNO3 or NaOH, and the systems were allowed to react for 3 h on a shaker. The pH ((0.3 pH units) was monitored every 15 min and adjusted as required with aliquots of 1 M HNO3 or NaOH. After 3 h of reaction, the suspensions were centrifuged, and the bacterial pellet was retained for analysis by X-ray absorption ﬁne structure (XAFS) spectroscopy. The supernatant was ﬁltered (0.45 μm) using nylon membrane (Millipore ﬁlter), acidiﬁed, and analyzed for dissolved HgII by inductively coupled plasmaoptical emission spectroscopy (ICP-OES; Perkin-Elmer) with matrix-matched standards. The amount of Hg adsorbed to bacteria was calculated by subtracting the concentration of Hg remaining in solution from the total Hg concentration in the experimental system. Reaction of Biomass-Bound HgII with Magnetite/Green Rust. Magnetite and hydroxysulfate green rust (GRSO4), a green rust containing SO42‑ as the interlayer anion, were synthesized as described by Cornell and Schwertmann.27 After 3 h of reaction time between HgII and the biomass, magnetite or green rust was added at a molar ratio of HgII:FeII = 1:50. The system was rotated end-over-end at 20 rpm. All reactions were carried out in an anoxic glovebox (Coy) containing an atmosphere of 5% H2 and 95% N2. After reaction for 2 h, 2 d, or 2 months, subsamples of the suspension were centrifuged under anoxic conditions. Pellets containing biomass and Fe oxides were retained for Hg XAFS analysis within 2 h. Hg XAS Measurements and Data Analysis. Hg LIII-edge X-ray absorption near edge structure (XANES) and extended X-ray absorption fine-structure (EXAFS) spectroscopy measurements were performed at the MRCAT sector 10-ID beamline,28 Advanced Photon Source, Argonne National Laboratory. Details of the XAS experiments, standards, and data analysis are in the Supporting Information.

’ RESULTS AND DISCUSSION HgII Complexation with Biomass. HgII concentrations in the

supernatants of samples without reductant were below the ICPOES detection limit (0.05 μM), indicating complete removal of Hg from solution by sorption to biomass. Figure 1a,b compares the XANES and k2-weighted χ(k) EXAFS data for Hg standards with HgII complexed to biomass at 5 μM (HgL-bio) and 50 μM (HgH-bio) HgII loadings, at pH 5.0. The spectra indicate that Hg is complexed via sulfhydryl groups in the HgL-bio sample. Spectral features supporting this conclusion are the small preedge peak and the slight dip at 12 300 eV in the XANES, as well as the large amplitude and the phase of oscillations in the k2weighted χ(k) data (which are similar for HgL-bio and Hgcysteine data). Similarly, a strong pre-edge peak at 12285 eV and a peak at 12300 eV in the XANES spectra, combined with smaller amplitudes of oscillation in the k2-weighted χ(k) data, suggest that Hg in the HgH-bio sample is predominantly complexed via carboxyl groups. The differences between the amplitudes and bond distances of HgL-bio and HgH-bio and their similarities with Hgcysteine and Hg-acetate solution standards, respectively, are further illustrated in the magnitude and real part of the Fourier transforms shown in Figure S1a,b (Supporting Information). 9598

dx.doi.org/10.1021/es201820c |Environ. Sci. Technol. 2011, 45, 9597–9603

Environmental Science & Technology

Figure 1. (a) Hg LIII-edge XANES spectra of HgII sorbed to biomass samples at high (HgH-bio) and low (HgL-bio) HgII loadings, with XANES spectra of Hg standards for comparison. (b) k2-weighted χ(k) spectra of Hg LIII-edge EXAFS for high and low loadings of HgII sorbed to biomass samples, with k2-weighted χ(k) spectra of Hg standards for comparison.

The first derivative of XANES data comapring HgH-bio and HgL-bio samples shown in Figure S1c of the SI also suggest that Hg in the HgH-bio and HgL-bio samples is predominantly complexed via carboxyl and sulfhydryl groups respectively.29 No differences in spectra are observed between pH 6.5 and 5.0 for the same metal to biomass ratio (Figure S2 in Supporting Information shows HgL-bio at pH 5 and 6.5), consistent with previous finding that the HgII binding mechanism to biomass does not change over this pH range.24 The EXAFS data from samples HgL-bio and HgH-bio and the Hg standards were modeled quantitatively as described in the SI, by using simultaneous multiple k-weight ﬁts and multiple sample ﬁts. Best-ﬁt values are in Table 1, and the ﬁts are shown in Figure S3 (SI). The best ﬁt for HgL-bio was with 1.85 ((0.18) S atoms at 2.32 ((0.01) Å in the ﬁrst shell. Inclusion of an O/N atom in the ﬁrst shell or a C atom in the second shell did not signiﬁcantly improve the ﬁt (see SI). The best ﬁt for HgH bio was with 1.65 ((0.24) O atoms at 2.06 ((0.01) Å in the ﬁrst shell. Inclusion of a C atom (1.58 ( 0.36) in the second shell signiﬁcantly improved the ﬁt. However, the Hg C distance for the HgH bio sample was 3.05 ((0.02) Å— much longer than the Hg C distance determined for Hg acetate

ARTICLE

solution standard (2.83 ( 0.01 Å)—suggesting the formation of a carboxyl with α-hydroxy carboxylic acid or a malate-type coordination geometry, consistent with previous ﬁndings.22,24 In summary, the EXAFS results suggest an inner-sphere binding mechanism of HgII to biomass. At low metal:biomass, HgII binds to sulfhydryl groups, followed by carboxyl groups on bacterial biomass at higher metal:biomass ratios, consistent with previous ﬁndings.24 Preferential binding of HgII to sulfhydryl groups at low metal:biomass, followed by carboxyl groups at higher metal: biomass, has also been observed for HgII complexation with NOM.22,30 32 Reaction of Biomass-Bound HgII with Magnetite. Hg sorbed to biomass under the conditions described above was reacted with magnetite (HgH/L-bio-magnetite). At pH 6.5, over 90% of the added HgII in the HgH-biomagnetite sample was reduced to Hg0 by magnetite after only 2 h (see Figures 2a and S4 of the SI), indicating a rapid, possibly minute-scale reaction rate. This result is consistent with a previous study of HgII reduction by magnetite in the absence of biomass.11 In contrast to the nearly complete reduction of HgII at pH 6.5 within 2 h, only 60% of the added HgII was reduced to Hg0 at pH 5.0 over the same reaction period (data not shown), indicating slower reaction kinetics at pH 5 than at pH 6.5. However, almost complete reduction was observed at pH 5.0 after 2 d, and the sample remained reduced after 2 months (see Figures 2a, 2b, and S4 of the SI). Slower kinetics of reduction of HgII by magnetite at pH 5.0 than at pH 6.5 have also been observed with HgIIin aqueous solution.11 The similarity of reduction of aqueous HgII and HgII sorbed to biomass under HgH bio-magnetite conditions suggests that complexation to biomass via carboxyl groups does not significantly affect the susceptibility of HgII to reduction by magnetite. HgII in the HgL-bio-magnetite sample was not reduced to Hg0 by magnetite after 2 d or 2 months of reaction time at pH 6.5 and 5.0 (Figures 2c and 2d). XANES spectra of the HgL-bio sample with magnetite after 2 d and 2 months at pH 5.0 and 6.5 match well with the HgL-bio spectrum, suggesting that HgII bound to the sulfhydryl groups on biomass was not reduced by magnetite after 2 months (Figure 2c and 2d). This is also conﬁrmed by the EXAFS data (Figures S2 and S4, Supporting Information). Effect of Magnetite Concentration. The stoichiometry of HgII:FeII was fixed at 1:50 for all experiments described above, while the concentration of biomass remained constant at 2 g 3 L 1 (wet mass). This resulted in a stoichiometric ratio of biomass: magnetite in the HgL bio-magnetite system ten times that of the HgH bio-magnetite system. To test the possibility that coating of the magnetite surface by biomass reduced reactivity in the HgL-bio-magnetite system, we increased the HgII:FeII stoichiometric ratio to 1:500 in the HgL-bio-magnetite system at pH 6.5, where magnetite effectively reduced HgII to Hg0 within 2 h. XANES spectra collected after 2 d reproduced the spectral features of the HgL bio or the HgL bio-magnetite (1:50) spectra, confirming that HgII reduction to Hg0 by magnetite is inhibited by binding of HgII to sulfhydryl groups on biomass, rather than by possible interaction between the bacteria and magnetite (Figure S5 of the SI). Reaction of Biomass-Bound HgII with Green Rust. In light of the inability of magnetite to reduce HgII bound to sulfhydryl groups on biomass, we repeated the experiment with green rust instead of magnetite as the reductant. Green rusts contain up to 75% FeII, are generally stronger reductants than magnetite,33 35 and readily reduce many heavy metals and radionuclides, 9599

dx.doi.org/10.1021/es201820c |Environ. Sci. Technol. 2011, 45, 9597–9603

Environmental Science & Technology

ARTICLE

Table 1. Best-Fit Values for Solution Standards and Hg-bio Samples sample

c

R(Å)

N

σ2(10

3

Å2)

ΔE0(eV)

Hg2+

Hg O

6.12 ( 0.65

2.30 ( 0.01

15.1 ( 3.5

2.0 ( 1.2

HgAc

Hg O

1.78 ( 0.32

2.06 ( 0.01

10.9 ( 0.9

3.2 ( 1.8

Hg C

1.78 a

2.83 ( 0.01

12.8 ( 4.0

Hg O

1.65 ( 0.24

2.06 b

10.9 b

HgH bio

a

path

3.2 b

Hg C

1.58 ( 0.24

3.05 ( 0.02

Hg cysteine

Hg S

1.88 ( 0.21

2.32 ( 0.01

10.5 ( 1.2

1.7 ( 0.9

HgL bio

Hg S

1.85 ( 0.18

2.32 c

10.5 c

1.7 c

12.8

b

Fixed this value to be the same as O based on crystallographic data. b This variable was set to be equal to the HgAc standard during the simultaneous ﬁt. This variable was set to be equal to the Hg-cysteine standard during the simultaneous ﬁt.

Figure 2. Top: Hg LIII-edge XANES spectra at high Hg:biomass ratio (HgH bio) reacted with magnetite at pH 6.5 for (a) 2 h or (b) 2 d and 2 months at pH 5.0, with data for Hg0, Hg2+, and HgH bio samples. Bottom: Hg LIII-edge XANES spectra at low Hg:biomass ratio (HgL bio) reacted with magnetite for 2 d and 2 months at (c) pH 6.5 and (d) pH 5.0, with data for Hg0, Hg2+, and HgH bio samples. The 2-d spectrum is not clearly visible, because 2-d and 2-month spectra overlap.

including HgII.10,36 38 At pH 6.5, partial reduction of HgII to Hg0 by green rust was observed after 2 d (Figure 3), under the experimental conditions where no reduction was observed with magnetite (5 μM HgII, 2 g 3 L 1 biomass, and 250 μM FeII as green rust). A linear combination fit of the XANES spectrum revealed that while 80% of the HgII added to the system remained bound to biomass as a Hg cysteine complex, about 20% of the HgII was reduced to Hg0. First derivative of the Hg XANES spectrum, which is usually more senstive to changes in oxidation state of Hg than Hg XANES, also confirm this observation (Figure S6 of the SI). We did not conduct this experiment at pH 5.0, because green rust becomes unstable at pH 5.39

Summary of Reduction Results. Results of the reactions of magnetite and green rust with HgII complexed to biomass under different conditions are compiled in Table 2. The uncertainty in the XANES analyses is about 10%; therefore, although XANES data indicate 100% reduction in the HgH bio-magnetite system, up to 10% of the Hg in the solid phase might remain oxidized. The results for the HgL biomagnetite system (5 μM added HgII) indicate that the same amount of sulfhydryl-bound Hg probably remains oxidized in the HgH bio-magnetite system (50 μM added HgII). The specific mechanism by which binding to sulfhydryl groups inhibits HgII reduction is not clear. Specifically, this study did not distinguish whether the inihibition is 9600

dx.doi.org/10.1021/es201820c |Environ. Sci. Technol. 2011, 45, 9597–9603

Environmental Science & Technology

ARTICLE

Figure 3. Linear combination ﬁt of the XANES for the low Hg:biomass sample (HgL bio) reacted with green rust at pH 6.5 ((0.2) for 2 d. Of the total HgII sorbed to biomass, 20% was reduced to Hg0, while 80% of the sorbed HgII remained as Hg-cysteine complex after 2 d.

Table 2. Hg XANES and EXAFS Analysis Results for High And Low Loadings of Biomass-Sorbed Hg Sample Reacted with Magnetite and Green Rust at pH 6.5 and 5.0 ((0.2) for Diﬀerent Reaction Timesa sample pH 5.0

50 (μMHg adsorbed to 2 g/L

5 μM Hg adsorbed to 2 g/L

Bacillus subtilis, and reacted with magnetite (2.5 mM FeII)

Bacillus subtilis and reacted with magnetite/GR(250 μM FeII)

2 h-60% reduced

2 days, not reduced

2 days, fully reduced

2 months, not reduced

2 months, fully reduced 6.5

2 h, fully reduced

2 days, not reduced )

2 months, not reduced 2 days (2.5 mM Fe ), not reduced 2 days (with Green Rust), 20% reduced a

The uncertainty in XANES analysis is about 10%.

because sulfhydryl-bound HgII cannot be reduced by magnetite or because the high binding constant of Hg cysteine complexes severely constrains the concentration of dissolved HgII. Additional studies are required to identify the exact mechanism of electron transfer for the reduction of HgII to Hg0. Implications for Subsurface Hg Biogeochemistry. The results of our study are relevant to the fate of HgII in the presence of FeII species in suboxic and anoxic environments. Reducing conditions are commonly encountered in natural aquatic environments. In addition, organic substrates have been injected into the subsurface and groundwater for the biostimulation of native metal-reducing bacteria, to promote in situ bioremediation by the reduction and potential immobilization of metals and radionuclides (e.g., CrVI, TcVII, UVI); however, when Hg is present as a cocontaminant, creation of reducing conditions may have undesired consequences for the speciation and mobility of Hg. Nonetheless, our results show that when conditions are favorable for Hg sorption to sulfhydryl groups on biomass, HgII is unlikely to be reduced to Hg0 by FeII species.

Our studies involved relatively high concentrations of Hg and biomass to enable spectroscopic analyses; however, concentrations of Hg in natural and contaminated geologic settings seldom exceed the nanomolar range. Biomass cell density in natural aquatic environments can also be orders of magnitude lower than those used in this study. However, since the ratio of Hg:biomass determines the nature of HgII complexation to biomass, the results of this study would be applicable under similar Hg:biomass ratios in the environment. For example, a natural environment with 5 nM HgII and 2 mg 3 L 1 biomass would likely exhibit same behavior as the 5 μM Hg and 2 g 3 L 1 biomass conditions in our study. At lower Hg concentrations or alternatively at higher biomass density, HgII forms more stable Hg(cysteine)2 and Hg(cysteine)3 complexes,24 which would likely further limit the availability of HgII complexed with biomass for reduction by mixed FeII/III phases. Hence, the biostimulation of a subsurface environment would likely inhibit the reduction of HgII to Hg0 by mixed FeII/III phases. The use of a model gram-positive aerobic bacterium in this study should not limit the applicability of our results to more complex systems. Sulfhydryl functional groups are ubiquitous in natural environments and have very high aﬃnity for Hg. The relative abundances of functional groups corresponding to the deprotonation constant of cysteine (8.5 ( 1.0), obtained from potentiometric titration data of B. subtilis, Shewanella oneidensis MR-1, and Geobacter sulfurreducens, are 1.0:1.5:2.0, respectively.25,26,40 Moreover, complexation of HgII with sulfhydryl groups is not unique to bacterial biomass. Previous studies have shown HgII binding with natural and dissolved organic matter to be dominated by sulfhydryl groups.17,22,29,41 Recent work has shown that complexation of HgII with sulfhydryl groups is also prevalent in sulﬁde-rich environments.42 Reduction of HgII to Hg0 is a complex biogeochemical phenomenon, with competing microbial and abiotic redox pathways playing a role in surface and subsurface environments. Our results provide new insight into aspects of Hg biogeochemistry necessary for an eﬀective assesment of HgII reduction and remobilization in surface and near-subsurface environments. Previous studies have shown a decline in the availability of HgII to mercuric reductase and in the rate of bacterial HgII reduction to Hg0 with increased cell density.43,44 Although both microbial and abiotic reductions of HgII are less likely under biostimulated conditions because of sorption of HgII to the high-aﬃnity binding sites of biomass—given variations in cell structure, aﬃnity, and biochemical processes—it might not be unreasonable to observe an increase in Hg0 budgets via enzymatic reduction under such conditions. Clearly, additional studies are required to assess the long-term stability of HgII bound to biomass in natural systems. Moreover, iron in the form of FeII sorbed to clays and other minerals is far more common in the environment than FeII mineral phases like magnetite and green rust; hence, it is also important to evaluate the potential for sorbed FeII to reduce HgII complexed with sulfhydryl ligands.

’ ASSOCIATED CONTENT

bS

Supporting Information. Details regarding the experimental procedures, EXAFS data collection and analysis; additional ﬁgures for XANES, derivative of XANES, and EXAFS data and their ﬁts. This material is available free of charge via the Internet at http://pubs.acs.org.

9601

dx.doi.org/10.1021/es201820c |Environ. Sci. Technol. 2011, 45, 9597–9603

Environmental Science & Technology

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT The authors thank Jeremy Fein and Jennifer Szymanowski for providing B. subtilis 168 strains and titration data for G. sulfurreducens. Help from Snow Rui (University of Notre Dame) and Tomohiro Shibata (MRCAT) with XAS data collection is also appreciated. This research is part of the Subsurface Science Scientiﬁc Focus Area at Argonne National Laboratory, supported by the Subsurface Biogeochemical Research Program, Oﬃce of the Biological and Environmental Research, Oﬃce of Science, U.S. Department of Energy (DOE), under contract DE-AC0206CH11357. MRCAT operations are supported by DOE and the MRCAT member institutions. Use of the Advanced Photon Source, an Oﬃce of Science User Facility operated for the DOE Oﬃce of Science by Argonne National Laboratory, was supported by the DOE under contract DE-AC02-06CH11357. ’ REFERENCES (1) 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: J. Human Environ. 2007, 36, 3–11. (2) Fitzgerald, W. F.; Engstrom, D. R.; Mason, R. P.; Nater, E. A. The case for atmospheric mercury contamination in remote areas. Environ. Sci. Technol. 1998, 32, 1–12. (3) Morel, F. M. M.; Kraepiel, A. M. L.; Amyot, M. The chemical cycle and bioaccumulation of mercury. Annu. Rev. Ecol. Syst. 1998, 29, 543–566. (4) Mason, R. P.; Morel, F. M. M.; Hemond, H. F. The role of microorganisms in elemental mercury formation in natural waters. Water, Air, Soil Pollut. 1995, 80, 775–787. (5) Alberts, J. J.; Schindler, J. E.; Miller, R. W.; Nutter, D. E. Elemental mercury evolution mediated by humic acid. Science 1974, 184, 895–897. (6) Skogerboe, R. K.; Wilson, S. A. Reduction of ionic species by fulvic acid. Anal. Chem. 1981, 53, 228–232. (7) Allard, B.; Arsenie, I. Abiotic reduction of mercury by humic substances in aquatic system—An important process for the mercury cycle. Water, Air, Soil Pollut. 1991, 56, 457–464. (8) O’Loughlin, E. J.; Kelly, S. D.; Kemner, K. M.; Csencsits, R.; Cook, R. E. Reduction of AgI, AuII, CuII, and HgII by FeII/FeIII hydrosulfate green rust. Chemosphere 2003, 53, 437–446. (9) Wiatrowski, H. A.; Das, S.; Kukkadapu, R.; Ilton, E. S.; Barkay, T.; Yee, N. Reduction of Hg(II) to Hg(0) by magnetite. Environ. Sci. Technol. 2009, 43, 5307–5313. (10) Amyot, M.; Mierle, G.; Lean, D.; McQueen, D. J. Eﬀect of solar radiation on the formation of dissolved gaseous mercury in temperate lakes. Geochim. Cosmochim. Acta 1997, 61, 975–987. (11) Krabbenhoft, D. P.; Hurley, J. P.; Olson, M. L.; Cleckner, L. B. Diel variability of mercury phase and species distribution in the Florida Everglades. Biogeochem. 1998, 40, 311–325. (12) Fredrickson, J. K.; Zachara, J. M.; Kennedy, D. W.; Dong, H.; Onstott, T. C.; Hinman, N. W.; Li, S. Biogenic iron mineralization accompanying the dissimilatory reduction of hydrous ferric oxide by a groundwater bacterium. Geochim. Cosmochim. Acta 2002, 62, 3239– 3257. (13) Ona-Nguema, G.; Abdelmoula, M.; Jorand, F.; Benali, O.; Gehin, A.; Block, J.-C.; Genin, J-M R. Iron(II,III) Hydroxycarbonate green rust formation and stabilization from lepidocrocite bioreduction. Environ. Sci. Technol. 2002, 36, 16–20.

ARTICLE

(14) Jardim, W. F.; Bisinoti., M. C.; Fadini, P. S.; Silva, G. S. Mercury Redox Chemistry in the Negro River Basin, Amazon: The role of organic matter and solar light. Aqua. Geochem. 2010, 16, 267–278. (15) Gu, B.; Bian, Y.; Miller, C. L.; Dong, W.; Jiang, X.; Liang. L.; Mercury reduction and complexation by natural organic matter in anoxic environments, Proc. Natl. Acad. Sci., 2011, doi:10.1073/pnas.1008747108 (16) Barringer, J. L; Szabo, Z. Overview of investigations into mercury in ground, water, soils, and septage, New Jersey coastal plain. Water, Air, \Soil Pollut. 2006, 175, 193–221. (17) Xia, K.; Skyllberg, U. L.; Bleam, W. F.; Bloor, P. R.; Nater, E. A.; Helmke, P. A. X-ray absorption spectroscopic evidence for the complexation of Hg(II) by reduced sulfur in soil humic substances. Environ. Sci. Technol. 1999, 33, 257–261. (18) Qian, J.; Skyllberg, U.; Frech, W.; Bleam, W. F.; Bloom, P. R.; Petit, P. E. Bonding of methyl mercury to reduced sulfur groups in soil and stream organic matter as determined by x-ray absorption spectroscopy and binding aﬃnity studies. Geochim. Cosmochim. Acta 2002, 66, 3873–3885. (19) Jay, J. A.; Morel, F. M. M.; Hemond, H. F. Mercury speciation in the presence of polysulﬁdes. Environ. Sci. Technol. 2000, 34, 2196–2200. (20) Skyllberg, U.; Xia, K.; Bloom, P. R.; Nater, E. A.; Bleam, W. F. Binding of mercury(II) to reduced sulfur in soil organic matter along upland-peat soil transects. J. Environ. Qual. 2000, 29, 855–865. (21) Skyllberg, U.; Qian, J.; Frech, W. Bonding of methyl mercury to thiol groups in soil and aquatic organic matter. Phys. Scr. 2005, T115, 894–896. (22) Skyllberg, U.; Bloom, P. R.; Qian, J.; Lin, C. M.; Bleam, W. F. Complexation of mercury(II) in soil organic matter: EXAFS evidence for linear two-coordination with reduced sulfur groups. Environ. Sci. Technol. 2006, 40, 4174–4180. (23) Skyllberg, U., Competition among thiols and inorganic sulﬁdes and polysulﬁdes for Hg and MeHg in wetland soils and sediments under suboxic conditions: Illumination of controversies and implications for MeHg net production, J. Geophys. Res., 2008, 113, (24) Mishra, B., Fein, J. B., Yee, N., Beveridge, T. J., Myneni, S. C. B., Cell surface bound Hg complexes inhibits the rate and extent of Hgmethylation, Nat. Geosci., in review. (25) Mishra, B.; Boyanov, M.; Bunker, B.; Kelly, S. D.; Kemner, K. M.; Fein, J. B. High- and low-aﬃnity binding sites for Cd on the bacterial cell walls of Bacillus subtilisand Shewanella oneidensis. Geochim. Cosmochim. Acta 2010, 74, 4219–4233. (26) Fein, J. B.; Boily, J. F.; Yee, N.; Gorman-Lewis, D.; Turner, B. F. Potentiometric titration of Bacillus subtilis cells to low pH and a comparison of modeling approaches. Geochim. Cosmochim. Acta 2005, 69, 1123–1132. (27) Cornell, R. M. Schwertmann, U. The Iron Oxides: Structure, Properties, Reactions, Occurrences, And Uses, 2nd ed., Willey-VCH: New York, 2003. (28) Segre, C. U., Leyarovsky, N. E., Chapman, L. D., Lavender, W. M., Plag, P. W., King, A. S., Kropf, A. J., Bunker, B. A., Kemner, K. M., Dutta, P., Duran, R. S., Kaduk, J., The MRCAT insertion device beamline at the Advanced Photon Source, CP521. In Synchrotron Radiation Instrumentation: Eleventh U.S. National Conference; Pianetta, P., Eds.; American Institute of Physics: NewYork, 2000, 419 422. (29) Rajan, M.; Darrow, J.; Hua, M.; Barnett, B.; Mendoza, M.; Greenﬁeld, B. K.; Andrews, J. C. Hg L3 XANES Study of Mercury Methylation in Shredded Eichhornia crassipes” . Environ. Sci. Technol. 2008, 42, 5568–5573. (30) Hesterberg, D.; Chou, J. W.; Hutchison, K. J.; Sayers, D. E. Bonding of Hg(II) to reduced organic sulfur in humic acid as aﬀected by S/Hg ratio. Environ. Sci. Technol. 2001, 35, 2741–2745. (31) Haitzer, M.; Aiken, G. R.; Ryan, J. N. Binding of Hg(II) to dissolved organic matter: The role of the mercury-to-DOM concentration ratio. Environ. Sci. Technol. 2002, 36, 3564–3570. (32) Drexel, R. T.; Haitzer, M.; Ryan, J. N.; Aiken, G. R.; Nagy, K. L. Mercury(II) sorption to two Florida Everglades peats: Evidence for strong and weak binding and competition by dissolved organic matter released from the peat. Environ. Sci. Technol. 2002, 36, 4058–4064. 9602

dx.doi.org/10.1021/es201820c |Environ. Sci. Technol. 2011, 45, 9597–9603

Environmental Science & Technology

ARTICLE

(33) Elsner, M.; Schwarzenbach, R. P.; Haderlein, S. B. Reactivity of Fe(II)-bearing minerals toward reductive transformation of organic contaminants. Environ. Sci. Technol. 2004, 38, 799–807. (34) Lee, W.; Batchelor, B. Reductive capacity of natural reductants. Environ. Sci. Technol. 2003, 37, 535–541. (35) O’Loughlin, E. J.; Kelly, S. D.; Kemner., K. M. XAFS investigation of the interactions of UVI with secondary mineralization products from the bioreduction of FeIII oxides. Environ. Sci. Technol. 2010, 44, 1656–1661. (36) Loyaux-Lawniczak, S.; Refait, P.; Ehrhardt, J.-J.; Lecomte, P.; Genin, J.-M. R. Trapping of Cr by formation of ferrihydrite during the reduction of chromate ions by Fe(II) Fe(III) hydroxysalt green rusts. Environ. Sci. Technol. 2000, 34, 438–443. (37) Myneni, S. C. B.; Tokunaga, T. K.; Brown, G. E., Jr. Abiotic selenium redox transformations in the presence of Fe(II,III) oxides. Science 1997, 278, 1106–1109. (38) O’Loughlin, E. J.; Kelly, S. D.; Cook, R. E.; Csencsits, R.; Kemner, K. M. Reduction of uranium (VI) by mixed iron(II)/iron(III) hydroxide (green rust): Formation of UO2 nanoparticles. Environ. Sci. Technol. 2003, 37, 721–727. (39) Genin, J-M.R.; Refait, P.; Bourrie, G.; Abdelmoula, M.; Trolard, F. Structure and stability of the Fe (II)-Fe (III) green rust. Appl. Geochem. 2002, 16, 559–570. (40) Private communication with Dr. Jeremy Fein, University of Notre Dame, Notre Dame, (U.S.A.) (41) Ravichandran, M. Interactions between mercury and dissolved organic matter—A review. Chemosphere 2004, 55, 319–331. (42) Miller, C. L.; Mason, R. P.; Gilmour, C. C.; Heyes, A. Inﬂuence of dissolved organic matter on the complexation of mercury under sulﬁdic conditions. Environ. Toxicol. Chem. 2007, 26, 624–633. (43) Rasmussen, L. D.; Turner, R. R.; Barkay, T. Cell-density dependent sensitivity of a mer-lux bioassay. Appl. Environ. Microbiol. 1997, 63, 3291–3293. (44) Wiatrowski, H. A.; Ward, P. M.; Barkay, T. Novel reduction of mercury(II) by mercury-sensitive dissimilatory metal reducing bacteria. Environ. Sci. Technol. 2006, 40, 6690–6696.

9603

dx.doi.org/10.1021/es201820c |Environ. Sci. Technol. 2011, 45, 9597–9603

ARTICLE pubs.acs.org/est

Effect of Radial Directional Dependences and Rainwater Influence on CVOC Concentrations in Tree Core and Birch Sap Samples Taken for Phytoscreening Using HS-SPME-GC/MS Olaf Holm† and Wolfgang Rotard*,† †

Department of Environmental Engineering, Technische Universit€at Berlin, Germany, Strasse des 17. Juni 135, D-10623 Berlin, Germany

bS Supporting Information ABSTRACT: Phytoscreening for chlorinated volatile organic compounds (CVOC) in tree core samples is inﬂuenced by many factors. For instance, greater ﬂuctuations are observed for CVOC concentrations in samples taken around the trunk at a ﬁxed height compared to samples taken directly next to each other. To avoid false negatives and inaccurate interpretation of the results, we investigated this radial directional dependence as well as the inﬂuence of rainwater on measured concentrations. CVOC analysis was performed by gas chromatography/mass spectrometry (GC/MS) following SolidPhase-Microextraction (SPME). Phytoscreening was successfully carried out at three sites using this method. In addition, sap samples were taken from white birches during their budding period as a novel phytoscreening approach. Birch sap sampling is shown to be a suitable means of characterizing contaminant distribution within the soil subsurface. Radial directional dependence of CVOC concentrations varies by almost 80% for tree core samples and 50% for birch sap samples. Variations in concentrations measured around the trunk do not, however, provide information on the inﬂow direction of contaminated groundwater. The weather conditions were shown to have a greater inﬂuence so that CVOC concentrations measured from samples taken during colder, rainier weather were, on average, a factor of 100 lower than those taken during a warm and dry period. Nevertheless phytoscreening is adequate for CVOC characterization in the soil subsurface if the campaign is carried out during a dry weather period, the results then can be taken as being semiquantitative.

’ INTRODUCTION The term phytoscreening is associated with the application of contaminant detection in plants as a means for extensive characterization of contaminant distribution in the subsoil.1 Work in the mid-1990s 2 4 established that volatile organic compounds (VOC) are incorporated by plants through their root uptake of water from the aquifer, soil or soil gas,5 and are transferred throughout the stems via the transpiration stream. Therefore, the use of xylem saps or plant tissues are suitable as sample sources for detection of contaminants. Taking drill cores from tree trunks is the most common sampling technique. Sampling of branches, leaves, and fruits as well as reeds and other types of plants is also possible.3,6,7 The ﬁrst phytoscreening applications based on tree cores were published as early as 1999.8 Since then, several examples of applications have been added documenting not only the diverse utilization but also the limitations of this procedure.1,6,9 For example tree core samples were used to monitor natural attenuation.9 Phytoscreening to trace subsurface contaminations of volatile chlorinated hydrocarbons (CVOC) is widely applied and also by commercial users. Vroblesky10 summarized several aspects of this procedure for VOC in a user guide in 2008, presenting advantages and drawbacks of the technique and its applicability for particular contaminants as well r 2011 American Chemical Society

as a list of factors inﬂuencing the sampling and analysis of tree cores. Various aspects which have a signiﬁcant inﬂuence on both the planning of sampling and the interpretation of results are considered in depth in this paper. These inﬂuential parameters are primarily the eﬀects of rainwater and the sampling location on the tree. High ﬂuctuations were observed for sampling at a given height around the tree trunk.1,5,8,11 The reasons for this radial directional dependence are headed by, among others, the inﬂow direction of contaminated groundwater,10 which should allow undetected points of contamination to be traced. Several of our investigations give additional support to the hypothesis that the contaminant inﬂow direction is decisive for analyte distribution within the tree. Therefore, more detailed examinations concerning radial directionality were performed. Because of the locally uniform groundwater ﬂow at the examined site, it is expected that maximum concentrations in the trees should be measured in direction of the deﬁned contaminant source. Received: June 14, 2011 Accepted: October 10, 2011 Revised: October 6, 2011 Published: October 10, 2011 9604

dx.doi.org/10.1021/es202014h | Environ. Sci. Technol. 2011, 45, 9604–9610

Environmental Science & Technology Additional objectives of this work are to present the application of birch sap samples for phytoscreening and the ﬂuctuations of rainwater inﬂuence and radial directional dependence of both, birch sap and tree core samples. Furthermore, we anticipate increasing concentrations in tree core samples with decreasing rainwater availability. The investigations presented in this paper make use of SolidPhase-Microextraction (SPME) for enrichment of CVOC. SPME combines sampling, analyte isolation, and enrichment and has been widely applied to the sampling and analysis of e.g. environmental samples.12 On the other hand the use of carboxen/ polydimethylsiloxane (PDMS) ﬁbers are not that common. This ﬁber shows very high sensitivity for TCE and cDCE but also a poorer repeatability and prolongation of equilibrium time.13 Other ﬁbers do not show these eﬀects14 but are not as sensitive.

’ EXPERIMENTAL SECTION Site Description. A dry cleaning plant in the northwest area of the former military base Potsdam-Krampnitz (west of Berlin, Germany) has caused considerable groundwater contamination, which distributes subsequently into the neighboring (bordering) wetlands. The geology on the base is predominantly simple with a fixed top of the aquifuge. The groundwater level in the existing wells ranged from 0.85 to 2.40 m. In the northeast of the base the geology is more complex and suggested the situation in the wetlands is complex as well. The groundwater flow is directed to the north into the wetlands. The area of investigation is dominated by two tree species: in the wetland almost exclusively by white willow (Salix alba) and on the base by white birch (Betula alba). Based on direct-push groundwater screening, the concentrations of CVOC reached a maximum of 122 mg CVOC/L.15 Main components are trichloroethene (TCE) and cis-1,2-dichloroethene (cDCE). More details on the site are given in the Supporting Information. The main focus is placed on this contaminated site in Krampnitz. However, the data of two more sites (Hamburg and Neuruppin) are referred for the statistical information only. Sampling and Sample Handling. To avoid any contamination by surface-adsorbed components, the outer layer of the bark was removed from the sampling sites using hammer and chisel before either birch sap or tree core samples were taken. For this investigation, samples were taken at a height of 50 cm above ground level. In addition to the sampling documentation (see Table S3 in the Supporting Information), the coordinates of the trees were determined by GPS. Tree core samples were taken with increment borers (Suunto, Finland; length 15 cm, internal diameter 0.5 cm). The bark was discarded, and the ﬁrst ﬁve centimeters of the xylem wood was collected. A second sample was obtained about two centimeters away from the ﬁrst sample location. Tree core sampling for weather dependency investigations took place in July 2007 on one willow, one poplar, and one birch tree, respectively. In the same month, the radial direction dependence was determined for three birch trees by taking samples at the same height at eight points around the tree. At this stage the investigations were not yet standardized, hence the bark was not discarded and the core lengths varied between 3.8 and 4.3 cm. Further investigations to evaluate variability due to radial direction dependence were undertaken between 22nd of May and 10th of July 2008. Two parallel samples were taken from each of the four compass points for each tree. On two white willows, ﬁve samples were taken in an

ARTICLE

x-shaped conﬁguration with a maximum distance of two centimeter between them. Altogether, 19 trees were sampled (eleven white willows, ﬁve birches plus one poplar, one oak, and one apple tree). Birch sap samples were taken by drilling an eight centimeter deep hole using a cordless drill (Bosch, model: PSR 14,4 Li-2). This was followed by pressing in a brass spreading dowel (eight mm with M6 internal thread) which had initially been covered with Teﬂon tape. Final insertion was completed with a hammer if necessary. A specially constructed aluminum tube with external thread, also covered with Teﬂon tape, was screwed in. To catch the birch sap, a vial was hung on a nut screwed onto the tube. The vial and the sampling syringe were ﬁrst rinsed. In general it takes only a few seconds, round about 10 s, to ﬁll the vial. Ten milliliters of birch sap was carefully transferred from the bottom of the vial into a second vial, while the sap is ﬂooding. The vial was immediately sealed. Therefore the potential loss of CVOC is negligible. In general, the ﬂow allowed the ﬁlling of at least two further vials. To close the wound after completion of sampling, the aluminum tube was unscrewed, and the hole was sealed with a threaded bolt covered with Teﬂon tape. To determine any radial direction dependence, six birches were sampled at eight points around the trunk, and six more birches were each sampled on sides facing and opposite to the contaminant source, all at the same height. To compare the results with those from tree cores, 31 holes were made with an increment borer as described above, and the respective tree cores and birch sap samples from each resulting core hole were analyzed separately. Data from all trees depicted in the graphics are compiled in Table S3, and their locations are shown in Figure S3 in the Supporting Information. All samples were transported and stored at room temperature. Analysis of the gas space in the sampling vial took place by means of SPME followed by GC/MS determination (see below) within 24 h of sampling. The fresh tree cores were subsequently weighed. Their water content was then determined by oven drying at 105 °C for at least 24 h and reweighing. Solid-Phase-Microextraction (SPME). The SPME system used consisted of a carboxen/PDMS coated fiber (Supelco) attached to a plunger within a protective needle, which directly pierced the septum of the vial thus being exposed to the headspace within the sample vial. The fibers showed slightly differences in the sensitivity to each other and a decrease in sensitivity when used repeatedly. Assuming a linear relation between the loss of ﬁber sensitivity and the number of measurements, it is possible to implement a drift factor based on aqueous standards measured at the beginning and end of a sample series. Applying the calculated drift, each peak area was corrected to the value that would have been obtained had the sample been on a fresh ﬁber. Using aqueous standards, it was likewise possible to calculate compensation factors for the diﬀerences between individual ﬁbers within a measurement campaign. Each sample peak area value was multiplied by a factor relating to the ratio between the peak areas of standards using a chosen reference ﬁber and the peak areas of the standards for the actual ﬁber used. Data collected in 2007 relating to weather dependency and radial direction dependence were neither corrected for drift, nor was a ﬁber comparison carried out. The number of measurements within the relevant series was suﬃciently low so that corrections were not required. The samples were conditioned and extracted each for 30 min at 35 °C. The ﬁber then is desorbed in the GC-injector for 1 min 9605

dx.doi.org/10.1021/es202014h |Environ. Sci. Technol. 2011, 45, 9604–9610

Environmental Science & Technology

ARTICLE

Table 1. Semiquantitative Evaluation Scale Including Peak Area Relationships and Corresponding Concentrations for TCE and cDCE in Aqueous Samplesa scale

a

peak area

TCE [ng/L]

cDCE [ng/L]

1

nd

2

uncertain

3

500 10,000

3 67

2 49

4

10,000 25,000

67 169

49 122

5 6

25,000 50,000 50,000 100,000

169 337 337 674

122 243 243 487 487 4870

7

100,000 1,000,000

674 6740

8

1,000,000 10,000,000

6740 67,400

4870 48,700

9

10,000,000 100,000,000

67,400 674,000

48,700 487,000

10

>100,000,000

>674,000

>487,000

nd: not detected.

at 250 °C (see Table S2 in the Supporting Information). The calibrations are made by external standards with aqueous standards under the same conditions, which is widely accepted for SPME.16 The conditions of the subsequent GC/MS measurements and annotations to the calibration (Table 1) are summarized in the Supporting Information. Procedures of Data Processing. Graphical representations of contaminant distribution were based on the semiquantitative results, taking the main mass peak areas from the parallel samples. Gradation of the semiquantitative evaluation scale is given in Table 1. Variations in the results from a single tree are represented by normalizing the measured values to the maximum concentration (c/cmax) or to the maximum content of the applicable series of measurements, respectively. Assuming that a tree core length correlates to its weight, the 2007 samples, which were taken before the sampling technique had been standardized, were weight corrected. The results of the radial directionality tests are presented in three forms. In the ﬁrst, a simple network diagram, the relative concentrations are shown in relation to compass points. For the second representation, the tree core sample results were evaluated using the semiquantitative scale and ﬁnally interpolated using the program Surfer (Version 8.05; Golden Software, Inc.) in kriging mode. This results in a two-dimensional representation of the isolines from all four compass points as well as of the previously averaged values. The third representation uses vectors to emphasize the direction for which the highest concentrations were measured. The vector representation uses the average from multiple tree core sampling on the south side, from which the value from the north is subtracted; similarly, the west from the east. These two vectors were then added. The resulting vector was then divided by the sum of all concentration values of the four compass points. The vector length thus becomes a relative value of the extent of concentration diﬀerences between the radially distributed sampling points on the tree. These radial concentration diﬀerence (RCD) vectors were recalculated into polar coordinates and described by the directional angle and length (see Figure 1). The maximum possible vector length value equal to one indicates that contaminants were only measured in one direction. The RCD vector is represented by an arrow, beginning at the point of the respective tree’s coordinates. Additionally, the RCD vectors were used to evaluate the correlation with respect to the four compass points as well as the direction of the contamination source. This required the

Figure 1. Schematic illustration for the determination of the radial concentration diﬀerence (RCD) vectors.

measurement of northings and eastings (positioning coordinates) of the trees and the angle with respect to the point of contamination, stated by the position of the storage tanks. Thus, the angular deviations relative to the contaminant source with respect to the compass points could be calculated for each tree. The smaller the resulting angle is, the greater was the agreement between the directions of the highest measured concentration to the reference direction. By using a factor which relates to the length of the RCD vector, these angles may be normalized so that the trees with especially high internal concentration diﬀerences may be more strongly weighted when determining a mean value of the angular deviation of all trees. From the reference-related averaging of these angles for all trees, it is possible to estimate the probability that the highest concentration will be measured in the reference direction.

’ RESULTS Radial Directionality for Samples Taken at the Same Height. Tree core samples (duplicate samples) of the phyto-

screening applications showed average variation coefficients for various CVOCs ranging from 15% to 22%, with the exception of PCE at the site in Hamburg with 40% (Table 2). Radial directionality examinations also support this variation range. In contrast, radial directionality values vary by 79% for cDCE and 63% for TCE. Variations in birch sap sample results show a similar trend. Variations from multiple single core samples show a much lower mean variation coefficient of 7% compared to radially distributed samples with 50%. Although they are somewhat lower than those for the tree cores, they nevertheless clearly show dependence on the radial direction. The network diagram shown in Figure 2 also supports the argument that this eﬀect is not caused by statistical errors. Moreover, it can be seen that an area of higher concentrations is formed with pronounced directionality, as illustrated by the relatively similar shapes of the structures depicted. Statistical errors would have resulted in much more diﬀuse forms. The very similar mean variation coeﬃcients for birch sap sampling from diametrically opposite sample sites and from the eight directions 9606

dx.doi.org/10.1021/es202014h |Environ. Sci. Technol. 2011, 45, 9604–9610

Environmental Science & Technology

ARTICLE

Table 2. Summary of the Mean Relative Standard Deviations for Substances and Campaignsb mean relative standard deviation number of campaign

maximum

average

in %

in %

substance trees/samples Duplicate Samples PCE

9

104

40

1,2-DCA

13

36

15

cDCE

19

57

15

TCE

23

83

22

phytoscreening

tDCE

12

40

17

Krampnitz

cDCE

35

58

20

TCE TCE

120 59

130 31

21 7

tree coring

cDCE

51

116

21

(radial samples)

TCE

60

88

16

tree coring

cDCE

8

40

17

(radial samples)

TCE

8

36

20

birch sap (total)

TCE

phytoscreening Hamburga phytoscreening Neuruppina

birch sap

Quintuplicate Samples

Radial Distributed Samples 12

132

50

birch sap (two Sides) TCE birch sap (radial) TCE

6 6

91 132

46 54

tree coring

cDCE

14

136

79

TCE

15

122

63

Samples were taken at a height of 100 cm. b The “number of trees/ samples” represent the number of samples in which the substances are detected.

a

Figure 2. Radial concentration diﬀerences from birch sap samples; above: relative concentrations related to the maximal value for the particular tree; below: semiquantitative evaluation.

also argue against statistical errors. Therefore, a systematic radial direction dependence is indicated. All three trees sampled in 2007 show highest core concentrations from the side nearest to the contamination source (see Figure S8 in the Supporting Information). The birch sap samples from early 2008 do not,

however, support this. Core samples were taken from Tree 01 in 2007 and birch sap in 2008. The TCE concentration maximum shows a clear shift from a southwest direction (contamination source) toward the north (compare Figure 2 with Figure S8 in the Supporting Information). The RCD vectors for TCE and cDCE resulting from the 2008 radial tree core sampling campaign are presented in Figure 3. The direction as well as the magnitude of the radial variations as depicted by RCD vector length appears to be somewhat random. No dependence on tree species or concentration gradient of subsurface contaminants (compare Figure 3 with Figure 4 or Figure S2 in the Supporting Information) can be observed. Completely diﬀerent vectors are seen for closely spaced single species trees of similar size or age. Angular variations with respect to the compass points and the contamination source are lower for the latter. However, the diﬀerences are small (Table 3). Figure 4 shows several isolines based on the semiquantitative evaluation (Table 1) from the radial tree core sampling campaign which enable interpolation in the case of a ﬁxed sampling direction. It should be considered that samples were taken under varying weather conditions. This has, however, no relevance for the eﬀects on semiquantitative evaluation. The diﬀerences for ﬁxed sampling directions (in this case the four compass points) are only small when applying semiquantitative evaluation. Comparison of Birch Sap and Tree Cores. Parallel sampling of ten milliliter birch sap and five centimeter diameter birch tree cores resulted in almost identical results for TCE. The ratio of peak areas in birch sap to tree cores for 31 samples was around 1.0 with a variation coefficient of 20%. Birch sap was, however, clearly more sensitive for cDCE, which could be detected in 29 of the 31 samples but in only 22 of the respective tree cores. The ratio for these 22 values was around 5.0, with a variation coefficient of 66%. With respect to radial directionality, the birch sap and tree core samples show similar distribution patterns (compare Figure 2 with Figure S8 in the Supporting Information). The Influence of the Weather. Figure 5 shows the influence of an extreme change in weather conditions on TCE concentrations. On the first day of sampling (11th of July, 2007), just as during the days preceding the sampling campaign, the daily maximum temperature was around 15 °C accompanied by continuous medium to heavy rainfall. On the 14th of July, 2007, the temperature climbed to a maximum of over 30 °C with no rain. The following days remained dry with similar temperatures. The values measured during this campaign on the colder, rainier days were, on average, a factor of 100 lower, single values being up to 1000 times lower. This trend was conﬁrmed for all three trees and also for cDCE (data not shown). The increase in concentrations occurred simultaneously for two of the trees, with the third following one day later.

’ DISCUSSION Radial directional dependence in sampling was investigated to assess possible ﬂuctuations in results obtained in relation to the context of phytoscreening and to verify whether the inﬂow direction of contaminated groundwater contributes to this. The results of radial directionality correspond closely to previously published values. Various studies1,5,8,11 have shown that the variations in amounts of contaminants from tree core samples taken at the same height around the trunk are signiﬁcantly greater than for sampling points placed directly next to each other (duplicate samples). 9607

dx.doi.org/10.1021/es202014h |Environ. Sci. Technol. 2011, 45, 9604–9610

Environmental Science & Technology

ARTICLE

Figure 3. Radial concentration diﬀerence (RCD) vectors for trichloroethene (TCE, left), for cis-1,2-dichloroethene (cDCE, right) and tree location. The lengths of the RCD vectors do not reﬂect concentration levels but rather the extent of variation between the radially distributed tree core abstractions for each tree.

Figure 4. Isolines from semiquantitative evaluation of values related to the four compass points and the previously averaged values; left: for trichloroethene (TCE); right: for cis-1,2-dichloroethene (cDCE).

Table 3. Average Angular Deviation between the RCD Vector and the Contaminant Source for the Four Compass Points substance

contamination

north

east

south

west

source [deg]

[deg]

[deg]

[deg]

[deg]

c-DCE

69

76

82

104

98

TCE

69

94

73

86

107

The published variations in tree cores around the trunk are presented in diﬀerent ways and show maxima of around 90% for TCE and cDCE8 and a factor of 5 for TCE1 and PCE,5 respectively. A single examination of birch sap samples from Tree 05 implies that the radial concentration diﬀerences converge with increasing sampling height, as discussed previously.8 The examinations presented here show that systematic radial dependence in tree cores at the same height exists. At this site this does not, however, allow any conclusions on the inﬂow direction of contaminated groundwater or the gradient of the contaminants in the subsurface. Concentration variations for radially distributed samples may involve a number of inﬂuencing factors, such as sorption,17 decomposition,4 and diﬀusion.18 The results

Figure 5. Relative concentrations (logarithmic scale) for Tree 04 related to the maximum of a set of measurements during a change in both temperatures from 15 °C to more than 30 °C and in daily rainfall.

of the birch sap sampling show that sap transport plays a substantial role. Exposure to the sun and heterogeneous soil and root structures are regarded to cause radial diﬀerences in sap ﬂow.1,10 Contaminated sites are often subject to anthropogenic inﬂuences, so that surface sealing or root capping can contribute to heterogeneous sap ﬂow. Also, each tree’s individual structure is especially relevant.19 In addition, it appears that the distribution characteristics of CVOC around the trunk are not constant over time. In Tree 01, the TCE concentration maximum shows a clear shift between tree core sampling in 2007 and birch sap sampling in 2008 (compare Figure 2 with Figure S8 in the Supporting Information). Both sampling campaigns were carried out at the same sampling height on the tree, although the tree cores from 2007 were not taken at exactly the four compass points. Any inﬂuence of spiral growth20 on the diﬀerent concentration distributions can thus be excluded. Compared to variations due to dependency on radial diﬀerence (Figure 2), the weather (Figure 5) has a much greater 9608

dx.doi.org/10.1021/es202014h |Environ. Sci. Technol. 2011, 45, 9604–9610

Environmental Science & Technology inﬂuence on overall variation and leads to a clear change of the categorization when semiquantitative evaluation is applied. In four of twelve samples, cDCE could not be detected attributable to the cold, wet weather. On dry days, the sampled trees could be allocated to categories 7 and 8 on the semiquantitative evaluation scale (Table 1). For TCE, the classiﬁcation ranged from categories 3 to 9. Vroblesky11 previously illustrated the impact of precipitation on VOC concentration through an irrigation experiment, which resulted in a factor of less than two. This is below variations due to radial diﬀerence dependence.1,5,8,11 A rapid response to irrigation was seen within one day during the above experiment, namely at all sampled heights up to 3.5 m above ground level. Diﬀerent inﬂuencing factors such as sorption,17 decomposition,4 and diﬀusion18 may well lead to changes in the spectrum of contaminants within the tree. In the period of active growth, these processes apparently play a minor role for CVOC. The concentration increase presented here shows the great dependence on weather conditions. Trees only take up groundwater from the saturated zone when there is insuﬃcient water in the unsaturated zone. At sites where, due to climatic inﬂuences, trees mainly take up groundwater, this inﬂuence is probably not as serious. Doucette21 documented factors of 10 to 100 for variations in contaminant levels at climatically diﬀerent sites with comparable groundwater concentrations. Volatile contaminants such as CVOC diﬀuse out of the tree trunk. Due to loss to air, concentrations usually decrease with height.5,8,18,22,23 The gradients shown by Vroblesky11 and Sorek et al.1 do not match this concept. Our own height proﬁles range from unclear to inverse gradients (data not shown). A change in contaminant concentration of the extracted water results in a likewise change of concentration in the tree. Depending on the retention, a step gradient may occur and migrate upward along the tree. In addition, a shift in contaminant spectrum with increased height may be expected. Ma and Burken18 assume that, based on losses by diﬀusion, the depth horizontical proﬁle will show higher concentrations inside the trunk. Our own examinations show no consistent depth proﬁles (data not shown). Migration gradients caused by slow diﬀusive transport processes are likely responsible for these ﬁndings. Based on the many factors inﬂuencing water uptake, deposition, and behavior of CVOC,10 only a semiquantitative evaluation of the measured concentration is meaningful. Nevertheless, these evaluations lead to meaningful contaminant images at all three sites, which coincide closely with those from direct push groundwater samples.15 Sampling at diﬀerent directions raises the probability of taking a sample from the side of highest concentration. As could be shown, additional eﬀort in taking radially distributed samples is actually not required for semiquantitative evaluations. The limitation to sampling of one point is reasonable, however critical for trees, whose analyte concentrations are close to the detection limit. Sampling from one ﬁxed direction, e.g. the sunny side 1 of a tree, is not decisive, considering the results presented here. Rainwater has a considerable inﬂuence on concentrations of TCE and cDCE measured in the tree. Sampling in the course of a phytoscreening should therefore be undertaken during dry weather. Implications for Phytoscreening. Birch sap sampling is a suitable means of characterizing contaminant distribution within the soil subsurface. The difference in sensitivity relative to cDCE and TCE from birch sap and tree cores from the same point on the tree confirms the dependence of the kind of substance. For better correlation factors to groundwater contents, birch sap

ARTICLE

sampling is an interesting option. However, the method is only applicable during a very narrow annual time window. The points of measurement cannot be used over periods of several years, since the tree reacts to such injuries by building various reaction or barrier zones24 thus sealing off this area. The sampling port already seals itself within a few days, so that it is not even usable for the roughly four week long budding period. The sap flow is certainly weather dependent. Birch sap sampling could be repeated at points which had already dried up.

’ ASSOCIATED CONTENT

bS

Supporting Information. Detailed site description. Data of tree location and sampling conditions. Conditions of SPME extraction, GC/MS measurements and annotations to the calibration with aqueous standards. Radial directional concentrations of tree cores in 2007. Height proﬁles. Geological data and groundwater levels. Interpolated groundwater concentrations. This material is available free of charge via the Internet at http:// pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: +49 30 314-25220/-21978. Fax: +49 30 314-29319. E-mail: [email protected].

’ ACKNOWLEDGMENT This work was funded by the German Federal Ministry of Education and Research, project SINBRA, contract no. 0330757D. The authors also thank A. Horn, F. Jaeger, S. Klemer, W. Seis, F. Zietzschmann, R. Hatton, and W. Frenzel for their assistance in preparing this manuscript. ’ REFERENCES (1) Sorek, A.; Atzmon, N.; Dahan, O.; Gerstl, Z.; Kushisin, L.; Laor, Y.; Mingelgrin, U.; Nasser, A.; Ronen, D.; Tsechansky, L.; Weisbrod, N.; Graber, E. R. ”Phytoscreening”: The Use of Trees for Discovering Subsurface Contamination by VOCs. Environ. Sci. Technol. 2008, 42 (2), 536–542. (2) Burken, J.; Dietz, A.; Jordahl, J.; Schnabel, W.; Thompson, P.; Licht, L.; Alvarez, P.; Schnoor, J., Phytoremediation at Hazardous Waste Sites. Proceedings - WEFTEC ’96, Annual Conference & Exposition, 69th, Dallas, Oct. 5 9, 1996 1996, 1, 327-332. (3) Schnabel, W. E.; Dietz, A. C.; Burken, J. G.; Schnoor, J. L.; Alvarez, P. J. Uptake and Transformation of Trichloroethylene by Edible Garden Plants. Water Res. 1997, 31 (4), 816–824. (4) Newman, L. A.; Strand, S. E.; Choe, N.; Duﬀy, J.; Ekuan, G.; Ruszaj, M.; Shurtleﬀ, B. B.; Wilmoth, J.; Heilman, P.; Gordon, M. P. Uptake and Biotransformation of Trichloroethylene by Hybrid Poplars. Environ. Sci. Technol. 1997, 31 (4), 1062–1067. (5) Schumacher, J. G.; Struckhoﬀ, G. C.; Burken, J. G. Contamination Using Tree Cores at the Front Street Site and a Former Dry Cleaning Facility at the Riverfront Superfund Site, New Haven, Missouri, 1999 2003; 2004 5049; Virginia, 2004; p 41. (6) Gopalakrishnan, G.; Negri, M. C.; Minsker, B. S.; Werth, C. J. Monitoring subsurface contamination using tree branches. Ground Water Monit. Rem. 2007, 27 (1), 65–74. (7) Chard, B. K.; Doucette, W. J.; Chard, J. K.; Bugbee, B.; Gorder, K. Trichloroethylene uptake by apple and peach trees and transfer to fruit. Environ. Sci. Technol. 2006, 40 (15), 4788–4793. 9609

dx.doi.org/10.1021/es202014h |Environ. Sci. Technol. 2011, 45, 9604–9610

Environmental Science & Technology

ARTICLE

(8) Vroblesky, D. A.; Nietch, C. T.; Morris, J. T. Chlorinated Ethenes from Groundwater in Tree Trunks. Environ. Sci. Technol. 1999, 33 (3), 510–515. (9) Larsen, M.; Burken, J.; Machackova, J.; Gosewinkel Karlson, U.; Trapp, S. Using Tree Core Samples to Monitor Natural Attenuation and Plume Distribution After a PCE Spill. Environ. Sci. Technol. 2008, 42 (5), 1711–1717. (10) Vroblesky, D. A. User s Guide to the Collection and Analysis of Tree Cores to Assess the Distribution of Subsurface Volatile Organic Compounds; 2008 5088; Virginia, 2008; p 59. (11) Vroblesky, D. A.; Clinton, B. D.; Vose, J. M.; Casey, C. C.; Harvey, G.; Bradley, P. M. Ground Water Chlorinated Ethenes in Tree Trunks: Case Studies, Inﬂuence of Recharge, and Potential Degradation Mechanism. Ground Water Monit. Rem. 2004, 24 (3), 124–138. (12) Ouyang, G. F.; Vuckovic, D.; Pawliszyn, J. Nondestructive Sampling of Living Systems Using in Vivo Solid-Phase Microextraction. Chem. Rev. 2011, 111 (4), 2784–2814. (13) Popp, P.; Paschke, A. Solid phase microextraction of volatile organic compounds using carboxen-polydimethylsiloxane ﬁbers. Chromatographia 1997, 46 (7 8), 419–424. (14) Avila, M. A. S.; Breiter, R.; Mott, H. Development of a Simple, Accurate SPME-based Method for Assay of VOCs in Column Breakthrough Experiments. Chemosphere 2007, 66 (1), 18–29. (15) Rein, A.; Popp, S.; Leven, C.; Bittens, M.; Dietrich, P., Comparison of approaches for the characterization of contamination at rural megasites. Environ. Earth Sci. 2010 (Online-First-Version), In Press. (16) Pawliszyn, J.; Ouyang, G. Recent developments in SPME for on-site analysis and monitoring. TrAC, Trends Anal. Chem. 2006, 25 (7), 692–703. (17) Trapp, S.; Miglioranza, K. S. B.; Mosbaek, H. Sorption of lipophilic organic compounds to wood and implications for their environmental fate. Environ. Sci. Technol. 2001, 35 (8), 1561–1566. (18) Ma, X.; Burken, J. G. TCE Diﬀusion to the Atmosphere in Phytoremediation Applications. Environ. Sci. Technol. 2003, 37 (11), 2534–2539. (19) Cohen, Y.; Cohen, S.; Cantuarias-Aviles, T.; Schiller, G. Variations in the Radial Gradient of Sap Velocity in Trunks of Forest and Fruit Trees. Plant Soil 2008, 305 (1 2), 49–59. (20) Vite, J. P.; Rudinsky, J. A. Water-conducting systems in conifers and their importance to the distribution of trunk-injected chemicals. Contrib. Boyce Thompson Inst. 1959, 20, 27–38. (21) Doucette, W. J.; Bugbee, B. G.; Smith, S. C.; Pajak, C. J.; Ginn, J. S., Uptake, metabolism, and phytovolatilization of trichloroethylene by indigenous vegetation: impact of precipitation. In Phytoremediation; McCutcheon, S. C.; J., S. J., Eds.; 2003; pp 561-588. (22) Baduru, K. K.; Trapp, S.; Burken, J. G. Direct Measurement of VOC Diﬀusivities in Tree Tissues: Impacts on Tree-Based Phytoremediation and Plant Contamination. Environ. Sci. Technol. 2008, 42 (4), 1268–1275. (23) Trapp, S. Fruit Tree Model for Uptake of Organic Compounds from Soil and Air. SAR QSAR Environ. Res. 2007, 18 (3 4), 367–387. (24) Shigo, A. L. Compartmentalization of Decay in Trees. Sci. Am. 1985, 252 (4), 96–103.

9610

dx.doi.org/10.1021/es202014h |Environ. Sci. Technol. 2011, 45, 9604–9610

ARTICLE pubs.acs.org/est

Interactive Priming of Biochar and Labile Organic Matter Mineralization in a Smectite-Rich Soil Alexandra Keith,† Balwant Singh,*,† and Bhupinder Pal Singh‡ † ‡

Faculty of Agriculture, Food, and Natural Resources, The University of Sydney, NSW, 2006, Australia NSW Department of Primary Industries, P.O. Box 100, Beecroft, NSW 2119, Australia

bS Supporting Information ABSTRACT: Biochar is considered as an attractive tool for long-term carbon (C) storage in soil. However, there is limited knowledge about the eﬀect of labile organic matter (LOM) on biochar-C mineralization in soil or the vice versa. An incubation experiment (20 °C) was conducted for 120 days to quantify the interactive priming eﬀects of biochar-C and LOM-C mineralization in a smectitic clayey soil. Sugar cane residue (source of LOM) at a rate of 0, 1, 2, and 4% (w/w) in combination with two wood biochars (450 and 550 °C) at a rate of 2% (w/w) were applied to the soil. The use of biochars (∼ 36%) and LOM (12.7%) or soil (14.3%) with isotopically distinct δ13C values allowed the quantiﬁcation of C mineralized from biochar and LOM/soil. A small fraction (0.41.1%) of the applied biochar-C was mineralized, and the mineralization of biochar-C increased signiﬁcantly with increasing application rates of LOM, especially during the early stages of incubation. Concurrently, biochar application reduced the mineralization of LOM-C, and the magnitude of this eﬀect increased with increasing rate of LOM addition. Over time, the interactive priming of biochar-C and LOM-C mineralization was stabilized. Biochar application possesses a considerable merit for long-term soil C-sequestration, and it has a stabilizing eﬀect on LOM in soil.

’ INTRODUCTION Biochar production, the centuries old tradition of heating organic residues under oxygen limited conditions for application to soil, is now the focus of a rapidly expanding area of research. The interest in biochar is manifold and stems from the observations on old agricultural soils, called Terra Preta, in the Amazon Basin. These soils were treated with charcoal (or biochar) creating much higher soil fertility and carbon (C) content than the neighboring natural soils.1 Based on these observations, biochars produced by pyrolyzing organic waste materials in thermal reactors have been promoted as soil amendments. Biochar application to the soil can reduce greenhouse gas emissions,2 decrease the availability of heavy metals,3 and beneﬁt soil fertility and plant productivity.1,4,5 Furthermore, the demonstrated large mean residence time (MRT) of natural char or biochar in soils and sediments68 has generated interest in the use of biochar for increasing C-sequestration in soil. Consequently, the production and application of biochar to soil is considered to possess considerable greenhouse gas emissions mitigation beneﬁts compared to conventional management of biomass feedstocks.9,10 Despite the recalcitrant nature of biochars, research shows that biochar oxidizes both by abiotic and biotic mechanisms.8,11,12 The stability of biochar-C in soils is dependent on several factors, including the properties of biochar and soil, and environmental conditions.7,13 Furthermore, biochar application may aﬀect the r 2011 American Chemical Society

mineralization rate of native soil organic matter (SOM), and similarly the addition labile organic matter (LOM), that is, organic material that mineralizes more rapidly than biochar or native SOM, may also impact biochar-C mineralization in soil.13,16 There are conﬂicting reports on the interactive priming eﬀects on mineralization of biochar-C and LOM-C in the soil.1316 Here, we deﬁne the interactive, positive or negative, priming eﬀect as the stimulation or suppression, respectively, of biochar-C or LOM-C mineralization above or below the respective control.17 Liang et al.13 studied the eﬀects of black C, a form of biochar-type C, in Anthrosols (Terra Preta soils) on the C mineralization and cycling of relatively labile plant residues, that is, sugar cane leaves. They reported that the presence of black C in the soil caused rapid incorporation of added LOM into aggregates and organo-mineral fractions and thus stabilizing LOM in soil. Contrary to the study of Liang et al.,13 Wardle et al.14 reported that charcoal presence promoted the loss of humus-C from the forest ﬂoor in a 10-year old study at three contrasting boreal forest sites in northern Sweden. However, neither of these studies13,14 observed the eﬀect of added or native organic matter on the mineralization of biochar-C. Black C in the Received: June 27, 2011 Accepted: September 28, 2011 Revised: September 13, 2011 Published: September 28, 2011 9611

dx.doi.org/10.1021/es202186j | Environ. Sci. Technol. 2011, 45, 9611–9618

Environmental Science & Technology

ARTICLE

Table 1. Important Properties of the Soil, Biochars, and Sugar Cane Residue Used in the Incubation Experiment property

soil

450 °C biochar

550 °C biochar

sugar cane residue

total C (%)

0.45

67.8

74.9

total N (%)

0.04

0.5

0.6

39.6 0.4

δ13C

14.3

36.3

36.4

12.7

pH (1:5 H2O)

8.10

8.60

9.90

Electrical conductivity (1:5, dS m1)

0.11

0.90

1.10

CECa (mmolc kg1)

347

11.4

54.0

clay (%)

53

silt (%) sand (%)

14 33

speciﬁc surface areab (m2 g1)

191.0

228.3

pore volume (%)

57.2

67.5

clay minerals

S****, K*, I*

a Cation-exchange capacity measured by the silver thiorurea method. b CO2 adsorption method: S**** = smectite (>80%); K* = kaolinite (<10%); and I* = Illite (<10%).

soils used by Liang et al.13 was several hundred to thousands of years old and thus likely to be deprived of relatively labile organic C components because of their mineralization over time, whereas Wardle et al.14 used litter bags and it was not possible to detect small changes in relatively labile charcoal-C using the mass balance approach.15 In contrast to these studies, Hamer et al.17 and Kuzyakov et al.18 found increased mineralization of biochar-C by repeated addition of glucose-C; this eﬀect was attributed to cometabolism resulting from increased microbial biomass and the simultaneous increase in enzyme production. Consistent with the study of Wardle et al.,14 increased mineralization of LOM in the presence of charred-/biochar-C has been observed by other researchers.17,19 The presence of charred-C enhanced the loss of added glucose-C from carbon-free sand.17 In another study, the addition of pecan-shell biochar increased the overall C mineralization from a loamy sand soil amended with switchgrass.19 More recently Zimmerman et al.16 observed both negative and positive priming eﬀects from incubation of grass and wood biochars in sandy soils. The positive priming (i.e., increased C mineralization) was mainly observed in the low temperature biochars in the ﬁrst 90 days of the experiment. Over a longer-term (250500 days) mineralization of both biocharand soil-C was suppressed; this eﬀect was attributed to sorption of native SOM to surfaces and pore spaces of biochars and C stabilization through enhanced organic-mineral interactions.16 The authors undertook a limited analysis, one measurement each at early and late stages, to determine the priming eﬀect through C source identiﬁcation in soils amended with grass biochars. It is clear from the above discussion that the direction and magnitude of the interactive priming eﬀects of biochar and LOM (native or added) on C mineralization are unclear and need urgent attention to realize the potential of biochar for long-term sequestration of C in the soil. In natural systems, variable amounts of organic residues are added to soils, which may impact biochar-C mineralization by sustaining high microbial growth and activity in soils. In the present study, we used diﬀerent rates of sugar cane residues (a relatively labile source of C compared to biochar-C or native SOM) in combination with two woody biochars of diﬀerent lability to determine the interactive priming eﬀects on organic C mineralization in soil. Unlike earlier studies16,17,19 using sand or sandy soils or biochar-humus mixtures without soil minerals,14 we have incubated biochar and

LOM mixtures (at increasing rates of LOM addition) in a smectite-rich clayey soil to incorporate the inﬂuence of organomineral interactions on organic C mineralization. We employed novel 13C-depleted biochars to unambiguously identify source of C (biochar or added LOM/native SOM) mineralized in the soil.

’ MATERIALS AND METHODS Soil, Biochar, and Organic Matter. Surface sample (010 cm) of a Vertisol was collected from an experimental plot that has been grown with C4 tussock Mitchell grass (Astrebla spp.) for over 100 years. The soil was air-dried, sieved through 2 mm, and analyzed for basic characteristics including pH, particle size, cation-exchange capacity, and exchangeable cations.20 Clay minerals were identified by X-ray diffraction analysis (details given in Supporting Information) of basally oriented specimens after various pretreatments.21 For separating C mineralization sources, we used wood biochar with contrasting δ13C values to soil and LOM (Table 1). The 13C-depleted woody biomass was collected from Eucalyptus salinga that had been grown for two years under elevated CO2 environment.22 The woody biomass was used to produce two biochars, one each at 450 and 550 °C, by slow pyrolysis (510 °C/min heating rate, 40 min residence time) at Paciﬁc Pyrolysis Australia. The biochars were ground to <2 mm before mixing with the soil and LOM. Sugar cane residue (C4 biomass), obtained from a local nursery, was used as the source of LOM. The dried residue was ground to <2 mm using a pestle and mortar before mixing with the soil and biochar. Total C and nitrogen contents of soil, biochar, and sugar cane residue were determined using a TruSpec CHN Leco analyzer. Inorganic carbon was absent in the soil, and therefore, the total C represents the organic carbon content of the soil. Important chemical and physical properties of the soil, biochars, and sugar cane residue are given in Table 1. Incubation Experiment. Treatments consisted of all combinations of two biochars, that is, 450 and 550 °C biochars, applied at 0 and 2% w/w and sugar cane residue added at four rates (0, 1, 2, or 4% w/w). The soil (200 g oven-dried basis) was homogenously mixed with the biochars and sugar cane residue, and all treatments were replicated three times. A nutrient solution23 and a microbial inoculum were mixed with deionized water, and the volume was made to obtain 60% water 9612

dx.doi.org/10.1021/es202186j |Environ. Sci. Technol. 2011, 45, 9611–9618

Environmental Science & Technology

ARTICLE

Figure 1. Cumulative carbon (C) mineralized (mg CO2C/g C), for 0, 1, 2, and 4% labile organic matter (LOM) with no biochar and in the presence of 450 °C (B450) and 550 °C (B550) biochars, during the 120 day incubation period. The plotted values are averages of the three replicates, and standard errors are represented by the error bars; the scales for the vertical axis are diﬀerent for 0% LOM and the other LOM rates graphs.

holding capacity of soil mixtures, which was maintained during incubation (more details provided in the Supporting Information). Each experimental unit was set up in 1.2 L sealed plastic containers, which held three small containers: a 200 mL container with soil, biochar, and/or LOM mixture, a 70 mL jar containing NaOH (2 or 2.5 M) to capture CO2 produced following C mineralization, and a 70 mL jar containing 40 mL of water to maintain a constant humidity. The incubation experiment was conducted over a period of 120 days, and CO2 traps were replaced after 1, 2, 5, 8, 13, 20, 30, 44, 65, 90, and 120 days of incubation. The containers were incubated in the dark at a constant temperature of 20 ( 1 °C in the laboratory. Three “blank” 1.2 L incubation chambers, without soil but with the sealed jars to make up the volume, were set up to account for the atmospheric CO2 present in the headspace of the incubation chambers. At each sampling day, the NaOH jars were sealed immediately upon removal from the incubation containers and subsequently replaced with the new ones. The NaOH jars were stored in sealed containers along with an open NaOH jar to absorb CO2 in the enclosed atmosphere. Carbon Mineralization Rate, Source, Priming Effects, and Mean Residence Time. The total C mineralized in various treatments was determined by titrating a known volume of NaOH against 0.1 M HCl (see the Supporting Information). To determine the amount of C mineralized from biochar or LOM (and/or soil), the trapped CO2 in an aliquot of NaOH was precipitated with SrCl2, and δ13C analysis was performed on the SrCO3 precipitate. The soil, fresh biochar, and sugar cane residue were also analyzed for δ13C (see the Supporting Information).

From the δ13C data, the proportion of CO2C derived from biochar (CBiochar) was determined by the following equation:24 CBiochar ð%Þ ¼

ðδT 13 CO2 δSLOM 13 CO2 Þ 100 ðδB 13 C δSLOM 13 CO2 Þ

ð1Þ

where δT13CO2 is the δ13C value of total CO2C evolved from biochar and LOM (sugar cane residue) amended soil, δSLOM13CO2 is the δ13C value for the CO2C evolved from LOM amended/ nonamended soil, and δB13C is the 13C isotopic composition of the fresh biochar. The proportion of CO2C derived from soil or soil + LOM was determined by subtracting CBiochar(%) from 100. In the soil + LOM treatments, the quantity of C mineralized from each source (soil-C or LOM-C) was indistinguishable because of their similar δ13C values (Table 1). The priming eﬀect of biochar-C (i.e., changes in soil C or soil + LOM C mineralization induced by biochar) on soil ( LOM-C ) was determined as follows:24 (PESLOM B PEBSLOM ¼ CB, SLOM CSLOM

ð2Þ

where, CB,SLOM is the amount (mg kg1 soil) of soil C or soil + LOM C mineralized from biochar-amended soil and CSLOM is the amount (mg kg1 soil) of soil C or soil + LOM C mineralized from control soil (without biochar). The priming eﬀect of LOM (i.e., changes in biochar-C mineralization induced by LOM addition) on biochar-C (PELOMB) was determined as follows: ¼ CB, SLOM CB PELOM B 9613

ð3Þ

dx.doi.org/10.1021/es202186j |Environ. Sci. Technol. 2011, 45, 9611–9618

Environmental Science & Technology

ARTICLE

Figure 2. Proportion (%) of the applied biochar carbon (C) mineralized during the 120 day incubation period in the presence of 0, 1, 2, and 4% labile organic matter (LOM). B450 and B550 represent 450 °C biochar and 550 °C biochar, respectively.

where CB,SLOM is the amount (mg kg1 soil) of biochar C mineralized from biochar-amended soil in the presence of LOM and CB is the amount (mg kg1 soil) of biochar C mineralized from biochar-amended soil in the absence of LOM. Biochar-C mineralization data were ﬁtted to the two-pool exponential model to determine the MRT of biochar-C in the soil (see the Supporting Information). Statistical Analysis. Significant differences between treatments were calculated using two-way analysis of variance (ANOVA) in GenStat 11th edition at different times of incubation.25 The treatments biochar and LOM were used as independent and interactive sources of variance. When significant F-tests were obtained (0.05 probability level), means separation was achieved using a least significant difference (LSD) test at the 0.05 probability level, unless otherwise stated.

’ RESULTS Total Carbon Mineralization. The total C mineralized (measured as CO2C produced on per unit C basis), both cumulative and daily rate, showed a significant (P < 0.001) positive increase with increasing addition rate of LOM (Figure 1 and Figure S1 of the Supporting Information). The quantity and rate of total CO2C produced were significantly decreased by the presence of biochars in all LOM treatments. The cumulative C mineralized from 550 °C biochar applied soil was significantly lower than the 450 °C biochar samples. The maximum C mineralization rate was observed on the second day of incubation in all treatments except for the 0% LOM treatments where the maximum CO2C flux occurred on the first day (Figure S1 of the Supporting Information). A second, smaller peak occurred

on day 8 in all treatments, but overall there was an exponential decline in the CO2C flux with time. In the absence of applied LOM (i.e., the 0% LOM treatment), biochar contributed on an average 31% (450 °C biochar) and 21% (550 °C biochar) to the total C mineralized (Figure S2 of the Supporting Information). In comparison, within the 1, 2, and 4% LOM treatments, biochars contributed only 4.46.7, 3.34.6, and 3.13.4%, respectively, to the total C mineralized. Albeit, the proportion of C mineralized from the 550 °C biochar was signiﬁcantly lower than that of the 450 °C biochar at all levels of LOM, with the exception of days 1, 2, and 3 where it contributed a greater proportion to the total C mineralized than the 450 °C biochar. Mineralization of Biochar and LOM. The proportion of biochar-C mineralized during the 120 day incubation period ranged between 0.4 and 1.1% of the total biochar-C added, and it was significantly higher for the 450 °C biochar than the 550 °C biochar at all levels of LOM (Figure 2). Within each LOM treatment, total mineralization of the 450 °C biochar was between 1.5 and 1.9 times that of the 550 °C biochar. BiocharC mineralization rate increased significantly with each addition of LOM during early stages of incubation (days 2, 3, and 5) and decreased or stabilized during later stages (Figure S3 of the Supporting Information). Consequently, the cumulative biochar-C mineralized over 120 days was significantly increased with each incremental LOM treatment. Similar to the total CO2C flux data, biochar-C mineralization occurred in two stages, the first peak occurred on the day 1 or 2, where the majority of mineralization occurred, and the second and a smaller peak occurred on the day 8 (Figure S3 of the Supporting Information). In all treatments, following the initial rapid mineralization of 9614

dx.doi.org/10.1021/es202186j |Environ. Sci. Technol. 2011, 45, 9611–9618

Environmental Science & Technology

ARTICLE

Figure 3. Cumulative priming eﬀect of labile organic matter (LOM) on the mineralization of biochar carbon (C) during the 120 day incubation period. Error bars represent standard error of the mean (n = 3).

biochar-C, the rate of biochar-C mineralization followed a strong, negative exponential trend, and biochar was being mineralized at a much lower rate toward the end of incubation period (Figure 2 and Figure S3 of the Supporting Information). The estimated MRT of biochars, derived from the short-term data in the study, ranged between 62 and 112 years for 450 °C biochar and between 100 and 248 years for 550 °C biochar across various LOM treatments (Table S1 of the Supporting Information). During the 120 day incubation period between 4.2 and 13.6% of the total soil and/or LOM was mineralized, and there was no signiﬁcant diﬀerence in the fraction of the soil and/or LOM-C mineralized with the biochar type at all addition rates of LOM (data not shown). The mineralization rate of LOM (and soil-C) increased signiﬁcantly with incremental addition of LOM (Figure S4 of the Supporting Information); it followed a trend similar to the biochar-C mineralization data (Figure S3 of the Supporting Information). Interactive Priming Effects of LOM and Biochar. The priming effect of LOM on biochar-C mineralization and vice versa (interactive priming) was variable both with respect to the application rate of LOM and the incubation time (Figure 3). On day 1, there was a negative priming effect of LOM at all levels on both biochars, which became positive from the day 2 and increased steadily for up to 3 weeks for the 1 and 2% LOM treatments, and for up to 46 weeks in the 4% LOM treatment. The higher the LOM addition rate the greater was its initial positive priming effect on biochar-C mineralization. The positive priming then reached a plateau and overturned in the 1 and 2% LOM treatments or stabilized in the 4% LOM treatment for the remaining incubation period. In the 1% LOM treatment, the overall priming effect of LOM on biochar mineralization became negative after approximately 9 weeks of incubation and was

slightly positive with 2% LOM at the end of the incubation period. There was no significant difference in the magnitude of priming effect of LOM between 450 and 550 °C biochar, except during early stages of incubation in the 2% and 4% LOM treatments. In the absence of LOM (0% LOM), biochar exhibited an overall positive priming eﬀect (i.e., increased mineralization) on the native soil C during 120 d of incubation (Figure 4). However in the LOM treated soils, biochar exhibited a net negative priming eﬀect on the mineralization of LOM-C. The negative priming eﬀect of biochar on LOM-C mineralization increased signiﬁcantly with increasing rate of LOM addition. Within the 1 and 2% LOM treatments, the priming eﬀect was similar for the two biochars whereas with the 4% LOM treatment the 550 °C biochar produced a more pronounced negative priming eﬀect (though statistically nonsigniﬁcant) than the 450 °C biochar. The majority of the LOM-C priming occurred within the ﬁrst 30 days, which appeared to have stabilized thereafter (Figure 4). This temporal pattern of the LOM-C priming by biochar corresponds well with that of biochar-C priming by LOM, albeit in opposite direction initially (Figure 3).

’ DISCUSSION Biochar-C Mineralization. The overall mineralization of biochar-C in various treatments was relatively small (0.41.1%) and is within the range (0.14.7%) reported for biochars, made from different feedstocks and produced at different pyrolysis temperatures, over the similar time period.1618,24 Hamer et al.17 reported 0.30.8% C mineralization for oak-wood and crop residue biochars (charred at 800 °C) in the absence of LOM, which increased to 0.61.2% with glucose addition. Kuzyakov et al.18 also reported enhanced C mineralization of ryegrass biochar 9615

dx.doi.org/10.1021/es202186j |Environ. Sci. Technol. 2011, 45, 9611–9618

Environmental Science & Technology

ARTICLE

Figure 4. Cumulative priming eﬀect of biochar on the mineralization of carbon (C) in soil, or soil plus labile organic matter (LOM), during the 120 day incubation period. Error bars represent standard error of the mean (n = 3).

(charred at 400 °C) in soils with repeated additions of glucose-C and following repeated disruption of soil aggregates to increase the availability of native soil C and stimulate soil microbial activity. These results highlight the importance of the availability of labile-C, pyrolysis temperature, and biomass type in relation to the stability of biochar-C.12,17,18 Our own data support these observations, as biochar-C mineralization increased with increasing addition of LOM, and there was consistently greater mineralization of C from the biochar produced at 450 °C than the 550 °C biochar with all combinations of LOM treatments. Biochars produced at higher pyrolysis/charring temperatures contain a greater proportion of condensed aromatic structures,26 which contribute to its greater recalcitrance.8,27 The initial (days 1 and 2) higher C mineralization rate from the 550 °C than the 450 °C biochar (Figure S3 of the Supporting Information) may be due its high specific area and pore volume, thereby providing greater accessibility to labile C in biochar by microbes compared to the 450 °C biochar (Table S1 of the Supporting Information). The rapid decrease in biochar-C mineralization rate after the first 2 days (Figure S3 of the Supporting Information) suggests that the biochars contain a small proportion of highly labile C that was mineralized or volatilized quickly following incorporation in the soil. An additional peak of biochar mineralization on Day 8 of the incubation could represent the peak of biotic mineralization of biochar as the lag time is indicative of establishing microbial populations. The biochar-C mineralization rate remained higher for the 450 °C biochar than the 550 °C biochar suggesting the presence of greater fraction of relatively easily mineralizable C (nonaromatic-C) in the low-temperature biochar.26

Interactive Priming of Biochar and LOM Carbon Mineralization. The observation on the increased rate of biochar-C

mineralization with increasing levels of LOM (i.e., positive priming) is consistent with earlier studies where a significant increase in biochar-C mineralization was reported after the addition of glucose-C.17,18 This had been attributed to cometabolic mineralization of biochar-C by the microbial enzymes that were produced to utilize the glucose. The increased mineralization of biochar-C in the presence of LOM in our study was sustained over a relatively longer period of time, which possibly occurred either due to (i) a greater quantity of labile C pool in the system and/or (ii) the C from LOM was relatively less labile and more complex than glucose-C. The observed reversal or stabilization of the positive priming effect of LOM on biochar-C mineralization with time may have occurred due to several reasons, such as (i) the labile C in biochar was used up, (ii) the most labile C pool of LOM that caused the priming effect was used up, (iii) organic matter derived from LOM covered biochar surfaces and clogged pores, (iv) biochar particles were entrapped within soil aggregates, and (v) biochar particles formed complexes with smectite. Contrary to our results, Liang et al.13 observed no priming eﬀect of LOM on biochar-C mineralization. However, in their study, native SOM and added biochar have coexisted for hundreds to thousands of years in the ﬁeld compared to our study where “fresh” biochar was applied to the soil. We suggest that the positive priming eﬀect of LOM on biochar is time dependent; the cometabolic priming eﬀect of LOM on biochar-C mineralization decreases concurrently with the possible decrease in the labile components such as, volatile organic C, including aliphatic and dissolved organic C in the biochar (Figures 3 and 4). 9616

dx.doi.org/10.1021/es202186j |Environ. Sci. Technol. 2011, 45, 9611–9618

Environmental Science & Technology As higher levels of LOM resulted in greater positive priming, it can be expected that, in these instances, the labile C components of biochar were depleted quicker; however, the cometabolic eﬀect of LOM on the mineralization of recalcitrant C components of biochar (aromatic-C) appeared to have continued and needs further investigation. Additionally, reactive surfaces of aged biochars may provide greater protection to LOM through chemical interactions as compared to relatively fresh biochars. Although native soil organic C mineralization was increased initially in the presence of biochars (Figure 4), our results on the stabilization of LOM-C mineralization in the presence of biochars are consistent with the ﬁndings of Liang et al.13 who also observed stabilization of LOM decomposition through its rapid incorporation into organo-mineral fractions in biochar-amended Anthrosols. Other studies14,17 have reported contrary results that show increased mineralization of labile forms of C in the presence of biochar; however, clay minerals were not present in the system used in these studies. Soil minerals play an important role in the stabilization of added organic matter due to organo-mineral interactions,28 and the addition of LOM in the smectite rich soil, a highly reactive clay mineral, is likely to enhance organo-mineral interactions and stabilize organic C in the biochar-amended soil. It is also possible that some of the metabolites produced during the decomposition of LOM may have diﬀused into the pore-space of biochar and adsorbed thus remained protected from biotic and abiotic decomposition.16 A similar stabilization mechanism has been suggested for the sequestration of hydrophobic organic contaminants by soot type materials.29 Organic metabolites may also remain protected in the interlayers of smectite mineral, which was abundant in the system. Major et al.30 also observed a consistently greater soil surface CO2 eﬄux in the presence of biochar; however, they attributed this eﬀect to the greater production of plant biomass following biochar application, leading to increased contribution of plant-derived C to total soil CO2C evolved. The positive priming of native soil organic C (0% LOM treatment) with biochar application suggests that the small proportion of labile carbon components in biochars have the potential to increase the mineralization of the native SOM. The results further suggest that the priming eﬀect of biochar is diﬀerent on SOM and LOM. The priming eﬀect of biochar on SOM may be signiﬁcant in soils with low organic C content and needs further research. C-Sequestration Implications. Several studies have recognized the potential of biochar as a suitable tool for C sequestration in soil; the MRT data from our study support this assertion. Considering the short-term duration and the ideal mineralization conditions existing in the laboratory study, care must taken to extrapolate the MRT of biochars under realistic field situations. Nevertheless, our data show the potential benefits of biochar application in the stabilization of LOM in the soil; this benefit is slightly offset by the increased mineralization of labile components of biochar in the presence of LOM. As organic matter addition to the soil is a regular occurrence, both naturally and within most agricultural situations, it is likely that over long-term the interactive priming effects between biochar and LOM and subsequent organo-mineral interactions will result in an overall increased C sequestration in the soil from biochar addition; this aspect is being investigated through a long-term incubation study. Biochar produced at the higher temperature (550 °C) resulted in greater soil- and LOM-C sequestration than at the lower

ARTICLE

(450 °C) temperature; thus, application of biochar would be desirable for increased soil C-sequestration, particularly combined with its high chemical recalcitrance.8,12,26

’ ASSOCIATED CONTENT

bS

Supporting Information. Additional information on the soil, incubation experiment, SSA and pore volume analyses, X-ray diﬀraction analysis, CO2 and δ13C analyses, the MRT of biochars, a table presenting the MRT values and the associated parameters (Table S1), and ﬁgures related to total soil C mineralization rate (Figure S1), the proportion of biochar-C to the total C mineralized (Figure S2), biochar-C mineralization rate (Figure S3), soil-C or soil- plus LOM-C mineralization rate (Figure S4), and the proportion of initial soil-C or soil- plus LOM-C mineralized (Figure S5). This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: +61 2 86271140; e-mail: [email protected].

’ ACKNOWLEDGMENT We express our gratitude to Claudia Keitel at the University of Sydney for the δ13C analysis; Yunying Fang (The University of Sydney) and Kamaljeet Kaur and David Giles (NSW Department of Primary Industries) for their assistance with the laboratory work; Craig Barton (NSW Department of Primary Industries) and Burhan Amiji (University of Western Sydney) for help with procuring the woody biomass; and Scott Donne (University of Newcastle) for the speciﬁc surface area and pore volume analyses of the biochars. We also thank David Phelps and former Peggy Olsson at the Queensland Department of Employment, Economic Development and Innovation for supplying the soil. ’ REFERENCES (1) Glaser, B.; Lehmann, J.; Zech, W. Ameliorating physical and chemical properties of highly weathered soils in the tropics with charcoal - a review. Biol. Fert. Soils 2002, 35, 219–230. (2) Singh, B. P.; Hatton, B. J.; Singh, B.; Cowie, A. L.; Kathuria, A. Inﬂuence of biochars on nitrous oxide emission and nitrogen leaching from two contrasting soils. J. Environ. Qual. 2010, 39 (4), 1224–1235. (3) Namgay, T.; Singh, B.; Singh, B. P. Inﬂuence of biochar application to soil on the availability of As, Cd, Cu, Pb, and Zn to maize (Zea mays L.). Aust. J. Soil Res. 2010, 48, 638–647. (4) Lehmann, J.; Gaunt, J.; Rondon, M. Biochar sequestration in terrestrial ecosystems: a review. Mitigation Adapt. Strategies Global Change 2006, 11, 403–427. (5) Kookana, R. S.; Sarmah, A. K.; Van Zwieten, L.; Krull, E.; Singh, B. Biochar application to soil: Agronomic and environmental beneﬁts and unintended consequences. Adv. Agron. 2011, 112, 103–143. (6) Schmidt, M. W. I.; Noack, A. G. Black carbon in soils and sediments: Analysis, distribution, implications, and current challenges. Global Biogeochem. Cycles 2000, 14, 777–793. (7) Cheng, C. H.; Lehmann, J.; Engelhard, M. H. Natural oxidation of black carbon in soils: Changes in molecular form and surface charge along a climosequence. Geochim. Cosmochim. Acta 2008, 72 (6), 1598–1610. (8) Zimmerman, A. R. Abiotic and microbial oxidation of laboratoryproduced black carbon (biochar). Environ. Sci. Technol. 2010, 44 (4), 1295–1301. 9617

dx.doi.org/10.1021/es202186j |Environ. Sci. Technol. 2011, 45, 9611–9618

Environmental Science & Technology (9) Roberts, K. G.; Gloy, B. A.; Joseph, S.; Scott, N. R.; Lehmann, J. Life cycle assessment of biochar systems; estimating the energetic, economic and climate change potential. Environ. Sci. Technol. 2010, 44 (2), 827–833. (10) Woolf, D.; Amonette, J. E.; Street-Perrott, F. A.; Lehmann, J.; Joseph, S. Sustainable biochar to mitigate global climate change. Nat. Commun. 2010, 1, 56. (11) Cheng, C. H.; Lehmann, J.; Thies, J. E.; Burton, S. D.; Engelhard, M. H. Oxidation of black carbon by biotic and abiotic processes. Org. Geochem. 2006, 37 (11), 1477–1488. (12) Nguyen, B. T.; Lehmann, J.; Hockaday, W. C.; Joseph, S.; Masiello, C. A. Temperature sensitivity of black carbon decomposition and oxidation. Environ. Sci. Technol. 2010, 44, 3324–3331. (13) Liang, B.; Lehmann, J.; Solomon, D.; Sohi, S.; Thies, J. E.; Skjemstad, J. O.; Luizao, F. J.; Engelhard, M. H.; Neves, E. G.; Wirick, S. Black carbon aﬀects the cycling of non-black carbon in soil. Org. Geochem. 2010, 41, 206–213. (14) Wardle, D. A.; Nilsson, M. C.; Zackrisson, O. Fire-derived charcoal causes loss of forest humus. Science 2008, 320, 629–629. (15) Lehmann, J.; Sohi, S. Comment on “Fire-Derived Charcoal Causes Loss of Forest Humus. Science 2008, 321, 1295. (16) Zimmerman, A. R.; Gao, B.; Ahn, M.Y. Positive and negative carbon mineralization priming eﬀects among a variety of biocharamended soils. Soil Biol. Biochem. 2011, 43, 1169–1179. (17) Hamer, U.; Marschner, B.; Brodowski, S.; Amelung, W. Interactive priming of black carbon and glucose mineralisation. Org. Geochem. 2004, 35 (7), 823–830. (18) Kuzyakov, Y.; Subbotina, I.; Chen, H.; Bogomolova, I.; Xu, X. Black carbon decomposition and incorporation into soil microbial biomass estimated by 14C labeling. Soil Biol. Biochem. 2009, 41, 210–219. (19) Novak, J. M.; Busscher, W. J.; Watts, D. W.; Laird, D. A.; Ahmedna, M. A.; Niandou, M. A. S. Short-term CO2 mineralization after additions of biochar and switchgrass to a Typic Kandiudult. Geoderma 2010, 154 (34), 281–288. (20) Rayment, G.; Higginson, F. Australian Laboratory Handbook of Soil and Water Chemical Methods; Inkata Press: Melbourne, Australia, 1992. (21) Brown, G.; Brindley, G. W. X-ray Diﬀraction Procedures for Clay Mineral Identiﬁcation. In Crystal Structures of Minerals and Their Identiﬁcation; Brindley, G. W., Brown, G., Eds.; Mineralogical Society: London, 1984; pp 305360. (22) Barton, C. V. M.; Ellsworth, D. S.; Medlyn, B. E.; Remko, A. D.; Tissue, D. T.; Adams, M. A.; Eamus, D.; Conroy, J. P.; McMurtrie, R. E.; Parsby, J.; Linder, S. Whole-tree chambers for elevated atmospheric CO2 experimentation and tree scale ﬂux measurements in south-eastern Australia: the Hawkesbury Forest experiment. Agric. For. Meteorol. 2010, 150, 941–951. (23) Chapman, S. J. Carbon substrate mineralization and sulphur limitation. Soil Biol. Biochem. 1997, 29, 115–122. (24) Kuzyakov, Y.; Bol, R. Sources and mechanisms of priming eﬀect induced in two grassland soils amended with slurry and sugar. Soil Biol. Biochem. 2006, 38, 747–758. (25) Payne, R. W.; Murray, D. A.; Harding, S. A.; Baird, D. B.; Soutar, D. M. GenStat for Windows Introduction, 12th, ed.; VSN International: Hemel Hempstead, U.K., 2009. (26) McBeath, A. V.; Smernik, R. J. Variations in the degree of aromatic condensation of chars. Org. Geochem. 2009, 40, 1161–1168. (27) Baldock, J. A.; Smernik, R. J. Chemical composition and bioavailability of thermally altered Pinus resinosa (Red pine) wood. Org. Geochem. 2002, 33 (9), 1093–1109. (28) Sollins, P.; Kramer, M. G.; Swanston, C.; Lajtha, K.; Filley, T.; Aufdenkampe, A. K.; Wagai, R.; Bowden, R. D. Sequential density fractionation across soils of contrasting mineralogy: evidence for both microbial- and mineral-controlled soil organic matter stabilization. Biogeochem. 2009, 96, 209–231. (29) Luthy, R. G.; Aiken, G. R.; Brusseau, M. L.; Cunningham, S. D.; Gschwend, P. M.; Pignatello, J. J.; Reinhard, M.; Traina, S. J.; Weber,

ARTICLE

W. J., Jr.; Westall, J. C. Sequestration of hydrophobic organic contaminants by geosorbents. Environ. Sci. Technol. 1997, 31 (12), 3341–3347. (30) Major, J.; Lehmann, J.; Rondon, M.; Goodale, C. Fate of soilapplied black carbon: downward migration, leaching and soil respiration. Global Change Biol. 2010, 16 (4), 1366–1379.

9618

dx.doi.org/10.1021/es202186j |Environ. Sci. Technol. 2011, 45, 9611–9618

ARTICLE pubs.acs.org/est

Chemical Looping Combustion in a Rotating Bed Reactor Finding Optimal Process Conditions for Prototype Reactor † Silje Fosse Hakonsen and Richard Blom*,† †

SINTEF Materials & Chemistry, P.O. Box 124 Blindern, N-0314 Oslo, Norway

bS Supporting Information ABSTRACT: A lab-scale rotating bed reactor for chemical looping combustion has been designed, constructed, and tested using a CuO/Al2O3 oxygen carrier and methane as fuel. Process parameters such as bed rotating frequency, gas ﬂows, and reactor temperature have been varied to ﬁnd optimal performance of the prototype reactor. Around 90% CH4 conversion and >90% CO2 capture eﬃciency based on converted methane have been obtained. Stable operation has been accomplished over several hours, and also stable operation can be regained after intentionally running into unstable conditions. Relatively high gas velocities are used to avoid fully reduced oxygen carrier in part of the bed. Potential CO2 purity obtained is in the range 30 to 65% mostly due to air slippage from the air sector which seems to be the major drawback of the prototype reactor design. Considering the prototype nature of the ﬁrst version of the rotating reactor setup, it is believed that signiﬁcant improvements can be made to further avoid gas mixing in future modiﬁed and up-scaled reactor versions.

’ INTRODUCTION A chemical looping combustion (CLC) process was already suggested in the 1950s as a way to produce pure carbon dioxide.1 It was further developed as a combustion technique in the 1980s2 and later in the 1990s presented as a possible way to separate CO2 during fossil fuel combustion.3 The interest in CLC has boosted during the past decade due to its relatively high net energy eﬃciency4,5 and potential low cost of CO2 capture.6 CLC is a cyclic process where a metal oxide ﬁrst is used to combust a fuel, and then the reduced metal oxide is reoxidized in air before a new cycle can be carried out. Such a red-ox cycle can in principle be carried out in two ways; either i) by moving the metal oxide between static gas streams or ii) by keeping the metal oxide static while switching the gas streams. Option i) is in most cases implemented by a circulating ﬂuidized bed (CFB) reactor setup where the metal oxide powder circulates between a fuel reactor in which the combustion takes place and an air reactor where reoxidation takes place.7,8 CFB reactors have recently gained far the most attention within the CLC community since this reactor type already is commercial for combustion processes (boilers) and within reﬁnery processes such as ﬂuidized catalytic cracking (FCC). Option ii) most often involves one or more ﬁxed bed reactors where complex valving sequences ensure cyclic gas feeding to the reactors and optimal gas separation. Early CLC experiments were carried out in single ﬁxed bed reactors.3,9 We have recently developed an alternative reactor concept for CLC which belongs to the option i) group above in which the metal oxide is kept in a doughnut shaped ﬁxed bed that rotates between the diﬀerent gas streams ﬂowing radially outward through the bed. A schematic drawing of the reactor along the rotation axis is shown in Figure 1. It is believed that a radial gas ﬂow is a r 2011 American Chemical Society

good choice for a CLC process due to gas expansion caused by temperature increase and increase in moles of gas in the system. This could be handled by exploiting the natural increase in bed volume by going from small to large radius. A similar reactor concept was already suggested for “conversion of organic reactants to other organic products” in 1955.10 Rotating bed reactors have also previously been suggested for CO2 temperature swing sorption processes, but, to our knowledge, only at a conceptual level.11 In addition, the use of rotating packed bed (RPB) reactors for CO2 separation using alkanolamine solutions has recently been demonstrated.12 We have in an earlier communication described the basics of our rotating bed reactor system showing that separation of the gases is possible, although some internal gas mixing does occur.13 The present paper results from the ﬁrst series of experiments carried out at elevated temperatures using methane as fuel. A supported copper oxide (CuO/Al2O3) oxygen carrier has been used. CuO is chosen for two reasons: First because of its fast red-ox kinetics already at relatively low temperatures (<800 °C).14 Secondly due to the exothermicity of both the reduction reaction (4CuO + CH4 f CO2 + 2H2O + 4Cu) and the reoxidation reaction (4Cu + 2O2 f 4CuO) with ΔHo of 206 and 596 kJ/mol, respectively.13 More extreme division of the total combustion energy between the two reactor sectors (fuel and air) might lead to larger temperature gradients within the reactor and thus more material stress and possibilities for lockage during bed rotation. It Received: June 30, 2011 Accepted: October 4, 2011 Revised: October 3, 2011 Published: October 04, 2011 9619

dx.doi.org/10.1021/es202244t | Environ. Sci. Technol. 2011, 45, 9619–9626

Environmental Science & Technology is expected that industrial CLC should operate at temperatures well above 1000 °C where Cu/CuO based oxygen carriers are useless, so the ultimate use of a rotating bed CLC reactor should also tolerate such temperatures.15 However, for demonstrating the concept without coming into unexpected thermal problems we have chosen to operate at relatively low temperatures (up to 800 °C). Increasing the temperature of the system will be important when going to the next version of the rotating bed reactor. To ﬁnd the optimal performance of the prototype reactor we have varied process parameters such as reactor temperature (650800 °C), total and relative gas ﬂow rates, and rotation frequency of the bed. The observed tendencies are discussed, and improvements in the reactor design are suggested. Finally, the feasibility of large scale reactors of this kind has brieﬂy been discussed.

’ EXPERIMENTAL SECTION Reactor Design, Construction, Infrastructure, and Oxygen Carrier Material Used. A prototype lab-scale reactor has been

designed and constructed based on typical capacity and kinetic data of the most known oxygen carrier materials.16,19 The central part is a rotating bed holder having the following dimensions: outer diameter 60 mm, bed thickness approximately 12 mm, holder height 60 mm. The holder is filled with 120 g CuO/ Al2O3 spheres having 10 wt % oxygen capacity and fast kinetics

Figure 1. Schematic drawing of the rotating reactor along the rotation axis. The gas is fed radially outward. The gases are fed in the diﬀerent inner sectors: 1) fuel sector, 2) steam sector, and 3) air sector. The light green part denotes the rotating oxygen carrier bed. The thick solid lines are separation walls, while the dotted lines indicate the various sectors.

ARTICLE

at temperatures around 700 °C as described elsewhere.13 The 1.5 mm (diameter) CuO/Al2O3 spheres have specific surface area of 115 m2/g (BET) indicating high rest porosity after Cuimpregnation. The reactor is divided into four feeding sections (see Figure 1): one air section (240°), one fuel section (60°), and two steam sections (2 30°). The steam sections are used to prevent air and fuel from mixing. For simplicity, argon has been used in the steam sectors in the following experiments. The two exit chambers are divided by walls positioned in the middle of the steam sections to reduce internal mixing. In the prototype these walls are simple single walled pieces. The position of the diﬀerent gas feeding sections can be adjusted relative to the walls dividing the two exit chambers, making it possible to compensate for the rotation of the reactor. The central part of the reactor containing the housing of the rotating bed is kept well isolated. Rotation of the oxygen carrier bed is done by coupling the bed holder to a shaft that can be rotated from the outside by a drive belt. Ball bearings between the static inner tube and the rotating part at the top and bottom of the reactor ensure smooth rotation. Water cooling above and below the central part assures low temperatures of the ball bearings containing polymeric sealing. Figure 2 shows the reactor infrastructure. The rig is automated using Labview. A 400W oven in the center of the reactor (the red central circle in Figure 1) is used to bring the reactor to reaction temperature. All ﬂows to the diﬀerent feeding zones can be controlled by mass-ﬂow controllers. The ﬂows through the two reactor exits are monitored by calibrated rotameters (gas ﬂow meters) on each line, and needle valves are used to regulate the exit ﬂows manually. Water cooled condensers are installed on the two exit lines prior to the two rotameters to remove produced water. On the fuel side, an online gas chromatograph (GC, Agilent 6890) was used to analyze the product gases, while on the air side both infrared (IR) and mass spectrometry (MS) analyzers were used. The GC and IR analyzers were calibrated before the experiments, while the MS was used to see rapid changes in gas composition in a qualitative way. While IR and MS analyses were recorded close to continuously, the GC sampled every 9 min. During testing parameters such as total ﬂow (12 L/min), rotation speed (14 min1), the ratio between the two exit ﬂows fuel air /Fout = 0.20.5), and the relative angle of the inner feeding (Fout sectors and the two walls in the outer chamber (030°) were varied within the limits noted in parentheses.

Figure 2. Experimental setup showing the infrastructure. Slipped species shown in brackets. RotA; rotameter with needle valve on fuel exit. RotB: rotameter with needle valve on air exit. 9620

dx.doi.org/10.1021/es202244t |Environ. Sci. Technol. 2011, 45, 9619–9626

Environmental Science & Technology

ARTICLE

Figure 3. Experimental details and derived results from experiment RUN 1.

Data Handling. Based on the gas analyses the CH4 conversion was estimated using the following formula

Fout ðCO2 Þ % CH4 conversion ¼ XðCH4 Þ ¼ 100 Fin ðCH4 Þ

ðiÞ

where Fin and Fout denote gas flow in and out of the reactor respectively. Fout includes gas detected both on fuel and air exits. No CO was detected. The capture eﬃciency was estimated based on converted methane using the following formula % CO2 capture efficiency on converted CH4 basis fuel

¼ 100

Fout ðCO2 Þ fuel

air ðCO Þ Fout ðCO2 Þ þ Fout 2

ðiiÞ

The degree of internal gas mixing is also measured in terms of slippage of CH4/CO2 to the air sector and, on the other side, slippage of air to the fuel sector. The %slippage of CO2 is estimated by the following equation air Fout ðCO2 Þ % CO2 slippage ¼ 100 fuel air Fout ðCO2 Þ þ Fout ðCH4 Þ

ðiiiÞ

The slippages of the other components are estimated using similar equations. As deﬁned, the %CO2 slippage will also include any CO2 formed by oxidation of carbon deposits formed in the fuel sector. The CO2 purity was taken directly from the percentage CO2 measured in the exit gas on the fuel side neglecting the argon present.

’ RESULTS AND DISCUSSION In the following the results from a series of experiments carried out to ﬁnd the optimal operating conditions for the prototype version of the rotating bed reactor are described.

Effect of Reactor Temperature, CH4 Flow and Total Flow (RUN 1). The main results from experiment RUN 1 are shown in

Figure 3. Initially the inlet gas flows were as follows: 1313 mL/ min air to the air sector, 167 mL/min argon in each of the inert sectors, and 332 mL/min CH4 + argon in the fuel sector (total feed flow of 1979 mL/min). The angle between the inner feed section and the outer compartment was fixed at 15° throughout the experiment. Introducing the reacting gases at 650 °C (time 0 min) with a low methane flow (18 mL/min), CH4 conversions fluctuating between 35 and 62% are observed. Increasing the temperature to 750 °C and at the same time increasing the CH4 flow to 48 mL/min lead to an increase in CH4 conversion to around 70% and the CO2 capture efficiency based on the converted CH4 is stable above 80%. Further increasing the temperature to 800 °C increases the CH4 conversion further to above 80%. At the end of the experiment (after 100 min), the total flow was reduced to 65% of the original flow keeping the CH4 flow constant at 48 mL/min. This led to a slight increase in CH4 conversion but a sudden drop in CO2 capture efficiency. Knowing that the reaction order of methane with the oxygen carrier typically is around unity,19 the increased conversion can be explained by the increased partial pressure of methane when reducing the inert gas flow. The “CO2 purity” data given in Figure 3 is the amount of CO2 present in the fuel exit gas taking all other components than argon into consideration. It is seen that the CO2 purity is low (around 20%) during most of the experiment but increasing to above 40% when the total feed ﬂow is reduced at the end. The main components beside CO2 in the fuel side eﬄuent are N2 and unconverted CH4. The obvious ways to increase the CO2 purity will be to reduce slippage of air into the fuel exit more eﬃciently and to work at higher CH4 conversions. In the present experiment, increased CO2 purity was obtained by reducing the total ﬂow since less air slips into the fuel exit side. However, as expected, 9621

dx.doi.org/10.1021/es202244t |Environ. Sci. Technol. 2011, 45, 9619–9626

Environmental Science & Technology

ARTICLE

Figure 4. Experimental details and derived results from experiment RUN 2.

the CO2 capture eﬃciency is reduced at the same time, since less argon then ﬂows in the inert sectors giving rise to higher slip of unconverted methane and CO2 into the air exit gas stream. The most challenging problem with a rotating bed reactor regarding the gas mixing inside the reactor becomes very apparent: Since the bed is rotating, there must necessarily be some space between the parts moving relative to each other to avoid too much friction and possible lockage. Within the reactor, only the rotating bed holder and the shafts driving it are rotating, while the inner axis containing the feeding sections and the outer housing are static during operation. Sideway diﬀusion of gaseous species leading to gas slippage is possible along both the inner and outer interfaces of the rotating bed. The major part of the slippage would be expected where the diﬀusion distance is shortest which is along the interface between the rotating bed holder and the inner axis. Any pressure drop over the oxygen carrier bed will also add to this. In the pilot setup, eﬀorts have been made to minimize sideways diﬀusion by keeping the space between the moving parts at a minimum. However, some slippage must be expected. On the positive side, since the slippage caused by sideway diﬀusion along the interfaces is dependent on diﬀusion distances it should be expected to be a minor problem for up-scaled versions of the rotating bed reactor. This will be further discussed in the up-scaling discussion below. In addition to sideways diﬀusion along the interfaces, dispersion of the gas ﬂowing through the oxygen carrier bed will occur adding to the total slippage. In the pilot setup, approximately 1.5 mm spherical CuO/Al2O3 particles have been used. Big particles are used to facilitate the gas ﬂow thus minimizing the pressure drop

over the bed. In addition, intraparticle diﬀusion will play a role with particles having surface area of around 115 m2/g, indicating signiﬁcant amounts of pores left after deposition of the CuO layer. Effect of Reactor Rotation Frequency and Balance between Exit Flows (RUN 2). Experiment RUN 2 was carried out to analyze the reactor performance using different reactor rotation frequencies and also optimizing the ratio between the two exit gas flows within reasonable values. Figure 4 gives the major results from this experiment. The inlet ﬂows were as follows: 1320 mL/min air in the air sector, 167 mL/min argon in each of the inert sectors, 40 mL/min CH4 and 393 mL/min argon to fuel sector (Total ﬂow 2087 mL/min). These ﬂows were kept until the end of the experiment (210 min) where the total ﬂow was reduced by 25% while keeping the CH4 ﬂow constant as indicated in Figure 4. The angle between the inner feed section and the outer compartment was also in this experiment ﬁxed at 15°. The relatively high total ﬂow of around 2 L/min assures fast exchange of gas within the volumes between the reactor and the analytical sampling points, giving rise to fast observable response to the manipulations done. The 15 min used between each change should thus be suﬃcient to observe the eﬀects. Variation of the reactor rotation speed between 1 and 4 rotations/min gave moderate changes in the reactor performance; CH4 conversion varied between 76 and 83%, the lowest conversion was obtained at rotation speed of 1 min1 and highest at around 2 min1. The highest CO2 capture eﬃciencies are obtained at 2 min1 which also seems to give the highest potential CO2 purity. A rotation speed of 2 min1 is therefore considered to be optimal for the setup at these conditions. 9622

dx.doi.org/10.1021/es202244t |Environ. Sci. Technol. 2011, 45, 9619–9626

Environmental Science & Technology

ARTICLE

Figure 5. Experimental details and derived results from experiment RUN 3.

After 120 min the ratio between the two exit gas ﬂows were varied somewhat around the theoretical value being 600 mL/min on the fuel side and 1487 mL/min on the air side: Switching the exit ﬂows to the air side (a 3% increase in the air exit ﬂow and around 7% reduction in the fuel side exit ﬂow) led to a signiﬁcant reduction in CO2 capture eﬃciency but an increase in potential CO2 purity. This is expected since less air will ﬁnd its way to the fuel side exit, but more CO2 will slip to the air side. Reducing the air exit ﬂow from the set point by a similar amount (at 140 min) was followed by an increase in CO2 capture eﬃciency and a reduced potential CO2 purity. At the end of the experiment (after 190 min) the total ﬂow was reduced by 25% while keeping the CH4 ﬂow at 38 mL/min. As in RUN 1, a slight increase in CH4 conversion, decreased CO2 capture eﬃciency and increase potential CO2 purity were observed. Effect of Changes in Inlet Gas Flows (RUN 3). RUN 3 was carried out to find the limit with respect to CO2 capture efficiency and potential CO2 purity. To find this limit, the experiment was run with different CH4 feed flows and different argon flows in the inert sectors. The reactor temperature was kept at 800 °C and the reactor rotation speed at 2 min1 throughout the run. The initial gas flows were as follows: 1320 mL/min air in the air sector, 167 mL/min argon in the two inert sectors and 333 mL/min CH4 and argon in the fuel sector (total flow 1987 mL/min). This is considered 133% of “normal” flow defined as a total flow of 1500 mL/min. The results are given in Figure 5. Initially, a CH4 ﬂow of 38 mL/ min was used. At these conditions the CH4 conversion was around 83% and the CO2 capture eﬃciency based on converted CH4 around 90%. The potential CO2 purity is around 35%.

Reducing the total ﬂow to “normal” (at 60 min) keeping the methane ﬂow constant gave no changes in CH4 conversion, while the CO2 capture eﬃciency based on converted CH4 was slightly reduced. Reducing the CH4 ﬂow to 18 mL/min (at 90 min) keeping the total ﬂow constant led to increased CH4 conversion (to around 89%) and also increased CO2 capture eﬃciency. The potential CO2 purity drops, however, to below 30%. The higher conversion observed is a consequence of the longer contact time between methane and the oxygen carrier; the gas hourly space velocity (GHSV), deﬁned by the volume of reactive gas fed divided on the bed volume, is halved, doubling the contact time. The CH4 ﬂow is then stepwise increased back to 38 mL/min. The CH4 conversion, the CO2 capture eﬃciency, as well as the potential CO2 purity then stepwise returns to their original values (at 140 min). Increasing the argon ﬂow in the inert sectors to 270% of “normal” resulted in a signiﬁcant increase in potential CO2 purity (to above 60%) maintaining the CH4 conversion at about 80% and the CO2 capture eﬃciency at 90% (at 150 min). More argon in the inert sectors will suppress mixing between the air and fuel sectors more eﬃciently. Increasing the CH4 ﬂow rate while keeping the argon ﬂow in the inert sectors high (at 180 min) led to a reduction in the CH4 conversion as a consequence of the reduced contact time. The CO2 separation eﬃciency was slightly reduced. Reducing the argon ﬂow in the inert sectors back to normal (at 210 min) resulted in reduced potential CO2 purity (around 50%) and a slight reduction in CH4 conversion and in CO2 separation eﬃciency. By further increasing the CH4 ﬂow to 100 mL/min (at 255 min) while keeping the “normal” total ﬂow gives some reduction in both conversion and capture capacity. While reducing the total ﬂow to 67% of normal (at 275 min) gives an abrupt 9623

dx.doi.org/10.1021/es202244t |Environ. Sci. Technol. 2011, 45, 9619–9626

Environmental Science & Technology

ARTICLE

Table 1. Parameters Used for Estimating up-Scaling of Rotating Reactora,b prototype lab

atmospheric

pressurized

scale

case

case

r (m) R (m)

0.01 0.02

1.5 5.0

0.3 3.0

h (m)

0.04

7.6

2.4

total bed volume (m3)

6.3 105

410

14

volume of fuel sector (m3)

1.05 105

68

2.3

fuel feed rate (m3/min, STP)

1.0 104

650

650

pressure (atm)

1

1

30

GHSV fuel (h1)

571

571

571

residence time fuel (s) MO mass (ton)

6.3 1.3 104

6.3 820

6.3 28

rotation speed (rotation/min) 0.1

0.1

0.1

energy output (Wth)

400M

400M

55

a

Abbreviations used: r = bed thickness, R = inner radius of rotating bed, h = bed height, MO = oxygen carrier material. b Assuming a bulk density of 2.0 ton/m3.

reduction in both conversion and capture eﬃciency. We believe this drastic change is a result of complete reduction of the inner part of the oxygen carrier bed leading to a signiﬁcant reduction in active bed length. At a certain limiting GHSV, full reduction of the oxygen carrier will occur that gradually moves radially outward. Fully reduced metal may lead to severe coking through CH4 decomposition to coke and H2. No H2 was, however, observed in the present experiment which is expected since any formed H2 would react to water on its way through the outer part of the bed still having reactive oxide available. Any coke formed would be oxidized to CO2 on the air side adding to the slippage of CO2 to the air side. Increasing the total ﬂow back to 133% of normal (at 300 min) regained stable conditions. When reducing the CH4 ﬂow to 38 mL/min keeping 133% total ﬂow (same as initial conditions) the CH4 conversion and CO2 capture eﬃciencies became the same as in the initial part of the experiment. Condensed water collected from the two exit gas ﬂows was 6.1 g (fuel side) and 7.2 g (air side), corresponding to a total of 0.74 mol H2O. Integrating the converted CH4 over the whole experiment gives a total CH4 converted of 0.64 mol. Assuming a temperature of around 10 °C in the water condensation units corresponding to a saturation pressure of H2O of around 1.0 kPa and knowing that the total amount of gas passed through the condensation units over the whole experiment is around 620 L, a total H2O production of around 1.04 mol can be estimated. From the stoichiometry of the reaction, one CH4 molecule gives two H2O molecules; 0.64 mols CH4 should thus give 1.28 mols H2O. We can account for around 81% of the water produced. We believe the discrepancy is a consequence of incomplete condensation and that some of the produced water might be sorbed within the oxygen carrier material. The fact that the water was only slightly selectively condensed from the fuel side eﬄuent indicates that water, at least to some extent, is delayed through the oxygen carrier bed due to sorption/desorption sequences. The three runs presented in the previous sections clearly show that it is possible to achieve CO2 separation with the prototype rotating bed CLC reactor. In the given examples the process parameters are only changed within limits that are reasonable

taking the reaction kinetics into consideration. The only real test touching the limit was carried out in experiment RUN 3 using a relatively high methane concentration where a sudden drop in performance indicated drastic changes within the reactor, changes most probably caused by full reduction of the inner part of the oxygen carrier bed. In general, the modest responses in reactor performance when changing the process parameters indicate that the major part of the performance is determined by the speciﬁc design chosen and its construction. In the given construction the major part of the internal gas mixing will be ﬁxed, and it is possible that varying the process parameters will only make minor perturbations in performance. The interesting fact is that the response in performance clearly shows the expected trends: When methane ﬂow is increased at constant total ﬂow, the methane conversion goes down as expected. When increasing the temperature from 650 to 800 °C the methane conversion is increased as expected. When increasing the inert ﬂow in the steam sectors the CO2 purity increased as expected. Also the dependence of the performance by the rotation frequency shows the expected correlation. So, in total, the reactor works and responds as expected. At the same time, the present implementation of the rotating bed reactor has limitations in its performance that is caused by the design and construction itself and certainly also its size. The major limitations being a consequence of the internal gas mixing observed mixing that is not possible to avoid in the present setup by changing the external process conditions. Estimation of up-Scaled Rotating Bed Reactor. The results obtained in lab scale given in the previous sections are promising, especially considering the fact that the internal gas mixing observed strongly will be determined by sideway diffusion along the interfaces of the moving parts. By properly designing upscaled versions of the reactor, sideway diffusions should be strongly limited since diffusion distances increase drastically, while the radial distances between the moving parts can be kept about the same as in the prototype reactor presented here. We also believe that the rotating bed reactor is well suited for operation at elevated pressures. To get an idea about the approximate size of an up-scaled version we have considered a 400 MWth natural gas (NG) fueled CLC power plant. Considering a typical NG having a heating value of around 37 MJ/m3, a 400 MW reactor will need a feeding rate of about 650 m3 NG/min (at STP). First we will discuss a plant operating at atmospheric pressures (which one may argue is not realistic, but at present all our data is at this pressure); second we will discuss a possible pressurized reactor system. As in the prototype, the gas ﬂows from the center radially outward through the oxygen carrier bed. The oxygen carrier material is ﬁxed in a doughnut-shaped bed with inner radius R, thickness of the bed r, and height h. The outer radius of the doughnut is then R + r. In the lab-scale prototype reactor we have chosen R = 2.0 cm, r = 1.0 cm, and h = 4.0 cm (see Table 1). The typical methane feeding rate is 100 mL CH4/min and 400 mL/ min air or depleted air. We need 100% conversion of methane, while the conversion of oxygen can be less than 100% the extra air acting as cooling agent. The volume of the methane sector of the doughnut bed will be Vbed ¼ 9624

j π h rð2R þ rÞ 360 3 3 dx.doi.org/10.1021/es202244t |Environ. Sci. Technol. 2011, 45, 9619–9626

Environmental Science & Technology where j is the angle of the methane sector of the reactor. The average residence time (τ) is given by the inverse of GHSV. It should here be pointed out that the kinetic parameters used are typical values derived from thermogravimetric experiments,13,16 a method that often underestimates kinetics.17 Faster real kinetics will give the possibility for higher GHSV and thus reduced reactor volume. The Atmospheric Case. For the atmospheric case, we assume the same gas/solid contact time of 6.3 s and a corresponding GHSV of 571 h1 as for the lab-scale prototype. From the needed amount of NG (650 m3/min) the volume of the bed can be estimated to 410 m3! Assuming a bulk density of 2.0 ton/m3 for the oxygen carrier, the bed will have a total mass of 820 ton. Assuming R = 5.0 m and r = 1.5 m, a bed height (h) of 7.6 m will give the needed total bed volume, the outer diameter of the rotating bed being 13 m. The size and mass of the rotating bed is large, in addition comes all housing and infrastructure needed. The Elevated Pressure Case. A similar exercise was carried out for the high pressure case assuming both the reduction and oxidation reactions to be of first order.16 A typical pressure of 30 atm was chosen. The only change from the atmospheric case is that the volume of natural gas at 30 atm is only 22 m3/min. The same gas/solid contact time then only needs a total bed volume of around 14 m3. Fixing R = 3.0 m and r = 0.3 m yields a height (h) of 2.4 m to obtain the needed bed volume. The total rotating bed has then an outer diameter of 6.6 m and a height of 2.4 m. The rotating bed volume scales directly with the pressure when first order reactions are assumed. However, most kinetic studies indicate that the reaction orders both with respect to hydrocarbon pressure upon metal oxide reduction and in oxygen pressure when reoxidizing are lower than unity,18,19 typically 0.60.8, leading to a pressure induced increase in reaction rate of 815 times relative to the atmospheric case which will increase the bed volume with a factor of 23. However, knowing that the kinetic parameters used are underestimated we leave this as an open question until proper tests at elevated pressures have been carried out for a second reactor version. From the experiments with the prototype version of the rotating reactor we have shown that it is possible to obtain around 90% CH4 conversion and >90% CO2 capture eﬃciency based on converted methane. It is natural to believe that the internal gas mixing observed is mainly caused by two factors: 1) by gas diﬀusion along the interfaces between the parts moving relative to each other inside the reactor and 2) by gas dispersion in the oxygen carrier bed. The former source for mixing should be strongly reduced when up-scaling the reactor since diﬀusion pathways then is strongly reduced. The second source for mixing will be strongly dependent on the shape of the oxygen carrier material used. Although the reactor has only been operated in periods of up to 6 h, we see that operation is possible over many hours, and also that stable operation can be regained after running into unstable conditions. Stable operation is obtained at relatively high gas velocities where fully reduced oxygen carrier in (the inner) part of the bed is avoided. Potential CO2 purities obtained are in the range 20 to 65% the lack of purity mostly being due to air slippage from the air sector. This seems to be the major drawback of the prototype reactor setup. We believe that signiﬁcant improvements can be made to further avoid gas mixing in future modiﬁed and upscaled reactor versions, mainly because diﬀusion distances along

ARTICLE

the interfaces then will be signiﬁcant longer, while the distance between the moving parts still can be kept at a minimum. It is also important to notice that the performance of the rotating bed reactor is strongly dependent on the choice of geometry and design of the individual parts. With the chosen design some adjustments of the performance can be done by varying the process conditions, but the optimal performance is limited by nonadjustable factors. The knowledge gained from the work with this ﬁrst prototype reactor should be used to design and construct an improved second version reactor with the main focus on high pressure operation at temperatures above 1000 °C, as well as on oxygen carrier stability over prolonged time-on-stream.

’ ASSOCIATED CONTENT

bS

Supporting Information. Figures. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: +47 90622647. Fax: +47 22067350. E-mail: richard.blom@ sintef.no.

’ ACKNOWLEDGMENT This publication forms a part of the BIGCO2 project, performed under the strategic Norwegian research program Climit. The authors acknowledge the partners: Statoil, GE Global Research, Statkraft, Aker Clean Carbon, Shell, TOTAL, ConocoPhillips, ALSTOM, the Research Council of Norway (178004/I30 and 176059/I30), and Gassnova (182070) for their support. ’ REFERENCES (1) Lewis, W. K.; Gilliland, E. R. Production of pure carbon dioxide. S.O.D. Company, US Patent: 2,665,971, 1954. (2) Ritcher, H.; Knoche, K. Reversibility of combustion process. ACS Symp. Ser. 1983, 235, 71–85. (3) Ishida, M.; Jin, H. A novel combustor based on chemical-looping combustion reactions and its reactions kinetics. J. Chem. Eng. Jpn. 1994, 27, 296–301. (4) Kvamsdal, H. M.; Jordal, K.; Bolland, O. A quantitative comparison of gas turbine cycles with CO2 capture. Energy 2007, 32, 10–24. (5) Erlach, B.; Schmidt, M.; Tsatsaronis, G. Comparison of carbon capture IGCC with pre-combustion decarbonisation and with chemicallooping combustion. Energy 2011, 36, 3804–3815. (6) Ekstrom, C.; Schwendig, F.; Biede, O.; Franco, F.; Haupt, G.; de Koeijer, G.; Papapavlou, C.; Røkke, P. E. Techno-Economic Evaluations and Benchmarking of Pre-combustion CO2 Capture and Oxy-fuel Processes Developed in the European ENCAP Project. Energy Proc. 2009, 1, 4233–4240. (7) Berguerand, N.; Lyngfelt, A. Batch testing of solid fuels with ilmenite in a 10 kWth chemical-looping combustor. Fuel 2010, 89, 1749–1762 and references therein. (8) Pr€ oll, T.; Kolbitsch, P.; Bolhar-Nordenkampf, J.; Hofbauer, H. A novel dual circulating ﬂuidized bed system for chemical looping combustion. Environ. Energ. Eng. 2009, 55, 3255–3266. (9) Jin, H.; Ishida, M. Reactivity study on natural-gas-fueled chemical-looping combustion by a ﬁxed-bed reactor. Ind. Eng. Chem. Res. 2002, 41, 4004–4007. (10) Thayer, C. H. Method and apparatus for conversion of organic reactants to other organic products, US Patent: 2,704,741, 1955. (11) Shimomura, Y. The CO2 wheel: a revolutionary approach to carbon dioxide capture. Modern Power Systems 2003, January, 15–17. 9625

dx.doi.org/10.1021/es202244t |Environ. Sci. Technol. 2011, 45, 9619–9626

Environmental Science & Technology

ARTICLE

(12) Cheng, H.-H.; Tan, C.-S. Carbon dioxide capture by blended alkanolamines in rotating packed bed. Energy Proc. 2009, 1, 925–932. (13) Dahl, I. M.; Bakken, E.; Larring, Y.; Spjelkavik, A. I.; Hakonsen, S. F.; Blom, R. On the development of novel reactor concepts for chemical looping combustion. Energy Proc. 2009, 1, 1513–1519. (14) Abad, A.; Adanez., J.; García-Labiano, F.; de Diego, L. F.; Gayan, P.; Celaya, J. Mapping of the range of operational conditions for Cu-, Fe-, and Ni-based oxygen carriers in chemical-looping combustion. Chem. Eng. Sci. 2007, 62, 533–549. (15) Naqvia, R.; Wolf, J.; Bolland, O. Part-load analysis of a chemical looping combustion (CLC) combined cycle with CO2 capture. Energy 2007, 32, 360–370. (16) Readman, J. E.; Olafsen, A.; Smith, J. B.; Blom, R. Chemical looping combustion using NiO/NiAl2O4: Mechanisms and kinetics of redox reactions from in-situ powder X-ray diﬀraction and TG experiments. Energy Fuel 2006, 20 (4), 1382–1387. (17) Mattisson, T.; Jerndal, E.; Linderholm, C.; Lyngfelt, A. Reactivity of a spray-dried NiO/NiAl2O4 oxygen carrier for chemical-looping combustion. Chem. Eng. Sci. 2011, 66, 4636–4644. (18) Moghtaderi, B.; Song, H. Reduction properties of physically mixed metallic oxide oxygen carriers in chemical looping combustion. Energy Fuels 2010, 24, 5359–5368. (19) Abad, A.; Garcia-Labiano, F.; de Diego, L. F.; Gayan, P.; Adanez, J. Reduction Kinetics of Cu-, Ni-, and Fe-Based Oxygen Carriers Using Syngas (CO + H2) for Chemical-Looping Combustion. Energy Fuel 2007, 21, 1843–1853.

9626

dx.doi.org/10.1021/es202244t |Environ. Sci. Technol. 2011, 45, 9619–9626

ARTICLE pubs.acs.org/est

Escherichia coli Inactivation by UVC-Irradiated C60: Kinetics and Mechanisms Min Cho,†,‡ Samuel D. Snow,† Joseph B. Hughes,† and Jae-Hong Kim*,† †

School of Civil and Environmental Engineering, Georgia Institute of Technology, 200 Bobby Dodd Way, Atlanta, Georgia 30332, United States ‡ Division of Biotechnology, Advanced Institute of Environmental and Bioscience, College of Environmental and Bioresource Sciences, Chonbuk National University, Iksan, Jeonbuk 570-752, South Korea ABSTRACT: Motivated by recent studies that documented changes in fullerene toxicity after chemical transformation, C60 aggregates (nC60) were subject to UVC irradiation at monochromatic 254 nm and subsequently evaluated for antibacterial and bactericidal properties against Escherichia coli. The nC60 treated with UVC irradiation, referred to herein as UVC-irradiated C60, did not directly inhibit bacterial growth at concentrations up to 20 mg/L. In the presence of UVA and visible light, however, UVC-irradiated C60 rapidly inactivated E. coli, suggesting that photochemical production of reactive oxygen species (ROS) was involved. The use of ROS scavengers and probes determined that hydroxyl radicals were the primary ROS responsible for the E. coli inactivation. Results from protein release, lipid peroxidation, cell permeability, and intracellular enzyme assays suggest that the inactivation mechanism involves UVC-irradiated C60 diﬀusing through E. coli cell membrane and producing hydroxyl radicals within the cell. Further study on water-soluble C60 derivatives and possible transformative processes is, therefore, recommended based on the environmental implications of results presented herein that nC60 exposed to UVC irradiation is more toxic than parent nC60.

’ INTRODUCTION The toxicity, or lack thereof, of nC60 (aggregate of pristine C60 in the aqueous phase) has been a topic of many recent studies. Early studies reported toxic eﬀects in various organisms, including mammalian cells, aquatic organisms, and bacteria.1 3 Contrary to these initial ﬁndings, some researchers reported much lower toxicity levels for nC60.4,5 Later, it was suggested that the method of nC60 formation could explain some of the initial toxicity reported; when tetrahydrofuran (THF) was used as an intermediate solvent, trace amounts of THF and its byproducts could have remained and exerted toxic eﬀects.6 8 In the absence of solvent eﬀect, researchers found nC60 to be only mildly toxic to Escherichia coli.8 This result was further supported by additional studies which showed that nC60’s crystalline structure caused self-quenching of photoexcitation,9 11 greatly inhibiting its ability to photochemically produce reactive oxygen species (ROS), which have been suggested as one of the C60’s primary toxicity mechanisms.3,12 Although nC60 themselves may not be of major environmental concern, few studies have examined the potential impacts of derivatized C60 in the aqueous environment. We have recently reported that nC60 is readily oxidized by ozone, forming watersoluble products containing as many as 29 oxygen atoms per C60,13 similar to fullerenol, a polyhydroxylated C60. Ozonated C60 was found to have low intrinsic antimicrobial properties yet a strong and unique photochemical toxicity pathway.14 Similarly, we also reported that nC60 could be readily oxidized by UVC irradiation, which is commonly used in water and wastewater r 2011 American Chemical Society

disinfection, resulting in the formation of highly oxidized, soluble C60 products (referred to herein as UVC-irradiated C60).15 A series of papers by Hou et al.16 18 also reported transformation of nC60 into water-soluble, multiple oxygenated products via sunlight. Yet, toxicological eﬀects of these transformation products are unknown. Functionalization of C60 is known to aﬀect its photochemical properties in a variety of ways,19,20 with a general decrease in photochemical reactivity with consecutive additions of functional groups13 caused by the disruption of C60’s pi-conjugated system.21 Fullerene derivatives, however, typically retain their ability to produce some ROS.16,22,23 Solubilizing C60 in the aqueous phase via chemical transformations with ozonation and UVC irradiation could in fact result in increased photochemical reactivity compared to nC60, since self-quenching is prevented. Additionally, functionalization may allow facile diﬀusion of individual C60 molecules (i.e., molecularly dispersed in the aqueous phase) through cell membranes, resulting in direct contact with intracellular materials. The objective of this study is to examine how UVC-irradiated C60 interacts with E. coli, a representative gram-negative bacterium, in order to gain a better understanding of environmental consequences of the phototransformation of nC60. Results suggest that UVC-irradiated C60 Received: July 1, 2011 Accepted: October 15, 2011 Revised: October 11, 2011 Published: October 15, 2011 9627

dx.doi.org/10.1021/es202269r | Environ. Sci. Technol. 2011, 45, 9627–9633

Environmental Science & Technology eﬀectively inactivates E. coli under UVA and visible light irradiation conditions through mechanisms that potentially involve transport of derivatized C60 through the cell membrane and subsequent photocatalytic production of 3 OH within the cell.

’ EXPERIMENTAL SECTION Materials. Water (>18 MΩ) produced from a Milli-Q Water Purification System (Millipore Co.) was used in preparation of all solutions and reagents. All chemicals used were analytical, reagent-grade. All glassware used in the experiments was washed with DI water, then autoclaved at 121 °C for 15 min. UVC-Irradiated C60 Preparation. An organic solvent-free preparation technique was used to avoid potential solvent interferences between prepared nC60 and microorganisms.8 Ultrasound (20 W) was continuously applied to 500 mL of ultrapure water containing 100 mg of C60 (99.9%, Materials Electronics Research Corporation, Tucson, AZ) in a sealed Pyrex bottle for 24 h at ambient temperature using an ultrasonicator (S-4000, Misonix Co.). The color of the solution became light orange, indicating that C60 had formed water-stable clusters. The resultant solution was filtered through a 0.45-μm cellulose filter (Millipore Co.) and stored in the dark. The filtrate containing 2 3 mg/L of nC60 was further concentrated to a stock solution of 30 mg/L using a rotary evaporator. The concentration of nC60 was measured using a UV/vis spectrophotometer at 350 nm (Agilent 8453, Agilent Co.)13 and a Shimadzu TOC-VWS analyzer. The nC60 was diluted in ultrapure water and subject to UVC-mediated transformation for 7 days in a magnetically stirred quartz reactor under water cooling using four UVC lamps emitting monochromatic radiation at 254 nm (Philips Co.). The incident light intensity of each lamp at the location of the reactor was measured at 11 mW/cm2 using a UVX Radiometer with 254nm sensor (UVP Co.). Preparation and Analysis of E. coli. E. coli was selected as a representative gram negative bacterium and was obtained from the American Type Culture Collection (ATCC). The bacteria were cultured and counted according to the method described by Cho et al.24 Briefly, E. coli (ATCC 8739) was inoculated in 50 mL of Tryptic Soy Broth in a 200-mL flask and grown for 18 h at 37 °C in a shaking incubator. The bacteria were harvested by centrifugation in a 50-mL conical tube at 1000g for 10 min and washed two times with 50 mL of phosphate buffered saline solution at pH 7.2 (PBS). An E. coli stock solution was prepared by resuspending the final pellets in 50 mL of PBS. The initial population of E. coli was approximately 105 cfu/mL and obtained by diluting the stock solution. The cell concentration was determined by a spreading plate method on nutrient agar, incubating at 37 °C for 24 h, with three replicate plates at each dilution. Inactivation Experiments. A suspension (20 mL) containing UVC-irradiated C60 and E. coli was placed in a 40-mL quartz reactor and vigorously stirred using a magnetic stirrer at 1100 rpm. Three blacklight blue (BLB) lamps (300 420 nm; Philips Co.) or six fluorescent lamps (FL, visible range with a small amount of UVA irradiation; Philips Co.) were placed 3 cm from the reactor surface. The emission spectra of these lamps were verified with an Acton Research Detection system (Spectrapro-500, USA). The incident light intensity by three BLB lamps was verified by ferrioxalate actinometry25 as 1.2 10 6 Einstein/L-s. The light intensity of six FLs at a representative wavelength of 365 nm on reactor surface was measured as 330 μW/cm2 using a UVX radiometer with a long-wave sensor (UVP Co.) at bandwidth of

ARTICLE

10 nm. The role of visible light was investigated after placing UV cutoff filter (Optivex Co.). Temperature and pH were adjusted to 19 ( 1 °C (air-cooling) and 7.1 (10 mM of phosphate buffer), respectively. Antibacterial Activity Assay. Antibacterial activity was assayed by determining the minimum inhibitory concentration (MIC).1 For assaying the MIC, modified minimal Davis (MD) media tubes containing 0 20 mg/L of UVC-irradiated C60 were inoculated overnight with a culture of E. coli (DH5α) and grown on Luria Bertani nutrient broth. Initial OD600 was controlled to 0.002 and the lack of growth of E. coli after overnight incubation indicated the MIC. Detection of Produced ROS. Furfuryl alcohol (FFA, 0.85 mM; Aldrich Co.) and p-chlorobenzoic acid (pCBA, 2 μM; Aldrich Co.) were used as indicators for detecting 1O2 and 3 OH, respectively.9 The concentrations were analyzed by means of HPLC (Agilent 1100; Agilent Co.). A C18 reverse-phase column (Agilent Zorbax RX-C18; Agilent Co.) was used with a diode-array UV detector at a wavelength of 230 and 237 nm for measuring the concentrations of FFA and pCBA, respectively. The production of O2 3 was measured by using XTT sodium salt (2,3-bis(2-methoxy-4-nitro5-sulfophehyl)-2H-tetrazolium-5-carboxanilide inner salt, 0.15 mM; Sigma Co.) as a probe which forms a purple product (λmax = 480 nm) from the reaction with O2 3 .9 Microbial Inactivation Mechanisms. The possible mechanisms of microbial inactivation were comparatively investigated by quantifying the degree of lipid peroxidation, the amount of protein oxidation or disruption, the change in cell wall permeability, and the degradation of intracellular enzymes during cell death. The time scale for all these experiments was adjusted in order to achieve 90% (1 log) inactivation of E. coli. All the experiments were conducted in triplicate and all the assay results were within a 90% confidence interval. Note that samples from photochemical experiments were collected and instantly transferred to amber bottles after centrifugation for all these analyses. Lipid peroxidation was analyzed by observing the quantity of malondialdehyde (MDA), a product of the lipid peroxidation, based on a reaction with thiobarbituric acid (TBA) which forms a pink MDA TBA product, following a method by Maness et al.26 The amount of proteins released from E. coli surface components was determined using a modified Bradford assay,27 while the amount of oxidized proteins (i.e., protein carbonyls) was measured using an OxyELISA oxidized protein quantification kit (Millipore Co.).28 To verify the possible change of cell permeability, the reaction between an intracellular enzyme, β-D-galactosidase, and its substrate, o-nitrophenyl-β-D-galactopyranoside (ONPG),29 was monitored following the method described by Zheng et al.29 The change of enzymatic activity after photocatalytic inactivation was investigated by means of the API-ZYM system (BioMerieux Co.).30 Additional experiments were also performed using a spheroplast of E. coli which was prepared by incubating fresh E. coli in a Tris-HCl buﬀered suspension containing sucrose, lysozyme, and ethylenediaminetetraacetic acid (EDTA) as described by Zheng et al.29

’ RESULTS AND DISCUSSION Inactivation Kinetics. Results from MIC tests suggested that addition of as high as 20 mg/L of UVC-irradiated C60 did not inhibit E. coli growth. These findings are consistent with previous results for ozonated C60 (no effect on the growth of E. coli up to 10 mg/L 14) as well as nC60 with varying degree of UVC 9628

dx.doi.org/10.1021/es202269r |Environ. Sci. Technol. 2011, 45, 9627–9633

Environmental Science & Technology

ARTICLE

Figure 1. Kinetics of E. coli inactivation by UVC-irradiated C60 ([UVCirradiated C60]0 = 15 mg/L; [ozonated C60]0 = 10 mg/L14).

irradiation (increased MIC with increasing UVC irradiation time from 0 to 110 h with the same UVC intensity15). This contrasts with the previous findings that nC60 exhibits a MIC of approximately 2.0 mg/L.1,3 The difference resulted presumably due to different methods employed for nC60 synthesis (i.e., nC60 samples used in this study were thoroughly washed to remove THF and THF derivatives8). Collectively, this oxidative transformation of nC60, either by UVC or ozone, results in decreased intrinsic (i.e., in the absence of light) bacteriostatic activity of nC60. Figure 1 shows the eﬀect of light sources on the inactivation of E. coli by the photocatalytic activity of UVC-irradiated C60. In control tests, E. coli was not inactivated by 15 mg/L nC60 under BLB or FL light irradiation within experimental time scale (results not shown). Consistent with the results of the MIC tests, E. coli was not inactivated when as much as 15 mg/L of UVC-irradiated C60 was applied for 150 min in the dark (results not shown). A minor level (0.1 log) of inactivation was observed under BLB light in the absence of C60, which may be attributed to action of UVA and UVB light. Note that 15 mg/L (and other concentrations in this study) refers to the initial mass concentration of nC60 before UVC irradiation, which would change due to addition of functional groups during UVC treatment. No E. coli inactivation was observed under ﬂuorescent light alone within the experimental time scale. In contrast, a signiﬁcant level of E. coli inactivation (2.4 log inactivation for 120 min) was observed when BLB light was irradiated. Figure 1 also shows a comparison with a previous study that reported the kinetics of E. coli inactivation by ozonated C60 (10 mg/L) which achieved only a 1.3 log inactivation for 150 min under the same light condition.14 Also note that nC60 under the same conditions would result in no measurable E. coli inactivation.14 Under FL light irradiation, 15 mg/L of UVC-irradiated C60 resulted in approximately 2 log inactivation (99%) in 135 min. When a UV cutoﬀ ﬁlter was used to remove the UVA below 400 nm, typically comprising about 4% of the total light energy emitted from the ﬂuorescence lamps, the inactivation kinetics

Figure 2. Eﬀect of (a) UVC exposure time for the preparation of UVCirradiated C60, (b) UVC-irradiated C60 concentration, and (c) FL light intensity on the kinetics of E. coli inactivation. Experimental condition: (a) UVC irradiation time = 3, 5, 7 days; FL irradiation intensity at 365 nm = 330 μW/cm2; [UVC-irradiated C60]0 = 15 mg/L; (b) UVC irradiation time = 7 days; FL irradiation (six lamps) intensity at 365 nm = 330 μW/cm2; [UVC-irradiated C60]0 = 5, 10, 15 mg/L; (c) UVC irradiation time = 7 days; FL irradiation (six lamps) intensity at 365 nm = 110, 220, 330 μW/cm2; [UVC-irradiated C60]0 = 15 mg/L.

was slightly decreased. The observed inactivation levels for 150 min FL exposure with and without UV cutoﬀ ﬁlters were 1.95 and 2.4 log, respectively. It is noteworthy that UVCirradiated C60 inactivates bacteria in response to visible light 9629

dx.doi.org/10.1021/es202269r |Environ. Sci. Technol. 2011, 45, 9627–9633

Environmental Science & Technology

Figure 3. Eﬀect of N2 bubbling and addition of ROS scavenger on the kinetics of E. coli inactivation. Inset: degradation of pCBA as a function of time. ([UVC-irradiated C60]0 = 15 mg/L, light intensity at 365 nm by six FLs = 330 μW/cm2).

and the observed inactivation primarily resulted from the visible light irradiation. Most experiments and discussions presented below, therefore, focus on visible light irradiation condition which is more environmentally relevant. At a constant concentration of UVC-irradiated C60 (15 mg/L), the E. coli inactivation rate increased as nC60 was treated for longer UVC exposure time (from 3 to 7 days) (Figure 2a). We have reported that as UVC exposure time increased, nC60 disaggregated into more hydrophilic products with concurrent decrease in aggregate size.15 The increased hydrophilicity and available surface area may have contributed to the enhanced E. coli inactivation rate. As the UVC-irradiated C60 concentration was increased from 0 to 15 mg/L, the inactivation at 150 min of exposure increased from 0 to 2.4 log (Figure 2b), exhibiting a linear relationship between the concentration and log inactivation achieved (R2 = 0.99 when log inactivation was plotted versus concentration; inset in Figure 2b). The E. coli inactivation rate also increased as the light intensity of the FLs was increased from 110 (2 lamps) to 330 μW/cm2 (6 lamps) (Figure 2c). A squareroot dependence was found for the observed E. coli inactivation for 150 min and the light intensity (R2 = 0.98 when log inactivation was plotted versus square root of intensity). This observation is in agreement with results of previous studies on photocatalytic microbial inactivation.31,32 The E. coli inactivation observed above was caused by a photocatalytic production of ROS by UVC-irradiated C60. When N2 gas was bubbled in order to remove the residual oxygen, only a slight level of E. coli inactivation was observed (Figure 3; 0.15 log for 150 min). The low level of inactivation observed in the N2 purged system might have been caused by residual oxygen which was not completely removed during N2 bubbling. Among ROS, 1 O2 and O2 3 were not responsible for E. coli inactivation. First, when excess (30 mM) 1O2 and O2 3 scavengers (L-histidine and superoxide dismutase (SOD)) were added, there was little change in the inactivation kinetics (Figure 3), suggesting that both were not a primary disinfection agent. Second, no measurable

ARTICLE

degradation of FFA, a probe for 1O2, was observed over the experimental time scale, while degradation of XTT (0.15 mM), a probe for O2 3 , as indicated by increase of absorption at 480 nm due to formation of purple-colored product, was observed. It should be noted that, although O2 3 is known to have a potential for inactivating E. coli,24,33 1O2 would not inactivate E. coli even if a detectable amount had been produced. Control tests were performed with Rose Bengal (10 μM) under FL irradiation for 60 min to independently produce 1O2 (e.g., Φ(1O2) of RB (550 nm) = 0.75 in water34). The results of control test suggested that as much as 6.4 10 5 mg-min/L28 of 1O2 exposure did not induce E. coli inactivation (data not shown), although the same level of exposure under identical condition would lead to 99% of MS2 bacteriophage inactivation.28 Detection of O2 3 indicates that O2 3 might have been produced as a precursor for 3 OH production. The fact that E. coli inactivation was not inhibited by SOD, therefore, may suggest that either 3 OH production is kinetically favored over quenching by SOD, or SOD did not eﬀectively quench O2 3 that was produced inside the cell. Regardless, the observed E. coli inactivation was directly caused by photochemically produced 3 OH, since E. coli inactivation was virtually prohibited in the presence of excess (30 mM) 3 OH scavenger, t-butanol (t-BuOH). Results obtained with pCBA as an 3 OH probe is shown in the inset of Figure 3 and suggests steady-state concentration of 3 OH is approximately 5.62 10 15 M.38 The Mechanism of Inactivation. Mechanism of E. coli inactivation by 3 OH has been delineated as a sequence of nonselective reactions of 3 OH with major cell wall components such as proteins and lipids35 before it reaches inner cell components, followed by increase in cell wall permeability and subsequent 3 OH penetration into cytoplasm, and ultimately cell death.36,37 Surface protein release during cell death, therefore, is indirect evidence for cell surface destruction by 3 OH. For example, a control experiment was performed to inactivate E. coli to achieve 0.5 log inactivation by 3 OH produced via a different pathway (i.e., mixing 0.1 mg/L of O3 and 1 mg/L of H2O2). It was found that approximately 0.03 ( 0.0015 mg/L of protein, measured by Bradford assay, were released from cell surface components of 2 107 cfu/mL of E. coli. Similar phenomena were observed when E. coli was inactivated using even weaker disinfectants (oxidants) such as ozone and chlorine dioxide38 compared to 3 OH. For the same level of inactivation, approximately 0.02 and 0.01 mg/L of protein was released from the same concentration of E. coli. Note that the initial lag phase in the kinetics (Figures 1 and 2) is commonly attributed to the time required for the chemical disinfectant to disrupt the cell surface components before penetrating into the cytoplasm and reacting with vital cell components.39 However, the following lines of evidence collectively suggest that the mechanism of E. coli inactivation by 3 OH, photochemically produced from UVC-irradiated C60, may be diﬀerent from the aforementioned pathway which involves cell surface disruption as a critical, initial step. First, E. coli inactivated up to 1 log showed no measurable protein release (i.e., below the detection limit of 0.001 mg/L) under the same condition in which protein release was observed by 3 OH produced in diﬀerent methods. Second, the concentration of protein carbonyls, a commonly used marker of protein oxidation,40 also did not increase as E. coli was inactivated up to 1 log. Third, MDA release was also not observed (below the detection limit of 0.01 nmol/(mg cell dry wt)) during E. coli inactivation up to 1 log. MDA is one of the 9630

dx.doi.org/10.1021/es202269r |Environ. Sci. Technol. 2011, 45, 9627–9633

Environmental Science & Technology

Figure 4. Degradation of intracellular β-D-galactosidase as a function of reaction time ([UVC- irradiated C60]0 = 15 mg/L, light intensity at 365 nm by six FLs = 330 μW/cm2).

most abundant aldehydes formed from the peroxidation of the lipid membranes,41 primarily consisting of polyunsaturated phospholipids,26 and MDA release provides direct evidence of cell membrane denaturation. Fourth, change in cell permeability, which would result from damage or alteration of cell surface components, was examined using ONPG as a probe.29,36 The ONPG hydrolysis, as indicated by the chromogenic product formation (420 nm), results only when ONPG penetrates into the cell and reacts with intracellular E. coli enzymes such as β-Dgalactosidase.29,42 No measurable ONPG hydrolysis was observed for E. coli inactivated up to 2 log by UVC-irradiated C60, similarly to intact E. coli. In contrast, ONPG hydrolysis was as high as approximately 1500 μmol/(min-mg cell dry wt) at 0.6 log inactivation using ozone, equivalent to 69% of that obtained with lysed cells.43 While no evidence points toward cell surface damage as a preceding step for cell death, E. coli inactivation was accompanied by degradation of intercellular enzyme, β-D-galactosidase (approximately 80% over 2 log inactivation) (Figure 4). The reaction between 3 OH and β-D-galactosidase was nearly instantaneous in an independent in vitro control experiment. When t-BuOH was added to the suspension, β-D-galactosidase degradation was prohibited, suggesting that enzyme was degraded by 3 OH. Similarly, other intracellular enzymes were found to be degraded during E. coli inactivation by UVC-irradiated C60. The API-ZYM enzyme assay performed with adding E. coli after 1 log inactivation by UVC-irradiated C60 also suggested that ﬁve intercellular enzymes, alkaline phosphatase, leucine arylamidase, acid phosphatase, β-galactosidase, and β-glucuronidase, showed decreased enzymatic activity. The above results collectively suggest that 3 OH was the primary agent for E. coli inactivation but the mechanism does not involve cell surface damage that would be expected when 3 OH were produced in the bulk phase. Consequently, penetration of UVC-irradiated C60 into E. coli and generation of 3 OH within the cytoplasm, causing internal damage and consequently cell death, is suspected. In such a case, the initial lag in the

ARTICLE

Figure 5. Inactivation kinetics of intact E. coli after initially exposing to UV irradiated C60 for 60 min and spheroplast E. coli (19 ( 1 °C, pH 7.1, UVC exposure: 7 days, [UV irradiated C60]0 = 15 mg/L, light intensity at 365 nm by FL (six lamps): 330 μW/cm2).

kinetics discussed above could be related to the time required for UVC-irradiated C60 to penetrate into the cytoplasm, not the time required for 3 OH to cause suﬃcient cell surface damage. Consistently, when E. coli was pre-exposed to UVC-irradiated C60 under dark for 60 min (during which no inactivation was observed) and then exposed to FL irradiation, the lag phase was signiﬁcantly reduced from approximately 40 to 15 min with little change in the postshoulder kinetics (Figure 5). Another experimental result obtained with E. coli spheroplast (i.e., E. coli with most outer membrane and peptidoglycan layer removed)44 also supports this hypothesis. The reduction in the lag phase in E. coli spheroplast was approximately the same as E. coli that was pre-exposed to UVC-irradiated C60. This mechanism partly explains why E. coli was inactivated by a much lower exposure to 3 OH compared to the reported 3 OH CT value. The CT stands for the product of disinfectant concentration and contact time and is widely used as a parameter to gauge the eﬃciency of disinfectant. The steady-state 3 OH concentration measured using pCBA was approximately 5.62 10 15 M45 (inset of Figure 3), which translates into an 3 OH CT (the product of radical concentration and contact time) to induce 1 log inactivation of E. coli of 0.86 10 8 mg/L-min.46 This value is 698 times less than the CT values obtained from 3 OH CT of 6.0 10 6 mg/ L-min measured via diﬀerent photocatalyst (TiO2).45 Alternatively, 0.86 10 8 mg/L-min of CT would not be suﬃcient to disrupt the surface structure of E. coli if produced in the bulk phase. Environmental Significance. This study suggests that nC60 in the aqueous phase, upon hydrophilic transformation by UVC irradiation, exhibits a markedly different biocidal activity compared to nC60. The results are similar to C60 transformed by ozone treatment. Ozonation of nC60 resulted in the formation of soluble, multiple-oxygenated products which readily inactivated E. coli in the presence of UVA light (by BLB lamps) and oxygen.14 Similar to the mechanism presented herein, penetration of ozonated C60 into E. coli cytoplasm and photochemical production of 3 OH was suspected as the primary reason for the 9631

dx.doi.org/10.1021/es202269r |Environ. Sci. Technol. 2011, 45, 9627–9633

Environmental Science & Technology cell death. The inactivation kinetics had similar characteristics (i.e., presence of lag phase), although a direct quantitative comparison is not possible, and the values of 3 OH CT were also very similar: 0.86 10 8 mg/L-min for UVC-irradiated C60 and 1.1 10 8 mg/L-min for ozonated C60. This study highlights that understanding the fate of nC60 with potential transformation by UVC as well as ozone, either in nature or in engineered systems such as water and wastewater treatment, and the ecological impact of the transformation products is critical to assess the overall environmental impact of C60. Whereas this study focused on the interaction of UVC transformation product with a representative microorganism, further studies are required to assess its interaction with other microorganisms and higher organisms as well as transport behaviors in the environment. Other oxidative transformation routes, particularly in engineered systems, are possible and have potential to transform C60 into potentially more toxic forms. Many C60 derivatives are intentionally being manufactured for particular applications and the concerns and potential impacts of these materials have yet to be assessed. Finally, we acknowledge that more direct evidence is required to firmly prove the hypothesis of E. coli inactivation by penetration of UVCirradiated C60, which is the focus of our currently ongoing study.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]; phone: 404-894-2216; fax: 404- 385-7087.

’ ACKNOWLEDGMENT This work was partly supported by the National Science Foundation (Award CBET-0932872) and Korean National Research Foundation (Korean Ministry of Education, Science and Technology, Award NRF-2011-35B-D00020). ’ REFERENCES (1) Lyon, D. Y.; Fortner, J. D.; Sayes, C. M.; Colvin, V. L.; Hughes, J. B. Bacterial cell association and antimicrobial activity of a C60 water suspension. Environ. Toxicol. Chem. 2005, 24 (11), 2757–2762. (2) Lovern, S. B.; Klaper, R. Daphnia magna mortality when exposed to titanium dioxide and fullerene (C60) nanoparticles. Environ. Toxicol. Chem. 2006, 25 (4), 1132–1137. (3) Sayes, C. M.; Fortner, J. D.; Guo, W.; Lyon, D.; Boyd, A. M.; Ausman, K. D.; Tao, Y. J.; Sitharaman, B.; Wilson, L. J.; Hughes, J. B.; West, J. L.; Colvin, V. L. The diﬀerential cytotoxicity of water-soluble fullerenes. Nano Lett. 2004, 4 (10), 1881–1887. (4) Oberdorster, E.; Zhu, S. Q.; Blickley, T. M.; McClellan-Green, P.; Haasch, M. L. Ecotoxicology of carbon-based engineered nanoparticles: Eﬀects of fullerene (C60) on aquatic organisms. Carbon 2006 44 (6), 1112–1120. (5) Lyon, D. Y.; Alvarez, P. J. J. Fullerene water suspension (nC60) exerts antibacterial eﬀects via ROS independent protein oxidation. Environ. Sci. Technol. 2008, 42 (21), 8127–8132. (6) Andrievsky, G.; Klochkov, V.; Derevyanchenko, L. Is the C60 fullerene molecule toxic?! Fullerenes, Nanotubes, Carbon Nanostruct. 2005, 13 (4), 363–376. (7) Henry, T. B.; Menn, F. M.; Fleming, J. T.; Wilgus, J.; Compton, R. N.; Sayler, G. S. Attributing eﬀects of aqueous C60 nano-aggregates to tetrahydrofuran decomposition products in larval zebraﬁsh by assessment of gene expression. Environ. Health Perspect. 2007, 115 (7), 1059–1065. (8) Zhang, B.; Cho, M.; Fortner, J. D.; Lee, J.; Huang, C. H.; Hughes, J. B.; Kim, J. H. Delineating oxidative processes of aqueous C60

ARTICLE

preparations: Role of THF peroxide. Environ. Sci. Technol. 2009, 43 (1), 108–113. (9) Lee, J.; Fortner, J. D.; Hughes, J. B.; Kim, J. H. Photochemical production of reactive oxygen species by C60 in the aqueous phase during UV irradiation. Environ. Sci. Technol. 2007, 41 (7), 2529–2535. (10) Lee, J.; Kim, J. H. Eﬀect of encapsulating agents on dispersion status and photochemical reactivity of C60 in the aqueous phase. Environ. Sci. Technol. 2008, 42 (5), 1552–1557. (11) Lee, J.; Yamakoshi, Y.; Hughes, J. B.; Kim, J. H. Mechanism of C60 photoreactivity in water: Fate of triplet state and radical anion and production of reactive oxygen species. Environ. Sci. Technol. 2008 42 (9), 3459–3464. (12) Isakovic, A.; Markovic, Z.; Nikolic, N.; Todorovic-Markovic, B.; Vranjes-Djuric, S.; Harhaji, L.; Raicevi, N.; Romcevic, N.; VasiljevicRadovic, D.; Dramicanin, M.; Trajkovic, V. Inactivation of nanocrystalline C60 cytotoxicity by gamma-irradiation. Biomaterials 2006, 27 (29), 5049–5058. (13) Fortner, J. D.; Kim, D. I.; Boyd, A. M.; Falkner, J. C.; Moran, S.; Colvin, V. L.; Hughes, J. B.; Kim, J. H. Reaction of water-stable C60 aggregates with ozone. Environ. Sci. Technol. 2007, 41 (21), 7497–7502. (14) Cho, M.; Fortner, J. D.; Hughes, J. B.; Kim, J. H. Escherichia coli inactivation by water-soluble, ozonated C60 derivative: Kinetics and mechanisms. Environ. Sci. Technol. 2009, 43 (19), 7410–7415. (15) Lee, J.; Cho, M.; Fortner, J. D.; Hughes, J. B.; Kim, J. H. Transformation of aggregate C60 in the aqueous phase by UV irradiation. Environ. Sci. Technol. 2009, 43 (13), 4878–4883. (16) Hou, W. C.; Jafvert, C. T. Photochemistry of aqueous C60 clusters: Evidence of 1O2 formation and its role in mediating C60 phototransformation. Environ. Sci. Technol. 2009, 43 (14), 5257–5262. (17) Hou, W. C.; Jafvert, C. T. Photochemical transformation of aqueous C60 clusters in sunlight. Environ. Sci. Technol. 2009, 43 (2), 362–367. (18) Hou, W. C.; Kong, L. J.; Wepasnick, K. A.; Zepp, R. G.; Fairbrother, D. H.; Jafvert, C. T. Photochemistry of aqueous C60 clusters: Wavelength dependency and product characterization. Environ. Sci. Technol. 2010, 44 (21), 8121–8127. (19) Guldi, D. M.; Prato, M. Excited-state properties of C60 fullerene derivatives. Acc. Chem. Res. 2000, 33 (10), 695–703. (20) Guldi, D. M.; Asmus, K. D. Photophysical properties of monoand multiply-functionalized fullerene derivatives. J. Phys. Chem. A 1997, 101 (8), 1472–1481. (21) Krusic, P. J.; Wasserman, E.; Keizer, P. N.; Morton, J. R.; Preston, K. F. Radical reactions of C60. Science 1991, 254 (5035), 1183–1185. (22) Isakovic, A.; Markovic, Z.; Todorovic-Markovic, B.; Nikolic, N.; Vranjes-Djuric, S.; Mirkovic, M.; Dramicanin, M.; Harhaji, L.; Raicevic, N.; Nikolic, Z.; Trajkovic, V. Distinct cytotoxic mechanisms of pristine versus hydroxylated fullerene. Toxicol. Sci. 2006, 91 (1), 173–183. (23) Bogdanovic, V.; Stankov, K.; Icevic, I.; Zikic, D.; Nikolic, A.; Solajic, S.; Djordjevic, A.; Bogdanovic, G. Fullerenol C60(OH)24 eﬀects on antioxidative enzymes activity in irradiated human erythroleukemia cell line. J. Radiat. Res. 2008, 49 (3), 321–327. (24) Cho, M.; Chung, H. M.; Choi, W. Y.; Yoon, J. Y. Diﬀerent inactivation behaviors of MS-2 phage and Escherichia coli in TiO2 photocatalytic disinfection. Appl. Environ. Microbiol. 2005, 71 (1), 270–275. (25) Hatchard, C. G.; Parker, C. A. A new sensitive chemical actinometer 0.2. potassium ferrioxalate as a standard chemical actinometer. Proc. R. Soc. London Ser. A 1956, 235 (1203), 518–536. (26) Maness, P. C.; Smolinski, S.; Blake, D. M.; Huang, Z.; Wolfrum, E. J.; Jacoby, W. A. Bactericidal activity of photocatalytic TiO2 reaction: Toward an understanding of its killing mechanism. Appl. Environ. Microbiol. 1999, 65 (9), 4094–4098. (27) Zor, T.; Seliger, Z. Linearization of the Bradford protein assay increases its sensitivity: Theoretical and experimental studies. Anal. Biochem. 1996, 236 (2), 302–308. (28) Cho, M.; Lee, J.; Mackeyev, Y.; Wilson, L. J.; Alvarez, P. J. J.; Hughes, J. B.; Kim, J. H. Visible light sensitized inactivation of MS-2 9632

dx.doi.org/10.1021/es202269r |Environ. Sci. Technol. 2011, 45, 9627–9633

Environmental Science & Technology

ARTICLE

bacteriophage by a cationic amine-functionalized C60 derivative. Environ. Sci. Technol. 2010, 44 (17), 6685–6691. (29) Zheng, H.; Maness, P. C.; Blake, D. M.; Wolfrum, E. J.; Smolinski, S. L.; Jacoby, W. A. Bactericidal mode of titanium dioxide photocatalysis. J. Photochem. Photobiol. A 2000, 130 (2 3), 163–170. (30) Chudnicka, A.; Matysik, G. Research of enzymatic activities of fresh juice and water infusions from dry herbs. J. Ethnopharmacol. 2005, 99 (2), 281–286. (31) Lee, S.; Nishida, K.; Otaki, M.; Ohgaki, S. Photocatalytic inactivation of phage Qβ by immobilized titanium dioxide mediated photocatalyst. Water Sci. Technol. 1997, 35 (11 12), 101–106. (32) Bekbolet, M. Photocatalytic bactericidal activity of TiO2 in aqueous suspensions of E. coli. Water Sci. Technol. 1997, 35 (11 12), 95–100. (33) van Hemmen, J. J.; Meuling, W. J. Inactivation of Escherichia coli by superoxide radicals and their dismutation products. Arch. Biochem. Biophys. 1977, 182 (2), 743–8. (34) Muraseccosuardi, P.; Gassmann, E.; Braun, A. M.; Oliveros, E. Determination of the quantum yield of intersystem crossing of rosebengal. Helv. Chim. Acta 1987, 70 (7), 1760–1773. (35) Berg, J. M.; Tymoczko, J. L.; Stryer, L. Biochemistry, 6th ed.; Freeman: New York, 2006. (36) Tsuchido, T.; Katsui, N.; Takeuchi, A.; Takano, M.; Shibasaki, I. Destruction of the outer-membrane permeability barrier of Escherichia coli by heat treatment. Appl. Environ. Microbiol. 1985, 50 (2), 298–303. (37) Arana, I.; Santorum, P.; Muela, A.; Barcina, I. Chlorination and ozonation of waste-water: Comparative analysis of eﬃcacy through the eﬀect on Escherichia coli membranes. J. Appl. Microbiol. 1999, 86 (5), 883–888. (38) Cho, M.; Yoon, J. Measurement of OH radical CT for inactivating Cryptosporidium parvum using photo/ferrioxalate and photo/TiO2 systems. J. Appl. Microbiol. 2008, 104 (3), 759–766. (39) Driks, A. Bacillus subtilis spore coat. Microbiol. Mol. Biol. R. 1999, 63 (1), 1–20. (40) Dean, R. T.; Fu, S. L.; Stocker, R.; Davies, M. J. Biochemistry and pathology of radical-mediated protein oxidation. Biochem. J. 1997, 324, 1–18. (41) Wheatley, R. A. Some recent trends in the analytical chemistry of lipid peroxidation. TrAC, Trends Anal. Chem. 2000, 19 (10), 617–628. (42) Matthews, B. W. The structure of E. coli beta-galactosidase. Cr. Biol. 2005, 328 (6), 549–556. (43) Cho, M.; Kim, J.; Kim, J. Y.; Yoon, J.; Kim, J. H. Mechanisms of Escherichia coli inactivation by several disinfectants. Water Res. 2010 44 (11), 3410–3418. (44) Sullivan, C. J.; Morrell, J. L.; Allison, D. P.; Doktycz, M. J. Mounting of Escherichia coli spheroplasts for AFM imaging. Ultramicroscopy 2005, 105 (1 4), 96–102. (45) Cho, M.; Lee, Y.; Chung, H.; Yoon, J. Inactivation of Escherichia coli by photochemical reaction of ferrioxalate at slightly acidic and nearneutral pHs. Appl. Environ. Microbiol. 2004, 70 (2), 1129–1134. (46) Elovitz, M. S.; von Gunten, U.; Kaiser, H. P. Hydroxyl radical/ ozone ratios during ozonation processes. II. The eﬀect of temperature, pH, alkalinity, and DOM properties. Ozone-Sci. Eng. 2000, 22 (2), 123–150.

9633

dx.doi.org/10.1021/es202269r |Environ. Sci. Technol. 2011, 45, 9627–9633

ARTICLE pubs.acs.org/est

Linking Denitrification and Infiltration Rates during Managed Groundwater Recharge Calla M. Schmidt,†,* Andrew T. Fisher,† Andrew J. Racz,† Brian S. Lockwood,‡ and Marc Los Huertos§ †

University of California, Santa Cruz, Santa Cruz, California 95064, United States Pajaro Valley Water Management Agency, Watsonville, California 95076, United States § California State University, Monterey Bay, Seaside, California 93955, United States ‡

bS Supporting Information ABSTRACT: We quantify relations between rates of in situ denitriﬁcation and saturated inﬁltration through shallow, sandy soils during managed groundwater recharge. We used thermal methods to determine time series of point-speciﬁc ﬂow rates, and chemical and isotopic methods to assess denitriﬁcation progress. Zero order denitriﬁcation rates between 3 and 300 μmol L1 d1 were measured during inﬁltration. Denitriﬁcation was not detected at times and locations where the inﬁltration rate exceeded a threshold of 0.7 ( 0.2 m d1. Pore water proﬁles of oxygen and nitrate concentration indicated a deepening of the redoxocline at high ﬂow rates, which reduced the thickness of the zone favorable for denitriﬁcation. Denitriﬁcation rates were positively correlated with inﬁltration rates below the inﬁltration threshold, suggesting that for a given set of sediment characteristics, there is an optimal inﬁltration rate for achieving maximum nitrate load reduction and improvements to water supply during managed groundwater recharge. The extent to which results from this study may be extended to other managed and natural hydrologic settings remains to be determined, but the approach taken in this study should be broadly applicable, and provides a quantitative link between shallow hydrologic and biogeochemical processes.

’ INTRODUCTION Global extraction of groundwater for agriculture, municipal, and industrial use has increased dramatically in recent decades, and aquifer overdraft is a common problem throughout the world.1,2 In addition, groundwater resources are increasingly threatened by nutrient contamination. Intense global use of nitrogen (N) fertilizers in the last half century has led to widespread nitrate (NO3) contamination of groundwater, particularly in regions of extensive agricultural development.36 Nitrate contamination in groundwater poses an immediate risk to human health, and can lead to eutrophication of surface water bodies receiving groundwater discharge. 7 Protecting the quality and quantity of groundwater supplies requires ﬂexible management strategies that promote replenishment of aquifers. Artiﬁcial recharge of groundwater using surface water, commonly known as managed aquifer recharge (MAR), is an important management strategy for augmenting supplies8 and potentially improving water quality through physical and biogeochemical processes during inﬁltration and subsurface transport.9,10 MAR can use nonpristine water sources such as stormwater runoﬀ, excess irrigation ﬂows, and treated wastewater.11,12 Many potential MAR sources have high r 2011 American Chemical Society

concentrations of nutrients, especially nitrate, requiring a better understanding of techniques and conditions that could be used to reduce nutrient loads in inﬁltrating water. Denitriﬁcation, a microbially mediated process by which nitrate (NO3) is converted to dinitrogen gas (N2), can remove nitrate during MAR. Denitriﬁcation is often carried out by facultative heterotrophic bacteria that utilize nitrate NO3 as a terminal electron acceptor when dissolved oxygen (DO), which is energetically favorable, is unavailable.13 In recent decades, there has been considerable interest in quantifying denitriﬁcation in aquatic systems because this process can help to reduce nitrogen loading, and because incomplete denitriﬁcation results in emissions of N2O which is a powerful greenhouse gas.14,15 Denitriﬁcation rates in shallow sediments are inﬂuenced by redox conditions,16 nutrient concentration,17,18 availability of dissolved and sedimentary organic matter,1922 soilwater content,10,23 soil texture,24 mineral composition of the sediments,25 water Received: July 8, 2011 Accepted: September 13, 2011 Revised: September 8, 2011 Published: October 12, 2011 9634

dx.doi.org/10.1021/es2023626 | Environ. Sci. Technol. 2011, 45, 9634–9640

Environmental Science & Technology

ARTICLE

temperature,26 and ﬂuid residence time.27 However, the eﬀects of ﬂuid ﬂow on denitriﬁcation rates remain poorly understood in many settings, in part because of the diﬃcultly of making contemporaneous, in situ measurements of both processes. Studies conducted at the mesocosm scale have limited ability to capture spatial and temporal heterogeneity in ﬂuid ﬂow, and studies conducted at larger scales generally use spatially averaged ﬂuid ﬂow rates, obscuring functional relationships. Laboratory studies also may not be representative of ﬁeld conditions. For example, Smith et al.28 estimated denitriﬁcation rates that were up to 26 times higher in incubation experiments using cores of aquifer material than rates determined in situ in the same aquifer. Similarly, laboratory experiments using intact stream sediment cores found denitriﬁcation rates of 100400 mg N m2 d1, whereas studies completed at the reach scale detected no denitriﬁcation once dilution of the stream by groundwater was taken into account.29 The primary objective of this study was to link the spatial and temporal dynamics of ﬂuid ﬂow and denitriﬁcation. We accomplished this goal by combining pore-ﬂuid nitrate concentration and isotopic data with point-speciﬁc inﬁltration rates. Fieldwork was completed within an operating MAR system, which has uniquely strong experimental control at a ﬁeld scale, with accurate accounting of ﬂuid and nutrient storage and ﬂuxes. An earlier study at this ﬁeld site showed that the nutrient load delivered to the aquifer during MAR is 3060% less than that in water entering the recharge pond, based on an assessment of whole-system inﬁltration rates and changes to local nitrate concentrations in shallow sediment pore ﬂuids.30 In this study, we combine newly developed methods to measure inﬁltration rates with contemporaneous and collocated determination of denitriﬁcation rates to relate the dynamics of shallow hydrologic and biogeochemical processes to nitrate load reduction.

’ EXPERIMENTAL SITE AND METHODS

Figure 1. Maps and diagrams showing ﬁeld site and methods used in this study. (a) Regional map showing the Pajaro Valley Groundwater Basin and the location of the recharge pond (labeled HS-MAR). Inset shows location of site in central coastal California, U.S.A. (b) Detailed site map showing instrument locations within the Harkins Slough artiﬁcial recharge pond. (c) Schematic illustration showing a cluster of instruments for thermal measurement and ﬂuid sampling in shallow soils below the base of the inﬁltration pond.

Field Site. Field work was completed using the Harkins Slough managed aquifer recharge system (HS-MAR), located in the Pajaro Valley of central coastal California (Figure 1). The HSMAR system is a three hectare infiltration pond that occupies a modified natural depression having a maximum depth of 6 m. Water is diverted to the recharge pond from nearby Harkins Slough (a wetland draining part of the 3400 km2 Pajaro River watershed) during the winter rainy season when flows are sufficiently high. As infiltration proceeds during each operating year (typically 100 days between January to May), a wetting front and inverted water table are driven downward into soils underlying the pond, eventually forming a 12 m thick saturated layer. Infiltrating water passes through a 2030 m thick vadose zone and creates a local water table mound in the underlying perched aquifer within eolian, fluvial, and alluvial sediments.31 Diversions from Harkins Slough to the infiltration pond continue each year until the rate of infiltration below the pond is greatly reduced because of a reduction in hydraulic conductivity.32 At the end of each operating season when the pond is dry, shallow soils are scraped and tilled in preparation for future MAR. Artificial recharge water is recovered from the shallow aquifer using dedicated wells that encircle the pond, blended with water from elsewhere in basin, and distributed to local growers for use in irrigation. 9635

dx.doi.org/10.1021/es2023626 |Environ. Sci. Technol. 2011, 45, 9634–9640

Environmental Science & Technology

ARTICLE

Figure 2. Inﬁltration rate and the percentage of initial nitrate concentration removed during 2008 operations at eight piezometer where collocated measurments were made. Water samples were collected at piezometers screened at 0.5 m depth beneath the base of the pond.

Point Measurements of Infiltration Rate. Point-specific infiltration rates during MAR were determined using heat as a tracer, applying time-series analysis to subsurface thermal data.3234 Temperature was recorded at 15-min intervals with a resolution of 0.02 °C using autonomous thermal loggers suspended on cables inside water-filled, 3.8 cm diameter polyvinyl chloride (PVC) tubes (Figure 1b,c). Tubes were installed in holes excavated with a hand auger, and the annulus around the tubes was backfilled with bentonite and native soil (2008), or silica slurry (2009), to ensure a good thermal contact with the surrounding soil. Time-series records of temperature were filtered to extract daily temperature fluctuations, and pairs of subsurface sensors were analyzed to resolve the amplitude reduction of this diurnal temperature signal with depth. These values were used to solve for fluid infiltration rates based on a one-dimensional (vertical) conservation of heat equation.34 This method yields daily values of the vertical component of infiltration representative of the interval between sensors pairs. Infiltration rates were interpreted only when soils surrounding the sensors were saturated,32 which corresponds to times when shallow piezometers were used to recover pore water samples. Fluid Sampling and Nitrate Measurements. Synoptic water samples were collected from the infiltration pond itself and from nests of piezometers screened beneath the base of the pond (Figure 1b,c). Twelve piezometer nests were installed along four transects during the 2008 operating season, with screen depths of 50 and 100 cm beneath the base of the pond. An additional profile was added in 2009 and fluid sampling piezometers were screened at depths of 30, 60, and 90 cm to improve vertical resolution. Data presented in this study are from piezometers

collocated with eight thermal probes in 2008 and three thermal probes in 2009. Fluid sampling piezometers were constructed from 1-cm diameter polycarbonate tubing, perforated at the base and wrapped with a fine mesh nylon screen. Piezometer holes were bored with a hand auger when the pond was dry, and a coarse sand filter (grain diameter = 0.71.7 mm, well rounded, >97% silica) was installed around the screen and capped with a 10 cm bentonite seal. The annulus of each borehole was backfilled with native soil, and a second bentonite seal was placed at the ground surface to prevent water from flowing down the piezometer tubing. Piezometers were developed after installation to ensure a good connection with the formation by saturating the sand filter and soil around the screens, then flushing water back and forth through the piezometer screen using a peristaltic pump. Fluid samples from piezometers were collected with nylon tubing that extended to the edge of the pond, allowing access throughout MAR operations. Additionally, pore water samples were collected throughout the season using dialysis samplers (peepers).35 Peepers were submerged in a container ﬁlled with deionized water and bubbled with N2 for a minimum of 10 days, then deployed in the base of the pond as it was being ﬁlled. Peepers were left deployed in the base of the pond for at least 14 days, then they were recovered sequentially at diﬀerent times during the MAR season, providing samples that are most representative of conditions for ∼24 h preceding recovery.36 Water samples were ﬁeld ﬁltered with 0.45 μmol cellulose acetate ﬁlters into acid washed polyethylene bottles. A subset of samples was ﬁltered through 0.22 μmol cellulose acetate ﬁlters for nitrate isotopic analysis. All samples were placed on ice immediately and returned to the lab. Samples collected for 9636

dx.doi.org/10.1021/es2023626 |Environ. Sci. Technol. 2011, 45, 9634–9640

Environmental Science & Technology

ARTICLE

Figure 3. Porewater concentrations of DO and NO3 at three diﬀerent inﬁltration rates at location 0B. (a) 1.4 m d1 on MAR day 10 (b) 0.9 m d1 on MAR day 21 and (c) 0.5 m d1, on MAR day 30.

nutrient analysis were stored at 4 °C and samples collected for isotopic analysis were frozen until analyzed. Dissolved and total nutrient concentrations were determined within one week of collection by colormetric ﬂow injection analysis on a Lachat Instrument QuickChem 800, with accuracy (based on ﬁeld and lab blanks, spikes, and standards) of 25%. The δ15N and δ18O of nitrate (relative to air and VSMOW, respectively) were measured at the UCSC Stable Isotope Facility using a procedure modiﬁed from McIlvin and Altabet (2005). Absolute accuracy for NO3 isotopic analyses was 0.2% and 0.3% for δ15N and δ18O, respectively. Identifying Denitrification and Quantifying its Rate. We used the difference between δ15N and δ18O of nitrate in pond surface water and residual nitrate in pore fluids to identify the occurrence of net denitrification. Isotopic fractionation during microbial denitrification causes residual nitrate to become progressively enriched in δ18O and δ15N, with a typical fractionation ratio (Δδ18O: Δδ15N) around 0.6.37,38 At times and subsurface locations where this characteristic enrichment of δ18O and δ15N of nitrate was observed, we used the gradient in NO3 concentration to quantify the extent of net denitrification during infiltration. Denitrification rates (r) were determined as follows: r ¼ f½NO3 initial ½NO3 piezometer g=t

ð1Þ

where initial refers to samples collected in pond surface water, piezometer refers to porewater sampled from piezometers screened in the saturated sediments beneath the pond, and t is the mean travel time between the bottom of the pond and the piezometer screen. The dominant flow direction through shallow, saturated soil below the pond is vertically downward (the pond is much wider than deep, and vertical head gradients are considerably larger than horizontal gradients),32 so the mean travel time from the pond to subsurface piezometer screens, and between screens, is approximated as a function of infiltration rate and spacing between sampling points. Water samples from the pond and piezometers were collected on the same day due to rapid rates of infiltration relative to the spacing between sample depths, and the comparatively slow rate of change of nitrate concentration in the pond.

’ RESULTS Infiltration Rates. Point measurements of daily infiltration rate indicate considerable temporal and spatial variability in the rate of infiltration through the base of the pond. Infiltration rates

Figure 4. Change in δ18O and δ15N of NO3 during inﬁltration in 2008 for the same locations shown in Figure 2. δ18O and δ15N values of nitrate in pond surface water represented initial nitrate isotopic compositions. The linear relationship and ratio of isotopic enrichment Δ δ18O:Δ δ15N of 0.64 (r2 = 0.96) suggest that net denitriﬁcation occurred in the saturated zone beneath the recharge pond. Samples are grouped according to denitriﬁcation rate (r), to show that a consistent relationship between denitriﬁcation rate and the magnitude of Δ δ15N and Δ δ18O was not observed in this study (Table S1, Supporting Information).

ranged from <0.1 to 4.4 m d1. The highest rates of infiltration occurred at piezometers located in Profiles 0, 1, and 2 (Figures 2a-e, and S1 in Supporting Information). In this portion of the pond, infiltration rates were higher for the first 40 days of operations, followed by a rapid decline. In contrast, infiltration rates measured in piezometers in profiles 3 and 4 ranged from <0.1 to 0.2 m d1 when the pond was first filled. In this portion of the pond, infiltration rates gradually rose throughout the MAR operating season, but never exceeded 0.5 m d1 (Figures 2f-h). Denitrification During Infiltration. Nitrate was the dominant form or inorganic nitrogen in pond water during this study. The concentration of nitrate in the pond ranged from 650 μmol L1 NO3 when the pond was first filled to 20 μmol L1 NO3 at the end of the season, with changes occurring relatively smoothly and systematically throughout the two operating seasons (Figure S2, Supporting Information). Nitrate concentration in the pond declined over time due to flushing of the slough by winter rain, with nitrate concentration generally <100 μmol L1 after the first 40 days of operations. Ammonium was present in the pond water at concentrations typically less than 5 μmol L1 NH4+. The concentration of dissolved organic carbon in the pond was considerably higher than that of nitrate, ranging from 0.5 to 2 mmol L1 C. DO and NO3 concentration proﬁles in pore water were determined with dialysis samplers at several locations collected when inﬁltration rates were between 0.5 to 1.4 m d1 (Figure 3). These data show that the zone of oxygen penetration into the subsurface deepened at higher ﬂow rates. Oxic conditions persisted to >15 cm when the inﬁltration rate was above 1 m d1, but shallowed to 5 cm at an inﬁltration rate of 0.5 m d1. The depth at which NO3 removal was initiated was also shallower at lower ﬂow rates. At the resolution of the dialysis samplers (2 cm), the reduction of [DO] and [NO3] was collocated when inﬁltration rates were <0.9 m d1. 9637

dx.doi.org/10.1021/es2023626 |Environ. Sci. Technol. 2011, 45, 9634–9640

Environmental Science & Technology

Figure 5. Cross plot of denitriﬁcation rate, r, normalized by initial nitrate concentration [NO3] versus inﬁltration rate into the saturated zone beneath the pond. Samples with initially higher and lower values of [NO3] show diﬀerent, but internally consistent, trends.

Nitrate concentrations in the saturated zone beneath the pond were lower than in pond surface water at times and locations where and when inﬁltration rates were less than 0.7 ( 0.2 m d1 (Figure 2). At locations where inﬁltration rates were consistently less than 0.7 ( 0.2 m d1 (e.g., PZ-3D and PZ-4D), the percentage of nitrate removed in the subsurface increased with time of MAR operations, and a 60% decrease from the initial nitrate concentration was common. The threshold inﬁltration rate above which denitriﬁcation was not observed varied somewhat between sampling locations. For example, nitrate removal was not observed at inﬁltration rates above 0.5 m d1 at piezometers in Proﬁle 2, whereas nitrate removal was observed at inﬁltration rates up to 0.8 m d1 at PZ-1C (Figure 2). The isotopic composition of nitrate in water diverted to the pond changed over the time, with both the δ15N and δ18O of nitrate in the pond decreasing during the operational season (Table S1, Supporting Information). Figure 4 shows the change in δ15N and δ18O (Δδ15N, Δδ18O) between pond water and residual nitrate in pore waters sampled on the same day at times and locations where NO3 concentrations decreased during inﬁltration. In many locations both δ15N and δ18O increased by more than 15 % in the ﬁrst 50 cm of inﬁltration, but values at individual locations vary considerably. The absolute magnitude of the change in δ15N and δ18O of nitrate varied over time and between sample locations, and we did not observe a relationship between denitriﬁcation rate and isotopic enrichment. However, despite variability in the magnitude of Δδ15N and Δδ18O we observed a relatively consistent ratio of Δδ18O: Δδ15N = 0.64 (Figure 4). Zero order denitriﬁcation rates in the saturated zone beneath the pond ranged from 3 to 300 μmol L1 d1. The highest rates of denitriﬁcation occurred when [NO3 ] in the pond was >100 μmol L1, and denitriﬁcation rates generally decreased with pond nitrate concentration, consistent with a ﬁrst-order behavior. When [NO3 ] was less than 50 μmol L1, near complete removal of NO3 during shallow inﬁltration was common. Because denitriﬁcation rates decreased with the pond nitrate concentration, it is useful to normalize by initial nitrate concentration when comparing denitriﬁcation rates between locations and sampling dates. There was a positive correlation between inﬁltration rate and ﬁrst order denitriﬁcation rate at all locations monitored in 2008 and 2009 (Figure 5).

ARTICLE

’ DISCUSSION Reductions in nitrate concentration and duel enrichment of δ15N and δ18O of residual nitrate were observed in the sediments beneath the recharge pond, consistent with microbial denitriﬁcation occurring during inﬁltration. Denitriﬁcation in the saturated zone below the inﬁltration pond occurred when inﬁltration rates were between 0.01 m d1 and 0.9 m d1. At locations in Proﬁles 0, 1, 2, where inﬁltration rates tended to be greatest, there was no measurable denitriﬁcation in the upper 1 m of soil when inﬁltration rates were above 0.7 ( 0.2 m d1. When the rate of inﬁltration was below the maximum threshold, denitriﬁcation rate was positively correlated with inﬁltration rate. This observation suggests that the higher ﬂux of NO3 into the subsurface may have stimulated denitriﬁcation. This interpretation is consistent with laboratory experiments during which increased ﬂuid ﬂow enhanced uptake of DO and NO3 by stimulating facultative aerobes which rapidly switch to NO3 once [DO] is lowered.39,40 The ability of the soil microbial community to denitrify available nitrate at higher inﬁltration rates was likely limited by deeper penetration of oxygen into the saturated zone (Figure 3). Deepening of the redoxocline reduced the thickness of the saturated zone that was favorable for denitriﬁcation. In addition, the residence time of inﬁltrating water within this zone was reduced at higher inﬁltration rates, limiting the time available for denitriﬁcation processes to occur. From the present study it is not possible to determine if denitriﬁcation occurred in the unsaturated zone beneath our sampling points at times when high rates of inﬁltration caused deeper penetration of oxic conditions. However, no additional denitriﬁcation beneath the shallow saturated zone was detected in a prior study of nitrate loading at this site.30 If denitriﬁcation does not occur in the unsaturated zone, then denitriﬁcation during MAR is limited mainly by the thickness of saturated zone, particularly at high rates of inﬁltration. The highest zero-order rate of denitriﬁcation observed in this study was 300 μmol L1 d1, whereas the average rate of denitriﬁcation calculated for shallow soils below the MAR pond system was 40 μmol L1 d1. Denitriﬁcation rates in artiﬁcial recharge systems are not widely reported, but rates between 40 and 500 μmol L1 d1 can be inferred from inﬁltration rates and gradients in nitrate concentrations reported by Greskowiak et al.41 for a recharge pond near Lake Tengel, Germany. Interestingly, in that system a reduction in nitrate concentration was observed at ﬂow rates as high as 3 m d1, but signiﬁcant nitrate removal at 150 cm depth beneath the pond was not achieved until inﬁltration rates declined to less than 1 m d1. Zero order denitriﬁcation rates calculated in this study and in Greskowiak et al.41 are within the range of values determined for small streams, lakes, and rivers reported in recent compilations (after converting MAR denitriﬁcation rates by volume to rates by area). 45,46 However, MAR denitriﬁcation rates reported here and in Greskowiak et al. are among the highest reported in the groundwater literature, where rates of less than 1 μmol L1 d1 are more common.42,43 Apparent denitriﬁcation rates are likely higher in MAR systems due to high ﬂow rates, and because measurements are made over short ﬂow paths that isolate the denitrifying zone. Denitriﬁcation rates in the Harkins Slough MAR system may also be comparatively high because there is no limitation on the availability of dissolved carbon as diverted water originates from a local wetland with high carbon content. 9638

dx.doi.org/10.1021/es2023626 |Environ. Sci. Technol. 2011, 45, 9634–9640

Environmental Science & Technology It is possible that similarly high rates of denitriﬁcation occur in natural groundwater systems where there is an adequate supply of electron donors and solutes migrate into zones of reducing conditions, but these elevated denitriﬁcation rates would be diﬃcult to detect in studies conducted over long ﬂow paths. For example Tesoriero et al.44 measured denitriﬁcation rates of 3.5 to 7 μmol L1 d1 along an approximately 9 month long groundwater ﬂow path to a creek, but in situ estimates of denitriﬁcation across the redoxocline adjacent to the creek were as high as 385 μmol L d1. The primary uncertainties in the denitriﬁcation rates determined in this study result from errors in inﬁltration rates calculated using thermal methods, and analytical errors associated with ﬂuid sampling, processing and lab analyses. The latter are relatively small based on the methods used in this study, generally <5%, but accurate determination of inﬁltration rates using thermal methods requires measurement or estimation of soil thermal properties (conductivity, heat capacity, dispersivity) that vary in natural sediments.47 For a true inﬁltration rate of 1 m d1, errors in calculated values on the order of (20% are expected based on a reasonable range of sediment properties.34 Errors of this magnitude would have little inﬂuence on the relative spatial and temporal trends documented in this study, although they would shift calculated denitriﬁcation rates proportionately. This study illustrates considerable spatial and temporal variability in denitriﬁcation rates during inﬁltration, even within a relatively homogeneous and controlled soilwater system. In addition, at 50% of the locations where point inﬁltration rate measurements were determined in 2008 and 2009, the rate of inﬁltration was too high to allow measurable denitriﬁcation for periods of 1 day (PZ-2B) to 17 days (PZ-1B). This suggests that limiting inﬁltration rates during MAR could be beneﬁcial for improving water quality in underlying aquifers. It remains to be determined how representative our results are for rates of denitriﬁcation in other MAR systems, and within a variety of natural hydrologic systems, when there is no limit to available organic carbon and other electron donors. That said, this work shows how quantitative links can be elucidated between hydrologic and biogeochemical processes during soil inﬁltration. Additional work is needed to determine the nature of microbial populations that are most important for facilitating denitriﬁcation in shallow soils during inﬁltration and recharge, and the extent to which conditions in managed systems can be optimized so as to achieve simultaneous water supply and water quality objectives.

’ ASSOCIATED CONTENT

bS

Supporting Information. A table (Table S1) of the data presented in Figure 3 and three ﬁgures which provide additional experimental details. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: 831-459-2838; fax: 831-459-3074; e-mail: cschmidt@ ucsc.edu

’ ACKNOWLEDGMENT We are grateful to the leadership and staﬀ of the Pajaro Valley Water Management Agency, who provided crucial support and

ARTICLE

encouragement for this work. Field and lab assistance and advice was provided by Rob Franks, Dyke Andreasen, Dan Sampson, Adina Paytan, Carol Kendall, Mark Altabet, Jonathan Lear, Joanna Hoﬀman, and Nic Massetani. This project was funded by awards from the UCSC Committee on Research, the UCSC Science, Technology, Engineering, Policy, and Society (STEPS) Institute, the US EPA STAR Fellowship program (FP-91679901), the California Water Resources Control Agency (through the Santa Cruz County Resource Conservation District, UCSCPAJARO013107), and the National Institutes for Water Resources (08HQGR0054).

’ REFERENCES (1) Wada, Y.; van Beek, L. P. H.; van Kempen, C. M.; Reckman, J.; Vasak, S.; Bierkens, M. F. P. Global depletion of groundwater resources. Geophys. Res. Lett. 2010, 37 (20), 1–5. (2) Konikow, L. F.; Kendy, E. Groundwater depletion: A global problem. Hydrogeol. J. 2005, 13 (1), 317–320. (3) Burow, K. R.; Nolan, B. T.; Rupert, M. G.; Dubrovsky, N. M. Nitrate in groundwater of the United States, 19912003. Environ. Sci. Technol. 2010, 44 (13), 4988–4997. (4) Zhang, W. L.; Tian, Z. X.; Zhang, N.; Li, X. Q. Nitrate pollution of groundwater in northern China. Agric. Ecosyst. Environ. 1996, 59 (3), 223–231. (5) Strebel, O.; Duynisveld, W. H. M.; Bottcher, J. Nitrate pollution of groundwater in Western-Europe. Agric. Ecosyst. Environ. 1989, 26 (34), 189–214. (6) Nolan, B. T.; Hitt, K. J.; Ruddy, B. C. Probability of nitrate contamination of recently recharged groundwaters in the conterminous United States. Environ. Sci. Technol. 2002, 36 (10), 2138–2145. (7) 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), 364–376. (8) Bouwer, H. P., D., Brown, Germain, J.; , D., Morris, T., Brown, C., Dillon, P. Rycus, M. Design, Operation, and Maintenance for Sustainable Underground Strorage Facilities; International Water Association. 2008. (9) Dillon, P. Future management of aquifer recharge. Hydrogeol. J. 2005, 13 (1), 313–316. (10) Fox, P.; Houston, A. R.; Westerhoﬀ, P., Eds. Advances in Soil Aquifer Treatment Research for Sustainable Water Reuse; Awwa Research Foundation: Denver, CO, 2006. (11) Ma, L.; Spalding, R. F. Eﬀects of artiﬁcial recharge on ground water quality and aquifer storage recovery. J. Am. Water Resour. Assoc. 1997, 33 (3), 561–572. (12) Drewes, J. E. Ground water replenishment with recycled waterwater quality improvements during managed aquifer recharge. Ground Water 2009, 47 (4), 502–505. (13) Knowles, R. Denitriﬁcation. Microbiol. Rev. 1982, 46 (1), 43–70. (14) Beaulieu, J. J.; Arango, C. P.; Tank, J. L. The eﬀects of season and agriculture on nitrous oxide production in headwater streams. J. Environ. Qual. 2009, 38 (2), 637–646. (15) Seitzinger, S. P.; Kroeze, C.; Styles, R. V. Global distribution of N2O emissions from aquatic systems: natural emissions and anthropogenic eﬀects. Chemosphere—Global Change Sci. 2000, 2 (34), 267–279. (16) Mariotti, A.; Landreau, A.; Simon, B. N-15 isotope biogeochemistry and natural denitriﬁcation process in groundwater—Application to the chalk aquifer of Northern France. Geochim. Cosmochim. Acta 1988, 52 (7), 1869–1878. (17) Fleischer, S.; Gustafson, A.; Joelsson, A.; Pansar, J.; Stibe, L. Nitrogen removal in created ponds. Ambio 1994, 23 (6), 349–357. (18) Duﬀ, J. H.; Jackman, A. P.; Triska, F. J.; Sheibley, R. W.; Avanzino, R. J. Nitrate retention in riparian ground water at natural and 9639

dx.doi.org/10.1021/es2023626 |Environ. Sci. Technol. 2011, 45, 9634–9640

Environmental Science & Technology elevated nitrate levels in north central Minnesota. J. Environ. Qual. 2007, 36 (2), 343–353. (19) Bradley, P. M.; Fernandez, M.; Chapelle, F. H. Carbon limitation of denitriﬁcation rates in an anaerobic groundwater system. Environ. Sci. Technol. 1992, 26 (12), 2377–2381. (20) Hume, N. P.; Fleming, M. S.; Horne, A. J. Denitriﬁcation potential and carbon quality of four aquatic plants in wetland microcosms. Soil Sci. Soc. Am. J. 2002, 66 (5), 1706–1712. (21) Hunter, R. G.; Combs, D. L.; George, D. B. Nitrogen, phosphorous, and organic carbon removal in simulated wetland treatment systems. Arch. Environ. Contam. Toxicol. 2001, 41 (3), 274–281. (22) Mulholland, P. J.; Valett, H. M.; Webster, J. R.; Thomas, S. A.; Cooper, L. W.; Hamilton, S. K.; Peterson, B. J. Stream denitriﬁcation and total nitrate uptake rates measured using a ﬁeld N-15 tracer addition approach. Limnol. Oceanogr. 2004, 49 (3), 809–820. (23) Groﬀman, P. M.; Gold, A. J.; Simmons, R. C. Nitrate dynamics in riparian forests—Microbial studies. J. Environ. Qual. 1992, 21 (4), 666–671. (24) Pinay, G.; Black, V. J.; Planty-Tabacchi, A. M.; Gumiero, B.; Decamps, H. Geomorphic control of denitriﬁcation in large river ﬂoodplain soils. Biogeochemistry 2000, 50 (2), 163–182. (25) Postma, D.; Boesen, C.; Kristiansen, H.; Larsen, F. Nitrate reduction in an unconﬁned sandy aquifer—Water chemistry, reduction processes, and geochemical modeling. Water Resour. Res. 1991, 27 (8), 2027–2045. (26) Stanford, G.; Dzienia, S.; Vanderpol, R. A. Eﬀect of temperature on denitriﬁcation in soils. Soil Sci. Soc. Am. J. 1975, 39 (5), 867–870. (27) Christensen, S.; Simkins, S.; Tiedje, J. M. Spatial variation in denitriﬁcation—Dependency of activity centers on the soil environment. Soil Sci. Soc. Am. J. 1990, 54 (6), 1608–1613. (28) Smith, R. L.; Garabedian, S. P.; Brooks, M. H. Comparison of denitriﬁcation activity measurements in groundwater using cores and naturalgradient tracer tests. Environ. Sci. Technol. 1996, 30 (12), 3448–3456. (29) Kellman, L. Nitrate removal in a ﬁrst-order stream: reconciling laboratory and ﬁeld measurements. Biogeochemistry 2004, 71 (1), 89–105. (30) Schmidt, C. M.; Fisher, A. T.; Racz, A. J.; Wheat, C. G.; Los Huertos, M.; Lockwood, B. Rapid nutrient load reduction during inﬁltration as part of managed aquifer recharge in an agricultural groundwater basin, Pajaro Valley, CA. Hydrol. Process., in press. (31) Hanson, R. T. Geohydrologic Framework of Recharge and Seawater Intrusion in the Pajaro Valley, Santa Cruz and Monterey Counties, California; 2003 (DOI: 10.1002/hyp.8320). (32) Racz, A. J.; Fisher, A. T.; Schmidt, C. M.; Lockwood, B.; Los Huertos, M. Quantifying the spatial and temporal dynamics of inﬁltration during managed aquifer recharge using mass balance and thermal methods. Ground Water, in press. (33) Constantz, J.; Thomas, C. L. The use of streambed temperature proﬁles to estimate the depth, duration, and rate of percolation beneath arroyos. Water Resour. Res. 1996, 32 (12), 3597–3602. (34) Hatch, C. E.; Fisher, A. T.; Revenaugh, J. S.; Constantz, J.; Ruehl, C. Quantifying surface water-groundwater interactions using time series analysis of streambed thermal records: Method development. Water Resour. Res. 2006, 42, 10. (35) Carignan, R. Interstitial water sampling by dialysis-methodological notes. Limnol. Oceanogr. 1984, 29 (3), 667–670. (36) Webster, I. T.; Teasdale, P. R.; Grigg, N. J. Theoretical and experimental analysis of peeper equilibration dynamics. Environ. Sci. Technol. 1998, 32 (11), 1727–1733. (37) Lehmann, M. F.; Reichert, P.; Bernasconi, S. M.; Barbieri, A.; McKenzie, J. A. Modelling nitrogen and oxygen isotope fractionation during denitriﬁcation in a lacustrine redox-transition zone. Geochim. Cosmochim. Acta 2003, 67 (14), 2529–2542. (38) Chen, D. J. Z.; MacQuarrie, K. T. B. Correlation of delta N-15 and delta O-18 in NO3 during denitriﬁcation in groundwater. J. Environ. Eng. Sci. 2005, 4 (3), 221–226. (39) John, P. Aerobic and anaerobic bacterial respiration monitored by electrodes. J. Gen. Microbiol. 1977, 98 (JAN), 231–238.

ARTICLE

(40) O’Connor, B. L.; Hondzo, M. Enhancement and inhihition of denitriﬁcation by ﬂuid-ﬂow and dissolved oxygen ﬂux to stream sediments. Environ. Sci. Technol. 2008, 42 (1), 119–125. (41) Greskowiak, J.; Prommer, H.; Massmann, G.; Johnston, C. D.; Nutzmann, G.; Pekdeger, A. The impact of variably saturated conditions on hydrogeochemical changes during artiﬁcial recharge of groundwater. Appl. Geochem. 2005, 20 (7), 1409–1426. (42) Green, C. T.; Puckett, L. J.; Bohlke, J. K.; Bekins, B. A.; Phillips, S. P.; Kauﬀman, L. J.; Denver, J. M.; Johnson, H. M. Limited occurrence of denitriﬁcation in four shallow aquifers in agricultural areas of the United States. J. Environ. Qual. 2008, 37 (3), 994–1009. (43) Groﬀman, P. M.; Altabet, M. A.; Bohlke, J. K.; Butterbach-Bahl, K.; David, M. B.; Firestone, M. K.; Giblin, A. E.; Kana, T. M.; Nielsen, L. P.; Voytek, M. A. Methods for measuring denitriﬁcation: Diverse approaches to a diﬃcult problem. Ecol. Appl. 2006, 16 (6), 2091–2122. (44) Tesoriero, A. J.; Liebscher, H.; Cox, S. E. Mechanism and rate of denitriﬁcation in an agricultural watershed: Electron and mass balance along groundwater ﬂow paths. Water Resour. Res. 2000, 36 (6), 1545–1559. (45) Pina-Ochoa, E.; Alvarez-Cobelas, M. Denitriﬁcation in aquatic environments: A cross-system analysis. Biogeochemistry 2006, 81 (1), 111–130. (46) Mulholland, P. J.; Helton, A. M.; Poole, G. C.; Hall, R. O.; Hamilton, S. K.; Peterson, B. J.; Tank, J. L.; Ashkenas, L. R.; Cooper, L. W.; Dahm, C. N.; Dodds, W. K.; Findlay, S. E. G.; Gregory, S. V.; Grimm, N. B.; Johnson, S. L.; McDowell, W. H.; Meyer, J. L.; Valett, H. M.; Webster, J. R.; Arango, C. P.; Beaulieu, J. J.; Bernot, M. J.; Burgin, A. J.; Crenshaw, C. L.; Johnson, L. T.; Niederlehner, B. R.; O’Brien, J. M.; Potter, J. D.; Sheibley, R. W.; Sobota, D. J.; Thomas, S. M. Stream denitriﬁcation across biomes and its response to anthropogenic nitrate loading. Nature 2008, 452 (7184), 202–U46. (47) Ferguson, G.; Bense, V. Uncertainty in 1D heat-ﬂow analysis to estimate groundwater discharge to a stream. Ground Water 2011, 49 (3), 336–347.

9640

dx.doi.org/10.1021/es2023626 |Environ. Sci. Technol. 2011, 45, 9634–9640

ARTICLE pubs.acs.org/est

Groundwater or Floodwater? Assessing the Pathways of Metal Exports from a Coastal Acid Sulfate Soil Catchment Isaac R. Santos,*,† Jason de Weys,† and Bradley D. Eyre† †

Centre for Coastal Biogeochemistry, School of Environmental Science and Management, Southern Cross University, Lismore, NSW 2480, Australia

bS Supporting Information ABSTRACT: Daily observations of dissolved aluminum, iron, and manganese in an estuary downstream of a coastal acid sulfate soil (CASS) catchment provided insights into how ﬂoods and submarine groundwater discharge drive wetland metal exports. Extremely high Al, Fe, and Mn concentrations (up to 40, 374, and 8 mg L 1, respectively) were found in shallow acidic groundwaters from the Tuckean Swamp, Australia. Signiﬁcant correlations between radon (a natural groundwater tracer) and metals in surface waters revealed that metal loads were driven primarily by groundwater discharge. Dissolved Fe, Mn, and Al loads during a 16-day ﬂood triggered by a 213 mm rain event were respectively 80, 35, and 14% of the total surface water exports during the four months of observations. Counter clockwise hysteresis was observed for Fe and Mn in surface waters during the ﬂood due to delayed groundwater inputs. Groundwater-derived Fe ﬂuxes into artiﬁcial drains were 1 order of magnitude higher than total surface water exports, which is consistent with the known accumulation of monosulﬁdic black ooze within the wetland drains. Upscaling the Tuckean catchment export estimates yielded dissolved Fe ﬂuxes from global acid sulfate soil catchments on the same order of magnitude of global river inputs into estuaries.

’ INTRODUCTION The relationship between riverine exports and biogeochemical processing of trace metals within estuaries has been studied in the last few decades.1 However, little is known about how ﬂoods and groundwater may contribute to trace metal cycling in estuaries. Since trace element concentrations in groundwater often exceed those in surface waters, groundwater can play a signiﬁcant role in regional and global metal budgets.2 The contribution of groundwater to estuarine budgets will likely increase as human activity on coastal watersheds increases.3 Groundwater may be especially important at sites where low pH increases the solubility of metals. Millions of hectares of coastal ﬂoodplains and wetlands worldwide are underlain by sediments rich in iron sulﬁde minerals.4 When drained for agriculture and grazing, iron sulﬁde minerals oxidize and produce acid that can be ﬂushed to adjacent waterways.5 In Australia and many other (sub-) tropical countries, thousands of kilometers of artiﬁcial drains have been constructed on coastal ﬂoodplains underlain by sulﬁdic soils.6 These drains are designed to lower the water table, but they also expose reducing soils to atmospheric oxygen. Previous studies have documented that low pH waters from coastal acid sulfate soils (CASS) are associated with trace metals concentrations exceeding water quality guidelines.7 Soluble metals from CASS, including Al3+ and Fe2+, can impact entire estuarine ecosystems8 and eventually be stored in coastal sediments.9 Episodic ﬂoods in tropical CASS catchments have been linked with severe estuarine degradation. During summer ﬂoods, ﬂoodplain r 2011 American Chemical Society

vegetation decomposition may lead to deoxygenation events that can result in ﬁsh kills.10 The discharge of acidic CASS groundwaters often follows the release of deoxygenated surface waters from the ﬂoodplain.11 There is a scarcity of research relating to the ﬂux of heavy metals from CASS. Previous dissolved metal work in CASS catchments were largely based on spatial surveys5 or time series observations following ﬂood events.12 To our knowledge, no studies have focused on the contribution of groundwater to surface water metal loads during a ﬂood. In this paper, we investigate the contribution of groundwater to dissolved Al, Fe, and Mn exports from a subtropical CASS catchment into an estuarine embayment during a ﬂood. We advance earlier investigations by (A) performing higher resolution, longer-term observations that captured contrasting hydrological conditions, and (B) linking groundwater discharge to surface water metal loads with concomitant radon, a natural groundwater tracer,13 measurements.

’ EXPERIMENTAL SECTION Experimental Site. Field observations were performed in the Tuckean Swamp in northern New South Wales, Australia (Figure 1). The swamp has been progressively cleared and drained to reduce periodic inundation, resulting in land suitable Received: July 25, 2011 Accepted: October 3, 2011 Revised: October 3, 2011 Published: October 03, 2011 9641

dx.doi.org/10.1021/es202581h | Environ. Sci. Technol. 2011, 45, 9641–9648

Environmental Science & Technology

ARTICLE

Figure 1. The distribution of selected variables in Tuckean Swamp waterways in June 2009. The numbers adjacent to selected samples show concentrations. Radon and pH values from ref 11a. The gray scale on the upper left map represents the distribution of soils in the Tuckean Swamp according to maps from the Department of Land and Water.35 The dark gray area around the Main Drain represents areas with CASS at the ground surface. The larger light gray area represents areas with CASS within 1 m of the ground surface. The white area represents CASS absent or below 1 m below the ground surface. Squares represent the location of groundwater samples.

for grazing and agriculture.14 Large areas developed into CASS after drains were constructed. The drains range from roadside table drains 0.5 to 1 m deep to main drains up to 5 m deep and 25 m wide10a and are thought to enhance groundwater discharge and the associated release of metals from CASS.7,12 The Swamp drains into the Tuckean Broadwater, a major tidal tributary of the Richmond River estuary. Average annual rainfall is about 1800 mm with ∼65% falling from December to April.15 Low pH waters can be observed year-round in the Tuckean Broadwater and are related to

groundwater inputs.11a The floodplain is a large back barrier lagoon infilled with Quaternary sediments mostly near 1 m AHD. Complex sequences of marine sands and pyritic estuarine clays are overlain by alluvial sediments. Experimental Approach. Our approach consisted of (A) sampling groundwaters to characterize the endmember concentrations entering surface waters; (B) conducting a spatial survey within the Tuckean Swamp drains and natural creeks to identify possible groundwater points of entry; and (C) performing long9642

dx.doi.org/10.1021/es202581h |Environ. Sci. Technol. 2011, 45, 9641–9648

Environmental Science & Technology

ARTICLE

Table 1. Average ((99% Conﬁdence Interval) pH, Dissolved Oxygen, and Metal Concentrations in Groundwater Samplesa n

a

pH

DO (mg L 1)

Al (mg L 1)

Fe (mg L 1)

Mn (mg L 1)

all samples

31

5.54 ( 0.64

1.89 ( 0.46

5.5 ( 5.1

70 ( 46

1.6 ( 1.0

non-CASS samples (pH > 5)

21

6.40 ( 0.45

1.83 ( 0.59

0.3 ( 0.3

24 ( 27

0.5 ( 0.5

CASS samples (pH < 5)

10

3.88 ( 0.48

2.06 ( 0.91

16.1 ( 12.4

132 ( 94

3.6 ( 1.7

Results for all groundwater samples are reported in Table S1.

term (i.e., 4 months) daily observations to assess the influence of recurrent floods on groundwater-derived metal exports from the catchment. A total of 31 groundwater samples were collected from shallow wells installed with a hand auger and monitoring wells installed by the Department of Environmental Protection in the 1990s. All wells were constructed with slotted PVC pipes with short screens (usually about 50 cm long) centered at the depths indicated in Table S1. In all cases, we used a peristaltic pump to retrieve samples after purging the wells. The spatial survey in surface waters was conducted in November 2009 about two weeks after a 200 mm rain event. The survey covered the Swamp drains and natural creeks as far upstream as possible. A metal sample was collected every 1 km or when noticeable changes in pH were observed. The time series observations in surface waters captured the drainage of the entire catchment and consisted of daily sampling at low tide in the Tuckean Broadwater immediately downstream of the Tuckean Swamp (Figure 1). Sampling started on 28 January 2010 during dry conditions, captured a ﬂood event, and stopped on 1 June 2010 when radon and pH returned to dry condition values. Analytical Methods. Dissolved element samples were immediately filtered with 0.45 μm disposable acetate filters. The filtered samples were acidified with high-purity HNO3 and stored in acid-cleaned vials. Samples were analyzed for dissolved metals using a Perkin-Elmer DRCe Inductively Coupled Plasma Mass Spectrometer (ICP-MS). Calibrations were performed before and after running the samples and concentrations covered at least 3 orders of magnitude. We accounted for instrument and background drift during analysis by following a standard sample standard bracketing scheme. System blanks were run by treating ultrapure DI water as a sample. The blank concentrations were at least 3 orders of magnitude lower than found in our samples. Water levels, pH, conductivity, and temperature were monitored at 1-h intervals using dataloggers maintained by the Richmond River County Council and calibrated biweekly. Radon (222Rn, t1/2 = 3.84 days) was determined using a continuous, automated radon-in-air monitor adapted for radon-in-water.16 Daily groundwater discharge rates into drains upstream of the surface water monitoring station were derived from a radon mass balance approach as described in detail in a companion paper.17

’ RESULTS AND DISCUSSION Metals in Groundwater. Dissolved metal and pH values in groundwater spanned over 3 orders of magnitude (Table S1). We define groundwater as any water within the saturated zone of geological material18 and group samples as CASS and non-CASS groundwaters (Table 1). The CASS samples were defined as samples with pH < 5, were collected from depths shallower than 2.5 m, and had extremely high metal concentrations and pH as low as 3.11. The non-CASS samples were collected from depths ranging between 0.5 and 33 m and had pH values as high as

Figure 2. Scatter plots between pH and depth and pH and metals in groundwater samples indicating shallow CASS groundwater as the main source of metals to surface waters.

7.61 but usually approaching neutral and metal concentrations 1 order of magnitude lower than the CASS samples. Radon and dissolved oxygen concentrations were very similar in the two groups of groundwater samples. The metals under investigations negatively correlated with pH (n = 31; p < 0.01; r2 > 0.41; Figure 2). These results imply that only shallow CASS groundwaters can be a concomitant source of radon, acid, and dissolved metals to surface waters. If deep non-CASS samples were a major source of groundwater to surface waters, we would not observe significant correlations between radon and trace metals in surface waters (shown later). The extremely high dissolved metal concentrations and low pH in shallow groundwaters from the Tuckean Swamp were consistent with high rates of weathering, iron sulfide mineral oxidation, production of acidity, and an associated release of trace elements from CASS.19 The observed groundwater concentrations were within the wide concentration range typically observed in other (sub-) tropical20 and boreal21 CASS catchments. Metal Distributions within the Tuckean Swamp. A spatial survey (Figure 1) revealed two locations where groundwater was entering surface waters: (1) Stibbards Creek, a natural creek surrounded by nonacidic sandy soils. This creek exhibited high radon concentrations, neutral pH, and relatively low trace metal concentrations. (2) Main Drain, a deepened and straightened drain surrounded by CASS. This area exhibited overall high radon and trace metal concentrations and pH as low as 4.3. The Main Drain captures the discharge from Meerchaum Vale, a groundwater-dominated acidified drain. The metal concentrations 9643

dx.doi.org/10.1021/es202581h |Environ. Sci. Technol. 2011, 45, 9641–9648

Environmental Science & Technology

Figure 3. Time series of trace metals and associate variables in surface waters of the Tuckean Broadwater. Radon data and groundwater ﬂuxes were originally reported in ref 17. Radon and pH values represent 24-h averages, while metal concentrations represent spot samples at low tide.

in this area were within the range observed in similar surveys performed in the 1990s,5,7 demonstrating that trace metal releases from CASS is a long-term problem. Interestingly, two samples collected in Hendersons Drain just upstream of the Main Drain had high Fe but relatively low radon, Mn and Al concentrations, and neutral pH, indicating a surface runoff-derived Fe source in this area. Samples collected from the Stony Island and Tucki Drains had low radon and metal concentrations and neutral pH, implying that these drains are simply a conduit for water from the upper catchment. The survey results demonstrate that not all groundwaters were a source of trace metals to surface waters. Surface water time series observations described below represent an integration of processes taking place within the swamp. Surface Water Time Series. The daily observations captured contrasting hydrological and chemical conditions (Figure 3). Following hydrological/biogeochemical conceptual models for nearby Australian tidal rivers and estuaries,22 we define four stages to describe trace metal distribution:

ARTICLE

1 Dry Period. During the ﬁrst month of observations, surface waters experienced “slug” ﬂow6 and the net water export during a complete tidal cycle approached zero. Radon concentrations and radon-derived groundwater discharge rates also approached zero. Overall, the water quality was reasonable with high pH and relatively low dissolved metal concentrations. With the onset of the ﬁrst rain early in February, Fe increased from nearly 0 to 0.5 mg/L, while Al and Mn did not show any change. 2 Flood. A sharp transition from extremely dry to wet conditions was observed in early March. A series of minor rain events in February preceded a strong precipitation (213 mm) event on 2 March. This rain inundated the swamp for approximately one week, increased surface ﬂows to about 120 m3 s 1, increased groundwater ﬂows to a maximum of 5 m3 s 1, and caused a well-deﬁned sequence of chemical changes. Immediately after the rain, during the rising limb of the hydrograph, pH dropped from >6 to about 4.5. During the falling limb of the hydrograph, radon concentrations sharply increased as a result of groundwater inputs into surface waters. While dissolved Al increased slightly during the ﬂood stage to a nearly constant value, Fe and Mn had more distinct changes. Both Fe and Mn peaked at 8.0 and 0.5 mg L 1 during the falling limb of the hydrograph 3 days after the peak surface discharge. The Fe and Mn peak coincided with the radon increase but not with the initial pH decrease. The ﬂood followed a long dry season, which probably allowed weathering products to accumulate in the soil proﬁle. 3 Postﬂood. The surface water chemistry during the postﬂood stage was clearly dominated by ﬂoodplain groundwater discharge. Radon concentrations remained very high (16 19 dpm L 1) for a week and slowly started to decrease. The radon peak coincided with pH dropping to about 4. The radon decrease toward the end of the postﬂood stage coincided with a pH increase and an overall decrease in metal concentrations. Fe was 1 order of magnitude lower, Al was 2-fold higher, and Mn was comparable during the postﬂood stage relative to the ﬂood stage. Radon-derived ﬂoodplain groundwater discharge rates reached 2.5 m3 s 1, equivalent to about 15% of the surface runoﬀ. 4 Minor Rains. The time series also captured two minor rain events after the postﬂood (37 mm on 20 April, and again 37 mm on 4 May 2010). A clear spike in radon and trace metal concentrations followed both rain events. We emphasize that the radon signal detected at the time series station most likely represents groundwater discharge from the CASS ﬂoodplain only. While the natural creeks and streams in the upper catchment (Alstonville Plateau) are fed by groundwater, waterfalls and turbulent ﬂow likely degas most of the radon from the upper catchment. The short residence times (about 1 day) and nonturbulent ﬂow in the ﬂoodplain drains prevents signiﬁcant radon degassing within the Tuckean Swamp.11a These observations demonstrate a suspected but poorly understood role played by groundwater and ﬂood events in dissolved metal exports from CASS wetlands. While previous work has not used radon as a groundwater tracer, the sequence of events described for the Tuckean Swamp seems similar to observations made for a CASS catchment in Europe.23 Under dry baseﬂow conditions, the runoﬀ from areas with CASS was low in comparison to that in non-CASS areas, resulting in small 9644

dx.doi.org/10.1021/es202581h |Environ. Sci. Technol. 2011, 45, 9641–9648

Environmental Science & Technology

ARTICLE

Figure 4. Scatter plots between dissolved metals in surface waters and possible controls. The lines on current plots show the hysteresis pattern observed during the ﬂood. The numbers near selected samples are days after the 213 mm rain event that triggered the ﬂood.

loads of trace elements. Rain events following extended dry periods increased both stream discharge and metal concentrations in boreal Finland,23 similar to our observations in warmer Australian conditions. Controls on Surface Water Metal Concentrations. Metal concentrations significantly correlated with radon and pH during the dry period, postflood, and minor rains stage (p < 0.01; Figure 4). While complex biogeochemical and physical processes can influence metal concentrations in surface waters and estuaries, groundwater is the only significant source of radon to surface waters. Other sources are negligible11a and cannot explain the radon concentrations higher than 2 dpm L 1 observed after the rain events in the Tuckean Broadwater. Both shallow and deep groundwaters have high radon concentrations. However, only shallow groundwaters are acidic and high in metals (Table S1). The significant correlations between radon and trace metals (Figure 4) in surface waters imply that their signal was derived mostly (if not all) from shallow groundwater discharge. The shallow soil profile in CASS catchments is often highly permeable due to interconnecting cracks and pores.24 Indeed, hydraulic conductivities in the Tuckean Swamp CASS (10 100 m d 1) are 1 order of magnitude higher than nearby soils,24 allowing groundwater to move quickly, uptake metals from CASS, and discharge to artificial drains. Our observations contrast to work in other systems where surface water discharge was considered the major control on

trace metal concentrations due to a diluting eﬀect.25 In the Tuckean Broadwater, metal concentrations may have been inﬂuenced by surface discharges only during the ﬂood. Inspection of scatter plots reveals Fe and Mn counter clockwise hysteresis during the ﬂood associated with a delayed release of metals from the catchment. Fe and Mn concentrations increased slowly during the rising limb of the hydrograph and reached a peak 8 to 10 days after the 213 mm rain event when groundwater discharge was the highest. Metal concentrations started to return to normal values after 11 days. To our knowledge, this is the ﬁrst observation of dissolved Fe and Mn hysteresis during a ﬂood. Previous studies have measured metal concentrations during rain events in CASS,12,26 but the sampling resolution may have not been enough to detect hysteresis. Total Surface Water versus Groundwater Fluxes. We estimate (1) total surface water metal fluxes at the surface water time series station and (2) groundwater-derived metal fluxes into surface waters upstream of the time series station (Table 2). The total surface water fluxes were estimated by simply multiplying measured dissolved concentrations in surface waters by surface water discharge. The groundwater metal fluxes are a component of the total surface water fluxes. Groundwater fluxes were estimated using the average metal concentrations in groundwater and the daily radon-derived groundwater discharge rates described in a companion paper.17 Large uncertainties may be associated with the groundwater fluxes. Similar to other investigations in coastal groundwaters,27 9645

dx.doi.org/10.1021/es202581h |Environ. Sci. Technol. 2011, 45, 9641–9648

Environmental Science & Technology

ARTICLE

Table 2. Average Total Surface Water and Groundwater Fluxes during the Four Diﬀerent Hydrological Stagesa n (days)

water ﬂux (m 3 s 1)

dry

30

4.35

12

129

22

ﬂood

16

89.24

1567

11164

1029

post ﬂood

27

13.98

951

393

320

minor rains

44

19.73

2868

709

487

dry

30

0.04

12

158

4

ﬂood

16

1.92

809

10315

242

post ﬂood minor rains

27 44

1.19 0.88

499 378

6366 4823

150 113

Al (kg d 1)

Fe (kg d 1)

Mn (kg d 1)

Total Surface Water

Groundwater

a The groundwater ﬂuxes were estimated by multiplying radon-derived groundwater discharge rates by the average metal concentration in all groundwater samples. The surface water ﬂuxes can be converted to aerial export rates using the CASS area in the Tuckean Swamp (4000 ha).

the main issue here is that metal concentrations in groundwaters are highly variable, complicating the definition of endmembers for flux estimates. We take average metal concentrations in all groundwaters as a conservative metal endmember, i.e., these averages provide minimum groundwater fluxes because the source of metals to surface waters is clearly shallow groundwater highly enriched in metals (Table 1). Using average metal concentrations in CASS samples only would triple Al and double Fe and Mn groundwater fluxes. In spite of uncertainties, Al and Mn groundwater fluxes were usually comparable to the total surface water fluxes (Table 2). These observations give confidence in our estimates, support the suggestion of shallow groundwater as the main pathway releasing metals from CASS, and indicate that most of the groundwater-derived Al and Mn were exported to the Richmond River estuary. In contrast to Mn and Al, groundwater-derived Fe ﬂuxes were at least 1 order of magnitude higher than total surface water ﬂuxes during the postﬂood and minor rains stages. This substantial discrepancy indicates that Fe was sequestered within the drainage system prior to discharge at the time series monitoring station. This observation is consistent with known diagenetic pathways for the formation and accumulation of Fe-rich monosulﬁdic black ooze (MBO) in the basal sediments of CASS drains.28 As MBO are organic oozes highly enriched in reactive iron mineral phases, their resuspension followed by oxidation during ﬂoods could be a source of dissolved Fe enrichment during ﬂoods.29 However, the observed delayed Fe input (i.e., counterclockwise hysteresis) implies that groundwater was the major source of Fe. If MBO resuspension was the major source of Fe, the dissolved Fe enrichment should have been observed in the earlier stages of the ﬂood (i.e., clockwise hysteresis). The highest metal concentrations (about 5 mg L 1) observed for 3 days after the surface water discharge peak may be related to an initial wash out of shallow groundwater that had accumulated exchangeable Fe during the dry season. Alternatively, reductive processes in surface waters after the ﬂood could also increase Fe and Mn concentrations. The ﬂood stage lasted for 16 of the 117 days of uninterrupted observations (Table 2). Dissolved Fe, Mn, and Al loads during those 16 days (14% of the time investigated) were respectively 80, 35, and 14% of the total exports from the Tuckean Swamp. These observations underscore the importance of ﬂoods in metal exports from coastal catchments and the need to include these events in monitoring programs. Floods can be especially important

in tropical and subtropical estuaries which are often characterized by a higher degree of variability in water ﬂows than the better studied temperate systems.30 Since ﬂoods are essentially an “above surface” process, the observed delayed groundwater inputs were previously overlooked during these events. While a number of previous investigations have assessed metal concentrations and loads in surface waters aﬀected by CASS, our link to a groundwater tracer and the high sampling resolution (i.e., daily for 4 months) are unprecedented. Previous work focused on hourly sampling for a few days12 or seasonal sampling for several years.31 For example, a 65-h monitoring of a rain event at a CASS catchment near the Tuckean Swamp demonstrated high and variable trace metal concentrations.12 Previous assessments of groundwater ﬂuxes to coastal waters in acidic environments relied on the assumption that the annual inﬁltration equals groundwater discharge.21 Our link to high resolution radon measurements revealed exactly when shallow groundwater was ﬂushed from CASS, and that this groundwater was the major source of trace metals to surface waters. A main advantage of using radon as a natural tracer is that radon integrates spatially heterogeneous groundwater pathways that may be diﬃcult to quantify when using conventional hydrological approaches. Global Perspectives. The concentrations of Fe in Tuckean Swamp surface water samples (average ∼1 mg L 1) were orders of magnitude higher than the average metal concentrations in the world’s rivers of <40 μg L 1.32 However, the extremely high dissolved metal concentrations and low pH observed in the Tuckean Swamp are consistent with previous work in this5,7,29 and other CASS catchments in Australia,12 Europe,19b,26 and Asia.33 While acid sulfate soils cover only about 0.1% of the global continental area,4 they can play a major but still unknown role on the global Fe cycle. We use back of the envelope calculations to illustrate the potential role of CASS on the global Fe cycle. Multiplying the daily surface water Fe ﬂuxes shown in Table 2 by the number of days of each hydrological stage yielded total dissolved Fe exports from the Tuckean Swamp of 222,300 kg in 114 days of continuous observations (or 699,800 kg yr 1). As the CASS area in the Tuckean Swamp is 4000 ha, we obtain an aerial dissolved Fe ﬂux of 175 kg yr 1 ha 1. Assuming the Tuckean Swamp ﬂuxes can be upscaled to all Acid Sulfate Soil areas worldwide (total of about 17 million ha4), these soils can potentially release about 3 Tg yr 1 of dissolved Fe to creeks and rivers. This estimate is on the same order of magnitude of the global dissolved Fe input from rivers to estuaries.32,34 The extremely high dissolved 9646

dx.doi.org/10.1021/es202581h |Environ. Sci. Technol. 2011, 45, 9641–9648

Environmental Science & Technology metal ﬂuxes are clearly related to high CASS weathering rates, acidic groundwater conditions, and eﬀective groundwater ﬂushing from highly permeable, drained CASS. While this rough estimate is subject to large uncertainties, it highlights the need for additional work on groundwater-derived metal exports from CASS catchments.

’ ASSOCIATED CONTENT

bS

Supporting Information. Table containing the results of groundwater observations. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: +61 2 66203494. Fax: +61 2 66212669. E-mail: isaac. [email protected].

’ ACKNOWLEDGMENT This project was supported by a Hermon Slade Foundation grant (09-01). Santos and Eyre acknowledge the support of ARC grants (LP100200731, LP110200975, and DP110103638). We thank Bruce Heynatz for taking metal samples daily. Ben Shepherd and Douglas Tait assisted with ﬁeld work, the Richmond River County Council supplied water quality information, and the NSW Oﬃce of Water allowed access to monitoring wells. Scott Johnston and three anonymous reviewers have provided constructive comments. ’ REFERENCES (1) Windom, H.; Schropp, S. J.; Calder, F. D.; Ryan, J. D.; Smith, R. G.; Burney, l. C.; Lewis, F. G.; Rawlinson, C. H. Natural trace metal concentrations in estuarine and coastal marine sediments of the southeastern United States. Environ. Sci. Technol. 1989, 23 (3), 314–324. (2) (a) Bone, S.; Charette, M. A.; Lamborg, C. H.; Gonneea, M. E. Has submarine groundwater discharge been overlooked as a source of mercury to coastal waters? Environ Sci. Technol. 2007, 41 (9), 3090–3095. (b) Beck, A. J.; Cochran, J. K.; Sa~nudo-Wilhelmy, S. A. The distribution and speciation of dissolved trace metals in a shallow subterranean estuary. Mar. Chem. 2010, 121 (1 4), 145–156. (c) Santos, I. R.; Burnett, W. C.; Misra, S.; Suryaputra, I. G. N. A.; Chanton, J. P.; Dittmar, T.; Peterson, R. N.; Swarzenski, P. W. Uranium and barium cycling in a salt wedge subterranean estuary: The inﬂuence of tidal pumping. Chem. Geol. 2011, 287 (1 2), 114–123. (3) Bowen, J. L.; Kroeger, K. D.; Tomasky, G.; Pabich, W. J.; Cole, M. L.; Carmichael, R. H.; Valiela, I. A review of land-sea coupling by groundwater discharge of nitrogen to New England estuaries: Mechanisms and eﬀects. Appl. Geochem. 2007, 22 (1), 175–191. (4) Ljung, K.; Maley, F.; Cook, A.; Weinstein, P. Acid sulfate soils and human health - A Millennium Ecosystem Assessment. Environ. Int. 2009, 35 (8), 1234–1242. (5) Sammut, J.; White, I.; Melville, M. D. Acidiﬁcation of an estuarine tributary in eastern Australia due to drainage of acid sulfate soils. Mar. Freshwater Res. 1996, 47 (5), 669–684. (6) Johnston, S. G.; Slavich, P. G.; Hirst, P. The impact of controlled tidal exchange on drainage water quality in acid sulphate soil backswamps. Agric. Water Manage. 2005, 73 (2), 87–111. (7) Ferguson, A.; Eyre, B. Behaviour of aluminium and iron in acid runoﬀ from acid sulphate soils in the lower Richmond River catchment. J. Aust. Geol. Geophys. 1999, 17 (5/6), 193–201. € (8) Nordmyr, L.; Osterholm, P.; Astr€om, M. Estuarine behaviour of metal loads leached from coastal lowland acid sulphate soils. Mar. Environ. Res. 2008, 66 (3), 378–393.

ARTICLE

(9) Preda, M.; Cox, M. Trace metals in acid sediments and water, Pimpama catchment southeast Queensland, Australia. Environ. Geol. 2001, 40, 755–768. (10) (a) Eyre, B. D.; Kerr, G.; Sullivan, L. A. Deoxygenation potential of the Richmond River Estuary ﬂoodplain, northern NSW, Australia. River Res. Applications 2006, 22 (9), 981–992. (b) Wong, V. N. L.; Johnston, S. G.; Bush, R. T.; Sullivan, L. A.; Clay, C.; Burton, E. D.; Slavich, P. G. Spatial and temporal changes in estuarine water quality during a post-ﬂood hypoxic event. Estuarine, Coastal Shelf Sci. 2010, 87 (1), 73–82. (c) Johnston, S. G.; Slavich, P. G.; Sullivan, L. A.; Hirst, P. Artiﬁcial drainage of ﬂoodwaters from sulﬁdic backswamps: eﬀects on deoxygenation in an Australian estuary. Mar. Freshwater Res. 2003, 54 (6), 781–795. (11) (a) Santos, I. R.; Eyre, B. D. Radon tracing of groundwater discharge into an Australian estuary surrounded by coastal acid sulphate soils. J. Hydrol. 2011, 396 (3 4), 246–257. (b) Johnston, S. G.; Slavich, P.; Hirst, P. The acid ﬂux dynamics of two artiﬁcial drains in acid sulfate soil backswamps on the Clarence River ﬂoodplain, Australia. Aust. J. Soil Res. 2004, 42, 623–637. (12) Macdonald, B. C. T.; White, I.; Astrom, M. E.; Keene, A. F.; Melville, M. D.; ReynoldS, J. K. Discharge of weathering products from acid sulfate soils after a rainfall event, Tweed River, eastern Australia. Appl. Geochem. 2007, 22 (12), 2695–2705. (13) 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 (2), 165–178. (14) Taﬀs, K. H.; Farago, L. J.; Heijnis, H.; Jacobsen, G. A diatombased Holocene record of human impact from a coastal environment: Tuckean Swamp, eastern Australia. J. Paleolimnol. 2008, 39 (1), 71–82. (15) Eyre, B. D.; Pont, D. Intra- and inter-annual variability in the diﬀerent forms of diﬀuse nitrogen and phosphorus delivered to seven subtropical east Australian estuaries. Estuarine, Coastal Shelf Sci. 2003, 56, 1–13. (16) 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 (3 4), 298–304. (17) 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 (8), 3310–3316. (18) Burnett, W. C.; Aggarwal, P. K.; Aureli, A.; Bokuniewicz, H.; Cable, J. E.; Charette, M. A.; Kontar, E.; Krupa, S.; Kulkarni, K. M.; Loveless, A.; Moore, W. S.; Oberdorfer, J. A.; Oliveira, J.; Ozyurt, I. N.; Povinec, P.; Privitera, A. M. G.; Rajar, R.; Ramessur, R. T.; Schollten, J.; Stieglitz, T.; Taniguchi, M.; Turner, J. V. Quantifying submarine groundwater discharge in the coastal zone via multiple methods. Sci. Total Environ. 2006, 367 (2 3), 498–543. (19) (a) Appleyard, S.; Wong, S.; Willis-Jones, B.; Angeloni, J.; Watkins, R. Groundwater acidiﬁcation caused by urban development in Perth, Western Australia: source, distribution, and implications for management. Aust. J. Soil Res. 2004, 42, 579–585. (b) Roos, M.; Astrom, M. Hydrochemistry of rivers in an acid sulphate soil hotspot area in western Finland. Agric. Food Sci. 2005, 14 (1), 24–33. (c) Åstr€om, M. Eﬀect of widespread severely acidic soils on spatial features and abundance of trace elements in streams. J. Geochem. Exploration 2001, 73 (3), 181–191. (20) Hinwood, A. L.; Horwitz, P.; Appleyard, S.; Barton, C.; Wajrak, M. Acid sulphate soil disturbance and metals in groundwater: Implications for human exposure through home grown produce. Environ. Pollut. 2006, 143 (1), 100–105. (21) Lavergren, U.; Astr€om, M. E.; Falk, H.; Bergb€ack, B. Metal dispersion in groundwater in an area with natural and processed black shale - Nationwide perspective and comparison with acid sulfate soils. Appl. Geochem. 2009, 24 (3), 359–369. (22) Eyre, B. D.; Ferguson, A. J. P. Impact of a ﬂood event on benthic and pelagic coupling in a sub-tropical east Australian Estuary (Brunswick). Estuarine, Coastal Shelf Sci, 2006, 66 (1 2), 111–122. (23) Åstr€om, M. The eﬀect of acid soil leaching on trace element abundance in a medium-sized stream, W. Finland. Appl. Geochem. 2001, 16 (3), 387–396. 9647

dx.doi.org/10.1021/es202581h |Environ. Sci. Technol. 2011, 45, 9641–9648

Environmental Science & Technology

ARTICLE

(24) Johnston, S.; Hirst, P.; Slavich, P.; Bush, R. T.; Aaso, T. Saturated hydraulic conductivity of sulfuric horizons in coastal ﬂoodplain acid sulfate soils: Variability and implications. Geoderma 2009, 151, 387–394. (25) (a) Canovas, C. R.; Hubbard, C. G.; Olías, M.; Nieto, J. M.; Black, S.; Coleman, M. L. Hydrochemical variations and contaminant load in the Río Tinto (Spain) during ﬂood events. J. Hydrol. 2008, 350 (1 2), 25–40. (b) Keith, D. C.; Runnells, D. D.; Esposito, K. J.; Chermak, J. A.; Levy, D. B.; Hannula, S. R.; Watts, M.; Hall, L. Geochemical models of the impact of acidic groundwater and evaporative sulfate salts on Boulder Creek at Iron Mountain, California. Appl. Geochem. 2001, 16 (7 8), 947–961. (26) Astrom, M.; Spiro, B. Sources of acidity andmetals in a stream draining acid sulphate soil, till and peat, western Finland, revealed by a hydrochemical and sulphur isotope study. Agric. Food Sci. 2005, 14, 34–43. (27) (a) Charette, M. A.; Sholkovitz, E. R. Trace element cycling in a subterranean estuary: Part 2. Geochemistry of the pore water. Geochim. Cosmochim. Acta 2006, 70, 811–826. (b) Swarzenski, P. W.; Baskaran, M. Uranium distribution in the coastal waters and pore waters of Tampa Bay, Florida. Mar. Chem. 2006, 102 (3 4), 252–266. (c) 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 (1 4), 37–48. (28) Burton, E. D.; Bush, R. T.; Sullivan, L. A. Sedimentary iron geochemistry in acidic waterways associated with coastal lowland acid sulfate soils. Geochim. Cosmochim. Acta 2006, 70 (22), 5455–5468. (29) Bush, R. T.; Fyfe, D.; Sullivan, L. A. Occurrence and abundance of monosulﬁdic black ooze in coastal acid sulfate soil landscapes. Aust. J. Soil Res. 2004, 42, 609–616. (30) (a) Eyre, B. D. Regional evaluation of nutrient transformation and phytoplankton growth in nine river-dominated sub-tropical east Australian estuaries. Mar. Ecol.: Prog. Ser. 2000, 205, 61–83. (b) McKee, L.; Eyre, B. D.; Hossain, S. Intra- and inter-annual export of nitrogen and phosphorus in the sub-tropical Richmond River catchment, Australia. Hydrol. Processes 2000, 14, 1787–1809. € (31) Osterholm, P.; Astr€om, M. Meteorological impacts on the water quality in the Pajuluoma acid sulphate area, W. Finland. Appl. Geochem. 2008, 23 (6), 1594–1606. (32) Haese, R., The biogeochemistry of iron. In Marine Geochemistry; Schulz, H. D., Zabel, M., Eds.; Springer: Berlin, Heidelberg, 2006; pp 241 270. (33) Nguyen, T. T.; Wilander, A. Chemical conditions in acidic waters in the plain of reeds, Vietnam. Water Resour. 1995, 29 (5), 1401– 1408. (34) Raiswell, R. Towards a global highly reactive iron cycle. J. Geochem. Exploration 2006, 88 (1 3), 436–439. (35) Naylor, S. D.; Chapman, G. A.; Atkinson, G.; Murphy, C. L.; Tulau, M. J.; Flewin, T. C.; Milford, H. B.; Morand, D. T. Guidelines for the Use of Acid Sulfate Soil Risk Maps; Department of Land and Water Conservation: Sydney, 1998.

9648

dx.doi.org/10.1021/es202581h |Environ. Sci. Technol. 2011, 45, 9641–9648

ARTICLE pubs.acs.org/est

The Atmospheric Photolysis of o-Tolualdehyde Grainne M. Cliﬀord,† Aurelie Hadj-Aïssa,† Robert M. Healy,† Abdelwahid Mellouki,‡ Amalia Mu~noz,§ Klaus Wirtz,§ Montserrat Martín Reviejo,§ Esther Borras,§ and John C. Wenger†,* †

Department of Chemistry and Environmental Research Institute, University College Cork, Cork, Ireland CNRS ICARE, F-45071 Orleans 2, France § Instituto Universitario UMH-CEAM, C/Charles R. Darwin 14, Parque Tecnologico, 46980 Paterna, Valencia, Spain ‡

bS Supporting Information ABSTRACT: The photolysis of o-tolualdehyde by natural sunlight has been investigated at the large outdoor European Photoreactor (EUPHORE) in Valencia, Spain. The photolysis rate coeﬃcient was measured directly under diﬀerent solar ﬂux levels, with values in the range j(o-tolualdehyde) = (1.622.15) 104 s1 observed, yielding an average value of j(o-tolualdehyde)/j(NO2) = (2.53 ( 0.25) 102. The estimated photolysis lifetime is 12 h, conﬁrming that direct photolysis by sunlight is the major atmospheric degradation pathway for o-tolualdehyde. Published UV absorption cross-section data were used to derive an eﬀective quantum yield (290400 nm) close to unity, within experimental error. Possible reaction pathways for the formation of the major photolysis products, benzocyclobutenol (tentatively identiﬁed) and o-phthalaldehyde, are proposed. Appreciable yields (513%) of secondary organic aerosol (SOA) were observed at EUPHORE and also during supplementary experiments performed in an indoor chamber using an artiﬁcial light source. Oﬀ-line analysis by gas chromatographymass spectrometry allowed identiﬁcation of o-phthalaldehyde, phthalide, phthalic anhydride, o-toluic acid, and phthalaldehydic acid in the particle phase.

’ INTRODUCTION o-Tolualdehyde is an aromatic aldehyde regularly detected in ambient air.13 It is emitted into the atmosphere as a primary pollutant from automobile exhausts 4,5 and can also be formed in situ from the hydroxyl radical (OH) initiated oxidation of o-xylene.3,6,7 The subsequent degradation of o-tolualdehyde in the atmosphere may contribute to the formation of secondary species such as ozone, nitrates, and organic aerosol, which are major constituents of polluted air in the troposphere.8 A detailed knowledge of the kinetics and mechanisms of these atmospheric degradation processes is therefore required to fully understand the environmental impact of o-tolualdehyde and its parent compound, o-xylene. The potential gas-phase removal processes for o-tolualdehyde are reaction with OH, the nitrate radical (NO3), ozone, chlorine atoms, and direct photolysis by sunlight.8 Laboratory kinetic studies indicate that the reactions with ozone and Cl atoms are of negligible importance and that reaction with NO3 is a minor pathway compared to OH-initiated oxidation.9,10 The gas-phase absorption cross section of o-tolualdehyde is relatively strong in the actinic region, indicating that photolysis may also be an important loss process if the quantum yield is close to unity.11 Some preliminary studies of the gas-phase photolysis of o-tolualdehyde have been performed at the outdoor European r 2011 American Chemical Society

Photoreactor (EUPHORE) in Valencia, Spain;12,13 however, the reported value for the photolysis rate coeﬃcient is somewhat uncertain and no mechanistic information was obtained. In this work a more detailed investigation of the photolysis of otolualdehyde by natural sunlight has been performed at EUPHORE. The photolysis rate coeﬃcient has been determined, reaction products identiﬁed and secondary organic aerosol formation observed. Supplementary experiments have also been performed in an indoor simulation chamber using an artiﬁcial light source. The results provide new information on the atmospheric degradation of o-tolualdehyde and its potential impact on the environment.

’ EXPERIMENTAL SECTION EUPHORE Chamber. The sunlight photolysis of o-tolualdehyde was investigated in Chamber B at the EUPHORE facility which consists of a 204.5 m3 hemispherical reactor made of FEP Teflon foil. Technical details concerning the facility and its Received: August 1, 2011 Accepted: October 18, 2011 Revised: October 17, 2011 Published: October 18, 2011 9649

dx.doi.org/10.1021/es2026533 | Environ. Sci. Technol. 2011, 45, 9649–9657

Environmental Science & Technology application for photolysis experiments have been previously reported in the literature.1418 The chamber was cleaned by ﬂushing with puriﬁed air overnight and ﬁlled to atmospheric pressure. o-Tolualdehyde was introduced into the chamber and allowed to mix for approximately one hour while its loss to the chamber walls was measured. Photolysis was initiated by opening the protective housing to expose the contents of the chamber to sunlight for 23 h. The temperature inside the chamber increased slightly as the experiments progressed but was always within the range 296308 K. The solar actinic ﬂux over the range 290520 nm was measured using a calibrated spectroradiometer (Bentham DM300). A full spectral scan took 420430 s. Chemical analysis was performed by in situ FTIR spectroscopy and gas chromatography (GC). The FTIR spectrometer (Nicolet Magna 550) was operated at 1 cm1 resolution and spectra over the range 6004000 cm1 were derived from the coaddition of 270 scans collected over 5 min. The GC (Fisons 8160), equipped with ﬂame ionization and photoionization detectors (FID and PID), was operated using a 30 m DB-624 fused silica capillary column (J&W Scientiﬁc, 0.32 mm i.d., 1.8 μm ﬁlm). Air was sampled from the chamber into a 3 cm3 sampling loop and injected onto the column operated at 150 °C. The reactants and products were quantiﬁed using calibrated FTIR spectra and GC sensitivity factors obtained by introducing known volumes of pure materials into the chamber. The leak rate was determined by adding about 20 ppbV of SF6 to the chamber and measuring its loss by FTIR spectroscopy. Additional chemical analysis was aﬀorded by gas chromatographymass spectrometry (GCMS) using a Varian GC 3400 interfaced to a Saturn 2000 ion trap mass spectrometer. The GCMS incorporated a cryogenic sample preconcentration trap (SPT) containing glass beads cooled to 160 °C. The SPT was operated for 5 min at a ﬂow of 40 cm3 min1. Desorption was performed at 270 °C onto a 30 m HP Innowax column (0.25 mm id, 0.25 μm ﬁlm thickness). The column was held at 20 °C for 7 min and increased to 250 at 10 °C min1. The GCMS was operated in electron ionization (EI) mode over the m/z range 46250. Products were identiﬁed by comparison with chromatographic retention times and mass spectra of authentic standards. A scanning mobility particle sizer (SMPS), comprising of a condensation particle counter (TSI 3022A) and diﬀerential mobility analyzer (TSI 3081), was used to measure particle size distribution, number and volume concentrations with a time resolution of ca. 5 min. Aerosol mass concentrations were measured using a tapered element oscillating microbalance (TEOM, Rupprecht and Patashnick 1400a) ﬁtted with a PM1 inlet and operated at a total ﬂow of 16.7 L min1. For stable operation of the TEOM system, the sampling line and the sensor unit were held at 27 °C. The cabinet temperature of the SMPS was 25 °C. Both the TEOM and SMPS systems were started prior to the introduction of reactants to conﬁrm that no particles were present in the chamber before photolysis was initiated. The instruments were also operated for several hours after the chamber was closed to measure the loss of particles to the chamber walls. Oﬀ-line chemical analysis of the particles was performed by GCMS using a method reported previously.19 Particles were collected onto a 47 mm quartz ﬁber ﬁlter at a ﬂow rate of 80 L min1 for 1 h and subsequently extracted under sonication in 5 mL of a CH2Cl2/CH3CN (1:1) mixture. The extract was derivatized with O-(2,3,4,5,6-

ARTICLE

pentaﬂuorobenzyl)-hydroxylamine (PFBHA) and N-methylN-trimethylsilyltriﬂuoro-acetamide (MSTFA) to react with carbonyls and hydroxyl-containing compounds respectively. One μL was injected into the GCMS (TRACE-DSQ II, Thermo Fisher Scientiﬁc Co., Waltham, MA) ﬁtted with an RTX-5MS column (30 m 0.25 mm i.d. 0.25 μm ﬁlm thickness, Thermo Fisher Scientiﬁc Co). Indoor Chamber. A number of experiments were also conducted in the 3.91 m3 FEP Teflon indoor chamber at University College Cork, described in detail elsewhere.20 The chamber was surrounded by 18 lamps (Philips TL05, 40 W; 320 nm < λ > 480 nm; λmax = 360 nm) which deliver a light intensity equivalent to j(NO2) = 1.56 103 s1. Photolysis of o-tolualdehyde was initiated by turning on the lamps for 23 h. Quantitative analysis of gas-phase species was performed by in situ FTIR spectroscopy using the procedures outlined above. A SMPS, consisting of a condensation particle counter (TSI 3010) and a differential mobility analyzer (TSI 3080), was used to measure particle properties at ca. 5 min intervals. Additional chemical analysis was provided by GCMS. The contents of the chamber were sampled using a 1 cm3 gastight syringe (Hamilton) and injected directly into the GCMS instrument (Varian GC 3800 interfaced to a MS Saturn 2000). Typically five samples were taken per experiment. The GCMS was operated in electron ionization (EI) mode over the m/z range 46250. Chromatographic separation was achieved using a CP-Sil8 fused silica capillary column (Varian, 30 m, 0.25 mm i.d, 0.25 μm film thickness) operated at 60 °C for 1 min and then increased to 290 °C at a rate of 10 °C per minute. A flow rate of 5 cm3 min1 was used and the injector temperature was 250 °C. Products were identified and quantified by comparison with chromatographic retention times and mass spectra of authentic standards. It should be noted however, that reliable quantitative information could not be obtained for all products due to sampling losses associated with use of the syringe. Further off-line detection of carbonyl reaction products in the gas and particle phase was performed using a denuder-filter sampling method coupled with PFBHA derivatization and GCMS analysis.21,22 Materials. All organic compounds were obtained from Aldrich Chemical Co. at the highest purity available (stated purities >97%) and used without further purification. Sulfur hexafluoride (99.9%) was obtained from Messer Griesheim, Germany.

’ RESULTS AND DISCUSSION Photolysis Rate Coefficient. Photolysis of o-tolualdehyde follows first order kinetics: j

o-tolualdehyde þ hν f products

ðIÞ

where j is the photolysis rate coefficient. Assuming photolysis is the only loss process, j can be determined from a simple first order kinetic plot: ln½o-tolualdehydet =½o-tolualdehyde0 ¼ jt

ðIIÞ

where the subscripts 0 and t refer to the concentrations at initial time 0 and elapsed time t, respectively. Test experiments carried out in the presence and absence of excess amounts (1025 ppmV) of OH radical scavenger (isopropanol and cyclohexane in the indoor and outdoor chambers, respectively) produced very similar decay rates, indicating that loss of o-tolualdehyde due to reaction with OH in the chambers was negligible. However, 9650

dx.doi.org/10.1021/es2026533 |Environ. Sci. Technol. 2011, 45, 9649–9657

Environmental Science & Technology

ARTICLE

Figure 1. Concentrationtime proﬁle and j(NO2) during the photolysis of o-tolualdehyde at EUPHORE on 3 July 2003. The vertical dotted lines indicate the time the chamber was opened (09:44) and closed (12:27).

Table 1. Experimental Details for the Photolysis of o-Tolualdehyde in the Outdoor (EUPHORE) and Indoor Simulation Chambersa EUPHORE 19 September 2002

indoor chamber

3 July 2003

4 July 2003

four experiments

initial concentration (ppbV)

525

508

256

364, 518, 712, 1354

irradiation period j(NO2)average (s1)

11.59 14:24 (6.41 ( 0.64) 103

09:4412:27 (8.13 ( 0.81) 103

11:3613:53 (8.44 ( 0.84) 103

13 h 1.56 103

kwall (s1)

(7.80 ( 0.47) 106

(1.03 ( 0.47) 105

(1.68 ( 0.50) 105

(7.03 ( 0.90) 106

6

(5.48 ( 0.90) 10

6

4

1

kSF6 (s )

(6.57 ( 0.72) 10

j(o-tolualdehyde)FTIR (s1)

(1.97 ( 0.05) 104

1

4

4

(4.16 ( 0.23) 105

j(o-tolualdehyde)GC‑PID (s )

(1.62 ( 0.05) 10

(2.12 ( 0.05) 10

(2.15 ( 0.06) 10

j(o-tolualdehyde)average (s1)

(1.62 ( 0.05) 104

(2.05 ( 0.05) 104

(2.15 ( 0.06) 104

(4.16 ( 0.23) 105

(2.53 ( 0.25) 10

2

2

(2.55 ( 0.26) 10

2

(2.66 ( 0.23) 102

(1.33 ( 0.13) 10 1.22 ( 0.03

4

(1.90 ( 0.19) 10 1.13 ( 0.05

4

j(o-tolualdehyde)/j(NO2) 1

maximum theoretical loss rate (s ) eﬀective quantum yield

(2.52 ( 0.25) 10

4

(1.85 ( 0.19) 10 1.11 ( 0.03

molar yield of benzocyclobutenol

0.77 ( 0.04

0.71 ( 0.06

molar yield of o-phthalaldehyde

0.21 ( 0.02

0.22 ( 0.02

yield of aerosol

0.104

0.049, 0.059, 0.080, 0.133

Except for j(NO2), quoted errors are twice the standard deviation arising from the least squares ﬁt of the data and include the uncertainty in calibration and response factors. For j(NO2) and the maximum theoretical loss rate, the estimated error is 10%. The molar yield of benzocyclobutenol is based on the use of 1-indanol as a surrogate compound. a

o-tolualdehyde was found to undergo a small amount of deposition to the walls of the reactor. The rate coefficient for this process (kwall) was determined by measuring the first order decay of the compound in the dark for about one hour prior to photolysis. This value was incorporated into the overall decay as follows: ln½o-tolualdehydet =½o-tolualdehyde0 kwall t ¼ jt

ðIIIÞ

Although dilution was also observed during the EUPHORE experiments, the rate determined by measuring the loss of SF6 from the chamber, kSF6, was lower than the wall loss and is therefore already incorporated into kwall. Thus a plot in the form of eq III should yield a straight line with gradient j.

Concentrationtime proﬁles and kinetic plots in the form of eq III were generated for all experiments. The concentration time proﬁle for the EUPHORE experiment conducted on third July is presented in Figure 1 and clearly shows the rapid decay of o-tolualdehyde following exposure of the chamber to sunlight. The light intensity is represented by the photolysis rate coeﬃcient for NO2, j(NO2), which was calculated from the solar ﬂux measurements of the spectroradiometer and recommended values for the absorption cross-section and quantum yield.23 The corresponding kinetic plot used to calculate j from these FTIR spectroscopic measurements is shown in Figure S1 (Supporting Information) and exhibits good linearity and a near-zero intercept. The value for j(o-tolualdehyde) is listed in Table 1 along 9651

dx.doi.org/10.1021/es2026533 |Environ. Sci. Technol. 2011, 45, 9649–9657

Environmental Science & Technology

ARTICLE

Figure 2. Yield plot of the products detected by FTIR spectroscopy during the photolysis of o-tolualdehyde at EUPHORE on 3 July 2003.

with a summary of the reaction conditions and results obtained for all of the photolysis experiments. As shown in Table 1, there is very good agreement between the j values determined using FTIR spectroscopy and GC-PID in the experiment performed on third July 2003. In the experiment performed on fourth July 2003, o-tolualdehyde was among several aromatic aldehydes subjected to photolysis in the chamber and FTIR spectroscopy could not be used for analysis due to overlapping absorption bands. The value for j(o-tolualdehyde) obtained by GC-PID in this experiment is very similar to that obtained on third July 2003, indicating that the presence of the other compounds (2,3-dimethylbenzaldehyde and 2,6-dimethylbenzaldehyde) did not aﬀect the photolysis rate. The average values for j(NO2) and hence j(o-tolualdehyde)/j(NO2) in these two EUPHORE experiments were also very similar, reﬂecting the fact that they were performed under almost identical, cloud-free conditions. As expected, the sunlight intensity was somewhat lower for the experiment performed on 19th September 2002 and although a slightly lower value was observed for j(o-tolualdehyde), the j(o-tolualdehyde)/j(NO2) ratio was virtually the same. This indicates that the average value of j(o-tolualdehyde)/j(NO2) = (2.53 ( 0.25) 102 is a useful parameter for calculating the photolysis rate of o-tolualdehyde under diﬀerent light conditions in chambers or in the real atmosphere. The eﬀective quantum yield for photolysis of o-tolualdehyde, jeﬀ was determined using the following expression; jeff ¼ jexp =jmax

ðIVÞ

where jexp and jmax are the experimentally observed and maximum theoretical values of the photolysis rate coeﬃcient respectively. The latter term was calculated using the solar ﬂux intensity measurements of the spectroradiometer, the absorption cross section data reported by Thiault et al.,11 and assuming a quantum yield of unity over the atmospheric absorption range of the compound. The values obtained for jeﬀ during the EUPHORE experiments are in reasonable agreement, yielding an average of jeﬀ = (1.15 ( 0.05). However, the calculated value of jmax does not include the uncertainty in the reported absorption cross

sections, which is estimated to be 15% below 340 nm and 20% in the range 340363 nm.11 The results therefore suggest that, within experimental error, the eﬀective quantum yield for photolysis of o-tolualdehyde by natural sunlight is unity. The results of this work can be compared to those obtained in the preliminary studies also performed at EUPHORE. A value of j(o-tolualdehyde) = (2.00 ( 0.10) 104 s1 was obtained by Volkamer et al.12 in one experiment during February, where the solar zenith angle was 50° and the UV ﬂux reduced by around a factor of 2 compared to midsummer. In contrast, Thiault et al.13 obtained a value of j(o-tolualdehyde) = (1.10 ( 0.20) 104 s1 during an experiment performed in April. Values of j(o-tolualdehyde)/j(NO2) = 0.032 and jeﬀ = 0.6 were also reported, but no information was provided on whether the wall loss of o-tolualdehyde was taken into account during data analysis. It is interesting note that in these preliminary studies,12,13 the photolysis of benzaldehyde, m- and p-tolualdehyde was found to be negligible, suggesting that the presence of the methyl group in the ortho position is a key factor in determining the photolysis eﬃciency of o-tolualdehyde. The photolysis rate of o-tolualdehyde in the indoor chamber was around a factor of 5 slower than in the outdoor chamber. This result was expected since the intensity and wavelength distribution of UV light produced from the TL05 lamps is quite diﬀerent from natural sunlight. However, a useful comparison between the two chambers can be made by examining the values determined for j(o-tolualdehyde)/j(NO2), shown in Table 1. The values are in very close agreement, indicating that the TL05 lamps used in the indoor chamber studies provide reasonably realistic light conditions for studies of the atmospheric photolysis of the aromatic aldehydes. Photolysis Products. Gas-phase products arising from the photolysis of o-tolualdehyde were determined by FTIR spectroscopy and GCMS. The FTIR spectra obtained in both chambers enabled identification and quantification of o-phthalaldehyde and carbon monoxide as reaction products. Phthalide and formic acid were also detected during the later stages of the experiments, but below the limits of quantification. The FTIR product spectra also contain significant absorption features around 1050 cm1 9652

dx.doi.org/10.1021/es2026533 |Environ. Sci. Technol. 2011, 45, 9649–9657

Environmental Science & Technology and just below 1400 cm1 and 750 cm1 (Figure S2, Supporting Information). The peak at 1050 cm1 is characteristic of the OH bending vibration in an alcohol, and inspection of the literature reveals that UV photolysis of o-alkyl aromatic carbonyl compounds in solution and in the condensed phase leads to the efficient formation of the corresponding dienols and benzocyclobutenols.24,25 The dienol and benzocyclobutenol expected from photolysis of o-tolualdehyde are not commercially available, however, their IR absorption spectra (not quantified) in a low temperature matrix have been reported previously.26 The main absorption bands of benzocyclobutenol in the fingerprint region, located at 1055 cm1 (OH bend), 1212 cm1 and 1401 cm1, are all apparent in the product spectra obtained here, while there is no evidence for the presence of the dienol (at 1256 cm1 (OH bend) and 1102 cm1 (CO stretch)). It is therefore proposed that benzocyclobutenol is a product of the gas-phase photolysis of o-tolualdehyde. Since benzocyclobutenol is not commercially available, 1-indanol, the similarly structured aromatic cyclic alcohol, was used as a surrogate compound:

The main absorption features of 1-indanol are virtually identical to those observed in the product spectra (Supporting Information Figure S2) thus providing further evidence to support the formation of benzocyclobutenol as a reaction product from photolysis of o-tolualdehyde. Calibrated infrared spectra of 1-indanol were subsequently used for quantiﬁcation of benzocyclobutenol and the concentrationtime proﬁle in Figure 1 shows that the aromatic cyclic alcohol is in fact the major reaction product. The corresponding product yield plots for benzocyclobutenol and o-phthalaldehyde displayed in Figure 2 are linear, indicating that these compounds are primary products of the reaction and are not removed to any signiﬁcant extent during the time scale of the experiments. Only trace amounts of carbon monoxide were observed and the yield plot in Figure 2 is curved, indicating that it is most likely a secondary product. Similar results were also obtained from analysis of the FTIR spectra obtained in the indoor chamber experiments, Table 1. A number of gas-phase photolysis products were also detected using GCMS. During the indoor chamber experiments, ophthalaldehyde, phthalide, phthalic anhydride, o-toluic acid, and o-cresol were identiﬁed by comparison of the retention times and mass spectra with those of standards (Figure S3 and Table S1, Supporting Information). The molar yields of o-phthalaldehyde and phthalide were (0.25 ( 0.04) and (0.02 ( 0.01), in good agreement with the results obtained by FTIR spectroscopy. Analysis of the denuder extracts showed the formation of ophthalaldehyde and very small amounts of glyoxal. Quantiﬁcation of the other products identiﬁed by direct injection GCMS proved diﬃcult due to losses during the sampling procedure. However, there was no evidence for the presence of benzocyclobutenol in the mass spectra of the products. Similar results were obtained at EUPHORE, where o-phthalaldehyde and phthalide were also detected as the major products, along with o-cresol and 2-hydroxymethylbenzaldehyde in trace amounts. A small peak, possibly due to benzocyclobutenol was also observed in the GCMS product spectra, but could not be conﬁrmed due to the lack of a standard. The benzocyclobutenols are known to undergo thermal decomposition above 80 °C24 and it therefore

ARTICLE

Scheme 1

seems very likely that benzocyclobutenol decomposed during the GCMS analysis, either in the heated injector or on the column itself. In fact, degradation of the aromatic cyclic alcohol could be responsible for generating some of the aromatic products (phthalide, phthalic anhydride, o-cresol) which were detected by GCMS but not by in situ FTIR spectroscopy. The information obtained from these product studies can be used to propose a mechanism for the atmospheric photolysis of o-tolualdehyde, shown in Scheme 1. The initial step in the mechanism is photoexcitation of the aldehyde through the nfπ* electronic transition followed by intramolecular hydrogen abstraction (Norrish Type II process) to produce a 1,4-biradical species.24,25 Based on the distribution of identiﬁed reaction products, two possible reactions for the 1,4-biradical are proposed, cyclization to form benzocyclobutenol and reaction with O2 to produce o-phthalaldehyde. The cyclization reaction was initially proposed by Yang27 to explain the formation of cyclic alcohols from photolysis of aliphatic ketones in solution. However, more recent studies on o-alkyl aromatic carbonyl compounds, in solution and in the solid phase, indicate that direct cyclization of the 1,4-biradical is unlikely, and that formation of the dienol, in both (Z)- and (E)-conﬁgurations is preferred.24,25 The (Z)-dienol is very short-lived and reverts to the starting aldehyde via a rapid 1,5-H atom shift, whereas the (E)-dienol undergoes thermal conrotatory ring closure to form benzocyclobutenol. As indicated above, there is no evidence for the formation of the dienol in these experiments, although its lifetime in the gas-phase may be too short to be detected. Nevertheless, the possibility that formation of benzocyclobutenol proceeds via the (E)-dienol cannot be ruled out. The formation of o-phthalaldehyde is postulated to occur via reaction of the 1,4-biradical with O2. It is possible that this pathway may involve a number of concurrent or subsequent steps; H-atom abstraction from the OH group and addition of O2 to the methylene unit to form a peroxy radical. In the absence of NO, the peroxy radicals react together, with the major reaction pathway resulting in oxy radicals which also react with O2 to form o-phthalaldehyde. The minor reaction pathway involves combination of the peroxy radicals to produce two molecular products in one step; o-phthalaldehyde and 2-hydroxymethylbenzaldehyde. A very small amount of the latter species was detected by GCMS in the EUPHORE experiment. However, for the sake of simplicity, this mechanistic detail is omitted from Scheme 1. 9653

dx.doi.org/10.1021/es2026533 |Environ. Sci. Technol. 2011, 45, 9649–9657

Environmental Science & Technology

ARTICLE

Figure 3. Number and size distribution of particles formed during the photolysis of o-tolualdehyde at EUPHORE on 3 July 2003. The numbers in the legend refer to time during the experiment.

Additional reaction pathways are required to explain the formation of the minor products detected by GCMS and FTIR spectroscopy. Phthalide and phthalic anhydride are known to be generated from the UV photolysis of the primary product, ophthalaldehyde.28 The formation of the other aromatic compounds, o-cresol and o-toluic acid, is more diﬃcult to explain and since these products were only detected in very small amounts, speculative mechanisms for their formation will not be considered here. As indicated above, carbon monoxide appears to be formed as a secondary product. Benzocylcobutenol is not expected to undergo rapid photolysis under the conditions employed in the chambers, suggesting that photolysis of ophthalaldehyde is the most likely source of carbon monoxide. However, preliminary experiments on the photolysis of o-phthalaldehyde in both chambers did not generate detectable levels of carbon monoxide. The origin of the secondary carbon monoxide therefore remains unknown. Secondary Organic Aerosol Formation. Secondary Organic Aerosol (SOA) formation was observed in all experiments performed in both chambers. The evolution of the aerosol produced from photolysis of o-tolualdehyde in the EUPHORE chamber is shown in Figure 3. A large number of particles with a mean diameter of around 20 nm were observed within 5 min of the start of photolysis. This initial “burst” of particles also corresponded to the greatest number present during the reaction. As photolysis continued, there was a gradual reduction in particle number and a corresponding increase in average particle diameter due to coagulation. Particles with a mean diameter of around 70 nm were present at the end of the experiment. Very similar results were obtained in the indoor chamber, although the greatest number of particles was typically observed around 10 min later when the mean diameter was 2535 nm. This slight difference in the particle-time profile is probably due to the fact that the photolysis process, and hence particle formation, was slower in the indoor chamber.

The mass and volume concentration of SOA generated in the EUPHORE experiment on third July 2003 is shown in Figure 4. The density of SOA formed in the experiment was determined to be (1.09 ( 0.01) g cm3 by plotting the measured mass (TEOM) versus the volume (SMPS), Figure 4 (inset). The calculated density is lower than the value of 1.35 g cm3 determined for SOA generated from photooxidation of benzene in the EUPHORE chamber using the same method,15 and also at the lower end of the range of eﬀective densities (1.06 1.45 g cm3) for laboratory-generated SOA from anthropogenic precursors.29 The maximum aerosol concentration was observed at the end of the experiment when all of the o-tolualdehyde had reacted. After closing the chamber, the aerosol was found to undergo a ﬁrst order decay, k = (1.46 ( 0.10) 105 s1, due to deposition at the walls of the reactor. This wall loss factor was used to provide a corrected value of 258 μg m3 for the maximum aerosol mass concentration, which when divided by the mass of o-tolualdehyde reacted (2491 μg m3) gives an aerosol yield of 0.104. SOA yields were also obtained from four diﬀerent experiments performed in the indoor chamber by converting the measured volume concentrations to mass concentrations using the density of 1.09 g cm3. As shown in Table 1, the SOA yields were found to increase with starting concentration of o-tolualdehyde. This eﬀect has been observed for many other SOA precursors and is consistent with gas/particle partitioning theory.29 The partitioning of semivolatile reaction products to the aerosol phase increases with the amount of available aerosol mass, resulting in higher yields of SOA when greater precursor concentrations are used. It is interesting to note that the yield obtained for the experiment with an initial concentration of 518 ppbV is considerably lower than that determined in the equivalent experiment at EUPHORE. The semivolatile reaction products would be expected to undergo a higher degree of wall loss in the smaller indoor chamber, resulting in less partitioning to the particle 9654

dx.doi.org/10.1021/es2026533 |Environ. Sci. Technol. 2011, 45, 9649–9657

Environmental Science & Technology

ARTICLE

Figure 4. Aerosol mass (TEOM) and volume (SMPS) concentrations measured during the photolysis of o-tolualdehyde at EUPHORE on 3 July 2003. The inset contains the plot of mass versus volume concentration used to determine the density of the aerosol.

phase and lower SOA yields.30 In addition, aerosol formation is slightly slower in the indoor chamber and this suggests that the photolysis rate may also inﬂuence SOA yield. A similar rate eﬀect has also been observed in previous studies of the photooxidation of aromatic hydrocarbons, where various factors including OH precursor type and concentration,31,32 light intensity33 and relative humidity34 have been shown to aﬀect OH levels, oxidation rate, and SOA mass yield. The results obtained here suggest that direct photolysis of otolualdehyde may contribute to SOA formation in chamber studies of o-xylene photooxidation35 and in the ambient atmosphere. Information on the chemical composition of the SOA can provide insights into the key species and processes involved in aerosol formation and also help to identify molecular markers for speciﬁc sources of ambient organic aerosol.29 The only carbonyl products identiﬁed in ﬁlter samples collected in the indoor chamber experiments were o-phthalaldehyde (approximate yield of 10%) and glyoxal (not quantiﬁed). The same carbonyl products were identiﬁed in particles collected at EUPHORE, along with o-toluic acid, a hydroxyl-containing compound with MW 120, tentatively attributed to benzocyclobutenol, phthalaldehydic acid, phthalide, and phthalic anhydride, Supporting Information Table S2. The latter three compounds were also detected in ﬁlter samples of SOA generated from the photooxidation of o-tolualdehyde, where direct photolysis and reaction with OH were both responsible for loss of the aromatic compound.36 Interestingly, a number of oxygenated polycyclic aromatic compounds were also identiﬁed in this study, indicating that intermolecular reactions could be involved in SOA formation and growth. There is evidence to suggest that biradical (Criegee) intermediates are involved in addition reactions with peroxy radicals leading to SOA formation during the ozonolysis of alkenes,29,37 and it is possible that the biradical species produced during the photolysis of o-tolualdehyde may also be involved in similar types of reactions leading to polycyclic compounds with low volatility and hence SOA formation.

Photolysis of aromatic carbonyls in the solid and liquid phases has been shown to generate a wide range of polycyclic aromatic products,24,25 and another intriguing possibility is that some of the photochemically active compounds present in the SOA, for example, o-phthalaldehyde, may undergo light-induced intermolecular reactions with other aromatic species to produce the oxygenated polycyclic aromatic compounds. This work is one of the ﬁrst studies to demonstrate that direct photolysis of a volatile organic compound can produce SOA. The formation of aerosol from photolysis of 2,4-hexadienedial38 and ortho-nitrophenols39 has been noted, although not investigated in detail. More recently, Kessler et al.40 used the photolysis of alkyl iodides to generate single organic radical precursors and proposed that this technique could be used as a simpliﬁed experimental approach to investigate SOA formation from alkanes. Similarly, the photolysis of o-tolualdehyde could also be considered as a good model system to investigate SOA formation mechanisms. The possible range of reaction pathways is considerably less complex than in OH-initiated photooxidation systems and does not involve the use of OH precursors. Although beyond the scope of the present work, a systematic study of the various parameters aﬀecting aerosol production from direct photolysis of o-tolualdehyde could prove to be beneﬁcial in elucidating key processes responsible for SOA formation. Atmospheric Implications. The rate coefficient for the sunlight photolysis of o-tolualdehyde can be used to calculate the tropospheric lifetime with respect to photolysis (τp) from the relationship: τp = 1/j. Using the values of j(o-tolualdehyde) obtained from the EUPHORE experiments yields photolysis lifetimes in the range 1.3 1.7 h. The average value of j(o-tolualdehyde)/j(NO2) = 2.53 102 can be used to provide an estimate of the photolysis lifetime under a variety of other solar irradiation conditions. The other atmospheric loss processes for o-tolualdehyde are reaction with OH and NO3, which have lifetimes of 13.6 and 56.7 h respectively.10 Thus photolysis by sunlight is clearly the dominant atmospheric loss process for 9655

dx.doi.org/10.1021/es2026533 |Environ. Sci. Technol. 2011, 45, 9649–9657

Environmental Science & Technology o-tolualdehyde. The short photolysis lifetime indicates that if released or formed in the atmosphere, it will readily photolyze in the troposphere and could contribute to the production of ozone and secondary organic aerosol. Atmospheric degradation of the reaction products could also contribute to the formation of ozone and other oxidants. The major atmospheric fate of benzocyclobutenol is likely to be gas-phase reaction with OH radicals, with an estimated lifetime of around 24 h,41 whereas ophthalaldehyde can undergo both photolysis and reaction with OH radicals.28 Finally, the kinetic and mechanistic information obtained for the photolysis of o-tolualdehyde could be included in photochemical degradation models, such as the Master Chemical Mechanism,42 that are used to predict secondary pollutant formation. It is envisaged that the inclusion of this chemistry will help to improve the prediction capability of the models.

’ ASSOCIATED CONTENT

bS

Supporting Information. Photolytic loss plot, FTIR spectra, GCMS data. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: +353 21 4902454; fax: +353 21 4903014; e-mail: [email protected].

’ ACKNOWLEDGMENT This work was supported by the Higher Education Authority in Ireland and the European Commission through the research project EUROCHAMP 2 (contract number 228335). The Instituto Universitario CEAM-UMH is partly supported by Generalitat Valenciana, Fundaci on Bancaja, and the projects GRACCIE (Consolider-Ingenio 2010) and FEEDBACKS (Prometeo - Generalitat Valenciana). ’ REFERENCES (1) Feng, Y.; Wen, S.; Chen, Y.; Wang, X.; L€u, H.; Bi, X.; Sheng, G.; Fu, J. Ambient levels of carbonyl compounds and their sources in Guangzhou, China. Atmos. Environ. 2005, 39, 1789–1695. (2) L€u, H.; Cai, Q.-Y.; Wen, S.; Chi, Y.; Guo, S.; Sheng, G.; Fu, J. Seasonal and diurnal variations of carbonyl compounds in the urban atmosphere of Guangzhou, China. Sci. Total Environ. 2010, 408, 3523–3529. (3) Obermeyer, G.; Aschmann, S. M.; Atkinson, R.; Arey, J. Carbonyl atmospheric reaction products of aromatic hydrocarbons in ambient air. Atmos. Environ. 2009, 43, 3736–3744. (4) Kean, A. J.; Grosjean, E; Grosjean, D.; Harley, R. On-road measurement of carbonyls in California light-duty vehicle emissions. Environ. Sci. Technol. 2001, 35, 4198–4204. (5) Jakober, C. A.; Robert, M. A.; Riddle, S. G.; Destaillats, H.; Charles, M. J.; Green, P. G.; Kleeman, M. J. Carbonyl emissions from gasoline and diesel motor vehicles. Environ. Sci. Technol. 2008, 42, 4697–4703. (6) Calvert, J. G. A.; Becker, K. H.; Kamens, R. M.; Seinfeld, J. H.; Wallington, T. J.; Yarwood, G., The Mechanisms of Atmospheric Oxidation of Aromatic Hydrocarbons; Oxford University Press: Oxford, UK, 2002. (7) Atkinson, R; Aschmann, S. M.; Arey, J. Formation of ringretaining products from the OH radical-initiated reactions of o-, m-, and p-xylene. Int. J. Chem. Kinet. 1991, 23, 77–97.

ARTICLE

(8) Atkinson, R.; Arey, J. Atmospheric degradation of volatile organic compounds. Chem. Rev. 2003, 103 (12), 4605–4638. (9) Thiault, G; Mellouki, A.; Le Bras, G. Kinetics of gas phase reactions of OH and Cl with aromatic aldehydes. Phys. Chem. Chem. Phys. 2002, 4, 2194–2199. (10) Cliﬀord, G. M.; Th€uner, L. P.; Wenger, J. C.; Shallcross, D. E. Kinetics of the gas-phase reactions of OH and NO3 radicals with aromatic aldehydes. J. Photochem. Photobiol., A 2005, 176, 172–182. (11) Thiault, G.; Mellouki, A.; Le Bras, G.; Chakir, A.; SokolowskiGomez, N.; Daumont, D. UV-absorption cross sections of benzaldehyde, ortho-, meta-, and para-tolualdehyde. J. Photochem. Photobiol., A 2004, 162, 273–281. (12) Volkamer, R.; Platt, U.; Wirtz, K. The European Photoreactor (EUPHORE), 3rd Annual Report 2000; Barnes, I., Sidebottom, H., Eds.; Bergische Universit€at Wuppertal: Wuppertal, Germany, 2001, 1. (13) Thiault, G.; Mellouki, A.; Le Bras, G.; Wirtz, K. The European Photoreactor (EUPHORE), 4th Annual Report 2001; Barnes, I., Ed.; Bergische Universit€at Wuppertal: Wuppertal, Germany, 2003, 29. (14) Klotz, B.; Sørensen, S.; Barnes, I.; Becker, K. H.; Etzkorn, T.; Volkammer, R.; Platt, U.; Wirtz, K.; Martín-Reviejo, M. Atmospheric oxidation of toluene in a large-volume outdoor photoreactor: In situ determination of ring-retaining product yields. J. Phys. Chem. 1998, 102, 10289–10299. (15) Martín-Reviejo, M; Wirtz, K. Is benzene a precursor for secondary organic aerosol?. Environ. Sci. Technol. 2005, 39, 1045–1054. (16) Wenger, J. C.; Le Calve, S.; Sidebottom, H. W.; Wirtz, K.; Martín-Reviejo, M; Franklin, J. A. Photolysis of chloral under atmospheric conditions. Environ. Sci. Technol. 2004, 38, 831–837. (17) Sellevag, S. R.; Kelly, T.; Sidebottom, H.; Nielsen, C. J. A study of the IR and UV-Vis absorption cross-sections, photolysis and OHinitiated oxidation of CF3CHO and CF3CH2CHO. Phys. Chem. Chem. Phys. 2004, 6, 1243–1252. (18) O’Connor, M. P.; Wenger, J. C.; Mellouki, A.; Wirtz, K.; Mu~ noz, A. The atmospheric photolysis of E-2-hexenal, Z-3-hexenal and E,E-2,4-hexadienal. Phys. Chem. Chem. Phys. 2006, 8, 5236–5246. (19) Borras, E.; Tortajada-Genaro, L. A. Determination of oxygenated compounds in secondary organic aerosol from isoprene and toluene smog chamber experiments. Int. J. Environ. Anal. Chem. 2011in press. (20) Th€uner, L. P.; Bardini, P.; Rea, G. J.; Wenger, J. C. Kinetics of the gas-phase reactions of OH and NO3 radicals with dimethylphenols. J. Phys. Chem. A. 2004, 108, 11019–11025. (21) Temime, B.; Healy, R. M.; Wenger, J. C. A denuder-ﬁlter sampling technique for the detection of gas and particle phase carbonyl compounds. Environ. Sci. Technol. 2007, 41, 6514–6520. (22) Healy, R. M.; Wenger, J. C.; Metzger, A.; Duplissy, J.; Kalberer, M.; Dommen, J. Gas/particle partitioning of carbonyls in the photooxidation of isoprene and 1,3,5-trimethylbenzene. Atmos. Chem. Phys. 2008, 8, 3215–3220. (23) DeMore, W. B.; Sander, S. P.; Golden, D. M.; Hampson, R. F.; Kurylo, M. J.; Howard, C. J.; Ravishankara, A. R.; Kolb, C. E.; Molina, M. J. In Chemical Kinetics and Photochemical Data for use in Stratospheric Modelling, JPL Publication 97-4; Jet Propulsion Laboratory, Pasadena, CA, 1997. (24) Wagner, P. J.; Subrahmanyam, D.; Park, B.-S. Mechanism for the photocyclization of o -alkyl ketones to cyclobutenols. J. Am. Chem. Soc. 1991, 113, 709–710. (25) Moorthy, J. N.; Mal, P.; Natarajan, R.; Venugopalan, P. Eﬃcient photocyclization of o-alkylbenzaldehydes in the solid state: Direct observation of E-xylylenols en route to benzocyclobutenols. J. Org. Chem. 2001, 66, 7013–7019. (26) Gebicki, J.; Krantz, A. Trapping of photoenols from o-tolualdehydes in gas matrices. Dependence of photoenol formation on the nature of the carbonyl function. J. Chem. Soc. Perkin Trans. 2 1984, 1623–1627. (27) Yang, N. C.; Yang, D-D. H. Photochemical reactions of ketones in solution. J. Am. Chem. Soc. 1958, 80, 2913–2914. (28) Wang, L.; Arey, J.; Atkinson, R. Kinetics and products of photolysis and reaction with OH radicals of a series of aromatic carbonyl compounds. Environ. Sci. Technol. 2006, 40, 5465–5471. 9656

dx.doi.org/10.1021/es2026533 |Environ. Sci. Technol. 2011, 45, 9649–9657

Environmental Science & Technology

ARTICLE

(29) 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.; Prev^ot, 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, 5155–5236. (30) Matsunaga, A; Ziemann, P. J. Gas-wall partitioning of organic compounds in a Teﬂon ﬁlm chamber and potential eﬀects on reaction product and aerosol yield measurements. Aerosol Sci. Technol. 2010, 44, 881–892. (31) Ng, N. L.; Kroll, J. H.; Chan, A. W. H.; Chhabra, P.; Flagan, R. C.; Seinfeld, J. H. Secondary organic aerosol formation from mxylene, toluene, and benzene. Atmos. Chem. Phys. 2007, 7, 3909–3922. (32) Song, C.; Na, K. S.; Warren, B.; Malloy, Q.; Cocker, D. R. Secondary organic aerosol formation from m-xylene in the absence of NOx. Environ. Sci. Technol. 2007, 41, 7409–7416. (33) Warren, B.; Song, C.; Cocker, D. R. Light intensity and light source inﬂuence on secondary organic aerosol formation for the mxylene/NOx photooxidation system. Environ. Sci. Technol. 2008, 42, 5461–5466. (34) Healy, R. M.; Temime, B.; Kuprovskyte, K.; Wenger, J. C. The eﬀect of relative humidity on gas/particle partitioning and aerosol mass yield in the photooxidation of p-xylene. Environ. Sci. Technol. 2009, 43, 1884–1889. (35) Song, C.; Na, K.; Warren, B.; Malloy, Q.; Cocker, D. R. Secondary organic aerosol formation from the photooxidation of pand o-xylene. Environ. Sci. Technol. 2007, 41, 7403–7408. (36) Webb, P. J.; Hamilton, J. F.; Lewis, A. C.; Wirtz, K. Formation of oxygenated-polycyclic aromatic compounds in aerosol from the photooxidation of o-tolualdehyde. Polycyclic Aromat. Compd. 2006, 26, 236–252. (37) Sadezky, A.; Winterhalter, R.; Kanawati, B.; R€ompp, A.; Spengler, B.; Mellouki, A.; Le Bras, G.; Chaimbault, P.; Moortgat, G. K. Oligomer formation during gas-phase ozonolysis of small alkenes and enol ethers: New evidence for the central role of the Criegee Intermediate as oligomer chain unit. Atmos. Chem. Phys. 2008, 8, 2667–2699. (38) Klotz, B.; Barnes, I.; Becker, K. H. Kinetic study of the gas-phase photolysis and OH radical reaction of E,Z- and E,E-2,4-hexadienedial. Int. J. Chem. Kinet. 1999, 31, 689–697. (39) Bejan, I.; Abd El Aal, Y.; Barnes, I.; Benter, T.; Bohn, B.; Wiesen, P.; Kleﬀmann, J. The photolysis of ortho-nitrophenols: A new gas phase source of HONO. Phys. Chem. Chem. Phys. 2006, 8, 2028–2035. (40) Kessler, S. H.; Nah, T.; Carrasquillo, A. J.; Jayne, J. T.; Worsnop, D. R.; Wilson, K. R.; Kroll, J. H. Formation of secondary organic aerosol from the direct photolytic generation of organic radicals. J. Phys. Chem. Lett. 2011, 2, 1295–1300. (41) US Environmental Protection Agency. Estimation Programs Interface (EPI) Suite v.3.20, 2007, http://www.epa.gov/oppt/exposure/pubs/episuitedl.htm (accessed October 7, 2011). (42) Bloss, C.; Wagner, V.; Jenkin, M. E.; Volkamer, R.; Bloss, W. J.; Lee, J. D.; Heard, D. E.; Wirtz, K.; Martin-Reviejo, M.; Rea, G.; Wenger, J. C.; Pilling, M. J. Development of a detailed chemical mechanism (MCMv3.1) for the atmospheric oxidation of aromatic hydrocarbons. Atmos. Chem. Phys. 2005, 5, 641–664.

9657

dx.doi.org/10.1021/es2026533 |Environ. Sci. Technol. 2011, 45, 9649–9657

ARTICLE pubs.acs.org/est

An Automated Platform for Phytoplankton Ecology and Aquatic Ecosystem Monitoring Francesco Pomati,†,* Jukka Jokela,†,‡ Marco Simona,§ Mauro Veronesi,§ and Bas W. Ibelings†,|| †

)

Department of Aquatic Ecology, Swiss Federal Institute of Aquatic Science and Technology (Eawag), Seestrasse 79, 6047 Kastanienbaum, Switzerland ‡ Department of Environmental Sciences, Aquatic Ecology, Institute of Integrative Biology (IBZ), ETH-Z€urich, € berlandstrasse 133, 8600 D€ubendorf, Switzerland U § Istituto Scienze della Terra, IST-SUPSI, 6952 Canobbio, Switzerland Netherlands Institute of Ecology (NIOO-KNAW), Droevendaalsesteeg 10, 6708 PB Wageningen, The Netherlands

bS Supporting Information ABSTRACT: High quality monitoring data are vital for tracking and understanding the causes of ecosystem change. We present a potentially powerful approach for phytoplankton and aquatic ecosystem monitoring, based on integration of scanning ﬂow-cytometry for the characterization and counting of algal cells with multiparametric vertical water proﬁling. This approach aﬀords high-frequency data on phytoplankton abundance, functional traits and diversity, coupled with the characterization of environmental conditions for growth over the vertical structure of a deep water body. Data from a pilot study revealed eﬀects of an environmental disturbance event on the phytoplankton community in Lake Lugano (Switzerland), characterized by a reduction in cytometry-based functional diversity and by a period of cyanobacterial dominance. These changes were missed by traditional limnological methods, employed in parallel to high-frequency monitoring. Modeling of phytoplankton functional diversity revealed the importance of integrated spatiotemporal data, including circadian time-lags and variability over the water column, to understand the drivers of diversity and dynamic processes. The approach described represents progress toward an automated and trait-based analysis of phytoplankton natural communities. Streamlining of high-frequency measurements may represent a resource for understanding, modeling and managing aquatic ecosystems under impact of environmental change, yielding insight into processes governing phytoplankton community resistance and resilience.

’ INTRODUCTION Freshwater ecosystems are characterized by high levels of biodiversity, and are among the most threatened ecosystems on earth 1,2 (Millennium assessment: http://www.maweb.org). Understanding and managing environmental change in aquatic ecosystems is complicated by co-occurring and interacting stressors like climate change, eutrophication, and pollution that, for example, can interact to favor harmful algal blooms.36 We suﬀer from a general lack of knowledge on the background rates and direction of change in pristine ecological systems, as well as in stressed ecological communities.7 These limits can hamper our ability to detect the signature of a range of anthropogenic impacts on ecosystems, or predict patterns of recovery. Phytoplankton communities are highly diverse and dynamic. They respond rapidly to climate change, eutrophication, and pollution, and play an important role in aquatic ecosystem biogeochemical processes.4,814 Phytoplankton density (algal blooms) and community composition (e.g., toxic cyanobacteria) are the prime agents impacting water quality, ecosystem and human r 2011 American Chemical Society

health,15 and have been suggested to be used as such for ecosystem assessment.1619 Monitoring, understanding, and predicting changes in structural (composition, diversity, evenness) and functional (phenotypic characteristics, growth rate, productivity) aspects of phytoplankton communities across space and over time represents however a challenge for aquatic ecology. The capturing of population dynamics, community succession and adaptation to environmental change requires: (1) high-frequency sampling to follow fast plankton ﬂuctuations20 and potential chaotic dynamics;21 (2) vertical (depth) distribution of algal taxa and their physio-morphological characteristics (traits);22 (3) a functional, trait-based assessment of communities and ecosystems based on the characteristics of the organisms’ phenotypes

Received: June 7, 2011 Accepted: October 5, 2011 Revised: September 9, 2011 Published: October 07, 2011 9658

dx.doi.org/10.1021/es201934n | Environ. Sci. Technol. 2011, 45, 9658–9665

Environmental Science & Technology that directly respond to environmental changes and determine eﬀects on aggregated processes.13,23,24 The goal of this article is to present an integrated platform able to (1) provide automated high-frequency measurements of phytoplankton at diﬀerent lake depths; (2) couple in situ biological monitoring with data about the physical environment; (3) provide a streamline of real-time data for modeling and forecasting phytoplankton dynamics. By integrating a Cytobuoy with an Idronaut vertical proﬁling system, we addressed the objective of increasing spatiotemporal resolution in ﬁeld data collection. It has been proposed that scanning ﬂow-cytometry, oﬀered by instruments like the commercially available Cytobuoy, may oﬀer advantages over microscopic methods for cell counting and classiﬁcation of phytoplankton, including the possibility of automation and high frequency ﬁeld measurements of phytoplankton physio-morphological characteristics.20,2527 A novel aspect of our monitoring approach, therefore, lays in the use of cytometrydata for a description of phytoplankton functional diversity and expressed phenotypic traits, which allow tracking phytoplankton responses at the functional group level. Trait-based approaches and functional groups are becoming increasingly important in understanding phytoplankton ecology.22,2830 In this study we tested our monitoring platform optimized for deep water bodies, designed to aﬀord comprehensive data to study phytoplankton ecology and to improve water resource management. To support the validity of our approach we report the results form a monitoring campaign (spanning roughly one month in May 2010) during which automated measurements were coupled by fortnightly limnological data (physics, chemistry, and biology).31

’ MATERIALS AND METHODS Automated Monitoring Platform. Phytoplankton counting, characterization, and classification were performed using a scanning flow cytometer Cytobuoy (Woerden, The Netherlands), designed to analyze the full naturally occurring range from small (e.g., picoplankton) to large (e.g., colonial cyanobacteria) planktonic particles (1700 μm in diameter and a few mm in length) and relatively large water volumes (http://www.cytobuoy.com)25 (Supporting Information (SI) Figure S1-e). In our instrument, particles were intercepted by two laser beams (Coherent solidstate Sapphire, 488 and 635 nm, respectively, 15 mW) at the speed of 2 m s1. In this study, digital data acquisition was triggered by the sideward scatter (SWS) signal (908 nm). The light scattered at two angles, forward (FWS) and SWS, provided information on size and shape of the particles. The fluorescence (FL) emitted by photosynthetic pigments was detected as red (FLR), orange (FLO) and yellow (FLY) signals collected in the wavelength ranges of 668734 (chlorophyll-a, Chl-a), 601668 (phycocyanin and phycoerythrin), and 536601 nm (degraded pigments), respectively. Laser alignment and calibration processes were done before the monitoring campaign using yellow FL beads of 1 and 4 μm diameter. Our Cytobuoy allowed automatic acquisition of particles in time-intervals, time-speciﬁc measurement, and ﬁxed-measurement on occurrence of a trigger signal (see below). This study was based on automated acquisition of 2 ﬁxed-measurements for every trigger-signal received in order to optimize the detection and quantiﬁcation of small and large particles in two separated analyzes, and on a scheduled time-speciﬁc background measurement per day with water being sampled at 25 m (no phytoplankton growth). Remote accessibility of the Cytobuoy via the

ARTICLE

Internet-UMTS network allowed unlimited data access and transmission rates along with increased location ﬂexibility. Further technical details on our Cytobuoy, measuring settings and conﬁgurations are reported in the SI. In order to accomplish depth resolution, we employed a vertical proﬁling system made up of three integral parts: Controller Module (SI Figure S1-a,-b), Proﬁler Module (SI Figure S1-b), and OCEAN SEVEN 316Plus CTD (O7) multiparameter probe (SI Figure S1-c) (Idronaut, Brugherio, Italy, www.idronaut.it). The O7-probe was equipped with seven sensors: pressure, temperature (°C), conductivity (μS, absolute and at 20 °C), pH, oxygen (mg/L and % saturation), and NO3 (μg/L) (Idronaut). An external TriLux ﬂuorimeter was interfaced with the O7 probe in order to quantify levels of Chl-a, phycoerythrin and phycocyanin (Chelsea Technologies Ltd., Surry, UK). More information on the Idronaut proﬁling system can be found in the SI. For automatic depth proﬁles, we allowed the Cytobuoy to accept an electric signal from the Idronaut Controller Module as a trigger to start the measurement cycle during O7 step-proﬁles. We ran two independent automatic monitoring programs, one with the Cytobuoy and one only with the O7-multiparameter probe, with separated proﬁle settings and diﬀerent monitoring frequencies. In this study we scheduled a step proﬁle involving six depths—covering the entire photic zone—with the Cytobuoy (2, 4, 6, 8, 10, and 12 m) and a continuous proﬁle with the O7-multiparameter probe from 1 to 20 m to be performed twice a day each, to catch diel variations in the temperature structure of the water column: the theoretical maximum and minimum daily stratiﬁcation at 3 p.m. and 3 a.m. (12 h frequency), respectively. For step-proﬁle phytoplankton measurements, we retrieved water from selected depths using an external pump (capacity 1 L min1), an antimicrobial silver-nanoparticle coated and shaded ﬂexible polyethylene tubing (Flexelene, Eldon James Corp., Loveland, CO), and a surface plexiglass chamber (250 mL) from which the Cytobuoy subsamples through a needle injector (SI Figure S1-e). The pump was placed downstream from the chamber in order to avoid damaging algal cells or colonies prior to measurements. More information on structural components of the monitoring platform, how we integrated our instruments to achieve depth proﬁles, and an example of automated operation using the integrated system and maintenance details are reported in the SI. Sampling. The automated monitoring platform was moored in Lake Lugano, at a site protected from strong winds and currents and close to the location of the routine historic lake monitoring program (coordinates 45°570 33.4300 N, 8°520 53.4900 E) (SI Figure S2). This site is representative for the most eutrophic of the lake’s three distinct basins31 (SI Figure S2). Data presented in this article refer to the monitoring period from the 28th of April to the 31st of May 2010 (with six depths over the photic zone and a frequency of two profiles per day). Independent limnological data were collected at 300 m distance from the platform with a fortnightly frequency. They included physical characteristics of the whole water column, chemical analyses on algal nutrients and integrated phytoplankton samples (from 0 to 20 m). Additional information on these data can be found in the SI. For comparison between cytometry-based richness and phytoplankton species richness (Table 1, SI Figure S6) we used additional samples from Lake Lugano collected between June and December 2010 and data from a study conducted in Lake Zurich during spring 200932 (SI). 9659

dx.doi.org/10.1021/es201934n |Environ. Sci. Technol. 2011, 45, 9658–9665

Environmental Science & Technology

ARTICLE

Table 1. Comparison of Selected Properties of Automated Measurements to Classical Phytoplankton Monitoring feature* number of samples year1 (n)

classical limnology 1218a

automated platform >700b

lag (Δ)

2 weeks 1 month

12 h

fundamental period (T0 = Δn)

12

>700

frequency (1/T0)

0.083

0.0014

nyquist frequency (1/2Δ), highest

12 months (612 cycles year1)

24 h (365 cycles year1)

possible frequency resolution of depth gradient

from 1 integrated to 10 samples over photic zone

from 6 to 12 samples over photic zone

phytoplankton density and physiomorphological traits

estimated from ca. 200500 counts/in 100200 mL

from ca. 30,000 counts/in 100400 μL volume

number of descriptors measured per individual

2 (size, volume)

54 (3D descriptors, pigment type, concentration etc.)

estimation of diversity

taxonomic, functional

Functional

number of taxa groups

14 to 61 per samplec

NA

number of functional groups

5 to 20 per samplec

4 to 53 per samplec

reproducibility/repeatability of data

lowd

high 27e

a

Considering one sample per month plus an extra fortnightly sample during productive seasons as in refs 14 and 31 (SI). b The automated system is currently producing data series across seasons. c Range in number of species and functional groups during intercalibration performed in Lake Zurich and Lake Lugano: Reynolds categories29 were utilized for functional grouping of microscopically identiﬁed species, for Cytobuoy-derived functional groups see the Materials and Methods, for a plot of Cytobuoy-derived versus taxonomic diversity see SI Figure S6, d Quality assessment trials highlighted that phytoplankton microscopic counts can be diﬃcult to reproduce across laboratories since they rely on human subjective assessment, biased by the experience/ability/condition of the operator, and that they suﬀer from low repeatability (high diﬀerences between replicated samples) (http://www. planktonforum.eu)26,50 (SI); e Five consecutive-replicated sampling cycles were performed in this study at the same depth and data assessed by canonical discriminate function analysis (SI). * From ref 34.

Data Analysis. Data manipulation, analysis and graphics were performed in the R programming language (www.r-project.org). The Cytobuoy provided 54 descriptors of 3D structure and FL profile for each particle.25 Data sets also included original sampled volume, date, time, and depth at which particles were taken. We visually inspected the distribution of raw data with regards to FL signals and set database-specific threshold levels to divide fluorescent (FL) from non-FL particles. The overall FL and nonFL databases comprised 1 and 5 million particles, respectively. Cytobuoy particle descriptors were standardized to zero mean and unit variance and, by principal component analysis, reduced to 33 orthogonal vectors covering 99% of total variance in the data (data not shown). Principal components were utilized for grouping particles into functional categories using K-means clustering. We compared several K values and selected the optimal number of K based on the within groups sum of squares.33 Phytoplankton densities were calculated by inferring the number of cells from the number of humps in the SWS signal of each particle to account for colonial species.20,25 O7 sensor data were organized in a separated database. Cyanobacterial-like particles were identiﬁed based on FLO and FLR emissions after excitation by the 495 and 635 nm lasers, respectively, after visual inspection. These signals are expected as a response to the presence of the cyanobacterial-speciﬁc pigment phycocyanin.25 We modeled richness of Cytobuoy-derived functional groups of phytoplankton (response variables) in the upper 12 m of the water column based on high frequency environmental data (explanatory variables). Explanatory variables included: water parameters (mean of top 12 m), coeﬃcient of variation (CV = SD/mean) of parameters over water-column and meteorological data at time-lag(0), -lag(1) (=24 h), and -lag(2) (=48 h). The response variables showed signiﬁcant temporal autocorrelation only at time-lag(1) (data not shown). We therefore included for each model the response variable at time-lag(1) as explanatory, in order to account for temporal autocorrelation of data. All variables

were scaled in order to standardize eﬀect sizes and let to compete in the same model. The best model was selected based on Akaike’s information criterion (AIC) with a stepwise procedure (alternation of forward selection and backward elimination of variables with p > 0.05).34 The relative importance of drivers was assessed by bootstrapping (999 times) the percentage contribution to the R2 of the model among the regressors, and extracting the relative 95% conﬁdence intervals.

’ RESULTS AND DISCUSSION Phytoplankton Depth Heterogeneity. Our monitoring approach was able to reveal fine changes in the relative depth distribution of phytoplankton functional-group richness, Chl-a concentration and cell density with statistically significant differences between day and night profiles (SI, Figure S3S4). Similar data have been observed using flow-cytometry in oceanic profiles of phytoplankton communities.3537 We did not observe a significant difference in the vertical physical structure of the water column between day and night profiles (SI Figure S3S4), and limited changes between day and night airtemperatures during the study period (data not shown). Our data suggest that depth-specific day-night dynamics in phytoplankton community composition and abundance are driven by biological factors, rather than environmental changes (SI Results and Discussion). Temporal Phytoplankton Dynamics. The frequency and intensity of phytoplankton blooms are key elements for ecological status definition.16,17,19 Considering that most algal taxa can reach bloom conditions and disappear within a few days (implying a maximum oscillation frequency of 23 density peaks per week), a minimum sampling frequency of 46 times per week would be needed to follow algal dynamics (Nyquist frequency, Table 1) and quantify their intensity adequately.20 9660

dx.doi.org/10.1021/es201934n |Environ. Sci. Technol. 2011, 45, 9658–9665

Environmental Science & Technology

ARTICLE

Figure 1. Automated measurements of phytoplankton density, diversity and associated changes in environmental heterogeneity. (A) Phytoplankton abundance (from Cytobuoy, solid line) compared to microscopic counts (black square), abundance of non-FL particles (dashed line, scaled to ﬁt graph by dividing values by 250) and Chl-a concentration (from O7-probe, gray line); (B) Richness of Cytobuoy-based functional groups (black line) compared to microscopic species counts (black square), and Pielou’s evenness (Shannon-diversity/Log(species richness)) of groups (gray line) compared to the same index derived by microscopic counts (gray square); C) CV over the water column in temperature (black line) and conductivity at 20 °C (gray line). The CV can be used as a proxy of environmental (depth) heterogeneity.14 In (A) and (B), data represent the average of the top 12 m of the water column. The gray vertical line highlights the mixing event.

Our automated monitoring platform was able to perform 2 vertical proﬁles per day (at a ﬁxed depth the maximum frequency could be of six samples per hr). Figure 1 reports results from daily monitoring samples (time is 3 pm, frequency = 1 day1) during the study. This frequency was capable of capturing ﬁne ﬂuctuations in FL particle density (phytoplankton) and total Chl-a concentration over the water column (Figure 1A). Our data were comparable to previous work using ﬂow-cytometry in the ﬁeld in terms of temporal resolution on algal dynamics (ref 27 and literature therein). Measured phytoplankton density was comparable with microscopic counts and correlated well with Chl-a concentration levels (Figure 1A) (R2-adjusted = 0.651, p = 4.32408), as also reported elsewhere.32 Our system was able to follow dynamics of non-FL particles (suspended solids, dead

cells, heterotrophic bacteria), which did not correlate with algal cell concentrations apart from a short period in the middle of the time-series (days 1518) (Figure 1A). Previous work using ﬂow cytometry in phytoplankton aimed at identifying broad functional groups (such as picoeukaryotes, microalgae, cyanobacteria, etc.) and some phytoplankton species with clearly distinguished morphology or pigmentation (such as Pseudonitzschia, Cryptomonas, Synura, Dinobryon)20,25,27,38 (and literature therein). This type of analysis lacked a proper measure of diversity. We used the Cytobuoy to describe key phytoplankton traits like size, coloniality, pigment type, and content, which we used to create groups of functionally similar individuals.29,30 The possibility of monitoring individually measured phytoplankton physio-morphological descriptors may oﬀer the best prospects in 9661

dx.doi.org/10.1021/es201934n |Environ. Sci. Technol. 2011, 45, 9658–9665

Environmental Science & Technology terms of objectivity, reproducibility, functional properties and prediction of algal assemblages.22,23,30 The number of Cytobuoyderived functional groups was comparable with the total number of species detected in the photic zone of the water column (Figure 1B, SI Figure S6), as also reported elsewhere.32 Generally, the number of functional groups in a community is smaller than the number of species, since in current functional classiﬁcation methods more than one species can be assigned to the same functional category.29,30 With our trait-based approach, however, it is also possible that individuals of the same species can be allocated to diﬀerent functional groups based on their expressed morphology (for example, colonial species can be assigned to two diﬀerent groups depending on whether they are present as single cells or colonies). The Cytobuoy description of the relative abundance of phytoplankton functional groups deviated from microscopically measured evenness (Figure 1B). This could be caused by superior precision of automated density measurements, and to the fact that the identity (and abundance) of Cytobuoy-derived functional groups does not fully reﬂect the identity (and abundance) of microscopically deﬁned taxonomic groups as reported above (several species can map into one functional category and individuals of the same species can be assigned to diﬀerent groups). We observed a strong decrease in phytoplankton functional richness and evenness in the middle of the time-series (Figure 1B), followed by a short recovery period that led to higher cell density (Figure 1A). These dynamics were completely missed by the fortnightly limnological sampling (Figure 1). Our approach offered the advantage of having automated measurements of environmental conditions for the observed algal dynamics (SI Figure S7). Six days of rainy and stormy weather (SI Table S1) were associated with a period of low phytoplankton diversity and productivity (with high levels of non-FL particles), and a decrease in CV in temperature and conductivity over the ﬁrst 12 m of the water column. This eventually led to a mixing event on day 19 (Figure 1C, SI Figure S7). The phytoplankton community in the days preceding disturbance (started at day 5) showed a gradual decline, reaching the minimum of evenness and richness just before the mixing event (on days 17 and 18, respectively). The mixing event re-established evenness in the community that fully recovered functional diversity in 6 days (Figure 1BC). Functional diversity, as opposed to taxonomic diversity, appears to be a better predictor of ecosystem functioning across a range of communities and measures of functional diversity may aﬀord a better description of the functionality of the ecosystem and its resilience to disturbance.12,13,23,24,39 Using Cytobuoy-Derived Phytoplankton Traits. Our approach allows tracking phytoplankton physio-morphological characteristics such as cell size and shape (which influence motility and nutrient uptake through surface/volume ratio), photosynthetic performance (driven by pigment type and concentration), active nutrient uptake and coloniality.22 Cell size and photosynthetic performance are key phytoplankton traits, affecting growth, metabolism, access to resources, susceptibility to grazing, and are extremely plastic responding to the environment and to species interactions.22,32 Analysis of dynamics and distributions of these focal phytoplankton traits could improve our forecasting capabilities of community structure and ecosystem functions.12,13,24,39 Pigment profiles can also be used to specifically target certain phytoplankton groups of interest in their spatiotemporal dynamics.20,25 We report temporal changes in mean and variance of phytoplankton size and suspended non-FL particles size (Figure 2A, SI

ARTICLE

Figure 2. Using phytoplankton traits such as size and pigment content to track community changes. A) Average size of FL (phytoplankton; black line) and non-FL (suspended solids, bacteria, dead cells; gray line) particles; B) Ratio between concentrations of phycocyanin and Chl-a (black line) and abundance of cyanobacterial-like cells (gray line) compared to microscopic counts of cyanobacteria (/). Phycocyanin is a cyanobacterial-speciﬁc pigment: the ratio between phycocyanin and Chl-a concentrations can be used as an indication of the dominance of cyanobacteria in the phytoplankton community. Data represent the average of the top 12 m of the water column. The gray vertical line highlights the mixing event.

Figure S8). In addition, we tracked the dynamics in abundance of cyanobacteria using Cytobuoy data and phycocyanin/Chl-a concentration ratios obtained with the O7-probe (Figure 2B). Shortly before “disturbance” (days 1517), a period characterized by low diversity and productivity (Figure 1), the study site was dominated by large cyanobacterial colonies (Figure 2A and B). Mean water column cyanobacterial density obtained by the Cytobuoy was almost identical to microscopic count levels (Figure 2B) and was likely associated with the presence of Planktothrix rubescens ﬁlaments (SI Table S2). The mixing event rapidly and dramatically reduced cyanobacterial abundance and the average size of the phytoplankton community (Figure 2).40 Variation in the dimensions of non-FL particles appeared to be very small compared to the dynamics in phytoplankton size (note the y-axis scales in Figure 2A). Compared to conditions before the disturbance, the ﬁnal days of our time-series were characterized by smaller size phytoplankton cells (Figure 2A), probably eukaryotic nanoplankton of genera Stephanodiscus and Melosira (SI Table S2), dominating a more productive (Figure 1A) and diverse community (Figure 1B, SI Figure S8). Our approach introduces the possibility of monitoring a large number of phytoplankton individuals and their traits per population or through the entire community. Individuals and populations should be the basic units of investigation to assess the status of communities and ecosystems, since they respond phenotypically 9662

dx.doi.org/10.1021/es201934n |Environ. Sci. Technol. 2011, 45, 9658–9665

Environmental Science & Technology

ARTICLE

Table 2. Multiple Linear Regression Model Describing Phytoplankton Richness (Cytobuoy-Derived Functional Groups) in Terms of Changes in Environmental Conditions over the Period of Study 95% conﬁdencec drivera

coeﬃcient p-value percentage of R2b lower upper

Air T-lag(1)

0.906

0.0000

22.7

0.113 0.277

Cond.-lag(1)

0.266

0.0282

16.3

0.067 0.230

Cond.-lag(2)

0.589

0.0000

15.7

0.096 0.193

CV-Cond.-lag(1)

0.751

0.0000

10.8

0.063 0.142

pH-lag(2)

0.709

0.0000

10.2

0.058 0.168

0.286 0.544

0.0000 0.0143

4.5 4.2

0.032 0.064 0.042 0.066

N-NO3 CV-pH-lag(1) N-NO3-lag(2)

2.246

0.0001

3.9

0.034 0.047

1.394

0.0000

3.8

0.025 0.073

Water T-lag(1)

0.932

0.0010

2.7

0.022 0.037

N-NO3-lag(1)

1.519

0.0035

2.6

0.031 0.037

CV-NO3-lag(1)

0.534

0.0097

1.7

0.016 0.040

Light -lag(1)

0.203

0.0012

0.9

0.015 0.079

CV-Water T-lag(1)

Drivers: T = temperature (°C); Cond. = conductivity at 20 °C; CV = coeﬃcient of variation over the sampled depths; Light = maximum irradiance (W/m2); lag(1) and (2) = time-lag 24 and 48 h, respectively. b Drivers are ordered based on their relative contribution to the R2 of the model, expressed as percentage of total. c Conﬁdence intervals refer to the bootstrapped relative contribution to the R2 of the model. a

(and genetically) to disturbance or stress and eventually evolve altering community processes and ecosystem functioning.41 Modeling High-Frequency Phytoplankton Dynamics. Our automated monitoring approach allows to better couple environmental forcing with phytoplankton community dynamics, in particular at the functional level (which may relate to crucial ecosystem services13,24,42). Using data from the period of study, we modeled the Cytobuoy-based phytoplankton functional richness in order to provide an example of how spatiotemporal measurements of environmental conditions, coupled with biological data, can provide insight into drivers of community responses and changes. Temperature (both atmospheric and water), conductivity (whose main contributors were carbonate and bicarbonate ions) and the heterogeneity of environmental conditions over the water column appeared to be the most important drivers of phytoplankton functional richness (Table 2). Most of the drivers appeared to inﬂuence the response variable with a time lag of 24 or 48 h (Table 2). Our modeling exercise highlights the importance of (i) time-lags between environmental change and response at the level of phytoplankton community, (ii) variability of parameters over the water column (depth heterogeneity), and (iii) in situ meteorological conditions for understanding and modeling phytoplankton community dynamics. Intensity of ﬂuctuations and heterogeneity by depth in key environmental variables may represent fundamental factors to understand and predict changes in plankton diversity.14 The collection of the above type of high-resolution data would be intractable without the aid of an in situ automated monitoring station like the one presented in this study. A similar approach can be used to model and forecast cyanobacterial blooms.

Toward an Adaptive, Integrated Approach to Aquatic Ecosystem Monitoring. Monitoring frameworks that evolve

along with our improved knowledge of ecosystem processes would strongly benefit ecosystem health assessment and management by allowing to assess the impact of ongoing environmental change, to study recovery processes, and to built more reliable forecasting models.43 Sophisticated monitoring approaches like the one that we have developed can offer the spatiotemporal resolution and flexibility necessary to capture and model natural phytoplankton responses to disturbance or stress, or to test ecological and evolutionary hypotheses including the mechanisms that lead to stable coexistence of species. For example, high-frequency data afford the possibility of studying niche processes and environmental filters on diversity and trait distribution patterns,44,45 while tracking the vertical distribution of functional groups and their abundance allow testing for the importance of dispersal limitation among patches in the assembly of the phytoplankton community.46 Table 1 summarizes some of the properties of our automated data-series compared to traditional monitoring, including diel temporal resolution in phytoplankton community dynamics and water column structure over the photic zone of the lake (Figure 1, 2, SI Figures S3, S4, S7). We were not able to capture horizontal spatial heterogeneity of phytoplankton and the associated environment. The lack of spatial information across the water surface may be solved by integrating our platform data with remote sensing from satellites or from local devices that use spectral information reﬂected from the water surface47 (http:// www.waterinsight.nl). Depth represents however the most heterogeneous aspect of the phytoplankton spatial environment, and our vertical proﬁles may be crucial to understand and model the eﬀects of disturbance, spatial heterogeneity and patch dynamics on phytoplankton community structure.48,49 Several phytoplankton groups are in fact capable of vertically migrating in the water column being motile (e.g., dinoﬂagellates) or able to regulate buoyancy (e.g., cyanobacteria).29 Depth resolution is therefore essential to track algal populations, which can be deﬁned as groups of similar organisms (for example belonging to the same cytometry-derived cluster) that coexist at the same time in the same water layer. The bottleneck in monitoring natural systems is the development of automated technologies for the identiﬁcation and counting of organisms.20,27,50,51 Our description of phytoplankton richness obtained by cluster analysis of automated ﬂowcytometry data appeared to closely match the taxonomic richness derived by microscopic analysis (Table 1, SI Figure S6). Technical repeatability and across-lab reproducibility currently represent disadvantages of classical microscopic counts. An automated monitoring station like the one that we developed may oﬀer the objectivity and reproducibility of a standardized measuring system that (1) reduces human error; (2) aﬀords a detailed description of individual algal features; (3) provides high data complexity; and (4) increase spatiotemporal resolution compared to manmade monitoring campaigns (Table 1).20,50 The temporal and spatial monitoring scales of our analysis (Table 1) were roughly equivalent since both of them reﬂected processes operating over day-night cycles across the water column. The beneﬁts of an integrated spatiotemporal approach to monitoring include52 (i) accounting for spatiotemporal coexistence mechanisms that purely spatial or temporal approaches would miss; (ii) generating new hypotheses and allowing rigorous testing of theoretical models; (iii) improving our descriptive 9663

dx.doi.org/10.1021/es201934n |Environ. Sci. Technol. 2011, 45, 9658–9665

Environmental Science & Technology power for developing forecasting models; and (iv) optimizing monitoring strategies by choosing appropriate scales for sampling. A ﬁne spatiotemporal resolution with regards to organisms and the environment may represent a critical resource for scientists and stakeholders challenged by understanding, modeling, and managing aquatic ecosystems.1719 The approach presented here can be applied to both freshwater and marine ecosystems, and to both natural and engineered environments such as drinking water reservoirs, water-treatment, and aquaculture plants.

’ ASSOCIATED CONTENT

bS

Supporting Information. Extended Materials and Methods and Results and Discussion Sections, Figures S1S8, and Tables S1S2. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone +41 58 765 2174; fax +41 58 765 2162; E-mail: francesco. [email protected].

’ ACKNOWLEDGMENT This research was funded by the Swiss National Science Foundation (SNSF) R’Equip program (project no. 316030_121331 to B.W.I.), Eawag (to F.P.), and Schure-Beijerinck Popping Fonds (to B.W.I.). We are grateful to Referees for their constructive comments and to D. Steiner, M. Schurter, Idronaut, Cytobuoy, and Chavanne Bootswerft for technical support and advice. ’ REFERENCES (1) Abell, R.; Allan, J. D.; Lehner, B. Unlocking the potential of protected areas for freshwaters. Biol. Conserv. 2007, 134 (1), 48–63. (2) Williamson, C. E.; Saros, J. E.; Schindler, D. W. Climate change: Sentinels of change. Science 2009, 323 (5916), 887–888. (3) Mooij, W. M.; H€ulsmann, S.; De Senerpont Domis, L. N.; Nolet, B. A.; Bodelier, P. L. E.; Boers, P. C. M.; Dionisio Pires, L. M.; Gons, H. J.; Ibelings, B. W.; Noordhuis, R.; Portielje, R.; Wolfstein, K.; Lammens, E. H. R. R. The impact of climate change on lakes in the Netherlands: A review. Aquat. Ecol. 2005, 39 (4), 381–400. (4) Paerl, H. W.; Scott, J. T. Throwing fuel on the ﬁre: Synergistic eﬀects of excessive nitrogen inputs and global warming on harmful algal blooms. Environ. Sci. Technol. 2010, 44 (20), 7756–7758. (5) Jak, R. G.; Maas, J. L.; Scholten, M. C. T. H. Ecotoxicity of 3,4dichloroaniline in enclosed freshwater plankton communities at diﬀerent nutrient levels. Ecotoxicol. 1998, 7 (1), 49–60. (6) Pomati, F.; Neilan, B. A.; Suzuki, T.; Manarolla, G.; Rossetti, C. Enhancement of intracellular saxitoxin accumulation by lidocaine hydrochloride in the cyanobacterium Cylindrospermopsis raciborskii T3 (Nostocales). J. Phycol. 2003, 39 (3), 535–542. (7) Magurran, A. E.; Baillie, S. R.; Buckland, S. T.; Dick, J. M.; Elston, D. A.; Scott, E. M.; Smith, R. I.; Somerﬁeld, P. J.; Watt, A. D. Long-term datasets in biodiversity research and monitoring: Assessing change in ecological communities through time. Trends Ecol. Evol. 2010, 25 (10), 574–582. (8) Johnston, E. L.; Roberts, D. A. Contaminants reduce the richness and evenness of marine communities: A review and meta-analysis. Environ. Pollut. 2009, 157 (6), 1745–1752. (9) Downing, A. L.; DeVanna, K. M.; Rubeck-Schurtz, C. N.; Tuhela, L.; Grunkemeyer, H. Community and ecosystem responses to a pulsed

ARTICLE

pesticide disturbance in freshwater ecosystems. Ecotoxicol. 2008, 17 (6), 539–548. (10) Echeveste, P.; Dachs, J.; Berrojalbiz, N.; Agusti, S. Decrease in the abundance and viability of oceanic phytoplankton due to trace levels of complex mixtures of organic pollutants. Chemosphere 2010, 81 (2), 161–168. (11) Fussmann, G. F.; Loreau, M.; Abrams, P. A. Eco-evolutionary dynamics of communities and ecosystems. Funct. Ecol. 2007, 21 (3), 465–477. (12) Hillebrand, H.; Bennett, D. M.; Cadotte, M. W. Consequences of dominance: a review of evenness eﬀects on local and regional ecosystem processes. Ecology 2008, 89 (6), 1510–1520. (13) Hillebrand, H.; Matthiessen, B. Biodiversity in a complex world: consolidation and progress in functional biodiversity research. Ecol. Lett. 2009, 12 (12), 1405–1419. (14) Pomati, F.; Matthews, B.; Jokela, J.; Schildknecht, A.; Ibelings, B. W. Eﬀects of re-oligotrophication and climate warming on plankton richness and community stability in a deep mesotrophic lake. Oikos Accepted for publication. (15) Ibelings, B. W.; Chorus, I. Accumulation of cyanobacterial toxins in freshwater “seafood” and its consequences for public health: A review. Environ. Pollut. 2007, 150 (1), 177–192. (16) Chorus, I.; Bartram, J. Toxic Cyanobacteria in Water. A Guide to Public Health Consequences, Monitoring and Management; E & FN Spon, WHO: London, 1999. (17) Borja, A.; Bricker, S. B.; Dauer, D. M.; Demetriades, N. T.; Ferreira, J. G.; Forbes, A. T.; Hutchings, P.; Jia, X.; Kenchington, R.; Marques, J. C.; Zhu, C. Overview of integrative tools and methods in assessing ecological integrity in estuarine and coastal systems worldwide. Mar. Pollut. Bull. 2008, 56 (9), 1519–1537. (18) Xu, F.-L.; Tao, S.; Dawson, R. W.; Li, P.-g.; Cao, J. Lake Ecosystem Health Assessment: Indicators and Methods. Water Res. 2001, 35 (13), 3157–3167. (19) EC. Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 establishing a framework for Community action in the ﬁeld of water policy. Oﬀ. J. Eur. Communities, 2000; 2212-2000 L 327/1. (20) Dubelaar, G. B. J.; Geerders, P. J. F.; Jonker, R. R. High frequency monitoring reveals phytoplankton dynamics. J. Environ. Monit. 2004, 6 (12), 946–952. (21) Beninca, E.; Huisman, J.; Heerkloss, R.; Johnk, K. D.; Branco, P.; Van Nes, E. H.; Scheﬀer, M.; Ellner, S. P. Chaos in a long-term experiment with a plankton community. Nature 2008, 451 (7180), 822–825. (22) Litchman, E.; Klausmeier, C. A. Trait-based community ecology of phytoplankton. Ann. Rev. Ecol. Evol. Syst., 2008; 39, 615639. (23) Suding, K. N.; Lavorel, S.; Chapin, F. S.; Cornelissen, J. H. C.; DIAz, S.; Garnier, E.; Goldberg, D.; Hooper, D. U.; Jackson, S. T.; Navas, M.-L. Scaling environmental change through the community-level: A trait-based response-and-eﬀect framework for plants. Global Change Biol. 2008, 14, 1125–1140. (24) Reiss, J.; Bridle, J. R.; Montoya, J. M.; Woodward, G. Emerging horizons in biodiversity and ecosystem functioning research. Trends Ecol. Evol. 2009, 24 (9), 505–514. (25) Dubelaar, G. B. J.; Casotti, R.; Tarran, G. A.; Biegala, I. C. Phytoplankton and their Analysis by Flow Cytometry. In Flow Cytometry with Plant Cells; Dolezel, J., Greilhuber, J., Suda, J., Eds.; Wiley-VCH Verlag GmbH & Co.: Weinheim, 2007. (26) Rutten, T. P. A; Sandee, B.; Hofman, A. R. T. Phytoplankton monitoring by high performance ﬂow cytometry: A successful approach? Cytometry Part A 2005, 64 (1), 16–26. (27) Sosik, H. M.; Olson, R. J.; Armbrust, E. V. Flow cytometry in phytoplankton research. In Chlorophyll a Fluorescence in Aquatic Sciences: Methods and Applications; Suggett, D. J., Prasil, O., Borowitzka, M. A., Eds.; Springer: Netherlands, 2010; Vol. 4, pp 171185. (28) Mieleitner, J.; Reichert, P. Modelling functional groups of phytoplankton in three lakes of diﬀerent trophic state. Ecol. Mod. 2008, 211 (34), 279–291. 9664

dx.doi.org/10.1021/es201934n |Environ. Sci. Technol. 2011, 45, 9658–9665

Environmental Science & Technology (29) Reynolds, C. S.; Huszar, V.; Kruk, C.; Naselli-Flores, L.; Melo, S. Towards a functional classiﬁcation of the freshwater phytoplankton. J. Plankton Res. 2002, 24 (5), 417–428. (30) Kruk, C.; Huszar, V. L. M.; Peeters, E. T. H. M.; Bonilla, S.; € Costa, L.; LURling, M.; Reynolds, C. S.; Scheﬀer, M. A morphological classiﬁcation capturing functional variation in phytoplankton. Freshwater Biol. 2010, 55 (3), 614–627. (31) UPDA (Uﬃcio Protezione e Depurazione Acque). Campagna 2007 e Rapporto quinquennale 20032007. In Ricerche sull’evoluzione del Lago di Lugano. Aspetti limnologici; Commissione Internazionale per la Protezione delle Acque Italo-Svizzere, 2008. (32) Pomati, F.; Posch, T.; Kraft, N. J. B.; Eugster, B.; Jokela, J.; Ibelings, B. W. Trait-based analysis of natural phytoplankton communities: spring bloom dynamics in Lake Zurich (Switzerland). In preparation. (33) Kaufman, L.; Rousseeuw, P. J., Finding Groups in Data: An Introduction to Cluster Analysis; Wiley: New York, 1990. (34) Legendre, P.; Legendre, L. Ecological data series. In Numerical ecology; Legendre, P., Legendre, L., Eds.; Elsevier, 1998; pp 637705. (35) Durand, M. D.; Olson, R. J. Contributions of phytoplankton light scattering and cell concentration changes to diel variations in beam attenuation in the equatorial paciﬁc from ﬂow cytometric measurements of pico-, ultra and nanoplankton. Deep-Sea Res., Part II 1996, 43 (46), 891–906. (36) Vaulot, D.; Marie, D. Diel variability of photosynthetic picoplankton in the equatorial Paciﬁc. J. Geophys. Res., C: Oceans Atmos. 1999, 104 (C2), 3297–3310. (37) Binder, B. J.; DuRand, M. D. Diel cycles in surface waters of the equatorial Paciﬁc. Deep-Sea Res., Part II 2002, 49 (1314), 2601–2617. (38) Li, W. K. W. From cytometry to macroecology: A quarter century quest in microbial oceanography. Aquat. Microb. Ecol. 2009, 57 (3), 239–251. (39) Litchman, E.; de Tezanos Pinto, P.; Klausmeier, C. A.; Thomas, M. K.; Yoshiyama, K. Linking traits to species diversity and community structure in phytoplankton. Hydrobiologia 2010, 653 (1), 15–28. (40) Walsby, A. E.; Avery, A.; Schanz, F. The critical pressures of gas vesicles in Planktorhrix rubescens in relation tothe depth of winter mixing in Lake Z€urich, Switzerland. J. Plankton Res. 1998, 20 (7), 1357–1375. (41) Harmon, L. J.; Matthews, B.; Des Roches, S.; Chase, J. M.; Shurin, J. B.; Schluter, D. Evolutionary diversiﬁcation in stickleback aﬀects ecosystem functioning. Nature 2009, 458 (7242), 1167–1170. (42) de Bello, F.; Lavorel, S.; Díaz, S.; Harrington, R.; Cornelissen, J.; Bardgett, R.; Berg, M.; Cipriotti, P.; Feld, C.; Hering, D.; Martins da Silva, P.; Potts, S.; Sandin, L.; Sousa, J.; Storkey, J.; Wardle, D.; Harrison, P. Towards an assessment of multiple ecosystem processes and services via functional traits. Biodiversity Conserv. 2010, 19 (10), 2873–2893. (43) Lindenmayer, D. B.; Likens, G. E. Adaptive monitoring: a new paradigm for long-term research and monitoring. Trends Ecol. Evol. 2009, 24 (9), 482–486. (44) Chesson, P. Mechanisms of maintenance of species diversity. Ann. Rev. Ecol. Syst. 2000, 31, 343–366. (45) Kraft, N. J. B.; Valencia, R.; Ackerly, D. D. Functional traits and niche-based tree community assembly in an Amazonian forest. Science 2008, 322 (5901), 580–582. (46) Hubbell, S. P. The Uniﬁed Neutral Theory of Biodiversity and Biogeography; Princeton University Press: Princeton, 2001. (47) Vos, R. J.; Hakvoort, J. H. M.; Jordans, R. W. J.; Ibelings, B. W. Multiplatform optical monitoring of eutrophication in temporally and spatially variable lakes. Sci. Total Environ. 2003, 312 (13), 221–243. (48) Fukami, T. Community assembly dynamics in space. In Community Ecology: Processes, Models and Applications; Verhoef, H. A., Morin, P. J., Eds.; Oxford University Press: Oxford, 2010; pp 4554. (49) Colwell, R. K.; Rangel, T. F. Hutchinson’s duality: The once and future niche. Proc. Natl. Acad. Sci. 2009, 106 (2), 19651–19658. (50) MacLeod, N.; Benﬁeld, M.; Culverhouse, P. Time to automate identiﬁcation. Nature 2010, 467 (7312), 154–155. (51) Preston, C. M.; Marin, R.; Jensen, S. D.; Feldman, J.; Birch, J. M.; Massion, E. I.; DeLong, E. F.; Suzuki, M.; Wheeler, K.; Scholin, C. A. Near real-time, autonomous detection of marine bacterioplankton

ARTICLE

on a coastal mooring in Monterey Bay, California, using rRNA-targeted DNA probes. Environ. Microbiol. 2009, 11 (5), 1168–1180. (52) White, E. P.; Ernest, S. K. M.; Adler, P. B.; Hurlbert, A. H.; Lyons, S. K. Integrating spatial and temporal approaches to understanding species richness. Philos. Trans. R. Soc., B 2010, 365 (1558), 3633–3643.

9665

dx.doi.org/10.1021/es201934n |Environ. Sci. Technol. 2011, 45, 9658–9665

ARTICLE pubs.acs.org/est

Effect of Particle Size on Droplet Infiltration into Hydrophobic Porous Media As a Model of Water Repellent Soil Christopher A. E. Hamlett,† Neil J. Shirtcliﬀe,†,§ Glen McHale,† Sujung Ahn,‡ Robert Bryant,‡ Stefan H. Doerr,‡ and Michael I. Newton.*,† † ‡

School of Science and Technology, Nottingham Trent University, Clifton Lane, Nottingham, NG11 8NS, United Kingdom College of Science, Swansea University, Singleton Park, Swansea, SA2 8PP, United Kingdom

bS Supporting Information ABSTRACT: The wettability of soil is of great importance for plants and soil biota, and in determining the risk for preferential ﬂow, surface runoﬀ, ﬂooding,and soil erosion. The molarity of ethanol droplet (MED) test is widely used for quantifying the severity of water repellency in soils that show reduced wettability and is assumed to be independent of soil particle size. The minimum ethanol concentration at which droplet penetration occurs within a short time (e10 s) provides an estimate of the initial advancing contact angle at which spontaneous wetting is expected. In this study, we test the assumption of particle size independence using a simple model of soil, represented by layers of small (∼0.22 mm) diameter beads that predict the eﬀect of changing bead radius in the top layer on capillary driven imbibition. Experimental results using a three-layer bead system show broad agreement with the model and demonstrate a dependence of the MED test on particle size. The results show that the critical initial advancing contact angle for penetration can be considerably less than 90° and varies with particle size, demonstrating that a key assumption currently used in the MED testing of soil is not necessarily valid.

’ INTRODUCTION The wettability of soil is of great importance for plants and soil biota, and in determining the risk for preferential ﬂow, surface runoﬀ, ﬂooding and soil erosion.13 There are a range of distinctive environmental conditions that can give rise to water repellent soil. It is well established that ﬁres can volatilize hydrophobic compounds in the vegetation, litter or soil and these vapors can then condense on the sandy particles producing a hydrophobic granular texture that can exhibit high levels of water repellency. Under these circumstances vegetation recovery can be delayed, which further increases rates of surface runoﬀ and erosion, and, on some slopes, the risk of debris ﬂows.4 Where land with naturally high levels of water repellency, such as eucalyptus forest, is cleared for farming, productivity can be aﬀected. This can be alleviated, but only at signiﬁcant cost to farmers, by mixing in substantial amounts of clay.5 Where gray water is used for irrigation, or soil has been used as a natural ﬁlter for wastewater disposal, crop productivity can be signiﬁcantly reduced due to the gradual increase in soilwater repellency.6 A less benign origin of soil or sediment water repellency is from hydrocarbon contamination. Such contaminated sites can show a long-term persistence in water repellency, which can sometimes re-establish itself after attempts to remediate the land.7 In these types of situations it is important to be able to monitor and classify water repellency. The molarity of ethanol droplet (MED) test,1 which is sometimes referred to as the critical surface tension2 and %ethanol3 r 2011 American Chemical Society

test is used widely to determine the severity of water repellency for soil and other porous or granular samples. It involves placing drops of aqueous ethanol solutions with decreasing surface tension on to diﬀerent areas of the sample surface until a solution of suﬃciently low surface tension is reached that just allows the drops to penetrate the soil within 310 s. The molarity (MED), concentration (%ethanol) of the solution, or the surface tension allowing the porous surface to be penetrated by the liquid, is then taken as being characteristic for that soil. This method has been shown to be quite reproducible and diagnostic of soilwater repellency, provided soil samples are reasonably dry, homogenized,and atmospheric conditions are controlled.3,810 The relationship between surface tension and the equilibrium contact angle, a concept that assumes there is contact angle hysteresis, is often described by Young’s equation. In the MED test it is assumed reducing the surface tension causes imbibition by reducing this contact angle to below 90° at which point a parallel walled capillary would spontaneously ﬁll. The surface tension of the solution that just penetrates the soil, γc (i.e., the critical surface tension for penetration), has been used to estimate the average surface energy of the soil.2,11 It should be noted that the critical surface tension deﬁned in this manner is not the same as Received: July 6, 2011 Accepted: October 19, 2011 Revised: October 18, 2011 Published: October 19, 2011 9666

dx.doi.org/10.1021/es202319a | Environ. Sci. Technol. 2011, 45, 9666–9670

Environmental Science & Technology the critical surface tension often referred to within surface science and typically obtained from a Zisman plot by extrapolating the results of contact angle measurements using a range of liquids to give an estimated surface tension at which a smooth ﬂat surface would be completely wetted.12,13 It has also been suggested14 that the (initial advancing) contact angle at the surface tension, which gives wetting into water repellent soil or other granular materials, is not 90° as often assumed, but is closer to 51° when the dominating forces are capillary,and a model of hexagonal close-packed spherical particles can be assumed. Experimental data suggested a contact angle of around 61°65° could describe the critical surface tension for penetration into water repellent sand. Given that the initial advancing contact angle for penetration can be considerably less than 90° according to these reports, it is important to assess whether the MED test is independent of particle size and whether there are consequences for the implied initial advancing contact angle. Here we extend this previous work by developing a model of the conditions under which a liquid will penetrate under capillary forces into a hexagonal symmetry pack of spherical particles (beads). This model uses a surface layer of beads that have a smaller radius than those upon which they rest, which themselves are in a close-packed arrangement. The model predicts that penetration will occur at a critical advancing contact angle for the liquid that depends on the ratio of the two sizes of beads. This implies that the advancing contact angle at the surface tension, which gives wetting into soil, can be above 51° when the surface layer of beads are smaller than subsequent layers. This implies that the MED test gives a critical advancing contact angle that is dependent on the arrangement of particles and their sizes. We then develop a systematic method of creating bead packs with the model geometry and use an MED test approach to assess the critical advancing contact angle at which ethanol solutions penetrate them. The experimental data is shown to follow the trend predicted by the model, but with penetration occurring at systematically lower concentrations corresponding to higher advancing contact angles. An implication of this work is that the MED test may give results that require subtle analysis to be able to classify the severity of water repellency for granular material such as soil.

’ GEOMETRIC MODEL To examine the eﬀect of particle size on capillary driven droplet inﬁltration into hydrophobic soil and granular systems sandy soil particles can be modeled using packs of spherical particles (beads). For the transition from a system in which water does not penetrate to one which it does, one can imagine the top (surface) layer composed of beads of a radius r laying on top of close-packed beads having a larger radius R, thereby introducing a loose-packing element to the surface layer arrangement (Figure 1a and b). This allows the separation between surface particles and the distance from the top of particles in the surface layer to the top of particles in the layer below to be altered while retaining a hexagonal symmetry of the top layer arrangement. This symmetry can be visualized by imagining a pyramid (tetrahedral) arrangement deﬁned by a bead in the top layer resting on the space deﬁned by three close-packed beads of the layer below (Figure 1). The base-to-apex height of the pyramid can be found from the geometry of Figure 1a,b. From consideration of the surface free energy, the condition for capillary driven imbibition of a liquid into the beads is when the wetting front of

ARTICLE

Figure 1. Schematic representations of bead packing of a surface layer on top of a layer of close packed beads (side view (a) and top view (b)), equations relating the relevant lengths and conﬁguration of bead beds investigated by MED tests (c).

an impinging liquid, touches the beads in the layer below; should this happen the liquid will then continue into the subsequent layers giving full penetration of the bead pack. This condition is met when the wetting front has an equilibrium position at a critical distance (dc) measured from the top of the beads in the top layer to the top of the beads in the layer below. Assuming that the liquid has a horizontal meniscus as it bridges between three adjacent beads deﬁning a pore, this has an associated critical angle of contact (θc), at the liquid/solid/vapor phase boundary. The critical depth is determined by the comparative sizes of the beads in the top layer and subsequent layers of the bead pack provided only capillary forces are important and gravity can be ignored. From the geometry of Figure 1, the model predicts, 2sﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 3 2 R r 4 15 1 þ ð1Þ cos θc ¼ 4 r R 3 For eq 1 to be valid, capillary forces must dominate over gravity, and this requires the separations between beads to be signiﬁcantly less than the capillary length of the liquid k1=(γ /Fg)1/2, where γ is the surface tension of the liquid, F is its density,and g = 9.81 ms2 is the acceleration due to gravity. In this model, any liquid with an advancing contact angle, θA, below that of the critical angle contact will penetrate into the bead pack. In the case of a uniform particle size throughout the particle bed (i.e., r/R = 1) the predicted critical angle is 50.73°, which is consistent with previous work reported in the literature.14,15

’ EXPERIMENTAL METHODS Glass beads of diﬀerent sieve fractions in the size ranges 0.180.21 mm up to 1.82.0 mm (General Purpose Glass Microspheres, Whitehouse Scientiﬁc, a full list of sizes are included in the Supporting Information), comparable to sizes found in sandy soils, were immersed in HCl (30 vol. %) for 24 h and then rinsed with UHQ H2O (resistivity = 18 MΩ 3 cm1) and dried for 4 h at 110 °C. The hydrophilic glass beads were then immersed in chlorotrimethylsilane (CTMS) (2 vol.% in toluene; CTMS purchased from Aldrich) for 48 h at room temperature then rinsed with toluene and allowed to air-dry. CTMS was chosen because it provides a high contact angle to solutions of ethanol and persistent repellency on contact with the 9667

dx.doi.org/10.1021/es202319a |Environ. Sci. Technol. 2011, 45, 9666–9670

Environmental Science & Technology

ARTICLE

angle (on CTMS modiﬁed glass surfaces). This can be achieved by creating interpolation equations through the data, θA ðcÞ ¼ 0:0000126c3 þ 0:0029998c2 0:090953c þ 87:022

ð2Þ

and 10 4 γ1 c þ 5:1594 108 c3 EtOH ðcÞ ¼ 1:2804 10

7:2619 106 c2 þ 6:5721 104 c þ 1:3954 102 ð3Þ Figure 2. Advancing contact angle (measured on CTMS modiﬁed glass slides) and surface tensions of ethanol solutions in water used for MED tests (Literature values shown are taken from ref 12).

liquid, thereby being suitable for MED type experiments seeking to determine the initial advancing contact angle. Bead packs were constructed by placing the glass beads into a triangular template etched into laser-cut acrylic sheet and agitated until the beads formed a close packed monolayer (Figure 1c). A layer (∼110 μm thick) of polyurethane adhesive (1A33, Humiseal) was applied to a glass microscope slide, which was then placed on top of the beads, removed and a thin triangular acetate frame placed around the beads. The adhesive was then cured at 80 °C for 16 h ﬁxing this initial layer of beads in place and so providing a hexagonal packing symmetry for registration of subsequent layers of beads. A second acetate frame, with a slightly smaller triangular hole than the ﬁrst, was then stuck to the ﬁrst frame using double sided tape and a second layer of beads was poured into the frame and agitated until close packed. This layer of beads therefore registered with the ﬁrst layer, but was loose and not ﬁxed by any adhesive. This process was repeated for a third bead layer but this time using a frame of approximately half the bead thickness so the top of the third layer was exposed. This method of constructing a three-layer bead pack ensured a ﬁxed base layer with a hexagonal close-packed structure acting as a template to ensure registration of subsequent layers of beads, which themselves did not include adhesive bonding that might interact with a penetrating liquid in the experiments (Figure 1). We also conducted tests with bead packs with more than three layers (top layer with beads of radius r and other layers with beads of radius R) and did not ﬁnd any signiﬁcant diﬀerences in the ethanol concentrations at which penetration began. The surface tensions, γetOH, of ethanol solutions, for use in the MED tests, were measured using a Du Nouy ring at 25 °C. The corresponding advancing contact angle, θA, on a CTMS treated ﬂat glass microscope slide was measured using a Kr€uss DSA 10 contact angle meter by depositing a droplet of ethanol solution and increasing its volume to 20 μL at a rate of 20 μL 3 min1. The observed contact angle was measured using Kr€uss DSA software and the value just prior to the droplet’s contact line moving was taken as the advancing contact angle. The surface tension of the ethanol solutions were consistent with those reported in the literature16 and exhibit a range of advancing contact angles suﬃcient to investigate a range of r/R values of up to 1 (θc = 50.73°)14,15 (Figure 2). Since an MED test uses a range of concentrations of ethanol to estimate the advancing contact angle at which an ethanol solution just penetrates a porous system it is useful to be able to transform numerically from ethanol concentration, c, to surface tension or advancing contact

where c is the ethanol concentration by volume percentage, θA is the advancing contact angle in degrees and γEtOH1 is the inverse surface tension in units of m mN1. The accuracy of these interpolation formulas compared to the data is shown in the Supporting Information and the surface tension interpolation predicts the data of Vasquez et al.16 to within 0.5%. An important parameter for knowing whether capillary forces dominate over gravitational forces is the capillary length, k1. In the MED test, both the surface tension and the density vary with ethanol concentration and so we also measured the changes in density and have constructed an interpolation formula for the (inverse) capillary length, kEtOH ðcÞ ¼ 3:9315 100 c4 þ 1:1489 106 c3 1:2949 104 c2 þ 7:6214 103 c þ 0:37077 ð4Þ where eq 4 gives the inverse capillary length in units of mm1. Thus, for a 0% v/v concentration of ethanol (i.e., pure water), the capillary length is k1 = 2.70 mm and as ethanol concentration increases the capillary length reduces, e.g, at 20% v/v ethanol the capillary length is k1 = 2.08 mm. MED tests were carried out by placing a single droplet (8 18 μL) of aqueous ethanol solution at various concentrations onto a bead pack with the use of a syringe controlled by a stepper motor. The imbibition time of the droplet into the bead pack was measured by the video system of the goniometer (frame rate of up to 25 fps). The drop volume depended on the concentration of EtOH in solution and a single droplet was used per MED test. To estimate the lowest concentration at which a solution penetrated within 10 s, we constructed a plot of imbibition time versus concentration of ethanol (see example MED curves in the Supporting Information). Steps of 5% in concentration were used over the wider range and this was narrowed to steps of 3% around the concentrations at which a solution penetrated a bead pack. For each bead pack we observed a step-like transition curve in whether or not penetration occurred as the ethanol concentration was increased. The value of concentration at the transition to imbibition provides estimates of the initial advancing contact angle, θAc, and the surface tension, γetOHc, for penetration via the interpolation eqs 2 and 3. The plots together with the interpolation equations also allow the uncertainty of these values to be quantiﬁed. The measured value of θAc can then be compared to the critical angle of contact, θc, from eq 1 predicted by the model on the basis of the relative bead sizes.

’ RESULTS AND DISCUSSION Figure 3 shows the measured threshold ethanol concentrations for penetration as a function of bead size for bead packs with r/R = 1 (i.e., top layer beads having the same diameter as the 9668

dx.doi.org/10.1021/es202319a |Environ. Sci. Technol. 2011, 45, 9666–9670

Environmental Science & Technology

Figure 3. Ethanol concentration and advancing contact angles for penetration as determined by MED tests, of various 3 layer bead packs and loose packed beds with r/R = 1. Solid circles show where bead lifting was observed and open circles where it was not observed. Solid diamond symbols show data from thick random loose-packed beds.

lower two layers) thereby investigating whether absolute bead size matters; the secondary y-axis shows the equivalent values of advancing contact angle θAc as deduced from eq 2. This tests whether or not the assumption that capillary forces are dominant is valid for the various bead sizes used to obtain r/R = 1. As bead size reduces the capillary forces can be suﬃciently strong that on contact with the lower surface of a droplet a bead is lifted up from the top layer of beads as the droplet surface is shaped by its surface tension. Bead lifting is caused by the strength of capillary forces relative to the force of gravity acting upon a loose bead and is related to the ability of droplets to encapsulate themselves with a shell of hydrophobic particles to create liquid marbles.17,18 The data for the three-layer close packed beads are shown as solid circles where bead lifting was observed and as open circles where it was not. In these experiments we also constructed and tested penetration of ethanol solutions into much thicker randomly packed beds of CTMS treated glass-beads from single size sieve fractions of beads, and these data points are shown as solid diamond symbols. From this ﬁgure we can see that the threshold ethanol concentration, and hence critical advancing contact angle θAc, increases as bead size increases. This is consistent with geometrical considerations for a hexagonal arrangement of spherical beads, which show that the radius of the meniscus, rgap, between three close-packed beads deﬁning a pore at an angle of contact θc is rgap = 0.866R(1 rsinθc/0.866R) so that the assumption rgap/k 1,1 needed for gravity to be ignored fails as the bead size increases. At larger bead sizes the results show some scatter with contact angles between 65° and 85° and this probably arises from small imperfections in the shape and monodispersity of the beads creating defects in the hexagonal symmetry of the packing and, hence, larger gaps through which imbibition can commence. The results of the MED tests on close packed hydrophobic beads with a range of r/R values are shown in Figure 4 with a comparison to the theoretical model (dashed line) and ﬁt to the data (solid line). Similar to Figure 3, eq 2 has been used to present that data with two y-axes thereby allowing a simple comparison between the percentage of ethanol and the equivalent advancing contact angle at which penetration occurs. This data contains data from two types of bead packs constructed using (i) only monodisperse beads and (ii) beads of sieved ranges (open circles where the beads showed no lifting and solid circles where bead lifting was observed). Details of the sieved ranges can

ARTICLE

Figure 4. Ethanol concentration and advancing contact angles for penetration into three-layer bead packs of varying r/R as determined by MED tests. Solid circles show where bead lifting was observed and open circles where it was not observed. The dashed line shows the theoretical curve and the solid line the ﬁt to the experimental points.

be found in the Supporting Information. The trend of the data follows that of the theoretical curve with the value of the critical advancing contact angle θAc, decreasing signiﬁcantly below 90° as the r/R ratio increases toward unity although the data tends to lie slightly above the theoretical curve. This may result from any defects in bead packing associated with variation in shape and/or size of adjacent beads within the packs producing defects and larger gaps between beads compared with model predictions and capillary length. The approach in this work uses a hexagonal symmetry packing model with hydrophobic spherical beads and is valid only under conditions where capillary forces are dominant. It does, however, provide a test of the prediction that under these conditions complete penetration occurs when the meniscus of the liquid advancing down the loose-packed surface layer comes into contact with particles from the layer below. This is a general principle for capillary driven penetration that should be applicable to other types of packing and to nonspherical particles. Should pores in the system approach in size to the capillary length or large defects with substantial pore size be present, penetration will occur at much lower ethanol concentrations. Similarly, the model does not include eﬀects of a hydrostatic pressure. The model is intended to improve understanding of the MED test, which itself provides an estimate only of the initial advancing contact angle. In any system where the hydrophobicity of the particle coating changes over time on contact with water, the test will not provide an accurate indication of the length of time over which soil will remain repellent as indicated in previous work empirically8 and by direct observation of soil particle coatings.19 Similarly, prolonged exposure will lead to adsorbed vapors on particles and this may itself cause changes to the contact angle and hence lead to penetration of water. This work shows that the critical initial advancing contact angle for penetration into particle beds taken to represent sandy soil can be considerably less than 90° and varies with particle size. If soils of diﬀerent particle size distributions are compared, the critical initial advancing contact angle θAc is likely to vary although some trends with both absolute size and packing of the layers can be expected. The results demonstrate that the widely held assumption11 that a liquid will just enter a porous substrate when it has a contact angle of 90° is not necessarily valid. This has important implications for evaluating the wettability 9669

dx.doi.org/10.1021/es202319a |Environ. Sci. Technol. 2011, 45, 9666–9670

Environmental Science & Technology of soils and other granular materials. Some soils or granular materials previously classiﬁed as fully wettable, based on MED, surface tension or %ethanol tests, may in fact, exhibit a signiﬁcant resistance to wetting, which in turn bear some of the environmental implications typically associated with the presence of water repellency.

’ ASSOCIATED CONTENT

bS

Supporting Information. (1) Table showing bead size and source, (2) example graph of MED data, (3) graph of data and interpolation for advancing contact angle as a function of ethanol concentration, (4) graph of data and interpolation for 1/surface tension as a function of ethanol concentration (5) graph of data and interpolation for 1/capillary length as a function of ethanol concentration. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: +44 115 8483365; fax: +44 115 8486636; e-mail: [email protected]. Present Addresses §

ARTICLE

(10) Roy, J. L.; McGill, W. B. Assessing soil water repellency using the molarity of ethanol droplet (MED) test. Soil Sci. 2002, 167 (2), 83–97. (11) Letey, J.; Carrillo, M. L. K.; Pang, X. P. Approaches to characterize the degree of water repellency. J. Hydrol. 2000, 231, 61–65. (12) Zisman, W. A. Relation of equilibrium contact angle to liquid and solid constitution. In Contact Angle Wettability and Adhesion. Advances in Chemistry Series; R. F.. Gould, ed.; American Chemical Society: Washington, DC, 1964; Vol. 43, pp 151. (13) Adamson, A. W. & Gast, A. Physical Chemistry of Surfaces; Wiley-Blackwell, 1997. (14) Shirtcliﬀe, N. J.; McHale, G.; Newton, M. I.; Pyatt, F. B. Critical conditions for the wetting of soils. Appl. Phys. Lett. 2006, 89 (9), 094101. (15) Ban, S.; Wolfram, E.; Rohrsetzer, S. The condition of starting of liquid imbibition in powders. Colloids Surf. 1987, 22 (24), 301–309. (16) Vazquez, G.; Alvarez, E.; Navaza, J. M. Surface tension of alcohol plus water from 20 degrees C to 50 degrees C. J. Chem. Eng. Data. 1995, 40 (3) 611614. (17) McHale, G.; Shirtcliﬀe, N.,J.; Newton, M.,I.; Pyatt, F.,B.; Doerr, S. H. Self-organization of hydrophobic soil and granular surfaces. Appl. Phys. Lett. 2007, 90 (5), 054110. (18) McHale, G.; Newton, M.,I. Liquid marbles: principles and applications. Soft Matter 2011, 7 (12), 5473–5481. (19) Cheng, S.; Bryant, R.; Doerr, S. H.; Wright, C. J.; Williams, R. Investigation of surface properties of soil particles and model materials with contrasting hydrophobicity using atomic force microscopy. Environ. Sci. Technol. 2009, 43 (17), 6500–6506.

Now at Rhine-Waal University of Applied Sciences, Germany.

’ ACKNOWLEDGMENT We thank the UK Engineering and Physical Sciences Research Council (EPSRC) for funding CAEH, SA and NJS under grants EP/H000704/1, EP/H000747/1 and EP/E063489/1 respectively. ’ REFERENCES (1) King, P. M. Comparison of methods for measuring severity of water repellence of sandy soils and assessment of some factors that aﬀect its measurement. Aust. J. Soil Res. 1981, 19 (4), 275–285. (2) Watson, C. L.; Letey, J. Indices for characterizing soil-water repellency based upon contact angle-surface tension relationships. Soil Sci. Soc. Am. J. 1970, 34 (6), 841–844. (3) Dekker, L. W.; Ritsema, C. J. How water moves in a water repellent sandy soil. 1. Potential and actual water repellency. Water Resour. Res. 1994, 30 (9), 2507–2517. (4) DeBano, L. F. The role of ﬁre and soil heating on water repellency in wildland environments: A review. J. Hydrol. 2000, 231, 195–206. (5) Cann, M. A. Clay spreading on water repellent sands in the South East of South Australia—Promoting sustainable agriculture. J. Hydrol. 2000, 231, 333–341. (6) Mataix-Solera, J.; García-Irles, L.; Morugan, A.; Doerr, S. H.; Garcia-Orenes, F.; Arcenegui, V.; Atanassova, I. Longevity of soil water repellency in a former wastewater disposal tree stand and potential amelioration. Geoderma 2011, 165 (1), 78–83. (7) Roy, J. L.; McGill, W. B. Investigation into mechanisms leading to the development, spread and persistence of soil water repellency following contamination by crude oil. Can. J. Soil Sci. 2000, 80 (4), 595–606. (8) Doerr, S. H. On standardizing the ’water drop penetration time’ and the ’molarity of an ethanol droplet’ techniques to classify soil hydrophobicity: A case study using medium textured soils. Earth Surf. Proc. Landforms 1998, 23 (7), 663–668. (9) Doerr, S. H.; Dekker, L.; W. Ritsema, C. J.; Shakesby, R. A.; Bryant., R. Water repellency of soils: The inﬂuence of ambient relative. Soil Sci. Soc. Am. J. 2002, 66 (2), 401–405. 9670

dx.doi.org/10.1021/es202319a |Environ. Sci. Technol. 2011, 45, 9666–9670

ARTICLE pubs.acs.org/est

Characterization of Aquatic Particles by Direct FTIR Analysis of Filters and Quantification of Elemental and Molecular Compositions Luc Tremblay* and Ghita Alaoui Department of Chemistry and Biochemistry, Universite de Moncton, Moncton, New Brunswick, Canada E1A 3E9

Marc N. Leger Department of Chemistry, St. Francis Xavier University, Antigonish, Nova Scotia, Canada B2G 2W5

bS Supporting Information ABSTRACT: This paper presents the ﬁrst characterization of aquatic particles and particulate organic matter (POM) by attenuated total reﬂectance infrared spectroscopy (ATR-FTIR) using particles deposited on ﬁlters. Particles from 30 water samples from the St. Lawrence System (Canada) were analyzed. ATR-FTIR spectra revealed changes in numerous organic and inorganic functional group contents. Particles from marine waters contained POM enriched in amide, NH, and aliphatic groups, while terrigenous POM had more COO/COOH and aromatic groups. The spectra showed the selective degradation of amide, NH, aliphatic, and carbohydrate-like structures during the sinking of the particles. Partial least-squares (PLS) regression of the ATR-FTIR spectra was used to quantify 12 important elemental and molecular parameters, such as amino acids, bacterial biomarkers, and degradation indices. Most parameters were quantiﬁed with good accuracy compared to conventional methods (<15% error). The spectral regions leading to the best quantiﬁcations and the PLS loadings revealed that aromatic cycles, other unsaturated structures, and COO/COOH groups were degraded at a much slower rate than N-molecules, such as amino acids, and carbohydrates. Marine POM was enriched in CH3 groups. CH3 groups appeared highly labile and abundant in bacterial POM. ATRFTIR represents a new and powerful method for a rapid, inexpensive, and nondestructive characterization of particles collected by ﬁltration revealing important biogeochemical processes involving POM.

’ INTRODUCTION The particulate matter found in natural waters plays several important roles in the cycles of vital elements such as carbon, nitrogen, and oxygen. Particulate organic matter (POM) represents a source of energy and nutrients for the organisms of the water column and sediments. In stratiﬁed waters, the sinking of POM toward deep waters, known as the “biological pump”, acts as a carbon sink. Every year, rivers bring approximately 2 1014 g of particulate organic carbon toward oceans; its fate during the transit in estuaries and coastal zones is still unclear.13 In addition, POM has a strong aﬃnity for pollutants and thus inﬂuences their mobility, toxicity, and bioavailability in ecosystems.4 POM composition inﬂuences each of these processes. Unfortunately, the study of POM fate, and processes involving POM, is limited by the lack of knowledge of POM composition.5 Following the death of the organisms, most of the POM that survived mineralization quickly becomes unrecognizable by current separation/identiﬁcation techniques, indicating that original biomolecules (e.g., proteins, carbohydrates, lipids) undergo alterations.5 It is therefore important to develop techniques that allow characterization of this complex mixture. Ideally, these techniques should be able to analyze bulk POM in its natural matrix. r 2011 American Chemical Society

Spectroscopic techniques such as nuclear magnetic resonance (NMR)6 and Fourier transform infrared (FTIR)79 spectroscopy can be used for quantitative functional group analyses of bulk POM. These two techniques provide complementary information, as they are sensitive to diﬀerent functional groups. FTIR has several advantages over NMR; it is not subject to interferences or sensitivity reduction from minerals,10,11 and it is much faster (<5 min per sample), simpler, and cheaper. FTIR spectra can be obtained in transmission and reﬂectance modes. Reﬂectance-mode techniques, in particular attenuated total reﬂectance (ATR), are becoming more popular because they require minimal to no sample preparation, allow the analysis of opaque samples, and are reproducible.12 ATR can be used to collect spectra of thin surfaces. In ATR, the path length of the infrared beam in the sample, or evanescent wave, is up to a few micrometers when the sample is in contact with a crystal having a high refractive index13 (Supporting Information, Figure S1). This characteristic makes ATR the only FTIR sampling mode suitable to collect spectra from particles deposited on a opaque Received: July 27, 2011 Accepted: October 3, 2011 Revised: September 23, 2011 Published: October 03, 2011 9671

dx.doi.org/10.1021/es202607n | Environ. Sci. Technol. 2011, 45, 9671–9679

Environmental Science & Technology ﬁlter. Since ATR-FTIR is nondestructive, ﬁlters can be reused after analysis. ATR-FTIR has been used with ﬁlters or membranes in studies of atmospheric aerosols1416 and of membrane fouling or ﬁltration performance in water treatments.1719 However, it appears to the best of our knowledge that this technique has not been used to characterize POM or used in a geochemical study of particles collected by routine ﬁltration of natural waters. The general objective of this study was to use ATR-FTIR to characterize the particulate matter and POM collected by the ﬁltration of 30 water samples from the St. Lawrence System (Canada). The sampling was done in the river, estuary, and gulf sections of the St. Lawrence, as well as in six of its major tributaries. A more speciﬁc objective was to assess the potential of partial least-squares (PLS) regression to provide qualitative and quantitative information on particle composition, and on geochemical processes involving POM. PLS can reveal information, hidden in the numerous overlapping bands of the FTIR spectra, without the need for a priori band assignment. Once a calibration model is generated with samples of known composition, PLS can then be used to quantify the same parameters in samples of unknown composition. Important markers of POM origin and diagenetic state, or transformations, were studied in this work: organic carbon (%C) and nitrogen (%N) contents; the atomic ratio of nitrogen to carbon (N:C); total hydrolyzable amino acids (THAA) and muramic acid (Mur) contents; degradation index (DI); proportions of organic carbon and nitrogen from amino acids (%CAA and %NAA), proportion of D-enantiomers of amino acids relative to THAA (%D-AA), mole percentage of individual amino acids including nonproteic amino acids and D-AA. The heterogeneity of the sample set oﬀers a challenging test for the ATR-FTIR and PLS methods. PLS has been coupled with FTIR spectroscopy to analyze various soil or sediment properties.7,9,2022 However, these studies used diﬀuse reﬂectance (DRIFT) for spectral acquisition and did not quantify molecular parameters, such as amino acids, DI, and muramic acid.

’ EXPERIMENTAL SECTION Study Sites and Sampling. The St. Lawrence is the secondlargest river system in North America, extending over 1197 km, covering an area of about 11 000 km2, and collecting the waters of a drainage basin of approximately 106 km2.23 In May 2007, nine stations were sampled along the St. Lawrence salinity gradient, from the river end-member near the city of Quebec to the marine end-member in the Gulf of St. Lawrence. Terrigenous inputs of POM are considered negligible in the gulf where all the POM appears of marine origin.23 During the same cruise, four stations in the Saguenay Fjord, the most important tributary of the St. Lawrence, were also sampled. A detailed description of these sites can be found in a previous study.24 Five other tributaries of the St. Lawrence were sampled in May 2008, more specifically the Montmorency, Chaudiere, Bersimis, Outardes, and Manicouagan rivers. These rivers were sampled near their mouths. A map of the sampling sites can be found in the Supporting Information (Figure S2). Particles from 30 water samples were sampled. The waters from the St. Lawrence and the Saguenay Fjord were sampled with 12 L Niskin bottles installed on a rosette equipped with a CTD probe (Sea-Birds Electronics Model 9plus). The sampling was done at diﬀerent depths in order to sample the diﬀerent water

ARTICLE

layers. The water samples from the ﬁve other rivers were collected near the surface with a clean polyethylene bucket. Water samples were then ﬁltered on combusted (500 °C, 16 h) and preweighted 47 mm GF/F ﬁlters (glass microﬁbers, binder free, 0.7 μm porosity, Whatman Co.). This type of ﬁlter is opaque to IR light. The water samples were frequently shaken before and during ﬁltration to ensure an homogeneous deposition. The ﬁlters were then rinsed three times with deionized water to remove the salts. Each water sample was simultaneously ﬁltered on four or ﬁve ﬁlters using a multiposition ﬁltration manifold, allowing parallel vacuum ﬁltration. The ﬁlters were preserved at 20 °C before freeze-drying. They contained 5 25 mg of particles (dry weight) depending on samples. Only one ﬁlter per sample was analyzed by ATR-FTIR. Spectroscopic Analysis. A Varian Scimitar 1000 Fourier transform infrared (FTIR) spectrometer was used. A Pike MIRacle ATR accessory equipped with a single-reflection diamond crystal (Supporting Information, Figure S1) was placed in the measurement chamber. The spectra were collected in the mid-IR range, (5004000 cm1) by averaging 200 scans with a 4 cm1 resolution. Before acquisition of the particle spectra, a background spectrum was collected with the same operational parameters. A new background spectrum was collected after every 2 h of analysis. The measurement area of the ATR accessory was purged with nitrogen, before and during all spectral acquisitions. Sample analysis was done by placing the freeze-dried filter directly onto the diamond crystal having a diameter of ∼3 mm. Maximal pressure, as set by Pike Co., was then applied to ensure good contact between the particles on the filter and the crystal. No significant alteration to the small area of sample in contact with the crystal was observed. As with other spectroscopic methods, the measured absorption depends on the optical path length. With ATR, the penetration depth of the evanescent wave increases with the wavelength (or with decreasing wavenumbers).13 As a result, the relative intensity of the absorption bands of spectra collected by ATR and transmission modes can be diﬀerent. Though the penetration depth is small, the diameter of the incident beam is much larger (∼1 mm). As a result, the signals from all the particles included in the irradiated surface are averaged. The quantity of particles (525 mg) on the ﬁlters produced a sample that was 212 μm thick, considering a measured density (particle compacted in air) of 1 g cm3. This is more than the eﬀective path length of the ATR evanescent wave measured for a diamond/ZnSe crystal.13 Elemental and Molecular Characterizations. Elemental and molecular data of the samples coming from the St. Lawrence salinity gradient and the Saguenay Fjord have been presented for other purposes in a previous study.24 These values were used here to build PLS calibration models. Organic carbon (%OC) and total nitrogen (%N) contents of the ﬁltered particles were measured with Costech ECS 4010 and an Elementar Micro elemental analyzers. Before analyses, carbonates were eliminated under HCl vapor.25 Calibration was done using acetanilide as a standard. The quantiﬁcation of L- and D- enantiomers of amino acids (L-AA, D-AA) and of muramic acid in the ﬁltered particles was done by acid hydrolysis, followed by reversed-phase HPLC.24,26 Brieﬂy, the freeze-dried ﬁlters were placed in 6 mol L1 HCl for 20 h at 110 °C for AA hydrolysis, or 3 mol L1 HCl for 5 h at 100 °C for Mur hydrolysis. Automated precolumn derivatization was performed before reversed-phase HPLC separation. The separation was done using an Agilent HPLC system (model 9672

dx.doi.org/10.1021/es202607n |Environ. Sci. Technol. 2011, 45, 9671–9679

Environmental Science & Technology 1200) equipped with a ﬂuorescence detector. A Merck C-18 LiChrosphere 100 RP-18 (4 250 mm, 5 μm) column was used for AA analysis while a Merck C-18 Supersphere 100 RP18 (4 125 mm, 4 μm) column was used for Mur analysis. Chemical racemization that occurr during hydrolysis was corrected using the method described by Kaiser and Benner.26 Asparagine (Asn) and glutamine (Gln) are converted to aspartic acid (Asp) and glutamic acid (Glu), respectively, during hydrolysis. In this work, “Asx” refers to both Asn and Asp, while “Glx” refers to both Gln and Glu. Degradation indices (DI), developed by Dauwe et al.27 based on changes in AA composition during POM degradation or diagenesis, were calculated. A high DI value indicates relatively fresh POM while a low DI value indicates altered POM. Partial Least-Squares Regression. A description of the PLS methods in the present context can be found in our previous study,22 while a more theoretical background can be found in the paper by Haaland and Thomas.28 Briefly, PLS is a chemometric method used to generate a regression model between spectra and defined properties of each sample. A set of n calibration spectra measured at m different wavenumbers, with known values for the parameter of interest, are laid into an n m spectral matrix. The PLS algorithm decomposes the spectral matrix into a series of latent variables (LVs), which consist of loadings (essentially describing a new axis system) and scores (the coordinates of each spectrum along that new axis system). The scores are used directly in the PLS regression model, while the loadings offer information about spectral characteristics that are correlated with the properties of interest. Loadings can be visualized in the same way as spectra and thus can be quite useful in assigning spectral bands in complex mixtures such as POM and in highlighting geochemical processes.22 The first few latent variables contain all useful information within the spectra, and they are retained to generate the calibration model. New samples of unknown composition are quantified by projecting their spectra onto the regression model. In the present case, a separate regression model was generated for each parameter of interest. PLS regression was done with GRAMS/AI software (version 8.0, Thermo Scientiﬁc) and its PLSplus/IQ add-on. The ﬁrst step was to build a data set containing the 30 sample spectra, along with the known values for the parameters to be quantiﬁed. The PLS-1 calibration and cross-validation modes were selected, and all spectra were mean-centered. The spectral regions used for quantiﬁcation were chosen with the “correlation” tool, which selects regions that appear to be best correlated to the parameter. This selection was later validated by calculating cross-validation error (see eq 1 below) and selecting regions that led to the lowest error. Among the calibration modes available in the software, the “quantiﬁcation” type was selected. Some samples were removed from some sample sets, when values were below or near detection limits (i.e., %N, Mur); in other rare occasions, outliers were detected and removed. For this reason, sample sets of less than 30 samples were used for some parameters. Atomic N:C ratio and %CAA are C-normalized parameters. For these parameters, predictions were done on C-normalized spectra, which consist of absorbance values divided by the sample’s %OC. This ensures the absorbance values are representative of the POM composition instead of the POM concentration in the particles made of mineral and organic materials.

ARTICLE

Cross-validation was used on a sample set of N samples: the ﬁrst sample was removed from the sample set, and a calibration model was built with the other N 1 samples. The parameter value for the left-out sample was then predicted while retaining diﬀerent numbers of latent variables (115 in the present study). The left-out sample was then replaced, the second sample was removed, and a new calibration model was built with these N 1 samples. This process was repeated until all N samples had been left out once. The best number of retained latent variables was chosen by minimizing the relative root-mean-squared error of cross-validation, or %RMSECV, calculated by the following equation:29,30 11=2 N ∑ ðc ^ c Þ i C Bi ¼ 1 i C ðcmax cmin Þ1 %RMSECV ¼ 100B @ A N 0

ð1Þ

where ci represents the actual value (e.g., %OC measured by elemental analyzer) for the ith sample; ^ci indicates the values predicted by ATR-PLS for that sample; and cmax and cmin represent the maximum and minimum actual values within the data set. Another approach, used to validate the chosen model, was to build a calibration model with samples from all stations except for one. The samples from the left-out station were then predicted. The relative root-mean-squared error of prediction (%RMSEP) was calculated using eq 1 with only samples from the prediction set.

’ RESULTS AND DISCUSSION ATR-FTIR Spectra. Figure 1 presents five typical ATR-FTIR spectra from our sample set. The region between 4000 and 1500 cm1 was enlarged because of the lower ATR optical path length at higher wavenumbers.13 Table 1 shows some characteristics of these five samples. Their spectra revealed both similarities and important differences in the composition of filtered particles. Absorption band assignments of organic functional groups were based on the work of Bellamy.31 A wide and intense band centered around 1000950 cm1 was apparent in nearly all ATR-FTIR spectra. This band is attributed to the mineral matrix, more speciﬁcally to vibrations from SiO and Si(Al)O.32 Another mineral band was observed around 3625 cm1, especially in particles with low POM such as those from the Chaudiere river. That band corresponds to OH stretching of aluminosilicates (SiOH + AlOH).33 In contrast, the spectra of particles from the surface waters of the Gulf of St. Lawrence did not show signiﬁcant mineral bands, as these particles were mostly organic (%OC > 25%, Table 1). In these cases, the mineral band at 998 cm1 was much weaker and was surpassed by a band centered at 1065 cm1 attributed to vibrations from CO and CH of carbohydrates and polysaccharides. These sugars come from the relatively high productivity of the gulf’s surface waters in spring.34 The absorption band at ∼2160 cm1 is typical of a bending vibration mode of adsorbed water. Organic OH groups were responsible for most of the absorption in the 35002600 cm1 region, with a maximum generally located near 34503350 cm1. The broadness of this band is caused by signiﬁcant intramolecular hydrogen bonding. NH groups also absorb in this region, but the corresponding 9673

dx.doi.org/10.1021/es202607n |Environ. Sci. Technol. 2011, 45, 9671–9679

Environmental Science & Technology

ARTICLE

band is generally centered at lower wavenumbers (e.g., ∼3200 cm1 for amino acids). ATR-FTIR was sensitive enough to measure a shift of the peak maxima of the OH/NH band in samples showing contrasting nitrogen contents. Marine particles (i.e., from the gulf), with %N more than 5 times higher than the other samples (Table 1), showed a peak maximum at ∼3290 cm1, while this maximum was at >3350 cm1 for the other samples (Figure 1). It was also possible to detect other changes related to particle POM content (or %OC), %N, and amino acid yield (%CAA). The bands centered at 2923 and 2845 cm1 correspond to

asymmetric and symmetric CH2 stretching, respectively. These bands were more intense in POM-rich particles (Figure 1, Table 1). This correlation has been reported in a previous study with sediments.8 The particles from the surface waters of the gulf also exhibited a band at 2962 cm1 caused by the asymmetric stretching of CH3. These particles may not only contain more POM, but their POM appeared to be more aliphatic (and less aromatic, see below). The weak 1425 cm1 band can also be attributed to aliphatic groups (i.e., deformation vibrations). The amide I (CdO) and II (NH and CN) bands centered at ∼1645 and ∼1545 cm1, respectively, were more important in

Figure 1. ATR spectra of ﬁltered particles from three stations. Spectra from particles collected at three diﬀerent depths in the Gulf of St. Lawrence are presented. A description of these sites is in Table 1.

Table 1. Salinity, Particle Concentration, and Elementary and Amino Acid Compositions of the Particles Associated with the Spectra Shown in Figure 1 sample sitesa Chaudiere river St. Lawrence river

salinity (g kg1)

particles (mg L1)

<0.5

70.0

0.6

12.6

%N

2.01

0.27

8.7

8.2

1.0

12.4

13.0

10.8

atomic C:N

%CAAb

%OC

St. Lawrence gulf 2 m

31.9

2.8

32.3

5.6

6.8

15.1

20 m

32.7

1.3

25.1

4.0

7.3

13.8

300 m

34.9

1.2

15.3

1.8

13.3

2.4

Chaudiere river: 46°400 3700 N/71°130 5100 W. St. Lawrence river (station 1): 46°440 4400 N/71°170 3800 W. St. Lawrence gulf (station 10): 49°250 1200 N/ 64°4500 2300 W. A map of these sites is available in the Supporting Information (Figure S2). b Percentage of total organic carbon (OC) from amino acids. a

9674

dx.doi.org/10.1021/es202607n |Environ. Sci. Technol. 2011, 45, 9671–9679

Environmental Science & Technology

ARTICLE

Table 2. Results from PLS Cross Validation for Various Elementary and Molecular Parameters of Filtered Particles parameter

Na

LVb

R2

%RMSECV

value range

selected spectral regions (cm1)

%OC (dry weight)

30

4

0.863

10.5

2.0132.3

%N (dry weight)

28

2

0.928

6.75

0.145.6

13001125 35503060 + 1364985

atomic N:C

27

13

0.882

12.0

0.040.15

2168985

THAA (nmol mg1)c

29

7

0.936

9.15

23.4894

36003030 + 17881412

Mur (nmol mg1)d

24

4

0.932

8.64

0.0110.97

35683019 + 2193563

DIe

30

5

0.842

11.2

0.131.43

23081854

%CAAf

28

3

0.628

20.5

2.1917.6

35882835 + 18001415

%NAAg %G-Aba + %B-Alah

24 30

4 14

0.791 0.675

18.4 15.7

6.8250.8 0.3811.4

35123015 + 18001415 + 1312730 35283015 + 2282554

%D-AAi

30

4

0.680

12.2

1.14.0

1055567

D-Asx (nmol mg1)j

27

10

0.971

6.55

0.103.01

35533055 + 17731404

D-Glx (nmol mg1)k

27

5

0.726

14.3

0.202.62

35763060

a

Number of samples. b Number of retained latent variables. c Total hydrolyzable amino acid content per mg of dry particle. d Muramic acid content per mg of dry particle. e Degradation index.27 f Percentage of total organic carbon from amino acids. g Percentage of total nitrogen from amino acids. h Mole percentage of γ-aminobutyric acid and β-alanine relative to the other amino acids. i Mole percentage of the sum of D-amino acids relative to the other amino acids. j D-Asparagine and D-aspartic acid content per mg of dry particles. k D-Glutamine and D-glutamic acid content per mg of dry particles.

Figure 2. Validation plots for ATR-PLS quantiﬁcations versus values obtained by conventional methods for several parameters of ﬁltered particles. See Table 2 for the description of acronyms and abbreviations.

samples that contained more in situ produced nitrogen and amino acids such as the POM from the surface waters of the gulf (Figure 1, Table 1). Considering that particles from the Chaudiere and St. Lawrence rivers contained much less nitrogen and amino acids, the absorption in the 16501500 cm1 region of their spectra can be assigned to COO and CdC groups absorbing these frequencies as well. This is supported by the strong absorption bands at ∼2035 and ∼1975 cm1 in these terrigenous samples, which are caused by aromatic rings and other unsaturated structures. Terrigenous POM is generally more acidic and aromatic than marine POM.35,36 The ATR-FTIR spectra also showed important composition changes in samples from diﬀerent depths, revealing diagenetic transformations during particle sinking. Examples from the Gulf

of St. Lawrence are presented in Figure 1. All bands related to POM became less intense in samples collected at greater depths. These changes can be associated to decreasing %OC or POM degradation during particle sinking (Table 1).24 The decrease of band intensity was particularly marked for the three aliphatic bands between 2970 and 2840 cm1. This ﬁnding indicates a preferential degradation of these structures during particle sinking. The preferential degradation of amide groups was also observed by the decrease in amide bands (1645 and 1545 cm1) with depth (Figure 1). Moreover, the rapid degradation of NH groups caused a shift in the peak maximum of the broad OH + NH band. This maximum was at 3294 cm1 in the surface sample, but toward 3400 cm1 in the deepest sample as the 9675

dx.doi.org/10.1021/es202607n |Environ. Sci. Technol. 2011, 45, 9671–9679

Environmental Science & Technology

ARTICLE

Table 3. Results for the PLS Quantiﬁcation of Various Elementary and Molecular Parameters of Filtered Particles from One Station in the Gulf of St. Lawrencea parameter

a

Ncal

Npred

LV

mean %RMSEP

value range

%OC(dry weight)

27

3

4

35.4

15.232.3

%N (dry weight)

25

3

2

6.96

1.345.55 0.0750.15

atomic N:C

24

3

10

39.6

THAA (nmol mg1)

26

3

5

7.24

70.9894.3

Mur (nmol mg1) DI

21 27

3 3

3 4

29.2 26.7

0.170.98 0.971.43

%CAA

25

3

2

24.3

2.4115.1

%NAA

21

3

3

214

8.6126.9

%G-Aba + %B-Ala

27

3

13

67.9

0.991.44

%D-AA

27

3

7

59.3

1.061.54

D-Asx (nmol mg1)

24

3

13

11.9

0.263.01

D-Glx (nmol mg1)

24

3

1

56.8

0.292.62

See Table 2 for a description of acronyms and abbreviations.

absorption of OH increased relative to that of NH. These changes are supported by the measured decrease of %N and % CAA in deeper samples (Table 1). Finally, the degradation of carbohydrates and polysaccharides explains the shift of the band centered at 1065 cm1 (CO and CH) for surface particles, toward 998 cm1 (SiO and Si(Al)O) in the spectrum of the particles collected at 300 m (Figure 1). The absorption of some functional groups may also decrease by bond formation between POM and the mineral matrix during sinking. Amines and oxygenated functional groups are thought to be important agents in organo-mineral association.37,38 Qualitative analysis of the ATR spectra revealed important changes in POM composition according to their origins and during degradation. In summary, particles from marine waters contained POM enriched in amide, NH, and aliphatic groups, while terrigenous POM had more COO/COOH and aromatic groups. The spectra showed the selective degradation of amide, NH, aliphatic, and carbohydrate-like structures. These changes can be quickly revealed by a simple comparison of the general shape of the FTIR spectra. Chemometric Analysis. More structural and quantitative information on elemental and molecular composition can be extracted from the ATR spectra by using PLS regression. In a first experiment, PLS calibration models were built for numerous parameters by using values obtained via conventional methods. These values were then quantified using the ATR-FTIR spectra of the samples and cross-validation. Table 2 summarizes the results obtained. Figure 2 shows validation plots for PLS quantifications of key parameters (other plots are available in the Supporting Information, Figure S3). These results show that it was possible to quantify most tested parameters from the ATR spectra with a reasonable accuracy (i.e., less than 15% error). %RMSECV values ranged from 6.55% for D-Asx content to 20.5% for %CAA. The coefficient of determination (R2) for the linear regression of the validation plots ranged from 0.628 to 0.971 (Table 2). These results are similar to those obtained with diffuse reflectance FTIR spectroscopy for the PLS quantification of elemental and humic contents in 150 sediment samples.22 The precision of the values measured with conventional methods, as well as the number of samples in the calibration set, influences the

calculated accuracy of the PLS quantification. In the present case, the %OC and %N were much higher than in the sediments,22 and thus the values measured by the elemental analyzer were more precise. This likely counterbalanced the fact that fewer samples were analyzed here (i.e., 30 vs 150). Table 2 also lists the spectral regions leading to the best predictions for each parameter. Although it was possible to use the entire spectrum to quantify %OC with reasonable prediction errors, because most bands are associated with POM, the best spectral region to quantify %OC was found to be at 13001125 cm1. This shoulder of the intense mineral (and sugar) band can be associated to ester, ether, and phenol groups. Not surprisingly, the best spectral regions to quantify %N and N-containing compounds (i.e., THAA, Mur, individual amino acids) included those attributed to amide and NH groups at around 1545, 1645, and 3200 cm1. Mur and D-AA contents can be used as a marker of bacterial contribution to POM.24,39 Atomic N:C, %CAA, %NAA, DI, %G-Aba + %B-Ala, and %D-AA can all be used as markers of the diagenetic state (i.e., maturation or alteration state) of the POM.24,27,3941 However, considering they are eﬀective at diﬀerent stages of diagenesis,24,41 the spectral regions leading to the best quantiﬁcation of these markers can reveal functional groups having diﬀerent reactivities. %CAA and %NAA are very sensitive markers and thus were best quantiﬁed with the same spectral regions as %N and THAA that are rapidly degraded during early diagenesis.24,40,41 In contrast, DI was correlated with a relatively narrow spectral region (23081854 cm1). This region is generally associated with aromatic cycles and other unsaturated structures absorbing at ∼2035 and ∼1975 cm1. Considering that DI is a marker of more advanced diagenetic stages,27,41 these groups appear to be degraded at a much slower rate than N-molecules such as amino acids. %D-AA, which appears to possess an intermediate sensitivity to diagenesis,24 was best quantiﬁed at 1055567 cm1. These wavenumbers are absorbed by numerous groups such as COOH, secondary amides, alkenes, and substituted benzenes. Atomic N:C and %G-Aba + %B-Ala were best quantiﬁed by spectral regions typical of both sensitive and less sensitive markers. A second experiment was done by removing one station’s samples from the calibration set. Each parameter was then quantiﬁed for these left-out samples, and a relative root-meansquared error of prediction (%RMSEP) was calculated. The accuracy of the quantiﬁcations varied depending on the analyzed station and parameter. Among the eight stations that had numerous samples, the station that generally gave the best quantiﬁcations was the one in the Gulf of St. Lawrence (station 10 in Figure S2 of Supporting Information). These results are shown in Table 3. Some stations gave better predictions for speciﬁc parameters; for example, station 6 in the estuary (Figure S2 in Supporting Information) gave a %RMSEP of 7.3% for D-Asx and 21.1% for D-Glx (not shown). The relatively high errors in many predictions (i.e., atomic N:C, %NAA, %G-Aba + %B-Ala, %D-AA, and D-Glx in Table 3) can be attributed to a lack of both quantity and representativity in the calibration set. The present study used only 30 samples, and these samples were from very diﬀerent locations receiving particles of diﬀerent compositions. This reduced the likelihood of having calibration samples with a composition representative of the samples to be quantiﬁed (or unknown samples).22 A larger calibration set, representative of all expected sources of variance in potential unknown samples, would reduce prediction errors. 9676

dx.doi.org/10.1021/es202607n |Environ. Sci. Technol. 2011, 45, 9671–9679

Environmental Science & Technology

ARTICLE

Figure 3. First two PLS loadings (i and ii) for diﬀerent parameters of the ﬁltered particles. Close-ups show enlarged regions along the y axis. See Table 2 for the description of acronyms and abbreviations.

The ﬁrst two loadings of PLS regression models obtained for %OC, N:C, THAA, %D-AA, DI, and Mur are shown in Figure 3. The loadings for the other parameters are available in the Supporting Information (Figure S4). Positive bands in a loading indicate functional groups that are directly correlated to that parameter, while negative bands imply an inverse correlation to that parameter. The loadings for %OC (Figure 3a) and %N were similar and showed the characteristic bands of marine POM, including the aliphatic bands between 2963 and 2850 cm1, the amide I and II bands at ∼1650 and ∼1545 cm1, the ester band at 1216 cm1, and the carbohydrates band at 1070 cm1. Thus, the proportion of marine POM in the particles appeared to be responsible for most of the variations of %OC and %N in the studied particles. A strong negative band associated to aluminosilicates (∼985 cm1) was observed, indicating that particles rich in carbon and nitrogen contain less minerals. The positive bands in the loadings for N:C ratio (Figure 3b) showed that high N:C ratios are associated with high amide, NH, aliphatic, and carbohydrates contents. Marine POM is generally enriched in these components and exhibits higher N:C ratios than terrigenous POM.2,36 Moreover, the band from CH3 at 2965 cm1 was relatively intense and indicates that marine POM is enriched in methyl groups, probably in branched and small aliphatic structures. Compared to other techniques such as

NMR, FTIR can more easily discriminate CH3 and CH2 contents. Relatively strong negative bands were observed at 2162, 2038, 1980, and 1590 cm1 (Figure 3b). These ﬁndings conﬁrm that particles having low N:C ratios (e.g., terrigenous particles) are relatively enriched in aromatic, carboxylate, and adsorbed water compared to marine particles. The absorption from NH groups (relative to OH) was more important for the quantiﬁcation of total nitrogen or N-molecules (i.e., THAA, D-AA, Mur) than for the quantiﬁcation of carbon as indicated by a peak maximum at ∼3285 cm1 (Figure 3bd,f) instead of 3321 cm1 (Figure 3a). The ﬁrst loadings for THAA (Figure 3c) were similar to those for %OC or %N except for few interesting points. Not surprisingly, the amide bands, located at 1650 and 1550 cm1, were more predominant. The other important diﬀerence was the negative band at 1004 cm1, a region that showed a positive peak in the loadings for %OC (and %N). POM generally shows no or very little absorption in this area. However, clay minerals, such as illite and kaolinite, show a strong and broad band around 1000 cm1.32 Clays are a very good matrix for the protective sorption of POM, and correlations between clays and POM have been reported.42,43 A negative correlation between clays and THAA, suggested by the negative band at 1004 cm1 in Figure 3c, suggests that THAA are not associated with clays 9677

dx.doi.org/10.1021/es202607n |Environ. Sci. Technol. 2011, 45, 9671–9679

Environmental Science & Technology (no protective sorption) which is consistent with the relatively high lability of THAA.40 The loading for %D-AA (Figure 3d) was almost the mirror image of the loading for %OC and THAA. This ﬁnding is not surprising considering that %D-AA increases as %OC and THAA content decreases during POM degradation.24 %D-AA is obtained by comparing D-AA content with THAA content. The loadings for DI (Figure 3e) exhibited only one important positive band at ∼1070 cm1. This band is typical of carbohydrates. High DI, indicative of fresh or unaltered POM, are thus associated with high carbohydrate content. In contrast, a broad negative band was observed at ∼1585 cm1. This band could comes from amides (sharp band) or from COO and CdC groups. Considering that unaltered POM (high DI) generally contains more amides,24,40 the inverse correlation between DI and this band is most likely due to COO and CdC groups. This suggests that altered POM is enriched in carboxylates and aromatics but depleted in carbohydrates compared to fresh POM. Particles with altered POM are also enriched in aluminosilicates as indicated by negative bands at 3621 and 980 cm1 (SiO). The formation of bonds between carboxylic groups and the mineral matrix, during the maturation of the POM, can promote the enrichment in COO.35,37,38,44 Davis et al.41 indicated that DI is an indicator of long-term diagenetic alteration and thus not sensitive to rapid alterations that occur in the ﬁrst weeks of degradation. The positive bands in the loadings for Mur (Figure 3f) indicated that NH (3294 cm1), CH3 (2963 cm1), amide (1590 cm1), and carbohydrate (1072 cm1) contents are correlated with Mur content. Mur is a very labile molecule that can be considered an indicator of living or relatively fresh bacterial POM associated with the particles.24,39,45 The predominance of CH3, relative to CH2, is thus an indication of the high lability of this group and/or of the high proportion of methyl groups in bacterial POM. The CH3 band was not as important in the loadings for D-Asx or D-Glx contents (Supporting Information, Figure S4), two other bacterial biomarkers, suggesting that diagenesis had an impact on the correlations observed here. Particles rich in Mur are relatively depleted in aluminosilicates as indicated by negative bands at 3620 and 981 cm1. This study presents the ﬁrst application of PLS regression with ATR-FTIR spectra of aquatic particles and the ﬁrst PLS quantiﬁcation of molecular parameters such as THAA, DI and Mur. The PLS loadings revealed information on the relative dynamic and reactivity of functional groups. This information is otherwise hidden in the numerous overlapping bands of the FTIR spectra. ATR-FTIR represents a new and powerful method for a rapid, inexpensive, and nondestructive characterization of particles collected by ﬁltration. Filtration is one of the most common and simple sampling procedures in environmental studies.

’ ASSOCIATED CONTENT

bS

Supporting Information. Additional material including a sketch of the ATR accessory used and infrared beam, a map of the sampling sites, validation plots for ATR-PLS quantiﬁcations of parameters not shown in the article, and the ﬁrst two PLS loadings of parameters not shown in the article. This material is available free of charge via the Internet at http://pubs.acs.org.

ARTICLE

’ AUTHOR INFORMATION Corresponding Author

*Phone: (506)858-4333. Fax: (506)858-4541. E-mail: luc.tremblay@ umoncton.ca.

’ ACKNOWLEDGMENT We thank Adeline Barreau for the quantiﬁcation of amino acids and muramic acid in river samples. Three anonymous reviewers provided valuable comments that improved the manuscript. This research was supported by grants from the Natural Science and Engineering Research Council of Canada (NSERC) and the New Brunswick Innovation Foundation (NBIF). ’ REFERENCES (1) Schlesinger, W. H.; Melack, J. M. Transport of organic carbon in the world’s rivers. Tellus 1981, 33 (2), 172–187. (2) Hedges, J. I.; Keil, R. G.; Benner, R. What happens to terrestrial organic matter in the ocean? Org. Geochem. 1997, 27 (5), 195–212. (3) Burdige, D. J. Preservation of organic matter in marine sediments: Controls, mechanisms, and an imbalance in sediment organic carbon budgets. Chem. Rev. 2007, 107, 467–485. (4) Luthy, R. G.; Aiken, G. R.; Brusseau, M. L.; Cunningham, S. D.; Gschwend, P. M.; Pignatello, J. J.; Reinherd, M.; Traina, S. J.; Weber, W. J., Jr.; Westall, J. C. Sequestration of hydrophobic organic contaminants by geosorbents. Environ. Sci. Technol. 1997, 31, 3341–3347. (5) Lee, C.; Wakeham, S.; Arnosti, C. Particulate organic matter in the sea: The composition conundrum. AMBIO 2004, 33 (8), 565–575. (6) Liu, Z.; Mao, J. D.; Peterson, M. L.; Lee, C.; Wakeham, S. G.; Hatcher, P. G. Characterization of sinking particles from the northwest Mediterranean Sea using advanced solid-state NMR. Geochim. Cosmochim. Acta 2009, 73 (4), 1014–1026. (7) Janik, L. J.; Skjemstad, J. O. Characterization and Analysis of Soils Using Midinfrared Partial Least-Squares. 2. Correlations with Some Laboratory Data. Aust. J. Soil Res. 1995, 33 (4), 637–650. (8) Tremblay, L; Gagne, J.-P. Fast quantiﬁcation of humic substances and organic matter by direct analysis of sediments using DRIFT spectroscopy. Anal. Chem. 2002, 74, 2985–2993. (9) 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 (4), 913–923. (10) Gelinas, Y.; Baldock, J. A.; Hedges, J. I. Demineralization of marine and freshwater sediments for CP/MAS 13C NMR analysis. Org. Geochem. 2001, 32 (5), 677–693. (11) Mao, J.-D.; Tremblay, L.; Gagne, J.-P.; Kohl, S.; Rice, J.; Schmidt-Rohr, K. Humic acids from particulate organic matter in the Saguenay Fjord and the St. Lawrence Estuary investigated by advanced solid-state NMR. Geochim. Cosmochim. Acta 2007, 71 (22), 5483–5499. (12) Griﬃths, P. R.; de Haseth, J. A. Fourier Transform Infrared Spectrometry; John Wiley & Sons: New York, 2007. (13) Averett, L. A.; Griﬃths, P. R.; Nishikida, K. Eﬀective Path Length in Attenuated Total Reﬂection Spectroscopy. Anal. Chem. 2008, 80, 3045–3049. (14) Shaka’, H.; Saliba, N. A. Concentration measurements and chemical composition of PM102.5 and PM2.5 at a coastal site in Beirut, Lebanon. Atmos. Environ. 2004, 38 (4), 523–531. (15) Coury, C.; Dillner, A. M. A method to quantify organic functional groups and inorganic compounds in ambient aerosols using attenuated total reﬂectance FTIR spectroscopy and multivariate chemometric techniques. Atmos. Environ. 2008, 42 (23), 5923–5932. (16) Coury, C.; Dillner, A. M.. ATR-FTIR Characterization of Organic Functional Groups and Inorganic Ions in Ambient Aerosols at a Rural Site. Atmos. Environ. 2009, 43 (4), 940–948. 9678

dx.doi.org/10.1021/es202607n |Environ. Sci. Technol. 2011, 45, 9671–9679

Environmental Science & Technology (17) Gustafson, J. P.; Renman, A.; Renman, G.; Poll, K. Phosphate removal by mineral-based sorbents used in ﬁlters for small-scale wastewater treatment. Water Res. 2008, 42 (12), 189–197. (18) Nanda, D.; Tung, K.-L.; Li, Y.-L.; Lin, N.-J.; Chuang, C.-J. Eﬀect of pH on membrane morphology, fouling potential, and ﬁltration performance of nanoﬁltration membrane for water softening. J. Membr. Sci. 2010, 349 (12), 411–420. (19) El-Azizi, I. M.; Schmalenberger, A.; Komlenic, R.; Edyvean, R. G. J. Study of a depth ﬁlter (DisruptorTM) for the novel application of reducing SWRO membrane fouling. Desalin. Water Treat. 2011, 29 (13), 20–28. (20) Zimmermann, M.; Leifeld, J.; Fuhrer, J. Quantifying soil organic carbon fractions by infrared-spectroscopy. Soil Biol. Biochem. 2007, 39 (1), 224–231. (21) Viscarra Rossel, R. A.; Walvoort, D. J. J.; McBratney, A. B.; Janik, L. J.; Skjemstad, J. O. Visible, near infrared, mid infrared or combined diﬀuse reﬂectance spectroscopy for simultaneous assessment of various soil properties. Geoderma 2006, 131, 59–75. (22) Alaoui, G.; Leger, M. N.; Gagne, J.-P.; Tremblay, L. Assessment of estuarine sediment and sedimentary organic matter properties by infrared reﬂectance spectroscopy. Chem. Geol. 2011, 286 (34), 290–300. (23) Gearing, J. N.; Pocklington, R. Organic geochemical studies in the St. Lawrence Estuary. In Oceanography of a large-scale Estuarine system, the St Lawrence; El-Sabh, M. I., Silverberg, N., Eds.; Springer: New York, 1990; pp 170201. (24) Bourgoin, L.-H.; Tremblay, L.. Bacterial reworking of terrigenous and marine organic matter in estuarine water columns and sediments. Geochim. Cosmochim. Acta 2010, 74 (19), 5593–5609. (25) Hedges, J. I.; Stern, J. H. Carbon and nitrogen determinations of carbonate-containing solids. Limnol. Oceanogr. 1984, 29 (3), 657–663. (26) Kaiser, K.; Benner, R. Hydrolysis-induced racemization of amino acids. Limnol. Oceanogr.: Methods 2005, 3, 318–325. (27) Dauwe, B.; Middelburg, J. J.; Herman, P .M. J.; Heip, C. H. R. Linking diagenetic alteration of amino acids and bulk organic matter reactivity. Limnol. Oceanogr. 1999, 44 (7), 1809–1814. (28) Haaland, D. M.; Thomas, E. V. Partial Least-Squares Methods for Spectral Analyses. 1. Relation to Other Quantitative Calibration Methods and the Extraction of Qualitative Information. Anal. Chem. 1988, 60, 1193–1202. (29) Brereton, R. G. Introduction to multivariate calibration in analytical chemistry. Analyst 2000, 125, 2125–2154. (30) Wentzell, P. D.; Montoto, L. V. Comparison of principal components regression and partial least squares regression through generic simulations of complex mixtures. Chemom. Intell. Lab. Syst. 2003, 65 (2), 257–279. (31) Bellamy, L. J. The Infra-red spectra of complex molecules: Chapman and Hall: London, 1975. (32) Salisbury, J. W.; Walter, L. S.; Vergo, N.; Daria, D. M. Infrared (2.125 μm) spectra of minerals; John Hopkins University Press: Baltimore, MD, 1991. (33) Koretsky, C. M.; Sverjensky, D. A.; Salisbury, J. W.; Daria, D. M. Detection of surface hydroxyl species on quartz, gamma-alumina, and feldspars using diﬀuse reﬂectance infrared spectroscopy. Geochim. Cosmochim. Acta 1997, 61 (11), 2193–2210. (34) Savenkoﬀ, C.; Vezina, A. F.; Roy, S.; Klein, B.; Lovejoy, C.; Therriault, J.-C.; Legendre, L.; Rivkin, R.; Berube, C.; Tremblay, J.-E.; Silverberg, N. Export of biogenic carbon and structure and dynamics of the pelagic food web in the Gulf of St. Lawrence Part 1. Seasonal variations. Deep Sea Res. Part II 2000, 47(34), 585-607. (35) Francois, R. Marine sedimentary humic substances - Structure, genesis, and properties. Rev. Aquat. Sci. 1990, 3, 41–80. (36) Malcolm, R. L. The Uniqueness of Humic Substances in Each of Soil, Stream and Marine Environments. Anal. Chim. Acta 1990, 232, 19–30. (37) Hedges, J. I.; Keil, R. G. Organic geochemical perspectives on estuarine processes: sorption reactions and consequences. Mar. Chem. 1999, 65 (12), 55–65. (38) Tremblay, L.; Gagne, J.-P. Organic matter distribution and reactivity in the waters of a large estuarine system. Mar. Chem. 2009, 116 (14), 1–12.

ARTICLE

(39) Tremblay, L; Benner, R. Organic matter diagenesis and bacterial contributions to detrital carbon and nitrogen in the Amazon River system. Limnol. Oceanogr. 2009, 54 (3), 681–691. (40) Cowie, G. L.; Hedges, J. I. Biochemical indicators of diagenic alteration in natural organic matter mixtures. Nature 1994, 369 (6478), 304–307. (41) Davis, J.; Kaiser, K.; Benner, R. Amino acid and amino sugar yields and compositions as indicators of organic matter diagenesis. Org. Geochem. 2009, 40 (3), 343–352. (42) Ransom, B.; Bennett, R. H.; Baerwald, R.; Shea, K. TEM study of in situ organic matter on continental margins: occurrence and the “monolayer” hypothesis. Mar. Geol. 1997, 138 (12), 1–9. (43) Keil, R. G.; Montluc-on, D. B.; Prahl, F. G.; Hedges, J. I. Sorptive preservation of labile organic matter in marine sediments. Nature 1994, 370 (6490), 549–551. (44) Thomas, J. E.; Kelley, M. J. Interaction of mineral surfaces with simple organic molecules by diﬀuse reﬂectance IR spectroscopy (DRIFT). J. Colloid Interface Sci. 2008, 322 (2), 516–526. (45) Moriarty, D. J. W. Improved method using muramic acid to estimate biomass of bacteria in sediments. Oecologia 1977, 26, 317–323.

9679

dx.doi.org/10.1021/es202607n |Environ. Sci. Technol. 2011, 45, 9671–9679

ARTICLE pubs.acs.org/est

High-Resolution, Two-Dimensional Measurement of Dissolved Reactive Phosphorus in Sediments Using the Diffusive Gradients in Thin Films Technique in Combination with a Routine Procedure Shiming Ding,*,† Fei Jia,‡ Di Xu,† Qin Sun,‡ Lei Zhang,† Chengxin Fan,† and Chaosheng Zhang§ †

State Key Laboratory of Lake Science and Environment, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing 210008, China ‡ College of Environmental Science and Engineering, Hohai University, Nanjing, China § GIS Centre, Ryan Institute and School of Geography and Archaeology, National University of Ireland, Galway, Ireland

bS Supporting Information ABSTRACT: Dissolved reactive phosphorus (DRP) is the most available P form in sediments and often directly controls phytoplankton blooms in aquatic systems. In this study, a novel procedure was developed for two-dimensional (2D) measurement of DRP in sediments at a spatial resolution of 0.45 mm using the diﬀusive gradients in thin ﬁlms (DGT) technique with a revised high-capacity binding phase (Zr oxide gel). This procedure involves DGT uptake of P in sediments, 2D slicing of the binding gel on a 0.45 0.45-mm grid system, elution of P from each gel square with 1 M NaOH, and microcolorimetric determination of DRP in each eluted solution using 384-microwell plates. Measurements of DRP via this procedure were tested in homogeneous solutions and sediments and produced an acceptable error (<20% relative standard deviation) for the analysis once the accumulated mass of P in each gel square reached 1.2 μg cm2. This method was successfully applied to produce 2D images of the DRP distribution in sediments with and without the inﬂuence of tubiﬁcid worm bioturbation, revealing a much more pronounced and localized impact from tubiﬁcid worms than that found using a one-dimensional measurement of pore water DRP concentrations at 1-cm resolution.

’ INTRODUCTION Sediments are traditionally treated as a one-dimensional (1D) system, with a general assumption of horizontal homogeneity.1 An understanding of element diagenesis and the related immobilization/remobilization processes is to a large extent based on vertical measurements at a common resolution (cm or lower). Recent developments in planar optodes have enabled the twodimensional (2D) measurement of O2,2 pH,3,4 and pCO25 in sediments at very high resolutions (∼100 μm). These measurements have revealed distinct vertical and horizontal heterogeneity features of chemical distributions in sediments even on a microscale. This is well demonstrated by the existence of micro niches in sediments, which have observable scales ranging from 400 μm to ∼1 cm.6 Observations of the 2D heterogeneity of chemicals in sediments necessitate their measurements at the 2D level and at higher resolution as well as the reassessment of their diagenetic processes based on these measurements. The diﬀusive gradients in thin ﬁlms (DGT) is a powerful technique comparable to planar optodes for measuring solutes in sediments. It relies on the diﬀusion-controlled ﬂux of a solute to a binding phase contained in the DGT probe.7 The mean concentration of the solute is measured over the deployment time, which results from the translocation of the solute from pore r 2011 American Chemical Society

water to the probe and the further release of the solute from the sediment solids to resupply the pore water.8,9 The measured concentration corresponds to the labile fraction of the solute, which equals to its true concentration in pore water when the resupply from the sediment solids is suﬃcient to maintain the pore water concentration.9 This concentration would change from local increase or decrease in pore water concentration as well as solid phase resupply.10,11 The DGT technique has been applied to 2D measurements at high resolution for the detection of trace metals and sulﬁdes in sediments in combination with analyses by proton induced X-ray emission (PIXE), laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS), and computer-imaging densitometry techniques (CID).1117 These measurements have revealed the heterogeneous nature of sediments, particularly on a microscale. For example, simultaneous measurements of multiple solutes using composite DGT probes have demonstrated the simultaneous release of metals and sulﬁdes in localized zones of Received: August 10, 2011 Accepted: October 11, 2011 Revised: October 7, 2011 Published: October 11, 2011 9680

dx.doi.org/10.1021/es202785p | Environ. Sci. Technol. 2011, 45, 9680–9686

Environmental Science & Technology sediments 18 and microscale biogeochemical heterogeneity in the presence of animal burrows and sea grass roots.16 Phosphorus is an essential and often limiting nutrient for biological productivity in aquatic ecosystems. Dissolved reactive phosphorus (DRP) is deﬁned as phosphate measured using the molybdenum blue method,19 which includes orthophosphate and a small quantity of acid hydrolyzable organic or condensed phosphorus compounds.20 A high-resolution and 2D measurement of DRP in sediment proﬁles is of vital importance for the assessment of its diagenesis and remobilization processes. Measurements by conventional techniques, such as coring-slicing-centrifuging, squeezing, and dialysis (peeper), are generally limited to 1D measurements at low spatial resolutions. Although a 2D peeper has been developed,21 its spatial resolution (9 mm) is not suﬃcient to reveal microscale changes in sediments. On the other hand, the DGT technique has been used for 1D measurement of DRP in sediments at higher resolutions (12 mm).22,23 2D measurements at high resolution (∼300 μm) with DGT have also been developed for sedimentary P, with analysis by LA-ICP-MS.11,24,25 Recently, Ding et al. (2010) 26 developed a new binding gel for the DGT measurement of DRP using amorphous zirconium (Zr oxide) as the binding agent. This Zr-oxide binding phase has a very high capacity (>100 μg P cm2) for DGT measurements in various environments. The aim of the present study was to develop and test a novel procedure for the 2D measurement of DRP in sediments at submm resolution using the Zr-oxide DGT technique.

’ EXPERIMENTAL SECTION Preparation and Characterization of the Zr-Oxide Binding Gel. The procedure for the preparation of the Zr-oxide binding

gel was modified, based on a report by Ding et al.,26 in order to enable a high-resolution measurement. Half-dried amorphous zirconium hydroxide (2 g) was added to 4 mL of gel solution composed of 28.5% acrylamide (w/v) and 1.5% methacrylamide (w/v). The mixture was thoroughly ground in an agate mortar, followed by further dispersion in an ultrasonic disruptor. The mixture was left to stand for 5 min to remove settled particles, and then 3.0 μL tetramethylethylenediamine (TEMED) catalyst and 75 μL freshly prepared ammonium persulfate initiator (10%, w/v) were added. After mixing, the solution was immediately cast between glass plates separated by 0.4-mm plastic spacers. The glass plate assembly was placed in an incubator at 10 ( 1 °C for 0.5 h, to allow the zirconium hydroxide to settle by gravity to one side of the gel, and was then transferred to an oven at 45 ( 1 °C to allow the gel to polymerize for 1 h. The gel sheet removed from the glass plates was soaked in deionized water for at least 24 h (the water was replaced 23 times) and stored in deionized water prior to use. The distribution of zirconium hydroxide on the modiﬁed binding gel was checked by scanning electron microscopy (SEM) and compared to a binding gel prepared according to the original procedure.26 The gels were placed on a gel dryer for 4 h at 60 °C under vacuum. The dried gels were analyzed on a Cambridge Stereoscan 120 instrument at an accelerating voltage of 20 kV after gold-coating using a Balzers Union SCD 40 sputter-coater. The DGT performance of this binding gel was evaluated using the DGT uptake of P with the deployment time in a well-stirred solution containing 21.5 mg L1 P (KH2PO4) and 0.03 M

ARTICLE

NaNO3 at pH 7.0 and 25 °C, with the diﬀusive gel (0.80 mm thickness) prepared from 15% acrylamide and 0.3% agarosederived cross-linker following the published procedure.7 A 0.13 mm cellulose nitrate ﬁlter membrane (Whatman, 0.45 μm pore size) was used as an extension of the diﬀusion layer in contact with the solution. Phosphorus accumulated in the binding gels was eluted with 1 M NaOH 26 and measured using the molybdenum blue method.19 The elution factor was detected at 0.95 (Figure S1), which is the same as reported previously.26 This elution factor was adopted for the calculation of the accumulated mass of P. Operational Procedure for the High-Resolution, 2D Measurement of DRP. A procedure for the high-resolution, 2D DGT measurement of DRP was developed in this study. This involved DGT uptake in various P-containing environments, 2D slicing of the binding gel on a submm scale, elution of P from each gel square, and microcolorimetric analysis of DRP in each eluted solution (Figure 1). A special cutter was made to slice the binding gel to yield a 0.45-mm resolution after DGT uptake. The cutter was made by stacking a total of 80 commercial Teflon-coated razor blades (0.10 mm thickness, Gillette, China), with each pair of neighboring blades separated by a 0.35-mm plastic spacer (see “The details about the special cutter and 2D slicing” and Figure S2A-D in the Supporting Information). Prior to cutting, the gel was fixed on one side of a commercial double-faced adhesive tape, with the Zr-oxide side facing outward. The other side of the tape was adhered to a clean perspex plate. The fixed gel was cut in one direction by pressing the cutter directly into the gel, followed by a second cut perpendicular to the first one. The twodimensional cuts produced a square array with each gel square having a size of 0.45 0.45 mm (Figure S2E). The gel square array was covered with a wet filter paper (made by immerging Whatman No. 40 into deionized water) for 5 min to remove the adhesive of the tape from the gel squares. Each gel square was then sequentially picked using a needle and transported into a polystyrene microwell composed of 16-microwell strips in a 384microwell plate holder (Jingrui, Shanghai GenoMintel Bioscience & Technique Development Co., Ltd.). 40 μL of 1 M NaOH was added to each microwell using a multichannel pipet, and the solutions were left standing for 16 h at room temperature to elute P from each gel square. The determination of DRP was based on the molybdenum blue method of Murphy and Riley.19 A volume of 24 μL elution solution was withdrawn from each microwell and transferred into a new microwell. A volume of 6 μLof 2 M H2SO4 was added, and the microwells were left to stand for 1 h to neutralize the alkali. Then a volume of 3 μL of reagent (see “The mixed reagent used for analysis of DRP” in the Supporting Information) was added, and the microwells were centrifuged at 2000 rpm for 5 min to remove any gas bubbles that were formed during the neutralization. The mixtures (33 μL in each microwell) were incubated at 35 °C for 1 h to allow colorimetric formation. The microtiter plates were ﬁnally read at 700 nm using an Epoch Microplate Spectrophotometer (BioTek, USA). Reproducibility of the DGT Measurement in Homogeneous Media. To test the reproducibility of the DGT measurement of DRP in the above procedure, DGT deployments were sequentially performed in P-containing solutions and aquatic microcosms with thoroughly homogeneous sediments. Standard piston-type DGT units (DGT Research Ltd.) containing the Zr-oxide gel were exposed to 4 L KH2PO4 solutions (pH 7.0) containing 0.03 M NaNO3 at 20 °C. The concentrations of 9681

dx.doi.org/10.1021/es202785p |Environ. Sci. Technol. 2011, 45, 9680–9686

Environmental Science & Technology

ARTICLE

Figure 1. Operational procedure for high-resolution, 2D measurement of DRP (CDGT) in sediments.

KH2PO4 varied between 1 and 20 mg P L1. These solutions at pH 7 contain roughly equal concentrations of the hydrogen phosphate ion and the dihydrogen phosphate ion. Both would diffuse through the diffusive layers to adsorb on the Zr-oxide gel. Each DGT unit was retrieved 2 h after deployment. The accumulated masses of P in the binding gels were characterized by random sampling of ∼80 gel squares within each gel using the above procedure. A total of 10 diﬀerent sediments were collected from several lakes in China. The fresh sediments were lyophilized at 80 °C and passed through a 0.15 mm-mesh sieve prior to a thorough combination and homogenization. A weight of 50 g of the dried sediment sample was transferred into a 100-mL beaker, followed by addition of 50 mL of deionized water. The mixture was shaken slightly by hand to produce a ﬂat interface between the sediment and aqueous phase. The beakers were covered with nylon nets (0.25 mm mesh) or aluminum foils to create oxygenated or oxygen-limited conditions, respectively. All beakers were left standing at room temperature for 1 month to enable suﬃcient settlement of the sediment particles. Standard piston-type DGT units containing the Zr-oxide gel were then placed into the beakers (one DGT unit per beaker) by pressing them gently onto the sediment-water interface (SWI) to create good contact between the membranes of the DGT units and the SWI. The beakers were further left to stand for diﬀerent periods ranging from 3 to 10 d. After retrieval, the accumulated masses of P in the binding gels were characterized by random sampling of ∼80 gel squares from each gel sheet using the above procedure.

DGT Measurements in Sediments. High-resolution, 2D measurements of DRP using the Zr-oxide DGT technique were performed to investigate the influence of tubificid worm bioturbation on DRP distributions in sediment profiles. An incubation microcosm was designed based on the report by Zhang et al.27 Two sediment cores as well as the overlying water from the estuary of the Dapu River, Lake Taihu (31°180 42.700 N, 119°560 52.200 E), were collected into plexiglass tubes (11 cm i.d., 50 cm long) using a gravity core sampler. The top 12-cm sediment sections were sliced at 2-cm intervals. Sediments of the corresponding intervals from the two cores were combined and sieved through a 0.6-mm mesh to remove macrofauna and large particles. Each sediment sample was then fully homogenized and put back into the two plexiglass tubes at 2-cm intervals according to the original sequence. Filtered overlying water was gently added to the top of the sediment using intravenous needles. The two tubes, each containing 12-cm sediment, were stored in a dark tank and further submerged in filtered lake water. The microcosm was preincubated at 15 ( 1 °C. Tubificid worms were collected from the same sampling area with a Peterson Grab. Active worms (3545 mm in body length) were selected, counted, and introduced into one tube on day 17 after the preincubation. The amount of worms added was on average 17.1 g ww/m2, which equaled the largest biomass found in the estuary of the Dapu River.27 Two Zr-oxide DGT probes assembled with standard DGT holders for sediment measurement (DGT Research Ltd.) were deoxygenated with nitrogen for 16 h and then 9682

dx.doi.org/10.1021/es202785p |Environ. Sci. Technol. 2011, 45, 9680–9686

Environmental Science & Technology

ARTICLE

Figure 2. SEM images of the binding gels prepared according to Ding et al.26 (left) and this study (right).

inserted across the SWI of each tube on day 30 after the preincubation. The DGT probes were deployed for 6 days, with the temperature maintained at 15 ( 1 °C. On retrieval, the SWI on the probe was immediately marked. Accumulated masses of P in the binding gel corresponding to the upper 3.0 cm of the sediment layers were measured using the high-resolution, 2D operation procedure described above. Concentrations of DRP (CDGT) were then calculated using the following equation7 CDGT

MΔg ¼ DAt

ð1Þ

where A is the surface area of the gel square (0.225 mm2), D is the diffusion coefficient of phosphate in the diffusive layer, t is the deployment time, and M is the corresponding accumulated mass of P over the deployment time.7 The 2D spatial distribution of DRP concentrations (with a total of ∼2222 data for one sediment profile) was plotted using the software Origin 8.0 (OriginLab Corporation, USA). Data Processing. As the sediments were fully homogenized prior to incubation, the 2D spatial distribution of DRP concentrations of the control could be regarded as the background without bioturbation. Taking the horizontal heterogeneity of DRP concentrations in sediments of the control into consideration, the influence of tubificid worm bioturbation on DRP concentration in a localized zone of sediments can be regarded as a probability event, which can be assessed by comparing its DRP concentration with those in all the localized zones of the control at the same horizontal level through meshing. Both 2D DRP distribution images obtained from DGT measurements (the control and the bioturbation treatment, with 64 rows and 32 columns and a square size of 0.45 0.45 mm) were first regridded at a 1.8 1.8-mm interval, yielding a total of 128 (16 columns 8 rows) subregions with each composed of 16 (4 4) DRP concentration data. The mean values of DRP in each subregion for the bioturbation treatment were compared to each of the 8 subregions of the control in the same horizontal layers using a nonparametric test (SPSS 10.5, USA). The number “0” or “1” was recorded, depending on whether the difference between the two mean concentrations was insignificant or significant (p < 0.05, 2-tailed). All numbers were summed for each subregion of the bioturbation treatment. The sums for the different subregions varied within a range of 0 to 8. A value of “0” means there was the least probability of an impact from tubificid worms, whereas “8” reflects the greatest probability of an impact.

Other sums indicate intermediate situations, with the probability increasing with an increase in the sum. The 2D spatial distribution of the sums for all subregions of the bioturbation treatment was plotted using Origin 8.0 (OriginLab Corporation, USA).

’ RESULTS AND DISCUSSION Performance of the Modified Binding Gel. To facilitate high-resolution and 2D DGT measurement, the particle size of the impregnated binding agent should be small enough to produce a homogeneous and compact distribution on the binding gel. For example, to measure trace metals with the DGT technique at high resolution, the suspended particulate reagent iminodiacetate is used instead of Chelex 100 as the binding agent for preparation of the binding gel, as it has a much smaller particle size (0.2 μm) than that of Chelex 100 (∼100 μm).28 The Zroxide binding gel that the authors initially developed was designed to measure DRP on a mm or lower resolution scale.26 Its SEM image shows obvious aggregation of the impregnated Zroxide to large particles (∼30 μm) with an uneven distribution (Figure 2), reflecting that this Zr-oxide binding gel is unsuitable for DGT measurements on a submm scale. Modification was performed to reduce the particle sizes of the Zr-oxide through grinding and ultrasonic disruption prior to gel casting. The SEM image of the modified binding gel showed no evident aggregation of the Zr-oxide particles, and their distribution was much more homogeneous and compact (Figure 2). This binding gel was thus considered suitable for high-resolution, 2D measurement of DRP with the DGT technique. The feasibility and capacity of the modiﬁed Zr-oxide binding gel for DGT measurements were examined by deployment in a solution containing a high concentration of phosphate (21.5 mg P L1). The measured mass of P accumulated in the binding gel increased linearly with an increase in deployment time over a 24-h experiment (r2 = 0.998) (Figure S3). The experimental data agreed well with the theoretical predictions calculated using eq 1, with measured-to-predicted ratios of 0.98 ( 0.05 (n = 21). The results validated the use of the Zr-oxide binding gel for DGT measurements. The capacity for DGT response was estimated at >100 μg P cm2 based on the response of the accumulation mass to the theoretical line, corresponding well to the authors’ previous report.26 This capacity is much greater than that of other binding gels used for DGT measurement of P, including conventional ferrihydrite gel (∼2 μg P cm2)23 and the recently 9683

dx.doi.org/10.1021/es202785p |Environ. Sci. Technol. 2011, 45, 9680–9686

Environmental Science & Technology developed precipitated ferrihydrite gel (7 μg P cm2)24 and Metsorb gel (∼12 μg P cm2).29 Analytical Errors from High-Resolution, 2D DGT Measurements. Prior to application, the analytical errors inherent in the high-resolution, 2D DGT measurement were estimated. Errors may result from all steps of the measurement, including DGT uptake, 2D slicing, P elution, and microcolorimetric determination (Figure 1). The error from DGT uptake was considered negligible with the use of the homogeneous Zr-oxide binding gel. For the slicing step, the distance between each pair of adjacent cutting edges in the cutter was examined using a microscope (Olympus BX51) and it varied within 2% (0.45 ( 0.01 mm) (Figure S2D). The resulting area of each gel square was further examined by randomly measuring ∼100 squares using the microscope, and it varied within 5% (0.2025 ( 0.001 mm2) (Figure S2E). The error from elution of P was also considered negligible since the eluting solution (1 M NaOH) had been mixed thoroughly prior to use. The error caused by

Figure 3. Analytical errors for high-resolution, 2D DGT measurement of DRP (CDGT) in solutions (0) and homogeneous sediments (9) at diﬀerent accumulated masses of P in the binding gel.

ARTICLE

microcolorimetric determination of P was investigated by detecting different concentrations of phosphate (0.010.6 mg L1) spiked in 1 M NaOH solutions, with 16 duplicates for each concentration. The results showed that the relative standard deviation (RSD) was <10% once the P concentration was >0.03 mg L1, corresponding to P accumulation masses of >0.6 μg cm2 in the binding gel. The total errors during the entire procedure were investigated through DGT deployments in phosphate-containing solutions of diﬀerent concentrations as well as in microcosms with homogeneously mixed sediments. The results showed that the RSD values were very similar for the two types of deployments, both exhibiting a sharp decrease and then remaining steady as the accumulated masses of P in the binding gels increased. The analytical errors were estimated to be within 20% and 10% once the mass of P was >1.2 μg cm2 and >4.0 μg cm2, respectively (Figure 3). These errors should have predominantly resulted from the 2D slicing and microcolorimetric determination processes as explained earlier. High-Resolution, 2D DGT Measurements in Sediments. As a pilot study, the newly designed high-resolution, 2D DGT operation procedure was used to investigate the impact of tubificid worms on DRP distribution in sediments sampled from a eutrophic lake. Tubificid worms represent an important fraction of the benthic community in eutrophic lake sediments.30 A similar study has been carried out by Zhang et al.27 via 1D measurements of DRP concentrations at 1-cm resolution. The field experimental design in this study was similar to that of Zhang et al.,27 but the measurements for DRP were performed at a finer resolution and at the 2D level. Two 2D DRP distribution profiles (for the control and bioturbation treatments) with a resolution of 0.45 0.45 mm were obtained in the upper 3.0 cm of the sediment layers (Figure 4), where the majority of tubificid worms were found (>60%).27 The analytical errors were controlled within 20% throughout the 6-d deployments, with most of the accumulated masses of DRP in the gel squares greater than 1.2 μg cm2.

Figure 4. 2D distribution images of DRP concentrations (CDGT) at a spatial resolution of 0.45 mm in sediments without (left) and with (right) tubiﬁcid worm bioturbation. The location of the sediment-water interface is represented by zero. 9684

dx.doi.org/10.1021/es202785p |Environ. Sci. Technol. 2011, 45, 9680–9686

Environmental Science & Technology

Figure 5. The eﬀects of tubiﬁcid worm bioturbation on DRP concentrations (CDGT) in sediments based on 1D (left) and 2D (right) analyses. The 1D data were the average concentrations of DRP at a given depth, with a 0.45 mm interval for the depths. “CK” and “Bio” point to the control and tubiﬁcid worm treatments, respectively, with the asterisks indicating their diﬀerence at signiﬁcance levels of p < 0.05 (*) and p < 0.01 (**). The 2D data were obtained through comparisons of DRP concentration in one localized zone (1.8 1.8 mm) of the tubiﬁcid worm treatment with those in all the localized zones of the control at the same horizontal level. The numbers of 0 to 8 indicate an increase in the probability of an impact of tubiﬁcid worms. The location of the sediment-water interface is represented by zero.

For the control, the 2D distribution of DRP showed systematic changes in the vertical direction, with low concentrations (<0.08 mg L1) in the upper ∼9 mm layers below the SWI, followed by increases to a depth of ∼15 mm, at which point the concentrations stabilized up to a depth of 30 mm. Horizontal variations were highly random due to mixing of the sediment prior to incubation. There were quite a few discrete localized elevations of DRP concentrations throughout the proﬁle. These small-scale hotspots, termed microniches, have previously been observed in sediments for P and other elements and are generally attributed to strong decomposition of enriched organic matter in localized zones.6 The presence of tubiﬁcid worms caused much more 2D heterogeneity in DRP distribution, reﬂected by a visibly more uneven distribution of DRP concentrations along the vertical and horizontal directions. In comparison to the same positions in the control, tubiﬁcid worm bioturbation mainly caused a negative eﬀect, reﬂected by lower concentrations of DRP in the upper layers above ∼7 mm and the deeper layers below ∼19 mm. Comparisons of DRP concentrations in the horizontal direction with the control using the averages from each corresponding depth showed that signiﬁcant eﬀects from tubiﬁcid worm bioturbation were exhibited at nearly 90% of depths (Figure 5), with most showing negative eﬀects. The observed negative impacts of tubiﬁcid worms in this study were much more pronounced than those in the report by Zhang et al.,27 in which only one depth (at 1.5 cm below the SWI) exhibited a signiﬁcantly lower concentration of DRP compared to the control. The results from this study provided more details about the DRP distribution and thus demonstrated an advantage of using a high-resolution measurement, as the increases and decreases at diﬀerent depths observed in this study may have canceled each other out if the measurements were made at lower resolutions (e.g., cm). Furthermore, the DRP measured with the DGT technique reﬂected not only

ARTICLE

the DRP in the pore water but also the resupply of DRP from the sediment solid phase to the pore water.22,26 A decrease of soluble ferrous ion was observed in pore water due to its oxidation to solid ferric oxyhydroxides.27 This occurred as a result of the introduction of dissolved oxygen from overlying waters when the worms irrigated their burrows. The newly formed iron oxyhydroxides would adsorb P ions and inhibit the resupply of DRP from the sediment solids. Comparisons of localized DRP concentrations with the control showed that the eﬀects caused by tubiﬁcid worm bioturbation were much more pronounced in the upper (above 7 mm depth) and deep layers (below 18 mm), which were consistent with the observed two regions on the 2D DRP distribution image of bioturbation with visibly lower DRP concentrations (Figure 5). The eﬀects were relatively weak in the intermediate region. The resulting dumbbell-shaped distribution was in good agreement with the bioturbation feature of tubiﬁcid worms as upward-conveyors. They generally burrow down into sediment vertically and ingest sediment there and then defecate the sediment to the surface.31 Correspondingly, the surface and deep sediment layers should be most aﬀected, as observed. To our knowledge, this is the ﬁrst report of such a localized feature of benthic bioturbation. Future Perspective. High resolution and 2D data can be obtained from DGT measurements in combination with microprobe techniques including PIXE, LA-ICP-MS, and CID.1117 Apart from the CID measurement of sulfides, the use of other techniques requires special equipment and is extremely expensive. The novel procedure developed in this study requires only routine equipments such as a microplate spectrophotometer. The cost is very low for the measurement with these equipments. These advantages make it feasible for further applications. To our knowledge, it is the first time that high-resolution, 2D spatial information of an important chemical (except for sulfides) was obtained using such a fairly routine procedure, which has the potential to be extended to other chemicals in combination with the use of other microdetection methods.

’ ASSOCIATED CONTENT

bS

Supporting Information. The details about the special cutter and 2D slicing; the mixed reagent used for analysis of DRP; elution eﬃciency of P from the modiﬁed Zr-oxide gel with 1 M NaOH; design and photo of the special cutter and a square array from gel slicing with it; relationship between accumulation mass of P in the modiﬁed Zr-oxide DGT devices and deployment time. This material is available free of charge via the Internet at http:// pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: 86-25-86882207. Fax: 86-25-86882207. E-mail: smding@ niglas.ac.cn.

’ ACKNOWLEDGMENT This study was sponsored by the Project of Knowledge Innovation for the third period, CAS (KZCX2-YW-JS304), the National Scientiﬁc Foundation of China (40730528, 40871220), and the Nanjing Institute of Geography and Limnology, CAS 9685

dx.doi.org/10.1021/es202785p |Environ. Sci. Technol. 2011, 45, 9680–9686

Environmental Science & Technology (NIGLAS2010KXJ01). We thank Dr. Jun Luo for his constructive suggestions in improving this manuscript.

’ REFERENCES (1) Berner, R. A. Early Diagenesis; Princeton University Press: Princeton, 1980. (2) Glud, R. N.; Kuhl, M.; Kohls, O.; Ramsing, N. B. Heterogeneity of oxygen production and consumption in a photosynthetic microbial mat as studied by planar optodes. J. Phycol. 1999, 35 (2), 270–279. (3) Hulth, S.; Aller, R. C.; Engstrom, P.; Selander, E. A pH plate ﬂuorosensor (optode) for early diagenetic studies of marine sediments. Limnol. Oceanogr. 2002, 47 (1), 212–220. (4) Zhu, Q. Z.; Aller, R. C.; Fan, Y. Z. Two-dimensional pH distributions and dynamics in bioturbated marine sediments. Geochim. Cosmochim. Acta 2006, 70 (19), 4933–4949. (5) Zhu, Q. Z.; Aller, R. C.; Fan, Y. Z. A new ratiometric, planar ﬂuorosensor for measuring high resolution, two-dimensional pCO2 distributions in marine sediments. Mar. Chem. 2006, 101 (12), 40–53. (6) Stockdale, A.; Davison, W.; Zhang, H. Micro-scale biogeochemical heterogeneity in sediments: A review of available technology and observed evidence. Earth-Sci. Rev. 2009, 92 (12), 81–97. (7) Zhang, H.; Davison, W. Performance characteristics of diﬀusion gradients in thin ﬁlms for the in situ measurement of trace metals in aqueous solution. Anal. Chem. 1995, 67 (19), 3391–3400. (8) Zhang, H.; Davison, W.; Mortimer, R. J. G.; Krom, M. D.; Hayes, P. J.; Davies, I. M. Localised remobilization of metals in a marine sediment. Sci. Total Environ. 2002, 296 (13), 175–187. (9) Harper, M. P.; Davison, W.; Zhang, H.; Tych, W. Kinetics of metal exchange between solids and solutions in sediments and soils interpreted from DGT measured ﬂuxes. Geochim. Cosmochim. Acta 1998, 62 (16), 2757–2770. (10) Naylor, C.; Davison, W.; Motelica-Heino, M.; Van Den Berg, G. A.; Van Der Heijdt, L. M. Simultaneous release of sulﬁde with Fe, Mn, Ni and Zn in marine harbour sediment measured using a combined metal/sulﬁde DGT probe. Sci. Total Environ. 2004, 328 (13), 275–286. (11) Stockdale, A.; Davison, W.; Zhang, H. High-resolution twodimensional quantitative analysis of phosphorus, vanadium and arsenic, and qualitative analysis of sulﬁde, in a freshwater sediment. Environ. Chem. 2008, 5 (2), 143–149. (12) Davison, W.; Fones, G. R.; Grime, G. W. Dissolved metals in surface sediment and a microbial mat at 100-μm resolution. Nature 1997, 387 (6636), 885–888. (13) Jezequel, D.; Brayner, R.; Metzger, E.; Viollier, E.; Prevot, F.; Fievet, F. Two-dimensional determination of dissolved iron and sulfur species in marine sediment pore-waters by thin-ﬁlm based imaging. Thau lagoon (France). Estuarine, Coastal Shelf Sci. 2007, 72 (3), 420–431. (14) Devries, C. R.; Wang, F. Y. In situ two-dimensional highresolution proﬁling of sulﬁde in sediment interstitial waters. Environ. Sci. Technol. 2003, 37 (4), 792–797. (15) Teasdale, P. R.; Hayward, S.; Davison, W. In situ, high-resolution measurement of dissolved sulﬁde using diﬀusive gradients in thin ﬁlms with computer-imaging densitometry. Anal. Chem. 1999, 71 (11), 2186–2191. (16) Robertson, D.; Teasdale, P. R; Welsh, D. T. A novel gel-based technique for the high resolution, two-dimensional determination of iron (II) and sulﬁde in sediment. Limnol. Oceanogr. Methods 2008, 6, 502–512. (17) Warnken, K. W.; Zhang, H.; Davison, W. Analysis of polyacrylamide gels for trace metals using diﬀusive gradients in thin ﬁlms and laser ablation inductively coupled plasma mass spectrometry. Anal. Chem. 2004, 76 (20), 6077–6084. (18) Motelica-Heino, M.; Naylor, C.; Zhang, H.; Davison, W. Simultaneous release of metals and sulﬁde in lacustrine sediment. Environ. Sci. Technol. 2003, 37 (19), 4374–4381. (19) Murphy, J.; Riley, J. P. A modiﬁed single solution method for the determination of phosphate in natural waters. Anal. Chim. Acta 1962, 26 (1), 31–36.

ARTICLE

(20) Karl, D. M.; Tien, G. MAGIC: a sensitive and precise method for measuring dissolved phosphorus in aquatic environments. Limnol. Oceanogr. 1992, 37, 105–116. (21) Lewandowski, J.; Ruter, K.; Hupfer, M. Two-dimensional small-scale variability of pore water phosphate in freshwater lakes: Results from a novel dialysis sampler. Environ. Sci. Technol. 2002, 36 (9), 2039–2047. (22) Monbet, P.; McKelvie, I. D.; Worsfold, P. J. Combined gel probes for the in situ determination of dissolved reactive phosphorus in porewaters and characterization of sediment reactivity. Environ. Sci. Technol. 2008, 42 (14), 5112–5117. (23) Zhang, H.; Davison, W.; Gadi, R.; Kobayashi, T. In situ measurement of dissolved phosphorus in natural waters using DGT. Anal. Chim. Acta 1998, 370 (1), 29–38. (24) Santner, J.; Prohaska, T.; Luo, J.; Zhang, H. Ferrihydrite Containing Gel for Chemical Imaging of Labile Phosphate Species in Sediments and Soils Using Diﬀusive Gradients in Thin Films. Anal. Chem. 2010, 82 (18), 7668–7674. (25) Stockdale, A.; Davison, W.; Zhang, H. 2D simultaneous measurement of the oxyanions of P, V, As, Mo, Sb, W and U. J. Environ. Monit. 2010, 12 (4), 981–984. (26) Ding, S. M.; Xu, D.; Sun, Q.; Yin, H. B.; Zhang, C. S. Measurement of Dissolved Reactive Phosphorus Using the Diﬀusive Gradients in Thin Films Technique with a High-Capacity Binding Phase. Environ. Sci. Technol. 2010, 44 (21), 8169–8174. (27) Zhang, L.; Gu, X. Z.; Fan, C. X.; Shang, J. G.; Shen, Q. S.; Wang, Z. D.; Shen, J. Impact of diﬀerent benthic animals on phosphorus dynamics across the sediment-water interface. J. Environ. Sci. 2010, 22 (11), 1674–1682. (28) Warnken, K. W.; Zhang, H.; Davison, W. Performance characteristics of suspended particulate reagent-iminodiacetate as a binding agent for diﬀusive gradients in thin ﬁlms. Anal. Chim. Acta 2004, 508 (1), 41–51. (29) Panther, J. G.; Teasdale, P. R.; Bennett, W. W.; Welsh, D. T.; Zhao, H. J. Titanium Dioxide-Based DGT Technique for In Situ Measurement of Dissolved Reactive Phosphorus in Fresh and Marine Waters. Environ. Sci. Technol. 2010, 44 (24), 9419–9424. (30) Krieger, K. A.; Ross, L. S. Changes in the benthic macroinvertebrate community of the Cleveland Harbor area of Lake Erie from 1978 to 1989. J. Great Lake Res. 1993, 19 (2), 237–249. (31) Nogaro, G.; Mermillod-Blondin, F.; Francois-Carcaillet, F.; Gaudet, J. P.; Lafont, M.; Gibert, J. Invertebrate bioturbation can reduce the clogging of sediment: an experimental study using inﬁltration sediment columns. Freshwater Biol. 2006, 51 (8), 1458–1473.

9686

dx.doi.org/10.1021/es202785p |Environ. Sci. Technol. 2011, 45, 9680–9686

ARTICLE pubs.acs.org/est

Differential Pair Distribution Function Study of the Structure of Arsenate Adsorbed on Nanocrystalline γ-Alumina Wei Li,*,†,^ Richard Harrington,*,†,‡ Yuanzhi Tang,*,†,z James D. Kubicki,§ Masoud Aryanpour,§,|| Richard J. Reeder,† John B. Parise,†,‡ and Brian L. Phillips† †

Department of Geosciences and Center for Environmental Molecular Science, Stony Brook University, Stony Brook, New York 11794-2100, United States ‡ Department of Chemistry, Stony Brook University, Stony Brook, New York 11794-3400, United States § Department of Geosciences, Pennsylvania State University, University Park, Pennsylvania 16802, United States

bS Supporting Information ABSTRACT: Structural information is important for understanding surface adsorption mechanisms of contaminants on metal (hydr)oxides. In this work, a novel technique was employed to study the interfacial structure of arsenate oxyanions adsorbed on γ-alumina nanoparticles, namely, diﬀerential pair distribution function (d-PDF) analysis of synchrotron X-ray total scattering. The d-PDF is the diﬀerence of properly normalized PDFs obtained for samples with and without arsenate adsorbed, otherwise identically prepared. The real space pattern contains information on atomic pair correlations between adsorbed arsenate and the atoms on γ-alumina surface (Al, O, etc.). PDF results on the arsenate adsorption sample on γ-alumina prepared at 1 mM As concentration and pH 5 revealed two peaks at 1.66 Å and 3.09 Å, corresponding to As O and As Al atomic pair correlations. This observation is consistent with those measured by extended X-ray absorption ﬁne structure (EXAFS) spectroscopy, which suggests a ﬁrst shell of As O at 1.69 ( 0.01 Å with a coordination number of ∼4 and a second shell of As Al at ∼3.13 ( 0.04 Å with a coordination number of ∼2. These results are in agreement with a bidentate binuclear coordination environment to the octahedral Al of γ-alumina as predicted by density functional theory (DFT) calculation.

’ INTRODUCTION Surface adsorption reactions occurring at mineral/water interface are of great importance for water treatment, nutrient management, and soil decontamination.1,2 To achieve comprehensive understanding of the adsorption mechanisms of aqueous solutes to solid surfaces, research tools that provide detailed structural information of species on solid surfaces are essential. In the past thirty years, successful applications of extended X-ray absorption ﬁne structure (EXAFS) spectroscopy to inorganic ion adsorption studies have demonstrated the power of spectroscopic techniques in providing the details of interfacial solutes adsorption at the atomic to molecular scale.3 However, EXAFS has some limitation, for example, in studying the systems containing light elements (i.e., P, Al, etc.). Therefore, the development of new experimental techniques that complement EXAFS for probing surface structures is an active area of research. X-ray diﬀraction (XRD) has been used to determine the atomic-scale structure of materials for a century by analyzing Bragg peaks arising due to translation symmetry in crystals. However, this traditional method breaks down when one attempts to study the structure of species with shorter length scales.4 7 One way around this is to take the Fourier transform of the total scattering pattern (both Bragg and diﬀuse scattering), correctly r 2011 American Chemical Society

normalized, to produce the pair distribution function (PDF). The PDF shows the probability of ﬁnding an atom at a given distance r from another atom.8,9 This technique does not require long-range order and, therefore, is ideal for the study of liquids, amorphous solids, and nanoparticles. Recently, with the high ﬂux and short wavelength X-rays available at third generation synchrotron sources and the use of fast read out area detectors that are able to simultaneously collect the total scattering pattern, diﬀerential pair distribution function (d-PDF) analysis becomes practical for a wide range of systems. This technique involves the subtraction of a reference PDF of the pristine sample from the PDF of the sample after reaction, leaving only the diﬀerence between two samples.10 14 Recently, this technique has been applied to the structures of adsorbates on the surface of nanomaterials.15,16 The correlations arising from the host guest relationship are obtained by subtracting the atomic correlations of the bulk material (Host) from the adsorption material (Host + Guest) as shown in Scheme 1. Harrington et al.,16 using high Received: March 4, 2011 Accepted: October 11, 2011 Revised: September 10, 2011 Published: October 11, 2011 9687

dx.doi.org/10.1021/es200750b | Environ. Sci. Technol. 2011, 45, 9687–9692

Environmental Science & Technology Scheme 1. Illustration of Diﬀerential Pair Distribution Function (d-PDF)

energy synchrotron radiation, investigated the structure of arsenate oxyanion adsorbed on the surface of 2-line ferrihydrite, a naturally occurring iron oxyhydroxide with particles typically 2 3 nm in size.17,18 Using the diﬀerential PDF technique, several correlations were observed, including an As Fe correlation of 3.26 Å. This As Fe correlation has a distance similar to that obtained from As K-edge EXAFS,19,20 which suggested a bridging bidentate binuclear surface structure of As absorbed on 2-line ferrihydrite. Further correlations were also observed in the d-PDF, corresponding to the correlations between the adsorbed As and the oxygen atoms from within the 2-line ferrihydrite particle, which were not previously observed using EXAFS. In this work, we expanded the application of d-PDF technique to arsenate adsorption on γ-alumina (γ-Al2O3). γ-Alumina is an important industrial nanomaterial and is extensively used in geochemistry and environmental chemistry research as a model compound for studying the surface reactivity of Al (hydr)oxides and clays, as well as for the exploration of the mechanisms of heavy metals/metalloids adsorption on metal oxides.21 25 Compared to 2-line ferrihydrite (2 nm in size), the size of γ-alumina is larger (10 20 nm); consequently, it is likely that there will be less surface interactions for the same volume illuminated by the X-ray beam. Another challenge for this system is the much weaker X-ray scattering property of Al atoms due to its lower Z compared to Fe. If d-PDF can successfully determine the arsenate bonding structure on γ-alumina, this technique can be applied to other engineered nanomaterials (e.g., nano-TiO2). The purpose of this study is to apply the novel d-PDF technique to study the mechanism(s) of arsenate adsorption on γ-alumina and to compare the results with other experimental (X-ray absorption) and theoretical (density function theory, DFT) techniques.

’ EXPERIMENTAL SECTION Preparation of Adsorption Samples. Adsorption samples were prepared by adding 0.25 g of dry γ-alumina powder (Aluminum Oxide C, Degussa) to 50 mL of vigorously stirring solution containing 0.01 M NaCl background electrolyte at pH 5. We chose pH 5 because at this pH As exists almost 100% as H2AsO4 (aq). The pH was carefully maintained using automatic titrators (Metrohm STAT 718) with 0.1 M HCl and 0.1 M NaOH. The adsorption isotherm experiments were carried out at pH 5 with initial As concentration of 0.1 10 mM. A short reaction time (15 min) was chosen to avoid precipitation. After reaction, the samples were centrifuged (10 000g, 15 min) to separate the solid and solution. The supernatants were analyzed for arsenate concentration using directly coupled plasma-atomic emission spectrometry (DCP-AES). Three wet paste samples with initial arsenate concentrations of 0.1, 0.4, and 1 mM were prepared for EXAFS measurement. One air-dried sample with 1 mM initial arsenate concentration was prepared for both EXAFS and d-PDF studies. For d-PDF, a

ARTICLE

control sample was prepared with identical treatment but without adding arsenate to the solution. Differential PDF. Powder diffraction experiments were carried out at beamline 11-ID-B at the Advanced Photon Source (APS; Argonne National Laboratory, Argonne, IL) using monochromatic X-rays with energy of ∼90 keV (λ = 0.12702 Å) in transmission mode with samples loaded in polyamide (kapton) capillaries. The diffraction pattern was collected using an amorphous Si detector manufactured by General Electric.26,27 The beamline was calibrated using a CeO2 standard (NIST 674b). The data was converted from 2D to 1D using the program Fit2D.28 Using the program PDFgetX2,29 PDFs were generated by Fourier transformation of the total structure function, S(Q ), with a Q max of 24 Å 1, after being corrected for background scattering, Compton scattering, and oblique incidence as described previously.26 Differential PDFs were obtained by subtraction of a reference PDF (γ-alumina with no arsenate loading) from a PDF of arsenate-sorbed γ-alumina in real space using Microsoft Excel. The control was multiplied by an appropriate constant to ensure that the scale of each PDF was the same. Interatomic distances were quantified by fitting the peaks with a Gaussian function. EXAFS Spectroscopy. Extended X-ray absorption fine structure (EXAFS) spectroscopy data were collected on the arsenate sorption samples and the reference compounds, a 10 mM As(V) solution and the mineral mansfieldite (AsAlO4 3 2H2O). Details of the data collection and analysis are provided in the Supporting Information. Quantum Chemical Calculations. The structures of adsorbed arsenate anions on the (010) and (001) surfaces of γ-alumina were calculated using quantum chemical calculation. Details on the cluster construction can be found in Supporting Information.

’ RESULTS Arsenate Adsorption on γ-Alumina. Systematic characterization of γ-alumina sorbent by powder X-ray diffraction, SEM, TEM, and 27Al NMR are provided in Figures S1 S4, Supporting Information. These results are consistent with previous studies of similar material30 and indicate no impurity. Arsenate adsorption isotherm was carried out within the range of 0.1 to 1 mM initial arsenate concentration (Figure 1). Over this range, the surface coverage increased from 19.1 μmol g 1 (0.13 molecules nm 2) to 187.8 μmol g 1 (1.25 molecules nm 2). A Langmuir isotherm equation, Q = (Q mKC)/(1 + KC), was used to fit the adsorption isotherm, where Q is the amount of adsorbed arsenate (μmol g 1), C is the equilibrium arsenate concentration (μM), Q m is the maximum adsorption amount, and K is the equilibrium constant for the sorption reaction. The fitting results (R2 = 0.97) give the Q m value as 319.2 μmol g 1 and K as 0.037. With higher arsenate concentration (up to 10 mM), an adsorption capacity of ca. 370 μmol g 1 can be achieved, but the extended isotherm (Figure 1 inset) can not be well fitted using a one-site Langmuir equation. Such sorption capacity of γ-alumina is comparable to several other reported adsorbents,31,32 with the comparison provided in Table S1, Supporting Information. The sample prepared for d-PDF investigation has an arsenate surface loading of 187.8 μmol g 1, much lower than the maximum adsorption capacity. Given the short reaction time and low arsenate concentration, 33 we consider it unlikely that any significant 9688

dx.doi.org/10.1021/es200750b |Environ. Sci. Technol. 2011, 45, 9687–9692

Environmental Science & Technology

ARTICLE

Figure 1. Adsorption isotherm of arsenate on γ-alumina at pH 5. Inset shows extend adsorption isotherm to higher concentration range. Arrow denotes the sample used for d-PDF analysis.

amount of aluminum arsenate precipitates would form during the sorption experiments. Differential PDF. The PDFs of bulk γ-alumina, γ-alumina with arsenate adsorbed on the surface, and the difference between them (the d-PDF) are shown in Figure 2a. The d-PDF comprises only correlations containing arsenate, as the other correlations are present in both the loaded sample and the control and are consequently subtracted out. The difference between the two PDFs is very small (Figure 2a); consequently, the d-PDF is noisy. However, two peaks at 1.66 and 3.09 Å are clearly observed in Figure 2b. The first correlation at 1.66 Å is in good agreement of the As O distance (1.68 Å) in most arsenate minerals (e.g., scorodite20), and the second peak at 3.09 Å is assigned as an As Al atomic pair correlation, the next logical correlation in this system. The value of r is similar to the As Al distance (3.11 ( 0.03 Å) determined by Arai et al.22 using EXAFS which suggests the formation of inner-sphere bidentate binuclear complexes. Figure S5 in the Supporting Information shows four d-PDFs calculated using diﬀerent values of Q max in the Fourier transform. The two aforementioned peaks remain at the same value of r no matter what value of Q max is used; other peaks in the d-PDF calculated using a Q max of 24 Å 1 (Figure S5, Supporting Information) are not present at the same r in the d-PDFs calculated using lower values of Q max, indicative of noise. Owing to the low signal-to-noise ratio, we cannot assign peaks at r values greater than 4 Å, even though such assignments were possible in a similar study of ferrihydrite.16 This is attributed to the higher electron density of Fe compared to Al. All features beyond 4 Å are well within noise according to the spectra processed using diﬀerent Q max for Fourier transform (Figure S5, Supporting Information). Arsenic K-Edge EXAFS. For comparison with the d-PDF results, we obtained As K-edge EXAFS for the same samples as well as the two model compounds (Figure 3a). The spectra of all samples are dominated by a strong oscillation from the backscattering of the first oxygen shell, leading to a subtle difference showing only at k ranges of ∼11 13 Å 1. The corresponding Fourier transforms (Figure 3b) for all samples are dominated by the contribution from the first oxygen shell, which is best fit with ∼4 O atoms at 1.69 ( 0.01 Å (Table S2, Supporting Information). Compared to the arsenate solution sample, all adsorption samples also show additional peaks at 2.5 3 Å in R space, which are best fit with ∼2 Al atoms at 3.13 ( 0.04 Å,

Figure 2. (a) The PDFs of γ-alumina (black line), γ-alumina loaded with As (red line), and the diﬀerence between these two spectra (multiplied 5 times for clarity; shown in blue line); (b) the smoothed d-PDF of arsenate adsorbed on γ-alumina, r-averaged over normalization ripples.

similar to the results of Arai et al.22 However, due to the weak backscattering property of Al atoms, the second shell is not well resolved;22,32 therefore, the Debye Waller factor was fixed at 0.006 to reduce the number of free parameters. This value is determined from fitting of the reference compound mansfieldite. No significant differences are observed between results obtained for samples analyzed as wet pastes and dried powders. Quantum Chemical Calculation. As previous studies22,25 only consider octahedral Al to interpret arsenate sorption mechanism, we conducted quantum chemical calculations to model the arsenate binding to γ-alumina (010) and (001) surfaces such that both tetrahedral and octahedral Al are taken into account. Tetrahedral Al on the (001) surface and octahedral Al on the (010) surface of γ-alumina were constructed as suggested by Pinto et al.,34 and arsenate was bonded to both types of sites. Only the monodentate mononuclear structure was considered for As atoms connected to tetrahedral Al sites (Figure 4a) because a bidentate structure (i.e., edge-sharing) is not likely. Both a monodentate (Figure 4b) and a bidentate binuclear structure (Figure 4c) were constructed for arsenate on the (010) surface, and a bidentate mononuclear structure was excluded due to the instability of this structure as previously suggested.20,35 An H2AsO4 group 9689

dx.doi.org/10.1021/es200750b |Environ. Sci. Technol. 2011, 45, 9687–9692

Environmental Science & Technology

ARTICLE

Figure 3. (a) k3-weighted EXAFS data of sorption samples and the reference compounds, As(V) solution, and mansﬁeldite; (b) corresponding Fourier transforms (not corrected for phase shift) showing both raw (black lines) and ﬁtted spectra (gray lines).

on γ-alumina was simulated in an attempt to mimic the reaction occurring at pH 5. The DFT calculations suggest the interatomic distances between As and Al for bidentate binuclear structures are 3.09 and 3.23 Å. This is distinct from the two monodentate models (Figure 4a,b), which give similar As Al distances (3.36 and 3.37 Å, respectively).

’ DISCUSSION Arsenate Bonding Structure on γ-Alumina. The interatomic distances of d(As O) and d(As Al) obtained by d-PDF, EXAFS, and DFT are compared in Table 1. The d(As O) determined by d-PDF is 1.66 Å, in good agreement with arsenate tetrahedron15,20,22 and EXAFS analysis (1.69 ( 0.01 Å). Unfortunately, it is difficult for d-PDF to provide coordination numbers, especially in a system with weak backscatterer (O and Al atoms). In principle, analyzing the peak intensities for As O and As Al correlations in the d-PDF could yield approximate coordination numbers. However, such analysis is prohibited due to the low signal-to-noise ratio. On the other hand, analysis of the EXAFS data suffers from the weak back scattering properties of Al atoms, resulting in large errors for the As Al coordination number ((40%). In such a situation, DFT calculations are useful to help the interpretation of d-PDF and EXAFS data. The predicted As O distances in all DFT calculated clusters (i.e., 1.68 1.79 Å) are in good agreement with those predicted by Ladeira et al.35 and Sherman and Randall,20 which are also similar to those experimentally measured by d-PDF and EXAFS. The calculated As Al distances for monodentate surface complexes connecting to either tetrahedral Al or octahedral Al are similar (∼3.36 Å) and are much larger than those measured by d-PDF and EXAFS (∼3.09 and 3.13 Å, respectively). In contrast, the bidentate structure (Figure 4c) gives two d(As Al) as 3.09 and

Figure 4. Optimized geometries of arsenate/alumina clusters calculated using density functional theory: (a) Biprotonated monodentate As coordinated with tetrahedral Al on (001) face; (b) biprotonated monodentate As coordinated with octahedral Al on (010) face; (c) biprotonated bidentate As coordinated with octahedral Al on (010) face. White balls represent H atoms, green for O, light blue for Al, and purple for As. Only a few atoms are shown for clarity. Graphics created with Materials Studio 5.5 (Accelrys Inc., San Diego CA).

3.23 Å. The averaged value (3.16 Å) is close to that from the EXAFS and d-PDF values (Table 1). The small differences between the averaged d(As Al) obtained by DFT and those from EXAFS/d-PDF are possibly due to the disordered nature of γ-alumina.32,33 In addition, the second peak at 3.09 Å in the d-PDF (Figure 2b) is as sharp as the first peak (As O correlation). This observation further implies that a monodentate structure is unlikely, 9690

dx.doi.org/10.1021/es200750b |Environ. Sci. Technol. 2011, 45, 9687–9692

Environmental Science & Technology

ARTICLE

interatomic distance method

sample/model

As O (Å)

As Al (Å)

d-PDF

1 mM, pH 5, dry

1.66

EXAFS

1 mM, pH 5, dry

1.69

3.13

0.1 mM, pH 5, wet 0.4 mM, pH 5, wet

1.69 1.69

3.12 3.15

1 mM, pH 5, wet

1.69

3.12

mono-Al[4]-(001)

1.79, 1.70, 1.78, 1.68

3.368

mono-Al[6]-(010)

1.70, 1.69, 1.79, 1.73

3.355

bi-Al[6]-(010)

1.69, 1.73, 1.77, 1.72

3.09, 3.23

DFT

3.09

Figure 5. Pair distribution functions of γ-alumina (γ-Al2O3), corundum (α-Al2O3), boehmite (γ-AlOOH), gibbsite (α-Al(OH)3), and bayerite (β-Al(OH)3).

since the As O Al bonds in a monodentate surface complex would have a large degree of freedom (i.e., rotation about the As O Al linkage) and lead to a large thermal factor and a broadening of this peak.16 It is worth noting that the average As Al distance (3.12 Å) measured for As adsorbed on γ-alumina is slightly shorter than those for As adsorbed on gibbsite35 and other Al-rich minerals36 (3.19 3.23 Å). This is possibly due to the differences in local structures among these Al (hydr)oxides as illustrated in the PDFs (Figure 5). The three samples with varied initial As concentrations have similar EXAFS features and fitting results, we therefore propose that arsenate forms bidentate binuclear structure from throughout the concentration range examined, i.e., from low surface loading (0.130 molecules per nm 2 or 0.216 μmol m 2) to high surface loading (1.25 molecules nm 2 or 2.08 μmol m 2). Comparison of d-PDF and EXAFS. In this work, similar interatomic distances were obtained from d-PDF and EXAFS measurement, providing strong support for our interpretation of arsenate bonding structure. As in inelastic X-ray spectroscopy, EXAFS suffers from signal loss with the increase of incident X-ray energy,37 resulting in a significant signal decay as the interatomic distance increases in the radial structural function (RDF), whereas PDF, intrinsically as a diffraction technique, is capable of avoiding such signal loss. This is one reason that the As Al correlation is more evident in the d-PDF spectra than that in EXAFS analysis.

The peaks in PDF have a full width at half-maximum (fwhm) of 0.22 0.24 Å 1, which are intrinsically much sharper than those in the RDF of EXAFS spectra (FMHW of 0.4 0.5 Å 1), such that the interatomic distances can be obtained straightforwardly. The drawback of d-PDF, especially at this developing stage, is the difficulty in estimating the errors associated with interatomic distances and coordination numbers, whereas EXAFS is a welldeveloped method that has systematic procedures for data analysis. Furthermore, EXAFS is exclusively sensitive to local structure,37 whereas d-PDF is sensitive to both local and intermediate structure.4 6 In most cases for interfacial studies, EXAFS can provide information for the first two shells (atomic correlations) in its RDF, with r less than 5 Å. Previous d-PDF study16 of arsenate adsorbed on 2-line ferrihydrite showed that atomic correlations can be observed to higher r (up to ∼7 Å). This present study is less favorable for two reasons: ﬁrst, the particles are larger (20 nm of γ-alumina particle vs 2 nm of 2-line ferrihydrite), so that less surfaces area is exposed to the X-ray beam for a given exposed volume. Second, Al has only one-half the scattering power of Fe (Z of 13 vs 26). Accordingly, signal-to-noise is much lower, reducing the amount of information that can be generated from the d-PDF in practice. From this, it can be deduced that there are three criteria that contribute to an optimal d-PDF experiment: (1) large speciﬁc surface area, (2) high scattering power of the atoms involved, and (3) a high surface loading of the adsorbed species (e.g., no signal was observed from a sample exposed to a 0.1 mM arsenate solution). As EXAFS suﬀers less from potential problems 1 and 3 and d-PDF has the potential to show correlations to higher r value, a combination of these techniques will be a good strategy to investigate environmental interfacial phenomena.

’ ASSOCIATED CONTENT

bS

Supporting Information. Detailed descriptions are available on EXAFS analysis and ﬁtting results, quantum chemical calculations, characterization of γ-alumina, comparison of diﬀerential PDFs obtained at diﬀerent Qmax values, and comparison of sorption capacities of various sorbants. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected] (W.L.); [email protected] (R.H.); [email protected] (Y.T.). Present Addresses

^ Environmental Soil Chemistry Group, Delaware Environmental Institute, University of Delaware, Newark, DE, 19716, USA. z School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, 02318, USA. Department of Mechanical Engineering, Stanford University, Stanford CA, 94305, USA.

)

Table 1. Comparison of the Interatomic Distances Obtained from Diﬀerential PDF, EXAFS, and DFT Calculation

’ ACKNOWLEDGMENT We sincerely appreciate three anonymous reviewers for their helpful comments. We thank Dr. Douglas B. Hausner for collecting the TEM images, Dr. Henry Pinto for supplying the structure ﬁle for the γ-Al2O3 crystal, and Dr. Wenqian Xu for discussions on the diﬀerence between EXAFS and d-PDF. Work done at 9691

dx.doi.org/10.1021/es200750b |Environ. Sci. Technol. 2011, 45, 9687–9692

Environmental Science & Technology Argonne and use of the Advanced Photon Source was supported by the U.S. Department of Energy, Oﬃce of Science, Oﬃce of Basic Energy Sciences, under Contract No. DE-AC02-06CH11357. Financial support was provided by the National Science Foundation (NSF) through Collaborative Research in Chemistry (CRC), Grants CHE0714183 and CHE0714173.

’ REFERENCES (1) Sparks, D. L. Environmental Soil Chemistry, 2nd ed.; Academic Press: Boston, 2002. (2) Brown, G. E., Jr.; Parks, G. A. Sorption of trace elements from aqueous media: Modern perspectives from spectroscopic studies and comments on adsorption in the marine environment. Int. Geol. Rev. 2001, 43, 963–1073. (3) Ginder-Vogel, M.; Sparks, D. L. The impacts of X-ray absorption spectroscopy on understanding soil processes and reaction mechanisms. In Developments in Soil Science 34: Synchrotron-based Techniques in Soils and Sediments; Singh, B., Grafe, M., Eds.; Elsevier: Burlington, 2010; pp 1. (4) Egami, T.; Billinge, S. J. L. Underneath the Bragg peaks: Structural analysis of complex materials; Pergamon Press/ Elsevier: Oxford, 2003. (5) Billinge, S. J. L.; Kanatzidis, M. G. Beyond crystallography: the study of disorder, nanocrystallinity and crystallographically challenged materials with pair distribution functions. Chem. Commun. 2004, 749– 760. (6) Petkov, V. Nanostructure by high-energy X-ray diﬀraction. Mater. Today 2008, 11, 28–38. (7) Harrington, R.; Neder, R. B.; Parise, J. B. The nature of X-ray scattering from geo-nanoparticles: practical considerations of the use of the Debye equation and the pair distribution function for structure analysis. Chem. Geol., doi:10.1016/j.chemgeo.2011.06.010. (8) Proﬀen, T.; Page, K. L. Obtaining structural information from the atomic pair distribution function. Z. Kristallogr. 2004, 219, 130–135. (9) Proﬀen, T.; Kim, H. Advances in total scattering analysis. J. Mater. Chem. 2009, 19, 5078–5088. (10) Chapman, K. W.; Chupas, P. J.; Maxey, E. R.; Richardson, J. W. Direct observtion of adsorbed H2-framework interactions in the Prussian Blue analogue MnII3[CoIII(CN)6]2: The relative importance of accessible coordination sites and van der Waals interactions. Chem. Commun. 2006, 4013–4015. (11) Chapman, K. W.; Chupas, P. J.; Kepert, C. J. Selective recovery of dynamic guest structure in a nanoporous Prussian Blue through in situ X-ray diﬀraction: A diﬀerential pair distribution function analysis. J. Am. Chem. Soc. 2005, 127, 11232–11233. (12) Kramer, M. J. A strategy for rapid analysis of the variations in the reduced distribution function of liquid metals and metallic glasses. J. Appl. Crystallogr. 2007, 40, 77–86. (13) Chupas, P. J.; Chapman, K. W.; Jennings, G.; Lee, P. L.; Grey, C. P. Watching nanoparticles grow: The mechanism and kinetics for the formation of TiO2-supported platinum nanoparticles. J. Am. Chem. Soc. 2007, 129, 13822–13824. (14) Chupas, P. J.; Chapman, K. W.; Chen, H.; Grey, C. P. Application of high-energy X-rays and pair-distribution-function analysis to nano-scale structural studies in catalysis. Catal. Today 2009, 145, 213–219. (15) Waychunas, G. A.; Fuller, C. C.; Rea, B. A.; Davis, J. A. Wide angle X-ray scattering (WAXS) study of ‘two-line’ ferrihydrite structure: eﬀect of arsenate sorption and counterion variation and comparison with EXAFS results. Geochim. Cosmochim. Acta 1996, 60, 1765–1781. (16) Harrington, R.; Hausner, D. B.; Bhandari, N.; Strongin, D. R.; Chapman, K. W.; Chupas, P. J.; Middlemiss, D. S.; Grey, C. P.; Parise, J. B. Investigations of surface structures by powder diﬀraction: a diﬀerential pair distribution function study on arsenate sorption on ferrihydrite. Inorg. Chem. 2010, 49, 325–330. (17) Michel, F. M.; Ehm, L.; Antao, S. M.; Lee, P. L.; Chupas, P. J.; Li, G.; Strongin, D. R.; Schoonen, M. A. A.; Phillips, B. L.; Parise, J. B. The structure of ferrihydrite, a nanocrystalline material. Science 2007, 316, 1726–1729.

ARTICLE

(18) Xu, W.; Hausner, D. B.; Harrington, R.; Lee, P. L.; Strongin, D. R.; Parise, J. B. Structural water in ferrihydrite and constraints this provides on possible structure models. Am. Mineral. 2011, 96, 513–520. (19) Waychunas, G. A.; Rea, B. A.; Fuller, C. C.; Davis, J. A. Surface chemistry of ferrihydrite: Part 1. EXAFS studies of the geometry of coprecipitated and adsorbed arsenate. Geochim. Cosmochim. Acta 1993, 57, 2251–2269. (20) 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. (21) Huang, C. P.; Stumm, W. Speciﬁc adsorption of cations on hydrous γ-Al2O3. J. Colloid Interface Sci. 1973, 43, 409–420. (22) Arai, Y.; Elzinga, E. J.; Sparks, D. L. X-ray absorption spectroscopic investigation of arsenite and arsenate adsorption at the aluminum oxide-water interface. J. Colloid Interface Sci. 2001, 235, 80–88. (23) Strawn, D. G.; Scheidegger, A. M.; Sparks, D. L. Kinetics and mechanisms of Pb(II) sorption and desorption at the aluminum oxidewater interface. Environ. Sci. Technol. 1998, 32, 2596–2601. (24) Fitts, J. P.; Brown, G. E., Jr.; Parks, G. A. Structural evolution of Cr(III) polymeric species at the γ-Al2O3-water interface. Environ. Sci. Technol. 2000, 34, 5122–5128. (25) Tang, Y.; Reeder, R. J. Enhanced uranium sorption on aluminum oxide pretreated with arsenate. Part I: batch uptake behavior. Environ. Sci. Technol. 2009, 43, 4446–4451. (26) Chupas, P. J.; Qiu, X.; Hanson, J. C.; Lee, P. L.; Gery, C. P.; Billinge, S. J. L. Rapid-acquisition pair distribution function (RA-PDF) analysis. J. Appl. Crystallogr. 2003, 36, 1342–1347. (27) Chupas, P. J.; Chapman, K. W.; Lee, P. L. Applications of an amorphous silicon-based area detector for high-resolution, high-sensitivity and fast time-resolved pair distribution function measurements. J. Appl. Crystallogr. 2007, 40, 463–470. (28) Hammersley, A. P.; Svenson, S. O.; Hanﬂand, M.; Hauserman, D. Two-dimensional detector software: from real detector to idealised image or two-theta scan. High Pressure Res. 1996, 14, 235–248. (29) Qui, X.; Thompson, J. W.; Billinge, S. J. L. PDFgetX2: a GUIdriven program to obtain the pair distribution function from X-ray powder diﬀraction data. J. Appl. Crystallogr. 2004, 37, 678. (30) Sun, M.; Nelson, A. E.; Adjaye, J. Examination of spinel and nonspinel structural models for γ-Al2O3 by DFT and Rietveld reﬁnement simulations. J. Phys. Chem. B 2006, 110, 2310–2317. (31) Mohan, D.; Pittman, C. U., Jr. Arsenic removal from water/ wastewater using adsorbents A critical review. J. Hazard. Mater. 2007, 142, 1–53. (32) Makris, K. C.; Sarkar, D.; Parsons, J. G.; Datta, R.; GardeaTorresdey, J. L. X-ray absorption spectroscopy as a tool investigating arsenic(III) and arsenic(V) sorption by an aluminum-based drinkingwater treatment residual. J. Hazard. Mater. 2009, 171, 980–986. (33) Stumm, W.; Morgan, J. J. Aquatic Chemistry: Chemical equilibria and rates in natural waters, 3rd ed.; John Wiley & Sons: New York, 1996. (34) Pinto, H. P.; Nieminen, R. M.; Elliott, S. D. Ab initio study of γ-Al2O3 surfaces. Phys. Rev. B 2004, 70 (12), 125402-1–125402-11. (35) Ladeira, A. C. Q.; Ciminelli, V. S. T.; Duarte, H. A.; Alves, M. C. M.; Ramos, A. Y. Mechanism of anion retention from EXAFS and density functional calculations: Arsenic (V) adsorbed on gibbsite. Geochim. Cosmochim. Acta 2001, 65, 1211–1217. (36) Beaulieu, B. T.; Savage, K. S. Arsenate adsorption structures on aluminum oxide and phyllosilicate mineral surfaces in smelter-impacted soils. Environ. Sci. Technol. 2005, 39, 3571–3579. (37) Rehr, J. J.; Albers, R. C. Theoretical approaches to X-ray absorption ﬁne structure. Rev. Mod. Phys. 2000, 72, 621–654.

9692

dx.doi.org/10.1021/es200750b |Environ. Sci. Technol. 2011, 45, 9687–9692

ARTICLE pubs.acs.org/est

Quantifying the Effects of 1,1,1-Trichloroethane and 1,1-Dichloroethane on Chlorinated Ethene Reductive Dehalogenases Winnie W. M. Chan,†,|| Ariel Grostern,†,^ Frank E. L€oﬄer,‡,§ and Elizabeth A. Edwards*,† †

Department of Chemical Engineering and Applied Chemistry, University of Toronto, 200 College Street, Toronto, ON, Canada, M5S 3E5 ‡ Department of Microbiology and Department of Civil and Environmental Engineering, University of Tennessee, Knoxville, Tennessee, 37996, United States § Biosciences Division, Oak Ridge National Laboratory, Oak Ridge, Tennessee, 37831, United States

bS Supporting Information ABSTRACT: Mixtures of chlorinated ethenes and ethanes are often found at contaminated sites. In this study, we undertook a systematic investigation of the inhibitory eﬀects of 1,1,1-trichloroethane (1,1,1-TCA) and 1,1-dichloroethane (1,1-DCA) on chlorinated ethene dechlorination in three distinct Dehalococcoides-containing consortia. To focus on inhibition acting directly on the reductive dehalogenases, dechlorination assays used cell-free extracts prepared from cultures actively dechlorinating trichloroethene (TCE) to ethene. The dechlorination assays were initiated with TCE, cis-1,2-dichloroethene (cDCE), or vinyl chloride (VC) as substrates and either 1,1,1-TCA or 1,1-DCA as potential inhibitors. 1,1,1-TCA inhibited VC dechlorination similarly in cell suspension and cell-free extract assays, implicating an eﬀect on the VC reductases associated with the dechlorination of VC to nontoxic ethene. Concentrations of 1,1,1-TCA in the range of 30 270 μg/L reduced VC dechlorination rates by approximately 50% relative to conditions without 1,1,1-TCA. 1,1,1-TCA also inhibited reductive dehalogenases involved in TCE and cDCE dechlorination. In contrast, 1,1-DCA had no pronounced inhibitory eﬀects on chlorinated ethene reductive dehalogenases, indicating that removal of 1,1,1-TCA via reductive dechlorination to 1,1-DCA is a viable strategy to relieve inhibition.

’ INTRODUCTION Chlorinated ethenes and ethanes are prevalent groundwater contaminants. Bioaugmentation using cultures with enhanced dechlorination capabilities has been used for the in situ biotransformation of chlorinated ethenes to nontoxic ethene1 3 and more recently has also been applied to chlorinated ethanes.1 However, bioremediation eﬀorts are confounded when both chlorinated ethenes and ethanes are present at a site. In the presence of 1,1, 1-trichloroethane (1,1,1-TCA), ethene production is slowed and toxic dechlorination intermediates cis-1,2-dichloroethene (cDCE) and vinyl chloride (VC) accumulate,2 4 considerably aﬀecting bioremediation performance. Bacteria from several genera, including Dehalobacter, Desulfuromonas, Desulﬁtobacterium, Sulfurospirillum, Geobacter, and Dehalococcoides, gain energy from reductive dechlorination of tetrachloroethene (PCE) and trichloroethene (TCE) to cDCE,8 11 but only certain Dehalococcoides strains have been shown to convert cDCE to VC and VC to ethene. Thus, bioremediation strategies that increase Dehalococcoides populations at a site, particularly those strains harboring VC reductase genes (vcrA or bvcA), are highly eﬀective.5 9 However, in the presence of 1,1,1-TCA, r 2011 American Chemical Society

chlorinated ethene dechlorination has been observed to stall at cDCE and VC, suggesting inhibition of Dehalococcoides.2,3 Previous modeling eﬀorts that applied Monod kinetics to TCE-to-ethene-dechlorinating enrichment cultures revealed that highly chlorinated ethenes competitively inhibited the transformation of cDCE and VC,10,11 and that substrate inhibition of PCE and TCE dechlorination occurred when the compounds were present at concentrations near aqueous solubility limits (i.e., near chlorinated solvent source zones).12 In this study, we investigated the eﬀects of 1,1,1-TCA and its daughter product 1,1-dichloroethane (1,1-DCA) on TCE, cDCE, and VC dechlorination in cell suspension and cell-free extract assays. This strategy permitted the determination of whether chlorinated ethanes inhibited Dehalococcoides dehalogenases catalyzing the reductive dechlorination of chlorinated ethenes directly, or if they exerted inhibitory eﬀect at the general cellular-level or on other, nondechlorinating, microorganisms in the cultures. Three distinct Received: April 13, 2011 Accepted: September 28, 2011 Revised: September 9, 2011 Published: September 28, 2011 9693

dx.doi.org/10.1021/es201260n | Environ. Sci. Technol. 2011, 45, 9693–9702

Environmental Science & Technology Dehalococcoides-containing mixed enrichment cultures and two subcultures containing either only Dehalococcoides or Geobacter as dechlorinators were compared. Unlike previous studies where inhibitors were also substrates, neither 1,1,1-TCA nor 1,1-DCA were substrates for the chlorinated ethene-dechlorinating cultures tested, facilitating quantiﬁcation of speciﬁc interactions.

’ MATERIALS AND METHODS Cultures. Three TCE-to-ethene dechlorinating consortia, KB-1, Bio-Dechlor Inoculum (BDI), and OW, enriched from geographically distinct locations, were used in these experiments. KB-1, which is sold commercially for bioaugmentation by SiREM (Guelph, ON), was originally enriched from a TCE-contaminated site in southern Ontario using TCE as electron acceptor and methanol as an electron donor.2 KB-1 contains a dechlorinating Geobacter strain and two strains of Dehalococcoides.13 Consortium BDI contains the PCE-to-cDCE-dechlorinators Geobacter lovleyi strain SZ, a Dehalobacter sp., as well as Dehalococcoides strains GT, BAV1, and FL2.14 In this study, the BDI culture was maintained with TCE and hydrogen and was composed of the following proportions of Dehalococcoides strains: 70% GT, 30% FL2, and less than 1% BAV1.20,21 Culture OW, originally established with sludge from a wastewater treatment facility in Texas, was enriched with PCE and methanol,15 but was maintained with TCE and methanol for seven feedings prior to biomass collection. Culture OW contains both Dehalobacter and Dehalococcoides spp.15 All three consortia contain vcrA, whose partially purified protein product was shown to catalyze cDCE-to-VC dechlorination and VC-to-ethene dechlorination.7 However, each culture contained a mix of different dechlorinating and nondechlorinating populations and a different combination of reductive dehalogenases (see Table 3). As such, the three cultures were selected to shed light on how a spectrum of different dehalogenases in different microbial contexts were affected by inhibitors. Cultures were grown in their respective defined mineral media as described for KB-1,16 BDI,14 and OW.15 Two KB-1 subcultures were also investigated. KB-1/Dhc18,24 was maintained on VC and H2 for 10 years, which eliminated Geobacter and therefore the only dechlorinating organism in this subculture was Dehalococcoides. This subculture was amended with TCE and methanol for one feeding prior to biomass collection, to mimic conditions in other experiments with KB-1. KB-1/ Geo is a subculture of KB-1 that was highly enriched (>99%) in a Geobacter sp. phylogenetically related to Geobacter lovleyi strain SZ, a PCE-to-cDCE dechlorinator.17 KB-1/Geo was enriched by ﬁve successive 1% transfers in mineral medium containing 10 mM sodium acetate, 0.54 mM PCE, and a H2-free, N2/CO2 (80:20) headspace, a method similar to that used to isolate strain SZ. Cell Suspensions. Cultures were prepared for cell suspension assays by purging each culture free of chlorinated ethenes with N2/CO2 (80:20) for 15 20 min. Setup of cell suspension assays involved a dilution of the cultures into mineral salts medium as described below. Cell-Free Extracts. All cell-free extracts were prepared under anoxic conditions from cultures actively dechlorinating TCE, which had been fed within 48 72 h. For KB-1, OW, and BDI, cell-free extracts were prepared once ethene production began. Centrifuge tubes used in the procedure were left in the anaerobic chamber (Coy Laboratory Products Inc., MI) for 48 h prior to use in order to remove oxygen, and anaerobic conditions during centrifugation were monitored by observing the color of the resazurin-containing medium. To prepare cell-free extracts of KB-1,

ARTICLE

600 mL of culture was centrifuged in three 250-mL polypropylene bottles with cap assembly (Beckman Coulter, Brea, CA) at 9900g for 30 min at 4 C. In the anaerobic chamber the triplicate cell pellets were resuspended in a single final volume of 40 mL of suspension buffer (100 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM titanium(III) citrate, 5% (v/v) glycerol). Cells were broken by sonication (Sonics Vibra-Cell sonicator, Sonics & Materials, Inc., Newton, CT) on ice in the anaerobic chamber using two 5-min cycles of 1-s pulses at 40% amplitude (23 W), separated by a 2-min break. Crude extracts were aliquoted into smaller volumes in 2-mL tubes with screw-top O-ring caps, flashfrozen in liquid N2, and stored at 80 C. For each assay, an aliquot of the extract was thawed on ice and centrifuged for 10 min at 12 500g at 4 C. Inside the anaerobic chamber, the thawed extract was passed through a 0.2-μm filter to remove unbroken cells. Cell-free extracts of BDI, OW, KB-1/Dhc, and KB-1/Geo were prepared by concentrating 40 mL of culture (80 mL for KB-1/ Geo) by centrifugation in 50-mL polypropylene tubes (Fisher Scientiﬁc, Pittsburgh, PA) sealed with vinyl tape (3M, St. Paul, MN) at 5500g for 30 min at 4 C. Pellets were suspended in suspension buﬀer to a ﬁnal volume of 5.5 mL. After sonication, the suspensions were centrifuged as described above for 15 min at 12 500g at 4 C to pellet unbroken cells. Inside the anaerobic chamber, the thawed extract was passed through a 0.2-μm ﬁlter to remove unbroken cells. The supernatant was aliquoted into smaller volumes, ﬂash-frozen, and stored at 80 C. Crude extracts were thawed on ice prior to the start of each assay. Cell Suspension and Cell-Free Extract Dechlorination Assays. Assays were performed in the anaerobic chamber in 2-mL glass vials with PTFE-lined caps (Supelco, Bellefonte, PA). The vials were filled completely to eliminate headspace partitioning of the chlorinated compounds. Cell suspension assays were conducted in 1.87 mL of mineral salts medium16 supplemented with 5 mM acetate and purged with H2/CO2 (80%/20%) immediately prior to use. Cell-free extract assays were conducted in 1.90 mL of assay buffer containing 100 mM Tris-HCl, pH 7.4, 2 mM methyl viologen, and 2 mM titanium(III) citrate. 1,1,1-TCA or 1,1-DCA and cell-free extracts or cell suspensions were added directly to the assay buffer, whereas 10 50 μL of chlorinated ethenes (from saturated aqueous stocks) were added at different concentrations directly to the vial to start each assay. Saturated aqueous stocks were prepared in 2-mL vials by adding to anaerobic water 5-fold greater mass of the chlorinated solvent needed to obtain the solubility limit. Vials were shaken by hand for 2 min and the aqueous and solvent phases were allowed to partition overnight. For cell suspension assays, purged culture was added to the assay buﬀer to obtain, after resuspension in buﬀer, 75 μL of purged culture per assay vial, for a ﬁnal total volume in the cell suspension assay vial of between 1.96 and 1.99 mL. For cell-free extract assays, cell-free extract was added to the assay buﬀer to obtain the equivalent of 10 30 μL of extract per assay vial, for a ﬁnal total volume in the cell suspension assay vial of between 1.92 and 1.98 mL. These amounts of culture or extract were chosen such that accurate determination of daughter products could be made within a 1 3 h incubation period while maintaining less than 10% dechlorination of parent compound. This strategy provided an accurate determination of initial dechlorination rates without a signiﬁcant change in initial chlorinated ethene concentration. Concentrations of TCE, cDCE, and VC tested (up to 10 concentrations per substrate/inhibitor combination) ranged from 2 to 600 μM, corresponding to 10 55 μL of aqueous stocks. Chlorinated ethane concentrations assayed ranged from 0 9694

dx.doi.org/10.1021/es201260n |Environ. Sci. Technol. 2011, 45, 9693–9702

Environmental Science & Technology

ARTICLE

Table 1. Kinetic (Vmax, Km) and Inhibition (KI) Parameters for VC and cDCE and TCE Dechlorination in Either Cell Suspensions or Cell-Free Extracts in the Presence of Chlorinated Ethanesa substrate VC

culture KB-1

OW

BDI

cDCE

KB-1 OW BDI

TCE

KB-1 KB-1/Geo

inhibitor

assay

Vmax (nmol 3 min

1

1 3 mg protein )

Km (μM)

KI (μM)

inhibition model

1,1,1-TCA

suspension

50 ((3)

27 ((8)

0.7 ((0.2)

C

1,1,1-TCA 1,1,1-TCA

extract extract

47 ((3) 49 ((4)

74 ((14) 83 ((15)

0.8 ((0.2) 2.0 ((0.3)

C N

1,1-DCA

extract

67 ((5)

63 ((15)

110 ((56)

C

1,1-DCA

extract

73 ((6)

76 ((14)

300 ((111)

N C

1,1,1-TCA

suspension

5.8 ((0.5)

7 ((2)

0.2 ((0.1)

1,1,1-TCA

extract

2.8 ((0.6)

151 ((54)

1.1 ((0.3)

C

1,1,1-TCA

extract

2.9 ( 0.5

156 ( 51

2.0 ( 0.4

N

1,1-DCA

extract

2.3 ((0.2)

105 ((23)

58 ((24)

C

1,1-DCA 1,1,1-TCA

extract suspension

2.5 ( 0.2 13 ((1)

116 ( 16 17 ((4)

104 ( 24 0.4 ((0.1)

N C

1,1,1-TCA

extract

9.1 ((0.6)

81 ((11)

0.5 ((0.1)

C

1,1,1-TCA

extract

9.2 ( 0.7

84 ( 13

1.2 ( 0.2

N

1,1-DCA

extract

9.3 ((0.5)

95 ((11)

80 ((22)

C

1,1-DCA

extract

9.8 ( 0.5

105 ( 11

162 ( 39

N

1,1,1-TCA

extract

91 ((7)

86 ((22)

19 ((4)

N

1,1-DCA

extract

91 ((4)

92 ((13)

830 ((280)

N

1,1,1-TCA 1,1-DCA

extract extract

26 ((1) 25 ((1)

42 ((5) 40 ((6)

86 ((17) 130 ((57)

N N

1,1,1-TCA

extract

15 ((1)

43 ((8)

5.5 ((0.8)

N

1,1-DCA

extract

18 ((1)

45 ((4)

110 ((18)

N N

1,1,1-TCA

extract

82 ((13)

40 ((19)

1.5 ((0.6)

1,1-DCA

extract

82 ((16)

98 ((41)

No inhibition

N

1,1,1-TCA

extract

84 ((10)

1.4 ((0.9)

5.1 ((2)

N

KB-1/Dhc

1,1,1-TCA

extract

179 ((20)

180 ((40)

2.2 ((0.6)

N

OW BDI

1,1,1-TCA 1,1,1-TCA

extract extract

25 ((3) 53 ((5)

54 ((14) 3 ((1)

40 ((9) 43 ((17)

N N

For the substrate VC in cell suspensions, the competitive model ﬁt the data best. For cell-free extracts, the noncompetitive model was usually the best ﬁt for all substrates (see SI Table S3 for model ﬁt data). Both competitive and noncompetitive model parameters are provided for VC extract data to enable comparison. Error values represent 95% conﬁdence intervals. C = competitive inhibition model; N = noncompetitive inhibition model. a

to 38 μM for 1,1,1-TCA and 0 to 225 μM for 1,1-DCA. All of the substrate and inhibitor combinations tested in dechlorination assays are listed in Table 1. Inhibitor and initial substrate concentrations for each experiment are provided in Supporting Information (SI) Table S1. Analytical Procedures. Following the 1 3 h incubation period, sacrificial liquid samples (0.15 1.0 mL) from each dechlorination assay vial were transferred to 10-mL autosampler vials containing 5.0 5.85 mL of acidified water (12 mM HCl) to stop further enzymatic activity, to a total volume of 6 mL. Samples were analyzed with a HP 7694 headspace sampler connected to a HP 5890A gas chromatograph coupled to a flame ionization detector (GC-FID), with a GSQ column (30 m 0.53 mm i.d. PLOT column; J&W Scientific, Folsom, CA). The headspace autosampler settings were as follows: 75 C oven temperature, 80 C loop temperature, 90 C transfer line temperature, 12 min GC cycle time, 45 min vial equilibration time, 0 min pressurization time, 0.2 min loop fill time, 0 min loop equilibration time, 3 min injection time, vial pressure at 17.3 psi, and carrier pressure at 9.4 psi. The GC oven temperature program used was as follows: hold at 50 C for 2 min, increase to 100 at 50 C/min, increase to 185 at 25 C/min, and hold at 185 C for 3 min. When dechlorination assays involved testing TCE as a substrate and 1,1-DCA as an inhibitor, the following program was used to resolve coeluting

1,1-DCA and daughter product cDCE peaks: hold at 50 C for 2 min, increase to 100 at 50 C/min, increase to 150 at 25 C/min, hold for 2.5 min, increase to 185 at 15 C/min, and hold at 185 C for 0.5 min. Protein concentrations were determined in cell-free extracts and cell suspensions using the Bradford assay.18 Kinetic and Inhibition Models. For each experiment at a different initial substrate concentration [S], an initial dechlorination rate v, normalized to the amount of protein per vial, in units of nmol substrate dechlorinated per min per mg protein, was calculated. The moles of substrate dechlorinated were calculated as the sum of all daughter products measured by GC-FID. The maximum dechlorination rate (Vmax) and half-velocity constant (Km) for each culture/substrate combination were calculated using a nonlinear regression method for the Michaelis Menten single-substrate model in the Enzyme Kinetics 1.3 Module for SigmaPlot 10 (Systat Software Inc., Chicago, IL). Haldane substrate inhibition19 for chlorinated ethene dechlorination was not observed at the concentrations tested. Data sets from experiments testing the effect of a chlorinated ethane were modeled using the competitive, uncompetitive, and noncompetitive inhibition equations (SI Table S2) as described previously.18 The model deemed most appropriate was chosen primarily based on the statistical parameters determined using SigmaPlot. In some cases, these parameters (highest coeﬃcient of determination 9695

dx.doi.org/10.1021/es201260n |Environ. Sci. Technol. 2011, 45, 9693–9702

Environmental Science & Technology

Figure 1. Kinetics of VC dechlorination in cell suspensions in the presence of increasing concentrations of 1,1,1-TCA. Three consortia are compared: (A) KB-1, (B) OW, and (C) BDI. The concentration of 1,1,1-TCA for each assay series is indicated on each graph (I = inhibitor concentration in μM). Solid lines represent the best ﬁt to each data set based on nonlinear regression using a competitive inhibition model (eq 2 in SI Table S2).

(R2), the lowest corrected Akaike’s Information Criterion (AICc), and the lowest standard deviation of the residuals (Sy.x)) were not able to provide enough resolution to choose a prevailing model. In these cases, we selected a model to maintain consistency among substrate inhibitor sets in order to provide a meaningful comparison between the inhibition coeﬃcients. All raw data (Table S1) and statistical parameters (Table S3) are provided as Supporting Information. For evaluation of the data from experiments involving KB-1 culture, KB-1/Geo, and KB-1/Dhc, an Eadie Hofstee transformation of the kinetic data was also generated. An Eadie Hofstee transformation plots the dechlorination rate as a function of the ratio of dechlorination rate to substrate concentration using a linearization of the Michaelis Menten equation, enabling easy visualization of model ﬁt.20

’ RESULTS AND DISCUSSION Cell-Free Extract and Cell Suspension Control Experiments. No differences in dechlorination rates were observed

ARTICLE

between freshly prepared cell-free extracts and cell-free extracts that had been stored at 80 C (data not shown). Therefore, all experiments with the different substrates and inhibitors could be completed with extract of identical protein content and activity. No TCE dechlorination was observed when hydrogen or methanol was added as electron donors to cell-free extracts, indicating the electron transport pathway leading to the reductive dehalogenase (which acts as a terminal reductase in respiration) was no longer intact. Furthermore, dechlorination was only observed in the presence of the artificial electron donor methyl viologen reduced with titanium citrate. The addition of up to 1 μM of vitamin B12 (cyanocobalamin) (over 200-fold higher than the cyanocobalamin composition in the mineral medium) to heatdenatured (10 min at 85 C) cell-free extract did not result in any reductive dechlorination, indicating that observed TCE dechlorination resulted from the activity of reductive dehalogenases and not simply from any dechlorinating activity of reduced vitamin B12.21 Finally, no transformation of 1,1,1-TCA or 1,1-DCA was observed in any of the dechlorination assays. Initial Rates of Dechlorination versus Initial Substrate Concentration. Data for cell suspension assays amended with the substrate VC and increasing concentrations of 1,1,1-TCA are shown in Figure 1. Data for cell-free extract assays with the substrate VC in the presence of 1,1,1-TCA and 1,1-DCA are presented in Figure 2. These graphs reveal reasonable agreement with Michaelis Menten kinetics (i.e., rate increasing to a maximum). Similar patterns were observed for cell-free extract assays with the substrate cDCE in the presence of 1,1,1-TCA and 1,1-DCA (SI Figure S1), and for assays with the substrate TCE in the presence of 1,1,1-TCA (SI Figure S2). The data plotted in Figures 1, 2, and S1 also reveal that chlorinated ethene dechlorination rates decreased with increasing 1,1,1-TCA concentration, but were not significantly affected by similar concentrations of 1,1-DCA. Quantification of Kinetic Parameters. The raw data (initial degradation rate (v) vs substrate concentration [S], see SI Table S1) were fit to the competitive, noncompetitive, and uncompetitive inhibition equations to extract kinetic parameters for quantitative comparisons. For data with VC as a substrate, the competitive model fit best, particularly for cell suspension assays (SI Table S3). For cDCE and TCE as substrates in cell-free extract assays, the noncompetitive model fit best in most cases (SI Table S3). The solid lines in Figures 1, 2, S1, and S2 represent these model equations using a single best set of parameters (Vmax, Km, and KI) for each complete data set (i.e., all concentrations for each substrate/inhibitor combination tested). The resulting kinetic parameters (Tables 1 and 2) enabled quantitative comparisons among cultures, substrates, and inhibitors. Comparisons between cell suspensions and cell-free extracts with VC as substrate were made using parameters from the competitive model because the competitive model clearly fit the cell suspension data better, but comparisons between all cell-free extract experiments were made using parameters from the noncompetitive model to compare these data for the same model. The choice of model (competitive vs noncompetitive) only affected the determination of the inhibition constant, KI, and did not significantly change the values of Vmax or Km (Table 1). Because the cell suspension assays were conducted over a short time frame (<3 h) during which negligible cell growth would occur (Dehalococcoides doubling time is approximately 2 days22), we deemed that it was reasonable to analyze both cell-free extract and cell suspension data with the Michaelis Menten Model. For this reason, Km, 9696

dx.doi.org/10.1021/es201260n |Environ. Sci. Technol. 2011, 45, 9693–9702

Environmental Science & Technology

ARTICLE

Figure 2. Kinetics of VC dechlorination in cell-free extracts in the presence of increasing concentrations of 1,1,1-TCA or 1,1-DCA. Cell-free extracts were prepared from (A B) KB-1, (C D) OW, and (E F) BDI. Solid lines represent the best ﬁt to each data set based on nonlinear regression using a noncompetitive inhibition model (eq 2 in SI Table S2). I = inhibitor concentration in μM.

rather than the Monod half-velocity constant, KS, for systems involving cell growth, was used in the analyses of cell suspension kinetic parameters. Dechlorination in the Absence of Inhibitor. The maximum rate of dechlorination Vmax and Michaelis Menten constant Km for each culture and chlorinated ethene substrate characterize the reaction in the absence of inhibitor. In mixed cultures, Vmax is a strong function of the cell numbers of the dechlorinating populations (and therefore enzyme concentration), so Vmax cannot be used to compare between different mixed cultures. However, Vmax values for different substrates can be compared within a given culture. For KB-1, Vmax values were similar for all three substrates tested (TCE, cDCE, and VC), whereas for BDI and OW Vmax was highest for TCE and lowest for VC (Table 1). Km values provide a measure of the aﬃnity of the enzyme for the substrate—the lower the Km, the higher the aﬃnity. We noted that the Km determined in cell suspensions for VC was signiﬁcantly lower (3 20 times) than the Km in corresponding cell-free extracts (Table 1), regardless of the choice of model for the cell-free extract data (competitive or noncompetitive). The apparent higher aﬃnity in the cell suspensions is puzzling, but could be the result of conformational changes in the enzymes. In Dehalococcoides, reductive dehalogenases are membrane-bound enzymes13 whose environment may be disrupted during preparation of the extract. Another signiﬁcant diﬀerence between cell-free extract and cell suspension assays was that the electron donor in cell suspension assays was hydrogen, while in cell-free

extracts it was reduced methyl viologen. Because electron transport to the reductive dehalogenase is not fully understood, these data are diﬃcult to interpret; however we observed a similar pattern of increased Km in cell-free extracts relative to cell suspensions in a Dehalobacter-containing 1,1,1-TCA dechlorinating culture.18 It would be interesting to test if this same trend holds up with the substrates cDCE and TCE as well. In the present work, the values of the inhibition constant KI were not signiﬁcantly diﬀerent between cell suspensions and cell-free extracts (Table 1; e.g., 0.7 vs 0.8 μM for KB-1) when calculated with the competitive kinetic model, even though the Km values were diﬀerent (Table 1; e.g., 27 vs 74 μM for KB-1). Additional work to determine the root cause of the variation of Km and KI between cell suspensions and cell-free extracts is needed. Focusing on the data from cell-free extract assays and only the noncompetitive model, since it ﬁt the data best (Table 1), the half-velocity constant Km values across cultures for the substrates VC and cDCE vary only within a factor of about two (76 156 μM for VC and 40 92 μM for cDCE). In contrast, Km values for TCE vary by a factor of about 100 between cultures (from 1.4 to 180 μM). The relative consistency of the VC and cDCE dechlorination results as compared to the TCE results agrees with cDCE and VC steps being catalyzed only by Dehalococcoides, whereas multiple genera, including Geobacter, Dehalobacter, and Dehalococcoides, with diverse reductive dehalogenases may be involved in TCE to cDCE dechlorination in these cultures. The lowest Km values for TCE were observed for KB-1/Geo and BDI; Geobacter 9697

dx.doi.org/10.1021/es201260n |Environ. Sci. Technol. 2011, 45, 9693–9702

Environmental Science & Technology

ARTICLE

Table 2. Summary of All Culture, Substrate, and Inhibitor Combinations Tested Ranked by KI from Lowest to Highest, in Both μM and μg/L (ppb)a KI ( 95%CI culture

substrate

inhibitor

μM

preparation

μg/L (ppb)

competitive model (cell suspensions) OW

VC

1,1,1-TCA

CS

0.2 ( 0.1

33 ( 11

BDI KB-1

VC VC

1,1,1-TCA 1,1,1-TCA

CS CS

0.4 ( 0.1 0.7 ( 0.2

58 ( 13 100 ( 27

BDI

VC

1,1,1-TCA

CFE

1.2 ( 0.2

160 ( 27

KB-1

TCE

1,1,1-TCA

CFE

1.5 ( 0.6

210 ( 74

noncompetitive model (cell-free extracts)

a

KB-1

VC

1,1,1-TCA

CFE

2.0 ( 0.3

270 ( 40

OW

VC

1,1,1-TCA

CFE

2.0 ( 0.4

270 ( 50

KB-1/Dhc

TCE

1,1,1-TCA

CFE

2.2 ( 0.6

300 ( 75

KB-1/Geo

TCE

1,1,1-TCA

CFE

5.1 ( 2

690 ( 260

BDI KB-1

cDCE cDCE

1,1,1-TCA 1,1,1-TCA

CFE CFE

5.5 ( 0.8 19 ( 4

730 ( 110 2,500 ( 600

OW

TCE

1,1,1-TCA

CFE

40 ( 9

5,300 ( 1,200

BDI

TCE

1,1,1-TCA

CFE

43 ( 17

5,800 ( 2,300

OW

cDCE

1,1,1-TCA

CFE

86 ( 17

11,500 ( 2,300

OW

VC

1,1-DCA

CFE

104 ( 24

10,300 ( 2,400

BDI

cDCE

1,1-DCA

CFE

110 ( 18

11,000 ( 1,800

OW

cDCE

1,1-DCA

CFE

130 ( 57

13,000 ( 5,600

BDI KB-1

VC VC

1,1-DCA 1,1-DCA

CFE CFE

162 ( 39 300 ( 111

16,000 ( 3,700 29,700 ( 11,000

KB-1

cDCE

1,1-DCA

CFE

830 ( 280

82,000 ( 28,000

KB-1

TCE

1,1-DCA

CFE

no inhibition

no inhibition

Error values represent 95% conﬁdence intervals. CFE = cell-free extract; CS = cell suspension.

may be responsible in both cases (see discussion below). In contrast, the KB-1/Dhc culture, which contains Dehalococcoides as the sole dechlorinator, had a much higher Km (180 ( 40 μM). These diﬀerences between KB-1/Geo, KB-1/Dhc, and the KB1 culture (Figure 3, panel A) are more evident in an Eadie Hofstee transformation of the kinetic data for each culture (Figure 3, panels B and C). Nonlinearity in the Eadie Hofstee transformation is indicative of a multiple-enzyme system catalyzing the activity of interest.23 This transformation provides a reasonable straight-line ﬁt to the single-dechlorinator enrichments KB-1/ Dhc and KB-1/Geo, but the multiple dechlorinator-containing KB-1 culture produced a curve on a graph of V vs V/[S]. In the case of TCE dechlorination by KB-1, this curve in the Eadie Hofstee plot likely indicates the involvement of both Geobacter and Dehalococcoides reductive dehalogenases and their diﬀering aﬃnities for TCE, both of which contribute to the overall mixed culture dechlorination kinetics. In KB-1, growth yield experiments have shown that Geobacter is responsible for approximately 80% of TCE to cDCE dechlorination.24 Although both Geobacter and Dehalococcoides use TCE as a terminal electron acceptor in the KB-1 culture, their coexistence may be based on a mutualistic relationship, rather than competition. The rapid conversion of TCE to cDCE by Geobacter provides an abundance of electron acceptor for Dehalococcoides to respire to ethene, while Geobacter may beneﬁt from the removal of cDCE, buildup of which could create daughter product inhibition for TCE dechlorination.11 Dechlorination in the Presence of Inhibitor. To identify trends in 1,1,1-TCA and 1,1-DCA inhibition of TCE, cDCE, and

VC reductive dechlorination, the data from Table 1 were sorted in order of increasing KI (Table 2) and converted to μg per L to provide units relevant to the remediation industry. 1,1,1-TCA inhibits VC dechlorination to ethene most significantly (KI from 30 to 100 μg/L in cell suspensions; 160 270 μg/L in cell-free extract assays) in all cultures tested, suggesting that 1,1,1-TCA affects the VC reductive dehalogenase in these cultures—most likely VcrA. TCE dechlorination in KB-1 was similarly sensitive to 1,1,1-TCA where VcrA is also thought to be involved.25 In KB-1 and particularly in OW, cDCE dechlorination was much less sensitive to 1,1,1-TCA, suggesting that enzymes distinct from VcrA dechlorinate cDCE in these cultures, as suggested in previous analyses of transcription in KB-1.25,26 In the BDI culture, cDCE dechlorination was more inhibited than TCE dechlorination. In all cultures, 1,1-DCA was less inhibitory than 1,1,1-TCA, with lowest inhibition constants above 10 mg/L, or over 100 times greater than inhibition constants measured for 1,1,1-TCA. Generally speaking, VC and cDCE dechlorination was most sensitive to 1,1-DCA. The effects of 1,1-DCA on TCE dechlorination were only tested with KB-1 and no inhibition was observed. Relationship to Reductive Dehalogenases. A summary of the information currently known about the reductive dehalogenases likely responsible for each dechlorination step, with corresponding measurements of Km, KI (1,1,1-TCA) and KI (1,1-DCA), which are characteristic properties of enzymes, is shown in Table 3. In some cases, the kinetic data have rather large confidence intervals. A likely explanation is the presence of 9698

dx.doi.org/10.1021/es201260n |Environ. Sci. Technol. 2011, 45, 9693–9702

Environmental Science & Technology

ARTICLE

Figure 3. Comparison of KB-1 Dehalococcoides and Geobacter reductive dehalogenase kinetic parameters. Data for the parent mixed culture KB-1, the Dehalococcoides-enriched subculture KB-1/Dhc, and the Geobacter-enriched subculture KB-1/Geo cultures are shown. (A) Michaelis Menten plot showing initial TCE dechlorination rates and derived kinetic constants. (B) Eadie Hofstee linear transformation of the three sets of data, showing good ﬁts to the cultures containing single dechlorinating population, and a poor ﬁt to the KB-1 culture. The lines show a linear regression through the data set, and the r2 value is the correlation coeﬃcient. V = initial dechlorination rate, in units of nmol TCE dechlorinated 3 mg protein 1 min 1; [S] = TCE concentration in μM. (C) Same plot as (B) showing the full range of V/[S] data.

multiple reductive dehalogenases in consortia or even within a single Dehalococcoides strain. Nonetheless, a few trends do emerge. As discussed above, KB-1/Geo and BDI showed the lowest values of Km for the TCE to cDCE step. This reaction is catalyzed by a reductive dehalogenase of a Geobacter sp. in KB-1/ Geo, and this is likely the case in BDI as well, which similarly contains the dechlorinator Geobacter lovleyi (Table 3). Several different bacterial genera dechlorinate TCE to cDCE, and the kinetic data reveal significant differences in both Km and KI between different cultures. The higher KI for 1,1,1-TCA in the OW and BDI cultures (Table 3) may be due to the presence of Dehalobacter strain(s) that have been implicated in TCE dechlorination15,27 and it is likely that certain TCE reductive dehalogenases are less affected by 1,1,1-TCA than others. The presence of multiple dechlorinating species carrying multiple yet distinct reductive dehalogenases with overlapping substrate ranges likely imparts robustness to these mixed consortia when challenged with mixtures of halogenated compounds often found at contaminated sites. The identity of the enzyme(s) responsible for the cDCE to VC step is not well established, and the data presented herein suggest that enzyme(s) other than VcrA are involved (Table 3). Indeed, because the VC to ethene dechlorination step is much more sensitive to the inhibitors than the cDCE to VC step, it is unlikely that the same enzyme catalyzes these two reactions. Data from Waller et al.25,26 revealed that multiple reductive dehalogenase genes, but not vcrA, were transcribed during cDCE dechlorination

in culture KB-1. Moreover, some strains of Dehalococcoides including strains 195 and FL222,28 can couple dechlorination of cDCE to VC with growth, but not VC to ethene. Thus, although puriﬁed VcrA catalyzes both steps in vitro, other enzymes in Dehalococcoides may catalyze the cDCE to VC step in vivo. The Km, KI(1,1,1-TCA) and KI(1,1-DCA) parameters for the VC to ethene step were similar in all three cultures within error (Table 3), consistent with the assumption that the same enzyme system, presumably VcrA, is responsible for this step in these cultures. Inhibition of the VC to ethene step by 1,1,1TCA also ﬁt signiﬁcantly better to the competitive inhibition model (particularly in the cell suspension assays), suggesting that 1,1,1TCA binds to the active site. In contrast, the cDCE and TCE data were often better modeled by the noncompetitive model, implying a more complicated scenario where the inhibitor binds to both free enzyme and the enzyme substrate complex, which could be an artifact of conditions in the cell-free extract assay. Both models assume that inhibition is reversible, and this was shown to be the case in experiments with mixed cultures dechlorinating both TCE and 1,1,1-TCA.3 Implications for Contaminated Sites. 1,1,1-TCA is known to interfere with microbial processes, particularly methanogenesis.29 1,1,1-TCA and carbon tetrachloride at concentrations of less that 20 μM inhibited methanogenic and hydrogenotrophic organisms.29 The presence of 1,1,1-TCA may affect multiple community members in dechlorinating consortia, for example by inhibiting fermentation reactions that provide 9699

dx.doi.org/10.1021/es201260n |Environ. Sci. Technol. 2011, 45, 9693–9702

Environmental Science & Technology

ARTICLE

Table 3. Summary of Measured Cell-Free Extract Aﬃnity (Km) and Inhibition (KI) Constants As a Function of Dechlorination Step, and Organisms and Reductive Dehalogenase (RDase) Enzymes Present

a

Organisms and likely enzymes were extracted from data from the following publications: OW;15 BDI;14,27,30 KB-1;29,30 KB-1/Geo (data not shown). In these cases there are two data sets without inhibitor: one in experiments with 1,1,1-TCA and one in experiments with 1,1-DCA. A combined Km was calculated as the median of the overlapping 95% conﬁdence interval ranges of the Km from the two data sets. c n.d.: not determined. b

hydrogen to dechlorinators. Given the potential for complicated inhibitory effects of 1,1,1-TCA on a range of microbial phylotypes and functions, we focused only on the effects on dechlorinating organisms. For this reason, the cell suspension experiments were amended with hydrogen to preclude the need for active fermentation. The inhibition constants, KI, for 1,1,1TCA acting on VC dechlorination were similar between cell suspensions and cell-free extracts, suggesting that 1,1,1-TCA acts directly on the reductive dehalogenase enzyme system(s) and does not exert a general toxic effect on Dehalococcoides cells. These data also suggest that it is reasonable to apply the values of KI obtained in these cell-free extract experiments to dechlorinating mixed cultures, keeping in mind that other microorganisms supporting the main dechlorinators may also be affected by these inhibitors’ compounding effects. The KI values in these experiments represent the inhibitor concentrations at which the rate of dechlorination is half the rate in the absence of inhibitor (this is mathematically true for the noncompetitive equation, and is true for the competitive equation as substrate concentrations decrease below Km). The KI values reported in Table 2 therefore provide useful guidelines to

assess whether 1,1,1-TCA or 1,1-DCA concentrations will aﬀect chlorinated ethene reductive dechlorination at a particular site cocontaminated with chlorinated ethanes. The data demonstrate that 1,1,1-TCA cocontamination should be a concern at all sites where practitioners seek to rely on microbial reductive dechlorination of VC to ethene as a remedial strategy. Fortunately, for all culture and chlorinated ethene combinations tested here, 1,1-DCA exerted low or negligible inhibition, which suggests that the removal of 1,1,1-TCA via reductive dechlorination to its daughter products (1,1-DCA or monochloroethane) would relieve inhibition to chlorinated ethene dechlorination. This observation agrees with previous ﬁndings that dechlorination of TCE past cDCE and VC only proceeded when cocontaminating 1,1,1-TCA was ﬁrst removed by the addition of a 1,1,1-TCAdechlorinating Dehalobacter-containing mixed culture.3

’ ASSOCIATED CONTENT

bS

Supporting Information. Table S1: Raw data of initial velocity versus concentration at various inhibitor concentrations from cell-free extract and cell suspension assays. Table S2: Inhibition models considered in this analysis. Table S3: Values

9700

dx.doi.org/10.1021/es201260n |Environ. Sci. Technol. 2011, 45, 9693–9702

Environmental Science & Technology of statistical parameters (R2, AICc, and Sy.x) used in ranking inhibition models. Figure S1: Kinetics of cDCE dechlorination in cell-free extracts in the presence of increasing concentrations of 1,1,1-TCA or 1,1-DCA. Figure S2: Kinetics of TCE dechlorination in cell-free extracts in the presence of increasing concentrations of 1,1,1-TCA. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]; phone: 416-946-3506; fax: 416-978-8605. )

Present Addresses

ENVIRON EC (Canada), Inc., Mississauga, ON, Canada, L5N 7G2. Department of Civil and Environmental Engineering, University of California, Berkeley, CA, 94720. ^

’ ACKNOWLEDGMENT This research was funded by the U.S. Department of Defense Strategic Environmental Research and Development Program (SERDP) under Contract W912HQ-10-C-0062 (project ER1586). W.C. was funded by an Ontario Graduate Scholarship in Science and Technology. A.G. was funded by a Natural Sciences and Engineering Research Council of Canada Graduate Scholarship. We thank Laura Hug for help in enrichment of the KB-1/ Geo subculture. ’ REFERENCES (1) Maes, A.; Van Raemdonck, H.; Smith, K.; Ossieur, W.; Lebbe, L.; Verstraete, W. Transport and activity of Desulﬁtobacterium dichloroeliminans strain DCA1 during bioaugmentation of 1,2-DCA-contaminated groundwater. Environ. Sci. Technol. 2006, 40, 5544–5552. (2) Duhamel, M.; Wehr, S. D.; Yu, L.; Rizvi, H.; Seepersad, D.; Dworatzek, S.; Cox, E. E.; Edwards, E. A. Comparison of anaerobic dechlorinating enrichment cultures maintained on tetrachloroethene, trichloroethene, cis-dichloroethene and vinyl chloride. Water Res. 2002, 36, 4193–4202. (3) Grostern, A.; Edwards, E. A. A 1,1,1-trichloroethane-degrading anaerobic mixed culture enhances biotransformation of mixtures of chlorinated ethenes and ethanes. Appl. Environ. Microbiol. 2006, 72, 7849–7856. (4) Scheutz, C.; Durant, N. D.; Hansen, M. H.; Bjerg, P. L. Natural and enhanced anaerobic degradation of 1,1,1-trichloroethane and its degradation products in the subsurface a critical review. Water Res. 2011, 45, 2701–23. (5) Krajmalnik-Brown, R.; Holscher, T.; Thomson, I. N.; Saunders, F. M.; Ritalahti, K. M.; L€oﬄer, F. E. Genetic identiﬁcation of a putative vinyl chloride reductase in Dehalococcoides sp. strain BAV1. Appl. Environ. Microbiol. 2004, 70, 6347–6351. (6) Lookman, R.; Paulus, D.; Marnette, E.; Pijls, C.; Ryngaert, A.; Diels, L.; Volkering, F. Ground water transfer initiates complete reductive dechlorination in a PCE-contaminated aquifer. Ground Water Monit. Rem. 2007, 27, 65–74. (7) M€uller, J. A.; Rosner, B. M.; Von Abendroth, G.; MeshulamSimon, G.; McCarty, P. L.; Spormann, A. M. Molecular identiﬁcation of the catabolic vinyl chloride reductase from Dehalococcoides sp. strain VS and its environmental distribution. Appl. Environ. Microbiol. 2004, 70, 4880–4888. (8) Scheutz, C.; Durant, N. D.; Dennis, P.; Hansen, M. H.; Jorgensen, T.; Jakobsen, R.; Cox, E. E.; Bjerg, P. L. Concurrent ethene generation and growth of Dehalococcoides containing vinyl chloride reductive dehalogenase genes during an enhanced reductive dechlorination ﬁeld demonstration. Environ. Sci. Technol. 2008, 42, 9302–9309.

ARTICLE

(9) Zaa, C. L. Y.; McLean, J. E.; Dupont, R. R.; Norton, J. M.; Sorensen, D. L. Dechlorinating and iron reducing bacteria distribution in a TCE-contaminated aquifer. Ground Water Monit. Rem. 2010, 30, 46–57. (10) Cupples, A. M.; Spormann, A. M.; McCarty, P. L. Vinyl chloride and cis-dichloroethene dechlorination kinetics and microorganism growth under substrate limiting conditions. Environ. Sci. Technol. 2004, 38, 1102–1107. (11) Yu, S. H.; Dolan, M. E.; Semprini, L. Kinetics and inhibition of reductive dechlorination of chlorinated ethylenes by two diﬀerent mixed cultures. Environ. Sci. Technol. 2005, 39, 195–205. (12) Yu, S.; Semprini, L. Kinetics and modeling of reductive dechlorination at high PCE and TCE concentrations. Biotechnol. Bioeng. 2004, 88, 451–464. (13) Duhamel, M.; Edwards, E. A. Microbial composition of chlorinated ethene-degrading cultures dominated by Dehalococcoides. FEMS Microbiol. Ecol. 2006, 58, 538–549. (14) Ritalahti, K. M.; Amos, B. K.; Sung, Y.; Wu, Q. Z.; Koenigsberg, S. S.; L€ oﬄer, F. E. Quantitative PCR targeting 16S rRNA and reductive dehalogenase genes simultaneously monitors multiple Dehalococcoides strains. Appl. Environ. Microbiol. 2006, 72, 2765–2774. (15) Daprato, R. C.; L€ oﬄer, F. E.; Hughes, J. B. Comparative analysis of three tetrachloroethene to ethene halorespiring consortia suggests functional redundancy. Environ. Sci. Technol. 2007, 41, 2261–2269. (16) Edwards, E. A.; Grbic-Galic, D. Anaerobic degradation of toluene and o-xylene by a methanogenic consortium. Appl. Environ. Microbiol. 1994, 60, 313–322. (17) Amos, B. K.; Sung, Y.; Fletcher, K. E.; Gentry, T. J.; Wu, W. M.; Criddle, C. S.; Zhou, J.; L€ oﬄer, F. E. Detection and quantiﬁcation of Geobacter lovleyi strain SZ: implications for bioremediation at tetrachloroethene- and uranium-impacted sites. Appl. Environ. Microbiol. 2007, 73, 6898–6904. (18) Grostern, A.; Chan, W. W. M.; Edwards, E. A. 1,1,1-Trichloroethane and 1,1-dichloroethane reductive dechlorination kinetics and cocontaminant eﬀects in a Dehalobacter-containing mixed culture. Environ. Sci. Technol. 2009, 43, 6799–6807. (19) Haldane, J. B. S. Enzymes; Longmans Green: London, 1930. (20) Hofstee, B. H. J. Non-inverted versus inverted plots in enzyme kinetics. Nature 1959, 184, 1296–1298. (21) Burris, D. R.; Delcomyn, C. A.; Smith, M. H.; Roberts, A. L. Reductive dechlorination of tetrachloroethylene and trichloroethylene catalyzed by vitamin B12 in homogeneous and heterogeneous systems. Environ. Sci. Technol. 1996, 30, 3047–3052. (22) He, J.; Sung, Y.; Krajmalnik-Brown, R.; Ritalahti, K. M.; L€ oﬄer, F. E. Isolation and characterization of Dehalococcoides sp. strain FL2, a trichloroethene (TCE)- and 1,2-dichloroethene-respiring anaerobe. Environ. Microbiol. 2005, 7, 1442–1450. (23) Bender, M.; Conrad, R. Kinetics of CH4 oxidation in oxic soils exposed to ambient air or high CH4 mixing ratios. FEMS Microbiol. Ecol. 1992, 101, 261–270. (24) Duhamel, M.; Edwards, E. A. Growth and yields of dechlorinators, acetogens, and methanogens during reductive dechlorination of chlorinated ethenes and dihaloelimination of 1,2-dichloroethane. Environ. Sci. Technol. 2007, 41, 2303–2310. (25) Waller, A. S. Molecular investigation of chloroethene reductive dehalogenation by the mixed microbial community KB-1. PhD Thesis, University of Toronto, Toronto, 2010. (26) Waller, A. S.; Krajmalnik-Brown, R.; L€oﬄer, F. E.; Edwards, E. A. Multiple reductive-dehalogenase-homologous genes are simultaneously transcribed during dechlorination by Dehalococcoides-containing cultures. Appl. Environ. Microbiol. 2005, 71, 8257–8264. (27) Amos, B. K.; Ritalahti, K. M.; Cruz-Garcia, C.; Padilla-Crespo, E.; L€oﬄer, F. E. Oxygen eﬀect on Dehalococcoides viability and biomarker quantiﬁcation. Environ. Sci. Technol. 2008, 42, 5718–5726. (28) Maymo-Gatell, X.; Chien, Y. T.; Gossett, J. M.; Zinder, S. H. Isolation of a bacterium that reductively dechlorinates tetrachloroethene to ethene. Science 1997, 276, 1568–1571. 9701

dx.doi.org/10.1021/es201260n |Environ. Sci. Technol. 2011, 45, 9693–9702

Environmental Science & Technology

ARTICLE

(29) Adamson, D. T.; Parkin, G. F. Impact of mixtures of chlorinated aliphatic hydrocarbons on a high-rate, tetrachloroethene-dechlorinating enrichment culture. Environ. Sci. Technol. 2000, 34, 1959–1965. (30) Ritalahti, K. M.; L€oﬄer, F. E.; Rasch, E. E.; Koenigsberg, S. S. Bioaugmentation for chlorinated ethene detoxiﬁcation: Bioaugmentation and molecular diagnostics in the bioremediation of chlorinated ethene-contaminated sites. Ind. Biotechnol. 2005, 1, 114–118.

9702

dx.doi.org/10.1021/es201260n |Environ. Sci. Technol. 2011, 45, 9693–9702

ARTICLE pubs.acs.org/est

Selection of Salt and Boron Tolerant Selenium Hyperaccumulator Stanleya pinnata Genotypes and Characterization of Se Phytoremediation from Agricultural Drainage Sediments John L. Freeman*,†,‡,§,|| and Gary S. Ba~nuelos†,§,|| †

Agricultural Research Service, U.S. Department of Agriculture, Parlier, California 93648, United States Department of Biology, California State University Fresno, Fresno, California 93740, United States § Center for Irrigation Technology, California State University Fresno, Fresno, California 93740, United States ‡

ABSTRACT: Genetic variation in salt (Na2SO4, NaCl) and boron (B) tolerance among four ecotypes of the selenium (Se) hyperaccumulator Stanleya pinnata (Pursh) Britton was utilized to select tolerant genotypes capable of phytoremediating Se from salt, B, and Se-laden agricultural drainage sediment. The few individual salt/B tolerant genotypes were successfully selected from among a large population of highly salt/B sensitive seedlings. The distribution, hyperaccumulation, and volatilization of Se were then examined in selected plants capable of tolerating the high salt/B laden drainage sediment. Salt/B tolerant genotypes from each of the four ecotypes had mean Se concentrations ranging from 2510 ( 410 to 1740 ( 620 in leaves and 3180 ( 460 to 2500 ( 1060 in seeds (μg Se g 1 DW ( SD), while average daily Se volatilization rates ranged from 722 ( 375 to 1182 ( 575 (μg Se m 2 d 1 ( SD). After two growing seasons (∼18 months), we estimated that hyperaccumulation and volatilization of Se by tolerant S. pinnata genotypes and their associated microbes can remove approximately 30% of the total soil Se in 0 30 cm sediment. The salt/B tolerant S. pinnata genotypes selected and characterized herein represent promising new tools for the successful phytoremediation of Se from salt/B and Seladen agricultural drainage sediments.

’ INTRODUCTION Excessive Se in water is a serious environmental concern in arid regions of the western United States.1 Selenium is a naturally occurring metalloid, which is primarily found as selenate (SeO42 ) in sedimentary shale, e.g., in the western side of the San Joaquin Valley (WSJV).2 Because SeO42 is highly watersoluble, bioavailable, and can be easily biomagniﬁed by aquatic organisms, Se toxicity has been observed in drainage-impacted agricultural regions where overirrigation, the presence of high water tables, high evaporation rates, and surface runoﬀ has concentrated SeO42 , salts, and B in water sources, causing deformities and death in ﬁsh and birds.3 5 Currently, phytoremediation strategies for managing soluble Se in salt, B, and Se-laden soils and waters in the WSJV utilize plant species such as moderately tolerant Indian Mustard (Brassica juncea) or canola (Brassica napus L.).6,7 These two Brassica crop plant species are secondary Se accumulating plants, while most plants are nonaccumulators of Se and accumulate only trace levels of nonessential Se.6 9 In contrast, there are hyperaccumulator species within the genera Stanleya and Astragalus that hyperaccumulate Se from 2000 to 15 000 μg Se g 1 DW, while growing on soils with 4 6 μg Se g 1.10 12 These types of arid western plants could be of greater value for improving the eﬃciency of Se phytoremediation. Interestingly, r 2011 American Chemical Society

Se hyperaccumulators show stunted growth in the absence of Se,13,14 and roots of S. pinnata are known to grow toward and proliferate in Se-rich soils.15 In Se hyperaccumulator plants, young leaves, ﬂowers, and seeds also accumulate much greater Se concentrations than medium or old aged leaves,11,16 while leaves concentrate Se inside speciﬁc tissues and organelles.14,16,17 Inherently, large amounts of variation in Se hyperaccumulation, biomass production, and Se tolerance exist among Stanleya species and between S. pinnata ecotypes,18,19 making S. pinnata an ideal model system for genetic, biochemical, and physiological studies on Se hyperaccumulation,14 but also for identifying plants potentially capable of eﬃcient Se phytoremediation.20 Stanleya pinnata not only extracts large amounts of Se from the soil solution but also biotransforms Se from toxic inorganic SeO42 into two relatively less-toxic, human health beneﬁcial organic forms, e.g., methylselenocysteine (MeSeCys) and selenocystathionine (SeCyst).16,21 A key factor responsible for Se biotransformation is the activity of the enzyme selenocysteine methyltransferase (SMT), which methylates selenocysteine Received: May 11, 2011 Accepted: October 11, 2011 Revised: September 14, 2011 Published: October 11, 2011 9703

dx.doi.org/10.1021/es201600f | Environ. Sci. Technol. 2011, 45, 9703–9710

Environmental Science & Technology (SeCys) into methylselenocysteine and eﬀectively prevents SeCys from mis-incorporation into proteins, circumventing Se toxicity.22,23 Methylselenocysteine may be further converted into γ-glutamylmethylselenocysteine, the dipeptide concentrated in trichomes and seeds of A. bisulcatus,16,24 or metabolized into the volatile organic dimethyl diselenide (DMDSe) found in A. racemosus.25 Phytovolatilization of Se was originally proposed as an additional means of Se removal to help optimize the eﬀectiveness of Brassica crop plant-based Se phytomanagement.6,7,26 Phytovolatilization is considered advantageous because Se is converted from watersoluble toxic inorganic SeO42 into less toxic plant volatile organic forms such as DMDSe or DMSe.27,28 Field-scale phytoremediation using S. pinnata has not been previously evaluated because of its nonagronomic status, coupled to the small biomass and slow growth of most ecotypes. However, in a ground-breaking paper by Feist and Parker in 2001, the Se phytoremediation potential of S. pinnata was ﬁrst realized after the authors tested sixteen diﬀerent S. pinnata ecotypes for their Se phytoremediation abilities in a sand culture greenhouse study.19 Leaf and seed samples were analyzed for Se from these naturally evolved distinct S. pinnata ecotypes, documented and named at each GPS coordinate location in the western U.S.19 Based on Se accumulation and biomass, one ecotype in particular, Colorado-4 (CO4), was tested and deemed the best suited for Se phytoremediation purposes, however, its salt/B tolerance was not tested19 (note: salt, B, and Se are all commonly found at high levels in drainage sediments originating from WSJV soils). Among these sixteen ecotypes, two poor performing ones, California-2 (CA2) and Plants of the Southwest (POSW), were then later screened for NaCl and B tolerance, but only B tolerance was observed.20 The combined abilities of a plant to tolerate high concentrations of soil salts and B while hyperaccumulating leaf Se and producing a high shoot biomass, are required for using plants to eﬃciently phytoremediate Se from Se-laden agricultural drainage sediments. The objectives of our current two-season ﬁeld study were as follows: (1) select for salt- and B-tolerant genotypes inherent in four distinct S. pinnata ecotypes of naturally segregating populations; (2) observe the reproduction and growth capability of the salt/B-tolerant genotypes; (3) monitor and nondestructively test individual plants growing on drainage sediments to determine Se distribution and accumulation in various plant parts and whole shoots of the ﬂowering salt/B-tolerant individuals; (4) allow seed shattered from the surviving tolerant genotypes to potentially reestablish under the high-salt/B/Se drainage sediment; and (5) determine the amount of Se that was hyperaccumulated and volatilized by the tolerant genotypes for estimating Se removed from the sediment soil. Over two growing seasons we selected, characterized, and quantiﬁed the most salt/B-tolerant individuals for their potential use in phytoremediating Se from salt-, B-, and Se-laden agricultural drainage sediments.

’ METHODS AND MATERIALS Field Site. Agricultural drainage sediment high in salt, B, and Se was collected from the San Luis Drain in Mendota, CA and placed in an excavated field plot located at the USDA ARS, Parlier, CA, as described by Ba~ nuelos and co-workers in 2005.7 Soil properties were recently reported29 and considered for water-extractable concentrations of Se, sulfur (S), B, Na, and Cl for 0 25 and 25 50 cm depths in the same sediment plots as follows: 0.9 1.2 μg Se mL 1, 764 940 μg S mL 1, 5.9 5.1 μg B mL 1, 525 624 μg Na mL 1, and 184 201 μg Cl mL 1. Soil

ARTICLE

EC (electrical conductivity; salinity) and soil pH were 4.8 5.8 dS m 1 and 7.9 8.0, respectively, for the two depths. In March 2009, two beds 33 m long and 1 m wide were then created on the Seladen drainage sediment, and one control bed of the same dimensions was created on sandy loam soil with low Se (<50 ng Se g 1 DW).7 On one sediment bed, a surface-drip irrigation system (DI sediment bed) was installed consisting of one in-line turbulent flow emitter (rate of 4 L h 1) spaced every 0.45 m apart. On the adjacent sediment bed, a surface microsprinkler or mister irrigation (MI sediment bed) was set up using the same irrigation rate and emitter spacing. Each sediment bed (33 m2) was divided into three plots (1 11 m), and within each plot a subplot (1 2.75 m) was created for planting four morphologically unique S. pinnata ecotypes; Colorado4 (CO4, collected just east of Denver, CO),19 commercially available Western Native Seed (WNS, collected just south of Coaldale, CO), Pine Ridge (PR, collected just west of Fort Collins, CO),11,16 and commercially available Plants of the Southwest (POSW, collected just southwest of Santa Fe, NM); resulting in four subplots per plot and twelve subplots per sediment bed. Soil samples were then collected at a depth of 10 cm from each subplot on both sediment beds and used to estimate total Se and S available at preplant to a root depth of 30 cm. In early April 2009, the DI sediment bed was seeded with three seeds 2.5 cm deep every 30.5 cm, while the MI sediment bed was seeded with 20 mL of seed per each ecotype and subplot. Seeds were spread out and dispersed evenly and then raked into the top 2.5 cm of soil because mister irrigation wets a larger surface area. The control soil bed with low Se (MI control bed) was also mister-irrigated and similarly seeded with the same ecotypes. One week after sowing, a further raking was required to loosen the crusty salty layer on the surface of all beds. The irrigated beds received good quality water (EC < 0.8 dS m 1), based on rate of evapotranspiration (ET) losses recorded at California Irrigation Management Information System (CIMIS), located 2 km away. As a secondary selection for salt/B tolerance, the tolerant S. pinnata genotypes that set seed in season one were allowed to naturally shatter (∼90% of their seed set) for reselection of salt/B tolerant offspring. Sampling, Harvest, Measurement, Calculations. At the onset of bloom (June 2009) in season one, the flowering salt/ B tolerant individuals were nondestructively subsampled to measure mean tissue Se concentrations and determine total Se accumulated within the shoot. To do this nondestructively, a subsample of ∼2 g fresh weight from each set of young, medium, and old aged leaves with intact petioles, flowers, and seeds was collected, while the main stems and inflorescence stems were not subsampled to allow for continued growth. Samples were immediately washed thoroughly with deionized water, blotted dry, weighed, and then dried at 50 °C for 3 days before total dry weights were taken. The following equation was used to determine the dry weight (DW) of shoot biomass per each flowering salt/B tolerant plant: (no. of young leaves the mean DW of a young leaf) + (no. of medium leaves the mean DW of a medium leaf) + (no. of old leaves mean DW of an old leaf) + (inflorescence length mean flower DW of a premeasured inflorescence length) = estimated DW of shoot biomass. Subsamples were collected from each part and analyzed for Se, as described later. Dry weights and tissue Se concentrations (μg Se g 1 DW) were then used to calculate the total Se accumulated in each biomass group. Separate values were then summed to yield an estimate of the total shoot Se per plant. Additionally, twelve plants were harvested, dried, and weighed for obtaining the average shoot biomass per plant from each ecotype (n = 3). 9704

dx.doi.org/10.1021/es201600f |Environ. Sci. Technol. 2011, 45, 9703–9710

Environmental Science & Technology In early spring of season two (April 2010), emerging young leaves (known to have the highest Se concentrations)11,16 were subsampled, pooled, and analyzed for Se from the surviving salt/ B tolerant genotypes originating from each ecotype on the MI and DI sediment beds. Once the ﬂowers were in full bloom (June 2010), all leaves and ﬂowers were again subsampled from the tolerant genotypes, pooled by ecotype separately on all three plots from both irrigated sediment beds, respectively, and Se concentrations were measured in leaves at various ages and in ﬂowers. In early July 2010, after only ∼10% of total seed set had naturally shattered, most of the total shoot biomass (∼95%) was clipped and harvested. Approximately 5% of the shoots (a short main stem with a few leaves) were left so that these hardy perennial plants (reported to be highly amenable to clipping19) could regrow for the purpose of continued Se phytoremediation and the production of salt/B tolerant seed stocks. The harvested shoot materials were dried at 50 °C, and total DW biomass was measured by ecotype from both irrigated sediment beds. Seeds were mechanically separated, cleaned, and weighed by ecotype at each respective subplot, and the remaining shoot biomass (leaves, a few ﬂowers, and stems) was then ground to powder. We calculated the total harvested Se from tolerant genotypes by ecotype via multiplying dry weight values with Se concentrations of ground shoots and ground seeds. Selenium and S Volatilization. Volatile Se and S were collected each week over a 24-h period from 0.5 m2 areas on the MI sediment bed and on the unvegetated MI sediment bed from August 2009 to August 2010. A complete description of the equipment and the protocol used herein is provided by Lin and co-workers.30,31 Volatile Se and S produced on each 0.5 m2 area over 24 h were normalized to the percent of total vegetation of the respective ecotype covering each 0.5 m2 area. Daily Se volatilization rates were averaged and multiplied by 365 days to estimate total average Se volatilized in grams over 1 year by the tolerant genotypes selected from each ecotype. Estimated Se Removal. We estimated preplant soil Se (grams) potentially available in the MI sediment bed to a 30-cm root depth by assuming the same Se concentration measured at 10 cm as follows: multiply mean total preplant soil Se concentration (9.0 ( 3.8 mg Se kg 1 soil) by the dry weight of soil per meter squared to a 30-cm depth (249.6 kg) by the total area of one sediment bed (33 m2), which equaled 74.13 g of Se available at preplant. After season two, we calculated the approximate amount of Se removed from the MI sediment bed by adding the total Se harvested in the shoot biomass from tolerant S. pinnata genotypes to the total amount of Se volatilized by the same plants as described above. Sample Se Analysis. Soil samples were cleaned of any residual plant material, thoroughly mixed, oven-dried at 50 °C for 7 d, and ground in a stainless steel Wiley mill equipped with a 1-mm screen. Total Se and S were determined using 500 mg ground soil or plant after digestion with HNO3, H2O2, and HCl.32 Total Se and S were analyzed after acid digestion by an inductively coupled plasma mass spectrometer (Agilent 7500cx, Santa Clara, CA). The National Institute of Standards and Technology (NIST) Wheat Flour (SRM 1567; Se content of 1.1 ( 0.2 μg g 1 DM) and two internal soil standards (sediment collected from Kesterson Reservoir, CA, with a total Se content of 7.5 ( 0.25 and 25 ( 0.87 mg kg 1) were used as the standardized quality control samples for plants and soils, respectively. In this study, the Se recovery rates of the standardized sample materials were over 94%, while Se detection limits were 50 ng Se g 1 DW.

ARTICLE

Figure 1. Photograph showing the salts precipitated on soil surface on the MI-bed and stunted growth of S. pinnata PR along with salt/Bsensitive dead seedlings after 2 months of stringent selection in season one (a). Example of the genotypic size variations present in two tolerant WNS individuals growing on DI sediment bed (b). Successfully selected salt/Btolerant WNS individuals pictured ﬂowering on the MI sediment bed (c).

The Statistical Analysis System (SAS, version 9.2) was used for the data analyses (SAS Inst., 2010).

’ RESULTS AND DISCUSSION At preplant, the irrigated sediment beds had mean total soil concentrations of Se at 9.0 ( 3.8 and S at 17 416 ( 5597 (μg g 1 DW ( SD). Soil salinity (EC) and B were assumed to be the 9705

dx.doi.org/10.1021/es201600f |Environ. Sci. Technol. 2011, 45, 9703–9710

Environmental Science & Technology

ARTICLE

Figure 2. Mean total Se (mg ( SD) summed for all young, medium, and old aged leaves (a c, respectively), ﬂowers, seeds, and shoot (d f, respectively), within each individual ﬂowering salt/B tolerant plant of each original ecotype selected on the sediment beds at season one (midsummer).

same as those recently reported by Ba~ nuelos and Lin.29 After spring germination of the four ecotypes in the sediment beds, we observed variability in seedling tolerance (Figure 1a). Visible symptoms of salt/B toxicity included small size, stunted growth, overproduction of anthocyanin leaf pigments, leaf chlorosis, and necrosis. Due to the high salt and B precipitation on the soil surface, the vast majority of seedlings from all ecotypes had wilted and died on both salt/B laden irrigated sediment beds shortly after germination in season one (Figure 1a). Overall, the rate of germination, tolerance selection, and thus biomass production was the best on the MI sediment bed. Hypothetically, the good quality water applied via mister irrigation may have leached more soluble salts (including Se) from a larger wetted surface area than did drip irrigation, and thereby further reduced the inhibiting eﬀect of salt/B on a greater number of germinating seedlings and actively growing plants. Compared to the other plants, a select group of relatively unstressed and healthy surviving set of individuals contained genotypes that exhibited rapid overall growth, had a larger size, and were without any visible anthocyanin production (Figure 1b). Western Native Seed had a total of 33 ﬂowering salt/B tolerant plants (pictured in Figure 1c), followed by PR with 8, CO4 with 4, and POSW with

only 1 plant. Thus, it appeared that POSW was initially the most sensitive to salts and B in season one, as only one plant ﬂowered and the relatively few other POSW survivors all exhibited dwarfed sizes and multiple symptoms of salt/B toxicity. Mean shoot weight biomass per mature plant harvested by ecotype was as follows (g DW ( SD): CO4 157.17 ( 60.37, WNS 102.24 ( 16.27, PR 31.64 ( 18.29, and POSW 41.40 ( 19.12. For comparison B. juncea weighed on average 65 ( 5 g DW, when grown earlier on the same sediment for a period of 6 months.33 In season one, young leaves subsampled from the ﬂowering tolerant genotypes by ecotype had the following Se concentrations (μg Se g 1 DW ( SD): CO4 1620 ( 600, WNS 1460 ( 1040, PR 1320 ( 1110, and POSW 9.0. For each ﬂowering tolerant individual, Se concentrations, subsampled fresh leaves, and estimated dry weight biomasses were used to then calculate an average amount of total Se contained in all young, medium, and old leaves (Figure 2a, b, c, respectively), ﬂowers and seeds, and in each individual shoot (Figure 2d, e, f, respectively). Large variability in size existed within each original ecotype, as indicated by the high SD values reported in Figure 2. The single POSW ﬂower was not subsampled to avoid excluding any potential POSW oﬀspring from seeds in the season two tests. 9706

dx.doi.org/10.1021/es201600f |Environ. Sci. Technol. 2011, 45, 9703–9710

Environmental Science & Technology

ARTICLE

Figure 3. After two seasons of repeated selection, the salt- and B-tolerant S. pinnata genotypes are pictured successfully growing and ﬂowering (June 2010) on both irrigated sediment beds (a). Selenium concentrations (μg Se g 1 DW) of the young (Y), medium (M), and old (O) aged leaves, and ﬂowers (F) pooled by original ecotype: CO4 (b), WNS (c), PR (d) and POSW (e), from both sediment beds. Data represent the mean Se concentrations ( SD.

The total Se accumulated within the entire shoot of each ﬂowering salt/B-tolerant individual plant was as follows (mg Se): CO4 27.9 ( 4.8, WNS 24.7 ( 8.8, PR 21.5 ( 10.6, and POSW 0.04. In regard to the average total Se accumulated per each shoot biomass after four months growth, CO4 appeared to be the best overall performer followed by PR, WNS, and POSW. For comparison, B. juncea accumulated ∼1.66 mg Se entire shoot 1, when grown earlier on the same sediment plot for a period of 6 months.33 In the early spring of season two (April 2010), the number of newly germinating tolerant seedlings was recorded and their averages ( SD for each of the three plots on the MI sediment bed were as follows: CO4 15.2 ( 3.3, WNS 15.7 ( 2.9, PR 5.3 ( 3.5, and POSW 9 ( 8.5. At this time, subsamples of the ﬁrst reemerging young leaves from established plants in season one were harvested and pooled together and their mean Se concentrations were from highest to lowest as follows (μg Se g 1 DW ( SD): CO4 2510 ( 410, PR 2240 ( 540, POSW 1740 ( 620, and WNS 1660 ( 200. By June 2010 of season two, only those selected individual plants with salt/B tolerant genotypes were left alive, and they are pictured growing successfully in full bloom on both irrigated sediment beds (Figure 3a). Leaves by age and ﬂowers were then subsampled and pooled by original ecotype on both sediment beds before total Se (Figure 3b e) and S were measured. In young leaves, both CO4 and WNS had signiﬁcantly higher (P < 0.05) mean Se concentrations (μg Se g 1 DW ( SD) at 1470 ( 360 and 1060 ( 64, respectively, than POSW at 570 ( 140 and PR at 340 ( 20. Colorado-4 and WNS also had

signiﬁcantly more Se (P < 0.05) concentrated than PR or POSW in young and medium aged leaves and in ﬂowers (Figure 3b e). For comparison, leaves from B. juncea grown earlier on the same sediment soils for 6 months accumulated 37 ( 9 μg Se g 1 DW ( SD.33 Selenium concentrations in S. pinnata seeds pooled from surviving genotypes were also hyper-elevated as follows (μg Se g 1 DW ( SD): PR 3180 ( 460, CO4 3080 ( 760, WNS 3060 ( 900, and POSW 2500 ( 1060. Irrespective of ecotype, all S. pinnata plants exposed to the natural ﬁeld environment on the control bed with low Se had died by the end of growing season two. This observation suggests S. pinnata, termed a Se indicator species, requires elevated soil Se concentrations, potentially for use as a Se-based elemental defense mechanism that is needed for its long-term survival in the natural environment. During season two, we found that S. pinnata accumulated high levels of Se (μg Se g 1 DW ( SD) in young leaves (2510 ( 410) and in seeds (3180 ( 460) despite the high concentrations of soluble S in drainage sediments. Usually high soil S concentrations occur as sulfate and inhibit Se phytoextraction through competition with selenate for root uptake through sulfate transporters located in root plasma membranes.6,7,26 The young leaves of the salt/B-tolerant Se hyperaccumulator genotypes accumulated the following S concentrations (μg S g 1 DW ( SD): CO4 13 420 ( 1950, WNS 13 250 ( 2830, POSW 13 430 ( 3280, and PR 5130 ( 2260. These S concentrations exemplify a unique diﬀerence in Brassica, e.g., between the Se hyperaccumulator 9707

dx.doi.org/10.1021/es201600f |Environ. Sci. Technol. 2011, 45, 9703–9710

Environmental Science & Technology

Figure 4. Daily average Se and S volatilization rates per 1 m 2 of the surviving salt/B-tolerant S. pinnata genotypes and bare unvegetated sediment soils were collected periodically over 1 year by original ecotype and normalized to the total percentage of vegetation coverage of 1 m 2. Volatilized Se (μg Se m 2 d 1) (a) and volatilized S (μg S m 2 d 1) (b). Data represent the mean daily volatilized Se and S concentrations collected periodically over 1 year ( SD.

S. pinnata and the secondary Se accumulating S loving species B. juncea. This is because after a year of growth, S. pinnata had a mean leaf S concentration of 11 300 ( 4119 μg S g 1 DW and accumulated 1.9-fold less S when compared with leaves from B. juncea (i.e., 21 500 ( 1600 μg S g 1 DW ( SD) when grown on the same agricultural sediment for a period of 6 months33 (P < 0.05). Furthermore, young leaves from CO4 had the lowest S to Se concentration ratio (S/Se g 1 DW) of 9.1, followed by WNS 12.5, PR 15.1, and POSW 23.5. The outlying high S/Se ratio in POSW may help explain why POSW accumulated the lowest Se concentrations. In the sediment soil, the Se/S concentration ratio for total μg Se/μg S g 1 DW was 0.0005 and 0.0031 for water-soluble μg Se/ μg S L 1. The Se enrichment factor in the CO4 plant leaves was ∼218 fold greater than total μg Se/μg S g 1 DW in soils and varied as follows among the S. pinnata ecotypes: CO4 0.109, WNS 0.080, PR 0.066, and POSW 0.042. In comparison, the Se enrichment factor in the leaves of B. juncea was only 0.0017 when grown earlier on the same sediment for a period of 6 months,33 which is ∼3.4 fold higher than this ratio in the soil and 64 times less than that observed in S. pinnata leaves. Previously reported preliminary observations suggested that elevated soil sulfate may actually stimulate Se volatilization in S. pinnata.20 In this study, the average Se and S volatilization

ARTICLE

results showed that collectively the salt/B-tolerant genotypes volatilized ∼3.6 times more Se than S (1002 ( 509 vs 279 ( 271 μg m 2 d 1 ( SD, respectively) (Figure 4a and b). The reported high SD values are because of the variable range of Se volatilization when continuously collected over 1 year. Volatile rates were as high as 2591 μg Se m 2 d 1 during the hot summer and as low as 287 μg Se m 2 d 1 during the cold winter. These rates of Se volatilization measured from plants grown in the drainage sediment demonstrated that even in the presence of high soil sulfate concentrations, S. pinnata has a preference for volatilizing Se over S. The average daily Se volatilization rates for salt/Btolerant genotypes by ecotype were from highest to lowest as follows (μg Se m 2 d 1 ( SD): PR 1182 ( 575, WNS 1094 ( 517, CO4 1010 ( 572 and POSW 722 ( 375. For comparison, B. juncea on the same soils volatilized Se less than 25 μg Se m 2 d 1.33 This was likely due to the high soluble sulfate levels in the sediment inhibiting selenate uptake and subsequent Se volatilization by B. juncea. The unvegetated bare MI sediment soil exhibited mean volatile rates of Se and S at 19 ( 7 and 36 ( 28 (μg m 2 d 1 ( SD), respectively, while the unvegetated low Se control bed exhibited volatile rates of Se and S at ∼2.5 and ∼20 (μg m 2 d 1 ( SD), respectively. The mean yearly volatilized Se was estimated for all tolerant genotypes and the entire area of the MI sediment bed as follows (g Se yr 1 ( SD): PR 3.88 ( 1.88, CO4 3.31 ( 1.88, WNS 3.04 ( 1.69, and POSW 2.37 ( 1.23. Summing rates of volatile Se, the salt/B-tolerant individuals together volatilized an estimated 12.6 g Se yr 1 from the vegetated MI sediment bed. Even though PR had the smallest above-ground shoot biomass, PR still volatilized a high amount of Se (similar to levels found in CO4) and lower levels of S compared to the other ecotypes (Figure 4b). During semidormant winter months, all ecotypes had a ground covering of only small green leaves, yet Se volatilization in CO4 was remarkably high at ∼390 μg Se m 2 d 1. These observations suggest that Se volatilization may in some part originate from S. pinnata roots and associated rhizospheric microﬂora. However, we did not attempt to diﬀerentiate between volatile Se produced from the plant shoot with that concomitantly produced by S. pinnata root associated microbes. After two seasons of growth and after seeds had set, the vast majority (∼95%) of total shoot biomass was harvested from all the tolerant genotypes by ecotype and dried at 50 °C for 3 days. The total dry shoot biomass, including seeds, harvested from both irrigated sediment beds was as follows (kg DW): WNS 8.31, CO4 7.54, POSW 4.93, and PR 3.10. Over two growing seasons, S. pinnata genotypes accumulated the following amounts of Se in shoot biomass from a 1-m2 vegetated area on the MI sediment bed (mg Se m 2): CO4 400, WNS ∼370, POSW ∼213, and PR ∼186, while in comparison B. juncea accumulated ∼57 when grown on the same soils combined over two growing seasons.33 Using the total biomass measurements and tissue Se concentrations in both shoots and seeds, the amount of Se accumulated by the tolerant genotypes from both MI and DI sediment beds was calculated as follows (g Se): CO4 4.96, WNS 4.61, PR 2.32, and POSW 2.65. A grand total of 14.55 g Se was harvested from both sediment beds by the tolerant genotypes. To estimate the potential Se phytoremediation by S. pinnata on the MI sediment bed, we added the mean total Se accumulated in shoots to the amount of volatilized Se emitted for 1 year. We calculate that the tolerant genotypes collectively removed ∼22.27 g Se over two growing seasons on the MI sediment bed. Thus, we estimate S. pinnata could remove 30% of the total soil 9708

dx.doi.org/10.1021/es201600f |Environ. Sci. Technol. 2011, 45, 9703–9710

Environmental Science & Technology Se (74 g Se), potentially available within the root depth of 30 cm on the 33 m 2 MI sediment bed over two growing seasons. The individual percentages of soil Se estimated to be removed by each original ecotype were as follows: CO4 35%, WNS 33%, PR 29%, and POSW 22%. We do not assume these rates of Se removal would be continuous over time under varying environmental and ﬁeld conditions. Moreover, potential unaccountable losses of soil Se (Se not accounted for in or by plant/microbe activity) may occur under ﬁeld conditions due to irrigation- or rain-induced leaching of soluble Se. Hence, mass balance determinations in future studies must include deeper soil sampling to determine the potential redistribution of soil Se to deeper depths. Environmental Considerations. Under these saline drainage sediment growing conditions, CO4 and WNS produced more salt/B-tolerant genotypes that appear to be more efficient at the phytoremediation of Se than any other plant previously tested. Future studies should investigate the selected S. pinnata genotypes tolerance to salt and B potentially through enhanced sequestration or exclusion of Na, Cl, or B ions. Repeated clippings of plants growing in salt/B-laden soils is one practical strategy to increase a plants ability to tolerate excessive ion accumulation, e.g., Na, B, and Cl, within leaves over long-term growth on such soils in the WSJV. Removing cut or fallen biomass is also critical so that Se accumulated by shoot tissues is not redeposited as Se-enriched plant litter. Field and greenhouse studies are in progress on reutilizing the harvested Seenriched S. pinnata shoot material as an organic fertilizer for the biofortification of more healthful Se-enriched nutraceutical food crops as previously achieved using other Brassica plant materials.34,35

’ AUTHOR INFORMATION Corresponding Author

*Telephone: +1-559-596-2729; fax: +1-559-596-2851; e-mail: [email protected]. )

Author Contributions

Authors contributed equally to this work

’ ACKNOWLEDGMENT Thank you to Dr. Parker at U.C. Riverside, for Stanleya pinnata (CO4) seeds, and Dr. Pilon-Smits at Colorado State University for (PR) seeds, Cynthia Rodriguez ACS SEED intern and Frank Dale for drawing the TOC art. Funding for this work was provided by CSU Fresno Agricultural Research Initiative and the California Department of Water Resources. USDA is an equal opportunity provider and employer. ’ REFERENCES (1) Letey, J.; Williams, C. F.; Alemi, M. Salinity, drainage, and selenium problems in the western San Joaquin Valley of California. Irrig. Drain. 2002, 16, 253–259. (2) Presser, T. Geologic origin and pathways of Se from the California Coast Ranges to the west central San Joaquin Valley. In Selenium in the Environment; Frankenberger, W. T., Jr., Benson, S., Eds.; Marcel Dekker Inc.: New York, 1994; pp139 156. (3) Zawislanski, P.; Benson, S.; Terberg, R.; Borglin, S. Selenium speciation, solubility, and mobility in land-disposed dredged sediments. Environ. Sci. Technol. 2003, 37, 2415–2430. (4) Ohlendorf, H. M.; Hoﬀman, D. J.; Saiki, M. K.; Aldrich, T. W. Embryonic mortality and abnormalities of aquatic birds; apparent

ARTICLE

impacts of selenium from irrigation drainwater. Sci. Total Environ. 1986, 52, 49–63. (5) Ohlendorf, H. M. The birds of Kesterson Reservoir: A historical perspective. Aquat. Toxicol. 2002, 57, 1–10. (6) Ba~ nuelos, G. S.; Meek, D. W. Accumulation of selenium in plants grown on selenium-treated soil. J. Environ. Qual. 1990, 19, 772–777. (7) Ba~ nuelos, G. S.; Lin, Z. Q.; Arroyo, I.; Terry, N. Selenium volatilization in vegetated agricultural drainage sediment from the San Luis Drain, Central California. Chemosphere 2005, 60, 1203–1213. (8) Zhang, Y.; Gladyshev, V. N. General trends in trace element utilization revealed by comparative genomic analyses of Co, Cu, Mo, Ni and Se. J. Biochem. 2010, 285, 3393–3405. (9) White, P. J.; Bowen, H. C.; Parmaguru, P.; Fritz, M.; Spracklen, W. P.; Spiby, R. E.; Meachan, M. C.; Mead, A.; Harriman, M.; Trueman, L. J. Interactions between selenium and sulphur nutrition in Arabidopsis thaliana. J. Exp. Bot. 2004, 55, 1927–1937. (10) Knight, S. H.; Beath, O. A. The occurrence of selenium and seleniferous vegetation in Wyoming. Wyo. AES Bull. 1937, 221, 2–64. (11) Galeas, M. L.; Zhang, L. H.; Freeman, J. L.; Wegner, M.; PilonSmits, E. A. H. Seasonal ﬂuctuations of selenium and sulfur accumulation in selenium hyperaccumulators and related non-accumulators. New Phytol. 2007, 173, 517–525. (12) Byers, H. G. Selenium occurrence in certain soils in the United States, with a discussion of related topics. USDA Tech. Bull. 1935, 482, 1–47. (13) Trelease, S. F.; Trelease, H. M. Selenium as a stimulating and possibly essential element for indicator plants. Am. J Bot. 1938, 25, 372–380. (14) Freeman, J. L.; Tamaoki, M.; Stushnoﬀ, C.; Quinn, C. F.; Cappa, J. J.; Devonshire, J.; McGrath, S.; Fakra, S.; Marcus, M. A.; Van Hoewyk, D.; Pilon-Smits, E. A. H. Molecular mechanisms of selenium tolerance and hyperaccumulation in Stanleya pinnata. Plant Physiol. 2010, 153, 1630–1652. (15) Goodson, C. C.; Parker, D. R.; Amrhein, C.; Zhang, Y. Soil selenium uptake and root system development in plant taxa diﬀering in Se-accumulating capability. New Phytol. 2003, 159, 391–401. (16) Freeman, J. L.; Zhang, L. H.; Marcus, M. A.; Fakra, S.; PilonSmits, E. A. H. Spatial imaging, speciation and quantiﬁcation of selenium in the hyperaccumulator plants Astragalus bisulcatus and Stanleya pinnata. Plant Physiol. 2006, 142, 124–134. (17) Pickering, I. J.; Hirsch, G.; Prince, R. C.; Yu, E. Y.; Salt, D. E.; George, G. N. Imaging of selenium in plants using tapered metal monocapillary optics. J. Synchrotron Radiat. 2003, 10, 289–290. (18) Beath, O. A.; Eppson, H. F.; Gilbert, C. S. Selenium and other toxic minerals in soils and vegetation. Wyo. AES Bull. 1935, 206, 1–55. (19) Feist, L. J.; Parker, D. R. Ecotypic variation in selenium accumulation among populations of Stanleya pinnata. New Phytol. 2001, 49, 61–69. (20) Parker, D. R.; Feist, L. J.; Varvel, T. W.; Thomason, D. N.; Zhang, Y. Selenium phytoremediation potential of Stanleya pinnata. Plant Soil 2003, 249, 157–165. (21) Shrift, A.; Virupaksha, T. K. Seleno-amino acids in seleniumaccumulating plants. Biochim. Biophys. Acta 1965, 100, 65–75. (22) Brown, T. A.; Shrift, A. Exclusion of selenium from proteins in selenium tolerant Astragalus species. Plant Physiol. 1981, 67, 1951–1953. (23) Neuhierl, B.; Bock, A. On the mechanism of selenium tolerance in selenium-accumulating plants: Puriﬁcation and characterization of a speciﬁc selenocysteine methyltransferase from cultured cells of Astragalus bisulcatus. Eur. J. Biochem. 1996, 239, 235–238. (24) Nigam, S. N.; McConnell, W. B. Seleno amino compounds from Astragalus bisulcatus isolation and identiﬁcation of g-glutamyl-Semethylselenocysteine and Se-methylselenocysteine. Biochim. Biophys. Acta 1969, 192, 185–190. (25) Evans, C. S.; Asher, C. J.; Johnson, C. M. Isolation of dimethyl diselenide and other volatile selenium compounds from Astragalus racemosus (Pursh.). Aust. J Biol. Sci. 1968, 21, 13–20. (26) Dhillon, S. K.; Dhillon, K. S. Phytoremediation of seleniumcontaminated soils; the eﬃciency of diﬀerent cropping systems. Soil Use Manage. 2009, 25, 441–453. 9709

dx.doi.org/10.1021/es201600f |Environ. Sci. Technol. 2011, 45, 9703–9710

Environmental Science & Technology

ARTICLE

(27) McConnell, K. P.; Portman, O. W. Toxicity of dimethyl selenide in the rat and mouse. P. Soc. Exp. Biol. Med. 1952, 79, 230–231. (28) Wilbur, C. G. Toxicology of Selenium: A review. Clin. Toxicol. 1980, 17, 171–230. (29) Ba~nuelos, G. S.; Lin, Z.-Q. Cultivation of the Indian ﬁg Opuntia in selenium-rich drainage sediments under ﬁeld conditions. Soil Use Manage. 2010, 26, 167–175. (30) Lin, Z.-Q.; Hansen, D.; Zayed, A.; Terry, N. Biological selenium volatilization: Method of measurement under ﬁeld conditions. J. Environ. Qual. 1999, 28, 309–315. (31) Lin, Z.-Q.; Schemenauer, R. S.; Cervinka, V.; Zayed, A.; Lee, A.; Terry, N. Selenium volatilization from the soil-plant system for the remediation of contaminated water and soil in the San Joaquin Valley. J. Environ. Qual. 2000, 29, 1048–1056. (32) Ba~nuelos, G. S.; Akohoue, S. Comparison of microwave digestion with block digestion for selenium and boron analysis in plant tissues. Commun. Soil Sci. Plan. 1994, 25, 1655–1670. (33) Ba~nuelos, G. S.; LeDuc, D. L.; Pilon-Smits, E. A. H.; Tagmount, A.; Terry, N. Transgenic Indian mustard overexpressing selenocysteine lyase, selenocysteine methyltranferase, or methionine methyltansferase exhibits enhanced potential for selenium-contaminated sediments. Environ. Sci. Technol. 2007, 41, 599–605. (34) Ba~nuelos, G. S. Phytoremediation of selenium-contaminated soil and water produces biofortiﬁed products and new agricultural byproducts. In Biofortiﬁcation and Development of New Agricultural Products; Ba~nuelos, G. S., Lin, Z.-Q., Eds.; CRC Press: Boca Roca, FL, 2009; pp 57 70. (35) Ba~nuelos, G. S.; Da Roche, J.; Robinson, J. Developing Seenriched animal feed and biofuel from canola planted for managing Seladen drainage waters in the Westside of central California. Int. J. Phytorem. 2010, 12, 243–253.

9710

dx.doi.org/10.1021/es201600f |Environ. Sci. Technol. 2011, 45, 9703–9710

ARTICLE pubs.acs.org/est

Evaluation of the Permeability of Agricultural Films to Various Fumigants Yaorong Qian,†,* Alaa Kamel,† Charles Staﬀord,† Thuy Nguyen,† William J. Chism,† Jeﬀrey Dawson,‡ and Charles W. Smith‡ †

Biological and Economic Analysis Division and ‡Health Eﬀects Division, Oﬃce of Pesticide Programs, U.S. Environmental Protection Agency

bS Supporting Information ABSTRACT: A variety of agricultural ﬁlms are commercially available for managing emissions and enhancing pest control during soil fumigation. These ﬁlms are manufactured using diﬀerent materials and processes which can ultimately result in diﬀerent permeability to fumigants. A systematic laboratory study of the permeability of the agricultural ﬁlms to nine fumigants was conducted to evaluate the performance of commonly used ﬁlm products, including polyethylene, metalized, and high-barrier ﬁlms. The permeability, as expressed by mass transfer coeﬃcient (cm/h), of 27 diﬀerent ﬁlms from 13 manufacturers ranged from below 1 10 4 cm/h to above 10 cm/h at 25 °C under ambient relative humidity test conditions. The wide range in permeability of commercially available ﬁlms demonstrates the need to use ﬁlms which are appropriate for the fumigation application. The eﬀects of environmental factors, such as temperature and humidity, on the ﬁlm permeability were also investigated. It was found that high relative humidity could drastically increase the permeability of the high-barrier ﬁlms. The permeability of some high-barrier ﬁlms was increased by 2 3 orders of magnitude when the ﬁlms were tested at high relative humidity. Increasing the temperature from 25 to 40 °C increased the permeability for some high-barrier ﬁlms up to 10 times more than the permeability at 25 °C, although the eﬀect was minimal for several of these ﬁlms. Analysis of the distribution of the permeability of the ﬁlms under ambient humidity conditions to nine fumigants indicated that the 27 ﬁlms largely followed the material type, although the permeability varied considerably among the ﬁlms of similar material.

’ INTRODUCTION Preplant soil fumigation is often used for broad spectrum pest control in many high-value crops. Soil fumigants are applied to control weeds, plant pathogens, nematodes and insects. However, soil fumigants are susceptible to rapid emission after being applied to soil due to their high volatility. High levels of fumigant emissions may endanger the health of workers and bystanders and also contribute to volatile organic compounds in the air that adversely aﬀect air quality.1 Use of agricultural plastic ﬁlms (tarps) to cover the treated ﬁeld after fumigation has been a common practice to reduce the fumigant emission and to minimize worker exposure to the fumigants. Recently, the U.S. Environmental Protection Agency (U.S. EPA) established a set of regulations requiring a suite of complementary mitigation measures to protect handlers, reentry workers, and bystanders from risks resulting from exposure to the soil fumigants.1 Among the requirements is the need for a buﬀer zone between treated and untreated sites to allow airborne fumigant residues to disperse before reaching bystanders. U.S. EPA is also granting “buﬀer zone credits”, which reduce buﬀer distances, to encourage users to employ practices that reduce emissions, such as the use of high-barrier tarps.1 This article not subject to U.S. Copyright. Published 2011 by the American Chemical Society

The beneﬁts of using high-barrier ﬁlms during fumigation have been documented in many studies.2 7 However, various agricultural ﬁlms from many diﬀerent manufacturers have been used by farmers and commercial organizations during the fumigation of soils to control the emission of fumigants and to retain fumigants in the soil for longer periods.5,8 11 Available ﬁlms include low-density polyethylene (LDPE), high-density polyethylene (HDPE), metalized ﬁlms, and multilayer ﬁlms with imbedded barrier materials (e.g., polyamide and ethylene vinyl alcohol-EVOH), the latter often designated as virtually impermeable ﬁlms (VIF) or totally impermeable ﬁlms (TIF). The permeability of the ﬁlms to fumigants varies widely with material composition, manufacturing technique, and manufacturer. In addition, the manufacturer assigned name designations (e.g., VIF, TIF) are somewhat arbitrary and permeability of these ﬁlms may not directly correlate with their designations. Many factors including ﬁeld conditions, temperature, fumigant type, ﬁlm Received: May 26, 2011 Accepted: October 5, 2011 Revised: September 27, 2011 Published: October 05, 2011 9711

dx.doi.org/10.1021/es201278p | Environ. Sci. Technol. 2011, 45, 9711–9718

Environmental Science & Technology

ARTICLE

Table 1. List of Agricultural Films Submitted by Various Manufacturers and Vendors Film Name

Description

Manufacturer

polyethylene ﬁlms (PE): AEP Sun Film High Barrier

1.0 mil, clear polyethylene

AEP Inc.

Cadillac HDPE

1.25 mil, clear

Cadillac Products Packaging Co.

Canslit Embossed HDPE

0.6 mil, black

Canslit Inc. /Imaﬂex Inc.

Canslit Embossed LDPE

1.25 mil, black

Canslit Inc. /Imaﬂex Inc.

Pliant Embossed LDPE

1.25 mil, embossed LDPE

Pliant Corp

Canslit Metalized Canslit Metalized

1.25 mil black/silver 1.25 mil white/silver

Canslit Inc. /Imaﬂex Inc. Canslit Inc. /Imaﬂex Inc.

Pliant Metalized

1.25 mil, black/silver

Pliant Corp

Cadillac VIF

1.25 mil, black

Cadillac Products Packaging Co.

Can-Block VIF

0.8 mil, black

Canslit Inc./Imaﬂex Inc.

FilmTech VIF

1.25 mil, black

FilmTech Corp.

Ginegar Ozgard

1.25 mil, black

Ginegar Plastic Products, Ltd.

Ginegar VIF Guardian Oleﬁnas VIF

1.25 mil, embossed black 1.2 mil, embossed black

Ginegar Plastic Products, Ltd. Guardian Agroplastics Oleﬁnas USA

MidSouth VIF

1.25 mil, embossed black

Mid South Extrusion

Pliant Blockade Black

1.25 mil, black

Pliant Corp

Pliant Blockade White

1.25 mil, white/black

Pliant Corp

Metalized Films:

Virtually Impermeable Films (VIF):

Totally Impermeable Films (TIF): AEP-One

EVOH barrier, 1.0 mil, Clear

AEP Inc.

BayFilm

2 mil, black, contains halosulfuron-methyl

Bayer Innovation

Berry EVOH-High Barrier Berry High Barrier w/improved toughness

EVOH barrier, black EVOH barrier, black

Berry Plastics Berry Plastics

Berry EVOH-Supreme Barrier

EVOH barrier, black

Berry Plastics

Dow SARANEX A

black

Dow Chemical Co.

Dow SARANEX B

black

Dow Chemical Co.

Klerks/HyPlast TIF

clear

Klerk’s Plastic/HyPlast

Raven TIF VaporSafe

1.0 mil, EVOH barrier, clear

Raven Industries Inc.

Raven TIF VaporSafe

1.4 mil, EVOH barrier, black

Raven Industries Inc.

stretching, and gluing in the ﬁeld, can potentially alter the permeability of the ﬁlms as well.12,13 Previous studies have described analytical approaches for determining the permeability of agricultural ﬁlms 14,15 and attempts have been made to characterize the ﬁlms based on their permeability.16 18 However, because of the lack of a standard method for classifying the available ﬁlms according to their permeability to fumigants and because of the limited available data on the permeability of these ﬁlms to fumigants, it has been diﬃcult for the regulatory agencies and fumigant applicators to systematically evaluate the permeability of the ﬁlms on the market and to reliably develop buﬀer zone credits. In order to mitigate this uncertainty, a systematic laboratory test for the ﬁlm permeability of commonly used agricultural ﬁlms against several fumigants was conducted to establish a database of the ﬁlm permeability for those fumigants. The eﬀects of temperature and humidity on the ﬁlm permeability to fumigants were also investigated. Permeability is expressed as a “mass transfer coeﬃcient” (MTC) which is dependent only upon the properties of the ﬁlm, the properties of the fumigant, and physical conditions of the environment (such as temperature and humidity) and are independent of compound concentrations within the test system.13 15 Results from this study could assist the management of fumigant emissions by growers and fumigant applicators

and help regulators in the evaluation of the beneﬁts of agricultural ﬁlm usage, assessment of fumigation risks, and reduction of the uncertainties in potential buﬀer zone credit calculations.

’ EXPERIMENTAL SECTION The test method used in this study was adapted from a previously published technique.14,15 The test apparatus consists of an airtight cylinder constructed from two stainless steel endcaps (chambers) separated by the test ﬁlm. The basis for the test is that as fumigants penetrate the ﬁlm over time, fumigant concentrations in the source chamber will decline while concentrations in the receiving chamber will increase until equilibrium is achieved. Test compounds were introduced into the chamber on one side of the ﬁlm (source chamber). The concentrations of the test compounds from both sides of the ﬁlm (source and receiving chambers) were monitored over time by analyzing the vapor in each chamber. The rate of change in the fumigant concentrations over time was used to calculate the mass transfer coeﬃcient (MTC) of the ﬁlm for each target compound as described previously.14,15 Agricultural Films and Fumigants. Twenty seven samples of agricultural films were obtained from 13 manufacturers or their authorized vendors in 2009 and 2010. The names of the tested 9712

dx.doi.org/10.1021/es201278p |Environ. Sci. Technol. 2011, 45, 9711–9718

Environmental Science & Technology

ARTICLE

Table 2. Mass Transfer Coeﬃcient (MTC, cm/h, Average of Triplicates) of 27 Films to Various Fumigants at 25°C and Ambient Humidity (Relative Humidity of 35%-45%) in the Laboratory MeBr

IOM

PPO

1,3 D, cis

1,3-D, trans

DMDS

MITC

PIC

SF

AEP Sun Film High Barrier Cadillac HDPE

0.8238 0.4800

1.4124 0.8157

0.5250 0.2896

4.2344 2.6985

5.6792 3.8916

3.8840 2.3719

8.5298 7.3816

1.5404 0.7104

0.0107 0.0060

Canslit LDPE

0.6269

1.0387

0.4189

3.2106

4.4027

2.9652

8.1979

1.2124

0.0089

Canslit HDPE

1.3611

2.1783

0.8553

6.3171

7.9845

5.6798

13.4456

2.2558

0.0200

Pliant Regular LDPE

1.1078

1.9215

0.7473

5.6506

7.3937

5.2480

11.7678

2.2710

0.0157

Canslit Metalized Black

0.0217

0.0360

0.0124

0.1286

0.1982

0.1107

0.3696

0.0362

0.0003

Canslit Metalized White

0.0185

0.0313

0.0111

0.1153

0.1796

0.1031

0.3030

0.0342

0.0002

Pliant Metalized

0.0570

0.0876

0.0359

0.3469

0.6251

0.2771

1.1435

0.0828

0.0006

PE Films

Metalized Films

VIFs Cadillac VIF

0.0085

0.0061

0.0053

0.0109

0.0232

0.0060

0.1093

0.0011

0.0020

Can-Block

0.0047

0.0018

0.0012

0.0012

0.0036

0.0008

0.0366

0.0001

0.0003

FilmTech VIF

0.0029

0.0015

0.0011

0.0019

0.0052

0.0009

0.0512

0.0001

0.0000

Ginegar VIF

0.0053

0.0030

0.0023

0.0033

0.0074

0.0017

0.0592

0.0005

0.0007

Ginegar Ozgard

0.0019

0.0007

0.0006

0.0005

0.0019

0.0003

0.0180

0.0001

0.0000

Guardian Oleﬁnas VIF

0.0151

0.0109

0.0089

0.0235

0.0523

0.0120

0.2928

0.0016

0.0000

Mid South VIF Pliant Blockade, Black

0.0017 0.0045

0.0008 0.0020

0.0019 0.0013

0.0024 0.0022

0.0047 0.0050

0.0021 0.0013

0.0097 0.0527

0.0000 0.0010

0.0000 0.0001

Pliant Blockade, White

0.0057

0.0027

0.0015

0.0025

0.0067

0.0010

0.0823

0.0001

0.0001

AEP-One

0.0001

0.0000

0.00004

0.0002

0.0003

0.0002

0.0003

0.0002

0.0000

BayFilm

0.0001

0.0001

0.0001

0.0007

0.0009

0.0006

0.0016

0.0005

0.0001

Berry High Barrier

0.0000

0.0000

0.0000

0.0001

0.0001

0.0000

0.0002

0.0001

0.0000

Berry High Barrier w/improv toughness

0.0000

0.0001

0.0000

0.0001

0.0002

0.0002

0.0004

0.0000

0.0000

Berry Supreme Barrier Dow Saranex (A)

0.0009 0.0006

0.0001 0.0006

0.0002 0.0003

0.0002 0.0008

0.0003 0.0011

0.0002 0.0004

0.0002 0.0051

0.0002 0.0007

0.0000 0.0000

Dow Saranex (B)

0.0002

0.0002

0.0001

0.0003

0.0005

0.0002

0.0021

0.0002

0.0000

Klerks/HyPlast

0.0010

0.0001

0.0002

0.0000

0.0002

0.0000

0.0097

0.0000

0.0000

Raven VaporSafe 1.0 mil

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

0.0000

Raven VaporSafe 1.4 mil

0.0001

0.0001

0.0001

0.0002

0.0003

0.0002

0.0005

0.0002

0.0000

TIFs

films and their manufacturers are listed in Table 1. Fumigant standards were obtained from several sources. Methyl bromide (MeBr) was obtained from Chemtura Corp (Williamston, SC). Iodomethane (IOM), 1,3-dichloropropene (1,3-D, mixture of cis and trans isomers), dimethyl disulfide (DMDS), propylene oxide (PPO), and methyl isothiocyante (MITC, transformation product of metam sodium or dazomet during fumigation) were purchased from Sigma-Aldrich. Chloropicrin (PIC) was obtained from Arysta LifeScience (Cary, NC). Sulfuryl fluoride (SF) was obtained from Dow Chemical (Indianapolis, IN). Procedures. Films were tested initially under ambient laboratory humidity conditions, with the relative humidity ranging from 35 to 45%, and repeated under high relative humidity conditions to simulate the near saturated moisture conditions expected under a tarp in the field. Tests were carried out in a temperature controlled environmental chamber at 25 °C ((0.5 °C). Select VIFs and TIFs were tested again at 40 °C ((0.5 °C) under ambient humidity conditions. To obtain the high humidity conditions, 1 mL of distilled water was added to the source side of the permeability test cells and equilibrated overnight before adding fumigants in the cells. A visible amount of liquid water remained present in the cells at the end of each test, indicating that high humidity was maintained for the duration of the test.

The effective relative humidity in the source side of the permeability chamber was determined in a separate apparatus using a hand-held NIST (National Institute of Standards and Technology) traceable hygrometer and the relative humidity was found to be approximately 90% ((2%). Samples of agricultural ﬁlms were placed between the two halfcells and the cells were joined together by epoxy glue to form a gastight seal. Aluminum tape was then applied to the outside of the cells to provide additional support and sealing of the apparatus. The cells were placed inside the temperaturecontrolled environmental chamber and equilibrated before fumigants were introduced. Generally, triplicate permeability cells were constructed for each ﬁlm type and the calculated MTCs were the average of the triplicates. At the beginning of each test, about 30 40 mL of premixed fumigant vapor was placed into the source side of each permeability cell. Excessive pressure was released through a valve installed on the permeability cell. During the course of testing, gas samples (250 μL) from both receiving and source sides of the permeability cells were collected periodically and fumigant concentrations were measured. The fumigant concentrations in collected samples were measured with an Agilent 6890 gas chromatograph/5973 mass spectrometer system (GC/MSD) 9713

dx.doi.org/10.1021/es201278p |Environ. Sci. Technol. 2011, 45, 9711–9718

Environmental Science & Technology

Figure 1. Cross plot of the MTCs for MeBr and IOM of the 27 ﬁlms at 25 °C and ambient humidity. Logarithmic scales are used to display the low MTC values. Films are grouped by manufacturer designation and reﬂect a range of MTC values within each group.

in the select ion monitoring (SIM) mode interfaced with an Agilent headspace autosampler. The temperature settings for the headspace autosampler were 80 °C (oven), 90 °C (loop), and 100 °C (transfer line). The GC column used was a DB-624 column (30 m 0.25 mm ID, 1.4 μm ﬁlm thickness). The GC oven temperature was programmed to hold for 3 min at the initial condition of 40 °C. The temperature was then increased to 50 at 10 °C/min, held for 10 min, followed by raising the temperature to 110 °C at a rate of 20 °C/min. The MTCs were calculated based on the compound responses at each sampling time using the ﬁlm permeability calculator (FilmPC v1.0.2), the permeability calculation software developed and provided by Scott Yates of the Agricultural Research Service, U.S. Department of Agriculture (ARS, USDA) at Riverside, CA. Most of the ﬁlms tested were designated as VIF and TIF. The testing period for each of these ﬁlms lasted up to 12 days. At the end of the testing period for some TIF ﬁlms, some fumigants were still not detected in the receiving chamber of the permeability cells, or the detectable amount was so low that the calculated MTC was below 1 10 4. The calculated MTCs in these cases were treated as zero. The relative standard deviation for replicate analyses was, in general, within 30% for the MTCs that were above 0.001 cm/h, but usually higher for ﬁlms with very low MTCs.

’ RESULTS AND DISCUSSION 1. Mass Transfer Coefficients of Films for Various Fumigants. The MTC results of the 27 films from thirteen manufacturers are listed in Table 2. The results clearly show that different films possess very different permeability for each compound. For example, the MTC for MeBr ranges from below 1 10 4 to 1.2 cm/h for the 27 different films tested, a range spanning 5 orders of magnitude (Table 2). The low density polyethylene (LDPE) films and high density polyethylene (HDPE) films, in general, have high permeability to all tested fumigants. Metalized PE films have lower permeability than the LDPE and HDPE films. The multilayered VIFs (usually with a polyamide barrier layer) generally have low permeability. The VIFs from different manufacturers, however, have vastly different MTC values. For example, the MTC for MeBr ranges from 0.0017 to 0.0151 cm/h for the nine VIFs, a 10-fold range, even though the MTC value is small (Table 2). The large

ARTICLE

Figure 2. Mass transfer coeﬃcients (MTCs) of 27 tarps for diﬀerent fumigants at 25 °C and ambient humidity. The MTCs of VIF and TIF ﬁlms are low and lay nearly along the x-axis.

range in the MTCs among the VIFs for all the fumigants suggests that the permeability of the films with similar name designations, and with supposedly similar material and construction, can differ substantially in their permeability. The MTCs for TIFs (typically with an EVOH barrier layer) are generally lower than that of VIFs, with the MTC for MeBr as an example ranging from 1 10 4 to 0.0010 cm/h, a range of more than 2 orders of magnitude. The permeability variations among the different films are in agreement with previous reports.13,18 Because of the vastly diﬀerent permeability of the ﬁlms to various fumigants, it is diﬃcult for users and regulators to estimate the emissions of fumigants in ﬁeld applications without conducting detailed permeability determination of each ﬁlm used in the ﬁeld. In order to simplify the eﬀort of estimating the fumigant emissions, managing the exposure risk, and granting relief from fumigant buﬀer zones (e.g., in the form of credits) if users opt to use less permeable ﬁlms, the ﬁlms may be broadly separated into diﬀerent categories based on the distribution and clustering of the MTCs of various ﬁlms to the fumigants. Figure 1 is a cross-plot of the MTC values of the 27 ﬁlms for MeBr and IOM. The separation of ﬁlms generally follows the manufacturer designations, for example, PE ﬁlms are within one group with the highest permeability, metalized ﬁlms are in another group with moderate permeability, and VIF ﬁlms are in one group with low permeability. Although the separation is not distinct in the values of MTC for MeBr or IOM alone between VIF and TIF, all TIF ﬁlms with the lowest MTC values can be grouped together using the combination of MeBr and IOM (Figure 1). The separations of these ﬁlms using diﬀerent fumigant combinations are generally similar. A more elaborated clustering analyses of the MTC values of individual fumigants and a composite value of all fumigants using a statistical approach produced similar groupings.19 Although the permeability of each ﬁlm to diﬀerent fumigants diﬀered from each other, they generally track with each other among the diﬀerent ﬁlms (Figure 2). For example, sulfuryl ﬂuoride was generally the slowest to permeate through a ﬁlm compared to other fumigants. The calculated MTC of the ﬁlms for SF ranged from below 1 10 4 to 0.0200 cm/h. MITC was the fastest fumigant to permeate through any ﬁlm compared to other fumigants. The MTC for MITC in the LDPE and HDPE was as high as 13 cm/h. Calculated ﬁlm permeability was generally similar for MeBr, IOM, PIC, and PPO. The MTCs for DMDS and both isomers of 1,3-D were generally higher than 9714

dx.doi.org/10.1021/es201278p |Environ. Sci. Technol. 2011, 45, 9711–9718

Environmental Science & Technology

ARTICLE

Figure 3. Examples of time needed for a fumigant (25 °C and ambient humidity) to reach equilibrium across a ﬁlm with MTCs of approximately (a) 10, (b) 1, (c) 0.1, and (d) 0.01 cm/h. The upper curve and data points indicate the amount of the fumigant in the source chamber (100% at time zero) at diﬀerent sampling times after fumigant is placed in the source chamber. The lower curve and data points indicate the amount of fumigant in the receiving chamber (0% at time zero).

other fumigants for the same ﬁlms (except MITC) though the trans-1,3-D generally moved faster through a ﬁlm than the cisisomer. Similar patterns in the rate of permeation among diﬀerent fumigants were also observed in another study.18 As a result, the groupings of ﬁlms can be represented by select fumigants, such as MeBr-speciﬁc grouping or a global grouping using composite value for all fumigants.19 MTC and Emission Rate. To better understand the relationship between the MTC and the emission of fumigants through the agricultural films, the time needed for a fumigant to reach near equilibrium across the film in the source and receiving chambers in the laboratory is shown in Figure 3 for films with MTC values of 10, 1, 0.1, and 0.01 cm/h, a typical range for PE film, metalized film and VIF film. In theory, it will take an infinite amount of time to reach true equilibrium on both sides of the film.15,18 In practice, when the process reaches about 95% of the true equilibrium, this status can be considered near equilibrium.18 This practically defined end point is generally within the current analytical limitations and uncertainty of the techniques used to determine the target compounds and the MTC of the films. For a compound to reach near equilibrium across a ﬁlm with a MTC of about 10 cm/h (e.g., the PE ﬁlms for MITC) the time needed was less than 1 h (about 0.8 h, Figure 3a). It took about 8 h for a compound to reach the near equilibrium state at both sides of the ﬁlm in a closed system if the MTC decreased to 1 cm/h (e.g., the PE ﬁlms for MeBr, IOM, and PIC, Figure 3b). This observation is consistent with the general belief that the PE ﬁlms are permeable to fumigants.13 When the MTC further decreased 10-fold to 0.1 cm/h (e.g., the metalized ﬁlms to DMDS, PIC, IOM) it took about 3 days (about 80 h, a 10-fold increase) for the compound to reach such a state (Figure 3c). As the MTC

decreased another 10-fold to 0.01 cm/h (e.g., most of the VIF ﬁlms for MeBr, IOM) it took about 800 h (a 10-fold increase) to reach near equilibrium (Figure 3d). The inversely proportional relationship between the increase in the time needed to reach the near equilibrium and the decrease in the MTC demonstrates the usefulness of this parameter as an indicator of ﬁlm property. The time needed to reach near equilibrium in Figure 3 was for a closed system and for approximately 50% of the fumigants to permeate through the ﬁlm. In actual ﬁeld settings, the “receiving side” is open atmosphere. The expected time needed to reach the near equilibrium state in an open system (i.e., no fumigant remaining under the ﬁlm) would be longer than that in a closed system because the entire amount of the fumigants would need to pass through the ﬁlms to reach equilibrium, whereas only 50% of the fumigants needs to permeate through a ﬁlm to reach equilibrium in a laboratory closed system. In one test with a PE ﬁlm, the time needed for about 95% of the fumigants in the source chamber to pass through a ﬁlm in an open system (simulated by using the test apparatus with no receiving chamber) was approximately 24 h for MeBr (Figure 4a and Table 1 in the Supporting Information). In comparison, it took approximately 7 h for MeBr to reach near equilibrium (about 53% in the source chamber) in the two chamber closed system described in this study (Figure 4b). In this case, it took approximately three to four times longer to reach near equilibrium in an open system. This could have implications when translating MTC values obtained from closed system tests in the laboratory to actual ﬁeld conditions. Because other types of ﬁlms have not been tested, it is unknown whether the time needed for about 95% of fumigants to permeate through in the simulated open system is longer than the closed system by similar extent. 9715

dx.doi.org/10.1021/es201278p |Environ. Sci. Technol. 2011, 45, 9711–9718

Environmental Science & Technology

Figure 4. Concentration change over time for MeBr in the source side in (a) an open system (MeBr permeating through the ﬁlm into air. 0% is the equilibrium concentration in the source); and (b) a closed system (50% is the equilibrium concentration in the source) for one HDPE ﬁlm. Symbols are the measured amount in triplicate tests. The measured MTC of this PE ﬁlm for MeBr is 0.824 cm/h. The time needed to reach 95% of the equilibrium concentration (about 53% in the source side) is approximately 7 h in the closed system. The time needed to reach 95% of the equilibrium concentration (about 5% in the source side) in the open system is about 24 h. Each test was conducted in triplicate. Other compounds exhibit similar changes (Table 1 in the Supporting Information).

In actual ﬁeld application of fumigants, the fumigants are injected into soil or drip applied into the soil. The amount of fumigant available for permeation through the ﬁlm probably is much less than the amount applied. In addition to the absorption and retention of fumigants by soil, fumigants in soil are degraded over time.20,21 For example, PIC, 1,3-D, and MITC are subject to fast degradation, either by biological processes in soil, or chemical processes in soil and vapor. The amount of emission of fumigants and the time needed for fumigants to dissipate under tarps is therefore much more complex to estimate than the ideal conditions in the laboratory. Effects of Environmental Conditions on the Permeability. Previous studies reported that gas transport through plastic films could be affected by relative humidity 18,22,23 and temperature.12,13,24,25 Because agricultural films can be exposed to high humidity during soil fumigation, it is important to characterize the effects of relative humidity on film permeability to fumigant vapors. Relative Humidity. The permeability of agricultural films, particularly those with low MTC values, can be greatly affected by the moisture content in the air.18 In this study, when VIF films were exposed to high relative humidity conditions, the MTCs increased by up to 3 orders of magnitude (Table 3). Similar effects by humidity on the permeability of some VIF films were also reported previously.18 The MTCs

ARTICLE

of many VIF films under high relative humidity conditions were close to the MTCs seen for regular LDPE and HDPE films (Tables 2 and 3 and Figures 1 and 2 in the Supporting Information). High humidity only had a slight effect on the permeability of the PE films. The increase in the MTC of PE films due to high humidity was less than 20% for most fumigants (Tables 2 and 3) and was generally within the range of the analytical method uncertainties. These results of the minor humidity effects on the permeability of PE films are similar to the results of a previous report.13 In contrast, the VIF tarps nearly lost all their original impermeability for fumigants when tested under high humidity. The effects of humidity on the permeability of metalized films are smaller than that of the VIF tarps. The implication of these humidity effects on the permeability of films is that high humidity in the air under a tarp in the field could potentially compromise the fumigant retaining ability of the VIF films. The EVOH layer in the TIF ﬁlms was an excellent barrier layer, reducing the MTCs to below 0.001 cm/h under ambient humidity conditions in laboratory tests. However, when the ﬁlms were exposed to high relatively humidity, the ability of this barrier layer to block the passage of fumigants was greatly reduced (Table 3). The MTCs of the TIF ﬁlms increased by 2 3 orders of magnitude under high relative humidity compared to that under ambient humidity conditions at the same temperature of 25 °C (Tables 2 and 3). However, the MTCs of the TIFs under high humidity remained low compared to the VIF under the same high humidity conditions. The MTC was below 0.1 cm/h for most of the TIFs for all the tested fumigants under high humidity conditions (Table 3), suggesting that the TIFs would perform better in reducing the emission of fumigants compared to other types of ﬁlms. In addition to its aﬀect on ﬁlm permeability, elevated humidity may also inﬂuence the stability of fumigants in the vapor. MITC appeared to be signiﬁcantly less stable under high humidity conditions during the tests, with over 80% of MITC degraded in two days (Figure 3 in the Supporting Information). In comparison, recoveries of SF, MeBr, IOM, and PPO from the cells remained largely between 70% and 120%. This eﬀect was observed during the tests of various ﬁlms. Any degradation of MITC in the source side of the high humidity test cell, relative to the receiving side, would have resulted in shorter apparent equilibrium time, and therefore the calculated MTC value for MITC was likely biased high. However, because MITC was the fastest compound to permeate any ﬁlm and equilibrium reached in a relatively short time, the bias in the MTC of MITC caused by the degradation of MITC under high relative humidity conditions may not be substantial. It is possible that the presence of liquid water in the source side of the permeability cells under high humidity conditions may aﬀect calculated MTC values if fumigants are partitioned into the liquid water. However, recoveries of most fumigants from the vapor phase suggest that the amount partitioned into the liquid phase is minimal. Recovery of MITC decreased rapidly to less than 20% of the initial amount within two days in all the tests (Figure 3 in the Supporting Information). If MITC had been sequestered by the small amount of liquid water (a few drops) in the source side, the decrease in the recovery would be sharp followed by a ﬂattened pattern. Two TIFs that do not use EVOH barrier material showed no apparent eﬀects from humidity. The MTCs of both Dow Saranex ﬁlms were similar under high relative humidity and 9716

dx.doi.org/10.1021/es201278p |Environ. Sci. Technol. 2011, 45, 9711–9718

Environmental Science & Technology

ARTICLE

Table 3. Mass Transfer Coeﬃcient (MTC, cm/h, Average of Triplicates) of Agricultural Films to Various Fumigants at 25°C and High Humidity (Relative Humidity of 90%) ﬁlms

MeBr

IOM

PPO

1,3-D, cis

1,3-D, trans

DMDS

MITC

PIC

SF

0.5000 1.2113

0.8686 2.1286

0.2913 0.7679

2.8621 8.8657

4.8446 12.9526

2.5951 7.8947

7.6397 23.9830

0.7754 2.6259

0.0072 0.0185

Canslit Metalized Black

0.0711

0.0828

0.0694

0.1340

0.1645

0.1257

0.3361

0.0919

0.0027

Pliant Metalized

0.1390

0.2030

0.0874

0.6072

0.9236

0.5360

2.1128

0.1713

0.0017

Cadillac VIF

0.1667

0.2066

0.2586

0.7273

1.3020

0.5438

6.6230

0.1563

0.0020

Can-Block

0.1430

0.1336

0.1395

0.3691

0.6684

0.2560

2.1650

0.0960

0.0010

FilmTech VIF Ginegar VIF

0.3542 0.2682

0.4533 0.3183

0.6339 0.3990

1.6355 1.1049

2.5607 2.3772

1.3169 0.8056

5.6756 10.7761

0.3887 0.2419

0.0026 0.0152

Ginegar Ozgard,

0.1631

0.2114

0.2460

0.8037

1.3461

0.6037

3.2284

0.1601

0.0010

Guardian Oleﬁnas VIF

0.2662

0.3236

0.3430

1.0608

2.3161

0.7783

11.5732

0.2637

0.0023

Mid South VIF

0.0878

0.0892

0.1300

0.1851

0.2878

0.1473

0.7659

0.0774

0.0012

Pliant Blockade, Black

0.2826

0.3441

0.3632

1.0154

1.4204

0.7971

3.2613

0.2811

0.0061

Pliant Blockade White

0.7421

1.0520

0.9096

4.7243

7.9022

3.6111

15.9521

0.9499

0.0051

0.0133 0.6182

0.0131 0.7838

0.0333 0.2897

0.0250 3.0451

0.0382 4.6441

0.0199 2.8390

0.1512 7.4034

0.0029 0.9628

0.0001 0.0067

PE Films Cadillac, HDPE Pliant Regular Black LDPE Metalized Films

VIFs

TIFs AEP-One BayFilm Berry EVOH High Barrier

0.0355

0.0289

0.0780

0.0535

0.0929

0.0482

0.5031

0.0038

0.0009

Berry High Barrier w/Improv toughness

0.0337

0.0293

0.0641

0.0524

0.0803

0.0458

0.1755

0.0108

0.0002

Berry EVOH Supreme Barrier

0.0446

0.0446

0.0818

0.0626

0.0806

0.0587

0.1063

0.0157

0.0009

Dow SARANEX A

0.0009

0.0006

0.0007

0.0013

0.0021

0.0009

0.0397

0.0000

0.0000

Dow SARANEX B

0.0002

0.0001

0.0002

0.0015

0.0013

0.0003

0.0099

0.0004

0.0000

Klerks/HyPlast

0.0834

0.0929

0.1429

0.2745

0.5024

0.1963

2.0422

0.0758

0.0031

Raven TIF VaporSafe 1.4 mil Raven TIF VaporSafe 1.0 mil

0.0069 0.0115

0.0047 0.0118

0.0179 0.0288

0.0129 0.0168

0.0213 0.0275

0.0099 0.0186

0.0796 0.0743

0.0007 0.0005

0.0000 0.0001

ambient humidity conditions at 25 °C (Tables 2 and 3) for all the fumigant except for MITC, further evidence suggesting that the presence of liquid water did not impact the calculated MTC values signiﬁcantly. The calculated MTC value for MITC was higher under high humidity conditions than ambient humidity conditions, likely the result of the bias caused by the degradation of MITC under high humidity conditions. Dow Saranex ﬁlms use a diﬀerent barrier layer, polyvinylidene chloride, which appeared to be not aﬀected by the moisture in the air. Temperature. The effects of rising temperature on the permeability, in contrast, were much smaller than the effects of humidity, as shown by the change of the permeability of twelve VIF and TIF tarps at 40 °C. Several films, such as Berry Supreme Barrier, Raven 1.4 mil TIF, Ginegar VIF and Cadillac VIF films, showed almost no change between the MTCs measured at 25 °C and at 40 °C for MeBr, SF and IOM (Table 2 in the Supporting Information). The measured MTCs at both temperatures were within the range of expected experimental variability. The MTCs of several films, such as FilmTech VIF, Pliant Blockade (black), Ginegar Ozgard, MidSouth VIF, Guardian Olefinas VIF, and AEPOne, increased approximately 2 5 times at 40 °C compared with those at 25 °C for MeBr and IOM. The increase in the permeability at 40 °C compared to that at 25 °C was between f5 and 10 times for Dow Saranex and Klerks/Hyplast TIF. Similar increases in the permeability with temperature were

also observed previously,13,18 with a 2 5 times increase in the permeability generally expected for every 10 °C increase.18 The calculated MTC values for MITC, DMDS, 1,3-D, and PIC were more variable at 40 °C than that for MeBr, IOM, and SF. The laboratory determined MTCs for ﬁlms were achieved under controlled, ideal conditions. During actual ﬁeld fumigant applications and tarp placement, many factors could potentially alter the emission rate of fumigants through agricultural ﬁlms, such as adsorption and retention of fumigants by soil, degradation of fumigants, variation of humidity under the tarp, air temperature, handling and installation techniques of the ﬁlms, and deterioration of the ﬁlms in the ﬁeld. 13,25,26 Some compounds, such as MITC, 1,3-D, PIC, are unstable under high moisture conditions, particularly when exposed to sunlight (refs 26,27 and Figure 3 in the Supporting Information). Some fumigants may be quickly degraded, thus reducing their potential emission to atmosphere. Because fumigants are retained in the soil, emission through the tarp is much less compared to the scenario that all the fumigants are placed under the tarp and are available for permeation through the tarp. High relative humidity under the tarps can further increase emission, particularly when the temperature is high during day time. Prediction of actual fumigant emission rates is therefore complicated by many interrelated processes. 9717

dx.doi.org/10.1021/es201278p |Environ. Sci. Technol. 2011, 45, 9711–9718

Environmental Science & Technology

’ ASSOCIATED CONTENT

bS

Supporting Information. Additional tables and ﬁgures 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: (410)305-2636; fax (410-305)3091; e-mail: Qian.yaorong@ epa.gov.

’ ACKNOWLEDGMENT We are grateful that Elizabeth Kolbe reviewed the quality of all the data for this study. Chemtura, Dow Chemical, and Arysta LifeScience provided methyl bromide, sulfuryl ﬂuoride, and chloropicrin standards, respectively. The MTC calculation application software was provided by Scott Yates. We thank the ﬁlm manufacturers for sending sample ﬁlms for testing. Detailed statistical analysis of the permeability and ﬁlm ranking are documented in a U.S. EPA report.19 Mention of trade names or commercial products in this publication is solely for the purpose of providing speciﬁc information and does not imply recommendation or endorsement by the U.S. EPA. The views expressed in this paper are those of the authors and do not necessarily represent those of the US EPA. No oﬃcial Agency endorsement should be inferred. ’ REFERENCES (1) U.S. EPA pesticide registration website. http://www.epa.gov/ oppsrrd1/reregistration/soil_fumigants/implementing-new-safetymeasures.html (accessed October 17, 2011). (2) Austerweil, M.; Steiner, B.; Gamliel, A. Permeation of soil fumigants through agricultural plastic ﬁlms. Phytoparasitica. 2006, 34 (5), 491–501. (3) Gao, S.; Trout, T. J. Surface seals reduce 1,3-dichloropropene and chloropicrin emissions in ﬁeld tests. J. Environ. Qual. 2007, 36 (1), 110–119. (4) Gao, S.; Hanson, B. D.; Qin, R; Wang, D.; Yates, S. Developing agricultural practices to reduce emissions from soil fumigation using ﬁeld plot tests. J. Environ. Qual. 2010, 0422, DOI:10.2134/jeq2009. (5) Wang, D.; Yates, S. R.; Ernst, F. F.; Gan, J.; Jury, W. A. Reducing methyl bromide emission with high-barrier ﬁlm and reduced dosage. Environ. Sci. Technol. 1997, 31 (12), 3686–3691. (6) Wang, D.; Yates, S. R. Methyl bromide emission from ﬁeld partially covered with a high-density polyethylene and a virtually impermeable ﬁlm. Environ. Sci. Technol. 1998, 32 (17), 2515–2518. (7) Samtani, J. B.; Ajwa, H. A.; Goodhue, R. E.; Daugovish, O.; Kabir, Z.; Fennimore, S. A. Weed control eﬃcacy and economics of 1,3dichloropropene and chloropicrin applied at reduced rates under impermeable ﬁlm in strawberry beds. HortScience 2010, 45 (12), 1841–1847. (8) Chellemi, D. O.; Mirusso, J. Optimizing soil disinfestation procedures for fresh market tomato and pepper production. Plant Disease 2006, 90 (5), 668–674. (9) Gamliel, A.; Grinstein, A.; Klein, L; Cohen, Y; Katan, J. 1998. Permeability of plastic ﬁlms to methyl bromide: Field study. Crop Prot. 1998, 17 (3), 241 248. (10) Gilreath, J. P.; Motis, T. N.; Santos, B. M. Cyperus spp. control with reduced methyl bromide plus chloropicrin doses under virtually impermeable ﬁlms in pepper. Crop Prot. 2005, 24 (3), 285–287. (11) Papiernik, S. K.; Yates, S. R.; Dungan, R. S.; Lesch, S. M.; Zheng, W.; Guo, M. Eﬀect of surface tarp on emissions and distribution of dripapplied fumigants. Environ. Sci. Technol. 2004, 38 (16), 4254–4262.

ARTICLE

(12) Wang, D.; Yates, S. R.; Gan, J.; Knuteson, J. A. Atmospheric volatilization of methyl bromide, 1,3-dichloropropene, and propargyl bromide through two plastic ﬁlms: Transfer coeﬃcient and temperature eﬀect. Atmos. Environ. 1999, 33 (3), 401–407. (13) Papiernik, S. K.; Yates, S. R. Eﬀect of environmental conditions on the permeability of high density polyethylene ﬁlm to fumigant vapors. Environ. Sci. Technol. 2002, 36 (8), 1833–1838. (14) Papiernik, S. K.; Yates, S. R.; Gan, J. An approach for estimating the permeability of agricultural ﬁlms. Environ. Sci. Technol. 2001, 35 (6), 1240–1246. (15) Papiernik, S. K.; Ernst, F. F.; Yates, S. R. An apparatus for measuring the gas permeability of ﬁlms. J. Environ. Qual. 2002, 31 (1), 358–361. (16) Yates, S. R., Chellemi, D.; Browne, G.; Wang, D.; Gao, G.; Hanson, B.; Ajwa, H.; Kluepfel, D. Update of Film Permeability Measurements for the USDA-ARS Area-Wide Research Project. 2008 Annual International Research Conference on Methyl Bromide Alternatives and Emissions Reductions. http://mbao.org/2008/018Yates. pdf (accessed October 17, 2011). (17) Ajwa, H. Testing Film Permeability to Fumigants Under Laboratory and Field Conditions. 2008 Annual International Research Conference on Methyl Bromide Alternatives and Emissions Reductions. http://mbao.org/2008/035Ajwa.pdf (accessed October 17, 2011). (18) Papiernik, S. K.; Yates, S. R.; Chellemi, D. O. Standardized approach for estimating the permeability of plastic ﬁlms to soil fumigants under various ﬁeld and environmental conditions. J. Environ. Qual. 2010, DOI: 10.2134/jeq2010.0118. (19) Second Update To Health Eﬀects Division Recommendations for Good Agricultural Practices and Associated Buﬀer Credits. 1/11/ 2011, Document ID: EPA-HQ-OPP-2005-0123-0748. http://www. regulations.gov/#!documentDetail;D=EPA-HQ-OPP-2005-0123-0748 (accessed October 17, 2011). (20) Ma, Q. L.; Gan, J.; Papiernik, S. K.; Becker, J. O.; Yates, S. R. Degradation of soil fumigants as aﬀected by initial concentration and temperature. J. Environ. Qual. 2001, 30 (4), 1278–1286. (21) Dungan, R. S.; Yates, S. R. Degradation of fumigant pesticides: 1,3-dichloropropene, methyl isothiocyanate, chloropicrin, and methyl bromide. Vadose Zone J. 2003, 2 (3), 279–286. (22) Johansson, F.; Leufven, A. Food packaging polymer ﬁlms as aroma vapor barriers at diﬀerent relative humidities. J. Food Sci. 1994, 59 (6), 1328–1331. (23) Lagaron, J. M.; Gimenez, E.; Catala, R.; Gavara, R. Mechanisms of moisture sorption in barrier polymers used in food packaging: Amorphous polyamide vs. high-barrier ethylene-vinyl alcohol copolymer studied by vibrational spectroscopy. Macromol. Chem. Phys. 2003, 204 (4), 704–713. (24) Muramatsu, M.; Okura, M.; Kuboyama, K.; Ougizawa, T.; Yamamoto, T.; Nishihara, Y.; Saito, Y.; Ito, K.; Hirata, K.; Kobayashi, Y. Oxygen permeability and free volume hole size in ethylene-vinyl alcohol copolymer ﬁlm: Temperature and humidity dependence. Radiat. Phys. Chem. 2003, 68 (3 4), 561–564. (25) Wang, D.; Yates, S. R.; Jury, W. A. Temperature eﬀect on methyl bromide volatilization: Permeability of plastic cover ﬁlms. J. Environ. Qual. 1998, 27 (4), 821–827. (26) Qin, R.; Gao, S.; McDonald, J. A.; Ajwa, H.; Shem-Tov, S.; Sullivan, D. A. Eﬀect of plastic tarps over raised-beds and potassium thiosulfate in furrows on chloropicrin emissions from drip fumigated ﬁelds. Chemosphere 2008, 72 (4), 558–563. (27) Yates, S. R.; Gan, J.; Papiernik, S. K. Environmental fate of methyl bromide as a soil fumigant. Rev. Environ. Contam. Toxicol. 2003, 177, 45–122.

9718

dx.doi.org/10.1021/es201278p |Environ. Sci. Technol. 2011, 45, 9711–9718

ARTICLE pubs.acs.org/est

Mechanisms of Efficient Arsenite Uptake by Arsenic Hyperaccumulator Pteris vittata Xin Wang,†,‡,||,z Lena Q. Ma,*,§,‡ Bala Rathinasabapathi,*,|| Yong Cai,^ Yun Guo Liu,z and Guang Ming Zengz †

College of Resources and Environmental Science, Hunan Normal University, Changsha, Hunan 410081, China Soil and Water Science Department and Horticultural Sciences Department, University of Florida, Gainesville, Florida 32611, United States § State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Jiangsu 210046, China ^ Department of Chemistry & Southeast Environmental Research Center, Florida International University, Miami, Florida 33199, United States z College of Environmental Science and Engineering, Hunan University, Changsha, Hunan 410082, China )

‡

ABSTRACT:

Arsenate (AsV) and arsenite (AsIII) are two dominant arsenic species in the environment. While arsenate uptake is via phosphate transporter in plants, including arsenic hyperaccumulator Pteris vittata, AsIII uptake mechanisms by P. vittata are unclear. In this study, we investigated AsIII uptake by P. vittata involving root radial transport from external medium to cortical cells and xylem loading. In the root symplastic solution, AsIII was the predominant species (90 94%) and its concentrations were 1.6 21 times those in the medium. AsIII inﬂux into root symplast followed Michaelis Menten kinetics with Km of 77.7 μM at external AsIII concentrations of 2.6 650 μM. In the presence of metabolic inhibitor 2,4-dinitrophenol (DNP), arsenic concentrations in the root symplast were reduced to the levels lower than in the medium, indicating that a transporter-mediated active process was mainly responsible for AsIII inﬂux into P. vittata roots. Unlike radial transport, AsIII loading into xylem involved both high- and low-aﬃnity systems with Km of 8.8 μM and 70.4 μM, respectively. As indicated by the eﬀect of 2,4-DNP, passive diﬀusion became more important in arsenic loading into xylem at higher external AsIII. The unique AsIII uptake system in P. vittata makes it a valuable model to understand the mechanisms of arsenic hyperaccumulation in the plant kingdom.

’ INTRODUCTION Arsenic is considered a class I human carcinogen.1 Worldwide arsenic contamination in the environment from both natural and anthropogenic sources is of increasing global concern.2,3 Being highly toxic to most forms of life, arsenic contamination in soil water plant systems has posed serious risks to food safety and public health. Arsenic occurs predominantly in inorganic form as arsenate (AsV) and arsenite (AsIII) in most ecosystems. Though AsIII is predominant in anaerobic conditions, it is also present in aerobic environments where AsV typically dominates. For example, as a result of biochemical transformation by plant roots4 and microbially mediated AsV reduction,5 AsIII accounts for an important part of arsenic pools in plant rhizosphere and aerobic soils. For r 2011 American Chemical Society

example, due to microbial reduction of AsV, AsIII was the predominant species eluted from oxic contaminated mine tailings.6 Both AsV and AsIII are present in surface water and groundwater.7 In South Asia, where the most extensive arsenic contamination in groundwater is observed, AsIII accounts for the dominant species (67 99%) in most groundwater samples.8,9 Even in oxygen-rich conditions, AsIII in water can exist as a metastable species with slow oxidation kinetics from AsIII to Received: May 26, 2011 Accepted: October 3, 2011 Revised: October 3, 2011 Published: October 26, 2011 9719

dx.doi.org/10.1021/es2018048 | Environ. Sci. Technol. 2011, 45, 9719–9725

Environmental Science & Technology AsV.7,10 However, the rate can be much faster due to the presence of AsIII-oxidizing microbes and catalytic metal oxides. Compared to AsV, AsIII is far more toxic to most organisms due to its high aﬃnity to bind with sulfhydryl groups, thus disrupting protein structure and functions involved in cellular metabolism.11 In addition, under circumneutral pH, AsIII exists as a neutral molecular species (pKa = 9.2) with relatively low chemical reactivity, leading to much more diﬃcult removal of AsIII from water by traditional treatment technologies.12 The arsenic hyperaccumulator Pteris vittata is eﬃcient in taking up both AsV and AsIII.13 On the basis of depletion techniques and radiotracer 73AsV uptake studies, it is established that AsV is taken up by P. vittata by phosphate transporters following Michaelis Menten kinetics, with P inhibiting AsV uptake in a directly competitive manner.14 16 Since AsIII is predominantly present as a neutral species in the environment, it is hypothesized that AsIII is taken up by plants via aquaglyceroporins,17 which have been shown to transport small uncharged solutes such as glycerol, silicic acid, and AsIII.18,19 Competitive inhibition study of AsIII uptake by glycerol (0.1 mM AsIII and 0.1 100 mM glycerol) in excised rice roots suggests that AsIII transport into rice is via aquaglyceroporins.18 A recent study10 where glycerol addition had no eﬀect on AsIII uptake by P. vittata indicate that the AsIII uptake system in P. vittata is diﬀerent from rice. Since AsIII is an analogue of silicic acid, it is understandable that silicic acid inhibits AsIII uptake by rice roots, consistent with the hypothesis that both silicic acid and AsIII are taken up by rice via aquaglyceroporins.20 However, addition of 0.5 mM silicic acid had no eﬀect on AsIII uptake by P. vittata roots at 15 μM AsIII during 1 day uptake study,21 again indicating that the AsIII uptake system by P. vittata is diﬀerent from that by rice. The fact that neither glycerol nor silicic acid inhibited AsIII uptake by P. vittata suggests that aquaglyceroporins may not be a major channel for AsIII transport into P. vittata roots. As a useful model to examine transporter-mediated uptake processes, a Michaelis Menten saturation curve can indicate either active uptake (against a concentration gradient by use of energy) or passive uptake (facilitated diﬀusion down a concentration gradient). Aquaglyceroporin-mediated transport of small neutral molecules such as AsIII, silicic acid, and glycerol across cellular membranes is via facilitated diﬀusion. When the solute is transported against a concentration gradient from external medium to living cells, the transport is active and energydependent. To further distinguish active transport from facilitated diﬀusion, a metabolic inhibitor, such as 2,4-dinitrophenol (uncoupler of oxidative phosphorylation), can be employed to inhibit active uptake by disabling cellular energy production.19 Previous study has shown the ability of P. vittata to accumulate AsV from contaminated drinking water.22,23 However, little is known about AsIII uptake kinetics by P. vittata and the underlying mechanisms. In this study, P. vittata were exposed to AsIII at low and high concentrations for 6 h in aerated hydroponic solutions. Our objectives were to (1) investigate the kinetics of AsIII uptake systems in P. vittata by use of a Michaelis Menten kinetic model, (2) determine arsenic speciation in root symplastic solution and xylem sap after AsIII exposure, and (3) determine the impact of metabolic inhibition on AsIII uptake in P. vittata. Results from the present study should provide important information on the mechanisms of AsIII uptake by P. vittata, which may serve as a model plant to better understand arsenic hyperaccumulation mechanisms in plants.

ARTICLE

’ MATERIALS AND METHODS Plant Growth. Healthy P. vittata plants with 4 5 fronds (4 months old) from Milestone Agriculture Inc. (Apopka, FL) were transferred to 0.2 strength Hoagland solution for acclimation. Plants were grown in a controlled room with 8-h photoperiod at light intensity of 350 μmol 3 m 2 3 s 1, 28/ 23 C day/night temperature, and 70% relative humidity. The nutrient solutions were aerated continuously and changed weekly. After 2 weeks of acclimation, ferns of similar size with sufficient new root tips were transferred to 0.1 mM CaCl2 for 12 h before the experiment. AsIII Uptake in Low and High Arsenite Concentration Ranges. Short-term AsIII uptake by P. vittata was tested with four replicates over both low (0 32.5 μM) and high (65 650 μM) AsIII concentration ranges. The ferns were transferred to 3.5 L of 0.1 mM CaCl2, which was spiked with 0, 2.6, 5.2, 7.8, 13, 32.5, 65, 130, 260, or 650 μM AsIII (as NaAsO2) (Sigma, St. Louis, MO). At the end of 6 h treatment under aeration, the growth medium was speciated by use of an As-speciation cartridge (Waters Corp., Milford, MA). Fern roots were thoroughly rinsed with deionized water, followed by 10 min of desorption in 1 mM K2HPO4. Metabolic Inhibitor Experiment. To determine the effects of metabolic inhibitor on AsIII uptake by P. vittata, 2 mM 2,4dinitrophenol (DNP) was employed. 2,4-DNP was dissolved in ethanol to make a stock solution before its addition to the uptake medium, with a final ethanol concentration of 0.3% (v/v). No effect at this concentration of ethanol on AsIII uptake was observed in our preliminary experiment. P. vittata plants were exposed to low and high AsIII concentrations (5.2, 65 and 650 μM AsIII) with and without 2 mM 2,4-DNP. After 6 h of treatment, root-cell symplastic solutions and xylem saps were extracted. Each treatment was repeated four times. Symplastic Solution and Xylem Sap Collection. Symplastic solution was extracted from the new root tips by centrifugation.19 Following 6 h of exposure to AsIII, plant roots were rinsed briefly with tap water and immersed in 1 mM K2HPO4 for ∼15 min to remove the adsorbed arsenic species. After that, ∼1.5 cm root tips were excised from P. vittata. After a quick wash with distilled water, root tips were blotted dry and placed into 0.22 μM filter unit (Ultrafreer-MC, Millipore Corp., Billerica, MA) with the cut ends facing down. By centrifugation at 2000g for 15 min at 4 C, apoplastic solutions were obtained and root segments were frozen at 80 C for 2 h, followed by thawing at room temperature for 10 min. After centrifugation at 2000g for 15 min at 4 C, symplastic solutions were collected from the frozen thawed root tips and stored at 80 C until analysis. To check the validity of the extraction method for both apoplastic and symplastic solution as well as the purity of each fraction, the activity of malic dehydrogenase (MDH) in the apoplastic and symplastic saps was determined in our preliminary experiment. The activity of MDH in the symplastic sap from root tips of P. vittata was ∼0.4 unit 3 mL 1 (1 unit of enzyme catalyzes the conversion of 1 μmol of substrate to product per minute). In contrast, the activity of MDH in the apoplastic solution was <2.5% that of the symplastic solution, indicating high purity of the symplastic solution from the method. For xylem sap, the roots were cut at ∼1 cm above the rhizomes and placed into a Scholander pressure chamber for 30 min to collect xylem sap ﬂowing from the rhizomes to the fronds. Samples were preserved in a 80 C freezer immediately. 9720

dx.doi.org/10.1021/es2018048 |Environ. Sci. Technol. 2011, 45, 9719–9725

Environmental Science & Technology

Figure 1. Concentrations of total As and AsIII in (a) uptake solution and (b) root biomass after exposure of P. vittata to 2.6 650 μM AsIII for 6 h. Total As = AsIII + AsV. Little organic arsenic species were detected in root biomass. Data are means and standard errors of four replicates.

Arsenic Analysis and Speciation. Arsenic concentration and speciation in the root symplastic solutions and xylem saps were determined by coupling high-performance liquid chromatography (HPLC) to inductively coupled plasma mass spectrometry (ICP-MS). Speciation of arsenic species (arsenite, arsenate, methylarsonic acid, and dimethylarsinic acid) was conducted on a Perkin-Elmer series 200 HPLC system hyphenated with a Perkin-Elmer Elan DRC-e ICP-MS for As detection. A Hamilton PRPX-200 cation-exchange column (250 4.1 mm in dimension and 10 μm particle size) was employed for the separation of arsenic species. The mobile phase was 0.1% formic acid with a flow rate of 1 mL/min. A nebulizer feed valve was used to direct the flow from HPLC into the ICP-MS, and an internal standard valve was used for postcolumn injection of internal standards. The m/z signal 75 was monitored for As. Data were collected and treated by use of Chromera software (Perkin-Elmer). The instrumental parameters were optimized against a tuning solution recommended by the manufacturer, with the nebulizer gas flow being optimized daily for maximum sensitivity. Quality assurance was obtained through the use of blanks, standard curves, standard check solutions and spiked samples, which were run during sample analysis. More detailed information for this cation-exchange HPLC-ICP-MS arsenic speciation method can be found elsewhere.24

ARTICLE

Figure 2. Concentrations of AsIII and AsV in (a) root symplastic solution and (b) xylem sap of P. vittata after exposure to 2.6 650 μM AsIII for 6 h. Total As = AsIII + AsV. Little organic arsenic species were detected in symplastic solution or xylem sap. Data are means and standard errors of four replicates.

For arsenic speciation in the root biomass, the bulk of root samples of P. vittata (stored at 80 C overnight) were cut into pieces, ground with liquid nitrogen in mortars, and extracted ultrasonically with a 1:1 methanol/water solution for 2 h.21 Prior experiment showed that arsenic species in organic forms in P. vittata biomass were negligible;25 therefore, only inorganic arsenic including AsIII and AsV was determined. AsV and AsIII in the plant extracts were separated by use of As-speciation cartridges (Waters Corp., Milford, MA). Total arsenic and AsIII concentrations in the extracts and uptake medium were determined by a graphite furnace atomic absorption spectrophotometer (GFAAS; AA240Z, Walnut Creek, CA). Preliminary experiment with control species of P. vittata showed that little arsenic was detected in root symplastic solution or xylem sap, and negligible arsenic was found in root biomass. Data Analysis. All values were expressed as means ( SE (n = 4). Two-parameter Michaelis Menten kinetic model was fitted to AsIII uptake data, and the kinetic parameters were calculated in SigmaPlot 11.0.

’ RESULTS Arsenic Speciation in the Growth Media, Root Symplastic Solutions, and Xylem Sap. During the 6 h of uptake, AsIII

accounted for 100% of total As in the growth medium (Figure 1a). Since a large volume of growth medium was used (3.5 L per plant), 9721

dx.doi.org/10.1021/es2018048 |Environ. Sci. Technol. 2011, 45, 9719–9725

Environmental Science & Technology

ARTICLE

AsIII Radial Transport from Media to the Root Symplast and DNP Inhibition. Arsenic concentrations in the root sym-

plastic solutions increased with increasing AsIII concentrations in the growth medium and were significantly greater (8 21fold) than those in the growth medium (Figure 3a). During the 6 h uptake, AsIII influx into the root symplast saturated gradually over external AsIII concentrations from 2.6 to 650 μM, and the uptake data can be well described by a Michaelis Menten model with an apparent Km of 77.7 μM and Vmax of 1.46 μmol of As 3 mL 1 3 (6 h) 1 (Table 1). Significantly higher arsenic concentrations in the root symplastic solutions than those in the uptake medium over the whole concentration range (Figure 3a) indicated that AsIII was transported from the external medium to the cortical cells against the external AsIII concentration gradient, which is an important characteristic of active uptake. To verify this, metabolic inhibitor 2,4-DNP was employed to investigate its eﬀect on AsIII transport into the roots of P. vittata. With 2 mM 2,4-DNP spiked into the growth medium containing 5.2, 65, or 650 μM AsIII, arsenic concentrations (with AsIII accounting for 90 94% in the root symplastic solutions) were decreased to levels lower than those in the growth medium, being only ∼4% of those in the controls without 2,4-DNP (Figure 4a). These results indicated that transporter-mediated active uptake played a dominant role in AsIII radial transport, with passive inﬂux being a minor component. Arsenic Loading to Xylem Sap via Dual Affinity Systems. With external AsIII concentrations at 2.6 32.5 μM and 65 650 μM, dual affinity systems exhibiting saturation kinetics were present for xylem loading of arsenic in P. vittata (Figure 3b,c). On the basis of the Michaelis Menten kinetic model, Km values of the high- and low-affinity systems were 8.78 μM and 70.4 μM, respectively (Table 1). In the presence of 2,4-DNP, arsenic loading to xylem sap was effectively inhibited with AsIII concentration being 13% of the controls at 5.2 μM AsIII (Figure 4b). However, the inhibitory effect of 2,4-DNP on arsenic loading to xylem sap was gradually decreased at high AsIII concentration range. As shown in Figure 4b, arsenic concentrations in xylem sap accounted for 27% and 46% of the corresponding controls at 65 μM and 650 μM, respectively, suggesting a more important role of passive diffusion in low-affinity loading system. Over the whole concentration range, arsenic concentrations in the xylem sap (dominated by AsIII) accounted for 10 12% of those in the root symplastic solutions (Figure 3).

Figure 3. Arsenic concentrations in (a) root symplastic solution (0 650 μM AsIII) and (b, c) xylem sap (0 32.5 and 65 650 μM AsIII) of P. vittata after exposure to AsIII for 6 h. The data were ﬁtted to the Michaelis Menten kinetic model. Data are means ( standard errors of four replicates.

AsIII concentrations in the medium changed little after 6 h of uptake (Figure 1a), which is an important prerequisite for AsIII uptake kinetic study. In the root biomass (including new root tips and mature roots), AsV was the predominant species, accounting for 82 87% of the extracted arsenic in all treatments (Figure 1b). By contrast, the symplastic solutions extracted from the new root tips and xylem saps were dominated by AsIII (90 94%) in all treatments (Figure 2). No methylated arsenic species were detected in root biomass, root symplastic solutions, or xylem saps (data not shown).

’ DISCUSSION Our data indicate that, at external AsIII of 2.5 650 μM, radial transport of AsIII from growth medium into P. vittata roots was mainly via an active process, with passive diﬀusion being a minor component. This is in contrast to AsIII inﬂux into rice roots, which is via aquaglyceroporin channels.20 For xylem loading of arsenic, both high- and low-aﬃnity systems were involved in P. vittata. Speciﬁcally, the role of passive diﬀusion in xylem loading becomes more important at higher external AsIII concentrations (65 650 μM) in the growth medium.

AsIII Dominated in the Growth Medium, Root Symplastic Solution, and Xylem Sap. During the 6 h uptake experiment,

little AsIII oxidation was observed in the growth medium, as only AsIII was detectable in the medium regardless of AsIII concentrations (Figure 1a), which is a critical prerequisite for the AsIII uptake kinetic study. 9722

dx.doi.org/10.1021/es2018048 |Environ. Sci. Technol. 2011, 45, 9719–9725

Environmental Science & Technology

ARTICLE

Table 1. Michaelis Menten Kinetic Parameters with External AsIII Supply to P. vittataa external AsIII concentration range (μM)

Km (μM)

Vmax [nmol of As 3 mL

1

1 3 (6 h) ]

R2

p

0.9746

<0.0001

Radial Transport from External Media into Root Symplastic Solution 0 650

77.7

1463

0 32.5

8.78

30.5

0.9978

<0.0001

65 650

70.4

164

0.9981

0.0009

Loading to Xylem Sap

Kinetic parameters were calculated by ﬁtting the inﬂux data in Figure 3 to a two-parameter hyperbola via nonlinear regression. Curve-ﬁtting was achieved by use of the ﬁtting regimes in SigmaPlot 11.0. a

Figure 4. Eﬀects of 2,4-DNP on arsenic concentrations in (a) root symplast and (b) xylem sap in P. vittata after exposure to 5.2, 65, and 650 μM AsIII for 6 h. Data are means and standard errors of four replicates.

In addition to the growth medium, AsIII was also the predominant species in the root symplastic solution, accounting for 90 94% of total arsenic (Figure 2a), which was documented for the ﬁrst time. The root symplastic solutions in the present study were extracted from the new root tips, which represent the actively transporting regions of the roots. Though no AsV was detected in the growth medium, a small portion of AsV (6 10%) was present in the root symplastic solution. This limited oxidation of AsIII to AsV may have occurred inside the roots. This is supported by data from Mathews et al.,26 who observed 35% AsV in P. vittata roots after exposure to 0.10 mM AsIII for 1 day under sterile conditions where no AsIII oxidation occurred in the growth medium. However, oxidation of

AsIII to AsV may have also occurred on the surface of roots, where microbes were present.26 Similar to root symplastic solution, AsIII also dominated in the xylem sap, accounting for 90 94% of the total As (Figure 2b), which is similar to that reported by Su et al.27 They observed that AsIII accounted for 93 98% of the total As in the xylem sap of P. vittata after exposure to 5 μM AsIII for 1 24 h. However, AsIII concentrations in the xylem sap in the present study were much less than those detected by Su et al.27 For example, the AsIII concentration in the xylem sap was 10.7 μM after exposure to 5.2 μM AsIII for 6 h in our study, compared to ∼220 μM after exposure to 5 μM AsIII for 8 h by Su et al.27 This diﬀerence can be attributed to diﬀerent AsIII concentrations in the root biomass, which seems to be the limiting factor for As translocation from the roots to the fronds, as AsIII is preferentially loaded into xylem sap rather than AsV.27 In their study,27 ∼75% of the total extractable As was present as AsIII in the roots, compared to 13 18% AsIII in this study (Figure 1b), which led to a signiﬁcantly reduced As concentrations (mainly as AsIII) in the xylem sap (Figure 3). The predominance of AsV in the root biomass (Figure 1b) may result from three processes: (1) microbial oxidation of AsIII to AsV on the root surface, especially mature root surface, before being taken up,21 (2) AsIII oxidation to AsV inside the roots,26 and (3) preferential translocation of AsIII from the roots to the fronds, leaving AsV in the roots.27 AsIII oxidation in the growth medium in the presence of P. vittata has been previously observed21 and is most likely microbially mediated.26 Since the AsIII oxidation most likely occurred on the root surface, the oxidized AsV was probably immediately taken up by P. vittata roots due to its high aﬃnity for AsV (Km value of 0.52 1.1 μM),14,16 making AsV undetectable in the uptake medium. AsIII Influx into P. vittata Roots Was a TransporterMediated Active Process. Kinetic data in this study (Table 1) showed that AsIII radial transport from external medium into the root symplastic solutions followed saturation kinetics and the uptake data can be well-fitted to a Michaelis Menten function. The significantly greater arsenic concentrations in the root symplastic solutions than in the external growth medium strongly suggested that AsIII transport into P. vittata roots was against concentration gradient, that is, active transport, which was further confirmed by the effective inhibition of AsIII influx into the roots by 2,4-DNP over the whole concentration range (Figure 4a). Suppressed AsIII uptake by metabolic inhibitor suggests that AsIII transport into the roots of P. vittata was dominated by the transporter-mediated active uptake, with the passive diffusion being a minor component, which is different from the AsIII uptake system in rice18,20 as suggested by a recent study.21 9723

dx.doi.org/10.1021/es2018048 |Environ. Sci. Technol. 2011, 45, 9719–9725

Environmental Science & Technology

ARTICLE

Table 2. Comparison of AsIII and AsV Uptake by the Roots of Rice and P. vittata external arsenic plant species excised rice roots

As species AsIII

uptake time

concn (μM) 50, 100, 200, 300,

experiment method concentration-dependent

400, and 500 intact P. vittata with

AsV

5.0

and medium 0.33 h, distilled water,

uptake

18

depletion technique

8 h, 0.5 mM CaCl2+ 5 mM

0.52

14

1.1

16

MES, pH 6.0 73

AsV

5 6 fronds intact P. vittata with 5 6 fronds

180

ref

pH 7.0

4 5 fronds intact P. vittata with

Km (μM)

0.1, 0.5, 1.5, 2.5,

73

AsV radioactive tracer

0.5 h, 0.1 mM CaCl2,

5.0, and 7.5 AsIII

2.6, 5.2, 7.8, 13, 32.5, 65, 130, 260, and 650

pH 6.5 concentration-dependent uptake

Compared to the Km of AsIII transporter in rice roots (180 μM)18 a 2.3 times lower Km value (77.7 μM) was observed for AsIII transporters in P. vittata roots (Table 1), indicating much higher aﬃnity of the AsIII transporters in P. vittata roots than those in rice. This is consistent with the extraordinary uptake ability of P. vittata for arsenic even in uncontaminated environments with low arsenic bioavailability.13 Compared to AsV uptake kinetics by P. vittata (Km = 0.52 1.1 μM; Table 2), the much higher Km value for AsIII uptake indicates higher aﬃnity of AsV transporter than AsIII transporter for its substrate during root uptake. This is consistent with the observations of Wang et al.,14 where AsIII uptake rate was 10% of that for AsV uptake in the absence of phosphate, although no kinetic data were provided for AsIII uptake in that study. For most nonhyperaccumulators, arsenic tolerance is generally achieved by reduced arsenic uptake with the absence of a highaﬃnity uptake system or the development of a low-aﬃnity system at higher arsenic concentrations, which are considered as important mechanisms for plants to survive arsenic toxicity in the environment.28,29 However, regarding P. vittata, a transportermediated active uptake with relatively low Km value was responsible for AsIII inﬂux into root symplast over a wide concentration range of external AsIII (Figure 3a), which makes an important contributing factor to the eﬃcient AsIII uptake by P. vittata. Considering AsIII concentrations in most contaminated water8,9 are within the concentration range in the present study (2.6 650 μM), eﬃcient root uptake of AsIII by P. vittata may provide an ecofriendly and cost-eﬀective cleanup technology for water treatment, especially for developing countries such as Bangladesh where extensive arsenic contamination occurs in drinking water. Dual-Affinity Systems Were Involved in Xylem Loading of Arsenic in P. vittata. As indicated by the kinetic data (Table 1), dual-affinity systems were involved in the xylem loading process in P. vittata (Figure 3b,c), with a high-affinity system being dominant at low AsIII concentrations (Km = 8.78 μM with 0 32.5 μM AsIII) and a low-affinity system operating at high substrate concentrations (Km = 70.4 μM with 65 650 μM AsIII). Furthermore, as indicated by the decreased effect of 2,4-DNP on arsenic concentrations in xylem sap (Figure 4b), the passive component played a more important role (27 46%) in xylem loading at higher external AsIII concentrations (65 650 μM). However, we are uncertain whether the passive loading of AsIII into xylem sap was mediated by aquaglyceroporins, which warrants further investigation at the molecular level.

6 h, 0.1 mM CaCl2, pH 6.0

77.7

this work

The unique AsIII transport system in P. vittata is not only important from a scientiﬁc standpoint to advance our understanding of AsIII hyperaccumulation mechanisms in P. vittata but also has potential implications to better understand AsIII uptake in other plants. By modifying the genes responsible for AsIII uptake system, genetic engineering may enhance arsenic uptake for phytoremediation application and hence mitigate arsenic contamination in waters and soils.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: lqma@uﬂ.edu (L.Q.M.), brath@uﬂ.edu (B.R.).

’ ACKNOWLEDGMENT This research was supported in part by the University of Florida IFAS Innovation Fund and a scholarship from the China Scholarship Council received by the senior author. We thank Ms. Rujira Tisarum and Dr. Shiny Mathews at University of Florida for help extracting xylem sap and carrying out arsenic analysis. We appreciate the helpful discussions with Dr. Jian Feng Ma at Okayama University. ’ REFERENCES (1) Williams, P. N.; Islam, S.; Islam, R.; Jahiruddin, M.; Adomako, E.; Soliaman, A. R. M.; Rahman, G. K. M. M.; Lu, Y.; Deacon, C.; Zhu, Y. G.; Meharg, A. A. Arsenic limits trace mineral nutrition in Bangladesh rice grain. Environ. Sci. Technol. 2009, 43, 8430–8436. (2) Oremland, R. S.; Stolz, J. F. The ecology of arsenic. Science 2003, 300, 939–944. (3) Zhao, F. J.; McGrath, S. P.; Meharg, A. A. Arsenic as a food chain contaminant: mechanisms of plant uptake and metabolism and mitigation strategies. Annu. Rev. Plant Biol. 2010, 61, 535–559. (4) Xu, X. Y.; McGrath, S. P.; Zhao, F. J. Rapid reduction of arsenate in the medium mediated by plant roots. New Phytol. 2007, 176, 590–599. (5) Yamamura, S.; Watanabe, M.; Kanzaki, M.; Soda, S; Ike, M. Removal of arsenic from contaminated soils by microbial reduction of arsenate and quinone. Environ. Sci. Technol. 2008, 42, 6154–6159. (6) Macur, R. E.; Wheeler, J. T.; McDermott, T. R.; Inskeep, W. P. Microbial populations associated with the reduction and enhanced mobilization of arsenic in mine tailings. Environ. Sci. Technol. 2001, 35, 3676–3682. (7) Cullen, W. R.; Reimer, K. J. Arsenic speciation in the environment. Chem. Rev. 1989, 89, 713–764. (8) Ahmed, K. M.; Bhattacharya., P; Hasan, M. A.; Akhter, S. H.; Alam, S. M. M.; Bhuyian, M. A. H.; Imam, M. B.; Khan, A. A.; Sracek, O. Arsenic enrichment in groundwater of the alluvial aquifers in Bangladesh: an overview. Appl. Geochem. 2004, 19, 181–200. 9724

dx.doi.org/10.1021/es2018048 |Environ. Sci. Technol. 2011, 45, 9719–9725

Environmental Science & Technology

ARTICLE

(9) Luu, T. T.; Sthiannopkao, S.; Kim, K. W. Arsenic and other trace elements contamination in groundwater and a risk assessment study for the residents in the Kandal Province of Cambodia. Environ. Int. 2009, 35, 455–460. (10) Eary, L. E.; Schramke, J. A. Rates of inorganic oxidation reactions involving dissolved oxygen. In Chemical Modeling of Aqueous Systems II; Melchior, D. C., Bassett, R. L., Eds.; American Chemical Society: Washington, DC, 1990; pp 379 396. (11) Ali, W.; Isayenkov, S. V.; Zhao, F. J.; Maathuis, F. J. M. Arsenite transport in plants. Cell. Mol. Life Sci. 2009, 66, 2329–2339. (12) Dodd, M. C.; Vu, N. D.; Ammann, A.; Le, V. C.; Kissner, R.; Pham, H. V.; Cao, T. H.; Berg, M.; von Gunten, U. Kinetics and mechanistic aspects of As(III) oxidation by aqueous chlorine, chloramines, and ozone: relevance to drinking water treatment. Environ. Sci. Technol. 2006, 40, 3285–3292. (13) Ma, L. Q.; Komar, K. M.; Tu, C.; Zhang, W.; Cai, Y. A fern that hyperaccumulates arsenic. Nature 2001, 409, 579. (14) Wang, J. R.; Zhao, F. J.; Meharg, A. A.; Raab, A.; Feldmann, J.; McGrath, S. P. Mechanisms of arsenic hyperaccumulation in Pteris vittata. Uptake kinetics, interactions with phosphate, and arsenic speciation. Plant Physiol. 2002, 130, 1552–1561. (15) Tu, S.; Ma, L. Q. Interactive eﬀects of pH, arsenic and phosphorus on uptake of As and P and growth of the arsenic hyperaccumulator Pteris vittata L. under hydroponic conditions. Environ. Exp. Bot. 2003, 50, 243–251. (16) Poynton, C. Y.; Huang, J. W.; Blaylock, M. J.; Kochian, L. V.; Elless, M. P. Mechanisms of arsenic hyperaccumulation in Pteris species: root As inﬂux and translocation. Planta 2004, 219, 1080–1088. (17) Zhao, F. J.; Ma, J. F.; Meharg, A. A.; McGrath, S. P. Arsenic uptake and metabolism in plants. New Phytol. 2009, 181, 777–794. (18) Meharg, A. A.; Jardine, L. Arsenite transport into paddy rice (Oryza sativa) roots. New Phytol. 2003, 157, 39–44. (19) Mitani, N.; Ma, J. F. Uptake system of silicon in diﬀerent plant species. J. Exp. Bot. 2005, 56, 1255–1261. (20) Ma, J. F.; Yamaji, N.; Mitani, N.; Xu, X. Y.; Su, Y. H.; McGrath, S. P.; Zhao, F. J. Transporters of arsenite in rice and their role in arsenic accumulation in rice grain. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 9931–9935. (21) Wang, X.; Ma, L. Q.; Rathinasabapathi, B.; Liu, Y. G.; Zeng, G. M. Uptake and translocation of arsenite and arsenate by Pteris vittata L.: Eﬀects of silicon, boron and mercury. Environ. Exp. Bot. 2010, 68, 222–229. (22) Tu, S.; Ma, L. Q.; Fayiga, A. O.; Zillioux, E. J. Phytoremediation of arsenic contaminated groundwater by an arsenic hyperaccumulating fern Pteris vittata L. Int. J. Phytorem. 2004, 6, 35–47. (23) Huang, J. K.; Poynton, C. Y.; Kochian, L. V.; Elless, M. P. Phytoﬁltration of arsenic from drinking water using arsenic-hyperaccumulating ferns. Environ. Sci. Technol. 2004, 38, 3412–3417. (24) Yehiayan., L.; Pattabiraman, M.; Kavallieratos, K.; Wang, X.; Boise, L. H.; Cai, Y. Speciation, formation, stability and analytical challenges of human arsenic metabolites. J. Anal. At. Spectrom. 2009, 24, 1397–1405. (25) Zhang, W.; Cai, Y.; Tu, C.; Ma., L. Q. Arsenic speciation and distribution in an arsenic hyperaccumulating plant. Sci. Total Environ. 2002, 300, 167–177. (26) Mathews, S.; Ma, L. Q.; Rathinasabapathi, B.; Natarajan, S.; Saha, U. K. Arsenic transformation in the growth media and biomass of hyperaccumulator Pteris vittata L. Bioresour. Technol. 2010, 101, 8024–8030. (27) Su, Y. H.; McGrath, S. P.; Zhu, Y. G.; Zhao, F. J. Highly eﬃcient xylem transport of arsenite in the arsenic hyperaccumulator Pteris vittata. New Phytol. 2008, 180, 434–441. (28) Meharg, A. A.; Macnair, M. R. An altered phosphate uptake system in arsenate-tolerant Holcus lanatus L. New Phytol. 1990, 116, 29–35. (29) Meharg, A. A.; Macnair, M. R. The mechanisms of arsenate tolerance in Deschampsia cespitosa (L.) Beauv. and Agrostis capillaris L. New Phytol. 1991, 119, 291–297. 9725

dx.doi.org/10.1021/es2018048 |Environ. Sci. Technol. 2011, 45, 9719–9725

ARTICLE pubs.acs.org/est

Electrochemical Carbon Nanotube Filter Oxidative Performance as a Function of Surface Chemistry Guandao Gao†,‡ and Chad D. Vecitis*,† † ‡

School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, United States Key Laboratory of Pollution Processes and Environmental Criteria (Ministry of Education), College of Environmental Science and Engineering, Nankai University, Tianjin 300071, China

bS Supporting Information ABSTRACT: An electrochemical carbon nanotube ﬁlter has been reported to be eﬀective for the removal and electrooxidation of aqueous chemicals and microorganisms. Here, we investigate how carbon nanotube (CNT) chemical surface treatments including calcination to remove amorphous carbon, acid treatment to remove internal residual metal oxide, formation of surﬁcial oxyfunctional groups, and addition of Sb-doped SnO2 particles aﬀect the electrooxidative ﬁlter performance. The various CNT samples are characterized by scanning electron microscopy (SEM), thermogravimetric analysis (TGA), and X-ray photoelectron spectroscopy (XPS) and electrochemically evaluated by cyclic voltammetry, open circuit potential versus time analysis, and electrochemical impedance spectroscopy. Voltammetry results indicate that the near CNT surface pH is at least two units lower than the bulk pH. The electrooxidative performance of the various CNT samples is evaluated with 1 mM of methyl orange (MO) in 100 mM sodium sulfate at a ﬂow rate of 1.5 mL min 1. At both 2 and 3 V, the eﬃcacy of electrochemical ﬁltration is observed to be function of CNT surface chemistry. The samples with the greatest electrooxidation were the calcinated then HCl-treated CNTs, i.e., the CNTs with the most surﬁcial sp2-bonded carbon, and the Sb SnO2-coated CNTs, i.e., the CNTs with the most electrocatalytic surface area. At 3 V applied voltage, these CNT samples are able to oxidize 95% of the inﬂuent MO within the liquid residence time of <1.2 s. The broader applicability of electrochemical ﬁltration is evaluated by challenging the C CNT HCl and C CNT HNO3 networks with various organics including methylene blue, phenol, methanol, and formaldehyde. At 3 V applied voltage, both CNTs are able to degrade a fraction of all the organics with the extent organic degradation dependent on both CNT and organic properties. The C CNT HCl network generally had the better oxidative performance than the C CNT HNO3 network with an exception being the positively charged methylene blue. The extent of MO degradation, steady-state current, anode potential, eﬄuent pH, and back pressure are also measured as a function of applied voltage (1 3 V) and CNT surface chemistry. Mass spectrometry of electrochemical CNT ﬁlter eﬄuent at 2 and 3 V is utilized to evaluate plausible electrooxidation products. Energy consumption as compared to state-of-the-art electrodes and strategies to tailor the CNT surface for a speciﬁc target molecule are discussed.

’ INTRODUCTION Electrochemistry, the interrelation of electrical and chemical eﬀects, has found utility for a wide-range of applications including electrophoretic separations, corrosion control, electroanalytical sensors, electroplating, batteries, and fuel cells.1 There are also opportunities for environmental applications of electrochemistry including wastewater treatment, metal recycling, and environmental sensing.2 In regards to electrochemical water treatment, the development of anodes with optimal geometries, high electrocatalytic activity, and extended operational lifetimes has resulted in electrooxidation eﬃciencies that are energetically comparable to conventional wastewater treatment3 and disinfection4 technologies. An electrochemical waste-to-energy process involving simultaneous anodic wastewater treatment and cathodic hydrogen production may lead to greater energy eﬃciencies.5 Recent research into electrooxidation for water treatment has focused on the design of novel anode materials and structures r 2011 American Chemical Society

based on boron-doped diamond (BDD),6 8 Sb-doped SnO2,9,10 and Bi-doped TiO2.5,11 These anode materials are exemplary due to a combination of properties including high O2 overpotential, corrosion stability, conductivity, and surface-bound hydroxyl radical yield. For example, BDD anodes are superior to platinum and glassy carbon toward phenol and formate oxidation6 and are able to mineralize atrazine.12 Novel three-dimensional anode nanoarchitechtures have also resulted in increases in electrooxidation. For example, the electrooxidative performance of BDD toward methanol has been improved by addition of a porous, three-dimensional (3D) platinum structure perpendicular to the BDD surface.7 BDD anode electrooxidative performance Received: July 1, 2011 Accepted: October 3, 2011 Revised: September 28, 2011 Published: October 03, 2011 9726

dx.doi.org/10.1021/es202271z | Environ. Sci. Technol. 2011, 45, 9726–9734

Environmental Science & Technology enhancements toward 2,4-dichlorophenoxyacetate oxidation and mineralization are observed when the BDD is coated with Sbdoped SnO2 nanoparticles.8 In both cases, the enhancements arise from the high surface area Pt/Sb SnO2 that increases the number of electrocatalytic surface sites in combination with the strongly oxidizing BDD that acts to limit Pt/Sb SnO2 passivating products. Another plausible material for novel 3D electrode nanoarchitechtures is carbon nanotubes (CNTs). Due to their physical,13 electrical,14 mechanical,15 and electrochemical16 properties, CNTs can be easily formed into stable, porous, and electrochemically active networks17 or ﬁlters.18 Recent studies have shown that CNT networks can be utilized as electrochemical ﬁlters that are eﬀective for the adsorptive removal and electrooxidation of aqueous dyes and anions19 and for the removal and inactivation of bacteria and viruses.20 The oxidative capacity of the electrochemical CNT ﬁlter was observed to be limited by the reactive CNT surface sites.19 Thus, there may be opportunities to enhance the anodic CNT ﬁlter electrooxidative performance by simply modifying the CNT surface chemistry. For example, CNT surface chemistry has been previously observed to aﬀect chemical adsorption,21 colloidal properties,22 antimicrobial properties,23 catalyst support performance,24 and photocatalytic nanocomposite performance.25 Here, we utilize a number of treatment methods to generate a set of multiwalled carbon nanotubes with varying surface chemistry including CNT (raw), C CNT (400 C for 1 h), CNT HCl (HCl at 70 C for 12 h), CNT HNO3 (HNO3 at 70 C for 12 h), C CNT HCl, C CNT HNO3, and C CNT SS (C CNT HNO3 coated with Sb-doped SnO2 particles). The materials are characterized by scanning electron microscopy (SEM), X-ray photoelectron spectroscopy (XPS), and thermogravimetric analysis (TGA). The electrooxidative performance of the CNTs is evaluated over a range of applied voltages by challenging the electrochemical ﬁlter with 1 mM methyl orange (MO) in 100 mM sodium sulfate (Na2SO4), and changes in surface chemistry are observed to signiﬁcantly aﬀect electrooxidative performance. Electrochemical parameters such as steadystate current, anode potential, and impedance as well as eﬄuent pH, back pressure, and product mass spectrum are also determined to relate CNT surface chemistry to electrochemical results.

’ MATERIALS AND METHODS Chemicals. Methyl orange (MO), hydrochloric acid (HCl; 36.5 38.0%), nitric acid (HNO3; 69.8%), sulfuric acid (H2SO4; 95.0 98.0%), phosphoric acid (H3PO4; g85.0%), tin chloride pentahydrate (SnCl4(H2O)5), antimony chloride (SbCl3), sodium hydroxide (NaOH), ethyl alcohol (EtOH; g95.0%), dimethyl sulfoxide (DMSO; g99.9%), potassium hydrogen phthalate (KHP), sodium sulfate (Na2SO4), sodium persulfate (Na2S2O8), sodium bicarbonate (NaHCO3), methylene blue (MB), cetylammonium bromide (CTAB), phenol (PhOH), methanol (MeOH), formaldehyde, formate, and sodium carbonate (Na2CO3) were purchased from Sigma-Aldrich. All chemicals were reagent grade except the DMSO that was spectrophotometric grade. CNT Selection. The multiwalled carbon nanotubes were purchased from NanoTechLabs, Inc. (Yadkinville, NC). The CNTs were characterized previously in detail26 to have a diameter distribution of 17 ( 9 nm and a length distribution of 91 ( 21 μm, in

ARTICLE

agreement with the manufacturer specifications of 5 7 walls, Ædæ = 15 nm, and Ælæ = 100 μm. CNT Calcination. To remove any amorphous or other carbon impurities,27 1 g of as-received CNTs was first calcinated in a tube furnace (Thermolyne, 21100) by increasing from room temperature to 400 C at a rate of 5 C/min and holding for 60 min at 400 C. If multiple CNT treatment steps were used, calcination was always completed first and the sample is given the C prefix. CNT Acid Treatment. Two types of acid treatment were completed depending on whether the goal was only to remove any residual metal catalyst impurities (concd HCl)27 or if the oxidative formation of surface carbonyl, hydroxyl, and carboxyl groups was also desired (concd HNO3).27,28 Both acid treatments were completed as follows: 0.5 g of CNT was placed into 0.5 L of acid and heated to 70 C in a round-bottom flask with stirring and a condenser for at least 12 h. After heating, the sample was cooled to room temperature and vacuum-filtered through a 5 μm PTFE membrane (Omnipore, Millipore) to collect the CNTs. The CNTs were then washed with Milli-Q deionized water (DI) until the filter effluent pH was near DI’s pH. The sample was then oven-dried at 100 C before use. Materials treated with HCl are labeled with the HCl suffix, and materials treated with HNO3 are labeled with the HNO3 suffix unless coated with metal oxide nanoparticles, see the following section. Sb-Doped SnO2 Particle Coatings. The Sb-doped SnO2 CNT were prepared by the hydrothermal method.29,30 Briefly, 50 mg of C CNT HNO3 was added to 30 mL of ethanol and 15 mL of DI water and dispersed by ultrasonication (Branson, Sonifier S450) for 5 min at an applied power of 400 W/L. Then, 27 mg of NaOH was added to the stirred mixture. Once dissolved, 117 mg of tin chloride pentahydrate (SnCl4(H2O)5) and 7.6 mg of antimony chloride (SbCl3) were slowly added to the stirred mixture. The final solution was then transferred to a Teflon-lined stainless steel autoclave and heated to 160 C for at least 12 h. CNTs prepared by this method are labeled C CNT SS. Electrochemical CNT Filter Preparation. The CNT filters were produced by first dispersing the CNTs in DMSO at 0.5 mg/mL and probe-sonicating (Branson, Sonifier S450) for 15 min at an applied power of 400 W/L. Then, 30 mL of the sonicated CNTs in DMSO was vacuum-filtered onto a 5 μm PTFE membrane (Millipore, Omnipore, JMWP), resulting in filter loadings of 1.5 1.6 mg/cm2. The CNT filters were washed with 100 mL of EtOH, 100 mL of 1:1 DI-H2O/EtOH, and 250 mL of DI-H2O to remove DMSO. Finally, the prepared filter was loaded into a filtration casing modified for electrochemistry, as described in our previous studies, Supporting Information Figure S1.19,20 Solution and Electrochemistry. Sodium sulfate (Na2SO4; 100 mM) was utilized as the background electrolyte, and methyl orange (MO; 1 mM) was used as the model pollutant unless otherwise noted. The influent MO electrolyte solution was peristaltically pumped (Masterflex) through the electrochemical CNT filter, Supporting Information Figure S1, and the electrochemistry was driven by a dc power supply (Agilent). Perforated stainless steel was used as the cathode, and an insulating silicone rubber O-ring separated the electrodes. The electrochemically active elements were incorporated into a modified polycarbonate 47 mm filter casing (Whatman). Before every experiment, the titanium ring was polished with sandpaper to optimize the electrical connectivity between the titanium and the CNTs. Bulk electrochemical ﬁltration of MO was completed at a number of rationally selected applied voltages, 2 and 3 V for comparison 9727

dx.doi.org/10.1021/es202271z |Environ. Sci. Technol. 2011, 45, 9726–9734

Environmental Science & Technology

ARTICLE

Table 1. Representative Carbon Nanotube Network Properties CNT type

weight (mg)

depth (μm)a

pore Ædæ (nm)

residual mass (%)b

low-T mass (%)c

O/C (%)d

ΔMO

2V (%)

CNT

18.8

62

98.2

4.03

0.49

1.88

30

C CNT

13.8

56

104.3

3.96

0.07

2.30

36

CNT HCl

16.1

57

114.9

3.94

0.19

2.53

54

CNT HNO3

15.1

58

118.3

2.64

1.19

4.67

24

C CNT HCl

14.8

65

109.5

2.36

0.11

2.24

72

C CNT HNO3

14.6

59

112.2

C CNT SS

38.5

82

136.1

2.16 52.8

0.76

4.08

28

3.02

12.6d

64

a Measured by microcaliper. b Measured by TGA and representative of percent metal catalyst impurity. c Measured by TGA. Mass loss between 150 and 400 C representative of amC or functionalization. d Measured by XPS; no surﬁcial Fe was detected in any of the samples; Sb and Sn detected in C CNT SS.

to previous reports,19,20 before the cyclic voltammetry (CV) peak, at the CV peak, and after the CV peak. The MO electrochemical ﬁltration experiments were completed for at least 30 min with at least three eﬄuent samples analyzed over this time period to ensure steady-state electrochemical ﬁltration was achieved. A number of parameters including eﬄuent pH (Corning 345), eﬄuent MO concentration (Agilent 8453 spectrophotometer; λmax = 464 nm; ε = 26 900 M 1 cm 1), steady-state current, anodic potential, and back pressure were all recorded. Bulk electrochemical ﬁltration at 3 V was also completed for methylene blue, phenol, CTAB, methanol, formaldehyde, and formate. Sodium sulfate at 100 mM was used as an electrolyte, and the inﬂuent concentration for all compounds was 1 mM with the exception of CTAB at 0.1 mM. Similar to MO, the electrochemical ﬁltration experiments were completed for at least 30 min with at least three eﬄuent samples analyzed over this time period to ensure steady-state electrochemical ﬁltration was achieved. The MB concentration was measured by spectrophotometer (λmax = 665 nm; ε = 74 100 M 1 cm 1). The phenol concentration was measured by UV vis and total organic carbon (TOC). The extent of electrochemical transformation of CTAB, methanol, formaldehyde, and formate to carbon dioxide was measured by eﬄuent total inorganic carbon TIC (inﬂuent TIC was ∼0). The CNT samples were also evaluated using a potentiostat (CHI604D) with the prepared sample as the working electrode, a stainless steel cathode as the counter electrode, and 1 M Ag/AgCl as the reference electrode in a ﬂow cell conﬁguration. Cyclic voltammetry, linear sweep voltammetry (LSV), and alternative current impedance (ACI) methods were used to electrochemically characterize the samples. SEM Analysis. Scanning electron microscopy was completed in Harvard’s Center for Nanoscale Systems on a Zeiss FESEM Supra55VP. Micrographs were analyzed with ImageJ software to determine aerial pore size that was an average of at least 100 measurements. TGA Analysis. Thermogravimetric analysis was completed in Harvard’s Material Research Science and Engineering Center on a Q5000-IR thermogravimetric analyzer (TA Instruments). Samples were heated from room temperature to 150 at 10 C min 1, held at this temperature for 30 min, then heated to 1000 at 10 C min 1 and held at this temperature for 30 min. A second run was completed immediately after the first and used as a background. The percent amorphous carbon and low-T combustibles was determined as the fraction burned between 150 and 400 C. The percent residual catalyst was determined using the initial mass and mass remaining after a complete thermal cycle.

XPS Analysis. X-ray photoelectron spectroscopy was completed on an ESCA SSX-100 in Harvard’s Center for Nanoscale Systems. For all samples, survey spectrum (0 1000 eV), C-1s (274 294 eV), O-1s (522 542 eV), and Fe-2p3 (700 720 eV) scans were completed. For the C CNT SS, Sn-3d5 (476 496 eV) and Sb-3d5 (520 540 eV) were also completed. Data was analyzed using CasaXPS. Mass Spectrometry Analysis. The influent and effluent MO samples oxidized at applied voltages of 2 and 3 V for all CNTs were analyzed by direct injection electrospray time-of-flight mass spectrometry (ESI-TOF-MS; Waters LCT Premier XE). The instrument was operated in negative-ion high-resolution mode (W ) with a capillary voltage of 3.5 kV. Every sample was continuously injected for at least 5 min at a flow rate of 10 μL min 1. TOC and TIC Analyses. Both TOC and TIC analyses were completed with a TOC analyzer (TOC-V; Shimadzu) with thermal persulfate oxidation. TOC measurements were used to analyze the extent of phenol removal. TIC measurements were used to analyze the electrochemical formation of carbon dioxide.

’ RESULTS AND DISCUSSION Physical Properties of Electrochemical CNT Filters. The physical properties of the CNT networks were similar for all the filter samples except the C CNT SS, Table 1. The CNT network mass was in the range of 14 19 mg, and the C CNT SS mass was 35 40 mg. The CNT network depth was in the range of 55 65 μm, and the C CNT SS depth was 80 85 μm. The Sb SnO2 particles were coated onto the CNTs prior to film formation; thus, these additional particles hindered the CNTs from forming a tightly packed network during vacuum filtration. Figure 1 and Supporting Information Figure S2 present SEM images of the CNT networks. The images look similar for all the samples except C CNT SS, Figure 1C, where micrometersized metal oxide particles are embedded in the CNT network supporting the conclusion that particle addition resulted in a more loosely packed CNT network. The pore size, as measured from SEM images, was also affected by the particle addition. The CNT networks had pore diameters in the range of 100 120 nm, whereas the C CNT SS had an average pore diameter of 136 nm. In all cases, the pore diameter was quite heterogeneous with a standard deviation of (50 60 and a range from 25 to 350 nm. Surface Chemistry of CNT Filters. In contrast to the CNT physical properties, the CNT surface chemistry was greatly affected by the treatments (Table 1, Figure 1D, and Supporting 9728

dx.doi.org/10.1021/es202271z |Environ. Sci. Technol. 2011, 45, 9726–9734

Environmental Science & Technology

ARTICLE

Figure 1. SEM images of representative CNT networks: (A) aerial image of C CNT HCl, (B) cross section of raw CNT network, and (C) aerial image of C CNT SS.

Information Figures S3 and S4). The residual mass, as determined by TGA, is a measure of the residual iron oxide catalyst (manufacturer specifications).31,32 In all cases, there was negligible Fe detected by XPS,33 indicating the Fe2O3 is attached to the inner CNT surface and thus beyond the X-ray analytical depth. The raw CNT samples had a residual mass around 4%, in agreement with manufacturer specifications. The C CNT and CNT HCl samples had similar, ∼4%, residual mass content. However, the C CNT HCl sample had a reduced residual mass, ∼2.4%, suggesting that amorphous carbon may have been blocking HCl from entering the ends of the CNT HCl tubes. For comparison, the CNT HNO3 sample also had a reduced residual content, ∼2.6%, likely due to the ability of HNO3 (NO2+) to oxidize the amorphous carbon. The sample with the lowest residual content was C CNT HNO3 at ∼2.2%, and the sample with the highest residual content was C CNT SS at >50% due to addition of noncombustible metal oxide. The mass loss over the temperature range of 150 400 C during thermogravimetric analysis gives insight into the amorphous and other non-sp2-bonded carbon content of the CNTs.32 The raw CNT sample had ∼0.5% mass loss over this range. The C CNT, CNT HCl, and C CNT HCl had reduced mass loss, 0.05 0.2%, over this range indicating these treatments reduced the amorphous carbon content. The CNT HNO3 and C CNT HNO3 had increased mass loss over this range, 0.8 1.2%, due to oxidative formation of easily combusted surface oxy-groups.34 The increased mass loss in the C CNT SS sample may be due to metal oxide catalyzed CNT oxidation. The CNT surface O/C ratios as determined by XPS are in agreement with the mass loss data. The raw CNT had an O/C ratio of 1.9%, which was only increased slightly in the C CNT, CNT HCl, and C CNT HCl samples to 2.2 2.5%. The CNT HNO3 and C CNT HNO3 samples had O/C ratios of ∼4.7% and ∼4.1% indicating signiﬁcant formation of carbonyl, hydroxyl, and carboxy groups on the CNT surface.22,28 The C CNT SS sample had an even greater O/C ratio due to addition of metal oxide particles. Signiﬁcant amounts of antimony and tin were also detected on the surface of the C CNT SS sample. The eﬀect of the CNT treatments utilized here is in agreement with previous studies22,31,32,34,35 and is well-described in the work of Pan and Xing.21 The three main surface features aﬀected by the treatments are amorphous carbon, internal Fe2O3 nanoparticles, and surface oxy-groups that are represented by thin gray surface coating, rust-colored internal spheres, and hydroxy and

Figure 2. Representative electrochemical CNT ﬁlter characterization. All ﬁgures were generated using the C CNT sample, 1 mM MO, 100 mM Na2SO4, and a ﬂow rate of 1.5 mL min 1. (A) cyclic voltammogram at completed at a scan rate of 10 mV s 1, (B) linear sweep voltammograms for all CNT ﬁlter samples, and (C) anodic and cathodic open circuit potential (V) over a range of applied voltages from 0 to 3 V.

carboxy groups, respectively (Figure 1D). The CNT network depictions are placed in order of increasing electrooxidative performance. The best performing CNT is the one that minimizes all of these features as will be discussed later. Electrochemical CNT Filter Characterization. A representative cyclic voltammogram, linear sweep voltammogram and, open circuit potential versus time for the C CNT sample are displayed in Figure 2. Identical measurements were completed for all CNT samples under influent conditions of 1 mM methyl orange, 100 mM Na2SO4, and flow rate of 1.5 mL min 1, Supporting Information Figure S5. The cyclic voltammogram in Figure 2A has two primary features. The 9729

dx.doi.org/10.1021/es202271z |Environ. Sci. Technol. 2011, 45, 9726–9734

Environmental Science & Technology first feature is the irreversible oxidation peak of methyl orange around 0.8 V versus Ag/AgCl, and the second feature is water oxidation (2H2O + 4h+ f O2 + 4H+) around 1.2 V versus Ag/ AgCl. The electrochemical irreversibility of the anodic CNT filter will be amplified over conventional bipolar electrodes since the electrooxidation products may be permanently carried away from the anode surface by the incident fluid flow. The MO oxidation potential suggests that the nearsurface pH of the CNT anode is ∼3,36 significantly lower than the influent pH ∼ 6. In the majority of the linear sweep voltammograms in Figure 2B, there is no distinct oxidation peak. Assuming a current similar to the C CNT peak is representative of peak potential, in all cases the near-surface pH will be <4. A decreased anode surface pH is also supported by the reversible Fe2O3 redox cycle at E1/2 ∼ 0.25 V in Figure 2A (black dashes). Extrapolation of pH-dependent CV of α-Fe2O3 nanoparticles37 indicates a near-surface pH of 3 4 supporting the MO data. Thus, the pH near the hydrophobic CNT interface may be significantly lower than that of the bulk solution, in agreement with recent results indicating an increased proton activity near the air water interface.38,39 The open circuit potential of the both the cathode and the anode as a function of applied voltage and time is displayed in Figure 2C. At applied voltages of both 0.5 and 1.0 V, the cathodic potential dominates over the anodic potential, in agreement with the observation that a negligible amount of MO is oxidized under these conditions. At 1.0 V applied voltage, the cathode potential is around 0.8 V, near the two-electron reduction potential of water to hydrogen (2H2O + 2e f H2 + 2OH ; E0 = 0.83 V).1 Further increases in applied voltage to 1.5, 2.0, 2.5, and 3.0 V results in greater increases to the anode potential as compared to cathode potential. Immediately after each increase in applied voltage up to 2.0 V, an exponential decay in anode potential is observed indicating the formation of a capacitive double layer and adsorption MO pseudocapacitance.40,41 It is promising that a low potential is required for the anodic oxidation of MO. However, it is dissatisfying that such a high fraction of potential, at least 50% at all voltages, e.g., 1.5 V at an 3.0 V applied voltage, is put toward the cathode when proton (E0 = 0.83 V), oxygen (O2 + e f O2• ; EpH7 = 0.33 V),42 and water reduction (2H+ + 2e f H2; E0 = 0 V) should occur at signiﬁcantly lower potentials. The energy put toward the cathode may be due to the disparity in surface area between the cathode and anode. The cathode has a surface area of at most 15 cm2 and a current density of 0.2 2.0 mA cm 2. The anode has an approximately 5000 cm2 of surface area19 and a current density of 0.006 0.0006 mA cm 2. The massive diﬀerence between cathode and anode current density suggests that an increase in cathode surface area may increase the extent of electrooxidation. Electrooxidative Performance of CNT Filters. The electrooxidative performance of the CNTs was evaluated under conditions of 1 mM MO, 100 mM Na2SO4, and J = 1.5 mL min 1 and is displayed in Figure 3, parts A and B, for applied voltages of 2 and 3 V, respectively. The current (mA; blue bars), MO degradation (%; red bars), and electrochemical impedance (ohm; gray bars; x-axis arc length in Supporting Information Figure S5C), are plotted versus CNT sample in order of their increasing electrooxidative performance: CNT HNO3 < C CNT HNO3 ∼ CNT < C CNT < C CNT HCl ∼ C CNT SS. The performance order is similar at 2 and 3 V with the range of percent MO degradation being greater at 2 V (24 72%) as compared to 3 V (66 95%). It is of note that the liquid residence

ARTICLE

Figure 3. Electrochemical CNT ﬁlter batch oxidative performance. Electrochemical ﬁlters were challenged with 1 mM MO in 100 mM Na2SO4 electrolyte at a ﬂow rate of 1.5 mL min 1. Steady-state current (mA; blue bars), MO degradation (%; red bars), and electrochemical impedance (ohm; gray bars) are plotted in rough order of increasing performance at an applied potential of (A) 2 V and (B) 3 V. (C) Oxidative performance at 3 V of the C CNT HCl (blue) and C CNT HNO3 (red) networks toward MO and MB decolorization, phenol TOC removal, and CTAB, methanol, formaldehyde, and formate conversion to carbon dioxide.

time in the electrochemical CNT filter is e1.2 s, and for C CNT HCl and C CNT SS ∼95% of the MO is oxidized. The steady-state current also roughly follows the MO degradation trend suggesting that most of the anodic current is toward MO oxidation. Both CNT HNO3 and C CNT HNO3 have slightly greater currents than expected indicating there is another anodic 9730

dx.doi.org/10.1021/es202271z |Environ. Sci. Technol. 2011, 45, 9726–9734

Environmental Science & Technology process contributing to the current such as corrosive elimination of the oxy-functional groups or increased O2 production. In contrast, the C CNT SS has a lower current than expected, due to SnO2’s high O2 overpotential of 2.2 V versus SCE.43,44 In terms of electrooxidation, the sample with the most surface sp2bonded carbon, C CNT HCl, and the sample modified with an electrocatalyst, C CNT SS, had the greatest efficacy. In contrast to current, the electrochemical impedance seems to have minimal correlation to electrooxidative performance. However, the measurement does yield insight into how the various CNT impurities and functionalizations aﬀect resistance to electron transfer. At both 2 and 3 V, the raw CNT has the greatest impedance, and the CNT treatment methods reduce the impedance following the order C < HNO3 < HCl indicating that interfacial amorphous carbon, oxy-functional groups, and metal oxides all act to impede electron transfer with amorphous carbon giving the least impedance. The greater improvement with HCl as compared to HNO3 treatment indicates that the oxidatively formed oxy-functional groups and defects also impede electron transfer likely by reducing the conductivity of the CNT network.14,45 The combined treatments used in C CNT HNO3 and C CNT HCl result in impedance reductions greater than the sum of the individual treatments. Since calcination was completed ﬁrst, the result suggests that the ends of the raw CNTs are not initially open, e.g., covered by amorphous carbon, and the acid cannot get inside the tubes to dissolve the internal Fe2O3. Thus, the CNT surface chemistry greatly aﬀects the electrochemical resistance toward interfacial charge transfer reactions; however, this resistance does not seem to have a great eﬀect on electrooxidative performance. The high sensitivity of electrochemical impedance toward CNT surface chemistry suggests it as a possibility for CNT surface analysis. The eﬀects of surface chemistry on electrooxidative performance may not be solely due to the electrochemical eﬀects since the adsorption of MO, which is a necessary ﬁrst step for direct electrooxidation, will also be aﬀected by surface chemistry.21,28 Previous studies have reported that aromatic molecules similar to methyl orange will strongly sorb to the extended sp2-CNT surface,46,47 and a CNT network with the greatest electrooxidation, C CNT HCl, has the greatest percentage of surﬁcial sp2bonded carbon. To further examine the importance of adsorption toward electrooxidative performance, the C CNT HCl and C CNT HNO3 networks were challenged at 3 V with a number of organics: methyl orange (negative aromatic), methylene blue (positive aromatic), phenol (neutral aromatic), CTAB (long-chain aliphatic), and methanol, formaldehyde, and formate (small, polar molecules). The percent degradation in terms of decolorization (MO and MB), TOC removal (phenol), and TIC formation, i.e., conversion to carbon dioxide (CTAB, MeOH, formaldehyde, formate) is presented in Figure 3C. Both networks were able to degrade some fraction of all of the organics, and the C CNT HCl network performed better in most cases with the exceptions being the positively charged MB and formate. It is of note that, although 10 20% of the phenol TOC is removed during electrochemical ﬁltration, there was no increase in eﬄuent TIC suggesting an electropolymerization mechanism is active.48 The complete electrochemical conversion of the small, polar organics to TIC and thus carbon dioxide is due to the relatively low number of electron transfers required for mineralization. For example, a 28-electron transfer is required to completely oxidize phenol, whereas a six-electron transfer is required to completely oxidize methanol. The minimal oxidation

ARTICLE

of formate is due to its high one-electron reduction potential, E0 ∼ 1.9 V.49 The varied extent of oxidation toward the target molecules indicates there are a number of factors aﬀecting electrochemical ﬁltration performance including CNT surface chemistry, molecule CNT interactions, and molecule oxidation potential. The two CNT networks evaluated are observed to perform best toward the molecules they adsorb the strongest. The oxidative performance of the C CNT HCl network toward MO oxidation is due to positive-charging of the CNT anode resulting in enhanced adsorption of the negatively charged MO.19 The strong performance of the C CNT HNO3 network toward MB oxidation is due to the electrostatic attraction between the negatively charged surface oxy-groups and the positively charged MB.28 A strongly adsorbed molecule will have a longer eﬀective retention time in the electrochemical CNT ﬁlter resulting in a greater number of chances to be electrochemically oxidized. Electrochemical and Effluent Characteristics versus Surface Chemistry and Voltage. As further evidence of the surface chemistry dependent electrochemical CNT filter performance, the extent of MO degradation (1 [MO]ef/[MO]in), steadystate current (mA), anode potential (V), effluent pH, and back pressure (kPa) are plotted as a function of applied voltage (V) and CNT in Figure 4. For most samples, the extent of MO oxidation, steady-state current, and anode potential increased monotonically with increasing applied voltage. Over the applied voltage range of 2.0 to 3.0 V, the CNT surface chemistry is observed to strongly affect both the extent of MO oxidation and the steady-state current. For the three best samples, CNT HCl, C CNT HCl, and C CNT SS, the extent of MO oxidation reached a plateau around 2.2 V. However, the steady-state current, which is also highest for these three materials, continues to increase indicating either an increase of anodic O2 production or a greater degree of molecular MO oxidation, i.e., a greater number of oxidized electrons per MO molecule. The anode potential is similar over the range of evaluated voltages for all CNT materials except the CNT and C CNT samples that are higher by 0.1 0.2 V. CNT and C CNT were the materials with the greatest electrochemical impedance suggesting there is some threshold value below which the impedance no longer affects the electrooxidation kinetics, in agreement with the lack of correlation between impedance and extent of electrooxidation. The eﬄuent pH is a strong function of both applied voltage and CNT surface chemistry. At 1.0 V, the eﬄuent pH is increased over the inﬂuent pH indicating that cathodic processes such as water reduction to hydrogen releasing hydroxide anions are controlling the pH, in agreement with Figure 2C where the cathodic potential dominates at e1.0 V. As the applied voltage is increased to 2.0 V, the eﬄuent pH approaches the inﬂuent pH indicating that the cathodic and anodic processes neutralize each other. As the applied voltage is increased further to 3.0 V, the eﬀect on eﬄuent pH is CNT surface chemistry dependent. For CNT, CNT HNO3, and C CNT HNO3, the eﬄuent pH tends to increase, whereas for the rest of the materials, the eﬄuent pH tends to decrease suggesting anodic processes are dominant. The increase of eﬄuent pH for the HNO3-treated CNTs suggests that electrooxidative cleavage of oxy-functional groups may result in a pH increase. The potential-dependent back pressure is observed to increase above baseline at applied voltages >2.0 V, Figure 4E, similar to the steady-state current response, Figure 4B, suggesting the back pressure is due to an electrochemical process. For example, both 9731

dx.doi.org/10.1021/es202271z |Environ. Sci. Technol. 2011, 45, 9726–9734

Environmental Science & Technology

ARTICLE

Figure 4. Electrochemical and eﬄuent characteristics vs surface chemistry and voltage: (A) MO degradation (1 [MO]ef/[MO]in), (B) steady-state current (mA), (C) anode potential (V), (D) eﬄuent pH, and (E) back pressure (kPa) are plotted as a function of applied voltage (V) for CNT (black squares), C CNT (red circles), CNT HNO3 (blue up triangles), C CNT HNO3 (green down triangles), CNT HCl (pink left triangles), C CNT HCl (yellow right triangles), and C CNT SS (navy diamonds). (F) Images of the gas bubbles produced during electrochemical CNT ﬁltration. Error bars were left oﬀ the ﬁgures for clarity, and typical standard deviations for the degradation, current, and potential plots were (5%, and for pH and back pressure they were (20%.

cathodic hydrogen production and anodic oxygen production and subsequent bubble formation within the ﬁltration device are responsible for the increased back pressure. To support this conclusion, images of bubbles being released into both the inﬂuent and eﬄuent and bubbles being formed on the electrodes are displayed in Figure 4F. Collection of cathodic hydrogen could result in increased energy eﬃciency of the electrochemical CNT ﬁltration process.50 Electrooxidative Mechanism. Although the electrochemical CNT filter is able to oxidize 95% of a 1 mM MO solution in a single-pass through the filter (τ < 1.2 s), there is still the question of the degree of molecular MO oxidation and the oxidation products. An estimation of the maximum oxidation can be made by comparing the MO molecular flux to the electron flux.

A 1 mM MO solution flowing at 1.5 mL min 1 would result in 1016 molecules s 1 flowing through the filter, and a current of 28 mA corresponds to 17 1016 e s 1 flowing through the anode. A maximum of 17 e could be anodically oxidized per MO molecule, which has a total of 80 e per molecule. Thus, only partial oxidation of an MO molecule is possible. Cathodic oxygen reduction to O2• , H2O2, and HO 3 could increase the degree of oxidation and anodic water oxidation to O2 could decrease the degree of oxidation. In an attempt to identify major MO electrooxidation products, inﬂuent and eﬄuent samples for all CNTs run at 2 and 3 V were analyzed by negative-ion direct injection mass spectrometry, Supporting Information Figure S7. A large number of the peaks 9732

dx.doi.org/10.1021/es202271z |Environ. Sci. Technol. 2011, 45, 9726–9734

Environmental Science & Technology did not change during electrooxidation. For example, peaks at m/z = 119/121 and m/z = 261/263 are characteristic of salt clusters NaSO4 and Na3(SO4)2 . The parent MO ion (m/z = 304/306), parent MO 16 (m/z = 288/289/290), parent MO + 16 (m/z = 320/322), and parent MO + 32 (m/z = 336/338) are observed in the inﬂuent sample and decrease signiﬁcantly in the 2 and 3 V samples indicating the parent MO molecule has been destroyed. Three new peaks in the 2 and 3 V samples appear. A peak at m/z = 290 representative of either CH2 or N loss appears at 2 V and disappears at 3 V. Two peaks at m/z = 173 and 189, indicated by arrows in the spectrum, appear at 2 V and grow further at 3 V. These peaks could possibly be aminobenzenesulfonate and hydroxyaminobenzenesulfonate. All three intermediates indicate an electrooxidative bond-breaking process is active. Environmental Implications. An electrochemical carbon nanotube filter has been shown to be effective for the oxidation of methyl orange and other organics. The energy efficiency of MO electrochemical filtration can be calculated in kW 3 hr kg 1 COD assuming 17 electrons transferred per molecule to be 4 (2 V) and 15 (3 V) for C CNT SS and 5 (2 V) and 16 (3 V) for C CNT HCl, and these values are similar to state-of-the-art electrochemical oxidation processes that are generally in the range of 5 100 kW 3 hr kg 1 COD.3 Alternatively, the energy per volume treated can be calculated in kW 3 hr m 3 to be 0.17 (2 V) and 0.93 (3 V) for C CNT SS and 0.22 (2 V) and 0.96 (3 V) for C CNT HCl, once again similar to other recently developed nanostructured electrodes at ∼0.7 kW 3 hr m 3.8 The efficiency and extent of degradation are both voltage-dependent with a greater efficiency at lower voltages and greater degradation at higher voltages. The efficiency and extent of degradation are also observed to be dependent on the CNT surface chemistry and the target molecule’s physical chemical properties. Thus, one strategy to increase electrooxidation is to add an electrocatalyst with a high O2 overpotential, C CNT SS, to increase the electron-transfer rate and reduce energy toward null reactions such as water oxidation. Another strategy is to tailor the CNT surface toward strong target molecule adsorption, C CNT HCl (MO) or C CNT HNO3 (MB), to increase the effective residence time of that molecule within the filter and in turn increase the possibility of oxidation.

’ ASSOCIATED CONTENT

bS

Supporting Information. Depiction of the electrochemical CNT ﬁlter, images of the modiﬁed ﬁltration casing, images of the CNTs before and after use, SEM images at both 10 k and 50 k magniﬁcation of the CNTs, thermogravimetric analysis data for all CNTs, XPS data for all CNTs, and direct injection mass spectrum of products at 2 and 3 V for all CNTs. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: (617) 496-1458; e-mail: [email protected].

’ ACKNOWLEDGMENT G.G. thanks the Harvard GSAS Visiting Fellow Scholarship. We thank Dr. Hao-Yu (Greg) Lin for assistance with the XPS analysis. We thank Dr. Philseok Kim for assistance with the

ARTICLE

TGA analysis. We thank Harvard’s Center for Nanoscale Systems for SEM and XPS. We thank Harvard’s Material Research Science and Engineering Center for TGA. We thank Dr. Soeren Zorn for assistance with the MS analysis. We thank the Martin Lab for MS.

’ REFERENCES (1) Bard, A. J.; Faulkner, L. R. Electrochemical Methods: Fundamentals and Applications, 2nd ed.; John Wiley & Sons: New York, 2001; p 833. (2) Rajeshwar, K.; Ibanez, J. G.; Swain, G. M. Electrochemistry and the environment. J. Appl. Electrochem. 1994, 24 (11), 1077– 1091. (3) Panizza, M.; Cerisola, G. Direct and mediated anodic oxidation of organic pollutants. Chem. Rev. 2009, 109 (12), 6541–6569. (4) Martinez-Huitle, C. A.; Brillas, E. Electrochemical alternatives for drinking water disinfection. Angew. Chem., Int. Ed. 2008, 47 (11), 1998–2005. (5) Park, H.; Vecitis, C. D.; Hoﬀmann, M. R. Solar-powered electrochemical oxidation of organic compounds coupled with the cathodic production of molecular hydrogen. J. Phys. Chem. A 2008, 112 (33), 7616–7626. (6) Zhi, J. F.; Wang, H. B.; Nakashima, T.; Rao, T. N.; Fujishima, A. Electrochemical incineration of organic pollutants on boron-doped diamond electrode. Evidence for direct electrochemical oxidation pathway. J. Phys. Chem. B 2003, 107 (48), 13389–13395. (7) Tong, X. L.; Zhao, G. H.; Liu, M. C.; Cao, T. C.; Liu, L.; Li, P. Q. Fabrication and high electrocatalytic activity of three-dimensional porous nanosheet pt/boron-doped diamond hybrid ﬁlm. J. Phys. Chem. C 2009, 113 (31), 13787–13792. (8) Zhao, G. H.; Li, P. Q.; Nong, F. Q.; Li, M. F.; Gao, J. X.; Li, D. M. Construction and high performance of a novel modiﬁed boron-doped diamond ﬁlm electrode endowed with superior electrocatalysis. J. Phys. Chem. C 2010, 114 (13), 5906–5913. (9) Li, P. Q.; Zhao, G. H.; Cui, X.; Zhang, Y. G.; Tang, Y. T. Constructing stake structured TiO2-NTs/Sb-doped SnO2 electrode simultaneously with high electrocatalytic and photocatalytic performance for complete mineralization of refractory aromatic acid. J. Phys. Chem. C 2009, 113 (6), 2375–2383. (10) Matyasovszky, N.; Tian, M.; Chen, A. C. Kinetic study of the electrochemical oxidation of salicylic acid and salicylaldehyde using UV/Vis spectroscopy and multivariate calibration. J. Phys. Chem. A 2009, 113 (33), 9348–9353. (11) Park, H.; Vecitis, C. D.; Hoﬀmann, M. R. Electrochemical water splitting coupled with organic compound oxidation: The role of active chlorine species. J. Phys. Chem. C 2009, 113 (18), 7935–7945. (12) Borras, N.; Oliver, R.; Arias, C.; Brillas, E. Degradation of atrazine by electrochemical advanced oxidation processes using a borondoped diamond anode. J. Phys. Chem. A 2010, 114 (24), 6613–6621. (13) Peigney, A.; Laurent, C.; Flahaut, E.; Bacsa, R. R.; Rousset, A. Speciﬁc surface area of carbon nanotubes and bundles of carbon nanotubes. Carbon 2001, 39 (4), 507–514. (14) Ebbesen, T. W.; Lezec, H. J.; Hiura, H.; Bennett, J. W.; Ghaemi, H. F.; Thio, T. Electrical conductivity of individual carbon nanotubes. Nature 1996, 382 (6586), 54–56. (15) Pantano, A.; Parks, D. M.; Boyce, M. C. Mechanics of deformation of single- and multi-wall carbon nanotubes. J. Mech. Phys. Solids 2004, 52 (4), 789–821. (16) Wang, X.; Li, W. Z.; Chen, Z. W.; Waje, M.; Yan, Y. S. Durability investigation of carbon nanotube as catalyst support for proton exchange membrane fuel cell. J. Power Sources 2006, 158 (1), 154–159. (17) Li, J.; Cassell, A.; Delzeit, L.; Han, J.; Meyyappan, M. Novel three-dimensional electrodes: Electrochemical properties of carbon nanotube ensembles. J. Phys. Chem. B 2002, 106 (36), 9299–9305. (18) Brady-Estevez, A. S.; Kang, S.; Elimelech, M. A single-walledcarbon-nanotube ﬁlter for removal of viral and bacterial pathogens. Small 2008, 4 (4), 481–484. 9733

dx.doi.org/10.1021/es202271z |Environ. Sci. Technol. 2011, 45, 9726–9734

Environmental Science & Technology (19) Vecitis, C. D.; Gao, G. D.; Liu, H. Electrochemical carbon nanotube ﬁlter for adsorption, desorption, and oxidation of aqueous dyes and anions. J. Phys. Chem. C 2011, 115 (9), 3621–3629. (20) Vecitis, C. D.; Schnoor, M. H.; Rahaman, M. S.; Schiﬀman, J. D.; Elimelech, M. Electrochemical multiwalled carbon nanotube ﬁlter for viral and bacterial removal and inactivation. Environ. Sci. Technol. 2011, 45 (8), 3672–3679. (21) Pan, B.; Xing, B. S. Adsorption mechanisms of organic chemicals on carbon nanotubes. Environ. Sci. Technol. 2008, 42 (24), 9005–9013. (22) Smith, B.; Wepasnick, K.; Schrote, K. E.; Bertele, A. H.; Ball, W. P.; O’Melia, C.; Fairbrother, D. H. Colloidal properties of aqueous suspensions of acid-treated, multi-walled carbon nanotubes. Environ. Sci. Technol. 2009, 43 (3), 819–825. (23) Kang, S.; Mauter, M. S.; Elimelech, M. Physicochemical determinants of multiwalled carbon nanotube bacterial cytotoxicity. Environ. Sci. Technol. 2008, 42 (19), 7528–7534. (24) Wang, X. M.; Li, N.; Webb, J. A.; Pfeﬀerle, L. D.; Haller, G. L. Eﬀect of surface oxygen containing groups on the catalytic activity of multi-walled carbon nanotube supported Pt catalyst. Appl. Catal., B 2010, 101 (1 2), 21–30. (25) Kim, Y. K.; Park, H. Light-harvesting multi-walled carbon nanotubes and CdS hybrids: Application to photocatalytic hydrogen production from water. Energy Environ. Sci. 2011, 4 (3), 685–694. (26) Kang, S.; Herzberg, M.; Rodrigues, D. F.; Elimelech, M. Antibacterial eﬀects of carbon nanotubes: Size does matter. Langmuir 2008, 24 (13), 6409–6413. (27) Kim, U. J.; Furtado, C. A.; Liu, X. M.; Chen, G. G.; Eklund, P. C. Raman and IR spectroscopy of chemically processed single-walled carbon nanotubes. J. Am. Chem. Soc. 2005, 127 (44), 15437–15445. (28) Cho, H. H.; Wepasnick, K.; Smith, B. A.; Bangash, F. K.; Fairbrother, D. H.; Ball, W. P. Sorption of aqueous Zn[II] and Cd[II] by multiwall carbon nanotubes: The relative roles of oxygen-containing functional groups and graphenic carbon. Langmuir 2010, 26 (2), 967–981. (29) Fujihara, S.; Maeda, T.; Ohgi, H.; Hosono, E.; Imai, H.; Kim, S. H. Hydrothermal routes to prepare nanocrystalline mesoporous SnO2 having high thermal stability. Langmuir 2004, 20 (15), 6476–6481. (30) Wen, Z. H.; Wang, Q.; Zhang, Q.; Li, J. H. In situ growth of mesoporous SnO2 on multiwalled carbon nanotubes: A novel composite with porous-tube structure as anode for lithium batteries. Adv. Funct. Mater. 2007, 17 (15), 2772–2778. (31) Moon, J. M.; An, K. H.; Lee, Y. H.; Park, Y. S.; Bae, D. J.; Park, G. S. High-yield puriﬁcation process of singlewalled carbon nanotubes. J. Phys. Chem. B 2001, 105 (24), 5677–5681. (32) Shi, Z. J.; Lian, Y. F.; Liao, F. H.; Zhou, X. H.; Gu, Z. N.; Zhang, Y. G.; Iijima, S. Puriﬁcation of single-wall carbon nanotubes. Solid State Commun. 1999, 112 (1), 35–37. (33) Briggs, D.; Seah, M. P. Practical Surface Analysis: Auger and X-ray Photoelectron Spectroscopy, 2nd ed.; John Wiley & Sons Limited: New York, 1990. (34) Hu, H.; Zhao, B.; Itkis, M. E.; Haddon, R. C. Nitric acid puriﬁcation of single-walled carbon nanotubes. J. Phys. Chem. B 2003, 107 (50), 13838–13842. (35) Rinzler, A. G.; Liu, J.; Dai, H.; Nikolaev, P.; Huﬀman, C. B.; Rodriguez-Macias, F. J.; Boul, P. J.; Lu, A. H.; Heymann, D.; Colbert, D. T.; Lee, R. S.; Fischer, J. E.; Rao, A. M.; Eklund, P. C.; Smalley, R. E. Large-scale puriﬁcation of single-wall carbon nanotubes: Process, product, and characterization. Appl. Phys. A 1998, 67 (1), 29–37. (36) Liu, L.; Li, F. B.; Feng, C. H.; Li, X. Z. Microbial fuel cell with an azo-dye-feeding cathode. Appl. Microbiol. Biotechnol. 2009, 85 (1), 175–183. (37) McKenzie, K. J.; Marken, F. Direct electrochemistry of nanoparticulate Fe2O3 in aqueous solution and adsorbed onto tin-doped indium oxide. Pure Appl. Chem. 2001, 73 (12), 1885–1894. (38) Enami, S.; Hoﬀmann, M. R.; Colussi, A. J. Proton availability at the air/water interface. J. Phys. Chem. Lett. 2010, 1 (10), 1599–1604. (39) Enami, S.; Stewart, L. A.; Hoﬀmann, M. R.; Colussi, A. J. Superacid chemistry on mildly acidic water. J. Phys Chem. Lett. 2010, 1 (24), 3488–3493.

ARTICLE

(40) Conway, B. E.; Tilak, B. V. Interfacial processes involving electrocatalytic evolution and oxidation of H-2, and the role of chemisorbed H. Electrochim. Acta 2002, 47 (22 23), 3571–3594. (41) Tilak, B. V.; Conway, B. E. Overpotential decay behavior 0.1. Complex electrode-reactions involving adsorption. Electrochim. Acta 1976, 21 (10), 745–752. (42) Wardman, P. Reduction potentials of one-electron couples involving free-radicals in aqueous-solution. J. Phys. Chem. Ref. Data 1989, 18 (4), 1637–1755. (43) Kotz, R.; Stucki, S.; Carcer, B. Electrochemical waste-water treatment using high overvoltage anodes 0.1. Physical and electrochemical properties of SnO2 anodes. J. Appl. Electrochem. 1991, 21 (1), 14–20. (44) Stucki, S.; Kotz, R.; Carcer, B.; Suter, W. Electrochemical wastewater treatment using high overvoltage anodes 0.2. Anode performance and applications. J. Appl. Electrochem. 1991, 21 (2), 99–104. (45) Fan, Y. W.; Goldsmith, B. R.; Collins, P. G. Identifying and counting point defects in carbon nanotubes. Nat. Mater. 2005, 4 (12), 906–911. (46) Cho, H. H.; Smith, B. A.; Wnuk, J. D.; Fairbrother, D. H.; Ball, W. P. Inﬂuence of surface oxides on the adsorption of naphthalene onto multiwalled carbon nanotubes. Environ. Sci. Technol. 2008, 42 (8), 2899–2905. (47) Yang, K.; Wu, W. H.; Jing, Q. F.; Jiang, W.; Xing, B. S. Competitive adsorption of naphthalene with 2,4-dichlorophenol and 4-chloroaniline on multiwalled carbon nanotubes. Environ. Sci. Technol. 2010, 44 (8), 3021–3027. (48) Iniesta, J.; Michaud, P. A.; Panizza, M.; Cerisola, G.; Aldaz, A.; Comninellis, C. Electrochemical oxidation of phenol at boron-doped diamond electrode. Electrochim. Acta 2001, 46 (23), 3573–3578. (49) Mrowetz, M.; Balcerski, W.; Colussi, A. J.; Hoﬀmann, M. R. Oxidative power of nitrogen-doped TiO2 photocatalysts under visible illumination. J. Phys. Chem. B 2004, 108 (45), 17269–17273. (50) Park, H.; Vecitis, C. D.; Choi, W.; Weres, O.; Hoﬀmann, M. R. Solar-powered production of molecular hydrogen from water. J. Phys. Chem. C 2008, 112 (4), 885–889.

9734

dx.doi.org/10.1021/es202271z |Environ. Sci. Technol. 2011, 45, 9726–9734

ARTICLE pubs.acs.org/est

Improving the Environmental Profile of Wood Panels via Co-Production of Ethanol and Acetic Acid J. Mason Earles,*,†,§ Anthony Halog,†,§ and Stephen Shaler†,‡,§ †

Forest Bioproducts Research Institute, University of Maine, 5737 Jennes Hall, Orono, Maine, United States Advanced Structures and Composites Center, University of Maine, 5793 AEWC Building, Orono, Maine, United States § School of Forest Resources, University of Maine, 5755 Nutting Hall, Orono, Maine, United States ‡

bS Supporting Information ABSTRACT: The oriented strand board (OSB) bioreﬁnery is an emerging technology that could improve the building, transportation, and chemical sectors’ environmental proﬁles. By adding a hot water extraction stage to conventional OSB panel manufacturing, hemicellulose polysaccharides can be extracted from wood strands and converted to renewably sourced ethanol and acetic acid. Replacing fossil-based gasoline and acetic acid has the potential to reduce greenhouse gas (GHG) emissions, among other possible impacts. At the same time, hemicellulose extraction could improve the environmental proﬁle of OSB panels by reducing the level of volatile organic compounds (VOCs) emitted during manufacturing. In this study, the life cycle signiﬁcance of such GHG, VOC, and other emission reductions was investigated. A process model was developed based on a mix of laboratory and industrial-level mass and energy ﬂow data. Using these data a life cycle assessment (LCA) model was built. Sensitive process parameters were identiﬁed and used to develop a target production scenario for the OSB bioreﬁnery. The ﬁndings suggest that the OSB bioreﬁnery’s deployment could substantially improve human and ecosystem health via reduction of select VOCs compared to conventionally produced OSB, gasoline, and acetic acid. Technological advancements are needed, however, to achieve desirable GHG reductions.

1. INTRODUCTION AND BACKGROUND Oriented strand board (OSB) is the most common structural wood composite panel used in the United States. In 1999, it held 50% of the market share,1 with North American production levels greater than 22 million m3.2 It has been proposed that the OSB manufacturing process can be modiﬁed to extract hemicellulose and other compounds from wood ﬂakes prior to panel manufacturing.3,4 This extract has the potential to be converted into combustible alcohols (e.g., ethanol, butanol, methanol) and industrial chemicals (e.g., acetic acid, polylactic acid, and furfural), among other useful products. Whereas combustible alcohols would likely replace fossil-based transportation fuels such as gasoline, biobased industrial chemicals can be used to manufacture polymers and other products. Experimental results suggest that the addition of a hot water extraction process could remove about 10% of the OSB strand mass in the form of hemicelluloses and other dissolved compounds.3 Paredes4 estimates that around 409.3 million liters of ethanol could be produced annually from these hemicelluloses. Although this represents a relatively small percentage (less than 0.1%) of total U.S. gasoline consumption on an energy basis, OSB extracted hemicellulose could become one option within a diverse portfolio of feedstocks used to produce biobased transport fuel, among other chemicals. r 2011 American Chemical Society

The OSB bioreﬁnery manufacturing process can be understood in three parts: shared, OSB, and ethanol/acetic acid pathways. Figure 1 illustrates these pathways. 1.1. Shared Pathway. Following harvest, logs are transported to an OSB mill in which they enter the debarking and flaking stage. Debarked logs are flaked by sets of rotating blades that cut strands of varying lengths, widths, and thicknesses. Bark is typically combusted to generate heat or sold as mulch.5 In conventional OSB manufacturing, OSB ﬂakes next move to the drying and screening stage. The OSB bioreﬁnery, however, adds an autohydrolysis, or hot water extraction, stage following debarking and ﬂaking. Flakes and water are inserted into a vessel and heated to 140190 °C which leads to autohydrolysis at a pH near 3.3,6,7 Various combinations of temperature and time can be utilized to obtain the desired composition of dissolved solids. In addition to carbohydrates, acetic acid, hydroxymethylfurfural, furfural, and other chemical compounds are contained in the extract. This unit process results in two coproducts, wood ﬂakes and hemicellulose extract, which are then converted into the ﬁnal Received: June 26, 2011 Accepted: October 3, 2011 Revised: September 23, 2011 Published: October 03, 2011 9743

dx.doi.org/10.1021/es202180m | Environ. Sci. Technol. 2011, 45, 9743–9749

Environmental Science & Technology

ARTICLE

Figure 1. OSB bioreﬁnery process diagram.

OSB and chemical products along two distinct pathways. As a result of extraction, wood ﬂake mass is reduced proportionally to the amount of solids removed. 1.2. OSB Pathway. The manufacturing processes along the OSB pathway after hot water extraction are the same as conventional manufacturing. Mass and energy flows, however, are altered. Wood flakes are dried following hot water extraction. In conventional OSB manufacturing, the drying process is energy intensive, often demanding up to 80% of the heat requirement in a mill.8 The primary heat source for drying is from wood fuel generated as a byproduct of OSB manufacturing.5 In an OSB biorefinery, the pressurized conditions during hot water extraction significantly increase strand moisture content to near full saturation conditions.3 Consequently, additional energy will likely be required for drying; although significant amounts of moisture may be flashed off as strands leave the extraction unit at high temperature.3 Because the drying stage is a major consumer of wood heat input, it contributes significantly to airborne volatile organic compound (VOC), particulate matter (PM), and greenhouse gas (GHG) emissions. To meet U.S. regulatory standards, the VOCs and PM released from drying must be reduced using emission control equipment such as a regenerative thermal oxidizer (RTO) and wet electrostatic precipitator (WESP). As VOCs are primarily produced via hemicellulose decomposition,9 and past research has shown significant VOC reduction from hot water extraction in the pressing process,3 it is expected that reductions will occur in the drying process as well. The presence of fewer VOCs consequently increases CO2 emissions from natural gas combustion (see Supporting Information, Section 1.2.4). The screening process follows drying. It aims to screen out ﬁne wood materials from the dried ﬂakes that are too small for

OSB mat formation. These screening ﬁnes are burned to generate heat. The blending process combines strands with resin binders and a small amount of wax. Then, these strands are oriented in the mat formation process before pressing. During the pressing stage the OSB mat is pressed under high heat and pressure, creating a rigid, dense structural panel. The press is also a signiﬁcant consumer of heat and emitter of VOCs. Importantly, hot water extraction is shown to reduce select VOCs between 15% and 75% during the pressing stage of OSB production.3 As with the drying process, VOC reduction oﬀers an opportunity to reduce energy consumption and related CO2 emissions. The ﬁnal stage of OSB manufacturing is ﬁnishing, in which panels are cooled, cut to size, grade stamped, stacked in bundles, and packaged for shipping.5 Scrap material in the form of sawdust, sander dust, and reject boards is used for heat generation. Regarding the ﬁnal panel, prior research suggests that following moderate hot water extraction mechanical properties of panels manufactured from extracted material are improved or not signiﬁcantly diﬀerent compared to conventional panels.3 The resultant panels also exhibit signiﬁcantly improved resistance to moisture.3 To maintain the same number of panels per unit wood input, it would be necessary to reduce panel density in proportion to the weight reduction associated with extraction. It is expected that such a density reduction would reduce transportation costs and environmental impacts related to transportation from the manufacturing facility to the construction site. 1.3. Ethanol and Acetic Acid Pathway. Following hot water extraction, the extract must meet a desirable concentration of hemicellulose. If the concentration is too low, evaporation, 9744

dx.doi.org/10.1021/es202180m |Environ. Sci. Technol. 2011, 45, 9743–9749

Environmental Science & Technology recirculation, or ultrafiltration can be used. Each of these techniques has unique technoeconomic strengths and weaknesses.10 Once the desired hemicellulose concentration is achieved, the extract enters the acid hydrolysis stage in which sulfuric acid (H2SO4) is added. Following hot water extraction (or autohydrolysis) the pH of the extracted liquid is about 34 due to the presence of acetic acid released from acetylated polysaccharides.7,11 For acid hydrolysis a pH of approximately 1 is desirable.10 Thus, the corresponding amount of sulfuric acid is added to lower the pH from approximately 3 to 1. It is assumed that all of the lignin in the extract is precipitated due to the low pH. Because it is desirable to minimize the production of gypsum during the liming stage, and as gypsum production is a function of sulfuric acid input, low sulfuric acid consumption is also preferable. One way to reduce overall sulfuric acid consumption is through an acid recovery process, such as membrane separation. Acid recovery using membrane separation can achieve at least 80% recovery and often higher.12 In the sensitivity section, we examine a scenario in which an acid recovery is performed following acid hydrolysis. The hydrolyzed extract, or hydrolyzate, enters the lignin ﬁltration stage in which lignin is removed, presumably at a rate of 100%.10 During the acid hydrolysis step, a considerable amount of acetic acid will be generated and can be sold as a coproduct. Additionally, acetic acid must be removed as it inhibits the fermentation of C5 and C6 sugars to ethanol.13 The liquidliquid extraction process removes acetic acid, and some furfural, using a suitable solvent.14,15 Consistent with Mao,10 it is assumed that ethyl acetate is used as the solvent for liquidliquid extraction which is recovered at a rate of 100%. Once the acetic acid is removed from the extract, the remaining solution enters the liming process in which the pH is raised from approximately 1 to about 6 by adding lime (calcium oxide, CaO). The liming process aims to (1) raise the pH of the extract, (2) precipitate sulfate ions as gypsum (CaSO4 3 H2O), and (3) detoxify the hydrolyzate.9 The gypsum is then removed by ﬁltration and typically disposed to a landﬁll.16 The hydrolyzate now enters the fermentation process. Assuming the sugar concentration is suitable, and inhibitors like furfural and hydroxyl methyl furfural are removed, micro-organisms such as E-coli B (KO11)17,18 or Z. mobiliz19 are used to convert ﬁve and six carbon sugars into ethanol and CO2. Consistent with Mao,10 this study assumes that fermentation takes place under anaerobic conditions with approximately 90% conversion eﬃciency of hexose and pentose sugars to ethanol (see Supporting Information). Because anerobic fermentation of glucoronic acid forms acetic acid, the acetic acid is sent to the liquidliquid extraction process. The ﬁnal output is a very dilute ethanol which is sent to ethanol distillation, where it is rectiﬁed and dehydrated to 99% concentration.10 1.4. Expected Environmental Benefits of the OSB Biorefinery. The OSB biorefinery has the potential to improve the environmental profile of OSB panels, ethanol, and acetic acid compared to conventional production systems. The primary environmental benefit of replacing fossil-based gasoline and acetic acid with hemicellulosic ethanol and acetic acid is the possibility of GHG emission reduction. More specifically, a GHG reduction can occur during the combustion of biobased ethanol compared to gasoline. Upon combustion biobased ethanol releases biogenic CO2 which is not typically considered as a contributor to global warming since it was originally sequestered during plant growth.20 Thus, assuming that GHG emissions

ARTICLE

Table 1. Functional Unit for OSB Bioreﬁnery and Conventional Production System OSB bioreﬁnery

a

quantity

conventional system

quantity

ethanol

1000 kg

gasoline

613 kg

acetic acid OSB panels

368 kg 55.3 MSFa

acetic acid OSB panels

368 kg 55.3 MSFa

MSF = 1000 square feet (3/800 panel thickness).

released in other parts of ethanol’s lifecycle are not substantially higher than gasoline, a reduction can occur. Acetic acid, on the other hand, is typically produced in the U.S. via the reaction of CO with natural gas-derived methanol.21 In both cases, switching from a fossil feedstock to renewable feedstock decreases depletion of fossil resources. The hemicellulose extraction process could also improve the environmental proﬁle of OSB panels. First, hemicellulose extraction is shown to reduce select VOCs between 15% and 75% in the pressing stage of OSB production.3 More speciﬁcally, Paredes4 found that methanol was reduced by about 26%, acetaldehyde by 74%, and formaldehyde by 15% (at varying degrees of signiﬁcance). Although no direct evidence exists to suggest that such reductions will occur in the drying process as well, the fact that VOCs are primarily produced by the decomposition of hemicellulose, much of which is removed during hot water extraction, implies this will be the case. Because select VOCs are damaging to human and ecosystem health, and the OSB manufacturing process is major emitter of VOCs,22 their removal could represent a substantial environmental improvement. Second, due to improved mechanical properties and ﬂake density reduction from hemicellulose extraction, a mass reduction of about 9.25% per panel could be achieved, which increases resource use eﬃciency and reduces overall transportation requirements per panel.3 Whereas the OSB bioreﬁnery holds many potential environmental beneﬁts, their relative signiﬁcance from a life cycle perspective is not obvious. This study aims to characterize these impacts with respect to all stages of the supply chain—including raw material extraction, manufacturing, use, and all intermediary transportation. A number of uncertain process parameters are included via sensitivity analysis. Finally, a target OSB bioreﬁnery process is presented that minimizes key environmental impacts. Due to potential GHG and VOC reductions associated with the OSB bioreﬁnery, emphasis is placed on climate change and toxicity-related impacts.

2. METHODOLOGY This study compares three OSB bioreﬁnery coproducts with a conventional manufacturing system as shown in Table 1. Detailed mass and energy ﬂow data for the shared, OSB, and ethanol/acetic acid pathways can be found in the Supporting Information. Several steps were taken to evaluate the environmental impacts of the OSB bioreﬁnery. First, since no existing process model was identiﬁed, a baseline mass and energy ﬂow model of an OSB bioreﬁnery was developed based on best available data. The process model integrated and modiﬁed data from existing studies on conventional OSB manufacturing,5 hot water extraction of hemicellulose,3 and hemicellulosic ethanol production,10 among other sources. To develop a life cycle inventory (LCI), Kline5 collected process-speciﬁc data via survey from four OSB 9745

dx.doi.org/10.1021/es202180m |Environ. Sci. Technol. 2011, 45, 9743–9749

Environmental Science & Technology

ARTICLE

Table 2. LCIA Comparison of Overall OSB Bioreﬁnery and Conventional Production Systems code GWP

impact category global warming

equivalence factor (kg)

conventional system

baseline OSB bioreﬁnery

% diﬀerence

CO2-Eq

1.89 104

2.16 104

14.6%

1.30 104

19.5%

HTP

human toxicity

1,4-DCB-Eq

1.62 10

FEP

freshwater eutrophication

P-Eq

9.89 101

1.08

FETP

freshwater ecotoxicity

1,4-DCB-Eq

5.35 101

4.82 101

MEP

marine eutrophication

N-Eq

1.77 102

1.77 102

0.1%

4

9.2% 9.8%

METP

marine ecotoxicity

1,4-DCB-Eq

5.85 10

5.24 101

10.5%

ODP

ozone depletion

CFC-11-Eq

3.87 106

1.40 105

261.9%

2

2.2% 0.3%

1

PMFP POFP

particulate matter formation photochemical oxidant formation

PM10-Eq NMVOC

3.43 10 1.42 103

3.51 102 1.41 103

TAP

terrestrial acidiﬁcation

SO2-Eq

8.81 102

9.13 102

TETP

terrestrial ecotoxicity

1,4-DCB-Eq

1.66

1.22

manufacturing plants in the southeastern United States. Kline5 was modiﬁed to include a hot water extraction process. Extraction conditions (i.e., time, temperature, liquid-to-wood ratio, and wood species) and dissolved solid composition were taken from laboratory scale results found by Paredes.4 The ethanol and acetic acid production pathway was based on a modiﬁed version of Mao.10 Modiﬁcations were made regarding extract composition (e.g., carbohydrates, acids, and lignin), acid hydrolysis conversion eﬃciency, and fermentation eﬃciency to account for the diﬀerence between hardwood modeled by Mao10 and the softwood species (i.e., southern yellow pine) assumed to be used in this study. Because an operational OSB bioreﬁnery does not currently exist, many of the assumptions made in the model above are uncertain. Thus, the model was parameterized to facilitate identiﬁcation of sensitive processes with respect to the system’s environmental impacts. The mass and energy ﬂow model was created such that thirteen parameters could be varied (listed in Supporting Information). Of these thirteen parameters, seven were identiﬁed as critical and tested for sensitivity in the following section. Next, mass and energy ﬂows for each scenario were brought into OpenLCA software23 to estimate the related life cycle environmental impacts using the ReCiPe 2008 impact method.24 Life cycle inventory (LCI) data originated from a variety of sources. Timber harvesting and production was based on LCI data collected by the Consortium for Research on Renewable Industrial Materials (CORRIM) on forest management and harvesting.25 LCI data associated with natural gas combustion in the OSB bioreﬁnery originated from the U.S. LCI database3 which was based on the GREET model.26 Ethanol transportation, distribution, and combustion in a vehicle also originated from Wu et al.26 More detailed system boundary diagrams are provided in the Supporting Information. Generally, most background data in this study originated from the U.S. LCI database.27 All energy related processes, such as electricity generation, natural gas production, and transport processes, were taken from the U.S. LCI database. Quicklime production is also available via the U.S. LCI database. Sulfuric acid production was taken from a report compiled by the Swiss Centre for Life Cycle Inventories entitled Ecoinvent Report No. 8, Chemicals (2007). Thus, sulfuric acid may not represent U.S. production technology. Data for potassium fertilizer production was acquired from an LCA report on U.S. pork production by Schenck.29 Finally, landﬁlling is modeled using one of two processes available through the European ELCD core database.30 More speciﬁcally, either a municipal solid waste or inert solid waste landﬁlling process was used since no better U.S. data are

3.6% 26.3%

publicly available. Such a limitation aﬀects this study’s ability to comment on the life cycle impacts of gypsum disposal with any certainty (see Supporting Information). LCI data for the conventional production system of gasoline, acetic acid, and OSB panels was taken almost exclusively from the U.S. LCI database; with the exception of gasoline transportation, distribution, and combustion which originated from the GREET model.26 For this study the ReCiPe 2008 Hierarchist impact assessment method was utilized.24 The impact categories included are contained within Table 2.

3. RESULTS AND DISCUSSION The following section presents the results for the baseline OSB bioreﬁnery impacts, tests for sensitive process parameters, and proposes a target process to minimize key environmental impacts. As shown in Table 2, compared to the conventional system the OSB bioreﬁnery oﬀers improvements with respect to toxicityrelated impact categories, such as HTP, FETP, METP, and TETP. The largest potential reductions occur in the FETP and HTP category at 26.3% and 19.5%, respectively. Decreased acrolein emissions from hot water extraction primarily drive such reductions in both cases. Barium from crude oil production and natural gas extraction undergoes the second largest reduction in the HTP category. Other VOCs, such as formaldehyde, acetaldehyde, and methanol have a less noticeable eﬀect on FETP and HTP. Some categories, most notably GWP and FEP, increase compared to the conventional production system. Increased CO2 from electricity and heat consumption almost exclusively explains the higher GWP impact of the OSB bioreﬁnery baseline compared to the conventional production system. Detailed process and ﬂow contributions for each impact category can be found in the Supporting Information. Whereas toxicity-related impact categories experience signiﬁcant reduction, GWP impacts should be lowered in order for ethanol produced at an OSB bioreﬁnery to qualify as a renewable fuel under U.S. biofuel policy as outlined in the Renewable Fuel Standard 2 (RFS2).31 Based on the results for the baseline scenario, seven process parameters were identiﬁed as potentially important in achieving this goal (see Table 3 and Figure 2). Three parameters pertain to the hot water extraction process. First, higher levels of heat recovery reduce energy consumption in the extraction process. Heat recovery was assumed to occur at 50% in the baseline scenario. For the alternative scenario an 9746

dx.doi.org/10.1021/es202180m |Environ. Sci. Technol. 2011, 45, 9743–9749

Environmental Science & Technology Table 3. Parameters Tested for Sensitivity Analysis

Figure 2. Sensitivity analysis of LCIA categories with respect to select process parameters.

arbitrary value of 80% was selected as a reasonable upper bound. Ultimately, actual heat recovery eﬃciency will be decided by process design and cost considerations. Second, depending on carbohydrate concentration levels following extraction an evaporation process could be required to concentrate these carbohydrates to desirable levels for the fermentation process. Evaporation of water is a highly energy intensive process. Whereas the baseline scenario assumes that no evaporation will be needed (due to the use of extract recirculation), the alternative scenario examined the case where it is required. Third, higher liquid to wood ratios result in greater energy consumption for achieving the desired cook time and temperature. Consistent with Paredes,4 a liquid to wood ratio of 4 to 1 was tested in the baseline scenario. For the alternative scenario, this ratio was lowered to 2 to 1. Prior research compared liquid-to-wood ratios of 4 to 1 and 8 to 1, ﬁnding that both liquid-to-wood ratios (at time and temperature conditions similar to this study) had a minor eﬀect on extract yields during autohydrolysis.32 Based on these ﬁndings, we assume a negligible loss in extract yield will occur at lower liquid-to-wood ratios given the presumed time and temperature extraction conditions. The baseline scenario assumes that recirculation will be used to concentrate the solids to a level of 8.5%. Presumably doing this would have a negligible eﬀect on overall mass and energy ﬂows. The fourth parameter describes the percent mass reduction permissible per panel. Whereas Paredes4 ﬁnds that manufacturing lower density panels that maintain/improve mechanical properties is possible, actual density will be determined by the manufacturer based on their own technical speciﬁcations. In the base scenario, it was assumed that each panel can weigh 4.62% less than conventional panels. In this alternative scenario, the permissible value was raised to 9.25%. Doing this increases the

ARTICLE

amount of panels that can be produced given the same input of logs. As a result, the functional unit for the OSB bioreﬁnery and conventional systems change from 55.30 MSF panels to 57.85 MSF panels per 1000 kg of ethanol produced. Essentially, this represents an eﬃciency gain spread across all stages of the OSB bioreﬁnery lifecycle. The ﬁfth parameter tests the presence of sulfuric acid recovery which is 80% eﬀective. As mentioned earlier, membrane separation technology has achieved acid recovery eﬃciencies of 80% for sulfuric acid, while continuous ion exchange has reached 97% with 2% sugar loss.12 This alternative scenario assumes that 80% acid recovery occurs with negligible heat input requirements. This can be compared to the baseline scenario which assumes that 0% acid recovery occurs as was modeled in Mao.10 The sixth parameter investigates the model’s assumption that only 10% additional heat is needed to dry ﬂakes following hot water extraction compared to conventional OSB manufacturing process (without hot water extraction). In this alternative scenario, the value of 10% was arbitrarily raised to 20% to test for parameter sensitivity. Finally, based on recent trends in the literature ultraﬁltration is tested as an alternative concentration process following hot water extraction.3336 Ultraﬁltration is used as a low-energy separation technology when isolating a substance at a low concentration. Although membrane fouling is generally a major concern with ultraﬁltration, past studies have observed almost no fouling when hydrophilic membranes were used to separate hemicelluloses from pulp mill water.37,38 Moreover, a recent study found that hemicelluloses could be concentrated nearly 15 times, from 0.7 to 1015 g/L, with only a 10% loss factor using ultraﬁltration.36 Recall that the baseline OSB bioreﬁnery performed well on toxicity measures, but poorly on GWP. The results above suggest that the greatest GWP reduction potential exists with respect to parameters 1 and 3. By increasing heat recovery from 50% to 80%, system-wide kg CO2eq. emissions can be lowered by 8.6%. Similarly, reducing the liquid to wood ratio from 4:1 to 2:1 results in a 9.4% reduction in kg CO2eq. emissions. Doing either of these options alone, however, does not reduce system-wide emissions below the conventional level—as the baseline scenario has over 14% greater kg CO2eq. emissions compared to conventional production. The third largest reduction in GWP can be made by adding a sulfuric acid recovery process, leading to 5.4% less system-wide emissions compared to the baseline scenario. The ﬁfth largest GWP improvement is associated with increasing the permissible mass reduction per panel from 4.62% to 9.25%. This results in a 3.5% decrease in system-wide kg CO2eq. emissions. A large increase in GWP (at 13.7%) results from adding an evaporation process after hot water extraction. Ultraﬁltration oﬀers a less energy- and GWP-intensive form of separation than evaporation, only resulting in about a 2.5% increase in GWP. Raising the drying heat requirement from 10% to 20% results in a rather modest GWP increase of 3.5%. Other impact categories exhibited some sensitivity to the parameters tested. Generally, the eﬀects were relatively small with the exception of ODP in relation to parameter 5. While the percent change was relatively large, the absolute release of kg CFC-11 eq. was very small. Thus, this impact category is not discussed in further detail. It is important to consider the interaction among these diﬀerent parameters which can have aggregate eﬀects. This is especially relevant for those parameters which pertain to the hot water extraction process. A liquid to wood ratio of 2 to 1 9747

dx.doi.org/10.1021/es202180m |Environ. Sci. Technol. 2011, 45, 9743–9749

Environmental Science & Technology

Figure 3. Percent change in each LCIA category for target and baseline scenarios with respect to conventional production system.

occurring with 80% heat recovery will have an even more dramatic eﬀect, for instance, than either of the parameters varied individually. The compounded eﬀect of changing parameters simultaneously was considered next by formulating a best environmental case (see Table 3). Essentially, the best case scenario assumes that process improvements are made to the baseline scenario with respect to extraction heat recovery, extraction liquid to wood ratio, percent panel mass reduction permissible, and sulfuric acid recovery. In making these changes, this analysis aimed to examine potential reductions and set a target for the OSB bioreﬁnery process development that minimizes environmental impacts. Figure 3 shows the percent change for each LCIA category compared to the conventional production system. The baseline system is also shown to demonstrate the reduction potential of achieving the parameter values set for the best case scenario. The graph above shows that a substantial reduction in kg CO2eq. is possible for the best case scenario. In fact, an absolute reduction of about 2367 kg CO2eq. can be achieved at the system-wide level. Compared to the emissions for 613 kg gasoline (the energy equivalent of 1000 kg ethanol) at 2157 kg CO2eq., this represents a major reduction, exceeding the requirement of 60% lifecycle GHG emission reductions set by the RFS2. Additionally, signiﬁcant reductions can be achieved across toxicity-related categories due to (1) the reduction of acrolein during OSB manufacturing by adding hot water extraction and (2) the reduction in barium due to decreased demand for crude oil reﬁning to produce gasoline (see Supporting Information for details). It is also worth noting that the best case scenario reduces LCIA impacts across every category measured with the exceptions of FEP and ODP. FEP is likely to increase compared to conventional production so long as fertilizer is used for the production of biomass. In the case where natural regeneration is utilized for timber production, this would probably not be the case. In summary, this study examined the potential of the OSB bioreﬁnery to improve the environmental proﬁle of OSB panels, ethanol, and acetic acid compared to conventional production systems. The general ﬁnding was that toxicity-related impacts can be substantially reduced in the baseline OSB bioreﬁnery system compared to conventionally produced OSB panels, gasoline, and acetic acid. Speciﬁcally, the impact categories of HTP and TETP could be reduced by about 20% and 25%, respectively, by switching from conventional products to those produced at an OSB bioreﬁnery. Other toxicity-based categories, or FETP and METP, experienced less dramatic reductions due to lowered

ARTICLE

crude oil reﬁning (for gasoline production) compared to the conventional system. Furthermore, after identifying and adjusting certain process parameters it is possible to achieve even greater impact reductions. Thus, one can conclude that the VOC reductions from hot water extraction appear to have the desired eﬀect of signiﬁcantly reducing lifecycle impacts related to VOC emissions. A second important ﬁnding deals with the lifecycle GHG impacts of the OSB bioreﬁnery. As formulated in the baseline scenario, the OSB bioreﬁnery system adds substantial GHG burdens compared to conventionally produced gasoline. However, through sensitivity analysis key process parameters are identiﬁed that could potentially reduce this burden, leading to an overall GHG reduction. Under the target scenario, this study shows that substantial reductions in GHGs compared to the conventional system is possible. Achieving such reductions, concurrent to large toxicity-related reductions from hot water extraction, should be a priority for future OSB bioreﬁnery research.

’ ASSOCIATED CONTENT

bS

Supporting Information. Additional details on the process model mass and energy ﬂows, system boundaries, and lifecycle impact assessment results. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]; phone: +1 314-479-6570; fax: +1 207-581-2875.

’ ACKNOWLEDGMENT We are grateful to Dr. Peter van Walsum, Dr. Adriaan van Heiningen, and Dr. Jonathan Rubin at the University of Maine for their valuable feedback. ’ REFERENCES (1) Bowyer, J. L.; Shmulsky, R.; Haygreen, J. G. Forest Products and Wood Science: An Introduction, 4th ed.; Iowa State University Press: Ames, IA, 2003. (2) Adair, C. Regional Production and Market Outlook: Structural Panels and Engineered Wood Products 20042009; APA E170; APAThe Engineered Wood Association: Tacoma, WA, 2004. (3) Paredes, J.; Shaler, S.; Edgar, R.; Cole, B. Selected volatile organic compound emissions and performance of oriented strandboard from extracted southern pine. Wood Fiber Sci. 201042 (4). (4) Paredes, J.; Jara, R.; Shaler, S.; van Heiningen, A. Inﬂuence of hot water extraction on the physical and mechanical behavior of OSB. For. Prod. J. 200858. (5) Kline, E. Gate-to-gate life cycle inventory of oriented strandboard production. Wood Fiber Sci. 2005, 37, CORRIM Special Issue. (6) Yoon, S. H.; Macewan, K.; van Heiningen, A. Hot water preextraction from loblolly pine (Pinus taeda) in an integrated forest products bioreﬁnery. TAPPI 2008June. (7) Tunc, M.; van Heiningen, A. R. P. Hemicellulose extraction of mixed southern hardwood with water at 150°C: Eﬀect of time. Ind. Eng. Chem. Res. 2008, 47 (18), 7031–7037, DOI: 10.1021/ie8007105. (8) Lees, A. Future OSB plants will require latest environmental technologies. Panel World 1993May. 9748

dx.doi.org/10.1021/es202180m |Environ. Sci. Technol. 2011, 45, 9743–9749

Environmental Science & Technology (9) Carlson, F. E.; Phillips, E. K.; Tenhaeﬀ, S. C.; Deltlefsen, W. D. A study of formaldehyde and other organic emissions from pressing of laboratory oriented strandboard. For. Prod. J. 199545 (3). (10) Mao, H.; Genco, M.; van Heiningen, A.; Pendse, H. Technical economic evaluation of a hardwood bioreﬁnery using the ‘near-neutral’ hemicellulose pre-extraction process. J. Biobased Mater. Bioenergy 2008, 2 (2), No. 10.1166/jbmb.2008.309. (11) Brasch, D. J.; Free, K. W. Prehydrolysis-kraft pulping of Pinus radiate grown in New Zealand. TAPPI 196548 (4). (12) Hamelinck, C. N.; van Hooijdonk, G.; Faiij, A. Ethanol from lignocellulosic biomass: techno-economic performance in short-, middle- and long-term. Biomass Bioenergy 2005, 28, No. 10.1016/j. biombioe.2004.09.002. (13) Takahashi, C. M.; Takahashi, D. F.; Carvalhal, M. L. C.; Alterthum, F. Eﬀects of Acetate on the Growth and Fermentation Performance of Escherichia coli KO11. Appl. Biochem. Biotechnol. 199981. (14) Jones, L. Economic Saving Through the use of Solvent Extraction. Chem. Ind-London 196712 (3). (15) Geankoplis, C. J. Transport Processes and Separation Process Principles, 4th ed.; Prentice Hall: Upper Saddle River, NJ, 2003. (16) Martinez, A.; Rodriguez, M.; Wells, M.; York, S. Detoxiﬁcation of Dilute Acid Hydrolyzates of lignocellulose with Lime. Biotechnol. Prog. 2001, 17 (4), No. 10.1021/bp0001720. (17) Lawford, H. G.; Rousseau, J. D. Fermentation of Biomass Derived Glucuronic Acid by per Expressing Recombinants of E. Coli B. Appl. Biochem. Biotechnol. 19976365 (73). (18) Balasubramanian, N.; Kim, J. S.; Lee, Y. Y. Fermentation of Xylose into Acetic Acid by Clostridium Thermoaceticum. Appl. Biochem. Biotechnol. 2001, 9193 (19), No. 10.1385/ABAB:91-93:1-9:367. (19) Wooley, R.; Ruth, M.; Sheehan, J.; Ibsen, K. Lignocellulosic Biomass to Ethanol Process Design and Economic Utilizing Co-current Dilute Acid Prehydrolysis and Enzymatic Hydrolysis Current and Futuristic Scenarios; Technology Report NREL/TP-580-26157; National Renewable Energy Laboratory (NREL), 1999. (20) Intergovernmental Panel on Climate Change. Climate Change 2007: The Physical Science Basis. Contribution of Working Group 1 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, 2007. (21) Cradle-to-gate life cycle inventory of nine plastic resins and four polyurethane precursors. Final report prepared for the Plastics Division of the American Chemistry Council; Franklin Associates, 2010; http:// greenbuildingsolutions.org/Main-Menu/Resources/White-Papers/Cradleto-Gate-Life-Cycle-Inventory-of-Nine-Plastic-Resins-and-Two-Polyurethane-Precursors.pdf. (22) National Emission Standards for Hazardous Air Pollutants: Plywood and Composite Wood Products; Eﬄuent Limitations Guidelines and Standards for the Timber Products Point Source Category List of Hazardous Air Pollutants, Lesser Quantity Designations, Source Category List; Final Rule, 40 Federal Register, Parts 63 and 429, July 30, 2004. (23) OpenLCA software; www.openlca.org. (24) Goedkoop, M.; 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; 1st edition, Report I: Characterisation, 2008; http://www.lcia-recipe.net. (25) Johnson, L.; Lippke, B.; Marshall, J. D.; Comnick, C. Forest Resources Paciﬁc Northwest and Southeast. Wood Fiber Sci. 2005, 37, CORRIM Special Issue. (26) Wu, M.; Wang, M.; Huo, H. Fuel-Cycle Assessment of Selected Bioethanol Production Pathways in the United States; Technical Analysis Report, Argonne National Laboratory, 2006; www.transportation.anl.gov/modeling_simulation/GREET/publications.html. (27) US Life Cycle Inventory Database; www.nrel.gov/lci. (28) Ecoinvent Centre. Life Cycle Inventories of Chemicals; Ecoinvent report no. 8; 2007.

ARTICLE

(29) Schenck, R. Life Cycle Assessment of USA Pork Production in 2004; Institute for Environmental Research and Education, 2006. (30) European LCI database; http://lca.jrc.ec.europa.eu/lcainfohub/datasetArea.vm. (31) U.S. EPA. Renewable Fuel Standard Program (RFS2) Regulatory Impact Analysis; EPA-420-R-10-006; February, 2010. (32) Testova, L.; Vilonen, K.; Pynnonen, H.; Tenkanen, M.; Sixta, H. Isolation of hemicellulose from birch wood: Distribution of wood components and preliminary trials in dehydration of hemicelluloses. Lenzinger Berichte 2009, 87, 58–65. (33) Persson, T.; Matusiak, M.; Zacchi, G.; Jonsson, A.-S. Extraction of hemicelluloses from process water from the production of masonite. Desalination 2006, 199, No. 10.1016/j.desal.2006.03.093. (34) Persson, T.; Nordin, A.-K.; Zacchi, G.; Jonsson, A.-S. Economic evaluation of isolation of hemicelluloses from process streams from thermomechanical pulping of spruce. Appl. Biochem. Biotechnol. 2007, 136140, No. 10.1007/s12010-007-9094-7. (35) al Manasrah, M. Recovery of hemicelluloses from wood hydrolysates by membrane ﬁltration. Master’s Thesis, Lappeenranta University of Technology, Finland, 2008. (36) Persson, T.; Jonsson, A.-S. Isolation of hemicelluloses by ultraﬁltration of thermomechanical pulp mill process water—Inﬂuence of operating conditions. Chem. Eng. Res. Des. 2010, 88, No. 10.1016/j. cherd.2010.04.002. (37) Maartens, A.; Jacobs, E. P.; Stewart, P. UF of pulp and paper eﬄuent: membrane fouling-prevention and cleaning. J. Membr. Sci. 2002, 209, No. 10.1016/S0376-7388(02)00266-1. (38) Persson, T.; Jonsson, A.-S.; Zacchi, G. Fractionation of hemicelluloses by membrane ﬁltration. In 14th European Biomass Conference and Exhibition, Paris, France, October 1721, 2005.

9749

dx.doi.org/10.1021/es202180m |Environ. Sci. Technol. 2011, 45, 9743–9749

ARTICLE pubs.acs.org/est

Impaired Gas Bladder Inflation in Zebrafish Exposed to a Novel Heterocyclic Brominated Flame Retardant Tris(2,3-dibromopropyl) Isocyanurate Juan Li,† Yong Liang,*,‡,§ Xu Zhang,† Jingyi Lu,‡ Jie Zhang,† Ting Ruan,§ Qunfang Zhou,§ and Guibin Jiang§ †

Key Laboratory of Subtropical Agriculture and Environment, Ministry of Agriculture, Huazhong Agricultural University, Wuhan 430070, China ‡ School of Medicine, Jianghan University, Hubei Province 430056, China § State Key Laboratory of Environmental Chemistry and Ecotoxicology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China ABSTRACT: The teleost gas bladder is a gas-ﬁlled internal organ that processes gas exchange and controls buoyancy. Here we report that an emerging heterocyclic brominated ﬂame retardant, tris(2,3dibromopropyl) isocyanurate (TBC), causes defects in the inﬂation of the gas bladder of zebraﬁsh larvae. This could cause impaired motility, which can ultimately lead to their death. Exposure to zebraﬁsh embryos revealed that TBC had the most signiﬁcant inﬂuence on the larvae at 7296 h postfertilization, which coincided with the time that the gas bladder ﬁrst inﬂates. Critical factors involved in early zebraﬁsh gas bladder development remained at normal levels, which indicated that TBC caused defects in the inﬂation of the gas bladder without disrupting early organogenesis. However, the ultrastructure of the gas bladder was altered in the TBC-treated groups: the electron density of cytoplasmic vesicles was changed and the mitochondria were damaged. We deduce that TBC causes damage to mitochondria that inﬂuences the secretion of mucus-like material, resulting in defects in gas bladder inﬂation. For the ﬁrst time, we report that the gas bladder could be a primary target organ for TBC, and assessment of the gas bladder should be included in toxicity testing protocols of zebraﬁsh embryos.

’ INTRODUCTION Brominated ﬂame retardants (BFRs) are added to many consumer products including carpets, electronic cables, polyurethane foams, television sets, and computers.1 Some of the BFRs are stable in the environment,2 bioaccumulative, capable of longdistance transport, and potentially harmful to ecosystems and human health.3 Although use of polybrominated diphenyl ethers (PBDEs) has declined due to bans and restrictions in the developed world, other groups of alternative BFRs, such as tetrabromobisphenol A (TBBPA) and hexabromocyclododecanes (HBCD), appear to be increasing in the environment.46 Tris(2,3-dibromopropyl) isocyanurate (TBC) is a heterocyclic hexabrominated chemical with high Kow (octanolwater partition coeﬃcient) and Koa (octanolair partition coeﬃcient) values.7 Due to its good ﬂame retardant properties, it is therefore used in certain glass ﬁber reinforced plastics.8 The annual production volume of TBC in China in the 1990s was more than 500 metric tons.9 Recently, TBC has been identiﬁed in soils, sediments, and earthworms around the Liuyang River in southern China. High relative levels of this chemical in biological samples suggested that TBC could bioaccumulate in some species.7 Also, high levels of TBC were detected in the intestine and brain of carp, which implies that this substance can pass r 2011 American Chemical Society

through the bloodbrain barrier.7 Although there are no current statistics on the overall production volume of TBC, increased production volumes are expected due to the enormous demand for electronic products. Furthermore, information about the toxicity of TBC is also limited. The only available data indicate that TBC exposure caused toxic eﬀects on adult zebraﬁsh (Danio rerio).10 Thus, further toxicological evaluation is urgently needed to evaluate the potential risks of this substance to wildlife and humans. Previous observations have shown that some BFRs, such as HBCD, could cause yolk sac edema, pericardial edema, axial spine curvature, and abnormal inﬂation of gas bladder of zebraﬁsh larvae,11 whereas BDE-47 induced dorsal curvature, curved tail, and defects in the inﬂation of the gas bladder of zebraﬁsh larvae.12 As a speciﬁc organ in teleosts, the gas bladder has important functions such as maintaining balance and producing space in the abdomen that protects other visceral organs from injury by external hydraulic pressure.13 In addition to its role in Received: December 13, 2010 Accepted: October 3, 2011 Revised: September 25, 2011 Published: October 03, 2011 9750

dx.doi.org/10.1021/es202420g | Environ. Sci. Technol. 2011, 45, 9750–9757

Environmental Science & Technology buoyancy, the gas bladder can provide oxygen when the ﬁsh is in an anoxic condition.14 Therefore, the gas bladder plays a vital role in maintaining the normal physiological activities of teleosts. This study was designed to investigate whether the gas bladder is a primary target organ of TBC in ﬁsh and, if so, what is the potential toxic mechanism of TBC on failed gas bladder inﬂation.

ARTICLE

Table 1. Primer Sequences for Real-Time PCR Used in This Study gene

’ MATERIALS AND METHODS

primer sequence

β-actin

β-actin F CAACAGAGAGAAGATGACACAGATCA

prolactin

β-actin R GTCACACCATCACCAGAGTCCATCAC prl F GTGGCTATTTTGATGTGTGC prl R

somatolactin

Chemicals. Tris(2,3-dibromopropyl) isocyanurate (TBC,

97% purity) was purchased from SigmaAldrich (St Louis, MO). Stock solutions (1 and 2.5 μg/mL), prepared by dissolving TBC in dimethyl sulfoxide (DMSO, Amresco), were stored at 4 °C. All the chemicals used in this study were analytical-grade. Animal Model. AB strain (wild type) zebrafish were kept under a 12-h light and 12-h dark photoperiod at 26 ( 1.5 °C. The fish were fed with live brine shrimp (Artemia nauplii) twice daily. Zebrafish were maintained and embryos were collected from tanks according to the procedures described by Westerfield.15 Normal fertilized eggs at 2 h postfertilization (hpf) were collected for exposure experiments after visual inspection with a stereomicroscope. TBC Exposure. The stock solution was diluted in water to create a graded series of TBC solutions. Zebrafish embryos were exposed to TBC at 0.5, 1, 2.5, 5, and 10 μg/mL with the final concentrations of DMSO in water less than 0.4%, or they were exposed to water containing 0.4% DMSO only (vehicle control). An Alliance 2695 model HPLC system (Waters, Milford, MA) equipped with a degasser and a quadruple pump was used for instrument analysis. To determine the actual concentration of TBC in the aqueous treatments, the TBC exposure solution in the 0, 1, 5, and 10 μg/mL groups was analyzed according to the method described by Ruan et al.7 Each exposure group was analyzed in triplicate. The embryos at 2 hpf were prepared for experimental treatments and were exposed to TBC or the vehicle control until 168 hpf. In addition, different stages of exposure were carried out in this study. The zebrafish embryos were exposed to TBC at 22, 46, 70, and 94 h intervals beginning at 2 hpf. Meanwhile, the embryos were also exposed to TBC at 166, 144, 120, 96, 72, and 48 h intervals beginning at 2, 24, 48, 72, 96, and 120 hpf respectively (Figure 3a). After each exposure, embryos from all treatments were removed, rinsed with water for 1 min over a plastic screen, and then transferred to clean water and observed periodically for development of the phenotype. For all exposure experiments, embryos and larvae were maintained at 26 ( 0.5 °C and reared in sterile 12 well cell culture plates (Corning International, Corning, NY) at a density of 20 embryos per well, each well containing 3 mL of treatment solution. Experiments were performed at least in triplicate, and the treatment solution of each well was replaced with fresh reagent every 24 h. At the end of each exposure period, the larvae were euthanized by an overdose of MS-222, and morphological observation was carried out under a stereomicroscope. Glass slides were used to hold the larvae so that the exact number of larvae that had defective gas bladders could be counted under the stereomicroscope. The scoring of the lethal and malformed end points was according to the methods described by Nagel16 and Hermsen et al.17 The gas bladder of zebrafish consists of posterior and anterior chambers that inflate at 4.5 and 21 days postfertilization, respectively.13 The region of the gas bladder hereinafter referred to in this paper is the posterior chamber. This study involving zebrafish larvae was conducted in accordance

primer name

Sonic Hedgehog

TTGGTGAGTGAGGTGCTGAG

sl-β F

GGAGTGTCCAGACCAAGAG

sl-β R

CCGAGAAGCGGTAAATGAG

shha F

TGTTTCCCAGGGTTCG

shha R

GGGTTCTTGCGTTTC

Indian Hedgehog ihha F

GCTCACGCCGAACTACAA

ihha R

TGCCGTCTTCATCCCAAC

with national and institutional guidelines for the protection of human subjects and animal welfare. Total RNA Extract. For each exposure experiment, zebrafish embryos were exposed to 0, 1, and 10 μg/mL TBC in two 12-well plates with 20 embryos per well, and each set of 40 embryos (from two wells) was pooled for RNA preparation. The larvae were washed twice in diethyl pyrocarbonate-treated water, and the total RNA was extracted from 40 homogenized zebrafish larvae (exposure durations were 072 and 096 hpf, respectively) by use of Trizol Reagent (Invitrogen) according to the manufacturer’s instructions. The purity of the total RNA was measured by a UV spectrophotometer (BioPhotometer plus, Eppendorf, Germany). The quality and integrity of the total RNA were verified by measuring the 260/280 nm ratios and by electrophoresis on a 1% agarose formaldehyde gel. The mortality, hatch rate, and gas bladder deformity rate were determined before the larvae were used for RNA extraction. Real-Time Polymerase Chain Reaction. Zebrafish larvae from 0, 1, and 10 μg/mL TBC exposure groups were chosen for gene detection. In this study, 2 μg of total RNA was reversetranscribed to cDNA by use of MMLV reverse transcriptase (Promega, Madison, WI) according to the manufacturer’s instructions. The primer sequences of β-actin, somatolactin (sl-β), prolactin (prl), Indian Hedgehog (ihha), and Sonic Hedgehog (shha) were designed with the assistance of the computer software Premier 5.0. All sequences of the primer are shown in Table 1. Each PCR reaction mixture in a total volume of 20 μL contained 1 μL of cDNA template, 0.1 μM primers, Milli-Q water, and 10 μL of 2 SYBR QPCR Master Mix (Toyobo). The thermal cycling program consisted of a denaturing step (94 °C for 5 min) followed by 45 cycles of denaturation (94 °C for 20 s), annealing (55 °C for 20 s), and extension (72 °C for 40 s) in a PTC-200 thermal cycler equipped with a Chromo4 real-time fluorescence detector (MJ Research). The gene expression levels were measured in four replicates for each treatment. The cycle threshold (CT) value was obtained from Opticon Monitor 3.0 software. The target gene expression level was normalized to that of β-actin. The x-fold change of the tested genes was analyzed by the 2ΔΔCt method.18 Transmission Electron Microscopy. Three zebrafish larvae were chosen randomly from each of the vehicle control, 1, and 10 μg/mL TBC treated groups for ultrastructural analysis. The whole zebrafish larvae exposed to TBC for 104 hpf were fixed in a solution of 2.5% glutaraldehyde adjusted to pH 7.4 with 0.1 M phosphate buffer, postfixed in 2% OsO4 in the same buffer, and then dehydrated and embedded in epoxy resin according to 9751

dx.doi.org/10.1021/es202420g |Environ. Sci. Technol. 2011, 45, 9750–9757

Environmental Science & Technology

ARTICLE

Figure 1. Toxic eﬀects of TBC on zebraﬁsh larvae following embryonic exposure from 0 to 168 hpf. (a) Mortality; (b) hatching rate. Each bar represents the mean ( SEM *Signiﬁcant diﬀerent from control group. P e 0.05.

Figure 2. Micrographs showing failed inﬂation of gas bladder induced by TBC exposure. (a) Top view and (c) lateral view of control (c) at 120 h postfertilization (hpf). (b, d) Lateral view of 10 μg/mL TBC-exposed larvae at 120 h postfertilization (hpf). Higher magniﬁcation views in panels c and d show the normal gas bladder and the failed inﬂation of the gas bladder, respectively. Arrows points to gas bladder of zebraﬁsh larvae. Scale bars = 300 μm.

Luft.19 Ultrathin sections obtained with an ultramicrotome (LKB-II, Sweden) were stained with uranyl acetate and lead citrate. The observations and recording of images were performed with an FEI TECNAI-G2 transmission electron microscope at 200 KV. Statistical Analysis. Mortality, hatch rate, and deformation rate were compared statistically between TBC and the control by use of a χ2 test. LC50 (95% CI) and EC50 (95% CI) were calculated by the trimmed SpearmanKarber test. For other time course experiments, one-way analysis of variance (ANOVA) and Tukey’s multiple range tests were used to determine whether the differences between the control and TBC exposure groups were significant. SPSS 13.0 and Origin 7.5 were used for the statistical analysis. For all experiments, p e 0.05 was used to determine significance.

’ RESULTS Quantification of TBC in Exposure Concentration. Liquid chromatography/mass spectrometry (LC/MS) was used to determine the actual concentration of TBC present in the aqueous treatments. The 0, 1, 5, and 10 μg/mL TBC exposure solutions were analyzed, and results were 0, 0.80 ( 0.16, 4.61 ( 0.73 and 9.58 ( 0.31 μg/mL, respectively (n = 3), which indicated that TBC was well dissolved in the aqueous treatment with DMSO.

Toxic Effects of TBC on Zebrafish Larvae. Mortality and hatching rate of TBC-treated embryos are shown in Figure 1. Exposure to TBC caused high mortality of larvae at the longer exposure periods (144 and 168 hpf) (Figure 1a). Larvae in 5 and 10 μg/mL TBC treatment groups showed significantly higher mortality than that of the control groups at 120 hpf. There was a significant difference in mortality between 2.5, 5, and 10 μg/mL TBC treatments and the control group after 144 and 168 hpf, and the values of LC50 were 6.85 ( 0.42 and 2.66 ( 0.23 μg/mL at 144 and 168 hpf, respectively. However, during the exposure duration from 0 to 168 hpf, there was a significant change in mortality between the 0.5 and 1 μg/mL TBC treatments and control groups (Figure 1a). Hatching rate declined significantly after TBC exposure 72 hpf (Figure 1b). Morphological Change of Zebrafish Larvae following TBC Exposure. Larvae exposed to TBC from 0 to 168 hpf showed highly reproducible defects in the inflation of the gas bladder, which mostly occurred after 96 hpf. Larvae in the control group swam normally and had a normal gas bladder (Figure 2a,c). As shown in Figure 2a,c, the normal gas bladder is oval, transparent, and filled with gas, which is easily detected under the microscope. After TBC exposure, the gas bladder inflation was defective, and instead of being transparent and oval, had the appearance of a “black line”. Less than 10% of larvae exposed to 1 μg/mL TBC exhibited abnormal swimming behavior. However, more than 9752

dx.doi.org/10.1021/es202420g |Environ. Sci. Technol. 2011, 45, 9750–9757

Environmental Science & Technology 80% of larvae lost their swimming ability and lay on their sides at the bottom of the well in the 2.5, 5, and 10 μg/mL TBC treatments. Because of the “side lying” phenomenon, we could not capture the top view of those larvae, so the lateral view of larvae that did not have inflated gas bladders was recorded (Figure 2b,d). Apart from the noninflation effect on the gas bladder, TBC did not induce other toxic effects, such as yolk sac edema, pericardial edema, or curved tail. The failed inflation of the gas bladder was dose-dependent and a significant increase was found in the 2.5, 5, and 10 μg/mL TBC-treated groups compared with the control group (Figure 3). The corresponding EC50 values for 96, 120, 144, and 168 hpf were 1.69 ( 0.11, 1.70 ( 0.13, 1.71 ( 0.09, and 1.72 ( 0.08 μg of TBC/mL, respectively. Toxic Window of TBC on Defects in the Inflation of Zebrafish Larvae Gas Bladders. To observe the narrow window

Figure 3. Failure in inﬂation of gas bladder in zebraﬁsh larvae following TBC exposure from 0 to 168 hpf.

ARTICLE

when TBC affected zebrafish larvae, we exposed them at various time frames (Figure 4a). At 5 and 10 μg/mL TBC exposure, gas bladder malformation occurred at exposure periods of 0168, 24168, 48168, and 72168 hpf (Figure 4c), and a positive correlation was found between malformation rate and exposure time. However, in the 5 and 10 μg/mL TBC treatments, the malformation rate decreased significantly during the exposure period of 96168 hpf and disappeared at 120168 hpf. A similar trend of malformation rate was observed, where no effect was found during 024, 048, and 072 hpf but significant effects were found in the 096 hpf exposure experiment (Figure 4b). Therefore, the most sensitive period during which TBC affects the gas bladder seems to be from 72 to 96 hpf, while no significant impact was observed on the gas bladder in the other periods. On the basis of the above observations, we further investigated whether TBC would have toxic effects on zebrafish larvae when the gas bladder was completely inflated. Zebrafish larvae were exposed to TBC from 120 to 288 hpf. The failed inflation rate in the 0, 1, 5, and 10 μg/mL TBC treatment groups was 0%, 0%, 0%, and 1.7%, respectively; and no mortality was observed for either the control or TBC treatment groups Gene Expression of Zebrafish Larvae following TBC Exposure. The larvae used for RNA extraction were affected by TBC exposure in a similar way to those used for morphological observation. We chose the time point between 72 hpf and 96 hpf to investigate the expression of gas bladder development related genes, including prl, sl-β, shha, and ihha. The expression of both prl and sl-β genes, which are members of the growth hormone and prolactin (GH/PRL) superfamily and play vital roles in gas bladder development, was detectable at 72 and 96 hpf in both the vehicle and TBC-treated groups. Neither prl nor sl-β showed any difference between the TBC-exposed group and the control group (Figure 5a, b). Both shha and ihha are crucial genes in the

Figure 4. (a) Exposure methods (based on the TBC exposure durations). (b, c) Defects in the inﬂation of the gas bladder, detected at the given end point (168 hpf). 9753

dx.doi.org/10.1021/es202420g |Environ. Sci. Technol. 2011, 45, 9750–9757

Environmental Science & Technology

ARTICLE

Figure 5. Gene expression in the GH/PRL superfamily and Hedgehog pathways under TBC treatment. Real-time PCR results of the tested genes in the control group and exposed to TBC groups until 72 and 96 hpf: (a) prl; (b) sl-β; (c) shha; (d) ihha. Each set of 40 embryos was pooled for RNA preparation, n = 4.

Hedgehog (hh) pathway, and the expression of both ihha and shha genes in TBC-treated larvae showed no significant difference compared with the control (Figure 5c,d). This result suggests that induction of severe malformations in the gas bladder by TBC was independent of the GH/PRL and hh gene pathways. TBC Induced Changes in Ultrastructure of Zebrafish Gas Bladder. The ultrastructure of the zebrafish gas bladder was examined by transmission electron microscopy (TEM), and by imaging directly through the cross section of the larval gas bladder (104 hpf). In the control group, a large number of cytoplasmic vesicles with an electron lucent appearance were found in the epithelial lining of the gas bladder (Figure 6ac). Some of the cytoplasmic vesicles were present in the lumen of the gas bladder (Figure 6c). The mitochondria of the control group had a normal appearance with intact cristae (Figure 6 b). However, cytoplasmic vesicles were atrophied and electron-dense in the 1 μg/mL TBC treated group (Figure 6df), and this phenomenon was more severe in the 10 μg/mL TBC group (Figure 6g, h). Furthermore, the mitochondrial cristae were disrupted in the 1 and 10 μg/mL TBC treated groups. This resulted in the cells having a vacuolar appearance and mitochondria with myelin figures (Figure 6d,fi). In addition, intact smooth muscle could be found in all experimental groups, which implied TBC might have no effect on muscle formation in the gas bladder.

’ DISCUSSION The gas bladder, commonly referred to as the swim bladder, is a vascularized internal organ that is found in almost all teleosts. It

controls the buoyancy and locomotory movements of ﬁshes.20,21 In this study, the most conspicuous development defect caused by TBC treatment was the failure of the gas bladder to inﬂate. This is similar to the eﬀect of some other BFRs such as HBCD and BDE-47.11,12 Results in this study showed that 0168 hpf TBC exposure caused defects in the inﬂation of the gas bladder of zebraﬁsh larvae, resulting in the impairment of free motility. The yolk sac, which provides nutritive material vital for the movement of larvae, plays an important role in the early development of zebraﬁsh.22 Inﬂation of the zebraﬁsh gas bladder began at around 72 hpf, coinciding with the time of yolk sac reabsorption and the beginning of feeding.23 Once gas bladder inﬂation occurred, zebraﬁsh larvae swam freely and began to seek nutrition. Our results showed that treatment with 0.5, 1, 2.5, 5, and 10 μg/mL TBC for 0168 hpf caused mortality of zebraﬁsh larvae in a dosedependent manner. We hypothesized that zebraﬁsh larvae are unable to feed themselves after TBC exposure, ﬁnally leading to their death once the yolk sac is depleted. In this study, TBC exposure (0168 hpf) induced gas bladder malformation in zebraﬁsh larvae at 96 hpf, with a 96 hpf EC50 of 1.69 ( 0.11 μg of TBC/mL. TBC exposure did not cause high mortality until 144 hpf, with a 144 hpf LC50 value of 6.85 ( 0.42 μg of TBC/mL. This means that the failure of gas bladder inﬂation appear earlier than the death of larvae in our study, suggesting failed gas bladder inﬂation is a sensitive index to evaluate the toxic eﬀects of TBC. With the increase of mortality, a decreasing EC50 would be expected, but results in this study showed that the EC50 values remained almost constant for all exposure periods, which indicated 9754

dx.doi.org/10.1021/es202420g |Environ. Sci. Technol. 2011, 45, 9750–9757

Environmental Science & Technology

ARTICLE

Figure 6. Ultrastructure of zebraﬁsh gas bladder after TBC exposure for 104 h. Numerous vesicles (V), mitochondria (M), myelin ﬁgure (F), and lumen (L), are shown. (c) Some of the vesicles have been released into the lumen (arrow). (f, i) Mitochondrial damage with the presence of myelin ﬁgures.

that there might be a deﬁned time window of toxicity. In addition, the hatch rate of TBC-treated larvae was signiﬁcantly lower at 72 hpf compared with the control group, indicating TBC might aﬀect early development of zebraﬁsh larvae. The results from diﬀerent exposure methods showed that when the exposure duration was from 0 to 72 hpf, the ﬁsh did not show obvious gas bladder defects, whereas when exposure duration was from 0 to 96 hpf, zebraﬁsh larvae exhibited abnormal gas bladders in the 5 and 10 μg/mL TBC exposure groups. When the exposure duration was from 72 to 168 hpf, zebraﬁsh larvae exhibited severe defects in their gas bladders, while exposure periods from 96 to 168 hpf resulted in few abnormalities in the gas bladders of ﬁsh treated with 10 μg/mL TBC. On the basis of these results, we deduce that the toxic window where TBC aﬀects the gas bladders of zebraﬁsh larvae is from 72 to 96 hpf, and this is consistent with development of the zebraﬁsh gas bladder, which begins to inﬂate at around 72 hpf. It has been reported that the posterior chamber of the gas bladder of zebraﬁsh is completely inﬂated at about 120 hpf, after which zebraﬁsh larvae can swim freely and feed themselves.13,23 Notably, when exposure duration was 120168 hpf and 120288 hpf in the present experiment, we found neither defects in gas

bladder nor death of larvae, suggesting that TBC has little eﬀect as long as the gas bladder has been completely inﬂated. This result strongly enhances our ﬁnding that the gas bladder is a target organ of TBC. The gas bladder of zebraﬁsh is made up of an epithelial, mesodermal, and outer mesodermal layer from the lumen to stratum externum, and the mesenchyme diﬀerentiates into the muscle layer.13 Hedgehog genes in the epithelia, such as shha and ihha, play an essential role in the formation and organization of all the three tissue layers of the swim bladder.13 The knockdown of shha and ihha genes led to failed inﬂation of the gas bladder at 72 hpf. Knockdown of ihha resulted in mild reduction of the epithelium and mesenchyme, absence of smooth muscle diﬀerentiation, and a reduction of the outer mesothelium. Similarly, shha knockdown caused prominent reduction of the epithelium, disorganization of the mesenchyme in the absence of smooth muscle diﬀerentiation, and disorganization of the outer mesothelium at 72 hpf.13 Members of the GH/PRL superfamily, such as prl and sl-β, play an important role in the development of the gas bladder of zebraﬁsh by regulating growth and osmoregulation. In addition, prl and sl-β knockdown led to failure of the gas bladder to inﬂate.24 The expression of those four genes showed no signiﬁcant 9755

dx.doi.org/10.1021/es202420g |Environ. Sci. Technol. 2011, 45, 9750–9757

Environmental Science & Technology diﬀerence between TBC-treated and control groups, indicating the defect in inﬂation of the gas bladder in TBC-treated larvae was independent of Hedgehog and GH/PRL gene pathways. Combined with our result that perfect smooth muscle was found in the gas bladder, it is suspected that TBC has a direct eﬀect on the inﬂation of the gas bladder instead of inﬂuencing its early formation and development. Teleosts are classiﬁed as physostomes or physoclists according to their gas bladder structure, and zebraﬁsh belong to physostomes.25 Brooks26 noted that there was mucus-like material along the surface of the physostomatous swim bladder. Such material, which originates from cytoplasmic vesicles and is commonly referred to as surfactant production, enables the ﬁsh to maintain the surface tension of the gas bladder and protect microbubbles from collapse until they are released into the lumen of the gas bladder.24 Furthermore, the ﬁnal act of gas secretion into the gas bladder occurs by formation of small cytoplasmic bubbles in the secretory epithelium coupled with relatively quick release into the lumen of the gas bladder.27,28 Thus, the cytoplasmic vesicles are closely related to the secretion of mucuslike material, which in turn is closely related to the surface tension of the gas bladder. Our TEM results showed a large amount of electron lucent cytoplasmic vesicles in the epithelium of the control group, and we also found that mucus-like material in cytoplasmic vesicles was released into the lumen of the gas bladder. In the TBC exposure groups, the cytoplasmic vesicles were electron-dense, which suggested that the mucus-like material might still reside in them. Mitochondria are the site for mitochondrial respiration and ATP synthesis. It is known that the formation of myelin ﬁgures is one of the indicators of mitochondrial damage, and the presence of myelin ﬁgures in mitochondria in TBC exposure groups indicated that mitochondrial elimination was ongoing in the epithelium of the gas bladder. Our data showed that TBC caused cytological damage to the gas bladder, including a decrease of cytoplasmic vesicles, change in the electron density of cytoplasmic vesicles, and mitochondrial damage. However, how exactly TBC gains access to and induces mitochondrial damage is currently unknown, and further study is needed. In this study, we have reported that TBC caused damage to the mitochondria of gas bladder epithelial cells, which might block the energy acquirement of the cells to release the mucus-like material from the cytoplasmic vesicles into the lumen of the gas bladder. This would result in a shortage of the surfactants that usually maintain the surface tension of the gas bladder and lead to defects in the inﬂation of the gas bladder as well as inhibition of larval motility, causing the ﬁsh to fail in swimming into surface waters to acquire food. As a result, zebraﬁsh larvae could die of starvation. The results from this study also indicate that the teleost gas bladder is an important target organ to evaluate toxic eﬀects and should be added as an end point in test protocols for zebraﬁsh embryos.

’ AUTHOR INFORMATION Corresponding Author

*Telephone: +86 27-8423-8886; e-mail: [email protected].

’ ACKNOWLEDGMENT We thank the National Natural Science Foundation of China (20890112, 20907017), the Major State Basic Research Development

ARTICLE

Program of China (973 Program) (2009CB421605), National Key Technology R&D Program of China (2007BAC27B01), and the State Key Laboratory of Environmental Chemistry and Ecotoxicology, RCEES, CAS (KF2009-05) for funding during the study.

’ REFERENCES (1) 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 US house dust. Environ. Sci. Technol. 2008, 42 (18), 6910–6916. (2) Birnbaum, L. S.; Staskal, D. F. Brominated ﬂame retardants: cause for concern?. Environ. Health.Perspect. 2004, 112 (1), 9. (3) De Wit, C. A. An overview of brominated ﬂame retardants in the environment. Chemosphere. 2002, 46 (5), 583–624. (4) Thomsen, C.; Lundanes, E.; Becher, G. Brominated ﬂame retardants in archived serum samples from Norway: a study on temporal trends and the role of age. Environ. Sci. Technol. 2002, 36 (7), 1414–1418. (5) Zhang, X. L.; Luo, X. J.; Chen, S. J.; Wu, J. P.; Mai, B. X. Spatial distribution and vertical proﬁle of polybrominated diphenyl ethers, tetrabromobisphenol A, and decabromodiphenylethane in river sediment from an industrialized region of South China. Environ. Pollut. 2009, 157 (6), 1917–1923. (6) Lam, J. C. W.; Lau, R. K. F.; Murphy, M. B.; Lam, P. K. S. Temporal trends of hexabromocyclododecanes (HBCDs) and polybrominated diphenyl ethers (PBDEs) and detection of two novel ﬂame retardants in marine mammals from Hong Kong, South China. Environ. Sci. Technol. 2009, 43 (18), 6944–6949. (7) Ruan, T.; Wang, Y. W.; Wang, C.; Wang, P.; Fu, J. J.; Yin, Y. G.; Qu, G. B.; Wang, T.; Jiang, G. B. Identiﬁcation and evaluation of a novel heterocyclic brominated ﬂame retardant tris(2,3-dibromopropyl) isocyanurate in environmental matrices near a manufacturing plant in southern China. Environ. Sci. Technol. 2009, 43 (9), 3080–3086. (8) Xiong, X. Y. Microencapsulated ﬂame retardant of TBC and its implication. Flame Retard. Mater. Technol. 1999, 3, 1–3 (in Chinese). (9) Cao, J. The developmental trend of plastics additives. China Chem. Ind. 1996, 8, 48–50 (in Chinese). (10) Zhang, X.; Li, J.; Chen, M. J.; Wu, L.; Zhang, C.; Zhang, J.; Zhou, Q. F.; Liang, Y. Toxicity of the brominated ﬂame retardant tris-(2,3-dibromopropyl) isocyanurate in zebraﬁsh (Danio rerio). Chin. Sci. Bull. 2011, 56, 1548–1555. (11) Deng, J.; Yu, L.; Liu, C.; Yu, K.; Shi, X.; Yeung, L. W. Y.; Lam, P. K. S.; Wu, R. S. S.; Zhou, B. S. Hexabromocyclododecane-induced developmental toxicity and apoptosis in zebraﬁsh embryos. Aquat. Toxicol. 2009, 93 (1), 29–36. (12) Lema, S. C.; Schultz, I. R.; Scholz, N. L.; Incardona, J. P.; Swanson, P. Neural defects and cardiac arrhythmia in ﬁsh larvae following embryonic exposure to 2,20 ,4,40 -tetrabromodiphenyl ether (PBDE 47). Aquat. Toxicol. 2007, 82 (4), 296–307. (13) Winata, C. L.; Korzh, S.; Kondrychyn, I.; Zheng, W.; Korzh, V.; Gong, Z. Development of zebraﬁsh swimbladder: The requirement of Hedgehog signaling in speciﬁcation and organization of the three tissue layers. Dev. Biol. 2009, 331 (2), 222–236. (14) Todd, E. S. Positive buoyancy and air-breathing: a new piscine gas bladder function. Copeia 1973, 461–464. (15) Westerﬁeld, M. The Zebraﬁsh Book: A Guide for the Laboratory Use of Zebraﬁsh (Brachydanio rerio); University of Oregon Press: Eugene, OR, 1995. (16) Nagel, R. DarT: The embryo test with the zebraﬁsh Danio rerio--a general model in ecotoxicology and toxicology. Altex 2002, 19, 38. (17) Hermsen, S. A. B.; Brandhof, E. J.; Van der Ven, L. T.; Piersma, A. H. Relative embryotoxicity of two classes of chemicals in a modiﬁed zebraﬁsh embryotoxicity test and comparison with their in vivo potencies. Toxicol. In Vitro 2011, 25, 745–753. (18) Ding, L.; Murphy, M. B.; He, Y.; Xu, Y.; Yeung, L. W. Y.; Wang, J.; Zhou, B. S.; Lam, P. K. S.; Wu, R. S. S.; Giesy, J. P. Eﬀects of brominated ﬂame retardants and brominated dioxins on steroidogenesis 9756

dx.doi.org/10.1021/es202420g |Environ. Sci. Technol. 2011, 45, 9750–9757

Environmental Science & Technology

ARTICLE

in H295R human adrenocortical carcinoma cell line. Environ. Toxicol. Chem. 2007, 26 (4), 764–772. (19) Luft, J. H. Improvements in epoxy resin embedding methods J. Biophys. Biochem. Cytol. 1961, 9 (2), 409. (20) Fange, R. Gas exchange in ﬁsh swim bladder. Rev. Physiol. Biochem. Pharmacol. 1983, 97, 111–158. (21) Teoh, P. H.; Shu-Chien, A. C.; Chan, W. K. Pbx1 is essential for growth of zebraﬁsh swim bladder. Dev. Dyn. 2010, 239 (3), 865–874. (22) Liu, Y. W.; Chan, W. K. Thyroid hormones are important for embryonic to larval transitory phase in zebraﬁsh. Diﬀerentiation 2002, 70 (1), 36–45. (23) Robertson, G. N.; McGee, C. A.; Dumbarton, T. C.; Croll, R. P.; Smith, F. M. Development of the swimbladder and its innervation in the zebraﬁsh, Danio rerio. J. Morphol. 2007, 268 (11), 967–985. (24) Zhu, Y.; Song, D.; Tran, N. T.; Nguyen, N. The eﬀects of the members of growth hormone family knockdown in zebraﬁsh development. Gen. Comp. Endocrinol. 2007, 150 (3), 395–404. (25) Finney, J. L.; Robertson, G. N.; McGee, C. A. S.; Smith, F. M.; Croll, R. P. Structure and autonomic innervation of the swim bladder in the zebraﬁsh (Danio rerio). J. Comp. Neurol. 2006, 495 (5), 587–606. (26) Brooks, R. E. Ultrastructure of the physostomatous swimbladder of rainbow trout (Salmo gairdneri). Cell. Tissue. Res. 1970, 106 (4), 473–483. (27) Wittenberg, J. B. The secretion of inert gas into the swimbladder of ﬁsh. J. Gen. Physiol. 1958, 41 (4), 783. (28) Copeland, D. E. Fine structural study of gas secretion in the physoclistous swim bladder of Fundulus heteroclitus and Gadus callarias and in the euphysoclistous swim bladder of Opsanus tau. Cell Tissue Res. 1969, 93, 305–331.

9757

dx.doi.org/10.1021/es202420g |Environ. Sci. Technol. 2011, 45, 9750–9757

ARTICLE pubs.acs.org/est

Combined NitritationAnammox: Advances in Understanding Process Stability Adriano Joss,*,† Nicolas Derlon,† Clementine Cyprien,† Sabine Burger,‡ Ilona Szivak,† Jacqueline Traber,† Hansruedi Siegrist,† and Eberhard Morgenroth†,§ †

Eawag, Swiss Federal Institute of Aquatic Science and Technology, Ueberlandstr. 133, 8600 Duebendorf, Switzerland ERZ, Entsorgung und Recycling Stadt Z€urich, WWTP Werdh€olzli, B€andlistrasse 108, 8010 Zurich, Switzerland § ETH Z€urich, Institute of Environmental Engineering, 8093 Z€urich, Switzerland ‡

bS Supporting Information ABSTRACT: Eﬃcient nitrogen removal from wastewater containing high concentrations of ammonium but little organic substrate has recently been demonstrated by several full-scale applications of the combined nitritationanammox process. While the process eﬃciency is in most cases very good, process instabilities have been observed to result in temporary process failures. In the current study, conditions resulting in instability and strategies to regain eﬃcient operation were evaluated. First, data from full-scale operation is presented, showing a sudden partial loss of activity followed by recovery within less than 1 month. Results from laboratory-scale experiments indicate that these dynamics observed in full scale can be caused by partial inhibition of the ammonia oxidizing bacteria (AOB), while anammox inhibition is a secondary eﬀect due to temporarily reduced O2 depletion. Complete anammox inhibition is observed at 0.2 mg O2 3 L1, resulting in NO2 accumulation. However, this inhibition of anammox is reversible within minutes after O2 depletion. Thus, variable AOB activity was identiﬁed as the key to reactor stability. With appropriate interpretation of the online NH4+ signal, accumulation of NO2 can be detected indirectly and used to signal an imbalance of O2 supply and AOB activity (no suitable online NO2 electrode is currently available). Second, increased abundance of nitrite-oxidizing bacteria (NOB; competing with anammox for NO2) is known as another cause of instability. Based on a comparison of parallel full-scale reactors, it is suggested that an infrequent and short-term increased O2 supply (e.g., for maintenance of aerators) that exceeds prompt depletion of oxygen by AOB may have caused increased NOB abundance. The volumetric air supply as a proxy for O2 supply thus needs to be linked to AOB activity. Further, NOB can be washed out of the system during regular operation if the system is operated at a sludge age in the range of 45 days and by controlling the air supply according to the NO3 concentration in the treated eﬄuent. Early detection of growing NOB abundance while the population is still low can help guide process operation and it is suggested that molecular methods of quantifying NOB abundance should be tested.

’ INTRODUCTION Combined nitritationanammox in a single sequencing batch reactor has been conﬁrmed as an attractive option for nitrogen removal in high-strength wastewater due to the low speciﬁc costs of nitrogen removal1,2 so that many full-scale projects are currently being realized. Now that anammox-based N-removal has been shown to reduce costs and energy consumption compared to denitriﬁcation-based processes, its process stability appears to be the major issue making operators reluctant to opt for this process. The start-up times required at full scale have decreased dramatically in recent years, conﬁrming the progress made in terms of better understanding of the process as well as by empirically testing new operation strategies: Abma et al.3 used a granular reactor to show the feasibility of anammox start-up with an increase of observed performance (0.055 d1) close to the maximum growth rate observed in the lab (0.06 d1 4). Additional full-scale r 2011 American Chemical Society

examples conﬁrm that start-up no longer requires years but can now be regarded as a well-controlled and reliable procedure lasting no more than several months.58 Further aspects currently under active discussion are a comparison of reactor conﬁgurations and process control options: single-stage versus two-stage reactors (i.e., with segregated nitritation and anammox stages), suspended growth versus attached bioﬁlm or granular reactors, selection of sensors required for process control, and a detailed strategy for embedding these sensors in automated process control systems are among the topics most actively pursued both in the laboratories and in Received: July 12, 2011 Accepted: October 7, 2011 Revised: September 29, 2011 Published: October 07, 2011 9735

dx.doi.org/10.1021/es202181v | Environ. Sci. Technol. 2011, 45, 9735–9742

Environmental Science & Technology full-scale operation. According to the authors’ knowledge, such direct comparisons of diﬀerent options are still diﬃcult, since the various treatment schemes are still being optimized. The present work discusses full-scale data of suspended sludge nitritationanammox reactors with the help of lab experiments as a contribution to understanding process stability issues. Several options for improved process control in treatments combining nitritation and anammox in a single reactor are presented. We aim to contribute to the applicability of the process as well as providing a sound basis for discussing the suitability of alternative processes.

’ MATERIALS AND METHODS Pilot and Full-Scale Sequencing Batch Reactor (SBR) Equipment. The pilot and full-scale sequencing batch reactor

equipment used was compatible and consists of a fully stirred reactor, a stirrer, a feed pump, a decanter, and an aeration unit. The process is controlled by a programmable logical controller equipped with online sensors for the water level, ammonium, nitrate, volumetric airflow control to the aeration unit, soluble oxygen, temperature, pH, and conductivity. The SBR cycle always comprises a feeding phase, an aeration phase, a mixing phase, a sedimentation phase, and a discharge phase; at full scale, a pause of up to several days was included between the discharge phase and the subsequent feeding phases to adapt to the incoming load, while the pilot-scale reactor was operated without pausing. A complete cycle typically lasts between 6 and 10 h under normal conditions, depending primarily on the amount of supernatant to be treated, parameter settings like the aeration rate, and the sludge conditions. During start-up or a washout phase of nitrogen oxidizing bacteria (NOB), a cycle can also last up to several days. Full-scale data are taken from the nitritation anammox reactor at the wastewater treatment facility (WWTP) of ZurichWerdholzli, which is equipped with two parallel reactors of 1400 m3 each. The operational parameters and influent of the two reactors were identical except for one detail: the reactor with growing NOBs was equipped with old aerator membranes, while the other one was fitted with new aerators prior to being put into operation. An aerator-cleaning procedure was consequently performed every 2 weeks only on reactor South (Figure 4): each cleaning event resulting in 2 h of increased aeration (26002800 m3 3 h1 after dosing with formic acid for scale removal). The pilot experiments were performed in a 400 L reactor fed with supernatant originating from the sludge dewatering plant at the WWTP Zurich-Werdholzli. The sludge originated from the full scale installation at the same WWTP (see ref 6 for more details). For the NOB washout experiment, sludge was taken from the full-scale reactor at WWTP Niederglatt (Switzerland), a 180 m3 reactor equipped identically with the one in Zurich, at a time (December 2009) when the sludge featured high NOB activity. Data Processing. Data processing was performed with Matlab (MathWorks Inc., Natick, USA) by directly accessing the raw data stored by the reactor’s supervisory control and data acquisition (SCADA) system, with a time resolution of 10 s for the pilotscale unit and 1 min for the full-scale installation. The ammonium depletion rate was obtained by linear regression of the ammonium online signal (Endress+Hauser, ionselective ammonium electrode Minical NAM 760) during the aeration phase, when ammonia oxidation and anammox occur simultaneously directly in the reactor.

ARTICLE

The biomass growth or net activity (N removal) increase rate μ is calculated as follows: rt ln r0 ð1Þ μ¼ t where r0 and rt are the rates of ammonium depletion at the beginning of the cycle and at a time t, respectively. Fluorescent in Situ Hybridization (FISH) and Confocal Laser Scanning Microscopy. Sludge samples, taken regularly over 40 days, were fixed by adding 100 μL of 37% (v/v) stabilized formaldehyde (Sigma Aldrich, Germany) to 1 mL of sludge and incubated on ice for 1 h. The samples were subsequently washed three times by centrifugation for 3 min at 4 °C and 10 000g and resuspended in 1 mL of phosphate-buffered saline solution (PBS pH 7.2). Fixed samples were added to a 1:1 (v/v) mixture of ethanol/PBS and stored at 20 °C. A quantity of 10 μL was then applied to a well of epoxy-masked slides (Marienfeld, LaudaK€onigshofen, Germany), dried for 1530 min at 46 °C and sequentially dehydrated for 3 min in 50%, 80%, and 100% (v/v) ethanol/PBS solutions on ice. Next, 10 μL of hybridization buffer (0.9 M NaCl, 20 mM Tris-HCl pH 8, 0.01% SDS, 35% formamide) and 1 μL of each fluorescent-labeled probe mix were added for staining all bacteria with 30 ng μL1 of Cy5-labeled EUB-mix probe,9,10 30 ng μL1 of Cy3 labeled AOB-mix probe,1114 and 50 ng μL1 of FLUO-labeled NOB-mix11,15 were added to the sample and incubated for 90 min at 46 °C in a humidified chamber (50 mL Greiner tubes). Following hybridization, a washing step was performed for 10 min at 48 °C in a buffer (0.07 M NaCl, 20 mM Tris-HCl (pH 8), 5 mM EDTA) and the slides were dipped for 2 s into ice-cold deionized water. The slides were immediately dried under compressed air and mounted with Citi-fluor AF1 (Citifluor Ltd., London, UK). Oligo-nucleotide probes were obtained from Thermo-Fisher Scientific (Ulm, Germany). For visualization, at least 20 randomly acquired fields of view were recorded by confocal microscopy (Leica, SP5, Germany). Three images per field of view were acquired sequentially by using the 488, 520, and 633 nm laser lines to detect Cy3, Cy5, and FLUO signals, respectively. For one channel the same setting of the detector sensitivity was maintained for every field of view. Around 5% of the 600 images were not considered for image analysis due to overexposure or low biomass. Quantification of FISH Images. Quantification of FISH images consisted in measuring the biosurfaces of each population. The images were treated using ImageJ (http://rsb.info.nih.gov/ij/) and a self-written macro. The threshold value was manually estimated for around 40 images (average = 30). It was then applied to all images to convert them from 8-bit to binary form. The suitability of the threshold was visually checked for each set of images. Biosurfaces of AOB to EUB, NOB to EUB, and AOB to NOB were then calculated. Analytical Methods for Liquid Samples. Commercial photochemical test kits (Hach Lange GmbH, D€usseldorf, Germany, Test LCK303, LCK339, LCK340, LCK342; spectrophotometer type LASA 26) were used to test the accuracy of the online ammonium sensors as well as for the offline measurement of ammonium, nitrite, nitrate, and COD. In the NOB washout experiment, colorimetric nitrite tests with test strips (Nitrite-test, 010 mg NO2N 3 L1, Merck KGaA, Darmstadt, Germany) were used. 9736

dx.doi.org/10.1021/es202181v |Environ. Sci. Technol. 2011, 45, 9735–9742

Environmental Science & Technology

ARTICLE

+

Figure 1. Overlay of a typical online NH4 signal of two complete batches without (December 5, 2008, solid line) and with (December 9, dashed line) NO2 accumulation, respectively. For better comparison, the time scale has been shifted to ﬁt zero to the end of the aeration phase. The gray area indicates the proposed observation time to check for NO2 accumulated during the preceding aeration. The dotted line shows the trigger level for switching to manual operation.

NO2

NH4+

Monitoring Accumulation with the Probe. Since an NO2 online sensor is currently not available, the NO2 concentration at the end of each batch was obtained by interpreting the online NH4+ signal (Figure 1): hereto the NH4+ depletion between 5 and 20 min after switching off the aeration was used, and the stoichiometric factor of 1.32 NO2 per NH4+ consumed by anammox16 was considered. NH4+ depletion requires either soluble O2 (by AOB) or NO2 (by anammox). After the blower is switched oﬀ, oxygen is still available for about 5 min: it takes several minutes to shut down the blowers and to relax the membrane aerator; it takes half a minute for the bubbles to rise to the top of the reactor and less than a minute for the biomass to deplete the typically <0.5 mg O2 3 L1 of soluble oxygen. During the following 20 min, the NH4+ depletion is monitored and assumed to depend principally on depletion of the residual NO2 accumulated in the reactor during the preceding aeration phase. NO2 formation due to heterotrophic denitriﬁcation is substrate-limited due to the low organic content of the inﬂuent and is thus low under normal operating conditions (as conﬁrmed by the constant NO3 values; data not shown). On the basis of screening the online data, a threshold value of 2 mg NH4+N 3 L1 depletion is proposed (corresponding to 2.6 mg NO2N 3 L1 according to the stoichiometry): if it is exceeded at the end of the aeration cycle, this indicates imbalanced O2 supply and NO2 depletion. If this occurs, the control algorithm should trigger an automatic feeding stop and switch to manual operation with reduced aeration.

’ RESULTS AND DISCUSSION Sudden Activity Loss. An event of uncontrolled activity loss and recovery during full-scale operation is shown in Figure 2: normal NH4+ depletion rates are seen until 6th August. On the 7th, the operator noticed a significant reduction of the ammonium depletion rate and consequently reduced the air supply by 40% to maintain typical dissolved oxygen levels during aeration

Figure 2. Ammonium depletion rate during aeration (blue, left axis) and nitrite concentration at the end of each aeration cycle (red, right axis; derived from the NH4+ signal) in the full-scale reactor (North lane) of Z€urich-Werdh€ olzli. One data point for each SBR cycle is plotted. The gray area indicates operation at reduced aeration. The horizontal line at 2.6 mg NO2 N 3 L1 indicates the proposed threshold value for triggering the end of the automatic reactor feeding.

(0.40.6 mgO2 3 L1, data not shown; the air supply was reduced from 0.86 to 0.5 m3air 3 m3reactor 3 h1; combined nitritation anammox is normally operated under oxygen-limiting conditions, so that the activity is proportional to the air supply). Nevertheless, a significant increase of the accumulated NO2 is observed at the end of the aeration phase, indicating significant inhibition of the NO2 depletion by anammox. For about 1.5 months the reactor was operated at reduced aeration and with one batch loading every 2.5 days on average (the rest being stirred idle time; under normal operating conditions, three batches are processed each day). Nevertheless, Figure 2 shows that the actual activity probably recovered faster: the 1.5 months taken to return to the previous operational settings represent a cautious choice of the operator having to handle an unexpected performance problem at full scale. On 17th September, regular operation was resumed and NH4+ depletion rates quickly returned to normal levels. No other interventions were carried out during this period. The direct cause of the activity loss has not yet been identified, but it is speculated that a toxic compound was contained in the influent supernatant. Diﬀerent eﬀects have been discussed to explain how anammox inhibition can occur during combined nitritationanammox. Kindaichi et al.17 describe inhibition of anammox within sections of a bioﬁlm reactor and speculate that organic compounds originating from the biomass may be the cause. Methanol has been shown to irreversibly inhibit anammox18 but is very unlikely to occur in digester supernatant. In presence of degradable organic carbon, denitriﬁers outcompete anammox for nitrite,19,20 but since this is leading to NO2 depletion rather than accumulation, this case is not deemed relevant of the performance loss discussed above. Nitrite toxicity has been discussed.21 Recently Abma et al.3 describe that up to 42 mg NO2-N 3 L1 did not aﬀect process performance of granular sludge. Suspended sludge systems have been successfully started up at nitrite concentrations in the range of 100200 mg NO2-N 3 L1,22 thus proving that nitrite toxicity is not relevant for long-term operation under typical conditions. Sulﬁde has been shown to aﬀect anammox activity <20 mg S2‑ 3 1 23 ; sulﬁde or compounds containing S2 can be contained in L 9737

dx.doi.org/10.1021/es202181v |Environ. Sci. Technol. 2011, 45, 9735–9742

Environmental Science & Technology

Figure 3. Experiment showing reversible inhibition of the anammox activity at bulk liquid oxygen concentrations <0.2 mg O2 3 L1. The gray areas indicate times of complete (left) or partial (right) anammox inhibition. The vertical red arrows show the addition of allylthiourea (ATU) with indication of the totalized soluble concentration in mg 3 L1; the vertical black arrows show nitrite addition with numerical indication of the mg NO2-N 3 L1 added at each event. The horizontal arrows in the legend indicate the respective Y scale.

digester supernatant (e.g., in bio-P plants where Fe3+ is not used to precipitate phosphorus). Partial Inhibition of AOB May Explain the Activity Loss. Pilot-scale experiments in the 400 L SBR reactor as shown in Figure 3 confirm that partial inhibition of the AOB can result in NO2 accumulation and thus explain the temporary activity loss shown in Figure 2. The chronology of the experiment of Figure 3 is given below. Prior to the experiment, the pilot SBR was operated continuously for several months with a normal process rate. • Time 12.45: start of a normal SBR cycle by feeding with untreated supernatant leading to the rise of the NH4+ concentration to ca. 137 mg NH4+-N 3 L1 • 12.45 to 13.30: regular operation with continuous aeration (at 12 L 3 min1; the higher speciﬁc aeration compared to the full-scale reactor is due to a reactor depth of only 1 m at pilot scale compared to 5.5 m at full scale) leading to normal NH4+ depletion via the nitritation-anammox process at a typical rate of ca. 20 mg N 3 L1 3 h1. • 13.30 to 14.05: stepwise addition of allylthiourea in three aliquots to reach a concentration of 1.25 mg ATU 3 L1. The oxygen increase conﬁrms the inhibition of the AOB leading to reduced depletion of the supplied O2. The very slow NH4+ depletion slope conﬁrms that the activity of the anammox biomass is also reduced. • 14.25 to 14.55: the air supply is reduced in steps from 10 to 2 L 3 min1 to avoid soluble oxygen concentrations >0.5 mg O2 3 L1. NO2 starts accumulating, since its depletion (anammox) is completely inhibited, while its formation (AOB) is only partly inhibited (note that NO2 concentrations were measured using colorimetric test strips with an assumed relative error of approximately 50%). • 15.45 to 16.45: the air supply is completely switched oﬀ. Within less than 20 min, the NH4+ depletion is back to its initial rate, thus conﬁrming that the anammox inhibition disappears as soon as (a) the soluble O2 concentration is

ARTICLE

depleted, and (b) NO2 is available.24,25 So the continued presence of ATU does not interfere with the anammox activity. To conﬁrm this, NO2 is added twice (20 mg NO2N 3 L1 at 16.05 and another 10 mg NO2-N 3 L1 at 16.25). • 16.45 to 17.40: the air supply is gradually increased without the soluble oxygen exceeding 0.2 mg O2 3 L1. The fact that the rate of increase of soluble O2 at an aeration rate of 6 L 3 min1 is now much slower than immediately after addition of the inhibiting ATU (i.e., between 14.05 and 15.00) shows that ATU has been partly degraded within 2.5 h so that the AOB inhibition has diminished signiﬁcantly. The last addition of 10 mgNO2-N 3 L1 at 17.05 conﬁrmed that the anammox activity is limited by the availability of NO2. • 17.40 to 18.05: further increase of aeration leads to complete inhibition of the anammox and thus to NO2 accumulation. • 18.05 onward overnight: reduced aeration (6 L 3 min1) without any NO2 accumulation. • Next day: the reactor returned to its usual aeration rate (12 L 3 min1) with a NH4+ depletion rate of ca. 20 mg N 3 L1 3 h1 as before the experiment, thus conﬁrming complete degradation of ATU and the reversibility of the inhibition of the anammox biomass (growing too slowly for signiﬁcant recovery overnight). The rationale of the interpretation of the experiment in Figure 3 is as follows: decreasing AOB activity resulted in lower O2 depletion, leading indirectly to the inhibition of anammox by increased oxygen concentrations inside the ﬂocs. Since anammox activity has been found in diﬀerent activated sludge samples,5,26 it can be expected that these anaerobic bacteria are also well adapted to survive prolonged exposure to oxygen. Thus, partial toxic inhibition of AOB is seen as a plausible explanation of the sudden performance loss and NO2 accumulation described previously at full scale. During start-up of the full scale installation at Zurich-Werdholzli, a comparable activity loss has been observed with the decrease in AOB activity preceding the drop in anammox activity by weeks.6 According to the modeling of Hao et al.,27 alternatively a signiﬁcant decrease in typical sludge particle diameter might also have caused the observed N removal performance loss and NO2 accumulation: to achieve optimal anammox activity inside ﬂocs of decreasing size, aeration must occur at a lower soluble oxygen set point; since typical ﬂoc diameter is currently not routinely monitored, such a change would go unnoticed (except in case of resulting in signiﬁcant impaired sludge settling which did not occur). Short-Term Aeration Pulses May Lead to Healthy NOB urich-Werdh€olzli, two full-scale reactors are Population. At Z€ operated in parallel. After 2 years of successful operation, one of the two reactors (South) was found to have developed quite a strong NOB population, as evidenced by the effluent nitrate values increasing from typically below 15 mg NO3-N 3 L1 (up to the end of June 2009) to over 200 mg NO3-N 3 L1 in midOctober 2009 (>30% of the influent NH4+), while the effluent of reactor North was 10 mg NO3-N 3 L1 on average, and never increasing above 35 mg NO3-N 3 L1 (data not shown). The operational parameters and influents of the two reactors were identical except for one detail: the reactor with growing NOBs was equipped with old aerator membranes requiring a cleaning procedure resulting in a significant increase of the aeration during 2 h every 2 weeks (Figure 4); during this time of increased aeration, the bulk O2 concentration did not increase significantly above 1 mgO2 3 L1, and so did not reach alarming levels of 9738

dx.doi.org/10.1021/es202181v |Environ. Sci. Technol. 2011, 45, 9735–9742

Environmental Science & Technology

Figure 4. Average air supply rate during aeration in the two reactors operated in parallel. The aeration spikes in reactor South show the aerator membrane cleaning procedure performed only in this reactor, on average every 2 weeks.

short-term exposure. It is assumed that during these cleaning events the oxygen supply exceeded the oxygen depletion rate of the AOB, thus leading to reversible inhibition of the anammox, and transiently to optimal growing conditions for NOB, with NO2 accumulating and O2 not completely depleted by AOBs. This growth is assumed to have eventually led to a healthy NOB population successfully competing with AOB for O2 and with anammox for NO2. The activity in reactor North also decreased gradually in July and August 2009 as previously described (Figure 2). During the same period in reactor South, the NO3 values started increasing at a rate of about 0.05 d1 without showing any NO2 accumulation. It is thus concluded that the postulated toxic influent led in both reactors to an inhibition NO2 depletion by anammox, but also provided further support to NOB growth in reactor South, where an increased abundance of NOB had previously developed, while NO2 accumulation was observed in reactor North. After the event, reactor South was emptied and its sludge discarded. By splitting the uncontaminated sludge of reactor North between both reactors, reactors South and North were back to full capacity within a few weeks. Stable operation of the very similar CANON system has been shown at lab scale with synthetic wastewater to require oxygen limiting conditions,25,28 i.e., to avoid inhibition of anammox as well as for allowing AOBs to outcompete NOBs. According to Wiesmann et al.,29 it is expected that the signiﬁcantly higher aﬃnity for oxygen of AOB compared to NOBs is responsible for this competition (half-saturation constant of 0.6 compared to 2.2 mg O2 3 L1). Similarly, Hawkins et al.30 showed in a nitriﬁcation reactor under saturated DO conditions that NOB activity was suppressed as long as AOBs were not substrate limited; the observed rise in NOB activity after depletion of ammonium indicates that diﬀusive mass transfer limitation combined with higher aﬃnity of AOB to oxygen is responsible for suppressing NOB activity. According to present experience from full-scale operation, the treatment of digester supernatant results in a different competition situation for NOBs: numerous full-scale combined nitritationanammox system face at times eﬄuent NO3 concentrations >100 mg NO3-N 3 L1 (thus achieving reduced

ARTICLE

Figure 5. Experimental washout of nitrite oxidizers under normal operating conditions. Over 45 days the air supply and the NH4+ removal rate (left axis) could be increased in steps without the ratio of NO3 formed per NH4+ removed (right axis) permanently exceeding 30%.

N removal), in spite of keeping O2 availability rate limiting and avoiding NH4+ limitation (personal communication from various operators). In several circumstances, the problem of NOB activity could be solved only by discarding the NOB containing sludge and reinoculation from another reactor devoid of NOBs. Thus, better understanding and strategies for reducing NOB activity are still a relevant issue for process stability. Hao et al.27 conﬁrm in their modeling eﬀort that the ratio of diﬀerent aﬃnity constants is crucial to NOB versus AOB/anammox population dynamics and currently still uncertain, thus explaining the discrepancy between the mentioned lab- and full-scale operation. NOB Washout during Regular Operation. Previously published experiments6 had shown that the decay rate of NOB is too slow to decrease the abundance of NOB without active sludge withdrawal. Unpublished results in our lab showed that NH3 concentration in a typical range (i.e., up to pH 8 and 200 mg NH4+-N/L equivalent to 10 mg NH3-N/L) do not to impact on NOB growth sufficiently to be suitable for controlling their growth. The experiment illustrated in Figures 5 and 6 demonstrates that under regular operating conditions a strong population of NOB can be washed out by sludge removal at a sludge age of around 45 days. In addition, the availability of O2 and NO2 was limited to restrict NOB growth by controlling the aeration. During the experiment, the aeration was slowly increased in steps in order to stop the eﬄuent NO3 concentration from permanently exceeding 30% of the removed NH4+-N: a step increase was performed only after the NO3 production had dropped below 20% of the NH4+ removal. At each step increase of the aeration rate, the NO3 production increased transiently, but was reduced to below 20% after a week or two of operation at an unchanged aeration rate. Prompt depletion kept the concentrations of O2 and NO2 below <0.5 mgO2 3 L1 and <1 mgNO2-N 3 L1, respectively, for the entire duration of the experiment. Since several groups had shown a higher aﬃnity of AOB toward O2 compared to NOBs, it is expected that competition for molecular oxygen rather than for NO2 is crucial for limiting NOB growth.5,30 Control of the Aeration Rate. The aeration must be controlled in order to match the O2 supply to the AOB activity, since providing more oxygen than the AOB are able to deplete promptly leads to NOB growth and may reversibly inhibit anammox 9739

dx.doi.org/10.1021/es202181v |Environ. Sci. Technol. 2011, 45, 9735–9742

Environmental Science & Technology

Figure 6. Confocal laser scanning microscopy images taken from nitritation anammox sludge on day 5 (A) and day 55 (B) of the NOB washout experiment. FISH probes used: AOB in red, NOB in green, transmitted light grayscale. The black bar indicates 50 μm. The graph on the right shows the image analysis conﬁrming the trend of decreased NOB to AOB activity ratio shown in Figure 5. Error bars indicate the standard deviation where multiple images have been taken (2 e n e 4, n indicating the mean of 2040 images analyzed per well).

activity.27 Figure 3 shows that the oxygen concentration as measured in the bulk may not be taken as a measure for maximum sustainable operation without considering further indications of activity: while a good N removal rate was achieved at 0.2 mgO2 3 L1 under normal operating conditions, at the same bulk oxygen concentration but with decreased activity of the AOB the anammox biomass was completely inhibited, resulting in optimal growth conditions for NOBs (i.e., sufficient O2 and NO2 were available simultaneously). The same conclusion is drawn from the oxygen concentrations monitored at full scale in the event illustrated in Figure 2: controlling the aeration according to the dissolved oxygen did not allow avoiding NO2 accumulation. It is thus suggested that the aeration rate be controlled according to the soluble NO2 concentration: as long as the NO2 formed by the AOB is promptly depleted by the anammox, the aeration rate is sustainable, but as soon as NO2 starts accumulating to >2 mg NO2-N 3 L1, this indicates that conditions propitious for NOB growth are present, so the aeration rate is too high. Since reliable online NO2 measurement in activated sludge with an accuracy of <1 mg NO2-N 3 L1 and a background nitrate concentration somewhere between 0 and 200 mg NO3N 3 L1 remain a challenge, the nitrite concentration can currently be assessed either by sampling and oﬄine measurement or indirectly by monitoring the NH4+ depletion in the absence of soluble oxygen (as discussed previously). Since anammox biomass exhibits rather slow growth of 0.06 d1,4 the aeration is normally also increased at a rather slow rate (i.e., increase of the aeration rate once weekly during the start-up phase), while downward corrections must be carried out signiﬁcantly more quickly (e.g., in case a toxicant is present in the inﬂuent). According to current experience with supernatant originating from digestion of municipal sludge (supplemented with up to 20% industrial waste), automatic monitoring of the NO2 at the end of each batch (i.e., every 8 h) suﬃces under normal operating conditions for SBR operation at an exchange volume of up to 25%. Finally, the experiment shown in Figures 5 and 6 suggests that it is appropriate to limit the aeration rate for transforming up to

ARTICLE

20% of the oxidized ammonium to nitrate, thus limiting the growth of the NOBs to an extent resulting in washout at a feasible sludge age of 45 days. Compliance with the above rules results in limiting the net activity of nitritationanammox by the air supply, so that the N removal rate is proportional to the volume of air supplied under normal operating conditions. The aeration of the system is consequently controlled more eﬀectively by the volumetric air ﬂow rate than by the soluble O2 concentration as measured in the bulk liquid (as is usual for activated sludge processes). Detecting NOB Growth. Process stability requires the NOB population to be kept constantly at a very low level. As discussed previously (Figure 4), NOB growth can be slow and go unnoticed for long time, since effluent NO3 may also result from anammox directly (transforming ca. 10% of the removed nitrogen to nitrate) and may be depleted by denitrification to an extent depending on the availability of degradable COD. In view of the process stability issues discussed in this article, it is assumed that operators are interested in methods with sufficiently low quantification limits, so that the NOB levels may be controlled on a monthly basis and remain well below those apparent from high NO3 production. Because of its better quantiﬁcation limit, the quantitative polymerase chain reaction (qPCR; detecting down to a few copies per sample31) is expected to be more suitable than FISH, since the latter has a typical quantiﬁcation limit of 25% of the biomass.32 Several authors have applied qPCR to anammox sludge, but for other purposes and mostly focusing on quantifying the anammox rather than the NOB.726,30,33,34,35 Thus, the applicability of qPCR for this purpose has not yet been conﬁrmed. Floc Structure. The reversible inhibition of the anammox biomass in the presence of molecular oxygen means that the biofilm may be clearly structured in terms of activity (i.e., AOB activity in the outer layer and anammox activity in the inner anaerobic core), but this need not translate to a corresponding stratification of the biomass itself: according to the reversibility of the inhibition of anammox, a homogeneous distribution of this biomass cannot be excluded, possibly resulting from flocs growing and falling apart randomly, thus not featuring clear structures in biomass stratification, as observed also by ref 5. The direct analysis of granular sludge by Vlaeminck et al.36 did not show a clear stratification (i.e., with anammox embedded as a core while AOB and NOB were located in the outer layers) in some reactors while it did in others (e.g., ref 37). This issue of relevance to understanding the conditions needed to avoid NOB growth has consequently not yet been resolved, at least to the author’s knowledge.

’ ASSOCIATED CONTENT

bS

Supporting Information. Table of the oligonucleotide probes used for FISH analysis of AOB, NOB and all bacteria. This material is available free of charge via the Internet at http:// pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]. Phone: +41 58 765 5408. Fax: +41 58 765 5389. 9740

dx.doi.org/10.1021/es202181v |Environ. Sci. Technol. 2011, 45, 9735–9742

Environmental Science & Technology

’ REFERENCES (1) Siegrist, H.; Salzgeber, D.; Eugster, J.; Joss, A. Anammox brings WWTP closer to energy autarky due to increased biogas production and reduced aeration energy for N-removal. Water Sci. Technol. 2008, 57 (3), 383–388. (2) Wett, B. Solved upscaling problems for implementing deammoniﬁcation of rejection water. Water Sci. Technol. 2006, 53 (12), 121–128. (3) Abma, W. R.; Driessen, W.; Haarhuis, R.; van Loosdrecht, M. C. M. Upgrading of sewage treatment plant by sustainable and costeﬀective separate treatment of industrial wastewater. Water Sci. Technol. 2010, 61 (7), 1715–1722. (4) Strous, M.; Heijnen, J. J.; Kuenen, J. G.; Jetten, M. S. M. The sequencing batch reactor as a powerful tool for the study of slowly growing anaerobic ammonium-oxidizing microorganisms. Appl. Microbiol. Biotechnol. 1998, 50 (5), 589–596. (5) Jeanningros, Y.; Vlaeminck, S. E.; Kaldate, A.; Verstraete, W.; Graveleau, L. Fast start-up of a pilot-scale deammoniﬁcation sequencing batch reactor from an activated sludge inoculum. Water Sci. Technol. 2010, 61 (6), 1393–1400. (6) Joss, A.; Salzgeber, D.; Eugster, J.; K€onig, R.; Rottermann, K.; Burger, S.; Fabijan, P.; Leumann, S.; Mohn, J.; Siegrist, H. Full-scale nitrogen removal from digester liquid with partial nitritation and anammox in one SBR. Environ. Sci. Technol. 2009, 43, 5301–5306. (7) Park, H.; Rosenthal, A.; Ramalingam, K.; Fillos, J.; Chandran, K. Linking Community Proﬁles, Gene Expression and N-Removal in Anammox Bioreactors Treating Municipal Anaerobic Digestion Reject Water. Environ. Sci. Technol. 2010, 44, 6110–6116. (8) Tokutomi, T.; Yamauchi, H.; Nishimura, S.; Yoda, M.; Abma, W. Application of the Nitritation and Anammox Process into Inorganic Nitrogenous Wastewater from Semiconductor Factory. J. Environ. Eng.ASCE 2011, 137 (2), 146–154. (9) Amann, R. I.; Binder, B. J.; Olson, R. J.; Chisholm, 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), 1919–25. (10) Daims, H.; Bruhl, A.; Amann, R.; Schleifer, K. H.; 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), 434–44. (11) Nielsen, P. H.; Daims, H.; Lemmer, H. FISH Handbook for Biological Wastewater Treatment. Identiﬁcation and quantiﬁcation of microorganisms in activated sludge and bioﬁlms by FISH; IWA Publishing: London, 2009. (12) Wagner, M.; Rath, G.; Amann, R.; Koops, H.-P.; Schleifer, K.-H. In situ identiﬁcation of ammonia-oxidizing bacteria. Syst. Appl. Microbiol. 1995, 18, 251–264. (13) Mobarry, B. K.; Wagner, M.; Urbain, V.; Rittmann, B. E.; Stahl, D. A. Phylogenetic probes for analyzing abundance and spatial organization of nitrifying bacteria. Appl. Environ. Microbiol. 1996, 62 (6), 2156–62. (14) Adamczyk, J.; Hesselsoe, M.; Iversen, N.; Horn, M.; Lehner, A.; Nielsen, P. H.; Schloter, M.; Roslev, P.; Wagner, M. The isotope array, a new tool that employs substrate-mediated labeling of rRNA for determination of microbial community structure and function. Appl. Environ. Microbiol. 2003, 69, 6875–6887. (15) Daims, H.; Nielsen, J. L.; Nielsen, P. H.; Schleifer, K. H.; Wagner, M. In situ characterization of Nitrospira-like nitrite-oxidizing bacteria active in wastewater treatment plants. Appl. Environ. Microbiol. 2001, 67 (11), 5273–84. (16) Vandegraaf, A. A.; Mulder, A.; Debruijn, P.; Jetten, M. S. M.; Robertson, L. A.; Kuenen, J. G. Anaerobic Oxidation of Ammonium Is a Biologically Mediated Process. Appl. Environ. Microbiol. 1995, 61 (4), 1246–1251. (17) Kindaichi, T.; Tsushima, I.; Ogasawara, Y.; Shimokawa, M.; Ozaki, N.; Satoh, H.; Okabe, S. In situ activity and spatial organization of anaerobic ammonium-oxidizing (anammox) bacteria in bioﬁlms. Appl. Environ. Microbiol. 2007, 73 (15), 4931–4939.

ARTICLE

(18) Guven, D.; Dapena, A.; Kartal, B.; Schmid, M. C.; Maas, B.; van de Pas-Schoonen, K.; Sozen, S.; Mendez, R.; Op den Camp, H. J. M.; Jetten, M. S. M.; Strous, M.; Schmidt, I. Propionate oxidation by and methanol inhibition of anaerobic ammonium-oxidizing bacteria. Appl. Environ. Microbiol. 2005, 71 (2), 1066–1071. (19) Kumar, M.; Lin, J. G. Co-existence of anammox and denitriﬁcation for simultaneous nitrogen and carbon removal-Strategies and issues. J. Hazardous Mater. 2010, 178 (13), 1–9. (20) deGraaf, A. A. V.; deBruijn, P.; Robertson, L. A.; Jetten, M. S. M.; Kuenen, J. G. Autotrophic growth of anaerobic ammoniumoxidizing micro-organisms in a ﬂuidized bed reactor. Microbiology-UK 1996, 142, 2187–2196. (21) Wett, B. Development and implementation of a robust deammoniﬁcation process. Water Sci. Technol. 2007, 56 (7), 81–88. (22) Lv, Y. T.; Wang, L.; Sun, T.; Wang, X. D.; Yang, Y. Z.; Wang, Z. Y. Autotrophic nitrogen removal discovered in suspended nitritation system. Chemosphere 2010, 79 (2), 180–185. (23) Dapena-Mora, A.; Fernandez, I.; Campos, J. L.; MosqueraCorral, A.; Mendez, R.; Jetten, M. S. M. Evaluation of activity and inhibition eﬀects on Anammox process by batch tests based on the nitrogen gas production. Enzyme Microbial Technol. 2007, 40 (4), 859–865. (24) Strous, M.; vanGerven, E.; Kuenen, J. G.; Jetten, M. Eﬀects of aerobic and microaerobic conditions on anaerobic ammonium-oxidizing (Anammox) sludge. Appl. Environ. Microbiol. 1997, 63 (6), 2446–2448. (25) Third, K. A.; Sliekers, A. O.; Kuenen, J. G.; Jetten, M. S. M. The CANON system (completely autotrophic nitrogen-removal over nitrite) under ammonium limitation: Interaction and competition between three groups of bacteria. Syst. Appl. Microbiol. 2001, 24 (4), 588–596. (26) Bae, H.; Park, K. S.; Chung, Y. C.; Jung, J. Y. Distribution of anammox bacteria in domestic WWTPs and their enrichments evaluated by real-time quantitative PCR. Process Biochem. 2010, 45 (3), 323–334. (27) Hao, X. D.; Heijnen, J. J.; Van Loosdrecht, M. C. M. Modelbased evaluation of temperature and inﬂow variations on a partial nitriﬁcation-ANAMMOX bioﬁlm process. Water Res. 2002, 36 (19), 4839–4849. (28) Sliekers, A. O.; Derwort, N.; Gomez, J. L. C.; Strous, M.; Kuenen, J. G.; Jetten, M. S. M. Completely autotrophic nitrogen removal over nitrite in one single reactor. Water Res. 2002, 36 (10), 2475–2482. (29) Wiesmann, U. Biological nitrogen removal from wastewater. In Advances in Biochemical Engineering Biotechnology; Biotechnics/ wastewater; Fiechter, A., Ed.; Springer: New York, 1994; Vol. 51, pp 113154. (30) Hawkins, S.; Robinson, K.; Layton, A.; Sayler, G. Limited impact of free ammonia on Nitrobacter spp. inhibition assessed by chemical and molecular techniques. Bioresour. Technol. 2010, 101 (12), 4513–4519. (31) Smith, C. J.; Osborn, A. M. Advantages and limitations of quantitative PCR (Q-PCR)-based approaches in microbial ecology. FEMS Microbiol. Ecol. 2009, 67 (1), 6–20. (32) Wagner, M.; Horn, M.; Daims, H. Fluorescence in situ hybridisation for the identiﬁcation and characterisation of prokaryotes. Curr. Opin. Microbiol. 2003, 6 (3), 302–309. (33) Huang, Z. H.; Gedalanga, P. B.; Asvapathanagul, P.; Olson, B. H. Inﬂuence of physicochemical and operational parameters on Nitrobacter and Nitrospira communities in an aerobic activated sludge bioreactor. Water Res. 2010, 44 (15), 4351–4358. (34) Tsushima, I.; Kindaichi, T.; Okabe, S. Quantiﬁcation of anaerobic ammonium-oxidizing bacteria in enrichment cultures by real-time PCR. Water Res. 2007, 41 (4), 785–794. (35) van der Star, W. R. L.; Abma, W. R.; Blommers, D.; Mulder, J. W.; Tokutomi, T.; Strous, M.; Picioreanu, C.; Van Loosdrecht, M. C. M. Startup of reactors for anoxic ammonium oxidation: Experiences from the ﬁrst full-scale anammox reactor in Rotterdam. Water Res. 2007, 41 (18), 4149–4163. (36) Vlaeminck, S. E.; Terada, A.; Smets, B. F.; De Clippeleir, H.; Schaubroeck, T.; Bolca, S.; Demeestere, L.; Mast, J.; Boon, N.; Carballa, M.; Verstraete, W. Aggregate Size and Architecture Determine Microbial 9741

dx.doi.org/10.1021/es202181v |Environ. Sci. Technol. 2011, 45, 9735–9742

Environmental Science & Technology

ARTICLE

Activity Balance for One-Stage Partial Nitritation and Anammox. Appl. Environ. Microbiol. 2010, 76 (3), 900–909. (37) Nielsen, M.; Bollmann, A.; Sliekers, O.; Jetten, M.; Schmid, M.; Strous, M.; Schmidt, I.; Larsen, L. H.; Nielsen, L. P.; Revsbech, N. P. Kinetics, diﬀusional limitation and microscale distribution of chemistry and organisms in a CANON reactor. FEMS Microbiol. Ecol. 2005, 51 (2), 247–256.

9742

dx.doi.org/10.1021/es202181v |Environ. Sci. Technol. 2011, 45, 9735–9742

ARTICLE pubs.acs.org/est

Novel Perspectives on the Bioaccumulation of PFCs the Concentration Dependency Changhui Liu,† Karina Y. H. Gin,‡ Victor W. C. Chang,†,* Beverly P. L. Goh,§ and Martin Reinhard|| †

School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798 Department of Civil and Environmental Engineering, National University of Singapore § Natural Sciences and Science Education, National Institute of Education, Singapore Civil and Environmental Engineering, Stanford University, California, United States

)

‡

bS Supporting Information ABSTRACT: The eﬀects of exposure concentration on the bioaccumulation of four perﬂuorinated chemicals (PFCs): perﬂuorooctanesulfonate (PFOS), perﬂuoroocanoic acid (PFOA), perﬂuorononanoic acid (PFNA), and perﬂuorodecanoic acid (PFDA), was investigated using green mussels, Perna viridis. Mussels were exposed to concentrations of 1 μgL1 and 10 μgL1 of each PFC for 56 days, and the bioaccumulation factors (BAF) were found to range from 15 to 859 L/kg and from 12 to 473 L/kg at 1 μgL1 and 10 μgL1, respectively. For all compounds, the BAF was larger at the lower dosage. Results suggest that the bioaccumulation of PFCs is concentration dependent. This concentration dependency can be explained by a nonlinear adsorption mechanism, which was further supported by the experimental results. The sensitivity of BAF to exposure concentration was found to be positively related to perﬂuorinated chain length and the binding aﬃnity of the compounds. Bioaccumulation of long chain carboxylates and sulfonates are more easily aﬀected by concentration changes. The validity of the conventional kinetic method was examined by comparing the results with the fundamental steady-state method: in addition to the above-mentioned batch test, mussels were also subject to 24-day exposure (1 μgL1 and 10 μgL1) followed by 24-day depuration. Contradictions were found in the resulting kinetic BAF and model curving ﬁttings. A new kinetic model based on adsorption mechanism was proposed, which potentially provide more accurate description of the bioaccumulation process of PFCs.

’ INTRODUCTION In recent years, there has been growing concern in perﬂuorinated chemicals (PFCs). PFCs are synthetic compounds that have been applied in a broad spectrum of commercial products and industrial processes in the past few decades. Due to their unique water and fat repellent properties, PFCs are widely used as surfactants, refrigerants and as components of pharmaceuticals, lubricants, paints, ﬁre ﬁghting foams, cosmetics, and food packaging. 13 PFCs are released to the environment either through usage of the PFCs-containing products or by degradation of their precursors. Due to the high energy of carbonﬂuorine covalent bonds, PFCs are thermally and chemically stable, and are also resistant to biodegradation. Hence, they are extremely persistent in the environment,48 and have been found extensively in wildlife and human bodies worldwide, even in remote regions such as the Arctic.917 A number of studies have also demonstrated that PFCs have adverse eﬀects on human and animals including possible carcinogenic eﬀects.3,1822 Both monitoring and controlled laboratory studies have demonstrated the potential of PFCs to bioaccumulate.1,9,23,24 Studying the bioaccumulation of PFCs requires approaches that are diﬀerent from those used for other persistent organic pollutants r 2011 American Chemical Society

(POPs), such as polychlorinated biphenyls (PCBs) because PFCs are not only hydrophobic but also oleophobic. They possess a high aﬃnity to protein albumin, and are sometimes referred to as “proteinophilic”.5,15 Both monitoring and laboratory studies have demonstrated that PFCs tend to accumulate preferentially in protein-rich tissues, such as the liver, and in blood.17,19 The accumulation mechanism and exposure routes of PFCs are therefore diﬀerent from other hydrophobic POPs. Thus, the commonly used octanolwater partition coeﬃcient, Kow, to predict bioaccumulation is inappropriate and inaccurate. Current bioaccumulation data shows discrepancies between diﬀerent laboratory studies.3,19 A possible explanation for these discrepancies is that the test species contained diﬀerent amounts of protein or that environmental conditions maintained during the tests diﬀered.25 Most studies adopted a kinetic approach that involved approximation and estimation in curve ﬁtting.19,2427 As the relative contribution of each of the aﬀecting factors of Received: April 4, 2011 Accepted: September 19, 2011 Revised: September 13, 2011 Published: October 11, 2011 9758

dx.doi.org/10.1021/es202078n | Environ. Sci. Technol. 2011, 45, 9758–9764

Environmental Science & Technology

ARTICLE

Figure 1. Bioaccumulation of PFCs in green mussels at exposure concentrations of 1 μgL1 and 10 μgL1 during 56 day exposure (n = 12). Data points are results from the steady-state experiment. Error bars represent standard deviation. Curves are the proposed kinetic model ﬁtted to eq 9.

PFCs bioaccumulation is still unclear, direct measurement of the bioaccumulation factor or the steady-state approach may elicit greater accuracy. Recent reviews suggested that PFCs bioaccumulation is concentration dependent,4,19 that is, it is a function of concentration. However, data that shows exposure concentration is a signiﬁcant parameter in the bioaccumulation assessment seems to be lacking. The purpose of this study was to examine the bioaccumulation mechanism of PFCs, and in particular, to characterize the eﬀects of exposure concentration on the bioaccumulation process. The validity of the commonly used approach to model bioaccumulation kinetics was evaluated and a new model was proposed. A local species of green mussels, Perna viridis, was chosen as the test organism. Bivalves, in particular mussels, have been commonly used as bioindicators in pollution monitoring because they are ﬁlter-feeding, sessile organisms and therefore can accumulate relatively large concentrations of pollutants.

’ MATERIALS AND METHODS Chemicals and Standards. Potassium perfluorooctanesulfonate (PFOS, 98%), perfluoroocanoic acid (PFOA, 96%), perfluorononanoic acid (PFNA, 97%), perfluorodecanoic acid (PFDA, 98%) were purchased from Sigma-Aldrich (St. Louis, MO).

The internal standards, perfluoro-n-[1,2,3,4- 13 C 4 ]octanoic acid (MPFOA, 99%), perfluoro-n-[1,2,3,4,5-13C5]nonanoic acid (MPFNA, 99%), perfluoro-n-[1,2-13C2]decanoic acid (MPFDA, 99%) and sodium perfluoro-1-[1,2,3,4-13C4]octanesulfonate (MPFOS, 99%) were purchased from Wellington Laboratories (Guelph, ON). The stock solutions were prepared with PFOS, PFOA, PFNA, and PFDA at 1000 mg L1 and 100 mg L1 in optima grade methanol. The stock solutions were stored at 4 °C. Experiment Set-Up. To examine the effects of exposure concentration on bioaccumulation, mussels were exposed to 1 μgL1 and 10 μgL1 of PFOS, PFOA, PFNA, and PFDA (total PFC concentration of the mixtures were 4 μgL1 and 40 μgL1, respectively). These concentrations were below the lowest no observed effect concentration (NOEC) of marine invertebrates,28 so as not to affect the general well-being of the test organisms. Seventy-liter polypropylene (PP) tanks were used as the test chambers. 6065 mussels with similar size were raised in artificial seawater. For each exposure concentration, two sets of duplicate tanks were used: in one set, mussels were exposed to PFCs for up to 56 days; in the other set, mussels were subject to a 24-day exposure followed by a 24-day depuration. The purpose of this experimental design was to determine and compare the bioaccumulation factors through both steady-state 9759

dx.doi.org/10.1021/es202078n |Environ. Sci. Technol. 2011, 45, 9758–9764

Environmental Science & Technology

ARTICLE

Figure 2. Depuration kinetic of PFCs after exposure to concentrations of (a)1 μgL1 and (b)10 μgL1.

(former set) and kinetic approaches (latter set). Another tank was also engaged as a control, where no PFCs were present. All tanks were cleaned and refilled every two days. Four mussels were sampled from each tank at each sampling time. 100 mL aqueous samples were also taken every four or eight days for PFCs concentration analysis. Additional information on experiment setup and mussel rearing is available in the Supporting Information (SI). Sample Preparation and Extraction. A few extraction methods have been described in the literature.11,29,30 In order to validate the precision and accuracy of PFC determination in mussel tissues, two commonly used extraction methods, the acetonitrile extraction30 and alkaline digestion with SPE11 were tested and compared. The latter (recovery 95110%) was selected in this experiment. Details of the extraction method and recovery tests are available in the SI. Briefly, homogenized dry mussel tissues was extracted using 30 mL of potassium hydroxide solvent (KOH 0.01 mol/L in methanol) in 50 mL PP tubes. The mixture was vortexed, shaken (300 rpm, 25 °C, 18 h), and centrifuged (4000 rpm, 15 min). 0.5 mL of supernatant was diluted (1:100 with Milli-Q water) and extracted by using Oasis HLB cartridges (0.2 g, 6 cm3; Waters). Prior to loading, cartridges were preconditioned by eluting with 5 mL methanol followed by 5 mL Milli-Q water. Cartridges were vacuum-dried before elution using 15 mL methanol. Elutes were dried by nitrogen gas and reconstituted to 2 mL with methanol. Prior to analysis, 200 mL of final extracts were transferred to the sample vial with 20 μL of internal standards. Instrumental Analysis. Details of analytical method and sample analysis are available in the SI. Briefly, concentrations of PFCs in mussel tissues were determined using high-performance liquid chromatography coupled with tandem mass spectrometry (LC MS/MS). The MS/MS was operated in negative electrospray ionization multiple reaction monitoring (MRM) mode. A volume injection of 20 μL was injected into a Targa Sprite C18 column (3.5 μm pore size, 40 mm 2.1 mm ID, Higgins Analytical, CA) using methanol gradient (initial and final eluent condition of 35% methanol) at a flow rate of 0.25 mL min1 with 2 mM ammonium acetate as the second mobile phase.27,30

Data Analysis. The bioaccumulation factor (BAF) was calculated according to

BAF ¼

Co Cw

ð1Þ

where Co is the PFC concentration in the organism at steady state(ng/g); Cw is the PFC concentration in water (μg/L); BAF is in L/kg. Steady state was determined using a previous apporach19 modified as follows: steady state was assumed when three or more consecutive measurements were not statistically different, or when the normalized slope of the fitted line of three or more consecutive measurements was less than 0.005 (1/day). One-way ANOVA was applied to determine the statistical significance.

’ RESULTS AND DISCUSSION Bioaccumulation Results. Among the tested compounds, the long-chain perfluorocarboxylate and perfluorosulfonate were found to have the highest bioaccumulation potential. PFDA possess the largest BAF followed by PFOS (Figure 1). PFOA, on the other hand, is the least accumulative compound with a steady state concentration about 20 times lower than PFOS. This is consistent with the observation that PFOS is generally detected at higher level in wildlife than PFOA, although the environmental concentrations of the two are comparable.16,31 Compounds with higher BAF take longer to reach steady state, which is also consistent with previous studies,4,24,26 and their depuration is generally slower too. PFC depuration follows a first-order (exponential) model with rates increasing in the order of PFDA < PFOS < PFNA < PFOA (Figure 2). The fast elimination may be facilitated by the presence of PFCs in the circulating blood, coupled with extensive blood-water exchange at mussel gills during respiration.15 It also implies that there should be a continuous exposure of PFCs at a certain level to maintain an observed tissue concentration. Concentration Dependency of Bioaccumulation. Figure 1 shows that when exposure concentration (Cw) changes from low (1 μgL1) to high (10 μgL1), steady state organism concentrations (Co) do not increase proportionally. The BAF values (eq 1) 9760

dx.doi.org/10.1021/es202078n |Environ. Sci. Technol. 2011, 45, 9758–9764

965 ( 297

Highlight that data in the columns with the same superscript are signiﬁcantly diﬀerent at p < 0.01. b A:ke estimated from elimination phase where Co = A 3 exp(ket); B: BAFss was the average of steady-state results; D: nku and (kuCw + ke) are directly obtained from curve ﬁtting results; E: k0 u estimated from initial uptake phase;25 n = 12; All curve ﬁttings were carried out by Matlab. Data are provided with standard error ((SE) and R2 in parentheses. a

473 ( 40 (0.96) 464 ( 25 9 PFDA

0.05 ( 0.01 (0.93) 0.04 ( 0.01 (0.95) 838 ( 66

0.10 ( 0.03

859 ( 93 (0.97)

112 ( 50

146 ( 64

18 ( 7

13.1 ( 0.0 (1.0) 10.4 ( 0.0 (1.0) 105 ( 7 (0.96) 149 ( 12 (0.97) 109 ( 14 8 PFNA

0.09 ( 0.04 (0.92) 0.09 ( 0.04 (0.93) 144 ( 14

0.05 ( 0.02

12 ( 1 (0.94) 15 ( 1 (0.97)

62.1 ( 16.8 (0.98) 40.6 ( 3.5 (0.99) 1375 ( 569

384 ( 126

1.9 ( 0.4 (0.99) 1.9 ( 0.4 (0.99)

236 ( 17 (0.97) 386 ( 37 (0.96)

12 ( 0

0.07 ( 0.02 235 ( 13

0.10 ( 0.04 (0.95) 0.10 ( 0.04 (0.96) 15 ( 1 7 PFOA

0.05 ( 0.01 (0.95) 0.05 ( 0.01 (0.96) 378 ( 29 8 PFOS

10 ppb perﬂuorinated chain length

1 ppb

10 ppb

1 ppb

0.03 ( 0.01

10 ppba#

19 ( 8

25.4 ( 22.9 (0.77) 19.3 ( 3.9 (0.99) 472 ( 447

10 ppba# 1 ppba*

ARTICLE

1 ppba*

(BAFkinetic = nku/(kuCw+ke)) (L/kg) BAFss = Co/Cw (L/kg) ke (1/d)

α (L/kg)

B A

Table 1. Bioaccumulation Factors and Parameters

C

D - proposed model

1 ppb

10 ppb

BAF0 kinetic = k0 u/ke (L/kg) k0 u (L/kg/d)

E - old model 25

Environmental Science & Technology

of each compound are different at the two exposure concentrations (p < 0.01, t test) and decrease with increasing exposure by a factor of 1.3 for PFOA and 1.8 for PFDA (Figure 1, Table 1-B). For each compound, the time to reach steady state is longer under the lower exposure concentration where the BAF is also higher. These results show that the bioaccumulation of PFCs is concentration dependent. Although the concentration dependency of PFCs bioaccumulation has been mentioned in previous studies,14,19 insight into the underlying factors have been lacking. A possible explanation of the observed results is that bioaccumulation of PFCs is an adsorption-like process in which PFC molecules adsorb to the surface of quasi-solid materials, and the rationale is that PFC molecules are surface active chemicals.4 The conventional bioaccumulation model only views tissues as a “bulk phase” and the biological uptake as simple partitioning process. The mechanism of chemical adsorption is shown as follows:32 ku

M þ Sf s s MS r

ð2Þ

ke

M = chemicals, S = free binding sites, MS = bonded chemicals, ku = uptake rate constant, ke = elimination rate constant For PFCs, the binding sites are most likely at the surface of hemocytes and liver cells. As mentioned previously, PFCs tend to accumulate in these protein-rich compartments. In adsorption, The binding sites limit the amount of adsorbate, and the fractional surface coverage of adsorbent, θ, depends on the concentration of adsorbate: θ¼

MS αM ku ¼ where α ¼ MS þ S 1 þ αM ke

ð3Þ

Therefore the amount of adsorption depends on the chemical concentration. From this adsorption model, the major ﬁndings in this experiment can be well explained: (1) exposure concentration dependency of BAF, (2) correlation between BAF sensitivity to exposure concentration and perﬂuorinated chain length, and (3) discrepancy between the kinetic and the steady-state approach. Additional information on derivations and calculations are available in the SI. (1) Concentration Dependency of BAF. As illustrated in Figure 1, the BAF decreases as the exposure concentration Cw increases. If n is the number of total binding sites per gram organism, Co is equivalent to nθ; Cw is equivalent to M. By substituting Co and eq 3 into the eq 1, we obtain BAF ¼

Co nθ nα ¼ ¼ Cw 1 þ αCw Cw

ð4Þ

Eq 4 shows that the BAF is an inverse function of the exposure concentration Cw. An increase in Cw will lead to a decrease in BAF. Therefore, when concentration Cw increased from 1 μgL1 to 10 μgL1, the BAF decreased from nα/(1+α 3 1 μgL1) to nα/(1+α 3 10 μgL1), which explains the observed results that BAF became lower at the higher exposure concentration. (2) Chain Length Effect. Binding Affinity Vs Chain Length. In eq 3, the constant α is determined by, and directly proportional to, the binding energy.32 The α value of each compound can be calculated by eq 3 when Co and Cw are known. The results obtained follow the order of PFDA > PFOS > PFNA > PFOA (Table 1-C). A linear relationship was found between α, the binding affinity, and perfluorinated chain length (Figure 3). This is 9761

dx.doi.org/10.1021/es202078n |Environ. Sci. Technol. 2011, 45, 9758–9764

Environmental Science & Technology

ARTICLE

Figure 3. Relationship between binding aﬃnity and perﬂuorinated chain length. Error bars represent standard deviation.

consistent with the previous conclusion that long chain PFCs have enhanced binding energy.8,13 Bioaccumulation of PFCs was shown to be related to the chemical’s hydrophobicity.24,26 Being the hydrophobic portion in the PFC molecule, the longer the perfluorinated chain, the stronger the hydrophobic interaction, and thus the higher the binding energy. In addition, it is also believed that bioaccumulation is governed to some degree by the contribution of ionic interactions of functional groups,33 which explains the higher binding affinity by sulfonate than carboxylate with the same perfluorinated chain (Figure 3: PFNA and PFOS). Concentration Sensitivity Vs Chain Length. The correlation of the exposure concentration induced change in log BAF (Δlog BAF) and chain length was demonstrated in Figure 4: among the carboxylates, BAF of long chain PFC is affected more significantly by concentration change. The magnitude of Δlog BAF follows the order of increasing chain length. The Δlog BAF of sulfonate is larger than that of carboxylate with the same perfluorinated chain length (Figure 4: PFNA and PFOS). These results can be well explained by involving the binding affinity α. In eq 4, dividing both the denominator and the numerator by α we have n BAF ¼ 1 þ Cw α

ð5Þ

As the number of total binding sites, n, is constant, the effect of Cw on BAF, or in other words the BAF’s sensitivity to Cw, depends on α value. When αv, 1/αV, therefore Cw will have greater influence on the BAF. Hence the larger the α value, the more sensitive the BAF to changes in Cw (i.e., the larger concentration induced change in log BAF). From a physio-chemical point of view, when concentration increases, compounds possessing higher binding affinity are more likely to be adsorbed than those with low binding affinity. Thus, the amount of accumulation with respect to concentration change is more significant. The concentration-induced change in log BAF was found to be linearly correlated with α, too (Figure 4). Since the binding affinity α is closely related with chain length as discussed earlier, it is feasible to relate chain length with Δlog BAF through α, that is, longer chain f larger α f greater influence by Cw.

Figure 4. Relationship between Δlog BAF with binding aﬃnity (lower curve, left axis); and with perﬂuorinated chain length (upper curve, right axis). Error bars represent standard deviation.

(3) Comparison of Kinetic Approach and Steady-State Approach. Besides the fundamental steady-state approach (eq 1) to determine BAF, the kinetic approach has been popular in laboratory bioaccumulation studies.2426 The kinetic BAF is estimated as the quotient of uptake and elimination rate constants (eq 7), with the assumption that both uptake and elimination of chemicals are first order reactions (eq 6).34 dCoðtÞ ¼ k0u Cw ke CoðtÞ dt BAF0kinetic ¼

k0u ke

ð6Þ ð7Þ

Controlled laboratory studies have shown that the kinetic approach can generate similar results as the steady-state method for many POPs.34 However, the validity of the kinetic approach had never been verified for PFCs before it was applied in several laboratory studies.13,2426 In this study, the kinetic method was for the first time compared with the steady-state method and the result shows that it is not suitable for the assessment of bioaccumulation of PFCs. As discussed previously, the BAF is exposure concentration dependent. However, the expression of eq 7 itself suggests that the kinetic BAF (BAF0 kinetic) should be a constant value independent of concentration. This fundamentally contradicts the experimental results. Moreover, if following the above-mentioned kinetic approach as described in a previous study,25 the resulting uptake rate constant in eq 7, k0 u, varies for the two exposure concentrations (Table 1-E), which suggests that k0 u here is not a constant as defined. Hence, the previous assumption of “first order uptake reaction” is inappropriate in the case of PFCs bioaccumulation. Even though two kinetic BAF values can be calculated from the variant k0 u, the results were still shown to deviate from the steady-state BAF (BAFss) (Table 1-B and E, SI Figure S1). Based on the experimental results and the special surfactant property of PFCs we hereby propose a new kinetic equation (eq 8) to describe the bioaccumulation of PFCs, incorporating the adsorption model (eq 2). Compared with the old kinetic model (eq 6), accumulation is no longer ﬁrst order reaction: the rate of accumulation of PFCs depends on both the exposure 9762

dx.doi.org/10.1021/es202078n |Environ. Sci. Technol. 2011, 45, 9758–9764

Environmental Science & Technology

ARTICLE

concentration and free binding sites: dCoðtÞ ¼ ku Cw SðtÞ ke CoðtÞ dt

ð8Þ

where S(t) is the free binding sites at time t (eq 2). The expression for organism concentration at time t (Co(t)) can be obtained by solving eq 8 as CoðtÞ ¼

nku Cw f1 exp½ ðku Cw þ ke Þtg ku Cw þ ke

ð9Þ

And the kinetic BAF can be obtained as BAFkinetic ¼

nku ku Cw þ ke

ð10Þ

Compared with eq 7, the exposure concentration eﬀect is incorporated in eq 10. This proposed expression of the kinetic BAF is consistent with the one obtained from the steady-state approach (eq 4). Curve ﬁttings using eq 9 show agreement with the steady-state experimental results with good reliability, and so does the resulted BAFkinetic with the steady-state BAF, the BAFss (R2 g 0.94; p < 0.01, F test) (Figure 1; Table 1-B and D). Time to Reach Steady State. As mentioned previously, the time to reach the steady state, tss, varies between the two exposure concentrations. Therefore, tss also appears to be concentration dependent. Mathematically the time to reach steady state is infinity. However, the time for Co to reach 95% of steady state concentration, t95% ss , can be used as a performance metric and derived from eq 9 as tss95% ¼

1 lnð1 0:95Þ ku Cw þ ke

ð11Þ

Equation 11 shows that when exposure concentration Cw increases, tss will decrease accordingly, which explains the observed results of longer tss at lower exposure. However, if the old kinetic method (eq 6) is followed, the t95% ss will be t 095% ¼ ss

1 lnð1 0:95Þ ke

ð12Þ

where tss depends only on ke and thus, is a constant value for each compound. Taken together, the bioaccumulation of PFC appears to follow an adsorption model. In consideration of the special partitioning behavior of PFCs, the conventionally used kinetic model appears to be inaccurate for this group of chemicals. The fundamental assumption that both uptake and elimination of PFCs are ﬁrst order reactions merits further scrutiny.

’ IMPLICATIONS To our knowledge, this is the ﬁrst study that demonstrates the concentration dependency of the bioaccumulation of perﬂuorinated compounds, and describes the relationships among various factors using mathematical models. Examination of the concentration dependency reveals the inadequacy of the conventional kinetic model and a new model based on the adsorption mechanism is accordingly proposed. This model may provide more accurate description of the bioaccumulation process and thus, the fate of PFCs. It is also noted that protein binding and adsorption have similar mechanisms, both of which can be described by eq 2.35 Protein-water partitioning has been suggested to be useful in

evaluating the bioaccumulation of PFCs.15 As proteins have been considered as major reservoirs for PFCs, it is expected that protein binding could dominate the bioaccumulation process. It is therefore also possible that the observed results in our study are attributed to this predominant process. Partitioning coeﬃcients used to predict the environmental distribution of chemicals, such as Kow, are independent of concentration. Although more species- and environmental factordependent, the BAF in many ways is similar to these partitioning coeﬃcients. Hence, in previous studies, the BAF of perﬂuorochemicals were always treated as a constant, regardless of the exposure concentration.13,2426 In other words, bioaccumulation was assessed without considering concentration as an inﬂuencing parameter. The unique properties of PFCs, however, suggest that this approach may not be appropriate. This argument is further supported by the results of this study which has shown the importance of specifying the environmental concentration. Literature reviews indicate that BAF data from both laboratory and ﬁeld studies are inconsistent, a fact that may possibly be explained by the concentration dependence of PFC. Although further research work is still needed to fully understand the mechanisms of bioaccumulation of PFCs, the concentration factor should be taken into consideration in this process and also in ecotoxicological assessment.

’ ASSOCIATED CONTENT

bS

Supporting Information. Additional information regarding the method, equation derivation, and model comparison. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: (65)6790 4773; e-mail: [email protected].

’ ACKNOWLEDGMENT Funding support for this study was provided by the Environment and Water Industry Development Council (EWI) (0601IRIS-031000). We also thank Nguyen Viet Tung for his assistance in PFCs analysis. ’ REFERENCES (1) Houde, M.; Martin, J. W.; Letcher, R. J.; Solomon, K. R.; Muir, D. C. G. Biological monitoring of polyﬂuoroalkyl substances: A review. Environ. Sci. Technol. 2006, 40 (11), 3463–3473. (2) Suja, F.; Pramanik, B. K.; Zain, S. M. Contamination, bioaccumulation and toxic eﬀects of perﬂuorinated chemicals (PFCs) in the water environment: A review paper. Water Sci. Technol. 2009, 60 (6), 1533–1544. (3) Lau, C.; Anitole, K.; Hodes, C.; Lai, D.; Pfahles-Hutchens, A.; Seed, J. Perﬂuoroalkyl acids: A review of monitoring and toxicological ﬁndings. Toxicol. Sci. 2007, 99 (2), 366–394. (4) Conder, J. M.; Hoke, R. A.; De Wolf, W.; Russell, M. H.; Buck, R. C. Are PFCAs bioaccumulative? A critical review and comparison with regulatory lipophilic compounds. Environ. Sci. Technol. 2008, 42 (4), 995–1003. (5) Rayne, S.; Forest, K. Perﬂuoroalkyl sulfonic and carboxylic acids: A critical review of physicochemical properties, levels and patterns in waters and wastewaters, and treatment methods. J. Environ. Sci. Health, Part A: Toxic/Hazard. Subst. Environ. Eng. 2009, 44 (12), 1145–1199. 9763

dx.doi.org/10.1021/es202078n |Environ. Sci. Technol. 2011, 45, 9758–9764

Environmental Science & Technology (6) Prevedouros, K.; Cousins, I. T.; Buck, R. C.; Korzeniowski, S. H. Sources, fate and transport of perﬂuorocarboxylates. Environ. Sci. Technol. 2006, 40 (1), 32–44. (7) Plumlee, M. H.; Larabee, J.; Reinhard, M. Perﬂuorochemicals in water reuse. Chemosphere 2008, 72 (10), 1541–1547. (8) Armitage, J. M.; MacLeod, M.; Cousins, I. T. Comparative assessment of the global fate and transport pathways of long-chain perﬂuorocarboxylic acids (PFCAs) and perﬂuorocarboxylates (PFCs) emitted from direct sources. Environ. Sci. Technol. 2009, 43 (15), 5830–5836. (9) Fernandez-Sanjuan, M.; Meyer, J.; Damasio, J.; Faria, M.; Barata, C.; Lacorte, S. Screening of perﬂuorinated chemicals (PFCs) in various aquatic organisms. Anal. Bioanal. Chem. 2010, 398 (3), 1447–1456. (10) Shaw, S.; Berger, M. L.; Brenner, D.; Tao, L.; Wu, Q.; Kannan, K. Speciﬁc accumulation of perﬂuorochemicals in harbor seals (Phoca vitulina concolor) from the northwest Atlantic. Chemosphere 2009, 74 (8), 1037–1043. (11) So, M. K.; Taniyasu, S.; Lam, P. K. S.; Zheng, G. J.; Giesy, J. P.; Yamashita, N. Alkaline digestion and solid phase extraction method for perﬂuorinated compounds in mussels and oysters from south China and Japan. Arch. Environ. Contam. Toxicol. 2006, 50 (2), 240–248. (12) Berger, U.; Glynn, A.; Holmstrom, K. E.; Berglund, M.; Ankarberg, E. H.; Tornkvist, A. Fish consumption as a source of human exposure to perﬂuorinated alkyl substances in Sweden—Analysis of edible ﬁsh from Lake Vattern and the Baltic Sea. Chemosphere 2009, 76 (6), 799–804. (13) Kwadijk, C.; Korytar, P.; Koelmans, A. Distribution of perﬂuorinated compounds in aquatic systems in The Netherlands. Environ. Sci. Technol. 2010, 44 (10), 3746–3751. (14) Morikawa, A.; Kamei, N.; Harada, K.; Inoue, K.; Yoshinaga, T.; Saito, N.; Koizumi, A. The bioconcentration factor of perﬂuorooctane sulfonate is signiﬁcantly larger than that of perﬂuorooctanoate in wild turtles (Trachemys scripta elegans and Chinemys reevesii): An Ai river ecological study in Japan. Ecotoxicol. Environ. Saf. 2006, 65 (1), 14–21. (15) 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. (16) Houde, M.; Czub, G.; Small, J. M.; Backus, S.; Wang, X. W.; 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. (17) Quinete, N.; Wu, Q.; Zhang, T.; Yun, S. H.; Moreira, I.; Kannan, K. Speciﬁc proﬁles of perﬂuorinated compounds in surface and drinking waters and accumulation in mussels, ﬁsh, and dolphins from southeastern Brazil. Chemosphere 2009, 77 (6), 863–869. (18) Moon, H. B.; Kannan, K.; Yun, S.; An, Y. R.; Choi, S. G.; Park, J. Y.; Kim, Z. G.; Moon, D. Y.; Choi, H. G. Perﬂuorinated compounds in minke whales (Balaenoptera acutorostrata) and long-beaked common dolphins (Delphinus capensis) from Korean coastal waters. Mar. Pollut. Bull. 2010, 60 (7), 1130–1135. (19) Giesy, J. P.; Naile, J. E.; Khim, J. S.; Jones, P. D.; Newsted, J. L. Aquatic toxicology of perﬂuorinated chemicals. Rev. Environ. Contam. Toxicol. 2010, 202, 1–52. (20) Review of EPA’s Draft Risk Assessment of Potential Human Health Eﬀects Associated with PFOA and Its Salts, EPA-SAB-06-006; U.S. Environemntal Protection Agency: Washington, DC, 2006; http:// www.epa.gov/sab/pdf/2006_0120_ﬁnal_draft_pfoa_report.pdf. (21) Fromme, H.; Mosch, C.; Morovitz, M.; Alba-Alejandre, I.; Boehmer, S.; Kiranoglu, M.; Faber, F.; Hannibal, I.; Genzel-Boroviczeny, O.; Koletzko, B.; Volkel, W. Pre- and postnatal exposure to perﬂuorinated compounds (PFCs). Environ. Sci. Technol. 2010, 44 (18), 7123–7129. (22) Long-Chain Perﬂuorinated Chemicals (PFCs) Action Plan; U.S. Environemntal Protection Agency: Washington, DC, 2009; http://www.epa. gov/oppt/existingchemicals/pubs/pfcs_action_plan1230_09.pdf. (23) de Vos, M. G.; Huijbregts, M. A. J.; van den Heuvel-Greve, M. J.; Vethaak, A. D.; de Vijver, K. I. V.; Leonards, P. E. G.; van Leeuwen, S. P. J; de Voogt, P.; Hendriks, A. J. Accumulation of perﬂuorooctane sulfonate (PFOS) in the food chain of the Western Scheldt estuary:

ARTICLE

Comparing ﬁeld measurements with kinetic modeling. Chemosphere 2008, 70 (10), 1766–1773. (24) 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. (25) Jeon, J.; Kannan, K.; Lim, H. K.; Moon, H. B.; Ra, J. S.; Kim, S. D. Bioaccumulation of perﬂuorochemicals in Paciﬁc Oyster under diﬀerent salinity gradients. Environ. Sci. Technol. 2010, 44 (7), 2695–2701. (26) Martin, J. W.; Mabury, S. A.; Solomon, K. R.; Muir, D. C. G. Dietary accumulation of perﬂuorinated acids in juvenile rainbow trout (Oncorhynchus mykiss). Environ. Toxicol. Chem. 2003, 22 (1), 189–195. (27) Higgins, C. P.; McLeod, P. B.; Macmanus-Spencer, L. A.; Luthy, R. G. Bioaccumulation of perﬂuorochemicals in sediments by the aquatic oligochaete Lumbriculds variegatus. Environ. Sci. Technol. 2007, 41 (13), 4600–4606. (28) Yamashita, N.; Kannan, K.; Taniyasu, S.; Horii, Y.; Petrick, G.; Gamo, T. A global survey of perﬂuorinated acids in oceans. Mar. Pollut. Bull. 2005, 51 (812), 658–668. (29) Wang, L.; Sun, H. W.; Yang, L. R.; He, C.; Wu, W. L.; Sun, S. J. Liquid chromatography/mass spectrometry analysis of perﬂuoroalkyl carboxylic acids and perﬂuorooctanesulfonate in bivalve shells: Extraction method optimization. J. Chromatogr., A 2010, 1217 (4), 436–442. (30) Stevenson, C. N.; MacManus-Spencer, L. A.; Luckenbach, T.; Luthy, R. G.; Epel, D. New perspectives on perﬂuorochemical ecotoxicology: Inhibition and induction of an eﬄux transporter in the marine mussel, Mytilus californianus. Environ. Sci. Technol. 2006, 40 (17), 5580–5585. (31) Hu, J.; Yu, J.; Tanaka, S.; Fujii, S. Perﬂuorooctane sulfonate (PFOS) and perﬂuorooctanoic acid (PFOA) in water environment of Singapore. Water, Air, Soil Pollut. 2011, 216 (1), 179–191. (32) Sylvin, M. Surface Chemistry (Adsorption); Sarup & Sons: New Delhi, India, 2005. (33) Woodcroft, M. W.; Ellis, D. A.; Raﬀerty, S. P.; Burns, D. C.; March, R. E.; Stock, N. L.; Trumpour, K. S.; Yee, J.; Munro, K. Experimental characterization of the mechanism of perﬂuorocarboxylic acids’ liver protein bioaccumulation: The key role of the neutral species. Environ. Toxicol. Chem. 2010, 29 (8), 1669–1677. (34) Tolls, J.; Kloepper-Sams, P.; Sijm, D. T. H. M. Surfactant bioconcentration—A critical review. Chemosphere 1994, 29 (4), 693–717. (35) Vuignier, K.; Schappler, J.; Veuthey, J. L.; Carrupt, P. A.; Martel, S. Drug-protein binding: A critical review of analytical tools. Anal. Bioanal. Chem. 2010, 398 (1), 53–66.

9764

dx.doi.org/10.1021/es202078n |Environ. Sci. Technol. 2011, 45, 9758–9764

ARTICLE pubs.acs.org/est

Impact of Porous Media Grain Size on the Transport of Multi-walled Carbon Nanotubes Nikolai T. Mattison,† Denis M. O’Carroll,†,||,* R. Kerry Rowe,‡ and Elijah J. Petersen§ †

)

Department of Civil & Environmental Engineering, The University of Western Ontario, London, ON, Canada N6A 5B8 Water Research Laboratory, School of Civil and Environmental Engineering, University of New South Wales, Manly Vale, NSW, 2093, Australia ‡ GeoEngineering Centre at Queen’s-RMC, Queen’s University, Kingston, Ontario, Canada K7L 3N6 § Material Measurement Laboratory, National Institute of Standards and Technology, Gaithersburg, Maryland 20899, United States

bS Supporting Information ABSTRACT: Nanoparticles possess unique physical, electrical, and chemical properties which make them attractive for use in a wide range of consumer products. Through their manufacturing, usage, and eventual disposal, nanoparticles are expected to ultimately be released to the environment after which point they may pose environmental and human health risks. One critical component of understanding and modeling those potential risks is their transport in the subsurface environment. This study investigates the mobility of one important nanoparticle (multiwalled carbon nanotubes or MWCNTs) through porous media, and makes the ﬁrst measurements on the impact of mean collector grain size (d50) on MWCNT retention. Results from onedimensional column experiments conducted under various physical and chemical conditions coupled with results of numerical modeling assessed the suitability of traditional transport models to predict MWCNT mobility. Findings suggest that a dual deposition model coupled with site blocking greatly improves model ﬁts compared to traditional colloid ﬁltration theory. Of particular note is that the MWCNTs traveled through porous media ranging in size from ﬁne sand to silt resulting in normalized concentrations of MWCNTs in the eﬄuent in excess of 60% of the inﬂuent concentration.

’ INTRODUCTION Increased production and utilization of engineered nanomaterials in consumer products will increase the potential for their release to the environment. Carbon nanotubes (CNTs) have garnered signiﬁcant attention due to their expected widespread usage (e.g., production of plastics, catalysts, water puriﬁcation systems, and components in the electronics, aerospace and automotive industries 1). CNTs are rolled up sheets of graphene composed of one or more concentric tubes (i.e., single-walled nanotubes (SWCNT) and multi-walled nanotubes (MWCNTs), respectively). Many MWCNTs are modiﬁed to make them stable in solution such as for biomedical applications,7,8 which may make them more mobile in the environment. Moreover, some forms of modiﬁed carbon nanotubes are already commercially available. Inevitably some CNTs will eventually enter the environment through their disposal or from incidental release, yet their environmental transport behaviors are not yet well understood. In particular, they may enter soil ecosystems at high concentrations through land application of sewage sludge or from disposal of CNT-containing waste in landﬁlls.2 Although modern landﬁlls have barrier systems, their eﬀectiveness at containing CNTs is presently unknown and hence the potential for their migration through porous media in the subsurface is r 2011 American Chemical Society

of considerable interest both with regards to potential ecotoxicological risks to soil organisms3 6 and human health risks (e.g., contamination of drinking water supplies). Traditional colloid ﬁltration theory (CFT) is used to predict retention of particles in porous media systems due to gravitational sedimentation, Brownian diﬀusion, and interception.9 This theory suggests that particles are transported to and become permanently deposited in the primary minimum well of a collector of uniform surface characteristics with inﬁnite retention capacity.9,10 Research suggests that additional removal mechanisms may be operative including deposition in the secondary minimum and surface charge heterogeneity.10 14 It has been suggested that physical removal mechanisms (e.g., straining which is a function of particle to pore throat size) also play a role in the removal of colloidal particles from suspension,15,16 and may be signiﬁcant for nonspherical particles with large aspect ratios like CNTs.17,18 Research on colloidal particles has also shown that the size of collectors may not only aﬀect the degree of Received: May 18, 2011 Accepted: September 28, 2011 Revised: September 25, 2011 Published: September 28, 2011 9765

dx.doi.org/10.1021/es2017076 | Environ. Sci. Technol. 2011, 45, 9765–9775

Environmental Science & Technology straining that may be operative but also aﬀect the attachment of particles.11,19 Evidence of site blocking, where the collector surface has a ﬁnite retention capacity, has also been observed in engineered nanoparticle studies.20,21 Transport studies involving CNTs thus far have shown that CFT cannot fully describe their deposition behavior and one or more additional mechanisms are needed to explain experimental ﬁndings.17,18,20,22 23 While a large number of studies have investigated transport of SWCNTs, 17,18,20,22 23 relatively little work has been conducted to investigate MWCNT mobility. All SWCNT studies have used porous media with a mean collector diameter greater than 263 μm, although one study had a signiﬁcant fraction of clay particles (29%).17 The two experimental studies investigating the mobility of MWCNTs suggest that the deposition of MWCNTs in porous media can be described using an extension of CFT.20,24 Enhanced mobility of acid-functionalized MWCNTs through porous media (quartz sand, d50 = 476 μm) was observed above a critical pore water velocity, but mobility was substantially decreased at pore water velocities similar to those expected in natural subsurface conditions.20 Wang et al.24 found that humic acidstabilized MWCNTs (d = 35 nm) were ﬁltered by quartz sand (d50 = 355 μm) to a greater extent than (humic acid-stabilized) SWCNTs (d = 1.4 nm) for all ranges of ionic strengths tested. The aforementioned studies investigating the mobility of CNTs show that their behavior is sensitive to many experimental conditions, but one important condition not yet investigated is the impact of porous media size on CNT transport. For example, no studies have investigated CNT mobility in porous media as ﬁne as silt yet a wide range of porous media types and sizes are present in the subsurface. The goal of this study was to investigate the impact of mean collector size, pore water velocity, and ionic strength on the mobility of MWCNTs. A series of one-dimensional column experiments were carried out at pore water velocities expected under both engineered and natural subsurface conditions using porous media ranging in classiﬁcation from ﬁne sand to silt. Manipulation of ionic strength facilitated an assessment of the relative importance of nontraditional ﬁltration mechanisms (e.g., straining). Multiple pulses of MWCNTs were injected into the column to establish whether site blocking was operative and evaluate its importance to MWCNT transport.

’ MATERIALS AND METHODS Multi-Walled Carbon Nanotubes. Multi-walled carbon nanotubes (MWCNTs) were purchased from Cheap Tubes Inc. (Brattleboro, VT). MWCNTs were functionalized through the addition of surface carboxylic and hydroxyl groups using a 3:1 v/v ratio of sulfuric (95 97%) and nitric acids (70%), respectively.25 This treatment enhanced their hydrophilicity and stability in the aqueous phase. The solution was placed in a bath sonicator (Aquasonic Ultrasonic Cleaner, VWR Scientific Products, West Chester, PA) for 2 h. The functionalized MWCNTs were then collected using a 0.45 μm polytetraflouroethylene (PTFE) membrane. Boiling deionized water was used to rinse the MWCNTs until the filtrate had a neutral pH.20 Finally MWCNTs were dried in a vacuum desiccator and stored until needed. For column experiments, a dispersion of functionalized nanotubes was prepared by placing 4 mg of MWCNTs in a 250 mL beaker containing 200 mL of aqueous solution. An ultrasonic probe (Fisher Sonic Dismembrator, ARTEK System Corporation, Farmingdale, NY) was placed in the beaker, which was

ARTICLE

placed in an icewater bath, at 210 W for 45 min. The dispersed nanotubes were then mixed with an additional 300 mL of aqueous solution to reach the desired concentration of 8 mg/L. The MWCNT dispersions were left to sit for at least 24 h to allow the solution to equilibrate. Experiments were conducted immediately following this period, and nanotube agglomerates were not observed on the sides or bottoms of the beakers. Thus, 8 mg/L should be near the exact concentration given the lack of settling, but other minor forms of nanotube loss such as aerosolization during the sonication process cannot be excluded. These MWCNTs were thoroughly characterized using thermal gravimetric analysis, X-ray photoelectron spectroscopy, and scanning electron microscopy (see Supporting Information (SI) for additional details on the characterization methods). Porous Media. Four quartz sands of varying grain size (Barco, Opta Minerals Inc. Waterdown, ON) were used (SI Table S1). The f32 and f71 sands (Barco #32 and Barco #71) size distributions were as received from the manufacturer. However, the 290 fractions 1 and 2 sands were obtained by sieving Barco #290 sand; 290 fraction 1 was the combination (1:4) of the fractions retained on the Tyler (Mentor, OH) no. 140 and Tyler no. 200 sieves while 290 fraction 2 was from sand retained on the Tyler no. 325 sieve. The sands are herein referred to by their d50 value to avoid confusion between the 290 fractions. The quartz sand was cleaned by washing with hydrochloric acid (0.1M) followed by hydrogen peroxide (5%) to remove any impurities from the surface of the sand grains. The sand was rinsed repeatedly with deionized water following both the acid and peroxide washing steps to ensure that all impurities had been removed from the sand and neutral pH had been achieved. The sand was then ovendried (90 °C) overnight, following rinsing, and stored. Zeta potential of the porous media, quantified by streaming potential measurement (Anton Paar SurPASS system, Saint Laurent, Canada) using aqueous solution AS1b, were negative (i.e., ∼ 50 mV). This suggests repulsive interactions between the CNTs and porous media. Aqueous Solution Chemistry. Three aqueous solutions were employed all with a pH of approximately 7.5. The first high ionic strength solution (AS1) had an ionic strength of 7.5 mmol/L and was buffered to pH 7.5 with 1.26 mmol/L monosodium phosphate (NaH2PO4.H2O), 1.73 mmol/L (Na2HPO4) and 1 mmol/L NaBr. The background aqueous solution (i.e., no MWCNTs) used in the high ionic strength column experiments (AS1b) had the same amount of phosphate as AS1 with the addition of 1 mmol/L NaCl instead of NaBr to act as a conservative tracer. The low ionic strength solution (AS2) (0.1 mmol/L) was obtained through dilution of AS1. In the column experiments at low ionic strength, the background solution was the same as the MWCNT solution, but without MWCNTs. Column Experiments. Glass columns (2.5 cm in diameter and 5 cm in length) were used in this study. A stainless steel mesh (66 μm openings) followed by a nylon screen (25 μm openings), at column ends, were used to support the porous media and evenly distribute aqueous flow through the column (see Supporting Information for additional details on column packing). Columns were flushed with the appropriate background solution for 10 pore volumes following 30 pore volumes flushing with DI water. Flow in the 100% water saturated columns was then reduced to the experimental flow rate for at least one pore volume and the direction of flow was switched to vertically downward. Following the flow reversal, MWCNT solutions were injected into the column. Multiple 60 mL plastic syringes and a 9766

dx.doi.org/10.1021/es2017076 |Environ. Sci. Technol. 2011, 45, 9765–9775

Environmental Science & Technology syringe pump (KD Scientific, Holliston, MA) were used for solution injection. Two pulses of MWCNTs were injected with sufficient time between pulses to allow column effluent absorbance to reach background levels. Once the absorbance of the column eﬄuent had reached background levels after the second MWCNTs pulse, deionised water was ﬂushed through the column to determine if the sudden drop in ionic strength would mobilize MWCNTs that may have been deposited on the sand grains. The column eﬄuent was collected using a CF-1 fraction collector (Spectrum Chromatography, Houston, TX) and the MWCNT concentration determined using a UV spectrophotometer (Helios Alpha, Thermo Fisher Scientiﬁc, Mississauga, ON) at a wavelength of 400 nm. A calibration curve of MWCNT concentration versus absorbance was highly linear at this wavelength with R2 > 0.997. Conservative tracer concentrations (NaBr) were quantiﬁed using high performance liquid chromatography (HPLC, Waters Company, Milford, MA) of the collected fractions. Mathematical Modeling of MWCNT Transport. A onedimensional finite element model was used to simulate MWCNT transport.20 The simulator was based on traditional colloid filtration theory but was modified to include site blocking and dual deposition. Please see the SI for additional numerical modeling details.

’ RESULTS AND DISCUSSIONS MWCNT Characterization. Thermal gravimetric analysis results indicated that the catalyst impurities were only (0.8 ( 0.1) % (n = 4) of the MWCNTs; thus, metal impurities were almost completely removed (SI Figure S1). The lack of a peak at a lower temperature than the principal peak indicated a lack of amorphous carbon. X-ray photoelectron spectroscopy (XPS) (see SI Figure S2) indicated that (6.2 ( 1.4) % of the MWCNTs was oxygen, a result similar to that obtained with MWCNTs treated with the same acid modification process in a prior study.26 These oxygen functional groups and a related scan of the C (1s) region indicated the presence of carboxyl and hydroxyl functional groups. A previous study indicated that ultrasonication of an MWCNT solution without an ice water bath for 6 h only caused a slight increase in the oxygen content from 7.5% to 8.6%.27 Thus, the 45 min sonication in an ice water bath used here is not expected to substantially increase the concentration of functional groups on the MWCNTs. According to scanning electron microscopy, the measured MWCNT outer diameter was 36 ( 11 nm (n = 132), and length was 540 ( 340 nm (n = 215) (see SI Figures S3 and S4). While the diameters matched the manufacturer’s specifications, the lengths were substantially shorter than the 10 20 μm indication by the manufacturer. It was unclear if this discrepancy was a result of the sonication process or acid functionalization or if the MWCNTs were originally shorter than indicated by the manufacturer. Regardless, this result further affirms the importance of in-depth nanoparticle characterization prior to environmental behavior experiments.28 30 Column Experiments. Column experiments were conducted to assess the impact of collector grain size on the transport and deposition of MWCNTs. Column experiments were first conducted at a higher pore water velocity (4.9 10 5 m/s or 4.2 m/d) using AS1 to disperse the MWCNTs and AS1b as the background solution. The first MWCNT pulse exited the column noticeably later than the conservative tracer (NaBr) with the extent of retardation increasing with decreasing collector size

ARTICLE

(Figure 1). Following breakthrough the MWCNT normalized effluent concentration increases relatively quickly but slows at a normalized effluent concentration of approximately 0.65. After approaching or reaching a plateau injection of MWCNTs ceased and background solution injection commenced. After MWCNT injections stopped, the normalized MWCNT effluent concentrations decreased with the conservative tracer for both pulses and for both pore water velocities (Figures 1 and 2). MWCNT breakthrough was signiﬁcantly diﬀerent for the second pulse. The maximum normalized concentrations for the (476, 175, and 80) μm sands were about 0.8 for both pulses, while the eﬄuent concentration for the 50 μm sand packed column reached an eﬄuent concentration of 0.65 with the conservative tracer and then gradually rose to a maximum value of 0.75 more slowly than the tracer (Figure 1). The results observed here diﬀer from those for humic acid-stabilized MWCNTs for which MWCNT breakthrough occurred ahead of the conservative tracer at all ionic strengths investigated.24 The pore velocity used in their study was similar (8.9 10 5 m/s or 7.7 m/d) to the higher velocity used in the current study and the mean grain size was in the same range (d50 = 350 μm). Given that their diameters ((35 ( 10) nm, length not reported) were similar to the MWCNT tested here, these observed diﬀerences are likely due to diﬀerent stabilization methods and the resultant diﬀerences in interaction energies. At the lower pore velocity (4.9 10 6 m/s or 0.42 m/d), MWCNT transport was similar to that observed in the high velocity experiments in the following ways: (i) the conservative tracer exits the column more quickly than the MWCNTs; (ii) the MWCNTs show increased retardation with decreasing collector grain size; (iii) MWCNT normalized eﬄuent concentration increased at a relatively fast rate initially until a steady state eﬄuent concentration was achieved for some experimental conditions, while in other cases eﬄuent concentrations increased at a much slower rate until injection of MWCNT ceased; and (iv) limited MWCNT retardation was observed for the second pulse but this was more observable than during the high velocity experiments (Figures 1 and 2). The maximum normalized concentration obtained for the (476, 175, and 80) μm sands after a second MWCNT pulse were again similar (0.65). For the second pulse, MWCNTs exhibited a slightly delayed breakthrough in the 50 μm sand compared to the conservative tracer and reached a maximum value of 0.5 (Figure 2d). It is important to note that the MWCNTs are still mobile in the ﬁnest porous media fraction, which would be classiﬁed as silt. Upon completion of the second pulse, deionized water was injected into the columns at the experimental ﬂow rate. For all experiments this resulted in a sharp and reproducible spike in eﬄuent absorbance (e.g., Figure 3) suggesting that a fraction of the MWCNTs that were loosely deposited on collector surfaces had been remobilized.18 This release has been attributed to the increase in electrostatic repulsion due to the elimination of the secondary energy minimum by lowering the ionic strength from 7.5 mM to that of deionized water.18 The amount of MWCNTs released from the porous media did not follow any speciﬁc trend with respect to collector diameter. Jaisi et al.18 reported that the concentration released from collector surfaces in this manner generally increased following experiments of increasing ionic strength (i.e., high experiment ionic strength leads to higher concentration of CNT in the eﬄuent following the deionized solution ﬂush).18 In the current study, the same trend was observed, because MWCNTs were not released upon injection 9767

dx.doi.org/10.1021/es2017076 |Environ. Sci. Technol. 2011, 45, 9765–9775

Environmental Science & Technology of deionized water in low ionic strength (0.1 mM) experiments (Figure 4). This result is likely attributable to the relatively small change in ionic strength between the 0.1 mM solution and deionized water and the negligible secondary minimum at the low ionic strength. A suite of experiments were conducted at the low pore velocity (4.9 10 6 m/s or 0.42 m/d) and low ionic strength (0.1 mM) to see if nonphysiochemical removal mechanisms (e.g., straining) were operative (Figure 4). In all cases MWCNTs exited the column at the same time and at a similar rate as the representative tracer in the high ionic strength experiments (Figure 4) (i.e., no retardation of either pulse). When the background solution ﬂush was initiated, MWCNT eﬄuent concentrations exited the

ARTICLE

column in a similar manner to representative tracer concentrations. The maximum normalized eﬄuent concentration was approximately 0.95 except in the case of the 50 μm sand, which had a maximum normalized eﬄuent concentration of approximately 0.9. Given the relatively large repulsive barrier that exists at the low ionic strength, any MWCNT removal was likely due mechanisms other than those associated with traditional ﬁltration theory. One possible nonphysiochemical removal mechanism that may be of importance is straining,31 although given that normalized eﬄuent concentrations are above 0.9 for the low ionic strength solution, this would suggest it is not a dominant retention mechanism. Liu et. al.20 observed a maximum normalized eﬄuent concentration of MWCNTs of approximately 0.65

Figure 1. Continued 9768

dx.doi.org/10.1021/es2017076 |Environ. Sci. Technol. 2011, 45, 9765–9775

Environmental Science & Technology

ARTICLE

Figure 1. Breakthrough curves of MWCNTs at the high pore water velocity, I = 7.5 mmol/L in (a) 476 μm, (b)175 μm, (c) 80 μm and (d) 50 μm sand packed columns. Open circles represent conservative tracer data, squares represent MWCNT data and solid lines represent the results of simulation (noted as sim.). Experiments A, B, and C are replicates. RMSE values provided in SI Table S2 provide estimates of the quality of the modeling ﬁts.

at an ionic strength of 0.1 mM which was similar to the maximum concentration achieved at the high ionic strength (10 mM) in that study. This diﬀerence from our results may be due to signiﬁcant nonphysiochemical removal occurring at the low ionic strength.20 The results obtained at the low ionic strength case (0.1 mmol/L) in this study were similar in shape and appearance to those obtained in a previous study using SWCNTs.18 In that study, SWCNTs exited the column with the conservative tracer and reached a maximum normalized eﬄuent concentration of approximately 0.93. In a diﬀerent study, higher than expected removal occurred at low ionic strengths which was attributed to straining rather than ﬁltration of the CNTs.17 Along these

lines, Jaisi et al.18 observed a decrease in the hydrodynamic diameter in the SWCNT fraction that passed through the column, and postulated that this decrease may be due to larger particles being removed in the column due to straining; while dynamic light scattering (DLS) has limitations for CNTs as a result of the assumption that the particles are monodisperse spheres in the ﬁtting algorithm, it was helpful for comparing relative diﬀerences among these fractions. In this study, the inﬂuent MWCNT suspension had an average eﬀective diameter of 188 nm, measured by DLS, and an average eﬀective eﬄuent diameter of 176 nm. Given the relatively small change in eﬀective diameter, straining appears to be a relatively minor mechanism. 9769

dx.doi.org/10.1021/es2017076 |Environ. Sci. Technol. 2011, 45, 9765–9775

Environmental Science & Technology Another method to assess if straining is a dominant removal mechanism is to quantify the ratio of a representative particle diameter and a representative pore throat diameter (L/L), usually taken to be d50.32 The critical ratio was previously proposed to be 0.003.31 A study using SWCNTs reported ratios of 0.0008 and 4.7 10 6 for the SWCNT eﬀective diameter and SWCNT diameter, respectively,18 whereas an MWCNT transport study had a ratio of 0.0001.20 The straining ratio using the length (0.0011 to 0.011) and the MWCNT diameter (0.00008 0.0007) in this study are below the critical ratio proposed of 0.003 for diameter but not length. Traditionally, the porous media is treated as a “clean bed” which means that particle deposition rate is constant.33 Site blocking, which occurs when the deposition rate decreases over

ARTICLE

time (i.e., less particles are deposited onto the collector surface as it becomes covered by particles),34 has been observed in other nanoparticle transport studies.20,21 Given the retarded MWCNT breakthrough for the ﬁrst pulse, and not the second pulse, in all high ionic strength experiments, this suggests that site blocking may be an operative mechanism (Figures 1 and 2). The shape of the breakthrough curves also provides evidence for the occurrence of site blocking as the deposition rate does change with continued MWCNT transport through the column (i.e., there is no stable steady state eﬄuent concentration in some experiments but rather the maximum normalized concentration slowly increases as the deposition rate drops). These observations are consistent with previous research that has suggested site blocking was operative.21,32,34 36

Figure 2. Continued 9770

dx.doi.org/10.1021/es2017076 |Environ. Sci. Technol. 2011, 45, 9765–9775

Environmental Science & Technology

ARTICLE

Figure 2. Breakthrough curves of MWCNTs at the low pore water velocity I = 7.5 mmol/L in (a) 476 μm, (b)175 μm, (c)80 μm, (d) 50 μm sand packed columns. Open circles represent conservative tracer data, diamonds represent MWCNT data and solid lines represent the results of simulation (noted as sim.) For Figure 2d simulated lines represent mechanisms associated with traditional colloid ﬁltration theory (CFT); traditional colloid ﬁltration theory with site blocking (CFT + site blocking) and dual deposition with site blocking (DDM + site blocking) Experiments A and B are replicates. RMSE values provided in SI Table S2 provide estimates of the quality of the modeling ﬁts.

’ RESULTS OF NUMERICAL SIMULATION A series of simulations were conducted to test the ability of the numerical simulator to reproduce observed high ionic strength experiments (I = 7.5 mmol/L), using traditional CFT and modiﬁed versions of CFT. It was assumed that there was no detachment of MWCNTs from the sand surface (i.e., kdet = 0) as no tailing was observed in any of the experiments. Initially simulations only incorporated mechanisms traditionally associated with CFT (i.e., assuming ψ = 1 and αii =0, eqs 2 and 4 in SI) and αi and α1

were ﬁtted. Model results suggest that MWCNTs break through with the conservative tracer and achieve an eﬄuent concentration plateau of 0.3 (Figure 2d, other results not shown due to space limitations). Observed behaviors, which exhibited signiﬁcant MWCNT retardation for the ﬁrst pulse, were signiﬁcantly diﬀerent than these simulation results. This suggests that clean bed behavior, where an eﬄuent steady state concentration is rapidly achieved and maintained for the duration of particle injection, could not adequately describe observed experimental 9771

dx.doi.org/10.1021/es2017076 |Environ. Sci. Technol. 2011, 45, 9765–9775

Environmental Science & Technology

ARTICLE

Figure 3. Breakthough curves of MWCNTs including elution with deionized water (third pulse on graph). Open circles represent conservative tracer data, diamonds represent MWCNT data.

Figure 4. Breakthrough curves of MWCNTS for low ionic strength (0.1 mmol/L) experiment through (476, 175, 80, and 50) μm sand packed columns.

results. For comparison with subsequent simulations, the rootmean-square error (RMSE) value of this analysis was 0.167 (SI Table S2). As previously discussed, observed MWCNT transport behaviors (Figures 1a d and 2a d) suggest site blocking may be operative. A site blocking function (ψ in eq 2) was therefore incorporated in the governing mass balance equations to account

for this observed behavior. This form of ψ suggests that as S approaches Smax the blocking function approaches zero. This means that once the surface becomes covered by deposited MWCNTs, attachment will cease and the normalized MWCNT eﬄuent concentration will increase to one. The blocking function would be particularly useful in describing the reduction in MWCNT retardation between the ﬁrst and second pulses, and 9772

dx.doi.org/10.1021/es2017076 |Environ. Sci. Technol. 2011, 45, 9765–9775

Environmental Science & Technology the non clean bed plateau observed in some cases where the normalized concentration of MWCNTs slowly increases with time (Figures 1 and 2). For these simulations the values of αi, Smax, and α1 were ﬁtting parameters (i.e., assuming αii = 0). The inclusion of site blocking signiﬁcantly improved the ﬁt to experimental results (RMSE = 0.117), but the model still did not predict MWCNT retardation (Figure 2d). Furthermore, MWCNT concentrations are overestimated until 6.7 pore volumes at which point the model underpredict the normalized eﬄuent concentration of MWCNTs in the ﬁrst pulse. The model also overpredicted the MWCNT concentration during the second pulse. Based on the site blocking function adopted, as more MWCNTs were injected into the column, the magnitude of the blocking term should decrease and thus the plateau of the second pulse should have been higher than the ﬁrst pulse due to lower deposition rate as fewer sites remained available for deposition. However, this was not the case, and thus these model results suggest that the site blocking function cannot account for the similar height of the ﬁrst and second pulses and that, along with retention at a ﬁnite number of sites, other deposition mechanisms may be operative. Straining is one such mechanism, but based on the results of the low ionic strength experiments, was likely not a dominant process. Dual deposition, a process which may be due to deposition of particles in the secondary energy minimum and surface charge heterogeneity,13 was thus considered. Results from the deionized water ﬂush, following the second MWCNT pulse, support the presence of a secondary energy minimum (Figure 3). Previous studies have used this type of model to simulate similar experimental behavior.13,37 Tufenkji et al.13 found that both favorable and unfavorable sites may be present on collector surfaces. As such favorable and unfavorable deposition of particles can occur simultaneously, where a fraction of the particles experience a fast deposition rate while others experience a slow deposition rate. To account for a dual deposition model (DDM), a second removal rate constant (katt,ii) was incorporated into the solid phase mass balance equation (eq 2). The site blocking function ψ was only applied to the fast deposition rate and it was assumed that site blocking did not play a signiﬁcant role in slow deposition. Inclusion of DDM and the site blocking term signiﬁcantly improved agreement between the numerical model and experimental results (RMSE = 0.053, Figure 2d). Model results are in similar good agreement with experimental observations for all of the experimental data (i.e., both velocities) when both DDM and a site blocking term are incorporated in the governing mass balance equations (Figures 1 and 2). Fitted parameters are presented in Table S2 in the SI. The values of Smax obtained in this study were similar in magnitude to previous results for fullerene particles,21 but were an order of magnitude greater than results obtained previously for MWCNTs.20 This suggests that the collector surfaces have signiﬁcantly larger capacities for the MWCNTs used in this study in comparison to the MWCNTs used previously.20 At both high and low pore water velocities, Smax increases as the average collector grain size decreases consistent with results reported in the fullerene nanoparticle transport study.21 The increase in Smax with decreasing grain size is likely due to the increased surface area of collectors and corresponding total number of depositional sites. Smax is an exponential function of speciﬁc surface area of collectors (surface area of sphere with diameter equal to the d50 of the soil divided by its volume) (SI Figure S5). Li et al.21

ARTICLE

reported an increase in Smax as pore velocity decreases, for the same collector sizes, consistent with the results reported here. The reduction in Smax associated with increased pore water velocity may be due to increased tangential velocities across the collector surface which creates a “shadow eﬀect” down gradient of the collector due to hydrodynamic scattering.34,38 It should be noted that Smax is a weaker function of pore water velocity than collector size. katt,i and katt,ii increase with increasing pore water velocity consistent with previous CNTs transport studies 17,20 (SI Table S2). katt,i and katt,ii also generally increased with decreasing mean collector size. The impact of collector size on katt,i and katt,ii was less than that of pore water velocity. Attachment eﬃciency (α), the total number of attachments to the total number of collisions between the MWCNT particles and collectors, was calculated from ﬁtted attachment rates (katt,i and katt,ii) and estimated single collector eﬃciencies formulated based on mechanisms associated with traditional CFT20 (SI Table S2). Attachment eﬃciency decreased with pore water velocity and with mean grain size. Values ranged between 0.01 and 1.20, with the value greater than unity for the 476 μm sand at 4.2 m/d. The value greater than unity may be attributed to shortcomings of the single collector eﬃciency (η0) relationship used in this study or that unique, mean values for key MWCNT and soil properties were used in the calculation. Historically, attachment eﬃciency was assumed to only be related to chemical factors (e.g., ionic strength, pH),9 but recent studies suggest that the attachment is also a function of hydrodynamic factors.10,39 41 Shen et al.10 showed that when hydrodynamic factors are not considered (i.e., only Derjaguin Landau Verwey Overbeek (DLVO) interactions and Brownian diﬀusion were considered), the attachment eﬃciency was overestimated. Additional mechanisms (hydrodynamic eﬀects and van der Waals interactions)42 should likely be included in η0 and may reduce α below unity. Although inclusion of a DDM and a site blocking term considerably improved model agreement with experimental results, the agreement is not perfect. A number of simplifying assumptions were made in conceptual model development (i.e., unique porous media grain size and carbon nanotube dimensions). Solution of a more complex conceptual model would likely improve agreement between model and experimental results. These ﬁtted parameters suggest that under the conditions investigated the maximum travel distance of MWCNTs (calculated as the distance needed to reduce the eﬄuent concentration of MWCNTs to 0.1% of the initial concentration) observed decreases from 1.2 m in ﬁne sand to 0.9 m in silts (SI Table S2) at the higher pore water velocity. These distances are obtained from values of katt.ii, but it is unknown if deposition sites associated with this term are subject to site blocking which could signiﬁcantly increase travel distances of MWCNTs.

’ ENVIRONMENTAL IMPLICATIONS Results from this study suggest that commercially available MWCNTs are mobile in porous media with collector sizes ranging from ﬁne sands to silt. Diﬀerences in breakthrough curves between the ﬁrst and second pulses indicate that the deposition rate decreases as the experiment progresses and that site-blocking may be operative. These results could not be predicted using traditional CFT. However, a dual deposition model with a site blocking term signiﬁcantly improved agreement between experimental observations and model results. 9773

dx.doi.org/10.1021/es2017076 |Environ. Sci. Technol. 2011, 45, 9765–9775

Environmental Science & Technology Increased MWCNT retardation with decreasing collector size was likely due to the increased surface area and depositional sites. Nonphysiochemical removal mechanisms (e.g., straining) may inﬂuence MWCNTs deposition, but low ionic strength (0.1 mmol/L) results suggest that this contribution is minor. These ﬁndings are based on fundamental experiments incorporating ideal porous media systems. The inclusion of organic matter, divalent cations (e.g., Ca2+ or Mg2+) and both chemical and physical heterogeneity, as would be expected in subsurface environments, would impact the mobility of MWCNTs and are important topics for additional research. Moreover, the CNTs used in this manuscript were functionalized to make them more stable in solution. Nonfunctionalized may exhibit diﬀerent transport behaviors. More research is needed to fully understand the complexities of CNT transport. Findings suggest that site blocking may increase the overall travel distances of MWCNTs in the subsurface beyond what is expected from traditional CFT. This study does highlight the importance of collector size on MWCNT transport and shows that MWCNTs are mobile through increasingly ﬁner material which are present in subsurface environments. Further work is necessary to assess the toxicity of CNTs that could be dispersed in aqueous solutions, be mobile in subsurface systems, as suggested in this study, and contaminate aquifers used as sources of drinking water.

’ ASSOCIATED CONTENT

bS

Supporting Information. Additional experimental methods, MWCNT characterization data, comparison of Smax and speciﬁc surface area of sands, summary ﬁgures for the ﬁrst pulse at the higher and lower velocities, summary of model parameters, and a summary of experiments. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: (519)-661-2198; e-mail: [email protected].

’ ACKNOWLEDGMENT Certain commercial equipment or materials are identiﬁed in this paper in order to specify adequately the experimental procedure. Such identiﬁcation does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the materials or equipment identiﬁed are necessarily the best available for the purpose. This research was supported by an Ontario Ministry of the Environment Best in Science Award, the Natural Sciences and Engineering Research Council of Canada and the Canadian Foundation for Innovation. We thank Howard Fairbrother for his helpful advice on MWCNT stability. ’ REFERENCES (1) Klaine, S. J.; Alvarez, P. J. J.; Batley, G. E.; Fernandes, T. F.; Handy, R. D.; Lyon, D. Y.; Mahendra, S.; McLaughlin, M. J.; Lead, J. R. Nanomaterials in the environment: Behavior, fate, bioavailability, and eﬀects. Environ. Toxicol. Chem. 2008, 27 (9), 1825–1851. (2) Gottschalk, F.; Sonderer, T.; Scholz, R. W.; Nowack, B. Modeled Environmental concentrations of engineered nanomaterials (TiO2, ZnO, Ag, CNT, fullerenes) for diﬀerent regions. Environ. Sci. Technol. 2009, 43 (24), 9216–9222.

ARTICLE

(3) Petersen, E. J.; Huang, Q. G.; Weber, W. J. Relevance of octanolwater distribution measurements to the potential ecological uptake of multi-walled carbon nanotubes. Environ. Toxicol. Chem. 2010, 29 (5), 1106–1112. (4) Petersen, E. J.; Huang, Q. G.; Weber, W. J. Bioaccumulation of radio-labeled carbon nanotubes by Eisenia foetida. Environ. Sci. Technol. 2008, 42 (8), 3090–3095. (5) Petersen, E. J.; Pinto, R. A.; Landrum, P. F.; Weber, W. J. Inﬂuence of carbon nanotubes on pyrene bioaccumulation from contaminated soils by earthworms. Environ. Sci. Technol. 2009, 43 (11), 4181–4187. (6) Petersen, E. J.; Pinto, R. A.; Zhang, L. W.; Huang, Q. G.; Landrum, P. F.; Weber, W. J. Eﬀects of Polyethyleneimine-mediated functionalization of multi-walled carbon nanotubes on earthworm bioaccumulation and sorption by soils. Environ. Sci. Technol. 2011, 45 (8), 3718–3724. (7) Shen, M. W.; Wang, S. H.; Shi, X. Y.; Chen, X. S.; Huang, Q. G.; Petersen, E. J.; Pinto, R. A.; Baker, J. R.; Weber, W. J. Polyethyleneiminemediated functionalization of multi-walled carbon nanotubes: Synthesis, characterization, and in vitro toxicity assay. J. Phys. Chem. C 2009, 113 (8), 3150–3156. (8) Shi, X. Y.; Wang, S. H.; Shen, M. W.; Antwerp, M. E.; Chen, X. S.; Li, C.; Petersen, E. J.; Huang, Q. G.; Weber, W. J.; Baker, J. R. Multifunctional dendrimer-modiﬁed multi-walled carbon nanotubes: Synthesis, characterization, and in vitro cancer cell targeting and imaging. Biomacromolecules 2009, 10 (7), 1744–1750. (9) Yao, K. M.; Habibian, M. M.; Omelia, C. R. Water and waste water ﬁltration—Concepts and applications. Environ. Sci. Technol. 1971, 5 (11), 1105–1112. (10) Shen, C.; Li, B.; Huang, Y.; Jin, Y. Kinetics of Coupled Primaryand Secondary-Minimum Deposition of Colloids under Unfavorable Chemical Conditions. Environ. Sci. Technol. 2007, 41 (20), 6976–6982. (11) Shen, C. Y.; Huang, Y. F.; Li, B. G.; Jin, Y., Predicting attachment eﬃciency of colloid deposition under unfavorable attachment conditions. Water Resour Res 2010, 46, -. (12) Elimelech, M.; Omelia, C. R. Eﬀect of particle-size on collision eﬃciency in the deposition of brownian particles with electrostatic energy barriers. Langmuir 1990, 6 (6), 1153–1163. (13) Tufenkji, N.; Elimelech, M. Deviation from the classical colloid ﬁltration theory in the presence of repulsive DLVO interactions. Langmuir 2004, 20 (25), 10818–10828. (14) Tufenkji, N.; Elimelech, M. Breakdown of colloid ﬁltration theory: Role of the secondary energy minimum and surface charge heterogeneities. Langmuir 2005, 21 (3), 841–852. (15) Bradford, S. A.; Yates, S. R.; Bettahar, M.; Simunek, J., Physical factors aﬀecting the transport and fate of colloids in saturated porous media. Water Resour Res 2002, 38, (12), 1327. (16) Bradford, S. A.; Bettahar, M.; Simunek, J.; van Genuchten, M. T. Straining and attachment of colloids in physically heterogeneous porous media. Vadose Zone J 2004, 3 (2), 384–394. (17) Jaisi, D. P.; Elimelech, M. Single-walled carbon nanotubes exhibit limited transport in soil columns. Environ. Sci. Technol. 2009, 43 (24), 9161–9166. (18) Jaisi, D. P.; Saleh, N. B.; Blake, R. E.; Elimelech, M. Transport of single-walled carbon nanotubes in porous media: Filtration mechanisms and reversibility. Environ. Sci. Technol. 2008, 42 (22), 8317–8323. (19) Torkzaban, S.; Bradford, S. A.; Walker, S. L. Resolving the coupled eﬀects of hydrodynamics and DLVO forces on colloid attachment in porous media. Langmuir 2007, 23 (19), 9652–9660. (20) Liu, X. Y.; O’Carroll, D. M.; Petersen, E. J.; Huang, Q. G.; Anderson, C. L. Mobility of multi-walled carbon nanotubes in porous media. Environ. Sci. Technol. 2009, 43 (21), 8153–8158. (21) Li, Y. S.; Wang, Y. G.; Pennell, K. D.; Abriola, L. M. Investigation of the transport and deposition of fullerene (C60) nanoparticles in quartz sands under varying ﬂow conditions. Environ. Sci. Technol. 2008, 42 (19), 7174–7180. (22) Lecoanet, H. F.; Wiesner, M. R. Velocity eﬀects on fullerene and oxide nanoparticle deposition in porous media. Environ. Sci. Technol. 2004, 38 (16), 4377–4382. 9774

dx.doi.org/10.1021/es2017076 |Environ. Sci. Technol. 2011, 45, 9765–9775

Environmental Science & Technology

ARTICLE

(23) Lecoanet, H. F.; Bottero, J. Y.; Wiesner, M. R. Laboratory assessment of the mobility of nanomaterials in porous media. Environ. Sci. Technol. 2004, 38 (19), 5164–5169. (24) Wang, P.; Shi, Q. H.; Liang, H. J.; Steuerman, D. W.; Stucky, G. D.; Keller, A. A. Enhanced environmental mobility of carbon nanotubes in the presence of humic acid and their removal from aqueous solution. Small 2008, 4 (12), 2166–2170. (25) Liu, J.; Rinzler, A. G.; Dai, H. J.; Hafner, J. H.; Bradley, R. K.; Boul, P. J.; Lu, A.; Iverson, T.; Shelimov, K.; Huﬀman, C. B.; RodriguezMacias, F.; Shon, Y. S.; Lee, T. R.; Colbert, D. T.; Smalley, R. E. Fullerene pipes. Science 1998, 280 (5367), 1253–1256. (26) Petersen, E. J.; Akkanen, J.; Kukkonen, J. V. K.; Weber, W. J., Jr. Biological uptake and depuration of carbon nanotubes by Daphnia magna. Env. Sci. & Tech. 2009, 43 (8), 2969–2975. (27) Zhang, L.; Petersen, E. J.; Huang, Q. G. Phase distribution of 14C-labeled multi-walled carbon nanotubes in aqueous systems containing model solids: Peat. Environ. Sci. Technol. 2011, 45 (4), 1356–1362. (28) Petersen, E. J.; Nelson, B. C. Mechanisms and measurements of nanomaterial-induced oxidative damage to DNA. Anal. Bioanal. Chem. 2010, 398 (2), 613–650. (29) Warheit, D. B. How meaningful are the results of nanotoxicity studies in the absence of adequate material characterization? Toxicol Sci. 2008, 101 (2), 183–185. (30) Park, H.; Grassian, V. H. Commercially manufactured engineered nanomaterials for environmental and health studies: Important Insights provided by independent characterization. Environ. Toxicol. Chem. 2010, 29 (3), 715–721. (31) Bradford, S. A.; Torkzaban, S.; Walker, S. L. Coupling of physical and chemical mechanisms of colloid straining in saturated porous media. Water Res. 2007, 41 (13), 3012–3024. (32) Bradford, S. A.; Simunek, J.; Bettahar, M.; van Genuchten, M. T.; Yates, S. R., Signiﬁcance of straining in colloid deposition: Evidence and implications. Water Resour Res 2006, 42, (12), W12S15. (33) Elimelech, M.; Gregory, J.; Jia, X.; Williams, R. A., Particle Deposition and Aggregation: Measurement, Modeling, and Simulation; Butterworth-Heinemann: Oxford, 1995; p 441. (34) Ko, C. H.; Bhattacharjee, S.; Elimelech, M. Coupled inﬂuence of colloidal and hydrodynamic interactions on the RSA dynamic blocking function for particle deposition onto packed spherical collectors. J. Colloid Interface Sci. 2000, 229 (2), 554–567. (35) Kuhnen, F.; Barmettler, K.; Bhattacharjee, S.; Elimelech, M.; Kretzschmar, R. Transport of iron oxide colloids in packed quartz sand media: Monolayer and multilayer deposition. J. Colloid Interface Sci. 2000, 231 (1), 32–41. (36) Kretzschmar, R.; Robarge, W. P.; Amoozegar, A. Inﬂuence of natural organic-matter on colloid transport through saprolite. Water Resour. Res. 1995, 31 (3), 435–445. (37) Tufenkji, N.; Elimelech, M. Spatial distributions of Cryptosporidium oocysts in porous media: Evidence for dual mode deposition. Environ. Sci. Technol. 2005, 39 (10), 3620–3629. (38) Johnson, P. R.; Elimelech, M. Dynamics of colloid deposition in porous-media—Blocking based on random sequential adsorption. Langmuir 1995, 11 (3), 801–812. (39) Tong, M. P.; Johnson, W. P. Excess colloid retention in porous media as a function of colloid size, ﬂuid velocity, and grain angularity. Environ. Sci. Technol. 2006, 40 (24), 7725–7731. (40) Johnson, W. P.; Tong, M. P. Observed and simulated ﬂuid drag eﬀects on colloid deposition in the presence of an energy barrier in an impinging jet system. Environ. Sci. Technol. 2006, 40 (16), 5015–5021. (41) Li, X. Q.; Zhang, P. F.; Lin, C. L.; Johnson, W. P. Role of hydrodynamic drag on microsphere deposition and re-entrainment in porous media under unfavorable conditions. Environ. Sci. Technol. 2005, 39 (11), 4012–4020. (42) Tufenkji, N.; Elimelech, M. Correlation equation for predicting single-collector eﬃciency in physicochemical ﬁltration in saturated porous media. Environ. Sci. Technol. 2004, 38 (2), 529–536. 9775

dx.doi.org/10.1021/es2017076 |Environ. Sci. Technol. 2011, 45, 9765–9775

ARTICLE pubs.acs.org/est

Toxicity of Individual Naphthenic Acids to Vibrio fischeri David Jones, Alan G. Scarlett, Charles E. West, and Steven J. Rowland* Petroleum and Environmental Geochemistry Group, Biogeochemistry Research Centre, University of Plymouth, Drake Circus, Plymouth, PL4 8AA, United Kingdom

bS Supporting Information ABSTRACT: Numerous studies have suggested that the toxicity of organic compounds containing at least one carboxylic acid group and broadly classiﬁed as “naphthenic acids”, is of environmental concern. For example, the acute toxicity of the more than 1 billion m3 of oil sands process-aﬀected water and the hormonal activity of some oﬀshore produced waters has been attributed to the acids. However, experimental evidence for the toxicity of the individual acids causing these eﬀects has not been very forthcoming. Instead, most data have been gathered from assays of incompletely characterized extracts of the water, which may contain other toxic constituents. An alternative approach is to assay the individual identiﬁed toxicants. Since numerous petroleum-derived naphthenic acids and some in oil sands process water, have recently been identiﬁed, we were able to measure the toxicity of some individual acids to the bioluminescent bacterium, Vibrio ﬁscheri. Thirty-ﬁve pure individual acids were either synthesized or purchased for this purpose. We also used the US EPA ECOSAR computer model to predict the toxicity of each acid to the water ﬂea, Daphnia magna. Both are wellaccepted toxicological screening end points. The results show how toxic some of the naphthenic acids really are (e.g., V. ﬁscheri Eﬀective Concentrations for 50% response (EC50) 0.004 to 0.7 mM) and reveal the inﬂuence of hydrophobicity and aqueous solubility on the toxicities. Comparison with measured toxicities of other known, but more minor, constituents of oil sands process water, such as polycyclic aromatic hydrocarbons and alkylphenols, helps place these toxicities into a wider context. Given the reported toxicological eﬀects of naphthenic acids to other organisms (e.g., ﬁsh, plants), the toxicities of the acids to further end points should now be determined.

’ INTRODUCTION The toxicity of the so-called naphthenic acids from commercial (petroleum) sources, from oﬀshore oil platforms and from oil sands, to a variety of organisms, has been widely studied, particularly recently.1 5 However, most experimental evidence for the toxicity has been gathered from assays of incompletely characterized extracts2 5 and it has proven diﬃcult to ascribe the toxicities to individual components or even to compound classes in such mixtures, except in one or two cases.1 Such mixtures may contain toxic components other than acids.2,6,7 Therefore, recent advances in the characterization of both oil sands and commercial petroleum-derived naphthenic acids,8 10 which have allowed some of the individual carboxylic acids to be identiﬁed, have assumed considerable importance. Toxicity assays can now, in principle, be carried out on acids which are actually known to be present in the mixtures. Such an approach may help to deﬁne which of the acids are most toxic, and/or whether components other than acids should also be studied. However, pure samples of individual acids in at least milligram quantities are required before such assays are possible. We synthesized numerous examples of individual acids recently11 and now report use of a well-accepted toxicity screening assay to determine the toxicity of r 2011 American Chemical Society

over thirty-ﬁve of these acids. We also show the relationship between the measured toxicity of the acids to the bioluminescent bacterium, Vibrio ﬁscheri, and the toxicity to the water ﬂea, Daphnia magna, predicted by the US EPA computer modeling program ECOSAR.12

’ MATERIALS AND METHODS Reference Acids. Alkylphenylalkanoate and cis/trans alkylcyclohexylalkanoate (alkyl = methyl to hexyl, nonyl; alkanoate = ethanoate and butanoate) acids were synthesized by a route based on the Kindler modification of the Willgerodt reaction.11 Other acids were obtained by Freidel-Crafts13 or Willgerodt chemistry.11 Straight chain and methyl branched, phytanic, citronellic, adamantane-1-carboxylic, adamantane-1-ethanoic, and 3,5-dimethyl-adamantane-1-carboxylic acids were purchased from Sigma (U.K.) with stated purities g97%. Decalin carboxylic, decalin Received: June 8, 2011 Accepted: September 26, 2011 Revised: August 26, 2011 Published: September 26, 2011 9776

dx.doi.org/10.1021/es201948j | Environ. Sci. Technol. 2011, 45, 9776–9782

Environmental Science & Technology

ARTICLE

Table 1. Measured EC50 Values (( Standard Error, n = 3) for the Toxicity of Individual Carboxylic Acids to Vibrio ﬁscheri (Microtox Assay) with Log Kow Values Used to Generate ECOSAR Predictionsa

9777

dx.doi.org/10.1021/es201948j |Environ. Sci. Technol. 2011, 45, 9776–9782

Environmental Science & Technology

ARTICLE

Table 1. Continued

9778

dx.doi.org/10.1021/es201948j |Environ. Sci. Technol. 2011, 45, 9776–9782

Environmental Science & Technology

ARTICLE

Table 1. Continued

a

Cn = carbon number. S = synthetic, C = commercial.

ethanoic, and decalin propanoic acids were synthesized from the aromatic analogues by hydrogenation.11,13 The 2,6-dimethylheptanoic and 2,6,10-trimethylundecanoic acids were synthesized previously and 3,7-dimethyloctanoic acid was obtained by hydrogenation of citronellic acid (PtO2, 8 psi H2). Purity of methyl esters of synthetic acids was assayed by gas chromatography mass spectrometry and was generally g99%. Microtox (Vibrio fischeri) Assay. Synthesized acids were carefully weighed on an Oxford five figure balance and placed into preweighed and prerinsed 7 mL vials. The acids (maximum 20 mg) were subsequently dissolved in 1 or 2 mL of 1 M NaOH and mixed on an autovortex mixer (Stuart Scientific) until completely dissolved. These solutions were then pH adjusted by dropwise titration with a Pasteur pipet with 1, 0.1, and 0.01 M HCL until a pH of between 6 and 8 was achieved. The amount of HCL/NaOH added to the solution by the titration was then calculated (on average 0.04 mL per drop) and a calculated amount was added to 10.0 mL of Microtox diluent which was then agitated by an autovortex mixer and titrated using 0.01 M HCL/NaOH until a pH of 7.5 ( 0.1 was achieved. The concentrations of these solutions ranged between 5 and 200 mg L 1 ((2%) depending on predicted solubility and toxicity. The acids tested were initially screened on the Microtox M500 analyzer (SDI Europe) using the 45% basic (15 min) test as a toxicity screening method. Briefly: Microtox glass cuvettes were placed in the M500 analyzer and an amount of diluent was added (either 1000 or 500 μL depending on position in analyzer). 2500 μL of acid solution was then added and osmotically adjusted, this was mixed and an amount was discarded before a 2 serial dilution was carried out. The subsequent test was performed (in triplicate) according to Microtox protocols. Phenol was tested as a positive control and found to be within the parameters set by Microtox (viz: EC50 between 13 and 26 mg L 1).

’ RESULTS AND DISCUSSION We measured the concentrations required to produce a 50% decrease (EC50) in the bioluminescence of the bacterium, Vibrio ﬁscheri, in triplicate experiments at pH 7.5 (Table 1). Given

the pKa of such acids, at this pH the EC50 values therefore relate to those of free acids and/or carboxylate ions. EC50 values ranged from 0.004 to about 0.7 mM, depending on both the structures and the carbon numbers of the acids (Table 1). The mechanism of action of this toxicity is widely accepted to be due to nonspeciﬁc narcosis12 and the decreases in EC50, representing increasing toxicity, with increased carbon number of the acids (Figures 1 and 2), were consistent with this. The toxicities of most of the acid classes approached an eﬀective limit at the higher carbon numbers assayed (Figures 1 and 2) and above these, the acids were insoluble at the pH of the assay. The maximum average toxicity was ca. 0.016 mM. So-called “straight-chain” or normal (n- or fatty) acids in oil sands process water extracts range from about C9 (nonanoic) to C15 (pentadecanoic) and in petroleum-derived acids from about C8 (octanoic) to C18 (octadecanoic).10,14 However, the proportions in the oil sands extracts appear to be low and are dominated by even numbered acids (Rowland et al., unpublished data and reference 14; earlier low resolution data may have been in error14), likely of biological origin. Our results (Table 1; Figure 1) show that n-C12 (dodecanoic) acid had the highest toxicity (EC50 0.019 ( 0.002 mM) of those n-acids tested and that acids larger than this were too insoluble to be assayed. For hexanoic acid and decanoic acid, Frank et al12 reported V. ﬁscheri EC50s of 19.1 and 0.33 mM respectively; a D. magna LC50 of 10.04 mM and ECOSAR predicted LC50 of 7.15 mM were also reported. Oil sands acids from Alberta, Canada, also include monomethyl branched acids over a similar carbon number range, (but again in low proportions).14 A more complex distribution was present in a sample of petroleum-derived acids.10 Our measurements indicated (Table 1; Figure 1) that C11 13 acids were most toxic, maximizing at 0.012 ( 0.001 mM, before a solubility limit was reached (Figure 1). Polymethyl branched, acyclic isoprenoid acids have yet to be reported in oil sands acids mixtures, but are common in biodegraded petroleum due to microbial oxidation of the related hydrocarbons which are virtually ubiquitous in petroleum. Unsurprisingly therefore, they were abundant in a sample of petroleum-derived naphthenic acids10 and are likely present in at 9779

dx.doi.org/10.1021/es201948j |Environ. Sci. Technol. 2011, 45, 9776–9782

Environmental Science & Technology

ARTICLE

Figure 1. Measured EC50 values ((standard error, n = 3) for the toxicity of individual carboxylic acids to Vibrio ﬁscheri (Microtox assay). Where error bars are not apparent the error was smaller than the symbol. The identities of the individual acids are shown in Table 1 along with the numerical EC50 values. R2 and P values represent the goodness of ﬁt of the polynomial trend lines.

least some oil sands acids, which are well-known to vary in composition with age. C10 14 acids were relatively toxic (Table 1: up to 0.015 ( 0.007 mM) but by C20 (phytanic acid) a solubility limit had been reached. Monocyclic acids are apparently not abundant in oil sands acids,14 but were present in petroleum-derived acids.10 Again, C12 14 acids were most toxic (Table 1; up to 0.012 mM).

Bicyclic acids are abundant (ca. 30%) in both oil sands and petroleum-derived naphthenic acids mixtures,10,14 but to date none have been ﬁrmly identiﬁed in the former. Bicyclic acids in a petroleum-derived naphthenic acids mixture however, included numerous decalin-type compounds10 and we therefore determined the toxicity of such species as the best models currently available. No doubt, more relevant acids can be synthesized, once 9780

dx.doi.org/10.1021/es201948j |Environ. Sci. Technol. 2011, 45, 9776–9782

Environmental Science & Technology

Figure 2. Summarized measured EC50 values ((standard error, n = 3) for the toxicity of all tested carboxylic acids to Vibrio ﬁscheri (Microtox assay). Depending on the individual acid class (e.g., n-acids, C13) there was a solubility limit beyond which the acids were too hydrophobic to dissolve at pH 7.5.

the individual bicyclic compounds in oil sands process water have been ﬁrmly identiﬁed. The bicyclic acids included the most toxic of the diverse ranges of acids tested (3-decalin-1-yl propanoic acid; Table 1; 0.004 mM). Tricyclic acids, along with the bicyclic compounds are often the most abundant species present in oil sands process-aﬀected water.8,14 Numerous structures have been identiﬁed as “nanodiamond” adamantane acids (Table 1; Figure 1) and some of these are also present in petroleum-derived acids.8,10 The maximum EC50 was 0.337 ((0.022 mM). Interestingly, the extrapolated trend was somewhat diﬀerent to those of the other acids (Figure 1). The trends of EC50 versus carbon number for the other acids approached asymptotes (Figure 2), whereas that of the diamondoid acids did not. Thus, C15 adamantane acids (if soluble) might exhibit toxicity at least comparable with the other classes of acids studied herein. Finally, we measured the toxicities of a range of monoaromatic acids found in petroleum-derived naphthenic acids from a commercial suppier,11 but not, to date, in oil sands acids from Alberta, Canada8 (Table 1; Figure 1), though some aromatics are likely present. These too followed a clear relationship with carbon number with a maximum toxicity of 0.023 ( 0.005 mM; C14). The eﬀective solubility limit observed in our assay with several acid classes (Figure 2) likely also applies to many of the constituents of the complex naphthenic acids mixtures in oil sands and petroleum, though overall lower concentrations of individual acids in such mixtures and possibly some synergistic or antagonistic cosolubilization, may extend or limit the eﬀective toxicity ranges of these somewhat. This solubility limit correlation with carbon number suggests that the gC15 pentacyclic diamondoid acids identiﬁed in oil sands process-aﬀected water,9 would probably be too insoluble to exhibit a toxic eﬀect to V. ﬁscheri. Unfortunately insuﬃcient amounts or diversities of such diamantane acids were available for assay, so we decided to estimate the toxicity using the U.S. Environmental Protection Agency ECOSAR program.12 This model is also based on action by narcotic activity12 and allows calculation of octanol water coeﬃcients, solubility and thereby toxicity, to various biological

ARTICLE

end points. We chose to compute the toxicities of the acids to the water ﬂea, D. magna (lethal response in 50% of a population of daphnids after 48 h (LC50)), since a previous study12 established a relationship between this widely used end-point and three acids. We noted (Figure 1S of the Supporting Information) that there were indeed, clear relationships between the measured (V. ﬁscheri, x axis) and modeled (D. magna, y-axis) end points for each acid class measured, though the responses were not always linear, nor 1:1, nor consistent between classes (R2 = 0.9095 to 0.9865, p < 0.001 to >0.1). Of course, there is no reason ipso facto, why the toxicities measured or predicted to these two end points, should be the same. Concentrations predicted to produce a LC50 ranged from 4 mM for the least toxic n-hexanoic acid to 0.004 mM for the most toxic, 3-decalin-1-yl propanoic acid. When the model was used to compute the toxicities of the pentacyclic diamantane acids, however, the results suggested the acids would be too insoluble to be reliably predicted. This data set may however be useful for future QSAR studies. Since the high concentrations of the acids in oil sands processaﬀected water have led others to conclude that such components are responsible for the measured toxicities to a range of biota,2 5,12 it would be useful to try to gauge what proportion of the observed toxicity might be assignable to the individual acids we have studied. However, the assays of oil sands or petroleum naphthenic acids published previously have all been carried out on mixtures in which, not only the individual acid constituents were unknown, but in which many toxic nonacids may have been present. For example, the acid extracts of oil sands are known to contain so-called O3, O4 and heteroatomic compounds15 and phenols and polycyclic aromatic hydrocarbons may also be present,16,17 as the emulsifying behavior of acids can make it diﬃcult to achieve clear separations of the acids from other toxic constituents. Similarly, some petroleum-derived naphthenic acids contain phenols and PAH.7,10 It is thus inappropriate to compare the measured toxicities with those of the acids measured here, in most instances. However, Frank et al.,18 reacted the acid extractables from oil sands process water from Alberta, Canada with diazomethane to convert those compounds containing esteriﬁable carboxylic acid groups to the corresponding methyl esters, separated these by distillation and resaponiﬁed the distilled ester fractions back to the carboxylic acids. The toxicities of the distilled, saponiﬁed fractions to V. ﬁscheri were then assayed. While compounds other than acids could still be present in such fractions, none were detected by comprehensive twodimensional gas chromatography mass spectrometrycf.8 Conversion of the toxicities of these distilled oil sands acids to mM concentrations by dividing concentrations by the measured18 median molecular weights of each distilled fraction, produced EC50 values of 0.14 to 0.28 mM. These values compare to our measured values of 0.004 to 0.7 mM for the individual acids. Thus, mixtures of various classes of the acids tested, especially in the range C10 14 (typically with EC50 values of <0.3 mM) and importantly, now including the n-, methyl branched, and adamantane acids known to be present in oils sands process-aﬀected water,8 appear to be suﬃcient to account for the toxicity of oil sands naphthenic acids to V. ﬁscheri. While numerous other compounds, such as alkylphenols and PAHs, are known to be present in oils sands process-aﬀected water16,17 and some are more toxic than most of the acids tested herein (EC50 V. ﬁscheri, 3-isopropylphenol 0.002 mM; 4-isopropylphenol 0.000 07 mM),19 their concentrations are reportedly very much lower than those of the acids (cf. <0.3 mg L 1 9781

dx.doi.org/10.1021/es201948j |Environ. Sci. Technol. 2011, 45, 9776–9782

Environmental Science & Technology alkylphenols;17 20 120 mg L 1 acids18) . Thus the contributions of these to the total toxic eﬀect on V. ﬁscheri should be low. Nonetheless the concentrations of these and of the individual acids assayed herein and their toxicities to other biological end points, should also be measured. Once further unknowns in oil sands process-aﬀected waters and oilﬁeld produced waters have been identiﬁed, their toxicities can also be assayed.

’ ASSOCIATED CONTENT

bS

Supporting Information. Graphs show predicted toxicities (LC50, 48 h assay, ECOSAR) of individual acids to Daphnia magna plotted versus measured toxicities to Vibrio ﬁscheri (15 min EC50). 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 technical oﬃcer A. Tonkin for help with supervision of syntheses and recrystallisation of alkylphenylalkanoic acids to white solids. We are grateful to Dr E. Teuten (University of Plymouth) and Professor G. Dacremont (University of Ghent) for gifts of synthetic 2,6-dimethylheptanoic and 2,6,10-trimethylsundecanoic acids respectively. Funding of this research was provided by an Advanced Investigators Grant (No. 228149) awarded to S.J.R. for project OUTREACH, by the European Research Council, to whom we are extremely grateful. ’ REFERENCES (1) Thomas, K. V.; Langford, K.; Petersen, K.; Smith, A. J.; Tollefsen, K.-E. Eﬀect-directed identiﬁcation of naphthenic acids as important in vitro xeno-estrogens and anti-androgens in North Sea oﬀshore produced water discharges. Environ. Sci. Technol. 2009, 43, 8066–8071. (2) Garcia-Garcia E.; Pun J.; Hodgkinson J.; Perez-Estrada L. A.; El-Din M. G.; Smith D. W.; Martin J. W.; Belosevic M. Commercial naphthenic acids and the organic fraction of oil sands process water induce diﬀerent eﬀects on pro-inﬂammatory gene expression and macrophage phagocytosis in mice. J. Appl. Toxicol., 2011, doi: 10.1002/jat.1687 (3) Young R. F.; Martinez M. L.; Fedorak P. M. Distribution of naphthenic acids in tissues of laboratory-exposed ﬁsh and in wild ﬁshes from near the Athabasca oil sands in Alberta, Canada. Ecotox. Environ. Safety, 2011, doi:10.1016/j.ecoenv.2010.12.009. (4) Kavanagh, R. J.; Frank, R. A.; Oakes, K. D.; Servos, M. R.; Young, R. F.; Fedorak, P. M.; MacKinnon, M. D.; Solomon, K. R.; Dixon, D. G.; Van Der Kraak, G. Fathead minnow (Pimephales promelas) reproduction is impaired in aged oil sands process-aﬀected waters. Aquat. Toxicol. 2011, 101, 214–220. (5) Zhang X.; Wiseman S.; Yu H.; Liu H.; Giesy J. P.; Hecker M. Assessing the toxicity of naphthenic acids using a microbial genome wide live cell reporter array system. Environ. Sci. Technol. 2011, dx.doi.org/ 10.1021/es1032579. (6) Peters, L. E.; MacKinnon, M.; Van Meer, T.; van den Heuvel, M. R.; Dixon, D. G. Eﬀects of oil sands process-aﬀected waters and naphthenic acids on yellow perch (Perca ﬂavescens) and Japanese medaka (Orizias latipes) embryonic development. Chemosphere 2007, 67, 2177–2183. (7) West, C. E.; Scarlett, A. G.; Jones, D.; Rowland, S. J. Compositional heterogeneity may limit usefulness of some commercial naphthenic acids for toxicity assays. Sci. Total Environ. 2011, 409, 4125 4131.

ARTICLE

(8) Rowland, S. J.; Scarlett, A. G.; West, C. E.; Jones, D.; Frank, R. A. Diamonds in the rough: identiﬁcation of individual naphthenic acids in oil sands process water. Environ. Sci. Technol. 2011, 45, 3154–3159. (9) Rowland, S. J.; West, C. E.; Scarlett, A. G.; Jones, D.; Frank, R. A. Identiﬁcation of individual tetra- and pentacyclic naphthenic acids in oil sands process water by comprehensive two-dimensional gas chromatography-mass spectrometry. Rapid Commun. Mass Spectrom. 2011, 25, 1198–1204. (10) Rowland, S. J.; West, C. E.; Scarlett, A. G.; Jones, D. Identiﬁcation of individual acids in a commercial sample of naphthenic acids from petroleum by two dimensional comprehensive gas chromatographymass spectrometry. Rapid Commun. Mass Spectrom. 2011, 25, 1741– 1751. (11) Rowland S. J.; West C. E.; Scarlett A. G.; Jones D.; Boberek M.; Pan L.; Ng M.; Kwong L.; Tonkin A. Monocyclic and monoaromatic naphthenic acids: Synthesis & characterisation. Environ. Chem. Lett. 2011, doi: 10.1007/s10311-011-0314-6. (12) Frank, R. A.; Sanderson, H.; Kavanagh, R.; Burnison, B. K.; Headley, J. V.; Solomon, K. R. Use of a (quantitative) structure-activity relationship [(Q)SAR] model to predict the toxicity of naphthenic acids. J. Toxicol. Environ. Health, A 2010, 73, 319–329. (13) Smith, B. E.; Lewis, C. A.; Belt, S. T.; Whitby, C.; Rowland, S. J. Eﬀects of alkyl chain branching on the biotransformation of naphthenic acids. Environ. Sci. Technol. 2008, 42, 9323–9328. (14) Martin, J. M.; Han, X.; Peru, K. M.; Headley, J. V. Comparison of high- and low-resolution electrospray ionization mass spectrometry for the analysis of naphthenic acid mixtures in oil sands process water. Rapid Commun. Mass Spectrom. 2008, 22, 1919–1924. (15) Barrow, M. P.; Witt, M.; Headley, J. V.; Peru, K. M. Athabasca oil sands process water: Characterization by atmospheric pressure photoionization and electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry. Anal. Chem. 2010, 82, 3727– 3735. (16) Geisy, J. P.; Anderson, J. C.; Wiseman, S. B. Alberta oil sands development. Proc. Natl. Acad. Sci. U.S.A. 2010, 107, 951. (17) Hargesheimer, E. E.; Coutts, R. T.; MacKinnon, M. D. Characterization of simple phenols in oil sands extraction-process water. Environ. Technol. Lett . 1984, 5, 433–440. (18) Frank, R. A.; Kavanagh, R.; Burnison, B. K.; Arsenault, G.; Headley, J. V.; Peru, K. M.; Van Der Kraak, G.; Solomon, K. R. Toxicity assessment of collected fractions from an extracted naphthenic acid mixture. Chemosphere 2008, 72, 1309–1314. (19) Choi, K.; Sweet, Li.; Meier, P. G.; Kim, P. G. Aquatic toxicity of four alkylphenols (3-tert-butylphenol, 2-isopropylphenol, 3-isopropylphenol, and 4-isopropylphenol) and their binary mixtures to microbes, invertebrates, and ﬁsh. Environ. Toxicol. 2004, 19, 45–50.

9782

dx.doi.org/10.1021/es201948j |Environ. Sci. Technol. 2011, 45, 9776–9782

ARTICLE pubs.acs.org/est

Environmental Risk Assessment of Fluctuating Diazinon Concentrations in an Urban and Agricultural Catchment Using ToxicokineticToxicodynamic Modeling Roman Ashauer,*,† Irene Wittmer,† Christian Stamm,† and Beate I. Escher†,‡ † ‡

Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 D€ubendorf, Switzerland The University of Queensland, National Research Centre for Environmental Toxicology (Entox), 39 Kessels Rd, Brisbane, Qld 4108, Australia

bS Supporting Information ABSTRACT: Temporally resolved environmental risk assessment of ﬂuctuating concentrations of micropollutants is presented. We separated the prediction of toxicity over time from the extrapolation from one to many species and from acute to sublethal eﬀects. A toxicokinetic toxicodynamic (TKTD) model predicted toxicity caused by ﬂuctuating concentrations of diazinon, measured by time-resolved sampling over 108 days from three locations in a stream network, representing urban, agricultural and mixed land use. We calculated extrapolation factors to quantify variation in toxicity among species and eﬀect types based on available toxicity data, while correcting for diﬀerent test durations with the TKTD model. Sampling from the distribution of extrapolation factors and prediction of time-resolved toxicity with the TKTD model facilitated subsequent calculation of the risk of undesired toxic events. Approximately one-ﬁfth of aquatic organisms were at risk and ﬂuctuating concentrations were more toxic than their averages. Contribution of urban and agricultural sources of diazinon to the overall risk varied. Thus using ﬁxed concentrations as water quality criteria appears overly simplistic because it ignores the temporal dimension of toxicity. However, the improved prediction of toxicity for ﬂuctuating concentrations may be small compared to uncertainty due to limited diversity of toxicity data to base the extrapolation factors on.

’ INTRODUCTION Background. Concentrations of micropollutants1 such as

pesticides and biocides in freshwater streams may strongly fluctuate over time, as demonstrated by an increasing number of chemical monitoring studies with high temporal resolution28 or predicted by pollutant fate models.9 Assessing the risk of adverse effects of such fluctuating concentrations to aquatic organisms is challenging,2,7,1013 because traditional risk assessment methods are based on ecotoxicological tests, employing constant exposure concentrations and fixed durations, that do not explicitly consider the temporal aspect of toxicity. As a result, the current risk assessment procedures lack methods to assess the potential toxicity of fluctuating and repeated pulsed exposures. Traditionally derived water quality criteria for short-term or long-term exposure do not suﬃce because it is unclear how to evaluate very short peaks above the short-term quality criterion or concentrations between the short-term and the long-term criterion.2,7,14 In addition, it is unclear how to deal with subsequent pollutant peaks, which raises the question whether exposed organisms have had enough time to recover in between exposures,7,14,15 and whether low concentrations between peaks contribute to the risk of adverse eﬀects.7,14 r 2011 American Chemical Society

Risk assessment of ﬂuctuating concentrations requires assessment of toxicity over time (prediction of toxic eﬀects beyond test conditions), and such a prediction requires a mechanistic model of toxicity.10,16 Toxicokinetic-toxicodynamic (TKTD) models 14,17 are suitable for this purpose as they can simulate time-dependent phenomena related to prediction of toxic eﬀects such as carry-over toxicity and cumulative eﬀects,15,16,1823 delayed and post exposure eﬀects 19,2427 and organism recovery.18,2831 Objectives. The objective of this study was to overcome the limitations of time-invariable toxicity data by implementing simulated time-variable toxicity into risk assessment of micropollutants. We demonstrate that prediction of toxic effects over time is feasible with a mechanistic model being used for two purposes: (i) the simulation of effects over time and (ii) the calculation of extrapolation factors for a set of ecotoxicological data. We used the insecticide diazinon as an example and assessed the contribution of urban and agricultural sources of diazinon toward the Received: July 13, 2011 Accepted: September 29, 2011 Revised: September 16, 2011 Published: September 29, 2011 9783

dx.doi.org/10.1021/es202413a | Environ. Sci. Technol. 2011, 45, 9783–9792

Environmental Science & Technology

ARTICLE

Figure 1. Calculation of the eﬀect and risk curves (middle and upper panels in Figure 2). Toxic eﬀects from time variable exposure patterns are predicted with a mechanistic eﬀect model (TKTD model). Extrapolation between species and from lethal to sublethal eﬀects is achieved by multiplying the exposure concentration (C(t)) with extrapolation factors. The extrapolation factors account for the variation in sensitivity of diﬀerent species and eﬀect types, based on the data in Table 1 (see Scheme 1 for explanation of the calculation steps).

ecotoxicological risk based on time-resolved exposure data, measured at different locations within a catchment. Outline of the Time-Resolved Risk Assessment Approach. The goal of the time-resolved risk assessment is to predict the risk posed to aquatic organisms as a function of cumulative exposure over time. We illustrate the approach using measured time series of diazinon exposure concentrations from three locations in a small stream network, which represent urban, agricultural and mixed land use, as inputs for the environmental risk assessment. Risk assessment of micropollutants encompasses three dimensions: (i) prediction of toxicity under long-term and ﬂuctuating exposure, (ii) extrapolation from lethal to sublethal end points and (iii) extrapolation from one (or a few) tested species to many (all) species in the water body of concern. Here, we separated the three dimensions and addressed the prediction of toxicity over time separately from the extrapolation to many species and sublethal eﬀects. Prediction of toxic eﬀects over time was achieved with simulations by a mechanistic TKTD model. Fixed assessment factors are traditionally used to address the above-mentioned extrapolation steps in risk assessment of chemicals.3236 Here, we replace these assessment factors with extrapolation factors that quantify the variation in the extrapolation of toxicity from one to many species and from lethal to

sublethal eﬀects based on available, relevant toxicity data. In the calculation of these extrapolation factors, we corrected eﬀect concentrations for diﬀerences in test durations by using the mechanistic TKTD model. Finally, we ﬁtted a distribution to the extrapolation factors, sampled a large number of extrapolation factors from that distribution and predicted the time course of toxic eﬀects with the TKTD model for each of these extrapolation factors. We then calculated the risk of undesired toxic events from the multitude of these simulations (Figure 1).

’ MATERIALS AND METHODS Diazinon Monitoring Data. Diazinon is a hydrophobic insecticide (log Kow 3.8137), which frequently occurs in pulses or fluctuating concentrations in streams3,5,6,8 and its metabolite diazoxon inhibits acetylcholinesterase. Next to its use in agriculture, diazinon is also used as insecticide in urban areas (e.g., against lice on roses, fish moths in wet rooms etc.). The diazinon data originate from the study described in Wittmer et al. (2010).8 The study area (Figure S1 in Supporting Information (SI)) is a subcatchment embedded in the catchment of Lake Greifensee located on the Swiss Plateau. The studied catchment covers 25 km2 and has a mixed urban and agricultural land use. We compared the 9784

dx.doi.org/10.1021/es202413a |Environ. Sci. Technol. 2011, 45, 9783–9792

Environmental Science & Technology temporal dynamics of diazinon concentrations in a subcatchment with only agricultural land use (AGR in Figure S1 of the SI) to one with predominant urban land use (URB in Figure S1 of the SI, including effluent from the wastewater treatment plant (WWTP)). Settlements located in the subcatchment AGR are connected to the sewer system in URB. There is no urban stormwater discharged into the river in AGR. We compared the dynamics of the two subcatchments to the dynamics at the outlet of the entire catchment (TOT in Figure S1 of the SI), which comprises 470 ha of arable land and 12 000 inhabitants (10 000 in the subcatchment URB). The three stations were monitored from 10th of March to 26th of June 2007. During several rain events throughout the monitoring period, samples were taken at a high resolution (15 min to hourly composite samples). Weekly grab samples were taken during dry weather conditions. Samples were stored at minus 20 °C in the dark to prevent degradation. Prior to analysis an isotope-labeled internal standard was added to 50 mL of ﬁltered sample. Analytes were enriched with online solid phase extraction, separated by liquid chromatography, and detected by tandem mass spectrometry (SPE-LC-MS/MS). For details on the analytical method, see Singer et al. (2010)38 and for details on sampling procedure and sites see Wittmer et al. (2010).8 Ecotoxicological Data. We demonstrate the method on the example of a catchment in Switzerland and Swiss protection goals. Swiss protection goals for surface water aim to protect fisheries, the health of animals and biological processes needed by plants and animals to fulfill their physiological needs, specifically metabolism, reproduction, and olfactory orientation.39,40 Thus we considered sublethal toxic effects as relevant and corresponding to the protection goals and collected such data for diazinon and aquatic organisms as input for this case study. Lethal data were not used in calculations, because assessment based on sublethal data is also protective for effects on survival. Risk assessments for different protection goals may use different criteria to select input data. Ecotoxicological data on sublethal eﬀects of diazinon in aquatic organisms were collected from two sources. The RIVM database contains data from the U.S. EPA and The Netherlands National Institute of Public Health and the Environment (RIVM)41 until the year 2000 (D. de Zwart Pers. comm.) and is quality controlled (see de Zwart41 for procedure). From this database we selected all entries for sublethal eﬀects on freshwater organisms (i.e., excluding mortality, immobility, abundance, and undeﬁned eﬀect types), which yielded twelve records. From these records, we deleted data on algae (1 record) because they do not carry the target enzyme, the acetylcholinesterase, and are not as sensitive as ﬁsh and invertebrates toward diazinon. Eﬀect sizes for NOEC values were taken from Crane and Newman42 and Suter et al.43 Additional data were collected from peerreviewed literature by searching the SCOPUS database. The query for “diazinon” “toxicity” “*water” AND (year after 2000) in title, abstract, or keywords yielded 102 abstracts (15 February 2011). From these, we selected studies with information relevant to the Swiss water protection goals yielding six additional records. Here, eﬀect sizes were taken from the original references. All studies are listed in Table 1. Toxicokinetic-Toxicodynamic Model. The TKTD model was previously parametrized to simulate survival of the stream dwelling arthropod Gammarus pulex.19 The freshwater arthropod G. pulex is frequently used in ecotoxicological studies,44 is very sensitive to diazinon (96 h-LC50 is 4.15 μg/L45) and exhibits similar sensitivity to environmental pollutants as

ARTICLE

Daphnia magna.46 G. pulex is abundant in headwater streams of our case study catchment. Furthermore, its longevity makes it a suitable test organism to parametrize a TKTD model, which captures the relationship between time and toxicity for diazinon, including carry-over toxicity.19 G. pulex requires approximately 28 days to recover from toxic stress caused by diazinon.19 If G. pulex is pre-exposed to diazinon, then a subsequent exposure can cause stronger toxic effects than what would be expected from a onepulse exposure alone.19 Such carryover toxicity was caused by slow organism recovery as shown by the TKTD model.19 The TKTD model for diazinon and G. pulex simulates the time-course of toxicant uptake, biotransformation and elimination as a first step (toxicokinetics, TK) and the development of damage within the organisms and subsequent increased mortality as a second step (toxicodynamics, TD).19 This TKTD model corresponds to the special case of stochastic death in the General Unified Threshold Model of Survival.17 The model was parametrized using measured internal concentrations of diazinon and its biotransformation products, in particular diazoxon, within G. pulex as well as long-term survival experiments with pulsed exposure patterns.19 This mechanistic model describes the processes leading to toxicity on a temporal scale and so captures the time-toxicity relationship for survival of G. pulex and diazinon (step 1 in Scheme 1). Thus, the model can be used to predict survival of G. pulex for other, untested diazinon exposure patterns. Here we also use the TKTD model as a proxy for the time-toxicity relationships of other combinations of species and effect type, because the TKTD model for survival of G. pulex is the best approximation for these unknown time-toxicity relationships that we have, even if that requires some bold assumptions. We used the TKTD model for two purposes. First, we calculate extrapolation factors for extrapolation from G. pulex to other species and eﬀect types. Second, we simulate the timecourse of risk of adverse toxic eﬀects for aquatic organisms. Details of both calculations follow below. Assumptions about the Relationship between Exposure Time and Toxicity. Assumptions about the relationship between exposure time and toxicity are rarely made explicit in current risk assessment schemes. The default method for dealing with timevariable exposure concentrations in risk assessment is to use time-weighted average concentrations.10,47,48 This method is based on Haber’s law and assumes that the product of exposure time and concentration determines toxicity, i.e., the same time integral of concentration yields the same toxicity,10,49 and that carry-over toxicity does not occur. Consequently, the average concentration of a fluctuating exposure pattern is assumed to result in the same toxicity as the fluctuating exposure. Deviations from Haber’s law can be caused by toxicokinetics (carry-over bioaccumulation), toxicodynamics (carry-over toxicity) or both. Relationships between time and toxicity can be modeled with TKTD models,17 such as the one we base the proposed risk assessment on.19 We assume that the relationship between time and toxicity in all the species that we address with our risk assessment is the same as that in G. pulex. This does not mean that we assume the same sensitivity, but we assume the same organism recovery time.18 Assuming that all assessed species resemble G. pulex in their time-toxicity relationship is the best currently possible approximation, because to date G. pulex is the only species for which a time toxicity relationship of diazinon was quantified with a TKTD model. We further assume that an extrapolation factor that we multiply exposure concentrations with, can account for sensitivity variation of different species 9785

dx.doi.org/10.1021/es202413a |Environ. Sci. Technol. 2011, 45, 9783–9792

9786

reproduction

reproduction

ﬁltration rate

reproduction reproductive rate,

Ceriodaphnia dubia

Ceriodaphnia dubia

Daphnia magna

Daphnia magna Daphnia pulex

population size

growth

growth

growth

reproduction

spinal deformities

swimming

reproduction

Daphnia pulex

Pimephales promelas

Pimephales promelas

Pimephales promelas

Pimephales promelas

Pogonichthys macrolepidotus

Oncorynchus mykiss

Salvelinus fontinalis NOEC

EC40

EC35

NOEC

NOEC

NOEC

NOEC

EC09

NOEC EC99

EC43

NOEC

MATC

14

40

35

14

10.5

10.5

10.5

9

10 99

50

28

43

25

25

50

10.5

(% aﬀected)

eﬀect size

43

65

64

43

42

42

42

63

42 62

EC50

61

61

11

1

4

168

32

32

7

10

21 2

1

7

7

7

2

estimatedb 42

2

2

(days)

test duration

EC50

42

eﬀect size

reference for

1.81

1643

4107

10.51

131.4

164.3

282.9

2.04

0.66 6.57

1.54

0.95

1.31

0.72

32 858

36 144

26 286

(nmol/L)

eﬀect concentration

7.8 7.8 7.0

RIVM database 41 RIVM database 41 RIVM database 41

999.2 13.6

65 RIVM database 41

58.2

19.3

RIVM database 41

64

13.7

63

8.9 1289.4

RIVM database 41 62

25.1 1279.4

RIVM database 41

61

26.8

24.2

61

150.1

RIVM database 41 RIVM database 41

96.5 262.3

RIVM database 41 RIVM database 41

factor (EF)

7.548

0.6082

0.0142

0.6657

0.0592

0.0474

0.0681

6.725

13.48 196.2

828.5

26.34

20.39

33.48

0.0046

0.0073

0.0037

extrapolation

same eﬀect size and duration in G. pulexnmol/L

ref

corresponding LCx for

0.878

0.216

1.849

0.177

1.228

0.828

1.130 2.293

2.918

1.421

2.340

log EFc

We calculate the concentration that causes the same percentage mortality in G. pulex (LCx) for the same duration and percentage eﬀect as in each of the ecotoxicological studies using the TKTD model. Extrapolation factors are the ratio of the LCx and the eﬀect concentration in the ecotoxicological study. Multiple values of the same combination of species and eﬀect type are replaced with their median. b As the MATC is the geometric mean between NOEC and LOEC a larger eﬀect size than the NOEC can be expected, thus we use the largest eﬀect size from all NOEC end points here, i.e. 25%. c A normal distribution with mean = 0.3325 and SD = 1.667 was ﬁtted to the log EFs in the last column.

a

EC50

reproduction

Ceriodaphnia dubia

population extinction

EC28

reproduction

Brachionus calyciﬂorus

EC50

reproduction

Brachionus calyciﬂorus

NOEC

measure

reproduction

eﬀect type

Brachionus calyciﬂorus

species

eﬀect

Table 1. Ecotoxicological Studies Relevant in the Context of the Protection Goals of the Swiss Water Protection Law39a

Environmental Science & Technology ARTICLE

dx.doi.org/10.1021/es202413a |Environ. Sci. Technol. 2011, 45, 9783–9792

Environmental Science & Technology Scheme 1. Step by Step Explanation of the Risk Assessment Calculationsa

a

Step (1) is the calibration of the toxicokinetic-toxicodynamic (TKTD) model.19 Step (2) generates the data for the probabilistic risk assessment, which is carried out in steps (3) to (6). The TKTD model is used in steps (3) and (5).

and effect types. This is necessary because we lack the knowledge to incorporate possible nonlinearity in that aspect of sensitivity variation. Note that applying species sensitivity distributions or assessment factors to averaged exposure concentrations, as in standard procedures,35,47,48 also assumes the same time-toxicity relationship for all organisms and it ignores the possibility of carry-over toxicity. Extrapolation Factors. We define extrapolation factors (EFs) as ratio between the lethal concentration for x% of G. pulex at duration t (LCx(t)) and the effect concentration for x% effect in the ecotoxicological studies with different species and effect types at duration t (ECx(t)) (Table 1, step 3 in Scheme 1). Multiplying the effect concentration ECx(t)i from a study in Table 1 with the extrapolation factor EFi, results in the corresponding lethal concentration LCx(t)i for G. pulex for the same duration t and same effect size (x%). Derivation of EFs for the ecotoxicological studies in Table 1 (step 2 in Scheme 1) requires ﬁrst calculating the LCx(t)i of G. pulex that corresponds to the same duration t and x percentage eﬀect of the sublethal eﬀects ECx(t)i on other species than G. pulex. This calculation was carried out using the TKTD model: survival of G. pulex under a constant exposure concentration LCx(t)i was simulated and the concentration adjusted such that the eﬀect level on mortality of G. pulex corresponded to the eﬀect

ARTICLE

level (x %) of the sublethal eﬀects ECx(t)i on the other species after the duration t. EFi was calculated as ratio LCx(t)i/ECx(t)i. All EFs were log transformed to achieve normally distributed data (passed KolmogorovSmirnov test, D’Agostino and Pearson omnibus normality test and Shapiro-Wilk normality test). Multiple values of EFs for the same combination of species and eﬀect type were replaced with their median (last column in Table 1). The log-transformed EFs in our case study are characterized by a normal distribution with mean 0.3325 and standard deviation 1.667 (n = 11). This is step 4 in Scheme 1. We also calculated EFs for mortality caused by diazinon in diﬀerent aquatic species in the RIVM database and compared their distribution to sublethal EFs (see SI). Here, only the distribution of sublethal EFs was used in further analyses (see rationale in section Ecotoxicological data). With the assumptions that the time-eﬀect relationship for G. pulex and diazinon resembles that of the other combinations of species and eﬀect type and that the EF captures sensitivity variation, one can do the following calculations: First multiply the concentration time series of diazinon in any exposure pattern with the EF for a given combination of species and eﬀect type. Then use that concentration time series as input in the TKTD model calibrated for G. pulex survival and simulate the time-course of eﬀects for the combination of species and eﬀect type of that EF. Effect Simulations. We repeatedly simulated survival of G. pulex in response to measured concentration time series in a small stream multiplied with the EF. Without multiplying exposure with the EF the simulation would predict survival of G. pulex. With the EF the simulated effect curve is the prediction for a sublethal effect on another species. If the simulation were done with one of the EFs from Table 1, then the simulated effect curve would be the time course of effect for that specific combination of biological species and sublethal effect type. The application of the extrapolation factor changed the interpretation of the model output from simulated time-course of survival of G. pulex into simulated toxic effects for the respective combination of aquatic organisms and effect type. We randomly sampled 1000 log EFs from the normal distribution of log-transformed sublethal EFs, back-transformed them to EFs (denoted EFn), multiplied the concentrations in the exposure time series with these EFs and simulated the time-course of eﬀect with the TKTD model for each of these instances (see Scheme 1, step 5). The same 1000 EFs were used for each of the three concentration time-series corresponding to the total catchment and its urban and agricultural subcatchments, respectively. The survival probability S(t), which is the output of the original TKTD model,19 was then converted to eﬀect as follows: Ef f ectn ðtÞ ¼ ð1 Sn ðt, CðtÞ EFn ÞÞ 100%

ð1Þ

where EFn [] is the extrapolation factor in simulation run n (out of 1000 runs), C(t) [nmol/L] is the measured concentration time series of diazinon8 and Eﬀectn [] is the predicted eﬀect size. Eﬀect size has the same interpretation as in the ecotoxicological study that corresponds to the EF, e.g., at the EF of 0.6082 a simulated eﬀect of 0.1 could be interpreted as swimming being aﬀected in 10% of Oncorhynchus mykiss (see Table 1). Calculation of Risk. Risk can be defined as the probability of an undesired event.50 Here we assume that the ecotoxicological data in Table 1 represents variability in sublethal toxicity of diazinon to fish and aquatic invertebrate species that the Swiss protection goal aims to protect. Thus we define an undesired event as one where Effect(t)n exceeds 50%, i.e., where 50% of 9787

dx.doi.org/10.1021/es202413a |Environ. Sci. Technol. 2011, 45, 9783–9792

Environmental Science & Technology

ARTICLE

Figure 2. Measured concentrations of diazinon (lower panels) in diﬀerent parts of the catchment, eﬀect curves simulated for diﬀerent percentiles (solid lines from left to right: 95th, 90th, 87.5th, 85th, and 80th percentile) of the extrapolation factor distribution (middle panels) and the risk (top panels). Risk is the fraction of probabilistic simulations that show toxic eﬀects, i.e. the fraction of aﬀected combinations of species and eﬀect type (see Figure 1 and Scheme 1 for calculation steps). In the lower panels, a peak originating from the agricultural part of the catchment is indicated by (a) and elevated concentrations originating from the urban part by (b). Note the diﬀerent scale of the y-axis in the lower panel (station AGR). Day 0 corresponds to the March 10, 2007. The eﬀect curve of the example AA-EQS value is plotted in the middle panels (dashed line).

the organisms of a species suffer from an undesired toxic effect (one could also chose a different effect size). As we simulated a sample of 1000 effect curves based on the distribution of EFs, we can calculate risk as the fraction of those simulations where Effect(t)n exceeds 50% (Scheme 1, step 6). As each of the 1000 EFs drawn from the distribution represents a single, possible combination of species and type of sublethal effect, this is the fraction of affected combinations of species and effect type, i.e., risk (top panels in Figure 2).

’ RESULTS AND DISCUSSION Extrapolation Factors. The toxicity data from literature and thereof derived EFs are listed in Table 1. The distributions of EFs for lethal and sublethal effects are plotted in Figure S1 of the SI. Both distributions span about 5 orders of magnitude. Even though EFs for sublethal effects are larger by approximately 1 order of magnitude, a considerable overlap in the two distributions is obvious. This comparison indicates that differences among species contribute more to variation than differences between lethal and sublethal effects. Time-Course of Effect and Risk to Aquatic Organisms. The measured diazinon concentrations fluctuated strongly (Figure 2, lower panels). At the outlet of the catchment (station TOT, Figure S2 of the SI), elevated background concentrations between 0.5 and 1 nmol/L were observed throughout most of the study with pronounced peaks occurring at irregular intervals.

Discharge of station URB (Figure S2 of the SI) consists mostly of urban stormwater runoff and effluent of the WWTP. There was no high peak at the beginning of the time series, as in the agricultural part of the catchment, but background concentrations were higher than at the outlet of the entire catchment (TOT) because there was less dilution of the WWTP discharge by discharge from agricultural areas. Using measured time-courses of diazinon concentrations as input for modeling resulted in a risk of adverse eﬀects to 17% (TOT), 20% (URB) and 24% (AGR) of aquatic organisms at the three monitoring stations (Figure 2, top panels). Interestingly, the temporal onset and contribution of urban and agricultural sources to the risk varied between sites. Selected percentiles of the eﬀect curves predicted using the TKTD model are plotted in Figure 2, middle panels. Eﬀect curves show the increasing eﬀects over time for the chosen percentiles of combinations of species and eﬀect type in relation to the concentration curve plotted below. From left to right (total, urban, agricultural part of the catchment) the exposure proﬁles aﬀect a larger fraction of combinations of species and eﬀect type and more severe eﬀects are seen earlier. In the agricultural exposure proﬁle the ﬁrst high peak is the dominant toxic event, whereas in the other two proﬁles toxicity results from combination of lower long-term exposure and short peaks. Low long-term and low-level pollution from urban sources of diazinon, which was observed at stations TOT and URB, caused risk for the 5% most sensitive combinations of species and 9788

dx.doi.org/10.1021/es202413a |Environ. Sci. Technol. 2011, 45, 9783–9792

Environmental Science & Technology sublethal eﬀect type (Figure 2, middle panels, 95th percentile curve). The high peak around day 8, coming from the agricultural part of the catchment (labeled (a) in Figure 2, lower panels), raised the risk in station TOT to 11%. The lower exposure levels between days 25 and 45 increased the risk by another 2% and then peaks around day 60 raised the risk to the ﬁnal total of 17% in station TOT. After day 50 risk at station URB increased to higher percentiles than at station TOT, i.e., toward the end of the monitoring period, we predict that more combinations of species and eﬀect type are aﬀected at station URB (20%) than station TOT (17%). A diﬀerent situation was observed at the monitoring station AGR, where discharge consists only of agricultural and some road runoﬀ. Hardly any constant background concentrations were observed. Instead short concentration peaks occurred (Figure 2, lower panels). Being about 1 order of magnitude higher than at the other two stations, the short pulse (1 to 2d) around day 8, caused most likely by agricultural activities alone, led to 24% of risk. Risk Assessment of Time-Varying Exposure. The concept of using fixed concentrations as water quality criteria is overly simplistic because it ignores the temporal dimension of toxicity that is evident from the examples above. Risk assessment based on TKTD models demonstrates that the risk of adverse effects can be caused by short, high peak exposures or by lower, long-term exposures or by a combination of both. Due to the mechanistic nature of the underlying TKTD model, which has been shown to properly predict carry-over toxicity,19 we are confident that the risk assessment procedure presented here captures the temporal aspects of toxicity. Alternative methods (reviewed elsewhere2,7,10), such as time-weighted averages, are less well qualified for risk assessment of fluctuating concentrations, because they are not able to predict carry-over toxicity.16 In this case study, the average concentration underestimated toxicity by up to a factor of 4 (see below). Thus a dynamic, mechanistic effect model complements classical risk assessment based on short- and longterm quality criteria. Quantiﬁcation of spatial and temporal exposure can be carried out together with toxicokinetictoxicodynamic modeling of eﬀects, which places the temporal extrapolation of toxicity on a mechanistic basis. The approach presented here separates the temporal extrapolation step from the extrapolation to other biological species and other end points (e.g., sublethal eﬀects). Consequently, the separated extrapolation steps can be tested, re-evaluated, improved, and validated independently of each other. The use of the EF for probabilistic risk assessment is in analogy to the species sensitivity distribution approach.51 As with species sensitivity distributions, the available relevant toxicity data determines the outcome of the risk assessment. The toxicity data (Table 1) as input for the risk assessment model, can be easily exchanged by a diﬀerent set of toxicity data, perhaps to reﬂect a diﬀerent protection goal and risk assessment context. The combination of TKTD models and probabilistic calculations allows quantiﬁcation of the time course of eﬀect curves for diﬀerent combinations of species and eﬀect type. Further, it allows calculation of the time course of risk in response to ﬂuctuating exposure concentrations. The proposed time-resolved risk assessment method is generic and can be adopted for risk assessment of other pesticides2 and biocides,52 or the assessment of pollution in the context of the Water Framework Directive.53 Comparison with Effects Based on Average Concentrations. We quantified the difference in toxicity of the fluctuating

ARTICLE

exposure patterns compared with the toxicity of their average concentrations. For each of the three different exposure patterns and for each of the three corresponding average exposure concentrations, we calculated the factor that the exposure time series must be multiplied with to result in 50% mortality of G. pulex at the end. The concentration time series from the total catchment had to be multiplied with 75 to result in 50% mortality at the end (107.4 d), whereas the corresponding average concentration of 0.0760 nmol/L required multiplication by a factor of 96 for the same toxic eﬀect. For the ﬂuctuating exposure time series from the urban part of the catchment the factor was 46 and for the corresponding average concentration of 0.1187 nmol/L the factor was 62. For the ﬂuctuating exposure time series from the agricultural part of the catchment the factor was 26 and for the corresponding average concentration of 0.0726 nmol/L the factor was 101. Thus, in all three exposure patterns, the ﬂuctuating concentrations were more toxic than their averages (1.28-, 1.35-, and 3.88-fold, respectively). Here, the TKTD model was used to simulate toxicity for the full length of the exposure proﬁle. In practice, the time-weighted average concentrations are calculated for a shorter period of time2 and how large the resulting error of these methods is needs to be systematically investigated. Comparison with Environmental Quality Standards. For comparison with environmental quality standards derived with the commonly applied method in the European Union,54 we took the Annual Average Environmental Quality Standard (AA-EQS) of 0.015 μg/L (0.049 nmol/L) from a report55 that compared different methods to derive environmental quality standards for Swiss surface waters. Assuming 365 days as a test duration and an effect size of 10% for the AA-EQS, which is generally derived from NOEC data, we calculated the corresponding EF as 141. This EF corresponds to the 88th percentile of the EF distribution, thus it protects 88% of combinations of species and effect type based on the data in Table 1. The corresponding effect curves are plotted in the middle panels in Figure 2 (AA-EQS, dotted line). The average concentrations in stations TOT, URB and AGR were 0.076 nmol/L, 0.119 nmol/L and 0.073 nmol/L, respectively. Thus comparison with the AA-EQS of 0.049 nmol/L already indicates a risk. The added value of the probabilistic assessment is the quantification of total risk from any exposure pattern and the possibility to investigate effects on specified percentiles of species effect combinations. Uncertainty and Limitations. Ecotoxicological data for this case study comprises only three taxonomic groups: rotifers (3 studies), crustaceans (7 studies), and fish (7 studies). We do not know how well this limited diversity of taxa in our data captures variation in sensitivity between different species. Similarly diversity of effect types in our data set is limited and biased toward reproduction in rotifers and crustaceans and growth in fish. As with any modeling study, quality of the results depends on the input data. In this case, the limited diversity of species and effect types in the ecotoxicity database poses the greatest source of uncertainty. The desired diversity in ecotoxicity data was also discussed in the context of species sensitivity distributions51 and these discussions suggest, that also for studies like the one presented here, a higher diversity of ecotoxicological data is needed. Variation in sensitivity of diﬀerent species and eﬀect types spans several orders of magnitude, whereas the diﬀerence between the toxicity of the ﬂuctuating exposure proﬁles and that of their corresponding average concentrations was less than 4-fold. Thus we improved prediction of toxicity for ﬂuctuating 9789

dx.doi.org/10.1021/es202413a |Environ. Sci. Technol. 2011, 45, 9783–9792

Environmental Science & Technology concentrations with our method, but these improvements may be small compared to the uncertainty due to limited diversity of toxicity data. Consequently, the potential of this new method can best be realized in combination with larger sets of ecotoxicity data. For small sets of ecotoxicity data large uncertainties remain. Relevant Sources and Priorities for Mitigation Measures. Urban and agricultural sources of diazinon contributed to the risk of adverse effects on aquatic organisms in the studied catchment. Their respective contribution varied and cannot be quantified a priori or as any fixed ratio. Rather, both sources contributed with variable proportions and the relationship by which long-term pollution and short concentration peaks induce adverse effects is nonlinear. From a practical point of view, the analysis demonstrated that, depending on the location in the river network, either urban or agricultural sources of diazinon or their interplay caused a risk of adverse effects. Without the TKTD model, just relying on the observed concentration data, it was difficult to judge the ecotoxicological relevance of the different exposure patterns and sources. Through the TKTD modeling, we learn that an integrated approach to risk mitigation that targets both urban and agricultural sources is most likely to efficiently reduce risk of adverse effects to aquatic organisms. Hence it is important to understand the dynamics of discharge from different sources.8 For diazinon, restrictions or mitigation measures on both urban and agricultural uses, are required. Spatial Aspects of Risk Assessment in Aquatic Systems. Availability of measured concentrations time-series at different locations in the catchment raises the question how to aggregate risks calculated at various locations in any given river network. Our set of three locations illustrates that the same event, for example the high peak originating from the agricultural area around day 8, may cause strong risk of adverse effects in one location (24% in AGR) but much less in another (11% in TOT). Integration of TKTD models with spatially explicit models for populations, such as individual or agent based models,56,57 may be one possible solution if the assessment is carried out at the population level. If however, the protection goal is aimed at the organism level, then the question of how to aggregate risk of adverse effects from different locations remains. Research Needs. TKTD models can handle fluctuating concentrations of micropollutants. However, the approach is currently limited by the availability of model parameters. These parameters are unique for each combination of chemical and biological species and more of these relationships of time and toxicity need to be quantified. The current reliance on one such relationship for all other assessed species carries uncertainty, which we need to quantify and reduce. If TKTD models for more species, compounds, and effect types were available, then these could ultimately replace the extrapolation factors used here. Raw data from standard acute toxicity tests can already be used for parametrization of TKTD models17 and selection of species and sublethal end points for TKTD model development58 could be prioritized based on ecological considerations (e.g., trophic position, ecosystem services) to maximize relevance. We assessed the risk from one compound alone, although it is known that multiple chemical stressors occur simultaneously or after each other, also in the catchment studied here.8 Risk assessment of mixtures in time is important as diﬀerent compounds can interact and contribute to mixture toxicity even if they occur days22 or weeks apart.15 TKTD models for mixtures of chemicals are available,15,59 but their relationship to standard models for toxicity of mixtures with simultaneous, constant exposure is not

ARTICLE

well established.60 TKTD models for temporal mixtures can explain phenomena such as the sequence eﬀect,15 which is caused by carry-over toxicity, however the large number of model parameters needed to assess the temporal mixtures of chemicals on a catchment scale are not available yet.

’ ASSOCIATED CONTENT

bS

Supporting Information. Details on the study catchment and extrapolation factors. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*Phone: +41-448235233; fax: +41-448235311; e-mail: roman. [email protected].

’ ACKNOWLEDGMENT This study was supported by the Swiss National Science Foundation (Grant No. 200021-119795), Swiss Federal Oﬃce for the Environment (Grant Nos. 09.033.PJ/I362-1602 and 09.0012.PJ) and the SETAC-CEFIC-LRI Innovative Science Award. Funding of the ﬁeld study by Eawag and AWEL is gratefully acknowledged. ’ REFERENCES (1) Schwarzenbach, R. P.; Escher, B. I.; Fenner, K.; Hofstetter, T. B.; Johnson, C. A.; von Gunten, U.; Wehrli, B. The challenge of micropollutants in aquatic systems. Science 2006, 313 (5790), 1072–1077. (2) Brock, T. C. M.; Alix, A.; Brown, C. D.; Capri, E.; Gottesb€uren, B.; Heimbach, F.; Lythgo, C. M.; Schulz, R.; Streloke, M. Linking Aquatic Exposure and Eﬀects; SETAC: Pensacola, FL, 2010; p 410. (3) Holmes, R. W.; De Vlaming, V. Monitoring of diazinon concentrations and loadings, and identiﬁcation of geographic origins consequent to stormwater runoﬀ from orchards in the Sacramento River watershed, U.S.A. Environ. Monit. Assess. 2003, 87 (1), 57–79. (4) Kreuger, J. Pesticides in stream water within an agricultural catchment in southern Sweden, 19901996. Sci. Total Environ. 1998, 216 (3), 227–251. (5) Pedersen, J. A.; Yeager, M. A.; Suﬀet, I. H. Organophosphorus insecticides in agricultural and residential runoﬀ: Field observations and implications for total maximum daily load development. Environ. Sci. Technol. 2006, 40 (7), 2120–2127. (6) Phillips, P. J.; Bode, R. W. In Pesticides in Surface Water Runoﬀ in South-Eastern New York State, USA: Seasonal and Stormﬂow Eﬀects on Concentrations; International Symposium on Non-agricultural Use of Pesticides, Copenhagen, DENMARK, May 0709, 2003; John Wiley & Sons Ltd: Copenhagen, DENMARK, 2003; pp 531543. (7) Reinert, K. H.; Giddings, J. A.; Judd, L. Eﬀects analysis of timevarying or repeated exposures in aquatic ecological risk assessment of agrochemicals. Environ. Toxicol. Chem. 2002, 21 (9), 1977–1992. (8) Wittmer, I. K.; Bader, H. P.; Scheidegger, R.; Singer, H.; L€uck, A.; Hanke, I.; Carlsson, C.; Stamm, C. Signiﬁcance of urban and agricultural land use for biocide and pesticide dynamics in surface waters. Water Res. 2010, 44 (9), 2850–2862. (9) FOCUS FOCUS Surface Water Scenarios in the EU Evaluation Process under 91/414/EEC; European Commission Health & Consumer Protection Diretorated-General Brussel: Belgium, 2003/05, 2001; pp 1238. (10) Ashauer, R.; Boxall, A. B. A.; Brown, C. D. Predicting eﬀects on aquatic organisms from ﬂuctuating or pulsed exposure to pesticides. Environ. Toxicol. Chem. 2006, 25 (7), 1899–1912. 9790

dx.doi.org/10.1021/es202413a |Environ. Sci. Technol. 2011, 45, 9783–9792

Environmental Science & Technology (11) Boxall, A. B. A.; Brown, C. D.; Barrett, K. L. Higher-tier laboratory methods for assessing the aquatic toxicity of pesticides. Pest Manag. Sci. 2002, 58 (7), 637–648. (12) Handy, R. D. Intermittent exposure to aquatic pollutants assessment, toxicity and sublethal responses in ﬁsh and invertebrates. Comp. Biochem. Physiol. CPharmacol. Toxicol. Endocrinol. 1994, 107 (2), 171–184. (13) Zhao, Y.; Newman, M. C. The theory underlying dose-response models inﬂuences predictions for intermittent exposures. Environ. Toxicol. Chem. 2007, 26 (3), 543–547. (14) Ashauer, R.; Escher, B. I. Advantages of toxicokinetic and toxicodynamic modelling in aquatic ecotoxicology and risk assessment. J. Environ. Monit. 2010, 12 (11), 2056–2061. (15) Ashauer, R.; Boxall, A. B. A.; Brown, C. D. Modeling combined eﬀects of pulsed exposure to carbaryl and chlorpyrifos on Gammarus pulex. Environ. Sci. Technol. 2007, 41 (15), 5535–5541. (16) Ashauer, R.; Boxall, A. B. A.; Brown, C. D. New ecotoxicological model to simulate survival of aquatic invertebrates after exposure to ﬂuctuating and sequential pulses of pesticides. Environ. Sci. Technol. 2007, 41 (4), 1480–1486. (17) Jager, T.; Albert, C.; Preuss, T. G.; Ashauer, R. General uniﬁed threshold model of survival—A toxicokinetic-toxicodynamic framework for ecotoxicology. Environ. Sci. Technol. 2011, 45 (7), 2529–2540. (18) Ashauer, R.; Boxall, A. B. A.; Brown, C. D. Simulating toxicity of carbaryl to Gammarus pulex after sequential pulsed exposure. Environ. Sci. Technol. 2007, 41 (15), 5528–5534. (19) Ashauer, R.; Hintermeister, A.; Caravatti, I.; Kretschmann, A.; Escher, B. I. Toxicokinetic-toxicodynamic modeling explains carry-over toxicity from exposure to diazinon by slow organism recovery. Environ. Sci. Technol. 2010, 44 (10), 3963–3971. (20) Boone, M. D.; Bridges, C. M. Eﬀects of carbaryl on green frog (Rana Clamitans) tadpoles: Timing of exposure versus multiple exposures. Environ. Toxicol. Chem. 2003, 22 (11), 2695–2702. (21) Buhl, K. J.; Hamilton, S. J.; Schmulbach, J. C. Chronic toxicity of the bromoxynil formulation BuctrilÒ to Daphnia magna exposed continuously and intermittently. Arch. Environ. Contam. Toxicol. 1993, 25, 152–159. (22) Macinnis-Ng, C. M. O.; Ralph, P. J. In situ impact of multiple pulses of metal and herbicide on the seagrass, Zostera capricorni. Aquat. Toxicol. 2004, 67 (3), 227–237. (23) McHenery, J. G.; Francis, C.; Davies, I. M. Threshold toxicity and repeated exposure studies of dichlorvos to the larvae of the common lobster (Homarus gammarus L). Aquat. Toxicol. 1996, 34 (3), 237–251. (24) Van Der Hoeven, N.; Gerritsen, A. A. M. Eﬀects of chlorpyrifos on individuals and populations of Daphnia pulex in the laboratory and ﬁeld. Environ. Toxicol. Chem. 1997, 16 (12), 2438–2447. (25) McWilliam, R. A.; Baird, D. J. Postexposure feeding depression: A new toxicity endpoint for use in laboratory studies with Daphnia magna. Environ. Toxicol. Chem. 2002, 21 (6), 1198–1205. (26) Cold, A.; Forbes, V. E. Consequences of a short pulse of pesticide exposure for survival and reproduction of Gammarus pulex. Aquat. Toxicol. 2004, 67 (3), 287–299. (27) Schulz, R.; Liess, M. Toxicity of fenvalerate to caddisﬂy larvae: chronic eﬀects of 1-vs 10-h pulse-exposure with constant doses. Chemosphere 2000, 41 (10), 1511–1517. (28) Hosmer, A. J.; Warren, L. W.; Ward, T. J. Chronic toxicity of pulse-dosed fenoxycarb to Daphnia magna exposed to environmentally realistic concentrations. Environ. Toxicol. Chem. 1998, 17 (9), 1860–1866. (29) Stuijfzand, S. C.; Poort, L.; Greve, G. D.; van der Geest, H. G.; Kraak, M. H. S. Variables determining the impact of diazinon on aquatic insects: Taxon, developmental stage, and exposure time. Environ. Toxicol. Chem. 2000, 19 (3), 582–587. (30) Fisher, D. J.; Burton, D. T.; Yonkos, L. T.; Turley, S. D.; Turley, B. S.; Ziegler, G. P.; Zillioux, E. J. Acute and short-term chronic eﬀects of continuous and intermittent chlorination on Mysidopsis bahia and Menidia beryllina. Environ. Toxicol. Chem. 1994, 13 (9), 1525–1534.

ARTICLE

(31) Naddy, R. B.; Klaine, S. J. Eﬀect of pulse frequency and interval on the toxicity of chlorpyrifos to Daphnia magna. Chemosphere 2001, 45 (45), 497–506. (32) Vaal, M. A.; Van Leeuwen, C. J.; Hoekstra, J. A.; Hermens, J. L. M. Variation in sensitivity of aquatic species to toxicants: Practical consequences for eﬀect assessment of chemical substances. Environ. Manage. 2000, 25 (4), 415–423. (33) Forbes, V. E.; Calow, P.; Sibly, R. M. The extrapolation problem and how population modeling can help. Environ. Toxicol. Chem. 2008, 27 (10), 1987–1994. (34) Chapman, P. M.; Fairbrother, A.; Brown, D. A critical evaluation of safety (uncertainty) factors for ecological risk assessment. Environ. Toxicol. Chem. 1998, 17 (1), 99–108. (35) van Leeuwen, C. J.; Vermeire, T. G. Risk Assessment of Chemicals—An Introduction, 2nd ed.; Springer: Dordrecht, The Netherlands, 2007; p 686. (36) Ahlers, J.; Riedhammer, C.; Vogliano, M.; Ebert, R. U.; Kuhne, R.; Schuurmann, G. Acute to chronic ratios in aquatic toxicity— Variation across trophic levels and relationship with chemical structure. Environ. Toxicol. Chem. 2006, 25 (11), 2937–2945. (37) Sangster, J. Octanol-water partition-coeﬃcients of simple organic-compounds. J. Phys. Chem. Ref. Data 1989, 18 (3), 1111–1229. (38) Singer, H.; Jaus, S.; Hanke, I.; L€uck, A.; Hollender, J.; Alder, A. C. Determination of biocides and pesticides by on-line solid phase extraction coupled with mass spectrometry and their behaviour in wastewater and surface water. Environ. Pollut. 2010, 158 (10), 3054–3064. (39) Swiss Confederation, Bundesgesetz €uber den Schutz der Gew€asser (Federal law on the protection of water bodies). In Die Bundesversammlung der Schweizerischen Eidgenossenschaft: Bern, 1991; Vol. 814.20, p 32. (40) Swiss Confederation, Gew€asserschutzverordnung (Ordonation on the protection of water bodies). In 814.201, Eidgenossenschaft, D. B. d. S., Ed. Bern, 1998; p 60. (41) De Zwart, D., Observed regularities in species sensitivity distributions for aquatic species. In Species Sensitivity Distributions in Ecotoxicology; Posthuma, L., Suter II, G. W., Traas, T. P., Eds.; Lewis Publishers: Boca Raton, 2002; pp 133154. (42) Crane, M.; Newman, M. C. What level of eﬀect is a no observed eﬀect? Environ. Toxicol. Chem. 2000, 19 (2), 516–519. (43) Suter, G. W., II; Rosen, A. E.; Linder, E.; Parkhurst, D. F. Endpoints for responses of ﬁsh to chronic toxic exposures. Environ. Toxicol. Chem. 1987, 6 (10), 793–809. (44) Kunz, P. Y.; Kienle, C.; Gerhardt, A. Gammarus spp. in aquatic ecotoxicology and water quality assessment: Toward integrated multilevel tests. Rev. Environ. Contam. Toxicol. 2010, 205, 1–76. (45) Ashauer, R. Toxicokinetic-toxicodynamic modelling in an individual based context - consequences of parameter variability. Ecol. Model. 2010, 221 (9), 1325–1328. (46) Ashauer, R.; Hintermeister, A.; Potthoﬀ, E.; Escher, B. I. Acute toxicity of organic chemicals to Gammarus pulex correlates with sensitivity of Daphnia magna across most modes of action. Aquat. Toxicol. 2011, 103, 38–45. (47) European and Mediterranean Plant Protection Organization. Environmental risk assessment scheme for plant protection products. Chapter 6: Surface water and sediment. EPPO Bull. 2003, 33 (2), 169–181. (48) European and Mediterranean Plant Protection Organization. Environmental risk assessment scheme for plant protection products. Chapter 7: Aquatic organisms. EPPO Bull. 2003, 33 (2), 183–194. (49) Rozman, K. K.; Doull, J. Dose and time as variables of toxicity. Toxicology 2000, 144 (13), 169–178. (50) van Straalen, N. M., Theory of Ecological Risk Assessment Based on Species Sensitivity Distributions. In Species Sensitivity Distributions in Ecotoxicology; Posthuma, L., Suter II, G. W., Traas, T. P., Eds.; Lewis Publishers: Boca Raton, 2002; pp 3748. (51) Posthuma, L.; Suter II, G. W.; Traas, T. P. Species Sensitivity Distributions in Ecotoxicology; Lewis Publishers: Boca Raton, 2002; p 587. 9791

dx.doi.org/10.1021/es202413a |Environ. Sci. Technol. 2011, 45, 9783–9792

Environmental Science & Technology

ARTICLE

(52) European Community, Directive 98/8/EC of the European Parliament and of the Council of 16 February 1998 concerning the placing of biocidal products on the market. Oﬀ. J. Eur. Commun. 1998, L123, 1-63. (53) European Community, Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 establishing a framework for Community action in the ﬁeld of water policy. Oﬀ. J. Eur. Commun. 2000, L 327, 1-72. (54) Lepper, P. Manual on the Methodological Framework to Derive Environmental Quality Standards for Priority Substances in accordance with Article 16 of the Water Framework Directive (2000/60/EC); FraunhoferInstitute Molecular Biology and Applied Ecology: Schmallenberg, 2005. (55) Junghans, M.; Chevre, N.; Di Paolo, C.; Eggen, R. I. L.; G€alli, R.; Gregorio, V.; H€aner, A.; Homazava, C.; Perazzolo, C.; Kase, R. Aquatic Risks of Plant Protection Products: A Comparison of Diﬀerent Hazard Assessment Strategies for Surface Waters in Switzerland; Swiss Centre for Applied Ecotoxicology, Eawag-EPFL, Duebendorf: 2011. (56) Van den Brink, P. J.; Baveco, J. M.; Verboom, J.; Heimbach, F. An individual-based approach to model spatial population dynamics of invertebrates in aquatic ecosystems after pesticide contamination. Environ. Toxicol. Chem. 2007, 26 (10), 2226–2236. (57) Wang, M.; Grimm, V. Population models in pesticide risk assessment: lessons for assessing population-level eﬀects, recovery, and alternative exposure scenarios from modelling a small mammal. Environ. Toxicol. Chem. 2010, 29 (6), 1292–1300. (58) Ashauer, R.; Agatz, A.; Albert, C.; Ducrot, V.; Galic, N.; Hendriks, J.; Jager, T.; Kretschmann, A.; O’Connor, I.; Rubach, M. N.; Nyman, A.-M.; Schmitt, W.; Stadnicka, J.; van den Brink, P. J.; Preuss, T. G., Toxicokinetic-toxicodynamic modeling of quantal and graded sublethal endpoints: A brief discussion of concepts. Environ. Toxicol. Chem. 2011, 30, (in press). (59) Baas, J.; Willems, J.; Jager, T.; Kraak, M. H. S.; Vandenbrouck, T.; Kooijman, S. Prediction of daphnid survival after in situ exposure to complex mixtures. Environ. Sci. Technol. 2009, 43 (15), 6064–6069. (60) Altenburger, R.; Greco, W. R. Extrapolation concepts for dealing with multiple contamination in environmental risk assessment. Int. Environ. Assess. Manage. 2009, 5 (1), 62–68. (61) Mahar, A. M.; Watzin, M. C. Eﬀects of metal and organophosphate mixtures on Ceriodaphnia dubia survival and reproduction. Environ. Toxicol. Chem. 2005, 24 (7), 1579–1586. (62) Stark, J. D.; Vargas, R. I. Demographic changes in Daphnia pulex (leydig) after exposure to the insecticides spinosad and diazinon. Ecotox. Environ. Safe. 2003, 56 (3), 334–338. (63) Stark, J. D. How closely do acute lethal concentration estimates predict eﬀects of toxicants on populations? Int. Environ. Assess. Manage. 2005, 1 (2), 109–113. (64) Teh, S. J.; Zhang, G. H.; Kimball, T.; Teh, F. C. Lethal and sublethal eﬀects of esfenvalerate and diazinon on splittail larvae. Am. Fish. Soc. Symp. 2004, 2004 (39), 243–253. (65) Brewer, S. K.; Little, E. E.; DeLonay, A. J.; Beauvais, S. L.; Jones, S. B.; Ellersieck, M. R. Behavioral dysfunctions correlate to altered physiology in rainbow trout (Oncorynchus mykiss) exposed to cholinesterase-inhibiting chemicals. Arch. Environ. Contam. Toxicol. 2001, 40 (1), 70–76.

9792

dx.doi.org/10.1021/es202413a |Environ. Sci. Technol. 2011, 45, 9783–9792

ARTICLE pubs.acs.org/est

Dietary Selenium Reduces Retention of Methyl Mercury in Freshwater Fish Poul Bjerregaard,*,† Susanne Fjordside,† Maria G. Hansen,† and Maya B. Petrova† †

Institute of Biology, University of Southern Denmark, Campusvej 55, DK-5230 Odense M, Denmark ABSTRACT: Adverse eﬀects from organic mercury transported along aquatic food chains are health issues in humans and other top predators. Methyl mercury in organisms at the lower food chain levels is eliminated slowly, and laboratory studies have not clariﬁed the role of selenium in the retention of methyl mercury in ﬁsh. Here, we investigated the eﬀects of dietary selenium on the retention of organic and inorganic mercury in freshwater ﬁsh. Addition of selenite to the food augmented elimination of methyl mercury (but not inorganic mercury) from goldﬁsh Carassius auratus in a dose dependent manner; selenite caused methyl mercury to be lost from the general body rather than from any speciﬁc organ. Seleno-cystine and seleno-methionine (but not selenate) likewise promoted elimination of methyl mercury from goldﬁsh. The threshold for the augmenting eﬀect of selenite on the elimination of methyl mercury in the zebra ﬁsh Danio rerio was 0.95 μg Se g 1 food; higher concentrations reduced retention of methyl mercury in a dose dependent manner. Selenium concentrations in the food approaching natural background levels increase the elimination of methyl mercury from ﬁsh. Thus, selenium levels in a given aquatic food chain may aﬀect mercury contamination along the food chain.

’ INTRODUCTION Anthropogenic mobilization of mercury has caused an increase in the ﬂux of mercury through the global atmosphere1 and subsequent elevation of methyl mercury concentrations in aquatic ecosystems also in remote areas far from direct discharges of mercury.2 Methyl mercury is readily assimilated from both food and water in most aquatic organisms, and once assimilated, methyl mercury is retained very eﬃciently, with biological half-lives in various aquatic organisms typically ranging from several weeks in daphnia3 to years in some ﬁsh e.g. pike Esox lucius4 and rainbow trout Oncorhynchus mykiss.5 The level of methyl mercury attained by predators at the top of aquatic food chains is generally determined by biomagniﬁcation processes up through the food chain, and the major amount of methyl mercury enters the aquatic food chains at the lower trophic levels.6 Hence the ability of the various species in the food chain to eliminate methyl mercury determines the levels of methyl mercury attained at the upper end of the food chain where methyl mercury may cause toxic eﬀects in top predators among wildlife7 and neurological symptoms in children of women with a high fraction of aquatic organisms in their diet.8 Selenium interacts with the accumulation and toxicity of mercury in aquatic organisms in fairly complicated ways. In some marine mammals, demethylation of methyl mercury in the liver leads to formation of insoluble HgSe9 giving rise to a 1:1 molar ratio Hg:Se.10 Contrary to this, ﬁsh11 13 and crayﬁsh11 in experimental lakes treated with selenium decreased their mercury contents, and ﬁsh in selenium contaminated areas have been shown to contain reduced amounts of mercury.14 16 r 2011 American Chemical Society

Laboratory investigations on the eﬀect of selenium on mercury handling in ﬁsh and aquatic invertebrates have shown highly variable results (reviewed by17,18), apparently depending on the form of the elements (organic/inorganic), exposure routes (injection, water, food), and timing and concentration of the exposure and the organism in question. Therefore, the results of the previous laboratory investigations are not able to explain the recent ﬁnding19 that ﬁsh in streams of the Western part of USA show lower concentrations of methyl mercury in areas with high selenium levels. However, the fact that selenium administered in the food reduces the concentrations of methyl mercury in liver, kidney, and muscle of rainbow trout Oncorhynchus mykiss20 raises the question if dietary selenium generally aﬀects the retention of methyl mercury in ﬁsh. Therefore, the purpose of the present investigation was to elucidate the eﬀect of dietary selenium on mercury retention in ﬁsh; goldﬁsh Carassius auratus and zebraﬁsh Danio rerio were used as test organisms.

’ EXPERIMENTAL SECTION Experimental Animals. Commercially bred fish were used in the experiments. The size ranges were 4 to 8 cm (average 6.4 g) for the goldfish Carassius auratus and 0.5 to 1.0 g for the zebrafish Received: July 25, 2011 Accepted: October 20, 2011 Revised: September 28, 2011 Published: October 20, 2011 9793

dx.doi.org/10.1021/es202565g | Environ. Sci. Technol. 2011, 45, 9793–9798

Environmental Science & Technology Danio rerio. The fish were acclimated to laboratory conditions for at least one week before experiments. Goldfish and zebrafish were chosen as experimental animals because of their experimental robustness and small enough sizes to fit into the gamma counters for live counting. Exposure to 203Hg. The fish received radioactively labeled mercury (203Hg) at activities of approximately 10.000 counts per minute (cpm), either by intraperitoneal injections (goldfish) or administered in the food (zebrafish). The goldfish were anesthetized in MS-222 (ethyl 3-aminobenzoate methenesulfonate salt; Sigma-Aldrich, Brøndby, Denmark) during injection, and the mercury was dissolved in 0.9% NaCl. Preparation of Food. For experiment 1, selenite enriched trout food was produced specially by Dansk Ørredfoder, Brande A/S. Chemical analysis showed the selenite enriched food to contain 4.7 ( 0.1 μg Se g 1 and the control food to contain 1.5 ( 0.1 μg Se g 1. For experiments 2 and 3, 9 g of commercially purchased food for goldfish (Vitakraft, Bremen) was homogenized in a blender together with 20 mL of water, and 2 g of gelatin (Fluka, BioChimika) was added. The mixture was heated until the gelatin dissolved, and the desired amount and form of selenium was added. The mixture was thoroughly mixed, poured on a glass plate, cooled, and cut into cubes of suitable sizes. For experiment 4, a similar procedure was used except soft parts of blue mussels (Mytilus chilensis) were homogenized instead of goldfish food and no water was added. The concentration of selenium in the food was determined. The Hg-containing food was prepared by the same procedure. Experimental Setups. In experiment 1, groups of 7 goldfish were kept in 40 L aquaria; different coloration allowed recognition of individual fish. In experiments 2 and 3, the fish were kept individually in 2 L aquaria. In experiment 4, the zebrafish were kept individually in 1 L aquaria. The goldfish were kept in tap water (groundwater) at 12.5 °C and the zebrafish at 26 °C in a mixture of 1/3 tap water (groundwater) and 2/3 deionized water, both with a 12:12 light:dark regime. The fish were fed between 1 and 2% of their body weight each day. Experimental Design. The fish received the 203Hg via injection or food where after they were fed control food for 1 to 7 days until feeding with selenium supplemented food began (day 0 of the experiment). The radioactivity of each individual fish was determined by live counting at day 0 and defined as 100%. The radioactivity of each fish was determined regularly and expressed as a percentage of the activity at day 0. Counts were corrected for the physical decay of 203Hg. Experiment 1. It was investigated if injected 203HgCl2 or CH3-203HgCl were retained differently in goldfish fed selenite amended (4.7 ( 0.1 μg Se g 1) or control (1.5 ( 0.1 μg Se g 1) food. Retention of methyl mercury and inorganic was recorded over 77 and 76 days, respectively. After the exposure the organs and a sample of dorsal muscle tissue were dissected out, and the level of 203Hg in the organs was determined. Experiment 2. The effect of the doses of selenium in the food (0, 11, 22, or 39 μg Se g 1) on the retention of injected CH3-203HgCl was studied. The retention of mercury was monitored over 31 days. Experiment 3. The effect of the form of selenium in the food on the retention of injected CH3-203HgCl was studied. The retention of mercury was registered over 28 days in goldfish fed 15 to 17 μg Se g 1 in the form of selenite, selenate, selenomethionine, or seleno-cystine. The control group received food containing 0.19 μg Se g 1.

ARTICLE

Figure 1. Carassius auratus. Retention of methyl mercury (A) and inorganic (B) mercury in whole body goldﬁsh after intraperitoneal injection of radioactively labeled inorganic Hg or methyl-Hg. Mean ( SEM for 7 8 ﬁsh in each group. O and dotted lines: Control food (1.5 μg Se g 1). b and solid lines: Selenite amended food (4.7 μg Se g 1).

Experiment 4. The purpose was to identify the threshold for the concentration of selenite in the food capable of augmenting the elimination of methyl mercury in zebrafish. Thirty-six zebrafish were fed food (based on mussel homogenate) containing CH3-203HgCl for a couple of days until they reached activities between 5000 and 21,000 cpm. The fish were then fed control food (0.408 ( 0.004 μg Se g 1) for 2 days where after the fish were separated into 6 groups, and feeding with the selenite enriched food (0.408 ( 0.004 [control], 0.51 ( 0.15, 0.95 ( 0.29, 1.37 ( 0.24, 2.1 ( 0.7, and 5.6 ( 1.4 μg Se g 1) was initiated. Chemicals. Selenite was purchased from Merck, Darmstadt, sodium selenate from Bie and Berntsen A/S. Seleno-L-methionine and seleno-DL-cystine (the diselenide form of selenocysteine) were obtained from Sigma-Aldrich. 203HgCl2 was obtained from Risø, Denmark, and CH3-203HgCl was synthesized from 203HgCl2 according to ref 21. Chemical Analysis. The radioactivity of the live goldfish was determined in a 7.6 cm well-type Bicron NaI(Tl) crystal gamma counter. The radioactivity of tissues and of the live zebrafish was determined in a Wizard TM3 automatic gamma counter. Selenium concentrations in the food were determined by hydride generation as described in ref 22; PerkinElmer 2380 and FIAS 100 atomic absorption spectrophotometers were used in experiments 1 and 2 to 4, respectively. The reliability of the selenium analysis was investigated by including standard material (DORM and TORT); the measured values were within the certified range. Data Treatment. The activity of each fish was set to 100%, the day the selenium exposure began. Repeated measures ANOVA analysis was used to check for difference in elimination of mercury between the groups (significance level: α = 0.05). The retention of mercury was fitted to first (Ct = C0*e k/t) or second 9794

dx.doi.org/10.1021/es202565g |Environ. Sci. Technol. 2011, 45, 9793–9798

Environmental Science & Technology

ARTICLE

Figure 3. Carassius auratus. Retention of organic mercury in goldﬁsh fed diﬀerent concentrations of selenite in the food. O: Control food (0.19 μg Se g 1). 9: 11 μg Se g 1. 2: 22 μg Se g 1. ): 39 μg Se g 1. Mean ( SEM for 5 or 6 ﬁsh in each group.

Figure 2. Carassius auratus. Distribution of methyl mercury (A) and inorganic mercury (B) after 11 weeks’ exposure to 1.5 (open bars) or 4.7 (solid bars) μg Se g 1 in the food. Mean ( SEM for 7 or 8 ﬁsh. * indicates that the diﬀerence between the two groups is statistically signiﬁcant at the 0.05 level. Percent distribution calculated at the end of the experiment.

(Ct = A*e a/t + B*e b/t) order kinetics. The half-life for the most rapidly exchangeable pool of methyl mercury was estimated from the tangent to the elimination curve at day zero.

’ RESULTS In all of the experiments the ﬁsh consumed both control food and the selenium amended food without any signs of reluctance to eat or adverse eﬀects. Retention of organic and inorganic mercury in goldﬁsh could both be described by two compartment models. Eighty and 69% of the methylmercury and inorganic mercury, respectively, were retained in compartments with no discernible elimination (Figure 1); half-lives for methylmercury and inorganic mercury in the exchangeable compartments were 77 and 57 days, respectively. Exposure to selenite in the food augmented the elimination of methyl mercury from the exchangeable pool of methyl mercury (t1/2 = 37 days; Figure 1A), whereas there was no signiﬁcant eﬀect on the elimination of inorganic mercury (Figure 1B). Seventy-six to 77 days after the injection of the radioactively labeled mercury, the methyl mercury was bound especially in muscle and residual tissues (Figure 2A), whereas the inorganic mercury was bound predominantly in liver, gut, kidney, and residual tissues (Figure 2B). Exposure to selenite in the food did not lead to any redistribution among the tissues of the methyl mercury retained in the goldﬁsh, so the selenium induced increase in the elimination of methyl mercury could not be attributed to augmented loss from any particular tissue or organ. The liver of the selenite exposed goldﬁsh contained a signiﬁcantly lower percentage of the body burden of inorganic mercury than the livers of the nonexposed group (Figure 2B).

Figure 4. Carassius auratus. Retention of organic mercury in goldﬁsh fed diﬀerent forms of selenium in the food. O: Control food (0.19 μg Se g 1). 2: Selenate (16 μg Se g 1). 9: Seleno-methionine (16 μg Se g 1). (: Seleno-cystine (15 μg Se g 1). b: Selenite (16 μg Se g 1). Mean ( SEM for 5 or 6 ﬁsh in each group.

Exposure to 11 and 22 μg Se g 1 in the food caused a dosedependent increase in the elimination of methyl mercury from the goldﬁsh, whereas exposure to 39 μg Se g 1 in the food caused no further increase in the elimination rate than exposure to 22 μg Se g 1 (Figure 3). Exposure to selenium in the form of selenite, seleno-cystine, and seleno-methionine in the food all increased the elimination rate for methyl mercury, while the elimination rate in the group exposed to selenate did not diﬀer from the elimination rate in the control group (Figure 4). In zebraﬁsh Danio rerio, administration of food with selenium concentrations between 0.95 and 5.6 μg Se g 1 caused a dose dependent decrease in the retention of methyl mercury; there was a trend that the group fed food with selenium concentrations at 0.51 μg Se g 1 eliminated methyl mercury faster than the 9795

dx.doi.org/10.1021/es202565g |Environ. Sci. Technol. 2011, 45, 9793–9798

Environmental Science & Technology

Figure 5. Danio rerio. Retention of organic mercury in zebraﬁsh fed diﬀerent concentrations of selenite in the food. O: Control food (0.41 μg Se g 1). 2: 0.51 μg Se g 1. 9: 0.95 μg Se g 1. (: 1.37 μg Se g 1. b: 2.1 μg Se g 1. 1: 5.6 μg Se g 1. Mean ( SEM for 5 or 6 ﬁsh in each group (except the group exposed to 5.6 μg Se g 1, where n = 3).

control group (0.41 μg Se g ), but the diﬀerence was not statistically signiﬁcant (Figure 5). Amendment of the food with selenite decreased the half-lives for methyl mercury in both goldﬁsh (Figure 6A) and zebraﬁsh (Figure 6B), but the eﬀect of small increases relative to the background selenium concentrations in the food appeared more pronounced in the zebraﬁsh than in the goldﬁsh. 1

’ DISCUSSION Exposure to selenium in the food reduces retention of methyl mercury, but not inorganic mercury, in ﬁsh. Methyl mercury appears to be lost from the entire organism, rather than from any speciﬁc tissue or organ. Mercury concentrations in ﬁsh have generally been shown to be reduced in freshwater environments contaminated with selenium or in experimental lakes where selenium concentrations have been modiﬁed. Addition of selenite to controlled ﬁeld enclosures reduced mercury accumulation in pike Esox lucius,23 pearl dace Semotilus margarita, yellow perch Perca ﬂavescens, and white sucker Catostomus commersoni11 and addition of selenite to experimental lakes in Sweden reduced the concentrations of mercury in perch Perca ﬂuviatalis, roach Leuciscus rutilus, and pike E. lucius.12,13 Mercury concentrations in largemouth bass Micropterus salmoides increased in a quarry from which ﬂy ash discharges high in selenium were eliminated,14 and mercury concentrations in freshwater organisms generally decreased with increasing selenium concentrations in a selenium contaminated area.16 Based on the inverse relationship between the contents of selenium and mercury in perch Perca ﬂavescens and walleye (Stizosedion vitreum), Chen et al.24 suggested that exposure to selenium reduces the assimilation of mercury in ﬁsh. These ﬁeld observations are, however, not readily explained by the knowledge on selenium mercury interactions obtained from laboratory experiments, since studies on the eﬀect of selenium on mercury handling in ﬁsh show highly variable results, depending on the form of the two elements, the exposure routes, and the timing and doses/concentrations of the exposures. When injected

ARTICLE

Figure 6. Carassius auratus and Danio rerio. Relations between half-lives for methyl mercury and selenium concentrations in the food amended with selenite. ): Exp. 1; 1: Exp. 2; 2: Exp. 3; 0: Exp. 4.

intramuscularly, selenite (0.4 mg Se kg 1) did not aﬀect the overall body retention of injected methyl mercury (1 mg Hg kg 1) in killiﬁsh Fundulus heteroclitus over 24 h, but the selenite exposure led to a redistribution of methyl mercury among the organs resulting in decreased concentrations in kidney and red blood cells.25 Exposure to 2.5 and 5 mg Se kg 1 wet wt food (as selenite) did not aﬀect accumulation from food (2.5 mg Hg kg 1 wet wt) of methyl mercury in muscle and liver of cod Gadus morhua over 32 days, whereas selenite exposure increased methyl mercury accumulation in the brain.26 Exposure to selenite and selenate in the water (2 and 200 μg Se L 1) did not aﬀect the accumulation of methyl mercury from water (1 μg Hg L 1) in tissues of plaice Pleuronectes platessa.27 Exposure of zebraﬁsh larvae to selenomethionine ameliorated some of the neurobehavioral eﬀects of concurrent exposure to methyl mercury but had no eﬀect on the accumulation of mercury.28 Waterborne selenite did not aﬀect methyl mercury accumulation in zebraﬁsh Brachydanio rerio29 and medaka Oryzias latipes.30 Exposure to 9.7 μg Se-SeO3— g 1 in the food reduced the concentrations of methyl mercury in liver, kidney, and muscle in rainbow trout.20 Concerning the eﬀects of selenium on the kinetics of inorganic mercury, it is also diﬃcult to obtain a clear picture from the experiments carried out with ﬁsh.27,31 34 In the majority of the investigations mentioned above, aspects of mercury uptake and accumulation processes were studied, whereas only the retention of accumulated mercury is considered in the present investigation. The results of the present investigation may indicate that in the in situ investigations in which selenium has been shown to lower the mercury concentrations in ﬁsh,11 13,15,16,23,24 selenium has aﬀected elimination rather than uptake processes. Selenium concentrations in freshwater ﬁsh generally increase with the concentration of selenium in the water24 although direct uptake from the water phase may diﬀer among species. Rainbow trout readily accumulates selenium in the tissues upon exposure 9796

dx.doi.org/10.1021/es202565g |Environ. Sci. Technol. 2011, 45, 9793–9798

Environmental Science & Technology to waterborne selenite,35 whereas turbot Scophthalmus maximus accumulates selenium in the tissues from selenite in the water very slowly or not at all.36 Although rainbow trout is able to accumulate selenium directly from water, food is considered to be the dominant source of selenium37,38 as it is indeed in many aquatic organisms.39,40 Furthermore,41 it is suggested that dietary selenium is metabolized in the liver of trout and stored in organic form, while selenite taken up from water is stored as inorganic selenium. It is thus conceivable that selenite given in the water phase may aﬀect mercury handling diﬀerently from dietary selenite, both in terms of the amount and the chemical form of selenium present in the tissues for interaction with mercury. Administration of selenate does not appear to aﬀect the retention of methyl mercury in goldﬁsh, and this is in accordance with the knowledge that selenate is taken up very slowly or not at all in aquatic organisms.22 The biochemical mechanisms underlying the eﬀects of selenium on the elimination of mercury from ﬁsh observed in the present laboratory study and in situ investigations11 13,23 are not known. Since the laboratory studies on the eﬀect of selenium on mercury retention indicate that selenium administered in the food is the only route of administration that has consistent eﬀects on methyl mercury retention, interactions between selenium and mercury within the digestive tract may be hypothesized. Entero-hepatic recirculation of methyl mercury in rainbow trout 42 has been suggested to explain the long half-life (>200 days) for methyl mercury observed,5,43 and it is possible that selenite in the food could react (e.g., microbiologically) in the intestine to form compounds that could bind methyl mercury and thus interrupt the entero-hepatic recirculation, leading to elimination of the methyl mercury via the faeces. In some organisms, selenium present in the tissues is known to ameliorate the toxic eﬀects of accumulated methyl mercury by interactions at the molecular level.44 46 This has generated interest in elucidating molecular ratios between selenium and mercury in tissues with the aim of identifying the amount of selenium needed to be present to protect against the adverse eﬀects of methyl mercury. In a recent investigation on the mercury and selenium concentrations in streams of the western USA, Peterson et al.19 concluded that ‘high Hg concentrations in ﬁsh tissue.... were found only when Se concentrations in the same tissue were low’. The results of the present investigation underline the need to consider the eﬀect of selenium on methyl mercury biokinetics as well as the interactions at the biochemical level in the tissues if the full mechanism underlying selenium’s inﬂuence on mercury biomagniﬁcation and toxicity is to be elucidated.

’ AUTHOR INFORMATION Corresponding Author

*Phone: +45 6552456. Fax: 45 6550 2786. E-mail: poul@ biology.sdu.dk.

’ ACKNOWLEDGMENT We thank Dansk Ørredfoder A/S, Brande, Denmark, for supplying the selenite amended food used in experiment 1. This investigation was supported by grants from the Danish Natural Science Research Council. ’ REFERENCES (1) UNEP. Global Mercury Assessment; United Nations Environment Programme Chemicals: Geneva, Switzerland, 2002; pp 1 270.

ARTICLE

(2) Gantner, N.; Power, M.; Babaluk, J. A.; Reist, J. D.; Kock, G.; Lockhart, L. W.; Solomon, K. R.; Muir, D. C. G. Temporal Trends of Mercury, Cesium, Potassium, Selenium, and Thallium in Arctic Char (Salvelinus alpinus) from Lake Hazen, Nunavut, Canada: Eﬀects of Trophic Position, Size, and Age. Environ. Toxicol. Chem. 2009, 28, 254–263. (3) Karimi, R.; Chen, C. Y.; Pickhardt, P. C.; Fisher, N. S.; Folt, C. L. Stoichiometric controls of mercury dilution by growth. Proc. Natl. Acad. Sci. USA 2007, 104, 7477–7482. (4) Tillander, M.; Miettinen, J. K.; Rissanen, K.; Miettinen, V.; Minkkinen, E. The excretion by ﬁsh, mussel, mollusc and crayﬁsh of methyl mercury nitrate and phenyl mercury nitrate, introduced orally or injected into musculature. Nord. Hyg. Tidskr. 1969, 50, 181–183. (5) Ruohtula, M.; Miettinen, J. K. Retention and Excretion of Hg-203Labeled Methylmercury in Rainbow trout. Oikos 1975, 26, 385–390. (6) Riget, F.; Moller, P.; Dietz, R.; Nielsen, T. G.; Asmund, G.; Strand, J.; Larsen, M. M.; Hobson, K. A. Transfer of mercury in the marine food web of West Greenland. J. Environ. Monit. 2007, 9, 877–883. (7) Scheuhammer, A. M.; Basu, N.; Burgess, N. M.; Elliott, J. E.; Campbell, G. D.; Wayland, M.; Champoux, L.; Rodrigue, J. Relationships among mercury, selenium, and neurochemical parameters in common loons (Gavia immer) and bald eagles (Haliaeetus leucocephalus). Ecotoxicology 2008, 17, 93–101. (8) Grandjean, P.; Weihe, P.; White, R. F.; Debes, F.; Araki, S.; Yokoyama, K.; Murata, K.; Sorensen, N.; Dahl, R.; Jorgensen, P. J. Cognitive deﬁcit in 7-year-old children with prenatal exposure to methylmercury. Neurotoxicol. Teratol. 1997, 19, 417–428. (9) Martoja, R.; Viale, D. Storage of Mercuric Selenide Concretions in Liver of Cetacean Mammals - A Possible Process for Detoxiﬁcation of Methylmercury by Selenium. C. R. Seances Acad. Sci., Ser. D 1977, 285, 109–+. (10) Koeman, J. H.; Peeters, W. H. M.; Koudstaa, C. H.; Tjioe, P. S.; Goeij, J. J. M. D. Mercury-Selenium Correlations in Marine Mammals. Nature 1973, 245, 385–386. (11) Turner, M. A.; Rudd, J. W. M. The English Wabigoon River System. 3. Selenium in Lake Enclosures - Its Geochemistry, Bioaccumulation, and Ability to Reduce Mercury Bioaccumulation. Can. J. Fish. Aquat. Sci. 1983, 40, 2228–2240. (12) Paulsson, K.; Lundbergh, K. Treatment of Mercury Contaminated Fish by Selenium Addition. Water, Air, Soil Pollut. 1991, 56, 833–841. (13) Paulsson, K.; Lundbergh, K. The Selenium Method for Treatment of Lakes for Elevated Levels of Mercury in Fish. Sci. Total Environ. 1989, 87 8, 495–507. (14) Southworth, G. R.; Peterson, M. J.; Ryon, M. G. Long-term increased bioaccumulation of mercury in largemouth bass follows reduction of waterborne selenium. Chemosphere 2000, 41, 1101–1105. (15) Southworth, G. R.; Peterson, M. J.; Turner, R. R. Changes in Concentrations of Selenium and Mercury in Largemouth Bass Following Elimination of Fly-Ash Discharge to A Quarry. Chemosphere 1994, 29, 71–79. (16) Belzile, N.; Chen, Y. W.; Gunn, J. M.; Tong, J.; Alarie, Y.; Delonchamp, T.; Lang, C. Y. The eﬀect of selenium on mercury assimilation by freshwater organisms. Can. J. Fish. Aquat. Sci. 2006, 63, 1–10. (17) Cuvinaralar, M. L. A.; Furness, R. W. Mercury and Selenium Interaction - A Review. Ecotoxicol. Environ. Saf. 1991, 21, 348–364. (18) Pelletier, E. Mercury-Selenium Interactions in Aquatic Organisms - A Review. Mar. Environ. Res. 1986, 18, 111–132. (19) Peterson, S. A.; Ralston, N. V. C.; Peck, D. V.; Van Sickle, J.; Robertson, J. D.; Spate, V. L.; Morris, J. S. How Might Selenium Moderate the Toxic Eﬀects of Mercury in Stream Fish of the Western US? Environ. Sci. Technol. 2009, 43, 3919–3925. (20) Bjerregaard, P.; Andersen, B. W.; Rankin, J. C. Retention of methyl mercury and inorganic mercury in rainbow trout Oncorhynchus mykiss (W): eﬀect of dietary selenium. Aquat. Toxicol. 1999, 45, 171–180. (21) Toribara, T. Y. Preparation of CH3(HgCl)-Hg-203 of High Speciﬁc Activity. Int. J. Appl. Radiat. Isot. 1985, 36, 903–904. (22) Sorensen, M.; Bjerregaard, P. Interactive Accumulation of Mercury and Selenium in the Sea Star Asterias rubens. Mar. Biol. 1991, 108, 269–276. 9797

dx.doi.org/10.1021/es202565g |Environ. Sci. Technol. 2011, 45, 9793–9798

Environmental Science & Technology (23) Turner, M. A.; Swick, A. L. The English Wabigoon River System 0.4. Interaction Between Mercury and Selenium Accumulated from Waterborne and Dietary Sources by Northern Pike (Esox lucius). Can. J. Fish. Aquat. Sci. 1983, 40, 2241–2250. (24) Chen, Y. W.; Belzile, N.; Gunn, J. M. Antagonistic eﬀect of selenium on mercury assimilation by ﬁsh populations near Sudbury metal smelters?. Limnol. Oceanogr. 2001, 46, 1814–1818. (25) Sheline, J. Schmidt-Nielsen, B. Methylmercury-selenium: Interaction in the killiﬁsh, Fundulus heteroclitus. In Physiology response of marine biota to pollutants; Vernberg, F. J., Ed.; Academic Press: New York, 1977; pp 119 130. (26) Ringdal, O.; Julshamn, K. Eﬀect of Selenite on the Uptake of Methylmercury in Cod (Gadus morhua). Bull. Environ. Contam. Toxicol. 1985, 35, 335–344. (27) Davies, I. M.; Russell, R. The Inﬂuence of Dissolved SeleniumCompounds on the Accumulation of Inorganic and Methylated Mercury-Compounds from Solution by the Mussel Mytilus edulis and the Plaice Pleuronectes platessa. Sci. Total Environ. 1988, 68, 197–205. (28) Weber, D. N.; Connaughton, V. P.; Dellinger, J. A.; Klemer, D.; Udvadia, A.; Carvan, M. J. Selenomethionine reduces visual deﬁcits due to developmental methylmercury exposures. Physiol. Behav. 2008, 93, 250–260. (29) Ribeyre, F.; Amiardtriquet, C.; Boudou, A.; Amiard, J. C. Experimental-Study of Interactions Between 5 Trace-Elements Cu, Ag, Se, Zn, and Hg Toward Their Bioaccumulation by Fish (Brachydanio rerio) from the Direct Route. Ecotoxicol. Environ. Saf. 1995, 32, 1–11. (30) Liao, C. Y.; Zhou, Q. F.; Fu, J. J.; Shi, J. B.; Yuan, C. G.; Jiang, G. B. Interaction of methylmercury and selenium on the bioaccumulation and histopathology in Medaka (Oryzias latipes). Environ. Toxicol. 2007, 22, 69–77. (31) Cuvin, M. L. A.; Furness, R. W. Uptake and Elimination of Inorganic Mercury and Selenium by Minnows Phoxinus phoxinus. Aquat. Toxicol. 1988, 13, 205–216. (32) Cuvinaralar, M. L. A.; Furness, R. W. Tissue Distribution of Mercury and Selenium in Minnows, Phoxinus phoxinus. Bull. Environ. Contam. Toxicol. 1990, 45, 775–782. (33) Kim, J. H.; Birks, E.; Heisinger, J. F. Protective Action of Selenium Against Mercury in Northern Creek Chubs. Bull. Environ. Contam. Toxicol. 1977, 17, 132–136. (34) Jorgensen, D.; Heisinger, J. F. The Eﬀects of Selenium on the Distribution of Mercury in the Organs of the Black Bullhead (Ictalurus melas). Comp. Biochem. Physiol., Part C: Pharmacol., Toxicol. Endocrinol. 1987, 87, 181–186. (35) Hodson, P. V.; Spry, D. J.; Blunt, B. R. Eﬀects on RainbowTrout (Salmo gairdneri) of A Chronic Exposure to Waterborne Selenium. Can. J. Fish. Aquat. Sci. 1980, 37, 233–240. (36) Nielsen, G.; Bjerregaard, P. Interaction Between Accumulation of Cadmium and Selenium in the Tissues of Turbot Scophthalmus maximus. Aquat. Toxicol. 1991, 20, 253–265. (37) Hilton, J. W.; Hodson, P. V.; Slinger, S. J. Absorption, Distribution, Half-Life and Possible Routes of Elimination of Dietary Selenium in Juvenile Rainbow-Trout (Salmo gairdneri). Comp. Biochem. Physiol., Part C: Pharmacol., Toxicol. Endocrinol. 1982, 71, 49–55. (38) Hilton, J. W.; Hodson, P. V.; Slinger, S. J. The Requirement and Toxicity of Selenium in Rainbow-Trout (Salmo gairdneri). J. Nutr. 1980, 110, 2527–2535. (39) Baines, S. B.; Fisher, N. S.; Stewart, R. Assimilation and retention of selenium and other trace elements from crustacean food by juvenile striped bass (Morone saxatilis). Limnol. Oceanogr. 2002, 47, 646–655. (40) Luoma, S. N.; Presser, T. S. Emerging Opportunities in Management of Selenium Contamination. Environ. Sci. Technol. 2009, 43, 8483–8487. (41) Hodson, P. V.; Hilton, J. W. The Nutritional Requirements and Toxicity to Fish of Dietary and Waterborne Selenium. Ecol. Bull. 1983, 335–340. (42) Giblin, F. J.; Massaro, E. J. Erythrocyte Transport and Transfer of Methylmercury to Tissues of Rainbow trout (Salmo gairdneri). Toxicology 1975, 5, 243–254.

ARTICLE

(43) Giblin, F. J.; Massaro, E. J. Pharmacodynamics of Methyl Mercury in Rainbow-Trout (Salmo gairdneri) - Tissue Uptake, Distribution and Excretion. Toxicol. Appl. Pharmacol. 1973, 24, 81–91. (44) Khan, M. A. K.; Wang, F. Y. Mercury-Selenium Compounds and Their Toxicological Signiﬁcance: Toward A Molecular Understanding of the Mercury-Selenium Antagonism. Environ. Toxicol. Chem. 2009, 28, 1567–1577. (45) Yang, D. Y.; Chen, Y. W.; Gunn, J. M.; Belzile, N. Selenium and mercury in organisms: Interactions and mechanisms. Environ. Rev. 2008, 16, 71–92. (46) Ralston, N. V. C.; Raymond, L. J. Dietary selenium’s protective eﬀects against methylmercury toxicity. Toxicology 2010, 278, 112–123.

9798

dx.doi.org/10.1021/es202565g |Environ. Sci. Technol. 2011, 45, 9793–9798

ARTICLE pubs.acs.org/est

Cr(VI)/Cr(III) and As(V)/As(III) Ratio Assessments in Jordanian Spent Oil Shale Produced by Aerobic Combustion and Anaerobic Pyrolysis Tayel El-Hasan,† Wojciech Szczerba,‡ G€unter Buzanich,‡ Martin Radtke,‡ Heinrich Riesemeier,‡ and Michael Kersten§,* †

Geology Department, Faculty of Science, Taibah University, Al-Madinah Al-Munawwarah, KSA BAM Federal Institute for Materials Research and Testing, Department 1: Analytical Chemistry and Reference Materials, Berlin 12200, Germany § Geosciences Institute, Johannes Gutenberg-University, Mainz 55099, Germany ‡

ABSTRACT: With the increase in the awareness of the public in the environmental impact of oil shale utilization, it is of interest to reveal the mobility of potentially toxic trace elements in spent oil shale. Therefore, the Cr and As oxidation state in a representative Jordanian oil shale sample from the El-Lajjoun area were investigated upon diﬀerent lab-scale furnace treatments. The anaerobic pyrolysis was performed in a retort ﬂushed by nitrogen gas at temperatures in between 600 and 800 °C (pyrolytic oil shale, POS). The aerobic combustion was simply performed in porcelain cups heated in a muﬄe furnace for 4 h at temperatures in between 700 and 1000 °C (burned oil shale, BOS). The high loss-on-ignition in the BOS samples of up to 370 g kg 1 results from both calcium carbonate and organic carbon degradation. The LOI leads to enrichment in the Cr concentrations from 480 mg kg 1 in the original oil shale up to 675 mg kg 1 in the g850 °C BOS samples. Arsenic concentrations were not much elevated beyond that in the average shale standard (13 mg kg 1). Synchrotronbased X-ray absorption near-edge structure (XANES) analysis revealed that within the original oil shale the oxidation states of Cr and As were lower than after its aerobic combustion. Cr(VI) increased from 0% in the untreated or pyrolyzed oil shale up to 60% in the BOS ash combusted at 850 °C, while As(V) increased from 64% in the original oil shale up to 100% in the BOS ash at 700 °C. No Cr was released from original oil shale and POS products by the European compliance leaching test CEN/TC 292 EN 12457-1 (1:2 solid/water ratio, 24 h shaking), whereas leachates from BOS samples showed Cr release in the order of one mmol L 1. The leachable Cr content is dominated by chromate as revealed by catalytic adsorptive stripping voltammetry (CAdSV) which could cause harmful contamination of surface and groundwater in the semiarid environment of Jordan.

1. INTRODUCTION Oil shale is a ﬁne-grained sedimentary rock containing relatively large amounts of kerogen from which it is possible to proﬁtably extract oil and combustible gas suitable for energy production.1 Oil shale is therefore considered an important alternative prospective source of energy in the world and especially in Jordan.2 However, due to still relatively low cost of crude oil compared to the cost of mining and extraction of oil shale, only few oil shale deposits have been exploited in the world. Nonetheless, the utilization plans of the oil shale as a result of the steady increase in oil prices are eminent. By 2010, oil shale extraction was being undertaken in Estonia, Brazil, and China, while Australia, the U.S., Canada, and Jordan are planning to start or restore shale oil production. A recent report reveals that about 34 billion barrels of shale oil can be r 2011 American Chemical Society

produced from the proven reserves at the main ﬁve locations on north, central, and southern parts of Jordan (El-Lajjoun, Sultani, Ed-Darawish, Attarat, and Maghar), which put Jordan at rank 7 of the world oil shale containing countries.3 On the other hand, Jordan is considered one of the ﬁve most water scarce countries in the world. There is concern that a sizable oil shale industry could negatively impact Jordan’s limited water supply. Studies have proven the feasibility of using the oil shale as a power generating fuel for direct combustion in a circulating Received: March 1, 2011 Accepted: October 5, 2011 Revised: September 26, 2011 Published: October 05, 2011 9799

dx.doi.org/10.1021/es200695e | Environ. Sci. Technol. 2011, 45, 9799–9805

Environmental Science & Technology ﬂuidized-bed (CFB) process at temperatures above 650 750 °C. However, a rough calculation suggests that a 300 MW thermal power plant would require 20 000 tons of oil shale, producing 11 000 tons of burned oil shale (BOS) ash annually,4 and even more if coupled with oil shale retorting facilities.5 There are plans to ﬁll up the open pit mines with the ash and spent shale.2 However, there is concern that through their interaction with rain and groundwater, the deposits could form leachate that could resemble toxic mine drainage causing degradation of surface and groundwater quality. This is of particular importance for semiarid environments such as in the Jordanian El-Lajjoun area which is rich in oil shale resources but have also a medium to high susceptibility to groundwater pollution.4 It is therefore crucial to analyze these wastes for their content and speciation of leachable toxic elements in order to take precautionary measures. The high Ca/S ratio renders the ashes rich in free lime (portlandite) which results in highly alkaline pH values of the leachates (pH 10 13). Classical heavy metals like Ni, Cu, Zn, Cd, and Pb are not very mobile under such conditions due to strong surface adsorption.6 On the other hand, oxyanion forming redox-sensitive metals and metalloids like V, Mo, As, Se, and Cr may well be mobilized under such conditions.7 Previous studies of BOS ash leachates have revealed concentrations in the order of 20 μM for Cr and 4 μM for As,4,8 which are well above the respective WHO threshold of groundwater utilization for drinking water (1 μM for Cr(VI)). No speciation analysis have been performed of the dissolved Cr, but chromium identiﬁed in alkaline leachates is almost always hexavalent because equilibrium with insoluble Ca Cr(III) minerals causes Cr(OH)4‑ species to be present at vanishingly small concentrations levels. The main concern to be veriﬁed is, therefore, that the Cr could occur in the highly toxic hexavalent form. To assess potential environmental impact of the BOS ashes and in case to develop and improve on ash stabilization and toxic element immobilization methodology, both solute and solid Cr speciation analysis are clearly desirable. For solid speciation analysis of trace elements in environmental samples on a molecular level, synchrotron beamlines have the potential of deploying in parallel X-ray ﬂuorescence, diﬀraction, and absorption spectroscopy.9 13 X-ray absorption near-edge spectroscopy (XANES) is extremely sensitive to the absorbing target atom oxidation state, site symmetry, and ligand covalence in solid samples. Much has been learned about chromium chemistry in coal combustion products utilizing such XANES measurements.11,12 XANES characterization of arsenic speciation in oil shale and its derivatives was ﬁrst performed by Cramer et al.13 However, they have not studied BOS ash for which to our knowledge there is yet no report on such toxic trace element speciation measurements. This baseline study aims at characterizing the potentially harmful oxidation state of Cr and As in the original oil shale rock upon combustion or pyrolysis. These results are essential in order to evaluate the potential leachability of these trace elements from the spent oil shale tailings traced by classical leaching tests, and hence, the toxicity to the environment and potential remediation measures.

2. MATERIAL AND METHODS 2.1. Sampling and Sample Preparation. The El-Lajjoun area is located in western part of Central Jordan about 110 km south of the capital city of Amman, with an estimated area of 34 km2.4

ARTICLE

The oil shale deposit is relatively shallow and near to the surface which favors an open pit mining, with a proven reserve of 1.2 billion tons of oil shale having an oil content of 125 million tons.2 Bender14 first used the term “oil shale” when studying this deposit and described it as an euxinic facies containing thick lenses of black, strongly bituminous shale, and brown-black, bedded, bituminous limestone. The sample was taken from an outcropping lithological unit of the main oil shale stratum with a thickness reaching 86 m.15 It is composed of bituminous marl that has thin-bedded and laminated features with intercalations of bituminous limestone of gray to light gray color. The lower portions are phosphatic marl intercalated with slightly bituminous thin black chert layers. The oil shale sample from El-Lajjoun area (10 kg) was crushed, ground, and sieved <63 μm to mimic the CFB furnace input. The ﬁrst subsample of 500 g was divided into four fractions, placed in open porcelain dishes and burned at 700 °C, 850 °C, or 1000 °C, respectively, during four hours using a muﬄe furnace. These aerobically burned oil shale (BOS) samples were allowed to cool down under open dry atmosphere for 24 h and sealed into polyethylene vials. A second oil shale subsample was treated by anaerobic pyrolysis. For this, 300 g batches of the sieved oil shale were placed into a cylindrical glass retort and ﬂushed with nitrogen at a ﬂow rate of 100 mL min 1. During heat-up, the generated hydrocarbons (“shale oil”) were passed into a water-cooled receiver. When the desired temperature set point was reached, the heater was switched oﬀ immediately, but the N2 ﬂow was maintained for cooling the retort down to 150 °C. The sample was then removed from the retort, cooled down to room temperature in a desiccator, weighted, and sealed into polyethylene vials. Five pyrolytic oil shale (POS) samples were thus obtained at respective set points of 600 800 °C in 50 °C intervals. A temperature range beyond the common Fischer Assay (ISO 647) was chosen to cover the full succession in shale oil decomposition reactions from pyrobitumen over asphaltenes and carboids up to methane formation. 2.2. Bulk Element Analysis. For analysis by bulk X-ray fluorescence spectrometry, sample pellets were prepared using a hydraulic press (Herzog, Germany), in which the finely powdered samples with a minor epoxy resin admixture were cast at 7-tonne pressure in a 30-mm DIE. A PANalytical MagiX Pro wavelength dispersive X-ray spectrometer (WD-XRF) was used for the total concentration determination of major and trace elements in the samples. The MagiX Pro is a sequential instrument with a single goniometer based measuring channel equipped with rhodium anode X-ray tube operated at 3.6 kW. The system analytical software SuperQ V4.0 was used for data collection, calibration, and analysis. Samples were run against a collection of hundred international standard reference materials in the databank to prepare the calibration curves for major and trace elements and to check for the accuracy of the analytical data. Loss-on-ignition (LOI) was determined after oven muffling at 1000 °C of subsamples and weighting after cooling in a desiccator over blue silicagel. Content of total carbon (C) and sulfur (S) was analyzed by using a Leco CS 244 combustion oven with infrared detector on gas emissions (ASTM D-1551). Relative differences in the analytical results between the original oil shale and the different BOS ash residuals after LOI determination for the major element concentrations were estimated at less than 5%, while that on the trace elements at less than 10%. 9800

dx.doi.org/10.1021/es200695e |Environ. Sci. Technol. 2011, 45, 9799–9805

Environmental Science & Technology

ARTICLE

Table 1. Bulk Chemical Characteristics of the Original Oil Shale and Its Spent Products from the Aerobic Combustion (BOS) and Anaerobic Pyrolysis (POS), and Dissolved As and Cr Concentrations in Their Leachates, Respectively LOI sample

a

C 1

(g kg )

S 1

(g kg )

Cr 1

Crtota

As 1

1

(g kg )

(mg kg )

(mg kg )

1

(mg L )

oil shale Ave. shale20

369

213 12

35.1 2.4

484 90

12 13

<0.003

39

Cr(III)b

Cr(VI)b

1

1

(mg L )

(mg L )

Astota (μg L 1) <20

BOS 700 °C

148

17.7

572

14

67.2 ( 0.4

6.9 ( 0.8

39.0 ( 0.7

BOS 850 °C

46

9.2

20.1

675

15

54.1 ( 0.3

6.5 ( 0.9

34.0 ( 0.9

187 ( 3 133 ( 24

BOS 1000 °C

14

8.8

18.9

672

17

38.6 ( 0.1

7.4 ( 0.5

25.9 ( 0.6

94 ( 7

POS 600 °C

336

153

22.2

513

12

<0.003

<20

POS 650 °C

331

150

22.6

498

12

<0.003

<20

POS 700 °C

337

149

22.2

503

12

<0.003

<20

POS 750 °C POS 800 °C

330 310

128 138

20.1 21.9

517 518

12 13

<0.003 <0.003

<20 <20

ICP-OES measurements of the leachates. b CAdSV measurements of the leachates.

2.3. Leaching Test. Many leaching tests, each characterized by different experimental conditions have been developed worldwide. For the present study, the European compliance leaching test CEN/TC 292 EN 12457-1 (L/S = 2) was preferred. Three subsamples of all pulverized dry samples were rotated for 24 h in 50 mL PE centrifuge tubes (17.5 g dry pulverized sample in 35 mL of water). The pH was not controlled because alkaline conditions minimize undesired Cr(III) Cr(VI) interconversion during digestion. The solid residue was then separated from the leachate by centrifugation for 30 min at >2000g. The pH values of the supernatants were measured using a Metrohm 827 pH meter. Upon decantation into a disposable plastic syringe, the leachates were additionally membrane-filtered (25 mm, 0.2 μm) and acidified by HCl (pH 2). Total dissolved As and Cr was measured by ICP-OES (Spectro Ciros Vision, Kleve, Germany) with a LOD of 0.2 μM for As and 0.05 μM for Cr (radial plasma viewing and vacuum sample injection unit). Once detected by ICP-OES, speciation of the dissolved chromium into Cr(III) and Cr(VI) was performed by catalytic adsorptive stripping voltammetry (CAdSV). A Metrohm 797 VA Computrace HDME measuring stand was used according to the Metrohm Application Note No. V-82. The speciation approach is based on the difference in the time dependence of the CAdSV signals of total Cr and Cr(VI) species obtained in acetate buffer solutions (0.1 M, pH 6) containing 10 mM DTPA and 10 mM NaNO3.16,17 Calibration was performed by standard additions of Cr(III) or Cr(VI) spikes. 2.4. XANES Measurements. X-ray absorption measurements were performed in the synchrotron facility BESSY II in Berlin at a beamline with a 7 T wavelength shifter (7T-WLS/1-bamlineKMC, “BAMline”).18 The beam energy was tuned using both the double-multilayer monochromator and the double-crystal monochromator installed at the BAMline. This setup enables a precise energy adjustment with a relative resolution of about 2 10 4, that is, 1.2 eV absolute at the energy of the Cr K-edge (5989 eV). For Cr the excitation energy was varied in the prepeak region from 5981 to 6005 eV in steps of 0.5 eV. Additional points for normalization far below (5900 to 5980 eV) and above the edge (6074 to 6089 eV) were measured. In the case of arsenic the measured energy range for the K-edge was from 11 848 to 11 907 eV. Same as in case of Cr, additional points for normalization were measured below the edge (11 778 to 11 830 eV) and above the white line (11 909 to 11 964 eV). Only X-ray absorption near edge spectra (XANES) measurements were performed because of the Mn

content and proximity of the Mn K-edge to the Cr K-edge, and the overall low As concentrations. Reference spectra were recorded from chromium metal foil for energy calibration, from Cr(NO3)3 3 9H2O for Cr(III), and from K2CrO4 for Cr(VI). Both chemicals were purchased from Merck in p.a. quality. For As(III), auripigment, As2S3, and for As(V), scorodite, Fe(AsO4) 3 2H2O, were measured as natural reference minerals. About 20 mg each of the reference samples were diluted by polyethylene powder, while the oil shale and BOS samples were used undiluted and pressed to pellets using a 13 mm DIE. The size of the beam spot was adapted to the Cr and As content, respectively, in the order of 0.3 1 mm2. Because the trace concentrations of Cr and As in the samples, they showed not enough absorption to be measured in transmission mode, and the XANES spectra were taken in ﬂuorescence mode. The ﬂuorescence radiation of the sample was detected at an angle of 90° with respect of the X-ray beam. The sample was tilted 45 degrees with respect to both the X-ray beam and the detector. The X-ray ﬂuorescence of the K lines was detected with a Si(Li)detector. A minimum of 10 000 counts was acquired in each measured point behind the edge, corresponding to an uncertainty of 1% (Poisson distribution). Several XANES spectra were collected for each sample and pooled to improve the statistical results. The ﬁrst and last spectra were compared to ensure that no beam damage has occurred, which may lead to oxidation state changes due to the relatively high organic carbon content. The measurements were performed at room temperature, since no such changes were detected. The Athena frontend from the software package IFEFFIT 1.2.11c was used for the XANES spectra deconvolution.19

3. RESULTS 3.1. Bulk Composition. The bulk chemical composition has already been discussed in the context of chemostratigraphic analysis to unravel the depositional and diagenetic history of the oil shale.15 The high LOI of 370 g kg 1 results from both calcium carbonate and organic carbon degradation (210 g kg 1 C, Table 1). The element concentrations on dry-weight basis are therefore increasing in the BOS ashes due to a significant loss in mass. Pyrolysis, on the other hand, did not reduce the LOI significantly because of relatively short heating times (Table 1). The main constituents are Ca (300 g kg 1 in LOI sample) and Si 9801

dx.doi.org/10.1021/es200695e |Environ. Sci. Technol. 2011, 45, 9799–9805

Environmental Science & Technology

ARTICLE

Figure 2. XANES Cr pre-edge features for untreated oil shale and the spent shale samples (BOS, POS) in comparison with the Cr(III) and Cr(VI) standards.

Figure 1. Cr K-edge XANES spectra for untreated oil shale and the spent shale samples (BOS, POS) in comparison with those of the Cr(III) and Cr(VI) standards.

(190 g kg 1 in LOI sample), by which the BOS residues can be characterized as calcareous siliceous ashes. Cr is enriched in the original oil shale (480 mg kg 1) compared to “average shale” (90 mg kg 1).20 Similar Cr enrichments have been found in Pennsylvanian Paleozoic black shales.21 On the other hand, the content of As in the oil shale is similar to the average shale reference (13 mg kg 1).20 The S content is enriched in the original oil shale (3.5 g kg 1), and a substantial part is known to be organically bound (30 80%), whereas only a minor part occurs as pyrite.2 Accordingly, the Fe content in the oil shale (0.75 g kg 1) is dominated by sulfidic Fe(II), while in the BOS ash it is probably oxidized to Fe(III) oxide compounds. 3.2. Cr Speciation. Figure 1 shows the Cr K-edge XANES spectra of all the samples and reference components. The Cr K-edge XANES spectrum may show a very distinct pre-edge structure and, if present, can favorably be exploited to determine

the Cr(III)/Cr(VI) ratio.11,12,22 In both chromate (CrO42‑) and dichromate (Cr2O72‑) oxyanions, chromium is in tetrahedral coordination with oxygen atoms, enhancing an otherwise forbidden s-d dipole transition by a dipole transition from 1s to d-p hybrid orbital. Thus, tetrahedral coordinated Cr shows an intense pre-edge peak. On the other hand, Cr(III) is octahedrally coordinated, resulting in a very weak s-d transition and a pre-edge structure with negligible intensity. The Cr(VI) content in the studied samples has been evaluated by a simple peak height ratio analysis of the characteristic Cr(VI) pre-edge peak as shown in Figure 2. The peak height of the Cr(VI) standard has been set to unity. The spectrum of the Cr(III) standard was assumed to have no Cr(VI). The detection limit for Cr(VI) with XANES in this case is at 2 3%, with an absolute error of (2%. Additionally, a linear combination of the standard spectra for Cr(III) and Cr(VI) has been fitted in order to reproduce the pre-edge peak shape of the samples. Such an approach is equivalent to evaluating peak area ratios. The fit range was set to include the pre-edge peak region only. Both evaluations were done only for the BOS ashes (Figure 3), since both the original oil shale and the POS ashes did not reveal any Cr(VI) feature in the XANES region (Figure 1), being very similar to the spectral shape of the Cr(III) standard. Thus, the relative Cr(VI) content in the samples amounts to 32% for BOS ash treated thermally at 700 °C, 57% for the BOS ash sample at 850 °C, and 58% for BOS ash heated at 1000 °C, respectively, with an absolute uncertainty for all these values of (3%. The results suggest therefore that no further increase of the Cr(VI) content in the BOS ash is present upon combustion at temperatures above 850 °C which is anyway above that common for CFB plants. The Mn content in the ElLajjoun shale hampered further Cr K edge EXAFS spectra analysis to elucidate its speciation in more detail. In the POS ashes, however, no Cr(VI) content could be detected which suggests that during anaerobic pyrolysis there is no oxidation of the Cr(III) inventory (Figure 1). With the CEN/TC 292 EN 12457 1 leaching tests of the original oil shale and all POS samples, no dissolved Cr have been found in any of the leachates above the LOD of 50 nM. However, in the alkaline (pH 12.4 ( 0.2) leachates of the BOS samples, signiﬁcant dissolved Cr concentrations have been detected by ICP-OES with concentrations in the range of 1 mmol L 1, much higher than what is known from previous BOS ash leaching studies.4,8 9802

dx.doi.org/10.1021/es200695e |Environ. Sci. Technol. 2011, 45, 9799–9805

Environmental Science & Technology

ARTICLE

Figure 3. Linear combination ﬁts of the standard spectra for Cr(III) and Cr(VI) reproducing the pre-edge peak shape of the BOS samples. Fit evaluation was limited to the window set by the vertical dashed lines because the white lines cannot be reproduced due to obvious diﬀerences in Cr(VI) phases of standards and samples.

Interestingly, the Cr released is not only in form of Cr(VI) (85%), but also in form of Cr(III) (15%) as detected by CAdSV (Table 1). However, there is some mismatch between total dissolved Cr concentrations measured by ICP-OES and those measured by CAdSV (the latter are systematically up to 25% lower), probably due to analytical interferences we could not resolve. 3.3. As Speciation. The As K-edge XANES spectrum of the oil shale sample is a mixture of both As(III) and As(V) features (Figure 4). The electronic configuration of As(V) is [Ar]3d104s04p0, and because of the empty 4p shell, it exhibits a strong 1s 4p transition at the absorption edge. Reduction to As(III) fills only the 4s subshell and, hence, does not affect the edge intensity. A lower edge intensity at the As(III) energy position therefore suggests in the original oil shale a higher proportion of the arsenic in the pentavalent oxidation state, probably in an organic binding form as already suggested earlier for Green River oil shale.13 In order to obtain quantitatively the ratio of both arsenic species, a linear combination of the two As(III) and As(V) standard spectra has been fitted to the XANES spectrum of the oil shale sample. The fit has been performed for the whole near-edge region stretching from 20 eV below the K-edge to 30 eV above. To compensate for possible normalization related errors, the fit has been done for the first derivative of the absorption function. The relative amount of As(III) in the oil shale sample is 64 ( 3%. Unlike the case of chromium, upon combustion at all temperatures (700 1000 °C) only the oxidized species As(V) was found to contribute to the XANES features of the BOS samples. This suggests that As in the host minerals was oxidized more readily and at lower temperatures than Cr which appears reasonable given the quite different standard redox potentials of the As(III/V) and Cr(III/VI) couples. Oxidation of the sulfidic As(III) host mineral during combustion may have led to the increase in As(V) as already known from bituminous coal combustion.23 However, the low arsenic concentration in the El-Lajjoun shale (13 mg kg 1) hampered further EXAFS spectra analysis to elucidate the new As(V) host phase in the

Figure 4. Arsenic XANES spectra for untreated oil shale and the BOS ashes in comparison with the As(III) and As(V) standards. The absorption edge shift of the ashes is the same as for the As(V) standard. The oil shale spectrum is much closer to As(V) showing some contributions of As(III), that is, the doubled white line and the post white line hump. The respective As(III) and As(V) white line peaks are marked by the vertical dotted lines to aid the eye.

BOS ashes. For the CEN/TC 292 EN 12457-1 leaching tests of the original oil shale and all POS samples, no dissolved As have been found in any of the leachates above the LOD of 0.2 μmol L 1. However, in the alkaline leachates of the BOS samples, significant dissolved As concentrations above 1 μmol L 1 were found by ICP-OES measurements (Table 1).

4. DISCUSSION 4.1. Speciation Changes during Combustion. In the oil shale, Cr and As were found in the lower oxidation states of Cr(III) and As(III), probably related to periodic or even permanently euxinic depositional environment and the thus more reducing diagenetic conditions during oil shale formation.24 The As(III) inventory in the oil shale is possibly bound in pyrite, while incorporation of Cr 9803

dx.doi.org/10.1021/es200695e |Environ. Sci. Technol. 2011, 45, 9799–9805

Environmental Science & Technology in clay minerals is made possible by Cr(III)/Al(III) substitutions. By burning under aerobic conditions, these elements tend to be oxidized and form Cr(VI) and As(V) species which can no more be bound by their original mineral hosts. Both the clay and pyrite hosts decompose during combustion and react with the silica and lime to form secondary (amorphous) Ca-silicate and Ca-ferrite phases. The environmental impact of the spent oil shale products appears not to be relevant on basis of their overall low total As concentrations (<20 mg kg 1). However, the dissolved As concentrations released by the CEN/TC 292 EN 12457-1 leaching tests of the BOS samples were about 10 times higher than the WHO threshold value for drinking water (Table 1). Nonetheless, the impact of the mobilizable As may not become serious because of readsorption onto Fe(III) oxides once released into an environment of circumneutral pH conditions such as soil and aquifer sediments.25 Chromium is shown to be solely in trivalent form in the original oil shale and the retorted POS samples. The Cr(VI) found in the BOS ashes must, therefore, be a result of the aerobic combustion process. One probable reaction scheme involving excess free lime was proposed in similar studies of chromium behavior during cocombustion of coal and biomass and thermal treatment of municipal solid waste incineration ashes.12,26 In these studies formation of Ca-chromate was suggested by the oxidative synthesis reaction 2Cr2O3 + 4CaO + 3O2 w 4 CaCrO4. Kinetic control of the slow solid surface reactions, which requires a relatively long residence time, may be the reason why the amount of hexavalent chromium has not reached 100%. Clearly, it is questionable as to how far test cups in a muﬄe furnace with intensive contact between particles and the relatively high residence time reﬂect those of real CFB boilers. Various oxidative and reductive processes are occurring simultaneously in a CFB boiler dependent on the local oxygen content in the ﬂue gas and fuel. Highly oxidized conditions may alternate with microenvironments of low oxygen partial pressure where reduced species may be formed during combustion of the carbonaceous materials. Moreover, Cr(VI) is easily reduced by fugitive SO2-containing gases.11 This process may occur during combustion due to the relatively high sulfur content in the oil shale, and would lead to formation of Cr(III) sulfate, Cr2(SO4)3.27 Cr(III) sulfate is readily soluble and may be the reason for the unexpectedly high Cr(III) concentrations found in the BOS leachates (Table 1). 4.2. Environmental Implications. From X-ray diffraction analysis it is known that besides the kerogen-type organic matter, oil shale is characterized by calcite, quartz, apatite, gypsum, and kaolinite as the main mineralogical phases. The BOS ash, on the other hand, contains typically anhydrite, lime (“portlandite”), and minerals from the apatite and Ca-silicate groups.28 Due to this mineralogy, the BOS ash has a moderate pozzolanic activity and can be utilized in cement and concrete production albeit somewhat limited by the high phosphate concentrations.29 Formation of primary Ca Cr(VI) compounds like CaCrO4 with its relatively high solubility (logKsp = 2.27) may render the hexavalent chromium to be readily mobilized. However, the Cr(VI) concentrations found in the leachates are highly undersaturated with respect to this phase. Subsequent hydrolysis and formation of less soluble secondary Ca-sulfoaluminate minerals is known from cement chemistry.7 The Cr(VI) solubility of the thus formed Cr(VI)-ettringite and Cr(VI)-hydrocalumite phases strongly depends on the CrO4/SO4 ratio and is in the low mM range.30,31 In fact, formation of up to 30% ettringite has been found in Al-rich oil shale ashes upon postdepositional hydration

ARTICLE

and early diagenetic transformation.32 The Cr(VI) solubility from these phases with a low CrO4/SO4 ratio is in the order of the Cr(VI) concentrations found in the BOS leachates. Moreover, under ambient atmospheric conditions, Ca-sulfoaluminates are metastable and degrade gradually into the more stable minerals calcite, gypsum and gibbsite.32,33 None of these tertiary phases found in old BOS ash dumps are efficient hosts for chromate. Alternatively, formation of a Ba(S,Cr)O4 solid solution may well render the leachability of chromate from alkaline ashes below the μM range on the long-term.34 In fact, this solid solution has been found as a minor mineral constituent in a rock formation near Maqarin, Jordan, and was named “hashemite” after the Hashemite Kingdom of Jordan.35 This rock formation widely known as the “Natural Maqarin Analogue” unit for a cementitious radioactive nuclide repository is characterized by a natural combustion event of bituminous limestone which occurred in geologic history.36 The mineral composition of the thus produced cement-marbles unit suggests that the temperature reached during this event was up to 1050 1100 °C. It may therefore be considered also as a natural analogue for the study of chromium behavior in BOS ash deposits. Chromate concentrations in the hyperalkaline spring waters (pH ≈ 12.5) draining from this rock formation are in the order of 0.1 μM.37 Similar Cr concentrations have been found in drainage waters of another such anomalous rock formation in Central Jordan.38 Unfortunately, the Ba content in the El-Lajjoun BOS ash is depleted (66 mg kg 1) in comparison to average shale (580 mg kg 1).20 Hence, the overall low concentrations of this element may not really help to immobilize the much higher Cr(VI) inventory. The leachate Cr(VI) concentrations and the above-mentioned concentration levels for solubility of potential Cr(IV) mineral hosts are orders of magnitude higher than the limit for chromate in tap water proposed recently as public health goal by California’s OEHHA (20 ng L 1).39 Clearly, environmental risk assessment and stabilization measures for BOS dumps, once established, should target primarily on the Cr(VI) inventory, which may necessitate leachate monitoring, collection, and eventually, treatment in the long term. Speciation and leaching data of chromium in fresh BOS ashes derived from a muﬄe furnace may represent a worst case scenario but does help to assess the initial conditions in BOS deposits. Due to the high reactivity and, hence, secondary phase formation in BOS ashes, further experiments with (ﬁrst) hydrated and (subsequently) carbonated ashes are warranted to elucidate the type and stability of the secondary host minerals potentially controlling the mobility of the hexavalent chromium in the deposits.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT T.E.-H. acknowledges the German Science Foundation (DFG), the Jordanian Higher Council for Science and Technology (HCST), and the Scientiﬁc Research Deanship of Mutah University (Jordan) for facilitating this cooperation by fellowships, and M.K. for writing the manuscript. The HelmholtzZentrum Berlin electron storage ring BESSY II provided synchrotron radiation at the BAMline. Omar Al-Ayed provided the POS samples. Nora Groschopf assisted in the bulk 9804

dx.doi.org/10.1021/es200695e |Environ. Sci. Technol. 2011, 45, 9799–9805

Environmental Science & Technology XRF measurements, Michael Maus in ICP-OES measurements, and Svetlana Karabacheva in voltammetry.

’ REFERENCES (1) Speight, J. G. Chapter 6—Hydrocarbons from oil shale. In Handbook of Industrial Processes; Elsevier: Amsterdam, 2011; pp 203 239. (2) Hamarneh, Y.; Alali, J.; Sawaqed, S. Oil Shale Resources Development in Jordan; Natural Resources Authority: Amman, Jordan, 2006; http://www.nra.gov.jo/images/stories/pdf_ﬁles/Updated_Report_2006. pdf (accessed February 18, 2011). (3) Knaus, E.; Killen, J.; Biglarbigi, K.; Crawford, P. An overview of oil shale resources. In Oil Shale: A Solution to the Liquid Fuel Dilemma, ACS Symposium Series; Ogunsola, O. I., Hartstein, A. M., Ogunsola, O., Eds.; American Chemical Society: Washington, DC, 2010; Vol. 1032, pp 3 20. (4) Al-Harahsheh, A.; Al-Adamat, R.; Al-Farajat, M. Potential impacts on surface water quality from the utilization of oil shale at Lajjoun Area/Southern Jordan using geographic information systems and leachability tests. Energy Sources, Part A 2010, 32, 1763–1776. (5) Jaber, J. O.; Probert, S. D.; Tahat, M. A. Predicted environmental and social impacts of the proposed oil shale integrated tri-generation system (OSITGS). Oil Shale 1999, 16, 2–29. (6) Adamson, J.; Irha, N.; Adamson, K.; Steinnes, E.; Kirso, U. Eﬀect of oil shale ash application on leaching behaviour of arable soils: An experimental study. Oil Shale 2010, 27, 250–257. (7) Cornelis, G.; Johnson, C. A.; Van Gerven, T.; Vandecasteele, C. Leaching mechanisms of oxyanionic metalloid and metal species in alkaline solid wastes: A review. Appl. Geochem. 2008, 23, 955–976. (8) Shirav, M.; Zimmels, Y. The pathway of trace elements during oil shale combustion—A clue to their availability for leaching processes. Environ. Geol. Water Sci. 1988, 11, 55–64. (9) Manceau, A.; Marcus, M. A.; Tamura, N. Quantitative speciation of heavy metals in soils and sediments by synchrotron X-ray techniques. In Applications of Synchrotron Radiation in Low-Temperature Geochemistry and Environmental Science, Reviews in Mineralogy & Geochemistry 49; Fenter, P. A., Rivers, M. L., Sturchio, N. C., Sutton, S. R., Eds.; Mineralogical Society of America: Washington, DC, 2002; pp 341 428. (10) G€orner, W.; Eichelbaum, M.; Matschat, R.; Rademann, K.; Radtke, M.; Reinholz, U.; Riesemeier, H. Non-destructive investigation of composition, chemical properties and structure of materials by synchrotron radiation. Insight 2006, 48, 540–544. (11) Huggins, F. E.; Najih, M.; Huﬀman, G. P. Direct speciation of chromium in coal combustion by-products by X-ray absorption ﬁnestructure spectroscopy. Fuel 1999, 78, 233–242. (12) Stam, A. F.; Meij, R.; Te-Winkel, H.; Van-Eijk, R. J.; Huggins, F. E.; Brem, G. Chromium speciation in coal and biomass co-combustion products. Environ. Sci. Technol. 2011, 45, 2450–2456. (13) Cramer, S. P.; Siskin, M.; Brown, L. D.; Georget, G. N. Characterization of arsenic in oil shale and oil shale derivatives by X-ray absorption spectroscopy. Energy Fuels 1988, 2, 175–180. (14) Bender, F. Geology of Jordan; Borntraeger: Berlin, 1974. (15) El-Hasan, T. Geochemistry of the redox-sensitive trace elements and its implication on the mode of formation of the Upper Cretaceous oil shale, Central Jordan. Neues Jahrb. Geol. Palaeontol., Abh. 2008, 249, 333–344. (16) Bobrowski, A.; Krolicka, A.; Zare) bski, J. Characteristics of voltammetric determination and speciation of chromium—A review. Electroanalysis 2009, 21, 1449–1458. (17) Grabarczyk, M. A catalytic adsorptive stripping voltammetric procedure for trace determination of Cr(VI) in natural samples containing high concentrations of humic substances. Anal. Bioanal. Chem. 2008, 390, 979–986. (18) Radtke, M.; Vincze, L.; G€orner, W. Quantiﬁcation of energy dispersive SRXRF for the certiﬁcation of reference materials at BAMline. J. Anal. At. Spectrosc. 2010, 25, 631–634. (19) Newville, M. IFFEFIT: Interactive XAFS analysis and FEFF ﬁtting. J. Synchrotron Radiat. 2001, 8, 322–324. (20) Turekian, K. K.; Wedepohl, K. H. Distribution of the elements in some major units of the earth’s crust. Bull. Geol. Soc. Am. 1961, 72, 175–192.

ARTICLE

(21) Schultz, R. B. Geochemical relationships of Late Paleozoic carbon-rich shales of the Midcontinent, USA: A compendium of results advocating changeable geochemical conditions. Chem. Geol. 2004, 206, 347–372. (22) Strub, E.; Plarre, R.; Radtke, M.; Reinholz, U.; Riesemeier, H.; Schoknecht, U.; Urban, K.; Jungel, P. Determination of Cr(VI) in wood specimen: A XANES study at the Cr K-edge. Nucl. Instrum. Methods Phys. Res., Sect. B 2008, 266, 2405–2407. (23) Kolker, A.; Huggins, F. E. Progressive oxidation of pyrite in ﬁve bituminous coal samples: An As XANES and 57Fe M€ ossbauer spectroscopic study. Appl. Geochem. 2007, 22, 778–787. (24) Alego, J. A.; Maynard, J. B. Trace element behavior and redox facies in core shales of upper Pennsylvanian Kansas-type cyclothem. Chem. Geol. 2004, 206, 289–318. (25) Smedley, P. L.; Kinniburgh, D. G. A review of the source, behavior and distribution of arsenic in natural waters. Appl. Geochem. 2002, 17, 517–568. (26) Kirk, D. W.; Chan, C. C. Y.; Marsh, H. Chromium behavior during thermal treatment of MSW ﬂy ash. J. Hazard. Mater. 2002, 90, 39–49. (27) Jiao, F.; Wijaya, N.; Zhang, L.; Ninomiya, Y.; Hocking, R. Synchrotron-based XANES speciation of chromium in the oxy-fuel ﬂy ash collected from lab-scale drop-tube furnace. Environ. Sci. Technol. 2011, 45, 6640–6646. (28) Sæther, O. M.; Banks, D.; Kirso, U.; Bityukova, L.; Sorlie, J. E. The chemistry and mineralogy of waste from retorting and combustion of oil shale. Geol. Soc. Spec. Publ. 2004, 236, 263–284. (29) Khedaywi, T.; Yeginobali, A.; Smadi, M.; Cabrera, J. Pozzolanic activity of Jordanian oil shale ash. Cem. Concr. Res. 1990, 20, 843–852. (30) Leisinger, S. M.; Lothenbach, B.; Le Saout, G.; K€agi, R.; Wehrli, B.; Johnson, C. A. Solid solutions between CrO4- and SO4-ettringite Ca6(Al(OH)6)2-[(CrO4)x(SO4)1‑x]3*26 H2O. Environ. Sci. Technol. 2010, 44, 8983–8988. (31) Perkins, R. B.; Palmer, C. D. Solubility of chromate hydrocalumite (3CaO 3 Al2O3 3 CaCrO4 3 nH2O) at 5 75 °C. Cem. Concr. Res. 2001, 31, 983–992. (32) Motlep, R.; Sild, T.; Puura, E.; Kirsim€ae, K. Composition, diagenetic transformation and alkalinity potential of oil shale ash sediments. J. Hazard. Mater. 2010, 184, 567–573. (33) Liira, M.; Kirsim€ae, K.; Kuusik, R.; Motlep, R. Transformation of calcareous oil-shale circulating ﬂuidized-bed combustion boiler ashes under wet conditions. Fuel 2009, 88, 712–718. (34) Kersten, M.; Schulz-Dobrick, B.; Lichtensteiger, T.; Johnson, C. A. Speciation of Cr in leachates of a MSWI bottom ash landﬁll. Environ. Sci. Technol. 1998, 32, 1398–1403. (35) P.L. Hauﬀ, P. L.; Foord, E. E.; Rosenblum, S.; Hakki, W. Hashemite, Ba(Cr,S)O4, a new mineral from Jordan. Am. Mineral. 1983, 68, 1223–1225. (36) Rose, J.; Crouzet, N.; Trotignon, L.; Susini, J.; Khoury, H.; Salameh, E.; Milodowski, A.; Mercier, F. Eﬀect of leaching on the crystallographic sites of trace metals associated with natural cements (site of Maqarin, Jordan): Case of Cr. J. Phys. IV Fr. 2003, 104, 447–450. (37) Clark, I. D.; Fritz, P.; Seidlitz, H. K.; Trimborn, P.; Milodowski, T. E.; Pearce, J. M.; Khoury, H. N. Recarbonation of metamorphosed marls, Jordan. Appl. Geochem. 1993, 8, 473–481. (38) Elie, M.; Techer, I.; Trotignon, L.; Khoury, H.; Salameh, E.; Vandamme, D.; Boulvais, P.; Fourcade, S. Cementation of kerogen-rich marls by alkaline ﬂuids released during weathering of thermally metamorphosed marly sediments. Part II: Organic matter evolution, magnetic susceptibility and metals (Ti, Cr, Fe) at the Khushaym Matruk natural analogue (Central Jordan). Appl. Geochem. 2007, 22, 1311–1328. (39) California Oﬃce of Environmental Health Hazard Assessment (OEHHA). Announcement of second public comment period draft technical support document on proposed public health goal for hexavalent chromium in drinking water. http://www.oehha.ca.gov/water/ phg/chrom123110.html (accessed September 15, 2011).

9805

dx.doi.org/10.1021/es200695e |Environ. Sci. Technol. 2011, 45, 9799–9805

ARTICLE pubs.acs.org/est

Steroidal Aromatic ‘Naphthenic Acids’ in Oil Sands Process-Affected Water: Structural Comparisons with Environmental Estrogens Steven J. Rowland,*,† Charles E. West,† David Jones,† Alan G. Scarlett,† Richard A. Frank,‡ and L. Mark Hewitt‡ †

Petroleum and Environmental Geochemistry Group, Biogeochemistry Research Centre, University of Plymouth, Drake Circus, Plymouth PL4 8AA, U.K. ‡ Aquatic Ecosystems Protection Research Division/Water Science & Technology Directorate, Environment Canada, 867 Lakeshore Road, Burlington, ON, Canada L7R 4A6

bS Supporting Information ABSTRACT: The large volumes, acute toxicity, estrogenicity, and antiandrogenicity of process-aﬀected waters accruing in tailings ponds from the operations of the Alberta oil sands industries pose a signiﬁcant task for environmental reclamation. Synchronous ﬂuorescence spectra (SFS) suggest that oil sands process-aﬀected water (OSPW) may contain aromatic carboxylic acids, which are among the potentially environmentally important toxicants, but no such acids have yet been identiﬁed, limiting interpretations of the results of estrogenicity and other assays. Here we show that multidimensional comprehensive gas chromatographymass spectrometry (GCxGC-MS) of methyl esters of acids in an OSPW sample produces mass spectra consistent with their assignment as C19 and C20 C-ring monoaromatic hydroxy steroid acids, D-ring opened hydroxy and nonhydroxy polyhydrophenanthroic acids with one aromatic and two alicyclic rings and A-ring opened steroidal keto acids. High resolution MS data support the assignment of several of the so-called ‘O3’ species. When fractions of distilled, esteriﬁed, OSPW acid-extractable organics were examined, the putative aromatics were mainly present in a high boiling fraction; when examined by argentation thin layer chromatography, some were present in a fraction with a retardation factor between that of the methyl esters of synthetic monoalicyclic and monoaromatic acids. Ultraviolet absorption spectra of these fractions indicated the presence of benzenoid moieties. SFS of model octahydro- and tetrahydrophenanthroic acids produced emissions at the characteristic excitation wavelengths observed in some OSPW extracts, consistent with the postulations from ultraviolet spectroscopy and mass spectrometry data. We suggest the acids originate from extensive biodegradation of C-ring monoaromatic steroid hydrocarbons and oﬀer a means of diﬀerentiating residues at diﬀerent biodegradation stages in tailings ponds. Structural similarities with estrone and estradiol imply that such compounds may account for some of the environmental estrogenic activity reported in OSPW acidextractable organics and naphthenic acids.

’ INTRODUCTION Widespread interest in reclaiming the large volumes of process-aﬀected water (OSPW) resulting from the surface mining of oil sands, particularly those contained within the industrial leases in northeastern Alberta, Canada, has catalyzed a spate of studies attempting to identify the toxic constituents13 and numerous others attempting to delimit the toxic eﬀects.46 Recently, we described the chromatographic resolution and mass spectral identiﬁcation of the methyl esters of some of the ﬁrst individual alicyclic acids in an acid-extractable OSPW mixture by multidimensional comprehensive gas chromatographymass spectrometry3,7 (GCxGC-MS). The toxicity of these individual trito pentacyclic diamondoid acids to the bacterium Vibrio ﬁscheri, measured in a screening assay and the modeled toxic eﬀects on the water ﬂea Daphnia magna, were those expected to result from nonspeciﬁc narcosis8 and exhibited a water solubility-controlled toxicity ‘cutoﬀ’ at higher carbon numbers. The toxicity of these r 2011 American Chemical Society

acids was suﬃcient to explain the narcotic toxicity of some OSPW fractions.9 Whether the same acids are responsible for more speciﬁc toxicological actions remains to be examined. Components other than the alicyclic acids so far identiﬁed in OSPW may be more toxicologically important.2,10,11 Numerous examinations of OSPW acid-extractable organic matter by electrospray ionization high and low resolution mass spectrometry, sometimes coupled with liquid chromatography,12,13 have shown that the acid-extractables of OSPW contain compounds with four or more double bond equivalents (DBE). These are designated by a z number of g 8 in the formula CnH2n+zO2 which is often used to describe the acids. While such acids undoubtedly include Received: July 27, 2011 Accepted: October 7, 2011 Revised: October 5, 2011 Published: October 20, 2011 9806

dx.doi.org/10.1021/es202606d | Environ. Sci. Technol. 2011, 45, 9806–9815

Environmental Science & Technology tetracyclic (z = 8) and pentacyclic (z = 10) compounds,7 no hexacyclic (z = 12) species have been identiﬁed to date, and an alternative explanation for some of the z g 8 (four or more double bond equivalents) species is that they include monoaromatic species (gfour DBE) with zero, one, or more alicyclic rings. Evidence for this also includes distinctive synchronous ﬂuorescence spectra (SFS) associated with OSPW and groundwater near a tailings pond.14 Characterization of these acids would begin to address important knowledge gaps about the toxicity and mobility of OSPW components and help to direct potential methods for remediating the toxicants. We now describe the chromatographic resolution and mass spectral identiﬁcation of some individual steroidal aromatic acids in the acid-extractables from an OSPW by GCxGC-MS. We provide supporting evidence for the identiﬁcations from argentation thin layer chromatography, ultraviolet (UV) and SFS, including for authentic monoaromatic acids and postulate a reasoned argument for the origins of the acids. The presence of such acids may help to explain some of the environmental estrogenic activity of OSPW and other naphthenic acid mixtures.

’ MATERIALS AND METHODS The OSPW acidic extract was obtained from a previous study where the organic acids were extracted from 3000 L of fresh OSPW collected from the discharge into West In-Pit settling basin at Syncrude Canada Ltd.15 Distilled fractions from the same organic acid extract were also generated from a previous study.9 In the present study, acids were derivatized by reﬂuxing with BF3-methanol or reaction with diazomethane (distilled fractions9). Authentic aromatic acids used were dehydroabietic acid (DHA) from Helix Biotech (Vancouver, BC, Canada) and 1,2,3, 4-tetrahydrophenanthrene-1,2-dicarboxylic acid dimethyl ester (TPDADE) and phenanthrene-4-carboxylic acid (PCA; both from Sigma-Aldrich (Oakville, ON, Canada). Methyl esters of cyclohexyl-3-propanoic and phenyl-3-propanoic acid used as TLC references were from previous studies.16 Argentation thin layer chromatography was conducted on the methylated OSPW extract with silica gel stationary phase containing 5% by weight silver nitrate to generate three TLC fractions. Synchronous ﬂuorescence spectroscopy was performed as described previously14 using a Perkin-Elmer Luminescence spectrometer LS50B paired with FL Winlab 3 software (PerkinElmer, Norwalk, CT) for data collection. In brief, samples were scanned in a 1 cm quartz cuvette with a PTFE stopper (Hellman, Concord, ON, Canada) at 20 °C. SFS were collected at a Δλ of 18 nm in the 250400 nm excitation wavelength range. The scan speed was 50 nm min1 at a resolution of 0.5 nm, and the excitation and emission monchromator slit widths were set at 5 nm. The spectrum of each sample was blank-corrected with HPLC water (Caledon Laboratory Chemicals, Georgetown, ON, Canada) and then smoothed with a 5-point averaging adjacent method using OriginPro software ver. 8.01 (OriginLab Corp., Northampton, MA). Two-dimensional comprehensive gas chromatographymass spectrometry (GCxGC-MS) analyses were conducted as described previously.3 Brieﬂy, an Agilent 7890A gas chromatograph (Agilent Technologies, Wilmington, DE) ﬁtted with a Zoex ZX2 GCxGC cryogenic modulator (Houston, TX, USA) interfaced with an Almsco BenchTOFdx time-of-ﬂight mass spectrometer

ARTICLE

(Almsco International, Llantrisant, Wales, UK) was operated in positive ion electron ionization mode and calibrated with perﬂuorotributylamine. For GC-MS, extracts were examined on an Agilent GC-MSD (Agilent Technologies, Wilmington, DE, USA). This comprised a 7890A gas chromatograph ﬁtted with a 7683B Series autosampler and a 5975A quadrupole mass selective detector. The column was a HP-5MS fused silica capillary column (30 m 0.25 mm internal diameter 0.25 μm ﬁlm thickness). The carrier gas was helium at a constant ﬂow of 1.0 mL min1. A 1.0 μL sample was injected into a 300 °C splitless injector. The oven temperature was programmed from 40 to 300 at 10 °C min1 and held for 10 min. High resolution MS accurate mass measurements were made using a Thermoﬁsher LTQ Orbitrap XL high resolution mass spectrometer. The mass range was m/z 1202000; mass accuracy <3 ppm rms with external calibration. For negative electrospray ionization the instrument was externally calibrated using sodium dodecyl sulfate and sodium taurocholate. For loopinjections a Thermo Scientiﬁc Surveyor MicroLC was used to provide solvent ﬂow at 20 μL min1 through a 2 μL sample loop. Solvents used were H2O:MeOH (1:1). For nanoelectrospray an Advion Triversa NanoMate was used to deliver samples diluted into MeOH ( 10% NH4OAc or NaOH at a ﬂow of approximately 0.25 μL min1. For UV spectra, solutions in dichloromethane were examined on an Agilent/Hewlett-Packard model 8453 (Agilent Technologies, Waldbronn, Germany), wavelength range 1901100 nm, slit width 1 nm. ADMET predictor software (SimulationsPlus Inc., Lancaster, CA; www.simulations-plus.com) was used to calculate possible human estrogenic and androgenic receptor activities.

’ RESULTS AND DISCUSSION When examined by GC-MS, the methyl esters of the OSPW were almost entirely unresolved and appeared as a bimodal ‘hump’ (nodes ca. 16 and ca. 19 min retention times); no useful spectra could be obtained (Figure 1A). One late-eluting resolved peak was observed (retention time 22 min; Figure 1 A), but even for this the mass spectrum suﬀered from interfering ions from coeluting compounds. When examined by GCxGCMS, however, the derivatized OSPW extract produced a complex chromatogram with well-resolved components eluting under these diﬀerent conditions from about 65125 min (GC1) and 15 s (GC2). The higher retention region (>∼100 min GC1) contained several very well resolved components with clear mass spectra and molecular weights of around 300 Da and apparently six (z = 12) to eight (Z = 16) DBE. One group of well-resolved compounds was highlighted by selected ion mass chromatography of the m/z 145 base peak ion (Figure 1B), others by m/z 237, 223 and by high mass (>300 Da) ions. When a distilled fraction9 of the methyl esters of the same OSPW extract boiling at 220 °C was examined by GC-MS, several resolved components were observed by selected ion mass chromatography (e.g., m/z 145; Figure 1C), and when examined by GCxGC-MS the same compounds were once again very well resolved. Previously, low resolution electrospray ionization mass spectrometry indicated that the major acids in this fraction possessed six DBE (z = 12) with a median mass of 288 (M+) and carbon numbers of ca. 1520.9 9807

dx.doi.org/10.1021/es202606d |Environ. Sci. Technol. 2011, 45, 9806–9815

Environmental Science & Technology

ARTICLE

Figure 1. A. Gas chromatographymass spectrometry total ion current chromatogram of OSPW acid extractable fraction (methyl esters). Bimodal nodes in the unresolved proﬁle occur at ca. 16 and 19 min retention time. One reasonably well resolved peak at 22 min. B. Comprehensive multidimensional gas chromatographymass spectrometry (GCxGC-MS) selected ion mass chromatogram (m/z 145) of OSPW acid extractable fraction (methyl esters) illustrating high chromatographic resolution of putative aromatic acids. C. Gas chromatographymass spectrometry selected ion mass chromatogram (m/z 145) of distilled OSPW acid extractable fraction (methyl esters) boiling at 220 °C, illustrating high chromatographic resolution of putative aromatic acids. D. Gas chromatographymass spectrometry total ion current chromatogram of OSPW acid extractable argentation thin layer chromatography fraction (Rf 0.740.82) fraction (methyl esters). Unimodal node in the unresolved proﬁle occurs at ca. 16 min retention time. Several reasonably well resolved peak at 2223 min. E. Gas chromatographymass spectrometry total ion current chromatogram of OSPW acid extractable argentation thin layer chromatography fraction (Rf 0.720.74) fraction (methyl esters). A unimodal node in the unresolved proﬁle occurs at ca. 19 min retention time. Several reasonably well resolved peaks occur at 2223 min. F. Gas chromatographymass spectrometry total ion current chromatogram of OSPW acid extractable argentation thin layer chromatography fraction (Rf 0.660.72) fraction (methyl esters). A unimodal node in the unresolved proﬁle occurs at ca. 22 min retention time. The well resolved peak at 2223 min is due to bis-ethylhexylphthalate, possibly arising from contamination during sampling or analysis.

When we examined the OSPW methyl esters by argentation thin layer chromatography, we isolated three fractions with retardation factors (Rf) of 0.740.82 (Fraction 1), 0.720.74 (Fraction 2), and 0.660.72 (Fraction 3). Under these conditions, methyl cyclohexyl-3-propanoate had a Rf of 0.760.83 and methyl phenyl-3-propanoate an Rf of 0.640.72. When examined by GC-MS, these TLC fractions comprised three separate unresolved ‘humps’; fraction 1 maximizing at ca. 16 min, fraction 2 at ca. 19 min, and fraction 3 at ca. 22 min retention time (Figure 1 D-F). GC-MS selected ion mass chromatography (e.g., m/z 145) of the three TLC fractions revealed that the components with apparent molecular weights of around 300 Da in the OSPW were in fractions 1 and 2 only, each with slightly diﬀerent distributions. The observations that fraction 2 and 3 components were more retained and that fractionation of the ‘hump’ of esters of OSPW into three separate nodes occurred by argentation TLC (Figure 1 D-F) suggest there was some aromatic (or double bond) character in the more retained acids. Examination of the unfractionated and TLC fractions of OSPW methyl esters by ultraviolet (UV) absorption spectroscopy revealed spectra with absorption maxima typical of benzenoid moieties at ∼265 nm and ∼295 nm (unfractionated, fractions 1 and 2) and ∼290 nm only (fraction 3). Steroidal Hydroxy Acids. Figure 2A and B show positive ion electron ionization mass spectra obtained by GCxGC-MS of the methyl esters of two apparently isomeric acids (I and II) in the unfractionated OSPW methyl esters, eluting about 16 s apart in

GC1 (see also, Table of Contents artwork). The apparent molecular ion m/z 310 is consistent with the methyl ester of either a monocyclic C19 acid or a C20 acid with eight DBE (z = 16). Previous high resolution mass spectrometry studies have shown that monocyclic acids are not abundant in OSPW,12,17 so we suggest the compounds are not monocyclic but are polycyclic C20 acids. Hexa- to octacyclic acids are unlikely since such hydrocarbons have rarely been reported in the oil sands bitumen or petroleum from which the oil sands acids are thought to originate.18 Rather, we favor the suggestion that the C20 acids comprise both aromatic (4 DBE) and alicyclic (1 DBE per ring) moieties. Other features of the spectra (Figure 2A and B) are the base peak ion at m/z 237, and abundant ions at m/z 221 and m/z 286. The latter ion suggests loss of 24 Da from the apparent molecular ion, which would be very unusual. These mass spectra (Figure 2A and B) show considerable similarities to the mass spectrum of 20 -[3β-hydroxy-5,7,9-estratrien-17-yl]propanoic acid (III; Figure 2C: reproduced with permission from NIST). The latter is also characterized by an apparent molecular ion at m/z 310, whereas the molecular weight is actually 328 Da. The m/z 310 ion is probably due to loss of water from the molecular ion. The intensity of the ion is approximately the same as that observed in the spectra shown in Figure 2A and B. Second, the spectrum of III is also characterized by a base peak ion at m/z 237 (Figure 2C) as shown by the spectra of the unknowns. The mass:charge ratios of the ions <m/z 237 are also quite similar to those of the unknowns, although the ion abundances are diﬀerent. 9808

dx.doi.org/10.1021/es202606d |Environ. Sci. Technol. 2011, 45, 9806–9815

Environmental Science & Technology

ARTICLE

Figure 2. A. Mass spectrum of methyl ester of putative C20 monoaromatic C-ring steroidal acid (I). B. Mass spectrum of methyl ester of putative C20 monoaromatic C-ring steroidal acid (II; stereoisomer of I). C. Mass spectrum of 20 -[3β-hydroxy-5,7,9-estratrien-17-yl]propanoic acid. Reproduced with permission from the National Institute of Standards & Technology mass spectral database ver. 2.0f Oct8 2008. D. Mass spectrum of methyl ester of putative C19 monoaromatic C-ring steroidal acid (IV).

These comparisons suggest to us that the unknowns I and II are in fact isomeric steroidal C20 hydroxy acids, with M+. 328 absent with seven DBE (z = 14). Thus we interpret the m/z 310 ion as due to M+.-H2O (viz. loss of 18 Da from m/z 328) and 286 ion (Figure 2A and B) as due to loss of propene from m/z 328. Formation of the m/z 237 ion in the spectrum of

III is best explained by fragmentation of the C1720 bond (Figure 2C) from ion m/z 310. In the unknowns I and II, which are methyl esters, this is possible if the side chain is an ethanoate group. This is also consistent with the ethanoate side chains observed in many of the co-occurring alicyclic acids.3,7 9809

dx.doi.org/10.1021/es202606d |Environ. Sci. Technol. 2011, 45, 9806–9815

Environmental Science & Technology Since we did not observe a molecular ion (m/z 328) under electron impact conditions, we also examined the nonmethylated free acids in the OSPW sample by electrospray ionization with a higher resolving power mass spectrometer (Orbitrap). This less energetic ionization method would be expected to result in the formation of molecular ions, rather than ions due to loss of water. The expected molecular weight of the underivatized C20 hydroxy acid with seven DBE would nominally be 314. Under negative ion electrospray conditions, this would be expected to appear as a deprotonated ion, m/z 313 ([M-H]). This ion was indeed observed, and the accurate mass was measured as 313.1795, which compares favorably to a theoretical value of 313.1809. This is a mass accuracy of <5 ppm, comparable to that achieved previously for OSPW17 and supports our contention that the unknown is a C20 hydroxy acid with seven DBE. Certainly, the possibility that the m/z 313 ion is due to the nominally isobaric C21H29O2 ion is ruled out (theoretical accurate mass 313.2168), although C17H29O3S would admittedly also have a similar mass to the unknown (313.1843). Also present in the OSPW extract, eluting a few minutes earlier than the isomers I and II above, was a component (IV) for which the mass spectral ions were 14 Da less. Thus, an apparent molecular ion, but again assigned to M+.-H2O, was observed at m/z 296 (Figure 2D) and ions at m/z 272, 223 (B+) assigned as above, due to losses of propene from m/z 310 and a methyl ethanoate moiety from m/z 296. We therefore conclude that unknown IV is a further demethylated analogue of the above isomers I and II, but apparently occurs as only one isomer, or is unresolved from other isomers. From the above data and interpretations we assign the three unknowns to C-10 demethylated (18-nor) C-ring monoaromatic C19 (Figure 2D) and C20 (Figure 2A and B) hydroxy steroid acids (IV, I, and II) and suggest they result from biotransformation of, for example, C21, 22 and C2729 C-ring steroidal hydrocarbons in the original petroleum from which the oil sands derive. As such these compounds appear to be the ﬁrst acids containing three oxygen atoms (so-called O3 species) to be tentatively identiﬁed in OSPW. B-ring (such as III) and A-ring monoaromatic steroids are uncommon and not abundant in petroleum, whereas C-ring monoaromatic steroids are widespread, including in oils from Alberta.19 The UV absorption spectra of the unfractionated and argentation ion TLC-fractionated acids (methyl esters) show evidence of benzenoid B-band ﬁne structure and compare quite well with those published for the methyl ester of III (280 nm EtOH),20 the methyl ester of a C24 analogue of III (278, 287 nm EtOH),21 and the acetoxy methyl ester of V (220, 225, 261, 268, 275 nm cyclohexane).22 Moreover, mass spectra (methyl esters) of A-ring hydroxy steroidal acids have also been published21,22 (e.g., VI; C23) and are dissimilar to those of the unknowns, whereas the mass spectrum of the methyl ester of a C24 C-ring acid22 (V) was characterized by ions due to M+., loss of water, loss of methoxy, and loss of side chains (with and without loss of water); many of the features observed for the mass spectra of the methyl esters of the unknowns I, II, and IV herein. The occurrence of isomers of the unknowns I and II is also consistent with this, since 5α and 5β isomers of C-ring monoaromatic steroid hydrocarbons in petroleum and C-ring acids are known19,22 and isomerization about C-17 or other positions is also possible.22 A notable diﬀerence between the spectra of the unknowns and that of III was the absence of the m/z 295 ion in the former, suggesting the absence of the C-13 methyl substituent in the unknowns. The presence of the

ARTICLE

suggested C-17 methyl substituent (e.g., I and II) would not require the presence of a strong m/z 295 ion since fragmentation in I and II is expected to be dominated by benzylic loss of the methyl ethanoate substituent from m/z 310 to produce m/z 237 (Figure 2A and B), rather than loss of the C-17 methyl substituent. This was also the case in the mass spectrum of the methyl ester22 of V. In IV, which is a demethylated analogue of I and II, we suggest the C-17 methyl group has already been lost due to microbial action. A further diﬀerence between IIII and the C-ring steroidal hydrocarbons known in petroleum is the absence of the C-10 methyl substituents in IIII. Biotransformation of steroids by processes involving microbial C1720 side chain oxidation and hydroxylation of the steroid nucleus, by both aerobic and anaerobic processes, is wellknown2326 and has been studied in order to optimize production of steroidal pharmaceuticals. Indeed “almost all positions in a steroid molecule have been hydroxylated by microbial strains”.24 We show one example position (C-4) for the hydroxyl group (I, II, and IV), which is also consistent with the possible loss of propene as a C-1 to C-3 moiety (Figure 2A, B, D). Biodegradation of the nonaromatic steranes of petroleum has also been shown to proceed via oxidation of the side chain and hydroxylation at C-4, thereby further supporting our assignment. Thus, Paulus27 incubated 5α-cholestane with a mutant strain of Nocardia sp. for 8 days and identiﬁed 4α-hydroxy-5α-pregnane20-oic acid as a metabolite by comparison of the mass spectrum of the acetoxy, methyl ester derivative with that of a synthesized sample.The biodegradation of aromatic steroids in petroleum, including oil sands bitumen, is also well studied19,28 and the relative orders of biodegradation established. Thus, C-ring monoaromatic steroids are more resistant than triaromatic steroids and within the former, the C2729 steroids are degraded before C2021 members. The latter are conventionally considered among the most resistant to microbial degradation.28 However, we have suggested previously that biodegradation of oil sands bitumen in the present sample has exceeded conventional scales, as evidenced by the oxidation of adamantane and diamantane hydrocarbons to the corresponding acids.3,7 Thus the postulation that the acids identiﬁed herein are degradation products of C2021 steroidal hydrocarbons is consistent with our earlier ﬁndings that OSPW represents a highly degraded residue in which conventional biodegradation scales may have been surpassed. We propose that the absence of the C-18 methyl substituent (at position C10) is also due to a demethylation known to occur during biotransformation of the hydrocarbons, akin to that observed at the analogous C-18 methyl substituent in the biotransformation of triterpenoids.27 Given the broad structural similarities of the proposed C-ring aromatic steroid acids I, II, and IV with known estrogens (although the latter are A-ring aromatics) it would be interesting to examine the estrogenic activities of I, II, and IV. These may help to explain the estrogenic activity of OSPW29,30 which has previously been unexplained owing to the complexity of the mixtures. GCxGC-MS has allowed us to unravel some of this complexity and to identify individual steroidal acids. Although acids related to estradiol are not estrogenic, many esters of the latter are potent estrogens.31 Indeed, when we modeled the possible human estrogenic and androgenic receptor activity of acid I (Figure 1) using ADMET predictor software, under conditions which correctly predicted the known estrogenic activity of nonylphenol, acid I was also predicted to aﬀect both receptors (Table S1). Experimental veriﬁcation of these predictions therefore now appears to be warranted. 9810

dx.doi.org/10.1021/es202606d |Environ. Sci. Technol. 2011, 45, 9806–9815

Environmental Science & Technology Hydroxy Octahydrophenanthroic Acids. Other components eluting in the same region of the GCxGC chromatogram produced quite different mass spectra from those shown in Figure 2 but also contained apparent molecular ions of ca. 300 Da. Thus, a component eluting at 120 min (GC1) produced a mass spectrum with an apparent molecular ion at m/z 330 and ion attributed to loss of water (M+.-18) at m/z 312 (Figure S1A, Supporting Information). The only other significant ion was at m/z 225 (B+) representing loss of 87 Da from M+.-18, suggesting additional loss of a methyl propanoate moiety. The molecular ion suggests the methyl ester of a C20 hydroxy acid with six DBE, which we assign to one aromatic and two alicyclic rings. A structure consistent with these features (Figure 2, VII: Figure S1A) is a hydroxy octahydrophenanthroic acid, conceivably originating from microbial opening of the D-ring of C-ring aromatic steroidal hydrocarbons, or of metabolites such as I, II, and IV. A component eluting some 30 s earlier (GC1) exhibited a mass spectrum containing a putative molecular ion at m/z 314, suggesting the methyl ester of a C20 acid with six DBE and base peak ion at m/z 199, consistent with loss of a methyl pentanoate moiety via benzylic cleavage. Structure VIII (Figures 2 and S1B) is consistent with these findings. Similarly, we identified a C19 hydroxy acid (IX) with six DBE from the mass spectral molecular ion (m/z 316) and M+.-18 (m/z 298), again with loss of a methyl propanoate side chain (Figures 2 and S1C; m/z 211), which we propose occurs via benzylic cleavage from M+.-18. Thus IX is a C10 demethylated homologue of VII. Further similar spectra corresponding to those of C19 and C18 pseudohomologues (e.g., Figure 2, X and Figure S1D) were also obtained. Unfortunately we could ﬁnd no published spectra of compounds similar to those we have postulated (Figure 2, VII-X and Figure S1). Numerous acids and hydroxy acids with polyhydrophenanthrene structures are known. These include carnosic (C20; dihydroxy), dehydroabietic (C20), podocarp-9(11),8(14),12-trien-15-oic (C17) and pisiferic acids (C20). Diaromatic analogues, such as the nonclassical estrogen, Z-bis-dehydrodoisynolic acid32 (C18) are also known. However, the spectra of none of these matched the spectra of the unknowns since the unknown acids were instead substituted, we propose, with alkanoate groups on the aromatic rings, prompting benzylic cleavages of the alkanoate groups. Additionally a number of polyhydrophenanthrene carboxylic acids were synthesized many years ago,33 before the advent of mass spectrometry. A number of derivatives of the latter showed estrogenic activity when intraperitoneal injections were made into mice,34 indicating that it might also be useful to synthesize and examine the estrogenic activities of unknowns VII-X. de-A Steroidal Keto Acids. The final group of about ten unknown components detected in the OSPW extract exhibited electron ionization mass spectra (Figure S2 A-G) characterized by apparent molecular ions (e.g., m/z 300, 314, 356) with about 520% abundance, confirmed by the slightly more abundant M+.-15 ions, indicative of methyl group substitution. The presence of M+.-32 ions in several cases, typically representing loss of methanol from methyl esters, appeared to confirm further the assignment of the molecular ions and the presence of a methylated carboxylic acid substituent. Other ions of significance in the spectra include m/z 171/2, 157, and 145 (B+). The compounds were present in the initial OSPW extract, argentation TLC fractions 1 and 2 and the 220 °C distillation fraction. The molecular ions suggest possibly methyl esters of C19, 20, 23 acids with several isomers, each with six DBE, suggesting (with the

ARTICLE

TLC behavior), one aromatic and two alicyclic rings, or of methyl esters of O3 species such as C18, 19 and C22 monoaromatic keto acids. To decide between these possibilities, we examined the accurate masses of the deprotonated free acids of the most abundant acids (viz. m/z 299, [M-H]). The accurate mass (299.1642) was within 1.7 ppm mass accuracy of that expected for the O3 species, comparable to accuracy achieved previously for OSPW17 and well outside that expected for the O2 species. The fragment ions observed in the electron ionization spectra (Figure S2 A-G) are present in the spectrum of a synthetic C23 B-ring monoaromatic de-A-steroid hydrocarbon35 and analogues in sediments36 but in diﬀerent relative abundances. The m/z 157 ion has been assigned previously to the hydroxybenzotropylium ion in the fragmentation of hydroxy A-ring aromatic steroids37 including methyl esters of steroidal acids.20 The m/z 145 ion has also been assigned previously, not only in the spectra of 14β(H) C23 B-ring monoaromatic de-A-steroid hydrocarbons35 where the ion presumably derives from D-ring fragmentation but also in the spectrum of androst-4,9-diene-17β-ol-3-one, by fragmentation of the B-ring to form a C2-substituted triunsaturated tetrahydroindane-type ion38 (C11H13+). In the mass spectrum of the methyl esters of the partial hydrogenation products of naphthyl-2-ethanoic acid (data not shown), the acid assigned as tetralin-2-ethanoic acid displayed a base peak ion at m/z 145, assigned to a C1-tetralin moiety (C11H13+). A corresponding keto ion (C10H9O+) would presumably be similar. Thus, we tentatively suggest that the unknown compounds (Figure S2 A-G) are keto acids resulting from A-ring-opening of the C-10 demethylated (18-nor) C-ring aromatic steroids, or more likely, of the A-ring hydroxylated metabolites of the latter (Figure 3). Biotransformation of some steroidal acids is known to proceed via A-ring and B-ring cleavage.2326 For example, the A-ring degraded cholestenone analogue known as Windaus’ keto acid (A-nor-3,5-secocholestan-5-on-3-oic acid39) was shown to be produced by degradation of cholestenone by Nocardia sp. and Proactinomycetes erythropolois.39 Oxidation of the A-ring by 3,4 cleavage is also known to facilitate the formation of the corresponding acids in hopanoids40 and has been speculated to occur in steroids also.3,36 Relationships between the classes of aromatic naphthenic acids we have tentatively identiﬁed herein and the C ring steroidal hydrocarbons, which are wellknown in petroleum, are summarized in Figure 3. (An order-ofmagnitude estimate of the concentrations of the acids obtained by simple, uncalibrated, summation of the GCxGC-MS peak areas of the peaks in the m/z 145 + 314 selected ion mass chromatogram, suggested these acids comprised about 0.7% of the total GC-detectable acids. However, until suitable compounds have been synthesized for calibration, this must remain a ‘ballpark’ estimate.) SFS of the acid extractable fraction of OSPW and of mono(DHA) and diaromatic (TPDADE) acids (the latter as the dimethyl ester) are shown in Figure 4. The three maxima observed for the OSPW extract (at 282, 320, and 333 nm) are consistent with those noted previously for OSPW from diﬀerent containments and a groundwater sample from near a tailings pond.14 The maximum at 282 nm is also observed in the spectrum of DHA, consistent with the suggestions herein from UV and MS data, of the presence of benzenoid steroidal acids in OSPW. The peak in the spectrum of the OSPW at 320 nm (Figure 4) coincides with that of TPDADE, suggesting perhaps the presence of diaromatic acids. GC-MS data for TLC fraction 3 were also consistent with the presence of diaromatic steroids 9811

dx.doi.org/10.1021/es202606d |Environ. Sci. Technol. 2011, 45, 9806–9815

Environmental Science & Technology

ARTICLE

Figure 3. Putative origin of products of oxidation of monoaromatic C-ring steroidal hydrocarbons. Each arrow encompasses multiple steps: a, C10 demethylation, A ring hydroxylation and side chain oxidation; b, C10, 17 demethylation, A ring hydroxylation and side chain oxidation; c, C10 demethylation; d, C10,17 demethylation, A ring hydroxylation and D ring-opening; e, C10 demethylation and D ring-opening; f, C10, 17 demethylation and A ring-opening; g, A ring hydroxylation and D ring-opening.

(data not shown).41 No evidence could be found for the presence of triaromatic acids such as PCA (341, 353 nm) from the SFS or mass spectra. Environmental Significance. Unlike in our previous studies,3,7,16 to date we have been unable to access samples of authentic acids for GCxGC retention time comparison with the acids we have tentatively identified herein, so the above suggestions must remain somewhat speculative until representative acids have

been synthesized and characterized. However, previously published mass spectra of the methyl esters of similar A-, B-, and C-ring aromatic steroidal hydroxy acids2126 support our contentions (e.g., Figure 2C). The aromatic steroidal structures, which are somewhat similar to those of known estrogens, and the known estrogenicity and antiandrogenicity of OSPW29,30 and modeled ER and AR activities (Table S1) suggest the estrogenicity of the acids should now be measured. It was reported recently30 that in contrast to 9812

dx.doi.org/10.1021/es202606d |Environ. Sci. Technol. 2011, 45, 9806–9815

Environmental Science & Technology

ARTICLE

Figure 4. Synchronous ﬂuorescence spectra of an acid extractable fraction (NAE) of OSPW, of dehydroabietic acid (DHA), and of 1,2,3,4tetrahydrophenanthrene 1,2-dicarboxylic acid, dimethyl ester (TPDADE).

OSPW acids, a commercial naphthenic acids mixture refined from petroleum produced only weak estrogenicity. As we have noted previously, this is consistent with the presence of small amounts of estrogenic alkylphenols and other nonacids in commercial mixtures 42 and with the very different individual acids in the commercial and OSPW samples.3,7,16 The steroidal acids we have identified herein in the OSPW, which is much more estrogenic,30 were absent from the commercial mixture when examined by GCxGC-MS. This is understandable since the degree of biodegradation (and hence production of steroidal acids) would be expected to be less in a slightly biodegraded petroleum such as that from which the commercial acids are obtained, compared to the highly degraded oil sands bitumen. Thus, our identification of the novel acids herein offers a method of distinguishing the degree of biodegradation of different highly biodegraded residues, including those found in oil sands tailings ponds of different ages and other heavy and extensively biodegraded crude oils. It was also noted recently30 that “if the NAs in OSPW are responsible for the estrogenic eﬀects and ozonation removes NAs, it is inconsistent that ozonation did not attenuate estrogenicity”. This is not necessarily the case: ozonation is dependent on structure. Ozonation of known tricyclic acids in OSPW, such as adamantane-1-carboxylic acid,3 would likely form very diﬀerent products (e.g., 3-hydroxy-adamantane-1-carboxylic acid, possibly followed by dehydration and ring-opening to more soluble and less toxic, polyacids),43,44 from those resulting from ozonation of the aromatic acids described herein. For example ozonation of I, II, and IV might initially result in preferential hydroxylation at the C12 and C13 positions of the aromatic ring. Such changes might not be expected to reduce estrogenicity. Prolonged or higher dose ozonation might lead to C ring fragmentation and formation of tricyclic acids, keto acids, and hydroxy acids, as observed for the A-ring in estradiol upon ozonation,45 which would likely reduce estrogenicity. Competing reactions whereby acids such as VII-X become hydroxylated in the aromatic ring might lead to production of further estrogens. Clearly, identiﬁcation of the individual chemicals in these mixtures is key to a better understanding of the toxicological and environmental eﬀects.

’ ASSOCIATED CONTENT

bS

Supporting Information. Figures of the mass spectra of putative D- and A-ring opened acids (methyl ester). This material is available free of charge via the Internet at http:// pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT Funding for this research was provided by an Advanced Investigators Grant (No. 228149) awarded to S.J.R. for project OUTREACH, by the European Research Council, to whom we are extremely grateful and by Environment Canada. We acknowledge the EPSRC National Mass Spectrometry Service Centre at Swansea University, UK for obtaining the accurate mass data. We thank Dr. M. Barrow (University of Warwick, UK) for advice on high resolution mass spectrometry and Prof. G. Wolﬀ (University of Liverpool, UK) for advice on the mass spectrometry of de-A steroids. We are grateful to Dr. C. Anthony Lewis (University of Plymouth) for constructive criticisms of an earlier draft manuscript. We thank Simulations Plus Inc., Lancaster, CA for use of a gratis copy of ADMET predictor software. ’ REFERENCES (1) Grewer, D. M.; Young, R. F.; Whittal, R. M.; Fedorak, P. M. Naphthenic acids and other acid-extractables in water samples from Alberta: What is being measured? Sci. Total Environ. 2010, 408, 5997–6010. (2) Headley, J. V.; Peru, K. M.; Janfada, A.; Fahlman, B.; Gu, C.; Hassan, S. Characterization of oil sands acids in plant tissue using Orbitrap ultra-high resolution mass spectrometry and electrospray ionization. Rapid Commun. Mass Spectrom. 2011, 25, 459–462. (3) Rowland, S. J.; Scarlett, A. G.; West, C. E.; Jones, D.; Frank, R. Diamonds in the rough: identiﬁcation of individual naphthenic acids in oil sands process water. Environ. Sci. Technol. 2011, 45, 3154–3159. (4) Garcia-Garcia, E.; Pun, J.; Hodgkinson, J.; Perez-Estrada, L. A.; Gamal El-Din, M.; Smith, D. W.; Martin, J. W.; Belosevic, M. 9813

dx.doi.org/10.1021/es202606d |Environ. Sci. Technol. 2011, 45, 9806–9815

Environmental Science & Technology Commercial naphthenic acids and the organic fraction of oil sands process water induce diﬀerent eﬀects on pro-inﬂammatory gene expression and macrophage phagocytosis in mice. J. Appl. Toxicol. 2011, doi 10.1002/jat.1687 (5) Young, R. F.; Martinez Michel, L.; Fedorak, P. M. Distribution of naphthenic acids in tissues of laboratory-exposed ﬁsh and in wild ﬁshes from near the Athabasca oil sands in Alberta, Canada. Ecotoxicol. Environ. Saf. 2011, 74, 889–896. (6) Zhang, X.; Wiseman, S.; Yu, H.; Liu, H.; Giesy, J. P.; Hecker, M. Assessing the Toxicity of Naphthenic Acids Using a Microbial Genome Wide Live Cell Reporter Array System. Environ. Sci. Technol. 2011, 45, 1984–1991. (7) Rowland, S. J.; West, C. E.; Scarlett, A. G.; Jones, D.; Frank, R. A. Identiﬁcation of individual tetra- and pentacyclic naphthenic acids in oil sands process water by comprehensive two-dimensional gas chromatography-mass spectrometry. Rapid Commun. Mass Spectrom. 2011, 25, 1198–1204. (8) Jones, D.; Scarlett, A. G.; West, C. E.; Rowland, S. J. The toxicity of individual naphthenic acids to Vibrio ﬁscheri. Environ. Sci. Technol. 2011, DOI: 10.1021/es201948j. (9) Frank, R. A.; Kavanagh, R.; Burnison, B. K.; Arsenault, G.; Headley, J. V.; Peru, K. M.; Van Der Kraak, G.; Solomon, K. R. Toxicity assessment of collected fractions from an extracted naphthenic acid mixture. Chemosphere 2008, 72, 1309–1314. (10) Rhodes, S.; Farwell, A.; Hewitt, L. M.; MacKinnon, M.; Dixon, D. G. The eﬀects of dimethylated and alkylated polycyclic aromatic hydrocarbons on the embryonic development of the Japanese medaka. Ecotoxicol. Environ. Saf. 2005, 60, 247–258. (11) Kelly, E. N.; Short, J. W.; Schindler, D. W.; Hodson, P. V.; Ma, M.; Kwan, A. K.; Fortin, B. L. Oil sands development contributes polycyclic aromatic compounds to the Athabasca River and its tributaries. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 22346–22351. (12) Martin, J. M.; Han, X.; Peru, K. M.; Headley, J. V. Comparison of high- and low-resolution electrospray ionization mass spectrometry for the analysis of naphthenic acid mixtures in oil sands process water. Rapid Commun. Mass Spectrom. 2008, 22, 1919–1924. (13) Barrow, M. P.; Witt, M.; Headley, J. V.; Peru, K. M. Athabasca oil sands process water: Characterization by atmospheric pressure photoionization and electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry. Anal. Chem. 2010, 82, 3727– 3735. (14) Kavanagh, R. J.; Frank, R. A.; Oakes, K. D.; Servos, M. R.; Young, R. F.; Fedorak, P. M.; MacKinnon, M. D.; Solomon, K. R.; Dixon, D. G.; Van Der Kraak, G. Fathead minnow (Pimephales promelas) reproduction is impaired in aged oil sands process-aﬀected waters. Aquat. Toxicol. 2011, 101, 214–220. (15) Frank, R. A.; Kavanagh, R.; Burnison, B. K.; Headley, J. V.; Peru, K. M.; Van Der Kraak, G.; Solomon, K. R. Diethylaminoethyl-cellulose clean-up of a large volume naphthenic acid extract. Chemosphere 2006, 64, 1346–1352. (16) Rowland, S. J.; West, C. E.; Scarlett, A. G.; Jones, D.; Boberek, M.; Pan, L.; Ng, M.; Kwong, L.; Tonkin, A. Monocyclic and monoaromatic naphthenic acids: Synthesis & Characterisation. Environ. Chem. Lett. 2011, doi: 10.1007/s10311-011-0314-6. (17) Barrow, M. P.; Headley, J. V.; Peru, K. M.; Derrick, P. J. Fourier transform ion cyclotron resonance mass spectrometry of principal components in oil sands naphthenic acids. J. Chromatogr. A 2004, 1058, 51–59. (18) Yang, C.; Wang, Z.; Yang, Z.; Hollebone, B.; Brown, C. E.; Landriault, M.; Fieldhouse, B. Chemical ﬁngerprints of Alberta oil sands and related petroleum products. Environ. Forensics 2011, 12, 173–188. (19) Wardroper, A. M. K.; Hoﬀmann, C. F.; Maxwell, J. R.; Barwise, A. J. G.; Goodwin, N. S.; Park, P. J. D. Crude oil biodegradation under simulated and natural conditions-II. Aromatic steroid hydrocarbons. Org. Geochem. 1984, 6, 605–617. (20) Sih, C. J.; Wang, K. C.; Tai, H. H. Mechanisms of steroid oxidation by microorganisms. XIII. C22 acid intermediates in the degradation of the cholesterol side chain. Biochemistry 1968, 7, 796–807.

ARTICLE

(21) Namba, T.; Hirota, T.; Hayakawa, S. Synthesis and mutagenicity of a ring-A-aromatized bile acid, 3-hydroxy-19-nor-1,3,5(10)-cholatrien-24-oic acid. J. Lipid. Res. 1988, 29, 809–814. (22) Meney, J.; Kim, Y.-H.; Stevenson, R.; Margulis, T. N. A new steroid aromatization rearrangement involving inversion of side chain conﬁguration. Tetrahedron 1973, 29, 21–30. (23) Mahato, S. B.; Mukherjee, A. Steroid transformations by microorganisms. Phytochemistry 1984, 23, 2131–2154. (24) Mahato, S. B.; Banerjee, S. Steroid transformations by microorganisms. Phytochemistry 1985, 24, 1403–1421. (25) Mahato, S. B.; Banerjee, S.; Podder, S. Steroid transformations by microorganisms. Phytochemistry 1989, 28, 7–40. (26) Mahato, S. B.; Garai, S. Advances in microbial steroid biotransformation. Steroids 1997, 62, 332–345. (27) Paulus, S. Biodegradation de steranes petroliers. These Doctorat, Universite de Strasbourg, France, 1996. (28) Peters, K. E.; Walters, C. C.; Moldowan, J. M. The biomarker guide. Vol. 2. Biomarkers and isotopes in petroleum exploration and earth history. Cambridge University Press: Cambridge, UK, 2007. (29) He, Y.; Wiseman, S. B.; Zhang, X.; Hecker, M.; Jones, P. D.; Gamal El-Din, M.; Martin, J. W.; Geisy, J. P. Ozonation attenuates the steroidogenic disruptive eﬀects of sediment free oil sands process water in the H295R cell line. Chemosphere 2010, 80, 578–584. (30) He, Y.; Wiseman, S. B.; Hecker, M.; Zhang, X.; Wang, N.; Perez, L. A.; Jones, P. D.; Gamal El-Din, M.; Martin, J. W.; Geisy, J. P. Eﬀect of ozonation on the estrogenicity and androgenicity of oil sands process water. Environ. Sci. Technol. 2011, doi.org/10.1021/es2008215. (31) Labardee, D. C.; Zhang, J.-x.; Harris, H. A.; O’Connor, C.; Reynolds, T. Y.; Hochberg, R. B. Synthesis and evaluation of B-, C-, and D-ring-substituted estradiol carboxylic acid esters as locally active estrogens. J. Med. Chem. 2003, 46, 1886–1904. (32) Meyers, C. Y.; Hou, Y.; Winters, T. A.; Banz, W. J.; Adler, S. Activities of a non-classical estrogen, Z-bis-dehydrodoisynolic acid, with ERα and ERβ. J. Steroid Biochem. Mol. Biol. 2002, 82, 33–44. (33) Fieser, L. F.; Holmes, H. L. The synthesis of phenanthrene and hydrophenanthrene derivatives .V. The addition of dienes to cyclic α, β-unsaturated esters. J. Am. Chem. Soc. 1936, 58, 2319–2322. (34) Pincus, G.; Werthessen, N. T. The oestrogenic activity of certain phenanthrene and hydrophenanthrene derivatives. Science 1936, 84, 45–46. (35) Peakman, T. M.; Farrimond, P.; Brassell, S. C.; Maxwell, J. R. De-A-steroids in immature marine shales. Org. Geochem. 1986, 10, 779–789. (36) van Graas, G.; de Lange, F.; de Leeuw, J. W.; Schenck, P. A. DeA-steroid ketones and de-A-aromatic steroid hydrocarbons in shale indicate a novel diagenetic pathway. Nature 1982, 299, 437–439. (37) Djerassi, C.; Wilson, J. M.; Budzikiewicz, H.; Chamberlin, J. W. Mass spectrometry in structural and stereochemical problems. XIV. Steroids with one or two aromatic rings. J. Am. Chem. Soc. 1962, 84, 4544–4552. (38) Thevis, M.; Sch€anzer, W. Mass spectrometric analysis of androstan-17β-ol-3-one and androstadiene-17β-ol-3-one isomers. J. Am. Soc. Mass Spectrom. 2005, 16, 1660–1669. (39) Perlman, D. Advances in Microbiology; Academic Press: New York, 1977; Vol. 22. (40) Corbet, B.; Albrecht, P.; Ourisson, G. Photochemical or photomimetic fossil triterpenoids in sediments and petroleum. J. Am. Chem. Soc. 1980, 102, 1171–1173. (41) Charrie-Duhaut, A.; Lemoine, S.; Adam, P.; Connan, J.; Albrecht, P. Abiotic oxidation of petroleum bitumens under natural conditions. Org. Geochem. 2000, 31, 977–1003. (42) West, C. E.; Jones, D.; Scarlett, A. G.; Rowland, S. J. Compositional heterogeneity may limit usefulness of some commercial naphthenic acids for toxicity assays. Sci. Total Environ. 2011. DOI: 10.1016/ j.scitotenv.2011.05.061. (43) Giamalva, D. H.; Church, D. F.; Pryor, W. A. Kinetics of ozonation. 6. Polycyclic aliphatic hydrocarbons. J. Org. Chem. 1988, 53, 3429–3432. 9814

dx.doi.org/10.1021/es202606d |Environ. Sci. Technol. 2011, 45, 9806–9815

Environmental Science & Technology

ARTICLE

(44) Perez-Estrada, L.; Han, X.; Drzewicz, P.; Gamal El-Din, M.; Fedorak, P. M.; Martin, J. W. Structure-reactivity of naphthenic acids in the ozonation process. Environ. Sci. Technol. 2011, 45, 7431–7437. (45) Huber, M. M.; Ternes, T. A.; von Gunten, U. Removal of estrogenic activity and formation of oxidation products during ozonation of 17α-ethinylestradiol. Environ. Sci. Technol. 2004, 38, 5177–5186.

9815

dx.doi.org/10.1021/es202606d |Environ. Sci. Technol. 2011, 45, 9806–9815

CORRESPONDENCE/REBUTTAL pubs.acs.org/est

Comment on “Photocatalytic Oxidation of Arsenite over TiO2: Is Superoxide the Main Oxidant in Normal Air-Saturated Aqueous Solutions?” ecently, Fei et al.1 reported a photoelectrochemical (PEC) study on the arsenite (As(III)) oxidation using illuminated TiO2 electrodes under aerated conditions, concluding that under open-circuit conditions both the photogenerated holes and superoxide radicals participate in the global photooxidation process, but holes have a higher contribution (57%) than superoxide and its derivatives (43%). After a thorough reading of the paper, we would like to comment on the following points. We are happy to see that the authors ﬁnally admitted that superoxide plays a signiﬁcant role in the photocatalytic oxidation (PCO) of As(III), as we claimed in our previous reports.2 4 However, in this paper1 and their previous comments,5,6 the authors seem to be unduly biased with PEC data in discussing the PCO mechanism in the TiO2 slurry system. The commented paper is entirely about a PEC study of arsenite oxidation but has a misleading title that states “photocatalytic oxidation...in airsaturated aqueous solution”. The photocatalytic and PEC reactions are closely related but they can be very diﬀerent in cases like the As(III) oxidation in which the biased potential sensitively inﬂuences the interfacial charge transfer (especially the critical recombination step: As(IV) + ecb f As(III)). Changing the electrode potential would eventually change both the oxidant and charge carriers concentrations (electrons, holes, superoxide, and derived radicals) and the kinetic constants of the elemental charge transfer steps to alter the rate determining step and hence the overall mechanism. What Fei et al.1 studied and concluded is about the PEC mechanism, not PCO mechanism. The PEC investigation provides useful data for understanding photocatalysis mechanism, but it does not accurately represent the photocatalytic system itself. Fei et al.1 failed to provide a balanced view on the PCO mechanism of As(III) with strong bias toward PEC data but against more direct photocatalytic evidence. They simply disregarded numerous evidence obtained from the slurry PCO systems2 4 without showing any solid scientiﬁc data to disprove them. For example, the direct evidence of transient spectroscopic data supporting the role of As(III) as a charge recombination center2 was simply dismissed as “questionable” despite our reply to their concerns.7 We wonder how they can be so sure about the credibility of our data even without trying to reproduce them. The claim by Fei et al. against our data and interpretation should be based on more complete scientiﬁc reasonings. A full mechanistic assessment should take into account all evidence coming from PCO experimental data, which Fei et al. failed to do. More speciﬁc comments are as follows. The authors compared the eﬀect of superoxide radicals in the dark and under illumination in Figures 2a and 3a,1 concluding that the photooxidation under illumination was accelerated by the presence of more eﬃcient oxidants than superoxide. This is not a fair comparison, as the concentration of superoxide should not be the same in the dark and under illumination. The claim

R

r 2011 American Chemical Society

Figure 1. The time proﬁles of As(III) oxidation to As(V) on UVirradiated TiO2 electrode, in the presence or absence of 0.1 M TBA (OH radical scavenger), at open-circuit condition and biased at 0.6 V vs SCE. Working electrode: P25 TiO2 deposited on 4 cm2 FTO. Counter electrode: Pt coiled wire. Reference electrode: SCE. Electrolyte: airequilibrated 0.5 M NaClO4 solution, buﬀered at pH 3; [As(III)]0 = 500 μM.

can be justiﬁed only if both systems contain a similar concentration of superoxides. Second, they claim that under illumination, the production of As(V) in aerated conditions (Figure 3a1) was accelerated even at negative bias (few holes available), due to the probable conversion of superoxide to hydroxyl radicals. To investigate the role of hydroxyl radicals in this case, we performed the similar PEC experiments in the presence and absence of tertbutyl alcohol (TBA, an OH radical scavenger). As shown in Figure 1, the presence of excess amount of TBA retarded the production of As(V) only by 34% and 19% (after 120 min irradiation) in the open-circuit and negatively biased condition, respectively, which indicates that superoxides should contribute more to the oxidation of As(III) than OH radicals. With a positive bias, the electrons are scavenged and the PEC oxidation of As(III) should be mediated by holes in the absence of O2 (Figure 3b1). On the other hand, it should be noted that the eﬀect of TBA in this PEC system is quite diﬀerent from the slurry Published: October 17, 2011 9816

dx.doi.org/10.1021/es202954j | Environ. Sci. Technol. 2011, 45, 9816–9817

Environmental Science & Technology PCO system in which the presence of excess TBA does not retard the oxidation rate at all.2 Even at the open-circuit condition which is the closest to the PCO system, a measurable inhibition eﬀect of TBA was observed (Figure 1, upper panel). This observation clearly indicates that the PEC system cannot faithfully represent the PCO system in the present case of As(III) oxidation. A direct comparison between the photocatalytic mechanism in slurry and PEC mechanism on the electrode should be very carefully carried out, considering that the photocatalytic mechanism sensitively changes depending on many experimental parameters (especially in the case of As(III) PCO) as we demonstrated in the previous reports.2 4 Between PCO and PEC systems, there are many diﬀerences in the number of irradiated particles, the number of charge carriers per particle, the available surface area, the mass transfer rate, the interparticle electron transfer rate, and the substrate/surface ratio, which eventually inﬂuence the overall mechanisms. As such conditions are diﬃcult to be balanced between both systems, all mechanistic studies must be taken with care as their nature is not universally conclusive. Hence, we strongly disagree with Fei et al.1 in their claim that their proposed PEC method could provide direct and undisputed evidence to reveal the true mechanism in PCO. The mechanism drawn from the PEC investigation cannot be always applied to the PCO system.

CORRESPONDENCE/REBUTTAL

(6) Leng, W. H.; Li, H.; Fei, H.; Zhang, J. Q.; Cao, C. N. Comment on “Photocatalytic oxidation mechanism of As(III) on TiO2: Unique role of As(III) as a charge recombinant species”. Environ. Sci. Technol. 2011, 45, 2028–2029. (7) Monllor-Satoca, D.; Tachikawa, T.; Majima, T.; Choi, W. Response to comment on photocatalytic oxidation mechanism of As(III) on TiO2: Unique role of As(III) as a charge recombinant species. Environ. Sci. Technol. 2011, 45, 2030–2031.

Damian Monllor-Satoca and Wonyong Choi* School of Environmental Science and Engineering, Pohang University of Science and Technology (POSTECH), Pohang 790-784, Korea

’ AUTHOR INFORMATION Corresponding Author

*Fax: +82-54-279-8299, e-mail: [email protected].

’ ACKNOWLEDGMENT This work was supported by the KOSEF NRL program (No. R0A-2008-000-20068-0) and KOSEF EPB center (Grant No. R11-2008-052-02002) ’ REFERENCES (1) Fei, H.; Leng, W.; Li, X.; Cheng, X.; Xu, Y.; Zhang, J.; Cao, C. Photocatalytic oxidation of arsenite over TiO2: Is superoxide the main oxidant in normal air-saturated aqueous solutions? Environ. Sci. Technol. 2011, 45, 4532–4539. (2) Choi, W.; Yeo, J.; Ryu, J.; Tachikawa, T.; Majima, T. Photocatalytic oxidation mechanism of As(III) on TiO2: unique role of As(III) as a charge recombinant species. Environ. Sci. Technol. 2010, 44, 9099–9104. (3) Ryu, J.; Choi, W. Photocatalytic oxidation of arsenite on TiO2: understanding the controversial oxidation mechanism involving superoxides and the eﬀect of alternative electron acceptors. Environ. Sci. Technol. 2006, 40, 7034–7039. (4) Ryu, J.; Choi, W. Eﬀects of TiO2 surface modiﬁcations on photocatalytic oxidation of arsenite: The role of superoxides. Environ. Sci. Technol. 2004, 38, 2928–2933. (5) Leng, W. H.; Cheng, X. F.; Zhang, J. Q.; Cao, C. N. Comment on “Photocatalytic oxidation of arsenite on TiO2: Understanding the controversial oxidation mechanism involving superoxides and the eﬀect of alternative electron acceptors”. Environ. Sci. Technol. 2007, 41, 6311–6312. 9817

dx.doi.org/10.1021/es202954j |Environ. Sci. Technol. 2011, 45, 9816–9817

CORRESPONDENCE/REBUTTAL pubs.acs.org/est

Response to Comment on “Photocatalytic Oxidation of Arsenite over TiO2: Is Superoxide the Main Oxidant in Normal Air-Saturated Aqueous Solutions?”

W

e appreciate the opportunity to respond to the comments of Choi et al.1 on our recent article.2 We believe that the comments are completely incorrect, irrelevant, and reﬂect a lack of understanding of our paper. The ﬁrst argument in the comment is not true where it is stated that we have ﬁnally admitted the superoxide plays a signiﬁcant role in the photocatalytic oxidation (PCO) of As(III) as they claimed in their previous four reports. We have never excluded that the superoxide may play a role in the system, and just disagree with their claim that it is the main oxidant because they failed to provide any convincing evidence to support this. Our conclusion is that the contribution of superoxide is not more than that of photohole even if that of its derivates is included.2 We are surprised that they said our conclusion was in agreement with their claim. The following argument in the comment where they argued that our mechanism is not a PCO mechanism and that it has a misleading title is also wrong. We stated clearly in our paper2 that the mechanism or contribution of superoxide and photohole was obtained under open-circuit (OC) and the title question was accurately answered by the results at OC. Yet this fact is overlooked, obviously reﬂecting a lack of understanding of our paper. Their comments about the role of potential are common sense in this ﬁeld. Note that the mentioned recombination process via As(IV) does not aﬀect the validness of the ﬂux-matching principle (eq 5 2). Potential does not aﬀect the validness of charge columbic eﬃciency (not quantity) though it aﬀects both charge density and rate constant, as only the transferred interfacial charge was needed for deriving the eﬃciency. They argued that we disregarded their evidence obtained from the slurry PCO systems without showing data to disprove them. This is also not true. For example, our results in both the previous3 and the commented paper go against their arguments. Importantly, their evidence was based on the addition of competitive additives that could alter the normal mechanism and is even contradictory.2 Thus, it is invalid. Actually, all their important evidence in the four papers cited by us in ref 2 was further questioned. Particularly, they argued that their reply to our concerns to the transient diﬀuse reﬂectance (TDR) results was ignored. This is also untrue as we have presented it already.2 Again, for instance, the TDR results are even in contrast with that of photocurrent which was actually increased with As(III) concentration but the untrue photocurrent data (due to sampling time too long, with O2, and etc.) were chosen to justify their claim of As(III) acting as a charge recombination center as we commented.2 We wonder why the actual mechanism depends on research methods. Furthermore, they even used misconception of lifetime of charge to explain their TDR results (lifetime depends the extent of the absolute TDR value decays.). We do not think it is necessary to explain more here within the limited journal space. In fact, as long as only one opposite fact like the r 2011 American Chemical Society

one given above cannot be explained by their arguments their conclusions should be invalid, thus we do not need to reproduce their data. Their two speciﬁc comments about the accelerated oxidation of superoxide by illumination are outside the scope of our article as we mentioned previously, and actually it has been brieﬂy addressed.2 For example, it is supported directly by a combination of much smaller columbic eﬃciency of electron (Figure 4 2), comparable current (Figure 1 2) or charge, and much smaller rate of As(III) oxidation in the dark compared to under illumination, and indirectly by the results of Figure S2.2 Importantly, these comments are irrelevant to the key point of our paper, partitioning the two mechanisms, because our conclusion does not depend how and whether it is accelerated. On the contrary, their claim that superoxide should contribute more to the oxidation than hydroxyl radicals from a small retardation of tert-butyl alcohol (TBA) under both OC and negatively biased condition is completely invalid, because it is based on (i) a misunderstanding that •OH was the only converted stronger oxidant than superoxide under a negative bias (we did not exclude others); (ii) an unproved assumption that the oxidation of TBA could fully compete with that of As(III) (probably not, possibly due to it is just a surface reaction), and the last but not the least (iii) a dogma that TBA is an •OH scavenger and consequently should retard the PCO of As(III) under OC (actually it depends on the diﬀerence in rate between that retarded by TBA through holemediated reactions and that enhanced by superoxide-mediated reactions because of an increased electron density as we mentioned2). Considering these, it is enough to conclude that their claim is wrong and even opposite to common sense. The argument is also untrue where it is stated that the photoelectrochemical system cannot faithfully represent the PCO system for As(III) oxidation, based on an observation that the oxidation was retarded by excess TBA with the ﬁlm electrode but not with slurry TiO2. Reemphasizing that our mechanism is the one under no bias. Also, we argue this is an unfair comparison because an improper electrode nanostructure TiO2/FTO instead of the polycrystalline and almost compact one we chose, was employed where there were many uncertainties such as a possible diﬀerence in the O2 reduction behavior (thus aﬀecting superoxide) over the TiO2 and FTO (exposed to solutions) that may aﬀect the As(III) PCO, yet these factors were not considered for comparison. The lengthy comments in the last paragraph are doctrines and cannot be used either to support their proposed superoxide mechanism or as an excuse that the PCO mechanism cannot be probed by proper photoelectrochemical methods. In summary, all their comments do not undermine our conclusions at all. Published: October 26, 2011 9818

dx.doi.org/10.1021/es2036042 | Environ. Sci. Technol. 2011, 45, 9818–9819

Environmental Science & Technology

CORRESPONDENCE/REBUTTAL

Wenhua Leng,* Hui Fei, and Jianqing Zhang Department of Chemistry, Yuquan Campus, Zhejiang University, Hangzhou, Zhejiang 310027, China

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT This project was supported by the National Basic Research Program of China (Grant No. 2011CB936003) and the National Science Foundation Council of China (NSFC, Grant No. 50971116). ’ REFERENCES (1) Monllor-Satoca, D.; Choi, W., Comment on “Photocatalytic oxidation of arsenite over TiO2: Is superoxide the main oxidant in normal air-saturated aqueous solutions? Environ. Sci. Technol. 2011, 45. (2) Fei, H.; Leng, W.; Li, X.; Cheng, X.; Xu, Y.; Zhang, J.; Cao, C. Photocatalytic oxidation of arsenite over TiO2: Is superoxide the main oxidant in normal air-saturated aqueous solutions? Environ. Sci. Technol. 2011, 45, 4532–4539. (3) Leng, W. H.; Cheng, X. F.; Zhang, J. Q.; Cao, C. N. Comment on “Photocatalytic oxidation of arsenite on TiO2: Understanding the controversial oxidation mechanism involving superoxides and the eﬀect of alternative electron acceptors. Environ. Sci. Technol. 2007, 41, 6311–6312.

9819

dx.doi.org/10.1021/es2036042 |Environ. Sci. Technol. 2011, 45, 9818–9819

ADDITION/CORRECTION pubs.acs.org/est

Correction to Environmental and Sustainability Factors Associated With Next-Generation Biofuels in the U.S.: What Do We Really Know? Pamela R. D. Williams, Daniel Inman, Andy Aden, and Garvin A. Heath It was recently brought to our attention that there is an error in the weblink in reference 103 of our article. The existing link is http://www.epa.gov/Region7/priorities/agriculture/ethanol_ plants_manual.pdf. The correct link should be http://www. epa.gov/Region7/priorities/agriculture/pdf/ethanol_plants_ manual.pdf.

Published: October 17, 2011 9820

dx.doi.org/10.1021/es2035043 | Environ. Sci. Technol. 2011, 45, 9820–9820

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

No title

Recommend Documents