WATER RESEARCH A Journal of the International Water Association
Editor-in-Chief
Associate Editors
Mark van Loosdrecht Delft University of Technology and KWR Watercycle Research The Netherlands E-mail:
[email protected]
Andrew Baker University of New South Wales Australia
Editors
Damien Batstone University of Queensland Australia
Jean-Claude Block Université H. Poincaré, Nancy I France
Nicolas Bernet INRA France
David Dixon University of Melbourne Australia
Guang-Hao Chen Hong Kong University of Science & Technology China
Hiroaki Furumai University of Tokyo Japan
Rob Eldridge Glen Waverley Australia
Xiaodi Hao Beijing University of Civil Engineering and Architecture China Gregory Korshin University of Washington USA Anna Ledin Formas Sweden Eberhard Morgenroth ETH Zürich and Eawag, Swiss Federal Institute of Aquatic Science and Technology Switzerland Wolfgang Rauch University of Innsbruck Austria
Monica B. Emelko University of Waterloo Canada Tim Fletcher University of Melbourne Australia Rosina Girones University of Barcelona Spain Stephen Gray Victoria University Australia Kate Grudpan Chiang Mai University Thailand
Dionisis Mantzavinos Technical University of Crete Greece Yoshihiko Matsui Hokkaido University Japan Armand Maul Université de Metz France Max Maurer Eawag Switzerland How Yong Ng National University of Singapore Singapore Jong Moon Park Pohang University of Science & Technology Korea Susan Richardson US Environmental Protection Agency USA Miguel Salgot University of Barcelona Spain David Sedlak University of California, Berkeley USA Masahiro Takahashi Hokkaido University Japan Nathalie Tufenkji McGill University Canada
Peter Hillis AECOM Australia Hong-Ying Hu Tsinghua University China
Vayalam Venugopalan BARC Facilities India
Ulf Jeppsson Lund University Sweden
Eduardo von Sperling Federal University of Minas Gerais Brazil
Jae-Hong Kim Georgia Institute of Technology USA
Julie Zilles University of Illinois at Urbana-Champaign USA
Thomas Ternes Bundesanstalt für Gewässerkunde Germany
Günter Langergraber University of Natural Resources and Applied Life Sciences, Vienna (BOKU) Austria
Anastasios Zouboulis Aristotle University of Thessaloniki Greece
Stefan Wuertz University of California, Davis USA
Shang-Lien Lo National Taiwan University Taiwan
Editorial Office
Maria Reis Universidade Nova de Lisboa/FCT Portugal Hang-Shik Shin Korea Advanced Institute of Science and Technology Korea
E-mail:
[email protected]
Publication information: Water Research (ISSN 0043-1354). For 2012, volume 46 is scheduled for publication. Subscription prices are available upon request from the publisher or from the Elsevier Customer Service Department nearest you or from this journal’s website (http://www.elsevier.com/locate/watres). Further information is available on this journal and other Elsevier products through Elsevier’s website (http://www.elsevier.com). Subscriptions are accepted on a prepaid basis only and are entered on a calendar year basis. Issues are sent by standard mail (surface within Europe, air delivery outside Europe). Priority rates are available upon request. Claims for missing issues should be made within six months of the date of dispatch. Author Enquiries: For enquiries relating to the submission of articles (including electronic submission) please visit this journal’s homepage at http://www.elsevier.com/ locate/watres. Contact details for questions arising after acceptance of an article, especially those relating to proofs, will be provided by the publisher. You can track accepted articles at http://www.elsevier.com/trackarticle. You can also check our Author FAQs at http://www.elsevier.com/authorFAQ and/or contact Customer Support via http://support.elsevier.com. Orders, claims, and journal enquiries: Please contact the Elsevier Customer Service Department nearest you: St. Louis: Elsevier Customer Service Department, 3251 Riverport Lane, Maryland Heights, MO 63043, USA; phone: (877) 8397126 [toll free within the USA]; (+1) (314) 4478878 [outside the USA]; fax: (+1) (314) 4478077; e-mail:
[email protected] Oxford: Elsevier Customer Service Department, The Boulevard, Langford Lane, Kidlington OX5 1GB, UK; phone: (+44) (1865) 843434; fax: (+44) (1865) 843970; e-mail:
[email protected] Tokyo: Elsevier Customer Service Department, 4F Higashi-Azabu, 1-Chome Bldg, 1-9-15 Higashi-Azabu, Minato-ku, Tokyo 106-0044, Japan; phone: (+81) (3) 5561 5037; fax: (+81) (3) 5561 5047; e-mail:
[email protected] Singapore: Elsevier Customer Service Department, 3 Killiney Road, #08-01 Winsland House I, Singapore 239519; phone: (+65) 63490222; fax: (+65) 67331510; e-mail:
[email protected] Application for membership of International Water Association should be made to: Executive Director, IWA, Alliance House, 12 Caxton Street, London SW1H 0QS, U.K. (Tel.: +44 207 654 5500; Fax: +44 207 654 5555; e-mail:
[email protected]; website: http://www.IWAhq.org.uk). Registered Charity (England) No. 289269. Individual membership is available from £30 upwards. For details contact IWA.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 5 9 e2 6 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A review of water recovery by vapour permeation through membranes Brian Bolto*, Manh Hoang, Zongli Xie CSIRO Materials Science and Engineering, Private Bag 33, Clayton South MDC, Victoria 3169, Australia
article info
abstract
Article history:
In vapour permeation the feed is a vapour, not a liquid as in pervaporation. The process
Received 26 May 2011
employs a polymeric membrane as a semi-permeable barrier between the feed side under
Received in revised form
high pressure and the permeate side under low pressure. Separation is achieved by the
21 October 2011
different degrees to which components are dissolved in and diffuse through the
Accepted 24 October 2011
membrane, the system working according to a solution-diffusion mechanism. The mate-
Available online 31 October 2011
rials used in the membrane depend upon the types of compounds being separated, so water transport is favoured by hydrophilic material, whether organic or inorganic. The
Keywords:
process is used for the dehydration of natural gas and various organic solvents, notably
Vapour permeation
alcohol as biofuel, as well as the removal of water from air and its recovery from waste
Pervaporation
steam. Waste steam can be found in almost every plant/factory where steam is used. It is
Gas drying
frequently contaminated and cannot be reused. Discharging the spent steam to the
Solvent dehydration
atmosphere is a serious energy loss and environmental issue. Recycling the steam can
Steam recovery
significantly improve the overall energy efficiency of an industry, which is responsible for massive CO2 emissions. Steam separation at high fluxes and temperatures has been accomplished with a composite poly(vinyl alcohol) membrane containing silica nanoparticles, and also, less efficiently, with an inorganic zeolite membrane. Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved.
Contents 1. 2.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Water permeability of membrane systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Dehydration of natural gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Drying of compressed air . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Flue gas dehydration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4. Dehydration of ethanol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5. Dehydration of isopropanol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6. Dehydration of acetonitrile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.7. Steam recovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.8. Miscellaneous water removal applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
* Corresponding author. Tel.: þ61 3 9545 2037; fax: þ61 3 9545 1128. E-mail address:
[email protected] (B. Bolto). 0043-1354/$ e see front matter Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.052
260 260 261 261 261 261 262 262 262 263
260
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 5 9 e2 6 6
3. 4. 5.
Membranes used in vapour permeation processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Comparison of vapour permeation and pervaporation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.
Introduction J¼
Vapour permeation employs a polymeric membrane as a semi-permeable barrier between a feed side under high pressure and a permeate side under low pressure (Brinkmann et al., 2003). Unlike pervaporation, the feed is a vapour and not a liquid, so there is no phase change or significant temperature difference across the membrane. Separation is achieved by the different degrees to which components are dissolved in and diffuse through the polymer. The polymers used in the membrane will hence depend on the types of compounds being separated (Leemann et al., 1996). The driving force can be approximated to the difference in partial pressures of the components in the feed. Compared to pervaporation, vapour permeation effectively increases the permselectivity of water (Fan et al., 2002) and is capable of higher fluxes (Sander and Janssen, 1991). Another advantage is that the membrane area required is less, and there is a lower likelihood of membrane damage by impurities in the feed. Although vapour permeation operates with the same type of membranes as pervaporation, it has taken longer for its technical realisation. The first industrial vapour permeation plant was built in Germany in 1989 for dehydrating 30 kL/d of ethanol (Sander and Janssen, 1991). More than 100 membrane vapour-gas separation systems have been installed worldwide for recovering high value solvents, liquefied petroleum gas, refrigerant gases, monomers such as ethylene, propylene and vinyl chloride, and for removing the acid gases H2S and CO2 from natural gas and hydrocarbon vapours from air streams in the petrochemical industry (Baker et al., 1998; Jonquie`res et al., 2002). The transport of water vapour and inert gases through polymeric membranes was reviewed in the literature some time ago (Metz et al., 2005). The authors noted that the use of membranes for this purpose is of major industrial importance, with applications in areas such as the drying of natural gas and compressed air, protective apparel, packaging materials, roofing covers and humidity control in confined spaces, involving air conditioning in buildings, aviation and space flight. Steam recovery is also in this category. Vapour permeation, like pervaporation, works according to a solution-diffusion mechanism. The molecular interaction between the membrane and the separated species is the determining factor rather than the molecular size. The main component of the process is the membrane material which determines the permeation and selectivity and hence the separation properties of the process. The driving force for the mass transfer of permeate from the feed side to the permeate side of the membrane is the chemical potential gradient established by applying a difference in the partial pressures of permeate across the membrane (Aptel et al., 1972; Neel, 1991). The membrane performance is expressed as the membrane flux J in
263 263 265 265
P$DP l
where P is the permeability coefficient for a vapour, l is thickness of the membrane, and DP is the difference in pressure between the feed and permeate sides (Scott, 1998).
2.
Water permeability of membrane systems
The water vapour and nitrogen permeability behaviour of 19 polymers reported in the literature has been summarised (Metz et al., 2005). The data extrapolated to zero water activity are given in Table 1. Many of the values were obtained from pure gas permeabilities by calculating the ratios for the mixtures. The permeabilties are given in Barrer, a non-SI unit of gas permeability (specifically, oxygen permeability) used in the contact lens industry (Alter, 1962): one Barrer ¼ 1011 (cm3 O2) cm cm2 s1 mmHg1. In real mixtures however, water may swell the membrane so that its effect on the slower
Table 1 e Water vapour permeabilities and water/ nitrogen selectivities at 30 C for various organic polymers (Metz et al., 2005; Sijbesma et al., 2008). Polymer
PEBAX 1074a PBT/PEO block copolymerb Sulphonated poly (ether ether ketone) Polydimethylsiloxane Sulphonated poly (ether sulphone) Ethyl cellulose Cellulose acetate Poly(phenylene oxide) Poly(ether sulphone) Natural rubber Polysulphone Polycarbonate Polystyrene Polyimide Polyacrylonitrile Poly(vinyl chloride) Polyamide 6 Polypropylene Poly(vinyl alcohol) Polyethylene
Water permeability, barrer
Selectivity, H2O/N2
160,000 85,500 61,000
200,000 40,500 10,200,000
40,000 15,000
143 214,000
20,000 6000 4060 2620 2600 2000 1400 970 640 300 275 275 68 19 12
6060 24,000 1068 10,500 299 8000 4670 388 5,330,000 1,880,000 12,500 11,000 227 33,300 6
a PEBAX 1074 is a blend of polyether block amide (nylon 12) and poly(ethylene oxide). b (Polybutylene terephthalate)/poly(ethylene oxide) block copolymer.
261
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 5 9 e2 6 6
species is not known with certainty. Generally, for binary mixtures of permanent gases a higher selectivity is accompanied by a lower permeability, but such a relationship does not hold for water vapour in a mixture with a permanent gas, where most of the highly selective polymers also have a very high permeability. There can be a variation of over seven orders of magnitude in selectivity and five orders of magnitude in permeability. The effect of polymer structure on performance is variable. Completely non-polar membranes such as those based on natural rubber, polystyrene, polypropylene and polyethylene tend to have poorer permeabilities. The low permeability of poly(vinyl alcohol) or PVA is surprising, as this is the basis of many commercial membranes in the area (Jonquie`res et al., 2002). The presence of some polarity such as ether or sulphonate groups is desirable. The introduction of sulphonate or carboxylate groups into the PVA structure would seem to be worthwhile. What is absent in the list of polymers tested here for vapour permeation are polysalts formed from cationic and anionic polyelectrolytes, which have good performance in pervaporation dehydration applications. For the removal of water from water/organic liquid or vapour mixtures, hydrophilic organic polymers are generally used because water is readily incorporated and diffused through these materials. The hydrophilicity is caused by groups present in the polymer chain that are able to interact with water molecules. Examples of hydrophilic polymers are: cellulose acetate, ionic polymers, PVA and polyacrylonitrile. Hydrophilic features are found to be essential for good pervaporation performance in the dehydration of ethanol (Bolto et al., 2011).
2.1.
increased by a factor of 2e5, and there is no adverse effect on selectivity (Van Wijk and Jansen, 1990).
2.2.
For the drying of compressed air the preferred membrane material is polydimethylsiloxane or cellulose acetate, because of their acceptable H2O/N2 selectivities and high permeabilities (Sijbesma et al., 2008). Increasing the level of sulphonate groups in poly(ether ether ketone) membranes has been shown to not only increase the permeability of water vapour, but to decrease gas permeability (Liu et al., 2001). The permeability towards water is doubled in poly(aryl ether sulphone) membranes by introducing carboxylate groups, and the H2O/N2 selectivity can be raised to more than 105 (Wang et al., 2001).
2.3.
For the dehydration of natural gas the preferred highly selective membranes are sulphonated poly(ether sulphone) and a (polybutylene terephthalate)/(poly(ethylene oxide) or PBT/PEO block copolymer of molecular weight 1000 Da (Metz et al., 2005). The latter polymer (Fig. 1) had soft rubbery amorphous hydrophilic and hard rigid crystalline hydrophobic segments. It has been found that the water vapour permeability decreases with increasing temperature, which was attributed to the decrease in water solubility on heating (Metz et al., 2002). A cellulose ether composite membrane of high water vapour permeability and selectivity is proposed for the removal of water vapour from pressurized gases or gas mixtures (Ohlrogge et al., 2002). The removal of water vapour from a gas mixture can be accelerated by impregnating a membrane composed of regenerated cellulose with a hygroscopic electrolyte such as lithium bromide. The flux is
C O
O
O
Flue gas dehydration
Appropriate polymers for flue gas dehydration are suggested to be sulphonated poly(ether ether ketone), or SPEEK, and the commercial polymer PEBAX 1074, which contains 45wt% of a polyether block amide (nylon 12) and 55wt% PEO (Sijbesma et al., 2008). Data for the latter material have been included in Table 1. Experiments have been conducted on this polymer using an artificial flue gas containing 11.2 vol% water, the remaining gases being nitrogen, CO2 and oxygen. A continuous removal of 0.6e1 kg/m2 h was achieved over a 150 h test at 50 C. Long term tests on a real, aggressive flue gas stream gave an average water vapour removal rate of 0.2e0.46 kg/ m2 h over 5300 h. Although the flux decreased with time because of the deposition of fly ash dust and gypsum crystals on the membrane surface, performance was judged to be adequate. The product water quality was not sufficient for reuse in the steam cycle, but was satisfactory for use as feed for a demineralised water plant.
Dehydration of natural gas
C
Drying of compressed air
2.4.
Dehydration of ethanol
The production of 99.9% ethanol from 94% ethanol has been achieved by vapour permeation (Sander and Janssen, 1991). Although the membrane material is not revealed, the authors remark on the use of the same materials for vapour permeation and pervaporation of aqueous alcohol, and in an earlier article quote PVA as the membrane of choice for this application (Sander and Soukup, 1988). Vapour permeation was claimed to have the advantages of a lower required membrane area, higher flow rates, and the avoidance of harsh chemical reaction conditions. The volumetric flow rate is about 400 times higher for a vapour than a liquid feed at a given mass flow rate and comparable temperature and pressure conditions. Nevertheless, pervaporation seems to be well established in
O
C
C O
O
O
X Fig. 1 e Structure of PBT/PEO block copolymer (Metz et al., 2005).
O
m Y
262
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 5 9 e2 6 6
this area. A composite PVA membrane commercialized for pervaporation when tested in vapour permeation mode has yielded fluxes as high as 5.5 kg/m2 h for 90 wt% aqueous ethanol at a pressure of 500 kPa, down to 0.7 kg/m2 h when the pressure was 100 kPa (Jansen et al., 1992). The addition of sodium montmorillonite clay to PVA decreased the water vapour permeation rate from 0.44 to 0.38 kg/m2 h for 90 wt% ethanol when there was 10 wt% clay present (Yeh et al., 2003). This was ascribed to an increase in the tortuosity of the diffusion pathway because of the barrier properties of the dispersed clay. Composite membranes of PVA coated on Nylon-4 had a flux of 0.088 kg/m2 h in the vapour permeation dehydration of 10 wt% aqueous ethanol at 25 C (Lee et al., 1992). In pervaporation mode the flux was 0.42 kg/ m2 h, but the respective separation factors were 94 and 14. Using microporous silica membranes formed by coating a microporous silica layer onto a ceramic porous tube (GallegoLizon et al., 2002), a water flux increase of approximately 100 times to 10 kg/m2 h has been reported for an increase in temperature from 25 to 120 C during the dehydration of a 96 wt % ethanol/water mixture, whilst the ethanol flux remained low at less than 0.1 kg/m2 h (van Veen et al., 2001). The dehydration of aqueous ethanol by vapour permeation has been studied using tubular NaY zeolite membranes at 90e110 C (Sato et al., 2008). For a 90 wt% ethanol feed at 110 C a flux of 20 kg/m2 h was obtained, to give a retentate that was 96.1 wt% ethanol. At a flow rate of 5 kg/m2 h the product was 98.5 wt% ethanol. For a pure water feed a flux of 80 kg/m2 h was achieved. With NaA zeolite membranes and 10 wt% water content a vapour permeation flux of 11 kg/m2 h at 125 C has been obtained (Richter et al., 2006). Further results have been obtained at lower temperatures: 2.5 kg/m2 h at 90 C and 5 kg/ m2 h at 110 C (Sommer and Melin, 2005). Some change in the crystal structure of LTA zeolite has been observed after its use in dehydrating aqueous ethanol (Kyotani et al., 2009). A small deterioration in the water permeance was observed, and some 10% of sodium ions were detached, indicating proton exchange. Membranes composed of aromatic polyamides made from the reaction of bis[4-(4-aminophenoxy) phenyl]diphenylmethane with the diacids terephthalic acid, 5-tert-butylisophthalic acid or 4,40 -hexafluoroisopropylidenedibenzoic acid have been used in vapour permeation experiments (Fan et al., 2002). The highest flux of 0.2 kg/m2 h was obtained with the fluorinated species. It was found that the permeation rate could be increased by introducing a bulky group into the polymer backbone. Other fluorinated polymers have given similar results, with best flux being 0.38 kg/m2 h (Teng et al., 2000). Polyimide membranes have been blended with sulphonated polyether sulphone to give enhanced results of 0.64 kg/m2 h for an air sweep mode of operating vapour permeation at 80e100 C (Wu et al., 2002). Earlier work on a series of polyimide membranes had shown large variations in performance in treating 10 wt% aqueous ethanol at 75 C, the permeation of water being some 10 times greater in the polyimide made from anhydride and dia3,30 ,4,40 -biphenyltetracarboxylic minodiphenylsulphone compared to that prepared from pyromellitic anhydride and 4,40 -oxydianiline (Okamoto et al., 1992). This was attributed to the high rigidity and bulkiness of the polymer backbone in the former case preventing close chain packing and giving a more open structure for water transport.
The vapour permeation performance of symmetrical and asymmetrical polycarbonate membranes, prepared by dryphase and wet-phase inversion methods, has been studied for aqueous ethanol (Wang et al., 2005). In the case of the symmetrical polycarbonate membrane, vapour permeation was found to have a significantly increased separation factor and a slightly decreased permeation rate compared to pervaporation, the respective permeation rates being 0.16 and 0.18 kg/m2 h for a 20 wt% aqueous ethanol at 25 C.
2.5.
Dehydration of isopropanol
The performance of commercially available crosslinked PVA and microporous silica membranes for the dehydration of isopropanol (IPA)/water mixtures by pervaporation/vapour permeation has been studied (Gallego-Lizon et al., 2002). The PVA membranes used in the study belong to a generic family of asymmetric composite membranes for water permeation. These membranes have a supporting layer of nonwoven porous polyester onto which an ultrafiltration membrane is cast and finally a layer of crosslinked PVA. For the range of conditions investigated in the study, water fluxes generally increased with water concentration and increased with the operating temperature from 70 to 105 C. The water flux through the microporous silica membrane was found be to up to three times higher than that through the PVA membrane. A water flux of 21.5 kg/m2 h across the microporous silica membrane was reported at 105 C for a 12.5 wt% water mixture. However, the IPA flux through the microporous silica membrane was much higher than that through the PVA membrane.
2.6.
Dehydration of acetonitrile
In a comparison of vapour permeation and pervaporation for dehydrating aqueous acetonitrile with a silica membrane, a total flux of 3.9 kg/m2 h at 70 C has been measured (Fontalvo et al., 2005). In an economical evaluation it was ascertained that at high water concentrations or just for overcoming the azeotropic composition, vapour permeation was preferred, but for water concentrations lower than the azeotropic amount, pervaporation was the better option.
2.7.
Steam recovery
Water vapour as steam is the most universal energy carrier. Its application is wide spread and can be found in all aspects of industrial processes. Industry converts more than 70% of the fuel it purchases for energy into steam. Waste steam can be found almost in every plant/factory where steam is used, from large industrial establishments such as refineries, power plants, chemical factories, steel makers, ore mining, to medium and small plants such as sugar mills and food processing facilities. It is usually contaminated and cannot be reused. A common practice in dealing with it is to use a condenser to collect water or to discharge the steam to the atmosphere. Discharging the spent steam to the atmosphere is not only an energy loss, but is at the same time an environmental issue as water vapour is a major driver of greenhouse gas-induced climate change (Hoang and Nguyen, 2009).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 5 9 e2 6 6
The steam consumption in a typical thermal power station of a 1000 MW capacity is about 2800 t/h, which translates to about 800 kg/s of condensate. A 10,000 tonne/day discharge of waste steam represents a loss of w0.7 million m3 of natural gas/day or w$40M/year (at 0.4 cent/MJ wholesale to industry). While no direct figures are available in Australia, recycling of steam has significant potential to improve the overall energy efficiency of industry which accounts for 25 Mt CO2/year emissions. With higher energy costs and a growing concern regarding environmental impact, it is highly desirable to recover the energy loss by recycling the spent steam. Membrane processes are being sought that effectively separate contaminants and recover the cleaned industrial waste steam. A high rate of diffusion of water vapour through a nonporous ionic membrane is reported as the key to producing high purity steam at temperatures approaching or exceeding 100 C (Spiegelman and Blethen, 2006). The preferred membrane polymer was a copolymer of perfluorinated ethylene and perfluorinated vinyl compound containing an acid group (sulphonic or carboxylic) or its salt. An example is the sulphonic acid copolymer Nafion, a copolymer of tetrafluoroethylene and perfluoro(4-methyl-3,6-dioxa-7-octene-1-sulphonic acid). A membrane that is substantially gas impermeable is desirable, and this copolymer has a permeability of water vapour that is more than three orders of magnitude greater than the permeability of CO2 or CO, and some six orders of magnitude greater than the permeability of oxygen or nitrogen. PVA/silica nanoparticle composite membranes have been tested for steam recovery (Hoang and Nguyen, 2009). Water vapour fluxes of 70e150 kg/m2 h were achieved at a differential pressure of 6 bars. A steam permeation plant that uses an LTA zeolite membrane has been installed at a sugar works. It treats 93 wt% bio-ethanol, obtaining a permeate that is below 0.1 wt% ethanol (Caro and Noack, 2008). Quite high fluxes are reported: 11.9, 14.9, 17.6 and 22.4 kg/m2 h at 100, 110, 120 and 130 C respectively.
2.8.
Miscellaneous water removal applications
Vapour permeation studies of mixtures propanol-methanolwater and propanol-methanol have been performed with a commercial hydrophilic PVA-polyacrylonitrile composite membrane (Will and Lichtenthaler, 1992). Also, the binary systems ammonia-water and methylamine-water were investigated using a commercial amine resistant PVApolysulphone composite membrane. With the exception of the non-aqueous propanol-methanol system the separation factors and fluxes obtained for the binary systems were sufficiently high for practical application. In the case of the ternary mixture, vapour permeation also showed a much better separation than pervaporation and a reasonable flux. This was not true for the non-aqueous system though, when the flux was extremely small at <1% of that for vapour permeation. In general, the vapour permeation separation factors were larger, but the flux was usually smaller than for pervaporation, so that separation via vapour permeation was possible when pervaporation failed. In practical applications, however, this is confined to those cases where the flux is sufficiently high.
263
3. Membranes used in vapour permeation processes A summary of the range of results viewed so far for water separation systems is given in Table 2. For enhanced passage of water vapour, introducing polar groups is a distinct advantage. There appears to be a further advantage in having an inorganic rather than a hydrophilic organic polymer membrane for the dehydration of ethanol, with silica and zeolite types being the best.
4. Comparison of vapour permeation and pervaporation Fig. 2 presents a schematic of the vapour permeation and pervaporation processes, which shows how closely aligned are the engineering aspects of the processes (Kujawaski, 2000). As mentioned earlier, vapour permeation is similar in principal to pervaporation, the only difference being that the feed for vapour permeation is a mixture of vapour or vapour and gases, whereas pervaporation involves a liquid feed and a liquidevapour phase change to achieve the separation (Neel, 1991). There is no phase change in vapour permeation, so the addition of heat equivalent to the enthalpy of vaporisation is not required in the membrane unit and there is no temperature drop along the membrane (Ito et al., 1997; Kujawaski, 2000). In vapour permeation, a slightly superheated vapour is usually employed to prevent condensation. However, the flux decreases with increasing superheating. Therefore, on the feed side of the membrane model, the local pressure should be as close as possible to the local saturation pressure in order to obtain a higher flux. This is very important with the feed side pressure losses of a vapour permeation model. Even a small degree of superheating may cause a significant decrease in flux, whereas selectivity seems not to be sensitive in this regard. Also, vapour condensation on the feed side of model cannot be avoided, depending on the operating conditions. The membrane hence has to be installed vertically so that if condensation occurs, the condensate flows downwards to the bottom of the cell instead of accumulating on the surface of the membrane. For a long membrane lifetime this may be essential (Chen and Lichtenthaler, 1995). A direct comparison of vapour permeation and pervaporation performance is not easy. The data available are summarised in Table 3. The application of NaA zeolite membranes from different sources using the two techniques for ethanol dehydration has been reported (Sato et al., 2008). With vapour permeation and 10 wt% aqueous ethanol a flux of 11 kg/m2 h is possible at 125 C, versus pervaporation mode results of 10 kg/ m2 h at 120 C (Richter et al., 2006). With vapour permeation, composite membranes of PVA coated onto Nylon-4 had a flux of 0.088 kg/m2 h in the dehydration of 10 wt% water/ethanol at 25 C, whereas in pervaporation mode the flux was much greater, at 0.42 kg/m2 h (Lee et al., 1992). There was a higher separation factor for vapour permeation (94 versus 14). The results can be explained by membrane swelling, which will not be significant for inorganic membranes, but for
264
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 5 9 e2 6 6
Table 2 e Water transfer data for membranes used for vapour permeation processes. Process Natural gas dehydration Drying of compressed air Flue gas dehydration Ethanol dehydration
Membrane
Water flux, kg/m2 h (Temp, C)
PEO/PBT copolymer Sulphonated poly (ether sulphone) Polydimethylsiloxane Cellulose acetate SPEEK PEBAX 1074 PVA PVA PVA/clay PVA on Nylon-4 PVA composite Polyamide & fluorinated version Fluorinated polyamide Modified polyimide Microporous silica
e e e e 0.2e0.46 (50) e e 0.44 0.38 0.088 (25) 0.7e5.5 (89) 0.20
NaY zeolite NaA zeolite
Isopropanol dehydration Acetonitrile dehydration Steam recovery
PVA Microporous silica Silica Nafion PVA inorganic hybrid LTA zeolite
membranes made from hydrophilic organic polymers direct contact of the membrane with the feed as in pervaporation will mean greater swelling and a higher flux rate. In some cases, as with certain polyimide membranes, the swelling can be so great that membrane performance is spoiled (Okamoto et al., 1992). This is in contrast with vapour permeation where the membrane is not in contact with the liquid feed. There is a larger permselectivity for water for vapour permeation, demonstrated by the significantly higher concentration of water in the permeate (Teng et al., 2000). A more detailed study of hybrid processes with distillation has shown that for water removal from acetonitrile with a silica membrane at high water concentrations or just for
Fig. 2 e Flow diagram for the vapour permeation (VP) and pervaporation (PV) processes (after Kujawaski, 2000).
Reference Metz et al., 2005 Sijbesma et al., 2008 Sijbesma et al., 2008 Sander and Soukup, 1988 Yeh et al., 2003 Lee et al., 1992 Jansen et al., 1992 Fan et al., 2002
0.38 (25) 0.64 (80e100) 0.1 (25) 10 (120) 20 (110) 11 (125) 2.5 (90) 5 (110) 7 (105) 21.5 (105) 3.9 (70)
Teng et al., 2000 Wu et al., 2002 van Veen et al., 2001
68 (106) 70e150 (153) 14.9 (110) 22.4 (130)
Spiegelman and Blethen, 2006 Hoang and Nguyen, 2009 Caro and Noack, 2008
Sato et al., 2008 Richter et al., 2006 Sommer and Melin, 2005 Gallego-Lizon et al., 2002 Fontalvo et al., 2005
overcoming the azeotropic composition, vapour permeation is preferred, but for water concentrations lower than the azeotropic amount, pervaporation is the better option (Fontalvo et al., 2005). Total and energy costs are slightly lower for vapour permeation than for pervaporation for the high water content mixtures for this application. In summary, vapour permeation, relative to pervaporation is capable of higher fluxes of water in certain systems (Sander and Soukup, 1988), but this may be diminished with highly swollen membranes, which favour greater fluxes for pervaporation (Fan et al., 2002) increases the permselectivity of water: for 10% aqueous ethanol at 25 C there is 97.7% water in the permeate, versus 54.5% for pervaporation (Teng et al., 2000) is more heat sensitive, as judged from Arrhenius plots (Fan et al., 2002) has a diffusion coefficient of the vapour phase that is about 400 times higher than that of the liquid phase (Sander and Janssen, 1991) a high selectivity is not necessarily accompanied by a lower permeability, as such a relationship does not hold always for water vapour in a mixture with a permanent gas, where most of the highly selective polymers also have a very high permeability the addition of heat equivalent to the enthalpy of evaporation is not required as there is no phase change (Ito et al., 1997)
265
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 5 9 e2 6 6
Table 3 e Comparison of water fluxes in vapour permeation and pervaporation processes. Process Ethanol dehydration, varying water contents
Membrane NaA zeolite PVA on Nylon-4 PVA composite Polyamide
Acetonitrile dehydration, distillation hybrid Acetic acid dehydration
Polyimide Polycarbonate Silica Silica
Vapour permeation water flux, kg/m2 h (Temp., C)
Pervaporation water flux, kg/m2 h (Temp., C)
10 wt%: 11 (125) 18 wt%: 15 (125) 10 wt%: 0.088 (25) 44 wt%: 0.087 (25) 10 wt%: 0.20 (55) 90 wt%: 0.14 (55) 10 wt%: 0.38 (25) 90 wt%: 0.285 (25) 10 wt%: 1.4a (75) 20 wt%: 0.16 (25) Higher for high water content 10 wt%: 3.6
10 wt%: 10 (120) 18 wt%: 13 (120) 10 wt%: 0.42 (25) 50 wt%: 0.90 (25) 10 wt%: 0.33 (55) 90 wt%: 0.22 (55) 10 wt%: 0.47 (25) 90 wt%: 0.293 (25) 10 wt%: 4.3a (75) 20 wt%: 0.18 (25) Higher for low water content 10 wt%: 7.2
Reference Richter et al., 2006 Lee et al., 1992 Fan et al., 2002 Teng et al., 2000 Okamoto et al., 1992 Wang et al., 2005 Fontalvo et al., 2005 Ishida et al., 2005
a Converted from reported permselectivity data.
requires a smaller membrane area (Sander and Janssen, 1991; Fontalvo et al., 2005) has less likelihood of membrane damage caused by impurities in the feed (Sander and Soukup, 1988) operation in the vapour phase eliminates concentration polarisation (Kujawaski, 2000) membrane life is expected to be longer because of less swelling of the membrane (Hayashi et al., 2000).
5.
Conclusions
Pervaporation is well established for dehydration applications, but it is likely that vapour permeation will intrude into this territory, its emphasis to date being on the extraction and recovery of organic species from a wide range of gas streams. The vapour permeation process is already used commercially for the dehydration of natural gas and organic solvents, especially alcohol, as well as for the removal of water from air. However, there may be further opportunities for the design of improved membranes of more ionic character, including polysalts. In a newer area, very high fluxes have been achieved now for steam separation and recovery at high temperatures using a composite PVA/silica nanoparticle membrane. Thus composite organic-inorganic membranes clearly have a promising future. The energy conservation and emission savings aspects of this process warrant further research to complete development. Although many advances have been made, the technology is still its early stages, and several challenges lie ahead, among them membrane stability at high temperatures. Again, composite membranes will have a role here.
references
Alter, H., 1962. A critical investigation of polyethylene gas permeability. Journal of Polymer Science 57, 925e935. Aptel, P., Cuny, J., Jozefowicz, J., Morel, G., Neel, J., 1972. Liquid transport through membranes prepared by grafting of polar
monomers onto poly(tetrafluoroethylene) films. I. Some fractionations of liquid mixtures by pervaporation. Journal of Applied Polymer Science 16 (5), 1061e1076. Baker, R.W., Wijmans, J.G., Kaschemekat, J.H., 1998. The design of membrane vapor-gas separation systems. Journal of Membrane Science 151 (1), 55e62. Bolto, B., Hoang, M., Xie, Z., 2011. A review of membrane selection for the dehydration of aqueous ethanol by pervaporation. Chemical Engineering and Processing: Process Intensification 50 (3), 227e235. Brinkmann, T., Dijkstra, M., Ebert, K., Ohlrogge, K., 2003. Improved simulation of a vapour permeation module. Journal of Chemical Technology & Biotechnology 78 (2e3), 332e337. Caro, J., Noack, M., 2008. Zeolite membranes - recent developments and progress. Microporous and Mesoporous Materials 115 (3), 215e233. Chen, Y., Lichtenthaler, R.N., 1995. In: Noble, R.D., Stern, S.A. (Eds.), Membrane Separations Technology: Principals and Application. Elsevier Science, pp. 101e105. Fan, S.-C., Li, C.-L., Wang, Y.-C., Lee, K.-R., Liaw, D.-J., Lai, J.-Y., 2002. Application of aromatic polyamide membranes for pervaporation and vapor permeation. Desalination 148 (1e3), 43e48. Fontalvo, J., Cuellar, P., Timmer, J.M.K., Vorstman, M.A.G., Wijers, J.G., Keurentjes, J.T.F., 2005. Comparing pervaporation and vapor permeation hybrid distillation processes. Industrial & Engineering Chemistry Research 44 (14), 5259e5266. Gallego-Lizon, T., Ho, Y.S., Freitas dos Santos, L., 2002. Comparative study of commercially available polymeric and microporous silica membranes for the dehydration of IPA/ water mixtures by pervaporation/vapour permeation. Desalination 149 (1e3), 3e8. Hayashi, Y., Yuzaki, S., Kawanishi, T., Shimizu, N., Furukawa, T., 2000. An efficient ethanol concentration process by vapor permeation through asymmetric polyimide membrane. Journal of Membrane Science 177 (1e2), 233e239. Hoang, M., Nguyen, C., 2009. Membrane and process for steam separation, purification and recovery. Australian Patent PCT/ AU2009/000386. Ishida, M., Tasaka, Y., Asaeda, M., 2005. A study on vapor permeation and pervaporation of acetic acid/water mixtures by porous silica membranes. Journal of Chemical Engineering of Japan 38 (1), 903e907. Ito, A., Feng, Y., Sasaki, H., 1997. Temperature drop of feed liquid during pervaporation. Journal of Membrane Science 133 (1), 95e102. Jansen, A.E., Versteeg, W.F., van Engelenburg, B., Hanemaaijer, J. H., ter Meulen, B.P., 1992. Methods to improve flux during
266
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 5 9 e2 6 6
alcohol/water azeotrope separation by vapor permeation. Journal of Membrane Science 68 (3), 229e239. Jonquie`res, A., Cle´ment, R., Lochon, P., Ne´el, J., Dresch, M., Chre´tien, B., 2002. Industrial state-of-the-art of pervaporation and vapour permeation in the western countries. Journal of Membrane Science 206 (1e2), 87e117. Kujawaski, W., 2000. Application of pervaporation and vapour permeation in environmental protetction. Polish Journal of Environmental Studies 9 (1), 13e26. Kyotani, T., Ikeda, T., Saito, J., Nakane, T., Hanaoka, T., Mizukami, F., 2009. Crystal structure of tubular Na LTA zeolite membrane used for a vapor permeation process: unusual distribution of adsorbed water molecules. Industrial & Engineering Chemistry Research 48 (24), 10870e10876. Lee, K.-R., Chen, R.-Y., Lai, J.-Y., 1992. Plasma deposition of vinyl acetate onto nylon-4 membrane for pervaporation and evapomeation separation of aqueous alcohol mixtures. Journal of Membrane Science 75 (1e2), 171e180. Leemann, M., Eigenberger, G., Strathmann, H., 1996. Vapour permeation for the recovery of organic solvents from waste air streams: separation capacities and process optimization. Journal of Membrane Science 113 (2), 313e322. Liu, S., Wang, F., Chen, T., 2001. Synthesis of poly(ether ether ketone)s with high content of sodium sulfonate groups as gas dehumidification membrane materials. Macromolecular Rapid Communications 22 (8), 579e582. Metz, S.J., Potreck, J., Mulder, M.H.V., Wessling, M., 2002. Water vapor and gas transport through a poly(butylene terephthalate) poly(ethylene oxide) block copolymer. Desalination 148 (1e3), 303e307. Metz, S.J., van de Ven, W.J.C., Potreck, J., Mulder, M.H.V., Wessling, M., 2005. Transport of water vapor and inert gas mixtures through highly selective and highly permeable polymer membranes. Journal of Membrane Science 251 (1e2), 29e41. Neel, J., 1991. Introduction to Pervaporation in: Pervaporation Membrane, Separation Process. Elsevier, Amsterdam. Ohlrogge, K., Nitsche, V., Wind, J., 2002. Arrangement for removing water vapour from pressurised gases or gas mixtures, US Patent 6,485,545. Okamoto, K., Tanihara, N., Watanabe, H., Tanaka, K., Kita, H., Nakamura, A., Kusuki, Y., Nakagawa, K., 1992. Vapor permeation and pervaporation separation of water-ethanol mixtures through polyimide membranes. Journal of Membrane Science 68, 53e63. Richter, H., Voigt, I., Ku¨hnert, J.-T., 2006. Dewatering of ethanol by pervaporation and vapour permeation with industrial scale NaA-membranes. Desalination 199 (1e3), 92e93. Sander, U., Janssen, H., 1991. Industrial application of vapour permeation. Journal of Membrane Science 61, 113e129.
Sander, U., Soukup, P., 1988. Design and operation of a pervaporation plant for ethanol dehydration. Journal of Membrane Science 36, 463e475. Sato, K., Sugimoto, K., Nakane, T., 2008. Mass-production of tubular NaY zeolite membranes for industrial purpose and their application to ethanol dehydration by vapor permeation. Journal of Membrane Science 319 (1e2), 244e255. Scott, K., 1998. Handbook of Industrial Membranes. Elsevier Advanced Technology, London. Sijbesma, H., Nymeijer, K., van Marwijk, R., Heijboer, R., Potreck, J. , Wessling, M., 2008. Flue gas dehydration using polymer membranes. Journal of Membrane Science 313 (1e2), 263e276. Sommer, S., Melin, T., 2005. Influence of operation parameters on the separation of mixtures by pervaporation and vapor permeation with inorganic membranes. Part 1: dehydration of solvents. Chemical Engineering Science 60 (16), 4509e4523. Spiegelman, J.J., Blethen, R.D., 2006. Method of producing high purity steam. US Patent PCT/US2006/035790. Teng, M.-Y., Lee, K.-R., Fan, S.-C., Liaw, D.-J., Huang, J., Lai, J.-Y., 2000. Development of aromatic polyamide membranes for pervaporation and vapor permeation. Journal of Membrane Science 164 (1e2), 241e249. van Veen, H.M., van Delft, Y.C., Engelen, C.W.R., Pex, P.P.A.C., 2001. Dewatering of organics by pervaporation with silica membranes. Separation and Purification Technology 22-23, 361e366. Van Wijk, H.F., Jansen, A.E., 1990. Method and Membrane for the Removal of Water Vapour from a Gas/Vapour Mixture by Means of Vapour Permeation. Wang, Z., Chen, T., Xu, J., 2001. Gas and water vapor transport through a series of novel poly(aryl ether sulfone) membranes. Macromolecules 34 (26), 9015e9022. Wang, Y.-C., Teng, M.-Y., Lee, K.-R., Lai, J.-Y., 2005. Comparison between the pervaporation and vapor permeation performances of polycarbonate membranes. European Polymer Journal 41, 1667e1673. Will, W., Lichtenthaler, R.N., 1992. Comparison of the separation of mixtures by vapor permeation and by pervaporation using PVA composite membranes. II. The binary systems ammoniawater, methylamine-water, l-propanol-methanol and the ternary system 1-propanol-methanol-water. Journal of Membrane Science 68, 127e131. Wu, Y., Peng, X., Liu, J., Kong, Q., Shi, B., Tong, M., 2002. Study on the integrated membrane processes of dehumidification of compressed air and vapor permeation processes. Journal of Membrane Science 196 (2), 179e183. Yeh, J.-M., Yu, M.-Y., Liou, S.-J., 2003. Dehydration of waterealcohol mixtures by vapor permeation through PVA/ clay nanocomposite membrane. Journal of Applied Polymer Science 89 (13), 3632e3638.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Review
State of the art and review on the treatment technologies of water reverse osmosis concentrates A. Pe´rez-Gonza´lez, A.M. Urtiaga, R. Iba´n˜ez, I. Ortiz* Dpto. Ingenierı´a Quı´mica y QI. ETSIIyT, Universidad de Cantabria, Av. de los Castros s/n, 39005 Santander, Spain
article info
abstract
Article history:
The growing demand for fresh water is partially satisfied by desalination plants that
Received 5 July 2011
increasingly use membrane technologies and among them reverse osmosis to produce
Received in revised form
purified water. Operating with water recoveries from 35% to 85% RO plants generate huge
19 October 2011
volumes of concentrates containing all the retained compounds that are commonly dis-
Accepted 20 October 2011
charged to water bodies and constitute a potentially serious threat to marine ecosystems;
Available online 31 October 2011
therefore there is an urgent need for environmentally friendly management options of RO brines.
Keywords:
This paper gives an overview on the potential treatments to overcome the environ-
Reverse osmosis concentrates
mental problems associated to the direct discharge of RO concentrates. The treatment
Volume reduction
options have been classified according to the source of RO concentrates and the maturity of
Pollutant load reduction
the technologies. For the sake of clarity three different sources of RO concentrates are
Salt recovery
differentiated i) desalination plants, ii) tertiary processes in WWTP, and iii) mining
Nutrient recovery
industries. Starting with traditional treatments such as evaporation and crystallization other technologies that have emerged in last years to reduce the volume of the concentrate before disposal and with the objective of achieving zero liquid discharge and recovery of valuable compounds from these effluents are also reviewed. Most of these emerging technologies have been developed at laboratory or pilot plant scale (see Table 1). With regard to RO concentrates from WWTP, the manuscript addresses recent studies that are mainly focused on reducing the organic pollutant load through the application of innovative advanced oxidation technologies. Finally, works that report the treatment of RO concentrates from industrial sources are analyzed as well. ª 2011 Elsevier Ltd. All rights reserved.
Abbreviations: AA, activated alumina; AOP, advanced oxidation process; BAC, biological activated carbon; BDD, boron doped diamond; BOD, biological oxygen demand; BWRO, brackish water reverse osmosis; CDI, capacitive deionization; CESP, chemically-enhanced seeded precipitation; COD, chemical oxygen demand; DOC, dissolved organic carbon; ED, electrodialysis; EDBM, electrodialysis with bipolar membranes; EDR, electrodialysis reversal; EFC, eutectic freeze crystallization; FBAR, fluidized bioactive adsorber reactor; FBC, fluidized bed crystallization; FO, forward osmosis; GAC, granular activated carbon; ICD, intermediate chemical demineralisation; MCr, membrane crystallizer; MDC, membrane distillation coupled with crystallization; PDO, Petroleum Development Oman; PLE, polimeric ligand exchange; RO, reverse osmosis; SCR, solid contact reactor; TAN, total ammoniacal nitrogen; TOC, total organic carbon; UF, ultrafiltration; VMD, vacuum membrane distillation; WAIV, wind aided intensified evaporation; WWTP, wastewater treatment plant; ZLD, zero liquid discharge. * Corresponding author. Tel.: þ34 942 20 15 85; fax: þ34 942 20 15 91. E-mail address:
[email protected] (I. Ortiz). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.046
268
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
Contents 1. 2.
3.
4. 5.
1.
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . RO concentrates from desalination plants: treatment alternatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Solar evaporation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Emerging technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1. Systems aiming to achieving Zero Liquid Discharge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2. Emerging technologies for salts recovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . RO concentrates from wastewater treatment plants: treatment alternatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1. Advanced oxidation processes (AOPs) applied to RO concentrates from WWTPs . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1. Ozonation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2. Fenton process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.3. Photocatalysis and photooxidation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.4. Sonolysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.5. Electrochemical oxidation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Recovery of nutrients and salts from RO concentrates from WWTP’s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . RO concentrates from industrial water sources: treatment alternatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Introduction
Reverse osmosis is a membrane technology widely applied in water desalination, production of potable water and more recently in tertiary wastewater treatment. This technology has the advantages of membrane processes such as modular construction and small footprint, which allow the combination with other treatment processes (Chelme-Ayala et al., 2009). The technology employs semi-permeable membranes that allow to separate a solution into two streams: permeate, containing the purified water that passes through the membrane, and concentrate, the portion that contains salts and retained compounds and therefore needs a suitable and environmentally friendly management option (Mauguin and Corsin, 2005). The characteristics of the waste stream, named concentrate, retentate or brine, depend on the quality of the feed water, the quality of the produced water (recovery varies from 35% to 85%), the pretreatment method (added chemicals) and cleaning procedures used (Chelme-Ayala et al., 2009; Greenlee et al., 2009; Squire et al., 1997; Watson, 1990). Then constituent concentrations in the retentate are found to be double or higher than that in feed water (Chelme-Ayala et al., 2009). Brine disposal in coastal desalination plants has been solved by direct discharge to seawater. In desalination plants, generation of brine is about 55% of collected seawater (Meneses et al., 2010). Recent estimates suggest that up to 25 million m3 of desalinated water is produced daily around the world (Lattemann and Ho¨pner, 2008). Representative examples of large membrane reverse osmosis seawater desalination plants with ocean outfalls for concentrate discharge are the 330,000 m3/day plant in Ashkelon, Israel; the 136,000 m3/ day Tuas Seawater Desalination Plant in Singapore; the 64,000 m3/day Larnaka Desalination Facility in Cyprus, and the majority of the large desalination plants in Spain, Australia and the Middle East (Voutchkov, 2011).
268 271 271 272 273 276 276 277 277 277 278 278 278 279 279 280 281 281
Roberts et al. (2010) have reported an outstanding literature review on the ecological impacts of desalination plants concluding that there is a widespread belief and recognition that brine discharge poses a potentially serious threat to marine ecosystems. Laboratory-based experiments, toxicological investigations and manipulative field experiments clearly demonstrate the potential for brines and their constituents to illicit adverse impacts on aquatic organisms when present at sufficient concentrations. In addition to the destructive saline properties of the concentrate, in the case of thermal desalination, the brine is usually hotter than the local recipient water body, a circumstance that has also been shown to cause further environmental damage, especially to fragile ecosystems such as corals. Furthermore, during pre- and post- treatment processes a variety of chemical agents are added to enhance flocculation, prevent foaming or to avoid membrane deterioration (Meerganz von Medeazza, 2005). Research on the development of effective anti-scalants with no biological effects may assist in the production of less toxic brines in the future (Roberts et al., 2010). But with the traditional option of direct release to seawater, those components are discharged along with the brine as well as certain metals (copper, nickel, iron, chromium, zinc, etc.) coming from thermal corrosion processes. Thus, as a result of brine direct discharge the most pronounced effects on receiving waters are eutrophication, pH value variations, and accumulation of heavy metals as well as sterilizing properties of disinfectants (Meerganz von Medeazza, 2005). Due to these negative effects, direct disposal to seawater of RO concentrates is doomed to disappear. One mechanism that has been applied to reduce adverse environmental effects of brine relies on its dilution with power plant cooling waters (Einav and Lokiec, 2003). Brines can be also diluted with natural seawater or municipal wastewaters to reduce salinity prior to discharge (Meneses et al., 2010; Roberts et al., 2010) although the effects of
269
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
disposal of diluted concentrates can also affect to sensitive species so the solution should be appropriate to local conditions (Meneses et al., 2010). To minimise areas of ecological impacts and limit the intrusion of brines into seawater intake areas the optimum infrastructure design and construction has been studied (Altayaran and Madany, 1992). Dispersion models are used to choose the discharge point with less environmental impact (Del Bene et al., 1994) but these structures appear to have limited success and the environmental damage is also incurred. Groundwater desalination has been reported as a more environmentally friendly alternative to seawater desalination because the discharged brines are less salty than those produced from seawater desalination (Mun˜oz and Ferna´ndez-Alba, 2008). Traditionally, options for disposal of the concentrates from inland desalination plants have been deep well injection and surface water discharge, or concentration of brines in evaporation ponds (Ahmed et al., 2001; Arnal et al., 2005; ChelmeAyala et al., 2009; Gabelich et al., 2010; Greenlee et al., 2009; Malaeb and Ayoub, 2011; Muniz and Skehan, 1990). Similar options of RO concentrates management have been used in the case of potable water production from brackish groundwater (Malaxos and Morin, 1990; Squire, 2000; Van Der Bruggen et al., 2003). Inland plants have to solve the problem of concentrate disposal without the possibility of their
discharge to seawater, so the development of other management options is an urgent demand. The cost of brine disposal is another important issue that must be taken into account. In coastal desalination plants the cost of brine disposal to the sea ranges from 5 to 33% of the total cost of desalination (Ahmed et al., 2001) being the disposal cost higher for inland desalination plants than for coastal plants where brine is discharged into the sea (Arnal et al., 2005). In both cases, the cost of brine disposal depends on the brine characteristics, level of treatment before disposal, disposal method and brine volume (Arnal et al., 2005; Malaeb and Ayoub, 2011). With regard to inland desalination plants, depending on the salinity of the concentrate, the cost associated to brine disposal could be considerably higher than in coastal plants (Stanford et al., 2010). Among the wide diversity of RO applications, this review is mostly focused on the conventional and emerging treatments for RO concentrates generated in desalination plants. Furthermore due to the increasing trend of using RO units in the tertiary treatments of WWTPs the most suitable treatment alternatives for this application are also addressed. Table 2 displays the average chemical characteristics of representative RO concentrates from both applications. Conductivity, total dissolved solids and chloride concentration present significantly higher values in RO brines from desalination
Table 1 e Evaluation of viability of treatment technologies applied to RO concentrates. RO concentrates source Desalination plants
Wastewater treatment plants
Desalination and wastewater treatment plants
Other industrial sources
Technology
Technological maturity
Solar evaporation (Evaporation ponds)
Industrial application
WAIV wind aided intensified evaporation Membrane distillation
Pilot plant scale
Forward osmosis
Laboratory level
Liquideliquid extraction
Laboratory level
Ozonation
Laboratory level
Fenton processes
Laboratory level
Photocatalysis and photooxidation
Laboratory level
Sonolysis Electrochemical oxidation
Laboratory level Laboratory level
Adsorption
Laboratory level
Electrodialysis
Pilot plant scale
Crystallization SAL-PROC
Laboratory level Patented process
EFC
Pilot plant scale
Laboratory level
Operation drawbacks and economic considerations Large land areas Low productivity Moderate investment and maintenance cost Industrial feasibility not proved Moderate investment cost Difficult operational control Scaling and fouling Moderate energy consumption Use of drawn solution Moderate energy consumption Several treatment stages Extractants consumption High chemical dosage High investment cost High chemical dosage Moderate investment cost High chemical dosage Moderate energy consumption High energy consumption High energy consumption Moderate investment cost Regeneration of exhausted resins (High chemicals consumption) Maintaining energy efficiency with high saline concentrates Precipitation on the membrane High capital and operation cost Stricted operational conditions Applicability to RO concentrates not completely proved Complex control of operation Moderate energy consumption
270
Table 2 e Characterization of RO concentrates. RO concentrates from desalination plants RO concentrates from wastewater treatment plants Lee Badruz- Kumar Ji Martinetti Korngold Macedonio Hajbi Ahmed Ersever Zhou Westerh- Van Van Pe´rez Radjenovic Chaplin Lee Hege Hege et al., et al., et al., off et al., et al., et al., et al., zaman et al., et al., et al., et al., et al., et al., et al., et al., et al., et al., 2010,a 2011,a 2010 2009a 2009b,a et al., 2007 2010 2009 2009 2011 2010 2003,b 2007a,a 2011,a 2009 2002 2004,a 2009 e e e 2 e e e e 50,200 e
e e e e e e e 7 e e
e e e e e e e 8 e e
e e e e e e e e e e
e e e e e e e 4 e 13,500
e e e e e e e 7.2 28,000 33,000
e e e e e e e 6.2 17,124 25,984
e e e 40 e e 80 e 3965 e
3.2 e 60 18 10 e e 6.9 1129 1705
e 40 138 e e e e 7 5560 10,000
e e 158 e e 37.5 e 8.2 e 3990
e e 184 e e 34 e 8.3 e 4340
2 23.3 e e e e 121.8 7.8 e 3250
e 57.2 e e e e e 7.6 e 4110
e 19.2 e e e e e e e 4450
e e 64.6 18.4 e e e 7.2 1218 1972
e e e 24.5 e e e 7.5 1276 1990
e e e 25 e e e 2950 3600
e e e e e e e e 2750 e
e
470
142
e
e
e
345
1006
e
1110
e
e
1233
e
289
e
e
520
e
e
116
72
103
56
e
15.6
e
e
e
e
e
e
e
e
e
e
120
120
Cations Naþ, mg/L Mg2þ, mg/L Kþ, mg/L Ca2þ, mg/L Fe2þ, mg/L Mn2þ, mg/L
15,500 2020 e 625 e e
991 318 e 1032 e e
5130 386 e 819 e e
e 468 e 1020 e e
2084 245 79 540 e e
5120 770 e 2080 e e
4160 370 134 1537 0.4 0.2
e e e e e e
203 7 62 65 e e
e e e e e e
e e e e e e
e 32 e 159 e e
e e e e e e
e e e e 0.3 230.5
435 122 22.6 306 e e
240.9 7.1 65.3 110.1 e e
226.9 11.5 38.4 63.8 e e
580 9.7 88 96 0.1 54
910 5 24 5 e e
Anions Cl, mg/L SO2 4 , mg/L NO 3 , mg/L NO 2 , mg/L PO3 4 , mg/L HCO 3 , mg/L
28,800 3060 e e e 199
2823 1553 e e 0.4 576
8960 1920 e e 2 223
6710 2688 e e e e
4068 2160 e e 0.04 e
14,170 5920 e e e e
8369 2334 14.6 e e 421
900 1240 e e 20 e
256 217 91 2 39 e
e e e e e e
592.9 e e e e e
700 e e e e e
479 443 e e e e
1.4 240 e e e e
220.1 1584 3.1 e e e
267.1 218.4 88.5 8.3 34.7 e
333.2 159.1 60.2 e 21.3 e
684 468 296 e 10 e
2811 1437 269 e e e
a Average value. b Average composition of reverse osmosis concentrates from desalination plants of Bahja, Rima, Nimr and Marmul of Petroleum Development Oman.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
Parameters Turbidity, NTU DOC, mg/L COD, mg/L TOC, mg/L TKN, mg/L TAN, mg/L NHþ 4 , mg/L pH TDS, mg/L Conductivity, mS/cm Alkalinity, mg/L as CaCO3 SiO2, mg/L
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
plants than in RO concentrates from wastewater treatment plants. Regarding sulfate, its concentration levels in concentrates from WWTP’s are similar to the chloride content while in concentrates from desalination plants, sulphate concentration is 2e8 times lower than the chloride one. Thus, treatment technologies must be adapted to the source of RO concentrates in order to eliminate the problem associated to their composition. For example, concentrates from wastewater treatment plants present higher organic pollutant load but lower salinity than desalination plants RO concentrates, therefore advanced treatments (like advanced oxidation processes (AOPs)) are required to destroy the most persistent pollutants. Thus, this paper gives an overview on the technologies aimed at reducing the environmental impact of the disposal of RO concentrates according to their origin, the level of development of the technologies and the final objective of the treatment.
2. RO concentrates from desalination plants: treatment alternatives Traditional management of RO concentrates from desalination plants is mainly conditioned by the location of the plant. In coastal desalination plants, RO concentrates are directly discharge to seawater, while in inland plants the traditional option consists on reducing the concentrate volume prior to disposal (Tang and Ng, 2008). Evaporation techniques have been widely applied to concentrate brines, since their application allow to obtain a solid waste easier to be managed than the original waste and a decontaminated liquid flow that can be directly discharged or even reused (Arnal et al., 2005). Due to the adverse effects of brine disposal together with its associated costs, current research is focused on reducing the impact of RO concentrates by reducing the volume and/or by diminishing the pollutant load of these concentrates. Besides, the beneficial use of brine byproducts is also studied and includes the technical feasibility of isolating salts of the required morphology and purity (Stanford et al., 2010). Recovering commercial byproducts from RO concentrates would be the optimum treatment option, as it solves the environmental problem of concentrate disposal, as well as the economic profitability of reverse osmosis is improved at the same time.
2.1.
Solar evaporation
Solar evaporation is one of the techniques considered as a common solution for brine disposal (Chelme-Ayala et al., 2009; Greenlee et al., 2009), especially for inland desalination plants in arid and semi-arid areas (Ahmed et al., 2000). The reverse osmosis concentrate is placed in a shallow lined pond which allows the water to evaporate naturally by using the solar energy; after water evaporates the salt is either left in the ponds or removed for disposal (Katzir et al., 2010). Evaporation ponds are relatively easy to construct, require little operation attention compared to mechanical equipment and except for pumps to convey the wastewater to the pond, no mechanical equipment is required (Mickley, 2001). Pond size includes two components: surface area and depth. Pond depths ranging
271
from 25 to 45 cm are considered optimal for maximizing the rate of evaporation (Ahmed et al., 2000). However, evaporation ponds are scarcely used because they require large land areas, especially if they are located in places with low evaporation rates, whereas they pose a high potential for contamination of groundwater coupled to the risk of leakage underneath the pond (Katzir et al., 2010). Evaporation ponds ranging from 13.6 to 34.3 ha are used for disposal purposes of the concentrate in the desalination plants located in the central region of Saudi Arabia (Ahmed et al., 2000). Besides this extensive land use, evaporation ponds have been criticized because they do not recover the evaporated water (Mickley, 2001). Additionally, the productivity of the process is quite low (around 4 L/m2d). This drawback can be overcome by using wet surfaces (capillaries or clothes) exposed to wind action to increase the evaporation surface. Arnal et al. (2005) tested different adsorbent materials in order to improve the water evaporation rate of brines from desalination plants. According to the experimental results, the most suitable adsorbent for natural evaporation was a rectangular cloth made of cellulose (65%) and cotton (35%). It was also concluded that air velocity improved natural evaporation, although the overall efficiency was limited by the blowing of solids onto the surface of water. Following this investigation line, a proprietary technique WAIV (Wind Aided Intensified eVaporation) was developed as alternative to natural evaporation. WAIV is a less land intensive method to reduce brine volumes by the use of the drying power of the wind without generating small droplets that can cause salt drift. This configuration employs recirculation of brines as falling films on vertical hydrophilic surfaces that are largely mounted parallel to the wind direction (Macedonio et al., 2011). When exposed to the dry winds of semi-arid regimes, these surfaces are cooled to near the wet bulb temperature and the difference of temperature between the warmer wind and the cold-water surface drives heat flux to the wetted surface. The vapour pressure gradient drives the evaporation mass transfer from the surface. Previous studies on normal desalination brines in a pilot unit with 31e43 m2 wetted evaporation surfaces showed that evaporation rates (L/D-m2) can be improved by 50e90% relative to open ponds while surface loading reached 15e30 m2/m2 footprint. For example, the Sabha seawater RO plant located near the Red Sea port of Eilat sends its rejected brine to salt production evaporation ponds. A pond area of 700,000 m2 is needed in order to handle about 5000 m3 per day of concentrate. It was estimated (Katzir et al., 2010) that a WAIV installation could reduce the required land area by an order of magnitude. To evaluate the performance of several evaporation surfaces, Gilron et al. (2003) tested a pilot unit using brines from Mekorot’s Sabha B desalination installation in Eilat. The total dissolved solids (TDS) content of the brine was around 16,000e18,000 ppm and it was supersaturated with respect to sparingly soluble calcium salts. Better results were obtained when using materials with no internal surfaces (netting) that are less susceptible to plugging than those with internal surfaces (nonwoven geotextiles). The WAIV technology was further studied in order to recover salts that can be potentially useful as raw materials (Katzir et al., 2010). Concentrates obtained in the RO and ED treatment of saline groundwater were
272
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
evaporated. Gypsum precipitation on the evaporation surfaces and in the feed basin was observed, leading to an enrichment of magnesium relative to sodium ion in the resulting concentrate brine. This technology can be an interesting option whenever precipitated salts are recovered. Although WAIV presents advantages with respect to evaporation ponds in land use, the availability of this technology has been only demonstrated at laboratory scale.
2.2.
Emerging technologies
Conventional treatments, like concentration in evaporation ponds, have several disadvantages such as extensive land use and low productivity. Thus, investigation on new options to improve the management of reverse osmosis concentrates is a current demand. In this section we review the potential of several emerging technologies focused on the treatment of RO concentrates from desalination plants, classified according to the operation objective. Membrane distillation has been studied as alternative for the processing of highly concentrated aqueous solutions. Vacuum Membrane Distillation (VMD) is an evaporative technology that uses a membrane to support the liquidevapour interface (Urtiaga et al., 2001). The main advantages of membrane distillation over conventional distillation processes are that the operating temperature is in the range of 60e80 C and that the membranes provide a high contact area per unit of equipment volume, allowing very compact installations and reduced footprint. Mericq et al. (2010) simulated an increase from 40% to 89% in the water recovery of a 40.000 m3/ day RO plant fed with seawater together with a reduction of the brine volume by a factor of 5.5 after coupling RO with VMD. The model used for this simulation was based on experimental results obtained at a bench-scale batch unit. In the experiments developed with real RO brines organic fouling or biofouling was not observed, at least at the time scale of a few hours. However, for high salt concentrations, calcium scaling occurred although its impact on the permeate flux was very limited. Scaling effects of CaCO3 and CaSO4 in membrane distillation operated at high seawater concentrations have been deeply studied (Curcio et al., 2010; He et al., 2008), being the main conclusions that scaling reduces the transmembrane permeate flux as membrane promotes heterogeneous nucleation mechanisms. Also, the detrimental effect of the presence of humic acid substances became significant at higher concentration factors. Nevertheless, calcium scaling was found to be reversible after appropriate washing and chemical cleaning. Calcium scaling could also be controlled by accelerated precipitation softening prior to the direct contact membrane distillation of the RO concentrate obtained in a drinking water treatment facility (Qu et al., 2009, 2010). Membrane distillation coupled with crystallization (MDC) was investigated in a bench-scale plant operated with brines discharged from a seawater reverse RO unit (Ji et al., 2010). The major advantage of the MDC process is its capability to concentrate the salt up to the supersaturated state, which allows its crystallization (Gryta, 2002). In MDC, the high contact area provided by hollow fiber membranes allows to achieve reliable evaporation fluxes at moderate temperatures
(40e50 C) with energy consumption of about 15e20 kW h/m3, while conventional evaporative equipments for NaCl crystallization need to operate at temperatures higher than 70 C, with specific energy consumption of about 30 kW h/m3. However the main problem of membrane distillation is that the technology is not yet commercially available at industrial scale (Singh, 2009). In a recent study (Ji et al., 2010) seawater was collected from the Tyrrenian Sea in Amantea (Calabria) and further processed in an RO lab-scale unit. Lime/soda ash softening was applied to the RO concentrate to reduce calcium and magnesium hardness and to limit scaling problems. Experiments were focused on NaCl crystallization, obtaining a production of 17 kg/m3 of NaCl crystals, representing 34% of the total content of dissolved solids in the brine. In conclusion, membrane distillation-crystallisation was reported to be a feasible technique to concentrate RO brines, achieving water recovery efficiency higher than 90%. Applying electrodialysis (ED) to brine effluents, their salt concentration is increased from 0.2 e 2% to 12e20% with an energy requirement of 1.0e7.0 kW h/m3 in contrast to approximately 25 kW h/m3 needed by thermal evaporation (Korngold et al., 2005, 2009). In ED operation, there are two major technical problems to be overcome: firstly to find solutions for operating at concentrations up to 20% of salts without significantly diminishing the energy efficiency and secondly to prevent CaSO4 precipitation on the membrane. This last drawback can be prevented by continuous removal of gypsum from the brine in a separate precipitator. ED experiments were carried out using the brine obtained from the RO desalination of brackish water as the feed to an ED pilot plant combined with a gypsum precipitation step (Korngold et al., 2005, 2009), as it is depicted in Fig. 1. After 600 h of operation, the amount of CaSO4 that precipitated was approximately 0.15 eq. CaSO4/h, and scaling on the membrane was prevented. Besides, significant acceleration of the precipitation can be achieved by seeding the oversaturated solution with 10e25 g CaSO4/L. A very low concentration (0.5e1 ppm) of antiscalant (Calgon) diminished the seeding effect on gypsum precipitation and increased the onset of precipitation time. Forward osmosis (FO) is presented as an innovative technique to reduce brine volume, having as main advantage lower energy requirement than RO. FO utilises a highly concentrated solution generally referred to as the draw
Fig. 1 e Simplified scheme of ED pilot plant combined with a gypsum precipitation step (Korngold et al., 2005).
273
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
has as inconvenient the need of a draw solute to create an effective driving force that allows water flux, and besides the economic feasibility of the technology has not been demonstrated (Singh, 2009).
2.2.1.
Fig. 2 e General scheme of Zero Liquid Discharge systems.
solution to generate an osmotic pressure differential across the membrane, thus resulting in the transport of water from the less concentrate feed stream to the highly concentrated draw solution. A wide range of draw solutions can be used: sulphur dioxide, aluminium sulfate, fructose, ammonium bicarbonate, etc (Neilly et al., 2009). Tang and Ng (2008) tested three types of membranes using a concentrate feed of 1 Me2 M NaCl, model brine concentrations, and a draw solution of fructose 5 Me6 M, which was kept constant by adding a suitable amount of fructose periodically. It was found that fructose is not an ideal draw solute as it is not efficient in establishing a high effective driving force. However, using as draw solution NaCl (50 g/L) with less saline concentrates (w0.25 M of chloride), recovery rates larger than 90% were achieved (Martinetti et al., 2009). In both cases, this technique
Systems aiming to achieving Zero Liquid Discharge
One of the most innovative investigation line focused on reducing the concentrate volume to the highest point is called Zero Liquid Discharge (ZLD), which is aimed to achieve the maximum water recovery, through several stages of treatment in order to avoid liquid effluent disposal. Studies related to this objective have been classified according to the treatment scheme as Basic ZLD systems, Type A, Type B and Type C scheme (see Fig. 2). Table 3 summarizes these schemes alongside the global recoveries achieved with the application of these treatment combinations. The basic combination is a RO tandem (Primary RO þ Secondary RO). Al-Wazzan et al. (2003) increased the product water recovery using brine staging, and besides a reduction of 37% in specific power consumption in a pilot scale unit was achieved. This brine staging unit was in operation during almost 5 months, being their availability of 97%. The use of tandem RO was also studied by several authors (Ning and Troyer, 2009; Ning et al., 2010; Singh, 2009), who asserted that there is a need of intermediate treatment in tandem processes to increase the recovery and avoid precipitation. These systems achieve high recoveries, although the zero discharge goal is not accomplished. Fouling and precipitation of scaling compounds are two of the main limiting factors for the RO recovery, as mentioned above. Antifouling and anti-scalants compounds are usually added to feedwaters in order to avoid the diminishing of
Table 3 e Treatment processes aimed at Zero Liquid Discharge. Treatment stages Basic ZLD scheme
Primary RO
Type A ZLD scheme
Primary RO
Type C ZLD scheme
Secondary RO
88e99%
References Al-Wazzan et al., 2003; Ning and Troyer, 2009; Ning et al., 2010; Singh, 2009
Ion exchange Resins for silica removal Precipitation treatment (alkaline-induced)
Secondary RO
Secondary RO
87e97%
Gabelich et al., 2007, 2011; Ning et al., 2006; Rahardianto et al., 2010
RO Primary RO
WAIV Secondary RO
77e89%
Macedonio et al., 2011 Ning and Tarquin, 2010
RO Primary RO
Secondary RO
Membrane crystallizer Precipitation treatment (calcium sulfate, magnesium, iron and silica) ED/EDR EDR þ UF
WAIV
92% >98%
Turek et al., 2009 Oren et al., 2010
Secondary RO
Evaporation
80e90%
Secondary RO
Brine concentrator þ Pond
w100%
Mohammadesmaeili et al., 2010a, b Bond and Veerapaneni, 2008
Primary RO
Type B ZLD scheme
Overall recovery
Primary RO Primary RO
Lime-soda treatment Intermediate treatment (several options)
Acevedo et al., 2010
274
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
Fig. 3 e Bipolar membrane electrodialysis scheme for obtaining mixed acid and basis.
membrane performance. However, the addition of these compounds can have negative effects in the post-treatment of RO concentrates. Greenlee et al. (2011, 2010a, 2010b) studied the effects of anti-scalants in the precipitation of scaling salts, like calcium carbonate, as intermediate stage in a tandem RO system. It was demonstrated that anti-scalants decreased the precipitate particle size and change the shape of the particles. Following the objective of increasing the global recovery of RO operation, several authors (Acevedo et al., 2010; Gabelich et al., 2007, 2011; Ning et al., 2006; Rahardianto et al., 2010) proposed RO tandem processes combined with an intermediate unit in order to eliminate the most problematic substances such as foulants and scalants (Type A scheme in Fig. 2). Acevedo et al. (2010) evaluated the possibility for silica concentration reduction by using ion exchange columns with strong basic anionic resins, and reported that the removal of silica strongly
depends on the pH being available at pH over 8.5, when it also takes place calcium and magnesium precipitation. Therefore, Acevedo proposed as another phase of the research a train of cationic and anionic resins to first remove hardness and allow to increase the pH without precipitation. Studies about salts and silica removal were also developed by Gabelich et al. (2007, 2011) who proposed intermediate chemical demineralization (ICD) that consisted on a solid contact reactor (SCR) followed by a filtration stage. Ning et al. (2006) used a limesoftening unit to reduce the limiting foulants of silica and barium sulphate. With the same objective of eliminating antiscalants and fouling compounds, Rahardianto et al. (2010) proposed a two step chemically-enhanced seeded precipitation (CESP) for accelerated desupersaturation. They reported that the resulting lab and pilot plant data would serve as possible scale-up information although control and operation strategy had to be developed to maintain consistently high demineralization efficiency at industrial level. Other schemes have been investigated with the aim of recovering both salts and water (Neilly et al., 2009). These systems are characterized for having one or more final steps of post-treatment of the concentrate obtained from the second RO stage (Type B scheme in Fig. 2) (Macedonio et al., 2011; Ning and Tarquin, 2010; Oren et al., 2010; Turek et al., 2009; Zhang et al., 2010). Combining WAIV technology with a membrane crystallizer (MCr), Macedonio et al., 2011 achieved recovery factors as high as 88.9% and capital cost of WAIV was reduced in a 64% in comparison to conventional ponds. Ning and Tarquin (2010) studied the fractional crystallization of salts before thermal evaporation of water. In turn, Turek et al. (2009) studied the combination of RO with electrodialysis reversal (EDR) achieving 91.6% of water recovery, although the synthetic water used had no comparable composition with real RO concentrates. Oren et al. (2010) proposed a system based on further concentration of the BWRO brine with EDR and WAIV. Scaling of ED unit was prevented by acidification, operating the electrodialysis in a reversal mode, and a side loop crystallizer which prevented build-up of scaling components. The super-concentrate from the EDR unit was further concentrate in a wind powered WAIV unit that brought final TDS to more of 30%, and showed
Fig. 4 e Experimental set-up for treatment of RO brine using ozone D BAC column D CDI (Lee et al., 2009a, b).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
275
Fig. 5 e Salt recovery scheme used by Mohammadesmaeili et al. (2010a, b).
promise as a method to recover mineral byproducts. Cost effectiveness of these process alternatives need to be assessed in order to determine their availability and profitability. The third type of ZLD scheme is based on combining type A and type B ZLD scheme. Mohammadesmaeili et al. (2010a, b) proposed a system with and intermediate stage of lime softening and several stages of evaporationecrystallization after the secondary RO (see Fig. 5). Results of this research are explained in more detail in Section 3.2. This type of scheme was also studied by Bond and Veerapaneni (2008); they affirmed that treatment costs and energy requirements of ZLD can be reduced by adding the intermediate steps of concentrate treatment and secondary RO. They evaluated different
options of intermediate treatment: chemical precipitation with sodium hydroxide (NaOH) or lime [Ca(OH)2], fluidized bed crystallization (FBC), adsorption on activated alumina (AA), chemical precipitation with alum [Al2(SO4)$14H2O] and ion exchange. FBC has been demonstrated to produce a much smaller volume of solids than that produced by conventional softening. Besides, the evaluation of treatment costs comparing the system without intermediate treatment (benchmark process) with the system with FBC and microfiltration (ZLD process) shows that treatment cost would be reduced by 50e70%, and energy consumption would be reduced by 60e75%. An important number of brine concentrators are in operation in United States. Operating
Fig. 6 e Electrochlorination system proposed by Badruzzaman et al., 2009.
276
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
experiences of plants using brine concentrators to concentrate reject streams have shown that this is a viable application and the system is highly reliable (Mickley, 2006). Treatment schemes based on ZLD objective are pointed to be a promising option for the treatment of RO concentrates. Initial estimation about treatment costs indicates that these systems would be feasible, but it would be necessary leap from pilot plant to industrial level to safely evaluate the applicability of these systems.
2.2.2.
Emerging technologies for salts recovery
Recovering commercial products is the final challenge to improve the management of RO concentrates whatever their water source is. If valuable substances are recovered from RO concentrates a double objective will be achieved: the reduction of the environmental impact of RO concentrates disposal added to the improvement of the economy of the global treatment process. Through combination of evaporation and crystallization technologies, valuable salts can be recovered. Hajbi et al. (2010) analysed the solubility diagrams of the hexary system Naþ, Kþ, Mg2þ, Ca2þ/Cl, SO2 4 //H2O to define crystallization paths that they used for the experimental design. Their results assessed the potential of recovering salts like NaCl, KCl, CaSO4$2H2O, MgSO4$7H2O, . through the isothermal and isobaric evaporation of the rejected brine. Several studies on the viability of producing salts from RO concentrates of desalination plants managed by Petroleum Development Oman (PDO) were developed by Ahmed et al. (2003). The viability study was focused on the application of the patented SAL-PROC process consisting on the sequential extraction of dissolved elements from inorganic saline waters in the form of valuable chemical products in crystalline, slurry and liquid forms. The potential products that can be recovered would be gypsumemagnesium hydroxide (mixture), magnesium hydroxide, sodium chloride, calcium carbonate, sodium sulfate and calcium chloride. In light of the estimations, processing the 405 106 L/year of the rejected brine generated in the four PDO desalination plants would make possible to produce commercial salts worth US $895,000 in the most optimistic scenario. Although the economic profit is not so high, it is indisputable that the recovery of potential commercial salts is a great option to improve the cost-effectiveness of desalination processes. Regarding the market of inorganic salts, a quantitative assessment of the global market of recovered salts is not available yet because technologies are in an early development stage, but it has been reported that market could hold the produced salts. According to Ahmad and Williams in the United States 45 million tons of salts are produced annually and about 70% of these salts are used by chemical industries (Ahmad and Williams, 2011.). On the other hand Van Houwelingen stated that in the Netherlands the market of recovered salts would not be a problem since all solids generated in fluidized bed crystallization (FBC) processes for the treatment of reverse osmosis concentrates are put to beneficial use by other industries, such as steel production (Bond and Veerapaneni, 2008). Production of mixed acids and mixed bases using Electrodialysis with Bipolar Membranes EDBM (see Fig. 3) is another viable option. Mavrov et al., 1999 obtained suitable acids and
bases mixtures for the regeneration of ion exchangers by processing RO concentrates from desalination of surface water. EDBM has been also applied to RO concentrates from wastewater treatment plants, so this technology will be widely explained in the Section 3.2. Jegatheesan et al. (2009) evaluated the recovery of valuable metals from concentrates by means of liquideliquid extraction. The recovery ratio for rubidium using BAMBP [4-tertbutyl-2-(a-methylbenzyl) phenol] or 4-sec-butyl-2-(a-methylbenzyl) phenol] as extractants was higher than 80% (Jeppesen et al., 2009). Liquideliquid extraction has been also applied by Le Dirach et al. (2005) to recover materials from the concentrated brine of the RO plant in Skhira (Tunisia). The treatment process is divided in several extraction stages: phosphates extraction using a blend of iron and aluminium sulphates, liquideliquid extraction to recover cesium, liquideliquid extraction to recover indium and gallium, rubidium extraction, evaporation of the remaining brine to recover carnallite crystals KMgCl3$6H2O and as final step, extraction of germanium and magnesium. Still, these separation protocols have been studied at viability scale and require final demonstration.
3. RO concentrates from wastewater treatment plants: treatment alternatives In recent years, reverse osmosis has been also applied to further treat the secondary effluents of wastewater treatment plants. These RO concentrates present less salinity than RO concentrates from desalination plants although larger amounts of organic matter, including persistent micropollutants, are contained. Solley et al. (2010) reported that the contaminants in these streams could be 6e7 times more concentrated than in the feed water. Therefore specific treatments, such as advanced oxidation processes (AOPs), have been researched in order to reduce the pollutant load. Most of the reported studies are focused on the reduction of a global parameter that resembles the organic contamination such as the chemical oxygen demand (COD) and the Total Organic Carbon (TOC), although the removal of other pollutants has been also explored. For example, Ersever et al. (2007a, b) studied the removal of nitrogen compounds from RO brines through biological nitrificationedenitrification and sulfate reduction via a fluidized bioactive absorber reactor (FBAR) process and using granular activated carbon (GAC). Experiments were conducted at different hydraulic retention times and nitrate concentrations, showing that the FBAR process could be an efficient technology for nitrogen removal of RO concentrates, although nitrogen was considered a minor pollutant, recent characterisation of RO concentrates from WWTPs shows concentrations of ammonia up to 120 mg/L (see Table 2), turning the interest to the referred technologies. Traditional treatments such as coagulation and activated carbon adsorption were tested for DOC (dissolved organic carbon) removal in RO concentrates (Dialynas et al., 2008). Experimental results showed that ferric chloride is a better coagulant than aluminium sulfate (Al2(SO4)3$18H2O), achieving 52% removal of DOC. Ferric coagulation tests effectively removed colour (79%) while DOC and COD mostly due to low molecular weight compounds were removed up to
277
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
34% and 49% respectively (Bagastyo et al., 2011b). However, other authors (Zhou et al., 2011) reported lower efficiencies of coagulation, attributed to the different characteristics of the wastewater investigated. Adsorption on activated carbon (AC) was studied at different carbon doses. The highest removal observed was 91% for the dose of 5 g/L, so results of DOC removal with GAC adsorption are more promising than in coagulation tests. Similar results have been reported by Zhou et al. (2011), and as in previous works, the organics remaining in the RO concentrate were hardly adsorbed at higher AC dosage, indicating the existence of non-adsorbable organic matter fraction.
3.1. Advanced oxidation processes (AOPs) applied to RO concentrates from WWTPs Table 4 summarizes the research on the application of AOPs to RO concentrates. The studies include the well known ozonation and Fenton technologies, the latest developments in photocatalysis and photooxidation, together with new applications of sonolysis and electrooxidation.
3.1.1.
Ozonation
Feasibility tests of advanced oxidation processes for the removal of organics from the RO concentrate obtained in a water reclamation plant located in Singapore were performed by Zhou et al. (2011). Simple ozone treatment could only remove a small fraction (22%) of DOC with marginal improvements at higher energy input. Nevertheless, O3 treatment provided better performances that UV irradiation or hydrogen peroxide addition. Meanwhile, it was observed that the combination of ozonation with photocatalysis (UVA/ TiO2) enhanced DOC removal up to 52%, while the highest overall efficiency was obtained by the combined UVA/TiO2/O3 with coagulation pretreatment. Moreover, it was found that the latter combination increased the biodegradability index (BDI ¼ BOD5/COD) of the raw RO concentrate by 7e20 times. The O3/H2O2 treatment applied by Westerhoff et al. (2009) achieved a 75% removal of DOC contained in RO concentrate from the Scottsdale Water Campus facility, although very high
ozone (1000 mg/L O3) and hydrogen peroxide (0.7 mol H2O2/ mol O3) dosages were required. Ozone pretreatment for a biological activated carbon unit increased the biodegradability of the RO brine obtained from a domestic wastewater treatment facility (Lee et al., 2009a). Laboratory scale batch experiments showed that ozonation alone achieved a maximum TOC removal rate of 24%, while Biological Activated Carbon (BAC) alone provided a limited 23% of TOC removal. However, the synergistic effect of combining ozone and BAC showed a good potential in reducing fouling problems for a subsequent capacitive deionization (CDI) stage (Lee et al., 2009a, b) (see Fig. 4), a method for removing salts contained in aqueous solutions by electroadsortion (Oren, 2008). A capacitive deionization cell unit consists of two electrodes made of activated carbon and separated by the circulating ion containing solution. Electrons are not transmuted by red-ox reactions but by electrostatic adsorption (Strathmann, 2010). On the basis of the results it was claimed that CDI might be an attractive option for desalting water due to its lower energy demand compared to RO. RO concentrates from wastewater treatment plants contain, among many other pharmaceuticals, beta blockers, that are classified as potentially toxic to aquatic organisms. Because of beta blockers molecules have moieties (amine groups, activated aromatic rings), which are reactive towards ozone, it was tested whether ozonation can be applied for their mitigation. Moderate ozone doses (5e10 mg/L) were found to be sufficient to remove beta blockers efficiently (Benner et al., 2008).
3.1.2.
Fenton process
Fenton process for the treatment of RO concentrates has not been widely studied Westerhoff et al. (2009) developed Fenton (Fe2þ/H2O2) and Fenton-like (Fe3þ/H2O2) experiments in a jartest apparatus. The Fenton process (pHw3.3; 10 mM Fe2þ and 10 mM H2O2) removed up to 50% of DOC; residual iron was precipitated by raising the pH back to 7.5e8.0. Higher DOC removals might have been achieved using higher chemical dosages. No more studies regarding application of Fenton process to RO concentrates have been found in literature.
Table 4 e Advanced oxidation processes (AOPs) applied to RO concentrates. AOP applied
Objective
Photocatalysis
TiO2 þ hv/e þ hþ hþ þ H2 O/HO þ Hþ
Photooxidation
H2 O2 þ hv/2HO
Sonolysis Electrooxidation
H2 OþÞÞÞ/H þ HO 2HO /H2 O2 H2 O þ Anode/Anode½HO ads þ Hþ þ e
Ozonation
O3 /Initiators ðH2 O2 ; UVÞ HO
Fenton8
Fe2þ þ H2 O2 /Fe3þ þ OH þ HO
DOC removal DOC removal with coagulation pretreatment (FeCl3) DOC removal 95% DOC removal DOC removal DOC removal (low) DOC removal COD and TAN removal COD and TAN removal COD, ammonium and emerging pollutants N-nitrosodimethylamine destruction DOC and pharmaceuticals and pesticides oxidation DOC removal Beta blockers molecules removal Improving biodegradability of RO brine DOC removal
Reference Dialynas et al., 2008 Zhou et al., 2011 Westerhoff et al., 2009 Bagastyo et al., 2011a, b Westerhoff et al., 2009 Dialynas et al., 2008 Zhou et al., 2011 Van Hege et al., 2002 Van Hege et al., 2004 Pe´rez et al., 2010 Chaplin et al., 2010 Radjenovic et al., 2011 Westerhoff et al., 2009 Benner et al., 2008 Lee et al., 2009a, b Westerhoff et al., 2009
278
3.1.3.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
Photocatalysis and photooxidation
From the various advanced oxidation processes, heterogeneous photocatalysis with TiO2 as catalyst brings the advantage of the possible use of solar radiation. Several recent works deal with the application of UV/TiO2 to the removal of organic load from RO concentrates. For a RO concentrate with an initial DOC concentration of 10.2 mg/L, dark adsorption and UVA light experiments were performed (Dialynas et al., 2008) using two different catalyst doses (0.5 and 1 g/L TiO2). In dark adsorption tests, 30% DOC removal was achieved within 50e60 min at both catalyst concentrations. When UVA light is used during 1 h, the DOC oxidation yield was between 49 and 41% at the high and low catalyst level respectively. The combination of photocatalytic oxidation with coagulation as pretreatment, has been also studied (Zhou et al., 2011). For a raw RO concentrate obtained from a water reclamation plant in Singapore the removal of DOC and COD became rather slow after 1 h of reaction time when using both UVC/TiO2 and UVA/ TiO2 treatment processes. This behaviour was assigned to the nature of the remaining organics that were very recalcitrant to both treatment processes. However, the coagulation pretreatment using ferric chloride (FeCl3), provided the conditions for the gradual removal of DOC and COD by photocatalysis. After a reaction time of 6 h, 95% and 72% of DOC could be removed by UVC/TiO2 and UVA/TiO2. Similarly, Westerhoff et al. (2009) achieved 95% DOC removal at a UV dose of 10.4 kW h/m3, being the removal rate nearly independent of the titanium dioxide dose that was varied between 1 and 5 g/L. When the system UV/H2O2 was used, 40% DOC removal with a UV dose of 11.8 kW h/m3 and H2O2 dose of 10 mM was achieved. Photooxidation (UV/H2O2), has been also tested in RO concentrates from a water treatment plant by Bagastyo et al. (2011a). Complete decolourisation and 50e55% of COD removal was achieved.
3.1.4.
Sonolysis
Organic contaminants can also be degraded by the ultrasonic technique (US) in the presence of a catalyst, acids and nonoxidant gases. The chemical effects of ultrasounds are due to the high temperatures and pressures produced during violent collapse of cavitation bubbles. In water, implosion and fragmentation of the bubble is the centre of energy phenomena: temperature, pressure and electric discharges give rise to H2O sonolysis with production of radical species (OH, H, HOO) and direct destruction of solutes (Hoffmann et al., 1996; Farooq et al., 2008). Results of DOC removal obtained with sonolysis show that this process is less efficient than other AOPs. Testing a RO concentrate with an initial DOC of 10.1 mg/L, only a 29% of the DOC was oxidized in 1 h at 67.5 W. If the energy level is duplicated (135 W) during 1 h of treatment, DOC removal is only increased to 34% (Dialynas et al., 2008). Due to the low DOC removal, the possibility of using sonolysis combined with other AOPs has been studied (Zhou et al., 2011). Using a RO concentrate from a wastewater treatment plant with an initial DOC of 18 mg/L, at 1 h of reaction time, the sequence of DOC removal efficiencies achieved was US < US/H2O2 < US/ O3 < US/H2O2/O3 achieving w31% DOC removal as the highest removal with the last combination.
3.1.5.
Electrochemical oxidation
Electrochemical oxidation is a very efficient alternative for the treatment of wastewater containing non-biodegradable organics and ammonium compounds. With regard to RO brine, electrochemical treatment seems a promising technology, as the high salinity of the RO concentrate ensures an excellent electric conductivity that could reduce the energy consumption. Moreover, the high chloride content could facilitate indirect bulk oxidation through the electrogenerated strong oxidants such as hypochlorite and oxidation of total ammonia nitrogen (TAN) and organics can be accomplished simultaneously (Van Hege et al., 2002; Pe´rez et al., 2010). Reverse osmosis concentrates from a pilot installation processing a mixed domestic and textile wastewater effluent were tested to determine the characteristics and efficiency of several anode materials (Van Hege et al., 2002, 2004). Comparing RuO2 and Boron Doped Diamond (BDD) anodes, results showed that current efficiency for BDD anode (35.2%) is more favourable than the efficiency obtained for a RuO2 anode (14.5%). The higher efficiency of BDD anodes can be attributed to the superior chlorine production rate of this material because of the high selectivity towards the chlorine evolution reaction. Longer treatment times showed the capacity of BDD anodes to provide total removal of ammonium and DOC contained in the RO concentrate generated in a reverse osmosis facility aimed at the industrial reuse of the reclaimed water (Pe´rez et al., 2010). Thanks to the good results obtained in COD and ammonia removal, electrochemical oxidation of RO retentates with BDD anodes was also studied for the removal of emerging micropollutants (Pe´rez et al., 2010) Membrane concentrates contain an increasing amount of salts, organics and biological constituents. Among those organic compounds appear the socalled “emerging pollutants” which include pharmaceuticals, personal care products and other metabolites. These compounds appear in RO concentrates because they are highly rejected by the reverse osmosis membranes, and their removal must be investigated because of the environmental risk associated to their emission to the receiving natural water bodies. RO concentrates from a UF/RO pilot plant that treats the secondary effluent of a WWTP were oxidised electrochemically using BDD anodes (Pe´rez et al., 2010). Ten emerging pollutants that were found as most prevalent in the secondary effluent of a WWTP were investigated achieving removal percentages higher than 92% after 2 h of electrooxidation. Ibuprofen appears like the most resistant compound to electrochemical treatment, thus needing longer times for complete removal. The influence of the applied current density and the initial concentration of the micropollutants, obtained after different recovery ratios of the RO operation, allowed to conclude that the controlling step of the process kinetics is the mass transfer from the liquid bulk to the surface of the BDD anode due to the low concentration of the target compounds. Moreover, the formation of organochloride compounds was kept at low levels. However, testing RuO2/IrO2-coated Ti electrodes for elimination of pharmaceuticals and pesticides, Radjenovic et al. (2011) concluded that this type of electrodes should not be used, because chlorine mediated indirect oxidation was the main
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
mechanism and the organic compounds were transformed into their halogenated derivatives, which are persistent and require extreme treatment conditions. Advantages of BDD anodes were also probed for the elimination of N-nitrosodimethylamine, a highly toxic disinfectant compound present in chlorinated wastewater effluents, and further concentrated in RO retentates (Chaplin et al., 2010). Research focused on the electrooxidation of this compound using BDD electrodes showed its advantage over other AOPs, as oxidation was carried out in the presence of high HCO 3 concentrations. This is an advantage over other advanced oxidation processes due to the fact that scavenging effect of HCO 3 over hydroxyl radical is not a big drawback in electrooxidation.
3.2. Recovery of nutrients and salts from RO concentrates from WWTP’s Evaporation and crystallization steps have been evaluated by Mohammadesmaeili et al. (2010a, b) to recover salts from RO concentrates. To maximize the salts recovery the authors proposed a combined system based on a RO tandem with an intermediate stage of lime softening, and several stages of evaporation-crystallization after the secondary RO system (see Fig. 5). Thanks to the lime-soda treatment, magnesium hydroxide with a purity of 51e58% was obtained. This precipitate also contained 19.5e23.3% of CaCO3 and 1.3e7.8% of CaSO4 as impurities. Besides, calcite (CaCO3) with a purity of 95% and calcium sulfate with a purity of 92% were obtained as byproducts (Mohammadesmaeili et al., 2010a). After this softening treatment, the brine was treated in a second reverse osmosis unit. Then, an evaporation and crystallization step was carried out and a mixture of Na2SO4 (86e88%) and NaCl (5e14%) was obtained. The final step of the treatment train is an evaporation to dryness of the left effluent and as a result, þ a mixture of Naþ, SO2 4 , Cl and K (35e40%, 17e33%, 25e41%, 1e5% respectively was obtained) (Mohammadesmaeili et al., 2010b). The combination of reverse osmosis with evaporation and crystallization has been applied at the Doswell Combined Cycle Power Plant in Hanover County (Virginia) (Seigworth et al., 1995) with the goal of achieving zero liquid discharge. However, in this case the filter cake from the press must be disposed off. Thus, salts recovery by combining evaporation and crystallization is highlighted as a feasible option, although further work is still needed to evaluate the economic aspects of salt production. Zhang et al. (2009) analysed the viability of recovering valuable compounds from RO concentrates by applying ED. They used ion exchange membranes for the separation of nutrient ions and organic compounds from salts contained in the RO concentrates coming from the treatment of wastewaters of a food company. RO concentrates contained 120 mg/L of phosphate and 120 mg/L of organic compounds measured as TOC. The experimental tests were carried out with synthetic solutions resembling the composition of this RO concentrate (0.45 g/L NaCl, 0.54 g/L MgSO4, 0.026 g/L NaNO3, 0.69 NaHCO3, 0.062 g/L Na2HPO4 and 0e120 g/L TOC). ED tests were carried out to evaluate the performance of two membranes: a non-selective ED membrane (SA) and an ED membrane selective for monovalent anions (MVA). Experimental results reflected that using the MVA membrane,
279
monovalent anions can be separated from multivalent ions, that later are retained in the diluate compartment. The separation factor was improved by increasing the pH or reducing the applied current. With regard to the organic compounds it was concluded that larger ions are retained more efficiently than the smaller ones, which points to a sizeexclusion mechanism, while zwitterions are retained almost completely in the diluate side. In the same study it was affirmed that in a test carried out with real RO concentrates, more than 85% of the organic fraction remained in the feed while salt concentrations (Cl and SO2 4 ) decreased, suggesting that the separation of salts from organics by electrodialysis is feasible. Later studies (Zhang et al., 2010) endorsed that high overall recovery using ED is achieved and a moderate selectivity was observed between SO2 4 /Cl and HCO3 /Cl and 2þ a moderate to high selectivity was found between Ca /Naþ by using the non-selective anion and cation exchange membranes. Production of mixed acids and mixed basis using Electrodialysis with Bipolar Membranes (EDBM) (see Fig. 3) is also a viable option. For conversion of RO concentrates from wastewater treatment plants into mixed acid and mixed base streams, Badruzzaman et al. (2009) evaluated the use of bipolar membrane electrodialysis (EDBM). Moreover, they proposed electrochlorination (see Fig. 6) for onsite chlorine generation. RO concentrates contain a significant amount of divalent metals, so a Naþ based cation exchange resin was used to remove them in order to avoid potential precipitation and scaling on the membrane surface. Acid and base production tests were conducted in a batch configuration. The salt concentration in the feed in terms of conductivity went down from 9 mS cm1 to less than 2 mS cm1, while the acid and base concentrations of the products was around 0.2 N. Regarding the separation of nutrients from RO concentrates, adsorption using polymeric resins has been investigated to recover phosphate. The technology seems feasible although phosphate concentrations are not very high, Kumar et al. (2007) analysed the separation of phosphates using polymeric ligand exchange (PLEs) resins and their recovery as struvite. RO concentrates were collected from two pilot integrated membrane systems at the wastewater treatment plant in Rio Rancho (New Mexico) and the North City water reclamation plant in San Diego. The RO concentrate with a low chloride concentration (0.684 g/L Cl) and containing 10 mg/L of phosphate was passed through a packed column loaded with a PLE resin. The regeneration of the exhausted column was conducted using NaCl. In the regenerant solution, P was recovered at a concentration of 78 mg/L. To achieve struvite precipitation, tests were conducted with deionized water spiked with 20 mg P/L to determine the best molar ratio (P:Mg2þ:NHþ 4 ::1:1.5:1) and then with the recovered solution. Phosphate can be potentially recovered as struvite, although nitrate and sulfate also precipitate.
4. RO concentrates from industrial water sources: treatment alternatives Hypersaline brine production is one of the most important environmental issues in mining industry. RO concentrates
280
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
have a conductivity similar to RO concentrates from desalination plants (22,000 mS cm1) and regarding anion composition, sulfate concentration is much higher than chloride (16,000 mg/L SO2 4 ; 955 mg/L Cl ) (Randall et al., 2011). According to this composition, crystallization techniques based on eutectic freeze diagrams were investigated. Eutectic freeze crystallization (EFC) is based on achieving the eutectic temperature as means to separate aqueous solutions into pure water and solid salts. Because the heat of fusion of ice (6.01 kJ/mol) is six times less than the heat of evaporation of water (40.65 kJ/mol), the energy required to separate the water as ice is significantly lower than that required to separate it by evaporation. In addition, the simultaneous production of pure ice and pure salt(s) is a major advantage. The economic advantage of EFC has been analyzed by Nathoo et al. (2009). The authors compared EFC in a tandem crystallizer (see Fig. 7) with an Evaporative Crystallization (EC) system to obtain NaCl$2H2O and Na2SO4$10H2O. Tests were carried out using reverse osmosis retentates from typical mine water brines containing high levels of sodium, sulfate and chlorine. The temperature in each crystallizer was selected according to the concentrates composition in order to obtain the maximum salt recovery. Taking into account the electricity cost and the salts production, a comparison between EFC and EC concluded that cost saving of using EFC is in the range of 80e85% for the brines under study. In addition, the revenue generated by the sale of the pure salts produced by EFC and the cost of disposal of the mixed salt waste produced by EC were not included in the cost evaluation developed in that study. The EFC technique has been also evaluated for the treatment of multicomponent hypersaline brine form the eMalahleni (South Africa) Water Reclamation Plant (EWRP), built to address the damage done to the environment and water systems through Acid Mine Drainage. It was concluded that the EFC process could increase the production of the water reclamation plant by 120 m3/d while a production of 476 kg/ d CaSO4$2H2O and 471 kg/d Na2SO4$10H2O would also be obtained (Randall et al., 2011). The recovery of pure salts is difficult because they usually have several hydrate forms. Sodium sulfate exists in three hydrate forms: anhydrous Na2SO4 (thenardite), Na2SO4$7H2O and Na2SO4$10H2O (mirabilite). Reddy et al. (2010) study the recovery and purity of sodium sulfate using synthetic
Fig. 7 e Simplified EFC process flow sheet (Nathoo et al., 2009).
solutions with a composition simulating brine from the mining industry. Pure water crystals were obtained (<20 ppm impurities) from a synthetic retentate stream. The presence of higher concentrations of NaCl lowers the solubility of Na2SO4 by the common ion effect. A recovery of >90% pure Na2SO4 crystals was obtained from a concentrated NaCl stream prior to any sodium chloride crystals being produced.
5.
Conclusions
Since direct disposal of RO concentrates from desalination plants is recognised as a practice with adverse impacts on marine ecosystems, the search for environmentally friendly management options is a technological challenge. Thus, in this review, treatment technologies of water reverse osmosis concentrates have been addressed, classified according to the source of RO concentrates and the level of development of each management technology. As the composition of the concentrate is closely related to its source, and the selection of the most suitable treatment is based on the concentrate composition, three different sources of RO concentrates have been reviewed i) desalination plants, ii) tertiary processes in WWTP and, iii) mining industries. With regard to RO concentrates from desalination plants traditional technologies, such as solar evaporation easy to operate but with a large requirement of land area, give way to more efficient emerging technologies that allow for considerable brine volume reductions such as electrodialysis, forward osmosis, membrane distillation or even its coupling with crystallisation, MDC, that facilitates easy recovery of high purity NaCl crystals. Recovery of salts, nutrients and valuable compounds from RO concentrates is a promising investigation line and thus related works have been collected. However, although thermal brine concentration technologies are well established, they are energy intensive and not cost effective for large-scale applications. In case of freezing processes the initial investment and complexity of operation limit their use. With regard to the treatment of RO concentrates from WWTPs studies are mainly focused on the reduction of pollutant load and removal of micro-pollutants that could accumulate and cause a high environmental impact. Good results have been obtained applying advanced oxidation technologies and among them electrochemical oxidation with BDD anodes for the removal of micro-pollutants and persistent compounds. Advanced oxidation processes are probably the most promising technologies to degrade and detoxify endocrine disrupting compounds, but the high cost of these technologies may limit their application. Alternatives aiming at zero liquid discharge, through the combination of individual technologies, are highlighted as the most promising management option in this investigated field. Although studies addressed to this objective are mostly related to RO concentrates from desalination plants, ZLD systems can be applied to any RO concentrate, regardless its origin source. The final objective of ZLD systems must be the recovery of valuable compounds from RO concentrates. Joining both objectives: ZLD and valuable compounds recovery, the problem of RO concentrates management would
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
be solved. However, the main drawback of ZLD systems is the feasibility of application to industrial level due to their investing and operating costs. Although first economic estimations are encouraging, the profitability of these systems has been demonstrated only at lab and pilot plant scale.
Acknowledgements Financial support from projects Consolider CSD2006-44, CTQ2008-00690/PPQ, ENE2010-15585, CTQ2008-03225/PPQ is gratefully acknowledged.
references
Acevedo, C.R., Gardea-Torresdey, J.L., Tarquin, A.J., 2010. Silica removal from brine by using ion exchange. In: World Environmental and Water Resources Congress 2010: Challenges of Change e Proceedings of the World Environmental and Water Resources Congress. Ahmad, M., Williams, P., 2011. Assessment of desalination technologies for high saline brine applications e discussion paper. Desalination and Water Treatment 30, 22e36. Ahmed, M., Arakel, A., Hoey, D., Thumarukudy, M.R., Goosen, M.F.A., Al-Haddabi, M., Al-Belushi, A., 2003. Feasibility of salt production from inland RO desalination plant reject brine: a case study. Desalination 158, 109e117. Ahmed, M., Shayya, W.H., Hoey, D., Al-Handaly, J., 2001. Brine disposal from reverse osmosis desalination plants in Oman and the United Arab Emirates. Desalination 133 (2), 135e147. Ahmed, M., Shayya, W.H., Hoey, D., Mahendran, A., Morris, R., AlHandaly, J., 2000. Use of evaporation ponds for brine disposal in desalination plants. Desalination 130 (2), 155e168. Altayaran, A.M., Madany, I.M., 1992. Impact of a desalination plant on the physical and chemical properties of seawater, Bahrain. Water Research 26 (4), 435e441. Al-Wazzan, Y., Safar, M., Mesri, A., 2003. Reverse osmosis brine staging treatment of subsurface water. Desalination 155 (2), 141e151. Arnal, J.M., Sancho, M., Iborra, I., Goza´lvez, J.M., Santafe´, A., Lora, J., 2005. Concentration of brines from RO desalination plants by natural evaporation. Desalination 182, 435e439. Badruzzaman, M., Oppenheimer, J., Adham, S., Kumar, M., 2009. Innovative beneficial reuse of reverse osmosis concentrate using bipolar membrane electrodialysis and electrochlorination processes. Journal of Membrane Science 326 (2), 392e399. Bagastyo, A.Y., Keller, J., Poussade, Y., Batstone, D.J., 2011a. Characterisation and removal of recalcitrants in reverse osmosis concentrates from water reclamation plants. Water Research 45 (7), 2415e2427. Bagastyo, A.Y., Keller, J., Batstone, D.J., 2011b. Size fractionation characterisation of removed organics in reverse osmosis concentrates by ferric chloride. Water Science and Technology 63 (9), 1795e1800. Benner, J., Salhi, E., Ternes, T., von Gunten, U., 2008. Ozonation of reverse osmosis concentrate: kinetics and efficiency of beta blocker oxidation. Water Research 42 (12), 3003e3012. Bond, R., Veerapaneni, S., 2008. Zeroing in on ZLD technologies for inland desalination. Journal American Water Works Association (AWWA) 100, 76e89. Chaplin, B.P., Schrader, G., Farrell, J., 2010. Electrochemical destruction of N-nitrosodimethylamine in reverse osmosis
281
concentrates using boron-doped diamond film electrodes. Environmental Science and Technology 44 (11), 4264e4269. Chelme-Ayala, P., Smith, D.W., El-Din, M.G., 2009. Membrane concentrate management options: a comprehensive critical review. Canadian Journal of Civil Engineering 36 (6), 1107e1119. Curcio, E., Ji, X., Di Profio, G., Sulaiman, A.O., Fontananova, E., Drioli, E., 2010. Membrane distillation operated at high seawater concentration factors: role of the membrane on CaCO3 scaling in presence of humic acid. Journal of Membrane Science 346 (2), 263e269. Del Bene, J.V., Jirka, G., Largier, J., 1994. Ocean brine disposal. Desalination 97, 365e372. Dialynas, E., Mantzavinos, D., Diamadopoulos, E., 2008. Advanced treatment of the reverse osmosis concentrate produced during reclamation of municipal wastewater. Water Research 42 (18), 4603e4608. Einav, R., Lokiec, F., 2003. Environmental aspects of a desalination plant in Ashkelon. Desalination 156, 79e85. Ersever, I., Ravindran, V., Pirbazari, M., 2007a. Biological denitrification of reverse osmosis brine concentrates: I. Batch reactor and chemostat studies. Journal of Environmental Engineering and Science 6 (5), 503e518. Ersever, I., Ravindran, V., Pirbazari, M., 2007b. Biological denitrification of reverse osmosis brine concentrates: II. Fluidized bed adsorber reactor studies. Journal of Environmental Engineering and Science 6 (5), 519e532. Farooq, R., Shaukat, S.F., Khan, A.K., Farooq, U., 2008. Ultrasonic induced decomposition of methidathion pesticide. Journal of Applied Sciences 8 (1), 140e145. Gabelich, C.J., Rahardianto, A., Northrup, C.R., Yun, T.I., Cohen, Y., 2011. Process evaluation of intermediate chemical demineralization for water recovery enhancement in production-scale brackish water desalting. Desalination 272, 36e45. Gabelich, C.J., Xu, P., Cohen, Y., 2010. Chapter 10 concentrate treatment for inland desalting. Sustainability Science and Engineering 2, 295e326. Gabelich, C.J., Williams, M.D., Rahardianto, A., Franklin, J.C., Cohen, Y., 2007. High-recovery reverse osmosis desalination using intermediate chemical demineralization. Journal of Membrane Science 301, 131e141. Gilron, J., Folkman, Y., Savliev, R., Waisman, M., Kedem, O., 2003. WAIV e wind aided intensified evaporation for reduction of desalination brine volume. Desalination 158, 205e214. Greenlee, L.F., Testa, F., Lawler, D.F., Freeman, B.D., Moulin, P., 2011. Effect of antiscalant degradation on salt precipitation and solid/liquid separation of RO concentrate. Journal of Membrane Science 366, 48e61. Greenlee, L.F., Testa, F., Lawler, D.F., Freeman, B.D., Moulin, P., 2010a. Effect of antiscalants on precipitation of an RO concentrate: metals precipitated and particle characteristics for several water compositions. Water Research 44 (8), 2672e2684. Greenlee, L.F., Testa, F., Lawler, D.F., Freeman, B.D., Moulin, P., 2010b. The effect of antiscalant addition on calcium carbonate precipitation for a simplified synthetic brackish water reverse osmosis concentrate. Water Research 44 (9), 2957e2969. Greenlee, L.F., Lawler, D.F., Freeman, B.D., Marrot, B., Moulin, P., 2009. Reverse osmosis desalination: water sources, technology, and today’s challenges. Water Research 43 (9), 2317e2348. Gryta, M., 2002. Concentration of NaCl solution by membrane distillation integrated with crystallization. Separation Science and Technology 37 (15), 3535e3558. Hajbi, F., Hammi, H., M’Nif, A., 2010. Reuse of RO desalination plant reject brine. Journal of Phase Equilibria and Diffusion 31 (4), 341e347.
282
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
He, F., Gilron, J., Lee, H., Song, L., Sirkar, K.K., 2008. Potential for scaling by sparingly soluble salts in crossflow DCMD. Journal of Membrane Science 311, 68e80. Hoffmann, M.R., Hua, I., Ho¨chemer, R., 1996. Application of ultrasonic irradiation for the degradation of chemical contaminants in water. Ultrasonics Sonochemistry 3 (3), S163eS172. Jegatheesan, V., Liow, J.L., Shu, L., Kim, S.H., Visvanathan, C., 2009. The need for global coordination in sustainable development. Journal of Cleaner Production 17 (7), 637e643. Jeppesen, T., Shu, L., Keir, G., Jegatheesan, V., 2009. Metal recovery from reverse osmosis concentrate. Journal of Cleaner Production 17 (7), 703e707. Ji, X., Curcio, E., Al Obaidani, S., Di Profio, G., Fontananova, E., Drioli, E., 2010. Membrane distillation-crystallization of seawater reverse osmosis brines. Separation and Purification Technology 71 (1), 76e82. Katzir, L., Volkmann, Y., Daltrophe, N., Korngold, E., Mesalem, R., Oren, Y., Gilron, J., 2010. WAIV e wind aided intensified evaporation for brine volume reduction and generating mineral byproducts. Desalination and Water Treatment 13, 63e73. Korngold, E., Aronov, L., Daltrophe, N., 2009. Electrodialysis of brine solutions discharged from an RO plant. Desalination 2428, 215e227. Korngold, E., Aronov, L., Belayev, N., Kock, K., 2005. Electrodialysis with brine solutions oversaturated with calcium sulfate. Desalination 172 (1), 63e75. Kumar, M., Badruzzaman, M., Adham, S., Oppenheimer, J., 2007. Beneficial phosphate recovery from reverse osmosis (RO) concentrate of an integrated membrane system using polymeric ligand exchanger (PLE). Water Research 41 (10), 2211e2219. Lattemann, S., Ho¨pner, T., 2008. Environmental impact and impact assessment of seawater desalination. Desalination 220, 1e15. Le Dirach, J., Nisan, S., Poletiko, C., 2005. Extraction of strategic materials from the concentrated brine rejected by integrated nuclear desalination systems. Desalination 182, 449e460. Lee, L.Y., Ng, H.Y., Ong, S.L., Hu, J.Y., Tao, G., Kekre, K., Viswanath, B., Lay, W., Seah, H., 2009a. Ozone-biological activated carbon as a pretreatment process for reverse osmosis brine treatment and recovery. Water Research 43 (16), 3948e3955. Lee, L.Y., Ng, H.Y., Ong, S.L., Tao, G., Kekre, K., Viswanath, B., Lay, W., Seah, H., 2009b. Integrated pretreatment with capacitive deionization for reverse osmosis reject recovery from water reclamation plant. Water Research 43 (18), 4769e4777. Macedonio, F., Katzir, L., Geisma, N., Simone, S., Drioli, E., Gilron, J., 2011. Wind-Aided Intensified eVaporation (WAIV) and membrane crystallizer (MCr) integrated brackish water desalination process: advantages and drawbacks. Desalination 273 (1), 127e135. Malaeb, L., Ayoub, G.M., 2011. Reverse osmosis technology for water treatment: state of the art review. Desalination 267 (1), 1e8. Malaxos, P.J., Morin, O.J., 1990. Surface water discharge of reverse osmosis concentrates. Desalination 78 (1), 27e40. Martinetti, C.R., Childress, A.E., Cath, T.Y., 2009. High recovery of concentrated RO brines using forward osmosis and membrane distillation. Journal of Membrane Science 331, 31e39. Mauguin, G., Corsin, P., 2005. Concentrate and other waste disposals from SWRO plants: characterization and reduction of their environmental impact. Desalination 182, 355e364. Mavrov, V., Chmiel, H., Heitele, B., Rogener, F., 1999. Desalination of surface water to industrial water with lower impact on the environment part 4: treatment of effluents from water
desalination stages for reuse and balance of the new technological concept for water desalination. Desalination 124, 205e216. Meerganz von Medeazza, G.L., 2005. “Direct” and socially-induced environmental impacts of desalination. Desalination 185, 57e70. Meneses, M., Pasqualino, J.C., Ce´spedes-Sa´nchez, R., Castells, F., 2010. Alternatives for reducing the environmental impact of the main residue from a desalination plant. Journal of Industrial Ecology 14 (3), 512e527. Mericq, J., Laborie, S., Cabassud, C., 2010. Vacuum membrane distillation of seawater reverse osmosis brines. Water Research 44 (18), 5260e5273. Mickley, M.C., 2006. Membrane Concentrate Disposal: Practices and Regulation. Report No.123. Bureau of Reclamation, Denver, CO, second ed. Water Treatment Engineering and Research Group; Mickley and Associates, Boulder, CO. Mickley, M.C., 2001. Membrane Concentrate Disposal: Practices and Regulation. Bureau of Reclamation, Denver, CO. Water Treatment Engineering and Research Group; Mickley and Associates, Boulder, CO. Mohammadesmaeili, F., Badr, M.K., Abbaszadegan, M., Fox, P., 2010a. Mineral recovery from inland reverse osmosis concentrate using isothermal evaporation. Water Research 44 (20), 6021e6030. Mohammadesmaeili, F., Badr, M.K., Abbaszadegan, M., Fox, P., 2010b. Byproduct recovery from reclaimed water reverse osmosis concentrate using lime and soda-ash treatment. Water Environmental Research 82 (4), 342e350. Muniz, A., Skehan, S., 1990. Disposal of concentrate from brackish water desalting plants by use of deep injection wells. Desalination 78 (1), 41e47. Mun˜oz, I., Ferna´ndez-Alba, A.R., 2008. Reducing the environmental impacts of reverse osmosis desalination by using brackish groundwater resources. Water Research 42 (3), 801e811. Nathoo, J., Jivanji, R., Lewis, A.E., 2009. Freezing your brines off: eutectic freeze crystallization for brine treatment. In: International Mine Water Conference, pp. 431e437. Neilly, A., Jegatheesan, V., Shu, L., 2009. Evaluating the potential for zero discharge from reverse osmosis desalination using integrated processes e a review. Desalination and Water Treatment 11, 58e65. Ning, R.Y., Tarquin, A.J., 2010. Crystallization of salts from superconcentrate produced by tandem RO process. Desalination and Water Treatment 16, 238e242. Ning, R.Y., Tarquin, A.J., Balliew, J.E., 2010. Seawater RO treatment of RO concentrate to extreme silica concentrations. Desalination and Water Treatment 22, 286e291. Ning, R.Y., Troyer, T.L., 2009. Tandom reverse osmosis process for zero-liquid discharge. Desalination 237, 238e242. Ning, R.Y., Tarquin, A., Trzcinski, M., Patwardhan, G., 2006. Recovery optimization of RO concentrate from desert wells. Desalination 201, 315e322. Oren, Y., Korngold, E., Daltrophe, N., Messalem, R., Volkman, Y., Aronov, L., Weismann, M., Bouriakov, N., Glueckstern, P., Gilron, J., 2010. Pilot studies on high recovery BWRO-EDR for near zero liquid discharge approach. Desalination 261 (3), 321e330. Oren, Y., 2008. Capacitive deionization (CDI) for desalination and water treatment e past, present and future (a review). Desalination 228, 10e29. Pe´rez, G., Ferna´ndez-Alba, A.R., Urtiaga, A.M., Ortiz, I., 2010. Electro-oxidation of reverse osmosis concentrates generated in tertiary water treatment. Water Research 44 (9), 2763e2772. Qu, D., Wang, J., Wang, L., Hou, D., Luan, Z., Wang, B., 2010. Concentrating primary reverse osmosis concentrate by direct
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 6 7 e2 8 3
contact membrane distillation. Water Science and Technology: Water Supply 10 (3), 403e410. Qu, D., Wang, J., Wang, L., Hou, D., Luan, Z., Wang, B., 2009. Integration of accelerated precipitation softening with membrane distillation for high-recovery desalination of primary reverse osmosis concentrate. Separation and Purification Technology 67 (1), 21e25. Radjenovic, J., Bagastyo, A., Rozendal, R.A., Mu, Y., Keller, J., Rabaey, K., 2011. Electrochemical oxidation of trace organic contaminants in reverse osmosis concentrate using RuO2/ IrO2-coated titanium anodes. Water Research 45 (4), 1579e1586. Rahardianto, A., McCool, B.C., Cohen, Y., 2010. Accelerated desupersaturation of reverse osmosis concentrate by chemically-enhanced seeded precipitation. Desalination 264 (3), 256e267. Randall, D.G., Nathoo, J., Lewis, A.E., 2011. A case study for treating a reverse osmosis brine using eutectic freeze crystallization-approaching a zero waste process. Desalination 266, 256e262. Reddy, S.T., Lewis, A.E., Witkamp, G.J., Kramer, H.J.M., van Soronsen, J., 2010. Recovery of Na2SO4$10H2O from a reverse osmosis retentate by eutectic freeze crystallisation technology. Chemical Engineering Research and Design 88 (9), 1153e1157. Roberts, D.A., Johnston, E.L., Knott, N.A., 2010. Impacts of desalination plant discharges on the marine environment: a critical review of published studies. Water Research 44 (18), 5117e5128. Seigworth, A., Ludlum, R., Reahl, E., 1995. Case study: integrating membrane processes with evaporation to achieve economical zero liquid discharge at the Doswell combined cycle facility. Desalination 102, 81e86. Singh, R., 2009. Brine recovery at industrial RO plants: conceptual process design studies. Desalination and Water Treatment 8, 54e67. Solley, D., Gronow, C., Tait, S., Bates, J., Buchanan, A., 2010. Managing the reverse osmosis concentrate from the Western Corridor recycled water scheme. Water Practice and Technology 5, 1e8. Squire, D., 2000. Reverse osmosis concentrate disposal in the UK. Desalination 132, 47e54. Squire, D., Murrer, J., Holden, P., Fitzpatrick, C., 1997. Disposal of reverse osmosis membrane concentrate. Desalination 108, 143e147. Stanford, B.D., Leising, J.F., Bond, R.G., Snyder, S.A., 2010. Chapter 11 inland desalination: current practices, environmental
283
implications, and case studies in Las Vegas, NV. Sustainability Science and Engineering 2, 327. Strathmann, H., 2010. Electrodialysis, a mature technology with a multitude of new applications. Desalination 264 (3), 268e288. Tang, W., Ng, H.Y., 2008. Concentration of brine by forward osmosis: performance and influence of membrane structure. Desalination 224, 143e153. Turek, M., Was, J., Dydo, P., 2009. Brackish water desalination in RO-single pass EDR system. Desalination and Water Treatment 7, 263e266. Urtiaga, A.M., Gorri, E.D., Ruiz, G., Ortiz, I., 2001. Parallelism and differences of pervaporation and vacuum membrane distillation in the removal of VOCs from aqueous streams. Separation and Purification Technology 22e23, 327e337. Van Der Bruggen, B., Lejon, L., Vandecasteele, C., 2003. Reuse, treatment, and discharge of the concentrate of pressuredriven membrane processes. Environmental Science and Technology 37 (17), 3733e3738. Van Hege, K., Verhaege, M., Verstraete, W., 2004. Electro-oxidative abatement of low-salinity reverse osmosis membrane concentrates. Water Research 38 (6), 1550e1558. Van Hege, K., Verhaege, M., Verstraete, W., 2002. Indirect electrochemical oxidation of reverse osmosis membrane concentrates at boron-doped diamond electrodes. Electrochemistry Communications 4, 296e300. Voutchkov, N., 2011. Overview of seawater concentrate disposal alternatives. Desalination 273, 205e219. Watson, I., 1990. Characterization of desalting concentrates. Desalination 78, 5e9. Westerhoff, P., Moon, H., Minakata, D., Crittenden, J., 2009. Oxidation of organics in retentates from reverse osmosis wastewater reuse facilities. Water Research 43, 3992e3998. Zhang, Y., Ghyselbrecht, K., Meesschaert, B., Pinoy, L., Van der Bruggen, B., 2010. Electrodialysis on RO concentrate to improve water recovery in wastewater reclamation. Journal of Membrane Science. doi:10.1016/j.memsci.2010.10.036. Zhang, Y., Van der Bruggen, B., Pinoy, L., Meesschaert, B., 2009. Separation of nutrient ions and organic compounds from salts in RO concentrates by standard and monovalent selective ionexchange membranes used in electrodialysis. Journal of Membrane Science 332, 104e112. Zhou, T., Lim, T.-T., Chin, S.-S., Fane, A.G., 2011. Treatment of organics in reverse osmosis concentrate from a municipal wastewater reclamation plant: feasibility test of advanced oxidation processes with/without pretreatment. Chemical Engineering Journal 166 (3), 932e939.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 8 4 e2 9 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Similarities in effluent organic matter characteristics from Connecticut wastewater treatment plants Matthew L. Quaranta, Mykel D. Mendes, Allison A. MacKay* University of Connecticut, Environmental Engineering Program, 261 Glenbrook Rd Storrs, CT 06269, USA
article info
abstract
Article history:
Effluent organic matter (EfOM) from five Connecticut (USA) municipal wastewater treat-
Received 13 August 2011
ment plants was isolated with DAX8 (hydrophobic fraction) and XAD4 (transphilic fraction)
Received in revised form
resins. Isolate recoveries ranged from 18 to 42% of the total organic carbon for DAX8 resin
7 October 2011
and from 6 to 12% for XAD4 resin. Isolated EfOM was characterized by traditional organic
Accepted 11 October 2011
geochemistry techniques. Weight-averaged molecular weights of extracted EfOM by size
Available online 28 October 2011
exclusion chromatography were 450e670 Da with higher weights observed for the hydrophobic fractions than the transphilic fractions. Fluorescence characterization showed both
Keywords:
humic- and fulvic-like fluorescence, as well as tryptophan- and tyrosine-like fluorescence,
Excitationeemission matrix
the latter not commonly observed for terrestrial organic matter. Fluorescence indices were
NOM
between 1.5 and 1.9 with lower values observed for hydrophobic EfOM fractions than for
SUVA
transphilic fractions. Specific ultraviolet absorbance was measured between 0.8 and
E2/E3 ratio
3.0 L mg1 m1 with higher values for the hydrophobic EfOM fractions. Together these results indicated that isolated EfOM is similar in characteristics to microbially derived organic matter from natural aquatic systems. Little variation in EfOM characteristics was observed between the five wastewater treatment plants, suggesting that the characteristics of EfOM are similar, regardless of treatment plant design. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Effluent discharges of treated municipal wastewater represent an important source of organic matter to aquatic ecosystems. In 2008, the estimated volume of wastewater treated was 32,345 million gallons per day in the US (EPA, 2008). Treated wastewater effluent typically contains dissolved organic matter (DOM) concentrations that range between 4 and 20 mgC/L, comparable to DOM concentrations in natural water sources (Thurman, 1985). Such effluent discharges can constitute a large fraction of streamflows in systems receiving treated wastewater. For example, nearly a quarter of all US National Pollutant Discharge Elimination System (NPDES) permitted effluent releases have in-stream dilutions of less than 10-fold under average flow conditions
(Brooks et al., 2006). The contribution of organic matter from wastewater effluent discharge to rivers and streams is particularly significant during low flow summer periods and in arid regions where wastewater effluent can contribute up to 100% of streamflow during dry months (Swayne et al., 1980). Little work has examined anthropogenic sources of organic matter, such as in wastewater effluent; thus, it is unknown how effluent organic matter (EfOM) contributes to ecosystem processes in streams receiving wastewater effluent discharges. The importance of dissolved organic matter from natural sources in stream systems is well established. DOM is an important carbon source in stream systems, supporting microbial growth as the base of the food web (Allan and Castillo, 2007). The presence of DOM affects the availability
* Corresponding author. Tel.: þ1 860 486 2450; fax: þ1 860 486 2298. E-mail address:
[email protected] (A.A. MacKay). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.010
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 8 4 e2 9 4
of dissolved metals and organic contaminants by serving as a sorptive phase that can be transported through aquatic environments (Croue et al., 2003; Haitzer et al., 2003; Pernetcoudrier et al., 2008; Polubesova et al., 2008). DOM is also an important mediator of indirect photolysis reactions that are important to the degradation of many organic contaminants, including pesticides, pharmaceuticals, and personal care products (Doll and Frimmel, 2003; Guerard et al., 2009; Lim et al., 2008). Efforts to probe more deeply the role of natural DOM in these ecosystem processes have emphasized bulk physical and chemical characteristics of these materials. Consequently, organic matter from natural systems has been extensively characterized in previous work (Thurman, 1985). The characterization of natural DOM isolated from many sources has indicated two end members that bound DOM characteristics for most aquatic systems (McKnight et al., 2001). These conclusions were drawn by isolating operationally defined humic and fulvic acid components using hydrophobic resin (Thurman and Malcolm, 1981), followed by characterization techniques, including size exclusion chromatography (Chin et al., 1994), ultravioletevisible light absorption- (Weishaar et al., 2003), fluorescence- (Cory et al., 2010; Hudson et al., 2007; McKnight et al., 2001), and nuclear magnetic resonance (Woods et al., 2010) spectroscopy. Findings from these studies highlighted DOM extracts from the Suwannee River, originating as allochthonous DOM from terrestrial sources, and from perennially ice-covered Antarctic lakes, originating as autochthonous DOM from microbial sources, to be end members, between which the characteristics of DOM from other sources fell (McKnight et al., 2001). In general, DOM of terrestrial origin, in contrast to DOM of microbial origin, has higher average molecular weight (Chin et al., 1994), higher specific UV absorbance (Weishaar et al., 2003), a higher fraction of aromatic carbon (Thurman, 1985), but lower fluorescence indices (McKnight et al., 2001). Importantly, these characteristics have been correlated with contaminant fate processes and thus, this continuum model with Suwannee River and Antarctic lake end members also applies to DOM contaminant sorption coefficients (Chin et al., 1997) and contaminant degradation rates (Guerard et al., 2009). In contrast to natural DOM, wastewater effluent organic matter (EfOM) has not been extensively characterized. One challenge to obtaining a comprehensive understanding of EfOM characteristics from existing studies is the lack of standard protocols between studies. For example, different resins have been used to extract DOM from effluent sources, including XAD4, anion and cation exchange resins, and some studies have used organic solvents as eluents (Baker, 2001; Imai et al., 2002; Leenheer, 1981; Pernet-coudrier et al., 2008; Wang et al., 2009). Often only a subset of typical natural DOM characterization techniques are applied to a given EfOM sample, making it difficult to determine whether EfOM characteristics have the same relative trends among characteristics as for natural DOM. Among the available data, EfOM appears be more similar to organic matter derived from natural microbial sources, despite the difference in microbial communities between wastewater treatment plants (predominantly bacteria and archaea) and natural aquatic systems (with a large algal contribution). EfOM appears to have lower molecular weight and lower SUVA than
285
terrestrially derived organic matter and fluorescence excitationeemission matrices that are more similar to microbial organic matter (Baker, 2001; Hudson et al., 2007; Imai et al., 2002; Park et al., 2010). However, on the basis of these few studies, it is unclear what similarities there are between EfOM characteristics from different treatment plants (Chen et al., 2003; Imai et al., 2002), and thus, where EfOM may group relative to the established natural DOM continuum. The purpose of this study was to characterize EfOM from five Connecticut (USA) municipal wastewater treatment plants. Our goals were to compare EfOM characteristics to natural DOM characteristics and to begin establishing the ranges in EfOM characteristic values between treatment plants. To this end, we used widely available characterization techniques e resin isolation, size exclusion chromatography and optical e and fluorescence spectroscopy e because extensive published datasets of NOM characteristics by these methods are available. We sampled EfOM from conventionally operated large (primary and secondary clarification) and small (no primary clarification) activated sludge plants, typical of designs throughout the US. Several plants had unique operating designs (advanced nitrogen removal, UV disinfection, powdered activated carbon) that could contribute to differing EfOM characteristics. This study examining the range of characteristics of EfOM allowed us to establish how EfOM relates to the natural organic matter continuum.
2.
Experimental methods
2.1.
Materials
Methanol (Certified ACS), acetonitrile (HPLC Grade), HCl (Trace Metal Grade), KOH (certified ACS) and acetone (Optima) were purchased from Fisher (Fair Lawn, NJ). DAX 8 Supelite Resin was purchased from Supelco (Bellafonte, PA), XAD4 Amberlite resin and cation exchange resin (DOWEX Marathon MSC, Hþform) were purchased from Sigma Aldrich. Glass Chromaflex chromatography columns with Teflon end fittings and 0.20 mm bed supports were purchased from Kontes (Vineland, NJ). Potassium hydrogen phthalate was purchased from Ricca Chemical Company (Pokomoke City, MD). Polystyrene sulfonic acids (PSS) in sodium salt form were purchased from Polysciences (Warrington, PA). All water was obtained from a high-purity water system (18 MU-cm, Diamond NANOpure, low TOC system, Barnstead-Thermolyne, Dubuque, IA).
2.2.
Treatment plants
Wastewater effluent was collected from five wastewater treatment plants in the state of Connecticut (USA) between November 2010 and April 2011 (Table 1). The sources of raw wastewater to all of the treatment plants consisted primarily of residential wastewater with some small contributions from light commercial businesses. CT1 was configured to receive stormwater inputs; however, we sampled effluent from this plant during dry weather conditions. CT1 is a mid-sized conventional treatment plant with primary and secondary clarification and aerated activated sludge tanks, as typical of many US treatment plants. CT2 is typical of many small-sized
286
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 8 4 e2 9 4
Table 1 e Overview of wastewater treatment plant operation. Treatment plant CT1
CT2
Influent sources
Ave flow rate
Primary clarification
Mainly residential, commercial office buildings Residential communities
60 MGD
Yes
0.4 MGD
No
CT3
Mainly residential, commercial office buildings
24 MGD
Yes
CT4
Mainly residential, some research laboratories
1.5 MGD
No
CT5
Mainly residential, commercial office buildings
3 MGD
Yes
treatment plants in the US, operating with aerated activated sludge, but no primary clarification. Some nitrogen removal is likely achieved in this plant because the aeration tanks are operated with lower than average dissolved oxygen levels. CT3 is representative of a mid-sized conventional treatment plant that practices nitrogen removal by carbon addition through the modified LudzackeEttinger process. Treated wastewater from CT3 is disinfected with UV light before discharge. CT4 also is typical of a small-scale conventional treatment plant with no primary clarification; however, the EIMCO carousel aeration tank design enables advanced nitrogen removal. CT5, a smallsized treatment plant that utilizes both primary and secondary clarification, is operated with the addition of powdered activated carbon to the aeration tanks that provides media for attached growth. With the exception of CT3, no other plants were operating disinfection treatments prior to effluent discharge at the time of our sample collection.
2.3.
Effluent collection and storage
Wastewater was collected as a grab sample at the discharge point of the treatment plant following disinfection (CT4, other plants were not operating seasonal disinfection). Effluent samples were transported to the lab in opaque 50 L Nalgene carboys, filtered using Whatman GF/A (Maidstone, England) glass fiber filters (1.6 mm), acidified to pH 2 using HCl, and stored in the refrigerator at 2 C within 24 h. Samples were isolated on the Amberlite resins within one week. The isolated fractions were stored acidified (pH 2) and at 2 C, and characterized within four weeks. Between sampling rounds, the carboys were washed with 3 consecutive high-purity water washes to remove residual water from the previous sample.
2.4.
Organic matter isolation
The resin cleaning and organic matter isolation methods using non-ionic DAX 8 and cation exchange resins were adopted from previous studies (Thurman and Malcolm, 1981), with the following changes: organic matter was only extracted
Treatment technology Traditional activated sludge treatment with coarse bubble diffusers Traditional activated sludge treatment with mechanical aeration Activated sludge and modified LudzackeEttinger process, fine bubble diffusers in aeration tanks Activated sludge with EIMCO carousel aeration tanks, mechanical aeration Activated sludge treatment with granular activated carbon suspended growth media
Denitrification
Disinfection
No designed denitrification process Nitrification and denitrification due to low DO in aeration tanks Nitrification and denitrification by modified Ludzacke Ettinger process Nitrification and denitrification by aerobic and anaerobic zones of carousel No designed denitrification process
None
None
UV
None
None
once on non-ionic resins, humic acids were not precipitated from the extracted eluent, and extracted eluent was analyzed directly without lyphilization. In addition, an XAD4 resin column was used to capture a portion of the less hydrophobic fraction of organic matter that is not retained on the DAX 8 resin. The XAD4 resin was prepared in the same manner as the DAX 8 resin. The DAX8 and XAD4 resin columns were connected in series using Teflon tubing. Effluent samples were pumped through the columns at a rate of 15 bed volumes per hour. Approximately 50 bed volumes of each effluent sample were extracted to ensure that a significant amount of organic matter was sorbed. The organic matter was back-eluted from each column individually using 0.1 M potassium hydroxide. Organic matter extracts were deionized to remove cations by pumping the eluted extracts through a strong Hþ cation exchange resin column. The organic matter eluted from the DAX8 resin is the hydrophobic (HPO) fraction, and the organic matter from the XAD4 resin is the transphilic (TPI) fraction. The organic matter that passed through both columns constituted the hydrophilic (HPI) fraction. The deionized, pH 2 EfOM extracts were used for the subsequent analyses. All glassware were cleaned using Liquinox laboratory soap, followed by repeated rinses with high-purity water.
2.5.
Dissolved organic carbon analysis
Dissolved organic carbon (DOC) was measured with a Tekmar Apollo 9000 combustion organic carbon analyzer. Carbon concentrations in the collected wastewater effluent samples were measured after the sample was filtered and acidified. Carbon concentrations of the HPO and TPI fractions were measured after cation exchange resin treatment. A homogenized sample of the HPI fraction was collected after the entire effluent sample had been extracted. The percentage of organic matter in each fraction was calculated on a per mass basis (Table 2). The total percent recovery of the CT3 treatment plant was much lower than the other four plants, possibly because the HPI fraction sub-sample for DOC analysis was not representative.
287
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 8 4 e2 9 4
Table 2 e Initial effluent organic carbon concentration and mass percentage of effluent organic matter in resin fractions. Location
Effluent DOC (mgC/L)
%Hydrophobic (HPO)
%Transphilic (TPI)
%Hydrophilic (HPI)
6.22 5.57 5.29 4.2 5.93
30.4 19 19.3 42.3 18.8
9.9 11.2 9.6 11.9 6.5
59.6a 79 40.3 57.1 80
CT1 CT2 CT3 CT4 CT5
% Recovered
109.2 69.2 111.3 105.3
a Calculated by difference.
2.6.
Fluorescence and ultravioletevisible spectroscopy
2.8.2.
Excitationeemission matrices (EEMS) were obtained for each sample using a Cary Eclipse (Australia) fluorescence spectrometer with a xenon flash lamp, slit widths of 5 nm and a scan rate of 1200 scans per second. The excitation wavelength was scanned from 200 to 450 nm in 10-nm increments, and the emission wavelength from 250 to 550 nm in 2-nm increments. Samples were analyzed on the fluorometer directly following isolation without dilution or pH adjustment (pH w2) (McKnight et al., 2001). Absorbance spectra for all isolated EfOM samples were obtained at wavelengths between 200 and 550 nm using a Cary 100 spectrophotometer.
2.7.
High pressure size exclusion chromatography
Extracted organic matter molecular weights were obtained using size exclusion chromatography on 50 mL aliquots (Hewlett Packard 1050 series HPLC). The mobile phase was 0.002 M, pH 6.8 phosphate buffer with 0.1 M NaCl (Cabaniss et al., 2000; Chin et al., 1994). A Protein Pak 125 column and matched guard column were purchased from Waters (Milford, MA). Five PSS standards (1 KDa, 1.8 KDa, 4.6 KDa, 18 KDa, 35 KDa) and acetone were used as calibration standards. The diode array detector was set to analyze the PSS standards and acetone at 224 nm and set to 254 nm for organic matter samples.
2.8.
Data analysis
2.8.1.
Organic matter molecular weight
The average molecular weights of the organic matter were determined from size exclusion chromatograms using the following two equations for number-averaged (Mn), and weight-averaged (Mw) molecular weight (Cabaniss et al., 2000; Chin et al., 1994): X N N X Mn ¼ hi ðhi =Mi Þ (1) i¼1
Mw ¼
N X i¼1
i1
X N hi ðMi Þ hi
(2)
i¼1
where hi is the height of the peak corresponding to the compound molecular weight Mi. To ensure that these equations gave accurate values of the average molecular weights: (1) the baseline was corrected by subtraction of a constant intensity across the entire chromatogram, and (2) the low molecular weight cutoff was set to 50 Da to minimize sensitivity to chromatogram tailing at low molecular weights (Cabaniss et al., 2000). The high molecular weight cutoff was set to 3200 Da for all samples, based upon the range of molecular weights observed among our sample set.
EEMS correction
Several corrections of EEMS data are necessary to obtain fluorescence spectra that are inter-comparable among datasets. Excitation intensity corrections were not necessary because our instrument was equipped with a red photomultiplier tube that provides excitation spectra that are not very distorted (Lakowicz, 1999). An emission intensity corrections array was made by comparing a solution of 103 M quinine sulfate solution in 0.05 M sulfuric acid that was scanned daily (Ex 346.5 nm and Em 384e667 by 1 nm) and compared with tabulated data (Lakowicz, 1999) Differences in intensity between our instrument and tabulated values were used to create a wavelength intensity correction matrix that accounted for changes in lamp emission intensity over time. Since excitation intensity corrections were not needed, the wavelength intensity matrix was the array of emission corrections multiplied by an array of ones. EEMS were corrected for the inner filter effect, according to (McKnight et al., 2001; Murphy et al., 2010): Fcorr ¼
Fmeas 10ðAexit þAemit Þ
(3)
where Fcorr and Fmeas are the corrected and measured fluorescences, respectively, and Aexit and Aemit are the sample absorbances at the excitation and emission wavelengths, respectively. The inner filter correction was applied to all fluorescence measurements in each EEM. Raman scatter was removed from fluorescence spectra by using a high-purity water blank analyzed using the same parameters as the samples. The blank was multiplied by the wavelength intensity correction matrix, and the resulting Raman scatter matrix was subtracted from each sample EEM. The subtraction of a water blank was not sufficient to remove scattered light when the excitation and emission wavelengths were equal and when the emission wavelength was twice the excitation wavelength. These light scattering interferences were removed manually by inserting a value of zero for intensity into an EEM when the excitation and emission wavelengths were within 8 nm of each other, and when the excitation wavelength multiplied by 2 was within 8 mm of the emission wavelength. Zero intensity values were also inserted into the EEM when the excitation wavelength was longer than the emission wavelength. Lastly, the EEM was normalized to Raman Units to remove any instrument-specific artifacts by dividing by the area under the Raman peak of the water blank at excitation 350 nm and emission 381e426 nm. Commonly, EEMS are measured with an excitation wavelength range from 200 nm to 450 nm; however, it was necessary to truncate data below 230 nm. At the concentrations of
288
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 8 4 e2 9 4
organic matter that were used in this study, the absorbance of light below 230 nm was above 1, indicating that less than 1% of light was transmitted through the sample. This high absorbance introduced a very large inner filter correction factor. Combining this large correction factor with the large amount of fluorometer noise at low wavelengths caused large noise peaks in the EEMS. A cutoff at 230 nm allowed the B, T1, and A peaks to be measured without excessive noise.
2.8.3.
Excitationeemission matrix analysis
The Fluorescence Regional Integration (FRI) method was used to quantitatively compare EEMS spectra among samples (Chen et al., 2003). The FRI method separates an EEM into five regions based on observed fluorescence peaks A, B, C, and the two T peaks (for example designation, see Fig. 2). The regions were separated with a horizontal line at 250 nm and two vertical lines at 330 and 380 nm. A diagonal line intersects all the points where emission wavelength equals excitation wavelength. The volume, Fi, under the EEM in each region was calculated using the following equation for discrete measurements (Chen et al., 2003): Fi ¼
XX ex
Iðlex ; lem ÞDlex Dlem
(4)
em
where I is the intensity at any excitationeemission pair, Dlex is the excitation wavelength interval (10 nm) and Dlem is the emission wavelength interval (2 nm) between observations. Since each of the five regions had a different area, the total volume of each region was normalized with the use of a multiplication factor given as the inverse of the area of that region divided by the total area (Table S3, Supplemental Information). The resulting values were average fluorescence response per unit area in each region. Lastly, the average fluorescence response of each region was converted to a percent fluorescence response.
2.8.4.
Optical methods
Three optical methods were used to characterize EfOM. Specific ultraviolet absorbance (SUVA, L mg1 m1) was calculated using the following formula (Weishaar et al., 2003): SUVA ¼
100 A254 DOC
(5)
where A254 is the sample absorbance measured at 254 nm measured in a 1-cm cuvette, and DOC is measured in mgC/L. Fluorescence index was calculated by dividing the fluorescence intensity measured at excitation 370 nm and emission 450 nm by the intensity measured at excitation 370 nm and emission 500 nm (McKnight et al., 2001). The E2/E3 ratio was calculated by dividing the sample light absorbance at 254 nm by the absorbance at 365 nm (Dalrymple et al., 2010).
3.
Results and discussion
3.1.
Effluent dissolved organic carbon
The percent of effluent organic matter (EfOM) isolated in the hydrophobic (HPO) fraction of the wastewater samples was less than typically found for natural organic matter isolates.
All of our effluent samples had similar concentrations of dissolved organic carbon (DOC), ranging from 4 to 6 mgC/L (Table 2). Between 18 and 42% of EfOM was extracted as the HPO fraction (Table 2). These fractions of HPO organic matter are considerably less than the 50e80% of natural DOM from rivers and lakes that is recovered as the HPO fraction (Thurman, 1985). However, the percentages of HPO EfOM extracted from our effluent samples are similar to previous studies with treated municipal wastewaters that showed the HPO fraction to account for 26e40% of the EfOM (Baker, 2001; Krasner et al., 2009; Park et al., 2010; Pernet-coudrier et al., 2008). An additional 9e12% of the EfOM in our samples was extracted in the transphilic (TPI) fraction (Table 2). The TPI fraction is typically not quantified during natural DOM characterization. Our distributions of EfOM in the TPI fractions were also consistent with previous observations that have shown TPI fractions in wastewater EfOM to be as high as 20% (Krasner et al., 2009; Pernet-coudrier et al., 2008). The majority (>50%) of our EfOM was not retained by either the HPO or the TPI resin, but recovered in the hydrophilic (HPI) fraction (Table 2). Thus, one primary difference between DOM in wastewater effluent and natural systems is the much smaller proportion of organic matter that is operationally defined as hydrophobic. Our separation of EfOM fractions highlights a shortcoming of conventional resin extraction techniques for full characterization of organic matter from effluent sources. Resin extraction techniques that have been so effective for isolating natural DOM only capture between 30 and 50% of the organic matter in effluent samples. We focused our EfOM characterization efforts on the HPO and TPI fractions so that we could undertake comparisons of EfOM characteristics with characteristics of natural DOM that had been isolated with the same technique. HPO constitutes the bulk of organic matter in natural systems and it has been well characterized because of its potential role in phototransformation reactions, photooxidant generation and contaminant transport (Dalrymple et al., 2010; Guerard et al., 2009; Haitzer et al., 2003). Our resin isolation results (Table 2) do highlight the large fraction of hydrophilic organic matter that is present in wastewater effluent and that ultimately may be extracted using preconcentration techniques (Pernet-coudrier et al., 2008). We note that cation exchange resin treatment of our organic matter extracts reduced the iron and aluminum concentrations to less than 2.6 mg/mgC. We anticipate that these amounts of bound metals will not impart a shift in the size or optical characteristics of our organic matter, based on their similarity to the background levels of iron and aluminum in the work of Pullin et al. (2007).
3.2.
Organic matter size
Size exclusion chromatography showed isolated EfOM to have distinctly different characteristics from natural DOM. First, the chromatograms for the wastewater samples were multimodal, with two characteristic features (Fig. 1): (i) the presence of 5 (HPO) or 6 (TPI) distinct ‘peaks’ between 100 and 1000 Da, and (ii) the presence of a separated peak at 50,000 Da in CT2 and CT4. In contrast, size exclusion chromatograms of natural DOM are typically unimodal, with a single broad peak between 100 and 10,000 Da (Cabaniss et al., 2000; Chin et al., 1994).
289
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 8 4 e2 9 4
Fig. 1 e Size exclusion chromatograms of hydrophobic (left panels) and transphilic (right panels) fractions of CT2 (upper panels) and CT3 (lower panels) EfOM. Note that the chromatograms have different y-axis scales to show details of the peak distributions. Peak numbers shown on graph, hydrophobic fractions did not show peak 4.
Second, the number-averaged (Mn) and weight-averaged (Mw) molecular weights were smaller than observed for natural DOM. Our number- averaged molecular weights ranged from 290 to 360 Da for the HPO fraction (Table 3) whereas hydrophobic natural DOM typically has Mn values between 713 and 1360 Da (Cabaniss et al., 2000). Likewise, the Mw values for HPO EfOM ranged from 570 to 670 Da (Table 3), lower than the 1000e2310 reported for NOM (Cabaniss et al., 2000). Our measured polydispersities were between 1.5 and 2.2 (Table 3) which are consistent with reported data for natural sources of DOM, despite our overall lower averaged molecular weights. There was little variation between the average sizes of organic matter in our HPO and TPI fractions with the exception that weight-averaged molecular weights were about 100 Da higher on average for the HPO fraction than for the TPI fraction. Our EfOM size characteristics were similar to previous studies of EfOM molecular weights that reported Mn and Mw molecular weights of 400e500 and 650e850, respectively (Imai et al., 2002; Park et al., 2010). Thus, the HPO and TPI fractions of EfOM have lower averaged molecular weights than the corresponding fractions of organic matter isolated from natural sources. We excluded the 50,000 Da peak from our calculations of molecular weights for the CT2 and CT4 HPO fractions reported in Table 3. Isolated high molecular weight peaks have been observed in previous size exclusion chromatograms for
wastewater organic matter and are thought to consist of polysaccharide materials (Her et al., 2002; Imai et al., 2002). Inclusion of the high molecular weight peak in our molecular weight calculations greatly skews the average values, yielding a molecular weight (Mw) value of 4160 Da and polydispersity of 13.51 for CT2 and respective values of 1530 Da and 5.25 for CT4. The multimodal distribution of EfOM sizes (Fig. 1) followed a similar pattern for all wastewater samples. Such multimodal
Table 3 e Size characteristics of effluent organic matter in the 50e3200 Da range. Mw (Da)
Polydispersity
Hydrophobic fraction CT1 302.9 CT2 293.5 CT3 359.1 CT4 287.2 CT5 346.2
664.9 566.9 641.4 586.6 669.8
2.2 2 1.79 2.04 1.93
Transphilic fraction CT1 304.5 CT2 300.5 CT3 391.9 CT4 300.0 CT5 374.1
501.2 493.7 578.9 458.4 557.3
1.65 1.64 1.48 1.53 1.49
Location
Mn (Da)
290
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 8 4 e2 9 4
distributions of organic matter size have been observed in previous work investigating wastewater organic matter characteristics (Her et al., 2002; Imai et al., 2002). In our study, the size exclusion chromatograms consistently showed five peaks in the HPO fractions for all samples and six peaks in the TPI fractions (Fig. 1) between 1000 and 100 Da. Certain chromatograms did not have a sharp peak at all locations, but did exhibit a shoulder, indicating that a small peak was present. We calculated the molecular weight at the maximum intensity of each peak to examine whether any trends in peak masses were observed between effluent locations (Table S1, Supporting Information). The molecular weights of each of the chromatogram peaks were all very similar (5 to 20 Da) across the different treatment plants within and between organic matter fraction types. The appearance of discrete peaks in the low molecular weight organic matter suggests the presence of distinct organic byproducts/biomolecules in the wastewater treatment process. Furthermore, the fact that the peak molecular weights were remarkably consistent among the samples, despite the variation in treatment plant size and technologies employed (Table 1), suggests common characteristics of organic matter in treated wastewater effluent sources. We examined absorbance spectra of each peak of the size exclusion chromatograms to identify whether they may suggest similarities in chemical characteristics of EfOM between treatment plants. We collected absorbance scans from 220 to 350 nm whenever a peak was detected in a chromatogram (peak numbers are shown in Fig. 1). Peak 6 was not detected by the instrument, so absorption spectra were not available. The shape of the absorbance curves for peaks 1 through 4 in both the HPO and TPI fractions were very similar (peak numbers shown on Fig. 1). Each showed a steep, exponential-like decrease in absorbance with increasing wavelength from 220 nm and with a small shoulder at about 272 nm (Fig. S1, Supporting Information). Peak 5 differed from peaks 1e4 in the absence of the 272 nm shoulder and a steeper monotonic decrease in absorbance from 220 nm to 350 nm. To quantify decreases in peak absorbance spectra, the absorbance at 220 nm was divided by the absorbance at 272 nm (Table S2, Supporting Information). Typical values for peaks 1 through 4 ranged from 1.9 to 3.4, whereas the ratio for peak 5 ranged between 1.7 and 60. A sharper slope in absorbance for peak 5 is consistent with the organic matter in peak 5 being characterized by less carbonecarbon bond conjugation (little absorbance at longer wavelengths). This interpretation is reasonable given that the organic matter in peak 5 had a molecular weight of only 170 Da.
3.3.
Organic matter fluorescence
The EEMS of wastewater effluent fractions showed fluorescence signatures that are not found in natural DOM extracts. The EEMS for all samples and fractions of EfOM showed fluorescence in regions A and C that are characteristic of natural DOM; however, EfOM EEMS also showed fluorescence in regions B and T1 (Fig. 2). Comparisons of our EEMS with natural DOM samples are hampered somewhat by the bounds of conventional analyses that truncate excitation wavelengths below 250 nm. Nevertheless, there is some evidence
that peaks T1 and B are not typically present in natural DOM since studies of organic matter without 250-nm wavelength truncation showed low fluorescence intensity in the regions of peaks B and T1 (Baker, 2001; Holbrook et al., 2005). Suwannee River shows very little fluorescence in the Peaks B and T1 regions (Chen et al., 2003). To our knowledge, all published EEM data of microbial end member DOM is truncated at excitation wavelengths above 250 nm, so we were not able to undertake comparisons with these natural DOM sources. We observed different characteristics between the EEMS of the HPO fractions and TPI fractions for our samples (Fig. 2). Within our samples, the EEMS spectra of all HPO fractions showed similar characteristics to each other, and likewise the TPI fractions showed similar characteristics to each other. One primary difference between the HPO and TPI fractions was the occurrences of peaks A and T. In the HPO fraction, EEMS peaks A and T1 did not show distinct separation. This was particularly evident for the HPO fraction from CT5 (Fig. 2) in which fluorescence intensity was high through the range of excitation wavelengths between 230 and 250 nm and emission wavelengths between 330 and 450 nm. In contrast to the HPO fraction, the EEMS for the TPI fractions had a reduced fluorescence in the T1 region, and a distinct peak in the A region. The qualitative differences in the EEMS matrix contours were substantiated quantitatively with the integration of the fluorescence peak regions (Table 4). The HPO fractions had a smaller percentage contribution of peak C, as compared to the TPI fraction. Previous work has shown that peaks A and C are the two fluorescence regions for fulvic acids (Baker, 2001; Chen et al., 2003; Hudson et al., 2007). Since peaks A and C are both larger for the TPI EEMS, two explanations are possible: either fulvic acids are more strongly retained by the XAD4 resin, or fulvic acids are retained equally by both resins, and other fluorophores are more strongly retained on the DAX8 resin. We were unable to compare our regional fluorescence distributions to prior studies of EfOM because our choices of wavelength boundary differed from prior studies (Chen et al., 2003) as a result of excitation wavelength truncation at 230 nm to eliminate excess noise. Fluorescence index measurements indicated that the extracted EfOM was similar in characteristics to microbially derived natural DOM. All fluorescence indices fall within the ranges of the natural organic matter continuum (Table 5). In addition, all fluorescence indices were above 1.5, indicating that the organic matter from the all five treatment plants are closer in characteristics to NOM at the microbial end of the continuum (fluorescence index, 1.9; McKnight et al., 2001), regardless of hydrophobicity. The indices of the TPI fractions of organic matter are larger than those of the HPO fraction, thus indicating that the TPI fraction had fluorescence index characteristics somewhat closer to natural DOM microbial source end members than did the HPO fraction.
3.4.
Optical analyses
Like the fluorescence characterization, SUVA values also indicated that the EfOM had similar aromaticity to microbially derived natural organic matter of microbial origin. The SUVA values for all of the organic matter fractions were between 1.5 and 3 L mg1 m1, with the exception of the CT2 TPI fraction
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 8 4 e2 9 4
291
Fig. 2 e EEMS of CT2 (upper panels) and CT5 (lower panels) treatment plants. Hydrophobic fractions are shown in the left panels and transphilic fractions are shown in the right panels. The letters on the CT2 hydrophobic EEM represent fluorescence regions (Coble, 1996).
(Table 5). SUVA values of natural DOM typically range from 1.7 to 5.2 L mg1 m1 with lower values reported for organic matter from microbial sources (Weishaar et al., 2003). The SUVA values for our extracted EfOM were consistent with SUVA measurements from prior EfOM characterization studies (Imai et al., 2002; Krasner et al., 2009; Park et al., 2010). We observed the SUVA value of the HPO fractions for each treatment plant to be larger than their corresponding TPI fraction. Such differences suggest that the HPO fractions may have more aromatic character than the TPI fractions.
Table 4 e Percentage distribution of different EEMs regions by the FRI method. Location
B (%)
T1 (%)
A (%)
T2 (%)
C (%)
Hydrophobic fraction CT1 10.31 CT2 10.81 CT3 9.46 CT4 7.86 CT5 7.85
36.12 45.31 39.85 34.4 40
37.03 27.27 34.2 37.88 35.49
5.49 6.07 5.42 6.04 5.11
11.06 10.54 11.07 13.81 11.56
Transphilic fraction CT1 10.41 CT2 11.86 CT3 10.72 CT4 8.52 CT5 9.55
21.18 31.85 24.87 26.47 27.53
47.2 33.46 41.36 44.5 41.11
5.57 7.8 6.43 4.83 6.58
15.64 15.29 16.62 15.68 15.24
While the SUVA values for individual EfOM samples indicated similarity to natural organic matter from microbial origins, the pooled data suggest some differences in optical characteristics between organic matter from effluent sources and natural sources. Prior work has indicated a strong correlation between natural DOM size (Mw) and molar absorptivity at 280 nm with a slope of 3.5 Da cm M1 (Aldrich humic acid omitted) (Chin et al., 1994). Our pooled effluent organic matter samples also showed a correlation between Mw and molar absorptivity at 280 nm, but exhibited a much lower slope of 0.9 Da cm M1 (Fig. S2). Such a lower slope suggests that EfOM has fewer chromophore groups per ‘molecule’ than does natural organic matter. The absorbance spectral slopes of our isolated EfOM, as quantified with the E2/E3 ratio, also were similar to natural DOM from microbial sources. Our EfOM E2/E3 ratios (Table 5) fell with the range of values between 3.1 and 7 that have been reported for natural DOM (Dalrymple et al., 2010). Unlike the other optical characteristics, E2/E3 ratios do not exhibit a clear separation between microbial and terrestrial end member sources; however, low E2/E3 ratios are consistent with organic matter structures, such as lignin precursors, that can form charge transfer complexes that are manifest in the long wavelength ‘tail’ of organic matter absorption spectra (Del Vecchio and Blough, 2004). Thus, the high E2/E3 ratios that we observed for our isolated EfOM samples would suggest the lack of terrestrial organic matter precursors and hence, characteristics that were more similar to natural organic matter of microbial origin.
292
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 8 4 e2 9 4
Table 5 e Fluorescence and optical properties, and metal contents of effluent organic matter. Location
pH
Fluorescence index
SUVA (L mg1 m1)
E2/E3
Fe (mg mg1 C )
Al (mg mg1 C )
Hydrophobic fraction CT1 CT2 CT3 CT4 CT5
3.93 2.43 3.18 1.9 2.57
1.57 1.69 1.57 1.71 1.64
2.09 2.08 3.04 2.09 2.81
5.44 4.85 5.33 5.69 4.98
NM 0.97 1.4 NM 0.64
NM 0.26 0.65 NM 0.22
Transphilic fraction CT1 CT2 CT3 CT4 CT5
2.92 2.55 3.11 2.54 2.57
1.83 1.94 1.7 1.93 1.82
2.06 0.86 1.58 1.71 2.33
5.73 4.05 4.79 5.28 4.82
NM 2.6 0.37 NM 0.45
NM 1.6 0.28 NM 1.7
3.5.
Treatment plant intercomparison
Overall, our characterization of isolated EfOM indicated great similarity between treatment plants. A few differences were observed. The CT1 and CT4 extracted effluent had a greater fraction of EfOM that was hydrophobic in character (Table 2); however, the physical and chemical characteristics of these EfOM fractions fell within the range of the other three plants. CT2 and CT4 did show the presence of a high molecular weight organic matter (50,000 Da) peak in the HPO fraction which may contribute to the greater fraction of organic matter mass in the HPO fraction for these samples. The EEMS fluorescence regional integration of CT2 showed a much higher T:A ratio in the HPO fraction and higher T:A ratio in the TPI fraction that the other plants (Table 4). Such a difference was not observed for CT4, suggesting that the differing fluorescence characteristics for CT2 were not a result of the apparent high molecular weight component. Fluorescence indices were somewhat lower for CT1 and CT3 in the HPO fraction than the other plants (Table 5). SUVA values were greater for the HPO fractions of the CT3 and CT5 plants (Table 5). These results suggest that there is a somewhat greater range in characteristics for the hydrophobic than the hydrophilic fractions of effluent organic matter. The small differences in EfOM characteristics among our samples could not be linked generally to plant operating conditions. CT2 and CT4 both utilize mechanical aerators in activated sludge treatment, whereas the other plants deliver oxygen via bubble diffusers. It is possible that differences in shear characteristics between the two aeration treatments could cause the release of extracellular polymeric substances from floc in the presence of mechanical aerators. High SUVA values have been associated with advanced nitrogen removal, however, we did not observe a clear trend in our dataset. CT3 and CT4 both employ advanced nitrogen removal, yet only CT3 showed high SUVA in the hydrophobic fraction. CT5 also had a high SUVA value for the HPO fraction, although advanced nitrogen removal is not an explicit treatment step at that plant. Finally, some change in EfOM characteristics might have been expected with ultraviolet disinfection at the CT3 treatment plant. Studies of drinking water disinfection suggest that UV treatment reduces the SUVA value of finished waters and may reduce the proportion of lower molecular
weight organic matter (Lamsal et al., 2011). No obvious trend of lowered SUVA values (Table 5) or higher molecular weights (Table 3) were observed for the CT3 plant, relative to the other treatment plants. Nor, did the use of primary clarification (or not) appear to indicate trends in any EfOM characteristics. Thus, wastewater treatment plant operating conditions appear to have only a minor effect on EfOM characteristics.
3.6.
Environmental significance
The results of our comprehensive characterization of extracted EfOM from a number of treatment plant sources highlight closer similarities in EfOM characteristics to natural DOM from microbial sources than from terrestrial sources. Consequently, by most of our organic matter characterization techniques, EfOM appears to group closer to the microbial end member of the natural organic matter continuum. The one exception is the occurrence of A and T peaks in EfOM that are apparently not observed in DOM from natural sources. That most EfOM characteristics are similar to microbial DOM sources, indicates that streams receiving significant (e.g., >10% total flow) effluent discharges will contain DOM with a mixture of characteristics since DOM in stream systems is typically dominated by allochthonous, terrestrial sources of organic matter. What implications that this may have with respect to fate of chemical contaminants released in the effluent is yet unknown, as little work has been done to evaluate how sorption coefficients or indirect photolysis rates may be similar to, or different from, natural sources of DOM. Certainly, the similarity between EfOM characteristics from treatment plants that are representative of those found nationwide in the US suggests that understanding of EfOM effects on stream ecological and contaminant fate processes may be readily translated to other systems with wastewater inputs.
4.
Conclusions
Our results show that effluent organic matter characteristics differ from terrestrial organic matter that is typically present in small to moderately sized rivers.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 8 4 e2 9 4
1. Wastewater effluent discharges introduce organic matter with distinctly different characteristics than typically present in small river systems. First, the hydrophobic and transphilic components of EfOM are closer in characteristics to natural DOM derived from microbial sources; whereas, natural stream water DOM tends to be dominated by terrestrial sources. 2. Second, wastewater EfOM contains between 60 and 80% material that is operationally defined as hydrophilic organic matter for which the environmental implications (e.g. carbon cycling, contaminant fates) are unknown. 3. Our results show that extracted effluent organic matter samples from plants have very similar characteristics. This work, which establishes beginning ranges of EfOM characteristics, highlights some questions about the ranges in EfOM characteristics that remain unanswered. For example, more study is required to evaluate the effects of factors such as influent type (e.g., large industrial contributions), treatment plant residence time, and stormwater contributions may have on the characteristics of EfOM.
Acknowledgment Funding for this project was provided by Connecticut Sea Grant project number PD-10-07 and by the UConn Environmental Engineering Program and Civil and Environmental Engineering Department. We thank the following wastewater treatment plant operators for permission and assistance in collecting samples that were used in this work: Arnie Bevins, Gene Ely, Kenneth Pelzar, Marylee Santoro, and Tom Tyler. We also thank two anonymous reviewers for their thoughtful comments and critique of an earlier version of this manuscript.
Appendix A. Supporting information Supporting information associated with this article contains: (i) organic matter sizes for SEC chromatogram peaks, (ii) absorption spectra at peaks of size exclusion chromatograms, (iii) absorbance intensity ratios for SEC chromatogram peaks, and (iv) parameters for fluorescence regional integration method, and (v) molar absorptivity at 280 nm plotted against weight-averaged molecular weight and can be found in the online version at doi:10.1016/j.watres.2011.10.010.
references
Allan, J.D., Castillo, M.M., 2007. Stream Ecology. Springer, Dordrecht, The Netherlands. Baker, A., 2001. Fluorescence excitation-emission matrix characterization of some sewage e impacted rivers. Environmental Science and Technology 35 (5), 948e953. Brooks, B.W., Riley, T.M., Taylor, R.D., 2006. Water quality of effluent-dominated ecosystems: ecotoxicological, hydrological, and management considerations. Hydrobiologia 556, 365e379.
293
Cabaniss, S.E., Zhou, Q., Maurice, P.A., Chin, Y.-P., Aiken, G.R., 2000. A log-normal distribution model for the molecular weight of aquatic fulvic acids. Environmental Science and Technology 34 (6), 1103e1109. Chen, W., Westerhoff, P., Leenheer, J.A., Booksh, K., 2003. Fluorescence excitation e emission matrix regional integration to quantify spectra for dissolved organic matter. Environmental Science and Technology 37 (24), 5701e5710. Chin, Y.-P., Aiken, G.R., O’Loughlin, E., 1994. Molecular weight, polydispersivity, and spectroscopic properties of aquatic humic substances. Environmental Science and Technology 28 (11), 1853e1858. Chin, Y.P., Aiken, G.R., Danielsen, K.M., 1997. Binding of pyrene to aquatic and commercial humic substances: the role of molecular weight and aromaticity. Environmental Science & Technology 31 (6), 1630e1635. Coble, P.G., 1996. Characterization of marine and terrestrial DOM in seawater using excitation emission spectroscopy. Marine Chemistry 51 (4), 325e346. Cory, R.M., Miller, M.P., McKnight, D.M., Guerard, J.J., Miller, P.L., 2010. Effect of instrument specific response on the analysis of fulvic acid fluorescence spectra. Limnology and Oceanography Methods 8, 67e78. Croue, J.-P., Benedetti, M.F., Violleau, D., Leenheer, J.A., 2003. Characterization and copper binding of humic and nonhumic organic matter isolated from the South Platte River: evidence for the presence of nitrogenous binding site. Environmental Science and Technology 37 (2), 328e336. Dalrymple, R.M., Carfagno, A.K., Sharpless, C.M., 2010. Correlations between dissolved organic matter optical properties and quantam yield of singlet oxygen and hydrogen peroxide. Environmental Science and Technology 44, 5824e5829. Del Vecchio, R., Blough, N.V., 2004. On the origin of the optical properties of humic substances. Environmental Science and Technology 38, 3885e3891. Doll, T.E., Frimmel, F.H., 2003. Fate of pharmaceuticals e photodegradation by simulated solar UV-light. Chemosphere 52, 1757e1769. EPA, 2008. Clean Watersheds Needs Survey. E.P.A, U.S, Washington DC. Guerard, J.J., Miller, P.L., Trouts, T.D., Chin, Y.-P., 2009. The role of fulvic acid composition in the photosensitized degradation of aquatic contaminants. Aquatic Sciences 71, 160e169. Haitzer, M., Aiken, G.R., Ryan, J.N., 2003. Binding of mercury(II) to aquatic humic substances: influence of pH and source of humic substances. Environmental Science and Technology 37 (11), 2436e2441. Her, N., Amy, G., Foss, D., Cho, J., Yoon, Y., Kosenka, P., 2002. Optimization of method for detecting and characterizing NOM by HPLC e size exclusion chromatography with UV and on-line DOC detection. Environmental Science and Technology 36 (5), 1069e1076. Holbrook, R.D., Breidenich, J., Derose, P.C., 2005. Impact of reclaimed water on select organic matter properties of a receiving stream e fluorescence and perylene sorption behavior. Environmental Science and Technology 39 (17), 6453e6460. Hudson, N., Baker, A., Reynolds, D., 2007. Fluorescence analysis of dissolved organic matter in natural, waste and polluted waters e a review. River Research and Applications 23, 631e649. Imai, A., Fukushima, T., Matsushige, K., Kim, Y.-H., Choi, K., 2002. Characterization of dissolved organic matter in effluents from wastewater treatment plants. Water Research 36, 859e870. Krasner, S.W., Westerhoff, P., Chen, B., Rittmann, B.E., Nam, S.-N., Amy, G., 2009. Impact of wastewater treatment processes on organic carbon, organic nitrogen, and DBP precursors in
294
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 8 4 e2 9 4
effluent organic matter. Environmental Science and Technology 43 (8), 2911e2918. Lakowicz, J.R., 1999. Principles of Fluorescence Spectroscopy, second ed. Kluwer Academic / Plenum Publishers, New York, NY. Lamsal, R., Walsh, M.E., Gagnon, G.A., 2011. Comparison of advanced oxidation processes for the removal of natural organic matter. Water Research 45, 3263e3269. Leenheer, J.A., 1981. Comprehensive approach to preparative isolation and fractionation of dissolved organic carbon from natural waters and wastewaters. Environmental Science and Technology 15 (5), 578e587. Lim, M.-H., Snyder, S.A., Sedlak, D.L., 2008. Use of biodegradable dissolved organic carbon (BDOC) to assess the potential for transformation of wastewater-derived contaminants in surface waters. Water Research 42, 2943e2952. McKnight, D.M., Boyer, E.W., Westerhoff, P., Doran, P.T., Kulbe, T., Anderson, D.T., 2001. Spectrofluorometric characterization of dissolved organic matter for indication of precursor organic matter and aromaticity. Limnology and Oceanography 46 (1), 38e48. Murphy, K.R., Butler, K.D., Spencer, R.G.M., Stedmon, C.A., Boehme, J.R., Aiken, G.R., 2010. Measurement of dissolved organic matter fluorescence in aquatic environments: an interlaboratory comparison. Environmental Science and Technology 44 (24), 9405e9412. Park, M.-H., Lee, T.-H., Lee, B.-M., Hur, J., Park, D.-H., 2010. Spectroscopic and chromatographic characterization of wastewater organic matter from a biological treatment plant. Sensors 10, 254e265. Pernet-coudrier, B., Clouzot, L., Varrault, G., Tusseauvuillemin, M.-H., Verger, A., Mouchel, J.-M., 2008. Dissolved organic matter from treated effluent of a major wastewater
treatment plant: characterization and influence on copper toxicity. Chemosphere 73, 593e599. Polubesova, T., Chen, Y., Navon, R., Chefetz, B., 2008. Interactions of hydrophobic fractions of dissolved organic matter with Fe3þ- and Cu2þ-montmorillonite. Environmental Science and Technology 42 (13), 4797e4803. Pullin, M.J., Anthony, C., Maurice, P.A., 2007. Effects of iron on the molecular weight distribution, light absorption, and fluorescence properties of natural organic matter. Environmental Engineering Science 24 (8), 987e997. Swayne, M.D., Boone, G.H., Bauer, D., Lee, J.S., 1980. Wastewater in Receiving Waters at Water Supply Abstraction Points. EPA, U.S., Washington, DC. Thurman, E.M., 1985. Organic Geochemistry of Natural Waters. Martinus Nijhoff/Dr W. Junk Publishers. Thurman, E.M., Malcolm, R.L., 1981. Preparative isolation of aquatic humic substances. Environmental Science and Technology 15 (4), 463e466. Wang, L., Wu, F., Zhang, R., Li, W., Liao, H., 2009. Characterization of dissolved organic matter fractions from Lake Hongfeng Southwestern China Plateau. Journal of Environmental Sciences 21, 581e588. Weishaar, J.L., Aiken, G.R., Bergamaschi, B.A., Fram, M.S., Fujii, R., Mopper, K., 2003. Evaluation of specific ultraviolet absorbance as an indicator of the chemical composition and reactivity of dissolved organic carbon. Environmental Science and Technology 37 (20), 4702e4708. Woods, G.C., Simpson, M.J., Kelleher, B.P., Mccaul, M., Kingery, W. L., Simpson, A.J., 2010. Online high-performance size exlusion chromatography e nuclear magnetic resonance for the characterization of dissolved organic matter. Environmental Science and Technology 44 (2), 624e630.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 9 5 e3 0 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Colloid retention at the meniscus-wall contact line in an open microchannel Yuniati Zevi a, Bin Gao b, Wei Zhang c, Vero´nica L. Morales a, M. Ekrem Cakmak d, Evelyn A. Medrano e, Wenjing Sang a, Tammo S. Steenhuis a,* a
Department of Biological & Environmental Engineering, Riley-Robb Hall, Cornell University, Ithaca, NY 14853-5701, USA Department of Agricultural and Biological Engineering, University of Florida, Gainesville, FL 32611-0570, USA c National Exposure Research Laboratory, Ecosystems Research Division, U.S. Environmental Protection Agency, Athens, GA 30605, USA d Department of Environmental Engineering, Cukurova University, Adana 01330, Turkey e Water Resources Engineering, Technical University of Delft, Delft, The Netherlands b
article info
abstract
Article history:
Colloid retention mechanisms in partially saturated porous media are currently being
Received 18 February 2011
researched with an array of visualization techniques. These visualization techniques have
Received in revised form
refined our understanding of colloid movement and retention at the pore scale beyond what
15 September 2011
can be obtained from breakthrough experiments. One of the remaining questions is what
Accepted 23 September 2011
mechanisms are responsible for colloid immobilization at the triple point where air, water,
Available online 25 October 2011
and soil grain meet. The objective of this study was to investigate how colloids are transported to the air-water-solid (AWS) contact line in an open triangular microchannel, and
Keywords:
then retained as a function of meniscus contact angle with the wall and solution ionic
Contact angle
strength. Colloid flow path, meniscus shape and meniscus-wall contact angle, and colloid
Retention
retention at the AWS contact line were visualized and quantified with a confocal micro-
Capillary forces
scope. Experimental results demonstrated that colloid retention at the AWS contact line was
Air-water-solid interface
significant when the meniscus-wall contact angle was less than 16 , but was minimal for the
Contact line
meniscus-wall contact angles exceeding 20 . Tracking of individual colloids and computa-
Colloids
tional hydrodynamic simulation both revealed that for small contact angles (e.g., 12.5 ),
Microchannel
counter flow and flow vortices formed near the AWS contact line, but not for large contact angles (e.g., 28 ). This counter flow helped deliver the colloids to the wall surface just below the contact line. In accordance with DLVO and hydrodynamic torque calculations, colloid movement may be stopped when the colloid reached the secondary minimum at the wall near the contact line. However, contradictory to the prediction of the torque analysis, colloid retention at the AWS contact line decreased with increasing ionic strength for contact angles of 10e20 , indicating that the air-water interface was involved through both counter flow and capillary force. We hypothesized that capillary force pushed the colloid through the primary energy barrier to the primary minimum to become immobilized, when small fluctuations in water level stretched the meniscus over the colloid. For large meniscus-wall contact angles counter flow was not observed, resulting in less colloid retention, because a smaller number of colloids were transported to the contact line. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ1 607 255 2489. E-mail addresses:
[email protected] (Y. Zevi),
[email protected] (B. Gao),
[email protected] (W. Zhang),
[email protected] (V.L. Morales),
[email protected] (M. Ekrem Cakmak),
[email protected] (W. Sang),
[email protected] (T.S. Steenhuis). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.09.046
296
1.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 9 5 e3 0 6
Introduction
Colloids facilitate the transport of a wide range of contaminants through the vadose zone and into groundwater. Some typical contaminants that experience colloid-facilitated transport and are of great concern for human and ecological health include: radionuclides, pesticides and trace metals (Saiers and Hornberger, 1999; Williams et al., 2006). While most colloid transport studies have been carried out under saturated conditions, fewer studies have characterized colloid retention processes in unsaturated soils. These studies found that, in addition to retention mechanisms occurring in saturated soils, e.g., pore straining, and attachment to grain surfaces by the overall Derjaguin-Landau-Verwey-Overbeek (DLVO) energy minima and/or surface chemical heterogeneity or roughness that produce locally attractive sites on an overall repulsive surface (Johnson et al., 2010), the presence of a gaseous phase in unsaturated soils forms other potential retention sites such as the air-water interface (AWI) and the air-water-solid (AWS) contact line (Bradford and Torkzaban, 2008). Early work in unsaturated processes reported colloid retention at the airwater interface (Wan and Wilson, 1994) and in thin water films enveloping grains (Wan and Tokunaga, 1997; Veerapaneni et al., 2000). Research in the last decade has utilized microscopy advances and identified two new retention sites: immobile water zones (Gao et al., 2006) and the AWS contact line, which is defined as the triple point where air, water and grain approach each other. Colloid attachment at the AWS contact line, which is the focus of this study, has been observed in soil media (Crist et al., 2004, 2005; Zevi et al., 2005, 2009) and in surrogate of soil pores (e.g., microchannels) (Lazouskaya et al., 2006; Lazouskaya and Jin, 2008). Capillary and DLVO forces have both been considered to explain colloid retention at this interface by Crist et al. (2005), Gao et al. (2008), and Shang et al. (2008). This study sets out to explain why colloids adhere at the AWS contact line. To date several studies have indicated that a variety of colloids attach at the AWS contact line, especially when the DLVO forces do not favor colloid attachment at the solid-water interface (SWI) (Scha¨fer et al., 1998a; Sirivithayapakorn and Keller, 2003; Auset and Keller, 2006; Lazouskaya et al., 2006; Crist et al., 2004, 2005; Zevi et al., 2005, 2009). The studies by Crist et al. (2004, 2005) were likely the first time that the colloids were visualized to attach at the AWS contact line. Other earlier studies might have also observed this phenomenon, but attributed it to the attachment at the air-water interface (Wan and Wilson, 1994; Sirivithayapakorn and Keller, 2003). Although a controversy exists regarding the role of evaporation at the AWS contact line (Steenhuis et al., 2005; Wan and Tokunaga, 2005), since then the colloid attachment at the AWS contact line has been observed by Lazouskaya et al. (2006), Lazouskaya and Jin (2008), Gao et al. (2008), Morales et al. (2009), and Zevi et al. (2009). Despite these observations there has not been a satisfactory explanation on how colloids can approach the often energetically repulsive solid-water and/or air-water interfaces. Hydrodynamic forces have been speculated to be responsible for supplying the necessary external energy to overcome the repulsive energy exerted at the AWS contact line (Lazouskaya et al., 2006). However, according to Shang et al.
(2008), drag forces at realistic groundwater flow rates are too weak to thrust suspended colloids through the energy barrier at the AWS contact line so that capillary force can more permanently immobilize the colloids. Although the source of energy that colloids use to approach this energetically unfavorable retention site remains uncertain, experimental observations indicate that colloid immobilization at the AWS contact line ubiquitously occurs. For now, its mechanistic understanding remains elusive. The studies of Gao et al. (2008) and Shang et al. (2008) indicate that capillary forces are generally at least two orders of magnitude greater than the system’s DLVO forces. A detailed theoretical description of DLVO and capillary forces is presented in the auxiliary material. O’Brien and van den Brule (1991) suggested that colloid attachment on a solid substrate is controlled by capillary force of the film covering the solid, and can be determined by the liquid-colloid and liquid-substrate contact angles. Despite this critical observation for colloid retention in unsaturated systems, many studies have explained colloid attachment by employing the DLVO theory exclusively (e.g., Auset and Keller, 2006; Crist et al., 2004, 2005; Lazouskaya et al., 2006; Scha¨fer et al., 1998a,b; Sirivithayapakorn and Keller, 2003). Numerical models for colloid retention in unsaturated porous media mainly consider the air-water interface as an additional retention site as compared to saturated media (Corapcioglu and Choi, 1996; Massoudieh and Ginn, 2007; Chen, 2008). Specific model for colloid attachment at the AWS contact line are unavailable; therefore, additional work in this area is needed. Shi et al. (2010) simulated a moving airwater interface in a microchannel. Other studies have used the finite element analysis software COMSOL in colloid transport studies (Cakmak et al., 2008; Kemps and Bhattacharjee, 2009; Torkzaban et al., 2007, 2008; Bradford and Torkzaban, 2008). The objective of this study was to experimentally visualize and quantify how colloids flowing through a simplified “soil pore” (i.e., microchannel) are retained at the AWS contact line under a wide range of DLVO forces (varied by solution ionic strengths) and capillary force (varied by the air-water interfacial shapes exhibiting different meniscus-wall contact angles).
2.
Materials and methods
2.1.
Colloid attachment experiment in a microchannel
The colloids used were hydrophilic (water contact angle ¼ 12 ), negatively charged, 1-mm yellow fluorescent synthetic polystyrene microspheres (Polysciences, Inc., Warrington, PA). Colloid suspensions of 9.1 106 microspheres/ mL were prepared in the solution of 0.001% Rhodamine B dye at three ionic strengths (i.e., 1 mM, 50 mM, and 100 mM NaCl), respectively. The very low concentration of Rhodamine B dye was added to allow for imaging the water phase and does not influence colloid behavior (Zevi et al., 2006, 2009). The colloidfree solutions of identical solution chemistry with that of the colloid suspensions were used as background solutions. Prior to each experiment the colloid suspensions were sonicated in
297
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 9 5 e3 0 6
an ultrasonic bath (Bransonic B-52, Branson Ultrasonics, Danbury, CT) for 5 min to ensure monodispersivity. The principal components of the experimental setup consisted of an open triangular microchannel, syringe pumps, and a confocal microscope, as illustrated in Fig. 1a. The microchannel made of borosilicate glass consisted of an isosceles triangular shape carved canal with a 90-degree angle, a length of 54-mm and a wall height of 5-mm (Fig. 1b and c). The channel piece was mounted on the confocal microscope stage with an inclination of 2 to prevent ponding of inflow liquid. The channel inlet and outlet were composed of small capillary reservoirs (10-mm long) that dispersed the flow from the tube to the channel cross-sectional area (Fig. 1b). Push and pull syringe pumps were used (Orion Sage Pump model 341B, Thermo Scientific; KDS model 120, KD Scientific, Inc., Holliston, MA) to maintain a constant and unpulsed flow into and out of the channel. The visualization system was a Leica TCS SP2 laser scanning confocal microscope equipped with an HC PL FLUOTAR CS objective (5.0 magnification) with a numerical aperture of 0.15, yielding a resolving power of 0.97 mm/pixel (Fig. 1a). Before commencing the experiment, the microchannel was cleaned by sonication in the ultrasonic bath with a low concentration of anionic glassware cleaner (Alconox Inc.) in distilled water for 15 min, rinsed with distilled water for five times followed by rinsing with ethanol and distilled water again, and finally dried at 65 C for 30 min. The cleaned microchannel mounted under the microscope was initially empty, and then the testing liquid was added and removed by the pumps, as illustrated in Fig. 1a. Two inlet syringes were
prepared containing the background solution and the colloid suspension, respectively. The microchannel was initially filled with the background solution using the inlet syringe pump at a high flow rate (10 mL/h) until the water front reached the outlet reservoir, whereupon the outflow pump was turned on to begin withdrawing the liquid until the water level reached equilibrium. Once the water level reached a uniform height of 2 mm along the channel, the flow rate of both pumps were reduced to 1 mL/h in order to further lower the water height to approximately 1.5 mm. Typically 15e30 min were required to obtain a continuous and uniform flow. The average linear flow velocity under a steady-state flow condition was 0.12 mm/s. Once at the steady-state, a pulse of the colloid suspension was introduced into the uninterrupted inflow by rapidly exchanging the inlet tubing from the syringe containing the colloid-free background solution to the second inlet syringe containing the colloid suspension. Initial experiments demonstrated that the contact angle of the meniscus (q) in the channel could be increased by covering the channel with a 24 50 mm glass microscope slide (Fisher Scientific). Therefore, for selected experiments, the contact angle at the AWS contact line was altered by covering the channel. For such experiments, the glass cover was placed on the top of the microchannel after the flow became uniform and continuous, and the rest of experimental procedures were the same as that described above. The length of the microchannel covered by the glass slide was 50 mm, representing 93% of the total coverable channel surface area. Experiments were conducted in duplicate or triplicate for conditions of open (without a cover) or closed channel (with a cover) at solution ionic strengths of 1, 50 and 100 mM NaCl (see Table 1 for a detailed experimental list). The parameter combinations resulted in six different treatments to compare a broad range of DLVO and capillary forces. To study the effect of covering the channel on the contact angle without flowing water, one static treatment was added where the channel was covered and the flow was stopped after the pulse of colloids was injected. With the confocal microscopy system, simultaneous imaging of the colloids, Rhodamine B dyed water and the microchannel was achieved using excitation/emission spectral channels set at 488/700 nm (Argon laser), 543/650 nm (HeNe laser), and transmitted light, respectively. A more detailed description of the instrumental setting is given by Zevi et al. (2005). During each experiment, two types of image sequences were recorded at a distance of 3 cm away from the inlet. Initially, a “xyz” image stack was recorded to construct a 3-D
Table 1 e Summary of experiments performed. Experiment
Fig. 1 e (a) Schematic diagram of the experimental setup, (b) Detailed view of the microchannel, (c) Schematic cross section of the microchannel.
Series 1 Series 2 Series 3 Series 4 Series 5 Series 6 Series 7
Solution
Condition
1 mM NaCl 50 mM NaCl 100 mM NaCl 1 mM NaCl 50 mM NaCl 100 mM NaCl 1 mM NaCl
Open channel Open channel Open channel Closed channel Closed channel Closed channel Open and closed channel
Flow Number of rate replicates 1 ml/h 1 ml/h 1 ml/h 1 ml/h 1 ml/h 1 ml/h Static
3 2 3 2 2 2 2
298
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 9 5 e3 0 6
cross section of the channel at the beginning of the experiment and subsequently used to measure the meniscus contact angle at the microchannel walls. From these images the meniscus-wall contact angle was determined using 2D tools, as described in detail by Zevi et al. (2009). The measurements were repeated for at least four different cross sections because of observed variation in the profile within a single channel. Additional measurements of the contact angle were made at succeeding stages of each experiment: prior to the injection of the colloid pulse, during the experiment if the meniscus shape was observed to have changed, and at the end of the experiment. During the experiment, but after the meniscus was well-established, a series of time sequenced images (xyt series) were collected to monitor the accumulation of colloids at the AWS contact line, with a frequency of 1e2 images per second. From these images, colloid retention was quantified by analyzing consecutive images in a given region using ImageJ software, as carried out by Zevi et al. (2005, 2009). In addition, two series of time sequenced images (xyt series) from two experiments described earlier (one with large contact angle of 28 at solution ionic strength 100 mM NaCl and one with a small contact angle of 12.5 at solution ionic strength 1 mM NaCl) were sufficiently clear to allow for tracing the path of the colloids near the AWS contact line with manual tracking by the ImageJ software. The period of particle tracking was 4 min for the experiment with the 12.5-degree contact angle and 40 min for the experiment with 28-degree contact angle. With the right calibration value for xy and interval time, the xy coordinates, travel distance and velocity of one colloidal particle between two successive images were automatic recorded by simply clicking the colloidal particle on the image window. Multiple colloids were tracked for each experiment. Based on these data, an overlay of the original stack and traced trajectories were generated, with a different color applied for each traced colloid.
2.2. Computational simulation of flow field in the microchannel The three-dimensional steady-state flow field in the triangular channel with 54 mm of length was simulated by numerically solving the Stokes equations using COMSOL Multiphysics v3.5a software package (COMSOL, Inc., Burlington, MA). Vp ¼ mV2 v
(1)
V$v ¼ 0
(2)
where p is the fluid pressure (Pa), n is the velocity of the water (m/s), m is the dynamic viscosity of the water (0.001 Pa s). The upper curved boundary, which accounts for the airwater interface (i.e., the water meniscus), was set as a slip boundary condition (i.e., the normal component of velocity is zero and the tangential component of total stress is zero, Zhang et al., 2010). The channel walls were specified as a noslip boundary condition (i.e., zero velocity). The contact angle between the meniscus and the channel wall (q) was set at 12.5 and 28 . The flow was induced by specifying a fluid velocity of 0.12 mm/s at the inlet and zero pressure at the outlet of the channel.
2.3. Calculations of DLVO and hydrodynamic forces and torques Colloid retention is greatly dependent on the forces and torques that act on the colloids (Bradford and Torkzaban, 2008; Bradford et al., 2011). Thus, DLVO forces (FDLVO) were calculated for colloids interacting with the SWI and AWI. The calculations of hydrodynamic drag force (FD), applied hydrodynamic torques (Tapplied), and DLVO adhesive torques (Tadhesive) were performed for the colloids attached on the SWI in a typical Poiseuille type flow field, and the detailed equations are provided in the auxiliary material. The balance of Tapplied and Tadhesive determines whether the colloids could remain immobilized on the SWI.
3.
Results
The results from a total of 14 experimental runs (6 treatments with 2 or 3 replicates, experimental series 1 to 6, Table 1) were used to analyze the shape of the meniscus by way of the meniscus-wall contact angle, and the degree of colloid attachment at the AWS contact line. The system variables included the solution ionic strength (1, 50, and 100 mM) and the meniscus-wall contact angle, which was controlled partly by the presence and absence of a channel cover. By employing this set of combinations, the DLVO forces were altered by the solution’s ionic strength and the capillary force by the meniscus-wall contact angle.
3.1.
Meniscus and contact angle effects
Examples of reconstructed cross section from the “xyz” image stack of the microchannel and fluid, depicting the shape of the meniscus 30 min after the colloid pulse was initiated, are given for the contact angle less than 10 (Fig. 2a), between 10 and 20 (Fig. 2b), between 20 and 40 (Fig. 2c), and greater than 40 (Fig. 2d). The red color in these figures was the Rhodamine B dyed water. Green yellowish colloids near the contact line in Fig. 2a and b were elongated, which is typical for bright objects in the reconstruction of cross-sectional image stacks. The yellow spots especially notable in the center of Fig. 2a were due to the artifacts of the confocal microscope. In Fig. 2eeh, the longitudinal channel view of colloid attachment near the AWS contact line is shown at the same 30-minute time period. Colloids can be found as green yellowish dots either in a straight line at the AWS contact line or scattered in the “red” water. The ones at the contact line were stationary, while the others were in motion. The experiment with a meniscus-wall contact angles less than 10 (Fig. 2e) had the greatest amount of colloids retained at the contact line. For the contact angle between 10 and 20 colloids were retained to a less extent (Fig. 2f), but for larger contact angles colloid attachment at the AWS contact line was almost negligible (Fig. 2g and h). Colloids were not deposited at any other locations in the microchannel.
3.2. Colloid retention at the AWS contact line as a function of contact angle and ionic strength Despite the careful experimentation and cleaning procedures, contact angles between the wall and the meniscus varied
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 9 5 e3 0 6
299
Fig. 2 e Examples of the shape of the meniscus and colloid retention at the AWS contact line 30 min after the colloid pulse was introduced for different meniscus-wall contact angles: (a, e) < 10 , (b, f) 10e20 , (c, g) 20e40 , (d, h) > 40 .
greatly even for the same treatments. In general the covered experiments had greater contact angles (ranging from 42 to 63 , Fig. 3aec) than uncovered experiments with contact angles between 7 and 29 (Fig. 3def). We cannot readily explain these differences in contact angles, but it is assumed to be a result of slight differences in the channels used, cleanliness and heterogeneity of the surface, and the hysteresis of the contact angle (Johnson and Dettre, 1964; Lam et al., 2002; Johnson et al., 2010). Even after the flow of water was stopped, the contact angle remained at 30 in the covered microchannel (Series 7, Table 1). We should also note that after running the experiments for several hours the contact angles decreased and eventually approached the static value of around 10 . Colloid retention (Fig. 3) at 30 min after the injection of colloid pulse in all experiments with the meniscus-wall contact angles over 20 was less than 0.3 mm2/mm (defined as the area occupied by colloids per length of the contact line).
These observations are independent whether or not the microchannel was covered. For two uncovered experiments at 1 mM and 100 mM ionic strengths where the contact angle was over 20 , the colloid retention was small (under 0.3 mm2/ mm), similar to all covered experiments. For contact angles less than 16 colloid retention was significant and up to 3.0 mm2/mm for the 1 mM experiments (Fig. 3d). The retention of colloids at the injection time of 30 min is re-plotted as a function of the average contact angle (Fig. 4a) and ionic strength (Fig. 4b). Fig. 4a indicates that for the meniscus-wall contact angles smaller than 29 the colloid retention decreased with increasing contact angles. Colloid retention at a contact angle greater than 20 was minimal and less than 0.3 mm2/mm as noted above. In addition, there was a linear decrease in colloid retention with increasing ionic strength for the subgroup of 10e20 (R2 ¼ 0.96) for which sufficient data is available. Zevi et al. (2009) observed a similar trend for colloid retention at the AWS contact line in partially saturated sand
300
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 9 5 e3 0 6
Fig. 3 e Retention of the colloids at the AWS contact line for experiments with a closed (aec) and open channel (def) in the 1, 50 and 100 mM NaCl solutions. The average contact angle, q, plus or minus the standard deviation is presented for each microchannel experiment.
pack, which was caused by increased attachment at the solidwater interface with increasing ionic strength. Since there was no colloid attachment at the microchannel wall, the above explanation of Zevi et al. (2009) is not applicable here. We will discuss possible causes later.
3.3. Flow field effects on colloid retention at the AWS contact line The flow paths of colloids approaching the AWS contact line were investigated experimentally and computationally. The experimentally observed flow patterns are given in Fig. 5a for the meniscus-wall contact angle of 28 and in Fig. 5b for the 12.5-degree contact angle. The movies with the real time pathways are included in the auxiliary materials (Section B). When the contact angle was 12.5 10 colloids approached the channel wall during a 4-minute observation time (Fig. 5b). Once the colloids arrived in the vicinity of the wall they moved in the opposite direction to the general flow field before becoming immobilized at the AWS contact line. In contrast, for the 28-degree contact angle only three colloids approached
the wall during an observation period of 40 min (Fig. 5a). Unlike the case with the smaller meniscus-wall contact angle (Fig. 5b) there was no counter flow, and the colloids settled at the wall while moving forward. The computational results are given at Fig. 6a and b for the same set of the meniscus-wall contact angles (i.e., 12.5 and 28 ). The normalized velocity field was illustrated by the red arrows. The boundaries of the channel were made visible by a gray color. The simulated flow for the contact angle of 12.5 showed the counter flow close to the AWS contact line, (Fig. 6a) whereas no counter flow was observed in the simulation for the contact angle of 28 (Fig. 6b).
3.4.
Calculated forces and torques acting on colloids
The calculated DLVO forces for colloid-SWI interactions indicated that with increasing ionic strength the magnitude of the attractive forces at the primary minimum ðFDLVO 1min Þ decreased, whereas the attractive forces at the secondary minimum ðFDLVO 2min Þ increased (Table 2 and Figure A2). There is a formidable repulsive force for the colloids interacting with the AWI
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 9 5 e3 0 6
301
maximum as discussed later. As the colloids approach the AWS contact line (i.e., decreasing H ), the ratio of decreases, which increases the colloid retenTapplied =Tadhesive 2min tion at the SWI (Table 3). For example, at H ¼ 50 mm, the colloids will not remain stationary on the SWI, but roll away at all ionic strengths because Tapplied is greater than Tadhesive 2min (Table 3), which explains the absence of colloid retention at the channel wall far away from the AWS contact line. The torque analysis also predicts that the colloid retention at the secondary minimum will increase with increasing ionic strength (Table 2 and Table 3). Since this prediction does not agree with the decreased colloid retention with increasing ionic strength at the AWS contact line for the meniscus-wall contact angles of 10e20 as discussed above (Fig. 4b), some of the assumptions should be adjusted. This discrepancy indicates that at the AWS contact line the secondary minimum is not the sole colloid retention mechanisms. More discussion is provided in a later section.
4.
Fig. 4 e Colloid retention at 30 min after colloid addition: (a) plotted as a function of average contact angle; (b) plotted as a function of ionic strength.
(Figure A2). The applied hydrodynamic forces (FD) and torques (Tapplied) for a Poiseuille type flow field decreases as the colloid approaches the AWS contact line (i.e., smaller depths of the AWS wedged pore space [H], Table 3), suggesting that the colloid will be more likely retained due to reduced hydrodynamic drag. The calculated Tapplied values in Table 3 are comparable to the adhesive torques at the secondary minimum ðTadhesive 2min Þ, but three to six orders of magnitude smaller than the adhesive torques at the primary minimum ðTadhesive 1min Þ (Table 2). However, colloids could not be retained at the primary minimum due to the repulsive force at the primary maximum (Table 2) unless other force (e.g., capillary force) is involved to push the colloid through the primary
Fig. 5 e Trajectories of colloidal particles for (a) 28-degree contact angle and (b) 12.5-degree contact angle.
Discussion
The effectiveness of capturing colloids at the AWS contact line decreased with increasing meniscus-wall contact angles (Fig. 4a). Analogous to colloid deposition to a collector under saturated conditions, the capture of colloids at the AWS contact line depends on how colloids approach the interface (collision efficiency) and how effective these colloids are retained by the interface (attachment efficiency) (Yao et al., 1971; Brown and Abramson, 2006; Cakmak et al., 2008). In the following sections, based on visual evidence, computational simulations, and theoretical considerations, we discussed first the transport of colloids to, and subsequently their retention at the AWS contact line. Colloid transport to the AWS contact line was enhanced by counter flow along the walls as shown by the computational simulations (Fig. 6) and experimental path tracking (Fig. 5 and the movies in the auxiliary material section B). Counter flow occurred for small contact angles but not for larger contact angles (Figs. 5a and b, 6a and b). These observations are in agreement with the experiments of Lazouskaya et al. (2006) and Lazouskaya and Jin (2008) and with the latticeBoltzmann simulations of Shi et al. (2010) for a moving contact line in a microchannel with a relatively small contact angle. Although their experiments were transient due to a moving wetting front and different with our steady flow, in both cases colloids reach the interface as a result of counter flow that exists in these microchannels as can be seen in Movie-3 of Lazouskaya and Jin (2008) and Movie 1 and Movie 2 in the auxiliary material section B. Supplementary video related to this article can be found at doi:10.1016/j.watres.2011.09.046. Additionally, the counter flow (e.g., backward flow, and flow vortices) was also observed in the COMSOL simulations (Torkzaban et al., 2008) at the grainegrain contacts and the AWS contact line (Zhang et al., 2010), and numerical simulation at the moving contact line in a capillary tube (Sheng and Zhou, 1992). The lattice-Boltzmann saturated flow simulations by Li et al. (2010) in pore structures derived from real
302
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 9 5 e3 0 6
Fig. 6 e Computational simulations of flow field in a 54-mm long microchannel: (a) 12.5-degree meniscus-wall contact angle and (b) 28-degree meniscus-wall contact angle. (i) The cross section of microchannel with water phase (in pink color) (ii) The flow field along the channel (iii) The enlargement of one flow field (red arrows represent flow direction). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
porous media indicates that counter flow not only occurs at the AWS contact line in our simplified model soil pore, but also at the grainegrain contacts in the realistic soil pore structures. Of interest is that Wan and Tokunaga (2005) attributed the observed differences of colloid retention between a closed and open porous media to the effect of evaporation. Based on the experiments that investigated the evaporation effect between covered and uncovered experiments, Steenhuis et al. (2005) found that the effect of evaporation was negligible. In addition, using the theory presented by Morris (2001) we can easily prove that evaporation cannot
account for the observed difference in contact angles between the closed and open channels in this study. Once the colloid is transported to the wall, it can only stay stationary when the hydrodynamic drag forces are less than a combination of other forces (e.g., DLVO, capillary and friction forces). As explained above, colloid retention at the AWS contact line is not solely determined by the DLVO forces, because an increased colloid retention with increasing ionic strength is not observed as predicted by the theory (Fig. 4b). Instead, there was a slight decrease of colloid retention with increasing ionic strength for the contact angle of 10e20 . The
Table 2 e DLVO forces (FDLVO) normalized with the colloid radius (r [ 0.5 mm) and adhesive torques (Tadhesive) for colloids interacting with the solid-water interface (SWI) (see Appendix for detailed calculations).a IS (mM)
Primary minimum FDLVO 1min =r
1 50 100
(mN/m)
11.6 1.67 1.30
x (nm) 0.35 0.35 0.35
Tadhesive 1min
Primary barrier (N m)
8.26 1017 6.23 1018 4.46 1018
FDLVO max =r
(mN/m)
1.56 3.85 3.80
Secondary minimum
x (nm) 5 1 1
FDLVO 2min =r
(mN/m)
6.48 105 7.71 103 1.79 102
x (nm)
Tadhesive (N m) 2min
123 13 8
5.16 1024 3.02 1021 9.28 1021
DLVO a FDLVO 1min ¼ attractive DLVO force at the primary energy minimum, Fmax ¼ repulsive DLVO force at the primary energy maximum, DLVO F2min ¼ attractive DLVO force at the secondary energy minimum, x ¼ separation distance, Tadhesive ¼ DLVO adhesive torque at the primary 1min energy minimum, and Tadhesive ¼ DLVO adhesive torque at the secondary energy minimum. 2min
303
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 9 5 e3 0 6
Table 3 e Applied hydrodynamic torque (Tapplied), hydrodynamic drag force (FD) normalized with the colloid radius (r), and adhesive the ratio of Tapplied =T2min for 1-mm colloid in different depths of the air-water-solid wedged pore space (the flow depth [H] ranging from 1 mm to 50 mm) (see Appendix for detailed calculations).a q ¼ 12:5
H (mm)
FD =r (mN/m)
q ¼ 28
v ¼ 1:19 103 m=s;
v ¼ 9:17 104 m=s;
DP=Lc ¼ 87:2 Pa=m
DP=Lc ¼ 48:0 Pa=m
Tapplied (N m)
1 mM 1 5 10 20 50
6.99 6.29 1.33 2.73 6.92
7
10 106 105 105 105
2.45 2.20 4.65 9.54 2.42
22
10 021 1021 1021 1020
FD =r (mN/m)
Tapplied Tadhesive 2min
47.4 427 901 1849 4695
50 mM 0.08 0.73 1.54 3.16 8.02
Tapplied (N m)
100 mM 0.03 0.24 0.50 1.03 2.61
3.85 3.46 7.31 1.50 3.81
7
10 106 106 105 105
1.35 1.21 2.56 5.25 1.33
22
10 1021 1021 1021 1020
Tapplied Tadhesive 2min 1 mM
50 mM
100 mM
26.1 235 496 1018 2580
0.04 0.40 0.85 1.74 4.42
0.01 0.13 0.28 0.57 1.44
a q ¼ water contact angle of the meniscus, v ¼ average water velocity in the channel, and DP/Lc ¼ pressure gradient.
meniscus-wall contact angle is a more important factor that determines the colloid retention at the AWS contact line, which for small contact angles produces a counter flow slowing down the colloids so that the hydrodynamic torque is less than what would be expected from the Poiseuille type of laminar flow. As observed in the movies, at the large distance (e.g., H > 20 mm) from the AWS contact line and large contact angle, the attractive DLVO forces at the secondary minimum are not sufficiently strong to permanently keep the colloids at the surface (Table 3). An exception existed for the 100 mM ionic strength featured in the Movie 2 where a few colloids that approached the glass wall were retained. This observation agrees with the three-dimensional particle tracking coupled with the lattice-Boltzmann flow simulations in porous media by Li et al. (2010), which showed that colloids may intermittently stop moving by a weak association with the grain surface via the secondary minimum. Applied hydrodynamic drag forces are not great enough to push the colloid through the primary energy barrier (Tables 2 and 3), and certainly not large enough to push against the capillary force (Figure A3 in Auxiliary material section A.4). Since colloids for the small contact angles were attached firmly on the wall after the experiments were completed, indicating that they had a strong bond with the glass surface that is the characteristics of the primary minimum. Colloids can reach the primary minimum by either a force on the colloid that is greater than the repulsive force of the primary energy barrier, or by reducing the energy barrier force. Since enhanced colloid retention due to a decrease in the primary barrier repulsive force has been reported for the cases of nanoscale surface chemical heterogeneity or roughness, we cannot rule out this attachment mechanism (Duffadar and Davis, 2007; Duffadar et al., 2009; Johnson et al., 2010). However, because the same glass surface was used in all the experiments, a more likely candidate for pushing the colloids through the existing energy barrier would be the capillary force associated with the meniscus. However, the capillary force only acts when the meniscus is “stretched” over the colloid as in Figure A1 (Auxiliary material section A.2). Since the drag forces are not sufficient, we hypothesized that small movements of the meniscus due to slight changes in flow are the main factor responsible for stretching the meniscus over
and thus initiating the action of the capillary force on the colloid. Since the triple point where the meniscus attach to the surface is reluctant to move (i.e., contact angle hysteresis; Johnson and Dettre, 1964; Lam et al., 2002), any decrease in flow will decrease the contact angle, but more importantly, it will force the meniscus to stretch over the colloid as shown in Figure A1. At this point capillary force will start acting. Capillary force can be decomposed into two force components, i.e., the lateral force that pushes the colloid back in solution and the normal force that pushes the colloid against the wall (see Auxiliary material Section A.4). The normal capillary force will push the colloid through the primary barrier ranging from 1.6 mN/m for the 1 mM ionic strength to 3.8 mN/m for the 100 mM ionic strength (Table 2). Even for small protrusions (i.e., the colloid protrudes 10% of its diameter through the meniscus) the normal capillary force is around 45 mN/m for contact angles ranging from 12 to 45 (Figure A3 in Auxiliary material section A.4). Once the colloid is pushed through the energy barrier it will reach the primary minimum at 0.35 nm from the wall. For the colloid to stay at the primary minimum, the lateral capillary force has to be smaller than the friction force of the colloid with the glass wall and the attractive force at the primary minimum. The friction force is equal to the product of the friction coefficient and the sum of the normal components of the capillary force and the DLVO force at 0.35 nm (Table 2, Figure A3). By setting the friction force equal to the lateral force (Auxiliary material section A.5 Eq. (A12)), we can calculate the minimum friction coefficient that is needed for the colloid to just stay at the primary minimum. The minimum friction coefficients for contact angles of 12.5 and 28 and the three ionic strengths are shown in Fig. 7. For the 12.5-degree contact angle the friction coefficient (between the colloid and glass) must be at least of 0.06 for the colloid to remain at the wall near the AWS contact line for an ionic strength of 1 mM and a protrusion distance through the meniscus (d ) at 3% of the colloid radius r (i.e., d/r ¼ 0.03) (Fig. 7). For the colloid to remain at the wall at greater ionic strengths, the friction coefficient has to be at least 0.22 when the colloid protrudes farther through the meniscus (Fig. 7). The Engineering Toolbox (2011) states that the friction coefficient of lubricated glass/ metal is between 0.2 and 0.3. In other words, it is plausible that
304
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 9 5 e3 0 6
that of Lazouskaya et al. (2006). They observed as well that the shape of meniscus is related to the amount of colloid retention with larger contact angles having less retention.
5.
Fig. 7 e Minimum friction coefficient required for colloid retention at the AWS contact line dependent on meniscuswall contact angles (q [ 12.5 and 28 ) and ionic strength (1 mM, 50 mM, and 100 mM NaCl solutions) at the primary minimum (0.35 nm). For the x-axis, r is the colloid radius, and d is the distance that the colloid protrudes out of the meniscus.
the friction coefficient for wet glass and microspheres falls between 0.2 and 0.3, indicating that the colloids would be retained at the AWS contact line when the contact angle is 12.5 . This is in agreement with the experimental data in Figs. 3 and 4a where colloids remain attached at the wall for the meniscuswall contact angles smaller than 16 . Moreover, since at low ionic strength, the minimum friction coefficients required for attachment are less than for greater ionic strength, we can expect the greatest attachment efficiency at smaller ionic strength. This is collaborated by experimental data in Fig. 4b. At the meniscus-wall contact angle of 28 , for colloids to remain at the wall the required minimum friction coefficient depicted in Fig. 7 are generally above 0.35 for ionic strength of 50 and 100 mM. Only at the 1 mM ionic strength, a lower required minimum friction coefficient makes it possible for colloids to be retained when they barely protrudes through the meniscus. Although the exact value of static glass/colloid friction coefficient is unknown, for the 28-degree contact angle the required minimum friction coefficients is much greater than for the smaller contact angles, and likely more than what can be expected for colloids with wet glass (Fig. 7). Therefore, the attachment efficiency could be close to zero. The data in Fig. 4 underpin these theoretical findings with independently measured experimental results. Additionally, although the above analysis was performed for a hydrophilic colloid, capillary force is also important for the retention of hydrophobic colloids since capillary force overwhelms any other surface forces. For increased hydrophobicity of the colloid (i.e., a greater colloid-water contact angle), the colloid may be required to protrude farther out of the meniscus so that the capillary force can pin the colloid onto the solid surface since the direction and magnitude of capillary force are dependent on the colloid-water contact angle and the protrusion distance (d ) (Auxiliary material section A.4 Eq. (A9) and (A11)). For extremely hydrophobic colloids, they may become more easily associated with the airwater interface. Finally, our results are also in agreement with
Conclusions
Retention of colloids at the AWS contact line was observed experimentally, which was attributed to the presence of the contact angles of water meniscus and solid surface less than 16 . The small meniscus-wall contact angle induced counter flow near the triple point where water, wall and air meets (i.e., the AWS contact line). For larger contact angles counter flow was not experimentally observed. Computational simulations were in agreement with the experimentally observed flow pattern. Once colloids were in the close proximity of the wall near the AWS contact line, colloids were retained in the secondary minimum when hydrodynamic drag was small. It was hypothesized that small changes in average flow rate decreased the contact angle, thus stretching the meniscus over the colloids located in the secondary minimum. Capillary forces associated with the stretched meniscus could push colloids through the energy barrier into the primary minimum.
Acknowledgments This research was supported by funding from the USDANational Research Initiative (project 2005-03929 and 200835102-04462), Binational Agricultural Research and Development Fund, Project No. IS-3962-07, and the National Science Foundation (project 2006-0635954). The authors acknowledge the expert guidance of Carol Bayles, the manager of the Cornell University Biotechnology Center’s Microscopy and Imaging Facility.
Appendix. Supplementary material Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.09.046.
references
Auset, M., Keller, A.A., 2006. Pore-scale visualization of colloid straining and filtration in saturated porous media using micromodels. Water Resources Research 42, W12S02,. doi:10. 1029/2005WR004639. Bradford, S.A., Torkzaban, S., 2008. Colloid transport and retention in unsaturated porous media: a review of pore-scale processes and models. Vadose Zone Journal 7, 667e681. doi:10. 2136/vzj2007.0092. Bradford, S.A., Torkzaban, S., Wiegmann, A., 2011. Pore-scale simulations to determine the applied hydrodynamic torque and colloid immobilization. Vadose Zone Journal 10 (1), 252e261. Brown, D.G., Abramson, A., 2006. A collision efficiency distribution of a bacterial suspension flowing through porous media and implications for field-scale transport. Water Research 40, 1591e1598.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 9 5 e3 0 6
Cakmak, M.E., Gao, B., Nieber, J.L., Steenhuis, T.S., 2008. Pore scale simulation of colloid deposition. In: Petrone, G., Cammarata, G. (Eds.), Modelling and Simulation. InTech Education and Publishing, p. P688. Chen, G., 2008. Bacterial interactions and transport in unsaturated porous media. Colloid and Surfaces B: Biointerfaces 67, 265e271. Corapcioglu, M.Y., Choi, H., 1996. Modeling colloid transport in unsaturated porous media and validation with laboratory column data. Water Resources Research 32 (12), 3437e3449. Crist, J.T., McCarthy, J.F., Zevi, Y., Baveye, P.C., Throop, J.A., Steenhuis, T.S., 2004. Pore-scale visualization of colloid transport and retention in partly saturated porous media. Vadose Zone Journal 3 (2), 444e450. Crist, J.T., Zevi, Y., McCarthy, J.F., Troop, J.A., Steenhuis, T.S., 2005. Transport and retention mechanisms of colloids in partially saturated porous media. Vadose Zone Journal 4, 184e195. Duffadar, R.D., Davis, J.M., 2007. Interaction of micrometer-scale particles with nanotextured surfaces in shear flow. Journal of Colloid and Interface Science 308, 20e29. Duffadar, R., Kalasin, S., Davis, J.M., Santore, M.M., 2009. The impact of nanoscale chemical features on micron-scale adhesion: crossover from heterogeneity-dominated to meanfield behavior. Journal of Colloid and Interface Science 337, 396e407. Engineering Toolbox, 2011. Web Accessed on June 2011. http://www. engineeringtoolbox.com/friction-coefficients-d_778.html. Gao, B., Saiers, J.E., Ryan, J.N., 2006. Pore-scale mechanisms of colloid deposition and mobilization during steady and transient flow through unsaturated granular media. Water Resources Research 44, W04504. doi:10.1029/2005WR004233. Gao, B., Steenhuis, T.S., Zevi, Y., Morales, V.L., Nieber, J.L., Richards, B.K., McCarthy, J.F., Parlange, J.Y., 2008. Capillary retention of colloids in unsaturated porous media. Water Resources Research 42, W01410. doi:10.1029/2006WR005332. Johnson, R.E., Dettre, R.H., 1964. Contact angle hysteresis. III. Study of an idealized heterogeneous surface. The Journal of Physical Chemistry 68 (7), 1744e1750. Johnson, W.P., Pazmino, E., Ma, H., 2010. Direct observations of colloid retention in granular media in the presence of energy barriers and pitfalls of inferring mechanisms from indirect observations. Water Research 44, 1158e1169. doi:10.1016/j. watres.2009.12.014. Kemps, J.A.L., Bhattacharjee, S., 2009. Particle tracking model for colloid transport near planar surfaces covered with spherical asperities. Langmuir 25, 6887e6897. Lam, C.N.C., Wu, R., Li, D., Hair, M.L., Neumann, A.W., 2002. Study of the advancing and receding contact angles: liquid sorption as a cause of contact angle hysteresis. Advances in Colloid and Interface Science 96 (1e3), 169e191. Lazouskaya, V., Jin, Y., Or, D., 2006. Interfacial interactions and colloid retention under steady flows in a capillary channel. Journal Colloid and Interface Science 303, 171e184. Lazouskaya, V., Jin, Y., 2008. Colloid retention at air-water interface in a capillary channel. Colloids and Surfaces A: Physicochemical and Engineering Aspects 325, 141e151. Li, X., Li, Z., Zhang, D., 2010. Role of low flow and backward flow zones on colloid transport in pore structures derived from real porous media. Environmental Science & Technology 44 (13), 4936e4942. Massoudieh, A., Ginn, T.R., 2007. Modeling colloid-facilitated transport of multi-species contaminants in unsaturated porous media. Journalof Contaminant Hydrology 92, 162e183. Morales, V.L., Gao, B., Steenhuis, T.S., 2009. Grain surfaceroughness effects on colloidal retention in the vadose zone. Vadose Zone Journal 8 (1), 11e20.
305
Morris, S.J.S., 2001. Contact angles for evaporating liquids predicted and compared with existing experiments. Journal of Fluid Mechanics 432, 1e30. O’Brien, S.B.G., van den Brule, B.H.A.A., 1991. A mathematical model for the cleansing of silicon substrates by fluid immersion. Journal Colloid and Interface Science 144 (1), 210e221. Saiers, J.E., Hornberger, G.M., 1999. The influence of ionic strength on the facilitated transport of cesium by kaolinite colloids. Water Resources Research 35, 1713e1727. Scha¨fer, A., Ustohal, P., Harms, H., Stauffer, F., Dracos, T., Zehnder, A.J.B., 1998a. Transport of bacteria in unsaturated porous media. Journal Contaminant of Hydrology 33, 149e169. Scha¨fer, A., Harms, H., Zehnder, A.J.B., 1998b. Bacterial accumulation at the air-water interface. Environmental Science & Technology 32 (23), 3704e3712. Sheng, P., Zhou, M., 1992. Immiscible-fluid displacement: contactline dynamics and the velocity-dependent capillary pressure. Physical Review A 45 (8), 5694e5708. Shang, J., Flury, M., Chen, G., Zhuang, J., 2008. Impact of flow rate, water content, and capillary forces on in situ colloid mobilization during infiltration in unsaturated sediments. Water Resources Research 44, W06411. doi:10.1029/ 2007WR006516. Shi, X.-Y., Gao, H., Lazouskaya, V.I., Kang, Q., Jin, Y., Wang, L.-P., 2010. Viscous flow and colloid transport near air-water interface in a microchannel. Computers & Mathematics with Applications 59, 2290e2304. Sirivithayapakorn, S., Keller, A., 2003. Transport of colloids in saturated porous media: a pore-scale observation of the size exclusion effect and colloid acceleration. Water Resources Research 39, 1109. Steenhuis, T.S., McCarthy, J.F., Crist, J.T., Zevi, Y., Baveye, P.C., Throop, J.A., Fehrman, R.L., Dathe, A., Richards, B.K., 2005. Reply to comments on ‘pore-scale visualization of colloid transport and retention in partly saturated porous media’. Vadose Zone Journal 4, 957e958. Torkzaban, S., Bradford, S.A., Walker, S.L., 2007. Resolving the coupled effects of hydrodynamics and DLVO forces on colloid attachment in porous media. Langmuir 23, 9652e9660. Torkzaban, S., Tazehkand, S.S., Bradford, S.A., Walker, S.L., 2008. Transport and fate of bacteria in porous media: coupled effects of chemical conditions and system hydrodynamic. Water Resources Research 44, W04403. doi:10.1029/ 2007WR006541. Veerapaneni, S., Wan, J.M., Tokunaga, T.K., 2000. Motion of particles in film flow. Environmental Science & Technology 34 (12), 2465e2471. Wan, J.M., Tokunaga, T.K., 1997. Film straining of colloids in unsaturated porous media: conceptual model and experimental testing. Environmental Science & Technology 31, 2413e2420. Wan, J.M., Tokunaga, T.K., 2005. Comments on pore-scale visualization of colloid transport and retention in partly saturated porous media. Vadose Zone Journal 4, 954e956. Wan, J.M., Wilson, J.L., 1994. Visualization of the role of the gaswater interface on the fate and transport of colloids in porous media. Water Resources Research 30 (1), 11e23. Williams, C.F., Letey, J., Farmer, W.J., 2006. Estimating the potential for facilitated transport of napropamide by dissolved organic matter. Soil Science Society of America Journal 70, 24e30. Yao, K.M., Habibian, M.T., O’Melia, C.R., 1971. Water and waste water filtration: concepts and applications. Environmental Science & Technology 5 (11), 1105e1112. Zevi, Y., Dathe, A., McCarthy, J.F., Richards, B.K., Steenhuis, T.S., 2005. Distribution of colloid particles onto interfaces in
306
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 2 9 5 e3 0 6
partially saturated sand. Environmental Science & Technology 39 (18), 7055e7064. Zevi, Y., Dathe, A., Gao, B., Richards, B.K., Steenhuis, T.S., 2006. Quantifying colloid retention in partially saturated porous media. Water Resources Research 42 (12), W12S03. doi:10. 1029/2006wr004929. Zevi, Y., Dathe, A., Gao, B., Zhang, W., Richards, B.K., Steenhuis, T. S., 2009. Transport and retention of colloidal particles in
partially saturated porous media: effect of ionic strength. Water Resources Research 45, W12403. doi:10.1029/ 2008WR007322. Zhang, W., Morales, V.L., Cakmak, M.E., Salvucci, A.E., Geohring, L.D., Hay, A.G., Parlange, J.-Y., Steenhuis, T.S., 2010. Colloid transport and retention in unsaturated porous media: effect of colloid input concentration. Environmental Science & Technology 44 (13), 4965e4972.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 0 7 e3 1 5
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Cation exchange during subsurface iron removal D. van Halem a,*, D.H. Moed a, J.Q.J.C. Verberk a, G.L. Amy a,b,1, J.C. van Dijk a a b
Delft University of Technology, Faculty of Civil Engineering and Geosciences, Stevinweg 1, 2628 CN Delft, The Netherlands UNESCO-IHE, Westvest 7, 2611 AX Delft, The Netherlands
article info
abstract
Article history:
Subsurface iron removal (SIR), or in-situ iron removal, is an established treatment tech-
Received 20 April 2011
nology to remove soluble iron (Fe2þ) from groundwater. Besides the adsorptive-catalytic
Received in revised form
oxidation theory, it has also been proposed that the injection of O2-rich water onsets the
11 October 2011
exchange of adsorbed Fe2þ with other cations, such as Ca2þ and Naþ. In sand column
Accepted 12 October 2011
experiments with synthetic and natural groundwater it was found that cation exchange
Available online 15 November 2011
(NaþeFe2þ) occurs during the injection-abstraction cycles of subsurface iron removal. The Fe2þ exchange increased at higher Naþ concentration in the injection water, but
Keywords:
decreased in the presence of other cations in the groundwater. Field results with injection
Cation exchange
of elevated O2 concentrations (0.55 mM) showed increased Fe removal efficacy; the oper-
Drinking water treatment
ational parameter V/Vi (abstraction volume with [Fe]<2 mM divided by the injection
Groundwater
volume) increased from an average 7 to 16, indicating that not the exchangeable Fe2þ on
Fe
2þ
the soil material is the limiting factor during injection, but it is the supply of O2 to the
Subsurface iron removal
available Fe2þ. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Subsurface iron removal (SIR), or in-situ iron removal, is an established treatment technology to remove soluble iron (Fe2þ) from groundwater. In Europe, numerous drinking water supply companies are operating (some of) their production wells with this technology (Boochs and Barovic, 1981; Braester and Martinell, 1988; Grombach, 1985; Hallberg and Martinell, 1976; Mettler, 2002; Rott and Friedle, 1985; van Beek, 1985; van Halem et al., 2011). The principle of subsurface iron removal is that aerated water is periodically injected into an aquifer through a tube well (Fig. 1A), partially displacing the original Fe2þ-containing groundwater. The O2-rich injection water onsets the oxidation of Fe2þ in the subsurface environment around the tube well. When the flow is reversed, groundwater with low Fe concentrations is abstracted (Fig. 1B). More water with reduced iron concentrations can be abstracted (volume V)
than was injected (volume Vi), and this volumetric ratio (V/Vi) determines the efficiency of the system. Every period of injection-abstraction is referred to as a cycle. This technology has been reported to be very effective in the removal of Fe and V/Vi ratios have been reported to increase after multiple cycles (Hallberg and Martinell, 1976; Mettler, 2002; Rott and Friedle, 1985; van Beek, 1985). Clogging of the aquifer, even after years of operation, has not been found to threaten its sustainability (Mettler et al., 2001; van Halem et al., 2011). Although SIR has been applied for many decades, the exact processes responsible for its effectiveness have not been unraveled. Initially it was proposed that an adsorptivecatalytic oxidation mechanism could explain the system’s efficacy (Rott and Friedle, 1985; van Beek, 1985), where adsorbed Fe2þ is oxidized to form new adsorption sites. The adsorption of Fe2þ occurs in the absence of oxygen during the abstraction phase, which can be schematically formulated as:
* Corresponding author. Tel.: þ31 152786588; fax: þ31 152784918. E-mail address:
[email protected] (D. van Halem). 1 Present address: King Abdullah University of Science and Technology, P.O. Box 55455, Thuwal 34, Saudi Arabia. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.015
308
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 0 7 e3 1 5
S OHo þ Fe2þ /S OFeðIIÞþ þHþ
(1)
It should be noted that the use of SeOH0 in the equation is simplified, as in reality adsorbed Fe2þ may transfer an electron to the solid, creating a trivalent ion that will hydrolyze and is equivalent to Fe(OH)2 (Hiemstra and van Riemsdijk, 2007). The adsorptive capacity of the soil depends on the type of iron hydroxides present in the soil, as adsorption capacities for amorphous Fe3þ hydroxides is larger than more crystalline mineral structures with lower surface areas (goethite, lepidocrocite). When O2-rich water is injected into the Fe2þ saturated subsurface environment the adsorbed Fe2þ will oxidize heterogeneously: S OFeðIIÞþ þ0:25O2 þ 1:5H2 O/S OFeðIIIÞðOHÞ02 þ Hþ
(2)
2þ
adsorption, this oxidation reaction will Like during Fe release a proton per oxidized S-OFe(II)þ complex at near neutral pH, i.e., the combined reaction of adsorptive-catalytic oxidation results in the release of two protons. The consumption of oxygen during this reaction has an effect on the penetration of the dissolved oxygen into the aquifer, as the oxygen front will lag behind the injected water front (Fig. 1A). The heterogeneous oxidation rate is catalysed by the presence of Fe3þ hydroxides (Tamura et al., 1980; Sung and Morgan, 1980), depending on the type of mineral phase, e.g., ferrihydrite or goethite (Mettler, 2002; Tamura et al., 1980). Besides the adsorptive-catalytic oxidation theory, it has also been proposed that the injection of O2-rich water onsets the exchange of adsorbed Fe2þ with other cations, such as Ca2þ and Naþ (Appelo et al., 1999; Appelo and Postma, 2005): XCa þ Fe2þ 5XFe þ Ca2þ
or
2XNa þ Fe2þ 5X2 Fe þ 2Naþ
(3)
With X being the exchange sites and exchange coefficients, following the Gaines-Thomas convention, of KCa\Fe ¼ 0.7 and KNa\Fe ¼ 0.6 (Appelo and Postma, 2005). Once Fe2þ has exchanged/desorbed, the dissolved O2 in the injection water oxidizes the soluble Fe2þ. In the presence of high Fe3þ oxide concentrations, as generally found in the aquifer, this oxidation reaction can also occur heterogeneously. Subsequent hydrolysis results in the formation of immobile iron oxides or even mobile colloidal material (Wolthoorn, 2003). During the abstraction phase, these iron hydroxides provide new surface sites for Fe2þ adsorption and/or cation exchange. The process of cation exchange (CIEX) during SIR has been schematized in Fig. 2 for a simplified system containing Fe2þ during abstraction and Ca2þ, Naþ and O2 during injection. Before the start of injection, both soluble and adsorbed Fe2þ are present (Fig. 2A). During injection the cations in the injection water, Ca2þ and Naþ, exchange with the adsorbed Fe2þ on the soil grains (Fig. 2B). The desorbed Fe2þ is then flushed deeper into the aquifer, partially mixing with O2 in the injection water,
resulting in hydrolyzed Fe3þ precipitates (Fe(OH)3; Fig. 2C). When abstraction is started, the flow is reversed and the Fe2þ in the groundwater is retained on the mineral surface and on the Fe3þ hydroxides, either through CIEX or adsorption (Fig. 2D). The contribution of cation exchange to the system’s efficiency depends on the water composition of the injection water and groundwater, but also on the exchangeable Fe2þ on the aquifer material. The cation exchange capacity (CEC) of the soil depends, in order of importance, on the clay, organic carbon and iron hydroxide content (Appelo and Postma, 2005). The occurrence of CIEX is difficult to extract from field data, as iron oxidizes after it has exchanged, and will not reach the well in its ferrous, soluble state. However, for the purpose of optimizing the subsurface iron removal process or the application for other inorganic constituents, such as arsenic (Rott et al., 2002; van Halem et al., 2010) it is vital to understand the underlying mechanisms. The objective of this study was therefore to simulate the process of oxidation, adsorption and cation exchange in injection-abstraction column studies, and to investigate to what extent the occurrence of cation exchange influences the subsurface iron removal process. The column studies have been performed with a Naþ exchanger during injection, either in the presence or absence of O2, with synthetic groundwater and in the more complex environment of natural groundwater.
2.
Materials and methods
The experimental set-up (Fig. 3) consisted of duplicate transparent PVC columns with a length of 30 cm and an inner diameter of 36 mm (wall thickness 2 mm). During all experiments, the columns were wrapped in aluminium foil to exclude light. The columns were filled with washed (24 h with 1 M HCl) filter sand (500 g; grain size ¼ 0.5e0.8 mm; D10 ¼ 0.58). The absence of other mineral structures than quartz on the sand material was checked with X-ray Powder Diffraction (Bruker D5005; Brain PSD). The push-pull operational mode of injection-abstraction was simulated in the 1D plug-flow environment of the columns with down flow for injection and up flow abstraction (1.1 L h1 0.05). An injectionabstraction cycle started with 14 (0.5) pore volumes of injection water, to allow for complete breakthrough of dissolved O2. Subsequently the influent was switched to groundwater to allow retention of Fe2þ. Electrical conductivity was used as a conservative tracer from which the pore volume could be calculated to be on average 0.12 L (0.002). The flow rate in the columns (2.7 m h1) was controlled with a multichannel pump and PVC tubing with low gas permeability. Anoxic conditions were maintained in the columns by using
Fig. 1 e Principle of subsurface iron removal with (A) injection of aerated water and (B) abstraction of groundwater.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 0 7 e3 1 5
309
Fig. 2 e Schematic presentation of cation exchange during subsurface iron removal on the sand grain surface.
an airtight FESTO system (6 1 PUN, I.D. 4 mm) with matching connectors and valves. The experiments were executed (i) in the laboratory with synthetic groundwater and (ii) at a drinking water treatment plant with natural groundwater. At the start of each experiment the columns were conditioned with synthetic or natural groundwater, until complete breakthrough of iron occurred, and the Eh potential stabilized. A normal injection mode consisted of demineralized water containing a pH buffer (5 mM NaHCO3) and 0.28 mM O2. However, to study the role of cation exchange, injection cycles have also been performed in the absence of O2 and/or Naþ. The water quality during the abstraction phase was constant during the experiments, being either synthetic or natural groundwater. A summary of the experiments is given in Table 1, including water quality of injection and groundwater. It should be noted that the sand columns were only refilled after experiment E, but reference cycles (A) were repeated to exclude the potential influence of subsequent cycles. The synthetic groundwater was produced by sprinkling demineralized water on a 6 m gas stripping column containing
stainless steel Pall Rings. From the bottom pure N2 was blown into the degassing column to sparge out all O2. Before entering the sand columns, the water was checked for O2 with the Orbiphere (HACH Lange; M1100 Sensor; 410 Analyser) to ensure concentrations below 1.5 mmol L1. Addition of stock solutions for FeSO4, NaHCO3 and/or NaCl was done with a dosing pump followed by a static mixer. pH correction was achieved by addition of HCl or NaOH and all stock solutions were sparged with N2 in order to ensure the absence of O2. The experiments were performed with synthetic groundwater of pH 6.9 (0.02), a temperature of 20 C (0.1), Fe concentration of 0.1 mmol L1 (0.01), pH buffer of 5 mM NaHCO3 and ionic strength buffer of 1.6 mM NaCl. To study the occurrence of cation exchange in the multi-component groundwater matrix, the column set-up was transported to a groundwater treatment plant (WTP Loosdrecht, Vitens Water Supply Company). The groundwater, naturally containing Fe2þ, was used as feed water for the column experiments. During the research period the groundwater had an average pH of 7.2 (0.01), a constant temperature of 11 C, 0.1 mmol Fe.L1 (0.01), 3.3 mmol Mn.L1, 1.07 mmol Ca.L1, 0.52 mmol Na.L1, 0.28 mmol Si.L1, and 0.14 mmol SO4.L1.
310
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 0 7 e3 1 5
Fig. 3 e Experimental sand column set-up.
Fe analysis of the water samples was done with an Atomic Absorption Spectrometer (PerkineElmer Flame AAS 3110). Inline measurements were done for dissolved oxygen (Orbisphere and WTW Cellox 325), Eh potential (WTW SenTix ORP), pH (WTW SenTix 41), and electrical conductivity (WTW TetraCon 325). Measurements were registered on a computer with Multilab Pilot v5.06 software.
3.
Results
3.1.
Injection-abstraction cycles
A regular injection phase during full-scale subsurface iron removal consists of the injection of aerated water, in most cases drinking water from the clean water reservoir. In the experiments, demineralized water was used for injection containing a pH buffer (5 mM NaHCO3) and a 0.28 mM O2
concentration. During injection the tracer passed the column at 1 pore volume (PV), but O2 concentrations were delayed in the columns, corresponding to an approximate O2 consumption of 0.15 mmol in the synthetic groundwater columns (results not shown). Based on the stoichiometric ratio for Fe2þ oxidation of 4, this could result in a total Fe removal of 0.6 mmol for one cycle. To what extent iron is delayed compared to the groundwater determines the efficacy of the subsurface Fe removal technology. In general, groundwater treatment plants operate a V/Vi ratio based on the moment Fe starts to arrive at the well, thus (C/ C0)Fe>0. In the experiments, [Fe2þ] is allowed to breakthrough, in order to calculate the dimensionless retardation factor R. RFe þ 1 is calculated from a Vi corresponding to the pore volume when the tracer (electrical conductivity) is C/C0 ¼ 0.5, and V is the number of pore volumes that can be abstracted with iron concentrations below (C/C0)Fe ¼ 0.5. In the case of a column study, the oxidation zone is limited to the size of the column, i.e., 1 pore volume, making the calculation of V/Vi redundant:
Table 1 e Summary of conducted experiments for synthetic and natural waters: The injection water quality and the groundwater water quality. Experiment
Location
Injection water
Groundwater
A B C D E A B C D E
Laboratory
0.28 mM O2, 5 mM NaHCO3 0.01 M NaHCO3, No O2 0.1 and 0.5 M NaCl, No O2 No Na, No O2 No Na, 0.28 mM O2 0.28 mM O2, 5 mM NaHCO3 5 mM NaHCO3, No O2 0.1 M NaCl, No O2 No Na, No O2 No Na, 0.28 mM O2
pH ¼ 6.9 (0.02), T ¼ 20 C (0.1), 0.1 mM Fe2þ (0.01), 5 mM NaHCO3
WTP
pH ¼ 7.2 (0.01), T ¼ 11 C, 0.1 mM Fe (0,01), 3.3 uMMn, 1.07 mMCa, 0.52 mM Na, 0.28 mM Si and 0.14 mM SO4
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 0 7 e3 1 5
311
Fig. 4 e Fe breakthrough in duplicate columns after injection of water containing 0.28 mM O2 and 5 mM NaHCO3 for columns loaded with synthetic and natural groundwater (C0 [ 0.1 mM Fe2D).
RWell
V Vi FeC=C ðPVÞFeC=C 0 0 ¼ 10RColumn ¼ 1 V ðPVÞTracerC=C 0 Vi TracerC=C
(4)
0
The Fe retardation during abstraction for the regular injection-abstraction cycle is shown for both the synthetic and natural groundwater columns in Fig. 4. The tracer passed the columns at PV ¼ 1, but elevated Fe concentrations were not observed until 30 and 8 pore volumes for the synthetic and natural columns, respectively. The Fe retardation in the natural groundwater columns (RFe ¼ 9) was measured to be lower than for synthetic water (RFe ¼ 42). The reduction in removal efficiency can be explained by the presence of competing cations in the natural groundwater, binding to the same sites as Fe2þ. For instance, Ca2þ is known to inhibit the Fe2þ adsorption onto virgin and iron hydroxide coated sand (Sharma, 2001). When considering the synthetic groundwater column results it was calculated from the retardation factor (RFe ¼ 42) that the Fe retention was approximately 0.6 mmol/cycle. This value correlates well to the 0.6 mmol Fe calculated from the O2 consumption in the columns. Based on this finding it may be concluded that all O2 consumption was used for Fe2þ oxidation, however, there is still the question of whether cation exchange plays a (catalyzing) role. Adsorbed Fe2þ could either directly (heterogeneously) oxidize on the surface, like proposed with the catalytic adsorptive-oxidation mechanism, or it could exchange with other cations before oxidizing to Fe3þ hydroxides. During a regular injection-abstraction cycle, the Fe:O2 mass balance does not differentiate between these two mechanisms, as Fe2þ will always oxidize in the presence of O2 and be retained in the columns.
3.2.
Fe2þeNaþ exchange
After reaching steady state with the synthetic groundwater the columns were loaded for an injection phase without O2, but with elevated Naþ concentrations. Naþ was added as
Fig. 5 e Fe measurements during (A) injection of 0.01 M NaHCO3, Na-free or 0.5 M NaCl without O2 and (B) abstraction of synthetic groundwater (C0 [ 0.1 mM Fe2D), dotted lines represent results from PHREEQC (v2.15).
0.01 M NaHCO3 or 0.5 M NaCl. Additionally a control cycle was tested without the addition of any Naþ. It is noteworthy that the pH during injection with water containing low buffer capacities (absence of HCO 3 ) remained above pH 7.5. Fig. 5 shows the Fe concentrations that were measured during the injection phases. The dotted lines represent the results from the geochemical surface complexation model PHREEQC (v2.15; Parkhurst and Appelo, 1999). Fe desorption was measured from the column material at concentrations 1.5e25 times the C0 of 0.1 mM. The PHREEQC model results even show a higher desorbed concentration, indicating that during sampling the high peak concentration was missed. The amount of exchanged Fe was higher in the case of 0.5 M Naþ, resulting in subsequent higher removal efficiencies during abstraction (Fig. 5). RFe was measured to be 11 and 30, for 0.01 M and 0.5 M Naþ, respectively. The same experiment was conducted with 0.1 M NaCl, showing the same retardation as for 0.5 M NaCl (not shown, confirmed with PHREEQC), which means that maximum Naþ regeneration has occurred at 0.1 M NaCl. The retardation of Fe during this particular cycle provides information on the exchangeable Fe2þ capacity of the sand, as retention of approximate 0.4 mM Fe can be recalculated to a Cation Exchange Capacity (CEC) of 1.55 meq.kg1 (0.78 mmol Fe.kg1) for the column sand. This CEC value is lower than for
312
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 0 7 e3 1 5
average aquifer sand (10 meq.kg1; Appelo and Postma, 2005), as can be expected based on the absence of clay, organic carbon and iron oxides on this clean (silica) filter sand. The CEC value correlates well with the Fe2þ adsorption capacity of virgin sand of 0.74 mmol Fe.kg1 observed by Sharma et al. (1999). In that study the adsorption of Fe2þ onto clean sand (D ¼ 0.7e1.25 mm) at pH 7.0 in the absence of oxygen was investigated. When using the measured exchange value of 1.55 meq.kg1 in PHREEQC, the measured results were simulated well. It should be noted, though, that there was slightly more tailing of the Fe2þ breakthrough curve during abstraction in the measurements than the model, potentially due to kinetics and/or stagnant zones. These results show that also in the absence of O2, but in the presence of abundant Naþ, a retardation of Fe can be achieved to a certain level. The desorption of Fe during injection shows the exchange of attached Fe2þ with soluble Naþ, and subsequent retardation of Fe points towards the vise versa exchange of Fe2þ and Naþ. The occurrence of Fe2þeNaþ exchange, even on clean filter sand, can potentially play an important role during subsurface iron removal. However, this depends strongly on cation concentrations in the injection water and on the CEC of the aquifer material. The control cycle with Naþ-free water confirmed that the Fe2þ exchange was very low during injection (Fig. 5). Fe was subsequently not significantly delayed in the columns, with an RFe of 1.
3.3.
CIEX in natural groundwater
The role of CIEX in the multi-component environment of natural groundwater was studied in experiments using groundwater from a water treatment plant (WTP Loosdrecht) instead of synthetic groundwater. Under these natural conditions the injection cycles in the presence and absence of O2 and/or Naþ were investigated. Fig. 6A depicts the Fe desorption during the injection in the absence of oxygen. A peak concentration almost reaching the initial concentration of 0.1 mM was detected. At 0.1 M NaCl the Fe leaching from the columns was higher than for 5 mM Naþ, resulting in a slightly better Fe retardation during the abstraction phase (Fig. 6B). The Fe desorption after injection in the absence of Naþ was most likely caused by the cations in the natural groundwater present in the columns just before injection. It is noteworthy that the leaching in these natural groundwater columns was up to 20 times lower than in the synthetic groundwater columns, as can be seen from Fig. 5, not exceeding the background Fe concentration of 0.1 mM. The Fe retardation factors that were measured in all experiments for both synthetic and natural groundwater are summarized in Fig. 7, showing the decrease in efficiency in the natural groundwater columns compared to the synthetic groundwater columns. The RFe after injection with or without oxygen show a reduced value in the multi-component environment of natural groundwater (injection modes A, B, C, E). The Fe2þ exchange/adsorption is clearly limited by the presence of other cations, such as Ca2þ, resulting in retardation values below 12. In the absence of oxygen, the RFe was somewhat similar for both 5 mM and 0.1 M Naþ, i.e., CIEX between Naþ and Fe2þ is not significantly enhanced at higher Naþ. Also in the absence of Naþ during injection, the Fe2þ retardation was limited in natural groundwater compared to synthetic water
Fig. 6 e Fe measurements during (A) injection of 5 mM NaHCO3, 0.1 M NaCl or Na-free water without O2 and (B) abstraction with natural groundwater.
(injection mode E), as the RFe dropped from 33 to 8 in the multicomponent environment of the natural groundwater columns. It should be noted that although RFe between injection modes A, C and E correlate well, the Fe breakthrough trend looks different (not shown). After injection of Na-free water, the Fe
Fig. 7 e The measured Fe retardation factors for cycles with different injection modes with abstraction of either synthetic or natural groundwater.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 0 7 e3 1 5
concentrations never reached below the detection limit, i.e., some Fe always passed the columns. This was observed in the synthetic and natural groundwater columns, indicating that some CIEX is needed to reach ultra low Fe concentrations at the beginning of an abstraction phase. In Fig. 8 the measurements for Na and Ca are depicted, showing some desorption of Ca during injection (Fig. 8A). For the duration of 9 pore volumes, the Na concentration increased during injection from the concentration in the groundwater (C/C0 ¼ 0.12) to the concentration in the injection water (C/C0 ¼ 1). The tailing of both curves indicates that CIEX proceeds between these cations. The abstraction period shows the reversed trend (Fig. 8B), as the Na concentration of the groundwater (C0) is reached after approximately 8 pore volumes. The Ca concentration is already restored after 6 pore volumes. It should be noted that Ca2þ complexation with natural organic matter (NOM) may have occurred as well, but because of the low binding contact time can be considered insignificant. It may be concluded that CIEX occurs, in the multi-component environment of natural groundwater, between a wide range of cations, resulting in retardation during the abstraction phase. The effect of CIEX is most predominant on Fe retardation, as Fe concentrations in the groundwater are relatively low compared to Ca2þ and Naþ.
3.4.
Fe2þ desorption theory
The column studies in the absence of O2 and/or Naþ confirmed that, in addition to the adsorptive-oxidation process (van Beek,
1985; Rott, 1985), CIEX occurs during subsurface iron removal. The Naþ concentrations of actual injection water will always be lower than the concentrations in the experiments. The experiments therefore overestimate the actual occurrence of CIEX during subsurface iron removal. However, the CEC of clean filter sand in the columns is much lower than that of actual aquifer sand (on average 10 times) resulting in lower CIEX. At the beginning of a full-scale injection cycle, the CIEX will happen in the presence of O2, resulting in immediate oxidation. While injection proceeds, the injection water will penetrate further into the aquifer, but the O2 front will lag behind, caused by the consumption of O2 during Fe2þ oxidation. In other words, the injected water front does not contain any O2, but contains cations for exchange with Fe2þ. The further the injection water flows into the aquifer, the larger the distance between the injected water front and the O2 front becomes. In this moving zone where O2 is absent, the process of CIEX prevails and Fe2þ desorbs from the soil material and travels deeper into the aquifer (Figs. 5A and 6A). In theory, these elevated Fe2þ concentrations never come in contact with the O2 in the injection water and will not oxidize in this cycle (Fig. 9). When abstraction starts and the flow direction is reversed, the desorbed Fe2þ passes the available adsorption sites on the soil grains closer to the production well and is removed through either adsorption or CIEX. In other words, the injection phase of subsurface iron removal mobilizes a part of the adsorbed Fe2þ through CIEX but does not subsequently oxidize the Fe2þ. One may state that this proportion of mobilized Fe2þ limits the system’s efficacy during the abstraction phase, as part of the adsorption sites will be occupied by the desorbed Fe2þ and not by the Fe2þ present in the groundwater. On the other hand, the exchange may also enhance SIR, as the desorbed Fe2þ is pushed deeper into the aquifer and concentrations may be lowered through the buffering capacity of the aquifer material. Whether the Fe2þ truly enhances or limits the efficacy of SIR is yet to be determined, because it all depends on the actual separation between the Fe2þ desorption front and the O2 front. The investigation of this front separation was not the focus of these column experiments and is recommended for future research.
3.5.
Fig. 8 e Na and Ca measurements during a normal injection (A) and abstraction (B) cycle with natural groundwater.
313
Injection elevated O2 concentrations
CIEX may play a role during subsurface iron removal, but it is the supply of O2 to the (im)mobile Fe2þ that determines the efficiency of the system. Appelo et al. (1999) concluded that increasing the oxidant concentration of the injected water would be useless as long as the efficiency is limited by the amount of exchangeable Fe2þ capable of consuming the oxidant. The observation that injected oxygen will not be used in the absence of available oxidizeable Fe2þ is valid, but there is also a reason why injection of higher O2 concentrations can increase the system’s efficacy. Namely, if the injection water contains higher O2 concentrations, the O2 front will not lag far behind the injected water front. The desorbed Fe2þ will then come in contact with O2 for oxidation and not leach out of the oxidation zone into the aquifer. Thus by injecting higher oxidant concentrations, the exchangeable Fe2þ fraction on the soil material can be utilized for oxidation and will contribute
314
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 0 7 e3 1 5
Fig. 9 e Schematic representation of the separation between O2 front and Fe2D desorption peak.
as hydrolyzed Fe3þ hydroxides with their exchangeable/ adsorptive surface area during the following abstraction phase. At water treatment plant Corle (Vitens Water Supply Company) there has been extensive experience with the injection of elevated oxygen concentrations into subsurface iron removal wells. In the past they injected 3000 m3 of drinking water containing 0.28 mM O2, but they have changed the operational mode to the injection of 2000 m3 with an O2 concentration of 0.55 mM. Although the O2 concentration increased with a factor of 1.9, the total O2 injection increased only with a factor 1.3, from 0.84 to 1.1 *103 mol. As a result of this operational change, the volumetric ratio for abstraction (V) and injection (Vi) has increased from an average 7 to 16. This immense and sudden efficiency increase was clearly caused by the operational change, as an increasing efficacy caused by successive cycles (Appelo et al., 1999) may not be expected at stabilized subsurface iron removal wells. WTP Corle calculates V/Vi based on the moment when Fe breakthrough starts, so the moment [Fe]>2 mM is registered.
Fig. 10 depicts the relation of the operational parameter V/ Vi for 12 subsurface iron removal wells after injection of 0.28 mM O2 and 0.55 mM O2. Although there was some variation between the results, considering these are operational data from 12 different wells, it can be concluded that on average the V/Vi increases by an approximate factor 2. This indicates that not the available (adsorbed and/or exchangeable) Fe2þ is limiting during injection, but the supply of O2 to the (in)soluble Fe2þ. The field measurements support the theory that increasing the oxidant concentration has a positive effect on the subsurface removal of Fe, but they do not prove that it is actually the Fe2þ desorption front that is targeted by the higher O2 dose. The field results point towards the assumption that Fe2þ is abundantly available on the soil material. In conclusion, CIEX may occur in the aquifer during injection, but does not seem to limit the system’s efficiency when injecting elevated oxygen concentrations.
4.
Conclusions
In sand column experiments with synthetic and natural groundwater it was found that cation exchange (Naþ-Fe2þ) occurs during the injection-abstraction cycles of subsurface iron removal. The Fe2þ exchange increased at higher Naþ concentration in the injection water, but decreased in the presence of other cations in the groundwater. Field results with injection of high O2 concentrations indicated that not the exchangeable Fe2þ on the soil material is the limiting factor during injection, but it is the supply of O2 to the available Fe2þ.
Acknowledgements
Fig. 10 e The operational V/Vi ratio of 12 wells at WTP Winterswijk with (0.55 mM) and without (0.28 mM) the injection of 47 molmL3 technical oxygen.
The research presented in this article was supported by the InnoWATOR grant (Innovation Program for Water Technology) of Agentschap NL. The authors would like to thank Erik van der Pol and Hans Bergevoet of Vitens Water Supply Company for their support.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 0 7 e3 1 5
references
Appelo, C.A.J., Drijver, B., Hekkenberg, R., de Jonge, M., 1999. Modeling in situ iron removal from ground water. Ground Water 37 (6), 811e817. Appelo, C.A.J., Postma, D., 2005. Geochemistry, Groundwater and Pollution, second ed. Balkema, Rotterdam. Boochs, P.W., Barovic, G., 1981. Numerical-Model Describing groundwater treatment by Recharge of Oxygenated water. Water Resources Research 17 (1), 49e56. Braester, C., Martinell, R., 1988. The Vyretox and Nitredox methods of in situ treatment of groundwater. Water Science and Technology 20, 149e163. Grombach, P., 1985. Groundwater treatment in situ in the aquifer. Water Supply 3 (1), 13e18. Hallberg, R.O., Martinell, R., 1976. Vyredox - in situ purification of groundwater. Ground Water 14 (2), 88e93. Hiemstra, T., van Riemsdijk, W.H., 2007. Adsorption and surface oxidation of Fe(II) on metal (hydr)oxides. Geochimica et Cosmochimica Acta 71, 5913e5933. Mettler, S., Abdelmoula, M., Hoehn, E., et al., 2001. Characterization of iron and manganese precipitates from an in situ groundwater treatment plant. Ground Water 6, 921e930. Mettler, S. (2002) In-situ removal of iron from groundwater: Fe(II) oxygenation, and precipitation products in a calcareous aquifer. PhD dissertation, Swiss Federal Institute of Technology, Zurich. Parkhurst D.L. and C.A.J. Appelo (1999) User’s guide to phreeqc (version 2) e a computer program for speciation, batchreaction, one-dimentional transport, and inverse geochemical calculations. Water-Resources Inverstigation Report 99-4259, US Geological Survey.
315
Rott, U., 1985. Physical, chemical and biological aspects of the removal of iron and manganese underground. Water Supply 3 (2), 143e150. Rott, U., Meyer, C., Friedle, M., 2002. Residue-free removal of arsenic, iron, mangenese and ammonia from groundwater. Water Science and Technology: Water Supply 2 (1), 17e24. Rott, R., Friedle, M., 1985. Physical, chemical and biological aspects of the removal of iron and manganese underground. Water Supply 3, 143e150. Sharma, S.K., Greetham, M.R., Schippers, J.C., 1999. Adsorption of iron(II) onto filter media. Journal of Water SRT-Aqua 48 (3), 84e91. Sharma S.K. (2001) Adsorptive iron removal from groundwater. PhD dissertation, Wageningen University. Sung, W., Morgan, J.J., 1980. Kinetics and products of ferrous iron oxygenation in aqueous systems. Environmental Science & Technology 14, 561e568. Tamura, H., Kawamura, S., Hagayama, M., 1980. Acceleration of the oxidation of Fe2þ ions by Fe(III)-oxyhhydroxides. Corrosion Science 20, 963e971. van Beek, C.G.E.M., 1985. Experiences with underground water treatment in the Netherlands. Water Supply 3 (2), 1e11. van Halem, D., Olivero, S., de Vet, W.W.J.M., Verberk, J.Q.J.C., Amy, G.L., van Dijk, J.C., 2010. Subsurface iron and arsenic removal for shallow tube well drinking water in rural Bangladesh. Water Research 44, 5761e5769. van Halem, D., de Vet, W., Verberk, J., Amy, G., van Dijk, H., 2011. Characterization of accumulated precipitates during subsurface iron removal. Applied Geochemistry 26, 116e124. Wolthoorn A. (2003) Subsurface aeration of anaerobic groundwater; iron colloid formation and the nitrification process. Ph.D. dissertation, Wageningen University.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 1 6 e3 2 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effects of ozone and ozone/peroxide on trace organic contaminants and NDMA in drinking water and water reuse applications Aleksey N. Pisarenko a, Benjamin D. Stanford a,b, Dongxu Yan a,c,d, Daniel Gerrity a,e, Shane A. Snyder a,c,* a
Southern Nevada Water Authority, Applied Research & Development Center, PO Box 99954, Las Vegas, NV 89193, USA Hazen and Sawyer, P.C., Raleigh, NC 27607, USA c University of Arizona, Chemical and Environmental Engineering, Tucson, AZ, USA d Layne Christensen Company, 3804 E. Watkins Street, Phoenix, AZ 85034, USA e Trussell Technologies, Inc., 6540 Lusk Blvd., Suite C274, San Diego, CA 92121, USA b
article info
abstract
Article history:
An ozone and ozone/peroxide oxidation process was evaluated at pilot scale for trace
Received 27 June 2011
organic contaminant (TOrC) mitigation and NDMA formation in both drinking water and
Received in revised form
water reuse applications. A reverse osmosis (RO) pilot was also evaluated as part of the
13 October 2011
water reuse treatment train. Ozone/peroxide showed lower electrical energy per order of
Accepted 15 October 2011
removal (EEO) values for TOrCs in surface water treatment, but the addition of hydrogen
Available online 25 October 2011
peroxide increased EEO values during wastewater treatment. TOrC oxidation was corre-
Keywords:
predicting contaminant removal. A decrease in N-nitrosodimethylamine (NDMA) forma-
N-Nitrosodimethylamine (NDMA)
tion potential (after chloramination) was observed after treatment with ozone and ozone/
Trace organic contaminant (TOrC)
peroxide. However, during spiking experiments with surface water, ozone/peroxide ach-
Ozone
ieved limited destruction of NDMA, while in wastewaters net direct formation of NDMA of
lated to changes in UV254 absorbance and fluorescence offering a surrogate model for
Ozone/H2O2
6e33 ng/L was observed after either ozone or ozone/peroxide treatment. Once formed
Ozone/peroxide
during ozonation, NDMA passed through the subsequent RO membranes, which highlights
Ozone/hydrogen peroxide
the significance of the potential for direct NDMA formation during oxidation in reuse
Pharmaceuticals
applications.
Endocrine
disrupting
compounds
ª 2011 Elsevier Ltd. All rights reserved.
(EDCs) Advanced oxidation process (AOP)
1.
Introduction
As the global population is projected to reach nine billion by 2050, water reuse and desalination will become a critical
water resource for much of the world (Bereta and Miller, 2010). As such, ensuring adequate removal of trace organic contaminants (TOrCs) of health concern in reclaimed waters, including endocrine disrupting compounds (EDCs) and N-
* Corresponding author. University of Arizona, Chemical and Environmental Engineering, 1133 E. James E. Rogers Way, Tucson, AZ 85721-0011, USA. E-mail address:
[email protected] (S.A. Snyder). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.021
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 1 6 e3 2 6
nitrosodimethylamine (NDMA), becomes increasingly significant. While most pharmaceuticals and personal care products (PPCPs) and EDCs are not believed to pose significant human health threats at concentrations currently found in water (EPHC et al., 2008; Snyder et al., 2010), their presence is still a concern for many individuals and is the foundation for much of the resistance to widespread implementation of potable water reuse. Thus, there exists a need to study TOrC removal through various multi-barrier water reuse treatment processes. Various oxidation processes such as ozone, ozone/ peroxide, UV/H2O2, and non-thermal plasma have been reported to be effective in destroying TOrCs (Benotti et al., 2009a; Gerrity et al., 2010, 2011; Rosario-Ortiz et al., 2010; Rosenfeldt and Linden, 2004; Snyder et al., 2007; Wert et al., 2009a), but these processes are not equally able to remove all compounds. For example, one of the advantages of UV/H2O2 is the ability to remove NDMA and NDMA precursors in water (Kruithof et al., 2007) via photolysis and hydroxyl radical (OH) oxidation. This may be especially important in reuse applications where the eventual chlorination and/or chloramination of the finished water may result in more NDMA formation (PehlivanogluMantas et al., 2006). In contrast, ozone and ozone/peroxide have been implicated in direct formation of NDMA during oxidation of dimethylamine (DMA) and may in fact exacerbate NDMA concerns (Andrzejewski et al., 2008). Furthermore, Schmidt and von Gunten reported direct NDMA formation due to oxidation of N,N-Dimethylsulfamide by ozone in natural waters (Schmidt and Brauch, 2008; von Gunten et al., 2010). Despite the direct formation of NDMA, ozonation is extremely effective in reducing formation potential, which is the formation of NDMA following chloramination. Assuming direct NDMA formation can be controlled, it may be worthwhile to ozonate water intended for reverse osmosis (RO), as this water will require chloramination to control biological fouling of the membranes. In addition to the reductions in NDMA formation potential, ozonation can also be used to reduce organic fouling on microfiltration, ultrafiltration, and RO membranes due to the transformation of organic matter (Park et al., 2007; Pisarenko et al., 2011; Stanford et al., 2011; You et al., 2007; Zhu et al., 2010). Since RO is such an energy intensive process, preozonation has the potential to yield sufficient energy and cost savings to warrant its inclusion in advanced treatment trains. Therefore, it is important to not only evaluate the direct benefits of contaminant oxidation but also the benefits realized after integration into the larger treatment train, notably reduced RO breakthrough of contaminants and improved quality of the RO concentrate. The primary objective of this study was to evaluate TOrC destruction in drinking water and wastewater using a pilotscale ozone and ozone/peroxide technology. The efficacy of both ozone and ozone/peroxide were monitored to evaluate NDMA destruction, direct NDMA formation, NDMA formation potential, and the destruction of other TOrCs (e.g., pharmaceuticals and personal care products (PPCPs)). In order to narrow the scope of this research, a subset of the numerous compounds detected in previous occurrence studies was selected for evaluation. The indicator compounds were selected based on their magnitude and frequency of occurrence in water and wastewater (Snyder et al., 2007), varying
317
physical/chemical characteristics and resulting susceptibility to treatment (Snyder et al., 2007; Ternes et al., 2002; Westerhoff et al., 2005) and ease of analytical methods (Trenholm et al., 2009). Since bromate formation has historically been the most significant concern related to the use of ozone in water treatment (Orlandini et al., 1997; Wert et al., 2007), bromate was also monitored during the study. With respect to the advanced treatment trains, the project team also evaluated the integration of this technology as a possible pretreatment to an RO membrane system to minimize transport of trace contaminants and consequent impacts on the RO permeate. In this way, this technology is broadly evaluated for its application in reuse applications.
2.
Materials and methods
2.1.
Pilot-scale equipment and tested waters
For the wastewater tests, a primary-treated wastewater from the Las Vegas Valley was treated using a pilot-scale HYDRAsub Membrane Bioreactor (MBR) system (Hydranautics, Oceanside, CA) with hollow-fiber vacuum-type polyvinylidene fluoride (PVDF) membranes with a nominal pore size of 0.40 mm. The general MBR operating parameters were consistent with those previously described (Stanford et al., 2011) and are provided in the supplementary information (Table SI-I). Briefly, the MBR was operated at a mixed liquor suspended solids (MLSS) concentration of approximately 8000 mg/L, a solids retention time (SRT) of 12 days, and a hydraulic retention time (HRT) of 4 h. For the target compounds that were not completely removed by biological treatment, contaminant concentrations were monitored before and after ozonation using the pilot-scale HiPOx system (APTwater, Pleasant Hill CA). The HiPOx system was used in this study because it recently received California Department of Public Health Title 22 certification for disinfection in wastewater and water reuse applications. In addition benchscale ozonation experiments were performed by spiking ozone into MBR filtrate or secondary effluent samples at similar ozone doses, following a previously reported method (Stanford et al., 2011). The pilot scale RO skid (Hydranautics, CA) was used to monitor breakthrough of organic contaminants through ESPA2-4040 RO membranes. The surface charge of a clean ESPA2 membrane is typically 20 to 40 mV with an isoelectric point at pH of 4.0. The pilot skid was operated at 52% recovery using a single array of 6 elements. The feed water was adjusted to pH 6.8 using 50% sulfuric acid (Brenntag, NV) and spiked with bulk sodium hypochlorite (obtained from the wastewater treatment plant) and food grade ammonium chloride, 99% (Brenntag, NV) to form residual monochloramine. For the drinking water tests, Colorado River water (CRW) from Lake Mead, NV was used as the source water for the pilot-scale HiPOx and bench-scale RO membrane testing. The bench-scale RO testing consisted of two identical GE Osmonics (Minnetonka, MN) Sepa-CF cross-flow membrane flat-sheet cell holders to process the raw and oxidized waters using a protocol previously described (Stanford et al., 2011). The Colorado River water received at the pilot plant was prechlorinated (0.8 mg/L of free chlorine) by the local water
318
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 1 6 e3 2 6
utility to control Quagga Mussel growth in the intake structures. The experimental water initially contained a free chlorine residual less than 0.2 mg/L, but the water was also recirculated for 24 h in a 3000-gallon batch tank to remove residual chlorine. The feed water was also filtered by a 5micron cartridge filter prior to filling the tank. A recirculation rate of 10 gpm allowed for complete turnover of the batch tank nearly 5 times in the 24-hour period. At the end of 24 h, a spike solution of the analytes dissolved in laboratory grade water was added to the batch tank, and the water was recirculated for an additional 24 h. The analytes were not dissolved in a solvent in order to prevent the introduction of potential OH scavengers. Control samples were collected at the beginning and end of the experiments to evaluate consistency in the target compound concentrations. The HiPOx pilot unit is capable of operating in a variety of modes and configurations, at flow rates of 10e25 gpm, and at ozone doses of up to 15 mg/L. The pilot was fed either using liquid oxygen feedstock or a high-purity (99.9%) oxygen gas to generate up to 10% ozone in dry gas. A concentrated hydrogen peroxide solution (34%) was purchased from (EnviroTech Chemical Services, Modesto, CA), diluted to a 1e2% working solution, and injected immediately before ozone injection in the HiPOx reactor. In addition to the injection ports and static mixers, the HiPOx pilot contains a 60-gallon pipeline contactor with numerous sampling ports that allow sample collection for hydraulic residence times ranging from 0 to 5.5 min at a flow rate of 10 gpm. The MBR filtrate and Colorado River water were exposed to the following ozone doses by adjusting oxygen flow, generator power, and monitoring percent ozone in the dry gas: 0.6, 1.5, 3.0, 6.0, and 10.0 mg/L. The transfer efficiency was typically high (>95%) so the applied ozone dose was approximately equal to the transferred ozone dose.
2.2.
Analytical methods and reagents
Samples were collected into 40 mL glass amber bottles containing 40 mg of sodium azide (preservative). Analysis of trace contaminants was determined based on a previously published rapid on-line solid phase extraction (SPE) and LC/MS/MS technique (Trenholm et al., 2009). Briefly, sample extraction and analysis of 1.5 mL was performed using a Symbiosis (Spark Holland) automated SPE, coupled to API4000 QTRAP (ABSCIEX) mass spectrometer. Oasis HLB cartridges were used for the online SPE. Separation was performed using a C18 column (Phenomenex) and with a mobile phase consisting of 5 mM ammonium acetate in DI water: methanol gradient. All samples were analyzed using positive electrospray ionization (ESI) and tandem mass spectrometry, or multiple reaction monitoring (MRM). Quantitation was performed using isotope dilution. For NDMA analysis, samples were collected into 1 L amber bottles, containing 1.0 g of sodium azide and 80 mg of sodium thiosulfate (for quenching any residual chlorine). NDMA quantitation was performed using a GC-MS/MS system and isotope dilution, using a method developed at the SNWA laboratories (Holady et al., 2012). NDMA standards were purchased from Ultra Scientific (Kingstown, RI, USA) and isotopically labeled NDMA was purchased from Cambridge Isotope Laboratories (Andover, MA, USA). For evaluating NDMA-formation potential,
samples were collected from the pilot system and spiked with preformed monochloramine before storing for 10 days at room temperature based on a previously published method (Mitch and Sedlak, 2004). Blank samples of deionized water were always below the method reporting limit (MRL) for NDMA and did not yield any measurable NDMA during the 10-day formation potential test. A monochloramine stock solution was prepared by rapidly mixing sodium hypochlorite into ammonium chloride solution following method described previously (Kumar and Margerum, 1987). Sodium hypochlorite, 10e14 wt% FAC, was obtained from VWR and standardized using iodometric titration prior to use. Ammonium chloride, 99% was obtained from Sigma Aldrich (St. Louis, MO, USA). In the NDMA destruction experiments with MBR filtrate, an NDMA spike solution was fed into the HiPOx with an in-line static mixer. As mentioned earlier, the NDMA was dissolved in water to prevent the introduction of any solvents that would contribute to OH-scavenging. For the CRW NDMA destruction experiment, NDMA was spiked at 2300 ng/L into a 3000 gallon tank. Ozone doses of 0.6, 1.5, 3, 6, and 10 mg/L were used with and without excess hydrogen peroxide at a 0.7 mol ratio to evaluate NDMA destruction. For the CRW TOrC destruction experiment, selected organic compounds were spiked into the batch tank at approximately 1000 ng/L, except for TCEP and TCPP which were spiked at approximately 5000 ng/L. For DOC analysis, samples were collected into glass vial and acidified to pH <3 with hydrochloric acid and filtered through 0.20 mm hydrophilic polypropylene filter (GHP Acrodisk, Pall Life Sciences). A Shimadzu total organic carbon analyzer was used for quantification. In addition to standard water quality analyses, sample absorbance at 254 nm was measured using PerkineElmer Lambda 45 UVeVIS Spectrometer, consistent with Standard Method 5910 B. Fluorescence excitation emission matrices (EEM) were developed using a PTI fluorometer (Birmingham, NJ, USA) for the data acquisition and processed by MatLab (Natick, Massachusetts, USA). A modified Fluorescence Regional Integration (FRI) method described elsewhere (Gerrity et al., 2011; Stanford et al., 2011) and Fluorescence Index (FI) (McKnight et al., 2001) were used to assess changes in dissolved organic matter (DOM). The EEM images were corrected for the Raman Scatter by subtracting emission of the blank and corrected for inner-filter effect, following a previously described method (MacDonald et al., 1997).
2.3.
Energy calculations
The EEO values were determined by plotting ln(C/C0) vs. energy, consistent with previously described methods (Benotti et al., 2009a; Gerrity et al., 2010). The energy required to generate 1 g of O3 was assumed to be 0.01232 kWh. Hydrogen peroxide cost was assumed as $0.68/kg (Rosenfeldt et al., 2006). Using an electrical energy cost of $0.07 per kWh, the cost of hydrogen peroxide addition was converted to units of energy per unit volume of treated wastewater. This additional energy cost was factored into calculation of EEO values for the ozone/peroxide process. The cost of pumping/pressurization was not considered in this calculation as this will vary from process to process.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 1 6 e3 2 6
3.
Results and discussion
3.1.
Water quality parameters
Representative MBR filtrate, secondary effluent from a neighboring wastewater reclamation facility, and CRW water quality parameters are provided in Table 1. Analysis of inorganic water constituents was performed by the water reclamation facility’s lab according to standard methods. The CRW had a lower DOC of 2.6 mg/L, however, a relatively high alkalinity. The MBR filtrate was typically fully nitrified, partially denitrified and equivalent to a full-scale tertiary effluent with a DOC of 6 mg/L, UV254 absorbance of less than 0.130, and turbidity of less than 0.10 NTU as indicated in Table 1.
3.2.
319
Fig. 1 e NDMA removal by ozone and ozone/peroxide in spiked CO river water.
Colorado river water testing
The data are summarized in Fig. 1. All replicate samples and controls were reproducible with a relative percent difference of less than 6.3% based on duplicate samples. As expected, ozone alone had little impact on the removal of NDMA (12% removal at 10 mg O3/L) but ozone with peroxide achieved a similar level of destruction with significantly less ozone (12% removal at 1.5 mg O3/L þ 0.5 mg peroxide/L and 46% removal at 10 mg O3/L þ 3.5 mg peroxide/L). These levels of degradation are consistent with the low second-order ozone (kO3 ¼ 5.3 102 M1 s1) and OH (k$OH ¼ 4.6 108 M1 s1) rate constants for NDMA (Lee et al., 2007b). Experiments with TOrCs showed similar removal trends between ozone and ozone/peroxide (Figs. 2 and 3), while most of the compounds were more amendable than NDMA to ozone oxidation. As expected, the two flame retardants (TCEP and TCPP) were considerably more resistant to oxidation than the other compounds. Both compounds have been reported to have a faster second-order rate constant with hydroxyl radical (TCEP ¼ 5.6 108 M1 s1 and TCPP ¼ 7.0 108 M1 s1) than that of ozone alone at <10 M1 s1 (Pocostales et al., 2010). These tests were performed in CRW with pH of 8.1. Previously it has been shown that decomposition of ozone in basic aqueous solutions
produces hydroxyl radicals (Staehelln and Hoigne´, 1982; Tomiyasu et al., 1985), which supports the observed degradation of these flame retardant compounds with ozone only. The addition of peroxide expedited the formation of OH and yielded greater removal of many compounds compared to ozone alone. In addition to the oxidation experiments above, several hundred gallons of spiked CRWdwith (3 mg/L) and without ozonationdwere collected for flat-sheet testing using ESPA-2 RO membranes. The feed water and RO permeate samples were then analyzed for the various target compounds to evaluate the effects of ozone pretreatment on subsequent RO processes, including RO breakthrough. The removal of all TOrCs through the RO membrane was >95% for the raw/ control water, however with exception of TCEP and TCPP, TOrCs were detected at reportable concentrations in this RO permeate (Table 2). In contrast, none of the target compounds were detected at reportable concentrations in the RO permeate with preozonation. Therefore, the use of preozonation lowers the concentration of the contaminants in the permeate and, by mass balance, also minimizes the concentration of TOrCs in the retentate.
Table 1 e Typical water quality parameters for MBR filtrate, secondary effluent, and CRW. Constituent BOD (mg/L) COD (mg/L) Ammonia (mg-N/L) Bromide (mg/L) Alkalinity (mg/L as CaCO3) Nitrate (mg-N/L) Nitrite (mg-N/L) Silica (Mg/L) TDS (mg/L) pH DOC (mg/L) UV254 absorbance (1/cm) Total coliforms (MPN/100 mL) Turbidity (NTU)
MBR filtrate Secondary CRW effluent <2 <50 <0.5 0.10 99 13.8 <0.10 12.5 1100 7.3 6.0 0.130 <1 <0.1
N/A N/A <0.1 0.15 123 14.0 <0.10 N/A 1050 6.9 7.1 0.132 7.3 104 N/A
N/A N/A N/A 0.09 138 0.6 <0.05 7.5 625 8.1 2.6 <0.050 N/A <0.5
Fig. 2 e TOrC removal by ozone in spiked CO river water.
320
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 1 6 e3 2 6
Fig. 3 e TOrC removal by ozone/peroxide in spiked CO river water.
Although these experiments were performed on spiked surface water, there may be significant implications for water matrices that contain high TOrC concentrations, such as wastewater effluent intended for indirect potable reuse (IPR). This is particularly important as regulators consider treatment goals for advanced water treatment trains in reuse and groundwater recharge applications. With respect to the standard treatment train for IPR, contaminant breakthrough will put an additional burden on the AOP process following RO, which might increase costs for the facilities and cause significant concern for regulators. Furthermore, preoxidation decreases the concentration of contaminants in the RO feed, thereby lowering the concentration in the concentrate/brine stream relative to systems with no preoxidation.
3.3. Wastewater treatment by MBR, ozone, and RO membranes 3.3.1.
TOrC removal by MBR, ozone, ozone/peroxide and RO
Trace organic contaminant concentrations were measured in hydraulically-linked MBR influent and filtrate samples. As expected, the MBR achieved significant reductions in the target compounds that are susceptible to biotransformation and biodegradation (e.g., naproxen, ibuprofen, and
gemfibrozil (SI Table SI-II)), consistent with previous work (Snyder et al., 2006). However, the more biologically recalcitrant compounds, including sulfamethaxozole, TCEP, DEET, primidone, and phenytoin, were persistent in the filtrate. In general, the use of ozone alone in the MBR-filtrate provided removal of most compounds to below detection limits. For easily oxidized compounds in wastewater such as carbamazepine, trimethoprim, and diclofenac, the ozone/peroxide process did not provide additional removal benefits over ozone alone. For persistent compounds such as DEET and TCEP, the addition of hydrogen peroxide achieved only marginal improvements in the oxidation of these contaminants, as shown in Fig. 4 (normalized concentration vs. O3:DOC ratio) and Fig. 5 (normalized concentration vs. O3/ H2O2:DOC ratio). This observation is likely due to the AOP inherent in a high EfOM-ozone system, thereby negating the need for addition of peroxide to promote hydroxyl radical formation (Buffle et al., 2006; Lee et al., 2010; No¨the et al., 2009; Pocostales et al., 2010). Tabulated data for various O3:DOC ratios is included in SI: Tables SI-III and SI-IV. In addition to monitoring the removal of these contaminants during oxidation, changes in the RO concentrate and permeate were determined by calculating the mass balance between RO feed, permeate, and concentrate. The feed and permeate samples were analyzed for the various target compounds, but the concentrate levels were determined based on the 52% recovery of the RO pilot. Without preozonation, RO breakthrough was observed only for carbamazepine (Table SIII), thereby indicating that the pilot-scale RO skid was effective in achieving the reporting limits for the target compounds. The use of preozonation with 1.5 mg/L ozone reduced carbamazepine breakthrough to the reporting limit of the assay.
3.3.2. Removal of TOrCs and correlations to changes in UV254 absorbance and fluorescence Previous studies have established that ozone and ozone/ peroxide are very effective at reducing UV254 absorbance and fluorescence in addition to the oxidation of trace contaminants (Stanford et al., 2011; Wert et al., 2009a). Several studies have also shown that reductions in UV254 absorbance can be correlated to TOrC removal (Gerrity et al., 2010; Nanaboina and Korshin, 2010; Rosario-Ortiz et al., 2010; Wert et al., 2009a, 2009b). Other studies have grouped compounds with similar chemical structures based on their removal efficiency
Table 2 e Summary of TOrC removal through RO (ESPA-2) and ozone D RO process. Analyte (ng/L)
SSW
SSW with RO
SSW with O3 (3 mg/L)
SSW with O3 (3 mg/L) and RO
Meprobamate Phenytoin Carbamazepine Atrazine Primidone TCPP TCEP Atenolol Trimethoprim
860 930 800 820 1000 3100 4100a 780 760
23 19 16 19 24 <200 <200 25 16
290 77 <10 290 70 2500 5000a <25 <10
<10 <10 <10 <10 <10 <200 <200 <25 <10
SSW ¼ Spiked Surface Water. a Precision of these measurements was 1000 ng/L (based on dilution factor and MRL).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 1 6 e3 2 6
321
Fig. 4 e TOrC removal by ozone in MBR filtrate.
during oxidative treatment (Benotti et al., 2009b; Dickenson et al., 2009). In this study it was observed that the percent removal of trace organic contaminants correlated well with the percent reduction in both UV254 absorbance and total integrated fluorescence. Figures SI-2 and SI-3 depict the data graphically, while Tables 3 and 4 summarize the linear regression fitting of the data. Since similar removals and correlations were observed with both ozone and ozone/ peroxide, the ozone/peroxide results have been omitted for brevity. In general, it can be observed from both Tables 3 and 4 (Figures SI-1 and SI-2) that the correlation between contaminant oxidation and changes in fluorescence regional area can be used to group contaminants by resistance to oxidation. Compounds with a low slope (<1.0) and a negative intercept (e.g., TCEP) represent a class of compound that is more resistant to oxidation, thereby requiring higher ozone and ozone/
peroxide doses. These compounds are identified as Group 1 in Tables 3 and 4 and Figures SI-1 & SI-2. Other compounds such as DEET or phenytoin have a moderate slope (>1.0) but a negative intercept. These types of compounds can be grouped as moderately oxidized compounds and constitute Group 2 in Figures SI 1 and 2. The easily oxidized compounds such as trimethoprim or sulfamethoxazole have slopes higher than 1.0 and a positive intercept (Group 3 in Figures SI 1 and 2). The values of the slopes and intercepts described in Tables 3 and 4 and the proposed grouping of compounds is consistent with the magnitude of the second-order ozone and OH rate constants. Thus, in principle, UV254 absorbance and fluorescence correlations for other compounds can be estimated based on their chemical structure and rate constants (Figs. 4 and 5 show removal of TOrCs by groups). In general, the R2 values were >0.90, and the correlations based on
Fig. 5 e TOrC removal by ozone/peroxide in MBR filtrate.
322
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 1 6 e3 2 6
Table 3 e % Removal of TOrCs by ozone vs. % removal of total integrated fluorescence. Group
Contaminant
Slope
Intercept
R2
N
kO3 (M1 s1)
kOH (M1 s1)
1
TCEP
0.37
9.0
0.904
4
n/a (<1)
5.6 108a
2
Meprobamate Primidone DEET Phenytoin
1.22 1.82 1.86 1.54
36.8 70.6 37.1 41.5
0.910 0.971 0.996 0.965
5 3 4 4
n/a (<1) 1.0b n/a (<10) n/a (<10)
n/a (w1 108) 6.7 109b 5.0 109c 6.28 109d
3
Trimethoprim Sulfamethaxozole Carbamazepine
1.39 1.50 1.93
43.4 23.7 18.6
0.992 0.970 0.974
3 4 3
2.7 105e 2.5 106f 3.0 105f
6.9 109e 5.5 109f 8.8 109f
n/a ¼ not available. a Watts and Linden, 2009. b Real et al., 2009. c Song et al., 2009. d Yuan et al., 2009. e Dodd et al., 2006. f Huber et al., 2005.
fluorescence integration were more consistent than those for UV254 absorbance. This level of accuracy may be acceptable for utilities that are unable to monitor for TOrCs but do have the ability to monitor for changes in UV254 absorbance and/or fluorescence, which require much less experience, time, and money. This strategy provides a potential online assessment of process performance and treatment efficacy that may be a significant component of future IPR regulations.
3.3.3. Impact of ozone and ozone/peroxide on NDMA and transport through RO Of concern during water reuse applications is the removal of NDMA through a given process, driven in large part by California’s Title 22 regulations. Additionally, the formation of bromate by ozone processes may be of further concern when potable reuse is considered and ozone is used in bromidecontaining waters. Table SI-V shows changes in the concentrations of NDMA, bromide, and bromate at various ozone and ozone/peroxide doses. Ironically, in the control sample, direct formation of NDMA (6e9 ng/L) was observed in the MBR filtrate treated with ozone and ozone/peroxide. Even at a higher ozone dose of 10 mg/L (O3:DOC ¼ 1.7) with the
addition of hydrogen peroxide, net NDMA destruction was not observed. This direct NDMA formation was re-evaluated during separate experiments (Table 5) resulting in NDMA formation of 13e33 ng/L. In addition, the formation of 36e48 ng/L was also observed in bench-scale ozonation experiments with secondary effluent samples from a separate water reclamation facility (Table 5). These results indicate that direct formation is reproducible regardless of wastewater and ozone system and that there may be variable factors responsible for NDMA formation during ozonation of wastewater. Formation of NDMA during ozonation has recently been reported elsewhere (Hollender et al., 2009; Zimmermann et al., 2011). These results indicate a potential direct path to NDMA formation from reactions of unknown precursors with ozone and/or OH in wastewaters. Without additional screening for factors and precursors, it is not yet clear whether this formation is due to the presence of a precursor material analogous to N,N-dimethylsulfamide (DMS), a previously unknown metabolite of the fungicide, tolylfluanid, which has been shown to form NDMA by reactions with ozone and catalyzed by presence of bromide (Schmidt and Brauch, 2008; von Gunten et al., 2010). This fungicide has reportedly not
Table 4 e % Removal of TOrCs by ozone vs. % removal of UV254 absorbance. Group
Contaminant
Slope
Intercept
R2
N
kO3 (M1 s1)
kOH (M1 s1)
1
TCEP
0.51
5.9
0.954
4
n/a (<1)
5.6 108a
2
Meprobamate Primidone DEET Phenytoin
1.87 2.30 2.00 2.33
16.7 22.5 18.8 14.9
0.971 1.000 0.987 0.993
5 3 4 5
n/a (<1) 1.0b n/a (<10) n/a (<10)
n/a (w1 108) 6.7 109b 5.0 109c 6.28 109d
3
Trimethoprim Sulfamethaxozole Carbamazepine
3.97 4.93 5.79
12.4 18.3 28.7
0.778 0.790 0.834
3 4 3
2.7 105e 2.5 106f 3.0 105f
6.9 109e 5.5 109f 8.8 109f
n/a ¼ not available, a (Watts and Linden, 2009), b (Real et al., 2009), c (Song et al., 2009), d (Yuan et al., 2009), e (Dodd et al., 2006), f (Huber et al., 2005).
323
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 1 6 e3 2 6
Table 5 e Direct formation (Day 0) and formation potential (Day 10eDay 0) of NDMA (ng/L). Source water MBR filtrate MBR filtrate MBR filtrate MBR filtrate MBR filtrate MBR filtrate MBR filtrate RO permeate RO permeate Secondary Secondary Secondary Secondary Secondary
effluent effluent effluent effluent effluent
O3 dose mg/L
H2O2 dose mg/L
Day 0a (ng/L)
Day 10 (ng/L)
NDMA-FPb (ng/L)
0.0 1.5 1.5 6.0 6.0 10.0 10.0 0.0 10.0c
0.0 0.0 0.5 0.0 2.1 0.0 3.5 0.0 0.0
4.3 21 18 31 33 30 29 2.9 13
1600 160 150 78 84 70 74 7.2 12
1596 139 132 47 51 40 45 4 None
0.0 3.55 3.55 7.1 7.1
0.0 0.0 1.3 0.0 2.5
<2.5 48 45 42 36
590 230 230 150 140
590 182 185 108 104
a Day 0 refers to ambient concentrations or direct formation of NDMA from ozonation. b NDMA-FP refers to Day 10 (after addition of preformed chloramine) e Day 0. c Pilot-scale ozonation.
been used in the United States. Other compounds that form NDMA during reactions with ozone were recently summarized (Nawrocki and Andrzejewski, 2011). Therefore, additional research is necessary to determine the exact precursors and reaction pathways responsible for direct NDMA formation during ozonation of wastewaters. In addition to NDMA removal, the impact of ozone and ozone/peroxide treatment on NDMA formation potential (NDMA-FP) was also evaluated. Table 5 summarizes the NDMA-FP results from the MBR-Ozone pilot tests. As indicated, both ozone and ozone/peroxide provided significant reductions in NDMA formation potential (after chloramination). Since, ozone and ozone/H2O2 generally provide similar overall OH exposure in wastewater when sufficient reaction
time is provided, the H2O2 addition is often unnecessary for ozone to qualify as an advanced oxidation process, as indicated by similar reductions in TOrCs and NDMA-FP when using ozone and ozone/peroxide. Since NDMA is hydrophilic (log ko/w ¼ 0.57), low pKa <1 (Lee et al., 2007a), and of low molecular weight (74.08 g/mol), NDMA is poorly rejected by polyamide RO membranes (Plumlee et al., 2008; Steinle-Darling et al., 2007), which is a significant issue in water reuse applications. NDMA rejection/breakthrough for the current study is described in Table SI-V. Although preozonation achieves significant reductions in NDMA-FP, the direct formation during ozonation, subsequent formation during chloramination, and only partial rejection by RO highlights the potential concern for this
Table 6 e Electrical energy per log-order removal (EEO) of TOrC by ozone (O3) and ozone/peroxide (O3/H2O2) in surface and wastewater. Contaminant
Surface water (CRW) 3
TCEP TCPP Ibuprofen Meprobamate Atrazine DEET Primidone Phenytoin Naproxen Sulfamethoxazole Atenolol Gemfibrozil Trimethoprim Carbamazepine Diclofenac Triclosan NDMA
Wastewater (MBR filtrate) 3
EEO (O3) kWh/m -log
EEO (O3/H2O2) kWh/m -log
EEO (O3) kWh/m3-log
EEO (O3/H2O2) kWh/m3-log
0.419a 0.226a N/A 0.142 0.073 N/A 0.055 0.055 N/A N/A <0.004 N/A <0.005 <0.004 N/A N/A 2.608a
0.384a 0.237a N/A 0.043 0.141 N/A 0.016 0.018 N/A N/A <0.007 N/A <0.004 <0.005 N/A N/A 0.603a
1.169a N/A 0.393 0.361a N/A 0.168 0.139a 0.110 0.055 0.039 0.038 0.033 0.030 0.028 0.022 <0.010 N/A
1.371a N/A 0.269 0.245a N/A 0.155 0.145a 0.112 0.060 0.047 0.074 0.034 0.016 0.036 0.051 0.037 N/A
N/A ¼ not available. a EEO value is based on extrapolation, since actual 90% removal was not achieved during the ozone and ozone/peroxide tests.
324
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 1 6 e3 2 6
contaminant in water reuse applications. Furthermore, this highlights the importance of multiple barriers in IPR treatment trains. Bromate formation was generally not a problem (<10 mg/L) for the waters tested in this study as the bromide concentration in the MBR filtrate was typically 0.1 mg/L, which was insufficient to induce significant bromate formation during ozonation. However, in recognition of the issue, when bromide was spiked to approximately 0.4 mg/L, bromate exceeded 10 mg/L for applied ozone doses higher than 3.0 mg/ L. At a dose of 10 mg/L (O3:DOC ratio of 1.7), as much as 42e50 mg/L of bromate was formed. However, as indicated by the results in Table SI-V, bromate was not detected in the RO permeate samples above 10 mg/L. This highlights an additional benefit of the RO system, which is very effective at bromate rejection.
3.4.
Energy
Table 6 provides a summary of the removal of trace contaminants at various O3:DOC ratios and the calculated electrical energy per log-order removal (EEO, kWh/m3-log) values for surface water and wastewater. As expected, the EEO values are lower for the removal of TOrCs in surface water than wastewater, likely due to the presence of higher organic matter content and ozone scavenging in the wastewater matrix than the surface water matrix. Similar trends in the EEO values were recently reported elsewhere (Katsoyiannis et al., 2011). In addition, comparable EEO values for sulfamethoxazole and atrazine are reported in the current study. However, for NDMA, that study reported EEO values of 0.5 kWh/m3 and 0.9 kWh/m3 per 90% removal (same as kWh/m3-log) using ozone treatment of surface and wastewater respectively (Katsoyiannis et al., 2011). As shown by Table 6, in surface water NDMA EEO value of 2.608 kWh/m3-log and 0.603 kWh/m3-log was determined for ozone and ozone/peroxide. This substantial disagreement may be attributable to the significant amount of extrapolation required to reach 90% NDMA removal in both studies. Therefore, these numbers should be used with caution due to the high level of uncertainty inherent in the calculations. Since NDMA has a relatively low rate constant with ozone, significantly higher EEO values for ozone would be expected over the ozone/peroxide, which is consistently reported in the current study. Despite the additional “energy” cost of the hydrogen peroxide, the advanced oxidation process (i.e., ozone/ peroxide) was more efficient than ozone alone for surface water treatment, as indicated by lower TOrC EEO values. In contrast, the addition of hydrogen peroxide did not enhance contaminant removal in the wastewater samples, and in some cases (e.g. atenolol, sulfamethoxazole, diclofenac), the EEO values were higher because of the additional peroxide cost. This may be explained by the fact that in wastewater ozone reacts directly with EfOM to produce hydroxyl radicals, thereby creating a pseudo-advanced oxidation process (Buffle et al., 2006; Lee et al., 2010; No¨the et al., 2009; Pocostales et al., 2010). When compared to previous studies utilizing UV/ peroxide or non-thermal plasma, the ozone and ozone/ peroxide EEO values obtained for CRW and wastewater are
significantly lower, emphasizing the cost-effectiveness of ozone-based treatment for TOrC mitigation (Benotti et al., 2009a; Gerrity et al., 2010; Katsoyiannis et al., 2011).
4.
Conclusions
The results presented here indicate the potential benefit of ozone and ozone/peroxide for TOrC removal in water treatment, wastewater treatment, and as part of an advanced IPR treatment train. Consistent with the literature, the data indicate that the addition of hydrogen peroxide does not provide significant benefits for TOrC mitigation in wastewater treatment, but the greater OH exposure in surface water treatment was observed to be beneficial for the oxidation of the recalcitrant compounds, such as TCEP and DEET. Upon chloramination, NDMA formation potential was significantly reduced by ozone and ozone/peroxide pretreatment. Despite this reduction in NDMA formation potential and the NDMA destruction benefits for surface water treatment, the direct formation of NDMA in wastewater applications may require further mitigation measures. Therefore, the oxidation of trace organic contaminants and potential microbial inactivation must be balanced with the formation of disinfection byproducts, including bromate. For the waters tested in this study, an O3:DOC ratio of 0.50 proved to be the optimal dosing condition to balancing TOrC oxidation, direct NDMA formation, and bromate formation. Furthermore, this study demonstrated the potential use of differential UV254 absorbance or fluorescence as a surrogate for TOrC oxidation.
Acknowledgments The authors thank the managers, operators, and laboratory personnel at the City of Las Vegas Water Pollution Control Facility for providing assistance with facilities, analyses, and operation of the pilot system. The authors also wish to thank Benjamin Freeman, Naomi Jones, Rich Franks, and Steve Peck (Hydranautics), Keel Robinson, and Ricky Villalobos (APTwater) for their technical assistance with running MBR-RO and ozone pilot systems. In addition, the work of the following SNWA staff is acknowledged: Janie Holady, Rebecca Trenholm, Jasmine Koster, Shannon Fergusson, Susanna Blunt, Sujanie Gamage, Jing Lin, Josephine Chu, Roxanne Phillips, and Brett Vanderford. Support for this research has been provided by the WateReuse Research Foundation, project number WRF-08-08 and by participating utilities and industry partners.
Appendix. Supplementary material Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.10. 021.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 1 6 e3 2 6
references
Andrzejewski, P., Kasprzyk-Hordern, B., Nawrocki, J., 2008. Nnitrosodimethylamine (NDMA) formation during ozonation of dimethylamine-containing waters. Water Research 42 (4e5), 863e870. Benotti, M.J., Stanford, B.D., Wert, E.C., Snyder, S.A., 2009a. Evaluation of a photocatalytic reactor membrane pilot system for the removal of pharmaceuticals and endocrine disrupting compounds from water. Water Research 43 (6), 1513e1522. Benotti, M.J., Trenholm, R.A., Vanderford, B.J., Holady, J.C., Stanford, B.D., Snyder, S.A., 2009b. Pharmaceuticals and endocrine disrupting compounds in U.S. drinking water. Environmental Science and Technology 43 (3), 597e603. Bereta, B., Miller, W., 2010. Dynamic growth in water reuse and desalination. Water Reuse & Desalination 1 (1), 6e7. Buffle, M.-O., Schumacher, J., Salhi, E., Jekel, M., von Gunten, U., 2006. Measurement of the initial phase of ozone decomposition in water and wastewater by means of a continuous quench-flow system: application to disinfection and pharmaceutical oxidation. Water Research 40 (9), 1884e1894. Dickenson, E.R.V., Drewes, J.E., Sedlak, D.L., Wert, E.C., Snyder, S. A., 2009. Applying surrogates and indicators to assess removal efficiency of trace organic chemicals during chemical oxidation of wastewaters. Environmental Science and Technology 43 (16), 6242e6247. Dodd, M.C., Buffle, M.O., Von Gunten, U., 2006. Oxidation of antibacterial molecules by aqueous ozone: moiety-specific reaction kinetics and application to ozone-based wastewater treatment. Environmental Science and Technology 40 (6), 1969e1977. EPHC, NHMRC, NRMMC, 2008. Australian Guidelines for Water Recycling: Managing Health and Environmental Risks (Phase 2): Augmentation of Drinking Water Supplies. Biotext Pty Ltd, Canberra, Prepared for Environment Protection and Heritage Council, National Health and Medical Research Council, and Natural Resource Management Ministerial Council, p. 174. Gerrity, D., Stanford, B.D., Trenholm, R.A., Snyder, S.A., 2010. An evaluation of a pilot-scale nonthermal plasma advanced oxidation process for trace organic compound degradation. Water Research 44, 493e504. Gerrity, D., Gamage, S., Holady, J.C., Mawhinney, D.B., Quinones, O., Trenholm, R.A., Snyder, S.A., 2011. Pilot-scale evaluation of ozone and biological activated carbon for trace organic contaminant mitigation and disinfection. Water Research 45 (5), 2155e2165. Holady, J.C., Trenholm, R.A., Snyder, S.A., 2012. Analysis of NDMA and other nitrosamines in water using automated solid phase extraction and gas chromatography-tandem mass spectrometry. American Laboratory, in press. Hollender, J., Zimmermann, S.G., Koepke, S., Krauss, M., McArdell, C.S., Ort, C., Singer, H., von Gunten, U., Siegrist, H., 2009. Elimination of organic micropollutants in a municipal wastewater treatment plant upgraded with a full-scale postozonation followed by sand filtration. Environmental Science & Technology 43 (20), 7862e7869. Huber, M.M., Goebel, A., Joss, A., Hermann, N., Loeffler, D., McArdell, C.S., Ried, A., Siegrist, H., Ternes, T.A., Von Gunten, U., 2005. Oxidation of pharmaceuticals during ozonation of municipal wastewater effluents: a pilot study. Environmental Science and Technology 39 (11), 4290e4299. Katsoyiannis, I.A., Canonica, S., von Gunten, U., 2011. Efficiency and energy requirements for the transformation of organic micropollutants by ozone, O3/H2O2 and UV/H2O2. Water Research 45 (13), 3811e3822.
325
Kruithof, J.C., Kamp, P.C., B.J., M., 2007. UV/H2O2 treatment: a practical solution for organic contaminant control and primary disinfection. Ozone Science & Engineering 29 (4), 273e280. Kumar, K., Margerum, D.W., 1987. Kinetics and mechanism of general-acid-assisted oxidation of bromide by hypochlorite and hypochlorous acid. Inorganic Chemistry 26 (16), 2706e2711. Lee, C., Schmidt, C., Yoon, J., von Gunten, U., 2007a. Oxidation of N-nitrosodimethylamine (NDMA) precursors with ozone and chlorine dioxide: kinetics and effect on NDMA formation potential. Environmental Science & Technology 41 (6), 2056e2063. Lee, C., Yoon, J., Von Gunten, U., 2007b. Oxidative degradation of N-nitrosodimethylamine by conventional ozonation and the advanced oxidation process ozone/hydrogen peroxide. Water Research 41 (3), 581e590. Lee, S., Boo, C., Elimelech, M., Hong, S., 2010. Comparison of fouling behavior in forward osmosis (FO) and reverse osmosis (RO). Journal of Membrane Science 365 (1e2), 34e39. MacDonald, B.C., Lvin, S.J., Patterson, H., 1997. Correction of fluorescence inner filter effects and the partitioning of pyrene to dissolved organic carbon. Analytica Chimica Acta 338 (1e2), 155e162. McKnight, D.M., Boyer, E.W., Westerhoff, P.K., Doran, P.T., Kulbe, T., Andersen, D.T., 2001. Spectrofluorometric characterization of dissolved organic matter for indication of precursor organic material and aromaticity. Limnology and Oceanography 46 (1), 38e48. Mitch, W.A., Sedlak, D.L., 2004. Characterization and fate of Nnitrosodimethylamine precursors in municipal wastewater treatment plants. Environmental Science & Technology 38 (5), 1445e1454. Nanaboina, V., Korshin, G.V., 2010. Evolution of absorbance spectra of ozonated wastewater and its relationship with the degradation of trace-level organic species. Environmental Science & Technology 44 (16), 6130e6137. Nawrocki, J., Andrzejewski, P., 2011. Nitrosamines and water. Journal of Hazardous Materials 189 (1e2), 1e18. No¨the, T., Fahlenkamp, H., Sonntag, C.V., 2009. Ozonation of wastewater: rate of ozone consumption and hydroxyl radical yield. Environmental Science & Technology 43 (15), 5990e5995. Orlandini, E., Kruithof, J.C., vanderHoek, J.P., Siebel, M.A., Schippers, J.C., 1997. Impact of ozonation on disinfection and formation of biodegradable organic matter and bromate. Journal of Water Supply Research and Technology-Aqua 46 (1), 20e30. Park, G., Yu, M., Go, J., Kim, E., Kim, H., 2007. Comparison between ozone and ferrate in oxidising geosmin and 2-MIB in water. Water Science and Technology 55 (5), 117e125. Pehlivanoglu-Mantas, E., Hawley, E.L., Deeb, R.A., Sedlak, D.L., 2006. Formation of nitrosodimethylamine (NDMA) during chlorine disinfection of wastewater effluents prior to use in irrigation systems. Water Research 40 (2), 341e347. Pisarenko, A.N., Yan, D., Snyder, S.A., Stanford, B.D., 2011. Comparing oxidative organic fouling control in RO membrane applications. IDA Journal: Desalination and Water Reuse 3 (2), 45e49. Plumlee, M.H., Lo´pez-Mesas, M., Heidlberger, A., Ishida, K.P., Reinhard, M., 2008. N-nitrosodimethylamine (NDMA) removal by reverse osmosis and UV treatment and analysis via LC-MS/ MS. Water Research 42 (1e2), 347e355. Pocostales, J.P., Sein, M.M., Knolle, W., von Sonntag, C., Schmidt, T.C., 2010. Degradation of ozone-refractory organic phosphates in wastewater by ozone and ozone/hydrogen peroxide (Peroxone): the role of ozone consumption by dissolved organic matter. Environmental Science & Technology 44 (21), 8248e8253.
326
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 1 6 e3 2 6
Real, F.J., Benitez, F.J., Acero, J.L., Sagasti, J.J.P., Casas, F., 2009. Kinetics of the chemical oxidation of the pharmaceuticals primidone, ketoprofen, and diatrizoate in ultrapure and natural waters. Industrial & Engineering Chemistry Research 48 (7), 3380e3388. Rosario-Ortiz, F.L., Wert, E.C., Snyder, S.A., 2010. Evaluation of UV/H2O2 treatment for the oxidation of pharmaceuticals in wastewater. Water Research 44 (5), 1440e1448. Rosenfeldt, E.J., Linden, K.G., 2004. Degradation of endocrine disrupting chemicals bisphenol A, ethinyl estradiol, and estradiol during UV photolysis and advanced oxidation processes. Environmental Science & Technology 38 (20), 5476e5483. Rosenfeldt, E.J., Linden, K.G., Canonica, S., von Gunten, U., 2006. Comparison of the efficiency of OH radical formation during ozonation and the advanced oxidation processes O3/H2O2 and UV/H2O2. Water Research 40 (20), 3695e3704. Schmidt, C.K., Brauch, H.J., 2008. N, N-dimethylsulfamide as precursor for N-nitrosodimethylamine (NDMA) formation upon ozonation and its fate during drinking water treatment. Environmental Science and Technology 42 (17), 6340e6346. Snyder, S.A., Adham, S., Redding, A.M., Cannon, F.S., DeCarolis, J., Oppenheimer, J., Wert, E.C., Yoon, Y., 2006. Role of membranes and activated carbon in the removal of endocrine disruptors and pharmaceuticals. Desalination 202 (1e3), 156e181. Snyder, S.A., Wert, E.C., Lei, H.D., Westerhoff, P., Yoon, Y., 2007. Removal of EDCs and Pharmaceuticals in Drinking and Reuse Treatment Processes. Awwa Research Foundation, Denver, CO, p. 372. Snyder, S.A., Stanford, B.D., Bruce, G.M., Pleus, R.C., drewes, J.E., 2010. Identifying Hormonally Active Compounds, Pharmaceuticals, and Personal Care Product Ingredients of Health Concern from Potential Presence in Water Intended for Indirect Potable Reuse (WRF Project 05-005). WateReuse Foundation, Alexandria, VA, p. 228. Song, W., Cooper, W.J., Peake, B.M., Mezyk, S.P., Nickelsen, M.G., O’Shea, K.E., 2009. Free-radical induced oxidative and reductive degradation of N, N-diethyl-meta-toluamide (DEET): kinetic studies and degradation pathways. Water Research 43, 635e642. Staehelln, J., Hoigne´, J., 1982. Decomposition of ozone in water: rate of initiation by hydroxide ions and hydrogen peroxide. Environmental Science and Technology 16 (10), 676e681. Stanford, B.D., Pisarenko, A.N., Holbrook, D.R., Snyder, S.A., 2011. Preozonation effects on the reduction of reverse osmosis membrane fouling in water reuse. Ozone Science & Engineering 33, 379e388. Steinle-Darling, E., Zedda, M., Plumlee, M.H., Ridgway, H.F., Reinhard, M., 2007. Evaluating the impacts of membrane type, coating, fouling, chemical properties and water chemistry on reverse osmosis rejection of seven nitrosoalklyamines, including NDMA. Water Research 41 (17), 3959e3967.
Ternes, T.A., Meisenheimer, M., McDowell, D., Sacher, F., Brauch, H.J., Haist-Gulde, B., Preuss, G., Wilme, U., ZuleiSeibert, N., 2002. Removal of pharmaceuticals during drinking water treatment. Environmental Science and Technology 36, 3855e3863. Tomiyasu, H., Fukutomi, H., Gordon, G., 1985. Kinetics and mechanism of ozone decomposition in basic aqueous solution. Inorganic Chemistry 24 (19), 2962e2966. Trenholm, R.A., Vanderford, B.J., Snyder, S.A., 2009. On-line solid phase extraction LC-MS/MS analysis of pharmaceutical indicators in water: a green alternative to conventional methods. Talanta 79, 1425e1432. von Gunten, U., Salhi, E., Schmidt, C.K., Arnold, W.A., 2010. Kinetics and mechanisms of N-nitrosodimethylamine formation upon ozonation of N, N-dimethylsulfamidecontaining waters: bromide catalysis. Environmental Science & Technology 44 (15), 5762e5768. Watts, M.J., Linden, K.G., 2009. Advanced Oxidation Kinetics of Aqueous Trialkyl Phosphate Flame Retardants and Plasticizers. Environmental Science & Technology 43 (8), 2937e2942. Wert, E.C., Rosario-Ortiz, F.L., Drury, D.D., Snyder, S.A., 2007. Formation of oxidation byproducts from ozonation of wastewater. Water Research 41, 1481e1490. Wert, E.C., Rosario-Ortiz, F.L., Snyder, S.A., 2009a. Using ultraviolet absorbance and color to assess pharmaceutical oxidation during ozonation of wastewater. Environmental Science and Technology 43 (13), 4858e4863. Wert, E.C., Rosario-Ortiz, F.L., Snyder, S.A., 2009b. Effect of ozone exposure on the oxidation of trace organic contaminants in wastewater. Water Research 43 (4), 1005e1014. Westerhoff, P., Yoon, Y., Snyder, S., Wert, E., 2005. Fate of endocrine-disruptor, pharmaceutical, and personal care product chemicals during simulated drinking water treatment processes. Environmental Science and Technology 39, 6649e6663. You, S.-H., Tseng, D.-H., Hsu, W.-C., 2007. Effect and mechanism of ultrafiltration membrane fouling removal by ozonation. Desalination 202 (1e3), 224e230. Yuan, F., Hu, C., Hu, X., Qu, J., Yang, M., 2009. Degradation of selected pharmaceuticals in aqueous solution with UV and UV/H2O2. Water Research 43 (6), 1766e1774. Zhu, H., Wen, X., Huang, X., 2010. Membrane organic fouling and the effect of pre-ozonation in microfiltration of secondary effluent organic matter. Journal of Membrane Science 352 (1e2), 213e221. Zimmermann, S.G., Wittenwiler, M., Hollender, J., Krauss, M., Ort, C., Siegrist, H., von Gunten, U., 2011. Kinetic assessment and modeling of an ozonation step for fullscale municipal wastewater treatment: micropollutant oxidation, by-product formation and disinfection. Water Research 45 (2), 605e617.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 2 7 e3 3 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Bench-scale study of active mine water treatment using cement kiln dust (CKD) as a neutralization agent Allison L. Mackie, Margaret E. Walsh* Department of Civil and Resource Engineering, Dalhousie University, 1360 Barrington St., Rm D215, PO Box 15000, Halifax, NS B3H 4R2, Canada
article info
abstract
Article history:
The overall objective of this study was to investigate the potential impact on settled water
Received 24 June 2011
quality of using cement kiln dust (CKD), a waste by-product, to replace quicklime in the
Received in revised form
active treatment of acidic mine water. Bench-scale experiments were conducted to eval-
22 September 2011
uate the treatment performance of calcium hydroxide (Ca(OH)2) slurries generated using
Accepted 15 October 2011
four different CKD samples compared to a control treatment with quicklime (CaO) in terms
Available online 19 November 2011
of reducing acidity and metals concentrations in acid mine drainage (AMD) samples taken from the effluent of a lead/zinc mine in Atlantic Canada. Results of the study showed that
Keywords:
all of the CKD samples evaluated were capable of achieving greater than 97% removal of
Mine water
total zinc and iron. The amount of solid alkaline material required to achieve pH targets
Cement kiln dust (CKD)
required for neutralization of the AMD was found to be higher for treatment with the CKD
Quicklime
slurries compared to the quicklime slurry control experiments, and varied linearly with the
Acid mine drainage (AMD)
free lime content of the CKD. The results of this study also showed that a potential benefit
Wastewater treatment
of treating mine water with CKD could be reduced settled sludge volumes generated in the active treatment process, and further research into the characteristics of the sludge generated from the use of CKD-generated calcium hydroxide slurries is recommended. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Mine water from ferrous- and base-metal mines is often contaminated with acidity and dissolved metals due to the weathering of sulphide ores containing minerals such as pyrite (FeS2) and sphalerite (ZnS) (Kuyucak, 2001; Brown et al., 2002; Younger et al., 2002; Johnson and Hallberg, 2005). This acidic effluent is termed acid rock drainage (ARD) or acid mine drainage (AMD), though the issue is not confined to the mining industry. If discharged without treatment, contaminated water from mining operations can release dissolved metals which precipitate once in the aquatic environment due to changes in solubility with pH, coating receiving
waterways with metal hydroxides and ochre from precipitating iron. In addition to being unpleasing aesthetically, elevated concentrations of metal precipitates will coat benthic organisms and fish gills, smothering them (Sengupta, 1993; Brown et al., 2002; Johnson and Hallberg, 2005; Bell and Donnelly, 2006). Metals such as copper, iron, lead, and zinc, among others, are toxic to aquatic species in their bioavailable form, often replacing necessary ions in biomolecules (i.e. enzymes). Suspended and dissolved solids in mine water can also interfere with natural ecosystems by increasing turbidity and reducing light penetration (CCME, 1987; Brown et al., 2002; Bell and Donnelly, 2006). Many jurisdictions have regulations in place to ensure that contaminated waters
* Corresponding author. Tel.: þ1 902 494 8430; fax: þ1 902 494 3108. E-mail addresses:
[email protected] (A.L. Mackie),
[email protected] (M.E. Walsh). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.030
328
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 2 7 e3 3 4
from mining operations are treated effectively prior to discharge. The current most widely used and reliable treatment for ARD is active treatment using alkaline addition, with the most cost effective alkaline material being quicklime or hydrated lime (Brown et al., 2002; Younger et al., 2002; Johnson and Hallberg, 2005). In active lime treatment, mine water is neutralized with a hydrated lime (Ca(OH)2) slurry which is made by slaking quicklime with water. The hydroxide ions in the slurry combine with metal ions in the mine water, resulting in the precipitation of dissolved metals as hydroxides which can then be removed through settling or filtration mechanisms (Brown et al., 2002; Younger et al., 2002). Mine water is usually aerated prior to addition of lime slurry to ensure oxidation of ferrous iron (Fe2þ) to the less soluble ferric iron (Fe3þ) (Younger et al., 2002). After the neutralization stage, a polymer is typically added to the mine water in a slow-mix flocculation stage to allow continued precipitation, adsorption and aggregation of particles into flocs which will settle out more quickly in the clarification step (Huck et al., 1977a). Cement kiln dust (CKD) is a fine-grained caustic material that is generated as a by-product of cement manufacturing. Cement is produced in a similar fashion to quicklime, by introducing raw materials into a kiln where they are heated and thus chemically transformed to produce cement clinker. As with quicklime, the main material used in cement manufacture is limestone (CaCO3), which is transformed into lime (CaO) within a kiln by driving off carbon dioxide (CO2). Cement manufacture includes other raw materials, like silica, alumina and iron, resulting in a product that contains approximately 65% CaO, compared to over 90% for quicklime (Kosmatka et al., 2002; ASTM, 2006). The exhaust gases from the cement kiln are entrained with the tiny particles of CKD that are removed in air pollution control devices. Characterization studies of this cement manufacturing waste product have shown CKD material can contain 8e61 % by weight of total CaO (Todres et al., 1992; Bhatty, 1995; Sreekrishnavilasam et al., 2006; Peethamparan et al., 2008; Mackie et al., 2010). Due to the high lime content of CKD, several studies have been conducted to evaluate the capabilities of this waste byproduct material to remove soluble metals from water through precipitation reactions and gravity settling (El-Awady and Sami, 1997; Pigaga et al., 2005; Zaki et al., 2007). Those studies examined the use of CKD in powder form or CKD leachate for removal of metal ions from synthetic solutions, and found reductions in metals of approximately 100% at optimal conditions (e.g., pH, mixing time). However, available information on the use of CKD for such applications is preliminary and isolated in that these bench-scale studies have been conducted only on selected CKD samples and with synthetic wastewater samples. The overall objective of this study was to investigate the potential impact on settled water quality of using CKD to replace quicklime in active mine water treatment. Bench-scale experiments were conducted to evaluate treatment efficacy by monitoring pH, metals precipitation and removal, and particle settling performance using advanced particle analysis to investigate potential differences in the sedimentation properties and settled water quality of acid mine water
treated with quicklime hydroxide slurries.
and
CKD-generated
2.
Materials and methods
2.1.
Cement kiln dust
calcium
CKD samples were obtained from four North American cement manufacturing facilities for the bench-scale mine water treatment experiments. These samples were selected based on the results of a previous CKD characterization study (Mackie et al., 2010) in order to evaluate CKD samples with varying particle sizes, surface areas and free lime contents, as outlined in Table 1. The properties of the commercial quicklime sample (Graymont, Inc., Havelock, NB, CA) used in this study are also included in Table 1 for comparison. The four CKD samples (CKD-A, CKD-B, CKD-C, and CKD-F) had particle sizes 34e73 % smaller and specific surface areas 2e3 times greater than the quicklime sample, as well as lower percentages of both free and total lime. Lime (calcium oxide or CaO) is the main constituent of both CKD and quicklime. Available lime, or free lime, is measured using a chemical titration that determines the actual amount of CaO and other oxides in a sample that will react in solution. Total lime is a measure of the total calcium in a sample, which includes free lime as well as unreacted limestone (CaCO3) and other calcium-containing minerals (e.g., CaSO4). Analysis of the CKD samples used in this study showed percentages by weight of total lime ranging from 40 to 57 %, compared to 90% for quicklime. The free lime content was found to vary more widely from 5 to 9% for CKD-C and CKD-B, to 15% for CKD-A and 37% for CKD-F versus 87% for quicklime. CKD samples from different cement manufacturing facilities have been shown to have variable physical and chemical characteristics related to kiln type, raw materials used in cement production and other process variables as outlined in previous studies (Todres et al., 1992; Bhatty and Todres, 1996; Peethamparan, 2006; Mackie et al., 2010).
2.2.
Mine water
Mine water samples were collected from the inlet to the effluent treatment plant at a lead/zinc mine located in Eastern Canada. At this mine, three wastewater streams are combined in a buffer pond prior to entering the treatment plant: water pumped out of the current underground workings, leachate
Table 1 e Physicochemical properties of CKD samples. Sample ID
CKD-A CKD-B CKD-C CKD-F Quicklime
Specific surface area (m2/g)
Median particle size (mm)
Total lime (wt%)
Free lime (wt%)
0.502 0.350 0.471 0.393 0.164
8.5 15.9 20.5 21.2 32.0
44 48 40 57 90
15 9 5 37 87
329
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 2 7 e3 3 4
and runoff collected at the base of the tailings impoundment, and collected overflow and runoff from an abandoned open pit mine on the site. The main contaminants of concern in this mine effluent are acidity (pH ¼ 2.4 0.1), total suspended solids (TSS ¼ 70 50 mg/L), iron (total ¼ 429 78 mg/L; dissolved ¼ 399 78 mg/L), and zinc (total ¼ 122 15 mg/L; dissolved ¼ 115 18 mg/L). Average values for these parameters were calculated from 24 samples taken at the time of each batch experiment. Concentrations of select trace metals present in the mine water are presented in Table 2. The mine where these samples were taken uses the conventional High Density Sludge (HDS) process to treat their mine water. The treatment plant uses quicklime slaked onsite and mixed with recycled sludge to increase the mine water pH to a set point of 9.4e9.5. Effluent discharged from the treatment plant’s polishing pond must meet the Canadian guidelines of the Metal Mining Effluent Regulations (MMER, 2002); this includes having a discharge pH between 6.5 and 9.5 and a total zinc concentration in effluent of less than 1.0 mg/L in a single grab sample and less than 0.5 mg/L averaged over 30 days. Treated mine water also cannot be acutely toxic to aquatic species like rainbow trout. The low pH of this mine water, combined with high concentrations of iron and zinc mainly in their dissolved form, signifies that the mine water requires treatment prior to discharge from the site in order to protect the receiving environment.
2.3.
Experimental procedures
Calcium hydroxide slurries were prepared from each CKD and quicklime sample by adding 1 L of Milli-Q water to 250 g of alkaline material. Milli-Q water was used in order to eliminate the potential for interference from ions in tap water. An electric stirrer (Model 1750, Arrow Engineering, USA) was used with a 10 cm diameter circular PVC mixing vessel to simulate slaking of the CKD and quicklime samples at the bench-scale. The temperature of the slurries was also monitored in a separate test for the initial 5 min of slaking. A standard jar tester (Phipps & Bird, Fisher Scientific) was used to simulate the active treatment of the mine water (i.e. neutralization and precipitation) in batch tests. One litre of the mine water samples were added to each of the six 2 L capacity jars and were then dosed with a volume of CKD or quicklime slurry necessary to bring the mine water sample to a pH of 9.5. A pH set point of 9.5 was chosen to target the
minimum solubility of zinc (Charerntanyarak, 1999; Zinck and Aube´, 1999; Kurniawan et al., 2006), and is also the target pH of the active lime treatment plant for the mine water used in this study. The neutralized mine water was then rapid mixed for 1 min at 150 rpm (G ¼ 140 s1) after which 50 mL samples were taken from the sample ports on the jars to determine the degree of metal precipitation. These samples were analysed for soluble (filtered) zinc and iron concentrations, in addition to pH. After neutralization and precipitation, anionic polymer (POLYFLOC AE1138, GE Water & Process Technologies) was added at a dose of 1 mg/L and the samples were rapid mixed for an additional 30 s. Anionic polymers have been shown to work best in mine water treatment plants (Huck and LeClair, 1974). Polymer doses used for flocculation of mine water typically range from 0.5 to 1.5 mg/L, although doses as high as 10 mg/L have been reported (Huck et al., 1977b; Bratby, 2006). The mixing speed was then reduced to 50 rpm, which translates to a velocity gradient of approximately 44 s1, and run for 2 min flocculation time. Samples were taken at this point for analysis of the flocculated particles. The flocculated mine water was then allowed to settle quiescently for 30 min, after which settled water samples were taken in order to characterize the quality of the final treated effluent. Four replicates of the batch jar tests were performed for each of the CKD and quicklime slurries examined in this study. All experiments were performed at room temperature, approximately 22 C. Mixing speeds and times were chosen based on previous work with similar mine effluents (Huck and LeClair, 1978) and with CKD treatment of metal-laden effluents (ElAwady and Sami, 1997). The effect of increasing settling time to 60 min was also investigated for mine water treated with CKD-B, CKD-F, and quicklime slurries.
2.4.
Analytical methods
pH was measured using a variable temperature electrode (accuFlow, Fisher Scientific) with an XL50 meter from Fisher Scientific. The total and soluble zinc and iron concentrations of the samples were measured by atomic absorption spectrometry (AAnalyst 200, PerkinElmer). Samples to be analysed for soluble metals were filtered through a 0.45 mm polysulfone membrane filter (GE Water & Process Technologies) prior to being acidified to a pH less than 2 using concentrated nitric acid for analysis (APHA/AWWA/WEF, 2005). Trace metal
Table 2 e Mine water and settled water treated with CKD and quicklime slurries trace metals analysis by ICP-OES. Element (mg/L) Aluminium Calcium Copper Potassium Magnesium Manganese Sodium Lead Sulphur
Symbol
Mine water (n ¼ 3)
CKD-B (n ¼ 1)
CKD-F (n ¼ 1)
Quicklime (n ¼ 1)
Al Ca Cu K Mg Mn Na Pb S
61 8 200 30 6.5 0.3 87 150 30 55 9 800 300 0.4 0.3 1700 500
1.4 569 0.02 166 115 3.3 918 0.01 1585
1.4 590 0.03 50 70 1.4 943 0.03 1522
0.3 541 0.01 13 2.9 0.1 953 0.02 1356
330
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 2 7 e3 3 4
concentrations were measured using ICP-OES (Vista-PRO Radial, Varian). Total suspended solids (TSS) concentrations of the treated, settled water samples were measured following the procedure outlined in Standard Methods for the Examination of Water and Wastewater (APHA/AWWA/WEF, 2005). Particle size analysis and images of treated mine water samples were analysed using a Micro-Flow Imaging (MFI) system (Brightwell Technologies Inc., Ottawa, ON). This system measures particles greater than 2 mm in diameter when at the low magnification set point, which is 5 times magnification. Wet sludge volumes generated during the settling period in each of the jar tests were estimated from gradations on the jars of the jar tester as outlined by Tatsi et al. (2003) and Amuda and Amoo (2007). In order to accurately calculate the percent precipitation and removal of metals from the treated mine water, samples of the untreated mine water were also analysed for zinc and iron as well as pH and TSS at the time of each jar test. The percent precipitation of zinc and iron was quantified as the decrease in soluble metal concentration of the mine water, calculated as the difference between the concentration measured in the untreated mine water sample and the concentration measured in the treated sample taken after the 1 min rapid mix stage of the jar tests. The final metal removal for either zinc or iron was calculated as the difference between the total metal concentration (i.e. acidification with no filtration) of the untreated mine water sample and the concentration measured in the final settled water sample. Total metals analyses were used for comparison with discharge regulations, which are based on total values (MMER, 2002). Error bars on graphs and error terms in text and tables represent one standard deviation from the mean, except where noted.
3.
Results and discussion
3.1.
Mine water pH adjustment and metals precipitation
The slaking of quicklime is a highly exothermic reaction (Boynton, 1980). In this study, the quicklime slurry temperature increased from approximately 21e64 C in the first 90 s of mixing, dropping to approximately 60 C after 5 min. Slaking of the CKD samples showed only a slight increase in temperature during the experiments, with a maximum temperature of 25 C observed after 5 min of mixing. The volumes of the 25% CKD or quicklime slurries used in the
bench-scale mine water treatment experiments to increase the pH to 9.5 are presented in Table 3, along with the corresponding equivalent weight of dry alkaline material added to the 1 L mine water test volumes. The increased amount of material required ranged from 1.6 times that required for treatment with quicklime for CKD-F, which had the highest free lime content (37%), to 14.3 times the amount of quicklime required for CKD-C, which had the lowest free lime content (5%). The lower free lime and reactive oxide content of the CKDs reduces the concentration of hydroxyl ions (OH) in solution per gram of material added, resulting in an increase in the amount of product required when neutralizing with CKDs compared to quicklime. The concentration of CKD required to increase the pH of the mine water to 9.5 was found to correlate linearly to lime content; the correlation to free lime gave an R2 value of 0.84, and to total lime, 0.81 (Fig. 1). Mine water samples treated with the 25% CKD and quicklime slurries resulted in more than 99% precipitation of soluble zinc and iron. Specifically, soluble concentrations of zinc decreased on average from 115 18 mg/L to 0.38 0.25 mg/L, and soluble iron from 399 78 mg/L to 1.3 1.3 mg/L, after treatment with the CKD and quicklime slurries. As expected, the degree of metal precipitation depended solely on the treatment pH, which was achieved based on the volume of CKD or quicklime slurry that was added to the 1 L mine water samples.
3.2.
Particle size analysis of flocculated water samples
Samples of the flocculated mine water treated with the CKD-B, CKD-F, and quicklime slurries with polymer addition were analysed using Micro-Flow Imaging (MFI). These samples were taken just prior to the beginning of the settling period. Analysis showed that although all three treatments resulted in similar particle size distributions, the two mine water samples treated with CKD-B and CKD-F contained approximately five times more individual particles than the quicklime-treated samples. Specifically, analysis of the CKD-B flocculated sample showed 3.5 105 particles per millilitre (particles/mL) and the flocculated sample analysis after treatment with CKD-F showed 3.6 105 particles/mL. In contrast, the flocculated sample after treatment with
Table 3 e Slurry volumes (25% solids) and equivalent amounts of dry material added to 1 L mine water samples (n [ 4). Sample ID CKD-A CKD-B CKD-C CKD-F Quicklime
Slurry added (ml) 53.8 2.5 70.0 0.0 115 10 13.0 0.0 8.0 0.0
Material added (g) 13.4 17.5 28.7 3.2 2.0
0.7 0.0 2.5 0.0 0.0
Fig. 1 e Correlations between free and total lime content of CKDs and material added to increase pH of mine water to 9.5.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 2 7 e3 3 4
quicklime showed a lower number of particles per mL of sample (0.7 105 particles/mL). Additionally, analysis of the flocculated samples showed the two CKD-treated samples had smaller mean particle sizes and smaller ranges of particle sizes. CKD-B had a mean flocculated particle size of 7.4 11.1 mm; CKD-F, 8.4 25.6 mm; and quicklime, 10.6 37.4 mm. The error term in these measurements indicates the spread of the particle size data from one analysis. All three treatments had the majority of their particles, over 60%, in the smallest size range of 2e5 mm. Fig. 2 presents images of the particles analysed using the MFI instrument. Fig. 2(a) illustrates that the CKD-B treated mine water samples contained many small particles and small, dense flocs. The relative density of the flocs can be seen from the variations in their transparency. The CKD-F treated mine water samples seen in Fig. 2(b) were shown to present larger flocs than the CKD-B treated samples, with visibly denser particles. Treatment of the mine water with quicklime was shown to produce larger and more amorphous flocs than either of the CKD-treated samples evaluated in this study (Fig. 2(c)). However, particles greater than 100 mm in size accounted for less than 2% of all particles detected in all samples analysed.
3.3.
Settled water quality
3.3.1.
Metals
Treatment with the 25% CKD and quicklime slurries removed greater than 98% of total zinc and 97% of total iron from the mine water samples from initial concentrations of 122 15 and 429 78 mg/L, respectively. The final total and soluble zinc and iron concentrations measured in the settled mine water samples showed that the residual metals remained primarily in the insoluble or particulate form (Table 4). The concentrations of soluble iron in the treated mine water samples were minimal, and below the method detection limit (MDL) of 0.05 mg/L for many of the individual samples. Residual soluble zinc concentrations were similarly low, with an MDL of 0.02 mg/L. Statistical analyses using Dunnett’s method (Dunnett, 1964) and Analysis of Variance Analysis (ANOVA) testing confirmed that there was likely no difference between the average zinc removal percentages of the CKD-treated and quicklime-treated mine water (95% confidence level). For iron, only the mine water samples treated with the CKD-C and CKD-F slurries showed no statistical difference in removal percentage to those treated with the quicklime slurry at the 95% CI, however all slurry treatments were equivalent at the 99% percent confidence interval. The percent total removal of the target metals was lower than the percent metal precipitation (presented in Section 3.1) for all treated mine water samples. This difference was greater for the samples treated with the lower free lime content slurries (e.g., CKD-A and CKD-B) than for those treated with the higher free lime content slurries (e.g., CKD-F and quicklime slurries). Collectively, these results demonstrate that although targets can be met in terms of the percent metal precipitation achieved with CKD treatment, longer settling times may be required to achieve more complete removal of target metals from the treated mine water samples. Increased settling time would
331
potentially increase removal since residual metal concentrations measured in the settled water samples were primarily in the particulate form (Table 4). A trace metals analysis was conducted on the settled mine water samples that had been treated with slurries generated
Fig. 2 e (a): Image of flocculated mine water treated with CKD-B slurry. (b): Image of flocculated mine water treated with CKD-F slurry. (c): Image of flocculated mine water treated with quicklime slurry.
332
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 2 7 e3 3 4
Table 4 e Settled water quality parameters (n [ 4). Sample ID
Zinc (mg/L) Total
CKD-A CKD-B CKD-C CKD-F Quicklime
2.1 1.2 2.8 1.5 2.4 1.2 0.9 0.2 1.0 0.3
Iron (mg/L)
Soluble 0.1 0.1 0.1 0.1 0.5
0.1 0.1 0.1 0.1 0.6
Total 11.8 14.4 8.3 3.2 2.9
6.9 7.7 1.4 1.2 0.5
from CKD-B, CKD-F, and quicklime (Table 2). Trace metals are of concern when discussing any potential reuse option for CKD as they can impact the quality of the water stream that is being treated. Previous CKD characterization studies (Haynes and Kramer, 1982; USEPA, 1993) have demonstrated that several trace metals (e.g., barium, manganese, lead and zinc) are found in CKD, but at variable levels. The concentrations of trace metals remaining in the settled mine water samples treated with the CKD-B and CKD-F slurries were similar to or only slightly higher than those measured in the quicklime-treated mine water sample. In addition to zinc and iron, aluminium was also present in high concentrations in the untreated mine water (Table 2), and was removed to less than 2.0 mg/L with treatment by all three slurries. Copper was also reduced to very low concentrations (i.e. <0.1 mg/L) from an initial concentration of 6.5 mg/L. All of the settled water samples showed trace amounts of cadmium, cobalt, chromium and phosphorus removed to less than their detection limits. Magnesium, manganese, and sulphur were found in higher concentrations in the CKD-B and CKD-F treated samples compared to the quicklime-treated samples, although both CKD-treated samples showed a reduction from the initial concentrations of these ions found in the untreated mine water.
3.3.2.
Total suspended solids
The mine water samples treated with the low free lime content CKD slurries (i.e. CKD-A, CKD-B and CKD-C) resulted in higher overall concentrations of TSS in the clarified water after 30 min of settling than those treated with the CKD-F and quicklime slurries (Table 4). In addition to lower free lime content (i.e. less soluble matter) in the raw material, CKD-A, CKD-B and CKD-C were shown in previous work to have smaller particle size distributions than CKD-F and quicklime (Mackie et al., 2010), which provides a possible explanation for the poorer settling ability of mine water treated with those CKD samples. Settled water from samples treated with CKD-B, CKD-F, and quicklime were also analysed for particle size and concentration using MFI equipment. Although the particle size distributions of all three samples analysed were found to be quite similar, the concentrations of particles remaining were not. The majority of particles remaining in the settled water after treatment, 74% for CKD-B, 67% for both CKD-F and quicklime, were of the smallest size fraction, with diameters between 2 and 5 mm. In terms of concentrations of particles, the sample treated with CKD-B had over 5 105 particles/mL in this size range, while the CKD-F treated sample had 3.6 105 particles/ mL, and the quicklime-treated sample, 1.6 105 particles/mL. Particles greater than 20 mm in diameter, which would
TSS (mg/L)
Sludge volume (ml/L)
Soluble 0.2 0.2 0.1 0.1 0.2 0.4 0.1 0.2 0.3 0.5
262 104 302 119 310 71 48 12 61 14
165 153 166 148 276
10 5 14 5 54
potentially settle faster compared to the smaller 2e5 mm particles according to Stoke’s Law, accounted for less than 3% of particles in the sample treated with CKD-B, and for 5% of particles in the samples treated with CKD-F and quicklime. Results of the experiments treating mine water samples with CKD-generated slurries indicated that further reductions in total metals and TSS may be possible with increased settling time. This was inferred from higher total metals, TSS and particle concentrations in settled mine water samples treated with CKD-A, CKD-B, and CKD-C versus those concentrations in samples treated with CKD-F and quicklime, and low residual soluble metals. The effect of increasing the settling time from 30 to 60 min on settled water quality was investigated for mine water treated with the CKD-B, CKD-F, and quicklime slurries. CKD-B was selected from the three low free lime CKDs due to its smaller particle size distribution compared to CKD-C, which was closest to that of CKD-F, and lower free lime content compared to CKD-A, which had the smallest particles. Particle size distributions for these CKDs are presented in Mackie et al. (2010). Increasing the settling time decreased the amount of total zinc in the settled mine water samples treated with the CKD-B slurry (Fig. 3). Specifically, the final total zinc concentration was reduced from 5 2 mg/L with 30 min settling time to 0.18 0.03 mg/L with 60 min settling time, meeting the Canadian discharge guideline. The difference in removal of zinc with increased settling time was not statistically significant for the CKD-F or quicklime-treated mine water samples, which were found to be low (i.e. 1.0 mg/L) after the 30 min settling period. Final total iron levels were also significantly reduced in the mine water samples treated with CKD-B, from
Fig. 3 e Effect of increasing settling time on final total zinc and suspended solids (n [ 3).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 2 7 e3 3 4
19.0 0.6 mg/L to 1.3 0.2 mg/L, when the settling time was increased from 30 to 60 min. Reductions in total iron were also found for samples treated with CKD-F, from 2.6 0.4 mg/L at 30 min to 1.4 0.3 mg/L at 60 min, and quicklime, from 2.7 0.2 mg/L to 0.7 0.4 mg/L. The TSS concentrations remaining in the treated mine water samples after 30 and 60 min of settling are also presented in Fig. 3. The mine water samples treated with slurries made from CKD-F and quicklime showed no change in TSS with additional settling time. However, the mine water samples treated with the CKD-B slurry showed a significant reduction in TSS concentrations, from 392 102 to 53 15 mg/ L (>86%), with an increase in settling time from 30 to 60 min. MFI analysis showed the concentration of remaining particles were reduced in each size range, by an average of 85% for samples treated with CKD-B slurry, 43% for CKD-F and 89% for quicklime. A higher concentration of particles remaining in the CKD-treated samples was found compared to the settled water after quicklime slurry treatment. The particle size distribution of all treated samples was not found to vary significantly with increased settling time. The data presented here show that treatment of mine water samples with the CKD slurries resulted in the generation of more particles than treatment with the quicklime slurry, and that these particles settled out more quickly in the CKD-F treated samples than the CKD-B treated ones. This difference is illustrated by the elevated concentrations of total metals and TSS, in addition to higher concentrations of particles present, after 30 min of settling in the CKD-B treated samples when compared to the CKD-F treated samples. Since the particle size distributions of CKD-B and CKD-F treated samples were similar after flocculation, another parameter may have influenced the settling velocity of these particles. According to Stokes’ Law, particle settling velocity is dictated by particle size, particle density, water viscosity and water density (Droste, 1997). Since the water temperature was maintained at approximately 22 C for all experiments, the density of the flocculated particles may have influenced settling velocities, but was not measured in this study. Particle density would vary depending on the chemical makeup of the solids generated during treatment, which is impacted by the dry alkaline material’s composition. The large amount of inert material in CKD samples would also possibly contribute to increased density of particles. Overall, the longer settling times required to achieve settled water TSS concentration targets when mine water is treated with a low free lime content CKD could potentially translate into increased costs for a treatment plant related to increased polymer doses, although these added costs may be recouped from lower alkaline chemical costs.
3.4.
Settled sludge volumes
As presented in Table 4, the mine water samples treated with the CKD slurries generated significantly lower sludge volumes in the 1 L batch experiments than the quicklime slurry-treated sample ( p-value ¼ 1.56 105). This suggests that the smaller and denser floc particles generated from treatment of mine water with CKD slurries resulted in less voluminous sludge generation.
333
These results are interesting in that they suggest that there may be advantages in using CKD derived calcium hydroxide slurries for the treatment of mine water in active systems over conventional quicklime from a sludge management perspective. The larger amount of dense, inert material and finer particles present in CKD versus quicklime most likely resulted in increased compression and compaction of the settled sludge. It has been demonstrated in other studies that smaller particles can pack in together more closely, thus increasing density and decreasing sludge volume (Jin et al., 2003). Further investigation is needed to determine the impact of CKD characteristics as well as that of precipitation products (e.g., ferrous versus ferric hydroxide) on sludge solids generation and quality.
4.
Conclusions
The results of this study indicate that CKD represents a viable active treatment option for mine water to reduce acidity and precipitate and remove metals, comparable to the performance of quicklime. Treatment of mine water with CKD required higher volumes of the CKD slurries (translating to higher weights of dry material) be added to reach the target pH of 9.5 compared to a control treatment with quicklime slurry. The amount of CKD slurry required to reach the target pH varied linearly with the free lime content of the base alkali material used to generate the calcium hydroxide slurries. All of the CKD samples tested in this study were capable of precipitating more than 99% of the soluble zinc and iron in the mine water samples. Treatment of mine water with the CKD-generated slurries removed over 98% of zinc and 97% of iron, the two soluble metals found in the highest concentrations in the untreated mine water. This was comparable to treatment with quicklime slurry at the 99% confidence level. The TSS concentrations remaining after 30 min in settled mine water samples treated with CKD-A, CKD-B and CKD-C were higher than that found in mine water treated with CKD-F and quicklime, indicating increased settling times or polymer doses would be required to meet treatment requirements. Additional analysis performed during the settling experiments showed that all mine water samples treated with the CKD slurries generated significantly less sludge by volume than samples treated with quicklime. Sludge volumes generated in CKD-treated samples were only 50e60 % of those generated with the quicklime slurry. Further investigation is required to determine if the decreased sludge volume translates to decreased mass and/or increased dewaterability of the sludge, and what effects CKD might have on sludge stability and leaching products. The use of CKD in place of quicklime for active mine water treatment would represent a new market opportunity for this waste product, and can be assumed to also represent a cost savings in chemicals for a treatment plant. This, as well as the potential for savings in sludge disposal, may or may not be outweighed by the increased costs associated with the poorer particle settling properties observed with the low free lime content CKDs. Further research is recommended to
334
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 2 7 e3 3 4
determine the relative impact of these variables on full-scale treatment plant capital and chemical costs.
Acknowledgements The authors would like to acknowledge the Natural Sciences and Engineering Research Council of Canada (NSERC), the Portland Cement Association (PCA), and the Cement Association of Canada (CAC) for funding related to this research. We would also like to acknowledge the contributions of Heather Daurie, Brian Kennedy, Brian Liekens and Blair Nickerson of Dalhousie University, who assisted with equipment and analysis.
references
Amuda, O.S., Amoo, I.A., 2007. Coagulation/flocculation process and sludge conditioning in beverage industrial wastewater treatment. Journal of Hazardous Materials 141 (3), 778e783. American Public Health Association, American Water Works Association & Water Environment Federation, 2005. Standard Methods for the Examination of Water and Wastewater. APHA, Washington. ASTM, 2006. Standard Specification for Quicklime, Hydrated Lime, and Limestone for Environmental Uses. C 1529e06a. Bell, F.G., Donnelly, L.J., 2006. Mining and Its Impact on the Environment. Taylor & Francis, New York. Bhatty, J.I., 1995. Alternative Uses of Cement Kiln Dust. Portland Cement Association, Skokie. Bhatty, J.I., Todres, H.A., 1996. Use of Cement Kiln Dust in Stabilizing Clay Soils. Portland Cement Association, Skokie. Boynton, R.S., 1980. Chemistry and Technology of Lime and Limestone, second ed. John Wiley and Sons, New York. Bratby, J., 2006. Coagulation and Flocculation in Water and Wastewater Treatment, second ed. IWA Publishing, Seattle. Brown, M., Barley, B., Wood, H., 2002. Minewater Treatment: Technology, Application and Policy. IWA Publishing, Dorchester. CCME (Canadian Council of Ministers of the Environment), 1987. Canadian Water Quality Guidelines, Water Quality Branch, Inland Waters Directorate. Environment Canada, Ottawa. Charerntanyarak, L., 1999. Heavy metals removal by chemical coagulation and precipitation. Water Science and Technology 39 (10e11), 135e138. Droste, R.L., 1997. Theory and Practice of Water and Wastewater Treatment. John Wiley & Sons, Hoboken. Dunnett, C.W., 1964. New tables for multiple comparisons with a control. Biometrics 20 (3), 482e491. El-Awady, M.H., Sami, T.M., 1997. Removal of heavy metals by cement kiln dust. Bulletin of Environmental Contamination and Toxicology 59 (4), 603e610. Haynes, B., Kramer, G., 1982. Characterization of U.S. CKD, Bureau of Mines Information Circular (IC) 8885. U.S. Department of Interior. Bureau of Mines, Office of Assistant Director, Minerals and Materials Research, Washington, D.C. Huck, P., LeClair, P.B., 1978. Treatment of Base Metal Mine Drainage at Pilot Scale. Minister of Supply and Services Canada, Ottawa. Huck, P.M., Murphy, K.L., LeClair, B.P., 1977a. Scavenging and Flocculation of Metal-bearing Wastewaters Using Polyelectrolytes. Minister of Supply and Services Canada, Ottawa.
Huck, P.M., LeClair, B.P., 1974. Polymer selection and dosage determination methodology for acid mine drainage and tailings pond overflows. In: Flocculation and Dispersion Symposium, Chemical Institute of Canada, Toronto. Huck, P.M., Murphy, K.L., Reed, C., LeClair, B.P., 1977b. Optimization of polymer flocculation of heavy metal hydroxides. Journal of the Water Pollution Control Federation 49 (12), 2411e2418. Jin, B., Wile´n, B.-M., Lant, P., 2003. A comprehensive insight into floc characteristics and their impact on compressibility and settleability of activated sludge. Chemical Engineering Journal 95 (1e3), 221e234. Johnson, D.B., Hallberg, K.B., 2005. Acid mine drainage remediation options: a review. Science of the Total Environment 338 (1e2), 3e14. Kosmatka, S.H., Kerkhoff, B., Panarese, W.C., MacLeod, N.F., McGrath, R.J., 2002. Design and Control of Concrete Mixtures, seventh Canadian ed. Cement Association of Canada (CAC), Ottawa. Kurniawan, T.A., Chan, G., Lo, W.-H., Babel, S., 2006. Physicochemical treatment techniques for wastewater laden with heavy metals. Chemical Engineering Journal 118 (1e2), 83e98. Kuyucak, N., 2001. AMD prevention and control. Mining Environmental Management, 12e15. Mackie, A., Boilard, S., Walsh, M.E., Lake, C.B., 2010. Physicochemical characterization of cement kiln dust for potential reuse in acidic wastewater treatment. Journal of Hazardous Materials 173 (1e3), 283e291. Metal Mining Effluent Regulations (MMER), 2002. Fisheries Act: Metal Mining Effluent Regulations, SOR/2002e222. Peethamparan, S., 2006. Fundamental study of clay-cement kiln dust (CKD) Interaction to determine the Effectiveness of CKD as a potential Clay soil Stabilizer. PhD Thesis, Purdue University, West Lafayette. Peethamparan, S., Olek, J., Lovell, J., 2008. Influence of chemical and physical characteristics of cement kiln dusts (CKDs) on their hydration behavior and potential suitability for soil stabilization. Cement and Concrete Research 38 (6), 803e815. Pigaga, A., Juskenas, R., Virbalyte, D., Klimantaviciute, M.G., Pakstas, V., 2005. The use of cement kiln dust for the removal of heavy metal ions from aqueous solutions. Transactions of the Institute of Metal Finishing 83 (4), 210e214. Sengupta, M., 1993. Environmental Impacts of Mining: Monitoring, Restoration, and Control. Lewis, Boca Raton. Sreekrishnavilasam, A., King, S., Santagata, M., 2006. Characterization of fresh and landfilled cement kiln dust for reuse in construction applications. Engineering Geology 85 (1e2), 165e173. Tatsi, A.A., Zouboulis, A.I., Matis, K.A., Samaras, P., 2003. Coagulationeflocculation pretreatment of sanitary landfill leachates. Chemosphere 53 (7), 737e744. Todres, H., Mishulovich, A., Ahmad, J., 1992. Cement Kiln Management: Permeability. Portland Cement Association, Skokie. USEPA (United States Environmental Protection Agency), 1993. Cement Kiln Dust, Retrieved September 19, 2011, from USEPA Wastes. http://www.epa.gov/epawaste/nonhaz/industrial/ special/ckd/cement2.htm. Younger, P.L., Banwart, S.A., Hedin, R.S., 2002. Mine Water: Hydrology, Pollution, Remediation. Kluwer Academic Publishers, Dordrecht. Zaki, N.G., Khattab, I.A., Abd El-Monem, N.M., 2007. Removal of some heavy metals by CKD leachate. Journal of Hazardous Materials 147 (1e2), 21e27. Zinck, J.M., Aube´, B.C., 1999. Optimization of lime treatment process. In: Proceedings, 31st Annual Canadian mineral Processors Conference, CIM, Ottawa, pp. 76e93.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 3 5 e3 4 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Dewatering model for optimal operation of sludge treatment wetlands Enrica Uggetti, Albert Argilaga, Ivet Ferrer, Joan Garcı´a* GEMMA, Department of Hydraulic, Maritime and Environmental Engineering, Universitat Polite`cnica de Catalunya, BarcelonaTech, c/Jordi Girona 1-3, Building D1, E-08034 Barcelona, Spain
article info
abstract
Article history:
Sludge treatment wetlands (STW) are used as a dewatering technology in some European
Received 15 July 2011
countries since the 80’s. Although the efficiency of this technology in terms of sludge
Received in revised form
dewatering and mineralisation is well known, design and operation parameters are yet to
1 October 2011
be standardised. The aim of this study is to develop a mathematical model capable of
Accepted 18 October 2011
predicting the water loss with time, in order to optimise the feeding frequency enhancing
Available online 6 November 2011
sludge dewatering and expanding the lifespan of the system. The proposed model is validated with experimental data from one pilot and two full-scale STW. The scenarios
Keywords:
considered indicate that the optimum feeding frequency decreases with the sludge layer
Drying reed beds
height. In this way, systems with a sludge layer of 20 cm, 40 cm and 80 cm (corresponding
Water loss
to 2, 4 and 8 years of operation), should be fed every 2.5, 10 and 30e40 days, respectively.
Water percolation
On the other hand, evapotranspiration (ET) has no effect on the feeding frequency,
Evapotranspiration
although it does increase the sludge dryness from 25% to 45% (for ET of 2.5 and 14.5 mm/
Feeding frequency
d in the case of 20 cm of sludge height). According to the model output, the sludge loading rate is determined as a function of evapotranspiration, feeding frequency and sludge height. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Sludge treatment wetlands (STW), also known as drying reed beds, were developed at the end of the 80’s. Sludge dewatering and mineralisation are the main features of this treatment. STW consist of sealed basins in which sewage sludge is spread onto the surface of a granular filter planted with wetland plants. The operation of these systems consists of feeding periods (from 1 to 10 days) followed by resting periods (from 4 days to 3 months). After feeding, part of the sludge water content is rapidly drained by gravity through the sludge residue and granular filter, while another part is evapotranspirated by the plants. A concentrated sludge layer remains on the surface of the bed. It increases by about
10 cm/year when a surface sludge load around 60 kg TS/ m2$year is applied (Nielsen, 2003). When the sludge layer approaches the top of the basin, feeding is stopped during a final resting period (from 1 e 2 months to 1 year) (Uggetti et al., 2010), aimed at improving the sludge dryness and mineralisation. The final product is subsequently withdrawn, starting the following operating cycle. During the last decades, sludge dewatering and mineralisation have been investigated in pilot and full-scale STW differing on operation parameters (influent sludge, plant species, organic loading rate) and climate conditions (Bianchi et al., 2010; Lienard et al., 1994; Magri et al., 2010; Melidis et al., 2010; Stefanakis et al., 2009; Stefanakis and Tsihrintzis, 2011; Uggetti et al., 2009; Vincent et al., 2011). These studies have
* Corresponding author. Tel.: þ34 934016464; fax: þ34 934017357. E-mail addresses:
[email protected] (E. Uggetti),
[email protected] (I. Ferrer),
[email protected],
[email protected] (J. Garcı´a). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.040
336
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 3 5 e3 4 4
shown that the feeding frequency is a key factor determining the sludge layer increase rate and, consequently, the duration of operating cycles. This is a matter of concern, since sludge withdrawal and transport are the main treatment costs (Uggetti et al., 2011). Indeed, optimisation of the resting period might reduce the system restoration cost by 25% (Giraldi and Iannelli, 2009). Despite the importance of the feeding frequency, currently this operation parameter is not standardised. For instance, some Danish systems are fed for 7e8 days and rest for 55e56 days, while others are fed for 2e3 days and rest for 14e21 days (Nielsen, 2005, 2007). There are systems in France which are fed for 2 weeks and rest for 14 weeks (Troesch et al., 2008). Even if in some cases STW are only fed 3e8 times per year (Summerfelt et al., 1999; Obarska-Pempkowiak et al., 2003). Assuming that the optimum feeding frequency corresponds to the period of maximum water loss, it could be optimised by modelling the water loss from the sludge layer with time. Such a dewatering model should integrate all elements contributing to the water balance of STW, namely: 1) water drainage, 2) evapotranspiration (ET) and 3) precipitation. Water drainage may be explained by the consolidation theory (Terzaghi and Peck, 1967). Consolidation is defined as the reduction of soil volume caused by the expulsion of water under long term static load. This process occurs when any stress (e.g. a load) is applied to a soil, causing water loss and bulk volume reduction. It refers to any process that involves a decrease in the water content of a saturated soil without replacement of water by air. If we assume that the consolidation theory can be applied to sludge (Chu and Lee, 1999; Chang and Lee, 1998), water drainage in STW can be explained by the pressure exercised by the residual sludge layer. Evapotranspiration is the main component in the hydrology of wetlands (Zhou and Zhou, 2009). Indeed, efforts have been done to measure and simulate the ET of wetland plants in different regions of the world (Baird and Maddock, 2005; Borin et al., 2011; Zhongpong et al., 2010). Precipitation could be an important water source, depending on the climate conditions. This work aims at developing a mathematical model to simulate sludge dewatering in STW, in order to optimase and standardise STW operation. The proposed model considers drainage (i.e. consolidation), ET and precipitation in a differential equation which is solved with a numerical method. Model validation is carried out with experimental data from one pilot and two full-scale STW. Different scenarios were then considered to get an insight on the optimum feeding frequency and sludge loading rate in different climate conditions. This model is a useful tool to optimise the feeding frequency, enhancing sludge dewatering and expanding STW operating cycles. To our knowledge this is the first time that a numerical model is developed and validated to improve STW operation.
sludge layer compared well to an unsaturated soil, the process could be simulated by means of a porous flow model using the Richards equation. However, if we assume that the sludge layer is saturated for some days after loading, then the consolidation theory can be applied. The process of consolidation is here explained in an ideal system composed of a spring, a container with a valve in its cover and water (Fig. 1). In this system, the spring represents the compressibility of the sludge and water represents pore water in the sludge. Initially, the container is full of water and the valve is closed, representing fully saturated sludge (Fig. 1a). If a certain load is applied to the cover when the valve is still closed, water pressure is developed (Fig. 1b). This corresponds to the pressure exercised by the weight of the sludge layer in STW. As soon as the valve is opened (Fig. 1c), water starts draining through the valve due to excess pore water pressure. This represents water percolation in STW. When excess pore water pressure is fully dissipated, water drainage stops and the spring alone resists the load (Fig. 1d). The one-dimensional consolidation theory of Terzaghi (Terzaghi and Peck, 1967) is based on the following assumptions: 1. Soil is homogenous (uniform in composition throughout). 2. Soil is fully saturated (zero air voids due to the high water content). 3. Solid particles and water are incompressible (water density is constant and any change of soil volume is only due to change in void ratio). 4. Compression and flow are one-dimensional. 5. Strains in the soil are relatively small. 6. Flow of water in the soil voids is one-dimensional (Darcy’s Law is valid for all hydraulic gradients). 7. The coefficient of permeability and the coefficient of volume compressibility remain constant throughout the process. 8. There is a unique relationship, independent of time, between the void ratio and effective stress. Equation (1) is derived from Terzaghi’s consolidation theory: cv
v2 u vu ¼ vz2 vt
(1)
where Cv is the consolidation coefficient (m2/s), u is the interstitial pressure (N/m s2), z the distance (in our case, sludge height) (m) and t is time (s). Equation (1) can be expressed as a dimensionless equation (Equation (2)) where U is the consolidation ratio, Z the point where consolidation is considered, T the time factor and Ds the pressure variation: v2 U vU ¼ vZ2 vT
(2)
where: U ¼ u=Ds Z ¼ z=H T ¼ cv t=H2
2.
Material and methods
2.1.
Model implementation
2.1.1.
Drainage
The main dewatering mechanism in STW immediately after loading is water percolation through the sludge residue. If the
2.1.2.
Evapotranspiration and precipitation
Evapotranspiration is calculated using the PenmaneMonteith equation (Eq. (3)) (ASCE-EWRI, 2005).
ET0 ¼
Cn u2 $ðes ea Þ 0:408$D$ðRn GÞ þ g T þ 273 D þ g$ð1 þ Cd $u2 Þ
(3)
337
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 3 5 e3 4 4
a
b
c
d
Fig. 1 e Diagram representing the process of consolidation. a) The container is completely filled with water and the hole (valve) is closed (fully saturated soil); b) a load is applied onto the cover, while the hole is still closed; c) the hole is opened and water drains out due to excess pressure; d) when excess pore water pressure is fully dissipated, flow through the soil pores ceases, and the system reaches an equilibrium position, with a compressed soil skeleton resisting the applied force (adapted from Lambe and Withman, 1979).
where ET0 is the standard reference for crop evapotranspiration (mm/d), 0.408 ¼ 1/2.45 converts the unit from MJ/m2 d to mm/d, Rn is the calculated net radiation at the crop surface (MJ/m2), G is the soil heat flow density at the soil surface (MJ/ m2 d, approximated to zero), T is the mean daily air temperature at a height of 1.5e2.5 m ( C), u2 is the daily wind speed at a height of 2 m (m/s), es is the saturation vapour pressure at a height of 1.5e2.5 m (kPa) calculated for daily steps as the average of saturation vapour pressure at the maximum and minimum air temperature, ea is the mean vapour pressure at a height of 1.5e2.5 m (kPa), D is the slope of the saturation vapour pressure-temperature curve (kPa/ C), g is the psychometric constant (kPa/ C). Cn and Cd are 900 and 0.34 (mm/d) for short vegetation (0.12 m), and 1600 and 0.38 (mm/d) for tall vegetation (0.50 m). Wetlands vegetation can be taller than 50 cm, therefore the values for tall vegetation were taken as representative of the upper limit of ET (Allen et al., 2011). If we convert ET and precipitation (P) into pressure and introduce them in Eq. (2), we obtain Eq. (4): v2 U vU s ¼ þ Q$ vZ2 vT u0
(4)
u0 Q ¼ q$ SN
(5)
where Q represents the pressure variation resulting from evapotranspiration and precipitation, contributing to water loss and water supply, respectively. In Eq. (5) the variable q includes both ET and P. Consequently, negative values of q indicate predominance of evapotranspiration, while positive ones indicate predominance of precipitation. u0 is the initial pressure and SN represents the consolidation settlement after infinite time, when all overpressure has been dissipated (U ¼ 1) (Eq. (6)): ZH SN ¼
εðz; NÞ dz ¼ 0
av $Ds$H 1 þ e0
(6)
where av is the compressibility coefficient and e0 the index of pore, which can be deduced by the porosity n0 according to Eq. (7)
e0 ¼
n0 1 n0
(7)
The oedometric modulus (Em) is defined by Eq. (8): Em ¼
av 1 þ e0
(8)
SN becomes dimensionless by integrating the porosity no (Eq. (9)): SN ¼
SN n0 $H
2.1.3.
(9)
Model solution
The consolidation theory can be solved by means of analytical and tabulated solutions. In this case, however, the integration of ET and P makes the Terzaghi’s model more complicated and a numerical method is required to solve Eq. (4). The software Matlab R2010 was here used to create a mesh representing the wetland section. The mesh had 120 nodes in the spatial direction and 240 time steps. This dimension was selected to set the relative error below 102 (moisture probe accuracy). The implicit method of finite difference was used to solve the differential equation at every time step. This method is stable for any time/space step relationship, meaning that the time step selected does not influence the solution. ET and P are applied uniformly in the space as source and sink terms, respectively. Indeed, plants roots are supposed to be uniformly distributed within the sludge layer, and so is precipitation. Bearing in mind that pressure is caused by the sludge layer, initial and boundary conditions are set as follows: initial pressure is a function of the sludge layer height, while surface and bottom pressures are zero. This is due to the fact that both sides have free drainage: at the surface water can drain through the sludge layer, while at the bottom the gravel filter ease water percolation discharging the excess pore water pressure. Boundary conditions were selected in accordance with typical geotechnical problems. In this case, the pressure difference caused by the sludge layer is the main mechanism causing water loss. The final equilibrium state is reached after a long time, when at any value of the excess hydrostatic pressure is zero.
338
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 3 5 e3 4 4
2.2.
Model validation
2.2.1.
Data collection
Moisture data were collected by means of soil moisture probes SM 200 (Delta-T Devices Ltd) connected to a data logger GP 1 (Delta-T Devices Ltd). These probes consist of a sealed plastic body attached to two sensing rods which transmit an electromagnetic field into the sludge. The volumetric water content is determined from the permittivity, which a measures the medium response to polarisation in an electromagnetic field. In the case of sludge, its water content dominates the permittivity, which is measured by the moisture probe with an accuracy of 3%. The probes measure the moisture of the medium in a volume of 0.5 dm3 surrounding the two sensing rods. The calibration provided by the manufacturer for organic soils was verified for sludge. Therefore, sludge moisture was measured with probes located in situ at a depth of 10 cm. Hourly measurements were recorded. The volumetric water content given by moisture probes was converted to pressure in the same way as ET and P. According to Eq. (5), the volumetric values were multiplied by the overpressure and divided by the consolidation settlement after infinite time. Meteorological data for ET calculation were gathered from the municipal meteorological stations of Barcelona Zona Universitaria and Perafita, Barcelona, Spain, located near the STW studied.
2.2.2.
Pilot and full-scale STW
Data for model calibration were collected from a pilot plant located outdoors at the roof of the Department of Hydraulic, Maritime and Environmental Engineering of the Universitat Polite`cnica de Catalunya, Barcelona, Spain. This pilot plant was set up in winter 2008. It consists of three PVC containers with a surface area of 1 m2 and a height of 1 m. Thickened activated sludge from a WWTP near Barcelona (extended aeration), was manually loaded since May 2008. The sludge loading rate was 0.025 m3/week per STW, corresponding to 40 kg TS/m2$year. The STW used in this study had a granular filter composed by 10 cm of stones (d50 ¼ 25 mm), 30 cm of gravel (d50 ¼ 5 mm) and 10 cm of sand (d50 ¼ 1 mm). Three perforated PVC pipes were located at the bottom of the filter to collect the leachate and ease filter aeration. The STW was planted with Phragmites australis. Data for model verification was collected from two fullscale systems located in Alpens (400 PE) and Sant Boi de Lluc¸ane`s (1500 PE) in the province of Barcelona, Spain. The main characteristics of both facilities are summarised in Table 1; additional details may be found in Uggetti et al. (2009). The pilot and full-scale STW were loaded once per week. In model validations the resting period ranged from 5 to 8 days, depending on moisture data availability.
2.2.3.
Table 1 e Main characteristic of the wastewater treatment plants of Alpens and Sant Boi de Lluc¸ane`s, Barcelona, Spain.
Population equivalent Type of treatment Sludge production (kg TS/d) Sludge flow rate (m3/day) Total surface area of the STW (m2) Sludge loading rate (kg TS/m2$year)
Alpens
Sant Boi de Lluc¸ane`s
400 (800 design) Extended aeration 30
600 (1500 design) Extended aeration 45
2
3
198
324
55
51
coeficient is the sludge rigidity, which influences the final consolidation settlement. Cv ¼
K$Em gw
(10)
where K is the permeability (m/s) and gw the water weight (N/m3). Cv (Eq. (1)) and Em (Eq. (8)) are usually determined by fitting a theoretical curve of consolidation with time deduced from the consolidation theory, with the curve obtained by laboratory tests. This method defines the compressibility curve and the consolidation coefficient of a soil sample subjected to a one-dimensional compression. In our case, the standard consolidation test was not possible due to the nature of the sludge. Thus, model calibration was made by setting the sludge height (H ), index of void (e0), porosity (n0), evapotranspiration and precipitation (q). H was measured in situ and n0 deduced from moisture data obtained with probes located in situ. Thus, Cv and Em values were determined by fitting the model simulation curve to moisture curves deduced from data collected in the pilot STW in May 2010. According to the data availability, the model was verified with data obtained mainly during the warm season (Alpens records from May and Sant Boi de Lluc¸ane`s records from February, May and June 2010). The calibration confidence interval corresponds to the probes accuracy (3%). The following statistical tests were used to test model performance: the Mean Absolute Error (MEA), the Mean Square Error Normalised (RMSEN), the Pearson’s Correlation Coefficient (PCC), the Normalised Mean Bias Error (NMBE) and the Nash-Sutcliffe Coefficient of Efficiency (CE). The test are described in detail by Akratos et al. (2009). They were performed with Microsoft Excel 2010 and Minitab 15.0 Statistical Software. The significance of the difference between experimental data and the model output was assessed by means of the ANOVA test using the Minitab 15.0 statistical software.
Model calibration and verification
Two parameters were calibrated in order to solve Eq. (4): the consolidation coefficient (Cv) and the oedometric coefficient (Em). Both parameters are related by Eq. (10). The consolidation coeficient defines the dissipation velocity of the pressure exercised by water in the sludge, which corresponds to the slope of the curve moisture vs. time. The oedometric
2.3.
Scenarios
2.3.1.
Feeding frequency
Once the dewatering model was validated, different scenarios were considered to determine the optimum feeding frequency
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 3 5 e3 4 4
in different climate conditions of the Mediterranean Region. Three climate zones corresponding to mountain, coastal and dry continental climates were selected. In each zone, evapotranspiration was calculated for summer and winter seasons according to the PenmaneMonteith equation (ASCE-EWRI, 2005). Meteorological data for ET calculation were gathered from municipal meteorological stations located in the zones considered. The lowest (2.5 mm/d), average (8.6 mm/d) and highest (14.5 mm/d) ET values where considered as representative of the mountain, coastal and dry continental climates, respectively. These scenarios, characterised by different ET, can also be representative of different seasons. Indeed, average potential ET in summer oscillates between 3.6 and 5.2 mm/d in the North and South of Italy, respectively (Borin et al., 2011). Precipitation was not considered in these simulations. For each scenario, the following sludge layer heights were studied: 20, 40 and 80 cm, according to the typical increase rate of 10 cm/year (Nielsen, 2003). To determine the optimum feeding frequency based on ET and sludge layer height, the derivative of the dewatering curve was used. In order to enhance sludge dewatering, the optimum feeding frequency should include the whole period of sludge dewatering, and not only the maximum instant. This period corresponds to the maximum average dewatering. For this reason, the optimum feeding frequency was determined by calculating the cumulative moving average of the derivative distribution from t ¼ 0. The average of the derivative was calculated as a function of the time elapsed from sludge loading, with the Eq. (11): average ¼ ðderivative ð1 : TÞ=TÞ
(11)
In this way, the peak of the cumulative moving average corresponds to the time interval of maximum water loss. Since the influence of ET on dewatering time is only minor, this calculation was made with ET ¼ 0.
2.3.2.
Sludge loading rate
The maximum water loss corresponding to each feeding frequency was calculated from the derivate of the dewatering curve. By means of the least squares regression, we found a logarithmic relationship between dewatering (Dw) and sludge height (H ) for each feeding frequency (T ) and evapotranspiration (ET). Thus 15 equations like Eq. (12) were found, DwðHÞ ¼ a1 $lnðHÞ þ b1
(12)
where a1 and b1 are numerical constants for each equation, and w is the volume of water lost. The numerical constants of the 15 equations obtained were correlated by means the least squares regression as a function of ET and 5 equations like Eq. (13) were derived, one for each feeding frequency (T ) DwðH; ETÞ ¼ ða2 $ET þ b2 Þ$lnðHÞ þ c2 $ET þ d2
(13)
where a2, b2 c2 and d2 are numerical constants for each equation. By establishing the sludge loading rate (SLR) as the
339
volume of water lost (w); SLR was determined by means of the least squares regression of Eq. (13).
3.
Results and discussion
3.1.
Model validation
The consolidation and oedometric coefficients obtained from model calibration were Cv ¼ 3$108 m2/s and Em ¼ 4$104 N/m2, respectively. These values are consistent with the literature, with Cv ranging from 3$1010 to 3.5$108 m2/s depending on permeability and compressibility (Lambe and Withman, 1979), and Em ranging from 5$105 N/m2 in low consolidated sludge and 5$108 N/m2 in sand. The lowest Em values from the model calibration are attributed to the sludge water content. Fig. 2 shows the model verifications, carried out with data from the two full-scale STW in warm and cold seasons. In this Figure, the decrease of sludge moisture (or water loss) measured after loading is compared with the model output. In general all tests show a good response of the model (Table 2). The Mean Square Error Normalised (RMSEN) and the Normalised Mean Bias Error (NMBE) with values near zero support the validations. The Pearson’s Correlation Coefficient (PCC), with values near 1, indicates a positive correlation between measured and predicted results. According to the Absolute Error (MEA) and the Nash-Sutcliffe Coefficient of Efficiency (CE) the model fits better with data from Sant Boi de Lluc¸ane`s (validations 2 and 3) than Alpens (validation 4). According to the one-way ANOVA, experimental data and the model simulation are not significantly different ( p > 0.01). Only in the case of Alpens (Fig. 2d) daily oscillations of experimental data caused problems in model adaptation, producing a significant difference between measured and simulated data ( p < 0.01). Such daily variations in sludge moisture are attributed to the thin sludge layer (15 cm) which is strongly influenced by temperature and evapotranspiration effects. Although in some cases the curves simulated do not reproduce accurately the trend followed by experimental data (Fig. 2a and d), the final moisture values simulated by the model differ from experimental moisture values by less than 1%. This means that the model is able to simulate water loss within the sludge layer, estimating rather well the sludge moisture at the end of the resting period (in this case 5e8 days). However, the model response to precipitations was not tested since no rainfalls occurred during the periods considered. Notice that the model is valid for different seasons due to its strict dependence on evapotranspiration, which reflects climate conditions. In fact, the final moisture value predicted by the model corresponds to the value recorded by the probe both in winter (Fig. 2a), when sludge moisture is reduced by 5%, and in spring (Fig. 2b, c and d), when sludge moisture is reduced by 10% in 8 days. The comparison between the curves measured in February and May (Fig. 2a and b) highlights the difference between the water loss in different seasons. In winter (Fig. 2a) the initial sludge moisture is high (around 92%) and the water loss is low due to a low evapotranspiration rate, which does not enhance sludge drying during the cold season.
340
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 3 5 e3 4 4
95
95
Water content (%)
b 100
Water content (%)
a 100 90 85 80
Model Moisture probe data
75 70
0
1
2
90 85 80 75
3
4
5
70
6
Time (d)
95
Water content (%)
95
Water content (%)
d 100
90 85 80
Model Moisture probe data
70
0
2
0
1
2
3
4
5
6
7
Time (d)
c 100
75
Model Moisture probe data
Model Moisture probe data
90 85 80 75
4
6
8
10
70
0
Time (d)
1
2
3
4
5
6
Time (d)
Fig. 2 e Model verifications corresponding to February (a), May (b) and (d) and June (c) with data from Sant Boi de Lluc¸ane`s (a, b and c) and Alpens (d).
3.2.
Scenarios
3.2.1.
Sludge dewatering
The main goal of developing this model was to improve and standardise STW operation. In order to determine the most appropriate feeding frequency, different scenarios were tested by evaluating the influence of two variables: evapotranspiration and sludge layer height. Fig. 3 shows the curves of water loss corresponding to these scenarios. Dewatering curves represent moisture decrease during some days after sludge loading (t ¼ 0), when moisture corresponds to the initial value (82% in this case). The sharp slopes of dewatering curves after sludge loading (t ¼ 0) indicate that water is mainly lost during the first days of resting. The curves become then more flat until the asymptote, when moisture remains constant. In the model, the temporal distribution of water loss is mainly governed by the consolidation of the sludge layer caused by its pressure, thus the sludge layer height is the main parameter
influencing the shape of the dewatering curves. This means that, after the same resting time water loss increases with the sludge layer thickness, during the first 2 years of operation the water loss (up to 25%, Fig. 3a) is lower than during the following years (>30%, Fig. 3b and c). As expected, evapotranspiration has a significant effect on the dewatering performance of STW. If we look at the scenario with a sludge layer of 20 cm (Fig. 3a), the water loss increases from 5 to 20% with ET of 2.5 and 14.65 mm/d, respectively. Indeed, the slope of the moisture curve and the sludge dryness achieved at the end of the resting period increases with ET. This is even more evident in a thick sludge layer, when the sludge dryness increases significantly (Fig. 3b and c).
3.2.2.
Feeding frequency
The dewatering curves obtained from the scenarios investigated are useful to determine the optimum feeding frequency, which is the main practical application of this model. Indeed,
Table 2 e Results from the statistical analysis assessing the model performance in the 4 verifications performed: Mean Absolute Error (MAE), Mean Square Error Normalised (RMSEN), Pearson’s Correlation Coefficient (PCC), Normalised Mean Bias Error (NMBE) and Nash-Sutcliffe Coefficient of Efficiency (CE). Error criterion MAE RMSEN PCC NMBE CE
Verification 1
Verification 2
Verification 3
Verification 4
Optimum value
1.12 0.01 0.73 0.01 0.05
0.34 0.01 0.99 0.00 0.97
0.75 0.01 0.95 0.00 0.60
1.23 0.01 0.79 0.02 0.39
0 0 1 or 1 0 1
341
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 3 5 e3 4 4
a
85 80
70 65 60 55
90 ET= 2.5 mm/d ET= 8.6 mm/d ET= 14.5 mm/d
80
Water content (%)
75
Water content (%)
b
ET= 2.5 mm/d ET= 8.6 mm/d ET= 14.5 mm/d
70
60
50
50 45
0
10
20
40
30
0
10
20
Time (d)
c
Time (d)
85 ET= 2.5 mm/d ET= 8.6 mm/d ET= 14.5 mm/d
80 75
Water content (%)
30
70 65 60 55 50 45 0
5
10
15
20
25
30
35
Time (d) Fig. 3 e Scenarios corresponding to different climates within the Mediterranean Region: mountain (2.5 mm/d), coastal (8.6 mm/d) and dry continental (14.5 mm/d). The sludge height is 20 cm (a), 40 cm (b) and 80 cm (c).
all curves corresponding to the same sludge height reach the asymptote at the same time (10 and 20 days for sludge height of 20 and 40 cm, respectively). This means that different ET influence only slightly the dewatering time. According to this, the sludge layer height influences the feeding frequency, while ET is responsible for the sludge dryness achieved during the resting period between feedings. As shown by the peak of the cumulative average of the derivative distribution (Fig. 4), a system with a sludge layer of 20, 40 and 80 cm, corresponding to 2, 4 and 8 years of operation, should be fed every 2.5, 10, and 30e40 days, respectively. This patter is confirmed by the curves simulated by the model (Fig. 3): with a sludge layer of 20 cm (Fig. 3a) the water content is reduced from 10 to 20% during the first 3 days after feeding; whereas with a sludge layer of 40 cm (Fig. 3b) 10 days are needed to reach the same water loss (10e20%). Concerning the feeding procedure, single loading event is recommended to increase the sludge pressure and enhance sludge dewatering, according to the hypothesis of this model. The optimum feeding frequency corresponding to each sludge height was extrapolated from the scenarios. 3 points were used to derive Eq. (14), and the optimum feeding frequency was determined as a function of the sludge height. This corresponds to the time interval of maximum water
loss. If not enough time is left between feedings, the sludge dryness achieved between feedings is reduced increasing the sludge height, decreasing the duration of operation cycles and increasing the treatment costs. (14) TðdÞ ¼ 62:5$H2 ðmÞ
3.2.3.
Sludge loading rate
The maximum water loss corresponding to each feeding frequency was calculated from the derivate of the dewatering curve (Table 3). SLR, determined by means of the least squares regression of Eq. (13), results as a function of ET, time between feedings (T ) and sludge height (H ) (Eq. (15)). Note that R2 > 0.98 in all the correlations tested. SLR ¼ T$½ð0;596$ET þ 3;863Þ$lnðHÞ þ ð 0;014$T þ 1;239Þ$ET þ ð0:014$T þ 7:638Þ
(15)
where SLR is the height of sludge loading (mm), H is the sludge layer height (m), T is the time between feedings (d) and ET is the evapotranspiration (mm/d). In this way the model can be used to determine operational criteria and improve STW operation. According to Nielsen (2005), the main STW operational problems are 1) overloading, 2) insufficient duration of resting periods and 3) inadequate operation criteria. Systems
342
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 3 5 e3 4 4
a
0.02
0.04
b
derivative
derivative average derivative
average derivative
Dderivative
Derivative
0.03
0.01
0.02
0.01
0
0
2
Time (d)
c
4
0
6
0
2
4
6
8
Time (d)
10
12
14
0.12 derivative average derivative
0.10
Derivative
0.08 0.06 0.04 0.02 0
0
10
20
30
40
50
Time (d) Fig. 4 e Intersection between the derivative of the dewatering curve and the mean of the derivative curve in each time step, the sludge height is 20 cm (a), 40 cm (b) and 80 cm (c).
with short resting periods show slow dewatering and permanent water content on the sludge surface, leading to a rapid growth of the sludge layer and anaerobic conditions with consequent odours emission. In this way, the standardisation of parameters such as the feeding frequency and loading rate can be helpful to improve the performance of STW resulting in a drier final product and reducing STW operation costs.
3.3.
Model reliability and further implementations
In this model, the sludge layer height increase caused by sludge feeding was not considered, thus the results are independent from the sludge volume fed. This assumption is acceptable considering that the sludge fed is rapidly drained and the increase of sludge layer height after each feeding is insignificant with respect to the total sludge thickness. However, the introduction of the sludge volume fed as model input would improve further the model performance. The model was validated with data from systems fed with secondary sludge from extended aeration, thus the sludge quality was not introduced as a variable of the model. However, the sludge dewaterability might change according to sludge properties. The introduction of this parameter could
improve the model performance and enlarge its range of application. Sludge permeability along the vertical sludge profile was considered constant, while after some days of resting the surface layers tend to be dryer and less permeable than the bottom. By considering the decrease in sludge permeability over the resting time, sludge drainage would be delayed during the first instants after loading, but the feeding frequency would probably not be affected. The model was calibrated by determining the consolidation and oedometric coefficients from data collected by probes located within the top layer of the sludge. Recorded values were assumed to be constant along the vertical sludge profile. However, the introduction of different sludge layers and the calibration with moisture probes at different depths would make the model more reliable. In the scenarios and in the determination of the time frequency, the consolidation and oedometric coefficients were considered constant over time. This assumption should be tested with experimental data from sludge with different age. This model can be used to estimate water loss during some days or weeks of resting between feedings. However, the model hypothesis should be reformulated for dry sludge, which can be found after months or years without fresh
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 3 5 e3 4 4
Table 3 e Sludge dewatering (mm/d) as function of sludge height (H ) and evapotranspiration (ET) calculated from the derivate of the dewatering curve shown in Fig. 4. Feeding frequency (d)
H (m)
ET (mm/d) 2.5
8.6
14.5
3.5
0.25 0.30 0.35
4.28 5.27 6.16
7.69 9.31 10.71
10.98 13.27 15.12
7
0.35 0.40 0.50
4.23 4.95 6.20
7.60 8.81 10.79
10.86 12.54 15.24
14
0.50 0.55 0.60
4.28 4.80 5.27
7.69 8.55 9.34
10.98 12.19 13.27
28
0.70 0.80 0.90
4.23 4.95 5.60
7.60 8.81 9.86
10.86 12.54 13.99
sludge loading. In fact, the hypothesis of this model are only valid for saturated conditions and sludge dryness implying a consolidation degree higher than the consolidation settlement after infinite time (SN) would invalidate the model. During the last decades, efforts have been done in order to understand consolidation in unsaturated soils (Conte, 2004). This could be useful in order to improve the model and enlarge its range of application. Model performance could also be improved by introducing the effect of plants. It is well known that plants contribute to sludge dewatering by evapotranspiration, but also by cracking the sludge surface with the movement of stems. Such phenomena may contribute positively to sludge dewatering by accelerating water percolation. This aspect was partially and indirectly considered in the model by means of the Cv term, which influences the consolidation time. However, the introduction of sludge cracking in the model is limited by the difficulty of determining stems distribution within the sludge. Regarding evapotranspiration, the standardised reference crop evapotranspiration was assumed in this study. In this sense, the introduction of the actual evapotranspiration of different seasons and climate conditions would improve the model reliability. Thus, ET calibration with real data is suggested prior to the model application. Due to its relevance in the water balance in constructed wetlands, precipitation was introduced into the model. However, the model response to precipitation was not tested here, since no rainfalls occurred during the periods considered. Consequently, rainfalls were not considered in the scenarios studied. Even though the model is focused on sludge dewatering, mineralisation is another key process in STW. By introducing a mineralisation term, the model would consist of a powerful tool to optimise STW operation, since it would include the main process of sludge treatment in constructed wetlands (dewatering and mineralisation). However, this requires a deep knowledge on the mineralisation process in STW, which is nowadays still not available.
4.
343
Conclusions
A dewatering model capable of simulating sludge dewatering in STW was developed and solved by means of a finite difference method. The model combined Terzaghi’s consolidation theory, representing water loss by percolation and ET. It was validated with moisture data from one pilot plant and two full-scale STW. According to the verification test, the model is able to predict water loss with time, differing from real moisture values by less than 1%. Different scenarios were considered to determine the optimal feeding frequency, which depends on the sludge layer height, hence on the years of operation. In this way, systems with a sludge layer of 20 cm, 40 cm and 80 cm (corresponding to 2, 4 and 8 years of operation), should be fed every 2.5, 10 and 30e40 days, respectively. On the other hand, evapotranspiration (ET) has no effect on the feeding frequency. According to the model output, the sludge loading rate is determined as a function of evapotranspiration, feeding frequency and sludge height.
Acknowledgements This work was funded by the Spanish Ministry of Environment (MMARM, Projects A335/2007 and 087/PC08). E. Uggetti kindly acknowledges the Universitat Polite`cnica de Catalunya for her PhD grant.
references
Akratos, C.S., Papaspyros, J.N.E., Tsihrintzis, V.A., 2009. Artificial neural network use in ortho-phosphate and total phosphorus removal prediction in horizontal subsurface flow constructed wetlands. Biosystems Engineering 102 (2), 190e201. Allen, R.G., Pereira, L.S., Howell, T.A., Jensen, M.E., 2011. Evapotranspiration information reporting: I. Factors governing measurement accuracy. Agricultural Water Management 98, 899e920. ASCE-EWRI, 2005. The ASCE Standardized Reference Evapotranspiration Equation Technical Committee Report Of the Environmental and Water Resource Institute of the American Society of Civil Engineers from the Task Committee on Standardization of Reference Evapotranspiration. Baird, K.J., Maddock, T., 2005. Simulating riparian evapotranspiration: a new methodology and application for groundwater models. Joutnal of Hidrology 312 (1e4), 176e190. Bianchi, V., Peruzzi, E., Masciandaro, G., Ceccanti, B., Ravelo, S., Iannelli, R., 2010. Efficiency assessment of a reed bed pilot plant (Phramites australis) for sludge stabilisation in Tuscany (Italy). Ecological Engineering. doi:10.1016/ j.ecoleng.2010.05.008. Borin, M., Milani, M., Salvato, M., Toscano, A., 2011. Evaluation of phragmites australis (Cav,) evapotranspiration in Northern and Southern Italy. Ecological Engineering 37, 721e728. Chang, I.L., Lee, D.J., 1998. Ternary expression stage in biological sludge dewatering. Water Research 32 (3), 905e914. Chu, C.P., Lee, D.J., 1999. Three stage of consolidation dewatering of sludges. Journal of Environmental Engineering 125 (10), 959e965.
344
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 3 5 e3 4 4
Conte, E., 2004. Consolidation analysis of unsaturated soils. Canadian Geotechnical Journal 41 (4), 599e612. Giraldi, D., Iannelli, R., 2009. Short-term water content analysis for the optimization of sludge dewatering in dedicated constructed wetlands (red bed systems). Desalination 246, 92e99. Lambe, W., Withman, R.V., 1979. Soil Mechanics. SI Version. John Wiley & Sons, New York, NY, 553 pp. Lienard, A., Duchene, P., Gorini, D., 1994. A study of activated sludge dewatering in experimental reed-planted or unplanted sludge drying beds. Water Science and Technology 32, 251e261. Magri, M.E., Suntti, C., Voltolini, C.A., Philippi, L.S., 2010. Performance of different macrophytes species in constructed wetlands systems for anaerobic sludge dewatering, experience from Southern Brazil. In: Proceeding of Conference: 12th International Conference on Wetland Systems for Water Pollution Control 4e8 October. Venice (Italy). Melidis, P., Gikas, G.D., Akratos, C.S., Tsihrintzis, V.A., 2010. Dewatering of primary settled urban sludge in a vertical flow wetland. Desalination 250 (1), 395e398. Nielsen, S., 2003. Sludge treatment in wetland systems. In: Dias, V., Vymazal, J. (Eds.), Proceedings of Conference: The Use of Aquatic Macrophytes for Wastewater Treatment in Constructed Wetlands (IWA) 8e10 May, Lisbon (Portugal). Nielsen, S., 2005. Mineralization of hazardous organic compounds in a sludge reed bed and sludge storage. Water Science and Technology 51 (9), 109e117. Nielsen, S., 2007. Helsinge sludge reed beds systems: reduction of pathogenic microorganisms. Water Science & Technology 56 (3), 175e182. Obarska-Pempkowiak, H., Tuszynska, A., Sobocinski, Z., 2003. Polish experience with sewage sludge dewatering in reed systems. Water Science & Technology 48 (5), 111e117. Stefanakis, A.I., Akratos, C.S., Melidis, P., Tsihrintzis, V.A., 2009. Surplus activated sludge dewatering in pilot-plant sludge drying reed beds. Journal of Hazardous Materials 172, 1122e1130.
Stefanakis, A.I., Tsihrintzis, V.A., 2011. Dewatering mechanisms in pilot-scale sludge drying reed beds: effect of design and operational parameters. Chemical Engineering Journal 172, 430e443. Summerfelt, S.T., Adler, P.R., Glenn, D.M., Kretschmann, R.N., 1999. Aquaculture sludge removal and stabilization within created wetlands. Aquacultural Engineering 19, 81e92. Terzaghi, K., Peck, R.B., 1967. Soil Mechanics in Engineering Practice, second ed. Wiley, New York. Troesch, S., Lie`nard, A., Molle, P., Merlin, G., Esser, D., 2008. Sludge drying reed beds: a full and pilot-scales study for activated sludge treatment. In: Proceeding of Conference 11th Int. Conf. Wetland Syst. for Water Pollut. Control. 1e7 November. Indore (India). Uggetti, E., Llorens, E., Pedescoll, A., Ferrer, I., Castellnou, R., Garcı´a, J., 2009. Sludge dewatering and stabilisation in drying reed beds: characterisation of three full-scale systems in Catalonia, Spain. Bioresource Technology 100 (17), 3882e3890. Uggetti, E., Ferrer, I., Llorens, E., Garcı´a, J., 2010. Sludge treated wetlands: a review on the state of the art. Bioresource Technology 101 (9), 2905e2912. Uggetti, E., Ferrer, I., Molist, J., Garcı´a, J., 2011. Technical, economic and environmental assessment of sludge treatment wetlands. Water Research 45 (2), 573e582. Vincent, J., Molle, P., Wisniewski, C., Lienard, A., 2011. Sludge drying reed beds for septage treatment: towards design and operation recommendations. Bioresource Technology 102 (17), 8327e8330. Zhou, L., Zhou, G., 2009. Measurement and modeling of evapotranspiration over a reed (Phragmites australis) marsh in Northeast China. Journal of Hydrology 372, 41e47. Zhongpong, S., Bin, W., Wei, S., Wenming, S., Changzuo, W., Daian, Y., Zheng, L., 2010. Evapotranspiration estimation based on the SEBAL model in the Nansi Lake wetland of China. Mathematical and Computer Modelling. doi:10.1016/ j.mcm.2010.11.039.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 4 5 e3 5 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Evapotranspiration from subsurface horizontal flow wetlands planted with Phragmites australis in sub-tropical Australia T.R. Headley a,b,c,*, L. Davison a, D.O. Huett d, R. Mu¨ller b a
School of Environmental Science and Management, Southern Cross University, PO Box 157, Lismore, NSW 2480, Australia Helmholtz Centre for Environmental Research (UFZ), Permoserstrasse 15, 04318 Leipzig, Germany c BAUER Nimr LLC, PO Box 1186, Al Mina 114, Muscat, Oman d Centre for Tropical Horticulture, NSW DPI, PO Box 72, Alstonville, NSW 2477, Australia b
article info
abstract
Article history:
The balance between evapotranspiration (ET ) loss and rainfall ingress in treatment wetlands
Received 12 April 2011
(TWs) can affect their suitability for certain applications. The aim of this paper was to inves-
Received in revised form
tigate the water balance and seasonal dynamics in ET of subsurface horizontal flow (HF) TWs in
20 September 2011
a sub-tropical climate. Monthly water balances were compiled for four pilot-scale HF TWs
Accepted 19 October 2011
receiving horticultural runoff over a two year period (Sep. 1999eAug. 2001) on the sub-tropical
Available online 6 November 2011
east-coast of Australia. The mean annual wetland ET rate increased from 7.0 mm/day in the first year to 10.6 mm/day in the second, in response to the development of the reed (Phragmites
Keywords:
australis) population. Consequently, the annual crop coefficients (ratio of wetland ET to pan
Constructed treatment wetland
evaporation) increased from 1.9 in the first year to 2.6 in the second. The mean monthly ET rates
Crop coefficient
were generally greater and more variable than the Class-A pan evaporation rates, indicating
Pan evaporation
that transpiration is an important contributor to ET in HF TWs. Evapotranspiration rates were
Reed bed
generally highest in the summer and autumn months, and corresponded with the times of peak
Water balance
standing biomass of P. australis. It is likely that ET from the relatively small 1 m wide by 4 m long
Water use efficiency
HF TWs was enhanced by advection through so-called “clothesline” and “oasis” effects, which contributed to the high crop coefficients. For the second year, when the reed population was well established, the annual net loss to the atmosphere (taking into account rainfall inputs) accounted for 6.1e9.6 % of the influent hydraulic load, which is considered negligible. However, the net loss is likely to be higher in arid regions with lower rainfall. The Water Use Efficiency (WUE ) of the wetlands in the second year of operation was 1.3 g of above-ground biomass produced per kilogram of water consumed, which is low compared to agricultural crops. It is proposed that system level WUE provides a useful metric for selecting wetland plant species and TW design alternatives to use in arid regions where excessive water loss from constructed wetlands can be problematic. Further research is needed to accrue long-term HF TW water balance data especially in arid climatic zones. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Treatment wetlands (TW) are increasingly being used for a range of water quality improvement applications.
Treatment wetlands with subsurface flow in a horizontal direction (defined as Horizontal Flow TWs by Fonder and Headley (2010)) are particularly well suited for the decentralized treatment of domestic wastewater because of their
* Corresponding author. Wetlands Competence Centre, BAUER Nimr LCC, PO Box 1186, Postal Code 114, Al Mina, Oman. Tel: þ968 24699505; fax: þ968 24699636. E-mail addresses:
[email protected] (T.R. Headley),
[email protected] (L. Davison), kdhuett@ optusnet.com.au (D.O. Huett),
[email protected] (R. Mu¨ller). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.042
346
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 4 5 e3 5 4
relatively low operation and maintenance requirements, robust treatment performance and low risk of human contact with effluent (Davison et al., 2001, 2005; Wallace and Knight, 2006). Despite the many inherent advantages of constructed wetlands for decentralized wastewater treatment, one potential disadvantage is the fact that they contain wetland plants which have inherently low water use efficiencies since they are adapted to growing in situations where soil moisture is not limiting. Thus, water losses to the atmosphere via evapotranspiration (ET ) can be high (Borin et al., 2011), especially under warm and windy conditions. In moist or environmentally sensitive situations where sustainable “disposal” of the effluent is the main objective, the transfer of water to the atmosphere can be desirable. For example, zero-discharge wetland systems using willows to evapotranspire all of the effluent have found a niche in sensitive locations of Denmark where stringent effluent nutrient standards can make treatment prohibitively expensive (Gregersen and Brix, 2001). However, in arid water scarce regions where potential ET rates are high and treated effluent is considered a valuable resource to be reused, the loss of significant amounts of water via ET can be undesirable (El Hamouri et al., 2007; Green et al., 2006; Masi and Martinuzzi, 2007). Excessive ET losses can lead to an increase in the effluent salt concentration (Morari and Giardini, 2009) and increase the risk of long term soil salinization problems in irrigation reuse areas. For these reasons, constructed wetlands are often declared to be inappropriate for arid climates, despite a scarcity of published research on ET rates from such systems. Hydrological processes play a key role in determining the type of biochemical conditions and biota that occur in wetlands (Mitsch and Gosselink, 2007). By definition, wetlands are transitional environments between well-drained terrestrial and deepwater aquatic ecosystems, and therefore rely heavily on particular hydraulic characteristics for their very existence. The flow through a wetland system can be heavily modified by the atmospheric fluxes of precipitation and ET, which can have significant affects on water quality. The loss of water through ET slows flow velocities, increases contact times, and causes an increase in pollutant concentration. Rainfall has the opposite effect, resulting in increased flow velocities, reduced contact times and dilution of pollutants (Kadlec, 1989). In a natural wetland the hydrology is influenced by the inputs of stream-flow, runoff, groundwater and precipitation, and the outputs of stream-flow, groundwater and evapotranspiration (Kadlec and Knight, 1996; Mitsch and Gosselink, 2007). Typically, constructed TWs, such as horizontal flow (HF) wetlands, have a hydrology which is modified from that of natural wetlands by the intentional exclusion of groundwater fluxes and surface runoff, and the substitution of stream-flow with wastewater influent and effluent. Thus, the HF TW water balance, for a given time period, can be represented by equation (1): In þ P ¼ Out þ ET
(1)
where In ¼ influent, P ¼ precipitation, Out ¼ effluent, ET ¼ evapotranspiration While influent and effluent volumes can be measured with flow metering devices, and precipitation with a rain gauge, the direct measurement of ET presents greater difficulties. Thus,
ET from TWs is typically calculated by measuring three out of the four components of the simplified water balance where, for a given period: ET¼ InþP Out
(2)
where, ET ¼ evapotranspiration loss from the wetland (mm), In ¼ wetland influent (mm), Out ¼ wetland effluent (mm), P ¼ precipitation (mm) Evapotranspiration is a complex and multi-faceted process. It includes the atmospheric losses from a wetland that occur as a result of direct evaporation from water and soil, and the moisture that passes through vascular plants to the atmosphere via transpiration (Kadlec, 1987; Kadlec and Wallace, 2009; Mitsch and Gosselink, 2007). Wetzel (1975) states that atmospheric losses from wetlands are greatly modified by the transpiration of emergent macrophytes, and in a majority of situations the transport of water from a water body to the air is greatly increased by a dense stand of actively growing littoral vegetation, when compared to the evaporation rates from open water. The effect of vegetation can be particularly pronounced in subsurface flow TWs. El Hamouri et al. (2007) reported average ET rates of 40 and 57 mm/day respectively for Arundo donax and Phragmites australis planted subsurface horizontal flow TWs (26 m2) in Morocco over the winter e spring period, compared with 7 mm/day for an unplanted gravel bed. The process of transpiration results in water passing from the substrate to the air surrounding the leaves of the plant along a gradient of water potential, along which several resistances are encountered (Stewart, 1984). In a wetland, where the supply of water is rarely limited, the rate of ET is proportional to the difference between the vapor pressure at the water or leaf surface and the vapor pressure in the overlying air (Mitsch and Gosselink, 2007). Thus, ET is affected by factors that modify these vapor pressures, such as solar radiation, surface temperature, humidity and wind speed. It is often desirable to be able to estimate the rate of ET from a wetland, either for modeling the hydrology before a wetland is built or in situations where measurement of the wetland outflow is not possible. Several methods exist for estimating wetland ET loss, ranging from simple empirical equations to more complex modeling approaches requiring the input of multiple meteorological data from the site under investigation. One of the more complex approaches that has been used to model wetland ET involves modeling the wetland energy balance, as described in detail by Kadlec (2006) and Kadlec and Knight (1996). Providing adequate data is available for the site, the energy balance approach enables ET to be predicted with a relatively high resolution. For example, Kadlec (2006) used an energy balance approach to predict ET rates from a group of surface flow wetlands ranging in size from 0.12 to 1.3 ha in the arid climate of Arizona, USA and was able to demonstrate that wetlands with a high loading rate (short retention time) will have a higher ET rate because of the excess sensible heat of the incoming water. Kadlec and Wallace (2009) point out that the energy balance is more difficult to parameterize for subsurface flow wetlands because of the more complicated pathway for the transfer of sensible heat and water vapor to and from the water surface through the overlying dry gravel and vegetation layers. The energy balance is further complicated by the potentially large heat
347
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 4 5 e3 5 4
storage capacity of the gravel-water system in subsurface flow wetlands. Wetland specific energy balance models are generally computationally complex and require large amounts of site specific meteorological and related data, precluding their use in many situations. For these reasons, an increasingly popular approach is to take a local estimate of the potential ET or potential evaporation (E ) and multiply it by a relevant crop coefficient (Kc) based on the approach that is used for scheduling irrigation in agriculture (Kadlec and Wallace, 2009). Historically, Class-A pan evaporation (Epan) has been the most widely available measure of E. However, reference crop ET (ETo) data based on energy balance calculations such as the PenmaneMonteith equation (Monteith, 1965) have become more widely available since the advent of affordable automatic weather stations. Either way, wetland water balance data is needed in order to calibrate Kc for the local conditions. Borin et al. (2011) monitored the water balance of pilot-scale subsurface HF TWs planted with P. australis at two locations in Italy and found that wetland ET was 6e7 times higher than ETo when averaged over the growing season. When considering wetland ET, scale needs to be taken into consideration, as very small wetlands will react strongly to the surrounding microclimate, with enhanced ET rates due to a strong influence of advection and the so-called “clothesline” and “oasis” effects (Borin et al., 2011; Kadlec, 1989; Kadlec and Wallace, 2009). While the crop coefficient approach is simple to use, coefficients will vary somewhat depending on the site, season, vegetation and wetland characteristics. There is currently a lack of adequate water balance data from HF TWs to be able to derive and calibrate wetland crop coefficients for different climates. In order to better understand the magnitude of evapotranspirative water losses from constructed wetlands used for decentralized wastewater treatment, the aim of this study was to compile water balances in four identical pilot-scale subsurface horizontal flow TWs in sub-tropical Australia over a two year period and quantify the evapotranspiration dynamics with regards to weather conditions, season and vegetation growth.
2.
Materials and methods
2.1.
Site and study description
Water balances were compiled for four identical HF TWs located at the New South Wales Centre for Tropical Horticulture in Alstonville on the Australian east coast (28.85 S, 153.46 E). Alstonville lies at an elevation of 140 m above sea level and experiences a sub-tropical climate, with average annual rainfall, air temperature and relative humidity of 1822 mm, 19.3 C and 73% respectively (Table 1). These water balance investigations formed part of a broader study of the capabilities of the wetlands to treat nursery runoff water. Detailed descriptions of the experimental set-up and design of the HF TWs are given in Headley et al. (2001) and Headley et al. (2003) and summarized briefly here. Each HF TW was 4 m long by 1 m wide with a 0.5 m water depth, and contained a 10 mm diameter basaltic gravel substrate into which P. australis was planted in April 1999. The outlet height of each wetland was set to maintain the water level approximately 50 mm below the upper surface of the gravel. The HF TWs were contained in sealed fiber-glass troughs, therefore eliminating any interactions with groundwater or surrounding soil moisture. The HF TWs were intermittently loaded six times per day with a nutrient solution made to replicate horticultural runoff (TSS < 10 mg/L; TOC < 10 mg/L, TN 10.5 mg/L, NO3eN 10.0 mg/ L, NH4eN 0.5 mg/L, TP 0.5 mg/L, PO4eP 0.5 mg/L plus micronutrients and trace elements). The four wetlands were grouped into two pairs which were operated at different hydraulic loading rates (HLRs) - and hence different hydraulic residence times (HRTs) - in order to assess the effect of loading rate on treatment performance as part of a parallel study. Both pairs of TWs were operated at the same loading rate (approximately 40e50 mm/day, equivalent to a nominal HRT of 5e6 days) for an initial 11 month equilibration period after planting (April 1999eMarch 2000). One pair of beds (“Pair A”) was then maintained at a relatively constant loading rate (circa 50 mm/day) throughout most of the study, while the loading rate on the other pair of beds (“Pair B”) was
Table 1 e Mean climate statistics for the New South Wales Centre for Tropical Horticulture in Alstonville, Australia (source: Bureau of Meteorology, 2011). In general, data from years up to and including 2010 have been used. Parameter
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Mean maximum temperature ( C) Mean minimum temperature ( C) Mean temperature ( C) Mean daily sunshine (hours) Mean 9am relative humidity (%) Mean 9am wind speed (km/h) Mean rainfall (mm) Mean daily A-pan evaporation (mm)
27.2
26.7
25.9
23.9
21.3
19
18.6
20
22.4
24.2
25.4
26.9
23.5
37
19.5
19.4
18.3
15.9
13.3
10.9
9.9
10.6
12.8
14.8
16.6
18.5
15
37
23.4 7.5
23.1 6.9
22.1 6.8
19.9 6.7
17.3 6.1
15.0 6.1
14.3 6.9
15.3 7.7
17.6 8.1
19.5 7.9
21.0 7.6
22.7 7.6
19.3 7.2
37 28
79
82
81
78
76
73
68
65
64
67
73
75
73
35
10 178 5.7
8.9 228 5
Annual
Years of record
8.8
8.9
10.2
10.8
10.7
10.9
11.2
11
10.4
9.7
10.1
34
262 4.3
194 3.5
184 2.7
163 2.4
94 2.7
75 3.5
56 4.4
109 5
133 5.4
153 5.9
1822 4.2
48 34
348
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 4 5 e3 5 4
periodically increased from approximately 50e155 mm/day (nominal HRTs of 2e5 days) from April 2000 through to March 2001. For the final period of the study (MarcheAugust 2001), both pairs of beds were operated at similar loading rates of approximately 100 mm/day, as the prior monitoring in the simultaneous study indicated that this was the appropriate loading rate to achieve the desired level of nutrient reductions. Because the water level in the wetlands remained constant and the influent water was typically at ambient temperature throughout the study, there is no reason to expect that the hydraulic loading rate will have an influence over the absolute rate of water loss from the TWs. Thus, the wetland ET data in this paper is generally presented as the average for all four TWs. However, the percentage of the wetland inflow that is lost to ET does vary depending on the hydraulic loading rate and is therefore presented separately for the two pairs of TWs.
2.2.
Water balance monitoring
Water balance data was collected during a two year period from September 1999 through to August 2001. Equation (2) was used to determine the ET rate. The weekly volumes of influent and effluent for each HF TW were recorded using mechanical flow meters and calibrated collection tanks respectively. Data from weeks containing a 24 h period where greater than 90 mm of rain fell were omitted from the ET calculations, because such heavy rainfall resulted in nonsensical data. This only occurred in three out of the 104 weeks of the study. Precipitation, Class A-pan evaporation and maximum air temperature ( C) were recorded daily from an adjacent weather station. Water balance data is generally presented here as the mean of the four TWs. However, where appropriate, the means for the two pairs of TWs (“A” and “B”) are expressed separately, particularly when discussing ET as a percentage of the influent hydraulic load. Wetland crop coefficients were determined for each HF TW using Equation (3), where for a given period, the crop coefficient is the ratio of ET lost from the wetland as a proportion of Class-A pan evaporation: Kpan ¼ ETw =Epan
(3)
where Kpan ¼ crop coefficient relative to Class-A pan evaporation, ETw ¼ wetland evapotranspiration (mm), Epan ¼ class-A pan evaporation (mm) In a number of other studies, a reference crop evapotranspiration (ETo), calculated using a modified PenmaneMonteith method, has been used in place of the pan evaporation to determine crop coefficients (KETo) (eg. Fermor et al., 2001). To enable comparison with such studies, an estimation of ETo has been made from the measured pan evaporation data using the conversion factors sited in Grayson et al. (1996) for Brisbane, located in a similar climatic zone 160 km north of the study site. For summer, autumn, winter and spring, the factors used to convert pan evaporation to ETo were 0.78, 0.75, 0.66 and 0.74 respectively. These conversion factors are based on the extensive regression analyses of over 14 years of meteorological data reported by Chiew et al. (1995), and have been found
to provide a reasonable prediction of ETo over time scales of 5 days or greater (Grayson et al., 1996). The mean monthly wetland atmospheric water flux was calculated as the difference between ET losses from the HF TWs and incident precipitation for a given month. A negative atmospheric flux equates to a net loss of moisture through ET during that month, while a positive flux means that precipitation exceeded ET.
2.3.
Plant biomass measurements
The above-ground biomass of P. australis in the four wetlands was monitored periodically throughout the study, as described in Headley et al. (2003). In summary, at various points throughout the study the above-ground biomass was estimated using a non-destructive method, where a regression equation was developed relating the shoot dry weight to shoot height. The number of shoots within sequential height increments (every 20 cm) was then counted in quadrats at the inlet, middle and outlet sections of each wetland. Complete harvests of the above-ground biomass were also conducted during the winter of 2000 and 2001 (June/July). For the determination plant dry weight, all samples were dried in an oven at 70 C for 48 h before being weighed.
2.4.
Wetland water use efficiency
Water Use Efficiency (WUE) is defined as the amount of plant matter produced per unit of water consumed (Mengel and Kirkby, 2001) and provides a useful metric for comparing the relative water consumption of different wetland plant species. The annual WUE was determined for each wetland for the second year of operation by dividing the dry weight of aboveground biomass produced (g/m2) by the total volume of water lost via ET (L/m2) over the period of measurement, resulting in a wetland WUE with the units of grams of above-ground biomass produced per kilogram of water consumed (g biomass/kg water lost), assuming a water density of 998.2 g/L at 20 C. This is what is considered an “ecosystem level” measure of WUE, as opposed to the intrinsic or instantaneous WUE which is measured at the leaf level as the ratio of photosynthetic CO2 assimilation to water lost through the stomata via transpiration (Bacon, 2004).
3.
Results and discussion
3.1.
Wetland evapotranspiration and crop coefficients
3.1.1.
Annual water balances
The ET rates of the four individual wetlands were generally very similar to each other for each month of the study. Hence, mean ET rates for the four wetlands are presented unless otherwise stated. On an annual basis, the wetlands lost 2551 mm and 3874 mm of water to evapotranspiration (ETw) in the first and second years of operation respectively (Table 2). In contrast, the annual Class-A Pan evaporation (Epan) did not vary substantially from the first (1357 mm) to the second (1512 mm) year. Likewise, the estimated annual reference ET (ETo) was relatively stable between the two years.
349
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 4 5 e3 5 4
Consequently, the annual mean crop factors relative to Epan (Kpan) and ETo (KETo) respectively increased from 1.9 and 2.5 in year one to 2.6 and 3.4 in year two. The increase in ETw and Kpan in the second year reflects the fact that the plant above-ground biomass almost doubled in the second year of growth (Fig. 1), therefore increasing the transpiration capacity. A similar trend was also observed by Borin et al. (2011) for similar sized HF TWs planted with P. australis in Italy. The total annual precipitation was 1552 mm and 1708 mm in the first and second years. As a reflection of the moist subtropical climate, precipitation exceeded Epan and ETo on an annual basis in both years. However, only 61% and 44% of ETw was replaced by rainfall in the first and second years respectively. Thus, even under a sub-tropical climate with relatively high rainfall, wetland evapotranspiration can result in a significant net loss of water.
3.1.2.
Monthly water balances
On a monthly basis, mean ETw rates ranged from 3.2 mm/day to 15.1 mm/day throughout the study (Table 2, Fig. 2). The lowest ETw rate occurred during the first month of
measurement (September 1999), when the wetlands were only four months old and the macrophytes were still in an establishment phase. After this time, the mean monthly ETw rate remained above 5.5 mm/day. Evapotranspiration rates were generally highest during the summer and autumn months (DeceMay), corresponding with the times of peak standing biomass of P. australis during the second half of the growing season (Fig. 1). This trend was also observed by Fermor et al. (2001) for a pair of larger HF TWs (940 m2 total surface area) at Himley in England. Moro et al. (2004) also observed that transpiration of P. australis in a natural marsh in southeastern Spain was highest in summer and corresponded with the peak in plant growth and maximum leaf area. The mean monthly Epan rate ranged from 2.2 to 6.4 mm/day throughout the study. In all but the first month of measurement (Sep. 1999), ETw was greater than the Epan. Consequently, Kpan, which ranged from 0.9 to 4.5, was greater than one in all but the first month of measurements (Fig. 3). The highest mean monthly Kpan for the wetlands (4.5) occurred in April and May of 2001 during the late autumn of the second year of plant growth. This was due to a combination of relatively high ETw rates
Table 2 e Mean daily water balance data (per month and per year) for the four HF TWs. Units for ETw, Epan, precipitation (P) and the net loss or gain are mean mm/day, except for the annual totals, which are in mm. Crop coefficients (K ) are dimensionless. A negative net flux equals a net loss of water, while a positive net flux means that precipitation exceeded ET. HLRs are given separately for the wetland pairs “A” and “B” because they received different loading rates during parts of the study. Month
Year
Sept Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug
1999
2000
Annual mean Annual total (mm) Sept Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Annual mean Annual total (mm)
HLR pair A mm/d
2001
ETw mm/d
P mm/d
Net flux mm/d
Epan mm/d
Kpan
ETo mm/d
KETo
39.1 39.2 43.3 35.6 43.9 46.6 53.6 48.8 48.3 52.7 46.6 49.8
37.1 38.3 43.6 34.7 42.3 46.9 53.3 58.2 60.7 54.3 60.7 68.1
3.2 5.9 9.7 6.6 6.3 9.1 11.1 6.5 7.6 6.2 5.9 5.5
1.4 5.0 4.7 4.4 10.1 2.8 3.4 6.1 4.2 7.0 0.9 0.9
1.8 0.9 5.0 2.2 3.8 6.3 7.7 0.4 3.4 0.8 5.1 4.6
3.4 4.3 4.8 5.2 4.9 5.0 3.6 3.1 2.5 2.3 2.6 3.3
0.9 1.4 2.0 1.3 1.3 1.8 3.1 2.1 3.0 2.8 2.3 1.7
2.5 3.2 3.5 4.1 3.8 3.9 2.7 2.3 1.9 1.5 1.7 2.2
1.3 1.9 2.7 1.6 1.7 2.3 4.1 2.8 4.0 4.2 3.5 2.6
45.6
49.9
7.0
4.3
2.7
3.7
1.9
2.8
2.5
18560
2000
HLR pair B mm/d
20528
2551
1552
999
1357
n/a
1016
n/a
48.4 47.8 36.0 53.0 49.6 51.0 53.1 49.9 95.8 74.7 96.3 98.7
65.5 91.5 82.1 94.2 97.6 117.1 156.8 93.0 96.6 77.9 94.3 95.0
10.3 8.9 6.4 15.1 9.1 14.4 13.9 13.9 13.1 6.6 9.0 6.8
0.1 2.7 3.1 3.7 1.7 17.5 13.5 4.0 5.9 1.2 2.5 1.3
10.2 6.2 3.3 11.4 7.5 3.1 0.5 9.9 7.2 5.4 6.5 5.5
5.3 5.0 4.4 6.4 6.3 5.1 4.4 3.1 2.9 2.2 2.4 2.9
1.9 1.8 1.4 2.4 1.4 2.8 3.1 4.5 4.5 3.0 3.8 2.3
3.9 3.7 3.3 5.0 4.9 4.0 3.3 2.3 2.2 1.5 1.6 1.9
2.6 2.4 1.9 3.0 1.9 3.6 4.2 6.0 6.0 4.5 5.7 3.5
62.9 22969
96.8 34205
10.6 3874
4.7 1708
5.9 2166
4.1 1512
2.6 n/a
3.1 1145
3.4 n/a
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 4 5 e3 5 4
2
5000 4000 3000 2000
Aug 01
Apr 01
Jun 01
Feb 01
Dec 00
Oct 00
Jun 00
Aug 00
Apr 00
Date
Fig. 1 e Mean standing stock of above-ground biomass (g/ m2) for the four HF TWs throughout the study. The reeds were planted in April 1999 and shoots harvested in July of 2000 and 2001. Error bars represent D/L one standard error of the mean.
(13.1e13.9 mm/day) and low Epan rates (2.9e3.1 mm/day) at that time and indicates that wetland plant transpiration can make a strong contribution to ETw, even when the prevailing conditions are not favorable for evaporative processes. The estimated mean monthly reference crop ETo ranged from 1.5 mm/ day (June of 2000 and 2001) to 5.0 mm/day (December 2000) throughout the study. Mean KETo values ranged from 1.3 in the first month of the study (September 1999) up to 6.0 in April and May of 2001 and followed a similar trend to Kpan. Borin et al. (2011) measured similarly high KETo values for HF TWs of similar size to the current study and planted with P. australis in two locations in Italy. They reported average KETo values for the growing season at the two sites of 6.8 and 6.3, while the evaporation from unplanted controls was generally similar to ETo. In general, ETw displayed more variation from month to month than Epan (Fig. 2). Regression analysis indicated that the relationship between mean monthly ETw and Epan was not significant (r2 ¼ 0.12, 22 d.f., p ¼ 0.09), as shown in Fig. 4. Consequently, the monthly crop coefficients (Kpan) also varied substantially over time (Fig. 3). Peacock and Hess (2004)
Epan
ETw
Precipitation
Mean Air Temp.
25 5.0
o
16
Mean Daily Air Temperature ( C)
Mean ETw , Epan and Precipitation (mm/day)
18
20
14 12
15
10 8
10
6 4
5
2 Sep-99 Oct-99 Nov-99 Dec-99 Jan-00 Feb-00 Mar-00 Apr-00 May-00 Jun-00 Jul-00 Aug-00 Sep-00 Oct-00 Nov-00 Dec-00 Jan-01 Feb-01 Mar-01 Apr-01 May-01 Jun-01 Jul-01 Aug-01
0
0
Month-year
Fig. 2 e Mean monthly wetland evapotranspiration (ETw), pan evaporation (Epan), precipitation and air temperature for the study site. Error bars represent D/L one standard error of the mean.
4.0 3.0 2.0 1.0 0.0 Sep-99 Oct-99 Nov-99 Dec-99 Jan-00 Feb-00 Mar-00 Apr-00 May-00 Jun-00 Jul-00 Aug-00 Sep-00 Oct-00 Nov-00 Dec-00 Jan-01 Feb-01 Mar-01 Apr-01 May-01 Jun-01 Jul-01 Aug-01
Feb 00
Dec 99
Oct 99
Aug 99
Apr 99
0
Jun 99
1000
measured ET at a large P. australis marsh in southern England over a three month period and reported that crop coefficients (ET against ETo) were inconsistent from day to day. Herbst and Kappen (1999) reported that the ratio of transpiration to evaporation in a belt of P. australis on the edge of a northern German lake varied significantly depending on weather conditions. This was because reed transpiration responded more sensitively than evaporation to meteorological variables. In the current study, mean monthly Epan was found to correlate strongly with the mean air temperature (r2 ¼ 0.67, 22 d.f., p ¼ 0.000001) resulting in a clear seasonal trend. However, the relationship between ETw and air temperature was weak (r2 ¼ 0.26, 22 d.f., p ¼ 0.01). This seems to contradict the findings of Herbst and Kappen (1999). However, it is conceivable that plant transpiration responded more strongly to meteorological factors other than air temperature, such as duration of sunshine, incident radiation, relative humidity and wind. Moro et al. (2004) reported a significant correlation between P. australis transpiration (based on sap flow measurements) and air temperature on clear days, but not on cloudy or rainy days, indicating the complexity of wetland transpiration response to weather conditions. In this regard, regulation of the stomatal openings by the wetland plants in response to micro-climatic variation may have contributed to the observed variability in ETw, even though they were growing with a constant water supply (Pauliukonis and Schneider, 2001; Peacock and Hess, 2004). The lack of temperature effect on ETw and the variability in Kpan may also partly be because ETw can be strongly influenced by other factors which affect plant transpiration (but not Epan), such as plant vigor and the amount of above-ground biomass and leaf area (Kadlec and Wallace, 2009; Kirkham, 2005; Moro et al., 2004; Zhou and Zhou, 2009). The ETw rates and crop coefficients recorded throughout this study were higher than those measured for P. australis in some other studies. For example, Fermor et al. (2001) reported ET rates and crop coefficients (relative to ETo) for the previously mentioned HF TWs at Himely in England of 0.18e6.30 mm/day and 0.50 to 3.15 respectively. By comparison, the lowest ET rate measured during the second year in the current study was 6.4 mm/day (Nov. 2000), while
MeanK pan
Mean above-ground biomass (g/m )
350
Month
Fig. 3 e Mean monthly crop coefficients relative to Class-A pan evaporation (Kpan) for the four HF TWs. The dashed line represents a crop coefficient of 1.0, where ETw [ Epan. Error bars represent D/L one standard error of the mean.
351
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 4 5 e3 5 4
Mean monthly ETw (mm/day)
16
mesocosms planted with different ligneous species in Belgium. This enhanced ET rate will be characteristic of many HF TW systems, as they tend to be relatively small (Kadlec and Wallace, 2009), particularly when used for the decentralized treatment of sewage from single dwellings or small communities.
14 12 10 8 6
3.2.
4 2
y = 0.9099x + 5.1776 2 R = 0.1249
2
3
4
5
6
7
Mean monthly Epan (mm/day)
Fig. 4 e Linear regression of mean monthly ETw against Epan for the four HF TWs.
the KETo ranged from 1.3 to 6.0 throughout (Table 2). The higher rates of ET observed during the current study would be partly related to climatic differences and scale effects. Temperature has an important influence over both evaporation and transpiration processes (Dingman, 1994), as represented by the dependence of many ET models on temperature (for example: the FAO-24 radiation method (FAO, 1984); the Penmanemonteith method (Monteith, 1965); the Hammer and Kadlec/Scheffe method (Mitsch and Gosselink, 2007); and the temperature based method of Thornthwaite (Thornthwaite, 1948)). The monthly mean air temperatures ranged from 14.3 C to 23.4 C throughout the current study. Although Fermor et al. (2001) did not report temperatures for their study at Himely in England, the mean maximum air temperature from 1971 to 2000 for the Midlands District was 9.2 C, with a monthly range of 3.6 Ce16.0 C (UK Met Office, 2003). The ETw rate of the HF TWs in the current study would have been enhanced by what Kadlec and Wallace (2009) describe as the “clothesline” and “oasis” effects which cause small wetlands to respond strongly to the surrounding microclimate, often leading to enhanced ET losses. This is because of advection in which the relatively warm dry air from the terrestrial surroundings can contribute heat and drive water loss in excess of that driven by radiation alone. The rate of ET from the 4 m long by 1 m wide HF TWs in the current study would also have been enhanced by the increased exposure to wind movement and interception of sunlight facilitated by the erova´ et al. (2001) reported similarly high large edge effect. Kuc ET rates and crop coefficients from small 1 m2 wetland
Wetland water use efficiency
The water use efficiency (WUE ) of the P. australis wetlands was calculated over four periods of different duration during the second year of operation (Table 3). Each period started in July 2000 when the above-ground biomass of the reeds had been cut and extended for different durations depending on when the standing stock of above-ground biomass was estimated. It can be seen that the wetland WUE changed depending on which duration of plant growth was used. The highest mean WUE (3.0 g biomass/kg water) was measured for the period ending in January 2001 (Summer) which also corresponded with the period of maximum above-ground biomass growth rate and peak standing stock (Fig. 1). The estimated WUE subsequently decreased as longer periods of measurement were used. This is a reflection of the fact that the standing stock of P. australis actually decreased slightly after January 2001, as the growth rate of the reeds declined and older leaves may have been shed from the plants as a result of damage from wind or animals. High rates of ETw (13.1e14.4 mm/day) were maintained throughout FebeMay 2001, despite the lack of plant growth, causing the cumulative WUE to decline over time. Based on the amount of above-ground biomass that was measured at the end of the 12 month period in winter 2001, the mean annual WUE was only 1.3 g/kg. This can be considered to be a slight under estimate of the mean annual WUE, because the standing stock of shoots measured at the end of the growth season under estimates the total amount of shoot biomass produced throughout the year (Santos and Esteves, 2004). This raises a question about the appropriate timing for measurement of plant biomass in order to quantify wetland WUE in such a way. The mean annual WUE of 1.3 g/kg measured for the P. australis HF TWs is relatively low when compared to other plants. Most of the published data on whole plant or ecosystem level WUE relates to plants of agricultural importance. One of the most relevant crops for comparison is rice (Oryza sativa L.), which is also a member of the Poaceae family which naturally occurs in wetland conditions and is often grown in water logged soils. Haefele et al. (2009) report mean
Table 3 e Mean water use efficiency of the four wetlands during the second year of operation. Data points show the end date of the periods of WUE measurement, corresponding to times when the standing stock of above-ground biomass was measured. All periods began on the 15th July 2000 when the above-ground biomass was harvested. Period start 15th 15th 15th 15th
Jul. Jul. Jul. Jul.
2000 2000 2000 2000
Period end
Duration (days)
Mean shoot standing stock (g/m2)
Cumulative ET loss (kg/m2)
Mean WUE (g biomass/kg water)
10th Sep. 2000 25th Jan. 2001 21st Mar. 2001 8th Jun. 2001
57 194 249 358
952 5061 4987 4781
348 1706 2460 3562
2.7 3.0 2.0 1.3
352
15 10 5
Sep-99 Oct-99 Nov-99 Dec-99 Jan-00 Feb-00 Mar-00 Apr-00 May-00 Jun-00 Jul-00 Aug-00 Sep-00 Oct-00 Nov-00 Dec-00 Jan-01 Feb-01 Mar-01 Apr-01 May-01 Jun-01 Jul-01 Aug-01
0
Month
Fig. 5 e Mean monthly percentage of the influent hydraulic load that was lost via ET from wetland Pairs A and B.
hydraulic load of 1e12%, depending on the magnitude of the influent HLR. However, there was strong evidence of a maturation effect, as the high rate of ET from the two year old beds during March 2001 (13.9 mm/day) was great enough to hydraulically neutralize the 418 mm of rain that fell during that month. Furthermore, the only net atmospheric influx in the second year occurred during February 2001, when the study site received an uncharacteristically high 490 mm of rainfall. An examination of climatic records for the study site, from 1963 to 2003, indicates that the mean precipitation for February is 234.3 mm, while the 90th percentile is 490.2 mm (Bureau of Meteorology, 2003). Thus, over the 40 years of records, there have only been three other years that have experienced as high or higher rainfall in February. Thus, it seems probable that a mature HF TW under such sub-tropical climatic conditions would typically experience a net loss of moisture to the atmosphere in all months of an average rainfall year. The net loss of moisture to the atmosphere on an annual basis has important implications for effluent management systems. For example, in effluent reuse schemes in hot arid climate regions where water conservation is the goal, the net
15
Wetland Pair A
10
Wetland Pair B
5 0 -5 -10 -15 -20 Aug-01
Jun-01 Jul-01
Apr-01 May-01
Mar-01
Feb-01
Oct-00
Nov-00 Dec-00 Jan-01
Sep-00
Jul-00 Aug-00
May-00 Jun-00
Mar-00 Apr-00
Dec-99 Jan-00 Feb-00
-25 Oct-99
On an average annual basis ETw exceeded precipitation. Consequently, the HF TWs experienced total net moisture losses of 999 and 2166 mm/year in the first and second years of measurement respectively (Table 2). This equated to an average net water loss of 2.7 and 5.9 mm/day for the wetlands during years one and two respectively. The substantial increase in the second year is due to the elevated ETw rates in response to the maturation of the macrophytes. The water loss rates observed in the second year are therefore more representative of mature phase operation. During this second year, the net water loss represented only 9.4% and 6.1% of the mean influent hydraulic loading rate received by wetland Pair A (62.9 mm/day) and wetland Pair B (96.8 mm/day) respectively. In most situations, annual water losses of such an order can be considered negligible. However, in arid regions with low rainfall, the net loss of water will form a greater proportion of the inflow. On a monthly basis, ETw accounted for 5e27% of the influent HLR received by the HF TWs (Fig. 5, Table 2). These values tended to be higher for wetland Pair A from June 2000 through to June 2001 due to the lower HLR incurred by these beds over this period when compared to wetland Pair B. The net atmospheric fluxes (taking into account the incident precipitation), as a proportion of the HLR, ranged from an increase of 11.9% (wetland Pair A in Feb. 2001) to a loss of 23.3% (wetland Pair A in Sep. 2000), highlighting the impact that rainfall can have in counteracting the effect of ET in a sub-tropical climate (Fig. 6; Table 2). Evapotranspiration losses from the HF TWs generally exceeded the incident rainfall in all months, except where the precipitation for a given month was greater than 210 mm. These high rainfall months resulted in a net increase in the
Wetland Pair B
20
Nov-99
Significance of wetland ET losses
Wetland Pair A
25
Sep-99
3.3.
30
Net atmospheric flux as % of influent hydraulic load
WUEs for rice (O. sativa L.) of between 1.8 and 4.7 g dry matter per liter of water transpired, which is slightly higher than the range measured for P. australis in the current study. As might be expected, the WUE of terrestrial crops is generally higher. For example, the WUE of barley and wheat (also in the Poaceae family) range from 3.2e5.7 and 3.1e9.2 respectively (Kemanian et al., 2005). The relatively low WUE of P. australis would be in part due physiological differences (lack of adaptation for conserving water) and the fact that water was continuously available (ET was not limited by water availability). To date, ecosystem level measures of WUE have not been reported for wetland plants, presumably because WUE has traditionally been of interest to agronomical and irrigation disciplines. Most wetland plants are of limited commercial agricultural value. Furthermore, because wetlands do not typically occur in water limited environments, the efficiency by which wetland vegetation uses water has historically been of little interest. However, as constructed wetlands are increasingly being used for wastewater treatment purposes, often in water scarce regions, questions are likely to arise as to their WUE. Interest in the relative WUE of different wetland plant species, especially those occurring in arid regions, is also likely to increase in the coming years. In this regard, wetland WUE as calculated in the current study provides a useful metric for comparing different wetland plants.
ET as % of influent hydraulic load
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 4 5 e3 5 4
Month
Fig. 6 e Mean monthly Net Atmospheric Flux as a proportion of the influent hydraulic loading rate for wetland Pairs A and B. Note: a negative atmospheric flux equates to a net loss of moisture through ET during that month, while a positive flux means that precipitation exceeded ET.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 4 5 e3 5 4
loss of water through ET may discourage the use of constructed wetlands as a treatment option. However, the results of the current study indicate that the percentage of wetland influent that is lost to the atmosphere through ET is relatively low on an annual and monthly basis, even when the ET rate is high as was the case for the small HF TWs in this study. In this regard, an important factor to consider is the residence time (or HLR) of the HF TW. At lower HLRs, the significance of wetland ET losses will be higher. Typical HLRs used for HF TWs treating primary settled domestic wastewater to secondary quality are in the order of 25e60 mm/day (Davison et al., 2005; Vymazal and Kro¨pfelova´, 2008). In comparison, the mean HLR of 63 mm/day received by wetland Pair A in the second year of the current study was at the upper end of this range. Horizontal flow TWs are increasingly being used for the onsite treatment of domestic wastewater from individual households or small communities (Reed et al., 1995; Davison et al., 2001; Vymazal, 2002). In such systems, particularly on small or constrained lots, it is often desirable to minimize the amount of wastewater that needs to be disposed of via land application on-site. In highly constrained or environmentally sensitive sites, the aim may be to design a system that is capable of achieving zero-discharge of effluent (Gregersen and Brix, 2001). To achieve this, the HF TW would need to be designed so that the influent hydraulic loading rate is less than the net loss of moisture to the atmosphere through ET.
4.
Conclusions
The reuse of treated effluent for irrigation is becoming increasingly common place in water scarce regions. In such situations, the loss of water from constructed wetlands is undesirable and can discourage their use, despite the many inherent advantages of constructed wetlands especially for low-income communities. Although the HF TW water balances presented here have indicated that the annual net loss of water to the atmosphere relative to typical loading rates is negligible under sub-tropical conditions, there is currently a need for such investigations under arid climatic conditions where ET losses are likely to be higher and rainfall inputs lower. Measures such as mulching the top of the bed may provide a means for reducing ET losses from HF TWs. However, this is unlikely to impact significantly on plant transpiration, which has been shown here to form a major component of the water loss from HF TWs. Thus, the most fruitful avenues for minimizing water loss from constructed wetlands will be through careful selection of appropriate wetland plants based on metrics such as the WUE presented here, management practices such as regular harvesting of the plant shoots (unproven to date), and optimization of the wetland design with regard to treatment efficiency in order to minimize the wetland foot-print.
Acknowledgments Funding for the experimental set-up, data collection and write-up for this research was provided through the
353
Horticultural Stock and Nurseries Act of NSW, with a matching contribution from Horticulture Australia Ltd (Project number: NY98008). The writing up of this manuscript was partly sponsored through the IWRM Helmholtz Dead Sea Project (“Decentralised Wastewater Treatment and Reuse in Arid Regions: Optimization of constructed wetlands for wastewater treatment in salinity-prone areas”) funded by the BMBF (Project number: 02WM0846).
references
Bacon, M.A., 2004. Water use efficiency in plant biology. In: Bacon, M.A. (Ed.), Water Use Efficiency in Plant Biology. Blackwell Publishing, Oxford, pp. 1e26. Borin, M., Milani, M., Salvato, M., Toscano, A., 2011. Evaluation of Phragmites australis (Cav.) Trin. Evapotranspiration in northern and southern Italy. Ecological Engineering 37 (5), 721e728. Bureau of Meteorology, 2011. Summary Statistics Alstonville Tropical Fruit Research Stat. Bureau of Meteorology, Commonwealth of Australia (accessed 25.03.11.). http://www. bom.gov.au/climate/averages/tables/cw_058131_All.shtml. Bureau of Meteorology, 2003. Climatic Averages for Alstonville Tropical Fruit Research Stat., Commonwealth of Australia (accessed 26.11.03.). http://www.bom.gov.au/climate/ averages/tables/cw_058131.shtml. Chiew, F.H.S., Kamaladasa, N.N., Malano, H.M., McMahon, T.A., 1995. Penman-Monteith, FAO-24 reference crop evapotranspiration and class-a pan data in Australia. Agricultural Water Management 28, 9e21. Davison, L., Headley, T., Edmonds, M., 2001. On-site domestic wastewater treatment by reed bed in the moist subtropics. Water Science and Technology 44 (11e12), 353e360. Davison, L., Headley, T., Pratt, K., 2005. Aspects of design, structure, performance and operation of reed beds e eight years’ experience in northeastern New South Wales, Australia. Water Science and Technology 51 (10), 129e138. Dingman, S.L., 1994. Physical Hydrology. Macmillan College Publishing Co., New York. El Hamouri, B., Nazih, J., Lahjouj, J., 2007. Subsurface-horizontal flow constructed wetland for sewage treatment under Moroccan climate conditions. Desalination 215, 153e158. Fermor, P.M., Hedges, P.D., Gilbert, J.C., Gowing, D.J.G., 2001. Reedbed evapotranspiration rates in England. Hydrological Processes 15, 621e631. Fonder, N., Headley, T.R., 2010. Systematic classification, nomenclature and reporting for constructed treatment wetlands. In: Vymazal, J. (Ed.), Water and Nutrient Management in Natural and Constructed Wetlands. Springer, Dordrecht, The Netherlands, pp, pp. 191e219. Food and Agriculture Organisation, 1984. Crop Water Requirements. FAO. FAO Irrigation and Drainage Paper, 24. Grayson, R.G., Argent, R.M., Nathan, R.J., McMahon, T.A., Mein, R.G., 1996. Hydrological Recipes e Estimation Techniques in Australian Hydrology. CRC for Catchment Hydrology, Melbourne. Green, M., Shaul, N., Beliavski, M., Sabbah, I., Ghattas, B., Tarre, S., 2006. Minimizing land requirement and evaporation in small wastewater treatment systems. Ecological Engineering 26, 266e271. Gregersen, K., Brix, H., 2001. Zero-discharge of nutrients and water in a willow dominated constructed wetland. Water Science and Technology 44 (11e12), 407e412. Haefele, S.M., Siopongco, J.D.L.C., Boling, A.A., Bouman, B.A.M., Tuong, T.P., 2009. Transpiration efficiency of rice (Oryza sativa L.). Field Crops Research 111 (1e2), 1e10.
354
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 4 5 e3 5 4
Headley, T.R., Huett, D.O., Davison, L., 2001. The removal of nutrients from plant nursery irrigation runoff in subsurface horizontalflow wetlands. Water Science and Technology 44 (11e12), 77e84. Headley, T.R., Huett, D.O., Davison, L., 2003. Seasonal variation in phosphorus removal processes within reed beds- mass balance investigations. Water Science and Technology 48 (5), 59e66. Herbst, M., Kappen, L., 1999. The ratio of transpiration versus evaporation in a reed belt as influenced by weather conditions. Aquatic Botany 63, 113e125. Kadlec, R.H., 1987. The hydrodynamics of wetland water treatment systems. In: Reddy, K.R., Smith, W.H. (Eds.), Aquatic Plants for Water Treatment and Resource Recovery. Magnolia Publishers Inc, Orlando, pp. 373e392. Kadlec, R.H., 1989. Hydrologic factors in wetland water treatment. In: Hammer, D.A. (Ed.), Constructed Wetlands for Wastewater Treatment e Municipal, Industrial and Agricultural, (3rd Print). Lewis Publishers, Boca Raton, pp. 21e40. Kadlec, R.H., 2006. Water temperature and evapotranspiration in surface flow wetlands in hot arid climate. Ecological Engineering 26, 328e340. Kadlec, R.H., Knight, R.L., 1996. Treatment Wetlands, first ed. CRC Press, Boca Raton. Kadlec, R.H., Wallace, S.D., 2009. Treatment Wetlands, second ed. CRC Press, Boca Raton. Kemanian, A.R., Sto¨ckle, C.O., Huggins, D.R., 2005. Transpirationuse efficiency of barley. Agricultural and Foretry Meterology 130, 1e11. Kirkham, M.B., 2005. Principles of Soil and Plant Water Relations. Elsevier Academic Press, London. erova´, A., Pokorny´, J., Radoux, M., Nemcova, M., Cadelli, D., Kuc Du sek, J., 2001. Evapotranspiration of small-scale constructed wetlands planted with ligneous species. In: Vymazal, J. (Ed.), Transformations of Nutrients in Natural and Constructed Wetlands. Backhuys Publishers, Leiden, The Netherlands, pp. 413e427. Masi, F., Martinuzzi, N., 2007. Constructed wetlands for the Mediterranean counties: hybrid systems for water reuse and sustainable sanitation. Desalination 215, 44e55. Mengel, K., Kirkby, E.A., 2001. Principles of Plant Nutrition, fifth ed. Springer, New York. Mitsch, W.J., Gosselink, J.G., 2007. Wetlands, fourth ed. John Wiley & Sons Inc., New York. Monteith, J.L., 1965. Evaporation and the environment. In: Proceedings of the 19th Symposium of the Society for
Experimental Biology, New York. Cambridge University Press, New York, pp. 205e233. Morari, F., Giardini, L., 2009. Municipal wastewater treatment with vertical flow constructed wetlands for irrigation reuse. Ecological Engineering 35, 643e653. Moro, M.J., Domingo, F., Lopez, G., 2004. Seasonal transpiration pattern of Phragmites australis in a wetland of semi-arid Spain. Hydrological Processes 18, 213e227. Pauliukonis, N., Schneider, R., 2001. Temporal patterns in evapotranspiration from lysimeters with three common wetland plant species in the eastern United States. Aquatic Botany 71, 35e46. Peacock, C.E.:, Hess, T.M., 2004. Estimating evapotranspiration from a reed bed using the Bowen ratio energy balance method. Hydrological Processes 18, 247e260. Reed, S.C., Crites, R.W., Middlebrooks, E.J., 1995. Natural Systems for Waste Management and Treatment, second ed. McGrawHill, New York. Santos, A.M., Esteves, F.A., 2004. Comparison of calculation procedures of primary productivity by aquatic macrophytes in a shallow tropical coastal lagoon. Acta Limnologica Brasiliensis 16 (3), 239e249. Stewart, J.B., 1984. Measurements and prediction of evaporation from forested and agricultural catchment. Agricultural Water Management 8, 1e28. Thornthwaite, C.W., 1948. An approach toward a rational classification of climate. Geographical Review 38, 55e94. UK Met Office, 2003. 1971e2000 Climate Averages (accessed 26.11.03.). http://www.metoffice.gov.uk/climate/uk/averages/ 19712000/areal/midlands.html. Vymazal, J., 2002. The use of sub-surface constructed wetlands for wastewater treatment in the Czech Republic: 10 years experience. Ecological Engineering 18, 633e646. Vymazal, J., Kro¨pfelova´, L., 2008. Wastewater Treatment in Constructed Wetlands with Horizontal Sub-Surface Flow. Springer, New York. Wallace, S.D., Knight, R.L., 2006. Small-scale Constructed Wetland Treatment Systems Feasibility, Design Criteria, and O&M Requirements. Water Environment Research Foundation, Virginia. Wetzel, R.G., 1975. Limnology. W.B. Saunders, Philadelphia. Zhou, L., Zhou, G., 2009. Measurement and modeling of evapotranspiration over a reed (Phragmites australis) marsh in Northeast China. Journal of Hydrology 372, 41e47.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Environmental occurrence, fate and transformation of benzodiazepines in water treatment a T. Kosjek a, S. Perko a, M. Zupanc a,b, M. Zanoski Hren c, T. Landeka Dragicevic c, D. Zigon , d a, B. Kompare , E. Heath *
Jo zef Stefan Institute, Department of Environmental Sciences, Ljubljana, Slovenia Ecological Engineering Institute IEI, Maribor, Slovenia c Faculty of Food Technology and Biotechnology, University of Zagreb, Zagreb, Croatia d Faculty of Civil and Geodetic Engineering, University of Ljubljana, Ljubljana, Slovenia a
b
article info
abstract
Article history:
Benzodiazepine derivatives are prescribed in large quantities globally and are potentially
Received 19 July 2011
new emerging environmental contaminants. Unfortunately, a dearth of data exists con-
Received in revised form
cerning occurrence, persistence and fate in the environment. This paper redresses this by
24 October 2011
reviewing existing literature, assessing the occurrence of selected benzodiazepine anxio-
Accepted 25 October 2011
lytics (diazepam, oxazepam and bromazepam) in wastewater influent and effluent and
Available online 7 November 2011
surface water from Slovenia, evaluating their removal during water treatment and identifying the transformation products formed during water treatment. Their occurrence was
Keywords:
monitored in hospital effluent, river water and in wastewater treatment plant influent and
Diazepam
effluent. The study reveals the presence of benzodiazepine derivatives in all samples with
Oxazepam
the highest amounts in hospital effluents: 111 ng L1, 158 ng L1 and 72 ng L1 for diaz-
Bromazepam
epam, bromazepam and oxazepam, respectively. Removal efficiencies with respect to
Environment
biological treatment of diazepam were 16e18% (oxic), 18e32% (anoxic / oxic), 53e76%
Treatment
(oxic / anoxic) and 83% (oxic / anoxic / oxic / anoxic cascade bioreactors), while the
Transformation
removal oxazepam was 20e24% under anoxic conditions. Coupled biological and photochemical treatment followed by the adsorption to activated carbon resulted in a removal efficiency of 99.99%. Results reveal the recalcitrant nature of benzodiazepine derivatives and suggest that only combinational treatment is sufficient to remove them. In addition, eight novel diazepam and four novel oxazepam transformation products are reported. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Benzodiazepines are the drugs of choice in the pharmacotherapy of anxiety and related emotional disorders, sleep disorders, status epileptics, and other convulsive states. They are also used as centrally acting muscle relaxants, for premedication and as inducing agents in anaesthesiology (Neumeyer and Booth, 1995). Their potential shortcomings
include tolerance, withdrawal symptoms, and their abuse potential (Riss et al., 2008; Kosjek and Heath, 2011). Diazepam (DZ, Valium) is perhaps the most known drug among this group and is long acting because of its active metabolites that have long half-lives (Baldessarini et al., 1996). In humans it is metabolized into either N-desmethyldiazepam or nordazepam (half-life of up to 100 h). Nordazepam is then further 3-hydroxylated into oxazepam (OXA). The presence of
* Corresponding author. Tel.: þ386 1 4773584. E-mail address:
[email protected] (E. Heath). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.056
356
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
the hydroxyl group enables rapid glucuronidation and excretion with the urine and explains its significantly shorter halflife (4e15 h) when compared to both DZ and nordazepam (Neumeyer and Booth, 1995; Benzodiazepine Equivalents Table, 2011). Nordazepam and OXA are marketed as Calmday and Serax, respectively. According to Fu¨rst et al. (2006), besides DZ, the most commonly prescribed anxiolytics in Slovenia are bromazepam (BZ, half-life: 10e20 h) and alprazolam (half-life: 6e12 h). Fu¨rst et al. (2006) also reported significant differences in the use of high doses of DZ and BZ between the individual regions of Slovenia, what is suggestive of their potential abuse. In environmental research, benzodiazepines are among those pharmaceuticals less commonly addressed (Ternes et al., 2001). Diazepam has been detected in wastewater (WW) originating from hospitals as well as in effluents from municipal wastewater treatment plants. Ternes et al. (2001) and Martı´nez Bueno et al. (2007) found levels of DZ as high as 53 ng L1 and 87 ng L1, respectively, in wastewater treatment plant effluent. DZ was further found at 33 ng L1 in German rivers (Ternes et al., 2001), up to 21 ng L1 in rivers in the region of Madrid (Martı´nez Bueno et al., 2010) and Zuccato et al. (2000) found up to 23.5 ng L1 DZ in drinking water. Further, Heberer (2002) reports the presence of OXA (0.25 mg L1) in a wastewater treatment plant (WWTP) effluent, while Baker and Kasprzyk-Hordern (2011) found OXA in majority of WWTP effluent, influent and river water samples. In addition, BZ, OXA and DZ were determined in the Llobregat River (north eastern Spain e a source of potable water at mean concentrations of 7 ng L1, 20 ng L1 and 3 ng L1, respectively) (Huerta-Fontela et al., 2011). Benzodiazepines are normally halogenated compounds, and it is suggested that the presence of a halogen in a chemical structure significantly reduces its susceptibility to biodegradation (Johnson et al., 2008). Most data concerning the efficiency of WW treatment is for DZ, and reveals that <10% is removed during classical biological treatment, while anaerobic sludge treatment is only slightly more efficient (10e50%); removal is a result of adsorption to activated sludge, rather than degradation (Ternes et al., 2004; Lo¨ffler et al., 2005). Similarly, OXA is persistent to both, aerobic or anaerobic biodegradation and field-based experiments also reveal its recalcitrance (Patterson et al., 2010, 2011). To our knowledge, other than the papers cited no other published degradation data on OXA is available. DZ is considered to undergo photochemical degradation under environmental conditions which may constitute a feasible mechanism for its removal from surface waters (Boreen et al., 2003). In comparison to classical treatment technologies, advanced oxidation methods are more efficient at eliminating DZ (Belden et al., 2007; Calisto and Esteves, 2009). Degradation of DZ was improved in the presence of ferrioxalate, either under black-light or solar irradiation and the removal efficiency was 80% after 60 min irradiation (Bautitz and Nogueira, 2010). DZ is relatively resistant to ozonation (Verlicchi et al., 2010), but it can be oxidised by ∙OH radicals during ozone treatment (Ternes et al., 2004). Unfortunately, there are no published data for the remaining benzodiazepines regarding their persistence to abiotic treatment methods.
Eventhough the Kow of these compounds do not indicate sorption to be an important removal process, recent researches suggest this is not the case. Ternes et al. (2004) classify DZ as easily adsorbable to activated carbon (99% removal, 0.2 mg L1 of activated carbon) while Calisto and Esteves (2009) report that temazepam and OXA may undergo more abiotic losses by sorption to humic substances. Lo¨ffler et al. (2005) found DZ to be a highly persistent pharmaceutical with rapid and extensive sorption onto sediments as well as being highly stable in soils, ground waters and during the WWs treatment. In the same study OXA was reported as moderately persistent in water/sediment systems. It is clear that available data regarding the presence, cycling and fate of benzodiazepine derivatives in the environment are insufficient. This study aims to address this deficiency by (i) identifying and determining the environmental concentrations of benzodiazepines residues, (ii) evaluating their removal during biological and photochemical water treatment and (iii) studying the formation of stable transformation products during wastewater treatment. In addition we also make suggestions about how to improve current wastewater treatment.
2.
Experimental
2.1.
Chemicals
Authentic compounds DZ and BZ were donated by a cooperating body, while OXA was curtsy of Belupo, d.d. (Croatia). The internal standard [2H5]-oxazepam (OXA-d5: hydrogen atoms of the C6H5 phenyl group were replaced by deuterium atoms, 99.9%) was purchased from SigmaeAldrich (St. Louis, MO, USA), and [2H5]-diazepam (DZ-d5: hydrogen atoms of the C6H5 phenyl group were replaced by deuterium atoms, 99.9%) was purchased from LGC Standards GmbH (Wesel, Germany). Both deuterated internal standards were used for the trace-level analytical method development and validation, and for the determination of occurrence of the selected pharmaceuticals in the environment. All standards were of highest available purity (>99%). For derivatisation we used acetic anhydride (Sigma Aldrich, USA) and pyridine (Merck, Germany). All applied solvents (ethylacetate, methanol) and chemicals (hydrogen peroxide 30%) were of analytical grade purity.
2.2.
Sample preparation
Fourhundred mililiters river water (RW) and 200 mL WW samples were first filtered through glass microfiber prefilters (Machery Nagel, Dueren, Germany) and afterwards through 1.2 mm cellulose nitrate (Whatman, Kent, UK) filters. After filtration the internal standards DZ-d5 and OXA-d5 were added. The concentration of the internal standards was 4.4 nM in the case of treatment experiments; 0.44 nM (DZ-d5) and 1.32 nM (OXA-d5) for determination of actual WW samples; and 0.22 nM (DZ-d5) to 0.44 nM (OXA-d5) for natural water samples. The sample pH was 6.5e7.0 and was not adjusted before the extraction. Samples were extracted by Oasis (Waters Corp., Milford, MA, USA) 60 mg/3 mL solidphase extraction (SPE) cartridges, previously conditioned
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
with 3 mL ethylacetate, 3 mL methanol and equilibrated by 3 mL of deionised water. After the enrichment phase the cartridges were dried under stream of nitrogen and the analytes were eluted by 1 mL acetone, 1 mL of acetone/ethylacetate (3:7) and 1 mL ethylacetate. The eluate was transferred to a glass vial and dried under a gentle stream of nitrogen. The derivatisation agents (15 mL acetic anhydride and 5 mL pyridine) were added to the dry extract, which was vortexed and left at 80 C for 15 h. After derivatisation the solvent was removed (N2). The derivatised extracts were then suspended in ethylacetate (0.1 mL for WW sample, 0.2 mL for RW samples and 0.5 mL in case of treatment experiments). For the determination of transformation products we applied a similar sample preparation procedure as that described above, excluding the following steps: internal standards were not used, derivatisation was not performed, and the dry extracts were dissolved in an initial composition of LC mobile phase.
2.3.
Analysis
2.3.1.
Quantitative analysis
HP 6890 series (HewlettePackard, Waldbron, Germany) gas chromatograph fitted with a single quadrupole mass selective detector (GC-MSD) was used. The GC oven was programmed as follows: an initial temperature of 80 C was ramped at 30 C/min to 250 C, held for 4 min, ramped at 5 C/min to 280 C, at 30 C/min to 300 C, and finally held for 1 min and 2 min post run at 300 C. The total GC run time was 19.34 min. A DB-5 MS 30 m 0.25 mm 0.25 mm (Agilent J&W, CA, USA) capillary column was used, with He as the carrier gas (37 cm s1). One-microlitre samples were injected at 250 C in splitless mode, and the transfer line was maintained at 280 C. The MS was operated in EI ionisation mode at 70 eV. In SCAN mode, masses from m/z 50 to 550 were scanned, while in SIM mode, the following ions were monitored: m/z 219, 253 and 254 for acetyl-OXA, m/z 224, 258 and 259 for acetyl-OXA-d5, m/z 220, 299 and 301 for acetyl-BZ, m/z 256, 283 and 284 for DZ and m/z 261, 288 and 289 for DZ-d5. The GC-MSD used Chemstation software for instrumental control and data processing.
2.3.2.
Identification of transformation products
The chromatographic separation was performed on a Waters Acquity ultra-performance liquid chromatograph (Waters Corp., Milford, MA, USA), equipped with a binary solvent delivery system and an autosampler. The injection volume was 5 mL. Separation was achieved using a 3-cm-long Acquity UPLC BEH C18 (Waters Corp., Milford, MA, USA) column with 1.7-mm particle size and 2.1-mm internal diameter. Compounds were analysed under positive [ESI(þ)] ion conditions. The mobile phases used were (A) 0.1% formic acid and (B) acetonitrile. The elution gradient was linearly increased from 10 to 80% B in 5 min, decreased back to 10% in 0.3 min and then finally kept isocratic for 1.2 min. The total runtime was 6.5 min. Flow rate was 0.3 mL min1 and the column temperature was maintained at 40 C. The UPLC system was interfaced to a hybrid quadrupole orthogonal acceleration time-of-flight mass spectrometer (QqToF Premier, Waters Corp.). The instrument was equipped with an electrospray ionisation interface. The capillary voltage was set to 3.0 kV,
357
while the sampling cone voltage was varied between 20 and 30 V. Source and desolvation temperatures were set to 130 and 250 C, respectively. The nitrogen desolvation gas flow rate was 530 L h1. For MS experiments, the first quadrupole was operated in rf-only mode, while detection was performed in the ToF mass analyser. MS data were acquired over an m/z range of 100e1000 at collision energy of 4 V. For MS/MS operation, the acquisition range was between m/z 50 and 400, and argon was used as the collision gas at a pressure of 4.5 103 mbar in the T-wave collision cell. The MS/MS experiments were performed with collision energy, varied between 15 and 40 V, to generate product ion spectra providing the most structural information. Data were collected in centroid mode, with a scan accumulation time set to 0.25 s and an interscan delay of 0.02 s. The data station operating software was MassLynx v4.1. Prior to analysis, the instrument was calibrated over a mass range 50e1000, using a sodium formate calibration solution. Reproducible and accurate mass measurements, at a mass resolution of 10 000, were obtained using an electrospray dual sprayer with leucine enkephalin ([M H] 554.2615, [M H]þ 556.2271) as the reference compound. The latter was introduced into the mass spectrometer alternating with the sample via a Waters Lock Spray device. Elemental composition of TPs was calculated from accurate masses determined by high resolution mass spectrometry (HRMS) at following conditions: C:0e25, H:0e30, N:0e3, O:0e10; Cl:0e1 5.0 ppm tolerance. For enhanced detection of the transformation products data were acquired in centroid mode and afterwards processed by the MetaboLynx application manager embedded into MassLynx v4.1. software (Waters Corp.). The algorithm was programmed to detect products of expected transformation pathways (i.e. hydroxylation and demethylation) and also to detect unexpected components. The latter were examined in m/z 100e400 scanning range with 10 Da size of a step scan. The presence of transformation products was investigated in treated samples, while untreated samples were used as control samples.
2.4.
Sampling
Sampling was performed in three Slovene towns in winter and spring 2011. Each batch comprised from two to six sampling points. Sampling point locations are schematically shown in the Supporting material. Town A has approximately 35 000 inhabitants and on its outskirts is a pharmaceutical industry marketing DZ and BZ. Grab sampling was performed prior and post the inhabitated area. Town B has a population of cca 105 000 with a WWTP and a hospital (approx 250 beds and 10 000 patients hospitalised annually). Time proportional 24hrs sampling included two hospital waste streams (1 and 2), a WWTP influent (3) and effluent (4) and receiving waters upstream (5) and downstream (6) of the WWTP outflow. In town C (360 000 inhabitants, one of the largest hospitals in Central Europe with psychiatric clinic, in total cca 3000 beds and cca 105 000 patients hospitalised annually), grab samples were taken before (1) and after (2) the discharge of the town’s main WWTP effluent (3) into a river. The WWTP (mechanical and biological treatment) treats WW for 360 000 population equivalents (PE).
358
2.5.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
Biodegradation experiments
2.5.1. Setup and operation of a pilot wastewater treatment plant (PWWTP) Biodegradation experiments were performed in flow-through pilot bioreactors, which were operated either as individual units or were coupled into cascades as described in Table 1. The individual unit is a 4L flow-through oxic bioreactor, which is in details described in Kosjek et al. (2007, 2009). The anoxic conditions were established by removing the air supply, while constantly mixing the bioreactors’ content using a magnetic stirrer bar. The bioreactor is shown in Kosjek et al. (2007). All bioreactors/cascades, except for “D” were operated in two parallels. The bioreactors (or in case of a sequence the first bioreactor) were fed with a mineral-nutrient medium (Kosjek et al., 2007) and 100 mg L1 DZ or OXA was continuously added to the bioreactor influents as described in Table 1. The bioreactors fed with DZ were operated in parallel, for the first 19 weeks as two individual oxic bioreactors (parallels X1 and X2, encoded “A”, Table 1), then three weeks were allowed for adaptation of the anoxic/oxic sequence (parallels A1 / X1 and A2 / X2, “B”, Table 1), and thereafter sampling was performed for the following 7 weeks (week 23e29). Further, the bioreactors were swapped into sequence oxic/anoxic (parallels X1 / A1 and X2 / A2, “C”, Table 1), and again after a three-week adaptation period, sampling was performed from week 33 to 45. Finally, the two oxic (X1 and X2) and two anoxic (A1 and A2) bioreactors were coupled into a single sequence X2 / A2 / X1 / A1 (code “D”). Separately, OXA was added into two oxic bioreactors O1 and O2, as evident from Table 1 (code “E”).
2.5.2.
Assessment of the PWWTP operation
The operation of reactors was assessed by monitoring the total chemical oxygen demand (CODt), NO3eN, NO2eN and NH4eN in the influents and effluents of bioreactors. Along with these parameters the concentration of mixed liquor suspended solids (MLSS), temperature and concentration of oxygen were monitored inside the bioreactors. All measurements were performed on the same day as the sampling for pharmaceuticals. COD was determined using a DR/2010 spectrophotometer (Hach, Du¨sseldorf, Germany) and a Hach measuring kit in an appropriate concentration range (0e1500 mg L1 for influent and 0e150 mg L1 for effluent samples). NO3eN was
determined using powder pillows NitraVer 5 MR (0e4.5 mg L1) with Method 353 on a DR/2010 spectrophotometer, and NO2eN by NitriVer 3 LR (0e0.3 mg L1) with Method 371. For NH4eN the Nessler method (Method 380) was employed (0e2.5 mg L1). Where necessary, samples were appropriately diluted to fit into a defined concentration range. To determine MLSS, 15 mL of a sample was filtered through a previously dried and weighed filter using a vacuum crucible and dried at 105 C to constant weight. Oxygen levels and temperature were measured simultaneously using a HQ30d probe (Hach, Du¨sseldorf, Germany).
2.6.
H2O2/UV treatment
UV treatment of DZ and OXA was performed using a monochromatic low pressure mercury lamp (LP), with a peak emission at 254 nm. The apparatus was composed of a glass reactor with a volume of 1.5 L. A constant temperature (20e22 C) was maintained by continuously cooling the treated solution with cold water, and by continuous mixing with a magnetic stirrer at 400 rpm. The pH (6.5e7.0) did not change during the course of the treatment. The UV treatment was performed during different time intervals (10 min, 30 min, 60 min and 120 min) and with and without the addition of H2O2 in different concentrations (33 mg L1, 330 mg L1, 1.6 g L1 and 3.3 g L1). The treatment was initially performed in 0.5 L distilled water at 100 mg L1 concentrations of individual parent compounds, first to determine the transformation products generated during treatment, and secondly, to set the operational parameters for further treatment of actual WW obtained from the bioreactor effluents described in Table 1.
2.7.
Adsorption to granulated activated carbon
Activated carbon (AC) with screen size 30 60 was provided by the Coast Engineering Laboratory (Redondo Beach, USA). Experiments were performed on effluents of DZ and OXA bioreactors (Table 1) prior or post UV/H2O2 treatment, at varying concentrations of AC (from 200 to 4000 mg L1) and durations of exposure (15 and 60 min). 1200 mg L1 of AC and 60 min of exposure time were found sufficient for elimination of residual DZ and OXA post UV/H2O2 treatment.
Table 1 e Bioreactor setup. Week (Ad)/(Samp) Bioreactor setup No. of bioreactors per sequence Parent comp. Bioreactor code
1
4
20
23
30
33
46
49
46
49
Ad
Samp
Ad
Samp
Ad
Samp
Ad
Samp
Ad
Samp
X1 X2 1 DZ A
A1 / X1 A2 / X2 2 DZ B
X1 / A1 X2 / A2 2 DZ C
X1 / A1 / X2 / A2 4 DZ D
O1 O2 1 OXA E
A1, A2, anoxic bioreactors fed with DZ; X1, X2, oxic bioreactors fed with DZ; O1, O2, oxic bioreactors fed with OXA; Ad, addition of parent compounds; Samp, sampling.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
3.
Results and discussion
3.1.
Method development and validation
To improve gas-chromatographic response, peak shape and to avoid thermal decomposition of OXA in the injector (Kinani et al., 2007), derivatisation was performed to transform the benzodiazepines into more volatile and thermally stabile derivatives. According to Borrey et al. (2001), acetylation of benzodiazepines with acetic anhydride and pyridine is superior to common derivatisation methods such as silylation or alkylation. The advantages of acetylation are that the reagents and the by-product (acetic acid) are easily evaporated with nitrogen, and any excess reagents will not damage the chromatographic column. In addition, acetylation targets the secondary N-1 atom of OXA and BZ (Fig. 1), whereas DZ which already shows good peak shape and sensitivity has a tertiary N 1, which will remain underivatised. Derivatisation is considered to be the most variable step in the sample preparation procedure. The optimization of the derivatisation conditions involved testing the acetic anhydride/pyridine ratio, reaction temperature and reaction time. Finally, 15 mL acetic anhydride and 5 mL pyridine at 80 C for 15 h were sufficient to yield stabile acetyl derivatives of OXA, OXA-d5 and BZ. The derivatives were found to be stabile for at least 14 days. Other validation parameters involved testing SPE efficiencies, which were 84e101% with standard deviations between 2.9 and 14%. Since the internal standards were added prior to the SPE there was no need for this variation to be taken into account in the overall method performance. The limit of detection of the analytical method (LOD) values were not higher than 3 ng L1 for all three target analytes. The calibration curves were obtained by plotting the target ion response ratio of the analyte to internal standard against the concentration ratio. Both internal standards OXA-d5 and DZ-d5 were taken into account. The method linearity was proven by regression coefficients >0.98 in the range of 10e300 ng L1. The method repeatability was determined based on three consecutive measurements of the same sample containing the target analytes in the low ng L1 level and ranged from 1.4 to 3.7% within-day and 2.9e11% between-day.
3.2.
Occurrence studies
To confirm the identity of target analytes and to enable their secure quantification on the single quadrupole mass detector,
359
we applied acceptability criteria as follows: (1) ion ratios for a given analyte between qualifier and quantification ion were required to be within 20% of the average ion ratios in respect to the control samples, and (2) retention time was within 0.20 min of the average for all respective controls (Brooks et al., 2005). The results of the occurrence studies on the presence of OXA, DZ and BZ in Slovene water environment are given in Table 2. Despite DZ and BZ being marketed as commercial products by the pharmaceutical industry located at River A, none of the selected benzodiazepine derivatives were present at the sampling locations (upstream and downstream of town A, for detailed locations see Supplementary material). There are several possible explanations for this: it is possible that the compounds were not manufactured at the time of sampling or that grab sampling may have missed the contaminants that time or flow proportional composite sampling would have captured. Another solution would be to analyse sediment samples downstream from the suspected point source (pharmaceutical plant) that could provide a historical record of pollution. River C was sampled in the same manner as River A (grab sampling). The results reveal contamination with two out of three selected benzodiazepine derivatives (OXA and DZ). While OXA was detected only once in a concentration above LOD before the WWTP effluent was discharged into the River C, DZ was detected in all the grab samples in concentrations up to 69 ng L1. The fact that DZ was detected in all the samples suggests a constant discharge of DZ into River C. Town B samples were collected in collaboration with a local WWTP. Each was a 24 h composite taken as part of a regular monitoring campaign performed at six locations (see Supplementary material) during two seasons (winter and spring 2011). Results show that OXA, with the exception of both hospital effluents collected during the winter, is present in all the other samples. Compound BZ was present in approximately 60% of samples, while DZ was present in all samples. The presence of BZ in the WW and RW samples may be related to the significantly higher consumption of this pharmaceutical in the area as determined by Fu¨rst et al. (2006). There is a lack of data available in the area of persistence or half-life times of benzodiazepine derivatives available. PBT profiler (www.pbtprofiler.net, accessed 21.10.2011) reports half-lives for DZ and OXA in water, soil and sediment to be 38, 75 and 340, respectively; while BZ reported half-lives are 60, 120 and 540. These data indicate DZ and OXA not to be persistent in water (<60 days) while BZ could already be
Fig. 1 e Benzodiazepines: (a) 5-phenyl-1,4-benzodiazepin-2-one: ring system, (b) DZ, (c) OXA, (d) BZ.
360
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
Table 2 e Concentrations (ng LL1) of OXA, DZ and BZ determined in the wastewater and river water samples. Sampling point
Sampling date
1-A 2-A
Winter 2011 Winter 2011
1-B 1-B 2-B 2-B 3-B 3-B 4-B 4-B 5-B 5-B 6-B 6-B
1-C 2-C 2-C 3-C 3-C
Town A River before municipality River after municipality and pharmaceutical industry Town B Hospital effluent 1 Hospital effluent 2 WWTP influent WWTP effluent Stream before effluent Stream after effluent
Town C River before WWTP discharge River after WWTP discharge WWTP effluent
c(OXA) ng/L
c(BZ) ng/L
c(DZ) ng/L
Grab Grab
Winter 2011 Spring 2011 Winter 2011 Spring 2011 Winter 2011 Spring 2011 Winter 2011 Spring 2011 Winter 2011 Spring 2011 Winter 2011 Spring 2011
Composite 24 h Composite 24 h Composite 24 h Composite 24 h Composite 24 h Composite 24 h Composite 24 h Composite 24 h Composite 24 h Composite 24 h Composite 24 h Composite 24 h
40
27 49 17 111 21 25 18 22 17 28 21 20
Spring 2011 Spring 2011 Winter 2011 Spring 2011 Winter 2011
Grab Grab Grab Grab Grab
9 69 13 21 22
classified as moderately persistent (60 days). DZ, OXA and BZ are classified as moderately persistent in soil (60 days) and very persistent in sediment compartment (>180 days). Basing on these data BZ shows slightly higher persistence when comparing three derivatives. On the other side Lo¨ffler et al. (2005) acknowledges the differences between persistence of DZ and OXA where DZ is reported to be highly and OXA moderately persistent pharmaceutical.
3.3.
Biodegradation experiments
3.3.1.
Assessment of performance of the PWWTP
Supplementary data gives the recorded values for CODt, NO3eN, NO2eN and NH4eN. As illustrated in cases B, C and E the CODt value declined by approximately 95% in the first reactor, disregarding the oxygen concentration. The CODt was further decreased in the subsequent reactor(s) in a cascade
Sampling approach
(examples B and C), but this decrease was less significant. High variability of CODt is observed in the middle container and in the effluent, which may be due to the dead biomass discharged from a reactor. Based on an increase in NO3eN concentration (Supplementary data) the nitrification process is confirmed in the oxic reactor “E” and in the oxic bioreactor (IN / M) of the cascade “C”, and the predominant nitrification process in the cascade “D”. Slovene guideline (2011) on the quality of WWTP discharge suggests the upper limit for COD at 100 L1 and NH4eN at 5 L1 (for WWTPs with 100 000 PE), and NH4eN at 10 L1 (for WWTPs from 2000 to 100 000 PE). Considering the COD both, the actual effluents and the middle container effluents, are acceptable for discharge, while the NH4eN concentrations slightly exceed the limiting values, and therefore an additional removal of NH4eN is necessary.
Table 3 e Removal of DZ and OXA in individual or cascade bioreactors; expressed as average removal ± standard deviation (number of samples). Bioreactor/cascade code
Parent compound
Conditions
A
DZ
Oxic
B
DZ
Anoxic / oxic
C
DZ
Oxic / anoxic
D E
DZ OXA
Oxic / anoxic / oxic / anoxic Oxic
Removal stdev (n) X1: 16 9 (15) X2: 18 8 (14) A1: 13 8 (7) X1: 14 5 (6) A2: 13 10 (7) X2: 24 10 (6) X1: 34 17 (13) A1: 32 13 (11) X2: 67 10 (13) A2: 29 13 (10) X2 / A2 / X1 / A1: tot: 83 O1: 20 10 (5) O2: 24 9 (6)
tot: 18 tot: 32 tot: 53 tot: 76 8.5 (6)
A1, A2, anoxic bioreactors fed with DZ; X1, X2, oxic bioreactors fed with DZ; tot, total removal; O1, O2, oxic bioreactors fed with OXA.
6 (7) 12 (7) 15 (12) 10 (12)
361
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
Table 4 e The results of UV treatment (% removal) of DZ and OXA in DIW and of UV treatment of residual DZ and OXA in biologically treated wastewater (filtered wastewater-FWW and raw wastewater-RWW). Matrix/addition of H2O2 (mg L1) DIW/ DIW/ DIW/ FWW/ FWW/ FWW RWW/ RWW/ RWW/ 0 mgL1 33 mgL1 330 mgL1 330 mgL1 1.6 gL1 /3.3 gL1 330 mgL1 1.6 gL1 3.3 gL1 UV treatment (min)
3.3.2.
10 30 60 120 10 30 60 120
e e e 8 e e e e
47 e e e e e e e
71 96 e e 53 93 e e
22 35 e 74 27 72 83 99
Removal of DZ and OXA
For DZ different biological treatment methods report poor removal efficiencies, being evident the large amount of diazepam passes unaffected through WWTPs (Ternes et al., 2004; Suarez et al., 2010). Correspondingly, as evident from Table 3 (codes A and B), during the first weeks after the bioreactor setup both the oxic and anoxic bioreactors showed low removal efficiency for DZ. Table 3 summarizes the results of biological treatment of DZ and OXA. Biological treatment was for both compounds performed in two parallels, except for the cascade “D”, where both oxic (X1 and X2) and anoxic (A1 and A2) bioreactors fed with DZ were connected into a single cascade. The variability of removal of the two parent compounds was approximately 10%, which is not surprising since bioreactors contain biomass, which is difficult to control. In general, oxic treatment was more efficient than anoxic treatment, even though we observed an increase in both, the anoxic and oxic removal, over one year of operation. This increase in removal is particularly notorious for the X2 oxic bioreactor, where the elimination efficiency increases from 18% to 67% (Table 3) in one year (see Supplementary data). It is assumed that this phenomenon is the consequence of biomass adaptation, but additional experiments such as enrichment of WW with DZ and the microbiological analysis of the biomass are needed to prove this. For OXA, elimination under oxic conditions (E, Table 3) was 20% and 24% in parallels O1 and O2, respectively. The poor biological elimination efficiency of OXA is also reported in two studies by Patterson et al. (2010, 2011), whereas to our knowledge no other reports on the environmental behaviour of OXA exists.
3.4. Photochemical treatment and sorption to activated carbon The potential of UV treatment was investigated for DZ and OXA in distilled water (DIW) and biologically treated WW. The
22 39 e 86 e e e e
16 e 52 72 e e e e
14 11 43 72 20 69 90 e
17 23 54 82 38 60 83 97
e e 44 62 e e e e
Tested compound
DZ
OXA
UV treatment of DZ in DIW (direct photolysis) was not efficient removing only 8% of the parent compound in 2 h. By the addition of a H2O2 as the source of radicals the removal efficiency was improved significantly, as previously shown in Klavarioti et al. (2009) and Kosjek et al. (2011). Table 4 shows how the addition of 0.01% H2O2 increases removal efficiency to 47% under 10 min of UV treatment. By increasing the concentration of H2O2 (0.1%) and by prolonging the duration of UV treatment (30 min) we were able to enhance removal efficiency to 96% (Table 4). However, it was observed that further increases in concentration of H2O2 up to 1% resulted in a negative effect on DZ removal. This phenomenon has been in the scientific literature explained by excess H2O2 playing a role as OH radical scavenger, thus making the photochemical treatment less effective (Pereira et al., 2007; Matilainen and Sillanpa¨a¨, 2010). Further, it can be observed in Table 4 that for more complex matrices such as filtered wastewater (FWW) and raw wastewater (RWW) a higher concentration of H2O2 and prolonged UV irradiation time are needed to achieve a comparable removal effect. Thus, the addition of 0.5% H2O2 at 120 min UV irradiation time was most favorable, achieving 86% and 82% removal of DZ in FWW and RWW, respectively. In comparison, when applying the same treatment conditions (0.1% H2O2 and 30 min UV treatment) as those necessary for 96% removal of DZ in DIW, we achieved only 35% and 11% removal of DZ in FWW and RWW, respectively (Table 4). Table 4 further shows the results of the UV/H2O2 treatment of OXA in DIW, FWW and RWW under varying conditions, i.e. addition of H2O2 and at UV irradiation time. By comparing the results of the photochemical treatment for DZ and OXA (Table 4) it can be concluded that OXA is more prone to oxidation by UV/ H2O2 treatment than DZ achieving 90% removal at 0.1% addition of H2O2 at 60 min UV irradiation time. The likely reason for this is the additional hydroxyl group on the position C-3, which makes OXA more susceptible to OH radical oxidations.
Table 5 e Removal of DZ and formation of OXA during coupled biological and abiotic treatment.
c (ng/L) DZ c (ng/L) OXA formed during treatment
Influent
X2 / A2 / X1 / A1 effluent
Effluent þ UV/H2O2
Effluent þ AC
Effluent þ UV/H2O2 þ AC
96.6 103 0
18.4 103 2.2 103
170 166
450 163
16
362
Table 6 e Transformation products of DZ: LC retention time (LC e tR), description of mass spectra (HR measurement and tandem mass fragmentation), treatment conditions and proposed chemical structures. LC e tR (min)
Accurate mass (calculated) [M H]þ
Elemental composition [M H]þ
Mass error
MS/MS
MW (nominal)
Treatment conditions
Chemical structure (proposed)
DZ
2.97
285.0795
C16H14N2OCl
0.5 ppm
284
Parent compound
Fig. 1(b)
OXA
2.40
287.0587
C15H12N2O2Cl
0.3 ppm
285/287, 257/259, 228/230, 222, 193, 154/156 287/289, 269/271, 257/259, 241/243, 205, 163/165, 151
286
1. From DZ by biotransformation; 2. Used as a parent compound
Fig. 1(c)
Compound/ abbreviation
H 3C
OH
Temazepam
2.72
301.0744
C16H14N2O2Cl
1.3 ppm
301/303, 283/285, 271/273, 255/257
300
Photocatalysis, biotransformation
Cl
N
H N
O
Cl
Nordazepam
2.51
271.0638
C15H12N2OCl
0.0 ppm
271/273, 243/245, 208, 165/167, 140/142
270
Photocatalysis, biotransformation
N
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
O N
H 3C O
N
TP-C-301: 3 isomers, hydroxylated DZ (ring “C”)
1.66 2.29 3.39
301.0744
C16H14N2O2Cl
0.0 ppm 0.1 ppm 1.0 ppm
301/303, 273/275, 238, 209, 182, 154/156
Cl
300
N
Photocatalysis
OH
H3C O
N
TP-A-301: 2 isomers, hydroxylated DZ (ring “A”)
TP-C-317: 2hydroxylated DZ
Cl
1.75 2.59
301.0744
C16H14N2O2Cl
2.0 ppm 2.7 ppm
301/303, 273/275, 238, 209, 198, 170/172, 105
300
Photocatalysis
2.28
317.0693
C16H14N2O3Cl
0.9 ppm
317/319, 289/291, 260/262, 254, 225, 179/181, 182/184, 154/156, 123
316
Photocatalysis
N
363
(continued on next page)
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
HO
Accurate mass (calculated) [M H]þ
Elemental composition [M H]þ
Mass error
TP-A/C-317: 2hydroxylated DZ
1.67
317.0693
C16H14N2O3Cl
0.9 ppm
317/319, 289/291, 260/262, 199, 182/184, 105
316
Photocatalysis
TP-303
1.80
303.0900
C16H16N2O2Cl
1.0 ppm
303/305, 246/248, 228/230, 193
302
Biotransformation
MS/MS
MW (nominal)
Treatment conditions
Chemical structure (proposed)
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
LC e tR (min)
Compound/ abbreviation
364
Table 6 e (continued )
Table 7 e Transformation products of OXA: LC retention time (LC e tR), description of mass spectra (HR measurement and tandem mass fragmentation), treatment conditions and proposed chemical structures. Compound/ abbreviation
LC e tR (min)
Accurate mass (calculated) [M H]þ
Elemental composition [M H]þ
Mass error
MS/MS
MW (nominal)
Treatment conditions
2.89
271.0638
C15H12N2OCl
1.8 ppm
271/273, 253/255, 218, 190
270
Biotransformation
TP-A/C-303: 3 isomers, hydroxylated OXA (ring “A” or “C”)
1.75, 1.99, 2.63
303.0536
C15H12N2O3Cl
0.0 ppm, 0.7 ppm, 0.3 ppm
303/305, 285/287, 257/259
302
Photocatalysis at pH2
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
TP-271
Chemical structure (proposed)
or
365
366
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
To further eliminate the residual DZ and OXA, as well as the transformation products, which are recalcitrant to either biological or photochemical degradation, we investigated their sorption to activated carbon (AC). In our experiment we considered the effect of water matrix, exposure time and concentration of AC. It was observed that the both prolonging exposure time (60 min) and increasing the amount of AC significantly increased the removal efficiency. The results are gathered in the Supplementary data.
3.5.
Coupling biological and abiotic water treatment
Removal efficiencies of 99.9% for DZ and OXA may not be satisfactory since detectable concentrations of parent compounds and their transformation products are still present in the effluents. It is important to emphasize that such harsh treatment conditions as presented herein are irrelevant for actual domestic water treatment since the highest determined concentrations of benzodiazepines in Slovene WWTP influents were <200 ng L1, and such treatment would not be cost-effective. However, we believe that concentrations as high as 100 mg L1 (initial concentration) may occur in treatment of WWs discharged from pharmaceutical production units. In this view, and based on the results of biodegradation and abiotic degradation experiments, we coupled both treatments in series to achieve a factor 104 removal in concentration of DZ and OXA. As illustrated in Table 5, we determined the concentration of DZ prior and post biological treatment (cascade D: X2 / A2 / X1 / A1, Table 1), and coupled to UV or AC only, and to both abiotic treatment methods. It can be noted only the sequenced treatment technology that combines all three treatment methods: biological degradation, photochemical treatment and sorption, results in final concentrations in the low ng L1 range. Table 5 shows that during the biological degradation of DZ substantial amounts of OXA are formed, which are afterwards removed by combining UV/H2O2 and AC treatment.
3.6.
Transformation during water treatment
By utilizing the capabilities of the QqTOF mass spectrometer, i.e. tandem mass fragmentation and accurate mass measurement, coupled to an ultra-performance liquid chromatography (UPLC), which shows superior performance in terms of separation ability and speed compared to classical high resolution liquid chromatography (HPLC), and by employing mass spectral algorithm (MetaboLynx) for enhanced detection of transformation products, we determined eleven transformation products (TPs) formed during DZ (bio)degradation (Table 6), whereas four TPs were generated during treatment of OXA (Table 7). As shown in Table 6, the biotransformation of DZ resulted in the formation of OXA, nordazepam, temazepam, which are also known to be human metabolites of DZ and are marketed as individual pharmaceuticals. Nordazepam is formed by N-demethylation on N 1 of DZ, and temazepam by hydroxylation of the C-3 atom, whereas OXA comprises both transformation reactions. In contrast to photochemical transformation reactions, biotransformation is less predictable resulting in less
common transformations such as hydroxylation on double bond (TP-303, Table 6) and loss of oxygen to form TP-271 (Table 7). Photochemical transformation generally results in the formation of ring-hydroxylated TPs, as previously shown for other pharmaceuticals such as ketoprofen (Kosjek et al., 2011). These ring-hydroxylated TPs often appear as structural isomers (compounds TP-C-301, TP-A-301, TP-C-317, TPA/C-317 in Table 6, and TP-A/C-303 in Table 7). Resolving the exact position of the eOH group is difficult by mass spectrometry, whereas NMR analyses can provide the missing information in such examples. However, despite its excellent identification capabilities, the NMR is rarely employed in the environmental analysis, since it requires considerably high amounts of analytes and an efficient separation of complex analyte mixtures. The structures of TPs comprised in Table 6 and Table 7 are proposed basing on comprehensive inspection of their tandem mass spectra, with an aid of HRMS to determine the elemental compositions of protonated analyte molecules and their ion fragments. Yet, to confirm with certainty the chemical structure of the TPs proposed in Table 6 and Table 7, further investigation is needed, which indeed involves the use of authentic standards of the proposed TPs. However, to the best of our knowledge most of the TPs proposed herein (except for the human metabolites: OXA, nordazepam and temazepam) have until now not been recognised, and authentic standards are yet to be synthesized.
4.
Conclusions
In conclusion, biodegradation experiments in flow-through pilot bioreactors reveal significant biomass adaptation during incubation during both oxic and anoxic conditions. Operating reactors in oxiceanoxiceoxiceanoxic cascade provides an additional 10% towards total removal of diazepam. Photochemical treatment of benzodiazepine derivatives at concentrations relevant for pharmaceutical industry effluents resulted in their efficient removal. Despite this, chronic exposure to the residual pharmaceuticals could result in adverse effects. For this reason further treatment by adsorption on activated carbon was applied and resulted in sufficient removal of the residual compounds. Such extensive treatment is unrealistic and more feasible combinations will have to be researched but the described treatment combination does highlight the recalcitrant nature of these compounds. Novel transformation products, eight DZ and four OXA, are reported. But their occurrence and toxicity in the environment are yet to be investigated. Their identification does show that simply reporting the removal efficiencies of just the parent compounds is no longer sufficient and a full assessment of the risks posed by their human metabolites as well as environmental transformation products is required. Finally, as virtually no information on the transformation of benzodiazepine derivatives is currently available, the proposed biotransformation products bring an important contribution to recognising the fate of DZ and OXA during biological WW treatment. Hopefully, because of the increasing attention given to the qualitative determination of pharmaceutical
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
degradation products in the literature and the development of identification tools in general, knowledge about the fate of this particular group of pharmaceuticals in the environment will become more comprehensive.
Acknowledgements The financial support from the Ministry of Higher Education, Science and Technology (Projects Z1-3677 and P1-0143) and Slovenian Technology Agency (Young Researcher in the Economy, Grant P-MR-09/26) is acknowledged. Operation is partly financed by the European Union, European Social Fund. The authors also wish to acknowledge oxazepam donation by Belupo, d.d. (Croatia) as well as WWTP Ptuj (Slovenia) and ef Stefan Ecological Laboratory with Mobile Unit from Joz Institute (Slovenia) for facilitating sampling.
Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.10.056.
references
Baker, D.R., Kasprzyk-Hordern, B., 2011. Multi-residue analysis of drugs of abuse in wastewater and surface water by solidphase extraction and liquid chromatographyepositive electrospray ionisation tandem mass spectrometry. Journal of Chromatography A 1218, 1620e1631. Baldessarini, R.J., 1996. Drugs and the treatment of psychiatric disorders: depression and mania. In: Hardman, J.G., et al. (Eds.), Goodman & Gilman’s the Pharmacological Basis of Therapeutics. McGraw-Hill, New York, pp. 431e459. Bautitz, I.R., Nogueira, R.F.P., 2010. Photodegradation of lincomycin and diazepam in sewage treatment plant effluent by photo-Fenton process. Catalysis Today 151 (1, 2), 94e99. Benzodiazepine Equivalents Table, 2011. T.R.A.P, The Tranquillizer Recovery and Awareness Place. Available from: http://www.non-benzodiazepines.org.uk/equivalents.html (accessed June 22, 2011). Belden, J.B., Maul, J.D., Lydy, M.J., 2007. Partitioning and photodegradation of ciprofloxacin in aqueous systems in the presence of organic matter. Chemosphere 66, 1390e1395. Boreen, A.L., Arnold, W.A., McNeill, K., 2003. Photodegradation of pharmaceuticals in the aquatic environment: a review. Aquatic Sciences 65, 320e341. Borrey, D., Meyer, E., Lambert, W., Van Calenbergh, S., Van Peteghem, C., De Leenheer, A.P., 2001. Sensitive gas chromatographicemass spectrometric screening of acetylated benzodiazepines. Journal of Chromatography A 910 (1), 105e118. Brooks, B.W., Chambliss, C.K., Stanley, J.K., Ramirez, A., 2005. Determination of select antidepressants in fish from an effluent-dominated stream. Environmental Toxicology and Chemistry 24 (2), 464e469. Calisto, V., Esteves, V.I., 2009. Psychiatric pharmaceuticals in the environment. Chemosphere 77 (10), 1257e1274. Fu¨rst, J., Peklar, J., Samaluk, V., 2006. Use of anxiolytics, hypnotics and sedatives in Slovenia. Zdrav Var 45, 150e157.
367
Heberer, T., 2002. Tracking persistent pharmaceutical residues from municipal sewage to drinking water. Journal of Hydrology 266 (3, 4), 175e189. Huerta-Fontela, M., Galceran, M.T., Ventura, F., 2011. Occurrence and removal of pharmaceuticals and hormones through drinking water treatment. Water Research 45 (3), 1432e1442. Johnson, A.C., Ju¨rgens, M.D., Williams, R.J., Ku¨mmerer, K., Kortenkamp, A., Sumpter, J.P., 2008. Do cytotoxic chemotherapy drugs discharged into rivers pose a risk to the environment and human health? An overview and UK case study. Journal of Hydrology 348 (1, 2), 167e175. Kinani, S., Bouchonnet, S., Milan, N., Ricordel, I., 2007. A sensitive and selective method for the detection of diazepam and its main metabolites in urine by gas chromatographyetandem mass spectrometry. Journal of Chromatography A 1141 (1), 131e137. Klavarioti, M., Mantzavinos, D., Kassinos, D., 2009. Removal of residual pharmaceuticals from aqueous systems by advanced oxidation processes. Environment International 35 (2), 402e417. Kosjek, T., Heath, E., 2011. Halogenated heterocycles as pharmaceuticals. In: Iskra, J. (Ed.), Topics in Heterocyclic Chemistry. Springer. Kosjek, T., Heath, E., Kompare, B., 2007. Removal of pharmaceutical residues in a pilot wastewater treatment plant. Analytical and Bioanalytical Chemistry 387 (4), 1379e1387. , M., Barcelo´, D., 2009. Kosjek, T., Heath, E., Pe´rez, S., Petrovic Metabolism studies of diclofenac and clofibric acid in activated sludge bioreactors using liquid chromatography with quadrupole e time-of-flight mass spectrometry. Journal of Hydrology 372 (1e4), 109e117. Kosjek, T., Perko, S., Heath, E., Kralj, B., Zigon, D., 2011. Application of complementary mass spectrometric techniques to the identification of ketoprofen phototransformation products. Journal of Mass Spectrometry 46 (4), 391e401. Lo¨ffler, D., Ro¨mbke, J., Meller, M., Ternes, T., 2005. Environmental fate of pharmaceuticals in water/sediment systems. Environmental Science and Technology 39 (14), 5209e5218. Martı´nez Bueno, M.J., Agu¨era, A., Go´mez, M.J., Hernando, M.D., Garcı´a-Reyes, J.F., Ferna´ndez-Alba, A.R., 2007. Application of liquid chromatography/quadrupole-linear ion trap mass spectrometry and time-of-flight mass spectrometry to the determination of pharmaceuticals and related contaminants in wastewater. Analytical Chemistry 79, 9372e9384. Martı´nez Bueno, M.J., Hernando, M.D., Herrera, S., Go´mez, M.J., Ferna´ndez-Alba, A.R., Bustamante, I., Garcı´a-Calvo, E., 2010. Pilot survey of chemical contaminants from industrial and human activities in river waters of Spain. International Journal of Environmental Analytical Chemistry 90, 321e343. Matilainen, A., Sillanpa¨a¨, M., 2010. Removal of natural organic matter from drinking water by advanced oxidation processes. Chemosphere 80, 351e365. Neumeyer, J.L., Booth, R.G., 1995. Neuroleptics and anxiolytic agents. In: Foye, W.O., et al. (Eds.), Principles of Medicinal Chemistry. Williams & Wilkins, Philadelphia, pp. 199e231. Patterson, B.M., Shackleton, M., Furness, A.J., Pearce, J., Descourvieres, C., Linge, K.L., Busetti, F., Spadek, T., 2010. Fate of nine recycled water trace organic contaminants and metal(loid)s during managed aquifer recharge into a anaerobic aquifer: column studies. Water Research 44 (5), 1471e1481. Patterson, B.M., Shackleton, M., Furness, A.J., Bekele, E., Pearce, J., Linge, K.L., Busetti, F., Spadek, T., Toze, S., 2011. Behaviour and fate of nine recycled water trace organics during managed aquifer recharge in an aerobic aquifer. Journal of Contaminant Hydrology 122 (1e4), 53e62.
368
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 5 5 e3 6 8
Pereira, V.J., Linden, K.G., Weinberg, H.S., 2007. Evaluation of UV irradiation for photolytic and oxidative degradation of pharmaceutical compounds in water. Water Research 41, 4413e4423. Slovene Guideline on The Quality of WWTP Discharge. Available from: http://www.uradni-list.si/files/RS_-2007-045-02451OBwP002-0000.PDF (accessed June 28, 2011). Riss, J., Cloyd, J., Gates, J., Collings, S., 2008. Benzodiazepines in epilepsy: pharmacology and pharmacokinetics. Acta Neurologica Scandinavica 118 (2), 69e86. Suarez, S., Lema, J.M., Omil, F., 2010. Removal of pharmaceutical and personal care products (PPCPs) under nitrifying and denitrifying conditions. Water Research 44 (10), 3214e3224. Ternes, T., Bonerz, M., Schmidt, T., 2001. Determination of neutral pharmaceuticals in wastewater and rivers by liquid
chromatographyeelectrospray tandem mass spectrometry. Journal of Chromatography A 938 (1, 2), 175e185. Ternes, T.A., Janex-Habibi, M.L., Knacker, T., Kreuzinger, N., Siegrist, H., 2004. Detailed report related to the overall project POSEIDON (contract no. EVK1-CT-2000-00047) duration (2001e2004). Assessment of technologies for the removal of pharmaceuticals and personal care products in sewage and drinking water facilities to improve the indirect potable water reuse. Verlicchi, P., Galletti, A., Petrovic, M., Barcelo´, D., 2010. Hospital effluents as a source of emerging pollutants: an overview of micropollutants and sustainable treatment options. Journal of Hydrology 389, 416e428. Zuccato, E., Calamari, D., Natangelo, M., Fanelli, R., 2000. Presence of therapeutic drugs in the environment. The Lancet 355, 1789e1790.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 6 9 e3 7 7
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effects of microcystin-LR on the metal bioaccumulation and toxicity in Chlamydomonas reinhardtii Ning-Xin Wang, Xue-Ying Zhang, Jun Wu, Lin Xiao, Ying Yin, Ai-Jun Miao*, Rong Ji*, Liu-Yan Yang State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing, Jiangsu Province 210093, PR China
article info
abstract
Article history:
Microcystin-LR (MC-LR) is one of the most notorious toxins liberated from cyanobacteria in
Received 5 July 2011
eutrophicated freshwater ecosystems. Its effects on the bioaccumulation and toxicity of
Received in revised form
2þ 2þ in a green alga Chlamydomonas reinhardtii were investigated in Cd2þ, CrO2 4 , Cu , and Zn
25 September 2011
the present study. The metal bioaccumulation in the alga was unaffected by MC-LR. The
Accepted 17 October 2011
surface-adsorbed and intracellular metal concentrations in the treatments with and without
Available online 25 October 2011
the addition of MC-LR could be well simulated by a single Freundlich isotherm for each metal
Keywords:
bioavailable metal concentrations measured by diffusion gradients in thin-films remained
Bioaccumulation
unchanged when MC-LR was applied. Accordingly, the growth of C. reinhardtii was similarly
Microcystin-LR
inhibited at the same metal concentration regardless of the addition of MC-LR. The metal
Toxicity
toxicity could also be well delineated with the classic free ion activity and biotic ligand
Trace metal
models. However, the intracellular metal concentration was found to have the best
with their accumulation ability following the order of Cu2þ > Cd2þ > Zn2þ > CrO2 4 . The
predictability suggesting its more direct relationship with metal toxicity. Metal exposure induced the accumulation of MC-LR in the alga, which was leveled off at high metal levels. The underlying uptake mechanisms need to be further examined. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Heavy metals with density higher than 5 g/cm3 are widely used in various industries and a significant fraction would inevitably get to the environment (Fu and Wang, 2011). Substantially high concentrations of cadmium, chromium, copper, lead, mercury, and zinc, belonging to the thirteen priority metal/metalloid pollutants categorized by US EPA, have been found in lakes, rivers, estuaries and coastal areas, where harmful algal blooms frequently occur as a result of eutrophication (Diao et al., 2004). It has been reported that algae in bloom may take up abundance of metals, float with water current, and further strikingly influence the metal
distribution in aquatic environments (Reynolds and Hamiltontaylor, 1992). On the other hand, various organic compounds (e.g., exopolymeric substances and toxins) could be released from the algae especially during the period of their decomposition, which may further play an important role in the metal bioavailability and toxicity to aquatic organisms (Haye et al., 2006). As one of the most prevalent cyanobacterial toxins, microcystins are small monocyclic hepapeptides sharing a general structure of five identical but two variable L-amino acids in positions 2 and 4 (Babica et al., 2006). They are secondary metabolites produced by enzymes like microcystin, peptide or polyketide synthetases. Microcystins are specific
* Corresponding authors. E-mail addresses:
[email protected] (A.-J. Miao),
[email protected] (R. Ji). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.035
370
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 6 9 e3 7 7
inhibitors of eukaryotic protein phosphatases 1 and 2A, thus could increase the protein phosphorylation in the cells, destruct their cytoskeleton and further deregulate the cell division. They may also incur oxidative stresses, disrupt the osmotic and ion regulation in the organisms (Martins and Vasconcelos, 2009). Most of the current microcystin researches focus mainly on their behavior and fate in the environment as well as their accumulation, distribution, metabolisms, and toxicity in various organisms. However, their potential functions in cyanobacteria remain largely unknown. Some putative roles of microcystins have been proposed such as the allelophathic effects on other algae/plants, cyanobacteria protection from their grazers, light absorption improvement (quorum sensing effects), toxic metal sequestration in the cells, and metal nutrient acquisition from the ambient environment (Hesse et al., 2001; Utkilen and Gjolme, 1995). The binding characteristics of microcystins with Cd2þ, Cu2þ, Hg2þ, Pb2þ, and Zn2þ have been measured by differential pulse polarography, cyclic and anodic stripping voltammetry with intermediate affinity observed (Humble et al., 1997; Yan et al., 2000). Several microcystinemetal complexes (e.g., microcystin-Fe (II), -Zn, -Cu, -Mg etc.) were also detected using cryospray ionization-Fourier transform ion cyclotron resonance mass spectrometry (Saito et al., 2008). Zeng et al. (2009) compared the metal tolerance of a non-toxic and toxic strain of Microcystis aeruginosa, finding that the non-toxic strain was more sensitive to Cd2þ but both strains had similar tolerance to Zn2þ. It implies that microcystins may be involved in metal detoxification in a metal-specific manner. In contrast, the cellular microcystin concentration was found to be concomitant with the specific growth rate of M. aeruginosa rather than being triggered in response to Zn2þ and Cu2þ exposure, purporting a constitutive production of microcystins (Gouvea et al., 2008). To further investigate how microcystins and trace metals may interact with each other, the most toxic and frequently observed microcystin, microcystin-LR (MC-LR), was chosen in the present study. The bioaccumulation and toxicity of the 2þ 2þ four metal ions, Cd2þ, CrO2 4 , Cu , and Zn , in a classic freshwater green alga Chlamydomonas reinhardtii were compared with and without the addition of 1 mM MC-LR. Although mainly located inside the cells, a substaintial amount of MC-LR could be released into the ambient environment with the dissolved concentration varying from traces up to 1800 mg/l (approximately 1.8 mM) or higher, immediately after the collapse of a highly toxic bloom (Babica et al., 2006). According to Pearson’s hard soft acid base principle (Pearson, 1997), Cd2þ is soft and would rather bind with sulfur containing functional groups than oxygen or nitrogen ones. Cu2þ and Zn2þ belong to the borderline group with intermediate affinity to all the three types of binding ligands above, whereas CrO2 4 has negative charges. Different effects of MC-LR (if any) on the bioaccumulation and toxicity of these metals with different physicochemical properties would thus be expected. As the natural dissolved organic compounds (e.g, humic substances) are well documented to affect the metal speciation, accumulation and further toxicity in various organisms, it would be interesting to illuminate how MC-LR may be involved in the metaleorganism interactions (Koukal et al., 2003; Lamelas and Slaveykova, 2007; Vigneault et al., 2000).
Both the free ion activity model (FIAM) and the biotic ligand model (BLM), in which the metal toxicity is predicted either by its free ion concentration in the media or by its adsorption on the biotic ligands (e.g., fish gills or cell membrane), have now been widely accepted (Campbell, 1995; De Schamphelaere et al., 2005; Slaveykova and Wilkinson, 2005). However, a number of exceptions were reported as both models are based on several assumptions and the metal toxicity was thus further related to their internal concentrations in the organisms or their subcellular distribution (Miao and Wang, 2007). In the present study, the metal bioaccumulation and their inhibition of algal growth were plotted against four types of metal concentrations (i.e., the total dissolved metal concentration [M]T, the free metal ion concentration [M]F, the surface-adsorbed ([M]ads) and intracellular metal concentrations ([M]intra)) to elicit whether the different models could still be applied in the presence of MC-LR. The overall objective is to examine whether the microcystins liberated from cyanobacteria could affect the metal bioavailability and toxicity to aquatic organisms.
2.
Materials and methods
2.1.
Phytoplankton culture conditions
An axenic culture of the Chlorophyta C. reinhardtii used in the present study was originally obtained from the Institute of Hydrobiology, Chinese Academy of Sciences, Wuhan. The algal cells were maintained in an artificial freshwater WC medium (Guillard, 1975) with its pH kept at 7.5 0.1 by the addition of 5 mM 3-(N-morpholino) propanesulfonic acid (MOPS). The temperature was 25 C with a light illumination of 50 mmol photons/m2/s in a 12:12 LighteDark cycle.
2.2. Effects of MC-LR on metal bioaccumulation and toxicity MC-LR (formula: C49H74N10O12, molecular weight: 995.2, purity: >95%) was bought from the Express Bio-technology Co., Ltd., Beijing, China. There were two toxicity tests (i.e., one with and the other without the addition of 1 mM MC-LR) for each of the four metals with seven concentration treatments in 150 ml duplicates on average. MC-LR alone at the abovementioned concentration had no notable effects on C. reinhardtii. A modified WC medium was used as the base of the toxicity media. Its components are shown in Table S1. Since a considerable amount of ethylenediaminetetraacetic acid (EDTA, 11.7 mM) was present in the original WC medium which could possibly hide the impacts of MC-LR, it was excluded from the toxicity media. The trace metal nutrient concentrations were thus reduced correspondingly to avoid their unnecessary precipitation (e.g., Fe3þ) or toxicity (e.g., Cu2þ) to the alga. All the containers were soaked in 1 N HCl and then rinsed with Milli-Q water (18.2 MU) for at least six times. Trace metal clean technique was applied throughout the whole experiment. In our preliminary experiment, the dissolved metal concentrations were found to decrease remarkably as a result of their adsorption onto the container wall in the absence of
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 6 9 e3 7 7
EDTA or other strong metal binding ligands. Therefore, each flask or beaker to be used was pre-equilibrated with the corresponding solution containing the same components as the toxicity medium of the following experiment. After the solution for pre-equilibration was decanted, another batch of the same medium was poured in and left overnight to equilibrate before each toxicity test. Right before the algal addition to the toxicity media on the second day, 1 ml aliquot from each container was withdrawn and evaporated to dryness. Then it was digested in 1 ml concentrated HNO3 (ultrapure) under 80 C and further diluted with the final HNO3 concentration 7% w/v. Trace metals were quantified by Thermo M6 atomic absorption spectrophotometer equipped with a GF95Z graphite furnace system (GFAAS, Thermo Fisher Scientific Inc., Waltham, MA, USA). [M]F of the four trace metals were then calculated using the MINEQLþ software package (Version 4.5 from Environmental Research Software, Hallowell, ME, USA) with thermodynamic constants updated and ionic strength calibrated. The conditional stability constants for MC-LR with different metals were obtained from the literature (Humble et al., 1997; Yan et al., 2000). [M]T and [M]F are given in Table S2. The algal cells were first acclimated in the modified WC medium under the same conditions as the following toxicity experiments except that no extra metals (i.e., Cd2þ, CrO2 4 , Cu2þ, Zn2þ) or MC-LR were added. After arriving at the middleexponential growth phase, the cells were collected, rinsed with the fresh medium and then resuspended into the different toxicity media. The initial cell density was in the range of 8 104 to 1 105 cells/ml. For each metal, the algal cells from the same batch of culture were used in the two toxicity tests with or without MC-LR to eliminate any unpredictable batch-specific differences. The whole toxicity tests lasted two days during which the cells were enumerated every 24 h. The cell specific growth rate was calculated as described in Miao et al. (2005). At the end of the experiment, 10 ml aliquot from each replicate was filtered through a 0.22 mm polycarbonate membrane (Millipore). The <0.22 mm filtrate was collected for the measurement of [M]T. Metals weakly adsorbed on the cell surface ([M]ads) were removed with 10 ml 100 mM EDTA (Hassler et al., 2004). The algal cells retained on the membrane were further digested with concentrated HNO3 and [M]intra was then quantified. In the meantime, the amount of metals adsorbed on the wall of the filtration bottle and the total metal concentration in 1 ml aliquot without any filtration were measured for mass balance calculation. In the toxicity tests with the addition of 1 mM MC-LR, another 10 ml sample from each replicate was filtered through a combusted GF/F membrane. The filtrate was collected and the filter was further rinsed with fresh modified WC medium without any MC-LR. Both the filter and filtrate were stored under 20 C for further MC-LR measurement. MC-LR either accumulated in the algal cells or left in the dissolved phase was extracted and purified following a similar procedure as described by Ramanan et al. (2000). Briefly, the cell-containing GF/F membrane was cut into pieces and soaked in 25 ml 75% v/v methanol aqueous solution with continuous stirring for 1 h at room temperature. The procedure above was repeated and extractants combined from both
371
steps were reduced to 10 ml by rotary evaporation at 35 C. A 10 ml aliquot from the cell extract or the GF/F filtrate obtained above was passed through a preconditioned Waters Sep-Pak C18 column (0.5 g/3cc) at 1 ml/min to remove impurities. The cartridge was then washed with 20% aqueous methanol and MC-LR was eluted by 5 ml of 80% methanol at 1 ml/min. The eluent was further concentrated and evaporated to dryness. After that, the residue was either redissolved in 2 ml 50% methanol or 1 ml Milli-Q water depending on the following MC-LR quantification method. In the samples (GF/F filtrates) where the MC-LR concentrations were between 0.1 and 10 mM, HPLC was used for the analysis (Song et al., 2007). The MC-LR concentrations in the other samples (cell extractants) were determined through an ELISA kit (the Institute of Hydrobiology, Chinese Academy of Sciences, Wuhan, China) with a linear concentration range from 0.2 to 4 nM. For the HPLC method, an Agilent ZORBAX Eclipse XDB-C18 column (4.6 250 mm) under 40 C was applied to separate the microcystins, which were then monitored by an Agilent diode-array detector at 238.8 nm. The mobile phase was made up of 43% 50 mM KH2PO4 (pH ¼ 3) and 57% pure methanol. The elution time was 20 min with an injection volume 20 ml.
2.3. Metal bioavailability measurement with diffusion gradients in thin-films (DGT) DGT technique was adopted to further examine whether MCLR could influence the metal bioavailability (Zhang and Davison, 2000). There was only one concentration treatment in duplicate for each metal (i.e., 0.89, 7.70, 0.79, and 3.06 mM for 2þ 2þ Cd2þ, CrO2 4 , Cu , and Zn ) with and without the addition of MC-LR. The metal concentrations were at the same order of magnitude as that of MC-LR (1 mM). Any possible effects from the microcystin could be clearly shown in this manner. The concentrations of the other components were the same as those in the toxicity media. The DGT setup was deployed in a polypropylene beaker containing 1 L medium for each replicate as described by Ernstberger et al. (2002). After being stirred for 4 h under 25 C, it was taken out, disassembled and the chelex resin membrane inside was soaked in 1 ml diluted HNO3 (1 N) for another 4 h. The concentration of the objective metal released into the acid was determined with GFAAS. Additionally, the total dissolved metal concentrations of the media for DGT deployment were quantified at the beginning and the end to ensure that they remained unchanged during the experimental period.
2.4.
Statistical analysis
Median (50%) effect concentrations (EC50s) based on the different types of metal concentrations ([M]T, [M]F, [M]ads, and [M]intra) were obtained by simulating the corresponding doseeresponse curves with the Logistic model (y ¼ min þ {maxmin/[1 þ (x/EC50)Hillslope]}). Any ‘significant’ difference (accepted at p < 0.05) was based on results of oneway or two-way analysis of variance with post-hoc multiple comparisons (Turkey or Tamhane) (SPSS 11.0 by SPSS, Chicago, USA). The normality (KolmogoroveSmirnov and ShapiroeWilk tests) and homogeneity of variance (Levene’s test)
372
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 6 9 e3 7 7
was taken up intracellularly for the other treatments. Furthermore, the bioaccumulation (including both intracellular absorption and cell surface adsorption) ability of the different metals in C. reinhardtii was found to follow the order Cu2þ > Cd2þ > Zn2þ > CrO2 4 in the concentration range of the present study. Therefore, [M]F for [M]intra ([M]ads) to achieve 104 ng/cell was 0.079 (0.20), 1.29 (0.91), 3.14 (1.46), and 26.8 (4.56) mM for Cu2þ, Cd2þ, Zn2þ, and CrO2 4 , respectively, as predicted by the Freundlich isotherm. The accumulation of the above four metals by C. reinhardtii was investigated before with similar results obtained (Sunda and Huntsman, 1998; Wang and Dei, 2006; Zerhouni et al., 2004). Only the total cellular metal concentrations, including those adsorbed on the cell surface, were quantified in most studies. The cellular Cd2þ (Cu2þ) concentrations in the same strain of alga also increased approximately from 1.35 104 ng/cell (2.54 104 ng/cell) when [Cd2þ]F ([Cu2þ]F) was 0.048 mM (0.001 mM) to 6.74 104 ng/cell (1.02 103 ng/ cell) with [Cd2þ]F ([Cu2þ]F) 30.0 mM (8.3 mM) (Wang and Dei, 2006). In the meantime, the intracellular accumulation of Cu2þ in the green alga Scenedesmus subspicatus was compared with their surface adsorption by Knauer et al. (1997). Only a small proportion (20%) of the total cellular Cu2þ was allocated on the surface, whereas the main fraction (80%) was inside the cells. Higher [Zn2þ]ads than [Zn2þ]intra was also observed by the same authors with [Zn2þ]F of the same range as the present study. In addition, more Cd2þ was surfaceadsorbed than what was internalized by Chlorella kesslerii when [Cd2þ]F was higher than 0.1 mM (Lamelas and Slaveykova, 2007). On the other hand, the uptake of various metals or metalloids was evaluated in several algae including C. reinhardtii by Mahan et al. (1989) with the biosorption
of the data were examined when performing the analysis of variance.
3.
Results and discussion
3.1. Surface adsorption and internalization of different metals In the present study, a good mass balance was observed (100 10%) as comparing the sum of [M]T, [M]ads and [M]intra to the total metal concentrations in the media without filtration at the end of each toxicity test. A negligible amount of metals was adsorbed on the filtration bottle and there was less than 30% reduction in [M]T as a result of cellular uptake during the 2-day period for most experiments. [M]F thus calculated was based on the initial [M]T. Both [M]intra and [M]ads increased linearly with that of [M]F in the log-scale for each of the four metal ions (Fig. 1 and Fig. S1) as followed the Freundlich isotherm (Miao et al., 2005), log ½Mintra or½Mads ¼ 1=nlog ½MF þk
(1)
where k and n are empirically determined constants related to the maximum complexing capacity and binding affinity, respectively. The simulation results are given in Table S3. More Cd2þ and Zn2þ but less Cu2þ were adsorbed on the cell surface as compared to those inside the cells with up to sixfold difference in most of the concentration treatments. 2 Contrastively, less CrO2 4 was internalized at higher ½CrO4 F and the trend was reversed at lower metal levels. Accordingly, 2 ½CrO2 4 ads was approximately 3e8% of ½CrO4 intra in the control treatment but was then 1.8e18 time higher than what 1e-2
1e-3
-MC +MC
1e-3
1e-4 1e-4 1e-5
[M] intra (ng/cell)
1e-5 Cd
2+
2-
CrO4
1e-6 1e-8
1e-7
1e-6
1e-5
1e-4
1e-3
1e-8
1e-7
1e-6
1e-5
1e-4
1e-1 1e-2
1e-4 1e-3 1e-5 1e-4 Cu
1e-6 1e-8
1e-7
2+
Zn
1e-5 1e-6
1e-5
2+
1e-4
[M] F (M) Fig. 1 e The increase in the intracellular metal concentration ([M]intra, ng/cell) with that of free metal ion ([M]F, M) for Cd2D, 2D , and Zn2D, respectively, in the toxicity experiments with (DMC) and without (LMC) the addition of 1 mM MC-LR. CrO2L 4 , Cu Dashed lines are the simulation of the linear correlation between [M]F and [M]intra in the log-scale with the Freundlich isotherm. Data presented as mean ± SD (n [ 2).
373
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 6 9 e3 7 7
1e-5 -Total -DGT
Metal Conc. (M)
affinity following the order of Pb, Fe, Cu, Cd, Zn, Mn, Mo, Sr, Ni, V, Se, As, Co from high to low, which was consistent with our was not investigated in their finding here. Although CrO2 4 study, its affinity should be located at the lower end considering the weak biosorption of anions like As, Se, and Co (Quintelas et al., 2009). More importantly, there was no significant difference ( p > 0.05) in metal internalization or adsorption between the treatments with and without the addition of MC-LR at similar [M]F. Their increase with that of [M]F could thus be simulated with the same Freundlich isotherm for each of the four metals. Similar trends were obtained even when [M]intra and [M]ads were plotted against [M]T (data not shown). Theoretically, MC-LR might affect metal bioaccumulation in several ways. First, given its point of zero charge in acidic range, MCLR was in anionic form at neutral pH. Therefore, it may be complexed with metallic cations and further alter their bioavailability (Humble et al., 1997). Second, MC-LR may possibly increase the membrane permeability or phagocytic ability of algal cells and thus induce their metal accumulation (Dixon et al., 2004; Sieroslawska et al., 2007). Third, MC-LR could be adsorbed on the cell surface and then either block the metal binding sites (decrease its bioaccumulation) or form ternary complexes (increase the uptake) as similar to the effects of humic substances on Pb2þ bioaccumulation (Koukal et al., 2003; Lamelas and Slaveykova, 2007). Furthermore, the MC-LR concentration used in the present study may be too low or the microcystinemetal complexes formed may be too labile to incur any notable effects on metal bioavailability as well as its bioaccumulation. When [M]F with and without the addition of MC-LR was compared through the metal speciation software MINEQLþ, no substantial difference was observed excluding the first possibility above. The second and third possibilities were also less likely considering the unchanged growth of C. reinhardtii in the presence of 1 mM MC-LR, which was further supported by the metal accumulation results. To investigate the lability of the possible microcystinemetal complexes and further to elicit how metal bioavailability may change when the concentrations of MC-LR and metals are comparable to each other (the fourth possibility above), DGT technique was applied. [M]T was kept relatively constant 2þ 2þ during the 4-h period with [Cd2þ]T, ½CrO2 4 T , [Cu ]T and [Zn ]T in the range of 0.80e0.77 mM, 7.78e8.16 mM, 0.78e0.75 mM, and 3.50e3.55 mM, respectively, as were similar to the nominal concentrations of each metal. Most of the Cd2þ (81.9e84.8%) and Zn2þ (89.9e94.8%) initially added were bioavailable (Fig. 2). In contrast, there were only 39.4e44.0% and 3.34e3.02% of the total dissolved Cu2þ and CrO2 4 that were labile based on the DGT results. Although the concentrations of the four metals and MC-LR were of the same order of magnitude, there was no significant difference ( p > 0.05) for the bioavailable metal concentrations in treatments with and without the addition of MC-LR. It suggests that the metal-MC-LR complexes possibly formed in this situation were quite labile and could thus be concentrated in the chelex resin of DGT. The metal species determined by DGT include the free metal ion, the inorganic and labile organic metal species. Therefore, there is no surprise that most of the Cd2þ and Zn2þ were bioavailable as was consistent with what was calculated by MINEQLþ in the toxicity tests (Table S2). The solubility of
+Total 5e-6
+DGT
0 Cd
2+
2+
2-
CrO4
Cu
2+
Zn
Metals Fig. 2 e Comparison of the total dissolved metal concentrations (DTotal and LTotal) with those of the bioavailable metal (DDGT and -DGT) as measured by DGT for 0.89, 7.70, 0.79, and 3.06 mM total concentrations of 2D , and Zn2D with (DTotal and DDGT) and Cd2D, CrO2L 4 , Cu without (LTotal and LDGT) the addition of 1 mM MC-LR. Data presented as mean ± SD (n [ 2).
Cu2þ in the modified WC media is approximately 0.40 mM. Therefore, most of the Cu2þ precipitated out with only one third of them directly bioavailable. Although 91.7e98.9% of the was present as free ions (Table S2), its bioavailable CrO2 4 concentration was quite low. It implies that CrO2 4 as an anion could not accumulate in the cation-exchange chelex resin. Its concentration in DGT was simply determined by the diffusional equilibrium between the hydrogel and the bulk solution as described by Ernstberger et al. (2002). The potential effects of various organic compounds on metal accumulation were widely investigated. Some organic ligands like EDTA could strongly bind with metals, decrease their free ion concentrations and further their surface adsorption as well as intracellular accumulation (Campbell, 1995). Other compounds such as citric acid could form lipophilic complexes with various metals, which could then passively diffuse across the cell membrane and facilitate their uptake (Campbell, 1995). In addition, dual effects were observed for humic substances, which could either protect the organisms from metal exposure by coating the cell surface or increase the metal uptake by forming ternary complexes on the membrane or altering its permeability (Koukal et al., 2003; Lamelas and Slaveykova, 2007; Vigneault et al., 2000). Although MC-LR could be complexed with various metals, their affinity was not high enough to trigger any considerable effects on metal bioaccumulation. Further, it could be speculated that MC-LR had no remarkable effects on the membrane permeability of C. reinhardtii at least at the concentration used in the present study.
3.2.
Accumulation of MC-LR in C. reinhardtii
MC-LR in the dissolved phase of the toxicity media remained constant during the 2-day period. Its concentrations were
374
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 6 9 e3 7 7
0.90e1.06, 0.73e0.95, 0.87e0.96, and 0.85e1.12 mM at the end of 2þ and Zn2þ toxicity tests with 1 mM MC-LR the Cd2þ, CrO2 4 , Cu applied initially (Table S4). There was a considerable bioaccumulation of MC-LR (Fig. 3), which increased proportionally with that of metal concentration first and stayed stabilized or even diminished after the maximum was arrived at. Accordingly, the cellular MC-LR content went up from 1.34 106 ng/cell in the control treatment to 2.99 106 ng/ cell when the [Cd2þ]F was 6.46 mM and decreased to 2.47 106 ng/cell in the highest concentration treatment. Furthermore, substantial difference in the cellular MC-LR 2þ 2þ toxicity contents between the Cd2þ, CrO2 4 , Cu , and Zn tests was observed even in the control treatments, where 1.34 106, 3.78 107, 6.62 107, and 2.59 107 ng/cell MCLR were accumulated. It implies that there may be a batch-tobatch difference in the intracellular chemical composition or in the cell surface charge that could considerably influence the uptake of MC-LR (Mohamed, 2008). The bioaccumulation of MC-LR in organisms at different trophic levels was widely reported with relatively few in plants especially in microalgae. Singh et al. (2001) used the 14C-labeled microcystin to trace its uptake kinetics by two cyanobacteria (Anabaena BT1 and Nostoc muscorum) and up to 30% of the toxin was taken up by 2 h. The internalization of microcystin was higher under the light condition suggesting that an active transport process may be involved. In another study by Mohamed (2008), the cellular MC-LR contents were found to be in the range of 23e51 ng/g dry weight for the two freshwater green algae Chlorella vulgaris and Scenedusmus quadricauda after 3-day exposure to 1, 10, and 100 mg/l MC-LR. They were lower than what were obtained in the present study possibly due to the higher MC-LR concentration we used. The possibility also exists that there might be an algal species-specific uptake of MC-LR. However, our results were consistent with the hypothesis that algae, aquatic plants, zooplankton, bivalves, and fish could retain similar levels of toxins in the range of 0.1e100 mg/g (Babica et al., 2006; Martins and Vasconcelos, 2009).
Cd
Cu
2+
4e-6
Cellular MC-LR (ng/cell)
2+
CrO4
Zn
2+
2+
2e-6
0 A
B
C
E D Treatments
F
G
H
Fig. 3 e The cellular MC-LR concentrations (ng/cell) in treatments A-H with metal concentrations from low to 2D , and Zn2D toxicity tests, high for Cd2D, CrO2L 4 , Cu respectively. The total dissolved metal concentrations in each treatment are given in Table S2. Data presented as mean ± SD (n [ 2).
Based on its chemical structure and amino acid composition, MC-LR is a spatially large molecule with high hydrophilicity. Hence, it is incapable of crossing the cell membrane via passive diffusion, but rather gets into the cells through specific transporters. The organic anion/bile acid transporting protein (OATP) is well known to be involved in the MC-LR uptake by hepatocytes since the compounds (rifampin, cholate, taurocholate etc.) that could block or compete with bile acid uptake were found to remarkably reduce the toxicity of MC-LR in various animal cells (Fischer et al., 2010). The substantial bioaccumulation of MC-LR in microalgae and plants as observed in the present study and previous researches further suggests that other microcystin transport systems may exist, which needs to be further investigated (Wiegand and Pflugmacher, 2005). The enhancement of MC-LR uptake by trace metal exposure might be caused by the increased membrane permeability or be because of the upregulation of microcystin transporter synthesis under this condition. Although microalgae with high specific surface area are the most likely target of potential microcystin allelophathy, relatively few research was performed with both growth stimulation and inhibition reported (Babica et al., 2006). A considerable bioaccumulation of MC-LR in C. reinhardtii was found. However, its cell growth was still comparable to that cultured in the medium without any addition of MC-LR (data not shown), suggesting a species-specific toxicity of MC-LR.
3.3.
Inhibition of cell growth and EC50s
Relative changes of the cell specific growth rate m were plotted against four different types of metal concentrations (i.e., [M]T, [M]F, [M]ads, and [M]intra), as shown in Figs. 4 and 5 and Figs. S2 and S3, to investigate which of them can predict the metal toxicity best and further to examine whether MC-LR altered the metal toxicity to C. reinhardtii. In all the eight toxicity tests with and without the addition of MC-LR, the cell growth remained unchanged as [M]F increased from those in the control treat2þ ments to 2.99, 5.72, 0.036, and 10.0 mM for Cd2þ, CrO24 , Cu , and 2þ Zn , respectively. Thereafter, it went down until the algal cells completely died when [M]F was 5.32, 29.5, 0.13, and 43.1 mM. Similar doseeresponse relationship was observed for the other types of metal concentrations. Although there was no notable difference in the cell growth between the treatments with and without the addition of MC-LR at similar metal levels for most metal concentration types, [M]intra was the best to predict metal toxicity under different conditions. Such phenomenon was clearer when the Logistic model was used to simulate the growth inhibition at different metal concentrations with the highest r2 obtained when the growth inhibition was correlated with [M]intra (Table 1). Furthermore, the toxicity of the four > Zn2þ metals followed the order of Cu2þ > Cd2þ > CrO2 4 according to their [M]F based EC50s. Similar toxicity order was reported in other studies with different freshwater algal species such as C. reinhardtii, Chlorella sp. and S. quadricauda (Fargasova, 1999; Franklin et al., 2002; Szivak et al., 2009). 2þ > Cd2þ > Zn2þ) However, the order changed a bit (CrO2 4 > Cu when [M]intra based EC50s were applied, suggesting that the less toxicity of CrO2 4 obtained in the first case was mainly a result of its lower ability to be taken up by the cells. The pH effects on Cu2þ toxicity to C. reinhardtii were investigated by De Schamphelaere and Janssen (2006) with
375
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 6 9 e3 7 7
Cd
2+
2-
CrO 4
1.0
Relative changes of µ
0.5 -MC +MC
0.0 1e- 8
1e- 7
1e- 6
1e- 5
1e-4
Cu
1e- 7
1e- 6
1e- 5
2+
1e-4
Zn
2+
1.0
0.5
0.0 1e- 8
1e- 7
1e- 6 [M]F (M)
1e- 5
1e-4
Fig. 4 e The relative changes of the cell specific growth rate m under different free metal ion concentrations ([M]F, M) for Cd2D, 2D , and Zn2D, respectively, in the toxicity experiments with (DMC) and without (L MC) the addition of 1 mM MCCrO2L 4 , Cu LR. Dashed lines are the simulated doseeresponse curves with the Logistic model. Data presented as mean ± SD (n [ 2).
Cd
2+
2-
CrO4
1.0
Relative changes of µ
0.5 -MC +MC
0.0 1e-6
1e-5
1e-4
1e-3 Cu
1e-5
1e-4
2+
Zn
2+
1.0
0.5
0.0 1e-6
1e-5
1e-4
1e-3
1e-4
1e-3
1e-2
[M] intra (ng/cell) Fig. 5 e The relative changes of the cell specific growth rate m under different intracellular metal concentrations ([M]intra, ng/ 2D , and Zn2D, respectively, in the toxicity experiments with (DMC) and without (LMC) the addition cell) for Cd2D, CrO2L 4 , Cu of 1 mM MC-LR. Dashed lines are the simulated doseeresponse curves with the Logistic model. Data presented as mean ± SD (n [ 2).
376
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 6 9 e3 7 7
Table 1 e The median (50%) effect concentrations (EC50s) based on the total dissolved metal concentration ([M]T, M), the free metal ion concentration ([M]F, M), the cell surface-adsorbed ([M]ads, ng/cell) and intracellular accumulated ([M]intra, ng/cell) 2D metal concentrations for the cell specific growth rate of Chlamydomonas reinhardtii exposed to Cd2D, CrO2 , and Zn2D, 4 , Cu respectively. Data presented as mean ± SD (n [ 2). [M]T-EC50s
Metal ions Cd
2þ
CrO2 4 Cu2þ Zn2þ
6
4.42 10 3.00 r2 ¼ 0.96, p < 0.01 1.21 105 1.45 r2 ¼ 0.93, p < 0.01 2.22 107 1.83 r2 ¼ 0.96, p < 0.01 1.76 105 4.94 r2 ¼ 0.94, p < 0.01
10
[M]F-EC50s 7
106 108 107
6
4.12 10 2.84 r2 ¼ 0.96, p < 0.01 1.11 105 1.36 r2 ¼ 0.93, p < 0.01 7.43 108 6.12 r2 ¼ 0.96, p < 0.01 1.65 105 4.69 r2 ¼ 0.94, p < 0.01
[Cu2þ]F based EC50 (72 h) around 0.032 mM which was close to 0.074 mM obtained in the present study at similar pH (7.5). The cell growth of the same algal species was inhibited by 50% when [Zn2þ]F was 15.8 mM during a 5-day exponential growth period (Knauer et al., 1997) as also agreed well with the values given in Table 1. In another Cd2þ and Cu2þ toxicity tests (48 h) with the same strain of C. reinhardtii (Wang and Dei, 2006), the [M]F based EC50s of Cd2þ (12.6 mM) and Cu2þ (15.8 mM) were much higher than what were observed here. It was possibly due to the less sensitivity of the toxicity endpoint they used (i.e., maximum photosynthetic system II quantum yield) or because of the different intracellular metal bioaccumulation at different phosphate regime. No chromium data was available for C. reinhardtii in the literature. However, the growth of the green alga Scenedesmus acutus was half inhibited by 48 mM (Gorbi et al., 2007), which was at the same order of CrO2 4 magnitude as that in the present study (12.1 mM). Although the four different types of metal concentrations were all able to predict metal toxicity more or less, [M]intra was the best. It suggests that [M]intra was more directly related to metal toxicity as compared to the other three types of metal concentrations and thus had a higher reliability in metal toxicity prediction. Internal copper was also found to be a better toxicity predictor than the free Cu2þ ion activity when pH was varied in the freshwater green microalgae Chlorella sp. and Pseudokirchneriella subcapitata (De Schamphelaere et al., 2005). More importantly, the presence of MC-LR did not change the toxicity of the four metals. Neither did it affect the metal bioaccumulation. It implies that MC-LR could not influence the metal toxicity through other ways than metal speciation alteration at least for the alga used in the present study.
4.
Conclusions
Overall, the linear correlation between [M]F and [M]ads or 2þ 2þ in [M]intra of the four metals ions Cd2þ, CrO2 4 , Cu , and Zn the log-scale was found to follow the Freundlich isotherm. MC-LR, as a metal binding ligand with intermediate affinity, had no considerable effects on metal bioavailability, their accumulation, and further toxicity in the green alga C. reinhardtii at the environmentally realistic concentration. The metal toxicity could be well described with the current models (e.g., FIAM, BLM, and [M]intra) no matter whether MC-LR was
[M]ads-EC50s
10
7
106 109 107
5
0.0003 2.0 10 r2 ¼ 0.92, p < 0.01 0.0004 0.0002 r2 ¼ 0.84, p < 0.01 2.19 105 6.63 106 r2 ¼ 0.93, p < 0.01 0.0033 0.0005 r2 ¼ 0.88, p < 0.01
[M]intra-EC50s 0.0003 5.70 106 r2 ¼ 0.98, p < 0.01 4.96 105 8.69 107 r2 ¼ 0.99, p < 0.01 9.43 105 2.29 105 r2 ¼ 0.97, p < 0.01 0.0011 8.23 105 r2 ¼ 0.98, p < 0.01
present or not. The underlying uptake mechanisms of MC-LR need to be further examined.
Acknowledgement The financial supports offered by National Basic Research Program of China (973 Program, Grant No. 2008CB418102) to Liu-Yan Yang, and by the National Natural Science Foundation of China (41001338) and the Natural Science Foundation of Jiangsu Province (BK2010371) to Ai-Jun Miao have made this work possible.
Appendix. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.10.035.
references
Babica, P., Blaha, L., Marsalek, B., 2006. Exploring the natural role of microcystins e a review of effects on photoautotrophic organisms. Journal of Phycology 42, 9e20. Campbell, P.G.C., 1995. In: Tessier, A., Turner, D.R. (Eds.), Metal Speciation and Bioavailability in Aquatic Systems. John Wiley & Sons Ltd., Chichester, UK, pp. 45e102. De Schamphelaere, K.A.C., Janssen, C.R., 2006. Bioavailability models for predicting copper toxicity to freshwater green microalgae as a function of water chemistry. Environmental Science & Technology 40, 4514e4522. De Schamphelaere, K.A.C., Stauber, J.L., Wilde, K.L., Markich, S.J., Brown, P.L., Franklin, N.M., Creighton, N.M., Janssen, C.R., 2005. Toward a biotic ligand model for freshwater green algae: surface-bound and internal copper are better predictors of toxicity than free Cu2þ-ion activity when pH is varied. Environmental Science & Technology 39, 2067e2072. Diao, W., Ni, W., Ni, T., Yang, X., 2004. The existing status and evaluation of heavy metal pollution in water environment. Guangdong Weiliang Yuansu Kexue 11, 1e5. Dixon, R.A., Al-Nazawi, M., Alderson, G., 2004. Permeabilising effects of sub-inhibitory concentrations of microcystin on the growth of Escherichia coli. Fems Microbiology Letters 230, 167e170. Ernstberger, H., Zhang, H., Davison, W., 2002. Determination of chromium speciation in natural systems using DGT. Analytical and Bioanalytical Chemistry 373, 873e879.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 6 9 e3 7 7
Fargasova, A., 1999. The green alga Scenedesmus quadricauda a subject for the study of inhibitory effects of Cd, Cu, Zn, Pb and Fe. Biologia 54, 303e308. Fischer, A., Hoeger, S.J., Stemmer, K., Feurstein, D.J., Knobeloch, D., Nussler, A., Dietrich, D.R., 2010. The role of organic anion transporting polypeptides (OATPs/SLCOs) in the toxicity of different microcystin congeners in vitro: a comparison of primary human hepatocytes and OATPtransfected HEK293 cells. Toxicology and Applied Pharmacology 245, 9e20. Franklin, N.M., Stauber, J.L., Lim, R.P., Petocz, P., 2002. Toxicity of metal mixtures to a tropical freshwater alga (Chlorella sp.): the effect of interactions between copper, cadmium, and zinc on metal cell binding and uptake. Environmental Toxicology and Chemistry 21, 2412e2422. Fu, F., Wang, Q., 2011. Removal of heavy metal ions from wastewaters: a review. Journal of Environmental Management 92, 407e418. Gorbi, G., Zanni, C., Corradi, M.G., 2007. Sulfur starvation and chromium tolerance in Scenedesmus acutus: a possible link between metal tolerance and the regulation of sulfur uptake/ assimilation processes. Aquatic Toxicology 84, 457e464. Gouvea, S.P., Boyer, G.L., Twiss, M.R., 2008. Influence of ultraviolet radiation, copper, and zinc on microcystin content in Microcystis aeruginosa (Cyanobacteria). Harmful Algae 7, 194e205. Guillard, R., 1975. In: Smith, W., Chanley, M. (Eds.), Culture of Marine Invertebrate Animals. Plenum, New York, pp. 29e60. Hassler, C.S., Slaveykova, V.I., Wilkinson, K.J., 2004. Discriminating between intra- and extracellular metals using chemical extractions. Limnology and Oceanography-Methods 2, 237e247. Haye, J.M., Santschi, P.H., Roberts, K.A., Ray, S., 2006. Protective role of alginic acid against metal uptake by American oyster (Crassostrea virginica). Environmental Chemistry 3, 172e183. Hesse, K., Dittmann, E., Borner, T., 2001. Consequences of impaired microcystin production for light-dependent growth and pigmentation of Microcystis aeruginosa PCC 7806. Fems Microbiology Ecology 37, 39e43. Humble, A.V., Gadd, G.M., Codd, G.A., 1997. Binding of copper and zinc to three cyanobacterial microcystins quantified by differential pulse polarography. Water Research 31, 1679e1686. Knauer, K., Behra, R., Sigg, L., 1997. Effects of free Cu2þ and Zn2þ ions on growth and metal accumulation in freshwater algae. Environmental Toxicology and Chemistry 16, 220e229. Koukal, B., Gueguen, C., Pardos, M., Dominik, J., 2003. Influence of humic substances on the toxic effects of cadmium and zinc to the green alga Pseudokirchneriella subcapitata. Chemosphere 53, 953e961. Lamelas, C., Slaveykova, V.I., 2007. Comparison of Cd(II), Cu(II), and Pb(II) biouptake by green algae in the presence of humic acid. Environmental Science & Technology 41, 4172e4178. Mahan, C.A., Majidi, V., Holcombe, J.A., 1989. Evaluation of the metal uptake of several algae strains in a multicomponent matrix utilizing inductively coupled plasma emissionspectrometry. Analytical Chemistry 61, 624e627. Martins, J.C., Vasconcelos, V.M., 2009. Microcystin dynamics in aquatic organisms. Journal of Toxicology and Environmental Health-Part B-Critical Reviews 12, 65e82. Miao, A.J., Wang, W.X., 2007. Predicting copper toxicity with its intracellular or subcellular concentration and the thiol synthesis in a marine diatom. Environmental Science & Technology 41, 1777e1782. Miao, A.J., Wang, W.X., Juneau, P., 2005. Comparison of Cd, Cu, and Zn toxic effects on four marine phytoplankton by pulseamplitude-modulated fluorometry. Environmental Toxicology and Chemistry 24, 2603e2611. Mohamed, Z.A., 2008. Polysaccharides as a protective response against microcystin-induced oxidative stress in Chlorella vulgaris
377
and Scenedesmus quadricauda and their possible significance in the aquatic ecosystem. Ecotoxicology 17, 504e516. Pearson, R.G., 1997. Chemical Hardness: Applications from Molecules to Solids. Wiley-VCH, Weinheim. Quintelas, C., Rocha, Z., Silva, B., Fonseca, B., Figueiredo, H., Tavares, T., 2009. Biosorptive performance of an Escherichia coli biofilm supported on zeolite NaY for the removal of Cr(VI), Cd(II), Fe(III) and Ni(II). Chemical Engineering Journal 152, 110e115. Ramanan, S., Tang, J., Velayudhan, A., 2000. Isolation and preparative purification of microcystin variants. Journal of Chromatography A 883, 103e112. Reynolds, G.L., Hamiltontaylor, J., 1992. The role of planktonic algae in the cycling of Zn and Cu in a productive soft-water lake. Limnology and Oceanography 37, 1759e1769. Saito, K., Sei, Y., Miki, S., Yamaguchi, K., 2008. Detection of microcystin-metal complexes by using cryospray ionizationFourier transform ion cyclotron resonance mass spectrometry. Toxicon 51, 1496e1498. Sieroslawska, A., Rymuszka, A., Bownik, A., Skowroiiski, T., 2007. The influence of microcystin-LR on fish phagocytic cells. Human & Experimental Toxicology 26, 603e607. Singh, D.P., Tyagi, M.B., Kumar, A., Thakur, J.K., 2001. Antialgal activity of a hepatotoxin-producing cyanobacterium, Microcystis aeruginosa. World Journal of Microbiology & Biotechnology 17, 15e22. Slaveykova, V.I., Wilkinson, K.J., 2005. Predicting the bioavailability of metals and metal complexes: critical review of the biotic ligand model. Environmental Chemistry 2, 9e24. Song, L.R., Chen, W., Peng, L., Wan, N., Gan, N.Q., Zhang, X.M., 2007. Distribution and bioaccumulation of microcystins in water columns: a systematic investigation into the environmental fate and the risks associated with microcystins in Meiliang Bay, Lake Taihu. Water Research 41, 2853e2864. Sunda, W.G., Huntsman, S.A., 1998. Interactions among Cu2þ, Zn2þ, and Mn2þ in controlling cellular Mn, Zn, and growth rate in the coastal alga Chlamydomonas. Limnology and Oceanography 43, 1055e1064. Szivak, I., Behra, R., Sigg, L., 2009. Metal-induced reactive oxygen species production in Chlamydomonas reinhardtii (Chlorophyceae). Journal of Phycology 45, 427e435. Utkilen, H., Gjolme, N., 1995. Iron-stimulated toxin production in Microcystis aeruginosa. Applied and Environmental Microbiology 61, 797e800. Vigneault, B., Percot, A., Lafleur, M., Campbell, P.G.C., 2000. Permeability changes in model and phytoplankton membranes in the presence of aquatic humic substances. Environmental Science & Technology 34, 3907e3913. Wang, W.X., Dei, R.C.H., 2006. Metal stoichiometry in predicting Cd and Cu toxicity to a freshwater green alga Chlamydomonas reinhardtii. Environmental Pollution 142, 303e312. Wiegand, C., Pflugmacher, S., 2005. Ecotoxicological effects of selected cyanobacterial secondary metabolites a short review. Toxicology and Applied Pharmacology 203, 201e218. Yan, F., Ozsoz, M., Sadik, O.A., 2000. Electrochemical and conformational studies of microcystin-LR. Analytica Chimica Acta 409, 247e255. Zeng, J., Yang, L.Y., Wang, W.X., 2009. Cadmium and zinc uptake and toxicity in two strains of Microcystis aeruginosa predicted by metal free ion activity and intracellular concentration. Aquatic Toxicology 91, 212e220. Zerhouni, R.A., Bouya, D., Ronneau, C., Cara, J., 2004. Phosphate, nitrogen, chromium and cadmium uptake by three green algae species isolated from an urban wastewater. Revue des Sciences de l’Eau 17, 317e328. Zhang, H., Davison, W., 2000. Direct in situ measurements of labile inorganic and organically bound metal species in synthetic solutions and natural waters using diffusive gradients in thin films. Analytical Chemistry 72, 4447e4457.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 7 8 e3 8 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Phosphatase activities of cultured phototrophic biofilms Neil T.W. Ellwood a,*, Francesca Di Pippo b, Patrizia Albertano c a
Department of Geological Sciences, University of Rome “Roma Tre”, Largo San Leonardo Murialdo 1, 00146 Rome, Italy CNReIAMC, National Research Council, Institute for Coastal Marine Environment, Localita` Sa Mardini, Torregrande, 09072 Oristano, Italy c LBA-Laboratory for Biology of Algae, Department of Biology, University of Rome “Tor Vergata”, Via della Ricerca Scientifica snc, 00133 Rome, Italy b
article info
abstract
Article history:
The responses of cultured phototrophic biofilms to diverse phosphorus (P) regimes were
Received 18 May 2011
assessed using a semi-continuous flow incubator. Three biofilms were grown over 18 days
Received in revised form
under three different P regimes: replete inorganic P, organic P-only and limited inorganic P.
20 September 2011
Assessing the response of the biofilms took into account the rate of phosphomonoesterase
Accepted 29 October 2011
and phosphodiesterase activities, biofilm nutrient contents and biomass accrual across the
Available online 6 November 2011
growth period. Phosphorus limitation was indicated by slower biomass accumulation and higher phosphatase activities of the organic P-only and P-limited biofilms compared to the
Keywords:
P-replete biofilms. The cyanobacterium Phormidium sp. dominated the later stages in all the
Phosphomonoesterase
treatments forming a dense layer at the biofilmemedium interface. This layer possibly led
Phosphodiesterase
to a reduction of light and nutrient diffusion to sub-surface cells and may account for the
Phosphorus content
production of phosphatases under P replete conditions. In addition, the Phormidium-layer
Wastewater treatment
possibly produced a top-heavy P (and N) distribution and could explain the large reductions
Phosphorus limitation
in areal nutrient concentrations. End-product repression and de-repression of phosphatase
Organic phosphorus
activity was suggested to be a main controlling factor of phosphatase activity. Consequently, it is proposed that for efficient nutrient removal from wastewaters that biofilms should be regularly removed to continually maintain biofilms at the initial stages (3e7 days). ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Many sewage treatment processes provide optimal conditions for nutrient reduction, but effluents from secondary treatments can still contain relatively high levels of organic and inorganic nutrients (Shon et al., 2006). Among these, phosphorus (P) has been most frequently implicated in the eutrophication of surface waters (Jarvie et al., 2006). Most of the present tertiary treatments for the removal of P, aimed at minimizing harmful impact of effluents on receiving waters, involve chemical and/or physical processes, which are expensive and result in increased sludge volumes. An
alternative approach is biological P removal that has mostly involved the use of bacteria and microalgae. However, there still remains the problem of biomass produced after the treatment (Wei et al., 2003), although this can now be viewed as a valuable substrate for the production of oils and the following transformation in biofuels (Pittman et al., 2010). Use of suspended algae can be severely limited by the difficulties of harvesting the biomass after treatment (Roeselers et al., 2008) and therefore biofilms, that are relatively easy to harvest, represent a promising alternative (Ragusa et al., 2004). Phototrophic biofilms can be complex microbial communities commonly composed of microalgae, as well as bacteria
* Corresponding author. Tel.: þ39 06 57338077; fax: þ39 06 57338201. E-mail address:
[email protected] (N.T.W. Ellwood). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.057
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 7 8 e3 8 6
and fungi and micro-invertebrates, enclosed in a polymeric matrix (Donlan, 2002). The extracellular polymeric substances (EPS), that comprise the matrix, while composed largely of polysaccharides and proteins can also contain nucleic acids and organic materials both particulate and dissolved of allochthonous and autochthonous origin (Hu et al., 2003; Flemming et al., 2007; Flemming and Wingender, 2010). The matrix interacts with the environment by attaching the biofilm to surfaces, maintaining the hydrated state of the biofilm, providing ion exchange at the outer and interstitial surfaces and the entrapment of soluble and particulate nutrients (Wetzel, 2001; Flemming et al., 2007). Moreover, the complex heterogeneous matrix architecture, with an internal channel network, enhances nutrient transfer from the bulk water to microorganisms (Wetzel, 2001). The biofilm matrix can also form a diffusion barrier that promotes retention of extracellular enzymes such as phosphatases and proteases along with their substrates and products (Flemming and Wingender, 2010). There is some evidence that the extracellular hydrolytic activity is retained and accumulates over time, contributing to the overall biofilm metabolism, acting as an external digestive system and rendering the products more available for uptake (Flemming and Wingender, 2010). Extracellular enzymes may become attached to EPS through processes similar to the formation of enzymeehumus complexes in soil (Lock et al., 1984). It has also been suggested that the binding of enzymes protects them from degradation and increases their exposure to substrates (Espeland and Wetzel, 2001). The resulting efficient recycling of organic matter and increasing concentration of nutrients within this matrix-bound hydrolytic system buffers the community metabolism to changes in ambient nutrient dynamics (Sekar et al., 2002). The phosphatase group of enzymes hydrolyzes phosphate from dissolved organic molecules, making it potentially available for cellular uptake. Phosphatases can hydrolyze a wide variety of phosphate esters as they are highly specific for the mono- or di-phosphate ester bonds, and do not exhibit specificity for the organic moiety (Perry, 1972). The phosphatase assay has frequently been used as a monitoring tool for phosphorus status, with increased activity under P-limitation and repressed activity when P is replete (Whitton et al., 2005 and references therein). Most ecological studies of phosphatases, including biofilms, have focused solely on phosphomonoesterase (PMEase) activity. However, phosphodiesterase (PDEase) activity has proved to be only slightly less widespread than PMEase activity (Whitton et al., 2005 and references therein). There is scant information on the freshwater organic P forms, but a significant diester composition has been determined by enzyme hydrolysis studies (e.g. Turner et al., 2002; Monbet et al., 2007). Therefore, when determining P status of organisms, PDEase activity should also be considered. The aim of this study was to determine the response of constructed biofilms, composed of phototrophs previously isolated from a wastewater treatment plant, to changes in phosphorus supply. The PMEase and PDEase activities of nutrient contents and biomass accrual of biofilms grown under three different P regimes: replete inorganic P, organic P-
379
only and limited inorganic P were measured. The data were assessed to determine if enzyme accumulation occurred and to further the knowledge into the potential of biofilms as a tertiary treatment of wastewaters.
2.
Materials and methods
2.1.
Experimental design
The biofilms used in this study were composed of the coccal cyanobacterium Synechocystis sp., the filamentous cyanobacterium Phormidium sp. and the green alga Chlorococcum sp. previously isolated from biofilms grown in the sedimentation tank of the wastewater treatment plant located in Fiumicino (Rome, Italy) for the Airport ‘Leonardo da Vinci’. Each nonaxenic stock culture was maintained in 1 L flasks containing 250 mL modified BG11 medium (Guzzon et al., 2005) in a controlled growth chamber at 20 C, 60% relative humidity, illuminated at 30 mmol photon m2 s1 under a 16:8 h light:dark regime. Growth was estimated by measuring optical density (Kontron Uvikon 860 spectrophotometer) at 730 nm for cyanobacteria and at 678 nm for the green alga. When a stationary phase in all three cultures was reached biovolume was determined and a three-species inoculum was made ensuring equal cell biovolumes. The inoculation procedure involved adding 100 mL of the inoculum to 3.9 L of modified BG11 medium and this was pumped through a continuous flow-lane incubator (Fig. 1) for 72 h, at 100 L h1 and 25 C. After this inoculation period the medium was substituted with fresh medium, the flow rate reduced to 25 L h1 and temperature kept at 25 C. The medium was refreshed twice per week and biofilms sampled on six occasions.
2.2.
Flow-lane incubator
The incubator (Fig. 1) (UFZ Centre for Environmental Research, Germany) consists of four separate chambers each having a 798 cm2 growing area made up of 42 removable polycarbonate slides each (7.6 2.5 cm). The 4 L inoculum was
Fig. 1 e Continuous flow incubator used to culture phototrophic biofilms without the light source included for clarity. 1. Flow is valve regulated; 2. Growth surface of 42 polycarbonate slides; 3. Medium is circulated from a reservoir of 4 L.
380
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 7 8 e3 8 6
pumped through the inlet device within each lane with a laminar flow of 25 L h1 over the slide surface. A turbulence reducer and a water temperature sensor were located at the inlet device of each chamber and the flow rate was valve regulated. Irradiance of 100 mmol photon m2 s1 was provided by fluorescent lamps (True-light 36 W AURALIGHT, Sweden) in a 16:8 h lightedark cycle. The incubator design is described in more detail by Zippel et al. (2007). The medium used during the inoculation phase was the same for all treatments to give similar attached inocula. The start of the experiment was when the media were changed to give the diverse P regimes, which were: KH2PO4 at 3.68 mg L1 P (P-rep); KH2PO4 at 0.368 mg L1 P (P-lim); organic P-only (glucose 6-phosphate, G 6-P) at 3.68 mg L1 P (P-org).
2.3.
Biovolume calculation
Preparation of samples for biovolume evaluation involved sonication twice for 3 min and aliquots of suspensions were then diluted in phosphate buffer (0.1 M Na2HPO4/NaH2PO4; pH 7.2) and left to settle for 24 h in 25 mL sedimentation chambers. To estimate the biovolume of single cells standard equations proposed for the different cyanobacterial and algal shapes were used (Hillebrand et al., 1999).
2.4.
Biofilm phosphatase assays
The substrates para-nitrophenyl phosphate ( pNPP) and bispara-nitrophenyl phosphate (bis-pNPP) were used for assays of PMEase and PDEase activities. The assay procedure broadly follows that of Turner et al. (2001). Briefly, on each sampling occasion one representative slide was taken from each treatment and the whole biofilm biomass was carefully scraped from the slides, divided into similar sized aliquots and placed into six 5-mL tubes containing 4.32 mL of buffered medium without combined N or P. The tubes were then placed in a shaking incubator at 25 C for 20 min before the addition of 0.18 mL substrate (final concentration of 250 mM). The samples were then incubated for 3 h, after which the assay reaction was terminated by the addition of 0.25 mL of 0.5 M NaOH. Biomass was removed from the solution, rinsed and dried (105 C for 24 h) and weighed to an accuracy of 0.001 g. Absorbance of the solution was read at 405 nm. To determine the activity of phosphatases released to the medium, 0.18 mL of each substrate was added to 4.32 mL samples of the circulating growth medium using the same assay conditions and termination procedure as the biofilm. To determine the location of activity, assays were performed on cells separated from the biofilm matrix and on intact biofilms. Biofilm matrix was removed following a modified procedure of de Brouwer et al. (2002). This method was best suited for both the removal of the EPS and the maintenance of cell integrity. Biofilms were incubated in distilled water at 50 C for 1 h followed by centrifugation (3500 rpm, 20 min, 15 C). The supernatant was then separated from the resulting pellet comprising the cells which was resuspended in buffered modified BG11 medium (without combined N and P). After centrifugation, parts of the pellet were stained with Alcian Blue at pH 2.5 and 0.5. The samples were observed using a light microscope (Zeiss Axioskop) at
400 and 1000 magnification to determine the matrix-cell separation efficiency and the intactness of the cells. Phosphatase activity was measured using the methodology of the biofilm assay. Matrix activity was estimated by subtraction of cellular from whole biofilm activity. Optimization of the assay conditions involved measurement of PMEase and PDEase activities across ranges of pH, substrate and time. When measured across a pH range of 3e11, PMEase showed peaks of activity in both acid (pH 4e5) and alkaline range (pH 9) while PDEase had neutral to alkaline pH optimum (pH 7e8). Activity was linear over the 16 h incubation test period and values at 3 h were considered robust enough to incorporate variability in activity. The chosen substrate concentration (250 mM) was approximately five times the Km values of PMEase and PDEase (Km ¼ 50 and 44 mM respectively) to avoid any diffusion hindrance of substrate into the EPS by creating a large gradient into the biofilm. There was no substrate inhibition within the concentration range of the saturation curves (5e2000 mM). The temperature used was the same as that of the growth conditions of the biofilm.
2.5.
Digestion of biofilms and N and P analyses
Analytical reagent grade chemicals and double-distilled deionized water (DDW) were used for all digests. Oxidizing reagent was prepared on the day of digestion made up by adding 15 mL of stock NaOH solution (3.75 M) to 50 mL DDW in a 100 mL volumetric flask then 10 g of potassium persulphate (K2S208) was dissolved in this solution and made up to 100 mL with DDW (Johnes and Heathwaite, 1992). In a 125 mL Erlenmeyer flask, 10 mL of reagent was added to a known weight of pulverized biofilm, this was then tightly capped with aluminum foil and autoclaved at 1 bar for 45 min. Analyses were made on the same biofilm (rinsed in DDW) used for the phosphatase assay. Standards were diluted from 100 mg L1 KH2PO4eP and KNO3eN stock solutions to produce a standard curve. Before performing the N and P analyses, the digested solutions were precisely diluted tenfold with DDW to give absorbances within the linear range of the spectrophotometer. The samples were then split into two aliquots for the separate analyses of N and P. Analysis of N involved the determination of NO3eN using a second derivative UV/visible spectroscopical method (Ferree and Shannon, 2001) while P was measured by molybdate colorimetry (Murphy and Riley, 1962). Data analyses were carried out in SPSS v.13.0 for Windows. Spearman’s rank correlation coefficients (r values) were carried out to determine relationships between biofilm nutrient contents and phosphatase activities. Differences in mean values were determined by paired and un-paired t-tests.
3.
Results
3.1.
Biomass total and community structure
All three biofilm dry weights, normalized to a unit area (mg cm2) increased for the duration of the experiment (Fig. 2). The growth of the P-rep biofilm was significantly higher than that of the P-lim and P-org treatments (P < 0.01), there was no
381
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 7 8 e3 8 6
3.2.
Phosphatase activities in medium
All medium PMEase and PDEase activities were below the levels of detection.
3.3.
100
P-replete
75 50
Biofilm species composition ( )
significant difference between the P-lim and P-org treatments (P > 0.05). Under all three phosphorus regimes Phormidium sp. dominated all the final communities (Fig. 3). In the P-lim cultures, Synechocystis sp. was the dominant taxa in the early to mid stages. During the inoculation period, Phormidium sp. rapidly dominated the biofilms incubated under P-org and P-rep conditions. The opposite occurred for Chlorococcum sp., where the biovolumes rapidly decreased and were only detectable in the first samples (day 1) under all three P regimes (Fig. 3).
25 0 100
P-organic
75 50 25 0 100
P-limited
75
Biofilm phosphatase activities
Phormidium sp. Synechocystis sp. Chlorococcum sp.
50
Comparison of cellular versus intact biofilm phosphatase assays showed that activity was mostly cell-associated, the values for PMEase and PDEase activities being 59% and 77% of the total activities of the respective enzymes in the whole biofilm (Fig. 4). There was no significant difference between cellular and whole biofilm PDEase activity (P > 0.05). The response to the three P regimes showed that the PMEase activities of the P-rep and P-lim biofilms both decreased initially, but after day 12 there was a large increase in activity of the P-lim biofilm (Fig. 5). The PMEase of the P-org biofilm initially increased followed by a general decline for the remainder of the experimental period. Initial PDEase activities of all biofilms were either very low or absent for all three treatments, this was followed by an increase with time, the obvious exception was the P-org biofilm which showed a rapid and large increase followed by a progressive decrease in
1.2
P-rep
0 1
P-org
4
9
12
15
19
Sampling event (d)
Fig. 3 e Changes of community structure during the development of biofilms over 18 days. Single species contributions (as percent) to the community based on biovolume calculations (mean values, n [ 3).
PDEase activity for the remainder of the experiment (Fig. 5). PMEase activities were significantly and positively related to biofilm P contents (P < 0.01) and negatively related to biofilm N:P and DW cm2 (P < 0.01 and 0.05, respectively). PDEase activity showed no significant relationships with biofilm nutrient contents or DWs.
3.4.
Biofilm nutrient status
Biofilm N content (% DW) was less variable than P content for all treatments with mean values (SD) 8.87 0.95 and
1.0 P-lim
0.4
0.2
0.0 01
05
09
13
17
Time (d) Fig. 2 e Dry weights of biofilm per cmL2, these values were derived from the combined weights of the biofilm samples used for the phosphatase analyses.
-1
0.6
120
-1
0.8
Phosphatase activity (µmol product g DW h )
-2
Biofilm DW (mg cm )
25
100 Biofilm (cells + matrix) Cells Matrix
80
60
40
20
0 PMEase
PDEase
Fig. 4 e Direct measure of phosphatase activities (PMEase and PDEase) of intact biofilms and separated cells and estimation of matrix activity (total-cellular activity; n [ 3).
382
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 7 8 e3 8 6
60
P-rep
18
50 40
12
30 20
6
0
-1
PDEase activity (µmol product g DW h )
0 60
18
P-org
-1
-1
-1
PMEase activity (µmol product g DW h )
10
50 40 30 20 10 0 60
12
6
0
P-lim
18
50 40
12
30 20
6
10 0 01
07
13
0 01
19
07
13
19
Time (days)
Time (days)
Fig. 5 e Changes in the PMEase and PDEase activities during the development of three cultured biofilms over 18 days in diverse P regimes (P-rep, KH2PO4 at 3.68 mg LL1 P; P-org, glucose 6-phosphate at 3.68 mg LL1, P-lim, KH2PO4 at 0.368 mg LL1 P) (mean values, n [ 3).
0.87 0.43%, respectively (Fig. 6a and b). From the onset the general trend was for biofilm P (% P of DW) to decrease in the P-rep and P-lim biofilms (85 and 75% reduction from max to min values), while the P-org biofilm which was more stable and was higher in P in the end (þ33%) (Fig. 6b). The biofilm P contents of the P-lim and P-org biofilms increased from day 1e4 and then declined. When converted to areal concentrations (mg N or P cm2) the values for both N and P contents decreased sharply (Fig. 6c and d). The biofilm N:P increased with time in all cultures. the increase was less in the P-org cultures (6.8e10.8) compared to P-lim (5e27) and P-rep (6e23) cultures. For P-rep there was a decrease in the N:P after day 13(Fig. 6e). Correlation analyses showed significant negative relationships between biofilm DW and biofilm N and P content (P < 0.01).
4.
Discussion
The study was designed to avoid procedures likely to interfere with realistic measurements of phosphatase activities of intact biofilms. Comparisons of the response of the biofilms under the three P regimes showed that the P-org and the P-lim biofilms were more P-limited than the P-rep biofilm. Compared to the P-rep biofilm, the P-lim and P-org biofilms had lower biomass accrual rates and higher phosphatase activities (PMEase and PDEase) indicative of P-limitation. There was a more rapid increase in the phosphatase acitivity in P-org biofilms implying a more immediate P-limitation response. The increase in phosphatase activity of the P-lim biofilm in the latter stages of experiment suggests that the
383
9
b
8
7
6
c
1.6
1.2
Biofilm P content (
10
Biofilm N content
a DW)
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 7 8 e3 8 6
0.8
0.4
0.0
d
0.8
0.12
mg P cm-2
mg N cm-2
P-rep 0.6
0.4
P-org 0.08
P-lim
0.04 0.2
0.0
0.00 01
e
07
13
19
30
Biofilm N:P
20
10
0 01
07
13
19
Days Fig. 6 e Changes in biofilm nutrient concentrations (N and P). (a and b) N and P values as percentage of the biofilm dry weight; (c and d) areal N and P values normalized to surface area, (e) N:P ratio (by mass) of the biofilm (mean values, n [ 3).
biofilm was initially able to meet its P requirement without initiating a phosphatase response. The increase in phosphatase activity was possibly triggered by the increase in P demand caused by the development the biofilm. Although it cannot be discerned if the P-lim phosphatase response was due to overall biofilm development or the late shift in the community from Synechocystis sp. to Phormidium sp. dominance. It would require characterization of the phosphatases of single species and in situ enzyme labeled fluorescence experiments to determine the location and quantification of species-specific activity. The final biofilms all showed a clear dominance by Phormidium sp.. This indicates that it competes well under high P
availability, and that it can also compete efficiently using glucose 6-phosphate. Preference for high P availability for Phormidium has also been shown in a simulated stream experiment (Horner et al., 1990) and in wastewater treatment (Laliberte´ et al., 1997). Phormidium has frequently been used successfully during tertiary water treatment, specifically for P removal (inorganic and organic) (Markou and Georgakakis, 2011 and references therein). Moreover, it has been shown elsewhere that cultures of mixed species biofilms were dominated by Phormidium and had maintained high P retention rates (Guzzon et al., 2008). One explanation of the reduction in P-rep and P-org biofilm P contents could be growth dilution, where the use of P for
384
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 7 8 e3 8 6
growth outstrips P uptake. The response of the P-org biofilm was typical of P-limitation, showing a reduced growth rate (Litchman et al., 2003) and increased phosphatase activity (Newman et al., 2003). Both responses may have resulted in a more stable cellular P. Another possible explanation for the reduction in biofilm P was gained from observations of the biofilms; all three biofilms showed at some stage a dense covering strata composed primarily of the filaments of Phormidium sp.. It has been shown that filamentous nonheterocystous cyanobacteria (Oscillatoriales), including Phormidium, often form surface layers of biofilms (de los Rı´os et al., 2004; Roeselers et al., 2008). The voids present throughout the biofilm permit the gliding movements and changes in orientation of filamentous cyanobacteria (de los Rı´os et al., 2004). The movement of Phormidium to the surface likely created a barrier that caused large reductions in light and nutrient diffusion to the rest of the biofilm. Observations of natural biofilms of streams with differing nutrient status also showed a tiered development and suggested that it was a result of differing abilities of species to obtain nutrients from overlying waters (Sabater et al., 2002). This enhanced uptake at the surface will result in reduced rates of diffusion and uptake by the rest of the biofilm with increasing biofilm thickness (Horner et al., 1990). This Phormidium-layer was present in all biofilms, yet there were higher levels of P in the P-lim and Porg biofilms compared to those of the P-rep biofilm. The reduced growth rates and increased phosphatase activities of these biofilms, that are typical responses of P-limitation (Litchman et al., 2003; Newman et al., 2003), could possibly explain this. PDEase activity has been indicated as a secondary response to more enhanced P-limitation (Ellwood et al., 2008). How this applies with regard to the P-org and P-rep biofilms is not known as the medium concentrations of P were relatively high. The cells below the Phormidium-layer could be experiencing enhanced P-limitation leading to an augmentation of PDEase (and PMEase) activity to scavenge P from sources such as organics originating from leaked cellular products and lysed cells (Whitton et al., 2005). Although intact, the phosphatase assays here were made on detached biofilms therefore revealing the under-part of the biofilm to the analog substrates, essentially removing the diffusion barrier caused by the Phormidium-layer. It would therefore be interesting to compare the activities of attached and detached biofilms to determine if substrates are able to pass freely through the dense Phormidium-surface barrier. Accumulation of matrix-bound enzymes and the retention of hydrolyzates within the matrix have been implicated in the competitive success of biofilms over their free living counterparts (Sinsabaugh et al., 1991). If however, accumulation is demonstrated by continually increasing hydrolytic capacity, only the PMEase and PDEase activities in the P-lim biofilms indicated this to some extent. In P-rep and P-org biofilms after an initial increase there was a tailing-off of activity, exemplified in the case of PDEase activity in the P-org biofilm (Fig. 5). In addition, the lack of phosphatase loss to the medium suggests that they were immobilized on the cell surface or within the EPS matrix. So, if there was an accumulation of enzymes, why was activity shown to decrease? There are two possible explanations of decreasing activity: firstly, it has been
shown that phosphatase activity of aquatic microalgae and bacteria is repressed by the accumulation of end-product (Pi) in the cell, cell wall or in the surrounding medium and derepressed upon its removal (Chro´st and Overbeck, 1987). The cycle of repression and de-repression of phosphatase activity in biofilms would be very closely coupled to the cellular uptake of P due to the short distances involved (de Beer and Stoodley, 2006). Secondly, phosphatases can be denatured by physical and chemical factors in the aquatic environment or hydrolyzed by proteases (Chro´st and Siuda, 2002). Although the denaturing of matrix-bound enzymes is likely to be limited as ambient conditions are buffered by the physical and chemical make-up of the matrix and the physiological activity of the biofilm cell component (Wolfaardt et al., 1999). Significant proteolysis of phosphatases was also unlikely as concentrations of medium inorganic N were relatively high as to not incur heterotrophic N limitation (Matin et al., 1989). Given the high cell-associated activity and the possible repression/de-repression of activity by inorganic P, it was not expected to find significant, positive relationships between phosphatase activities and biofilm P content (P < 0.05). This may be accounted for by changing of the relative proportion of matrix and cellular activity as the biofilm matures, as found for other enzymes of natural biofilms (Romanı´ et al., 2008). This signifies that as the biofilm matures and enzymes accumulate within the EPS there is an increased proportion of the total biofilm activity that is removed from cell regulation mechanisms within the biofilm (Romanı´ and Sabater, 1999). In addition, differing P requirements and acquirement processes of the different growth stages (and associated EPS production) of the biofilm may also be responsible. Direct measurements of cellular and EPS activities and P contents during biofilm development are the next logical steps in order to determine the origins and the regulation processes of the total biofilm activity. This study gives initial results into the potential of cultured biofilms as a tertiary treatment for the removal of dissolved organic and inorganic P. The product inhibition (repression) of activity by Pi is most likely the major factor that limits the hydrolytic capacity of the biofilm. The evidence shown here indicates that water treatment using younger biofilms of 3e7 days would be most efficient. These biofilms, in general, had the highest P contents and phosphatase activities. On this basis, it can be suggested that for efficient water treatment biofilms should be regularly harvested to maintain high P removal rates. Simple scraping technologies would leave behind an inoculum for continual biofilm development.
5.
Conclusions
Adaptation to changes in ambient phosphorus regimes by phototrophic biofilms includes amending growth rates, utilization of stored P and production of phosphatase enzymes and the determination of both PMEase and PDEase activities suggests that these biofilms can make use of a wide range of organic P compounds. More phosphatase activity was associated with cells than the matrix, no activity was ever detected in the medium at any point
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 7 8 e3 8 6
Accumulation of enzymes in the EPS matrix during biofilm development was not possible to determine by the enzyme assays used because of masking by factors such as substrate (inorganic P) repression/de-repression of activity within the matrix. A tiered development of the biofilm comprised of a concentration of Phormidium filaments at the biofilmewater interface possibly caused nutrient limitation to the other parts of the biofilm. The use of biofilms as a tertiary treatment of wastewaters is promising especially under conditions that maintain continual development, such as harvesting more frequently, to sustain a high hydrolytic capacity and nutrient requirement. The acceptance of biological remediation of wastewaters by planktonic species is often hindered by the cost of harvesting of the resulting biomass. The use of biofilms can help solve this, because they can be harvested by simple physical methods.
references
Chro´st, R.J., Overbeck, J., 1987. Kinetics of alkaline phosphatase activity and phosphorus availability for phytoplankton and bacterioplankton in Lake Plußsee (North German eutrophic lake). Microb. Ecol. 13, 229e248. Chro´st, R.J., Siuda, W., 2002. Ecology of microbial enzymes in lake ecosystems. In: Burns, R.G., Dick, R.P. (Eds.), Enzymes in the Environment: Activity, Ecology and Applications. Marcel Dekker Inc., New York, pp. 35e72. de Beer, D., Stoodley, P., 2006. Microbial biofilms. Prokaryotes 1, 904e937. de Brouwer, J.F.C., Wolfstein, K., Stal, L.J., 2002. Physical characterization and diel dynamics of different fractions of extracellular polysaccharides in an axenic culture of a benthic diatom. Eur. J. Phycol. 37, 37e44. de los Rı´os, A., Ascaso, C., Wierzchos, J., Ferna´ndez-Valiente, E., Quesada, A., 2004. Bottom of form microstructural characterization of cyanobacterial mats from the McMurdo Ice Shelf, Antarctica. Appl. Environ Microbiol. 70 (1), 569e580. Donlan, R.M., 2002. Biofilms: microbial life on surfaces. Emerg. Infect. Dis. 8, 881e890. Ellwood, N.T.W., Whitton, B.A., Haile, S.M., 2008. Aquatic plant nutrients, moss phosphatase activities and tissue composition in four upland streams in northern England. J. Hydrol. 350 (3e4), 246e260. Espeland, E.M., Wetzel, R.G., 2001. Complexation, stabilization, and UV photolysis of extracellular and surface-bound glucosidase and alkaline phosphatase: implications for biofilm microbiota. Microb. Ecol. 42, 572e585. Ferree, M.A., Shannon, R.D., 2001. Evaluation of a second derivative UV/visible spectroscopy technique for nitrate and total nitrogen analysis of wastewater samples. Water Res. 35, 327e332. Flemming, H.C., Neu, T.R., Wozniak, D.J., 2007. The EPS matrix: the "house of biofilm cells". J. Bacteriol. 189, 7945e7947. Flemming, H.C., Wingender, J., 2010. The biofilm matrix. Nat. Rev. Microbiol. 8, 623e633. Guzzon, A., Congestri, R., Albertano, P., 2005. Light-induced changes in photosynthesis and structure of cyanobacterial cultured biofilms from an Italian wastewater treatment plant. Arch. Hybrobiol. Algol. Stud. 117, 223e238. Guzzon, A., Bohn, A., Diociaiuti, M., Albertano, P., 2008. Cultured phototrophic biofilms for phosphorus removal in wastewater treatment. Water Res. 42, 4357e4367.
385
Hillebrand, H., Du¨rselen, C.D., Kirschtel, D., Pollingher, U., Zohary, T., 1999. Biovolume calculation for pelagic and benthic microalgae. J. Phycol. 35, 403e424. Horner, R.R., Welch, E.B., Seeley, M.R., Jacoby, J.M., 1990. Responses of periphyton to changes in current velocity, suspended sediment and phosphorus concentration. Freshwater Biol. 24 (2), 215e232. Hu, C., Liu, Y., Paulsen, B.S., Petersen, D., Klaveness, D., 2003. Extracellular carbohydrate polymers from five desert soil algae with different cohesion in the stabilization of fine sand grain. Carbohyd. Polym. 54, 33e42. Jarvie, H.P., Neal, C., Withers, P.J.A., 2006. Sewage-effluent phosphorus: a greater risk to river eutrophication than agricultural phosphorus? Sci. Total Env. 360 (1e3), 246e253. Johnes, P.J., Heathwaite, A.L., 1992. A procedure for the simultaneous determination of total nitrogen and total phosphorus in freshwater samples using persulphate microwave digestion. Water Res. 26, 1281e1287. Laliberte´, G., Lessard, P., De La Nou¨e, J., Sylvestre, S., 1997. Effect of phosphorus addition on nutrient removal from wastewater with the cyanobacterium Phormidium bohneri. Bioresour. Technol. 59, 227e233. Litchman, E., Steiner, D., Bossard, P., 2003. Photosynthetic and growth responses of three freshwater algae to phosphorus limitation and daylength. Freshwater Biol. 48, 2141e2148. Lock, M.A., Wallace, R.R., Costerton, J.W., Ventullo, R.M., Charlton, S.E., 1984. River epilithon: toward a structuralefunctional model. Oikos 42, 10e22. Markou, G., Georgakakis, D., 2011. Cultivation of filamentous cyanobacteria (blue-green algae) in agro-industrial wastes and wastewaters: a review. App. Energ. 88 (10), 3389e3401. Matin, A., Auger, E.A., Blum, P.H., Schultz, J.E., 1989. Genetic basis of starvation survival in non-differentiating bacteria. Annu. Rev. Microbiol. 43, 293e316. Monbet, P., McKelvie, I.D., Saefumillah, A., Worsfold, P.J., 2007. A protocol to assess the enzymatic release of dissolved organic phosphorus species in waters under environmentally relevant conditions. Environ. Sci. Technol. 41, 7479e7485. Murphy, J., Riley, J.P., 1962. A modified single solution method for the determination of phosphate in natural waters. Anal. Chim. Acta 12, 162e176. Newman, S., McCormick, P.V., Backus, J.G., 2003. Phosphatase activity as an early warning indicator of wetland eutrophication: problems and prospects. J. Appl. Phycol. 15, 45e59. Perry, M.J., 1972. Alkaline phosphatase activity in subtropical Central North Pacific waters using a sensitive fluorometric method. Mar. Biol. 15, 113e119. Pittman, J.K., Dean, A.P., Osundeko, O., 2010. The potential of sustainable algal biofuel production using wastewater resources. Bioresour. Technol. 102, 17e25. Ragusa, S.R., McNevin, D., Qasem, S., Mitchell, C., 2004. Indicators of biofilm development and activity in constructed wetlands microcosms. Water Res. 38 (12), 2865e2873. Roeselers, G., van Loosdrecht, M.C.M., Muyzer, G., 2008. Phototrophic biofilms and their potential applications. J. Appl. Phycol. 20, 227e235. Romanı´, A.M., Fund, K., Artigas, J., Schwartz, T., Sabater, S., Obst, U., 2008. Relevance of polymeric matrix enzymes during biofilm formation. Micob. Ecol. 56, 427e436. Romanı´, A.M., Sabater, S., 1999. Effect of primary producers on the heterotrophic metabolism of a stream biofilm. Freshwater Biol. 41, 729e736. Sabater, S., Guasch, H., Romanı´, A., Mun˜oz, I., 2002. The effect of biological factors on the efficiency of river biofilms in improving water quality. Hydrobiologia 469, 149e156. Sekar, R., Nair, K.V.K., Rao, V.N.R., Venugopalan, V.P., 2002. Nutrient dynamics and successional changes in a lentic freshwater biofilm. Freshwater Biol. 47, 1893e1907.
386
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 7 8 e3 8 6
Shon, H.K., Vigneswaran, S., Snyder, S.A., 2006. Effluent organic matter (EfOM) in wastewater: constituents, effects, and treatment. Crit. Rev. Env. Sci. Tec. 36 (4), 327e374. Sinsabaugh, R.L., Golladay, S.W., Linkins, A.E., 1991. Comparison of epilithic and epixylic biofilm development in a boreal river. Freshwater Biol. 25 (1), 179e187. Turner, B.L., Baxter, R., Ellwood, N.T.W., Whitton, B.A., 2001. Characterization of the phosphatase activities of mosses in relation to their environment. Plant Cell Environ. 24, 1165e1176. Turner, B.L., McKelvie, I.D., Haygarth, P.M., 2002. Characterisation of water-extractable soil organic phosphorus by phosphatase hydrolysis. Soil Biol. Biochem. 34 (1), 27e35. Wei, Y., Van Houten, R.T., Borger, A.R., Eikelboom, D.H., Fan, Y., 2003. Minimization of excess sludge production for biological wastewater treatment. Water Res. 37 (18), 4453e4467.
Wetzel, R.G., 2001. Land-water interfaces: attached microorganisms, littoral algae, and zooplankton. In: Wetzel, R.G. (Ed.), Lmnology: Lake and River Ecosystems. Academic Press, San Diego, pp. 577e623. Whitton, B.A., Al-Shehri, A.M., Ellwood, N.T.W., Turner, B.L., 2005. Ecological aspects of phosphatase activity in cyanobacteria, eukaryotic algae and bryophytes. In: Turner, B.L., Frossard, E., Baldwin, D.S. (Eds.), Organic Phosphorus in the Environment. CAB International, Wallingford, pp. 205e241. Wolfaardt, G.M., Lawrence, J.R., Korber, D.R., 1999. Function of EPS. In: Wingender, J., Neu, T.R., Flemming, H.-C. (Eds.), Microbial Extracellular Polymeric Substances; Characterization, Structure and Function. Springer-Verlag Berlin, Heidelberg, New York. Zippel, B., Rijstenbil, J., Neu, T.R., 2007. A flow-lane incubator for studying freshwater and marine phototrophic biofilms. J. Microbiol. Meth. 70, 336e345.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 8 7 e3 9 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Occurrence and fate of quinolone and fluoroquinolone antibiotics in a municipal sewage treatment plant Ai Jia, Yi Wan, Yang Xiao, Jianying Hu* MOE Laboratory for Earth Surface Processes, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China
article info
abstract
Article history:
This study developed a method for analysis of nineteen quinolone and fluoroquinolone
Received 12 May 2011
antibiotics (FQs) in sludge samples, and investigated the occurrence and fate of the FQs in
Received in revised form
a municipal sewage treatment plant (STP) with anaerobic, anoxic, and aerobic treatment
4 October 2011
processes. Eleven compounds, including pipemidic acid, fleroxacin, ofloxacin, norfloxacin,
Accepted 25 October 2011
ciprofloxacin, enrofloxacin, lomefloxacin, sparfloxacin, gatifloxacin, moxifloxacin, and
Available online 15 November 2011
sarafloxacin (only in sludge), were detected in the STP. The predominance of ofloxacin and norfloxacin, followed by lomefloxacin, ciprofloxacin, gatifloxacin, and moxifloxacin, were
Keywords:
found in wastewater, suspended solids, and sludge. The total concentrations of FQs were
Antibiotics
2573 241 ng/L, 1013 218 ng/L, and 18.4 0.9 mg/kg in raw sewage, secondary effluent,
Quinolone
and sludge, respectively. Extremely low mass change percentages were observed for FQs in
Fluoroquinolone
anaerobic, anoxic, and aerobic treatment units, suggesting biodegradation to be of minor
Sorption
importance in the removal of FQs in STPs. 50e87% of the initial FQs loadings (except for
Mass balance
pipemidic acid (36%)) were ultimately found in the dewatered sludge. Mean removal effi-
Sewage treatment plant
ciencies of FQs in the STP were 56e75%, except for new generation drugs such as moxifloxacin (40 5%) and gatifloxacin (43 13%). A significant positive correlation was found between removal efficiencies and Kd of FQs. The major factor in the removal of FQs in the STP was sorption to sludge, which was not governed by hydrophobic interactions. The long-term cycling and persistence of FQs in the STP has made activated sludge as a huge reservoir of FQ antibiotics. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Antibiotics are commonly used in human and veterinary medicine, and the presence of these compounds in the environment is of concern due to their role in the development of antimicrobial resistance among microorganisms (Daughton and Ternes, 1999). Quinolones and fluoroquinolones (FQs) are a group of antibiotics widely used to treat a broad variety of Gram (þ) and Gram () bacterial infections since their derivation from nalidixic acid in 1962 (Ball, 2000). The FQs are among the five classes of antibiotics (b-lactam, macrolides,
fluoroquinolones, sulfonamides, and tetracyclines) frequently detected in the environment in relatively high concentrations (Diaz-Cruz and Barcelo, 2006; Khetan and Collins, 2007), and their ubiquitous presence has been reported in wastewater (Gros et al., 2007), surface water (Kolpin et al., 2002), ground water, and even in drinking water (Barnes et al., 2008; Ye et al., 2007). Extensive usage and wide occurrences have led to increasing bacterial resistances to FQs in wastewater from hospitals and livestock feedlots, effluent and sewage sludge from municipal sewage treatment plants (STPs), and rivers (Hu et al., 2008; Polk et al., 2004; Reinthaler et al., 2003; Taylor
* Corresponding author. Tel./fax: 86 10 62765520. E-mail address:
[email protected] (J. Hu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.055
388
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 8 7 e3 9 4
et al., 2008), making human and animals more susceptible to these microbes. A broad number of FQ antibiotics (more than twenty chemicals cross four generations) have been developed and are commercially available around the world (Martinez et al., 2006). Except for the most concerning FQs (e.g. ciprofloxacin) in previous studies (Castiglioni et al., 2005, 2006; Ghosh et al., 2009; Golet et al., 2001; Miao et al., 2004; Renew and Huang, 2004; Shi et al., 2009; Vieno et al., 2007, 2006; Xu et al., 2007), a new generation FQs such as gatifloxacin and moxifloxacin has been introduced to provide better therapeutic effects and counteract bacterial resistance (Kowalski et al., 2003). The significant disadvantages of the new generation, however, have resulted in their restricted use (Faich et al., 2004) due to serious side effects (e.g. hyperglycemia (Yip and Lee, 2006) and hallucination (Adams and Tavakoli, 2006)), and the genotoxic potentials of these compounds have been reported to be higher than those of other FQs based on in vitro bioassay (Hartmann et al., 1998; Hu et al., 2007). Therefore, the occurrence of a broad range of FQs is important for understanding their environmental risks. In most previous investigations, of all the FQs only ciprofloxacin, norfloxacin, and ofloxacin have been included as major target compounds possibly due to a lack in analytical methods (Castiglioni et al., 2005, 2006; Ghosh et al., 2009; Golet et al., 2001; Miao et al., 2004; Renew and Huang, 2004; Shi et al., 2009; Vieno et al., 2007, 2006; Xu et al., 2007). Although sensitive liquid chromatographyeelectrospray tandem mass spectrometry (LC-MS/MS) has been established for 20 quinolone and fluoroquinolone antibiotics in various water matrices (Xiao et al., 2008), there is still no method available for these compounds in solid samples as it is difficult to develop a method with high extraction efficiencies and low matrix effects for so many antibiotics in solid samples. Removal of FQ antibiotics in STPs plays a crucial role in their pollution control, since most FQs used to treat humans or livestock are ultimately discharged into the aquatic environment via STPs. While the presence of some FQs in STP influents and effluents have been widely investigated (Castiglioni et al., 2005, 2006; Ghosh et al., 2009; Golet et al., 2001; Miao et al., 2004; Renew and Huang, 2004; Shi et al., 2009; Vieno et al., 2007, 2006; Xu et al., 2007), little is known about the fates and major removal units of these antibiotics. Recently, the fates of norfloxacin and ciprofloxacin in two European STPs (Sweden and Switzerland) with mechanical and anaerobic treatment processes were reported (Golet et al., 2003; Lindberg et al., 2006). It is well known that anoxic and aerobic treatment units are very common in STP systems, and Wu et al. (2008) reported that some antibiotics are more persistent under anaerobic conditions than aerobic conditions. However, the fate of norfloxacin and ciprofloxacin in anoxic and aerobic units remains unclear, and no research has been conducted on the fates of other FQs. Thus, assessing the removal efficiencies of all FQs in anoxic or aerobic treatment units could provide basic information for the improvement of STP performance. In this study, we developed a single solid-phase extraction (SPE) method that allowed the simultaneous analysis of nineteen FQ antibiotics (cinoxacin, lomefloxacin, pipemidic acid, ofloxacin, danofloxacin, enrofloxacin, ciprofloxacin,
sarafloxacin, difloxacin, sparfloxacin, moxifloxacin, fleroxacin, norfloxacin, oxolinic acid, pefloxacin, flumequine, nalidixic acid, piromidic acid, and gatifloxacin, shown in Supplementary Data Fig. S1) in sewage sludge. The method was then applied to investigate the fate of each compound during STP treatment with complete mechanical, anaerobic, anoxic, and aerobic treatment processes. The removal efficiencies of detected FQs were evaluated, and their potential removal mechanisms in the STP were explored based on the mass balance analysis in anaerobic, anoxic, and aerobic treatment units.
2.
Materials and methods
2.1.
Sample collection
Qinghe STP (Beijing, China) serves a residential population of about 814,000, and its incoming raw sewage consists mainly of domestic wastewater at a rate of 400,000 m3/day. The sewage is firstly treated with a screen and an aerated grit chamber as the primary clarification. The primary sludge is pumped into the dewater room, while the primary effluent flows directly through the activated sludge system, which is comprised of anaerobic, anoxic, and aerobic units. After a final secondary clarification step, the effluent of the activated sludge reactor is discharged into the receiving river. The scheme of the STP and sampling locations are shown in Supplementary Data (Fig. S2). The hydraulic retention times in the aerated grit chamber, anaerobic tank, anoxic tank, and aerobic tank were 15, 2, 3, and 11 h, respectively, and total solid retention time was 20e25 d. To investigate the fates of quinolones and fluoroquinolones in the Qinghe STP, three-day sampling was carried out in dry weather at the outlet of each treatment step on July 7, 8, and 11, 2008. During this period, the sewage for treatment was around 200,000 m3/d, and water temperatures were 23e25 C. Several regularly measured parameters (e.g., BOD5, COD, and DO, NH4eN, and NO3eN) are displayed in Table S3 (Supplementary Data). Raw sewage and primary and secondary effluents were taken as 24 h flow-proportional composite samples, and the suspended solids were collected by filtering these water samples. All water samples were collected in 10-L amber glass bottles, which were washed with 0.5 g/L ethylenediamine tetraacetic acid disodium (Na2EDTA) water solution, followed by methanol and purified water before use. The sludge samples were generally taken at the outlet of every treatment step. Both the water and sludge samples were prepared for FQ analysis to assess the mass balance of the compounds in the STP. Water samples were extracted on the same day after being filtered by a glass microfiber filter GF/C 1.2 mm (Whatman, Maidstone, UK), and sludge samples were centrifuged and stored at 20 C until analysis.
2.2.
Sample preparation
Reagents and materials used in the analysis were shown in the Supplementary Data. The methods used to quantify 19 FQs in wastewater samples have been published previously (Xiao et al., 2008). Briefly, after filtration, wastewater samples
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 8 7 e3 9 4
were added with Na2EDTA (0.5% w/v) and adjusted to pH 3 with hydrochloric acid (HCl). The HLB cartridge (Waters Corporation, Milford, MA, USA) was preconditioned by 6 mL of methylene chloride, 6 mL of methanol, and 6 mL of purified water (0.5% Na2EDTA w/v, pH ¼ 3). Approximately 250 mL of water spiked with 40 ng of surrogate (norfloxacin-d5) was extracted with HLB cartridge at a flow rate of 5e10 mL/min. The cartridges were rinsed with 10 mL of distilled water, and were then dried under a flow of nitrogen. All target compounds were eluted with 6 mL of methanol containing 0.1% formic acid (v/v). The elute was evaporated to dryness under a gentle stream of nitrogen, and dissolved in 500 mL of a mixture of methanol and pure water containing 0.1% formic acid (1:9, v/v). Analyses of all target compounds were carried out using ultra performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS, Waters Acquity UPLC, Milford, MA, USA). After freeze-dried in a freeze dryer (FDU-200, EYELA, Japan), about 0.1 g of sludge was spiked with 40 ng of norfloxacin-d5 in a 50 mL centrifuge tube and stayed overnight before extraction. A total of 10 mL of triethylamine/methanol/ purified water (5/25/75, v/v/v) was added to the tube, and the mixture was shaken for 20 min at 300 rpm. After 20 min sonication, the sample was centrifuged for 10 min at 3000 rpm. Following centrifugation, the supernatant was transferred to an amber glass bottle. The extraction was repeated twice and all three extracts were combined. The extract was diluted with 1 L of pure water, and then extracted by HLB cartridge with the same procedures of water samples described above. The detail of UPLC-MS/MS analysis was provided in the Supplementary Data.
2.3.
Quantitation and quality control
Identification of the target quinolone and fluoroquinolone antibiotics was accomplished by comparing the retention time (within 2%) and the ratio (within 20%) of the two selected multiple-reaction monitoring (MRM) ion transitions with those of standards. Laboratory blanks made every day from purified water were analyzed to assess potential sample contamination. In this study, concentrations of laboratory blank samples were all lower than detection limits. To compensate for the loss of target compounds during the extraction process and correct the variation of instrument response and matrix effect, norfloxacin-d5 surrogate was spiked to samples prior to extraction. The same instrumental method which has been reported to determinate FQs in wastewater samples (Xiao et al., 2008) was used for quantifying 19 FQs in sludge samples. Namely, 11 FQs (pipemidic acid, fleroxacin, ofloxacin, pefloxacin, enoxacin, norfloxacin, ciprofloxacin, danofloxacin, enrofloxacin, lomefloxacin and difloxacin) were quantified relative to norfloxacin-d5 surrogate, and the method of external calibration was applied for quantification of the other nine target antibiotics. Recovery experiments were conducted to further assess method accuracy and precision. Samples collected from raw sewage, secondary effluent, and excess sludge samples were spiked with standard solutions of at least three times the original concentration, which was determined prior to the fortification experiment. Recoveries for spiked samples were 79e106%,
389
75e97% and 67e98% for raw sewage, secondary effluent, and excess sludge samples, respectively. Method detection limits (MDLs) were based on the peak-to-peak noise of the baseline near the analyte peak obtained by analyzing field samples and on a minimum value of 3 for signal-to-noise. For those nondetected chemicals, samples were spiked using a mixture of standard solution. The MDLs for the target antibiotics ranged from 0.0085 to 0.085 mg/kg in the sludge samples, and were 0.5e37 ng/L in wastewater samples.
2.4.
Data analysis
To assess the contribution of sorption and degradation of FQ antibiotics in the STP, we took initial raw sewage loading (including dissolved and suspended solid phases) as the system input (100%), while the system output consisted of (i) secondary effluent, and (ii) dewatered sludge (Fan et al., 2011). The third part was expressed as (iii) lost, due to the total effect of degradation or transformation mechanism in each treatment unit within the STP, and was calculated as W Lost ¼ W Influent W Effluent W Sludge
(1)
where the “W” was the mass of total FQs within the STP. The mass change percentage (%) in each treatment unit was calculated using ðWInflow WOutflow Þ=WInflow 100% to assess the mass variations of FQs under different treatment processes, where WInflow and WOutflow respectively represent the total mass flow (aqueous phase and sorbed phase) of the detected compound in the inflow and outflow of the individual treatment unit. The calculations of mass flow and solidewater partition coefficient (Kd) in STP treatment units were provided in the Supplementary Data.
3.
Results and discussion
3.1.
Method performance
To explore the fates of FQ antibiotics within STPs, several studies have developed analytical methods for norfloxacin, ciprofloxacin, and ofloxacin in sludge samples (Golet et al., 2002, 2003; Lindberg et al., 2006, 2005). Extraction efficiencies and matrix effects were reported to greatly influence the analysis of antibiotics in sludge samples (Gobel et al., 2002, 2005). In the current study, different extraction solvents were optimized for simultaneously analyzing 19 quinolone and fluoroquinolone antibiotics from sludge samples with high extraction efficiencies and low matrix effects. Of the nineteen target compounds, seven FQs had very poor recoveries (<40%) when extracted with methanol/citrate buffer (1/1, v/v, pH 2.6) reported in previous studies (Jacobsen et al., 2004; Kim and Carlson, 2007), and even low recoveries were observed with acetonitrile/phosphate buffer (1/1, v/v, pH 2.2) used previously for FQs analysis (Golet et al., 2002, 2003) (Table S2 in Supplementary Data). Following methods reported for norfloxacin, ciprofloxacin, and ofloxacin in STPs (extracted with phosphate buffer (pH 6.0) and triethylamine/methanol/purified water (5/25/75, v/v/v, pH 11.0)) (Lindberg et al., 2005), relatively high recoveries in sludge samples (61e80%) were observed; however, recoveries of
390
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 8 7 e3 9 4
pipemidic acid, pefloxacin, and norfloxacin-d5, which have not been analyzed previously, were limited to 49e55%. Since norfloxacin-d5 was used as a surrogate in the current study, its good recovery was important in the quantification of all target compounds. Since FQs were positively-charged compounds, relatively high recoveries of FQs were obtained by basic extraction solutions compared to acidified solutions (Golet et al., 2002, 2003; Lindberg et al., 2005). Thus, we extracted sludge samples three times with triethylamine/methanol/purified water (5/25/75, v/v/v, pH 11.0). The recoveries of all target compounds including pipemidic acid, pefloxacin, norfloxacin, ciprofloxacin, enrofloxacin, and norfloxacin-d5 improved greatly (67e98%) (Table S2). To assess matrix effects of the extraction solvent and purification steps, the percentages of signal intensity in the extracted sludge sample versus the signal of the same concentration in pure solvent (methanol) were calculated. The matrix effects for all target antibiotics ranged from 13 to 26%, and the MDLs of the current method in sludge samples ranged from 0.0085 to 0.085 mg/kg (Table S2 in Supplementary Data). MDLs of ofloxacin (0.028 mg/kg), norfloxacin (0.036 mg/kg) and ciprofloxacin (0.028 mg/kg) were improved compared with those (0.1 mg/kg) in previous studies (Lindberg et al., 2006, 2005).
3.2.
Occurrence
Of the 19 target antibiotics, 10 compounds, including pipemidic acid, fleroxacin, ofloxacin, norfloxacin, ciprofloxacin, enrofloxacin, lomefloxacin, sparfloxacin, gatifloxacin, and moxifloxacin, were detected in wastewater samples from Qinghe STP (Table 1). Ofloxacin (1287 97 ng/L) and norfloxacin (775 87 ng/L) were the dominant FQs in the raw sewage, accounting for 50 1% and 30 0.6% of total concentrations, respectively (Table 1). Relatively low concentrations were detected in the raw sewage for lomefloxacin (162 4 ng/L), ciprofloxacin (99 21 ng/L), pipemidic acid (86 17 ng/L), moxifloxacin (72 34 ng/L), gatifloxacin (66 7 ng/L), fleroxacin (14 1 ng/L), enrofloxacin (8.3 3.2 ng/L), and sparfloxacin (4.4 0.3 ng/L). The profile of detected FQs was constant throughout the STP treatment units, and concentrations of individual compounds ranged from 1.1 ng/L (sparfloxacin) to 528 ng/L (ofloxacin) in secondary effluent (Table 1). It should be
noted that concentrations of most FQs in secondary effluent were higher than those in the aerobic effluent sample, which were possibly due to desorption of the antibiotics from sludge in the secondary clarifier. Of all the target compounds in the current study, three FQs (ofloxacin, norfloxacin, and ciprofloxacin) have been extensively investigated by previous studies, and the predominance of ofloxacin and norfloxacin were also observed in STP wastewater in most developed countries (Castiglioni et al., 2005; Ghosh et al., 2009; Golet et al., 2001, 2003; Lindberg et al., 2006, 2005; Miao et al., 2004; Renew and Huang, 2004; Vieno et al., 2007, 2006). In the present study, raw sewage concentrations of ofloxacin and norfloxacin were higher than those in most developed countries (norfloxacin: 155e486 ng/L in Japan (Ghosh et al., 2009), ofloxacin: 19e287 ng/L and norfloxacin: 72e174 ng/L in Sweden (Lindberg et al., 2005), ofloxacin: 20e350 ng/L and norfloxacin: 13e960 ng/ L in Finland (Vieno et al., 2007)). However, ciprofloxacin, generally reported with similar or higher levels compared to norfloxacin in the USA, Finland, Sweden, and Switzerland (Golet et al., 2003; Lindberg et al., 2006, 2005; Renew and Huang, 2004; Vieno et al., 2007, 2006), was detected at concentrations of only about 13% of norfloxacin in raw sewage from the Qinghe STP. Similar concentration ratios between ciprofloxacin and norfloxacin were also reported in five STPs in Shanghai, China (Shi et al., 2009). Such a profile of FQs in Chinese wastewater is in accordance with the low production and consumption of ciprofloxacin compared with other countries. It has been documented that the production of ciprofloxacin was about 40% that of norfloxacin in China in 1999 (Chemical and Industry Association of China, 2000), and consumption of ciprofloxacin was about 20% that of norfloxacin in Beijing hospitals between 2003 and 2008 (Liu et al., 2010), while similar or higher production and consumption for ciprofloxacin compared with norfloxacin has been reported in Europe and North America (Golet et al., 2001, 2002; Vieno et al., 2007). It should be noted that fourth generation FQs, including gatifloxacin and moxifloxacin, were detected in raw sewage (66 7 and 72 34 ng/L) and secondary effluents (40 8 and 40 20 ng/L), suggesting that these compounds should not be neglected in future studies. In the suspended solids and sludge samples, ten antibiotics in wastewater and sarafloxacin (only in sludge) were
Table 1 e Concentrations of quinolone and fluoroquinolone antibiotics in sewage water (ng/L). Chemicals
Pipemidic acid Fleroxacin Ofloxacin Norfloxacin Ciprofloxacin Enrofloxacin Lomefloxacin Sarafloxacin Gatifloxacin Sparfloxacin Moxifloxacin
Raw sewage
Primary effluent
Anaerobic
Anoxic
Aerobic
Secondary effluent
Return sludge
Excess sludge
Supernatant of excess sludge
86 17
78 12
28 3
22 6
22 8
33 12
18 8
52 7
33 12
15 1 1575 101 831 98 105 13 6.6 3.2 186 26 <MDL 72 9 3.9 0.3 98 45
7.2 0.7 546 165 249 29 37 5 2.6 1.1 85 21 <MDL 37 11 1.0 0.2 36 3
4.9 0.4 384 22 207 38 30 5 2.5 0.7 71 16 <MDL 29 4 0.7 0.1 31 4
5.2 1.0 336 71 190 52 28 7 2.6 1.1 62 17 <MDL 25 5 0.7 0.2 33 7
5.2 0.3 528 89 256 64 37 16 2.1 0.6 71 14 <MDL 40 8 1.1 0.2 40 20
3.6 0.5 238 83 155 85 23 12 0.7 0.4 52 24 <MDL 20 6 0.6 0.3 21 11
11 1 1139 58 610 14 88 5 4.4 1.4 295 2 <MDL 111 1 5.5 0.3 85 24
5.2 0.3 528 89 256 64 37 16 2.1 0.6 71 14 <MDL 40 8 1.1 0.2 40 20
14 1 1287 97 775 87 99 21 8.3 3.2 162 4 <MDL 66 7 4.4 0.3 72 34
391
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 8 7 e3 9 4
Table 2 e Concentrations of Quinolone and Fluoroquinolone Antibiotics in Sewage Sludge (mg/kg).
Mass flow and mass balance
Mass flows and mass balance of individual FQs were determined to assess their potential removal mechanisms in the STP (Fig. 2 and Supplementary Data Table S4). In raw sewage, the combined aqueous and solid phase mass flows of all detected FQs were about 612 g/d in the Qinghe STP, and mass flows entering the STP were 515 48 g/d and 97 14 g/d in the dissolved and sorbed fraction, respectively. The proportion of
0.32 0.09 7.57 7.25 1.15 0.10 1.18 0.62 0.48 0.04 0.57
0.10 0.01 0.61 1.05 0.14 0.01 0.23 0.11 0.12 0.01 0.13
0.27 0.09 7.40 6.40 0.96 0.07 0.98 0.39 0.42 0.03 0.56
0.04 0.02 0.57 0.46 0.07 0.01 0.05 0.30 0.05 0.01 0.01
0.27 0.08 7.79 7.23 1.04 0.07 1.00 0.53 0.49 0.04 0.53
0.03 0.01 0.55 0.22 0.14 0.03 0.08 0.03 0.03 0.01 0.10
FQs sorbed to particles in the raw sewage ranged from 7% (pipemidic acid, ofloxacin) to 39% (moxifloxacin), and those of norfloxacin (23%) and ciprofloxacin (39%) were similar to that reported in an STP in Switzerland (33%) (Golet et al., 2003), but lower than that in Sweden (80%) (Lindberg et al., 2006). The different sorption proportions were possibly due to the different characterizations of the sewage in the various cities. Total FQs mass flow entering the activated sludge system was about 595 38 g/d, with individual chemicals varying from 0.8 (sparfloxacin) to 315 g/d (ofloxacin). In the activated sludge system, mass change percentages were 13 to 16%, 4 to 22%, and 33 to 2% in anaerobic, anoxic, and aerobic treatment units, respectively. The negative mass change percentages were due to the potential variations between the three sampling days and the analytical RSD. The low mass change percentages and variability in the three treatment units 100
4th
80
60
40
20
0
Gatifloxacin
3.3.
0.30 0.05 0.10 0.01 7.35 0.79 7.55 0.70 1.05 0.13 0.09 0.04 1.06 0.12 0.36 0.27 0.48 0.03 0.03 0.01 0.56 0.06
Moxifloxacin
detected (Table 2). Ofloxacin and norfloxacin were also dominant in the suspended solids and sludge samples. Relatively high concentrations of ofloxacin (7.29 0.45 mg/ kg) and norfloxacin (7.01 0.51 mg/kg) in the sludge samples were found in the current study compared with other countries (0.1e4.2 mg/kg), but concentrations of ciprofloxacin (0.20 0.05 mg/kg) were lower than those reported previously (0.5e7.7 mg/kg) (Golet et al., 2003; Lindberg et al., 2006, 2005). The profile of FQs in sludge samples was consistent with that in wastewater. In all the treatment units, concentrations of detected FQs in excess sludge were 3e9 folds higher than those in return sludge (Table 1), which were possibly due to the fact that excess sludge stored at least three days for thickening before further treatment, but return sludge was rapidly pump back to the active sludge treatment units. Mean removal efficiencies, calculated by comparing concentrations in the raw sewage and secondary effluent, ranged from 40% to 75% for the ten detected FQ antibiotics (Fig. 1). The removal efficiencies of some FQs in the current study were in the range of those previously reported for ciprofloxacin (37e86%), ofloxacin (33e66%), norfloxacin (58e87%), and lomefloxacin (21e72%) (Castiglioni et al., 2005; Lindberg et al., 2005; Shi et al., 2009; Xu et al., 2007). Compound-specific removal efficiencies were firstly observed for the detected FQs, and it should be noted that the lowest removal efficiencies were found for the new generation drugs, moxifloxacin (40 5%) and gatifloxacin (43 13%) (Fig. 1). Since moxifloxacin and gatifloxacin have been withdrawn or classified for restricted use in North American due to their serious side effects (Adams and Tavakoli, 2006; Faich et al., 2004; Yip and Lee, 2006), the presence and low removal efficiencies of these compounds in STPs indicates the importance of strict control of these antibiotics in China.
0.01 0.01 0.65 0.15 0.07 0.01 0.01 0.05 0.01 0.01 0.02
Dewatered sludge
Lomefloxacin
0.27 0.08 7.24 6.67 0.99 0.08 0.97 0.49 0.40 0.03 0.46
Excess sludge
Ofloxacin
0.29 0.01 0.08 0.01 6.36 0.36 6.39 0.33 0.93 0.02 0.10 0.01 1.01 0.05 0.22 0.30 0.39 0.03 0.03 0.01 0.51 0.02
Return sludge
Fleroxacin
0.04 0.05 0.02 0.01 0.33 0.07 1.06 0.36 0.22 0.08 0.02 0.01 0.06 0.03 <MDL 0.09 0.04 0.01 0.01 0.17 0.04
Aerobic
Ciprofloxacin
0.02 0.01 0.02 0.01 0.35 0.07 0.86 0.40 0.20 0.05 0.02 0.01 0.06 0.01 <MDL 0.07 0.03 0.01 0.01 0.17 0.03
Anoxic
Pipemidic acid
Anaerobic
Norfloxacin
Primary sludge
Enrofloxacin
Pipemidic acid Fleroxacin Ofloxacin Norfloxacin Ciprofloxacin Enrofloxacin Lomefloxacin Sarafloxacin Gatifloxacin Sparfloxacin Moxifloxacin
Raw sludge
Sparfloxacin
Chemicals
Fig. 1 e Removal efficiencies (%) of detected FQ antibiotics in Qinghe STP. Numbers indicate the new generation of the FQ antibiotics used in human and veterinary medicine.
392
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 8 7 e3 9 4
101 Internal recirculation
425 51095 515
Aerated grit Chamber 595 48
Anaerobic 413 38
97 14 Raw sewage
29874
2273
Anoxic
Aerobic 786
91
79291
2790
110 63 33622 3
67 8 Primary sludge
1.4
708
89
Secondary clarifier 203
168
Secondary effluent
85158 3788
6262
44
Return Sludge
3030
Supernatant of excess sludge
5 410
Dewatered sludge 439 155
2.0
Excess sludge
148 dissolved sorbed
Fig. 2 e Total mass flows (g/d) of detected quinolone and fluoroquinolone antibiotics in STP processes.
Since it has been reported that adsorption of chemicals was correlated with the cation exchange capacity (CEC) (Hang and Brindley, 1970; Jaynes and Boyd, 1991), CEC values of sludge samples from the three treatment units were measured to assess the variation of Kd. The CEC values were 149 3.7, 211 1.3, and 164 7.1 cmolc/kg dw for sludge samples collected from aerobic, anoxic and anaerobic units, respectively, and no significant correlation was found between CEC and Kd of all detected FQs. On the other hand, it was found that the hydraulic retention times in aerobic (11 h), anoxic (3 h) and
Psludge
Peffluent
Plost
120 100 80 60 40 20 0
Gatifloxacin
Moxifloxacin
Lomefloxacin
Ofloxacin
Fleroxacin
Ciprofloxacin
Pipemidic acid
Norfloxacin
Enrofloxacin
-20 Sparfloxacin
suggested that biodegradation in anaerobic, anoxic, and aerobic treatment units was of minor importance in the removal of FQs. For filtered secondary effluents, the mass flows of FQs were 203 44 g/d in the aqueous phase and varied from 0.2 g/d (sparfloxacin) to 106 g/d (ofloxacin), while the sorbed amount was ignored due to the low concentrations of suspended solids in the effluent. The excess sludge was then moved through a dewaterer without further digestion, and the total FQs mass in the dewatered sludge in the Qinghe STP was 439 155 g/d, varied from 263 to 552 g/d during the three sampling days. The mass balance of FQs in the Qinghe STP were expressed in chemical mass fractions (%) detected in (i) secondary effluent, (ii) dewatered sludge, and (iii) lost, relative to the calculated initial loading (100%) (Fig. 3). We observed that the calculated fraction of most FQs in the dewatered sludge accounted for 50e87% (except for pipemidic acid, 36%) of the initial loadings, while less than 47% of the total amount of FQs was found in the effluents. Mass balance of ciprofloxacin and norfloxacin in anaerobic treatment process has only been reported in Switzerland and Sweden (Golet et al., 2003; Lindberg et al., 2006). In those studies, the mass fractions for ciprofloxacin and norfloxacin were 72e83% in dewatered sludge (Golet et al., 2003; Lindberg et al., 2006), which are comparable to those in the present study (85e87%). These results further confirmed that sorption via sludge was the major removal mechanism of FQs in the STP. To further understand compound-specific removal efficiencies, pKa, octanol-water coefficient (log Pow, calculated by ACD/log Pow ver. 1.0 (Advanced Chemistry Development, Inc.)), and the apparent solidewater partition coefficients (Kd) were calculated (Table S5 in Supplementary Data). The average Kd values of target FQs in current study (12,300e37,700, Table S5) were consistent with those of ciprofloxacin, norfloxacin, trovafloxacin, gemifloxacin, sarafloxacin and enrofloxacin in a previous study (12,600e39,800) (Golet et al., 2003). The Kd values of the FQs in sludge obtained in this study were slightly higher in aerobic units than in the anoxic and anaerobic units.
Fig. 3 e Mass proportions of detected quinolone and fluoroquinolone antibiotics in (i) secondary effluent (Peffluent), (ii) dewatered sludge (Psludge) and (iii) total lost (Plost) relative to the calculated initial loading (100%) in Qinghe STP.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 8 7 e3 9 4
anaerobic (2 h) units were consistent with the Kd variations within the three treatment units, thus would influence the Kd values of FQs in the STP. The relatively low log Pow values (<2.5) of the detected FQs except for sarafloxacin indicated that these compounds have a weak sorption potential by hydrophobic interactions (Golet et al., 2003; Hu et al., 2007). A recent study suggested that electrostatic interactions is involved and, in some cases, may be dominating for positively charged compounds absorbed to sludge solid surfaces (StevensGarmon et al., 2011). Given the fact that all the target FQs contained nitrogen as positively-charged moiety (Fig. S1), the high sorption potentials of FQs were possible due to sorption through electrostatic interactions involved with the positively charged locations of these compounds (Golet et al., 2003; Stevens-Garmon et al., 2011). And the positively charged moieties for each chemical should be occurred at the nitrogen atoms, of which the atom partial charges were calculated by MOPAC (version 6, CAChe Scientific Inc., Oxford, U.K.) and showed in Fig. S1. The Kd of moxifloxacin and gatifloxacin were notably lower than the compounds with high removal efficiencies, especially for sparfloxacin, enrofloxacin, norfloxacin, and ciprofloxacin (about 2e3 times lower). The Kd of detected FQs also showed a significant positive correlation with their removal efficiencies (r2 ¼ 0.54, p ¼ 0.02), suggesting that high absorption to the activated sludge resulted in the high removal efficiencies of these compounds in the STP (Fig. 4). This further confirmed that sorption associated with sludge was a major removal mechanism of FQs in raw sewage (Golet et al., 2003; Lindberg et al., 2006). It should be noted that the mass load of FQs in the raw sewage was less than 1% the recycling sludge in the activated sludge treatment (Fig. 2). The high mass load of FQs in sludge in the Qinghe STP could result from the long-term cycling and accumulation of FQs in the STP. This also indicated that FQs
100
80
60
40
20
y = 0.0008x + 41.304 r2 = 0.5412 , p=0.02
0 0
10000
20000
30000
40000
50000
Kd (L/kg) Fig. 4 e Relationships between removal efficiencies (%) and Kd (L/kg) of detected quinolone and fluoroquinolone antibiotics in Qinghe STP.
393
were quite persistent in activated sludge, which is supported by previous research on the long-term persistence of FQs in sludgetreated soils (Golet et al., 2003). Therefore, the long-term cycling and persistence of FQs in STPs indicated that activated sludge was a huge reservoir of FQ antibiotics, emphasizing the importance of sludge management.
Acknowledgements Financial support from the National Nature Science Foundation of China (20837003) and the Education Committee of Beijing (YB20081000103) are gratefully acknowledged.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.10.055.
references
Adams, M., Tavakoli, H., 2006. Gatifloxacin-induced hallucinations in a 19-year-old man. Psychosomatics 47, 360. Ball, P., 2000. Quinolone generations: natural history or natural selection? J. Antimicrob. Chemoth. 46, 17e24. Barnes, K.K., Kolpin, D.W., Furlong, E.T., Zaugg, S.D., Meyer, M.T., Barber, L.B., 2008. A national reconnaissance of pharmaceuticals and other organic wastewater contaminants in the United States e I. Groundwater Sci. Total Environ. 402, 192e200. Castiglioni, S., Bagnati, R., Calamari, D., Fanelli, R., Zuccato, E., 2005. A multiresidue analytical method using solid-phase extraction and high-pressure liquid chromatography tandem mass spectrometry to measure pharmaceuticals of different therapeutic classes in urban wastewaters. J. Chromatogr. A 1092, 206e215. Castiglioni, S., Bagnati, R., Fanelli, R., Pomati, F., Calamari, D., Zuccato, E., 2006. Removal of pharmaceuticals in sewage treatment plants in Italy. Environ. Sci. Technol. 40, 357e363. Chemical and Industry Association of China, 2000. 19990 Statistics Information of Chemial and Pharmaceutical Industry. Chemical and Industry Association of China, China. Daughton, C.G., Ternes, T.A., 1999. Pharmaceuticals and personal care products in the environment: agents of subtle change? Environ. Health Persp. 107, 907e938. Diaz-Cruz, M.S., Barcelo, D., 2006. Determination of antimicrobial residues and metabolites in the aquatic environment by liquid chromatography tandem mass spectrometry. Anal. Bioanal. Chem. 386, 973e985. Faich, G.A., Morganroth, J., Whitehouse, A.B., Brar, J.S., Arcuri, P., Kowalsky, S.F., Haverstock, D.C., Celesk, R.A., Church, D.A., 2004. Clinical experience with moxifloxacin in patients with respiratory tract infections. Ann. Pharmacother 38, 749e754. Fan, Z.L., Wu, S.M., Chang, H., Hu, J.Y., 2011. Behaviors of glucocorticoids, androgens and progestogens in a municipal sewage treatment plant: comparison to estrogens. Environ. Sci. Technol. 45, 2725e2733. Ghosh, G.C., Okuda, T., Yamashita, N., Tanaka, H., 2009. Occurrence and elimination of antibiotics at four sewage treatment plants in Japan and their effects on bacterial ammonia oxidation. Water Sci. Technol. 59, 779e786. Gobel, A., Thomsen, A., McArdell, C.S., Alder, A.C., Giger, W., Theiss, N., Loffler, D., Ternes, T.A., 2005. Extraction and
394
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 8 7 e3 9 4
determination of sulfonamides, macrolides, and trimethoprim in sewage sludge. J. Chromatogr. A 1085, 179e189. Golet, E.M., Alder, A.C., Hartmann, A., Ternes, T.A., Giger, W., 2001. Trace determination of fluoroquinolone antibacterial agents in solid-phase extraction urban wastewater by and liquid chromatography with fluorescence detection. Anal. Chem. 73, 3632e3638. Golet, E.M., Strehler, A., Alder, A.C., Giger, W., 2002. Determination of fluoroquinolone antibacterial agents in sewage sludge and sludge-treated soil using accelerated solvent extraction followed by solid-phase extraction. Anal. Chem. 74, 5455e5462. Golet, E.M., Xifra, I., Siegrist, H., Alder, A.C., Giger, W., 2003. Environmental exposure assessment of fluoroquinolone antibacterial agents from sewage to soil. Environ. Sci. Technol. 37, 3243e3249. Gros, M., Petrovic, M., Barcelo, D., 2007. Wastewater treatment plants as a pathway for aquatic contamination by pharmaceuticals in the Ebro river basin (northeast Spain). Environ. Toxicol. Chem. 26, 1553e1562. Hang, P.T., Brindley, G.W., 1970. Methylene blue absorption by clay minerals. Determination of surface areas and cation exchange capacities (clay-organic studies XVIII). Clays Clay Minerals 18, 203e212. Hartmann, A., Alder, A.C., Koller, T., Widmer, R.M., 1998. Identification of fluoroquinolone antibiotics as the main source of umuC genotoxicity in native hospital wastewater. Environ. Toxicol. Chem. 17, 377e382. Hu, J.Y., Wang, W.F., Zhu, Z., Chang, H., Pan, F., Lin, B.L., 2007. Quantitative structure - activity relationship model for prediction of genotoxic potential for quinolone antibacterials. Environ. Sci. Technol. 41, 4806e4812. Hu, J.Y., Shi, J.C., Chang, H., Li, D., Yang, M., Kamagata, Y.C., 2008. Phenotyping and genotyping of antihiotic-resistant Escherichia coli isolated from a natural river basin. Environ. Sci. Technol. 42, 3415e3420. Jacobsen, A.M., Halling-Sorensen, B., Ingerslev, F., Hansen, S.H., 2004. Simultaneous extraction of tetracycline, macrolide and sulfonamide antibiotics from agricultural soils using pressurised liquid extraction, followed by solid-phase extraction and liquid chromatography-tandem mass spectrometry. J. Chromatogr. A 1038, 157e170. Jaynes, W.F., Boyd, S.A., 1991. Clay mineral type and organic compound sorption by hexadecyltrimethylammoniumexchanged clays. Soil Sci. Soc. Am. J. 55 (1), 43e48. Khetan, S.K., Collins, T.J., 2007. Human pharmaceuticals in the aquatic environment: a challenge to green chemistry. Chem. Rev. 107, 2319e2364. Kim, S.C., Carlson, K., 2007. Temporal and spatial trends in the occurrence of human and veterinary antibiotics in aqueous and river sediment matrices. Environ. Sci. Technol. 41, 50e57. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S.D., Barber, L.B., Buxton, H.T., 2002. Pharmaceuticals, hormones, and other organic wastewater contaminants in US streams, 1999e2000: a national reconnaissance. Environ. Sci. Technol. 36, 1202e1211. Kowalski, R.P., Dhaliwal, D.K., Karenchak, L.M., Romanowski, E.G., Mah, F.S., Ritterband, D.C., Gordon, Y.J., 2003. Gatifloxacin and moxifloxacin: an in vitro susceptibility comparison to levofloxacin, ciprofloxacin, and ofloxacin using bacterial keratitis isolates. Am. J. Ophthalmol. 136, 500e505. Lindberg, R.H., Wennberg, P., Johansson, M.I., Tysklind, M., Andersson, B., 2005. Screening of human antibiotic substances and determination of weekly mass flows in five sewage treatment plants in Sweden. Environ. Sci. Technol. 39, 3421e3429. Lindberg, R.H., Olofsson, U., Rendahl, P., Johansson, M.I., Tysklind, M., Andersson, B., 2006. Behavior of
fluoroquinolones and trimethoprim during mechanical, chemical, and active sludge treatment of sewage water and digestion of sludge. Environ. Sci. Technol. 40, 1042e1048. Liu, Z.J., Chen, D.K., Fu, D.X., Du, G.H., Hu, X., 2010. Evaluation and analysis on the utilization of quinolones for the inpatients of Beijing hospital of Ministry of Health during 2003w2008. Eval. Anal. Drug-Use Hospitals China 10, 201e203. Martinez, M., McDermott, P., Walker, R., 2006. Pharmacology of the fluoroquinolones: a perspective for the use in domestic animals. Vet. J. 172, 10e28. Miao, X.S., Bishay, F., Chen, M., Metcalfe, C.D., 2004. Occurrence of antimicrobials in the final effluents of wastewater treatment plants in Canada. Environ. Sci. Technol. 38, 3533e3541. Polk, R.E., Johnson, C.K., McClish, D., Wenzel, R.P., Edmond, M.B., 2004. Predicting hospital rates of fluoroquinolone-resistant Pseudomonas aeruginosa from fluoroquinolone use in US hospitals and their surrounding communities. Clin. Infect. Dis. 39, 497e503. Reinthaler, F.F., Posch, J., Feierl, G., Wust, G., Haas, D., Ruckenbauer, G., Mascher, F., Marth, E., 2003. Antibiotic resistance of E-coli in sewage and sludge. Water Res. 37, 1685e1690. Renew, J.E., Huang, C.H., 2004. Simultaneous determination of fluoroquinolone, sulfonamide, and trimethoprim antibiotics in wastewater using tandem solid phase extraction and liquid chromatography-electrospray mass spectrometry. J. Chromatogr. A 1042, 113e121. Shi, L., Zhou, X.F., Zhang, Y.L., Gu, G.W., 2009. Simultaneous determination of 8 fluoroquinolone antibiotics in sewage treatment plants by solid-phase extraction and liquid chromatography with fluorescence detection. Water Sci. Technol. 59, 805e813. Stevens-Garmon, J., Drewes, J., Khan, S., McDonald, J., Dickenson, E., 2011. Sorption of emerging trace organic compounds onto wastewater sludge solids. Water Res. 45, 3417e3426. Taylor, N.M., Davies, R.H., Ridley, A., Clouting, C., Wales, A.D., Clifton-Hadley, F.A., 2008. A survey of fluoroquinolone resistance in Escherichia coli and thermophilic Campylobacter spp. on poultry and pig farms in Great Britain. J. Appl. Microbiol. 105, 1421e1431. Vieno, N.M., Tuhkanen, T., Kronberg, L., 2006. Analysis of neutral and basic pharmaceuticals in sewage treatment plants and in recipient rivers using solid phase extraction and liquid chromatography-tandem mass spectrometry detection. J. Chromatogr. A 1134, 101e111. Vieno, N., Tuhkanen, T., Kronberg, L., 2007. Elimination of pharmaceuticals in sewage treatment plants in Finland. Water Res. 41, 1001e1012. Wu, C.X., Spongberg, A.L., Witter, J.D., 2008. Determination of the persistence of pharmaceuticals in biosolids using liquidchromatography tandem mass spectrometry. Chemosphere 73, 511e518. Xiao, Y., Chang, H., Jia, A., Hu, J.Y., 2008. Trace analysis of quinolone and fluoroquinolone antibiotics from wastewaters by liquid chromatography-electrospray tandem mass spectrometry. J. Chromatogr. A 1214, 100e108. Xu, W.H., Zhang, G., Li, X.D., Zou, S.C., Li, P., Hu, Z.H., Li, J., 2007. Occurrence and elimination of antibiotics at four sewage treatment plants in the Pearl river Delta (PRD), South China. Water Res. 41, 4526e4534. Ye, Z.Q., Weinberg, H.S., Meyer, M.T., 2007. Trace analysis of trimethoprim and sulfonamide, macrolide, quinolone, and tetracycline antibiotics in chlorinated drinking water using liquid chromatography electrospray tandem mass spectrometry. Anal. Chem. 79, 1135e1144. Yip, C., Lee, A.J., 2006. Gatifloxacin-induced hyperglycemia: a case report and summary of the current literature. Clin. Ther. 28, 1857e1866.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 9 5 e4 0 2
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Caffeine as an indicator for the quantification of untreated wastewater in karst systems Olav Hillebrand*, Karsten No¨dler, Tobias Licha, Martin Sauter, Tobias Geyer Geoscience Center, University of Go¨ttingen, Goldschmidtstrasse 3, D-37077 Go¨ttingen, Germany
article info
abstract
Article history:
Contamination from untreated wastewater leakage and related bacterial contamination
Received 15 August 2011
poses a threat to drinking water quality. However, a quantification of the magnitude of
Received in revised form
leakage is difficult. The objective of this work is to provide a highly sensitive methodology
25 October 2011
for the estimation of the mass of untreated wastewater entering karst aquifers with rapid
Accepted 1 November 2011
recharge. For this purpose a balance approach is adapted. It is based on the mass flow of
Available online 9 November 2011
caffeine in spring water, the load of caffeine in untreated wastewater and the daily water consumption per person in a spring catchment area. Caffeine is a source-specific indicator
Keywords:
for wastewater, consumed and discharged in quantities allowing detection in a karst
Groundwater quality
spring. The methodology was applied to estimate the amount of leaking and infiltrating
Karst vulnerability
wastewater to a well investigated karst aquifer on a daily basis. The calculated mean
Direct sewage leakage
volume of untreated wastewater entering the aquifer was found to be 2.2 0.5 m3 d1
Human impact
(undiluted wastewater). It corresponds to approximately 0.4% of the total amount of wastewater within the spring catchment. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Karst aquifers represent important drinking water resources supplying up to 25% of the world’s population with fresh water (Ford and Williams, 2007). Flow and transport in these types of aquifers can be very rapid due to the presence of highly permeable flow paths, e.g. karst conduits in the subsurface (Geyer et al., 2007). Because of thin soil coverage with a low field capacity and highly permeable local infiltration pathways, groundwater recharge generally occurs very rapidly after precipitation events. Consequently, karst springs are highly vulnerable and show strong variations in spring discharge and water quality (Heinz et al., 2009). Point-source input of wastewater has been shown to be a serious threat to groundwater quality, as it is related to
contamination with fecal bacteria (Heinz et al., 2009; Hrudey et al., 2003). Despite the short survival time of coliform bacteria in the environment (McFeters and Stuart, 1972), rapidly transported, untreated wastewater can pose a substantial health risk. Contamination of karst groundwater from point-sources can be triggered when single massive inputs of wastewater reach the aquifer in a short time by e.g. overflow of retention basins or failure of wastewater lagoons (Heinz et al., 2009, 2006; Memon and Azmeh, 2001). Furthermore, leaky sewers or the overflow from undersized sewer networks may be a threat to groundwater quality (Eiswirth and Ho¨tzl, 1997; Rutsch et al., 2006; Seiler et al., 1999; Gasser et al., 2010). To assess the magnitude of anthropogenic impact on water sources, indicators can be employed. An ideal indicator for
* Corresponding author. Tel.: þ49 551 39 9267; fax: þ49 551 39 9379. E-mail address:
[email protected] (O. Hillebrand). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.003
396
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 9 5 e4 0 2
anthropogenic source water has to be source-specific and released in sufficient quantities to allow detection after dilution in the environment (Takada et al., 1997). The detection of the anthropogenic impact on water bodies using different indicators has been studied before, proposing and using conservative wastewater-specific indicators, such as carbamazepine (Bahlmann et al., 2009; Gasser et al., 2010; Takada et al., 1997), human-specific antibiotics (Glassmeyer et al., 2005) or artificial sweeteners (Buerge et al., 2009). The lack of specificity of conservative indicators for treated or untreated wastewater is a drawback of those anthropogenic indicators, i.e. a distinction between these two types of wastewater is not possible. To assess the magnitude of pollution solely from untreated wastewater, a specific indicator for untreated wastewater is required. Apart from the traditionally used microbial indicators, which have been discussed controversially (Boehm et al., 2002; Heinz et al., 2009; Ogunseitan, 1996), caffeine was found to be a suitable wastewater indicator that allows even the quantification of wastewater burdens in surface waters (Buerge et al., 2006). It is found in several beverages (Siegener and Chen, 2002) and pharmaceuticals. These are the only sources for caffeine in Germany. Seiler et al. (1999) stated, that the main source of caffeine is likely not to be the excreted part of consumed caffeine, but the disposal of unconsumed caffeine-containing beverages, by disposing them down the sink or from rinsing coffee cups. Caffeine is wastewater-specific (Buerge et al., 2003), mobile (Gardinali and Zhao, 2002), shows excellent elimination rates during wastewater processing (Buerge et al., 2006), is degradable in the environment (e.g. Hakil et al., 1998) and has a high detection frequency (this study). Typical loads of caffeine in untreated wastewater are in the range of 16 mg d1 person1 (Buerge et al., 2003). Caffeine is metabolized in the human body mainly to paraxanthine, which is not found in plants nor food (Schmidt and Schoyerer, 1966; Stavric, 1988; Lelo et al., 1986). Another primary metabolite is theobromine. It is not only a metabolites but is also found in beverages, such as tea or cocoa (Srdjenovic et al., 2008). Both metabolites occur in wastewater, but they can as well be produced by microorganisms. However, since 80% of the ingested caffeine is metabolized to paraxanthine in humans (Lelo et al., 1986), its correlation with caffeine can qualitatively be used to support the specificity of caffeine for wastewater. The objective of this work is to transfer the methodological approach of wastewater quantification with caffeine in surface waters (Buerge et al., 2006) to groundwater karst systems. For this purpose a karst spring was sampled during a period of several months with a high temporal resolution. Karst springs are often connected to extended conduit networks and drain large catchment areas. The sampling of a single karst spring therefore allows an integral characterization of a catchment area. Thus, wastewater indicators in springs could play an important role for the identification and quantification of untreated wastewater entering a groundwater system. We hypothesize that caffeine degradation rates and the residence time distribution of water in karst systems allow the indication of rapidly transported, untreated wastewater. Treated wastewater or untreated wastewater that is not transported directly to the spring over highly conductive flow
paths, i.e. with higher residence times, is not expected to affect the indication or the quantification.
2.
Materials and methods
2.1.
Field work
A long-term sampling campaign was conducted with increased event-based sampling frequency.
2.1.1.
Study area
The investigated spring (Gallusquelle) is situated in Southwest Germany (Fig. 1). The spring has an average annual discharge of 0.5 m3 s1 draining a catchment area of 45 km2. The total population in the catchment is ca. 4000 inhabitants. Untreated wastewater can reach the karst aquifer by overflows from a retention basin at approximately 9 km distance to the spring or from sewer leakage. The retention basin belongs to a regional combined wastewater drainage and treatment system in the area. The wastewater from the Gallusquelle catchment is treated in wastewater treatment plants outside of the catchment area.
2.1.2.
Sampling campaign
Due to the high variability of spring discharge and water quality (Geyer et al., 2007) a highly time resolved sampling campaign with 157 water samples was undertaken. Spring water was sampled between March, 3rd and May, 24th 2010. The temporal resolution of sampling was strongly increased during spring discharge events (maximum sampling resolution: 8 samples per day). During recession periods of spring discharge the sampling interval was reduced to one sample per day. The samples were collected in glass bottles, stored at 4 C and preconcentrated within 24 h by solid phase extraction (SPE). A continuous monitoring system with hourly data for electrical conductivity (ref. T of 20 C), turbidity and water
Fig. 1 e The area of investigation. All possible sources for caffeine, settlements and the wastewater retention basin (WWRB), are located in a well defined area (modified after Geyer et al., 2007).
397
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 9 5 e4 0 2
levels was installed at the spring. Measured water levels were translated to spring discharge applying a rating curve. Hourly rainfall data, air temperature and depth of snow were obtained from weather stations (German Meteorological Service, DWD) located close to the catchment area.
2.2.
Laboratory analyses
2.2.1.
Analysis of carbamazepine and methylxanthines
An analytical method based on SPE and high-performance liquid chromatographic separation with tandem mass spectrometric detection (HPLC/MS-MS) was used for the analysis of selected methylxanthines (caffeine and its metabolites paraxanthine, theobromine, theophylline, 1-methylxanthine and 3-methylxanthine) and the anticonvulsant carbamazepine. Details have been published previously (No¨dler et al., 2010). In brief, a 500 mL sample volume was buffered at neutral pH (phosphate buffer) and extracted by SPE (500 mg OASIS HLB, Waters, Eschborn, Germany). Prior to extraction, 200 ng of paraxanthine-D6 and 100 ng of carbamazepine-D10 were added as internal standards for the quantification of xanthines and carbamazepine, respectively. After extraction the sorbent was rinsed with ultrapure water and dried. The cartridge was wrapped in aluminum foil and kept frozen (18 C) until analysis. The analytes were eluted with methanol and ethyl acetate, consecutively. The solvents were evaporated and the dry residue was re-dissolved in an aqueous 5 mM ammonium acetate solution, containing 4% methanol. In contrast to the volume described by No¨dler et al. (2010), only 0.8 mL were used to re-dissolve the analytes, resulting in a higher enrichment factor and thus lower method detection and quantification limits (Table 1).
2.2.2.
Chemicals
Methanol (LC/MS grade) and caffeine were purchased from Fisher Scientific (Schwerte, Germany), ethyl acetate and ammonium acetate (all analytical grade) were purchased from VWR (Darmstadt, Germany). Carbamazepine, paraxanthine, paraxanthine-D6, theobromine, theophylline, 1-methylxanthine and 3-methylxanthine were obtained from Sigma Aldrich (Steinheim, Germany). Carbamazepine-D10 was purchased from Promochem (Wesel, Germany). Potassium dihydrogen phosphate and disodium hydrogen phosphate dihydrate were obtained from VWR (Darmstadt, Germany).
3.
Results and discussion
3.1. Occurrence and variation of caffeine together with related methylxanthines and carbamazepine During the sampling campaign highly variable caffeine concentrations of 10.3 6.3 ng L1 (mean standard deviation) were found in the spring water. High concentrations were observed during winter and as response to precipitation events. Caffeine was detected in 95 of 157 samples. During the sampling campaign a snowmelt event and two rainfall events were observed. The snowmelt event (Fig. 2; t1) resulted in a significant decrease of caffeine concentrations in spring water, while after each rainfall event (Fig. 2; t4a/b) a positive caffeine signal in spring water was observed. Concentrations of caffeine in spring water before the snowmelt are relatively high. However, after the snowmelt caffeine concentrations were frequently lower than the limit of detection. This may be related to a lower input and/or a higher dilution of caffeine in the aquifer system by large quantities of melt water. The comparison of caffeine concentrations and turbidity at the Gallusquelle spring demonstrates that the turbidity curve is only of limited use for wastewater detection. The peak concentration of caffeine after the first rainfall occurs slightly delayed to the turbidity peak and the minimum of the electrical conductivity. It indicates that the caffeine enters the karst aquifer with concentrated point recharge, i.e. recharge through highly conductive flow paths, and is rapidly transported through the aquifer system to the spring. The decrease of electrical conductivity is attributed to the mixing of preevent groundwater and low mineralized event water. The second rainfall does not appear as clear signal in the data of the turbidity. The electrical conductivity develops a plateau instead of a peak, as the rainfall occurred within the recession period of the first rainfall. Apart from the two turbidity peaks produced by the melt event and the two consecutive rainfall events, a third peak (Fig. 2; t3) appears between the melt event and the precipitation events, when a water pipe, belonging to the local water supply company burst. The introduced water was processed drinking water and therefore not related to a contamination with wastewater. Two overflow events of the local wastewater retention basin occurred at the end of March and the beginning of April
Table 1 e Concentrations and detection frequencies of caffeine, two metabolites and carbamazepine in spring water for the method limit of quantification. In total 157 samples were taken and analyzed.
Caffeine Paraxanthine Theobromine Carbamazepine a b c d
MDLa (ng L1)
MQLb (ng L1)
Min (ng L1)
Max (ng L1)
Median (ng L1)
DFc (%)
QFd (%)
1.0 0.7 1.2 0.6
3.4 2.6 4.1 1.8
3.5 2.6 4.1 0
27.0 17.5 29.8 0
7.8 5.2 9.4 0
60.5 56.7 45.9 57.3
38.5 33.7 36.2 0
method detection limit. method quantification limit. detection frequency, n ¼ 157 samples. detection frequency above method quantification limit, n ¼ 157 samples.
398
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 9 5 e4 0 2
Fig. 2 e Concentrations of caffeine and paraxanthine at the spring. The illustrated daily precipitation amounts are calculated from hourly measurements. Caffeine and paraxanthine concentrations in grey show samples below the methods limit of quantification, concentrations on the abscissa represent samples below method’s limit of detection. t1 indicates the commencement of the snowmelt event, t2 the caffeine signal injected by an overflow of the local wastewater retention basin, t3 a burst of a water pipe and t4a/b two rainfall events.
(only 3 days apart), resulting in a single high peak of caffeine (Fig. 2; t2). Only one sample indicates the recharge of wastewater for these events. Which one of the overflow events resulted in a caffeine peak is not known, since the events occurred chronologically very close to each other. As no second peak developed, it is to assume, that one of the two overflow events did not result in wastewater recharge. As shown by Heinz et al. (2009), water from the overflows of the WWRB percolates to the groundwater and reaches the spring within a few days, producing a turbidity peak. It is noteworthy, that the turbidity may be insensitive for small wastewater quantities and did not evolve a peak for any of the two overflow events mentioned above, whereas the caffeine indicated at least one. Apart from the two overflow events all wastewater intrusions are related to sewer leakage. Two other dimethylxanthine compounds, paraxanthine and theobromine, were detected frequently during the sampling campaign and demonstrated reasonable occurrence above their method quantification limits (Table 1). The other analyzed methylxanthines appeared only rarely, which is likely related to their higher method detection limits. We tried to verify our results using carbamazepine, as an established
marker for wastewater, but it was only detected during spring at very low concentrations that never exceeded the method quantification limit. Applying carbamazepine as indicator for small amounts of wastewater in the investigated highly diluting system is therefore invalid, even if treated wastewater can be excluded as a source. Concentrations of caffeine in untreated domestic wastewater can exceed those of carbamazepine by several orders of magnitude (Miao et al., 2005), emphasizing the superior detectability of caffeine in such systems. Due to its relatively high detection frequency and its indicator properties for untreated wastewater caffeine can be used as wastewater indicator. Its degradability and hence non-refractory behaviour allows even the quantification of untreated domestic wastewater from caffeine concentrations.
3.2.
Estimation of the amount of domestic wastewater
Since caffeine is wastewater-specific, the measured concentrations of caffeine in spring water allow the estimation of wastewater effectively and directly reaching the karst spring. The specificity of caffeine for domestic wastewater in the
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 9 5 e4 0 2
spring is verified by the positive correlation with its predominantly human metabolite paraxanthine (Fig. 3; Schmidt and Schoyerer, 1966), while caffeine concentrations do not correlate with those of theobromine (data not shown). The exact origin of theobromine in spring water cannot be deduced and a contribution of each possible source (beverages, human metabolism, bacterial degradation) is probable. Groundwater contaminations from wastewater with long residence times do not affect the results because of the restricted life time of caffeine and paraxanthine in the groundwater system related to their high degradability. Apart from four farms and two inns all households in the catchment of the Gallusquelle spring are connected to the regional sewer system (personal communication, Mr Leins, municipality Bitz). Those, not connected to the sewer system dispose their wastewater to septic tanks, which are regularly emptied. Treated wastewater as a source of caffeine can be neglected as the wastewater treatment plants of the region are located outside of the catchment of the spring. Since karst springs are often the point outlet of the catchment, the calculated amounts of wastewater can be interpreted as an integrated signal. The results therefore refer to the whole catchment. The calculation for the amount of untreated wastewater appearing in spring water is as follows: WW ¼
c P WC cmax
(1)
PI Q
(2)
with cmax ¼
where WW is the amount of undiluted, rapidly transported and untreated wastewater per day discharging at the spring; P the total population of the catchment of the Gallusquelle is 4000 persons; WC is the daily water consumption per person. It is calculated to be 134 L d1 person1 (yearly data from the sanitary district ‘Scher-Lauchert Abwasserverband’). I is the load of caffeine in untreated wastewater. It is stated to be relatively stable at 15.8 3.8 mg d1 person1 (Buerge et al., 2003). This mean value and the standard deviation are used for further calculations as reasonable approximation. c and cmax are the actually observed concentrations of caffeine in the spring water and the maximum concentration to be expected for the given population and discharge conditions, if 100% of the wastewater was released and rapidly transported to the spring, respectively. Q is the spring discharge. Only concentrations exceeding the method quantification limit
399
(3.4 ng L1) are used for the calculations of wastewater volumes. Samples with lower concentrations are set to 0 m3 d1 of wastewater. During days with high sampling resolution daily mean values of the recharging wastewater were calculated. This leads to a slight bias of the calculation towards an underestimation of the true amounts of untreated wastewater at the spring. Note that the resulting equation (Eq. (3)) for the amount of untreated wastewater at the spring becomes independent from the number of inhabitants within the catchment. WW ¼
c WC Q I
(3)
For the calculation, the relative standard deviation of 6.1% (No¨dler et al., 2010) of the analytical method and the relative uncertainty of the caffeine load (24.1%) are taken into account, considering error propagation. The mean amount of domestic wastewater discharging at the spring is calculated to be 2.2 0.5 m3 d1 (mean standard deviation). Note that this volume refers to normalized wastewater, which is calculated as the product of P and WC. The volume is unaffected by dilution from e.g. precipitation events. The calculated volume corresponds to approximately 0.4% of the total amount of normalized wastewater in the groundwater catchment. The temporal distribution of wastewater load on a daily basis is shown in Fig. 4. It is evident that the influx of wastewater to the aquifer is highly variable with a maximum estimated volume of untreated wastewater of 14.0 3.5 m3 d1, corresponding to 2.6% of the total wastewater. Rutsch et al. (2006) stated the leakages in urban environments to be in the order of 1e10%. The calculated volumes in this study therefore represent a relatively low fraction of the total wastewater. Heinz et al. (2009) stated, that during overflow events on average 23,000 m3 of combined wastewater are released. These values cannot be compared to the calculated volumes from this study, as the effective dilution of the wastewater is not known. Taking into account the method quantification limit of caffeine (cmql ¼ 3.4 ng L1) and a spring discharge of 0.5 m3 s1, the minimum daily amount of untreated wastewater (WWmql) which can be quantified is 1.3 0.3 m3 d1 (mean standard deviation). The relative amount of wastewater to spring discharge, a, at which caffeine is still quantifiable is approximately 2.9 105. a¼
WWmql cmql WC ¼ I Q
(4)
As can be seen from Eq. (4), the detection limit to quantify wastewater in spring water is inversely proportional to the
Fig. 3 e Correlation of caffeine with paraxanthine. The correlation demonstrates that caffeine originates from domestic wastewater.
Fig. 4 e Volumes of leaked wastewater. Calculated volumes of wastewater at the spring Gallusquelle with the respective standard deviation.
400
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 9 5 e4 0 2
spring discharge. The highest accuracy for the calculation is given during low flow periods. However, as concluded by Musolff et al. (2010) from their findings in an urban area, sewer leakage does not occur evenly distributed over time, but simultaneously to natural recharge, i.e. precipitation events, implying an increase of spring discharge. This means, that the detection of wastewater is, in all cases, impeded by dilution. The above stated result reflects the lower boundary estimation for the amount of untreated domestic wastewater at the spring. The degradation of caffeine in the aquifer during transport is neglected here. This simplification is introduced since only limited information on the effect of the significant dilution and fast transport (i.e. low residence times) on biodegradation exists. However, a more realistic picture may be derived when degradation is considered. For this purpose residence times of wastewater in the aquifer resulting from concentrated recharge events and in-situ degradation rates are required.
3.3.
Mean residence time of caffeine in the aquifer
In a single event of point source recharge of untreated wastewater at a known location and time, the resulting breakthrough curve of caffeine at the spring can be used to identify the residence time of wastewater and thus of caffeine from that recharge event in karst systems. The residence time of recharging water directly flowing to the spring can be estimated from the time interval between the precipitation event and the occurrence of chemical signals transported by the percolating rainwater at the spring (Hunkeler and Mudry, 2007). The residence time of water in groundwater systems depends on the point of recharge and groundwater flow velocity. In karst systems groundwater flow velocity is related to the spring discharge (Heinz et al., 2009). Since all possible wastewater sources in the investigated area, both small towns and the retention basin (see Fig. 1) are located in a similar region, estimated residence times of wastewater can be considered to be representative for infiltrating wastewater. Two recharge events from point sources were selected for estimating the mean residence time (Fig. 2; t4a, t4b). Each of the two rainfall events were of sufficient intensity to result in caffeine peaks at the spring. The mean spring discharge during the two investigated recharge events is similar to the mean spring discharge during the whole sampling campaign. The time of the precipitation events is known at a resolution of 1 h (German Meteorological Service, DWD). Since the two rainfall events occurred over periods of 10 and 4 h, respectively, the effective resolution is adjusted appropriately. The respective beginning and the end of the rain events are used as earliest and latest possible points in time inducing the recharge event. The peak concentrations of caffeine in the spring water after rain events are used as the mean point in time for the arrival of caffeine at the spring. The sampling resolution is 3 and 6 h for the first and second precipitation event, respectively. The estimated residence time after the first rainfall event is 99 8 h. The second rainfall event yields 130 8 h. The mean spring discharge was 0.64 m3 s1 and 0.58 m3 s1, respectively. The mean spring discharge during the whole period of investigation is 0.6 m3 s1 and therefore similar to the mean
spring discharge of the two analyzed precipitation events. As representative residence time for wastewater during the period of investigation the mean of both estimations 115 20 h (mean standard deviation) can be assumed. Previous investigations of the residence time, using turbidity, bacteria and uranine as tracer of recharge occurring close to the settlements and from the wastewater retention basin, yielded residence times of approximately 55e238 h depending on the spring discharge (Heinz et al., 2009).
3.4.
Degradation
Buerge et al. (2003) indicated that biodegradation and photodegradation may be relevant degradation processes in surface water bodies. In aquifers the latter can be neglected. However, to improve the estimation on the relative contribution of untreated domestic wastewater in spring discharge over concentrated flow using caffeine as indicator requires a consideration of the in-situ biodegradation of caffeine during transport. Estimations on caffeine degradation in the environment are scarce. Swartz et al. (2006) estimated the in-situ degradation rate in a porous aquifer from a mass balance to be in the order of 0.07e0.014 d1. This results, for the above stated residence time, in an estimation errors between 6 and 48% for the volume of untreated wastewater at the spring. Since the dilution in karst aquifers is larger than in porous aquifers, a hindered degradation is expected. In contrast to the relatively high degradation rates stated above, Buerge et al. (2003) performed incubation experiments with lake water (20 C, dark), that yielded degradation rates of 0.004e0.006 d1, while they proposed, that the actual degradation at temperatures below 10 C is expected to be 0.003 d1. The in-situ degradation rate of caffeine in a lake was estimated to be in the range of 0.003e0.007 d1 for different temperatures and depths (Buerge et al., 2006). The temperature of water in the karst aquifer Gallusquelle is almost constant at 8.2 C. The corresponding half-lives are therefore in the range of 100e240 d. These values refer to a strongly diluted system and are believed to be a better approximation for the in-situ degradation rate of caffeine in karst aquifers, than the above mentioned rates of Swartz et al. (2006). The estimated residence time of signal-producing wastewater in the conduit system is lower than the half-life of caffeine in the aquifer by more than one order of magnitude. The effect of biodegradation on the estimations of wastewater can therefore be neglected. Note that the estimated amounts of wastewater should be understood as a lower boundary estimation, since the dilution of the wastewater occurs along the whole flow path and a gradually decreasing degradation of caffeine might occur with reasonable degradation close to the input. The mean residence time of water in the fissured matrix of karst aquifers (e.g. Maloszewski et al., 2002) is in the order of years to decades and hence high compared to the above stated half-lives in the karst conduit system. Caffeine degradation rates in the environment are therefore high enough not to show refractory behaviour and low enough to reach the spring after infiltration. Due to the good degradability in WWTPs and rather stable behaviour in the environment, caffeine is an excellent indicator for untreated wastewater in rapid systems,
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 9 5 e4 0 2
such as karst. As the caffeine concentrations show high variations after recharge events a high sampling rate is required to assess the amount of recharged domestic wastewater. In combined sewer systems during rain events mixing of wastewater with rainwater occurs, which dilutes the input signal. This does not affect the calculated volumes, but it may happen that caffeine concentrations in spring water do not exceed the method quantification limit. Further, constant caffeine loads in untreated wastewater are assumed for the calculations, which may not necessarily be applicable because of seasonal and temporal variations. However, as the exact point in time of the wastewater release is unknown a consideration of short-term variations (e.g. diurnal) in the calculation is not reasonable. However, long-term-trends in caffeine loads may be considered. Deviations of the assumed caffeine load or water consumption result in an inversely proportional and proportional deviation of the calculated wastewater volume respectively. For example, a caffeine load of 20 mg d1 p1 (instead of 15.8 mg d1 p1, i.e. 21% higher caffeine load) would result in a calculated wastewater volume lower by 21%. Assumptions, deviating from the real situation result in deviations of the wastewater volume appropriately. Literature values can be used in order to approximate the magnitude of contamination. For the estimation of in-situ degradation during transport, a limited source area is required to reliably assess the actual residence time of injected untreated domestic wastewater.
4.
Conclusions
1) Caffeine was detected in a karst spring with a higher frequency than the established wastewater marker carbamazepine, which fails as indicator in this karst system. 2) The degradation rates of caffeine and the residence time distribution of water in karst systems allows to infer to untreated and rapidly transported wastewater from caffeine concentrations in spring water. 3) A quantification of untreated wastewater at a karst spring is possible, from measured caffeine concentrations, caffeine load in untreated wastewater, water consumption rates and spring discharge. Even literature values can be used to estimate volumes of untreated wastewater, yielding reasonable approximations. 4) The degradation of caffeine, which may affect the calculation of wastewater volumes, in karst aquifers is uninvestigated. A tracer test could enhance the estimation, using caffeine and a reference tracer.
Acknowledgements We thank the staff of the water supply company Hermentingen/Swabian Alb, especially Mr. Peter Knaus for advice, providing data and field equipment. The German Meteorological Service (DWD, Gewa¨sserdirektion Riedlingen) supplied data on weather parameters, the sanitary-district ‘ScherLauchert Abwasserverband’ supplied data on wastewater
401
volumes. The presented study was funded by the German Federal Ministry of Education and Research (promotional reference No. 02WM1081, SMART, “Sustainable and Integral Management of Available Water Resources Using Innovative Technologies” and promotional reference No. 02WRS1277A, AGRO, “Risikomanagement von Spurenstoffen und Krankheitserregern in la¨ndlichen Karsteinzugsgebieten”).
references
Bahlmann, A., Weller, M.G., Panne, U., Schneider, R.J., 2009. Monitoring carbamazepine in surface and wastewater by an immunoassay based on a monoclonal antibody. Analytical and Bioanalytical Chemistry 395 (6), 1809e1820. Boehm, A.A., Grant, S.B., Kim, J.H., Mowbray, S.L., McGee, C.D., Clark, C.D., Foley, D.M., Wellman, D.E., 2002. Decadal and shorter period variability of surf zone water quality at Huntington Beach, California. Environmental Science and Technology 36 (18), 3885e3892. Buerge, I.J., Poigner, T., Mu¨ller, M.D., Buser, H.-R., 2003. Caffeine, an anthropogenic marker for wastewater contamination of surface waters. Environmental Science and Technology 37 (4), 691e700. Buerge, I.J., Poigner, T., Mu¨ller, M.D., Buser, H.-R., 2006. Combined sewer overflow to surface waters detected by the anthropogenic marker caffeine. Environmental Science and Technology 40 (13), 4096e4102. Buerge, I.J., Buser, H.-R., Kahle, M., Mu¨ller, M.D., Poigner, T., 2009. Ubiquitous occurence of the artificial sweetener acesulfame in the aquatic environment: an ideal chemical marker of domestic wastewater in groundwater. Environmental Science and Technology 43 (12), 4381e4385. Eiswirth, M., Ho¨tzl, H., 1997. The impact of leaking sewers on urban groundwater. In: Chilton, J., et al. (Eds.), Groundwater in the Urban Environment. Problems, Processes and Management, vol. 1. A.A. Balkema, Rotterdam, pp. 399e404. Ford, D., Williams, P., 2007. Karst Hydrogeology and Geomorphology. Wiley, Chichester. Gardinali, P.R., Zhao, X., 2002. Trace determination of caffeine in surface water samples by liquid chromatographyeatmospheric pressure chemical ionizationemass spectrometry (LC-APCI-MS). Environment International 28 (6), 521e528. Gasser, G., Rona, M., Voloshenko, A., Shelkov, R., Tal, N., Pankratov, I., Elhanany, S., Lev, O., 2010. Quantitative evaluation of tracers for quantification of wastewater contamination of potable water sources. Environmental Science and Technology 44 (10), 3919e3925. Geyer, T., Birk, S., Licha, T., Liedl, R., Sauter, M., 2007. Multitracer test approach to characterize reactive transport in karst aquifers. Ground Water 45 (1), 36e45. Glassmeyer, S.T., Furlong, E.T., Kolpin, D.W., Cahill, J.D., Zaugg, S.D., Werner, S.L., Meyer, M.T., Kryak, D.D., 2005. Transport of chemical and microbial compounds from known wastewater discharges: potential for use as indicators of human fecal contamination. Environmental Science and Technology 39 (14), 5157e5169. Hakil, M., Denis, S., Viniegra-Gonza´lez, G., Augur, C., 1998. Degradation and product analysis of caffeine and related dimethylxanthines by filamentous fungi. Enzyme and Microbial Technology 22 (5), 355e359. Heinz, B., Birk, S., Liedl, R., Geyer, T., Straub, K.L., Bester, K., Kappler, A., 2006. Vulnerability of a karst spring to wastewater infiltration (Gallusquelle, southwest Germany). Austrian Journal of Earth Sciences 99, 11e17.
402
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 3 9 5 e4 0 2
Heinz, B., Birk, S., Liedl, R., Geyer, T., Straub, K.L., Andresen, J., Bester, K., Kappler, A., 2009. Water quality deterioration at a karst spring (Gallusquelle, Germany) due to combined sewer overflow: evidence of bacterial and micro-pollutant contamination. Environmental Geology 57 (4), 797e808. Hrudey, S.E., Payment, P., Huck, P.M., Gillham, R.W., Hrudey, E.J., 2003. A fatal waterborne disease epidemic in Walkerton, Ontario: comparison with other waterborne outbreaks in the developed world. Water Science and Technology 47 (3), 7e14. Hunkeler, D., Mudry, J., 2007. In: Goldscheider, N., Drew, D. (Eds.), Methods of Karsthydrogeology. International Contributions to Hydrogeology, vol. 26. Taylor and Francis group, London, pp. 93e121. Lelo, A., Miners, J.O., Robson, R.A., Birkett, D.J., 1986. Quantitative assessment of caffeine partial clearances in man. British Journal of Clinical Pharmacology 22, 183e186. Maloszewski, P., Stichler, W., Zuber, A., Rank, D., 2002. Identifying the flow systems in a karstic-fissured-porous aquifer, the Schneealpe, Austria, by modelling of environmental 18O and 3 H isotopes. Journal of Hydrology 256 (1e2), 48e59. McFeters, G.A., Stuart, D.G., 1972. Survival of coliform bacteria in natural waters: field and laboratory studies with membranefilter chambers. Applied Microbiology 24 (5), 805e811. Memon, B.A., Azmeh, M.M., 2001. Failure of an industrial wastewater lagoon in a karst terrain and remedial action. Environmental Geology 40 (11e12), 1424e1432. Miao, X.-S., Yang, J.-J., Metcalfe, C.D., 2005. Carbamazepine and its metbaolites in wastewater and in biosolids in a municipal wastewater treatment plant. Environmental Science and Technology 39 (19), 7469e7475. Musolff, A., Leschik, S., Reinstorf, F., Strauch, G., Schirmer, M., 2010. Micropollutant loads in the urban water cycle. Environmental Science and Technology 44 (13), 4877e4883. No¨dler, K., Licha, T., Bester, K., Sauter, M., 2010. Development of a multi-residue analytical method, based on liquid chromatographyetandem mass spectrometry, for the simultaneous determination of 46 micro-contaminants in
aqueous samples. Journal of Chromatography A 1217 (42), 6511e6521. Ogunseitan, O.A., 1996. Removal of caffeine in sewage by pseudomonas putida: implications for water pollution index. World Journal of Microbiology and Biotechnology 12 (3), 251e256. Rutsch, M., Rieckermann, J., Krebs, P., 2006. Quantification of sewer leakage: a review. Water Science and Technology 54 (6e7), 135e144. Schmidt, G., Schoyerer, R., 1966. On the detection of caffeine and its metabolites in urine (German: Zum Nachweis von Coffein und seinen Metaboliten im Harn). Deutsche Zeitschrift fu¨r gerichtliche Medizin 57 (3), 402e409. Seiler, R.L., Zaugg, S.D., Thomas, J.M., Howcroft, D.L., 1999. Caffeine and pharmaceuticals as indicators of wastewater contamination in wells. Ground Water 37 (3), 405e410. Siegener, R., Chen, R.F., 2002. Caffeine in Boston harbor seawater. Marine Pollution Bulletin 44 (5), 383e387. Srdjenovic, B., Djordjevic-Milic, V., Grujic, N., Injac, R., Lepojevic, Z., 2008. Simultaneous HPLC determination of caffeine, theobromine, and theophylline in food, drinks, and herbal products. Journal of Chromatographic Science 46 (2), 144e149. Stavric, B., 1988. Methylxanthines: toxicity to humans. 3. theobromine, paraxanthine and the combined effect of methylxanthines. Food and Chemical Toxicology 26 (8), 725e733. Swartz, C.H., Reddy, S., Benotti, M.J., Yin, H., Barber, L.B., Brownawell, B.J., Rudel, R.A., 2006. Steroid estrogens, nonylphenol ethoxylate metabolites, and other wastewater contaminants in groundwater affected by a residential septic system on Cape Cod. MA. Environmental Science and Technology 40 (16), 4894e4902. Takada, H., Satoh, F., Bothner, M.H., Tripp, B.W., Johnson, C.G., Farrington, J.W., 1997. Anthropogenic molecular markers: tools to identify the sources and transport pathways of pollutants. In: Eganhouse, R.P. (Ed.), Molecular Markers in Environmental Geochemistry. American Chemical Society, Washington, DC, pp. 178e195. ACS Symposium Series 671.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 0 3 e4 1 2
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effects of Fenton treatment on the properties of effluent organic matter and their relationships with the degradation of pharmaceuticals and personal care products Wei Li a, Venkateswarlu Nanaboina b, Qixing Zhou a,*, Gregory V. Korshin b a
Key Laboratory of Pollution Processes and Environmental Criteria at Ministry of Education, College of Environmental Science and Engineering, Nankai University, Weijin Road 94, Tianjin 300071, China b Department of Civil and Environmental Engineering, Box 352700, University of Washington, Seattle, Washington 98195-2700, USA
article info
abstract
Article history:
This study examined effects of Fenton oxidation on trace level pharmaceuticals and
Received 13 April 2011
personal care products (PPCPs) commonly occurring in wastewater. The tested PPCPs
Received in revised form
included acetaminophen, atenolol, atrazine, carbamazepine, metoprolol, dilantin, DEET,
25 October 2011
diclofenac, pentoxifylline, oxybenzone, caffeine, fluoxetine, gemfibrozil, ibuprofen, iopro-
Accepted 1 November 2011
mide, naproxen, propranolol, sulfamethoxazole, bisphenol-A and trimethoprim. Trans-
Available online 15 November 2011
formations of effluent organic matter (EfOM) caused by Fenton oxidation were also quantified. All tested PPCPs, except atrazine and iopromide, were completely removed by
Keywords:
Fenton treatment carried out using a 20 mg/L Fe (II) concentration and a 2.5 H2O2/Fe (II) molar
Fenton oxidation
ratio. Up to 30% on the total carbon concentration was removed during Fenton treatment
Pharmaceutical and personal care
which was accompanied by the oxidation of EfOM molecules and formation of oxidation
products
products such as oxalic, formic and acetic acids and, less prominently, formaldehyde,
Effluent organic matter
acetaldehyde, propionaldehyde and glycolaldehyde. The absorbance of EfOM treated with
Absorbance spectroscopy
Fenton reagent at varying Fe (II) concentration and contact time underwent a consistent
Size exclusion chromatography
decrease. The relative decrease of EfOM absorbance was strongly and unambiguously correlated with the removal of all tested PPCPs. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The occurrence of pharmaceuticals and personal care products (PPCPs) in aquatic systems and their complex, yet to be ascertained environmental effects are a matter of increasing concern. Several comprehensive studies have determined that a great variety of PPCPs are present in surface water and wastewater (Kasprzyk-Hordern et al., 2008; Yoon et al., 2010; Kosma et al., 2010). Although environmental concentrations of PPCPs are typically in the nanogram to microgram per liter
range, its continuous input and persistence may constitute potential long-term risks for aquatic organisms (Klavarioti et al., 2009). Previous studies have shown that PPCPs cannot be removed completely by conventional wastewater treatment plants (Lishman et al., 2006; Kasprzyk-Hordern et al., 2009). In contrast, advanced oxidation processes (AOP) effectively degrade most PPCPs occurring in wastewater (Ikehata et al., 2006; Klavarioti et al., 2009; Rosario-Ortiz et al., 2010). Among various AOP variants, Fenton oxidation is an attractive option
* Corresponding author. Tel.: þ86 22 23507800. E-mail address:
[email protected] (Q. Zhou). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.002
404
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 0 3 e4 1 2
because of its low cost, lack of toxicity of its reagents (Fe (II) and H2O2) and relative, simplicity of this technology (Wang, 2008). In Fenton processes, hydroxyl radicals ( OH) are rapidly generated through the decomposition of H2O2 catalyzed by Fe (II) (Eq. (1)). The Fe (II) catalyst can be regenerated at a slow rate (K ¼ 0.01e0.02 M1 s1) through reaction (2) (Burbano, et al., 2008; Bautista et al., 2008). The oxidation of organic pollutants (denoted as RH) proceeds via reaction between OH radicals and RH (Eq. (3)). Generally, Fenton treatment of wastewater proceeds via four stages: oxidation, neutralization, coagulation/flocculation and solid-liquid separation (Deng, 2007). While both coagulation and oxidation can remove target contaminants, the role of the latter process in Fenton treatment is deemed to be by far prevalent.
FeðIIÞ þ H2 O2 /FeðIIIÞ þ$ OH þ OH ;
K ¼ 76 M1 s1
(1)
FeðIIIÞ þ H2 O2 /FeðIIÞ þ HO2 $ þ Hþ ; K ¼ 0:01e0:02 M1 s1
(2)
RH þ$ OH /$ R þ H2 O
(3)
The efficiency of Fenton process is affected by H2O2 and Fe (II) dosage, concentration and properties of effluent organic matter (EfOM) and the solution pH, with optimal pH values between 2 and 4 (Trapido et al., 2009). Furthermore, the efficiency of Fenton treatment of wastewater can be enhanced by UV or solar irradiation. This process known as photo-Fenton oxidation is characterized by photo-reduction of ferric to ferrous ions that is promoted concomitantly with the generation of additional OH radicals. The photo-Fenton process can be effective even at neutral pH whereas the Fenton oxidation per se is typically limited only to acidic pH (Bandala et al., 2004; Rozas et al., 2010). Considerable efforts have been made to examine the degradation of representative PPCPs by Fenton processes (Ay and Kargi, 2010; Elmolla and Chaudhuri, 2009; Pe´rez-Moya et al., 2010; Me´ndez-Arriaga et al., 2010; Trovo´ et al., 2009). For instance, Ay and Kargi (2010) investigated effect of Fe (II) and H2O2 concentrations on the degradation of amoxicillin at pH 3 and determined that amoxicillin could be completely degraded, with 37% of its initial amount mineralized at the optimum H2O2/Fe/amoxicillin weight concentrations of 255/ 25/105 mg/L. A complete degradation of another antibiotic, sulfamethazine at its initial concentration of 50 mg/L was observed in Fenton oxidation with 600 mg/L H2O2 and 50 mg/L Fe (II) concentrations and a 2 min exposure time and pH 3 (Pe´rez-Moya et al., 2010). The outcome of Fenton treatment of wastewater is influenced by EfOM that interacts readily with hydroxyl radicals, with apparent EfOM/ OH reaction rates exceeding 109 M1 s1 (Rosario-Ortiz et al., 2008; Wert et al., 2009a, Dong et al., 2010). As a result, EfOM can be expected to be the primary reaction substrate in all practically relevant conditions of Fenton treatment. Still, most of the experiments reported in prior literature were carried out using PPCPs dissolved in deionized water, with only a few studies reporting the degradation of PPCPs in real wastewater (Klamerth et al., 2010a, b). While the involvement of EfOM in these processes has not been examined in adequate detail, relevant changes of EfOM properties can be ascertained using a number of methods. For instance, high-performance size exclusion chromatography
(HPSEC) has been used to quantify contributions of different apparent molecular weight (AMW) EfOM fractions (Her et al., 2003; Nam and Amy, 2008; Korshin et al., 2009). Fluorescence spectroscopy has also been used to determine contributions of dissimilar class of compounds constituting EfOM (Her et al., 2003; Nam and Amy, 2008). Differential absorbance spectroscopy (DAS) has been shown to be capable of detecting changes of EfOM properties caused by water treatment processes and correlating them with the degradation of PPCPs caused by ozonation (Nanaboina and Korshin, 2010). In this study, we pursued a hypothesis that the oxidation of EfOM during Fenton treatment (quantified, for instance, by HPSEC and/or absorbance spectroscopy) proceeds concurrently with and is indicative of the degradation of multiple PPCPs present in the wastewater being treated. Conceptually, this hypothesis is similar to that successfully employed by Wert et al., (2009b), Nanaboina and Korshin (2010) in the studies of the removal of PPCPs in ozonated wastewater. This approach has not been applied to Fenton oxidations, nor have the relevant changes of EfOM properties and attendant generation of characteristic EfOM oxidation products such as carboxylic acids and aldehydes been ascertained. These issues are addressed in detail in this paper.
2.
Methods and materials
2.1.
Reagents
Acetaminophen, atenolol, atrazine, carbamazepine, metoprolol, dilantin, DEET, diclofenac, pentoxifylline, oxybenzone, caffeine, fluoxetine, gemfibrozil, ibuprofen, iopromide, naproxen, propranolol, sulfamethoxazole, bisphenol-A and trimethoprim were acquired from SigmaeAldrich. These compounds were selected based on their frequent occurrence in surface water and wastewater (Ellis, 2006; KasprzykHordern et al., 2008; Yoon et al., 2010; Kosma et al., 2010). Stock solutions were prepared by weighing and dissolving requisite amounts of each compound in Milli Q water to yield a 2 mg/L concentration. Other chemicals and solvents were purchased from various commercial suppliers and detailed information is provided in Text S1 of Supplementary Material (SM section).
2.2.
Wastewater characterization
A sample of secondary effluent (before disinfection) from West Point Wastewater Treatment plant in Seattle was collected in a pre-cleaned 10-L polypropylene container. Upon arrival at the laboratory, the wastewater was filtered through a 0.45 mm filter and stored at 4 C for no more than a week until used in experiments. The wastewater had the following chemical parameters: DOC 9.6 mg/L, pH 7.2, ammonia 13.6 mg/L NH3eN, nitrate 13.7 mg/L.
2.3.
Fenton oxidation experiments
The wastewater was spiked using the stock solution to reach a 1 mg/L concentration of each individual PPCP compound. Fenton treatments were conducted at room temperature, in
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 0 3 e4 1 2
the presence of low-intensity interior lighting and utilized varying Fe (II) and H2O2 concentration and reaction times. The pH of the wastewater was first adjusted to 3 with H2SO4 to prevent the formation of iron oxyhydroxides. Then, requisite amounts of Fe (II) and H2O2 were added to the solution to initiate the oxidation. The reagents were mixed by a magnetic stirrer to ensure complete homogeneity. After certain contact time, Fenton oxidation was stopped by increasing the pH to ca. 8. At this pH, the generation of hydroxyl radical is effectively prevented (Muruganandham and Swaminathan, 2004; Zhang et al., 2006). The treated wastewater was filtered through a 0.45 mm filter for analyses that followed. The detailed experiment methods are provided in Text S2 (SM section).
2.4.
Analytical methods
Dissolved organic carbon (DOC) concentrations in the samples were determined with a Shimadzu TOC-Vcsh analyzer. Absorbance spectra were recorded using a Perkin Elmer Lambda 18 spectrophotometer using 1 cm quartz cells. Nitrate concentrations were determined with a Dionex ICS-3000 ion chromatograph. PPCPs analyses were performed according to the method established by Nanaboina and Korshin (2010) using a Prominence Shimadzu HPLC coupled with an Applied Biosystems 4000 Q Trap tandem mass spectrometer (LC-MS/ MS) equipped with electrospray ionization (ESI) operated in both positive and negative ion modes. All relevant details of chromatographic separations and MS/MS quantitation of the analytes are reported in Nanaboina and Korshin (2010). Aldehydes concentrations were determined by a GC-2010 Shimadzu gas chromatograph (GC) equipped with an electron capture detector (ECD) according to Method 6252 (Disinfection by-products: aldehydes) (APHA, 1998). Concentrations of carboxylic acids were determined using an ICS-3000 ion chromatograph. High performance size exclusion chromatography (HPSEC) was carried out using a Dionex HPLC system with Ultimate 3000 diode array detector. Details of the analysis methods are provided in the Text S3 (SM section).
3.
Results and discussion
3.1.
Effect of Fenton treatment on PPCPs concentrations
Initial experiments were performed with varying H2O2/Fe (II) molar ratios at a constant Fe (II) concentration of 10 mg/L, a 30 min fixed treatment time and pH 3. The H2O2/Fe (II) molar ratios used in these experiments were 0.5, 1, 1.5, 2, 2.5 and 3. Their effects on the degradation of selected PPCPs and the radical probe pCBA (Yao and Haag, 1991) is shown in Fig. 1(a). In all cases, the removal of pCBA and PPCPs increased with the increase of H2O2/Fe (II) molar ratio but little change was observed when this ratio was above 1.0. For instance, the degradation of acetaminophen and diclofenac was 90.9% and 91.1%, respectively, at a 1.0 H2O2/Fe (II) molar ratio, while oxybenzone and fluoxetine were completely degraded at a 0.5 H2O2/Fe (II) molar ratio. When H2O2/Fe (II) molar ratio were increased from 1 to 3, degradation levels of almost all compounds were practically stable, except those for atrazine and iopromide for which the removal increased with the
405
a
b
Fig. 1 e (a) Degradation of pCBA and PPCPs by Fenton oxidation carried out using varying H2O2/Fe (II) molar ratios. Fe (II) concentration 10 mg/L, pH 3, reaction time 30 min. (b) Degradation of PPCPs by Fenton oxidation at varying Fe (II) concentrations. H2O2/Fe (II) molar ratio 2.5, pH 3, reaction time 30 min.
H2O2/Fe (II) molar ratio but did not exceed 56.1% and 61.9%, respectively. Although not much difference existed among the degradation of PPCPs when H2O2/Fe (II) molar ratios increased from 1 to 3, a 2.5 H2O2/Fe (II) molar ratio was employed in the experiments that followed to ensure that the removal of most PPCPs be at least 80%. In the experiments that employed a constant H2O2/Fe (II) molar ratio of 2.5, Fe (II) concentrations were set at 0.625, 1.25, 2.5, 3.75, 5, 7.5, 10, 15 and 20 mg/L. The pH was 3 and the treatment time was 30 min. Fig. 1(b) and Fig. S1 (SM section) show results of these experiments. The degradation of pCBA gradually increased with the Fe (II) concentration to reach a 97.3% level at a 20 mg/L Fe (II) dose. The removal of organic species other than pCBA also increased with the Fe (II) concentration but the patterns were somewhat different for different groups of compounds. For instance, acetaminophen, diclofenac, oxybenzone, sulfamethoxazole, naproxen, bisphenol-A, fluoxetine were largely removed at Fe concentrations <7.5 mg/L as shown in Fig. 1(b). Another group of species that included dilantin, DEET, trimethoprim, propranolol, ibuprofen, carbamazepine, caffeine, gemfibrozil, metoprolol, pentoxifylline and atenolol was removed less readily but their oxidation was nearly complete at Fe concentrations <15 mg/L (Fig. S1(a), SM section). Finally, atrazine, iopromide and pCBA were most resistant toward Fenton oxidation, especially iopromide and atrazine, whose degradation was
406
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 0 3 e4 1 2
86.6% and 88.9%, respectively, at a 20 mg/L Fe (II) concentration (Fig. S1(b), SM section). Possible effects of coagulation on PPCPs removal by Fenton process were explored using different concentration of ferric chloride (0.625, 1.25, 2.5, 3.75, 5, 7.5, 10, 15 and 20 mg Fe3þ/L) in the absence of H2O2; all the other experiment conditions were the same as those used in the Fenton oxidations. Results of these experiments demonstrated that coagulation did not cause any significant removal (<5%) of nearly all PPCPs, except propranolol (40%) and metoprolol (20%). This result is in accordance with the data of prior research (Ternes et al., 2002; Westerhoff et al., 2005) demonstrating that coagulation did not significantly affect the levels of diclofenac, carbamazepine, atrazine, DEET and several other PPCPs. The low removal efficiency can be explained by their hydrophilicity, although propranolol and metoprolol are more hydrophobic (Carballa et al., 2005). Additional measurements examined effects of treatment time variations at a constant Fe (II) concentration of 10 mg/L and H2O2/Fe (II) molar ratio of 2.5. In all cases, PPCPs oxidation was very rapid during the initial 2 min reaction. Most PPCPs can be degraded up to 80% of the initial concentration (Fig. S2, SM section). Only atrazine and iopromide were degraded slowly (36% and 35%, respectively after 2 min of reaction) and their oxidation continued in some extent beyond the initial rapid phase.
3.2.
Effect of Fenton oxidation on EfOM
3.2.1.
DOC removal during Fenton process
DOC concentrations in the Fenton-treated wastewater were measured at varying Fe (II) concentration at a constant H2O2/ Fe (II) molar ratio of 2.5. Similar experiments were also performed for Fenton treatment that utilized a constant Fe (II) concentration (10 mg/L) and H2O2/Fe (II) molar ratio (2.5) and varying contact times. In experiments with varying Fe (II) doses, the DOC concentrations decreased gradually from 9.6 mg/L in the untreated wastewater to 6.8 mg/L at a 15 mg/L Fe (II) dose due both to EfOM removal by coagulation and possibly its mineralization to CO2. Little additional DOC removal took place when the Fe (II) concentration increased from 15 to 20 mg/L (Table S3, SM section). Similarly to the observations concerning the degradation of PPCPs, the removal of DOC was very rapid during initial 4 min of Fenton oxidation (Table S3, SM section). Approximately 17% of DOC was removed during the first 4 min of contact time following which the DOC removal slowly increased to ca. 25%. Because little change occurred after reaction time exceeding 30 min that contact time was judged to be optimal for Fenton oxidations.
3.2.2. EfOM oxidation and formation of aldehydes and carboxylic acids The role of oxidative process in EfOM transformations caused by Fenton treatment was explored based on the measurements of the formation of aldehydes and carboxylic acids, typical products of EfOM oxidation by AOP (Swietlik et al., 2004). Four aldehydes (formaldehyde, acetaldehyde, propionaldehyde and glycolaldehyde) were determined to be present
a
b
Fig. 2 e Formation of (a) aldehydes and (b) carboxylic acids in wastewater treated by Fenton process using varying Fe (II) concentrations. H2O2/Fe (II) molar ratio 2.5, pH 3, reaction time 30 min.
in the treated wastewater. Fig. 2(a) demonstrates that their concentrations increased with the Fe (II) concentration and that formaldehyde was the most prevalent among these species. Similarly to the degradation of PPCPs and removal of DOC, the formation of aldehydes was fast, with their concentrations reaching a plateau within 15 min (Fig. S3(a), SM section). The behavior of glycolaldehyde was somewhat different, with its concentration reaching a maximum value (18.0 mg/L) at 8 min and then decreasing. The concentrations of formate, acetate and oxalate found in the Fenton-treated wastewater were much higher than those of aldehydes, reaching as much as 1.7 mg/L in the case of oxalate (Fig. 2(b) and Fig. S3(b) in the SM section). This result is similar to the formation of oxidation byproducts in ozonated wastewater, in which carboxylic acids concentration was ca. 4.3 times higher than aldehydes concentration (Wert et al., 2007). The acetate and oxalate levels increased monotonically with Fe (II) concentrations and/or reaction time, while the behavior of formate was somewhat different, with its concentration reaching a maximum value of 1.2 mg/L at mid-range Fe (II) doses and then decreasing. Examination of contributions of formate, acetate and oxalate in the total of these three carboxylic acids generated at varying Fe (II) concentrations and reaction times showed that at low Fe (II) concentrations and/or short reaction times, formate accounted for up to 61% of the total carboxylic acids (Table S4, SM section). As Fe (II) concentration and/or contact
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 0 3 e4 1 2
time increased, the percentage of formate decreased while that of oxalate increased to reach ca. 41%. The yields of aldehydes and carboxylic acids generated via the oxidation of the EfOM substrate were calculated by Eqs. (4) and (5), respectively; P Yield ¼
Caldehydes 100% DOCinitial
P
(4)
Cacids 100% (5) DOCinitial P where Caldehydes represents the sum of the equivalent concentration of formaldehyde, acetaldehyde, propionaldeP hyde and glycolaldehyde, Cacids represents the sum of the equivalent concentration of formate, acetate and oxalate (expressed in mg/L as carbon) and DOCinitial represents the initial DOC concentration of the wastewater. Yield ¼
407
Fig S4(a) (SM section) demonstrates that the yields of aldehydes and carboxylic acids increased with Fe(II) concentration and achieved 1.01% and 11.62% (compared with the initial concentration of organic carbon in the wastewater), respectively, at Fe(II) concentration of 20 mg/L. Strong linear correlations were found to exist between DOC removal and the yields of the aldehydes (R2 ¼ 0.99) and carboxylic acids (R2 ¼ 0.96) during Fenton treatment (Fig. S4(b), SM section). These correlations demonstrate that for every 1 mg of EfOM removed during the Fenton treatment, on the average 0.03 and 0.3 mg (as dissolved organic carbon) of aldehydes and carboxylic acids, respectively, were formed. While the yields of the aldehydes did not change significantly with variations of Fe concentrations, those of the carboxylic acids were <0.25 mg per mg of EfOM removed for Fe concentrations <5 mg/L while for Fe concentrations >7.5 mg/L, they were as high as 0.35e0.4 mg per mg of removed EfOM. This trend is indicative of an increasingly deep oxidation of EfOM and high
Fig. 3 e SEC elution profiles for the wastewater before and after Fenton oxidation with varying Fe (II) concentration at wavelengths of 254 nm (a) and 220 nm (b).
408
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 0 3 e4 1 2
yields of formation of its identified degradation products. Fig. S5 (SM section) presents an additional summary of the contributions of the removed DOC, identified aldehydes and carboxylic acids and unidentified DOC fractions remaining in wastewater treated with the Fenton reagent at varying Fe (II) concentrations. It demonstrated that the percentage of the sum of aldehydes and carboxylic acids in the remaining organic carbon was about 18% at Fe (II) concentration of 20 mg/L.
3.2.3.
a
Changes of EfOM properties in Fenton process
The prominent formation of carboxylic acids and aldehydes in the Fenton-treated wastewater is indicative of the oxidation of EfOM by the Fenton reagent. HPSEC analysis provided additional data related to this observation. Fig. 3 shows the SEC profiles for the wastewater before and after Fenton oxidation with varying Fe (II) concentration at H2O2/Fe (II) molar ratio of 2.5. The wavelengths of 220 and 254 nm were selected because prior research established that the absorbance of NOM at 220 nm is associated with both carboxylic and aromatic chromophores, while EfOM absorbance at 254 nm is primarily determined by aromatic groups in EfOM (Korshin et al., 2009). The SEC profiles generated using the absorbance data at 254 nm have three major peaks with maxima corresponding to elution times 14.1, 15.1 and 16.0 min (Fig. 3(a)). The feature with a 16.0 min maximum had an additional shoulder located at ca. 16.5 min elution time. As high molecular weight molecules are eluted first in HPSEC separations, we divided the observed HPSEC profiles into four segments corresponding to varying elution times. These HPSEC segments corresponded to EfOM species with operationally defined very high apparent molecular weight, AMW (VHMW, 12.0e14.0 min), high AMW
b
Fig. 4 e (a) Changes of relative contributions of EfOM fraction caused by Fenton treatment with vayring Fe (II) concentrations and (b) correlation between the contributions of EfOM fractions and SUVA254.
Fig. 5 e Correlations between relative changes of absorbance of treated wastewater and degradation of representative PPCPs by Fenton oxidation that utilized varying Fe (II) concentrations (B) and contact times (C).
409
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 0 3 e4 1 2
(HMW, 14.0e14.7 min), intermediate AMW (IMW, 14.7e15.5 min) and low AMW (LMW, 15.5e18.0 min). The contributions of different EfOM fractions were quantified by the integration of absorbances within each a priori defined range of the elution times (Fig. 4(a)). Examination of the HPSEC profiles of Fenton treated EfOM showed that the contribution of the VHMW fraction decreased while that of IMW fraction increased with the increase of Fe (II) concentration. This demonstrated that Fenton oxidation causes VHMW fraction to break down to form small fragments. SUVA254 (Table S3, SM section) was also correlated with the contributions of VHMW and IMW (Fig. 4(b)) demonstrating that the breakdown of the large AWM fractions of EfOM involved the preferential oxidation of its aromatic chromophores. Fig. 3(b) shows the SEC profile for wavelength of 220 nm. Its shape is different from that recorded for the wavelength of 254 nm due to the presence of intense features corresponding to the absorbance by carboxyl- and carbonyl-rich molecules present in the IMW and LMW fractions of EfOM. In these profiles, the intensity of the absorbance associated with the EfOM fraction eluted at 15.5e16.3 min decreased rapidly, while that having lower molecular weights and eluted at 16.1e8.0 min increased.
These observations indicate that the low AMW fraction of EfOM associated with a prominent HPSEC peak at 16.5 min is likely to include both the identified (e.g., aldehydes and carboxylic acids quantified in this study) and unidentified EfOM oxidation products (e.g., EfOM molecules enriched with carbonyl and carboxylic functionalities). This point is confirmed by the existence of strong correlations between the intensity change of SEC feature at 16.5 min (measured at 220 nm) and total concentrations of aldehydes (R2 0.99) and carboxylic acids (R2 0.99) formed upon Fenton oxidation (Fig. S6, SM section).
3.2.4. Changes of wastewater absorbance induced by Fenton treatment and their relationships with PPCPs degradation Examination of the absorbance spectra of EfOM affected by Fenton treatment that utilized varying Fe (II) concentrations and reaction times showed that the absorbance of EfOM decreased at all wavelengths exceeding 240 nm (Fig. S7 and Fig. S8, SM section). Evolution of the absorbance spectra showed patterns similar to those observed in the degradation of PPCPs. For example, the difference between the absorbance of untreated wastewater and that after 1 min of Fenton oxidation was relatively large but its increments at increasing
Table 1 e DA254/A0254 values and Fe(II) doses corresponding to the removal of 50% and 90% of selected PPCPs and relevant degradation rate constants KOH. Compounds Oxybenzone Acetaminophen Fluoxetine Naproxen Diclofenac Propranolol Sulfamethoxazole Bisphenol-A Trimethoprim Dilantin Carbamazepine DEET Caffeine Ibuprofen Atenolol Metoprolol Pentoxifylline Gemfibrozil Atrazine Iopromide pCBA
(DA254/A0254)50%
(DA254/A0254)90%
[Fe(II) dose]50% (mg/L)
[Fe(II) dose ]90% (mg/L)
pKaa
KOH (M1 s1)
0.18 0.23 0.24 0.25 0.26 0.28 0.25 0.26 0.30 0.31 0.31 0.32 0.32 0.33 0.33 0.33 0.35 0.35 0.44 0.45 0.36
0.27 0.32 0.42 0.35 0.42 0.43 0.46 0.46 0.46 0.46 0.46 0.46 0.48 0.49 0.51 0.51 0.51 0.5 0.58 0.58 0.55
0.5 1 1.2 1.5 2.4 2.5 2 2 3.2 3.5 3.7 4 4.1 4 4.4 3.1 5 5.3 8.5 9.2 5.6
2.2 3.8 7.5 5 7.5 8.2 10.7 10.2 10.9 9.8 10 10 11.1 12 12.6 12.4 13 12.4 >20 >20 15.3
NA NA NA 4.2b 4.2c 9.5 5.7c 9.6e10.2 3.2, 7.1g 8.3 <2 <2 6.1 4.9 9.6 NA NA 4.7 <2(1.6) <2 and >13 NA
NA NA NA 9.6 109b 7.5 109c 1.0 1010d 8.5 109e 1.0 1010f 6.9 109g NA 8.8 109c 5.0 109h NA 6.5 109b 7.1 109i NA NA 1.0 1010j 3.0 109k 3.3 109c 5.0 109c
a Yoon et al. (2010). b Packer et al. (2003). c Huber et al. (2003). d Benner et al. (2008). e Mezyk et al. (2007). f Rosenfeldt and Linden (2004). g Dodd et al. (2006). h Song et al. (2009). i Song et al. (2008). j Razavi et al. (2009). k Acero et al. (2000).
410
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 0 3 e4 1 2
1.20E+10
Degradation rate constant (KOH , M-1 S-1 )
Bisphenol-A Propranolol
1.00E+10
Gemfibrozil
Naproxen Carbamazepine Sulfamethoxazole
8.00E+09
Diclofenac Trimethoprim Caffeine
6.00E+09
Metoprolol Atenolol Ibuprofen DEET
pCBA
4.00E+09
R² = 0.61
Atrazine Iopromide
2.00E+09
0.00E+00 0.2
0.25
0.3
0.35
0.4
0.45
0.5
( A 254 /A254 0 )50% Fig. 6 e Correlation between (DA254/A0254)50% values and intrinsic rates of oxidation of the tested PPCPs by hydroxyl radicals.
contact times were much less prominent (Fig. S8, SM section). This was consistent with the fast degradation of PPCPs during the initial fast phase of Fenton oxidation (Fig. S2, SM section). The differential spectra generated using the data shown in Fig. S7(a) and Fig. S8(a) are presented in Fig. S7(b) and Fig. S8(b) (SM section), respectively. They demonstrate that the intensity of differential absorbance increased monotonically as a function of Fe (II) concentration and reaction time. The differential absorbance spectra of EfOM oxidized by Fenton’s reagent had a relatively prominent maximum in the wavelength range of 265e275 nm. The broad absorbance band observed in Fenton oxidation of EfOM was similar to that observed for EfOM affected by ozone (Nanaboina and Korshin, 2010). Correlations between the removal of PPCPs and changes of EfOM absorbance were examined based on absorbance measurements at the wavelength of 254 nm that is frequently used in practical applications (Abella´n et al., 2007; Korshin et al., 1997). Typical features of relationships between relative changes of EfOM absorbance (denoted as DA254/A0254) of the wastewater and the degradation of several important PPCPs such as sulfamethoxazole, naproxen, atrazine and iopromide are shown in Fig. 5. Similar patterns were observed for all the other PPCPs (Figs. S9 and S10, SM section). The relationships between the degradation of individual PPCPs and relative changes of absorbance at 254 nm were essentially the same for Fenton oxidations carried out at varying reaction times and Fe (II) concentrations. In all cases, little or no oxidation of the PPCPs took place at DA254/A0254 values <0.1. This indicates that the initial stage of Fenton oxidation of the wastewater corresponds to the engagement of highly reactive EfOM chromophores that consume all oxidants released in reactions (1)e(3) while less reactive EfOM fractions and individual PPCPs remain intact in these conditions. DA254/A0254 values exceeding the threshold of 0.1 were indicative of the commencement of PPCPs breakdown. For
instance, 50% and 90% of the initial sulfamethoxazole concentration were removed at DA254/A0254 values of 0.25 and 0.46, respectively. Further increases of DA254/A0254 values were accompanied by a complete removal of all the PPCPs except atrazine and iopromide. The DA254/A0254 values corresponding to the removal of 50% and 90% of PPCPs’ initial concentrations are compiled in Table 1. Compounds listed in it can be operationally separated into three groups. For the readily degradable PPCPs group (e.g. oxybenzone, acetaminophen, fluoxetine), (DA254/A0254)50% values were in the range 0.18e0.28, while the (DA254/A0254)90% values varied from 0.27 to 0.46. For the moderately degradable PPCPs group, the (DA254/A0254)50% values were quite similar ranging from 0.30 to 0.35. The DA254/A0254 value of the refractory PPCPs such as atrazine and iopromide were relative high, up to 0.58. For all examined PPCPs, major features of DC/C0 vs. DA254/ A0254 relationships obtained in the experiment with varying Fe (II) concentrations or contact time were close (Fig. 5). This observation is similar to that made in the studies of the oxidation of PPCPs by ozone (Nanaboina and Korshin, 2010). This result can be considered as a further confirmation that the DA254/A0254 parameter is useful for estimating the removal of PPCPs in wastewater treated using widely varying reaction conditions. Table 1 also shows the Fe (II) concentrations required to remove 50% and 90% of the PPCPs used in this study, as well as the reported rate constants kOH determined in prior researches for PPCPs oxidations by OH. A strong correlation (R2 0.95 and 0.93, respectively) was observed to exist between DA254/A0254 and Fe (II) concentrations required to attain 50% and 90% PPCPs removal thresholds. A noticeable correlation (R2 0.61) also existed between (DA254/A0254)50% and kOH values (Fig. 6). This correlation supports the point that the extent of the degradation of PPCPs is intrinsically correlated with the attendant changes of EfOM absorbance
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 0 3 e4 1 2
4.
Conclusions
This study quantified effects of Fenton treatment on the degradation of trace level PPCPs and concurrent transformations of EfOM. Results of the study can be outlined as follows: All examined PPCPs except atrazine and iopromide could be completely removed by Fenton oxidation carried out at a 20 mg/L Fe (II) dose and a 2.5 H2O2/Fe (II) molar ratio. Fenton treatment caused a considerable degree of DOC removal (up to 30%), while SEC data demonstrated that preferential removal of high molecular weight EfOM molecules and their breakdown into smaller fragments took place. Fenton treatment caused 13% of DOC to be oxidized to yield formate, acetate, oxalate and, less prominently, formaldehyde, acetaldehyde, propionaldehyde and glycolaldehyde. The formation of these oxidation products was strongly associated with changes of DOC and intensities of prominent features in the SEC profiles measured at 220 nm. Fenton oxidation caused the absorbance of EfOM to decrease consistently at all wavelengths >240 nm. The relative changes of absorbance at 254 nm (denoted as DA254/A0254) were strongly correlated with Fe (II) concentrations introduced into the system and also with the relative changes of concentrations of all tested PPCPs. Similar DC/C0 vs. DA254/ A0254 correlations were observed to exist in experiments with varying Fenton reagent doses and reaction times. DA254/A0254 values corresponding to 50% or 90% of removal of the tested PPCPs by Fenton reagent were well correlated with the intrinsic rates of interactions of these compounds with hydroxyl radical.
Acknowledgments Wei Li would like to thank the China Scholarship Council for supporting her studies at the Department of Civil and Environmental Engineering of the University of Washington. All authors wish to acknowledge the University of Washington and its VISIT program for their support of student exchange with Nankai University. The authors are deeply grateful to the Murdock Foundation whose support was critical for this study. The authors also want to thank the National Natural Science Foundation of China (grant No. 21037002) for their kind support and express their appreciation to the personnel of West Point Treatment Plant, Seattle and especially to Robert Bucher of King County Wastewater Technology Assessment Program for helpful discussions and their interest in and support of this study.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.11.002
411
references
Abella´n, M.N., Bayarri, B., Gime´nez, J., Costa, J., 2007. Photocatalytic degradation of sulfamethoxazole in aqueous suspension of TiO2. Appl. Catal. B: Environ. 74 (3, 4), 233e241. Acero, J.L., Stemmler, K., Von Gunten, U., 2000. Degradation kinetics of atrazine and its degradation products with ozone and OH radicals: a predictive tool for drinking water treatment. Environ. Sci. Technol. 34 (4), 591e597. APHA (American Public Health Association), 1998. In: Standard Methods for The Examination of Water and Wastewater, 20th ed. APHA, Washington, DC. Ay, F., Kargi, F., 2010. Advanced oxidation of amoxicillin by Fenton’s reagent treatment. J. Hazard. Mater. 179 (1e3), 622e627. Bandala, E.R., Martı´nez, D., Martı´nez, E., Dionysiou, D.D., 2004. Degradation of microcystin-LR toxin by Fenton and photoFenton processes. Toxicon 43 (7), 829e832. Bautista, P., Mohedano, A.F., Casas, J.A., Zazo, J.A., Rodriguez, J.J., 2008. An overview of the application of Fenton oxidation to industrial wastewaters treatment. J. Chem. Technol. Biotechnol. 83 (10), 1323e1338. Benner, J., Salhi, E., Ternes, T., Von Gunten, U., 2008. Ozonation of reverse osmosis concentrate: kinetics and efficiency of beta blocker oxidation. Water Res. 42 (12), 3003e3012. Burbano, A.A., Dionysiou, D.D., Suidan, M.T., 2008. Effect of oxidant-to-substrate ratios on the degradation of MTBE with Fenton reagent. Water Res. 42 (12), 3225e3239. Carballa, M., Omil, F., Lema, J.M., 2005. Removal of cosmetic ingredients and pharmaceuticals in sewage primary treatment. Water Res. 39 (19), 4790e4796. Deng, Y., 2007. Physical and oxidative removal of organics during Fenton treatment of mature municipal landfill leachate. J. Hazard. Mater. 146 (1, 2), 334e340. Dodd, M.C., Buffle, M.O., Von Gunten, U., 2006. Oxidation of antibacterial molecules by aqueous ozone: Moiety-specific reaction kinetics and application to ozone-based wastewater treatment. Environ. Sci. Technol 40 (6), 1969e1977. Dong, M.M., Mezyk, S.P., Rosario-Ortiz, F.L., 2010. Reactivity of effluent organic matter (EfOM) with hydroxyl radical as a function of molecular weight. Environ. Sci. Technol. 44 (15), 5714e5720. Ellis, J.B., 2006. Pharmaceutical and personal care products (PPCPs) in urban receiving waters. Envrion. Pollut. 144 (1), 184e189. Elmolla, E., Chaudhuri, M., 2009. Optimization of Fenton process for treatment of amoxicillin, ampicillin and cloxacillin antibiotics in aqueous solution. J. Hazard. Mater. 170 (2, 3), 666e672. Her, N., Amy, G., McKnight, D., Sohn, J., Yoon, Y., 2003. Characterization of DOM as a function of MW by fluorescence EEM and HPLC-SEC using UVA, DOC, and fluorescence detection. Water Res. 37 (17), 4295e4303. Huber, M.M., Canonica, S., Park, G.Y., Von Gunten, U., 2003. Oxidation of pharmaceuticals during ozonation and advanced oxidation processes. Environ. Sci. Technol. 37 (5), 1016e1024. Ikehata, K., Naghashkar, N.J., EI-Din, M.G., 2006. Degradation of aqueous pharmaceuticals by ozonation and advanced oxidation processes: a review. Ozone Sci. Eng. 28 (6), 353e414. Kasprzyk-Hordern, B., Dinsdale, R.M., Guwy, A.J., 2008. The occurrence of pharmaceuticals, personal care products, endocrine disruptors and illicit drugs in surface water in South Wales, UK. Water Res. 42 (13), 3498e3518. Kasprzyk-Hordern, B., Dinsdale, R.M., Guwy, A.J., 2009. The removal of pharmaceuticals, personal care products, endocrine disruptors and illicit drugs during wastewater treatment and its impact on the quality of receiving waters. Water Res. 43 (2), 363e380.
412
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 0 3 e4 1 2
Klamerth, N., Rizzo, L., Malato, S., Maldonado, M.I., Agu¨era, A., Ferna´ndez-Alba, A.R., 2010a. Degradation of fifteen emerging contaminants at mg L1 initial concentrations by mild solar photo-Fenton in MWTP effluents. Water Res. 44 (2), 545e554. Klamerth, N., Malato, S., Maldonado, M.I., Agu¨era, A., Ferna´ndezAlba, A.R., 2010b. Application of photo-Fenton as a tertiary treatment of emerging contaminants in municipal wastewater. Environ. Sci. Technol. 44 (5), 1792e1798. Klavarioti, M., Mantzavinos, D., Kassinos, D., 2009. Removal of residual pharmaceuticals from aqueous systems by advanced oxidation processes. Environ. Int. 35 (2), 402e417. Korshin, G., Chow, C.W.K., Fabris, R., Drikas, M., 2009. Absorbance spectroscopy-based examination of effects of coagulation on the reactivity of fractions of natural organic matter with varying apparent molecular weights. Water Res. 43 (6), 1541e1548. Korshin, G.V., Li, C.W., Benjamin, M.M., 1997. Monitoring the properties of natural organic matter through UV spectroscopy: a consistent theory. Water Res. 31 (7), 1787e1795. Kosma, C.I., Lambropoulou, D.A., Albanis, T.A., 2010. Occurrence and removal of PPCPs in municipal and hospital wastewaters in Greece. J. Hazard. Mater. 179 (1e3), 804e817. Lishman, L., Smyth, S.A., Sarafin, K., Kleywegt, S., Toito, J., Peart, T., Lee, B., Servos, M., Beland, M., Seto, P., 2006. Occurrence and reductions of pharmaceuticals and personal care products and estrogens by municipal wastewater treatment plants in Ontario, Canada. Sci. Total Environ. 367 (2, 3), 544e558. Me´ndez-Arriaga, F., Esplugas, S., Gime´nez, J., 2010. Degradation of the emerging contaminant ibuprofen in water by photoFenton. Water Res. 44 (2), 589e595. Mezyk, S.P., Neubauer, T.J., Cooper, W.J., Peller, J.R., 2007. Freeradical-induced oxidative and reductive degradation of sulfa drugs in water: absolute kinetics and efficiencies of hydroxyl radical and hydrated electron reactions. J. Phys. Chem. A 111 (37), 9019e9024. Muruganandham, M., Swaminathan, M., 2004. Decolourisation of reactive orange 4 by Fenton and photo-Fenton oxidation technology. Dyes Pigm. 63 (3), 315e321. Nam, S.N., Amy, G., 2008. Differentiation of wastewater effluent organic matter (EfOM) from natural organic matter (NOM) using multiple analytical techniques. Water Sci. Technol. 57 (7), 1009e1015. Nanaboina, V., Korshin, G.V., 2010. Evolution of absorbance spectra of ozonated wastewater and its relationship with the degradation of trace-level organic species. Environ. Sci. Technol. 44 (16), 6130e6137. Packer, J.L., Werner, J.J., Latch, D.E., McNeill, K., Arnold, W.A., 2003. Photochemical fate of pharmaceuticals in the environment: Naproxen, diclofenac, clofibric acid, and ibuprofen. Aquat. Sci. 65 (4), 342e351. Pe´rez-Moya, M., Graells, M., Castells, G., Amigo´, J., Ortega, E., Buhigas, G., Pe´rez, L.M., Mansilla, H.D., 2010. Characterization of the degradation performance of the sulfamethazine antibiotic by photo-Fenton process. Water Res. 44 (8), 2533e2540. Razavi, B., Song, W., Cooper, W.J., Greaves, J., Jeong, J., 2009. Freeradical-induced oxidative and reductive degradation of fibrate pharmaceuticals: kinetic studies and degradation mechanisms. J. Phys. Chem. A 113 (7), 1287e1294. Rosario-Ortiz, F.L., Mezyk, S.P., Doud, D.F.R., Snyder, S.A., 2008. Quantitative correlation of absolute hydroxyl radical rate
constants with non-isolated effluent organic matter bulk properties in water. Environ. Sci. Technol. 42 (16), 5924e5930. Rosario-Ortiz, F.L., Wert, E.C., Snyder, S.A., 2010. Evaluation of UV/H2O2 treatment for the oxidation of pharmaceuticals in wastewater. Water. Res. 44 (5), 1440e1448. Rosenfeldt, E.J., Linden, K.G., 2004. Degradation of endocrine disrupting chemicals bisphenol A, ethinyl estradiol, and estradiol during UV photolysis and advanced oxidation processes. Environ. Sci. Technol. 38 (20), 5476e5483. Rozas, O., Contreras, D., Mondaca, M.A., Pe´rez-Moya, M., Mansilla, H.D., 2010. Experimental design of Fenton and photo-Fenton reactions for the reatment of ampicillin solutions. J. Hazard. Mater. 177 (1e3), 1025e1030. Song, W., Cooper, W.J., Mezyk, S.P., Greaves, J., Peake, B.M., 2008. Free radical destruction of beta-blockers in aqueous solution. Environ. Sci. Technol. 42 (4), 1256e1261. Song, W., Cooper, W.J., Peake, B.M., Mezyk, S.P., Nickelsen, M.G., O’Shea, K.E., 2009. Free-radical-induced oxidative and reductive degradation of N, N0 -diethyl-m-toluamide (DEET): kinetic studies and degradation pathway. Water Res. 43 (3), 635e642. Swietlik, J., Dabrowska, A., Raczyk-Stanisławiak, U., Nawrocki, J., 2004. Reactivity of natural organic matter fractions with chlorine dioxide and ozone. Water Res. 38 (3), 547e558. Trapido, M., Kulik, N., Goi, A., Veressinina, Y., Munter, R., 2009. Fenton treatment efficacy for the purification of different kinds of wastewater. Water Sci. Technol. 60 (7), 1795e1801. Ternes, T.A., Meisenheimer, M., McDowell, D., Sacher, F., Brauch, H., Haist-Gulde, B., Preuss, G., Wilme, U., ZuleiSeibert, N., 2002. Removal of pharmaceuticals during drinking water treatment. Environ. Sci. Technol. 36 (17), 3855e3863. Trovo´, A.G., Nogueira, R.F.P., Agu¨era, A., Fernandez-Alba, A.R., Sirtori, C., Malato, S., 2009. Degradation of sulfamethoxazole in water by solar photo-Fenton. Chemical and toxicological evaluation. Water. Res. 43 (16), 3922e3931. Wang, S., 2008. A comparative study of Fenton and Fenton-like reaction kinetics in decolourisation of wastewater. Dyes Pigm. 76 (3), 714e720. Wert, E.C., Rosario-Ortiz, F.L., Drury, D.D., Snyder, S.A., 2007. Formation of oxidation byproducts from ozonation of wastewater. Water. Res. 41 (7), 1481e1490. Wert, E.C., Rosario-Ortiz, F.L., Snyder, S.A., 2009a. Effect of ozone exposure on the oxidation of trace organic contaminants in wastewater. Water. Res. 43 (4), 1005e1014. Wert, E.C., Rosario-Ortiz, F.L., Snyder, S.A., 2009b. Using ultraviolet absorbance and color to assess pharmaceutical oxidation during ozonation of wastewater. Environ. Sci. Technol 43 (13), 4858e4863. Westerhoff, P., Yoon, Y., Snyder, S., Wert, E., 2005. Fate of endocrine-disruptor pharmaceutical, and personal care product chemicals during simulated drinking water treatment. Environ. Sci. Technol. 39 (17), 6649e6663. Yao, C.C.D., Haag, W.R., 1991. Rate constants for direct reactions of ozone with several drinking water contaminants. Water Res. 25 (7), 761e773. Yoon, Y., Ryu, J., Oh, J., Choi, B.G., Snyder, S.A., 2010. Occurrence of endocrine disrupting compounds, pharmaceuticals, and personal care products in the Han River (Seoul, South Korea). Sci. Total Environ. 408 (3), 636e643. Zhang, H., Zhang, D., Zhou, J., 2006. Removal of COD from landfill leachate by electro-Fenton method. J. Hazard. Mater. 135 (1e3), 106e111.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 1 3 e4 2 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A continuous-flow approach for the development of an anaerobic consortium capable of an effective biomethanization of a mechanically sorted organic fraction of municipal solid waste as the sole substrate Lorenzo Bertin a,c,*, Cristina Bettini a, Giulio Zanaroli a, Dario Frascari b, Fabio Fava a,c a
Department of Civil, Environmental and Materials Engineering (DICAM), Faculty of Engineering, University of Bologna, via Terracini 28, 40131 Bologna, Italy b Department of Chemical, Mining and Environmental Engineering (DICMA), Faculty of Engineering, University of Bologna, via Terracini 28, 40131 Bologna, Italy c INCA - Interuniversitary Consortium “Chemistry for the Environment”, via delle Industrie 21/8, 30175 Marghera (VE), Italy
article info
abstract
Article history:
An effective mesophilic continuous anaerobic digestion process fed only with a mechanically
Received 27 May 2011
sorted organic fraction of municipal solid waste (MS-OFMSW) was developed. During
Received in revised form
a preliminary 3-month experimental phase, the microbial consortium was acclimated
30 October 2011
toward MS-OFMSW by initially filling the reactor with cattle manure and then continuously
Accepted 1 November 2011
feeding it with MS-OFMSW. The Hydraulic Retention Time (HRT) and Organic Loading Rate
Available online 11 November 2011
(OLR) were 23 days and 2.5 g/L/day, respectively. After 4 weeks, the reactor reached stationary performances (84% COD removal yield, 0.15 LCH4 /gCODremoved methane production yield). The
Keywords:
acclimated consortium was then employed in a second run in which the reactor was operated
Anaerobic digestion
under steady state conditions at the previous HRT and OLR for 73 days. The COD removal and
Mechanically sorted municipal solid
the methane production yield increased up to 87% and 0.25 LCH4 /gCODremoved, respectively.
waste
The capability of the acclimated consortium to biomethanize MS-OFMSW was further
Bioreactor
studied via batch digestion experiments, carried out by inoculating the target waste with
Continuous culture
reactor effluents collected at the beginning of first run and at the end of the first and second
Archaeal community
run. The best normalized methane production (0.39 LCH4 /ginitial COD) was obtained with the
Bacterial community
inoculum collected at the end of the second run. Molecular analysis of the microbial community occurring in the reactor during the two sequential runs indicated that the progressive improvement of the process performances was closely related to the selection and enrichment of specific hydrolytic and acidogenic bacteria in the reactor. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Waste management is an issue of great relevance under both economical and environmental standpoints. The European
Union Directive 2008/98/EC promotes a more sustainable management of all waste types through three principles: waste prevention, recycling and reuse (EC, 2008a). Nowadays, according to European Commission statistics, more than
* Corresponding author. Department of Civil, Environmental and Materials Engineering (DICAM), Faculty of Engineering, University of Bologna, via Terracini 28, 40131 Bologna, Italy. Tel.: þ39 (0) 51 2090317; fax: þ39 (0) 51 2090322. E-mail addresses:
[email protected] (L. Bertin),
[email protected] (C. Bettini),
[email protected] (G. Zanaroli),
[email protected] (D. Frascari),
[email protected] (F. Fava). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.001
414
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 1 3 e4 2 4
500 kg per capita of municipal solid waste (MSW) are produced annually in the European Union, with an overall generation of over 260 Mt (Skovgaard et al., 2008). The recycling capacity of different fractions of MSW was reported to concern less than one third of the total MSW production, while about 108 Mt are disposed in landfills and 58 Mt are incinerated each year (EC, 2008b). However, the use of the latter approaches is decreasing, due to their low environmental sustainability (Eriksson et al., 2005; ISPRA, 2010; Stroot et al., 2001). Thus, innovative and more environmentally sustainable solutions for biowaste management and exploitation are under identification and assessment. According to the goals of Directive 2008/98/EC, the recovery of the MSW dry matter by means of mechanical separation processes represents a valuable solution, since its sequential disposal in landfills or incinerators would result, respectively, in lower waste volumes to be dumped or higher combustion efficiencies, with respect to those of traditional MSW management. At the same time, the resulting wet fraction of MSW - the so called mechanically sorted OFMSW (MSOFMSW) - could be valorized by means of an anaerobic digestion (AD) treatment to provide additional economical and environmental benefits to the new management strategy. However, the efficiency of AD of MS-OFMSW is very often limited by the occurrence of toxic pollutants, such as heavy metals (Bolzonella et al., 2006; Hartmann and Ahring, 2006, 2005; Hartmann et al., 2004). For this reason, AD is currently mostly applied to the biomethanization of source sorted OFMSW (SS-OFMSW), which however represents only a small amount of the entire MSW production. Recent studies showed that the inhibition of methane production due to MS-OFMSW pollutants can be mitigated through co-digestion approaches (Goberna et al., 2010; Bertin et al., 2008; Hartmann and Ahring, 2005). However, low amounts of MS-OFMSW per digester volume can be processed via this approach. The aim of this work was to develop an effective microbial consortium able to carry out a continuous AD process fed only with an MS-OFMSW. To achieve this goal, an anaerobic consortium was adapted to an experimental MS-OFMSW in a column recycled bioreactor by initially filling the reactor with cattle manure and then continuously feeding it with only MS-OFMSW. The acclimated consortium was then employed as the inoculum in a second run carried out under the same loading conditions. The evolution of both the bacterial and archaeal communities was monitored throughout the whole study by means of DGGE analyses. To the best of our knowledge, this study represents the first attempt to obtain and characterize a microbial consortium capable of an effective biomethanization of an MS-OFMSW as the sole substrate.
2.
Materials and methods
2.1.
Substrates
The MS-OFMSW used in the experiments was obtained from the municipal solid waste treatment plant of Ostellato (Ferrara, Italy). It was obtained by applying the following mechanical pre-treatments of unsorted MSW: chopping, screening, magnetic deferrization, pressing (via a Doppstaadt extruder DSP 20-5 supplied with a rotating cone and mounted to mixer Doppstaadt DM 215, http://www.doppstadt.com/ dsp_205_en/dsp-20/) and separation of the aqueous organic suspension from the dry fraction, mainly constituted by plastic waste. The resulting suspension was diluted with the leacheate of a non-pressed fraction of the same unsorted MSW, which underwent a conventional composting process. Before its employment, the resulting stream was milled and sieved (cut off: 3 mm). The obtained experimental MS-OFMSW had a 5% (w/w) total solid (TS) content and a 2.5% volatile solid (VS) content (Table 1). Although these values are rather low for an MSW-derived waste stream, the dilution with leachate released from the untreated MSW ensures the presence of any compound e included the toxic ones e that can be released in water, and that therefore can interact with the microbial processes. The soluble COD of the experimental MS-OFMSW was about 57 g/L, a large part of which was represented by volatile fatty acids (VFAs) (30 gCOD/L). Acetic, propionic, butyric and caproic acid together contributed to the whole VFA mixture for more than 60% (Table 1). The concentration of the other VFAs were: isobutyric 2.3 g/L, isovaleric 2.2 g/L, valeric 2.4 g/L, isocaproic 2.2 g/L, eptanoic 2.3 g/L. The high VFA concentration was responsible for the low pH (4.8). The COD/ VS ratio, equal to 2.3 g/g, is slightly higher than the values typical of agro-industrial wastes (1e2 g/g). The MS-OFMSW was contaminated by heavy metals, mainly Mn (25 mg/kgTS), Fe (13 mg/kgTS) and Al (12 mg/kgTS); B, Sr, Cu, Zn and Pb were also detected (7, 7, 4, 3 and 1 mg/kgTS, respectively). The reactor was initially filled with cattle manure (CM), after milling, sieving (cut off: 3 mm) and diluting it in tap water. The resulting experimental CM contained 4.5% w/w of TS, 3.8% w/w of VS, 6.2 g/L of COD and 3.7 g/L of VFAs, and had a pH of 6.7 (Table 1).
2.2.
Reactor features and set up
The continuous-flow tests were conducted in an upflow recycled anaerobic reactor, consisting of a thermostated glass column (inner diameter: 38 mm; height: 350 mm; reaction
Table 1 e Main chemicalephysical parameters of the organic wastes employed in the study, with 95% confidence intervals. pH
CM MS-OFMSW
6.7 0.1 4.8 0.1
TS
VS
COD
Total VFAs
Acetic acid
Propionic acid
Butyric acid
Caproic acid
%
%
g/L
gCOD/L
gCOD/L
gCOD/L
gCOD/L
gCOD/L
4.5 0.1 4.8 0.3
3.82 0.06 2.5 0.2
6.2 0.3 58 1
3.7 0.2 30 1
0.96 0.05 4.2 0.1
0.48 0.02 2.4 0.1
0.40 0.01 6.1 0.5
0.32 0.02 5.4 0.4
415
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 1 3 e4 2 4
5
was at atmospheric pressure. When biogas was produced, a corresponding volume of water was transferred from the bottle to the cylinder: the produced biogas was measured as the increment of the water volume observed in the cylinder. A pH probe (81-04 model, ATI Orion, Boston, MA) was placed at the top of the bioreactor. The operational temperature was 35 C. The reactor was started under batch condition after completely filling it with CM. It was switched to a continuous feed of only MS-OFMSW after 7 days.
4
2.3.
3
1 2
Fig. 1 e Reactor scheme: peristaltic pumps regulating the inlet (1), outlet (2) and recycle (3) liquid flows; outlet gas flow connected to the biogas measurement system (4); pH probe (5).
volume 0.430 L, including the external recycle line; Fig. 1). The recycling ratio, expressed as the ratio between the flow of the recycled liquor and the whole flow entering the column, was equal to 0.97, corresponding to a ratio between the recycled and the inlet flows of 32.3. The inlet and outlet lines intersected the recycle line in the proximity of the column bottom. The reactor headspace (0.055 L) was connected to a “Mariotte” system, which was utilized to measure the volume of the produced biogas. It consisted of a hermetically closed 2-L bottle filled with 1 L of water hydraulically connected to a graduated cylinder, so that the heights of the bottle and cylinder water levels coincided when the anaerobic reactor
Experiments with the continuous-flow bioreactor
The above described reactor was employed in two consecutive experimental runs. Run 1 was dedicated to the selection and enrichment of an effective anaerobic consortium and it was carried out by continuously feeding the reactor with the experimental MS-OFMSW (x 2.1) at a Hydraulic Retention Time (HRT) of about 23 days for 3 months, with a 2.5 gCOD/L/day Organic Loading Rate (OLR) (Table 2). The selected anaerobic consortium was then employed as the inoculum for a conventional continuous process fed with the same loading conditions for 73 days (Run 2). To this aim, 50 mL of residue mixed liquor were left in the reactor and mixed with 380 mL of fresh experimental MS-OFMSW under nitrogen flux. Before the beginning of the second experiment, the reactor was operated under batch conditions for 2 months. The main process parameters of both continuous runs are reported in Table 2. Since the recycling ratio was equal to 0.97, the column reactor was approximated from a fluid-dynamic standpoint to a perfectly mixed reactor. Thus, the COD mass balance in the reactor was written as: dCOD CODin COD ¼ þ rCOD dt HRT
(1)
where CODin (g/L) represents the COD concentration in the inlet stream, and rCOD (g/L/day) the COD reaction term. Eq. (1) was utilized to evaluate the average COD removal rate for each period Dtj,jþ1 comprised between two consecutive monitoring times: rCODj;jþ1 ¼
CODin CODj;jþ1 CODjþ1 CODj HRT tjþ1 tj
(2)
To evaluate the temporal trend of the COD contained in CM (CODCM), Eq. (1) was re-written for CODCM, assuming a firstorder kinetic for both the total COD and CODCM (rCOD ¼ k,COD, rCODCM ¼ kCM ,CODCM ), supposing that k ¼ kCM and calculating, for each period Dtj,jþ1, kj,jþ1 as ðrCODj;jþ1 =CODj;jþ1 Þ, with rCODj;jþ1 obtained from Eq. (2). This procedure is likely to lead to an underestimate of rCODCM .
Table 2 e Main process parameters related to the acclimatization (Run 1) and validation (Run 2) experiments: averages of the daily performances ± 95% confidence intervals.
Run 1 Run 2
Experimental period
Inlet flow
HRT
Inlet COD
OLR
COD removal percentage
COD removal rate
CH4 production yield
Day
mL/day
Day
g/L
g/L/day
%
g/Lreactor/day
LCH4 /gCODremoved
87 73
18.6 0.3 19.2 0.4
23.1 0.4 22.4 0.4
57 1 57 1
2.49 0.08 2.54 0.09
84 2 89 1
2.15 0.06 2.32 0.09
0.15 0.01 0.25 0.02
416
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 1 3 e4 2 4
Indeed, on the basis of the Ks values reported in the literature on anaerobic digestion (Ako et al., 2008; Bryer, 1985; Wiesmann et al., 2007), the high total CODs tested in this work (6e10 g/L) are likely to be way above the first-order region; this determines an underestimate of the rate of the CODCM removal process, which occurred in the 0e5.8 g/L range, as the COD data utilized to estimate the first-order constant are in a higher concentration range (7e10 g/L). On the basis of the results of the biomethanization experiments (x 2.4), the hypothesis k ¼ kCM determines a further underestimate of rCODCM , as will be discussed in x 3.3. Considering that CODCM,in ¼ 0, the CODCM mass balance becomes:
column initial temperature 40 C; 1 min isotherm; temperature rate 25 C/min; final temperature 150 C; 6 min isotherm; temperature rate 4 C/min; final temperature 180 C; temperature rate 25 C/min; final temperature 240 C; detector temperature 280 C). The analytical samples were diluted with an oxalic acid solution (60 mM) in the ratio 1:1 in order to reduce all carboxyl terminations of VFAs. This method allowed the monitoring of the following VFAs: acetic, propionic, isobutyric, butyric, isovaleric, valeric, isocaproic, caproic and eptanoic acid. Heavy metals were determined by inductively coupled plasma optical emission spectroscopy (ICP-OES) after digestion with nitric acid 65%.
dCODCM CODCM ¼ k,CODCM dt HRT
2.6.
(3)
The CODCM was thus evaluated by means of a numerical integration of Eq. (3).
2.4.
Microcosm-based biomethanization experiments
Batch AD processes were set up in 0.1 L-anaerobic serum bottles, in order to evaluate if the above described experimental approach allowed an improvement of the capability of the anaerobic consortium acclimated throughout Run 1 and Run 2 to biomethanize MS-OFMSW without the addition of any external substrate. To this goal, reactor effluents collected under anaerobic conditions at the beginning and at the end of Run 1 and at the end of Run 2 were employed to inoculate (6 mL) three sets of duplicate batch microcosms, hermetically closed with a double Teflon-coated cap and filled with 46 mL of MSOFMSW. Three further sets of MS-OFMSW-containing duplicate microcosms were set up as controls: the 1st set was inoculated (6 mL) with undigested CM, the 2nd was inoculated (6 mL) with the effluent from an AD process fed with olive mill wastewater, and the 3rd was not inoculated. Furthermore, a last set of duplicate microcosms was set up with only CM, in order to compare the normalized biomethane production obtained from CM with that achieved with MS-OFMSW. In all microcosms, the pH was initially set at 7 with drops of NaOH 10 M. The biogas volume and composition were periodically determined all along the 3-month experiments. After each biogas sampling, the medium pH was re-adjusted to 7 while purging the microcosms with N2. The batch tests were run at 35 C.
2.5.
Analytical methods
TS and VS were analyzed according to Standard Methods for the Examination of Water and Wastewater (1998). COD values were obtained via catalytic oxidation and spectroscopic analysis following the Hach Mn(III) method (Miller et al., 2001). The volume of the produced biogas was determined via water displacement in a Mariotte flask, while the biogas composition was analyzed daily through an Agilent microGC 3000 coupled with a TCD detector (injector temperature 90 C; column temperature 60 C; sampling time 20 s; injection time 50 ms; column pressure 25 psi; run time 44 s; carrier gas, N2). VFAs were analyzed by an Agilent GC-FID (model 7890A) equipped with a Supelcowax-10 30 m 0.25 mm 0.25 mm capillary column (SigmaeAldrich, Milano, Italy) (injection volume 1 ml; injector temperature 250 C; column head pressure 5 psi;
Molecular analysis
The composition of the microbial community was investigated throughout the Runs 1 and 2, and compared with that of the cattle manure (CM) used as an inoculum to start up the MSOFMSW AD process. In Run 1, samples of microbial consortium were collected from the bioreactor after 7, 28, 42, 63 and 84 days of operation (samples 1d7, 1d28, 1d42, 1d63 and 1d84, respectively). During Run 2, microbial samples were taken from the bioreactor after 10 (2d10) and 73 (2d73) days. Genomic DNA was extracted from approximately 60 mg of pellet obtained after centrifugation at 13,000 rpm for 10 min. DNA was extracted with the UltraClean Soil DNA kit (Mo Bio Laboratories, Carlsbad, CA, USA) according to the manufacturer’s instructions. DNA quality was checked on 1.0% (w/v) agarose gel stained with ethidium bromide. For Bacterial DGGE analysis, PCR amplification was performed with primers GC-357f (50 CGCCCGCCGCGCCCCGCGCCCGGCCCGCCGCCCCCGCCCCCTACGGGAGGCAGCAG-30 ) and 907r (50 -CCGTCAATTCCTTTGAGTTT-30 ) (Sass et al., 2001) in 50 mL reaction mixtures containing 1 PCR buffer (Invitrogen, Paisley, UK), 1.5 mM MgCl2, 0.2 mM each dNTP, 0.4 mM each primer, 1.0 U of Taq polymerase (Invitrogen, Paisley, UK) and 2 mL of template DNA. The reaction began with an initial denaturation at 95 C for 5 min, followed by 30 cycles of 95 C for 30 s, 55 C for 30 s, 72 C for 45 s and a final extension at 72 C for 7 min. Archeal 16S rRNA gene DGGE analysis was performed with primers GC-344f (50 CGCCCGCCGCGCCCCGCGCCCGTCCCGCCGCCCCCGCCCCACGGGG(C/T)GCAGCAGGCGCGA-30 ) (Raskin et al., 1994) and 915r (50 GTGCTCCCCCGCCAATTCCT-30 ) (Stahl and Amann, 1991) in 50 mL reactions as described above. PCR products were resolved with a D-Code apparatus (Bio-Rad, Milan, Italy) on a 7% (w/v) polyacrylamide gel (acrylamide-N,N0 -methylenebisacrylamide, 37:1) in 1 TAE with a denaturing gradient from 40% to 60% denaturant, where 100% denaturant is 7 M urea and 40% (v/v) formamide. The electrophoresis was run at 55 V for 16 h at 60 C. The gel was stained in a solution of 1 SYBR-Green (Sigma Aldrich, Milwaukee, WI) in 1 TAE for 30 min and its image captured in UV transillumination with a digital camera supported by a Gel Doc apparatus (Bio-Rad, Milan, Italy). Subsequently, a matrix of similarities between the densitometric curves of the band patterns was calculated based on the Dice coefficient with Quantity One 4.6.5 software (Bio-Rad), and the DGGE patterns were clustered based on the unweighted pairegroup arithmetic average (UPGAMA) clustering algorithm. Finally, bands were excised from the gel with a sterile scalpel and DNA was eluted in 50 mL of sterile deionized water at 4 C for
417
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 1 3 e4 2 4
B
10
9
9
8
8
7
7
6
6
5
5
4
4
3
3
2
2
1
1
COD removal rate (g/L/day)
COD and CODCM (g/L)
A 10
0
0 0
15
30
45
60
75
90
0
15
30
45
60
75
90
Time (day)
Time (day)
Fig. 2 e Measured COD concentrations and simulated trend of CODCM (> and thick black line, respectively; left axis, g/L) along with COD removal rates (-, right axis, g/L/day) related to Runs 1 (A) and 2 (B). Both the experimental COD values and the COD removal rates are interpolated with a 7-point moving average (thin black lines).
3.
Results
3.1. Culture adaptation to MS-OFMSW under continuous conditions: Run 1
Methane production rate (LCH4/ Lreactor /day)
The initial 7-day batch incubation of the CM-filled bioreactor led to the onset of methane production after a 4-day lag phase.
The initial methane production rate was equal to 0.18 LCH4 /Lreactor/day. The batch test led to a decrease of CODCM from 6.2 to 5.8 g/L. The latter value was utilized as the initial concentration in the integration of Eq. (3) aimed at estimating the temporal trend of CODCM during the subsequent continuous-flow process. The COD concentration as a function of the time during the first continuous-flow test (Run 1) is reported in Fig. 2A, together with the corresponding 7-point moving average. Starting from the initial value of 5.8 g/L, the reactor COD increased during the initial 30 days of operation, remained roughly constant (8e9 g/L) until day 77, and decreased down to 6.5 g/L during the final 10 days. The average COD removal resulted equal to 84% (Table 2). As displayed in Fig. 2A, the COD removal rate, calculated according to Eq. (2), was roughly constant during Run 1, and equal to 2.15 0.06 g/L/day (Table 2). The methane volumetric production, shown in Fig. 3A, reached 0.37 LCH4 /Lreactor/day at the 20th day of the experiment, and remained almost constant afterward (0.33 0.01 LCH4 /Lreactor/day). The relative amount of methane in the biogas followed a similar profile: it increased until day 20 (when it was 82% of the biogas), than it remained almost constant (69 1%). A significant biomethanization of the
B
A
0.8
0.8
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
15
30
45
Time (day)
60
75
90
0
15
30
45
60
Time (day)
75
90
0
Methane production yield (LCH4 /gCOD removed)
16 h. Four mL of the solution were then used as template to reamplify the band fragment using the same primers without the GC-clamp (357f-907r and 344f-915r) and the same PCR conditions described above. The obtained bacterial and archaeal amplicons were then sequenced with primers 357f and 344f, respectively. Sequencing was performed after amplicon purification with EXOSAP (USB Corporation, Cleveland, Ohio, US) according to the manufacturer’s instructions. Sequencing reactions and runs were performed by BMR Genomics (Padova, Italy). For each 16S rRNA gene sequence, the most closely related sequence and the sequence of the most closely related cultured bacterial strain were retrieved from the Ribosomal Database Project-II website with the SEQMATCH tool. The phylogenetic affiliation of each sequence was retrieved from the same website with the CLASSIFIER tool.
Fig. 3 e Methane production rate (>, left axis, LCH4 /Lreactor/day) and yield (-, right axis, LCH4 /gCODremoved) related to Runs 1 (A) and 2 (B), with 7-point moving averages (thin black lines).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 1 3 e4 2 4
3.2. Performances of the acclimated culture under continuous feeding with MS-OFMSW: Run 2 Starting from the third week of Run 2, all the performance parameters resulted quite stable: COD concentration (7.2 0.9 g/L, Fig. 2B), COD removal (89 1%), COD depletion rate (2.32 0.09 g/L/day, calculated according to Eq. (2); Fig. 2B and Table 2), methane volumetric production (0.59 0.01 LCH4 /Lreactor/day, Fig. 3B) and methane production yield (0.25 0.02 LCH4 /gCODremoved, Table 2 and Fig. 3B). The relative amount of methane in the produced biogas was 46 1%. The improvement in process performance from the 1st to the 2nd Run (methane production rate þ79%, methane/COD yield þ67%) indicates that the full acclimation to MS-OFMSW of the initial CM-derived consortium required a long-term AD process. The contribution of the total VFAs to the effluent COD was averagely 66%. Propionic acid was the main detected VFA and its concentration was about one fifth of the total VFA concentration. Acetic acid represented about 15% of the VFA mixture. The concentrations of the other seven monitored VFAs were comparable (generally lower than 10% of the VFA mixture), and together contributed to the remaining 65% of the VFAs.
3.3.
Biomethanization experiments in microcosms
The normalized methane production (LCH4 /ginitial COD) obtained in the biomethanization microcosm tests is shown in Fig. 4. The 3 sets of MS-OFMSW-containing control microcosms (inoculated with undigested CM, inoculated with effluent from an AD process fed with olive mill wastewater, non-inoculated) are not shown in Fig. 4, since they did not lead to any methane production during the 3-month monitoring period. This outcome indicates that, as expected, MS-OFMSW is not biomethanizable with “conventional” inocula. The non-inoculated MS-OFMSW-containing microcosm represents a control test also for the MS-OFMSW-fed continuous-flow process. Fig. 4 shows that the continuous AD process fed with MSOFMSW, started up with the help of a 100% inoculum of digested CM, led to an increasing capacity of the anaerobic consortium to biomethanize MS-OFMSW. Indeed, the 85-day MS-OFMSW normalized methane production resulted close to zero with the reactor effluent sampled at the beginning of Run 1 (digested CM), close to that of CM with the effluent sampled at the end of Run 1, and 2.5 times higher than that of CM with the effluent sampled at the end of Run 2. It is noteworthy that the 85-day batch test inoculated with reactor effluent sampled at the onset of Run 1, consisting of partly digested CM and containing an active CH4-producing population, resulted in a very poor methane production. This outcome can be explained by considering that, while the batch test contained a 13% inoculum of CH4-producing CM (with the remaining 87% made of MS-OFMSW), the continuous-flow process was started with a 100% load of CM, which was gradually replaced by MS-OFMSW. The gradual replacement probably allowed the CM-derived bacterial and archeal population to slowly adapt to the new, poorly fermentable waste, while continuing to produce CH4 with satisfactory rates and yields. Interestingly, while the methane production yield of Run 2 (with a 22-day HRT) was equal to 0.25 LCH4 /gCODremoved, the 85day microcosm test inoculated with reactor effluent sampled at the end of Run 2 resulted in a normalized methane of 0.39
0.4
(LCH4 /ginitial COD)
removed organic matter was observed: the methane production yield increased up to 0.21 LCH4 /gCODremoved during the initial 40 days, then it slowly decreased until the end of the experiment (Fig. 3A), with an overall average value during Run 1 equal to 0.15 0.01 LCH4 /gCODremoved (Table 2). Fig. 2A shows the simulated trend of the CODCM evaluated according to Eq. (3). After an experimental period equal to the process HRT (23 days), the calculated CODCM was less than 0.1% of the total measured COD, indicating that the methane produced after day 23 entirely derived from the AD of MSOFMSW. The progressive replacement of CODCM with CODMSOFMSW was accompanied by a constant COD removal rate and by a slight decrease of the CH4 production rate, rapidly followed by a sustained increase of the same parameter. These results indicate that the initial consortium, obtained from the batch digestion of CM, rapidly developed the capability to methanize MS-OFMSW without any long-term decrease of the methanization rate or yield. The total VFA concentration, which represented over 80% of the CODCM (Table 1), was around 5 g/L. Acetic acid was the main VFA detected during the first 3 weeks, when its relative concentration among the detected VFAs was about 30%, while the amounts of all the other VFAs (propionic, isobutyric, butyric, isovaleric, valeric, isocaproic, caproic and eptanoic acid) were comparable and in the range 7e13% of the whole VFA mixture. Conversely, starting from the 4th week, the main contribution to the VFA mixture was given by propionic acid (24e37%), whereas acetic acid slightly decreased (20e27%). The relative concentration of butyric acid ranged between 10% and 13% during Run 1. At the end of Run 1, 88% of the reactor mixed liquor was replaced with fresh MS-OFMSW. The reactor was then operated in batch mode for 60 days, with an initial COD equal to 51.6 g/L. The batch test resulted in a final COD of 5.9 g/L, corresponding to an average COD removal rate of 0.77 g/Lreactor/ day.
Methane production yield
418
0.3 0.2 0.1 0
0
15
30
45
60
75
90
Time (day) Fig. 4 e Normalized methane production (LCH4 /ginitial COD) as a function of the time, in non-inoculated CM-containing microcosms (X) and in MS-OFMSW containing microcosms inoculated with reactor effluent sampled at the beginning of the first run (:), at the end of the first run (A) and at the end of the second run (-).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 1 3 e4 2 4
LCH4 /ginitial COD, which corresponds to a nearly identical value in terms of LCH4 /gCODremoved, since COD removal in the latter microcosms was basically complete (data not shown). This outcome indicates that the COD of the reactor effluent carried a not negligible methanization potential, and therefore that a higher HRT would have probably resulted in a higher methane/COD yield. Fig. 4 shows that the methane production rate of the CMcontaining microcosms is significantly higher than that of the MS-OFMSW-containing microcosms inoculated with reactor effluent sampled at the beginning of Run 1. This result indicates that, in the evaluation of the temporal trend of CODCM performed as described in x 2.3 and shown in Fig. 2A, the hypothesis kCM ¼ k determines an underestimate of the rate of CODCM depletion.
3.4. Main features of the microbial consortium enriched in the two successive runs The bacterial community of the cattle manure (CM) used as inoculum mainly consisted of five phylotypes belonging to the Firmicutes (bands 8, 13, 14, 15 and 16) and two Bacteroidetes (bands 7 and 22) (Fig. 5A, Table 3), whereas its archaeal community (Fig. 5B) was composed by only two phylotypes (bands 1A and 2A) having 98% sequence identity to each other and 93% and 94% sequence identity with Methanosarcina mazei, respectively. The same archaeal phylotypes initially occurring in CM were detected during the continuous AD process all along Run 1 and Run 2 (Fig. 5B). Conversely, the reactor’s bacterial community resulted to be significantly different from that of the inoculated CM already after 1 week of continuous feeding with MS-OFMSW in Run 1 (1d7, Fig. 5A), when bands 7, 8, 16 and 22 disappeared and two major phylotypes distantly related to the genera Anaerorhabdus (band 3) and Clostridium (band 18) and a Bacteroidetes (band 11) were detected (Fig. 5A, Table 3). Other remarkable changes occurred during the successive weeks. In particular, several phylotypes enriched only transiently (band 1 at day 28, bands 2 and 10 at day 42, band 3 at day 63, band 11 at days 28 and 42, and band 13 at days 42 and 63; Fig. 5A), whereas four Firmicutes phylotypes (bands 6, 8, 16 and 18), two Bacteroidetes (bands 8 and 11) and a bacterium distantly related to the genus Desulfomicrobium (band 5) were detected all along Run 1 (Fig. 5A, Table 3). At the end of the enrichment process (day 84), the highest bacterial biodiversity was observed, along with the enrichment of two Bacteroidetes, namely an Alkaliflexus sp. (band 4) and a phylotype distantly related to the genus Alistipes (band 9), an Aminobacterium sp. (band 19) and a Bacillus sp. (band 20). Based on relative intensities of bands, these phylotypes represented, along with a Sedimentibacter sp. (band 8) and a Firmucutes having low sequence similarity to Clostridium clariflavum (band 18), the most abundant eubacteria of the microbial community occurring in the bioreactor at the end of Run 1 (Fig. 5A, Table 3). Some of the most prominent bacterial bands enriched in Run 1 (bands 18, 19 and 20) had very low intensity at the beginning of Run 2 (day 10, Fig. 5A), when bands 3, 4, 5, 6, 7, 8, 9 and 16 represented the most abundant bacterial phylotypes of the microbial population. Apart from the enrichment of a few bands (bands 13, 17 and 18), the composition of the bacterial community did not
419
change significantly up to the end of Run 2 (Fig. 5A), when only three phylotypes initially occurring in the cattle manure, namely a Sedimentibacter sp., a Clostridium maritimum strain and a Proteiniphilum sp. (bands 7, 8 and 16), were detected. Cluster analyses (Fig. 6) based on the DGGE profiles showed that CM did not cluster with any of the samples collected during the continuous AD process. In addition, the bacterial communities detected during Run 1 clustered together, with the exception of that enriched at the end of Run 1 (1d84), which clustered with the Run 2 communities sub-cluster.
4.
Discussion
The present work was aimed at developing an anaerobic consortium capable of an effective biomethanization of the organic matter occurring in an MS-OFMSW, which was demonstrated to be not biomethanizable in preliminary batch tests carried out without inoculation and with unacclimated inocula. This fact could be ascribed to the presence of heavy metals, which typically inhibit anaerobic consortia, and in particular methanogenic populations, at concentrations comparable to those observed in the feed utilized in this study (Chen et al., 2008; Colussi et al., 2009; Sarioglu et al., 2010). However, we cannot exclude that other chemicals and/or factors could have adversely affected methane production in the batch control tests. Significant COD removal and methane production yields were observed all along Run 1 (Figs. 2A and 3A, respectively), also when the organic matter existing in the reactor was reasonably due only to MS-OFMSW (Section 3.1). This result demonstrates a remarkable adaptation of the AD microbial population toward the experimental MS-OFMSW, since no methane was produced in the non-inoculated batch tests of MSOFMSW biomethanization used as control. Importantly, biomethane production increased during the first 25 days, indicating an increasing acclimatization of the anaerobic consortium throughout the same period (Fig. 3A). Furthermore, a consistent trend of COD concentration was observed during Run 1: COD slowly increased during the initial 25 days, as a result of the high COD contained in the fed MS-OFMSW (58 g/L); then, after an intermediate period of constant COD, this parameter decreased during the last two weeks of Run 1 (Fig. 2A). The anaerobic consortium adaptation was also demonstrated by changes of the concentration of some of the most prominently produced VFAs during the same period. In particular, acetic acid, the substrate for acetoclastic methanogens, accumulated during the first two weeks, then its concentration decreased, probably as the consequence of the activity of the acetoclastic methanogenic bacteria. As a matter of fact, PCRDGGE analysis of the archaeal community evidenced the occurrence of two acetoclastic methanogens of the Methanosarcina genus. However, no significant changes occurred in the composition of the archaeal community throughout Run 1 and Run 2 (Fig. 5B). Methanosarcina spp. were already found to be predominant among archaea in digesters which were operated under difficult or unstable start-up conditions (McMahon et al., 2004, 2001), whereas they were not detected in the AD of a readily fermentable substrate, such as source-sorted organic household waste (Cardinali-Rezende et al., 2009).
420
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 1 3 e4 2 4
On the contrary, remarkable changes in the composition of the bacterial community were observed during Run 1 (Fig. 5A), thus suggesting that the increased capability of the AD microbial community to biodegrade MS-OFMSW was mainly due to the enrichment of specific bacterial phylotypes rather than archaeal ones. This is also supported by the evidence that prominent bacterial phylotypes detected at the end of Run 1 are related to several Firmicutes (bands 8, 14, 16, 17, 18 and 20), to two Bacteroidetes (bands 4 and 9) and to a Synergistetes (band 19) with well known hydrolytic and fermentative activities leading to acetate as the main product (Fig. 5A, Table 3). Among these, Aminobacterium colombiense, originally isolated from the anaerobic lagoon of a dairy wastewater treatment plant, ferments amino acids to acetate and propionate (Baena et al., 1998); Alkaliflexus imshenetskii ferments carbohydrates to propionate (Zhilina et al., 2004); Sedimentibacter saalensis ferments amino acids to acetic and butyric acids (Breitenstein et al., 2002); C. clariflavum hydrolyzes cellulose and ferments cellobiose to acetate and lactate (Shiratori et al., 2009); Clostridium disporicum is a starch hydrolyzing bacterium that ferments sugars to acids (Horn, 1987); Moorella perchloratireducens ferments monosaccharides to acetate (Balk et al., 2008). The selection of an effective MS-OFMSW biomethanizating consortium continued throughout Run 2, during which the reactor was fed only with MS-OFMSW. Increasing COD removal rates and biometanization yields were observed during the 3 first weeks, up to the achievement of the steady state (Figs. 2B and 3B, respectively). The normalized methane production obtained in the batch MS-OFMSW-containing microcosms inoculated with the consortium obtained at the end of Run 2, equal to 0.39 LCH4 /ginitial COD, represents an interesting performance in comparison with the values typically obtained for easily fermentable matrixes and source sorted OFMSW (Dong et al., 2010; Bolzonella et al., 2006, 2003). The continuous-flow MS-OFMSW-fed process performed a little worse (with a steady-state methane production yield of 0.25 LCH4 /gCODremoved in Run 2), probably as a result of the lower retention time (22.4 days in Run 2 versus 85 days in the batch tests). The comparison of the results obtained in Runs 1 and 2 indicates that the two-month batch period occurred between the two runs led to an increase of the process performances (methane production rate þ79%, methane/COD yield þ67%; Table 2). Although higher relative amounts of methane were observed for Run 1, the higher methane productivities and the higher methane/COD yields of Run 2 indicate a higher enhancement of the metabolic activities of the hydrolytic and fermentative bacteria with respect to those of methanogenic populations. The observed improvement of process performance suggests that the intermediate batch operation induced a further acclimatization of the anaerobic consortium selected during Run 1. Actually, some of the phylotypes
Fig. 5 e PCR-DGGE analysis of the bacterial (A) and archaeal (B) community in the CM used as inoculum and in samples taken from the reactor during Run 1 (1d7, 1d28, 1d42, 1d63 and 1d84) and Run 2 (2d10 and 2d77). Arrows indicate bands that were excised from the gel and sequenced. Band numbers and positions are indicated on the right hand side of the picture.
Table 3 e Phylogenetic identification of Bacteria represented by the excised DGGE bands. Band # Phylogenetic Closest classified relative group (% certainty)
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Bacteroidetes Bacteroidetes Firmicutes Bacteroidetes Unclassified bacteria Firmicutes Bacteroidetes Firmicutes Bacteroidetes Bacteroidetes Bacteroidetes Firmicutes Firmicutes Firmicutes Firmicutes Firmicutes Firmicutes Firmicutes Synergistetes Firmicutes Firmicutes Bacteroidetes Actinobacteria
% identity
Closest described bacterium [accession #]
% identity
Paludibacter (97%) Alkaliflexus (100%) Solobacterium (27%) Alkaliflexus (100%)
Uncultured bacterium HAW-RM37-2-B-1600d-A11 [FN563309] Ruminofilibacter xylanolyticum S1 [DQ141183] Uncultured bacterium ATB-KS-1431 [EF686955] uncultured Rikenellaceae bacterium 139 [GQ468575] Uncultured bacterium 42B_BS1_4 [FJ825527]
99% 100% 99% 100% 96%
Paludibacter propionicigenes WB4 (T) [AB078842]
92%
Anaerorhabdus furcosa [HM038002] Alkaliflexus imshenetskii Z-7010 (T) [AJ784993] Syntrophothermus lipocalidus DSM 12680 [CP002048]
90% 95% 77%
Sedimentibacter (100%) Proteiniphilum (67%) Sedimentibacter (100%) Rikenella (30%) Alkaliflexus (22%) Rikenella (17%) Anaerotruncus (18%) Dehalobacter (24%) Clostridium (100%) Fastidiosipila (15%) Sporacetigenium (92%) Natranaerobius (16%) Fastidiosipila (43%) Aminobacterium (93%) Sporosarcina (84%) Anaerovorax (99%) Salinibacter (58%) Corynebacterium (100%)
Uncultured bacterium 190_BE1_11 [FJ825495] Uncultured compost bacterium 1B27 [DQ346459] Uncultured Sedimentibacter sp. SSCP 74-10DNT-1 [EF990173] Uncultured bacterium BS08 [EU358683] Bacterium enrichment culture clone BBMC-11 [GU476611] Uncultured bacterium 5A_BS1_1 [FJ825515] Uncultured Clostridiales bacterium 69 [GU112189] Uncultured Clostridia bacterium G3 [EU551098] Clostridium disporicum DSM 5521 (T) [Y18176] Uncultured bacterium G35_D8_L_B_F07 [EF559144] Uncultured bacterium E118 [AM500797] Uncultured bacterium MBA02 [AB114312] Uncultured bacterium A35_D28_L_B_B01 [EF559207] Uncultured bacterium ATB-KS-1011 [EF686933] Uncultured bacterium F3_105X [GQ263157] Uncultured bacterium 167_BE2_17 [FJ825481] Uncultured bacterium CSN-NBRI047 [EU442030] Corynebacterium vitaeruminis NCTC 20294 [X84680]
99% 100% 97% 100% 100% 99% 99% 97% 100% 100% 99% 99% 100% 99% 99% 98% 94% 98%
Sedimentibacter hydroxybenzoicus JW/Z-1 (T) [L11305] Proteiniphilum acetatigenes TB107 (T) [AY742226] Sedimentibacter saalensis ZF2 (T) [AJ404680] Alistipes massiliensis 3302398 [AY547271]
95% 94% 95% 86%
Alistipes massiliensis 3302398 [AY547271] Clostridium straminisolvens CSK1 [AB125279] Selenomonas ruminantium ES3 [AF221600]
87% 87% 86%
Clostridium clariflavum EBR45 (T) [AB186359] Clostridium maritimum G12 [EU089965] Moorella perchloratireducens An10 [EF060194] C. clariflavum EBR45 (T) [AB186359] Aminobacterium colombiense DSM 12261 [CP001997] Bacillus sp. CHNTR52 [DQ337594] Anaerovorax odorimutans NorPut (T) [AJ251215] Rubricoccus marinus SG-29 [AB545808]
87% 98% 90% 86% 92% 99% 95% 86%
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 1 3 e4 2 4
1 2 3 4 5
Closest match [accession #]
421
422
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 1 3 e4 2 4
Fig. 6 e Cluster analysis of DGGE banding patterns of the microbial communities in CM, Run 1 (1d7, 1d28, 1d42, 1d63 and 1d84) and Run 2 (2d10 and 2d77).
enriched at the end of Run 1 were poorly detected at the beginning of Run 2 (Fig. 5A). These changes might be due to the exposure of the microbial community to high COD concentrations during the batch run (initial COD of the 2nd batch ¼ 51.6 g/L), which might have limited the growth of some bacterial species adapted to lower COD concentrations. On the other hand, the transient exposure to high COD concentrations might also have favored the further acclimatization of the bacteria that were not inhibited, as reported for sequencing batch AD processes, which are characterized by fluctuating COD concentrations (Alkaya and Demirer, 2011). The observed changes of the microbial population were however limited, as evidenced by the clustering of the community enriched at the end of Run 1 (1d84 in Fig. 6) with that of the onset of Run 2 (2d10 in Fig. 6). Minor changes in the bacterial community were also evidenced during Run 2. This was probably due to the achievement of a steady state after the second week of treatment. The bacterial population prevailing in reactor throughout Run 2 was mainly composed by phylotypes related to A. imshenetskii, S. saalensis and M. perchloratireducens. The latter were also among the major phylotypes enriched at the end of Run 1, along with phylotypes related to Sedimentibacter hydroxybenzoicus, an amino acid to butyrate fermenting bacterium (Breitenstein et al., 2002), Proteiniphilum acetatigenes, a proteolytic and amino acids fermenting bacterium that produces mainly acetate and propionate (Chen and Dong, 2005) and a phylotype distantly related to Syntrophothermus lipocalidus, a syntrophic fatty-acid-oxidizing bacterium isolated from a wastewater treating upflow anaerobic sludge blanket reactor (Sekiguchi et al., 2000). The acclimatization of the anaerobic population to MSOFMSW was further confirmed in batch microcosms filled with MS-OFMSW and inoculated with samples of the reactor effluents collected in Run 2. Importantly, the inoculum collected at the end of the study allowed a normalized biomethane production from MS-OFMSW higher than that obtained with CM, and even higher than that reported for Run 2. This evidence, together with the lack of methane production in MS-OFMSW microcosms inoculated with the anaerobic effluent collected at the beginning of the first run, clearly demonstrated that an effective acclimation of the microbial community to the target waste was obtained (Fig. 4).
The rates and yields of methane production observed in the continuous-flow and in the batch tests, in agreement with the results of the microbiological characterization, indicate the crucial role played by the 7-day initial batch incubation with 100% CM. Indeed, no methane was produced in the MSOFMSW-containing batch microcosms, even when they were supplemented with a 13% inoculum consisting of undigested CM or of digested CM (sample taken from the bioreactor at the onset of Run 1), whereas the gradual replacement of digested CM with MS-OFMSW led in the bioreactor to an AD process characterized by satisfactory performances. In agreement with this result, the DGGE analysis indicate that both the archeal phylotypes detected in CM, as well as 3 bacterial phylotypes of CM, were found in the bioreactor at the end of Run 2 with a significant band intensity. The lack of a replicate continuous-flow bioreactor does not allow to state that the observed adaptation of the AD consortium to MS-OFMSW is a reproducible process. However, this aspect of the study does not affect the attainment of the main goal of the work, which was to develop a microbial consortium able to effectively carry out a continuous biomethanization process of MS-OFMSW.
5.
Conclusions
An anaerobic consortium capable of an effective biomethanization of an MS-OFMSW as the sole substrate was obtained by means of a continuous culture enrichment approach. The consortium was composed by a large variety of AD typical bacteria along with a few archea acetoclastic species. To the very best of our knowledge, this is the first biologicallycharacterized, efficiently MS-OFMSW biomethanizing microbial consortium described so far in the literature. The results of this research are of special interest in the perspective of developing intensified large-scale process for the industrial biomethanization of MS-OFMSW.
Acknowledgments Lorenzo Bertin particularly thanks Professor Michele Pinelli (Engineering Department, University of Ferrara, Italy) for his collaboration proposal related to the present research field. The authors sincerely thank Dr. Andrea Simoni and Professor Claudio Ciavatta (Department of Agroenvironmental Science and Technologies, University of Bologna) for conducting the heavy metal analyses and Filippo Mingozzi (Recupera S.r.l., Ostellato (FE), Italy) for providing the experimental MS-OFMSW.
references
Ako, O.Y., Kitamura, Y., Intabon, K., Satake, T., 2008. Steady state characteristics of acclimated hydrogenotrophic methanogens on inorganic substrate in continuous chemostat reactors. Bioresource Technology 99, 6305e6310.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 1 3 e4 2 4
Alkaya, E., Demirer, G.N., 2011. Anaerobic-Fed and sequencingbatch treatment of Sugar-Beet processing wastes: a Comparative study source. Water Environment Research 83 (9), 247e255. Baena, S., Fardeau, M.L., Labat, M., Ollivier, B., Thomas, P., Garcia, J.L., Patel, B.K.C., 1998. Aminobacterium colombiense gen. nov. sp. nov., an amino acid-degrading anaerobe isolated from anaerobic sludge. Anaerobe 4, 241e250. Balk, M., van Gelder, T., Weelink, S.A., Stams, A.J.M., 2008. (Per)chlorate reduction by the thermophilic bacterium Moorella perchloratireducens sp. nov., isolated from underground gas storage. Applied Environmental Microbiology 74, 403e409. Bertin, L., Todaro, D., Bettini, C., Fava, F., 2008. Anaerobic codigestion of the mechanically sorted organic fraction of a municipal solid waste with cattle manure in packed microcosms under batch conditions. Water Science and Technology 58 (9), 1735e1742. Bolzonella, D., Innocenti, L., Pavan, P., Traverso, P., Cecchi, F., 2003. Semi-dry thermophilic anaerobic digestion of the organic fraction of municipal solid waste: focusing on the start-up phase. Bioresource Technology 86 (2), 123e129. Bolzonella, D., Pavan, P., Mace, S., Cecchi, F., 2006. Dry anaerobic digestion of differently sorted organic municipal solid waste: a full-scale experience. Water Science and Technology 53 (8), 23e32. Breitenstein, A., Wiegel, J., Haertig, C., Weiss, N., Andreesen, J.R., Lechner, U., 2002. Reclassification of Clostridium hydroxybenzoicum as Sedimentibacter hydroxybenzoicus gen. nov., comb. nov., and description of Sedimentibacter saalensis sp. nov. International Journal of Systematic and Evolutionary Microbiology 52, 801e807. Bryer, J.D., 1985. Structured modelling of the anaerobic digestion of biomass particulates. Biotechnology and Bioengineering 27, 638e649. Cardinali-Rezende, J., Debarry, R.B., Colturato, L.F.D.B., Carneiro, E.V., Chartone-Souza, E., Nascimento, A.M.A., 2009. Molecular identification and dynamics of microbial communities in reactor treating organic household waste. Environmental Biotechnology 84, 777e789. Chen, S., Dong, X., 2005. Proteiniphilum acetatigenes gen. nov., sp. nov., from a UASB reactor treating brewery wastewater. International Journal of Systematic and Evolutionary Microbiology 55, 2257e2261. Chen, Y., Cheng, J.J., Creamer, K.S., 2008. Inhibition of anaerobic digestion process: a review. Bioresource Technology 99, 4044e4064. Colussi, I., Cortesi, A., Della Vedova, L., Gallo, V., Cano Robles, F.K., 2009. Start-up procedures and analysis of heavy metals inhibition on methanogenic activity in EGSB reactor. Bioresource Technology 100, 6290e6294. Dong, L., Zhenhong, Y., Yongming, S., 2010. Semi-dry mesophilic anaerobic digestion of water sorted organic fraction of municipal solid waste (WS-OFMSW). Bioresource Technology 101, 2722e2728. EC, 2008a. Directive 2008/98/EC of the European Parliament and of the Council of 19 November 2008 on waste and repealing certain Directives. Official Journal of the European Union L 312, 3e30. 22.11.2008. EC, 2008b. DG Environment of the European Commission. Analysis of the Evolution of Waste Reduction and the Scope of Waste Prevention. http://ec.europa.eu/environment/waste/ prevention/pdf/draft3_2.pdf Bruxelles. Eriksson, O., Reich, M.C., Frostell, B., Bjo¨rklund, A., Assefa, G., Sundqvist, J.O., Granath, J., Baky, A., Thyselius, L., 2005. Municipal solid waste management from a systems perspective. Journal of Cleaner Production 13 (3), 241e252. Goberna, M., Gadermaier, M., Garcı´a, C., Wett, B., Insam, H., 2010. Adaptation of methanogenic communities to the
423
cofermentation of cattle excreta and olive mill wastes at 37 C and 55 C. Applied Environmental Microbiology 19, 6564e6571. Hartmann, H., Ahring, B.K., 2005. Anaerobic digestion of the organic fraction of municipal solid waste: influence of codigestion with manure. Water Research 39 (8), 1543e1552. Hartmann, H., Ahring, B.K., 2006. Strategies for the anaerobic digestion of the organic fraction of municipal solid waste: an overview. Water Science and Technology 53 (8), 7e22. Hartmann, H., Moller, H.B., Ahring, B.K., 2004. Efficiency of the anaerobic treatment of the organic fraction of municipal solid waste: collection and pre-treatment. Waste Management and Research 22, 35e41. Horn, N., 1987. Clostridium disporicum sp. nov., a Saccharolytic species able to form two spores per cell, isolated from a rat cecum. International Journal of Systematic and Evolutionary Microbiology 37, 398e401. ISPRA, 2010. Rapporto Rifiuti Urbani e Edizione 2009. Rapporti e N. 108/2010. ISPRA e Istituto Superiore per la protezione e la ricerca ambientale, Roma, ISBN 978-88-448-0430-5. McMahon, K.D., Stroot, P.G., Mackie, R.I., Raskin, L., 2001. Anaerobic codigestion of municipal solid waste and biosolids under various mixing conditionsdII: microbial population dynamics. Water Research 35 (7), 1817e1827. McMahon, K.D., Zheng, D., Stams, A.J.M., Mackie, R.I., Raskin, L., 2004. Microbial population dynamics during startup and overload conditions of anaerobic digesters treating municipal solid waste and sewage sludge. Biotechnology and Bioengineering 87 (7), 823e834. Miller, D.G., Brayton, S.V., Boyles, W.T., 2001. Chemical oxygen demand analysis of wastewater using trivalent manganese oxidant with chloride removal by sodium bismuthate pretreatment. Water Environment Research 73 (1), 63e71. Raskin, L., Stromley, J.M., Rittmann, B.E., Stahl, D.A., 1994. Groupspecific 16S rRNA hybridization probes to describe natural communities of methanogens. Applied Environmental Microbiology 60, 1232e1240. Sass, A.M., Sass, H., Coolen, M.J.L., Cypionka, H., Overmann, J., 2001. Microbial communities in the chemocline of a hypersaline deep-sea basin (Urania basin, Mediterranean Sea). Applied Environmental Microbiology 67 (12), 5392e5402. Sekiguchi, Y., Kamagata, Y., Nakamura, K., Ohashi, A., Harada, H., 2000. Syntrophothermus lipocalidus gen. nov., sp. nov., a novel thermophilic, syntrophic, fatty-acid-oxidizing anaerobe which utilizes isobutyrate. International Journal of Systematic and Evolutionary Microbiology 50, 771e779. Shiratori, H., Sasaya, K., Ohiwa, H., Ikeno, H., Ayame, S., Kataoka, N., Miya, A., Beppu, T., Ueda, K., 2009. Clostridium clariflavum sp. nov. and Clostridium caenicola sp. nov., moderately thermophilic, cellulose-/cellobiose-digesting bacteria isolated from methanogenic sludge. International Journal of Systematic and Evolutionary Microbiology 59, 1764e1770. Skovgaard, M., Hedal, N., Villanueva, A., Andersen, F.M., Larsen, H., 2008. Municipal Waste Management and Greenhouse Gases. ETC/RWM Working Paper 2008/1. European Environment Agency, Bruxelles. Sarioglu, M., Akkoyun, S., Bisgin, T., 2010. Inhibition effects of heavy metals (copper, nickel, zinc, lead) on anaerobic sludge. Desalination and Water Treatment 23, 55e60. Stahl, D.A., Amann, R.I., 1991. Development and application of nucleic acid probes. In: Stackebrandt, E., Goodfellow, M. (Eds.), Nucleic Acid Techniques in Bacteria Systematics. Wiley, New York, pp. 205e248. Standard Methods for the Examination of Water and Wastewater, 1998. twentieth ed. American Public Health Association/ American Water Works Association/Water Environment Federation, Washington DC. Stroot, P.G., McMahon, K.D., Mackie, R.I., Raskin, L., 2001. Anaerobic codigestion of municipal solid waste and biosolids
424
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 1 3 e4 2 4
under various mixing conditions - I. Digester performance. Water Research 35 (7), 1804e1816. Wiesmann, U., Choi, I.S., Dombrowski, E.M., 2007. Fundamentals of Biological Wastewater Treatment, first ed. Wiley, New York, pp. 169e194.
Zhilina, T.N., Appel, R., Probian, C., Llobet Brossa, E., Harder, J., Widdel, F., Zavarzin, G.A., 2004. Alkaliflexus imshenetskii gen. nov. sp. nov., a new alkaliphilic gliding carbohydratefermenting bacterium with propionate formation from a soda lake. Archives of Microbiology 182, 244e253.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 2 5 e4 3 2
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Improvement of activated sludge dewaterability by mild thermal treatment in CaCl2 solution Baohong Guan*, Jie Yu, Hailu Fu, Minhui Guo, Xinhua Xu Department of Environmental Engineering, Zhejiang University, Hangzhou 310058, China
article info
abstract
Article history:
Activated sludge dewatering is of great importance in sludge treatment and disposal. To
Received 9 September 2011
enhance the dewaterability, a novel method was performed by treating the sludge under
Received in revised form
mild temperature (50e90 C) in CaCl2 solution (3.7e1110.0 mg/g dry sludge). The capillary
2 November 2011
suction time, zeta potential, Fourier-transformed infrared spectra, concentration of soluble
Accepted 3 November 2011
protein and carbohydrates were employed to characterize the dewaterability and influ-
Available online 13 November 2011
encing mechanism. The sludge dewaterability was deteriorated with single thermal treatment, but significantly promoted in CaCl2 solution and advanced further together with
Keywords:
thermal treatment. An increasing CaCl2 dosage reduced the surface charge remarkably,
Activated sludge
and a higher temperature could strengthen this impact. The spectra indicate that Ca2þ
Dewaterability
could interact with the protein, phenols and OeH functional group in the flocs. The
Mild thermal treatment
thermal treatment could cause the solubilization of protein and carbohydrates, providing
CaCl2
more binding sites for Ca2þ to establish a strong bridging among the flocs. As CaCl2 dosage elevated, the soluble carbohydrates showed a reduction trend, while the soluble protein lowered firstly and then bounced back except that remained unchanged at room temperature. A bridging equilibrium is presumed to exist between Ca2þ and the soluble protein. And the bridging between Ca2þ and the soluble carbohydrates plays a more important role in the dewatering. The sludge dewaterability was successfully and economically improved by thermal treatment in CaCl2 solution. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The activated sludge process is widely used for wastewater treatment. Large amounts of sludge are generated, which poses a threat to the environment. Dewatering is essential to reduce the volume of the sludge, and still a bottleneck for the sludge treatment. Many kinds of technologies have been developed to enhance the sludge dewaterability, including thermal treatment under neutral, acid or alkaline conditions (Neyens et al., 2004), freezing and thawing (Diak et al., 2011), sonication (Huan et al., 2009), ozonation (Ahn et al., 2002) and biological hydrolysis (Lu et al., 2011). Thermal treatment is one of the
most simple, efficient and cost-effective technologies for sludge dewatering. It causes the solubilization of flocs, reducing EPS water retention properties and thus leading to the release of bound water (Bougrier et al., 2008). Temperature, the most important parameter of thermal treatment, ranges widely from 60 to 275 C (Bougrier et al., 2007, 2008; Prorot et al., 2011). Bougrier et al. (2008) brought forward a threshold temperature of 150 C for sludge dewatering. The dewaterability could be significantly enhanced above the temperature and deteriorated below it. Thermochemical treatment provides an alternative to improve the sludge dewaterability and gives an advantage of lowering the temperature. Neyens et al. (2003b) treated the
* Corresponding author. Tel.: þ86 571 88982026; fax: þ86 571 88273687. E-mail address:
[email protected] (B. Guan). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.014
426
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 2 5 e4 3 2
sludge with H2SO4 at temperatures between 120 and 155 C, resulting in a significant reduction in the amount of dry sludge (DS) and an increase in the DS content of the dewatered cake. The dewaterability could also be promoted remarkably by Ca(OH)2 under 100 C for 60 min (Neyens et al., 2003a). Metal salts play an important role in sludge flocculation, settling and dewatering, as the cations could change sludge properties, for instance, viscosity, surface charge (Pevere et al., 2007), capillary suction time (CST) and floc strength (Sobeck and Higgins, 2002). Lots of researchers have investigated the effect of cations on sludge dewatering and presented several theories to explain the mechanism. The Derjaguin-Landau-Verwey-Overbeek (DLVO) theory, a classical colloidal theory, is used frequently to set forth the cations effect of compressing the double layer of sludge colloids (Cousin and Ganczarczyk, 1998; Pevere et al., 2007). The Divalent Cation Bridging (DCB) theory provides a further insight into the function of cations (Tezuka, 1969). It describes that divalent cations could bridge negatively charged surfaces of adjacent cells promoting floc formation (Nguyen et al., 2007), while monovalent cations may cause a deterioration in floc properties by displacing divalent cations from within the flocs (Novak et al., 1998). The Alginate theory, a subset of the DCB theory, focuses especially on the interaction of Ca2þ and alginate (Bruus et al., 1992). Sobeck and Higgins (2002) tested the effect of cations on sludge biofocculation and suggested the DCB theory could best interpret the cations role among the three theories mentioned above. In addition, Hþ Translocation-Dehydration theory pays attention to Ca2þ effect of inducing the dehydration of bacterial surfaces (Teo et al., 2000). For the thermal treatment, there is still a room to lower the temperature and meanwhile improve the dewatering efficiency. However little research has attempted metal salts in this process for sludge dewatering. Among the common cations (Naþ, Kþ, Mg2þ, Ca2þ et al.), Ca2þ shows a more marked effect in improving the sludge dewaterability (Bruus et al., 1992; Novak et al., 1998). So the study here employs the mild thermal treatment cooperated with CaCl2 to change the sludge dewaterability. The purpose is to explore the feasibility of this treatment and investigate the possible influencing mechanism.
2.
Materials and methods
2.1.
Activated sludge
Activated sludge was obtained from a thickening tank of the wastewater treatment plant in Xihu Beer Chaori Limited Company, Hangzhou, China. The sludge was subjected to a serial of tests according to CJ/T 221-2005 (China standard for municipal sludge analysis). It contains about 87.4% water with a pH of 7.20 0.17 and mixed liquor suspended solids of 123.2 42.4 g/L. Table 1 presents the main characteristics of the sludge.
2.2.
Experimental procedures
The sludge was pretreated by rinsing twice with deionized water and sieving through a 24 mesh screen to get rid of
Table 1 e Main characteristics of the activated sludge at 25 C. Parameter Water content (% (w/w)) pH Soluble chemical oxygen demand (mg/L) Soluble protein (mg/L) Soluble carbohydrates (mg/L) Total carbohydrates (mg/g DS) Mixed liquor suspended solids (g/L) Mixed liquor volatile suspended solids (g/L)
Value 87.4 7.20 108.8 5.13 20.2 10.9 123.2 35.2
4.3 0.17 18.3 2.87 6.3 8.0 42.4 1.9
suspended residues and solid impurities. The screened sludge was filtered with a vacuum pump (0.1 MPa) and immediately stored in a refrigerator at 4 C. The experiments were proceeded in a 1.9 L double-jacketed glass reactor equipped with a thermometer, electric stirrer and glass condenser. The sludge was maintained at an expected temperature with a deviation of 1.0 C by circulating hot oil through the double walls and monitored by the thermometer. In each run, the pretreated sludge was diluted with deionized water to obtain 3.0% (w/w) DS content. First, a certain amount of deionized water was preheated to a certain temperature in the reactor along with CaCl2 at dosage of 3.7e1110.0 mg/g DS. Then the sludge was added and the temperature was raised quickly to the expected one (50, 60, 70, 80 and 90 C). The mixture was approximately 1.2 L and was homogenized by stirring at 260 rpm in the first 1 min and then at 160 rpm. Samples were withdrawn and immediately cooled to 25 C within 10 min before test. The control test was operated at 25 C with no CaCl2 addition.
2.3.
Analytical methods
The CST was measured using a 304M CST equipment (Triton, UK) by adding about 5.0 mL sample into the tube. The micrographs of flocs were taken under a B series biological microscope (Chongqing Optec Instrument Co., Ltd, China) at a magnification of 10 10. The sludge with CaCl2 was naturally precipitated for 5 min and the supernatant was collected for zeta potential measurement (Li and Yang, 2007). The sludge without CaCl2, which always possessed poor settling characteristic, was diluted 20 times with the supernatant obtained by centrifuging the sample at 5000 rpm for 5 min. The measurement was conducted at a Malvern Zetasizer Nano ZS (Malvern Instruments Ltd., UK) after adjusting sample to pH 7.0 0.1 by NaOH or HCl. A sample of 40 mL sludge was filtrated to remove most of water. The filter cake was dried at 105 C for 24 h, then ground into powder, and dried again under the same conditions. The Fourier-transformed infrared (FTeIR) spectra were recorded using an ALPHA FTeIR spectrometer (Bruker Optics, Germany) in the range of 4000e400 cm-1. Soluble fraction is the filtrate passing 0.45 mm filter membrane. Soluble protein was determined using the corrected Lowry method (Frølund et al., 1995; Raunkjær et al., 1994) with bovine serum albumin as the standard. As Ca2þ
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 2 5 e4 3 2
427
could react with the reagent Na2CO3 and form white sediment, the filtrate containing Ca2þ was homogenized with 6.5 g/L NaCO3 solution for 10 min and then filtered to remove Ca2þ. The absorbance of sample was determined at 750 nm using an ET99730 SpectroDirect (Tintometer GmbH, Germany). Soluble carbohydrates were measured by the anthrone method at 625 nm absorbance using glucose as the standard (Dreywood, 1946).
3.
Results and discussion
3.1.
Sludge dewaterability
The sludge dewaterability measured as CST presented a notable difference under various treatment conditions (see Fig. 1). The CST varied a little (from 227.3 to 291.9 s) within 120 min in the control test. However, it increased apparently and reached 1.8 and 3.0 times of the control test value as the temperature elevated to 60 and 80 C, respectively, indicating a significant deterioration in the dewaterability. This degeneration is possibly resulted from the breakdown of sludge and the rise of smaller flocs that are hard to be dewatered (Bougrier et al., 2008; Neyens et al., 2004). The addition of 37.0 mg/g DS CaCl2 to the sludge resulted in a fleetly decrease in the CST from 239.5 to 61.3, 33.3 and 20.8 s at 25, 60 and 80 C, respectively, within the first 10 min. Then a stable plateau presented with no more than 5.8 s variation over the following 110 min. That is, the sludge dewaterability is improved remarkably by CaCl2 and the change seems to occur within a relatively short time. A bit of improvement could be achieved by raising the temperature from 25 to 60 and 80 C. The CST decreased sharply with increasing CaCl2 dosage, and then reached a relative stabilization, exhibiting a parallel trend at various temperatures as presented in Fig. 2. Since the CST variation was no more than 9.8 s when the CaCl2 dosage exceeded 37.0 mg/g DS, the optimal CaCl2 dosage is supposed
Fig. 1 e CST evolution of sludge with time under mild thermal treatment with or without CaCl2.
Fig. 2 e CST evolution of sludge as a function of temperature and CaCl2 dosage.
to be 37.0 mg/g DS. Overdosed CaCl2 brought little benefit to the dewatering. Higher temperature led to a better dewaterability at a certain CaCl2 dosage (Fig. 2). Especially when the temperature elevated from 60 to 70 C, the CST dropped 48.3% at 185.0 mg/g DS CaCl2 dosage. Further increase in temperature only resulted in slight decrease in the CST. Thus, 70 C is considered suitable from a view of efficiency and economy. Compared to the threshold temperature of 150 C in thermal treatment (Bougrier et al., 2008), the mild thermal treatment cooperated with CaCl2 provides an alternative approach with a much lower temperature to enhance the dewaterability.
3.2.
Surface charge
According to the DLVO theory, the colloidal stability, which plays an important role in sludge dewatering (Mikkelsen et al., 1996), has a strong relation with the surface charge. The evolution of surface charge in terms of zeta potential is shown in Fig. 3. The sludge was originally negatively charged with 21.00 mV zeta potential at 25 C. The zeta potential of sludge without CaCl2 decreased from 19.40 to 23.90 mV as the temperature increased from 50 to 90 C. As plenty of charges in the floc interior are inaccessible, the disintegration of flocs by thermal treatment could cause an exposure of more surfaces and then a rise in negative charges (Mikkelsen, 2003). Interestingly, the zeta potential showed no obvious shift with increasing temperature at 18.5 mg/g DS CaCl2, while a rise trend was observed at 185.0 mg/g DS CaCl2. The adoption of CaCl2 could alter the zeta potential evolution via neutralizing the negative surface charges, and higher temperature could strengthen the impact. Increasing CaCl2 dosage induced an evident elevation in the zeta potential which can be seen in Figs. 3 and 4. The zeta potential increased sharply from 26.20 to 7.32 mV at 25 C, from 26.40 to 7.25 mV at 60 C and from 28.90 to 6.85 mV at 80 C as the CaCl2 dosage elevated from 0 to 185.0 mg/g DS, then rose gently and tended to an isoelectric point with increasing CaCl2 dosage (Fig. 4). Since the optimal dewaterability did not occurred at the isoelectric point, the DLVO
428
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 2 5 e4 3 2
Fig. 5 e FTeIR spectra of the sludge with and without CaCl2 at 25 C. CaCl2 dosage: (1), 0 mg/g DS; (2), 37.0 mg/g DS; (3), 370.0 mg/g DS; (4), 1110.0 mg/g DS. Fig. 3 e Zeta potential evolution of sludge as a function of temperature treated with or without CaCl2 for 20 min. Hollow square: 0 mg/g DS CaCl2; solid circle: 18.5 mg/g DS CaCl2; hollow triangle: 185.0 mg/g DS CaCl2.
theory can not give a full explanation of the interaction between Ca2þ and the flocs (Pevere et al., 2007).
3.3.
FTeIR analysis
To further determine the interaction between Ca2þ and the flocs, the FTeIR spectra of the sludge with different CaCl2 dosages are presented in Fig. 5. The strong band at 3402 cm1 corresponds to the OeH stretching vibration. The bands at 2926 and 2854 cm-1 are asymmetric and symmetric vibration of CH2 of aliphatic structures and lipids, respectively (Amir et al., 2005). The band at 1739 cm1 can be attributed to the C]O vibration of carboxylic acids. The typical bands of protein have been demonstrated: a C]O and CeN stretching band at 1642 cm1, CeN stretching vibration and NeH deformation vibration at 1563 and 1552 cm1. The bands at 1443 and 1423 cm1 are phenolic OeH plane variable-angle vibration
and CeO stretching vibration. The band at 1377 cm1 is assigned to the OeH vibration of phenols and CeO vibration of carboxylates. The intense band at 1091 cm1 is linked to the CeOeC and CeO vibration of polysaccharides. The functional groups are similar to those reported by Gulnaz et al. (2006) and Laurent et al. (2009a). The details of the shifts of characteristic bands are listed in Table 2. There was no obvious transformation of bands with 37.0 mg/g DS CaCl2 addition, indicating no strong interaction between CaCl2 and the functional groups. As CaCl2 dosage increased to 370.0 and 1110.0 mg/g DS, the bands of OeH stretching vibration and OeH plane variable-angle vibration of phenols shifted right, indicating the replacement of Hþ by Ca2þ (Chen et al., 2002). Moreover, Ca2þ reacted with the functional groups of protein, leading to the transformation of CeN and NeH vibration bands. The spectra indicate that Ca2þ interacted with protein, phenols and OeH functional group in the flocs. This interaction could precisely diminish the surface charge of flocs, as presented in Fig. 4.
3.4.
Soluble protein and carbohydrates
Protein and carbohydrates, the main components of EPS, play important role in the dewatering (Bruus et al., 1992; Higgins and Novak, 1997; Neyens et al., 2004). To get an insight into the dewaterability, the impact of temperature on the soluble
Table 2 e Main shifts of infrared absorption bands of sludge at different CaCl2 dosages. Band 1
Fig. 4 e Zeta potential evolution of sludge as a function of CaCl2 dosage under mild thermal treatment at 25, 60 and 80 C for 20 min.
Assignment 2
3
4
3417.87 3420.70 3396.61 3402.28 OeH stretching vibration 1558.73 1558.73 1562.98 1562.98 CeN stretching vibration 1540.31 1540.31 1553.06 1551.64 and NeH deformation vibration of protein 1456.00 1456.70 1459.54 1442.53 OeH plane variable-angle vibration and CeO stretching vibration of phenols Sludge treated with 0 (1), 37.0 (2), 370.0 (3), and 1110.0 mg/g DS CaCl2 (4) at 25 C.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 2 5 e4 3 2
Fig. 6 e Soluble protein of sludge as a function of temperature treated with or without CaCl2 for 20 min.
protein is given in Fig. 6. As the temperature increased, more protein was released. For example, about eight times of soluble protein was found in the sludge by raising the temperature from 50 to 90 C. Since the protein is mainly distributed in the pellet (Shao et al., 2009), besides the flocs breakage, the considerable solubilization of protein is probably resulted from the cell disruption. The rise in soluble protein could contribute to the increasing surface charge under thermal treatment (Fig. 3) (McKinney, 1952). The thermal treatment had been demonstrated to cause the flocs disruption and cell lysis, releasing biopolymers and some inorganic species (Laurent et al., 2009b; Prorot et al., 2011). The rise in biopolymers at mild temperature is often related with poorer dewaterability (Bougrier et al., 2008). But in the presence of CaCl2, to the opposite, better dewaterability was observed, which can be attributed to the emergence of more available functional groups for Ca2þ. The release of protein-like biopolymers transfers the functional groups from the particulate fraction to the soluble phase, providing more available binding sites for Ca2þ, for instance, OeH and
429
functional groups of protein and phenols as proved by FTeIR spectra, and strengthening the bridging between Ca2þ and the flocs. This also well interprets the zeta potential evolution with rising temperature in Fig. 3, as more released negative functional groups are bound by Ca2þ at higher temperature. Accordingly, the variation of flocs in CaCl2 solution with or without thermal treatment is schematically presented in Fig. 7. When CaCl2 is added at 25 C, the negatively charged flocs are bridged by Ca2þ (path a). However, the floc matrix breaks up initially when the sludge is thermally treated in CaCl2 solution (path b1). Then Ca2þ interacts with the exposed binding sites and forms a more compact and powerful matrix (path b2). Part of water which is captured in the floc matrix and cells could also become free during path b. The dewaterability of sludge in CaCl2 solution is promoted with increasing temperature consequently. The micrographs of sludge in Fig. 7 illustrate the variation of flocs by treatment. The raw sludge was relatively dispersed, without any aggregates. After the addition of 370.0 mg/g DS CaCl2, some aggregates were formed while plenty of dispersed flocs distributed among them. Once the mild thermal treatment was carried out at 80 C with 370.0 mg/g DS CaCl2, the sludge presented a strong aggregation and the dispersed flocs almost disappeared. The influences of CaCl2 dosage on the solubilization of protein and carbohydrates are shown in Fig. 8. As seen in Fig. 8a, the soluble protein concentration varied little with increasing CaCl2 dosage at 25 C. Nevertheless, at 60 and 80 C, the soluble protein firstly decreased, then dramatically increased and exceeded the original value as CaCl2 dosage increased. The minimal soluble protein was found at 18.5 mg/g DS CaCl2 for 60 C and 37.0 mg/g DS CaCl2 for 80 C. Differently, the soluble carbohydrates followed a reduction trend with increasing CaCl2 dosage (Fig. 8b). Especially when the CaCl2 dosage elevated from 0 to 18.5 mg/g DS, a sharp decrease was observed. Nguyen et al. (2008) also reported a decrease in the soluble polysaccharide with an increase in the concentration of Ca2þ. However, in the work of Higgins and Novak (1997), the bound extracellular protein elevated as the divalent cations increased while the bound extracellular polysaccharide varied
Fig. 7 e Schematic diagram of dispersion and compaction of flocs with Ca2D (a) at 25 C and (b) under mild thermal : Ca2D. treatment, accompanied with micrographs (10 3 10) of sludge. b1: Thermal treatment effect; b2:CaCl2 effect;
430
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 2 5 e4 3 2
Fig. 8 e Concentration of (a) soluble protein and (b) soluble carbohydrates of sludge as a function of CaCl2 dosage under mild thermal treatment at 25, 60 and 80 C for 20 min.
little, which differs from the results obtained from Fig. 8. Higgins and Novak (1997) employed laboratory continuousflow reactors where new biopolymers and flocs could form and interact with divalent cation according to Sobeck and Higgins (2002). The binding sites of polysaccharide are supposed to be saturated at lower divalent cation concentration so that no shift of the bound polysaccharide was observed. This could attribute to the affinity of different polysaccharides toward divalent cation. Certain species of bacteria could synthesize some protein and polysaccharide which have marked affinity for the cation (Dugan, 1970; Higgins and Novak, 1997). Thus, the effect of Ca2þ on the soluble protein and carbohydrates concentration differs when the bacteria species varies. Calcium could bridge with the functional groups of soluble protein and carbohydrates, as previously described, to lower their concentration in the solution (Bruus et al., 1992; McKinney, 1952). But when the CaCl2 dosage exceeded a certain point, the protein released again (Fig. 8a). Then we can assume a bridging equilibrium point which corresponds to the minimal soluble protein concentration. When the point is overruned, the protein bound with carbohydrates and other biopolymers tends to be exchanged by Ca2þ and thus releases into the supernatant (Higgins and Novak, 1997). The protein solubilization may do harm to the sludge dewatering. Considered the reduction trend of soluble carbohydrates (Fig. 8b) and CST (Fig. 2) with elevating CaCl2 dosage, the bridging between Ca2þ and the soluble carbohydrates is supposed to offset the inverse effect caused by protein solubilization. Once the inverse effect can not be counteracted, the dewaterability degeneration will occur. Based on this, the bridging between Ca2þ and the soluble carbohydrates is suggested to play a more important role in the sludge dewatering in CaCl2 solution with thermal treatment. Correlations between the CST and soluble protein as well as soluble carbohydrates are established in Fig. 9. Rough linear dependencies of soluble protein and carbohydrates on the CST were found, and low CST values corresponded to high soluble protein and carbohydrates concentrations. This result differs from that of single thermal treatment (Bougrier et al., 2008).
The interaction between Ca2þ and soluble biopolymers as indicated in Fig. 5 and the forming of new powerful matrix could account for this. Since the slope of solid line is lower than that of dash line, the variation of soluble carbohydrates concentration is considered to have a more remarkable influence on the CST, which again identifies the importance of soluble carbohydrates in the dewatering. As the main factor that leads to the protein and carbohydrates solubilization, temperature also has a significant effect on the CST during the process.
3.5.
Economic evaluation
Economy is a crucial factor to be considered when a method for sludge dewatering is under developing. An economic evaluation is conducted by calculating the energy and chemical consumptions and compared to those of thermal treatment under neutral, acid or alkaline conditions (Bougrier et al., 2008; Neyens et al., 2003a, b). Because the energy and chemical consumptions vary with the sludge source and treatment process, it is hard to make an exact comparison.
Fig. 9 e Variations of soluble protein and carbohydrates with CST under mild thermal treatment (50e90 C) with 18.5e1110.0 mg/g DS CaCl2.
431
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 2 5 e4 3 2
Table 3 e Economic evaluations of thermal treatment processes under neutral, acid, alkaline conditions or with the addition of CaCl2. Treatment conditions 190 C (Bougrier et al., 2008) 140 C, pH 2 (Neyens et al., 2003b) 100 C, pH 10 (Neyens et al., 2003a) 70 C, 37.0 mg/g DS CaCl2 (this research)
Initial DS content (%)
Qwater (kJ/kg)
Qsolids (kJ/kg)
Qsludge (kJ/kg)
Chemical cost (USD/t DS)
CST variation (s)
3.0 6.0 6.0 3.0
721.9 460.5 298.1 183.2
7.2 9.7 6.3 1.9
729.1 470.2 304.4 185.1
2.5 6.8 6.1
1300.031.0 33.128.9 34.022.0 239.522.3
The calculation here gives a primarily reliable estimate based on the heating temperature and chemical consumption. The chemical cost is determined on the chemicals’ average market price. The energy consumption is calculated according to Eqs. (1e3) (Kim and Parker, 2008). Qwater ¼ W Cpwater DT
(1)
Qsolids ¼(1-W ) Cpsolids DT
(2)
Qsludge ¼ Qwater þ Qsolids
(3)
where, Qwater, Qsolids and Qsludge are the energy consumptions for heating water, solids and sludge respectively. W is the water content of sludge, Cpwater and Cpsolids are the heat capacities of water and solids respectively (kJ/kg C), DT is the difference between control and heating temperature ( C). The heat capacities of water under different pressures are obtained from Perry’s Chemical Engineers’ Handbook (Perry and Green, 1997). A constant value of 1.41 kJ/kg C, calculated from the organic fraction proportion of the dried sludge (see Table 1) as well as the heat capacities of organic and inorganic matter, is used as the heat capacity of solids in the sludge (Kim and Parker, 2008). The calculation results are shown in Table 3. The energy consumptions are 729.1, 470.2, 304.4 and 185.1 kJ/kg for thermal treatment processes under neutral, acid, alkaline conditions and with the addition of CaCl2, respectively. If natural gas is taken as heating source, its industrial price is 5.40 USD/thousand cu ft, and heating value is about 1086.65 kJ/ cu ft. If the thermal efficiency of heating vessel is 90%, the fuel costs for heating sludge will be 4.0 103, 2.6 103, 1.7 103 and 1.0 103 USD/kg correspondingly. The method of mild thermal treatment with CaCl2 has a pretty lower energy requirement and fuel cost than other treatment processes. Moreover, compared with the fuel cost, the chemical cost is minor. So the method is much effective and economical to enhance the sludge dewaterability.
4.
Conclusions
The major conclusions are drawn as follows: 1. It is an efficient and energy-saving way to improve the activated sludge dewaterability under mild thermal treatment (50e90 C) in CaCl2 solution.
2. The mild thermal treatment enhanced the solubilization of protein and carbohydrates, resulting in the exposure of negatively charged surfaces and the deterioration of dewaterability. Calcium chloride could interact with the protein, phenols and OeH functional group, neutralizing the surface charge of flocs markedly. The emergence of more available binding sites due to thermal treatment could thus strengthen the bridging between Ca2þ and the flocs, promoting the dewaterability of sludge in CaCl2 solution. 3. Increasing CaCl2 dosage presented a prohibition at first and subsequently a promotion on the protein solubilization but a restraint on the carbohydrates solubilization. It is assumed that a bridging equilibrium exists between Ca2þ and the soluble protein. The bridging between Ca2þ and the soluble carbohydrates plays a more significant role in the sludge dewatering.
Acknowledgments The authors gratefully appreciate the Project 20100933B17 supported by the Sci & Tech Development Program of Hangzhou Government, China.
references
Ahn, K.H., Park, K.Y., Maeng, S.K., Hwang, J.H., Lee, J.W., Song, K.G., Choi, S., 2002. Ozonation of wastewater sludge for reduction and recycling. Water Science and Technology 46 (10), 71e77. Amir, S., Hafidi, M., Merlina, G., Revel, J., 2005. Sequential extraction of heavy metals during composting of sewage sludge. Chemosphere 59 (6), 801e810. Bougrier, C., Delgene`s, J.P., Carre`re, H., 2007. Impacts of thermal pre-treatments on the semi-continuous anaerobic digestion of waste activated sludge. Biochemical Engineering Journal 34 (1), 20e27. Bougrier, C., Delgene`s, J.P., Carre`re, H., 2008. Effects of thermal treatments on five different waste activated sludge samples solubilisation, physical properties and anaerobic digestion. Chemical Engineering Journal 139 (2), 236e244. Bruus, J.H., Nielsen, P.H., Keiding, K., 1992. On the stability of activated sludge flocs with implications to dewatering. Water Research 26 (12), 1597e1604. Chen, J.P., Lie, D., Wang, L., Wu, S., Zhang, B., 2002. Dried waste activated sludge as biosorbents for metal removal: adsorptive characterization and prevention of organic leaching. Journal of Chemical Technology and Biotechnology 77 (6), 657e662.
432
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 2 5 e4 3 2
Cousin, C.P., Ganczarczyk, J.J., 1998. Effects of salinity on physical characteristics of activated sludge flocs. Water Quality Research Journal of Canada 33 (4), 565e587. Diak, J., Ormeci, B., Proux, C., 2011. Freeze-thaw treatment of RBC sludge from a remote mining exploration facility in subarctic Canada. Water Science and Technology 63 (6), 1309e1313. Dreywood, R., 1946. Qualitative test for carbohydrate material. Industrial and Engineering Chemistry Analytical Edition 18 (8), 499. Dugan, P.R., 1970. Removal of mine water ions by microbial polymers. Proceeding Third Symposium on Coal Mine Drainage Research, 279e283. Frølund, B., Griebe, T., Nielsen, P.H., 1995. Enzymatic activity in the activated-sludge floc matrix. Applied Microbiology and Biotechnology 43 (4), 755e761. Gulnaz, O., Kaya, A., Dincer, S., 2006. The reuse of dried activated sludge for adsorption of reactive dye. Journal of Hazardous Materials 134 (1e3), 190e196. Higgins, M.J., Novak, J.T., 1997. Characterization of exocellular protein and its role in bioflocculation. Journal of Environmental Engineering 123 (5), 479e485. Huan, L., Yiying, J., Mahar, R.B., Zhiyu, W., Yongfeng, N., 2009. Effects of ultrasonic disintegration on sludge microbial activity and dewaterability. Journal of Hazardous Materials 161 (2e3), 1421e1426. Kim, Y., Parker, W., 2008. A technical and economic evaluation of the pyrolysis of sewage sludge for the production of bio-oil. Bioresource Technology 99 (5), 1409e1416. Laurent, J., Casellas, M., Pons, M.N., Dagot, C., 2009a. Flocs surface functionality assessment of sonicated activated sludge in relation with physico-chemical properties. Ultrasonics Sonochemistry 16 (4), 488e494. Laurent, J., Pierra, M., Casellas, M., Dagot, C., 2009b. Fate of cadmium in activated sludge after changing its physicochemical properties by thermal treatment. Chemosphere 77 (6), 771e777. Li, X.Y., Yang, S.F., 2007. Influence of loosely bound extracellular polymeric substances (EPS) on the flocculation, sedimentation and dewaterability of activated sludge. Water Research 41 (5), 1022e1030. Lu, J., Rao, S., Le, T., Mora, S., Banerjee, S., 2011. Increasing cake solids of cellulosic sludge through enzyme-assisted dewatering. Process Biochemistry 46 (1), 353e357. McKinney, R.E., 1952. A fundamental approach to the activated sludge process: II. a proposed theory of floc. Water Environment Federation 24 (3), 280e287. Mikkelsen, L.H., 2003. Applications and limitations of the colloid titration method for measuring activated sludge surface charges. Water Research 37 (10), 2458e2466. Mikkelsen, L.H., Gotfredsen, A.K., Agerbæk, M.L., Nielsen, P.H., Keiding, K., 1996. Effects of colloidal stability on clarification
and dewatering of activated sludge. Water Science and Technology 34 (3e4), 449e457. Neyens, E., Baeyens, J., Creemers, C., 2003a. Alkaline thermal sludge hydrolysis. Journal of Hazardous Materials 97 (1e3), 295e314. Neyens, E., Baeyens, J., Weemaes, M., De heyder, B., 2003b. Hot acid hydrolysis as a potential treatment of thickened sewage sludge. Journal of Hazardous Materials 98 (1e3), 275e293. Neyens, E., Baeyens, J., Dewil, R., De heyder, B., 2004. Advanced sludge treatment affects extracellular polymeric substances to improve activated sludge dewatering. Journal of Hazardous Materials 106 (2e3), 83e92. Nguyen, T.P., Hankins, N.P., Hilal, N., 2007. A comparative study of the flocculation behaviour and final properties of synthetic and activated sludge in wastewater treatment. Desalination 204 (1e3), 277e295. Nguyen, T.P., Hilal, N., Hankins, N.P., Novak, J.T., 2008. Determination of the effect of cations and cationic polyelectrolytes on the characteristics and final properties of synthetic and activated sludge. Desalination 222 (1e3), 307e317. Novak, J.T., Love, N.G., Smith, M.L., Wheeler, E.R., 1998. The effect of cationic salt addition on the settling and dewatering properties of an industrial activated sludge. Water Environment Research 70 (5), 984e996. Perry, R.H., Green, D.W., 1997. Perry’s Chemical Engineers’ Handbook, seventh ed. McGraw-Hill, New York, pp. 2e313. Pevere, A., Guibaud, G., van Hullebusch, E.D., Boughzala, W., Lens, P.N.L., 2007. Effect of Naþ and Ca2þ on the aggregation properties of sieved anaerobic granular sludge. Colloids and Surfaces A: Physicochemical and Engineering Aspects 306 (1e3), 142e149. Prorot, A., Laurent, J., Dagot, C., Leprat, P., 2011. Sludge disintegration during heat treatment at low temperature: a better understanding of involved mechanisms with a multiparametric approach. Biochemical Engineering Journal 54 (3), 178e184. Raunkjær, K., Hvitved-Jacobsen, T., Nielsen, P.H., 1994. Measurement of pools of protein, carbohydrate and lipid in domestic wastewater. Water Research 28 (2), 251e262. Shao, L., He, P., Yu, G., He, P., 2009. Effect of proteins, polysaccharides, and particle sizes on sludge dewaterability. Journal of Environmental Sciences 21 (1), 83e88. Sobeck, D.C., Higgins, M.J., 2002. Examination of three theories for mechanisms of cation-induced bioflocculation. Water Research 36 (3), 527e538. Teo, K.C., Xu, H.L., Tay, J.H., 2000. Molecular mechanism of granulation. II: proton translocating activity. Journal of Environmental Engineering 126 (5), 411e418. Tezuka, Y., 1969. Cation-dependent flocculation in a flavobacterium species predominant in activated sludge. Applied Microbiology 17 (2), 222e226.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 3 e4 4 1
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
The influence of hydraulic loads on depth filtration Jinkeun Kim a,b,*, Desmond F. Lawler b a b
Dept. of Environmental Engineering, Jeju National University, Jeju 690-756, Republic of Korea Dept. of Civil, Architectural and Environmental Engineering, The University of Texas at Austin, C1786 Austin, TX 78712, USA
article info
abstract
Article history:
The influence of hydraulic loads on the detachment of particles from the collector surface
Received 28 June 2011
or from previously retained particles was observed in a packed glass beads column. A
Received in revised form
hydraulic shock load (i.e., 20% increase of flow rate) was applied after 4 h of particle
16 September 2011
attachment at a constant flow rate. A single type of particle suspension (Min-U-Sil 5, nearly
Accepted 30 October 2011
pure SiO2) and three different chemical conditions (pH control, alum and polymer desta-
Available online 15 November 2011
bilization) were utilized. The magnitude of particle detachment increased with increasing particle size for non-Brownian particles because more shear force was applied to large
Keywords:
particles due to their large surface area. More favorable particles (i.e., particles with small
Filtration
surface charge) were detached to a lesser extent than unfavorable particles during the
Detachment
hydraulic shock loads application. This phenomenon can be caused by floc strength. In
Particle size
some cases, when the zeta potential of influent particles was relatively high, the magni-
Surface charge
tude of detachment of bigger particles (e.g., 4.0e5.0 mm) was less than that of smaller
Hydraulic loads
particles (e.g., 3.0e4.0 mm). This can be attributable to the breakup of detached flocs as an
Water treatment
individual particle. It was also found that the shape of the curve relating the magnitude of particle detachment and particle size can be concave, linear, or convex depending on physicochemical conditions such as floc strength. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
A wide variety of particles including microorganisms, other organic materials, and inorganics are present in water as a result of either natural events or human activities. As higher quality drinking water is demanded, the need to remove those particles in drinking water treatment processes is increased. Some particles of particular concern are microorganisms such as viruses, bacteria, and protozoa, because incomplete removal and inactivation of these particles may cause an acute threat to public health. In the treatment of surface water used for drinking water, filtration in a bed of granular media, i.e., depth filtration, is used almost universally. As filtration is usually the last
particle removal process in water treatment processes, reliable particle removal efficiency is desired (e.g., 3 log removal requirement for Cryptosporidium in conventional filtration process) during this process for both esthetic and public health reasons (USEPA, 2011). In recent years, much attention has been focused on the removal of small particles from low turbidity water supplies to ensure the sufficient removal of protozoa like Giardia or Cryptosporidium, and thereby improve the safety of drinking water. Many researchers believe that the removal of 2e13 mm particles is a good surrogate measure of the removal of these protozoa (Kawamura, 2000). The removal of microorganisms in filtration is perhaps the most important reason for its application in water treatment plants (WTPs) (Tobiason et al., 2011).
* Corresponding author. Dept. of Environmental Engineering, Jeju National University, Jeju 690-756, Republic of Korea. Tel.: þ82 64 754 3448; fax: þ82 64 725 2483. E-mail address:
[email protected] (J. Kim). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.059
434
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 3 e4 4 1
Filtered water with a turbidity as low as 0.1 NTU, which is considered safe for drinking, can contain hundreds of particles per milliliter (McTigue et al., 1998). As a suspension passes through the filter, some particles attach to the collector surface or to previously-attached particles, but others remain in suspension, depending on their physicochemical characteristics. At the same time, some of the previously-captured particles can be detached either as individual particles or flocs. Detachment, i.e., reentrainment of particles after they have been attached to collectors, is the reverse process of attachment (i.e., deposition). The duration and pattern of particle detachment or breakthrough can be influenced by many parameters including filter influent water characteristics, backwash procedure, coagulation efficiency, floc strength, operating conditions such as filtration velocity or hydraulic loading rate (flow rate normalized by filter surface area). These factors could either increase or decrease the effect of the different mechanisms involved and significantly affect the duration and severity of particle breakthrough. Particle detachment is still not completely understood, and particle breakthrough into the filtered water is not always well managed at water treatment plants (WTPs) (Ryan and Gschwend, 1994; Bai and Tien, 1997; Bergendahl and Grasso, 1999; Amburgey, 2005; Williams et al., 2007; Han et al., 2009). The fundamental mechanisms of the particle detachment process have not been explored in as much detail as that for particle attachment. In the past, researchers had investigated particle detachment processes in terms of macroscopic parameters such as effluent concentration. Only recently have researchers investigated particle detachment from a fundamental approach, i.e., consideration of various forces such as hydrodynamic and thermodynamic forces (Raveendran and Amirtharajah, 1995; Bai and Tien, 1997; Bergendahl and Grasso, 1999; Han et al., 2009). The forces that usually control the detachment step are chemical and colloidal forces (i.e., van der Waals, electrostatic, and other non-DLVO forces), while the transport step is governed by forces controlling Brownian motion of the particles, hydrodynamic drag forces, and gravitational forces (Ahmad and Amirtharajah, 1998). For attached particles in porous media to become detached, a disturbance to the system must occur. This perturbation can be a change in the chemistry of the solution or the hydrodynamics of the system (Bergendahl and Grasso, 2000). Chemical aspects of particle detachment have been studied experimentally using the packed column technique by several researchers. Most researchers identified ionic strength and pH of the solution as the major parameters for affecting detachment when a constant hydrodynamic force was used. They found that high pH and decreasing ionic strength increased the particle detachment because it increased the electric double layer repulsive force (Tobiason, 1989; Sharma et al., 1992; Elimelech, 1994; Ryan and Gschwend, 1994; Raveendran and Amirtharajah, 1995; Nocito-Gobel and Tobiason, 1996; Bai and Tien, 1999; Jegatheesan and Vigneswaran, 2000; Kuznar and Elimelech, 2007). One of the most important causes of particle detachment in WTPs is a sudden change of flow shear force caused by hydraulic shock loads. A hydraulic increase occurs whenever one filter is taken out of service (e.g., for backwashing) and the
overall plant flow remains constant, or when the plant flow is increased to meet increasing demand. Analysis of filtrate turbidity and particle counts for the in-service filters frequently shows an intermittent increase when other filters are backwashed (Kawamura, 2000; Han et al., 2009; Tobiason et al., 2011). These increasing filtration velocities will result in increasing drag forces on the deposited particles. Eventually the drag forces reach a magnitude equal to the adhesive forces, and after that point, particles are detached. The amount of particle detachment depends on the magnitude and duration of the hydraulic loads (Adin and Rebhun, 1987; Amirtharajah, 1998; Bergendahl and Grasso, 2000; Han et al., 2009). Under the action of hydrodynamic forces caused by the flow of water through the collector, which increases with increasing head loss, the structure is partially destroyed. Some of the previously attached particlesdthose that are less strongly linked to the others–can be detached from the collectors. The surface deposit on the filter media builds up to the point that increased shear forces in the pore spaces cause some of the more weakly held particles (e.g., particles with a more negative surface charge) to become detached (Ives, 1989; McDowell-Boyer, 1992). Moran et al. (1993) noticed particle detachment under reduced particle input with identical chemical and hydraulic conditions at the end of long filtration experiment, and they concluded that hydraulic shear stress caused particle breakthrough. Meanwhile, Kim and Tobiason (2004) documented that, for much of a filter run, filtrate particles are dominated by detached particles compared with non-attached influent particles. Several researchers attempted to predict the influence of flow shear on particle detachment. Bai and Tien (1997) postulated that net tangential force (Fn) can be calculated by subtracting the hydrodynamic shear force from the frictional force against sliding. They concluded that Fn would be dependent on particle and collector diameter, flow rate, and hydraulic gradient in a filter bed layer. Bergendahl and Grasso (2000) also introduced a similar idea, and they tried to predict particle detachment based on two parameters: the depth of the primary minimum (DGmin) that can be calculated from the gamma distribution and the necessary fluid shear to overcome DGmin. Many researchers have investigated various aspects of particle detachment (breakthrough) with limited success, yet few has investigated physicochemical characteristics of detached particles in terms of the zeta potential distribution (ZPD) as well as particle size distribution (PSD). The object of this research is to further elucidate the physicochemical aspects of particle detachment. The present research is motivated by the assumption that, by understanding these aspects, particle breakthrough in WTPs can be minimized through appropriate treatment.
2.
Materials and methods
2.1.
Particles and Porus media
Min-U-Sil 5 (U.S. Silica Company, Berkeley Springs, WV) particles were used throughout this research. These particles are white, natural crystalline silica powder (>99.2% SiO2) with virtually all particles below 5.0 mm in size (diameter). The PSD
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 3 e4 4 1
was measured by using a Coulter Counter (Coulter Multisizer, Coulter Electronics Inc., Hialeah, FL), and the majority of particles (88.4%) have diameters in the range of 0.8e2.5 mm (0.1 < log dp< 0.4) as shown in Fig. 1. Scanning electron microscopy (SEM) (Jeol T330A, Jeol USA Inc., Peabody, MA) images reveal that these particles are quite angular. Spherical soda-lime glass beads (Potters Industries, Inc., NJ) in a size range of 0.5e0.6 mm were utilized for filter media, and SEM measurements confirmed their sphericity. The manufacturer claims that glass beads are at least 72.5% SiO2. The glass beads were thoroughly cleaned to remove impurities by rinsing sequentially with deionized water from a MilliQ water system (Millipore Corp., Bedford, MA); 0.01 N NaOH; 1 N HNO3; deionized water (repeated rinsing). The beads were dried in an oven and stored in a desiccator until use. Clean glass beads were used for each set of column experiments.
2.2.
Experiments
Depth filtration experiments were performed for 5 h in a down-flow column with an inner diameter of 3.8 cm and
Fig. 1 e Particle size distribution of Min-U-Sil 5 Particles (C [ 100 mg/L) (top) and schematic of experimental filtration system (not to scale) (bottom).
435
media depth of 10 cm. This shallow bed depth was chosen for optimization of particle counting characterization and maintenance of the supply tank. With this configuration, complete particle removal was avoided, and yet the filter bed was deep enough to capture the majority of particles that would be captured in a much deeper column (e.g., a full scale column). A schematic diagram of the experimental filtration system utilized in this research is presented in Fig. 1. Each experiment consisted of a three-step procedure. In stage 1, a suspension of 20 mg/L Min-U-Sil 5 was fed continuously to the filter column for 4 h with filtration velocity of 5 m/h for particle attachment. In stage 2, the column was rinsed at the same flow rate as in stage 1 with a particle-free solution of the same chemical conditions as the solution used in phase 1 for 6 min. Finally, in stage 3, filtration velocity was increased by 20% for 1 h to detach particles that were previously attached to collectors. Han et al. (2009) also set the filtration velocity as 5 m/h and investigated particle detachment in terms of particle concentration and head loss in granular filtration under 20% increase of hydraulic loads. Separate gear pumps were provided for the water and the high-concentration suspension (i.e., 800 mg/L); after they were blended in-line, a syringe pump was used to inject the destabilizing chemical immediately prior to the filter influent, thereby ensuring that very little change of the size distribution occurred by flocculation prior to the filter. Aqueous solutions were prepared from a Milli-Q water system. A more detailed explanation of the filter system can be found elsewhere (Kim, 2004). To analyze the magnitude of solids detachment during the hydraulic shock loads, the total mass of solids attached during 4 h was calculated, and the solids mass of the effluent after the hydraulic shock loads was normalized by the total accumulated mass before the hydraulic shock loads were applied. During the hydraulic loads experiments, samples were taken every minute for 5 min, and less frequently thereafter. Usually, it took 1 min of sampling to have enough volume of sample to analyze PSD, ZPD, and solids concentrations. Solids concentrations were measured using a PerkineElmer UV/VIS (lambda 3) spectrophotometer, based on a calibration curve with known concentrations. ZPDs were obtained from electrophoretic mobility measurements made with a Zetaphoremeter IV (CAD, France), a microscopic electrophoresis instrument equipped with an automatic tracking function made possible by digital image processing. To investigate particle detachment under various chemical conditions, three different methods (pH control, polymer and alum destabilization) were used. For each method of particle destabilization (i.e., mitigation of particles stabilities), three different conditions were chosen; one of which was expected to have much better particle removal than the others based on surface charge considerations. For the destabilization by pH control, experiments were performed at a constant ionic strength (102 M) and the ionic strength was bolstered as necessary by KCl. At the chosen pH values (3.0, 4.0, and 5.0), no other buffering was provided. Alum and a polyamine (Superfloc C-572, Cytec Industries Inc., IN, dimethyl amine polymerized with epichlorohydrin) that is commonly used at WTPs were applied to destabilize particles for attachment prior to the detachment phase of the
436
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 3 e4 4 1
Results and discussion
3.1.
Tracer test
To determine the flow characteristics of the filter system, conservative tracer tests were performed using 1% NaCl solution. During tracer tests, the filter system was operated using the same procedure as in the main experiments. The relative concentration of NaCl in the effluent was measured by a conductivity meter (CDM 230, MeterLab, France). The tracer test was implemented for 10 min, and this time was sufficient to have a complete breakthrough curve. Samples were taken every 10 s before ultimate breakthrough occurred. Based on the three identical tracer tests results, it was deduced that complete particle breakthrough happened within approximately 4 min, i.e., at 0.33 m3/m2 cumulative hydraulic loading (CHL) after filtration started. During this test, 1 min filtration time corresponded to 0.083 m3/m2 in terms of CHL because the filtration velocity was 5 m/h. The CHL is defined as the volume throughout per cross sectional filter area; using this measure rather than time allows comparisons based on similar amounts of water treated to be made between the experiments.
3.2.
pH control
The solids concentration remaining during the entire 5 h filtration experiments (25 m3/m2 CHL) at pH 3.0 is shown in Fig. 2. After 4 h of normal filtration (stage 1) and the 6 min rinse with no particles (stage 2), the filtration velocity was increased 20%, i.e., from 5 m/h to 6 m/h. To exclude any influence of influent particles, the hydraulic shock loads were applied 0.50 m3/m2 CHL after the particle suspension line was closed. As 0.33 m3/m2 CHL was needed for thorough flushing of the filter volume, 0.50 m3/m2 CHL (i.e., 6 min) was thought to be enough to have particle-free influent. A small amount of particle detachment was noticed even during the 6 min with no particle input, presumably caused by particle detachment as documented by Moran et al. (1993). After that period, a substantial amount of particle detachment happened within 1 m3/m2 CHL (i.e., quite rapidly) after the increased hydraulic loads were applied. Even though the influent concentration was essentially zero during the hydraulic shock loads, the results were normalized by the pre-shock loads influent concentration. Three different pH conditions (pH ¼ 3.0, 4.0 and 5.0) were chosen and corresponding mean zeta potentials ( one standard deviation) were 26.4 7.1 mV, 38.7 9.2 mV, and
Solids Conc. Remaining (C/Co)
0.8 0.6
Shut-off particle feeding
0.4 0.2 0
Increase flow rate 0
5
10
15
20 3
25 2
Cumulative Hydraulic Loading (m /m ) Fig. 2 e Solids concentration remaining (C/Co) during the whole experiment (5 h) at pH 3.0.
54.7 9.3 mV, respectively. Detailed test results on the zeta potential of Min-U-Sil 5 at these pH values can be found elsewhere (Kim et al., 2006, 2008). Solids detachment after the hydraulic shock loads is shown in Fig. 3(A) for the set of pH experiments. Particle detachment was more severe at higher pH, i.e., the more
A Solids Mass Detachment (%)
3.
1
2 pH-3.0 pH-4.0 pH-5.0
1.5
1
0.5
0
0-1
1-2
2-3
3-4
4-5
Time (min)
B
80 Infleunt S-0 min S-1 min
70 60
Fraction (%)
experiment. In neutralizing the surface charge of particles by alum, the pH was set to 5.2 as recommended by Amirtharajah and Mills (1982), and this same pH was used in the polymer experiments. To supply alkalinity that was reasonably consistent with conditions in WTPs, particles were dispersed in 2 103 M NaHCO3 throughout the alum and polymer experiments. Doses of coagulants were chosen based on jartest results, and more detailed description and results of particle attachment portion of the experiments can be found elsewhere (Kim et al., 2008).
50 40 30 20 10 0
1.0-2.0
2.0-3.0
3.0-4.0
4.0-5.0
Particle size (µm) Fig. 3 e Solids mass detachment during the hydraulic shock loads application at pH control (normalized by total mass accumulated during 4 h filtration) (A) and particle size fraction (number base) at pH 4.0 during the hydraulic shock loads application (B) (S-0 min and S-1 min represent samples taken from 0 to 1 min and 1e2 min, respectively).
437
3.3.
Alum destabilization
Three alum doses (i.e., 0.06, 0.2, and 0.8 mg/L) were chosen to make negative, near zero, and positive surface charge
Particle Number Detachment (%)
150
Particle Number Detachment (%)
repulsion dominant region, even though less solid mass was removed during the attachment phase. It is thought that attachment forces were stronger at lower pH (where the absolute value of the zeta potential was lower), so that less particle detachment happened in the more destabilized conditions. These results suggest that the magnitude of particle detachment is inversely related to the floc strength. Fig. 3(B) shows the fraction of particles (on a number basis) in the standard influent and in two samples of effluent after the increase of hydraulic load at pH 4.0. The typical particle size range of Min-U-Sil 5 is between 0.6 and 5.0 mm. During these experiments, only non-Brownian particles (1.0e5.0 mm) were analyzed, and the majority of influent particles were between 1.0 and 2.0 mm. After the hydraulic shock loads were applied, the relative magnitude of particles in the various size ranges is substantially different, with the greatest number in the 2.0e3.0 mm size clas. Relative to the influent, the fraction in the largest two size ranges is also higher after the change in hydraulic load. Clearly, bigger particles were more easily detached than small particles under the increased hydraulic load. A higher shear force is applied to large particles because of their large surface area (Bai and Tien, 2000), and that is the most likely explanation for the size trend in the detachment. One other possibility is that larger particles have more negative surface charges, so particles with large size can be easily detached because attachment between particles and collectors is weak. But this possibility can be assumed to be relatively low, because most of the detached large particles may not be individual particles but flocs. Flocs that were once attached to the filter media but were broken off can be assumed to have less negative surface charge than particles in the suspension that were never caught (non-attachment particles); otherwise flocs would not have been made. A broad distribution of the zeta potential of Min-U-Sil 5 particles was reported by Kim et al. (2008). From Fig. 4, which illustrates particle detachment on the basis of particle number concentration at different pH values (i.e., pH 3.0, 4.0 and 5.0), three things can be deduced. First, particle detachment was more severe at higher pH (i.e., higher particle surface charge). It can be thought that attachment forces were stronger at lower pH (i.e., smaller particle surface charge), so that less particle detachment happened in more destabilized conditions. Second, larger particles are more liable to be detached than smaller particles during the hydraulic shock loads. However, at pH 5.0, particles in the 4.0e5.0 mm size range showed slightly less increase of particle number than the 3.0e4.0 mm group did, as shown in Fig. 4(C). It can be considered that some of the large particles were not a single particle but a floc, and during the hydraulic shock loads some of the larger flocs may have broken into separate particles, so that the number of the larger particles decreased. Third, the most severe particle detachment happened immediately after the application of the hydraulic shock loads, and its magnitude gradually decreased with time.
150
Particle Number Detachment (%)
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 3 e4 4 1
500
A
S-0 min S-1 min S-5 min
100
50
0
B
S-0 min S-1 min S-4 min
100
50
0
C
S-0 min S-1 min S-4 min
400 300 200 100 0 1.0-2.0
2.0-3.0
3.0-4.0
4.0-5.0
Particle size (µm) Fig. 4 e Particle number detachment during the hydraulic shock loads application at pH values 3.0 (A), 4.0 (B) and 5.0 (C) (normalized by influent PSD) (S-0 min, S-1 min and S-4 min represent samples taken from 0 to 1 min, 1e2 min, and 4e5 min, respectively).
conditions based on the preliminary jar tests, and the corresponding mean zeta potentials (one standard deviation) of particles were 38.9 12.1 mV, 7.9 12.2 mV, and 54.3 9.9 mV, respectively. Fig. 5 illustrates the particle number detached at alum dose of 0.06 mg/L (below optimum dose) and 0.2 mg/L (optimum dose) during the hydraulic shock loads application. The results are expressed as a percent of the influent PSD in four size classes. In general, particle detachment increased with the increase of particle size in both experiments. However, more particle detachment was noticed in the 3.0e4.0 mm range in comparison to the biggest particle size range (4.0e5.0 mm) in the low dose experiment (Fig. 5(A)). This result is thought to mean that bigger particles (i.e., flocs) were broken off as smaller individual particles or flocs.
150
A
S-0 min S-1 min S-5 min
120 90 60 30 0
6
1 10
Before After
5
8 10
5
6 10
5
4 10
5
2 10
0
0
0.1
150
0.3
0.4
0.5
0.6
0.7
p
B
S-0 min S-1 min S-2 min
120 90 60 30 0
0.2
Log of Particle Diameter (d in µm)
1.0-2.0
2.0-3.0
3.0-4.0
4.0-5.0
Particle size (µm) Fig. 5 e Particle number detachment during the hydraulic shock loads application at alum dose of 0.06 mg/L (A) and 0.2 mg/L (B) (normalized by influent PSD).
Ultrasonication (Branson Model: 13-22-4, Shelton, CT) of the detached particles during the hydraulic shock loads were conducted for 10 min to investigate whether bigger particles in the effluent were individual particles or flocs made up of two or more particles. The PSD of the S-0 min sample during the hydraulic shock loads at alum dose of 0.06 mg/L before and after sonication is shown Fig. 6; the results indicate that the PSD after 10 min of sonication shifted slightly to smaller sizes. This result suggests that larger particles were broken into several smaller particles when the higher shear field of the hydraulic shock loads was applied. A paired t-test showed significant statistical difference between two PSDs. The possibility of breakup of Min-U-Sil 5 particles themselves during sonication was investigated, but there were no statistically significant differences between the PSDs of the influent before and after sonication. That is, unflocculated Min-U-Sil 5 particles did not breakup during the sonication.
3.4.
Number Distribution (number/mL)
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 3 e4 4 1
Polymer destabilization
Three different polymer doses (i.e., 0.001, 0.01, 0.1 ppmv) were chosen based on jar-test results, and the corresponding mean zeta potentials (one standard deviation) of particles were 53.3 13.1 mV, 1.0 18.1 mV, and 50.9 7.9 mV. ZPDs of the effluent at the polymer dose of 0.001 ppmv during normal filtration and hydraulic loads application are shown in Fig. 7. The ZPD of effluent at 180 min during the deposition stage (i.e., Eff.-180 min) was far to the left of (more
Fig. 6 e PSDs of samples before sonication and after sonication at alum dose of 0.06 mg/L (sample of S-0 min., duration-10 min).
negative than) the influent. On the other hand, the ZPDs of S0 min and S-1 min (after the increase in the hydraulic loading) were between the influent and Eff.-180 min. It can be deduced that the zeta potentials (absolute values) of detached particles during the hydraulic shock loads were higher than those of the influent but less than those of the effluent. Two things can be deduced from Fig. 7. First, less negative (more destabilized) particles were well attached during the normal filtration, because the ZPD of Eff.-180 min during the attachment stage was more negative than the ZPD of influent. Second, when the ZPDs of Eff-180 min, S-0 min and S-1 min were compared, it was seen that the mean zeta potential moved slightly from more to less negative, which means more weakly held particles were detached right after the hydraulic shock loads. It can be considered that, as soon as hydraulic shock loads were applied, more stable particles were detached, and then less stable particles were subsequently detached. This sequence means the ZPDs were shifting from more to less negative under the hydraulic shock loads application when the surface charges of particles were negative. All four of the ZPDs in Fig. 7 were statistically different with 95% confidence based on t-test results. 30 Influent Eff.-180 min S-0 min S-1 min
25
Fraction (%)
Particle Number Detachment (%)
Particle Number Detachment (%)
438
20 15 10 5 0 -90
-80
-70
-60
-50
-40
Zeta Potential (mV)
-30
-2
0
Fig. 7 e Zeta potential distributions of Min-U-Sil 5 during the hydraulic shock loads application at polymer dose of 0.001 ppmv.
439
Solid Mass Detachment (%)
1.2 0.001 ppm 0.01 ppm 0.1 ppm
1
Particle Number Detachment (%)
Particle Number Detachment (%)
On the other hand, one of the possibilities of this movement was a change of chemical conditions. During the hydraulic shock loads, the polymer dose was the same as during the normal filtration (i.e., polymer flow increased proportionally to the increase of supply water flow) but there was no particle input (i.e., no consumption of positive charges by the particle suspension before the filtration). This difference might cause the movement of the ZPDs of effluent from more to less negative during the hydraulic shock loads experiment. Further research is needed to elucidate the characteristics of ZPDs of detached particles under various chemical conditions. The detachment on a solids mass basis under the hydraulic shock loads at various polymer doses is shown in Fig. 8. The least detachment and greatest solids accumulation were noticed at the optimum dose. It can be assumed that floc strength was too great to allow much detachment at the optimum dose (0.01 ppmv). The particle number detached during the hydraulic shock load at polymer dose of 0.001 ppmv is shown in Fig. 9(A). The amount of detached particles in the size range 4.0e5.0 mm was approximately the same as that of 3.0e4.0 mm particles at this polymer dose. It can be considered that flocs in the size range 4.0e5.0 mm were broken into two or more smaller particles as previously noticed in pH control and alum destabilization experiments. The particle detachment patterns under the hydraulic shock loads at polymer doses of 0.01 ppmv and 0.1 ppmv are shown in Fig. 9(B) and (C). Under pH control and alum destabilization experiments, particle detachment was increased with increasing particle size, and its slope was linear or convex. It can be thought that the shape of the slope had a close relationship with floc strength. If floc strength was high, then it was hard to break the floc when the hydraulic shock load was applied, so the slope would be concave. However, if floc strength was low, then the floc could easily be broken off the filter and also broken into several individual particles, which made the slope convex. From Fig. 9(B), it can be deduced that the floc strength at the optimum dose was very strong as it had a concave shape. Polymer destabilization
Particle Number Detachment (%)
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 3 e4 4 1
300
A
S-0 min S-1 min S-2 min
250 200 150 100 50 0 300
B
S-0 min S-1 min S-3 min
250 200 150 100 50 0 300 250 200
C
S-0 min S-1 min S-3 min S-5 min
150 100 50 0
1.0-2.0
2.0-3.0
3.0-4.0
4.0-5.0
Particle Size (µm) Fig. 9 e Particle number detachment during the hydraulic shock loads application at polymer dose of 0.001 ppmv (A), 0.01 ppmv (B) and 0.1 ppmv (C) (normalized by influent PSD).
0.8 0.6 0.4 0.2 0
0-1
1-2
2-3
Time (min)
3-4
4-5
Fig. 8 e Solids mass detachment during the hydraulic shock loads application at polymer destabilization (normalized by total mass accumulated during 4 h filtration).
is known to increase floc strength and is widely practiced in WTPs for that reason (Fabrizi et al., 2010). From the comparison of the particle number detached at S0 minute under the three different polymer doses, two things were observed. First, particle detachment was a function of particle size; that is, bigger particles can be more easily detached than smaller particles. Second, the slope of the graph was a function of floc strength. When the floc strength was small (0.001 ppmv), then the shape of graph would be convex, while if the floc strength was intermediate (0.1 ppmv), then its slope would be linear. And, if the floc strength (0.01 ppmv) was high, then its shape would be concave. In a real WTP, achieving a higher floc strength by proper charge neutralization is very important to minimize particle
440
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 3 e4 4 1
detachment; furthermore, detached particles can be reattached at subsequent filter depth assuming that the filter depth is sufficient, because particle transport efficiency increases with increasing particle size for non-Brownian particles.
3.5.
4.
Full scale observations
Sudden increases of filtration velocity (i.e., hydraulic loads) in WTPs can be caused by abrupt increases of inflow rate as well as by filter backwashing. Fig. 10 shows the influence of influent flow variation (relative to the mean flow for the day) on filtrate turbidity for 24 h at a conventional full scale WTP (Q ¼ 200,000 m3/day) that treats typical surface water in Korea. The influent flow was proportional to the number of operating intake pumps. Inflow rate and filtrate turbidity were measured using on-line instruments. It is clear that hydraulic loads significantly influenced filtrate turbidity, and the lag time between increase of inflow rate and turbidity breakthrough was less than 40 min, even though retention time of flocculation and rectangular sedimentation basin was over 4 h. The highest turbidity increase was noticed between hours 13 and 14; during this period, intake flow was increased 40% but turbidity increased 123% (from 0.026 NTU to 0.058 NTU). The duration of particle detachment lasted for about 40 min, and then the filtrate turbidity returned to a normal value, even though the influent flow rate remained high. It can be recommended that the change of filtration velocity should be controlled within relatively narrow limits to minimize particle detachment. This allowable filtration velocity change limit can be varied depending on floc strength (i.e., chemical conditions), filter depth and stage of filtration (i.e., head loss). Kawamura (2000) recommended that no more than 33%, and preferably a maximum of 15e20%, of hydraulic surcharge be allowed to flow to the remaining filters during backwashing; the result from the Korean plant shown in Fig. 10 reinforce this recommendation. The necessity for tight control of hydraulic changes in the plant might require changes in the design and operation of conventional water treatment plants. For example, the array of pumps at a large plant should be sufficient to make any hydraulic change be
40
Turbidity
0.06
20 0
0.05
-20
0.04
-40 0.03
-60 -80
0
4
When increased hydraulic loads were applied after a substantial period of depth filtration, the magnitude of particle detachment was proportional to the absolute value of the influent zeta potential. This can be attributable to the floc strength, which suggests that less favorable particles (i.e., particles with large surface charge) are more likely to be detached due to weak floc strength. At the same time, the amount of particle detachment (when the effluent PSD was normalized by the influent PSD) was closely related to the particle size; bigger particles were more liable to be detached than were smaller particles due to larger particles’ greater surface area. On the other hand, the detachment of larger particles (e.g., 4.0e5.0 mm) was less than that of smaller particles (e.g., 3.0e4.0 mm) when the influent zeta potential was too high. It is thought that bigger particles (i.e., flocs) were broken into small several individual particles due to weak attachment force of flocs. Meanwhile, the shape of the curve relating the magnitude of particle detachment to particle size can be concave, linear, or convex depending on physicochemical conditions such as floc strength. If the floc strength is small, then the slope can be convex; if it is intermediate, then the slope can be linear; and, if it is large, then the shape can be concave. When particle detachment at the optimum doses of both alum and polymer was compared, less particle detachment was noticed with polymer destabilization, although more solids mass was removed during 4 h of filtration. Therefore, to reduce possibility of particle detachment or to improve floc strength, polymer can be used.
Acknowledgment
0.07
8
12
Time (hr)
16
20
24
0.02
Fig. 10 e Influence of influent flow variation on filtrate turbidity.
Turbidity (NTU)
Inflow Flow Variation (%)
Inflow
Conclusions
The authors thank Prof. Jeffery A. Nason (Oregon State University) for helpful encouragement, discussions and stimulating comments during this research.
0.08
80 60
gradual, and the design and operation of the clearwell should provide sufficient equalization capacity to prevent the need for large increases in the flow rate through the plant.
references
Adin, A., Rebhun, M., 1987. Deep-Ded filtration: accumulationdetachment model parameters. Chem. Eng. Sci. 42 (5), 1213e1219. Ahmad, R., Amirtharajah, A., 1998. Detachment of particles during Biofilter backwashing. J. Am. Water Works Assoc. 90 (12), 74e85. Amburgey, J.E., 2005. Optimization of the extended terminal subfluidization wash (ETSW) filter backwashing procedure. Water Res. 39 (2e3), 314e330. Amirtharajah, A., 1998. Some Theoretical and Conceptual Views of filtration. J. Am. Water Works Assoc. 80 (12), 36e46.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 3 3 e4 4 1
Amirtharajah, A., Mills, K.M., 1982. Rapid-mix design for mechanisms of alum coagulation. J. Am. Water Works Assoc. 74 (4), 210e216. Bai, R., Tien, C., 1997. Particle detachment in deep bed filtration. J. Colloid Interface Sci. 186 (2), 307e317. Bai, R., Tien, C., 1999. Particle deposition under unfavorable surface Interactions. J. Colloid Interface Sci. 218 (2), 488e499. Bai, R., Tien, C., 2000. Effect of deposition in deep-bed filtration: determination and search of rate parameters. J. Colloid Interface Sci. 231 (2), 299e311. Bergendahl, J., Grasso, D., 1999. Prediction of colloid detachment in a model porous media: thermodynamics. AIChE J. 45 (3), 475e484. Bergendahl, J., Grasso, D., 2000. Prediction of colloid detachment in a model porous media: hydrodynamics. Chem. Eng. Sci. 55, 1523e1532. Elimelech, M., 1994. Effect of particle size on the Kinetics of particle deposition under Attractive double layer Interactions. J. Colloid Interface Sci. 164 (1), 190e199. Fabrizi, L., Jefferson, B., Parsons, S.A., Wetherill, A., Jarvis, P., 2010. The Role of polymer in improving floc strength for filtration. Environ. Sci. Tech. 44 (16), 6443e6449. Han, S., Fitzpatrick, C., Wetherill, A., 2009. The impact of flow surges on rapid gravity filtration. Water Res. 43 (5), 1171e1178. Ives, K.J., 1989. Filtration studied with Endoscopes. Water Res. 23 (7), 861e866. Jegatheesan, V., Vigneswaran, S., 2000. Transient stage deposition of Submicron particles in deep bed filtration under unfavorable conditions. Water Res. 34 (7), 2119e2131. Kawamura, S., 2000. Integrated Design and Operation of Water Treatment Facilities. John Wiley & Sons, Inc., New York. Kim, J., 2004. Physicochemical Aspects of Particle Breakthrough in Granular Media Filtration Ph.D. dissertation. The University of Texas at Austin. Kim, J., Nason, J.A., Lawler, D.F., 2006. Zeta potential distributions in particle treatment processes. J. Water Supply: Res. and Tech.dAQUA 55, 7e8. Kim, J., Nason, J.A., Lawler, D.F., 2008. Influence of surface charge distributions and particle size distributions on particle
441
attachment in granular media filtration. Environ. Sci. Tech. 42 (7), 2557e2562. Kim, J., Tobiason, J.E., 2004. Particles in filter effluent: the Roles of deposition and detachment. Environ. Sci. Tech. 38 (22), 6132e6138. Kuznar, Z.A., Elimelech, M., 2007. Direct microscopic observation of particle deposition in porous media: role of the secondary energy minimum. Colloid Surf. A 294, 156e162. McDowell-Boyer, L.M., 1992. Chemical Mobilization of MicronSized particles in Saturated porous media under Steady flow conditions. Environ. Sci. Tech. 26 (3), 586e593. McTigue, N.E., LeChevallier, M., Arora, H., Clancy, J., 1998. National Assessment of Particle Removal by Filtration. AWWARF and AWWA, Denver. Moran, M.C., Moran, D.C., Cushing, R.S., Lawler, D.F., 1993. Particle Behavior in deep bed filtration: Part 2-Particle detachment. J. Am. Water Works Assoc. 85 (12), 82e93. Nocito-Gobel, J., Tobiason, J.E., 1996. Effects of ionic strength on colloid deposition and Release. Colloid Surf. A 107, 223e231. Raveendran, P., Amirtharajah, A.A., 1995. Role of Short-range forces in particle detachment during filter backwashing. J. Environ. Eng. 121 (12), 860e868. Ryan, J.N., Gschwend, P.M., 1994. Effects of ionic strength and flow rate on colloid Release: relating Kinetics to Intersurface potential energy. J. Colloid Interface Sci. 164 (1), 21e34. Sharma, M.M., Chamoun, H., Sarma, D.S., Schechter, R.S., 1992. Factors controlling the hydrodynamic detachment of particles from Surfaces. J. Colloid Interface Sci. 149 (1), 121e134. Tobiason, J.E., 1989. Chemical effects on the deposition of nonBrownian particles. Colloids Surf. 39 (1), 53e77. Tobiason, J.E., Cleasby, J.L., Logsdon, G.S., O’Melia, C.R., 2011. Granular media filtration. In: Edzwald, J. (Ed.), Water Quality and Treatment, sixth ed. McGraw-Hill, pp. 10.45e10.46. USEPA, 2011. http://www.epa.gov. Williams, G.J., Sheikh, B., Holden, R.B., Kouretas, T.J., Nelson, K.L., 2007. The impact of increased loading rate on granular media, rapid depth filtration of wastewater. Water Res. 41 (19), 4535e4545.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 2 e4 5 2
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Contributions of meteorology to the phenology of cyanobacterial blooms: Implications for future climate change Min Zhang, Hongtao Duan, Xiaoli Shi, Yang Yu, Fanxiang Kong* State Key Laboratory of Lake Science and Environment, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, 73 East Beijing Road, Nanjing 210008, PR China
article info
abstract
Article history:
Cyanobacterial blooms are often a result of eutrophication. Recently, however, their
Received 9 August 2011
expansion has also been found to be associated with changes in climate. To elucidate the
Received in revised form
effects of climatic variables on the expansion of cyanobacterial blooms in Taihu, China, we
26 October 2011
analyzed the relationships between climatic variables and bloom events which were
Accepted 4 November 2011
retrieved by satellite images. We then assessed the contribution of each climate variable to
Available online 17 November 2011
the phenology of blooms using multiple regression models. Our study demonstrates that retrieving ecological information from satellite images is meritorious for large-scale and
Keywords:
long-term ecological research in freshwater ecosystems. Our results show that the
Climate change
phenological changes of blooms at an inter-annual scale are strongly linked to climate in
Cyanobacterial blooms
Taihu during the past 23 yr. Cyanobacterial blooms occur earlier and last longer with the
Taihu
increase of temperature, sunshine hours, and global radiation and the decrease of wind
Satellite image
speed. Furthermore, the duration increases when the daily averages of maximum, mean,
Temperature
and minimum temperature each exceed 20.3 C, 16.7 C, and 13.7 C, respectively. Among these factors, sunshine hours and wind speed are the primary contributors to the onset of the blooms, explaining 84.6% of their variability over the past 23 yr. These factors are also good predictors of the variability in the duration of annual blooms and determined 58.9% of the variability in this parameter. Our results indicate that when nutrients are in sufficiently high quantities to sustain the formation of cyanobacterial blooms, climatic variables become crucial in predicting cyanobacterial bloom events. Climate changes should be considered when we evaluate how much the amount of nutrients should be reduced in Taihu for lake management. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Cyanobacterial blooms have become ubiquitous in many freshwater ecosystems affected by human activities, especially since the 1940s (Fogg, 1969; Huisman et al., 2005; Paerl and Fulton, 2006). Previous research has identified a host of causative factors spanning many ecological levels: bottom-up factors, such as nutrients; physiological capability of
cyanobacteria, including buoyancy regulation and low light requirements (Fogg, 1969; Zevenboom, 1982); the lack of or decrease in top-down predators on the community level (Porter, 1973); long water residence time (Paerl, 1996); and the altered structure and function of a whole water ecosystem (Elser, 1999). Recently, another factor, global warming, was found to promote the incidence of cyanobacterial blooms (Mooij et al., 2005; Jeppesen et al., 2007).
* Corresponding author. Tel.: þ86 25 86882183; fax: þ86 25 87714759. E-mail address:
[email protected] (F. Kong). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.013
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 2 e4 5 2
Anthropogenic climate change leads to warm weather, which in turn may promote the occurrence of cyanobacterial blooms directly and indirectly (Paerl and Huisman, 2009). High temperature (often above 25 C) can directly promote the growth of cyanobacteria relative to other phytoplankton groups, such as diatoms and green algae (Reynolds, 2006; Jo¨hnk et al., 2008; Paerl and Huisman, 2009). Indirectly, warm weather induces many changes that may affect ecosystem processes and characteristics. First, warm weather, by heating the surface water, creates a stable water column that is advantageous for cyanobacterial growth in stratified water (Charpin et al., 1998; Dokulil and Teubner, 2000). Second, warm weather can result in either less or more precipitation, either of which can be beneficial to the development of blooms. For example, high winter rainfall may increase nutrient supply to surface waters and promote summer blooms, and increased summer drought will decrease flushing and promote the persistence of blooms (Paerl and Huisman, 2008). Third, despite some uncertainties, models have predicted that global warming will weaken the equator-to-pole temperature gradient and thus lessen mean summer wind speeds (Wentz et al., 2007). Slow wind promotes water stability and facilitates the gathering of cyanobacteria at the water surface to form blooms (Cao et al., 2006). Finally, global warming and associated changes in local climate alter light availability to phytoplankton. For example, increased cloud cover with increasing aerosol absorptivity decreased the surface solar radiation for the period of approximately 1960e1990. In the following years, however, solar radiation increased at most of the locations for which good records exist due to the interplay of direct and indirect aerosol effects (Wild et al., 2005; Perlwitz and Miller, 2010). However, all these factors would considerably affect the photosynthetic physiology of phytoplankton (Charpin et al., 1998). Compared with other phytoplankton groups, most planktonic cyanobacterium are known for adaptation to low light conditions (Richardson et al., 1983; Reynolds, 1984) and for their ability to gain more light by migrating up to the water surface by buoyancy (Sherman and Webster, 1994), where they can more effectively resist photoinhibition than other algae (Zhang et al., 2008). Therefore, global warming would have the potential to change the timing and scale of cyanobacterial bloom events. Changes of biological events in aquatic ecosystems have been particularly apparent according to long-term ecological records and could serve as sensitive indicators of climate change (Winder and Schindler, 2004; Adrian et al., 2006; Blenckner et al., 2007; Smol, 2010). Climate-induced changes in the phenology of diatom blooms have been reported in Mu¨ggelsee, Germany (Huber et al., 2008). The shift in the timing of spring blooms is attributed to an advance in the timing of diatom dominance, mediated by an increase in growth rate driven by temperature (Meis et al., 2009). Quantitative analyses in Mu¨ggelsee suggest that climate warming enhances the probability of cyanobacterial dominance, and their incidence will certainly increase in aquatic systems in future warming climates (Wagner and Adrian, 2009). However, this lack of long-term ecological datasets in many lakes presents difficulties in proving the hypothesis that cyanobacterial blooms benefit from increased temperatures or other
443
climate factors under climate changes (Wagner and Adrian, 2009). Traditional methods of taking ship-borne water samples and analyzing them in a laboratory and/or performing on-site measurements are also incapable of such a spatial scale and frequency (Wang and Shi, 2008). Recently, remote sensing has become a powerful tool in understanding the long-term dynamics of algal blooms (Wang and Shi, 2008; Duan et al., 2009). From satellite-derived data, Duan et al. (2009) found that the onset time of cyanobacterial blooms has occurred earlier in the last decade, and their annual duration has extended in Taihu, China. Qin et al. (2010) noted that the advancement of cyanobacterial bloom events might be attributable to warming weather in the context of high nutrient levels. However, we sill need long-term observation data and statistical analysis results to confirm this point and to obtain detailed insight into the contribution of the meteorological factors associated with climate change to the phenology of cyanobacterial blooms. In this study, we tested the hypothesis that climatic variables play an important role in mediating bloom events within the context of high nutrient levels using long-term data derived from satellite images. To elucidate the relationship between climate changes and cyanobacterial onset timing as well as annual duration in Taihu, we first characterized climate changes in terms of temperature, precipitation, wind speed, global radiation, and sunshine hours from local weather stations for 1987e2009 along with the onset time and duration of annual blooms in the same period. Next, we analyzed the relationship between the climatic variables and bloom events using linear correlation analysis and the contribution of climatic variables to the changes in bloom events by multiple regression.
2.
Materials and methods
2.1.
Study site
Taihu is in southeast Jiangsu Province, China (latitude 30 550 4000 e31 320 5800 N; longitude 119 520 3200 e120 360 1000 E, Fig. 1). The lake has a surface area of 2338 km2, a maximum and average depth of 2.6 and 1.9 m, respectively, and a mean water residence time of approximately 309 days (Qin et al., 2004). Taihu receives inflows from nearby riverine networks, including over 200 streams, canals and rivers (Chen et al., 2003). Blooms of Microcystis have been dominant in the past few decades (Qin et al., 2007).
2.2.
Data acquisition, processing, and analyses
Events (i.e., the onset time and annual duration) of cyanobacterial blooms were obtained from the satellite images from 1987 to 2009, except 1988 and 1999. A total of 418 remote sensing images over Taihu from 1987 to 2009 were obtained, including 178 scenes of Landsat TM/ETM covering nearly all cloud-free periods since 1987 and 240 scenes of MODIS images from 2002 to 2009. The MODIS images were downloaded from the NASA EOS Data Gateway (EDG), and the Landsat data were provided by the China Remote-Sensing Satellite Ground Station. The bloom information was retrieved with the
444
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 2 e4 5 2
Fig. 1 e Location of Taihu in China. The weather station from which the climate data were collected is shown as a dot.
detectable reflectance signals from cyanobacterial cells in water primarily as described (Duan et al., 2009). Detectable reflectance signals were observed at a wavelength of 900 nm (TM and MODIS) from the algal blooms floating on the water surface and at wavelengths of 550 nm (TM) and 650 nm (TM and MODIS) from water with an abundance of algae (no scum). The distinguishable reflectance signal (different from the signal of clear water) can be used to delineate the areas of algal blooms by a threshold. To avoid underestimating the bloom information due to the long revisiting period (16 days) of Landsat and the unavoidable cloudy conditions, MODIS data with daily revisiting intervals were used to improve frequency and spatial accuracy, which was feasible to reconstruct the long-term information of cyanobacterial blooms (Duan et al., 2009). Although the threshold value based on visible images may have some uncertainties in determining the information of the blooms, it provides the possibility of mapping blooms over longer time scales and larger spatial scales without in situ data. In this study, we focused on two bloom aspects: the onset time and the annual duration of blooms. We analyzed the relationship between the onset time and the climate changes during the period from recruitment initiation (when cyanobacteria start to migrate from sediments after growth therein) to the onset of the blooms. The onset time was defined as the date when cyanobacterial blooms were first recorded by remote sensing. Changes in climatic variables, including maximum temperature, mean temperature, minimum temperature, wind speed, precipitation, sunshine hours, and global radiation, were each calculated as the difference between their daily average in the given year and the standard
value (the mean of the daily average of each variable from 1971 to 1980) during the period between the initiation of cyanobacterial recruitment and the bloom’s onset; these variables are represented as DTmax, DTmean, DTmin, DW, DP, DS and DG, respectively. Microcystis aeruginosa grow and transfer from the sediment surface to the water column when the water temperature reaches 9 C in early spring (Latour et al., 2004; Cao et al., 2008). Meanwhile, water temperature responds rapidly to changes in the air temperature in shallow lakes (Carpenter et al., 1992), so we approximately considered the air temperature as the water temperature. We also found that if the temperature exceeded 9 C for 10 consecutive days, it would hardly decrease to <9 C in the following days of the year. Therefore, we defined the recruitment initiation time as the day when the air temperature exceeded 9 C followed by 10 consecutive days, each with temperature no lower than 9 C. We also analyzed the relationships between annual bloom duration and annual climatic variables. The annual bloom duration was defined as the number of months per year during which the algal blooms were detected. Annual mean concentrations of total nitrogen and total phosphorus were derived from the statistical yearbooks of Jiangsu Province, which were analyzed and calculated by the Environmental Protection Agency of Jiangsu Province according to samples from different lake zones once a month (APHA, 1985). The data of nutrient loadings (1998e2007), primarily originated from Jiangsu, were obtained from Ma et al. (2010). The climate data, including temperature, sunshine hours, wind speed, precipitation, and global radiation, were obtained from meteorology station #58358 of the China Meteorological Administration (Fig. 1).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 2 e4 5 2
The relationships between the potential explanatory variables and the events variables of cyanobacterial blooms were analyzed using multiple ordinary least squares (OLS) regression. Before the analysis, all data were standardized at a mean of 0 and a standard deviation of 1. Analysis of Variance (ANOVA) was used to test the colinearity of variables. We tested the variance inflation factor (VIF), which is a good indicator of whether one predictor has a strong linear relationship with other predictors. Where the VIF was <10, we concluded that the predictor variables were not strongly related (Myers, 1990). The best models were identified using Akaike’s information criterion (AICc) (Fotheringham et al., 2002). The AICc values of the models with all possible combinations of predictors were compared as a model set. The OLS was performed using the software SAM 4.0. Residual analysis was then used to check the influential points and outliers and the verification of the applicability of the regression models.
3.
Results
3.1.
Changes in climates and nutrients
The annual maximum, mean, and minimum temperatures all showed warming trends, with an asymmetric increase in maximum and minimum temperature, each rising by 1.40 C, 1.41 C, and 1.42 C, respectively, from 1987 to 2009 (P < 0.01, Table 1). The annual precipitation had no significant trend (P > 0.05). The daily mean wind speed decreased dramatically, by 0.47 m/s, in the same period (P < 0.05, Table 1). The sunshine hours over the past 23 yr showed two patterns of change: from 1987 to 1999, it declined at a rate of 49.76 h per year until 2000 (P < 0.01, Table 1), but increased since 2000 by 50.08 h per year (P < 0.01, Table 1). The global radiation increased by 144.25 W/m2 after a steady rise from 1987 to 2009 (P < 0.01, Table 1).
Table 1 e Linear regression parameters of climate and nutrient variables against years.
Tmax Tmean Tmin Precipitation Wind speed Sun hours
Global radiation TN TP TN/TP TN loading TP loading
Years
Intercept
Slope
n
P
R2adj
1987e2009 1987e2009 1987e2009 1987e2009 1987e2009 1987e2009 1987e1999 2000e2009 1987e2009
1066.71 1116.95 1160.69 122,933.37 458.15 3036.44 84,429.93 98,506.88 11,891.87
0.64 0.64 0.65 55.63 0.21 2.44 41.47 50.08 6.56
23 23 23 23 23 23 13 10 23
0.000 0.000 0.000 0.405 0.000 0.705 0.001 0.008 0.005
0.46 0.48 0.46 0.01 0.54 0.04 0.59 0.52 0.29
1988e2008 1988e2008 1988e1996 1988e2008 1988e1996 1998e2007 1998e2007
38.17 2.05 21.659 497.07 11,428.73 1.82 22,126.71
0.02 0.01 0.01 0.23 5.72 927.88 11.88
19 20 9 19 7 10 10
0.386 0.242 0.001 0.593 0.004 0.310 0.712
0.01 0.02 0.77 0.04 0.80 0.13 0.02
445
The changes in climatic variables showed that the temperatures for most of each year studied were higher than the standard values. The DTmax, DTmean and DTmin temperatures, all showed significant increases by 4.14 C, 3.26 C, and 2.66 C, respectively, over the 23 yr, except for 1997 (P < 0.01, Fig. 2aeC). There was little change in DP, which was only significant at a ¼ 0.1 (P ¼ 0.09, Fig. 2d). Compared with the standard values, DW showed an overall decrease of 0.64 m/s over the 23 yr (P < 0.01, Fig. 2e). Changes in DS and in annual sunshine hours in the same period both followed a similar pattern: a decrease occurred from 1987 to 1999 (P < 0.01), followed by an increase since 2000 (P ¼ 0.02, Fig. 2f). DG showed a significant increase of 70.30 W/m2 from 1987 to 2009 (P < 0.01, Fig. 2g). Total nitrogen (1.64e3.53 mg/L), total phosphorus (0.05e0.14 mg/L), and the TN:TP ratio during the past two decades fluctuated with economic development and lake management (Supporting Information Fig. S1), and there were no clear trends from the correlation analyses. From 1988 to 1996, however, there was a significant increase in TP concentration (R2Adj ¼ 0.77, P ¼ 0.001, Table 1) and a significant decrease in the ratio of TN:TP from 1988 to 1996(R2Adj ¼ 0.80, P ¼ 0.004, Table 1). TN and TP loading did not increase significantly since the late 1980’s (Table 2), nor did they show any trends in the most recent ten years (Supporting Information Fig. S2).
3.2. Changes in the observed events of cyanobacterial blooms, 1987e2009 Over the 23 yr, the onset time showed two different trends (Fig. 3a): from 1987 to 1997, the bloom onset was delayed by 5 d per year (P < 0.01); and from 1998 to 2009, the bloom onset advanced by approximately 10 d per year (P ¼ 0.01). In addition, the recruitment initiation time of cyanobacteria advanced significantly with increases in mean annual temperature, by 0.9 d each year (P < 0.05). The time from recruitment initiation to bloom onset became increasingly shorter, especially in the recent 5 yr (2005e2009), when none were beyond 45 d. In 2006, in particular, blooms were found within nine days since recruitment initiation. The annual duration of cyanobacterial blooms only lasted for one month before 1998, but extended to more than two months since 1998, and ended up lasting over eight months from 2005 to 2009 (Fig. 3b).
3.3. Relationship between climate changes and cyanobacterial bloom events The onset time of cyanobacterial blooms was highly correlated with DTmax, DTmean, DTmin, DW, and DS (Fig. 4aec, e and f, P < 0.01), and slightly correlated with DG (Fig. 4g, P ¼ 0.04). No significant correlation was found between the onset time and DP (Fig. 4d, P > 0.05). Blooms duration had no remarkable linear relationship with the annual averages of climatic variables. However, the duration increased when the daily averages of maximum, mean, and minimum temperatures each exceeded 20.3 C, 16.7 C, and 13.7 C, respectively, and the annual sunshine hours exceeded 1650 h. The duration also increased as precipitation and wind speed
446
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 2 e4 5 2
Fig. 2 e Trends in daily average anomalies of maximum (a), mean (b), and minimum (c) temperatures, precipitation (d), wind speed (e), sunshine hours (f), and global radiation (g) with respect to the 1971e1980 mean values during the period from cyanobacterial recruitment initiation to bloom formation in Taihu. The line shows the linear regression against year.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 2 e4 5 2
Table 2 e Loading of total nitrogen and total phosphorus in Taihu from late 1980’s to early 2000’s.
Late 1980’s Early 1990’s Late 1990’s Early 2000’s
TN loading (103 t/a)
TP loading (103 t/a)
Source
20.24 31.00 25.34 28.66
1.55 1.75 1.33 1.03
(Sun and Huang, 1993) (SEPA, 2000) (Huang, 2004) (Xu and Qin, 2005)
decreased and as global radiation increased (Supporting Information Fig. S3). The two predictors we selected, DW and DS, in the predictive model for the onset time by multiple regression analysis, could explain 84.6% of the variability in the onset time (Table 3). The predicted values of the onset time using the
447
model agreed well with the observed values (Fig. 5a). The multiple regression for bloom duration showed that wind speed and sunshine hours were the main contributors, accounting for 58.9% of the variation in duration (Table 3). Other variables, including TN and TP concentration, loading and TN:TP ratio, annual maximum, mean, and minimum temperatures, annual precipitation, and global radiation, were not significantly relevant (P > 0.05). Furthermore, wind speed had a negative impact on bloom duration, while sunshine hours had a positive impact. In other words, the bloom duration would increase when wind speed decreased and sunshine hours were prolonged (Fig. 5b). The results of the residual analysis were then used for the verification of the applicability of the two regression models. The existence of influential points and outlier observations were checked, but not found, in the models.
Fig. 3 e The intervals between cyanobacterial recruitment initiation (bottom point of each bar) and bloom onset (top point of each bar) (a) and the duration of cyanobacterial blooms (b) in Taihu from 1987 to 2009. The dash dot lines are the regression lines of bloom onset (1987e1996, R2adj [ 0.501, P [ 0.013; 1997-2009, R2adj [ 0.560, P [ 0.003), and the solid line is the regression line of recruitment initiation (R2adj [ 0.154, P [ 0.044).
448
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 2 e4 5 2
Fig. 4 e Correlation between mean daily anomalies of climatic variables (maximum (a), mean (b) and minimum (c) temperatures, precipitation (d), wind speed (e), sunshine hours (f), and global radiation (g)) and onset time of cyanobacterial blooms.
4.
Discussion
The formation of cyanobacterial blooms is a result of complex and synergistic environmental factors rather than a single
dominant variable (Dokulil and Teubner, 2000). First, nutrients are considered the foundation for bloom formation (Huisman et al., 2005; Paerl and Fulton, 2006), which could influence long-term cyanobacterial relative biomass dynamics (Chen
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 2 e4 5 2
Fig. 5 e Scatterplots of the multiple linear regression analysis on onset time (a) and duration (b) of cyanobacterial blooms. The solid line represents regression, and the dotted line indicates the 95% confidence interval. All of the variables were standardized (mean [ 0; SD [ 1).
Table 3 e Relationships between initial blooming time, duration time of cyanobacteria blooms and potential explanatory variables modelled using multiple ordinary least squares (OLS) regression from 1987 to 2009. The best models were identified using Akaike’s information criterion (AICc). All of the variables were standardized (mean [ 0; SD [ 1). Model
R2
R2adj
Initial blooming time model Constant 0.846 0.839 DSunshine hours DWind speed Blooms duration model Constant 0.589 0.568 Wind speed Sunshine hours
AICc
n
b
P
29.173
23
<0.001 0.609 0.446
<0.001 <0.001
<0.001 0.598 0.419
<0.001 0.006
51.884
23
449
et al., 2003). Moreover, there is a general consensus that the magnitude and duration of the blooms increase with increasing nutrient loads (Rydin et al., 2002) and are controlled by limiting nutrients, which are exhausted by the growth of blooming algae (Xu et al., 2010; Paerl et al., 2011). In our case, however, the correlation analyses showed that TN and TP, along with their loadings, had no significant relationship with the onset time or duration of annual blooms. In fact, from 2005 to 2009, cyanobacterial blooms lasted more than eight months per year and became invisible until early winter, which indicates that a bloom’s duration is not shortened by the seasonal limitation of nutrients showed by Xu et al. (2010). Furthermore, multiple linear regression indicates that the TN and TP concentrations and loadings, and even the TN:TP ratios, could not predict the two cyanobacterial bloom events. Even, though the nutrient concentrations and loadings in Taihu increased dramatically from the1960s to 1980s, their increasing trends have slowed down since the 1990s. In the last ten years, the nutrient loads and concentrations have already been maintaining high levels (TP: 0.09 0.02 mg/L; TN: 2.63 0.55 mg/L), and have even exceeded cyanobacterial growth requirements in some lake regions (Qin et al., 2010). Furthermore, blooming algae also obtain nutrients to keep the persistence of blooms from the degradation of blooms (Sfriso et al., 1987) and the internal release of nutrient loads (Jo¨hnk et al., 2008). Thus, although nutrient enrichment is a prerequisite to bloom formation, the contribution of nutrients to the changes in the phenology of cyanobacterial blooms might be weak in Taihu over the 23 yr because the nutrients had reached high levels with low inter-annual variation. When nutrients are high enough to sustain algal blooms, the magnitude, spatial extent and duration of blooms are mainly modulated by physical factors (Qin et al., 2010). In our case, in the context of the high nutrient level, climatic variables were crucial modulating factors of cyanobacterial bloom events in Taihu over the 23 yr. The onset time of cyanobacterial blooms has a negative correlation with the changes in daily maximum, mean, and minimum temperature, sunshine hours, and global radiation, and it has a positive correlation with the changes in wind speed and precipitation. Moreover, the daily averages of minimum temperature, wind speed and sunshine hours are primary contributors to the advance of the onset time and the extension of the duration of cyanobacterial blooms. Among these climate factors, increased temperature promotes the growth of cyanobacteria and allow them to develop earlier (Wiedner et al., 2007). It has been predicted that global warming will increase the frequency of cyanobacterial blooms through the direct effects of high water temperatures (Jo¨hnk et al., 2008; Paerl and Huisman, 2008). However, in situ observational evidence is still required to confirm that cyanobacteria profit directly from increased water temperature (Adrian et al., 2006). In the present study, temperature displays a negative correlation with the onset time of cyanobacterial blooms, implying that low temperatures early in the year could delay cyanobacterial blooms. The formation of cyanobacterial blooms is also favoured by warm and calm weather, and disfavoured by windy and precipitating conditions (Kanoshina et al., 2003). Reduced wind speeds and precipitation are thus favourable for buoyant
450
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 2 e4 5 2
Microcystis to accumulate on the water surface (Jo¨hnk et al., 2008). A low wind speed of approximately 3e4 m/s (George and Edwards, 1976; Cao et al., 2006) is favourable for Microcystis to accumulate at the water surface, mainly at the surface to 0.3 m depth (Zhang et al., 2008). In Taihu, the falling wind speed increases the stability of the water column, thereby reducing vertical turbulent mixing, which may shift the competitive balance in favour of the buoyant cyanobacteria, promoting the formation of cyanobacterial blooms. Additionally, increasing global solar radiation may be more favourable for cyanobacteria because they have the major advantage of accessing light by accumulating on the water surface and casting shade upon other competitors (Walsby et al., 1997; Klausmeier and Litchman, 2001; Huisman et al., 2004). Furthermore, even at the water surface, cyanobacteria are a good competitor compared with other phytoplankton because they can resist photoinhibition from increasing global solar radiation more effectively than other algae by nonphotochemical quenching, when solar radiation intensity exceeds the light saturation point of phytoplankton (Zhang et al., 2008). The negative correlation between the sunshine hours and the onset time also indicates that increasing light duration favours the development of cyanobacterial dominance. Therefore, light conditions are one of the crucial determinants of the onset time and duration of cyanobacterial blooms.
5.
Conclusion
In the studied case, we used satellite images as a data source to obtain historical cyanobacterial bloom information in Taihu, and we analyzed their correlation with climatic variables. Our study demonstrates that retrieving ecological information from satellite images is meritorious for large scale and long-term ecological research in freshwater ecosystems, despite uncertainties due to technological limitations and the complexity of aquatic systems. Our findings highlight the importance of meteorological factors associated with climate change on the phenology of cyanobacterial blooms in Taihu during the past 23 yr. Within the context of sufficiently high nutrient loadings and concentrations for the formation of cyanobacterial blooms, climatic variables play an important role in mediating bloom events. Increased temperature directly affects cyanobacterial bloom events by advancing its onset time and extending its duration. Furthermore, wind speed and sunshine hours are also important contributors for advancing the onset time and prolonging the duration of cyanobacterial blooms in Taihu from 1987 to 2009. Therefore, our study offers evidence that although nutrients must be substantially reduced, climate changes should be considered when evaluating how much the amount of nutrients should be reduced in Taihu for future lake management.
Acknowledgements We thank Dr. Ronghua Ma for retrieving cyanobacterial bloom information from the satellite images, Dr. Huansheng Cao at
Fordham University for reviewing the manuscript, and the anonymous reviewers for their crucial comments. The work was supported by the project of Jiangsu Province National Science Foundation (BK2011877), the National Basic Research Program of China ‘973’ (No. 2008CB418005), and the project of the Front Fields Program (CXNIGLAS200810).
Appendix. Supplementary material Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011. 11.013.
references
Adrian, R., Wilhelm, S., Gerten, D., 2006. Life-history traits of lake plankton species may govern their phenological response to climate warming. Global Change Biology 12 (4), 652e661. APHA, 1985. Standard Methods for the Examination of Water and Wastewater, sixteenth ed. American Public Health Association, Washington, D. C. Blenckner, T., Adrian, R., Livingstone, D.M., Jennings, E., Weyhenmeyer, G.A., George, D., Jankowski, T., Jrvinen, M., Aonghusa, C.N., Noges, T., 2007. Large-scale climatic signatures in lakes across Europe: a meta-analysis. Global Change Biology 13 (7), 1314e1326. Cao, H.S., Kong, F.X., Luo, L.C., Shi, X.L., Yang, Z., Zhang, X.F., Tao, Y., 2006. Effects of wind and wind-induced waves on vertical phytoplankton distribution and surface blooms of Microcystis aeruginosa in Lake Taihu. Journal of Freshwater Ecology 21 (2), 231e238. Cao, H.S., Tao, Y., Kong, F.X., Yang, Z., 2008. Relationship between temperature and cyanobacterial recruitment from sediments in laboratory and field studies. Journal of Freshwater Ecology 23 (3), 405e412. Carpenter, S.R., Fisher, S.G., Grimm, N.B., Kitchell, J.F., 1992. Global change and freshwater ecosystems. Annual Review of Ecology and Systematics 23, 119e139. Charpin, M.F., Maurin, N., Amblard, C., Devaux, J., 1998. Seasonal variations of phytoplankton photosynthate partitioning in two lakes of different trophic level. Journal of Plankton Research 20 (5), 901e921. Chen, Y.W., Qin, B.Q., Teubner, K., Dokulil, M.T., 2003. Long-term dynamics of phytoplankton assemblages: Microcystisdomination in Lake Taihu, a large shallow lake in China. Journal of Plankton Research 25 (4), 445e453. Dokulil, M.T., Teubner, K., 2000. Cyanobacterial dominance in lakes. Hydrobiologia 438 (1), 1e12. Duan, H.T., Ma, R.H., Xu, X.F., Kong, F.X., Zhang, S.D., Kong, W.J., Hao, J.Y., Shang, L.L., 2009. Two-decade reconstruction of algal blooms in China’s Lake Taihu. Environmental Science & Technology 43 (10), 3522e3528. Elser, J.J., 1999. The pathway to noxious cyanobacteria blooms in lakes: the food web as the final turn. Freshwater Biology 42 (3), 537e543. Fogg, G.E., 1969. The physiology of an algal nuisance. Proceedings of the Royal Society of London Series B e Biological Sciences 173, 175e189. Fotheringham, A.S., Brunsdon, C., Charlton, M., 2002. Geographically Weighted Regression: The Analysis of Spatially Varying Relationships. John Wiley & Sons Inc, Chichester, UK.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 2 e4 5 2
George, D.G., Edwards, R.W., 1976. The effect of wind on the distribution of chlorophyll a and crustacean plankton in a shallow eutrophic reservoir. Journal of Applied Ecology 13 (3), 667e690. Huang, W., 2004. The pollutant budget in Lake Taihu. In: Qin, B.Q., Hu, W.P., Chen, W.M. (Eds.), Process and Mechanism of Environmental Changes of Lake Taihu. Science Press, Beijing, pp. 21e28 (in Chinese). Huber, V., Adrian, R., Gerten, D., 2008. Phytoplankton response to climate warming modified by trophic state. Limnology and Oceanography 53 (1), 1e13. Huisman, J., Matthijs, H.C.P., Visser, P.M., 2005. Harmful Cyanobacteria. Springer, Dordrecht, Netherlands. Huisman, J., Sharples, J., Stroom, J.M., Visser, P.M., Kardinaal, W.E.A., Verspagen, J.M.H., Sommeijer, B., 2004. Changes in turbulent mixing shift competition for light between phytoplankton species. Ecology 85 (11), 2960e2970. Jo¨hnk, K.D., Huisman, J., Sharples, J., Sommeijer, B., Visser, P.M., Stroom, J.M., 2008. Summer heatwaves promote blooms of harmful cyanobacteria. Global Change Biology 14 (3), 495e512. Jeppesen, E., Meerhoff, M., Jacobsen, B., Hansen, R., Søndergaard, M., Jensen, J., Lauridsen, T., Mazzeo, N., Branco, C., 2007. Restoration of shallow lakes by nutrient control and biomanipulationdthe successful strategy varies with lake size and climate. Hydrobiologia 581 (1), 269e285. Kanoshina, I., Lips, U., Leppa¨nen, J.M., 2003. The influence of weather conditions (temperature and wind) on cyanobacterial bloom development in the Gulf of Finland (Baltic Sea). Harmful Algae 2 (1), 29e41. Klausmeier, C.A., Litchman, E., 2001. Algal games: the vertical distribution of phytoplankton in poorly mixed water columns. Limnology and Oceanography 46 (8), 1998e2007. Latour, D., Sabido, O., Salencon, M.J., Giraudet, H., 2004. Dynamics and metabolic activity of the benthic cyanobacterium Microcystis aeruginosa in the Grangent reservoir (France). Journal of Plankton Research 26 (7), 719e726. Ma, Q., Liu, J., Gao, M., 2010. Amount of pollutants discharged into Lake Taihu from Jiangsu province, 1998e2007. Journal of Lake Science 22 (1), 29e34 (in Chinese). Meis, S., Thackeray, S.J., Jones, I.D., 2009. Effects of recent climate change on phytoplankton phenology in a temperate lake. Freshwater Biology 54 (8), 1888e1898. Mooij, W., Hu¨lsmann, S., De Senerpont Domis, L., Nolet, B., Bodelier, P., Boers, P., Pires, L., Gons, H., Ibelings, B., Noordhuis, R., Portielje, R., Wolfstein, K., Lammens, E., 2005. The impact of climate change on lakes in the Netherlands: a review. Aquatic Ecology 39 (4), 381e400. Myers, R., 1990. Classical and Modern Regression with Applications. PWS and Kent Publishing Company, Boston. Paerl, H.W., 1996. A comparison of cyanobacterial bloom dynamics in freshwater, estuarine and marine environments. Phycologia 35, 25e35. Paerl, H.W., Fulton, R.S., 2006. In: Grane´li, E., Turner, J.T. (Eds.), Ecology of Harmful Algae. Springer, Berlin Heidelberg, pp. 95e109. Paerl, H.W., Huisman, J., 2008. Climate: blooms like it hot. Science 320 (5872), 57e58. Paerl, H.W., Xu, H., McCarthy, M.J., Zhu, G., Qin, B., Li, Y., Gardner, W.S., 2011. Controlling harmful cyanobacterial blooms in a hyper-eutrophic lake (Lake Taihu, China): the need for a dual nutrient (N & P) management strategy. Water Research 45 (5), 1973e1983. Paerl, W.H., Huisman, J., 2009. Climate change: a catalyst for global expansion of harmful cyanobacterial blooms. Environmental Microbiology Reports 1 (1), 27e37. Perlwitz, J., Miller, R.L., 2010. Cloud cover increase with increasing aerosol absorptivity: a counterexample to the conventional
451
semidirect aerosol effect. Journal of Geophysical Research 115 (D8), D08203. Porter, K.G., 1973. Selective grazing and differential digestion of algae by zooplankton. Nature 244, 179e180. Qin, B.Q., Zhu, G.W., Gao, G., Zhang, Y.L., Li, W., Paerl, H., Carmichael, W., 2010. A drinking water crisis in Lake Taihu, China: linkage to climatic variability and lake management. Environmental Management 45 (1), 105e112. Qin, B.Q., Hu, W.P., Chen, W.M., 2004. Process and Mechanism of Environmental Changes of the Taihu Lake. Science Press, Beijing. Qin, B.Q., Xu, P.M., Wu, Q.L., Luo, L.C., Zhang, Y.L., 2007. In: Qin, B., Liu, Z., Havens, K. (Eds.), Eutrophication of Shallow Lakes with Special Reference to Lake Taihu, China. Springer, Netherlands, pp. 3e14. Reynolds, C.S., 1984. The Ecology of Freshwater Phytoplankton. Cambridge Univ. Press, Cambridge, UK. Reynolds, C.S., 2006. Ecology of Phytoplankton (Ecology, Biodiversity and Conservation). Cambridge Univ. Press, Cambridge, UK. Richardson, K., Beardall, J., Raven, J.A., 1983. Adaptation of unicellular algae to irradiance: an analysis of strategies. New Phytologist 93 (2), 157e191. Rydin, E., Hyenstrand, P., Gunnerhed, M., Blomqvist, P., 2002. Nutrient limitation of cyanobacterial blooms: an enclosure experiment from the coastal zone of the NW Baltic proper. Marine Ecology-progress Series 239, 31e36. Sfriso, A., Marcomini, A., Pavoni, B., 1987. Relationships between macroalgal biomass and nutrient concentrations in a hypertrophic area of the Venice Lagoon. Marine Environmental Research 22 (4), 297e312. Sherman, B.E., Webster, I.T., 1994. A model for the light-limited growth of buoyant phytoplankton in a shallow, turbid waterbody. Australian Journal of Marine and Freshwater Research 45 (5), 847e862. Smol, J.P., 2010. The power of the past: using sediments to track the effects of multiple stressors on lake ecosystems. Freshwater Biology 55, 43e59. State Environmental Protection Administration (SEPA), 2000. Water Protection in the Three Rivers and Three Lakes of China. Chinese Environmental Press, Beijing, pp. 151e175 (in Chinese). Sun, S.C., Huang, Y.P., 1993. Lake Taihu. China Ocean Press, Beijing, pp. 23e89 (in Chinese). Wagner, C., Adrian, R., 2009. Cyanobacteria dominance: quantifying the effects of climate change. Limnology and Oceanography 54, 2460e2468. Walsby, A.E., Hayes, P.K., Boje, R., Stal, L.J., 1997. The selective advantage of buoyancy provided by gas vesicles for planktonic cyanobacteria in the Baltic Sea. New Phytologist 136 (3), 407e417. Wang, M., Shi, W., 2008. Satellite-observed algae blooms in China’s Lake Taihu. Eos Transactions, American Geophysical Union 89 (22), 201. Wentz, F.J., Ricciardulli, L., Hilburn, K., Mears, C., 2007. How much more rain will global warming bring? Science 317 (5835), 233e235. Wiedner, C., Ru¨cker, J., Bru¨ggemann, R., Nixdorf, B., 2007. Climate change affects timing and size of populations of an invasive cyanobacterium in temperate regions. Oecologia 152 (3), 473e484. Wild, M., Gilgen, H., Roesch, A., Ohmura, A., Long, C.N., Dutton, E.G., Forgan, B., Kallis, A., Russak, V., Tsvetkov, A.H., 2005. From dimming to brightening: decadal changes in solar radiation at earth’s surface. Science 308 (5723), 847. Winder, M., Schindler, D.E., 2004. Climate change uncouples trophic interactions in an aquatic ecosystem. Ecology 85 (8), 2100e2106.
452
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 2 e4 5 2
Xu, H., Paerl, H.W., Qin, B., Zhu, G., Gao, G., 2010. Nitrogen and phosphorus inputs control phytoplankton growth in eutrophic Lake Taihu, China. Limnology and Oceanography 55 (1), 420e432. Xu, P.Z., Qin, B.Q., 2005. Water quantity and pollutant fluxes of the surrounding rivers of Lake Taihu during the hydrological year of 2001e2002. Journal of Lake Science 17 (3), 213e218 (in Chinese).
Zevenboom, W., 1982. N2-fixing cyanobacteria: why they do not become dominant in shallow hypertrophic lakes. Aquatic Ecology 16 (2), 289e290. Zhang, M., Kong, F.X., Wu, X.D., Xing, P., 2008. Different photochemical responses of phytoplankters from the large shallow Taihu Lake of subtropical China in relation to light and mixing. Hydrobiologia 603 (1), 267e278.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 5 3 e4 6 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Complete nutrient recovery from source-separated urine by nitrification and distillation K.M. Udert*, M. Wa¨chter Eawag, Swiss Federal Institute of Aquatic Science and Technology, U¨berlandstrasse 133, 8600 Du¨bendorf, Switzerland
article info
abstract
Article history:
In this study we present a method to recover all nutrients from source-separated urine in
Received 9 June 2011
a dry solid by combining biological nitrification with distillation. In a first process step,
Received in revised form
a membrane-aerated biofilm reactor was operated stably for more than 12 months,
12 August 2011
producing a nutrient solution with a pH between 6.2 and 7.0 (depending on the pH set-
Accepted 5 November 2011
point), and an ammonium to nitrate ratio between 0.87 and 1.15 gN gN1. The maximum
Available online 17 November 2011
nitrification rate was 1.8 0.3 gN m2 d1. Process stability was achieved by controlling the pH via the influent. In the second process step, real nitrified urine and synthetic solutions
Keywords:
were concentrated in lab-scale distillation reactors. All nutrients were recovered in a dry
Membrane-aerated biofilm reactor
powder except for some ammonia (less than 3% of total nitrogen). We estimate that the
(MABR)
primary energy demand for a simple nitrification/distillation process is four to five times
Evaporation
higher than removing nitrogen and phosphorus in a conventional wastewater treatment
Primary energy demand
plant and producing the equivalent amount of phosphorus and nitrogen fertilizers.
Nutrient recovery
However, the primary energy demand can be reduced to values very close to conventional
Fertilizer
treatment, if 80% of the water is removed with reverse osmosis and distillation is operated
Ammonium nitrate
with vapor compression. The ammonium nitrate content of the solid residue is below the
NoMix technology
limit at which stringent EU safety regulations for fertilizers come into effect; nevertheless, we propose some additional process steps that will increase the thermal stability of the solid product. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Providing adequate sanitation is a major challenge for fast growing cities, especially in emerging and developing countries. However, the current technology of choice e waterborne conveyance of human excreta and centralized wastewater treatment e is inappropriate for many cities, because it requires large initial investments and high amounts of water (Corcoran et al., 2010). Many researchers and practitioners agree that new designs and technologies are needed, which focus on the recovery of resources such as water, nutrients and energy instead of merely preventing pollution (Guest
et al., 2009). One very promising approach is the separation of wastewater streams and their specific treatment in decentralized reactors. Urine is of particular interest, because it contains most of the nutrients, which could be used in fertilizers (Larsen et al., 2009). Since fertilizers have a market value, separate collection and treatment of urine has the potential to stimulate private business initiatives, which help to promote sanitation (Kone´, 2010). Direct application of urine as a fertilizer has been common in many rural societies around the world, but this straightforward technique is unsuitable for modern cities: the high water content makes the transport to agricultural land costly,
* Corresponding author. Tel.: þ41 44 823 53 60; fax: þ41 44 823 53 89. E-mail addresses:
[email protected] (K.M. Udert),
[email protected] (M. Wa¨chter). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.020
454
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 5 3 e4 6 4
ammonia volatilization from stored urine is unpleasant and causes high nitrogen losses, and pathogens pose a health risk to farmers and consumers of agricultural products (Udert et al., 2006). Maurer et al. (2006) reviewed a multitude of technologies, which could be used to produce a more convenient urine fertilizer. So far, struvite (MgNH4PO4$6H2O) has received the most attention (Etter et al., 2011): by adding a suitable magnesium source, more than 90% of the phosphate and some ammonia are recovered, but most ammonia and basically all potassium and sulfate stay in solution. Based on fertilizer prices in Nepal, Etter et al. (2011) calculated that struvite production only allows to recover 13% of the monetary fertilizer value of urine. Physical water removal is the only process that allows to recover all urine nutrients in one final product. Maurer et al. (2006) described four processes to separate water and dissolved compounds: electrodialysis, reverse osmosis, freeze/ thaw concentration and distillation. In the present paper, we will show that distillation is the most promising of the four processes. In previous studies, distillation was studied with fresh and stored urine. With both solutions, the urine was pretreated with strong acid to prevent ammonia volatilization. The acid addition to fresh urine inhibited the urea hydrolyzing enzyme urease (Maurer et al., 2003). In stored urine, acid converted free ammonia into non-volatile ammonium (Ek et al., 2006; Tettenborn et al., 2007). To recover 95% of the nitrogen during distillation, Ek et al. (2006) added 13 g H2SO4 to 1 L of stored urine and reached a pH value of 4.5. Since urea hydrolysis increases the alkalinity approximately by a factor of 20 (Udert et al., 2006), significantly more acid has to be added to stored urine than to fresh urine. However, urea hydrolysis is a very fast process in urine-collection systems (Udert et al., 2003a), so that the strong acid would already have to be added in the toilet, which is as dangerous as technically challenging. An alternative method for stabilizing nitrogen is biological nitrification. The alkalinity of stored urine allows to nitrify about 50% of ammonium to nitrate, while the pH decreases to values around 6.3 (Udert et al., 2003b). Nearly all ammonium can be oxidized to nitrate if additional alkalinity is added (Oosterhuis and Loosdrecht, 2009; Feng et al., 2008). Biological nitrification requires a well tuned interplay of ammonia oxidizing bacteria (AOB) and nitrite oxidizing bacteria (NOB), which is very challenging in high strength ammonia solutions such as urine (Udert et al., 2003b). Alternatively, nitrification can be split up into two processes: biological ammonium oxidation and chemical nitrite oxidation. Chemical nitrite oxidation with oxygen occurs at low pH values, which can either be reached with acid-tolerant ammonia oxidizers (Udert et al., 2005) or by adding acid (Verhave et al., 2007). The drawback of the process is the substantial volatilization of nitrogen oxides. Another alternative is the electrochemical oxidation of nitrite on graphite electrodes. This method has been successfully tested in the lab, though as yet, no longterm experiments exist (Faraghi et al., 2011). The aim of this paper is to test the hypothesis that a combination of biological nitrification and distillation is a technically suitable and energetically sensible process for complete nutrient recovery from urine. We will also discuss the chemical stability of the final product, which mainly consists of ammonium nitrate.
2.
Materials and methods
2.1.
Urine solutions
The urine for the nitrification experiments was taken from the men’s tank of the NoMix system in the Eawag main building (Forum Chriesbach). The average concentration of the urine solution is given in Table 1. Although the urine was practically not diluted with flushing water, nitrogen concentrations were significantly lower than in fresh urine due to ammonia volatilization (Goosse et al., 2009). Distillation experiments were conducted with effluent from the nitrification reactor, spiked effluent and synthetic solutions. Typical concentrations are given in Table 3. In the spiked solution, ammonium nitrate was added to correct for ammonia volatilization. Synthetic urine solutions were used to evaluate the effect of the pH value on nitrogen loss during distillation. The basic synthetic urine had a pH of 6.0. The recipe for the synthetic solution is given in the Supplementary Information (Table S1). From this stock four additional solutions with pH values of 6.3, 6.6, 6.9 and 7.2 were prepared. The pH values were adjusted with 2 M NaOH. All chemicals used for the synthetic solutions were of pro analysi grade.
2.2.
Nitrification reactor
A hybrid membrane-aerated biofilm reactor (MABR) was used for the nitrification experiment. The reactor had a liquid volume of 2.6 L. Silicon tubing coiled on a PVC rack (Supplementary Information, Figures S1 and S2) served for diffusive aeration and as substratum for biofilm. The tubing had a length of 21.7 m, a diameter of 4 mm and a wall thickness of 0.5 mm. The total surface available for biofilm growth in the reactor was 0.39 m2, thereof 0.25 m2 or 65% on silicone tubing. The surface to volume ratio was 149 m1. In addition to the membrane aeration, air was also supplied as bubbles through an air diffuser, which was attached to the lower end of the silicone tubing. The aim of combining silicone and bubble aeration was to minimize denitrification in the biofilm by
Table 1 e Average concentrations in the nitrification reactor influent (real stored urine). Average 1
Total Ammonia-N Total Phosphate-P Calcium Magnesium Potassium Sodium Sulfate Chloride Total inorganic C Diss. organic C Dissolved COD COD/N ratio [gO2 gN1] pH [e]
Std. Dev.
[mg L ]
[mg L1]
2390 208 16 <5 1410 1740 778 3210 1210 1830 4500 1.9 8.69
250 49 3 e 320 360 184 530 220 360 910 0.5 0.11
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 5 3 e4 6 4
supplying air from the substratum and the bulk. During the experiment reported in this paper, the air flow was fixed to 450 mL min1. The temperature in the reactor (23 2 C) was not controlled. The oxygen transfer coefficient KLa was determined to be 430 20 d1 (average and standard deviation of three experiments).
2.3.
Nitrification experiments
After some preliminary tests, the reactor was emptied, cleaned and reinoculated with activated sludge from the Eawag experimental wastewater treatment plant. Stored urine from the Forum Chriesbach was supplied continuously and half of the ammonium was oxidized to nitrate, but the pH had to be controlled frequently with carbon dioxide aeration to prevent process instabilities due to high pH values. Two months after inoculation, an alternative pH control scheme was started: the high buffer capacity and pH value of the influent urine was used to correct for the pH decrease caused by nitrification: as soon as the pH raised 0.1 units below a certain set-point, the influent pump was switched on, until the set-point was reached again. This control mechanism was automated with the process control tools of the data-logger (see below). After an initial phase with set-point 6.1 the actual experiment phase e a series with pH set-points 6.9, 6.7, 6.5, 6.3 and 6.1 e was run to test the influence of the pH value on the nitrification performance. The reactor was operated at each pH set-point for at least three hydraulic retention times. The water loss in the system was kept low (4.3 6.7%) by moisturizing the aeration air in a wash bottle before introducing it into the reactor.
To estimate the boiling point elevation during the distillation process, 1 L (1010.2 g solution) of a synthetic solution of nitrified urine (pH 6.0, Supplementary Information Table S1 and Table 3, but with pH adjusted to 6.0) was distilled in a custom-made glass reactor equipped with a reflux condenser, a condenser to trap the distilled water, a thermometer inside the reactor and a balance to weigh the distilled water (Labo-Tech LTS AG, Muttenz, Switzerland) at ambient pressure (approximately 970 mbar in Zurich). The water temperature during distillation was measured frequently until 99.2% of the water had evaporated.
2.5.
Thermal stability measurements
The thermal stability of the solid residues of a spiked nitrified urine sample, a commercial ammonium nitrate fertilizer (27% nitrogen, Landor, Birsfelden, Switzerland) and pure ammonium nitrate (Catalogue number 1.01188.1000, pro analysi, Merck, Darmstadt, Germany) was compared with differential scanning calorimetry (DSC). The measurements were carried out on an HP DSC827e (Mettler Toledo, Greifensee, Switzerland) under argon atmosphere at ambient pressure. The temperature range was 25e500 C and the heating rate 4 C min1. To homogenize the samples, the dry solids were milled and sieved. The diameter of the final particles was smaller than 100 mm. About 1 mg of these particles was placed in gold coated crucibles (M20, Swissi, Basel, Switzerland) and compacted with a closing force of at least 4 kN. The measured power curves were integrated with the integral tangential method implemented in the STARe Software Version 8.10 (Mettler Toledo, Greifensee, Switzerland).
2.6. 2.4.
455
Chemical and physical analyses
Distillation experiments
All distillation experiments except for the boiling point elevation experiment (see below) were performed at low pressure in a Bu¨chi Rotavapor R-124 with a Bu¨chi Water bath B-480 (Bu¨chi Labortechnik AG, Flawil, Switzerland). Tapwater was used for the cooling coil. The initial volume in the distillation reactor was approximately 200 mL, for both synthetic and real urine solutions. To guarantee the reproducibility and comparability of the results, the experiments in the Rotavapor were carried out according to the following procedure: 1.) the pressure in the system was lowered to 200 mbar and then the water bath was started to heat up from approximately 20 Ce78 C (duration: 19 min); 2.) water was distilled at 78 C and 200 mbar for 1 h (real urine) or 2 h (synthetic urine); 3.) the pressure was lowered for 5 min to 150 mbar; 4.) the pressure was lowered to 100 mbar for another 5 min; 5.) the water bath was removed and the system was brought to ambient pressure. In all experiments, the solids crystallized in step 4.) or earlier. The humid end product, which had approximately 2% of the initial volume, was dried completely in a desiccator under reduced air pressure and with silica gel as desiccant. Later, the dried solids were dissolved in nanopure water and samples were taken from this solution and the condensate to analyze their composition wetchemically.
The concentrations of nitrate, nitrite, phosphate, sulfate and chloride were measured with an ion chromatograph (IC 881 Compact IC pro, Methrom, Zofingen, Switzerland). Sodium, potassium, calcium and magnesium were analyzed with optical emission spectroscopy (OES) with inductively coupled plasma (ICP) (ICP-OES, Ciros, Spectro Analytical Instruments, Kleve, Germany). Cuvet tests (LCK 614, Hach-Lange, Berlin, Germany) were used to measure chemical oxygen demand (COD), while total organic carbon (TOC), dissolved organic carbon (DOC) and total inorganic carbon (TIC) were determined on a TOC-TN analyzer (IL 550 TOC-TN, Hach-Lange, Berlin, Germany). Ammonium was measured photometrically on a flow injection analyzer (Application Note 5520, FOSS, Hillerød, Denmark). The free ammonia (NH3) concentrations were simulated with the computer program PHREEQC Version 2.15.0.2697 using the “phreeqc.dat” database. Electrodes and handheld meters were used to measure pH (pH 191 and pH 340 meters, Sentix 20 electrode, WTW, Weilheim, Germany) as well as oxygen and temperature (Oxi 340 meter, CellOx 325 electrode, WTW, Weilheim, Germany). Online measurements of pH and oxygen were recorded on a data-logger (Memograph S, EndressþHausser, Reinach, Switzerland). The standard deviations of all physical and chemical analyses were 2% or less.
456
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 5 3 e4 6 4
3.
Nitrification
3.1.
Stable nitrate production
The nitrification reactor was operated stably for more than twelve months. The only exception was one short period with nitrite accumulation, which will be discussed in Section 3.5. Since nitrification generally consumes twice as much alkalinity (2 mol alkalinity per 1 mol ammonium) as is produced by urea hydrolysis (Udert et al., 2003a), about 50% of the ammonia was oxidized to nitrate. The ratio of nitrate to ammonium was slightly lower above pH set-point 6.5 and slightly higher below (Table 2, Fig. 1). Independent of the pH set-point, 90% of the organic substances were degraded (90 26% TOC, and 89 27% COD). Influent control alone was a good means to keep the pH stable at a certain value in the reactor. In a previous study (Udert et al., 2003b), we reported that stable nitrification is possible without any pH control at all, but process instabilities are more likely. Influent control can also increase the nitrification rate by raising the pH value in the reactor. Other studies described the use of base for pH control (Oosterhuis and Loosdrecht, 2009; Feng et al., 2008). In these studies, stable nitrification was only one of the two motivations for the use of chemicals: adding base also increased the alkalinity, so that all ammonia could be oxidized. However, the use of chemicals is costly and requires additional instrumentation. Both requirements make the process more complex and prone to malfunction. Therefore, a reactor setup without continuous base dosage is preferable.
3.2.
pH influence on nitrification rate
The nitrification rate correlated strongly with the pH value in the reactor (Fig. 1). This pH dependency can be explained with substrate limitation for AOB. Since free ammonia NH3 is generally assumed to be the true substrate of AOB (Suzuki
et al., 1974) we fitted the nitrification rates as function of free ammonia with Monod kinetics. The fit was very good (R2 ¼ 0.98, Figure S3 Supplementary Information), with a half saturation constant of 0.66 mgN L1 (maximum nitrification rate of 2.3 gN m2 d1), which is close to the half saturation constant (0.46 mgN L1) that Hellinga et al. (1999) proposed in their model for biological nitrogen transformation in high-strength ammonium wastewaters. The higher nitrification rates at higher pH values did not substantially change the product quality. While the nitrification rate doubled between pH 6.2 and pH 7.0, the average nitrate to ammonium ratio decreased only by 24% and was always in a range of 1.00 0.15 gN gN1. Higher pH set-points, however, had other adverse effects, which will be discussed in the following section.
3.3.
Nitrogen loss
Substantial nitrogen losses occurred at high pH set-points. The maximum nitrogen loss was 24% at a pH set-point of 6.7 (Table 2). Ammonia volatilization cannot explain this high nitrogen loss, because other aerated bioreactors treating urine did not experience nitrogen losses in the same pH range (Udert et al., 2003b). The most likely cause is heterotrophic denitrification. At high pH values, the ammonia and COD load in the inflow was so high that the oxygen supply became limiting in some parts of the biofilm. Oxygen limitation can also explain the nitrite accumulation, which we observed at high pH set-points. Nitrogen loss is one of the disadvantages of this reactor setup: to prevent anoxic zones, oxygen was supplied concomitantly from the substratum and the bulk, but the shear stress was too small to prevent the growth of thick biofilms with anoxic zones between substratum and bulk.
3.4.
Comparison with a moving bed biofilm reactor
A moving bed biofilm reactor (MBBR) in an earlier study (Udert et al., 2003b), performed significantly better than the MABR
Table 2 e Performance data of the nitrification reactor for different pH set-points. The nitrification rate is given per total surface available for biofilm growth (on tubing, stand and reactor wall). The nitrification rate was calculated using the difference between ammonia load in the inflow and in the outflow. Multiplication with 149 mgN LL1 m2 gNL1 gives the nitrification rate per fluid volume (in mgN LL1 dL1). The nitrogen loss was calculated by comparing the ammonia load in the influent and the sum of nitrite, nitrate and ammonia load in the effluent. pH setpoint
Measured pH [e] Average
Std. dev.
6.9 6.7 6.5 6.3 6.1 pH setpoint
6.99 6.77 6.56 6.36 6.17 Nitrogen loss [%]
6.9 6.7 6.5 6.3 6.1
Average 20 24 23 17 0
0.08 0.10 0.12 0.07 0.10
Std. dev. 14 9 8 4 9
Nitrification rate [gN m2 d1]
Nitrate/Ammonia [gN gN1]
Average
Average
1.8 1.6 1.4 1.2 0.9 Nitrite in bulk [mgN L1] Average 37 51 17 13 5
Std. dev. 0.3 0.4 0.1 0.2 0.2
Std. dev. 21 26 1 3 1
0.87 0.92 1.00 1.04 1.15 Oxygen in bulk [mg L1] Average 3.0 3.1 4.0 5.0 5.5
Std. dev. 0.06 0.06 0.06 0.06 0.14
Std. dev. 0.7 0.9 0.5 0.7 0.5
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 5 3 e4 6 4
457
storage and urine preparation. The high COD/N ratio for the MABR is due to ammonia volatilization, while the low value for the MBBR was caused by addition of ammonia to diluted urine. In stored urine without ammonia loss, one would expect a COD to ammonia ratio of about 1.2 gCOD gN1, which lies between the values for the MABR and the MBBR (Udert et al., 2006).
3.5.
Fig. 1 e Comparison of the nitrification rate (R2 [ 0.99) and the nitrate to ammonium ratio (R2 [ 0.97) in the nitrification reactor at different pH set-points. Error bars depict standard deviations.
presented in this paper. At an average pH value of 6.3, the MBBR achieved a surface specific nitrification rate of 1.7 0.1 gN m2 d1, while the MABR only reached 1.2 0.2 gN m2 d1. The volumetric nitrification rate was higher as well: 380 mgN L1 d1 compared to 180 30 mgN L1 d1. Additionally, no nitrogen loss was detected in the MBBR. The lower performance of the MABR is partly due to the reactor setup, which allowed biomass to accumulate. In the MBBR, strong mixing prevented thick biofilms on the biofilm carriers. Another reason for the lower performance of the MABR, especially for the high nitrogen loss, was the higher COD to ammonia ratio in the influent. The MABR was fed with urine that had a ratio of 1.9 0.5 gCOD gN1, while the urine used in the MBBR experiment had a COD/N ratio of only 0.47 0.27 gCOD gN1. The high COD/N ratio in the influent to the MABR supported strong heterotrophic growth. The strong differences in the COD/N ratio can be explained with processes during urine
Process instabilities
The activities of NOB and AOB strongly depend on the pH value, because the concentrations of their substrates of AOB (NH3) and NOB (HNO2) are in a pH dependent equilibrium with their acid (NHþ 4 ) or base (NO2 ). NH3 and HNO2 cannot only be limiting but also inhibiting (Anthonisen et al., 1976; Wiesmann, 1994). Additionally, AOB can be inhibited by their product HNO2 (Hunik et al., 1992), while inhibition of NOB by their product NO 3 is usually negligible (Hunik et al., 1993). The complex pH dependency of nitrifiers is particularly pronounced in urine treatment, because of the high substrate concentrations and the large pH difference between influent and effluent. Fig. 2 exemplifies that a sudden increase of the pH can be detrimental for urine nitrification. After the reactor had been operated for several weeks at pH 6.1, the pH set-point was raised to 6.9. Due to the higher availability of NH3 at higher pH values, AOB activity increased immediately. NOB reacted too slowly to the rise of nitrite, probably because of oxygen limitation. Nitrite kept on increasing and the HNO2 concentrations approached values that would have caused complete NOB inhibition. To prevent process breakdown, inflow and aeration were switched off and acetate added to remove nitrite (and nitrate) by denitrification. The concomitant pH increase was counteracted by dosing hydrochloric acid. As soon as all nitrite was removed, inflow and aeration were switched on again. In the following days, the process recovered and nitrate approached former levels, while nitrite concentrations stayed low. As this example shows, a sudden increase of the reactor pH value endangers process stability. Therefore, the pH
Fig. 2 e Nitrite build-up after a sudden increase of the pH set-point from 6.1 to 6.9. The nitrite was removed and the process stabilized by switching off the air and adding acetate and acid.
458
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 5 3 e4 6 4
control via the influent has to be very reliable. Furthermore, an automated monitoring system would be useful, which detects and announces sudden increases of the influent rate and of the pH value. pH decreases, for example caused by an inflow stop, are less critical, because ammonium oxidation usually comes to a halt at pH values of 6 or slightly below (Udert et al., 2003b). Rare exceptions are reactors, where acidtolerant AOB are present (Udert et al., 2005). However, according to our current knowledge, the acid-tolerant AOB only occur, when the reactor is operated for several weeks at pH values below 6.
4.
Distillation
4.1.
Mass balance
The recovery of most dissolved compounds in the solid residue was very high, regardless of whether synthetic solutions or real nitrified urine was used (Table 3). Only total inorganic carbon was lost completely due to the strong pH decrease to values around 4 (data not shown). In the real nitrified urine, organic compounds were reduced by 23%. Visual inspection of the glass bulb used in the experiment revealed a dark brown, presumably organic layer on the glass wall during the distillation. The transformation of the organic substances could cause scaling problems in large scale operation, but it also shows the possibility of targeted removal of organic substances during the distillation step. For all other dissolved compounds, losses due to distillation or scaling were negligible, considering that the minimum standard deviation for all analytical methods was 1%. However, in experiments with solutions that had a higher initial pH value, ammonia volatilization was relevant, as we will discuss in Section 4.2. Foaming was not a problem, neither with synthetic nor with real urine. Other researchers (Tettenborn et al., 2007; Ek et al., 2006) experienced foaming when distilling urine that had not been treated biologically. We assume that foaming is mainly caused by organic substances, which are removed during the nitrification step (see Section 3.1).
4.2.
Ammonia loss
The composition of the condensate was analyzed for trace compounds. As was expected, only ammonia and carbonate were detected (Fig. 3 and Figure S5 in the Supplementary Information). Although the mass balances in Table 3 did not reveal that significant amounts of ammonia volatilized, ammonia concentrations in the condensate were above 40 mgN L1. The ammonia concentrations in the condensate correlated strongly with the pH value in the initial synthetic nitrified urine (Fig. 3). When the initial pH value was 7.2, about 3% of the total nitrogen was captured in the condensate, while at an initial pH of 6.0, only 1.5% was captured. The low losses can be explained with a steady pH decrease to values below 4 (data not shown). Experiments with real nitrified urine solutions confirm the values obtained with synthetic nitrified urine. For example, the condensate of the experiment with real urine given in Table 3, contained
Fig. 3 e Ammonia in the condensate as function of the pH value in the initial synthetic nitrified urine. Initial composition of the synthetic nitrified urine according to Table 3 and Table S1 (Supplementary Information) except for pH and sodium (NaOH used for pH adjustment). The slope of the linear regression curve is 65 ± 25 mgN LL1. A correlation between the ammonia in the condensate and the initial pH of the nitrified urine is highly probable ( pvalue 0.085 for zero slope hypothesis). The dashed lines indicate the 95% confidence intervals.
approximately 40 mgN L1, which corresponds to less than 0.6% nitrogen loss. The nitrogen loss we recorded for nitrified urine was lower than in experiments with acidified stored urine. Using reverse osmosis to reduce the volume by a factor of five, Ek et al. (2006) reported 9% nitrogen loss with stored urine acidified to pH 7, and 2% nitrogen loss, when the pH value was lowered to 6. The same authors found higher nitrogen loss when more water was removed with the help of distillation (volume reduction by a factor 20). Here, the nitrogen losses were 11% (acidification to pH 5.5) and 5% (acidification to pH 4.5). More significant than the losses during urine treatment is the ammonia volatilization during urine storage. A study in Forum Chriesbach showed that more than 50% of the ammonia was lost by volatilization (Goosse et al., 2009). Ammonia volatilization and not dilution is probably the reason why nitrogen concentrations in stored urine are generally much lower than in fresh urine (see e.g. Etter et al., 2011). The condensate can be a second valuable product besides the nutrient-rich solid residue, because it contains hardly any dissolved compounds. Possibly, purified water for technical applications could be produced with only little additional treatment. According to our current knowledge, the main impurity is ammonia, which could be removed by for example electrolysis (Kapalka et al., 2010a, b) or stripping.
4.3.
Energy demand
The minimum energy demand for evaporative water removal consists of the energy required to heat the water to the boiling point and the energy to evaporate the water at the boiling point:
459
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 5 3 e4 6 4
DHboil ¼ mwater $cp $DT
(1)
DHevap ¼ mvapor $Dhv
(2)
DHboil: energy required to heat the water to the boiling point from room temperature (20 C) [kJ] DHevap: energy required to evaporate the water [kJ] mwater: mass of heated water [kg] mvapor: mass of vapor produced [kg] cp: heat capacity of water, 4.19 0.01 kJ kg1 K1 (average between 20 C and 100 C, Marsh, 1987) DT: temperature increase [K] Δhv heat of vaporization [kJ kg1], 2259 kJ kg1 (for ambient pressure of 1 atm, Lide, 2009) Due to the high amount of dissolved compounds the boiling point increases during evaporation. The increase of the boiling point is important, because it changes the amount of energy required to heat up a certain volume of water. At ambient pressure, the boiling point of synthetic urine increased only by 1.4 C up to a water removal of 82%, and an additional 9.0 C for a water removal of 97% (Figure S4, Supplementary Information). For the remaining 2.2% water removal the boiling point temperature increased exponentially, reaching 130.1 C. Mayer (2002) found a similar boiling point elevation for acidified fresh urine (ca. 9 C for a final 57% total solids which corresponds to 97.5% evaporated water) as we did for synthetic nitrified urine (12.5 C for 57% total solids content), if total solids are taken as reference. We therefore conclude that the boiling point elevation is mainly determined by the total solids in the liquid and does not differ substantially for differently treated urine. The energy demand for removing 99.2% water from synthetic nitrified urine at ambient pressure was calculated to be 710 W h L1 (90 W h L1 (13%) for heating the solution from ambient temperature (20 C) to boiling point, 620 W h L1 (87%) for steam production), which corresponds to 44 W cap1 (assuming a urine production of 1.5 L cap1 d1). With vapor compression, about 85% of the energy can be recovered at
small scale (Wood, 1982), which means that the energy requirement could be reduced to 110 W h L1 (6.9 W cap1), but electric energy is required. If one assumes a conversion efficiency of 31% for electricity production (average European electricity mix, UCPTE, 1994), the primary energy demand amounts to340 W h L1 or 21 W cap1. Additionally, nitrification in a small-scale MBBR (100 person equivalent) has a primary energy demand of about 8.9 W cap1 (Supplementary Information, Section S6). The total energy demand of nitrification and distillation (with energy recovery) amounts to 30 W cap1, which is about 0.6% of the average energy consumption rate in Switzerland (4690 W cap1 in the year 2009, calculated from BFE, 2010). The energy demand is two and a half times the energy demand estimated for removing nitrogen and phosphorus in a centralized wastewater treatment plant and producing the equivalent amount of nitrogen and phosphorus fertilizer: 190 W h L1 or 12 W cap1 (calculated based on Maurer et al. (2003) assuming phosphorus precipitation with iron (II) sulfate, phosphorus fertilizer production in Europe, nitrification/pre-denitrification, urea production, a urine composition according to spiked nitrified urine (Table 3) and a urine production of 1.5 Lurine cap1 d1). This energy comparison shows that a nitrification/distillation process would probably require more energy than conventional wastewater treatment, but it also shows that the energy demand is still low compared to the overall energy demand in a country like Switzerland. This comparison alone is not a sufficient basis to decide, whether a sanitation system based on nitrification/distillation requires more energy then conventional wastewater management. For a proper comparison, one has to include the energy demand for additional system components, such as the transport of urine, water and wastewater, the production of synthetic potassium and sulfate fertilizers and the handling of the feces and gray water.
4.4.
Alternative processes for water removal
To the best of our knowledge, lyophilization is the only other industrial technology for complete water removal from salt
Table 3 e Recovery of major compounds in two distillation experiments. On the left synthetic nitrified urine, on the right real nitrified urine (spiked with ammonium nitrate). The total inorganic solids are calculated from analyses. To measure the pH of the solids after distillation, the solid residue was diluted in 1 L of nanopure water. Synthetic nitrified urine
Water Phosphate-P Carbonate-C Sulfate Chloride Potassium Sodium Ammonia-N Nitrate-N Tot. inorg. solids DOC pH
Real nitrified urine
Input [g]
Product [g]
Diff. [%]
Input [g]
Product [g]
Diff. [%]
986 0.52 0.20 1.46 3.79 2.20 2.79 3.28 3.22 30.6 e 6.60
0.00 0.53 0.00 1.45 3.76 2.21 2.86 3.28 3.21 30.4 e 4.03
100 1.3 100 0.8 0.7 0.4 2.3 0.1 0.5 0.6 e e
989 0.22 <0.002 0.83 3.39 1.54 1.82 3.23 3.65 28.5 0.13 6.00
0.0 0.21 0.00 0.83 3.36 1.56 1.84 3.22 3.60 28.3 0.10 3.84
100 0.5 100 0.3 1.1 1.6 1.2 0.3 1.2 0.7 23 e
460
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 5 3 e4 6 4
solutions. During lyophilization, moist material is frozen at atmospheric pressure; later the water is sublimated at low temperatures (40 C and below in the condenser) and low pressure (millibars to microbars) (Rowe, 1960). This technology is often used to dry heat sensitive products in the food and pharmaceutical industry, but consumes high amounts of energy. According to Gehrmann et al. (2004) (cited in Claussen et al. (2007)) an electrical energy input of 1000 W h is needed to remove 0.4 kg water. This corresponds to 8100 W h L1 primary energy, which is 22 times higher than the energy demand for distillation with energy recovery. Maurer et al. (2006) described three additional processes that can be used to remove a particular amount of water from urine: electrodialysis, freeze and thaw, and reverse osmosis. In the electrodialysis process, ions are concentrated in an electric field by interrupting the ion flow with alternating anion and cation exchange membranes. For the treatment of stored urine, Dodd et al. (2008) estimated a gross primary energy demand of 145 W h cap1 d1, which corresponds to 97 W h L1 (assuming a urine production of 1.5 L cap1 d1 and exclusive energy for pumping). This is substantially lower than the estimated primary energy demand for distillation with energy recovery. However, the achievable concentration is low because water is also transported into the concentrate by osmosis and electro-osmosis. Furthermore, some dissolved compounds remain in the diluate. For stored urine, Pronk et al. (2006) reported a maximum concentration factor of 3.3 in continuous mode with 15% of the ions remaining in the diluate. The freeze and thaw method is based on the physical effect that during freezing most dissolved substances are excluded from the crystal lattice of ice and instead are concentrated in the remaining liquid phase. Lind et al. (2001) applied this method to fresh urine and synthetic fresh urine. They achieved 75% water removal, but only 80% of the nutrients were recovered. Gulyas et al. (2004) reported that the nutrient recovery from fresh urine decreased when more water was removed. In this study, the authors cited that a small scale falling film freezing reactor (250 L h1, Type “W33”, Niro Process Technology bv, ‘s-Hertogenbosch, The Netherlands) requires 244 W h L1. Assuming that this is electrical energy, the primary energy demand is 790 W h L1, which is in the same range as the energy demand for distillation without energy recovery. Today, reverse osmosis is the leading technology for water production from seawater. Due to the increasing osmotic pressure the typical water removal in seawater treatment is between 35 and 45% (Greenlee et al., 2009). Ek et al. (2006) used commercially available reverse osmosis units to concentrate untreated and acidified stored urine by a factor of 5 (80% water removal, 5 MPa, 29 C). With urine of an initial pH value of 9.2 (probably untreated urine) the nitrogen recovery was only 79%, while with acidified urine the recovery was increased to 91% and 98% at pH values of 7.0 and 6.0, respectively. The overall phosphorus loss was 10%, because some of the phosphorus, probably in the form of particulate compounds, was restrained in the pre-filtration step. For 80% water removal from acidified urine (pH 7), Ek et al. (2006) reported an energy demand of 8 W h L1 electricity and 4 W h L1 heat, which amounts to 30 W h L1 primary energy. This is considerably
lower than the energy demand we estimated for distillation. Fig. 4 summarizes the energy demand estimations for the various water removal technologies. Reverse osmosis has by far the lowest energy demand, but it is unsuitable for complete water removal. However, it can be combined with distillation to save energy. In a reactor setup that uses reverse osmosis to remove the first 80% of the water, the primary energy demand could be reduced to 170 W h L1 (11 W cap1 or 76% energy reduction) without energy recovery and 96 W h L1 (6.0 W cap1, 86% energy reduction) with energy recovery. Together with the energy for nitrification (8.9 W cap1, Section 4.3), this is only 25% higher than the energy demand for today’s wastewater treatment and fertilizer production (12 W cap1, see Section 4.3).
5.
Product stability
5.1.
Critical compounds
According to Table 3, ammonium nitrate accounts for 61% of the solid residue of the synthetic nitrified urine and 71% of the solid residues of the spiked nitrified urine. As ammonium nitrate is the most important nitrogen fertilizer in Europe (IFA, 2011), the use of the solid residue in agriculture seems to be an obvious choice for nutrient recycling. However, it has to be considered that ammonium nitrate is unstable at increased temperatures and pressures (Saunders, 1922).
Fig. 4 e Estimated primary energy demand for different processes for water removal from nitrified urine (1.5 L capL1 dL1). Dist.: distillation, ER: energy recovery (with vapor compression), RO: reverse osmosis. See text for a description of the processes. 1 W capL1 equals 16.0 W h LL1urine.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 5 3 e4 6 4
Two types of compounds in the solid residue can decrease the thermal stability even further: The first compound is chloride, which catalyzes the decomposition of ammonium nitrate by reducing the necessary activation energy (or onset temperature) of the exothermic ammonium nitrate decomposition (Keenan et al., 1968). The second set of compounds are organic substances, which can act as additional electron donors besides ammonium (Lurie and Lianshen, 2000) and thereby reduce the activation energy for the decomposition. To determine whether the solid residue is sufficiently stable to be used as a fertilizer, we conducted calorimetric measurements and consulted the legal regulations for ammonium nitrate fertilizers.
5.2.
Differential scanning calorimetry
We used differential scanning calorimetry (DSC) to assess the thermal stability of the solid residue. In a DSC analysis, a sample is heated at a constant rate and the energy amount taken up (endothermic peak) or released (exothermic peak) is measured. The onset temperature of the exothermic peak is an indicator for the sensitivity of a given compound, or in other words, for the amount of energy that is required to initiate a self-accelerating decomposition. Thermally stable materials generally have high onset temperatures. The area under the exothermic peak corresponds to the energy, which is released during the decomposition event. The amount of energy and the rate of heat release are indicators for the severity of the thermal decomposition (Ando et al., 1991). Endothermic peaks indicate phase transitions, which can reduce the stability of the mineral, for example by increasing the shock sensitivity of ammonium nitrate (Hahnefeld et al., 1983). Three materials were compared with DSC (Fig. 5): the solid residue of spiked nitrified urine (24% N content), a commercial ammonium nitrate fertilizer (27% N, Landor, Birsfelden, Switzerland) and pure ammonium nitrate (35% N, Catalogue
461
number 1.01188.1000, pro analysi, Merck, Darmstadt, Germany). Judging from the onset temperature and the released energy per gram N, the thermal stabilities of the solid residue and pure ammonium nitrate was similar. We assume that the advantage of less ammonium nitrate in the solid residue was counteracted by the presence of chloride and organic substances. The commercial fertilizer had a significantly higher onset temperature than the other the two solids. This may be due to powdered dolomite (CaMg(CO3)2), limestone (mainly calcite and fewer dolomite, silicates, quartz and gypsum) or pure CaCO3, which are usually added to ammonium nitrate fertilizers (EU, 2003). Wood and Wise (1955) showed that nitric acid accelerates the decomposition of ammonium nitrate. We assume that the carbonates help to neutralize the nitric acid, thereby increasing the activation energy for the thermal decomposition.
5.3.
Legal regulations
Since ammonium nitrate is the main nitrogen fertilizer in Europe, much experience exists for its safe handling and storage. Since disastrous accidents have happened and ammonium nitrate has been misused as explosive, there are strict regulations for the use of ammonium nitrate. The EU Decision 1348/2008/EC (EU, 2008) states that ammonium nitrate fertilizers with a nitrogen content higher than 16% are limited to farmers and other professional users, such as gardeners. Special regulations are given for ammonium nitrate fertilizers with more than 28% nitrogen content (EU, 2003). The solid residue of spiked urine (Table 3) and the theoretical solid residue of unspiked urine (composition as for spiked urine, but only 1.2 gNH4eN and 1.2 gNO3eN) have ammonium nitrate contents (as N) of 24% and 16%, respectively. The solid residue of undiluted stored urine without ammonia loss (composition according to Udert et al., 2006, all nitrogen hydrolyzed to ammonium, 50% ammonium oxidized to nitrate, no carbonate) would have an ammonium nitrate content (as N) of 24% per
Fig. 5 e DSC of three different ammonium nitrate (AN) powders with heating rate of 4 C per minute. All powders were ground by hand with a pestle before the analysis.
462
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 5 3 e4 6 4
mass. All these products have an ammonium nitrate content (as N) below 28%, but concentrations in urine vary widely and it is possible that this limit could be exceeded. In this case, the fertilizer will have to conform to the more stringent provisions of the European Regulation (EC) No. 2003/2003 (EU, 2003). Three of them will be critical: first, the content of combustible components (measured as C) should not exceed 0.4% by mass: the solid residue of spiked urine given in Table 3 has a DOC content of 0.35%, which is only slightly below this limit. Therefore, TOC removal from nitrified urine (e.g. adsorption on activated carbon) should be considered as an additional treatment step. Second, a solution of 10 g of fertilizer in 100 mL of water must have a pH of at least 4.5. We measured a pH value of 3.84, when diluting the solid residue of the spiked nitrified urine (about 28 g) in 1 L of nanopure water. Since the pH value in a mixture with less water is presumably even lower, a pH correction with alkaline minerals (e.g. calcite or dolomite) has to be considered. The third provision specifies a maximum chlorine (Cl) content of 0.02% by mass. This is the most critical provision, since the chlorine content (as chloride) in the solid residues was much higher (e.g. 11% in the solid residue of spiked nitrified urine). To the best of our knowledge, it will not be possible to push the chloride content of our products below this limit without removing significant amounts of nutrients. To comply with the EU regulations, the best approach will be to add fillers such as calcite, ground limestone or ground dolomite to the solid residue. Such an addition will reduce the ammonium nitrate content per gram of product, so that the strict rules for high-strength ammonium nitrate fertilizers will not be applicable. The addition of such fillers will also help to stabilize the fertilizer by raising the pH value. Nevertheless, even if the ammonium nitrate content is below 28% (as N), national regulations will require that special precautionary measures are taken for the handling of the ammonium nitrate containing fertilizer (see e.g. TRGS 511, 2008). Another approach to stabilize the solid residue of our process is to remove sodium chloride. Since sodium chloride has a comparatively low solubility, this mineral could be removed by sequential distillation or in an industrial post-treatment process.
6.
Conclusions
Our study has shown that biological nitrification with consecutive distillation of the effluent can be a stable and efficient process for the concentration and recovery of nutrients from urine. The solid residue contains high amounts of nutrients, such as ammonium nitrate, potassium, sulfate and phosphate, which makes it an interesting product for use in agriculture. Controlling the pH in the reactor via the inflow pump can prevent instabilities of the biological transformation processes. This control mechanism also allows for the control of the nitrification rate. MABRs are prone to strong biofilm growth, which can result in nitrogen loss by denitrification. A biofilm reactor with frequent biofilm erosion, such as an MBBR is supposedly more suitable for complete nitrogen recovery. We estimate that a very simple nitrification/distillation setup for urine treatment needs about four to five times the primary energy that is required to remove the same amount
of nitrogen and phosphorus in a conventional WWTP and produce equivalent amounts of synthetic phosphorus and nitrogen fertilizers. However, the energy demand can be significantly reduced by using vapor compression and reverse osmosis for water removal. An important advantage of a simple distillation process is the possibility to use solar energy for water heating. This is especially interesting for regions with high solar irradiation. The product of our process can be thermally instable, because of its high ammonium nitrate content. The ammonium nitrate content of the solid residues is below the limit for stringent EU regulations (28% N). Nevertheless, measures should be explored that render the product more stable, so that the process can also be used in small facilities, which might not be able to conform to the strict safety rules. Possible measures are adsorptive removal of organics before distillation or the addition of stabilizing fillers such as limestone, dolomite or pure calcite. An alternative process to ammonium nitrate production is the complete nitrification to nitrate, which requires an increase of the alkalinity by adding a base. The disadvantage of chemical dosage would be compensated by obtaining a thermally stable product. A possible setup for the nitrification reactor is the use of limestone granules or other alkaline particles as biofilm carriers.
Acknowledgments The project was supported by the Bill & Melinda Gates Foundation. Much of this work would not have been possible without contributions by Alexandra Hug (Eawag), Anabi Maria Assumpta Ngozi (Eawag) and several members of the research group of Prof. Dr. Mazzotti (Separation Processes Laboratory, ETH Zurich). A special thank you goes to Dr. Philippe Mauron (Empa) for helping with the DSC measurements, to Dr. Alexandre Sarbach (armasuisse) for his advice about safety assessment and to Max Reutlinger and Samuel Derrer (Eawag) for providing distillation equipment. The authors would also like to thank Claudia Ba¨nninger and Karin Rottermann (Eawag) for chemical analyses.
Appendix. Supplementary material Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.11.020.
references
Ando, T., Fujimoto, Y., Morisaki, S., 1991. Analysis of differential scanning calorimetric data for reactive chemicals. Journal of Hazardous Materials 28, 251e280. Anthonisen, A.C., Loehr, R.C., Prakasam, T.B.S., Srinath, E.G., 1976. Inhibition of nitrification by ammonia and nitrous acid. Journal of the Water Pollution Control Federation 48 (5), 835e852. BFE, 2010. Schweizerische Gesamtenergiestatistik 2009. (Swiss Total Energy Statistics 2009). Bundesamt fu¨r Energie, Berne, Switzerland (in German).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 5 3 e4 6 4
Claussen, I.C., Ustad, T.S., Strommen, I., Waide, P.M., 2007. Atmospheric freeze drying e a review. Drying Technology 25 (4e6), 947e957. Corcoran, E., Nellemann, C., Baker, E., Bos, R., Osborn, D., Savelli, H., 2010. Sick Water? The Central Role of Wastewater Management in Sustainable Development. A Rapid Response Assessment. United Nations Environmental Programme, UNHABITAT, GRID-Arendal, ISBN 978-82-7701-075-5. Dodd, M.C., Zuleeg, S., Von Gunten, U., Pronk, W., 2008. Ozonation of source-separated urine for resource recovery and waste minimization: process modeling, reaction chemistry, and operational considerations. Environmental Science and Technology 42 (24), 9329e9337. Ek, M., Bergstro¨m, R., Bjurhem, J.E., Bjo¨rlenius, B., Hellstro¨m, D., 2006. Concentration of nutrients from urine and reject water from anaerobically digested sludge. Water Science and Technology 54 (11e12), 437e444. Etter, B., Tilley, E., Khadka, R., Udert, K.M., 2011. Low-cost struvite production using source-separated urine in Nepal. Water Research 45 (2), 852e862. EU, 2008. Restrictions on the Marketing and Use of 2-(2methoxyethoxy)ethanol, 2-(2-butoxyethoxy)ethanol, Methylenediphenyl Diisocyanate, Cyclohexane and Ammonium Nitrate Decision No 1348/2008/EC of the European Parliament and of the Council. EU, 2003. Regulation (EC) No 2003/2003 of the European Parliament and of the Council of 13 October 2003 Relating Fertilisers. European Parliament and European Council. Faraghi, N., Ebrahimi, S., Udert, K.M., 2011. Electrochemical oxidation of nitrite to nitrate at graphite electrodes below the chlorine evolution potential. In preparation for submission to Water Research. Feng, D., Wu, Z., Xu, S., 2008. Nitrification of human urine for its stabilization and nutrient recycling. Bioresource Technology 99 (14), 6299e6304. Gehrmann, D., Zank, J., Kempkes, D.I., 2004. In: Eikevik, T.M. (Ed.), Gentle Drying Methods in the Pharmaceutical and Food Industry. Dewatering Laboratory NTNUdSINTEF, Trondheim, Norway. Goosse, P., Steiner, M., Udert, K.M., Neuenschwander, W., 2009. NoMix-Toilettensystem: Erste Monitoringergebnisse im Forum Chriesbach. (NoMix-toilet system: first results of the monitoring in Forum Chriesbach). Gas Wasser Abwasser (7), 567e574 (in German). Greenlee, L.F., Lawler, D.F., Freeman, B.D., Marrot, B., Moulin, P., 2009. Reverse osmosis desalination: water sources, technology, and today’s challenges. Water Research 43 (9), 2317e2348. Guest, J.S., Skerlos, S.J., Barnard, J.L., Beck, M.B., Daigger, G.T., Hilger, H., Jackson, S.J., Karvazy, K., Kelly, L., Macpherson, L., Mihelcic, J.R., Pramanik, A., Raskin, L., Van Loosdrecht, M.C.M., Yeh, D., Love, N.G., 2009. A new planning and design paradigm to achieve sustainable resource recovery from wastewater. Environmental Science and Technology 43 (16), 6126e6130. Gulyas, H., Bruhn, P., Furmanska, M., Hartrampf, K., Kot, K., ˜ ¼ttenberg, B., Mahmood, Z., Stelmaszewska, K., LA Otterpohl, R., 2004. Freeze concentration for enrichment of nutrients in yellow water from no-mix toilets. Water Science and Technology 50 (6), 61e68. Hahnefeld, K., Gill, R., Buske, G., 1983. Einflussgro¨ssen auf die Detonationsfa¨higkeit von Ammoniumnitrat. Wirtschaftsverlag NW, Bremerhaven. Hellinga, C., Van Loosdrecht, M.C.M., Heijnen, J.J., 1999. Model based design of a novel process for nitrogen removal from concentrated flows. Mathematical and Computer Modelling of Dynamical Systems 5 (4), 351e371. Hunik, J.H., Meijer, H.J.G., Tramper, J., 1992. Kinetics of Nitrosomonas europaea at extreme substrate, product and salt
463
concentrations. Applied Microbiology and Biotechnology 37 (6), 802e807. Hunik, J.H., Meijer, H.J.G., Tramper, J., 1993. Kinetics of Nitrobacter agilis at extreme substrate, product and salt concentrations. Applied Microbiology and Biotechnology 40 (2e3), 442e448. IFA, 2011. International Fertilizer Industry Association (2011). www.fertilizer.org (accessed 30.04.11.). Kapalka, A., Katsaounis, A., Michels, N.L., Leonidova, A., Souentie, S., Comninellis, C., Udert, K.M., 2010a. Ammonia oxidation to nitrogen mediated by electrogenerated active chlorine on Ti/PtOx-IrO2. Electrochemistry Communications 12 (9), 1203e1205. Kapalka, A., Joss, L., Anglada, A., Comninellis, C., Udert, K.M., 2010b. Direct and mediated electrochemical oxidation of ammonia on boron-doped diamond electrode. Electrochemistry Communications 12 (12), 1714e1717. Keenan, A.G., Notz, K., Franco, N.B., 1968. Synergistic catalysis of ammonium nitrate decomposition. Journal of the American Chemical Society 91 (12), 3168e3171. Kone´, D., 2010. Making urban excreta and wastewater management contribute to cities’ economic development: a paradigm shift. Water Policy 12, 602e610. Larsen, et al., 2009. Lide, D.R. (Ed.), 2009. CRC Handbook of Chemistry and Physics, eightynineth ed. CRC Press/Taylor and Francis, Boca Raton, Florida, USA (Internet Version). Lind, B.B., Ban, Z., Byde´n, S., 2001. Volume reduction and concentration of nutrients in human urine. Ecological Engineering 16 (4), 561e566. Lurie, B.A., Lianshen, Ch., 2000. Kinetics and mechanism of thermal decomposition of ammonium nitrate powder under the Action of Carbon Black 36(5), 607e617. Marsh, K.N. (Ed.), 1987. Recommended Reference Materials for the Realization of Physicochemical Properties. Blackwell Scientific Publications, Oxford, UK. Maurer, M., Schwegler, P., Larsen, T.A., 2003. Nutrients in urine: energetic aspects of removal and recovery. Water Science and Technology 48 (1), 37e46. Maurer, M., Pronk, W., Larsen, T.A., 2006. Treatment processes for source-separated urine. Water Research 40 (17), 3151e3166. Mayer, M., 2002. Thermische Hygienisierung und Eindampfung von Humanurin. Diplomarbeit des Institut fu¨r Umweltechnikder Fachhochschule beider Basel, Muttenz, Schweiz (Thermal disinfection and evaporation of human urine. Diploma work of the Institute for Environmental Technology, Fachhochschule beider Basel, Muttenz, Switzerland). Oosterhuis, M., Loosdrecht, M.C.M.v., 2009. Nitrification of urine for H2S control in pressure sewers. Water Practise and Technology 4 (3). doi:10.2166/wpt.2009.059. Pronk, W., Biebow, M., Boller, M., 2006. Electrodialysis for recovering salts from a urine solution containing micropollutants. Environmental Science and Technology 40 (7), 2414e2420. Rowe, T.W.G., 1960. The Theory and Practice of Freeze-Drying. Drying and Distillation Research Division, Edwards High Vacuum Ltd., Crawley, Sussex, England. 641e681. Saunders, H.L., 1922. The decomposition of ammonium nitrate by heat. Journal of the Chemical Society 121, 698e711. Suzuki, I., Dular, U., Kwok, S.C., 1974. Ammonia or ammonium ion as substrate for oxidation by Nitrosomonas europaea cells and extracts. Journal of Bacteriology 120 (1), 556e558. Tettenborn, F., Behrendt, J., Otterpohl, R., 2007. Resource Recovery and Removal of Pharmaceutical Residues. Treatment of Separate Collected Urine. Institute of Wastewater Management and Water Protection, Hamburg University of Technology, Hamburg, Germany.
464
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 5 3 e4 6 4
TRGS 511, 2008. Technische Regel fu¨r Gefahrstoffe. TRGS 511 Ammoniumnitrat. (Technical Guideline for the Handling of Hazardous Materials TRGS 511 Ammonium Nitrate). Ausschuss fu¨r Gefahrenstoffe, Bundesanstalt fu¨r Arbeitsschutz und Arbeitsmedizin (BAuA), Dortmund, Germany (in German). UCPTE, 1994. Yearly Report 1993. Union pour la coordination de la production et du transport de l’e´lectricite´, Vienna, Austria. Udert, K.M., Larsen, T.A., Biebow, M., Gujer, W., 2003a. Urea hydrolysis and precipitation dynamics in a urine-collecting system. Water Research 37 (11), 2571e2582. Udert, K.M., Fux, C., Mu¨nster, M., Larsen, T.A., Siegrist, H., Gujer, W., 2003b. Nitrification and autotrophic denitrification of sourceseparated urine. Water Science and Technology 48 (1), 119e130. Udert, K.M., Larsen, T.A., Gujer, W., 2005. Chemical nitrite oxidation in acid solutions as a consequence of microbial
ammonium oxidation. Environmental Science and Technology 39 (11), 4066e4075. Udert, K.M., Larsen, T.A., Gujer, W., 2006. Fate of major compounds in source-separated urine. Water Science and Technology 54 (11e12), 413e420. Verhave, W.A., Frediansyah, R., van Loosdrecht, M.C.M., 2007. Process for the conversion of liquid waste biomass into a fertilizer product. Patent, WO 2007/097612 A2. Wiesmann, U., 1994. Biological nitrogen removal from wastewater. Advances in Biochemical Engineering/ Biotechnology 51, 113e154. Wood, F.C., 1982. The changing face of desalination e a consulting engineers viewpoint. Desalination 42 (1), 17e25. Wood, B.J., Wise, H., 1955. Acid catalysis in the thermal decomposition of ammonium nitrate. The Journal of Chemical Physics 23 (4), 693e696.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 6 5 e4 7 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Development of predictive models for determining enterococci levels at Gulf Coast beaches Zaihong Zhang, Zhiqiang Deng*, Kelly A. Rusch Department of Civil & Environmental Engineering, Louisiana State University, Baton Rouge, LA 70803-6405, USA
article info
abstract
Article history:
The US EPA BEACH Act requires beach managers to issue swimming advisories when water
Received 18 June 2011
quality standards are exceeded. While a number of methods/models have been proposed to
Received in revised form
meet the BEACH Act requirement, no systematic comparisons of different methods against
17 September 2011
the same data series are available in terms of relative performance of existing methods. This
Accepted 5 November 2011
study presents and compares three models for nowcasting and forecasting enterococci levels at
Available online 19 November 2011
Gulf Coast beaches in Louisiana, USA. One was developed using the artificial neural network (ANN) in MATLAB Toolbox and the other two were based on the US EPA Virtual Beach (VB)
Keywords:
Program. A total of 944 sets of environmental and bacteriological data were utilized. The data
Coastal beaches
were collected and analyzed weekly during the swimming season (MayeOctober) at six sites of
Enterococci levels
the Holly Beach by Louisiana Beach Monitoring Program in the six year period of May
ANN model
2005eOctober 2010. The ANN model includes 15 readily available environmental variables
Virtual Beach model
such as salinity, water temperature, wind speed and direction, tide level and type, weather type, and various combinations of antecedent rainfalls. The ANN model was trained, validated, and tested using 308, 103, and 103 data sets (collected in 2007, 2008, and 2009) with an average linear correlation coefficient (LCC) of 0.857 and a Root Mean Square Error (RMSE) of 0.336. The two VB models, including a linear transformation-based model and a nonlinear transformation-based model, were constructed using the same data sets. The linear VB model with 6 input variables achieved an LCC of 0.230 and an RMSE of 1.302 while the nonlinear VB model with 5 input variables produced an LCC of 0.337 and an RMSE of 1.205. In order to assess the predictive performance of the ANN and VB models, hindcasting was conducted using a total of 430 sets of independent environmental and bacteriological data collected at six Holly Beach sites in 2005, 2006, and 2010. The hindcasting results show that the ANN model is capable of predicting enterococci levels at the Holly Beach sites with an adjusted RMSE of 0.803 and LCC of 0.320 while the adjusted RMSE and LCC values are 1.815 and 0.354 for the linear VB model and 1.961and 0.521 for the nonlinear VB model. The results indicate that the ANN model with 15 parameters performs better than the VB models with 6 or 5 parameters in terms of RMSE while VB models perform better than the ANN model in terms of LCC. The predictive models (especially the ANN and the nonlinear VB models) developed in this study in combination with readily available real-time environmental and weather forecast data can be utilized to nowcast and forecast beach water quality, greatly reducing the potential risk of contaminated beach waters to human health and improving beach management. While the models were developed specifically for the Holly Beach, Louisiana, the methods used in this paper are generally applicable to other coastal beaches. ª 2011 Elsevier Ltd. All rights reserved.
* Corresponding author. Tel.: þ1 225 5786850. E-mail addresses:
[email protected] (Z. Zhang),
[email protected] (Z. Deng),
[email protected] (K.A. Rusch). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.027
466
1.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 6 5 e4 7 4
Introduction
The US federal BEACH (Beaches Environmental Assessment and Coastal Health) Act requires beach managers to monitor bacterial water quality and issue swimming advisories when water quality criteria are violated. The main problem with the current beach monitoring program is that the level of water quality indicators like enterococci (ENT) may change between the time of sampling and reporting of results because current analysis methods commonly require an incubation step of 24e48 h. This incubation step makes protective actions such as preemptive beach closures impossible. This time lag of 24e48 h can lead to beach advisories and closures that cause unwarranted loss of valuable recreation access or to permit swimming when conditions present an unacceptable level of risk. People swimming during the time between sample collection and test results may be unnecessarily exposed to microbial pollutants at peak contamination times. Essentially, the efficacy of current beach monitoring procedure, called persistence model, generally depends on steadiness of bacterial concentration in beach water while the steady condition rarely occurs in coastal beach waters (Kim et al., 2004; Boehm et al., 2005; Hou et al., 2006; Heberger et al., 2008). Therefore, an alternative model is needed to meet the BEACH Act requirements. Extensive efforts have been made to develop predictive models for nowcasting and forecasting the level of fecal indicator bacteria in beach waters (Heberger et al., 2008; Lin et al., 2008; Sanders et al., 2005; Kelsey et al., 2004; Kay et al., 1994). Hou et al. (2006) presented a Dynamic Partial Least Square Regression (DPLSR) model for predicting ENT levels at the Huntington State Beach (HSB) and the Huntington City Beach (HCB), California. Parameters involved in the DPLSR model included storm water discharge, rainfall, sea surface temperature, upwelling index, wind velocity, wave height and direction, visitor number, atmospheric pressure, solar insolation, sampling time, tide level and range, and rainfall. A total of 703 sets of ENT and environmental data from October 1999 through December 2000 (one swimming season) were used. Results showed that the DPLSR model performed better than the persistence model. Frick et al. (2008) proposed a multiple linear regression (MLR) model, called Virtual Beach (VB) model. Variables used in the VB model included air temperature, dewpoint temperature, cloud cover, precipitation potential, wave height, wind direction, wind speed, alongshore wind component, and cross-shore wind component. The adjusted coefficient of determination of the VB model for Escherichia coli nowcasting for a 42 day swimming period in the summer 2006 was about 0.40. He and He (2008) introduced artificial neural network models for predicting fecal indicator bacterial concentrations at the Torrey Pines State Beach and the San Elijo State Beach in southern California. Model variables involved temperature, conductivity, pH, turbidity, river flow, rainfall, and time lapse after a rainstorm. A total of 184 data sets collected from March 15 to April 14, 2003 were used to train, validate, and test the models. Results showed that the linear correlation coefficients for training, validation, and testing of the models were 0.883, 0.878, and 0.789, respectively. No independent data were used to test the ANN models. In
addition, typical coastal water parameters (such as tide level and type) were not included in the ANN models. It is generally recognized that statistical models should be based on longterm data (Nevers and Whitman, 2005; Francy and Darner, 2006). However, most existing beach water quality models were trained and tested using the same data collected in a single/short swimming period. Few existing models were tested against beach water quality data observed over two swimming seasons, making it difficult for beach monitoring programs to find a reliable predictive model especially for coastal beaches with unknown bacterial sources. Therefore, more modeling efforts are needed to improve beach monitoring programs and to implement the BEACH Act. The overall goal of this study was to enhance beach monitoring programs by finding a more reliable model for predicting the level of fecal indicator bacteria in coastal beach waters impaired frequently by unknown bacterial sources. The US Environmental Protection Agency (US EPA) recommended that the enterococci (ENT) be used as a bacteriological water quality indicator for marine waters (USEPA, 1986). Therefore, specific objectives of this study were (1) to develop an ANN model for nowcasting and forecasting the ENT level at the Holly Beach, Louisiana, USA, (2) to construct a VB model using the linear transformation and another VB model using the nonlinear transformation for the same purpose, and (3) to compare the three models against the same data series and thereby to find and recommend a more reliable model to meet the BEACH Act requirements under different environmental conditions. While exhibiting typical features of coastal beaches, the Holly Beach also has some unique characteristics. Unique features of the models developed in this study include: (1) they are developed using a relatively long time series of data observed over three swimming seasons; (2) they are further tested through a hindcast procedure using additional three years of independent data which are not used in the model development; (3) bacterial sources to the Holly Beach are unknown; and (4) annual average bacterial level at the Holly Beach keeps increasing or displays a dynamic behavior. Due to the unique features and highly variable and dynamic bacterial levels, no existing models are applicable to the Holly Beach.
2.
Study area and data collection
2.1.
Study area
The Holly Beach is located in the Cameron Parish along the southwest Louisiana shoreline, as shown in Fig. 1. The Holly Beach stretches along the Gulf of Mexico from the Calcasieu River Outlet in the East toward the Sabine River Outlet in the West in the Calcasieu River Basin, Louisiana, US. There are three national wildlife refuges (NWR) close to the Holly Beach in the Calcasieu River Basin. The use of Holly Beach is very high during the swimming season from MayeOctober, with approximately 150 people using the beach on a typical weekday, 1000 people on a typical weekend, and 6000 people on a typical holiday. Six water quality sampling sites at the Holly Beach are shown in Fig. 1 and their coordinates are listed in Table 1.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 6 5 e4 7 4
467
Fig. 1 e Holly Beach and beach shed containing three national wildlife refuges (NWR) and six sampling sites (Holly 1eHolly 6) (Modified from LDHH, 2009).
A field survey was conducted in 2008 by the Louisiana BEACH Monitoring Program and the Louisiana Department of Environmental Quality to investigate bacterial sources to the Holly Beach (LDHH, 2009). Results of the survey are shown in Fig. 1 in terms of enterococci geometric mean and salinity. The figure indicates that the ENT levels in both the Calcasieu River and the offshore water are lower than those at the six beach sites, implying that the primary bacterial sources are from
LnðENTÞ ¼ 0:3113 ðYearÞ 620:77 R2 ¼ 0:8612
Table 1 e Six sampling sites along Holly Beach. Sampling site Holly Holly Holly Holly Holly Holly
1 2 3 4 5 6
neither the river nor the offshore oil platforms in the Gulf of Mexico. The population in the Cameron Parish has decreased since 2005 due to the destruction caused by Hurricane Katrina to the area. However, the annual average ENT level at the Holly Beach keeps increasing, as shown in Fig. 2. It is not clear what sources cause the increasing bacterial level at the Holly Beach. The three wildlife refuges may be potential sources but there are no known flow connections between the beach and the wildlife refuges, making the bacterial sources to the Holly Beach unknown. The trendline in Fig. 2 can best be described by the following regression equation:
Latitude
Longitude
29.7689 29.7692 29.7694 29.7694 29.7694 29.7697
93.4375 93.4442 93.4494 93.4542 93.4594 93.4642
(1)
Fig. 2 shows that beach water quality may exhibit a long-term variation trend and this long-term trend can only be found with at least multiple years of data. It is impossible to make predictions of beach water quality with a model developed using the data collected in one swimming season or even a shorter sampling period. The long-term variation trend in beach water quality may be caused by land use/land cover
468
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 6 5 e4 7 4
NOAA National Weather Service website (http://water. weather.gov/precip/) for the Holly Beach. Tidal water level data were obtained from the NOAA website for Tides and Currents (http://tidesandcurrents.noaa.gov/) for Calcasieu Pass, LA (station ID: 8768094). The 15 variables were selected based on the results of our analyses of processes and variables controlling ENT variations in beach waters, including principal component analysis and stepwise regression analysis.
3. Variables controlling bacterial transport and survival 3.1. Processes and mechanisms responsible for bacterial transport and survival Fig. 2 e Variation trend in annual log-mean ENT levels at Holly Beach (Standard deviations of LnENT for 2005e2010 were 4.5005, 5.8327, 5.2475, 5.5667, 6.0359, and 6.9855, respectively).
change or global climate change. Therefore, beach water quality models should be developed using long-term monitoring data.
2.2. data
Gathering and processing of beach water quality
The data used in this paper include ENT data and the data for 15 environmental variables. The ENT data for the period of May 2005eOctober 2010 were provided by the Beach Monitoring Program of Louisiana Department of Health and Hospitals (LDHH), US. Water samples have been collected by LDHH from the six sites at Holly Beach (Fig. 1) on a weekly basis during the swimming season May 1eOctober 31 since 2005. Samples were analyzed for fecal indicator bacteria (including ENT) using US EPA standard methods. The environmental data include the time series of 15 independent variables including salinity, water temperature, wind speed type (6 categories), tide type (9 categories), tide type (3 categories), tide water level, weather type (sunny Y/N), wind direction (off/on shore), daily rainfall, rainfall one day before, rainfall two days before, rainfall three days before, rainfall in last 48 h, rainfall in last 72 h, and rainfall in last 96 h. It should be pointed out that the tide type (9 categories) is used to describe the effect of individual tide categories (extremely high-9, high rising-8, high-7, high falling-6, normal-5, low rising-4, low-3, low falling-2, and extremely low-1) on bacterial transport while the tide type (3 categories: high-3, normal2, and low-1) represents the cumulative effect of the individual tide categories. The tide type (3 categories) is similar to the cumulative rainfall parameters (rainfall in last 48 h, rainfall in last 72 h, and rainfall in last 96 h) while the tide type (9 categories) is similar to the antecedent rainfall in an individual day such as rainfall 1 day before and rainfall 2 days before. The data for salinity, water temperature, wind speed, wind direction, tide, and weather were obtained from LDHH along with the ENT data. Rainfall data were obtained from
Due to the uncertainty in bacterial sources to Holly Beach, it is essential to understand all potential processes and mechanisms along with parameters responsible for the transport and survival of ENT in coastal beach waters so that important environmental variables affecting ENT levels can be included in the models. The variation in ENT levels in coastal beach waters may be affected by various physical and biogeochemical processes (Grant and Sanders, 2010) such as ENT transport (advection and dispersion), growth and inactivation processes. Transport processes may include the processes in coastal waters due to wind/wave-induced currents and the washing-off processes on beach due to tides/rainfalls. Dominant transport mechanisms vary with location in the marine environment (Mill et al., 2006). Beach sand/sediment is a natural filter that traps environmental particulates, organic matter and microorganisms (Ahammed and Chaudhuri, 1996; Hua et al., 2003; Bonilla et al., 2007; Halliday and Gast, 2011). The large surface area of beach on the shore provides microbes with the unique microhabitat within the cracks and crevices and a variety of potentially suitable environments for growth and enhanced survival (USEPA, 1999). Beach sand accumulates microorganisms from humans and animals (including bird droppings) when it naturally filters water washing ashore by waves or from the land after rain events (Papadakis et al., 1997; Elmir et al., 2007). The ENT can persist and potentially multiply in tropical soil and sand as well (Carrillo et al., 1985; Davies et al., 1995; Byappanahalli and Fujioka, 1998; Solo-Gabriele et al., 2000; Craig et al., 2004; Anderson et al., 2005). Therefore, beach sand acts as an important reservoir for microbial contaminants (USEPA, 1999). When there is a rain and/or tide event, the ENT will be washed off from the beach sand into the beach water. Rainfall can also cause overflow of sewage from septic tanks or soakaway pits which are the best accommodation for fecal indicator bacteria (Chambers et al., 2008). ENT levels in beach water also vary due to the mixing across the width and depth of the surf zone driven by wave-induced turbulence, parallel transport to the shore by wave-driven long shore currents, and dilution by rip cell exchange of ocean water between the surf zone and offshore (Boehm et al., 2005). Onshore wind may also prevent ENT from dispersing from the beach while offshore wind forces this dispersion. In addition, strong wind may produce a high tide. Onshore wind may bring the effluent from a nearby submarine outfall to the beach, implying that
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 6 5 e4 7 4
tide type, tide level, wind speed, and wind direction are important factors for the wash-off process of bacteria. Strong tides and winds (An et al., 2002; Obiri-Danso and Jones, 2000) as well as rainfall/storm events may also cause change in ENT levels in beach waters through bottom sediment re-suspension (Davies et al., 1995; Crabill et al., 1999; An et al., 2002). The ENT can attach to fine particles and flocs and settle-out of the water column to form a semi-permanent store on seabed. Bottom sediment particles may also coat and protect the ENT cells (Roper and Marshall, 1978). Thus, ENT can often accumulate in bottom sediment (Crabill et al., 1999; Irvine and Pettibone, 1993; Fries et al., 2006). The bottom sediment re-suspension can release the ENT attached to sediment particles. ENT growth is controlled by many factors such as water temperature, salinity, and nutrients. ENT activities increase with water temperature, and there is a positive correlation between the ENT level and water temperature at coastal beaches. Low salinity favors ENT growth. The salinity of coastal waters usually varies around 33 ppt, and it may drop to below 10 ppt after a heavy rainfall due to freshwater input. Therefore, the low salinity in coastal waters is commonly an indication of the freshwater discharge carrying high levels of bacteria and nutrients and thereby enhances population growth and/or survival of ENT in beach waters (Evanson and Ambrose, 2006). ENT inactivation happens when ENT experience mutation, lysis, dwarfing, or oxidative-damage due to environmental stress (Nystro¨m, 2004). Temperature, salinity, solar radiation, and lack of nutrients can cause the environmental stress and thus contribute to inactivation of ENT in natural waters (Noble et al., 2004; Hanes and Fragala, 1967; Boehm et al., 2002). Previous studies showed that solar radiation has significant effect on ENT inactivation (Burkhardt et al., 2000; Davies and Evison, 1991; Davies-Colley et al., 1994, 1997, 1999; Dura´n et al., 2002; Evison, 1988; Fujioka et al., 1981; Gameson and Gould, 1975; McGuigan et al., 1998; Pommepuy et al., 1992; Sinton et al., 1999, 2002). However, the solar radiation parameter was not explicitly included in our models because all ENT data used in this study were based on the samples taken in the early morning by 9:00AM.
3.2.
Statistical analyses
An important step involved in the development of the predictive models is the selection of explanatory variables for the ENT level based on the variables involved in the processes and mechanisms responsible for the transport and survival of
469
ENT in coastal beach waters. To that end, a principal component analysis and a stepwise regression analysis were conducted to identify the environmental parameters controlling the ENT transport and survival processes. The statistical analyses were performed using the SAS system. Results of the statistical analyses show that the natural log-transformed ENT (LnENT) is significantly correlated with the 15 parameters listed in Section 2.2. Based on results of Principle Component Analysis, it was noticed that most of the 15 parameters are uncorrelated except tide type (9 categories) and tide type (3 categories) as they use the same raw data. The seven rainfall parameters are also correlated to each other with a correlation coefficient greater than 0.5. In fact, among the 15 variables seven are derived from various combinations of antecedent rainfalls and three are related to the tide. Therefore, the rainfall is by far the most important physical driver of ENT concentrations at Holly Beach, followed by the tide.
4.
Development of predictive models
4.1.
Artificial neural network (ANN) model
The artificial neural network (ANN) analysis has been found to be able to describe nonlinear relationships between independent and dependent variables and provide a much higher accuracy than multiple linear regression analysis (Keiner and Yah, 1998). The role of ANN model is to develop a response by assigning weights in such a way that it represents the true relationship that really exists between model inputs and output (Zakiuddin and Modak, 2010). Therefore, a feedforward ANN model with the back propagation training algorithm was presented for prediction of ENT level in beach waters. The ANN model architecture consists of one input layer, five hidden layers consisting of 20 hidden neurons, and one output layer, as shown in Fig. 3. The model output is predicted ENT levels (Ln(ENT)) in the beach water. The input layer requires values for the 15 explanatory variables. The MATLAB Neural Network Toolbox (V. 7) was used for training, validation, and testing of the ANN model. A total of 514 data sets (each data set consists of the data for enterococci levels and for the 15 model input parameters measured on a specific day) collected in 2007, 2008, and 2009 were employed as the input data and split into three data groups for training (Group-1: 60% or 308 data sets), validation (Group-2: 20% or 103 data sets), and testing (Group-3: 20% or 103 data sets). The MATLAB Program randomly selects the
Fig. 3 e ANN model architecture.
470
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 6 5 e4 7 4
data sets from the 514 data sets for each group based on the percentage predefined by the modeler, and then trains, validates, and tests the ANN model. The training, validation, and testing results are displayed in terms of linear correlation coefficient (LCC) and mean squared error (MSE), as shown in Table 2. The ANN model may be trained numerous times until the overall performance of the model is satisfactory. It should be pointed out that the model with the highest LCC values or the lowest MSE values is not necessarily the best model when it is employed to nowcast or forecast ENT levels using new independent data. In this study, the model with the highest LCC values or the lowest MSE values is shown in Table 2. However, we finally selected another model, called model 0.893, with the linear correlation coefficient of 0.893 (LCC value for training), 0.811 (LCC value for validation), and 0.866 (LCC value for testing) as this model with lower LCC values actually performs better than the model with the highest LCC values when they are applied to additional data collected in 2005, 2006, and 2010. Fig. 4 shows the performance of the finally selected ANN model for the 2007e2009 data sets. The figure indicates that the Ln(ENT) levels predicted with the model 0.893 are very close to observed ones with an RMSE value of 0.336. The finally selected ANN model was then saved as a MATLAB project and can be used as a predictive tool for nowcasting and forecasting ENT levels in beach waters using the sim function in the MATLAB.
4.2.
from 1.3022 to 1.3026. The linear VB model (Model 1) with the lowest RMSE (1.3022) and the highest LCC (0.230) is written as: LnðENTÞ ¼ 4:2615 0:032254½SAL þ 0:13297½WSTy þ 0:082758½WDTyOnShore þ 0:39407½RFlag3
Virtual Beach (VB) models
0:53395½RF48 þ 0:18879½OWL
Virtual Beach (VB) is a free and open-source software package developed by US EPA to assist beach managers in constructing site-specific Multiple Linear Regression (MLR) models for the prediction of levels of fecal indicator bacteria at recreational beaches (http://www.epa.gov/ceampubl/swater/vb2/). An advantage of the VB program is that it enables automated selection of model input variables from a wide array of explanatory variables. In this study, the 15 input variables used in the development of ANN model were provided for the VB program. The VB program includes exhaustive and genetic algorithm (GA) search routines for finding the ‘best’ models from a large array of possible choices (various combinations of the 15 input variables through MLR equations). In fact, the VB program only selected 5e7 variables from the 15 input variables for both the linear and the nonlinear transformation options.
4.2.1.
Fig. 4 e Comparison between the Ln(ENT) levels predicted using the trained ANN model with R [ 0.893 and observed in swimming seasons 2007e2009 (1e197 are data from 2007 swimming season; 198e354 from 2008; and 355e514 from 2009).
(2)
where Ln(ENT) ¼ natural log-transformed enterococci level, SAL ¼ salinity, WSTy ¼ wind speed type (6 categories), WDTyOnShore ¼ wind direction (off/on shore), RFlag3 ¼ rainfall 3 days before, RF48 ¼ Rainfall in last 48 h, and OWL ¼ tidal water Level. While Models 2e8 may involves different variables, there performances are similar to that of Model 1, as shown in Fig. 5.
Linear transformation-based VB model
For the linear transformation option, the VB program found 8 ‘best’ models with comparable RMSEs varying in the range
Table 2 e ANN model performance for training, validation and testing data.
Training Validation Testing
Sample number
MSE
LCC
308 103 103
0.321 0.401 0.500
0.921 0.908 0.902
Fig. 5 e Comparison between the Ln(ENT) levels predicted using the best linear and nonlinear VB models and observed in swimming seasons 2007e2009 (1e197 are data from 2007 swimming season; 198e354 from 2008; and 355e514 from 2009).
471
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 6 5 e4 7 4
predicted using Eqs. (2) and (3) and observed in swimming seasons 2007, 2008 and 2009. While the nonlinear VB model indeed shows a quantitative improvement over the linear VB model in terms of RMSE and LCC, it is very hard to see the difference between the two VB models visually in Fig. 5. However, it is clear from Fig. 5 that both VB models underpredict the observed variability of ENT concentrations.
5.
Fig. 6 e Comparison between adjusted Ln(ENT) levels from ANN and VB models against observed Ln(ENT) level for swimming seasons 2005, 2006 and 2010 (1e157 are data from 2005 swimming season; 158e256 from 2006; and 257e430 from 2010).
4.2.2.
Nonlinear transformation-based VB model
The nonlinear transformation option in the VB program is intended to describe nonlinear relationships between dependent and independent variables and is often accomplished through the use of square, square root, inverse, log10, natural log, and polynomial transformations. In order to develop a nonlinear VB model for the Holly Beach, all nonlinear transformations available in the VB program were tried for the same data sets used in the development of the ANN model. The best nonlinear VB model with the lowest RMSE (1.205) and the highest LCC (0.337) is of the following form: LnðENTÞ ¼ 7:4999 0:02921 ½SAL þ 0:099795 ½WSTy þ 0:81877 ½POLY½TdHNLOrd þ 1:0582 ½POLY½RF þ 1:0870 ½POLY½RFlag3
(3)
where POLY ¼ polynomial transformation, TdHNLOrd ¼ tide type (3 categories), and RF ¼ daily rainfall. Both the RMSE (1.205) and the LCC (0.337) values for the nonlinear VB model show a smaller error than the corresponding values for the linear VB model with RMSE ¼ 1.3022 and LCC ¼ 0.230, indicating the improvement in the performance of the nonlinear VB model over the linear VB model. Fig. 5 shows a comparison between the Ln(ENT) levels
Results and discussion
In order to assess the nowcasting and forecasting performance of the ANN and VB models developed using the data from 2007 to 2009, the models were utilized to hindcast ENT levels in 2005, 2006, and 2010. The Ln(ENT) levels predicted with the three models are compared against the observed one in Fig. 6. It should be noted that the ENT data for 2005, 2006, and 2010 are independent data which were not used in the development of the models. The model predicted Ln(ENT) values were multiplied by the annual factor 1.00 for 2005, 1.09 for 2006, and 1.46 for 2010 to take into account the increasing trend in ENT level shown in Fig. 2. The annual factors were determined by dividing the annual averaged Ln(ENT) values (for 2005, 2006, and 2010) calculated from Eq. (1) with that of 2005. The adjusted ANN model is capable of predicting enterococci levels at the Holly Beach sites with the RMSE of 0.803 and LCC of 0.320 while the RMSE and LCC values are 1.815 and 0.354 for the adjusted linear VB model and 1.961and 0.521 for the adjusted nonlinear VB model. Obviously, the LCC value (0.521) of the nonlinear VB model is significantly higher than those of ANN and linear VB models. However, the RMSE value of the ANN model (0.803) is much smaller than those (1.815 and 1.961) of the VB models. It can be seen from Fig. 6 that the observed ENT level generally experiences a seasonal variation characterized by a peak in summer (JuneeAugust) and relatively low levels in spring (May) and fall (SeptembereOctober). The ENT levels predicted by the ANN model fit the observed ones reasonably well, as compared to the performance of the linear and nonlinear VB models. In addition, it can be seen from Figs. 5 and 6 that the Ln(ENT) values predicted by the VB models don’t follow the overall variation trend in observed ENT levels and vary in a narrow range in most cases. To further compare performances of the three models against the observed data, Table 3 lists statistical measures including mean, standard deviation (SD), and skewness of the Ln(ENT) values for the observed data (2005, 2006, and 2010) and predicted values from the ANN and VB models. Table 3 clearly shows that the ANN model produces the smallest errors among the three
Table 3 e Comparison between ANN and VB models against statistical measures. Model
Observed data ANN model Linear VB model Non-linear VB model
Statistical parameters Mean
% Error of mean
Standard deviation
% Error of SD
Skewness
% Error of skewness
4.256 4.600 4.826 5.534
0 8.083 13.379 30.016
1.729 1.796 1.221 1.132
0 3.892 29.372 34.508
0.374 0.010 0.100 0.319
0 97.269 73.222 14.786
472
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 6 5 e4 7 4
Fig. 7 e Comparison between simulated and observed ENT levels at six sites along Holly Beach for swimming seasons 2005, 2006 and 2010 (Holly 1: 1e24 are data from 2005, 25e38 from 2006, 39e66 from 2010; Holly 2: 1e25 are data from 2005, 26e41 from 2006, 42e69 from 2010; Holly 3: 1e22 are data from 2005, 23e44 from 2006, 45e72 from 2010; Holly 4: 1e26 are data from 2005, 27e43 from 2006, 44e72 from 2010; Holly 5: 1e29 are data from 2005, 30e48 from 2006, 49e76 from 2010; Holly 6: 1e27 are data from 2005, 28e42 from 2006, 43e75 from 2010).
models in terms of the two most important statistical measures (mean and standard deviation). The standard deviation (SD) is a measure of the variability in Ln(ENT) values. SD values of the two VB models are markedly smaller than that of observed ENT data, indicating the less variability in VB model predictions as compared to the observed data and ANN model predictions. As a result, Ln(ENT) predictions from the VB models yield significantly high errors in the SD. The smallest error in the SD of ANN model predictions indicates that the ANN model is able to simulate the measured variability in ENT levels best among the three models. It is surprising that the nonlinear VB model produces a higher error in the mean and the SD than does the linear VB model even though the nonlinear VB model has the least error in the skewness among the three models. The under-prediction of the observed variability of ENT concentrations by the VB models may be caused by the insufficient number (5e7) of independent variables selected by the VB program. The number (5 or 6) of independent variables currently involved in the VB models is simply too small to capture the natural variability in coastal beach water quality no matter the linear or the nonlinear transformation option is employed. The performance of VB models may be significantly improved if the VB program is revised to include more independent variables. This could be a future direction for improvements of the
VB program. The VB program may also be improved by developing an optimization algorithm for the nonlinear transformation option. Even though the under-prediction problem with the VB models, the VB program is still a useful tool for the development of predictive models for beach water quality because the VB models requires much less input data as compared to the ANN model. Therefore, the VB models and particularly the nonlinear VB model may be used in case that some input data required in the ANN model are not available. In order to further understand the performance of the ANN model for individual beaches, Fig. 7 shows a comparison between the Ln(ENT) values observed in 2005, 2006, and 2010 at the six sampling sites along Holly Beach and predicted by the ANN model. It can be seen from the figure that there are no apparent differences among the six beach sites in terms of the model performance. A significant feature shown in both Figs. 6 and 7 is the markedly elevated ENT levels in 2010. Consequently, the ANN model is capable of capturing both the seasonal and the annual variation trends in the ENT level.
6.
Conclusions
The paper presents three predictive models, an ANN model and two VB models, for nowcasting and forecasting enterococci
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 6 5 e4 7 4
levels at coastal beaches. The models were tested using a total of 944 sets of environmental and bacteriological data collected over 6 swimming seasons. Findings from the development and applications of the models can be summarized as follows: (1) The ANN model with 15 input parameters has an average linear correlation coefficient (LCC) of 0.857 and a Root Mean Square Error (RMSE) of 0.336 for a total of 514 sets of data used in model training, validation, and testing. The best linear VB model with 6 input variables has an LCC of 0.230 and an RMSE of 1.302 while the best nonlinear VB model with 5 input variables has an LLC of 0.337 and an RMSE of 1.205 for the same data sets used in the ANN model. (2) The ANN model is capable of predicting enterococci levels at the Holly Beach sites over additional three years (2005, 2006, and 2010) with an LCC of 0.320 and an RMSE of 0.803. The nonlinear VB model has the highest LCC value (0.521) among the three models but it also produces the highest errors in the mean and the standard deviation of predicted ENT concentrations for 2005, 2006, and 2010. The ANN model is able to capture both seasonal and annual variation trends in enterococci levels but it requires more input data as compared to the VB models. While the VB models only require 5e7 input variables, their predictive performance is not so good due to the less variability in predicted ENT levels as compared to the high variability in observed ENT levels in coastal beach waters. (3) The predictive models (especially ANN model) in combination with readily available real-time environmental and weather forecast data can be utilized to nowcast and forecast beach water quality, greatly reducing the potential risk of contaminated beach waters to human health and providing an efficient and effective tool for beach management and use. (4) While the ANN and VB models were specifically developed for the Holly Beach, Louisiana, the methods used in this paper are generally applicable to other coastal beaches.
Acknowledgments Support for this research by the US NASA (National Aeronautics and Space Administration) and the Beach Monitoring Program of Louisiana Department of Health and Hospitals is gratefully acknowledged.
references
Ahammed, M., Chaudhuri, M., 1996. Sand-based filtration/ adsorption media. Aqua e Journal of Water Supply: Research and Technology 45 (2), 67e71. An, Y.-J., Kampbell, D.H., Peter Breidenbach, G., 2002. Escherichia coli and total coliforms in water and sediments at lake marinas. Environmental Pollution 120 (3), 771e778. Anderson, K.L., Whitlock, J.E., Harwood, V.J., 2005. Persistence and differential survival of fecal indicator bacteria in subtropical waters and sediments. Applied and Environmental Microbiology 71 (6), 3041e3048.
473
Boehm, A.B., Grant, S.B., Kim, J.H., Mowbray, S.L., McGee, C.D., Clark, C.D., Foley, D.M., Wellman, D.E., 2002. Decadal and shorter period variability of surf zone water quality at Huntington beach, California. Environmental Science and Technology 36 (18), 3885e3892. Boehm, A.B., Keymer, D.P., Shellenbarger, G.G., 2005. An analytical model of enterococci inactivation, grazing, and transport in the surf zone of a marine beach. Water Research 39 (15), 3565e3578. Bonilla, T.D., Nowosielski, K., Cuvelier, M., Hartz, A., Green, M., Esiobu, N., McCorquodale, D.S., Fleisher, J.M., Rogerson, A., 2007. Prevalence and distribution of fecal indicator organisms in South Florida beach sand and preliminary assessment of health effects associated with beach sand exposure. Marine Pollution Bulletin 54 (9), 1472e1482. Burkhardt, W., Calci, K.R., Watkins, W.D., Rippey, S.R., Chirtel, S.J., 2000. Inactivation of indicator microorganisms in estuarine waters. Water Research 34 (8), 2207e2214. Byappanahalli, M.N., Fujioka, R.S., 1998. Evidence that tropical soil environment can support the growth of Escherichia coli. Water Science and Technology 38 (12), 171e174. Carrillo, M., Estrada, E., Hazen, T.C., 1985. Survival and enumeration of the fecal indicators Bifidobacterium adolescentis and Escherichia coli in a tropical rain forest watershed. Applied and Environmental Microbiology 50 (2), 468e476. Chambers, M.K., Ford, M.R., White, D.M., Barnes, D.L., Schiewer, S., 2008. Distribution and transport of fecal bacteria at spring thaw in a rural Alaskan Community. Journal of Cold Regions Engineering 22 (1), 16e37. Crabill, C., Donald, R., Snelling, J., Foust, R., Southam, G., 1999. The impact of sediment fecal coliform reservoirs on seasonal water quality in Oak Creek, Arizona. Water Research 33 (9), 2163e2171. Craig, D.L., Fallowfield, H.J., Cromar, N.J., 2004. Use of microcosms to determine persistence of Escherichia coli in recreational coastal water and sediment and validation with in situ measurements. Journal of Applied Microbiology 96 (5), 922e930. Davies, C., Evison, L., 1991. Sunlight and the survival of enteric bacteria in natural waters. Journal of Applied Bacteriology 70 (3), 265e274. Davies, C., Long, J., Donald, M., Ashbolt, N., 1995. Survival of fecal microorganisms in marine and freshwater sediments. Applied and Environmental Microbiology 61 (5), 1888e1896. Davies-Colley, R.J., Bell, R.G., Donnison, A.M., 1994. Sunlight inactivation of enterococci and fecal coliforms in sewage effluent diluted in seawater. Applied and Environmental Microbiology 60 (6), 2049e2058. Davies-Colley, R.J., Donnison, A.M., Speed, D.J., 1997. Sunlight wavelengths inactivating faecal indicator microorganisms in waste stabilisation ponds. Water Science and Technology 35 (11e12), 219e225. Davies-Colley, R.J., Donnison, A.M., Speed, D.J., Ross, C.M., Nagels, J.W., 1999. Inactivation of faecal indicator microorganisms in waste stabilisation ponds: interactions of environmental factors with sunlight. Water Research 33 (5), 1220e1230. Dura´n, A.E., Muniesa, M., Me´ndez, X., Valero, F., Lucena, F., Jofre, J., 2002. Removal and inactivation of indicator bacteriophages in fresh waters. Journal of Applied Microbiology 92 (2), 338e347. Elmir, S.M., Wright, M.E., Abdelzaher, A., Solo-Gabriele, H.M., Fleming, L.E., Miller, G., Rybolowik, M., Peter Shih, M.-T., Pillai, S.P., Cooper, J.A., Quaye, E.A., 2007. Quantitative evaluation of bacteria released by bathers in a marine water. Water Research 41 (1), 3e10. Evanson, M., Ambrose, R.F., 2006. Sources and growth dynamics of fecal indicator bacteria in a coastal wetland system and potential impacts to adjacent waters. Water Research 40 (3), 475e486.
474
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 6 5 e4 7 4
Evison, L., 1988. Comparative studies on the survival of indicator organisms and pathogens in fresh and sea water. Water Science and Technology 20, 11e12. Francy, D.S., Darner, R.A., 2006. Procedures for developing models to predict exceedances of recreational water-quality standards at coastal beaches. U.S. Geological Survey Techniques and Methods 6eB5, 34. Frick, W., Ge, Z.F., Zepp, R.G., 2008. Nowcasting and forecasting concentrations of biological contaminants at beaches: a feasibility and case study. Environmental Science and Technology 42 (13), 4818e4824. Fries, J.S., Characklis, G.W., Noble, R.T., 2006. Attachment of fecal indicator bacteria to particles in the Neuse River Estuary, N.C. Journal of Environmental Engineering 132 (10), 1338e1345. Fujioka, R.S., Hashimoto, H.H., Siwak, E.B., Young, R.H., 1981. Effect of sunlight on survival of indicator bacteria in seawater. Applied and Environmental Microbiology 41 (3), 690e696. Gameson, A.L.H., Gould, D.J., 1975. Effects of solar radiation on the mortality of some terrestrial bacteria in sea water. In: Gameson, A.L.H. (Ed.), Discharge of Sewage from Sea Outfalls. Pergamon Press, Oxford, pp. 209e219. Grant, S.B., Sanders, B.F., 2010. Beach boundary layer: a framework for addressing recreational water quality impairment at enclosed beaches. Environmental Science and Technology 44 (23), 8804e8813. Halliday, E., Gast, R.J., 2011. Bacteria in beach sands: an emerging challenge in protecting coastal water quality and bather health. Environmental Science and Technology 45 (2), 370e379. Hanes, N.B., Fragala, R., 1967. Effect of seawater concentration on survival of indicator bacteria. Journal of the Water Pollution Control Federation 39 (1), 97e104. He, L.M., He, Z.L., 2008. Water quality prediction of marine recreational beaches receiving watershed baseflow and stormwater runoff in southern California, USA. Water Research 42 (10e11), 2563e2573. Heberger, M.G., Durant, J.L., Oriel, K.A., Kirshen, P.H., Minardi, L., 2008. Combining real-time bacteria models and uncertainty analysis for establishing health advisories for recreational waters. Journal of Water Resources Planning and Management 134 (1), 73e82. Hou, D.Y., Rabinovici, J.M.S., Boehm, B.A., 2006. Enterococci predictions from partial least squares regression models in conjunction with a single-sample standard improve the efficacy of beach management advisories. Environmental Science and Technology 40 (6), 1737e1743. Hua, J., An, P., Winter, J., Gallert, C., 2003. Elimination of COD, microorganisms and pharmaceuticals from sewage by trickling through sandy soil below leaking sewers. Environmental Science and Technology 37 (18), 4395e4404. Irvine, K.N., Pettibone, G.W., 1993. Dynamics of indicator bacteria populations in sediment and river water near a combined sewer outfall, Buffalo River, New York. Environmental Technology 14, 531e542. Kay, D., Jones, F., Wyer, M.D., Fleisher, J.M., Salmon, R.L., Godfree, A.F., Zelenauch-Jacquotte, A., Shore, R., 1994. Predicting likelihood of gastroenteritis from sea bathing: results from randomised exposure. The Lancet 344 (8927), 905e909. Keiner, L.E., Yah, X.H., 1998. A neural network model for estimating sea surface chlorophyll and sediments from thematic mapper imagery. Remote Sensing of Environment 66, 153e165. Kelsey, H., Porter, D.E., Scott, G., Neet, M., White, D., 2004. Using geographic information systems and regression analysis to evaluate relationships between land use and fecal coliform bacterial pollution. Journal of Experimental Marine Biology and Ecology 298 (2), 197e209. Kim, J.H., Grant, S.B., McGee, C.D., Sanders, B.F., Largier, J.L., 2004. Locating sources of surf zone pollution: a mass budget
analysis of fecal indicator bacteria in Huntington Beach, California. Environmental Science and Technology 38, 2626e2636. Lin, B., Syed, M., Falconer, R.A., 2008. Predicting faecal indicator levels in estuarine receiving waters e an integrated hydrodynamic and ANN modelling approach. Environmental Modelling and Software 23 (6), 729e740. Louisiana Department of Health and Hospitals (LDHH), 2009. Louisiana BEACH Grant Report 2008 Swimming Season. Louisiana Department of Health and Hospitals, Office of Public Health, Center for Environmental Health. McGuigan, Joyce, Conroy, Gillespie, Elmore, M., 1998. Solar disinfection of drinking water contained in transparent plastic bottles : characterizing the bacterial inactivation process. Journal of Applied Microbiology 84 (6), 1138e1148. Mill, A., Schlacher, T., Katouli, M., 2006. Tidal and longitudinal variation of faecal indicator bacteria in an estuarine creek in south-east Queensland, Australia. Marine Pollution Bulletin 52 (8), 881e891. Nevers, M.B., Whitman, R.L., 2005. Nowcast modeling of Escherichia coli concentrations at multiple urban beaches of Southern Lake Michigan. Water Research 39 (20), 5250e5260. Noble, R.T., Lee, I.M., Schiff, K.C., 2004. Inactivation of indicator micro-organisms from various sources of faecal contamination in seawater and freshwater. Journal of Applied Microbiology 96 (3), 464e472. Nystro¨m, T., 2004. Stationary-phase physiology. Annual Review of Microbiology 58 (1), 161e181. Obiri-Danso, K., Jones, K., 2000. Intertidal sediments as reservoirs for hippurate negative campylobacters, salmonellae and faecal indicators in three EU recognised bathing waters in North West England. Water Research 34 (2), 519e527. Papadakis, J.A., Mavridou, A., Richardson, S.C., Lampiri, M., Marcelou, U., 1997. Bather-related microbial and yeast populations in sand and seawater. Water Research 31 (4), 799e804. Pommepuy, M., Guillaud, J., Dupray, E., Derrien, A., Le Guyader, F., Cormier, M., 1992. Enteric bacteria survival factors. Water Science and Technology 25 (12), 93e103. Roper, M.M., Marshall, K.C., 1978. Effects of a clay mineral on microbial predation and parasitism of Escherichia coli. Microbial Ecology 4 (4), 279e289. Sanders, B.F., Arega, F., Sutula, M., 2005. Modeling the dry-weather tidal cycling of fecal indicator bacteria in surface waters of an intertidal wetland. Water Research 39 (14), 3394e3408. Sinton, L.W., Finlay, R.K., Lynch, P.A., 1999. Sunlight inactivation of fecal bacteriophages and bacteria in sewage-polluted seawater. Applied and Environmental Microbiology 65 (8), 3605e3613. Sinton, L.W., Hall, C.H., Lynch, P.A., Davies-Colley, R.J., 2002. Sunlight inactivation of fecal indicator bacteria and bacteriophages from waste stabilization pond effluent in fresh and saline waters. Applied and Environmental Microbiology 68 (3), 1122e1131. Solo-Gabriele, H.M., Wolfert, M.A., Desmarais, T.R., Palmer, C.J., 2000. Sources of Escherichia coli in a coastal subtropical environment. Applied and Environmental Microbiology 66 (1), 230e237. United States Environmental Protection Agency (USEPA), 1986. Bacteriological Ambient Water Quality Criteria for Marine and Freshwater Recreational Waters US EPA-PB86-158-045, Springfield, VA. United States Environmental Protection Agency (USEPA), 1999. Review of Potential Modeling Tools and Approaches to Support the BEACH Program EPA823-R-99e002, Washington, D.C. Zakiuddin, K.S., Modak, J.P., 2010. Formulation of data based ANN model for the human powered Fodder-Chopper. Journal of Theoretical and Applied Information Technology 15 (2), 78e85.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
SANI process realizes sustainable saline sewage treatment: Steady state model-based evaluation of the pilot-scale trial of the process Hui Lu a, George A. Ekama b, Di Wu a, Jiang Feng a,c, Mark C.M. van Loosdrecht d,e, Guang-Hao Chen a,* a
Department of Civil & Environmental Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, PR China b Water Research Group, Department of Civil Engineering, University of Cape Town, Rondebosch 7701, South Africa c School of Chemistry & Environment, South China Normal University, Guangzhou, PR China d Department of Biotechnology, Delft University of Technology, Julianalaan 67, NL-2628 BC Delft, The Netherlands e KWR Watercycle Research Institute, The Netherlands
article info
abstract
Article history:
A steady state model was developed for evaluating the sulfur cycle based SANI process.
Received 20 January 2011
The model comprises: 1) a COD-based anaerobic hydrolysis kinetics model to determine
Received in revised form
removal of biodegradable COD and sulfate under different hydraulic retention time (HRT)
27 October 2011
and sludge retention time (SRT), 2) an element (C, H, O, N, P, S), COD and charge mass
Accepted 6 November 2011
balanced stoichiometric part for prediction of the concentrations of alkalinity (H2 CO3
Available online 18 November 2011
alkalinity þ H2S alkalinity), COD, sulfate, sulfide, nitrate and free saline ammonia in anaerobic sulfate reduction, anoxic autotrophic denitrification and aerobic autotrophic
Keywords:
nitrification, and 3) an inorganic carbon (HCO 3 ) and sulfide (H2S/HS ) mixed weak acid/
SANI pilot system
base chemistry part for pH prediction. Through characterization of the sewage organic
Saline water supply
matter and determination of the anaerobic hydrolysis kinetic rate and other relevant
Saline sewage
parameters, the steady state model was calibrated to a pilot plant for the SANI process.
Steady state model
The model predictions agreed well with the experimental data of the pilot-scale trial, demonstrating that the model developed from this study can explain the causes and conditions for the different bioprocesses and minimal sludge production in the SANI process. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
As an alternative natural water resource, seawater toilet flushing has been practiced in Hong Kong since 1950s (WSD, 2010). Seawater toilet flushing generates saline sewage which possesses a typical ratio of COD to sulfate sulfur at 2.4 mg COD/mg S. It enables us to develop a novel Sulfate
reduction, Autotrophic denitrification and Nitrification Integrated (SANI) process (Lau et al., 2006; Lu et al., 2009; Tsang et al., 2009; Wang et al., 2009). This innovative process removes organic matter by sulfate reducing bacteria (SRB) under anaerobic conditions and nitrate by autotrophic denitrifiers under anoxic conditions using the completely dissolved sulfide from the biological sulfate reduction (BSR). Our
* Corresponding author. Tel.: 852 23588752; fax: 852 23581534. E-mail address:
[email protected] (G.-H. Chen). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.031
476
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
Abbreviations ADN AHRT ALR AN ANR bADN
Autotrophic denitrifiers Actual hydraulic retention time (h) Ammonia loading rate (kg FSA-N/m3/d) Autotrophic nitrifiers Ammonia nitrification rate Endogenous respiration rate of autotrophic denitrifiers (d1) Endogenous respiration rate of autotrophic bAN nitrifiers (d1) Endogenous respiration rate of anaerobic bAD acidogens (d1) BAR1 Anoxic bioreactor BAR2 Aerobic bioreactor BPO Biodegradable particulate organics BSO Biodegradable soluble organics C Carbon COD Chemical oxygen demand CxHyOzNaPb Biodegradable particulate organics in the system influent CkHlOmNnPp Biomass composition in the SRUSB; Ck0 Hl0 Om0 Nn0 Pp0 Biomass composition in the anoxic reactor (BAR1) Ck00 Hl00 Om00 Nn00 Pp00 Biomass composition in the aerobic reactor (BAR2) Electrons available for redox reaction per mole of g0B autotrophic denitrifier biomass Ck0 Hl0 Om0 Nn0 Pp0 g00B Electrons available for redox reaction per mole of autotrophic; nitrifier biomass Ck00 Hl00 Om00 Nn00 Pp00 0 E Mass of COD existing in the anoxic reactor as autotrophic denitrification biomass and endogenous sludge per day as a fraction of the mass of nitrate reduced in the anoxic filter per day at the steady state 00 Mass of COD existing in the aerobic reactor as E autotrophic nitrification biomass and endogenous sludge per day as a fraction of the mass of ammonia reduced in the aerobic reactor per day at the steady state f Fraction of H2 PO 4 in the orthophosphate species formed Fraction of bypass flow (hydraulic shortfb circuiting) in the SRUSB FBSO Fermentable biodegradable soluble organics Fraction of biodegradable particulate organics in fBPO influent Ratio of COD-to-VSS or COD-to-Mass for soluble fcv organics Ratio of TOC-to-VSS or TOC-to-Mass for soluble fc organics Fraction of dead space in the SRUSB fd Fraction of endogenous residual fED Fraction of fermentable biodegradable soluble fFBSO organics in influent Ratio of TKN-to-VSS or TKN-to-Mass for soluble fn organics
fp
Ratio of TP-to-VSS or TP-to-Mass for soluble organics Fraction of un-biodegradable particulate organics fUPO in influent Fraction of un-biodegradable soluble organics in fUSO influent FSA Free saline ammonia fVSS_Biomass Fraction of biomass VSS in the combined VSS of the SRUSB fVSS_BPO Fraction of BPO VSS in the combined VSS of the SRUSB fVSS_UPO Fraction of UPO VSS in the combined VSS of the SRUSB H Hydrogen Hydrolysis rate (mg COD/L/d) rh HRT Hydraulic retention time (d) Rate of autotrophic denitrification KDN Maximum specific hydrolysis rate of acidogens KADm (mg COD/mg COD/d) Rate of autotrophic nitrification KN Half saturation constant of hydrolysis reaction by KADs acidogens (mg COD/L) mADmax Maximum specific growth rate of acidogens (d1) Mw Molar weight (g/mole) N Nitrogen NDR Nitrate denitrification rate NHRT Nominal hydraulic retention time (h) NLR Nitrate loading rate (kg NO3 eN/m3/d) O Oxygen 2 OP Orthophosphate (OP ¼ H2 PO 4 þ HPO4 ) P Phosphorus Influent flow rate (m3/d) Qi Effluent flow rate (m3/d) Qe Waste flow rate (m3/d) Qw RO Reverse osmosis Sludge age or sludge retention time in SRUSB RS reactor (d) Sludge age or sludge retention time in anoxic filter R0S (d) Sludge age or sludge retention time in aerobic R00S filter (d) S Sulfur Biodegradable particulate COD (mg COD/L) Sbp Initial biodegradable particulate COD (mg COD/L) Sbpi Hydrogen sulfide concentration (mg COD/L) SH Particulate COD in the reactor of hydrolysis batch Sp test Initial particulate COD concentration in the Spi reactor of hydrolysis batch test SRB Sulfate reducing bacteria SRT Sludge retention time (d) SRUSB Sulfate reduction up-flow sludge bed SWTF Seawater toilet flushing TN Total nitrogen TP Total phosphorous TSS Total suspended solid (g TSS/L) UPO Un-biodegradable particulate organics USO Un-biodegradable soluble organics
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
VFA VSS YADN YAN
Volatile fatty acids Volatile Suspended Solids The yield coefficient of autotrophic denitrifying biomass (mg COD biomass/mg NO3eN reduced) The yield coefficient of autotrophic nitrification biomass (mg COD/mg N oxidized)
500-day lab-scale study (Tsang et al., 2009; Wang et al., 2009) has demonstrated no emission of H2S and CH4 as well as little production of excess sludge. Following the successful labscale study, a steady state model for the SANI lab-scale system was developed which comprises a stoichiometric part for each of sulfate reduction in the anaerobic reactor, autotrophic denitrification in the anoxic bioreactor and nitrification in the aerobic bioreactor (Lu et al., 2009). We have recently conducted a pilot-scale trial of this novel process for treating 10 m3/d screened real saline sewage at the Tung Chung Sewage Pumping Station in Hong Kong near the Hong Kong International Airport (Fig. 1). Some aspects of the experimental evaluation of this pilot trial are given by Lu et al. (2011). This paper evaluates the pilot-scale trial of the SANI process by (1) applying the steady state model of Poinapen and Ekama (2010a) for biological sulfate reduction (BSR) with primary sewage sludge in an up-flow anaerobic sludge bed (UASB) reactor to BSR with the full raw wastewater flow also in sulfate reduction up-flow sludge bed (SRUSB) reactor and (2) measuring the volumetric and surface specific denitrification and nitrification rates in the anoxic and aerobic biofilm reactors (BAR1 and BAR2). The calibrated models can be used to optimize the process reactor design and operation of a fullscale SANI process for Hong Kong, which has been scheduled by the Drainage Services Department of the Hong Kong SAR Government (Fig. 2). This will be presented in a future project.
YAD ZAD ZED
477
The yield coefficient of anaerobic acidogens (mg COD biomass/mg COD degraded) Anaerobic acidogens concentration (mg COD/L) Endogenous residual (mg COD/L)
Mass balanced steady state models are based on the slowest process rate that governs the overall behavior of the system. This process rate determines the system design and operating parameters. Calibrated steady state models enable the system design and suitable operating parameters to be determined, such as reactor volume and hydraulic retention time. For established systems with known design and operation parameters, like the SANI pilot plant, mass balance steady state models can be used to determine bioprocess kinetic rates for application in design and provide a basis for cross-checking kinetic simulation model results (So¨temann et al., 2005; Poinapen and Ekama, 2010a,b). Based on sulfur, chemical oxygen demand (COD) and nitrogen mass balances, the calibrated steady state model can be used to predict sludge production and effluent quality, which are important for process evaluation of the SANI pilot plant. The lab-scale SANI system was operated and continuously fed with synthetic saline sewage containing the main components of glucose, sodium acetate, yeast extract (Wang et al., 2009). Since almost all the influent COD was biodegradable soluble organics (BSO), the anaerobic hydrolysis kinetics of biodegradable particulate organics (BPO) and accumulation of un-biodegradable particulate organics (UPO) were not included in the previous steady state model of the SANI system. For the steady state model of the SANI pilotscale system fed with real saline sewage, separation of particulate organics into BPO and UPO and BPO hydrolysis
Fig. 1 e Photograph of the SANI pilot plant.
478
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
Q SRUSB (Anaerobic)
BAR1 (Anoxic) Q
3Q SRT=90d NHRT=16.3hb AHRT=4.1hb
BAR2 (Aerobic) 3.5Q
SRT=110d NHRT=9.4hc AHRT=2.7hc
SRT=42d NHRT=9.4hd AHRT=2.7hd
2.5Q
Qa
Note: a: influent flow rate Q=10 m3/d; b: NHRT=VSRUSB/Q=6.8 m3/10 m3/d=16.3 h and AHRT=VSRUSB/(3Q+Q)=6.8 m3/40 m3/d=4.1 h, where VSRUSB is effective volume of the SRUSB and 3Q means the internal recirculation of the SRUSB; c: NHRT=VBAR1/Q=3.9 m3/10 m3/d= 9.4 h and AHRT=VBAR1/(2.5Q+Q)=3.5 m3/35 m3/d=2.7 h, where VBAR1 is effective volume of the BAR1 and 2.5Q means the recirculation from BAR2 to BAR1; d: NHRT=VBAR2/Q=3.9 m3/10 m3/d= 9.4 h and AHRT=VBAR2/(2.5Q+Q)=3.5 m3/35 m3/d=2.7 h, where VBAR2 is effective volume of the BAR2 and 2.5Q means the recirculation from BAR2 to BAR1;
Fig. 2 e Flow diagram of the SANI pilot system.
need to be included, because the total influent COD of the SANI pilot plant contains about 50.5% of BPO (see Fig. 3) that has to be hydrolyzed prior to utilization by the sulfate reducing bacteria (SRB). Therefore, compared with the previous steady state model, this revised SANI BSR model includes determination of an influent UPO fraction, anaerobic hydrolysis kinetics of real wastewater BPO and, also characterizes the organics compositions of influent wastewater and volatile suspended solids (VSS) in the SANI pilot plant in terms of CxHyOzNaPb to establish the CHONPS, COD and charge mass balanced stoichiometry. Also added are the stoichiometry of autotrophic denitrification with the HS as electron donor, which is slightly different from the stoichiometry with H2S as electron donor (see Section 3.3) and the stoichiometry of hydrogen sulfide (H2S and HS) oxidation to SO2 4 by dissolved oxygen (DO) because these bioprocesses were observed to take place in the aerobic bioreactor (BAR2). Therefore, this study is aimed at extending the steady state SANI system model for a detailed evaluation of the SANI pilot plant operated with 10 m3/day of real screened saline sewage for 225 days in Hong Kong (Lu et al., 2011). The main tasks of this study include hydrolysis model calibration and influent
UPO fraction determination with anaerobic hydrolysis batch testing to determine the hydrolysis kinetics with real saline sewage and measurements of the volumetric and media surface specific denitrification and nitrification rates. The model predictions will be compared with the experimental data measured during the steady state operation of the SANI pilot plant.
2.
Materials and methods
2.1.
Pilot plant information
The pilot-scale SANI plant system, as shown in Figs. 1 and 2, was established and stably operated for 225 days (Lu et al., 2011). The whole system was continuously fed with 6-mm screened saline sewage taken from the Tung Chung sewage pumping station, which contained on average 431 mg/L COD (organic only), 44.8 mg N/L ammonium, 587.4 mg/L sulfate or 195.8 mg sulfate-S/L, 4.0 mg H2SeS/L and 5000 mg/L chloride. The plant was built with a sulfate reducing up-flow sludge bed (SRUSB) reactor, and fixed media filled anoxic (BAR1) and
Fig. 3 e Characterization of influent and VSS organics of the SRUSB at the steady state (Notes 1e3 are described in the text).
479
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
aerobic bioreactors (BAR2) for autotrophic denitrification and nitrification, respectively. At steady state operation of the pilot plant, the actual Hydraulic Retention Times (HRTs) (with recirculation flow considered) and nominal HRTs (without recirculation flow considered) of the SRUSB, the anoxic bioreactor (BAR1), and the aerobic bioreactor (BAR2) were maintained at 4.1 and 16.3, 2.7 and 9.4, and 2.7 and 9.4 h, respectively (Fig. 2). The recirculation flow ratio between the aerobic bioreactor and the anoxic bioreactor was controlled at 2.5 Qi (Qi ¼ 10 m3/d, the plant influent flow rate), while the SRUSB had an internal recirculation flow of 3Qi to improve the mixing condition in this bioreactor. Details of the SANI pilot plant, its operating conditions and some sludge production results are reported by Lu et al. (2011).
are shown in Table 3. Details on how this characterization was determined from the SANI system measurements are presented in this paper. For instance, the particulate COD in the final effluent was determined to be 23.2 mg COD/L, and particulate COD increase at the bottom of the SRUSB was 323.6 g COD/d daily at the plant capacity of 10 m3/d (equivalent to 32.4 mg COD/L). So the COD concentration of the influent un-biodegradable particulate organics (UPO) was determined to be 23.2 þ 32.4 ¼ 55.6 mg COD/L or a UPO fraction ( fUPO) ¼ 55.6/431 ¼ 0.13, which appeared reasonable for an unsettled domestic wastewater compared with activated sludge treatment ( fUPO ¼ 0.10 w 0.15) (Ekama and Wentzel, 1999). The same reasoning was applied to determine the VSS, TOC, OrgN and OrgP of the influent UPO (Table 3).
2.2. Characterization of the performance of the SANI pilot plant
2.3.
During the steady state operation of the pilot plant, influents and effluents of all bioreactors were characterized regularly through 24-hr composite sampling and sample analysis (Lu et al., 2011). The characteristics of the influent and effluent for each bioreactor of the SANI pilot plant are summarized in Table 1 and COD, N, P and S mass balances for the individual reactors of the pilot plant and system as a whole are given in Table 2aed. In order to conduct mass balance analysis of the plant, we assumed: 1) the particulate organics and filtered organics in the final effluent are un-biodegradable, 2) the particulate organics accumulating in the bottom sludge of the SRUSB are un-biodegradable (UPO), and 3) the organics in 1) and 2) above originate from the influent to the SANI pilot plant (i.e. no bioprocess generated soluble and particulate unbiodegradable organics). Based on these assumptions, the raw wastewater influent was characterized, the results of which
Anaerobic hydrolysis batch testing
Batch testing was conducted to determine the anaerobic hydrolysis kinetic parameters: KADm, the maximum specific hydrolysis rate constant, KADs, the Monod half saturation constant, and mADmax, the maximum specific growth rate of acidogens (So¨temann et al., 2005; Ristow et al., 2006). The anaerobic hydrolysis batch testing was conducted using a completely mixed Perspex digester with a working volume of 5 L under a controlled temperature at 25 C for 24 h. Concentrations of unfiltered and filtered COD, hydrogen sulfide and sulfate in the influent and in the SRUSB sludge were measured, respectively, before mixing in the batch testing reactors. This allowed determination of the initial concentrations of biodegradable and un-biodegradable particulate COD, initial active acidogenic biomass and endogenous residue COD, initial hydrogen sulfide COD and sulfate (SO4eS) in the mixture at the start of the hydrolysis batch test. After mixing, the headspace was purged with
Table 1 e Characteristics of the influent and effluent for each reactor in the SANI pilot plant. BAR1 influent ③a
BAR1 effluent/ BAR2 influent ④a
BAR2 effluent ⑤a
Calculated concentration ③a
52.1 22.3 98.8 60.9 39.7 5.6
23.5 7.1 65.9 17.4 30.9 10.3
12.5 4.3 55.0 18.7 30.2 11.6
9.0 2.4 53.9 10.8 30.7 9.2
21.3 66.7 33.3
195.7 18.0 4.0 4.7
65.4 15.2 124.1 14.4
142.3 25.7 31.3 4.8
169.4 34.3 3.0 1.2
172.3 34.4 0
141.7 35.5
87.5 8.5 70.5 5.7 44.8 6.6 0
85.3 8.6 70.9 5.6 45.4 6.0 0
7.8 1.2 5.8 0.8 5.3 0.7
7.6 1.0 5.7 0.6 5.2 0.8
5.9 0.7 5.8 0.7 5.3 0.7
223.6 100.5 219.0 104.3 4.6 1.0
736.9 170.1 576.7 174.0 160.2 11.2
281.4 75.2 252.8 75.1 28.6 4.6
Parameters
SRUSB a
Effluent ②
VSS (mg/L) Unfiltered COD (mgCOD/L) Filtered COD (mgCOD/L)b
186 55.3 431 132.6 157.9 86.2
SO2 4 (mgS/L) H2S (mgS/L)
Influent ①
Unfiltered TKN (mgN/L) Filtered TKN (mgN/L) FSA (mgN/L) NO 3 (mgN/L) Unfiltered TP (mgP/L) Filtered TP (mgP/L) OP (mgP/L) Total Alkalinity (mgCaCO3/L) H2CO3* Alkalinity (mgCaCO3/L) H2S Alkalinity (mgCaCO3/L)c
a
39.2 36.5 16.9 13.5
4.4 4.4 4.5 1.8
35.7 35.3 16.4 2.2
5.6 3.8 2.7 2.4
23.4 23.2 3.4 16.8
5.5 4.3 2.4 3.2
41.1 36.8 15.4 12.0
5.7 0.5 5.6 0.8 5.3 0.5
5.6 0.6 5.6 0.6 5.3 0.5
6.2 5.6 5.3
247.7 70.5 245.7 71.6 2.0 0.2
120.5 62.4 120.5 62.4 0
296.6 250.8 45.8
a Numbers, ① ② ③ ④ ⑤, correspond to those in Fig. 2; b Excludes sulfide COD; and c The H2S Alkalinity (¼[HS] 50,000 mg/L as CaCO3) was calculated from the total dissolved hydrogen sulfide (ST) and pH value according to the weak acid/base chemistry.
480
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
Table 2 e (a) COD mass balances over individual reactors of the SANI pilot plant and the system as a whole from results in Table 1 and Fig. 2. (b) N mass balances over individual reactors of the SANI pilot plant and the system as a whole from results in Table 1 and Fig. 2. (c) P mass balances over individual reactors of the SANI pilot plant and the system as a whole from results in Table 1 and Fig. 2. (d) S mass balances over individual reactors of the SANI pilot plant and the system as a whole from results in Table 1 and Fig. 2. (a)
[5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24]
Influent COD and H2S ¼ 10 (431 þ 2 4) COD accumulation in SRUSB COD exiting SRUSB as H2S ¼ 10 124.1 2 COD exiting SRUSB as unfiltered COD ¼ 10 98.8 Total COD exiting SRUSB ([2] þ [3] þ [4]) % COD balance over SRUSB ¼ [5]/[1] Sulfide COD into BAR1 ¼ 35 31.3 2 Unfiltered COD into BAR1 ¼ 35 65.9 Total COD into BAR1 ¼ ([7] þ [8]) Unfiltered COD out BAR1 ¼ 35 55 Sulfide COD out BAR1 ¼ 35 3 2 COD equivalent of NO3 denitrification ¼ 35 (13.5e2.2) 2.86 COD utilized in SO4 production ¼ 35 (169.4e142.3) 2 Total COD out BAR1 ¼ [10] þ [11] þ [13] % COD balance over BAR1 ¼ [14]/[9] Unfiltered COD into BAR2 ¼ 35 55 Sulfide COD into BAR2 ¼ 35 3 2 Unfiltered COD out BAR2 ¼ 35 53.9 Sulfide COD out BAR2 ¼ 0 COD utilized in SO4 production ¼ 35 3 2 COD balance over BAR2 ¼ ([18] þ [20])/([16] þ [17]) Unfiltered COD from system ¼ 10 66.7 Total COD exiting system ¼ [2] þ [13] þ [20] þ [22] COD balance over system ¼ [23]/[1]
g COD/d 4390 323.6 2482 988 3794 86% 2191 2307 4498 1925 210 1131
Components of the N balance
[3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]
Components of the N balance [15] [16] [17]
Total N exiting system ¼ [2] þ [8] þ [14] % N balance over system ¼ [15]/[1] % Influent N denitrified ¼ [8]/[1]
Influent TKN ¼ 10 87.5 N in UPO accumulation in SRUSB ¼ 323.6/1.471 0.22 g N/g VSS TKN in effluent of SRUSB ¼ 10 85.3 Total TKN exiting SRUSB ¼ [2] þ [3] % N balance over SRUSB ¼ [4]/[1] TKN and nitrate into BAR1 ¼ 35 (39.1 þ 13.5) TKN and nitrate out BAR1 ¼ 35 (35.7 þ 2.2) Nitrate denitrified in BAR1 ¼ 35 (13.5e2.2) Total N out BAR1 ¼ [7] þ [8] % N balance over BAR1 ¼ [9]/[6] TKN and nitrate into BAR2 ¼ 35 (35.7 þ 2.2) TKN and nitrate out BAR2 ¼ 35 (23.4 þ 16.8) % N balance over BAR2 ¼ [12]/[11] N in system effluent ¼ 10 (23.4 þ 16.8)
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]
Influent TP ¼ 10 7.8 P in UPO accumulation in SRUSB ¼ 323.6/1.471 0.024 g P/g VSS P exiting SRUSB ¼ 10 7.6 Total P exiting SRUSB ¼ [2] þ [3] % P balance over SRUSB ¼ [4]/[1] TP into BAR1 ¼ 35 5.9 TP out BAR1 ¼ 35 5.7 % P balance over BAR1 ¼ [7]/[6] TP into BAR2 ¼ 35 5.7 TP out BAR2 ¼ 35 5.6 % P balance over BAR2 ¼ [10]/[9] TP in system effluent ¼ 10 5.6 TP exiting system ¼ [2] þ [12] % P balance over system ¼ [13]/[1]
(d)
4032 90% 1925 210 1886 0 210 98%
Components of the S balance
667 3098 71%a
g N/d 875 48.4 853 901 103% 1845 1327 396 1723 93% 1327
g N/d 846 97% 45%b
(c)
1897
(b)
[1] [2]
(b)
Components of the P balance
Components of the COD balance [1] [2] [3] [4]
Table 2 e (continued )
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11]
Influent SO4 and H2S ¼ 10 (195.7 þ 4.0) S exiting SRUSB ¼ 10 (65.4 þ 124.1) % S balance over SRUSB ¼ [2]/[1] S into BAR1 ¼ 35 (142.3 þ 31.3) S out BAR1 ¼ 35 (69.4 þ 3.0) % S balance over BAR1 ¼ [5]/[4] S into BAR2 ¼ 35 (169.4 þ 3.0) S out BAR2 ¼ 35 (172.3 þ 0) % S balance over BAR2 ¼ [8]/[7] S in system effluent ¼ 10 (172.4 þ 0) % S balance over system ¼ [10]/[1]
g P/d 78.0 5.3 76.0 81.3 104% 206 199 97% 199 196 98% 56 61.3 79%
g S/d 1997 1895 95% 6076 6034 99% 6034 6031 100% 1724 86%c
a The low COD balance over the system is mainly due to the loss of sulfide between the SRUSB and BAR1 reactors from elemental sulfur formation on the interconnecting pipe wall. b Low due to high un-biodegradable soluble organic N in effluent 23.2 mg OrgN/L. c Most S not accounted for in S balance was lost between SRSUB effluent and BAR1 influent, where elemental sulfur was observed on the pipe wall.
nitrogen gas in order to remove oxygen and the batch digester sealed. During the anaerobic hydrolysis testing, the completely mixed liquor samples were taken from the batch reactor every 2 h, which were used to determine the concentrations of unfiltered and filtered COD, hydrogen sulfide (SH), and sulfate (SO4eS).
3.
Steady state model developments
3.1.
Introduction
1407 106% 402
A steady state model for biological sulfate reduction (BSR) in a UASB reactor fed with primary sewage sludge (PSS) was
481
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
Table 3 e Influent characterization of the SANI pilot system. Parameters
VFA
FSA
OP
FBSO
BPO
USO
UPO
Soluble
Particulate
Total
COD (mg COD/L) TOC (mg C/L) TKN (mg N/L) TP (mg P/L) VSS (mg VSS/L)
34.5 12.9 N/A N/A N/A
N/A N/A 44.8 N/A N/A
N/A N/A N/A 5.3 N/A
92.7 40.9 2.5 0.2 N/A
217.5 82.2 8.7 1.1 148.2
30.7 11.5 23.2 0.3 N/A
55.6 20.2 8.3 0.9 37.8
157.9 65.3 70.5 5.8 N/A
273.1 102.4 17 2.0 186
431 167.7 87.5 7.8 186
Note: N/A means not applicable.
developed by Poinapen and Ekama (2010a). This model consists of three sequential parts: 1) a COD mass balanced anaerobic hydrolysis kinetics part which allows determination of the influent un-biodegradable particulate COD fraction ( fUPO) of the feed organic, and the rate of hydrolysis of biodegradable particulate organics (BPO), from which are determined the biodegradable COD removal, hydrogen sulfide production and acidogen biomass concentration (i.e. the BSR products that have COD), 2) COD, C, H, N, O, P, S and charge mass balance stoichiometry, which uses the biodegradable COD removal from part 1 and its composition (i.e. x, y, z, a and b in CxHyOzNaPb) as inputs to determine all the BSR products, those that have COD and those that do not, such as the acidogen biomass and sulfide (H2S and HS) production, ammonia and orthophosphate release and alkalinity gener 2 ation (HCO 3 , H2 PO4 , HPO4 and HS ), and 3) a mixed weak acid base chemistry part from which the pH in the bioreactor is determined from the aqueous BSR products determined with part 2. Poinapen and Ekama (2010a) validated their steady state model for a UASB BSR fed with primary sewage sludge (PSS). They further demonstrated that most organics, including PSS, are carbon deficient in that they can donate more electrons for BSR than supply C for the required increase in alkalinity. Consequently all CO2 produced stays dissolved as HCO 3 , and the alkalinity deficit is supplied by the sulfide system as HS. Without CO2 gas production and low orthophosphate (H2 PO 4 ), the sulfide system controls the UASB reactor pH. This model will be applied to the SRUSB reactor and the composition of the real wastewater organics (x, y, z, a and b in CxHyOzNaPb) determined from the measured pilot plant results. The details of this model can be found in Poinapen and Ekama (2010a) and Lu et al. (2009). The results of applying this model to the SANI pilot plant system (Fig. 2) and characterized wastewater (Table 3) are given in Table 4. Attention in this paper is mainly focused on (1) demonstrating that practically all biodegradable organics (influent BPO þ FBSO) were hydrolyzed and, together with the influent VFA, utilized for BSR and (2) determining the anoxic autotrophic denitrification and nitrification bioprocess rates in the fixed media biofilm part of the SANI system (BAR1 and BAR2).
3.2.
Anaerobic hydrolysis kinetic parameters
The equations below describe the rate of change of biodegradable particulate organics (BPO, Sbp) concentration with time in the batch reactor during the anaerobic BSR controlled by the hydrolysis bioprocess:
dZAD ¼ YAD $rh bAD $ZAD dt
(1a)
dSbp ¼ rh þ bAD $ZAD dt
(1b)
dSH ¼ ð1 YAD Þ$rh dt
(1c)
Table 4 e COD balanced hydrolysis kinetics part 1: acidogenesis (AD) model results as applied to the SANI pilot plant system (Fig. 2) and sewage characteristics (Table 2). Parameter
Value
Comment/Equation
Influent USO COD (Susi) (mg COD/L) Influent UPO COD (Supi) (mg COD/L) Influent VFA COD (Sbsai) (mg COD/L) Influent FBSO COD (Sbsfi) (mg COD/L) Influent BPO COD (Sbpi) (mg COD/L) Influent biodeg. organics (mg COD/L) Residual biodeg. Organics in SRUSB (Sbp) (mg COD/L with respect to sludge bed) Biodeg. COD utilized (DCOD) (mg COD/L) AD biomass conc. (ZAD) (mg COD/L) UPO conc. (Sup) in the SRUSB (mg COD/L) SRUSB bed conc. (mg COD/L) E factor
31
USO in system final effluent Supi ¼ fUPO Sti
Hydrolysis rate (rh) (mg COD/L/h) Sulfide COD conc. (SH) (mg COD/L) H2S as COD produced from VFA (mg S/L) Total H2S produced as COD (mg COD/L) Total H2SeS produced (mg S/L) Total SO4eS reduced (mg S/L) Sulfate reduced (mg SO4/L)
56 34
VFA does not require fermentation
93 217 310 0
344
Sbsfi (FBSO) included with BPO All (>99%) biodeg. organics utilized
500
At long sludge age of 90 days Eq. (3) in Poinapen and Ekama (2010a) Eq. (4) in Poinapen and Ekama (2010a) ZAD þ Sup Eq. (8) in Poinapen and Ekama (2010a) rh ¼ Qi (DCOD)/VSRUSB
302
SH ¼ (1YAD) rh Rhn
1085 7360 8445 0.0264
34 336 168 168 504
336 32/64 Sulfide produced 168 96/32
482
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
dSO4 1 dSH 1 ¼ $ ¼ $ð rh þ bAD $ZAD Þ dt 2 dt 2
(1d)
where rh ¼ volumetric hydrolysis rate (mg COD/L/d) ZAD ¼ acidogenic biomass COD in batch reactor (mg COD/L) YAD ¼ yield coefficient of acidogenic biomass (increased to take account of acetoclastic and hydrogenotrophic sulfidogens) (0.113 mg COD/mg COD) (So¨temann et al., 2005) bAD ¼ endogenous respiration rate of acidogenic biomass (0.041 d1) (So¨temann et al., 2005) Sbp ¼ biodegradable particulate organics (BPO) COD in batch reactor (mg COD/L) SH ¼ hydrogen sulfide COD in batch reactor (H2S þ HS) (mg COD/L) SO4 ¼ sulfate concentration in batch reactor (mg SO4eS/L) 1/2 ¼ mg SO4eS/mg Sulfide COD. After mixing the wastewater feed with the SRUSB sludge in the anaerobic batch test, the concentrations of ZAD and Sbp cannot be separately determined because both are particulate. Therefore, the Sp is used to express the combined particulate COD in the batch testing reactor during the anaerobic hydrolysis (and subsequent “instantaneous” sulfidogenesis), i.e. Sp ¼ ZAD þ ZED þ Sbp þ Sup
(2)
where, Sp ¼ combined particulate COD in the reactor of batch test (mg COD/L) ZAD ¼ acidogen biomass COD in the reactor of batch test (mg COD/L) ZED ¼ endogenous residue COD in the reactor of batch test (mg COD/L) Sbp ¼ residual biodegradable particulate COD in the reactor of batch test (mg COD/L) Sup ¼ un-biodegradable particulate COD in the reactor of batch test (mg COD/L) Since the endogenous residue ( fED ¼ 0) and unbiodegradable particulate COD are constant, the change of particulate COD (Sp) is only caused by the changes of biomass (ZAD) and biodegradable particulate COD (Sbp), as shown in the following Equations achieved by adding Eqs. (1a) and (1b): dSp dZAD dSbp ¼ þ ¼ ðYAD 1Þ$rh or ¼ ð1 YAD Þ$rh dt dt dt
(3)
The change of Sp with time (or the volumetric hydrolysis rate rh) can be obtained by curve fitting model products to measured data. To do this, the determination of starting value (Sp) is important, which comprises ZADi, ZEDi, Sbpi, and Supi, as shown in Eq. (4). Spi ¼ ZADi þ ZEDi þ Sbpi þ Supi
ZEDi ¼ initial endogenous residue COD in the mixture of feed and sludge at beginning of the hydrolysis batch test ( fED ¼ 0 so ZEDi ¼ 0 mg COD/L); Sbpi ¼ initial biodegradable particulate COD in the mixture of feed and sludge at beginning of the hydrolysis batch test (mg COD/L); Supi ¼ initial un-biodegradable particulate COD in the mixture of feed and sludge at beginning of the hydrolysis batch test (mg COD/L) The sludge used in the hydrolysis batch testing was taken from the SRUSB reactor at the steady state and the feed was taken from the influent of the SANI pilot plant. The characterization of the wastewater organics and the VSS of the SRUSB are shown in Table 3 and Fig. 3 respectively and are based on complete hydrolysis and utilization of biodegradable organics (BPO and BSO) (more details in Section 4.2). Thus, the concentrations of ZADi, ZEDi, Supi and Sbpi can be determined. Consequently, the initial combined particulate COD (Spi) is available. The starting values for H2S and SO4 in Equations of (1c) and (1d) can be determined from the concentrations of hydrogen sulfide and sulfate in the mixture of feed and sludge at the beginning of the hydrolysis batch testing. The volumetric anaerobic hydrolysis rate (rh) can be described with Monod kinetics (So¨temann et al., 2005),
(4)
where, Spi ¼ initial combined particulate COD in the mixture of feed and sludge at beginning of the hydrolysis batch test (mg COD/L); ZADi ¼ initial acidogen biomass COD in the mixture of feed and sludge at beginning of the hydrolysis batch test (mg COD/L);
rh ¼
KADm $Sbp mADmax $Sbp $ZAD ðmgCOD=L=dÞ $ZAD ¼ KADs þ Sbp YAD KADs þ Sbp
(5)
where, rh ¼ volumetric hydrolysis rate (mg COD/L/d) KADm ¼ maximum specific hydrolysis rate constant of acidogens in Monod kinetics (mg COD/mg COD/d) KADs ¼ half saturation constant of acidogens (mg COD/L) mADmax ¼ maximum specific growth rate of acidogens (d1) Sbp ¼ biodegradable particulate COD (mg COD/L) ZAD ¼ concentration of acidogen biomass (mg COD/L) YAD ¼ yield coefficient of acidogen biomass (0.113 mg COD/ mg COD) (So¨temann et al., 2005) Since the anaerobic hydrolysis rate is described with Monod kinetics, which is a non-linear equation (see Eq. (5)), the curve fitting method of least square regression between predicted and measured data is used for the estimation of hydrolysis kinetic parameters (KADm and KADs) in this study. The predicted values for Sp, H2S and SO4 can be obtained based on the above differential equations if KADm (or mADmax) and KADs are known. The experimentally measured data set of Sp, H2S and SO4 is used for the non-linear curve regression, and the sum of square error of all data pairs between predicted and measured results can be minimized by the Solver routine in Microsoft Excel (Nikitas and Pappa-Louisi, 2000; Riefler and Smets, 2003). As mentioned above, the initial ZADi and Sbpi concentrations for the batch tests were determined assuming (1) an influent raw wastewater UPO fraction fUPO ¼ 0.13; and (2) all the BPO (and BSO and VFA) were utilized in the SRUSB. The KADm and KADs obtained from the batch test regression were then applied to the steady state SRUSB system to estimate the residual (non-utilized) BPO (Sbp) in the SRUSB, from which revised initial ZADi and Sbpi concentrations for the batch tests were calculated. These revised ZADi and Sbpi
483
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
concentrations in the batch test regression led to negligible change to the KADm and KADs constants and correlation coefficient (R2). Finally the UPO fraction ( fUPO) was varied around the initially estimated 0.13 value. Again there was negligible change in the R2 value in the batch test regression. So the 0.13 fraction was retained. The hydrolysis kinetic parameters finally determined were KADm ¼ 3.25 mg COD/(mg COD/d), KADs ¼ 557 mg COD/L and mADmax ¼ KADm*YAD ¼ 0.37 d1 at 25 C. The KADm and KADs were then applied to calculate and draw the prediction curves of Sp, H2SeS and SO4eS for crosschecking (see Fig. 4). The negligible changes in KADm, KADs and fUPO in batch test results regression confirmed as valid the two assumptions above for calculating the initial ZADi and Sbpi concentrations for the batch tests and it was concluded that practically all (99%) influent BPO (and so FBSO also) were hydrolyzed in the SRUSB and utilized for BSR.
3.3. Stoichiometry of autotrophic denitrification in the anoxic bioreactor In our SANI pilot system, the influent to the Anoxic Bioreactor (BAR1) received the effluent from the SRUSB and the recycled effluent of the Aerobic Bio-reactor (BAR2), as shown in Fig. 2. This mixture contained sufficient sulfide and nitrate for the autotrophic denitrification in the anoxic bioreactor (Table 1). Sulfide donates electrons to nitrate for denitrification and thus the CHONPS, COD and charge mass balanced stoichiometry of the autotrophic denitrification can be derived from element and charge balancing (Ekama, 2009) and is expressed by Eq. (6):
2000
E0 ¼
where 0 ¼ sludge waste flow from anoxic bioreactor (BAR1) (m3/ QW day) Qi0 ¼ influent flow rate to BAR1 (m3/day) R0S ¼ sludge age of BAR1 (d) R0hn ¼ nominal hydraulic retention time of BAR1 (1/day) ZADN ¼ concentration of autotrophic denitrifiers in BAR1 (mg COD/L)
140 120
1850
100
1800
80
1750 1700
60
1650
40
1600
20
1550
0
1500 0.0
0.2
0.4
0.6
0.8
Sp
2300
H2S-S
140
SO4-S
120
2250 2200
S (mgCOD/L)
SO4-S
160
(Sbpi/ZADi=0.13 mgCOD/mgCOD)
2350
100
2150
80
2100 2050
60
2000
40
1950 1900
20
1850
0
1800
1.0
0.0
Time (d)
Sulfur concentration(mgS/L)
H2S-S
0 QW ðZADN þ ZEDN Þ R0hn ðZADN þ ZEDN Þ YADN ¼ ¼ ; 1 þ bADN R0S Qi0 ðNni Nne Þ R0S ðNni Nne Þ
2400
Sulfur concentration(mgS/L)
Sp
1900
S (mgCOD/L)
where, f ¼ fraction of H2 PO 4 in the OP species formed (OP ¼ H2 PO4 þ HPO2 ), 4 g0B ¼ electron donating capacity per mole of autotrophic denitrifier biomass, Ck0 Hl0 Om0 Nn0 Pp0 (¼4k þ l 20m 30n þ 50p ), 0 E ¼ flux of COD exiting from the anoxic bioreactor as the autotrophic denitrification biomass and endogenous residue sludge as a fraction of the flux of nitrate reduced in the anoxic bioreactor at a steady state (net yield of the autotrophic denitrification biomass and endogenous residue), i.e. from the equation of
160
(Sbpi/ZADi=0.76 mgCOD/mgCOD)
1950
0.2
0.4
0.6
0.8
1.0
Time (d)
2800
160
(Sbpi/ZADi=0.42 mgCOD/mgCOD)
2750
Sp
H2S-S
140
SO4-S
2700
120
2650
100
2600
80
2550
60
2500
40
2450
20
2400
Sulfur concentration(mgS/L)
S (mgCOD/L)
2 2 E nE 2 þ ð1EÞ 0 ½nþpð2f Þ HCO ð1EÞNO 3 þ 0 NH4 þ 3 8 10 gB gB 10 E E 2 4 2 1 þf $p 0 H2 PO ð1EÞ 4 þð1f Þ$p 0 HPO4 þ H2 Sþ gB gB 8 8 10 E E 1 0 ½2kmþnþpð2þf Þ H2 O/ 0 Ck0 Hl0 Om0 Nn0 Pp0 þ SO2 gB gB 8 4 1 2 2 E þ ð1EÞN2 þ ð1EÞ 0 ½knþpð2f Þ CO2 ð6Þ 10 8 10 gB
0 0.0
0.2
0.4
0.6
0.8
1.0
Time (d)
Fig. 4 e Correlation between experimental data (points) and prediction curves (lines) for the KADm (3.25 mg COD/mg COD/d) and KADs (557 mg COD/L) determined by least squares regression on anaerobic BSR batch test results.
484
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
ZEDN ¼ concentration of endogenous residue in BAR1 (mg COD/L) Nni, Nne ¼ influent and effluent nitrate concentration of BAR1 with respect to system influent flow (Qi) (mg NO3eN/L) YADN ¼ yield coefficient of the autotrophic denitrifying microorganisms (mg COD biomass/mg NO3eN reduced) bADN ¼ endogenous respiration rate of autotrophic denitrifying microorganisms (1/d). The SRUSB reactor also produces dissociated sulfide (HS) to meet the alkalinity requirement of the BSR. The stoichiometry of autotrophic denitrification with HS is slightly different from Eq. (6), i.e. the 2/8 in the HCO 3 term changes to 1/8, the 2/8 in the H2O term to 3/8 and the 2/8 in the CO2 term to 1/8, as shown in Eq. (7). 1 2 E nE 2 þ ð1 EÞ 0 ½ n þ pð2 f Þ HCO ð1 EÞNO 3 þ 0 NH4 þ 3 8 10 gB gB 10 E E 1 3 4 2 þ f $p 0 H2 PO ð1 EÞ 4 þ ð1 f Þ$p 0 HPO4 þ HS þ gB gB 8 8 10 E E 1 0 ½2k m þ n þ pð2 þ f Þ H2 O/ 0 Ck0 Hl0 Om0 Nn0 Pp0 þ SO2 gB gB 8 4 1 1 2 E þ ð1 EÞN2 þ ð7Þ ð1 EÞ 0 ½k n þ pð2 f Þ CO2 10 8 10 gB Simplifying Eqs (6) and (7) by assuming zero biomass growth (E ¼ 0) and negligible P ( p ¼ 0) yields: For H2S: 1 2 1 3 1 1 1 þ N2 þ CO2 HCO NO H2 O þ SO2 3 þ 3 þ H2 S/ 20 10 8 20 8 4 10 20
(8)
And for HS: 3 2 1 2 1 1 3 þ N2 þ HCO CO2 þ NO H2 O/ SO2 3 þ HS þ 3 40 10 8 40 8 4 10 40
(9)
Eq. (8) shows that there is a net alkalinity consumption of 0.893 mg/L as CaCO3 per mg NO3eN/L denitrified or 0.625 mg/L as CaCO3 per mg H2SeS/L oxidized. Eq. (9) shows that there is a consumption of 6.25 (¼50/8) mg/L as CaCO3 sulfide alkalinity but a production of 3.75 (¼50 3/40) mg/L as CaCO3 H2CO3 alkalinity in the form of HCO 3 alkalinity. Therefore, Eqs. (8) and (9) show that although the autotrophic denitrification with HS consumes the sulfide alkalinity (HS) and generates the H2 CO3 alkalinity (HCO 3 ), the net total alkalinity consumption is exactly the same as denitrification with H2S, i.e. Eq. (8): (50,000/20)O(2/ 10 14,000) ¼ 0.893 mg/L as CaCO3 per mg NO3eN/L denitrified and from Eq. (9): [(1/8 þ 3/40) 50,000]O(2/ 10 14,000) ¼ 0.893 mg/L as CaCO3 per mg NO3eN/L denitrified.
3.4. Stoichiometry of nitrification in the aerobic bioreactor The main reaction in the aerobic reactor (BAR2) was autotrophic nitrification. As outlined by Ekama (2009), the stoichiometry of the nitrification in this reactor is described as, 1 E00 1 E00 n00 1 þ þ 14 00 ðn þ pÞ HCO O2 3 þ NH4 þ 14 00 7 8gB 7 4 gB 8 00 00 00 00 00 14p E 14E 1 n E NO þ H2 PO Ck00 Hl00 Om00 Nn00 Pp00 þ 14 4/ 3 00 00 00 8gB 8gB 14 8gB 00 00 1 E 3 E þ 14 00 ðk00 þ n00 þ p00 Þ CO2 þ 14 ð2k00 m00 þ 7 8gB 14 8g00B E 2n00 þ 3p00 Þ ð10Þ H2 O 16
Where, g00B ¼ electron donating capacity per mole of autotrophic nitrifier biomass, Ck00 Hl00 Om00 Nn00 Pp00 ð¼ 4k00 þ l00 2m00 3n00 þ 5p00 Þ, 00 E ¼ flux of COD exiting from the aerobic filter as the autotrophic nitrification biomass and endogenous residue sludge as a fraction of the flux of ammonia oxidized at a steady state, i.e. from the N-based kinetic equation of E00 ¼
00 QW ðZAN þ ZEN Þ R00hn ðZAN þ ZEN Þ YAN ¼ 00 ¼ 1 þ bAN R00S Qi00 ðNai Nae Þ RS ðNai Nae Þ
where 00 ¼ sludge waste flow from aerobic bioreactor (BAR2) QW 3 (m /day) Qi00 ¼ influent flow rate to BAR2 (m3/day) R00S ¼ sludge age of BAR2 (d) R00hn ¼ nominal hydraulic retention time of BAR2 (1/day) ZAN ¼ concentration of autotrophic nitrifiers in BAR2 (mg COD/L) ZEN ¼ concentration of endogenous residue in BAR2 (mg COD/L) Nai, Nae ¼ influent and effluent ammonium concentration of BAR2 with respect to system influent flow (Qi) (mg NH4eN/L) YAN ¼ yield coefficient of the autotrophic nitrifying microorganisms (mg COD/mg NH4eN nitrified) bAN ¼ endogenous respiration rate of autotrophic nitrifying microorganisms (1/d) In order to model the change in alkalinity, the stoichiometry of aerobic sulfide oxidation to SO2 4 with dissolved oxygen in BAR2 also needs to be modeled because both nitrification and H2S oxidation consume alkalinity. H2S and HS oxidation to SO2 4 with oxygen and HCO3 HCO3 consumption and sulfide oxidizing biomass growth is given by Eq. (11) and Eq. (12): 00 1 E00 2 E00 00 00 E H PO H2 S þ n00 00 NHþ þ 00 ðn p00 Þ HCO 4 þp 4 þ 3 00 2 gB g 8 8 gB B 00 00 1E 2 E 1 O2 / þ ðk00 n00 þ p00 Þ CO2 þ SO2 4 8 gB 8 4 E00 1 E00 2 E00 þ 00 ð2k00 m00 þ 00 Ck00 Hl00 Om00 Nn00 Pp00 þ gB 2 8 gB þ n00 þ 3p00 Þ H2 O
ð11Þ
00 1 E00 1 E00 00 00 E 00 þ p H PO þ ðn p Þ HCO HS þ n00 00 NHþ þ 2 4 4 3 gB g00B 8 8 g00B 1 E00 1 E00 00 1 O2 / þ ðk n00 þ p00 Þ CO2 þ SO2 4 8 g00B 8 4 00 00 00 E 1E 3 E þ 00 ð2k00 m00 þ 00 Ck00 Hl00 Om00 Nn00 Pp00 þ gB 2 8 gB 00 00 þ n þ 3p Þ H2 O
ð12Þ
Simplifying Eqs. (11) and (12) for negligible biomass growth 00 (or small, E ¼ 0) and zero (or low) P yields: 1 1 1 1 2 1 1 H2 S þ HCO 3 þ O2 / SO4 þ H2 O þ CO2 8 4 4 8 4 4
(13)
1 1 1 1 2 1 1 HS þ HCO 3 þ O2 / SO4 þ H2 O þ CO2 8 8 4 8 8 8
(14)
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
From Eq. (13), the alkalinity consumption (HCO 3 ) in the sulfide oxidation is 50/4 8/32 ¼ 3.13 mg/L as CaCO3 per mg H2SeS oxidized to SO2 4 eS. Eq. (14) shows the same alkalinity consumption. The only difference between Eqs. (13) and (14) is that in Eq. (14) H2S alkalinity (HS) and H2 CO3 alkalinity (HCO 3 ) are consumed. Because the primary reaction of BAR2 is nitrification, alkalinity also decreases due to this process. Alkalinity consumption of NHþ 4 to NO3 (or NO2 ) is the well known 7.14 mg/L as CaCO3 per mg FSA-N/L nitrified.
4.
Model parameters
4.1.
Yield coefficients and endogenous respiration rates
Even though the ambient temperature of the SRUSB reactor was 25 C, the same constant values of YAD and bAD at 35 C of 0.113 mg COD/mg COD and 0.04 d1 respectively were used here (Lu et al., 2009; Poinapen and Ekama, 2010a). Even significant changes in these constants make a very small difference to the sludge production from the SRUSB, which is mainly set by the UPO fraction ( fUPO) or concentration (Supi) in the influent wastewater (Table 4). Since the yield coefficient (YADN) of autotrophic denitrifiers is 0.57 g VSS/g NO3eN denitrified (Claus and Kutzner, 1985; Oh et al., 2003) and the measured conversion factor of VSS to COD is 1.44 mg COD/mg VSS, the yield coefficient in terms of COD is 0.82 mg COD biomass/mg NO3eN denitrified. The endogenous respiration rate (bADN) of denitrifiers follows the default value (0.05 d1 at 20 C) of autotrophic bacteria anoxic respiration rate of Activated Sludge Model No.3 (ASM3) (Gujer et al., 1999). The yield coefficient (YAN) of 0.15 mg COD/mg NH4eN and endogenous respiration rate (bAN) of 0.04 d1 for autotrophic nitrifiers in BAR2 reported by Ekama (2009) were adopted in this study.
4.2. Elemental compositions of influent organics and biomass The elemental compositions of the SANI pilot plant influent biodegradable organics and the biomasses in the SRUSB, Anoxic Bioreactor and Aerobic Bioreactor are required in the stoichiometric equations. These compositions were determined based on the equations of Ekama (2009). For 1 g VSS organics (or biomass), CxHyOzNaPb can also be written as, i.e. Cx Hy Oz Na Pb hCfc =12 Hfh =1 Ofo =16 Nfn =14 Pfp =31 hC1 Hy=x Oz=x Na=x Pb=x (15a) !
fo ¼
16 fcv 8fc 17fn 26fp 1 8 12 14 31 18
fh ¼
2 44fc 10fn 71fp þ 1 þ fcv 12 14 31 18
(15b) ! (15c)
where, x, y, z, a, b ¼ elemental compositions for CxHyOzNaPb; fcv, fc, fh, fo, fn, fp ¼ COD, TOC, H, O, OrgN, OrgP-to-VSS mass ratio for particulate organics or COD, TOC, H, O, OrgN, OrgP-toMass ratio for filtered organics.
485
For particulate organics the mass ratios fcv, fc, fn and fp can be measured via COD, TOC, TKN and FSA, TP and OP and VSS tests. The fh and fo mass ratios do not need to be measured because they can be replaced by the measured COD/VSS ratio ( fcv) and mass balance ( fc þ fh þ fo þ fn þ fp ¼ 1). So with measured fcv, fc, fn and fp, the fh and fo can be calculated (Eqs. (15b) and (15c)). With fc, fh, fo, fn and fp known, the stoichiometric composition for 1 g of VSS can be found from Eq. (15a), where x ¼ fc/12, y ¼ fh/1, z ¼ fo/16, a ¼ fn/14 and b ¼ fp/31. The composition formula can then be normalized with respect to one of its components, e.g. C ¼ 1 from C1Hy/xOz/xNa/xPb/x (Eq. (15a)). Because a mass measurement for the soluble organics (FBSO, USO) cannot be made (like VSS for particulate organics), one of the mass ratios has to be assumed, i.e. in this paper fcv ¼ 1.42 mg COD/mg was accepted for both fermentable biodegradable (FBSO) and un-biodegradable soluble (USO) organics in the raw wastewater. The assumption fcv ¼ 1.42 for the influent FBSO had a negligible effect on the accuracy of the BSR model of Poinapen and Ekama (2010a) because the FBSO make up less than 0.1% of the COD in primary sewage sludge (PSS). However, the assumption that fcv ¼ 1.42 for the FBSO can have some impact on the accuracy of the BSR model applied here to raw wastewater because the influent FBSO is a larger proportion of the hydrolyzed biodegradable organics (30%, Fig. 3). Fig. 3 illustrates the characterization of the different components of the raw influent wastewater and VSS organics of the SRUSB and the determinations of their respective elemental compositions from the measured mass fraction ratios ( fcv, fc, fn, fp) which follows the procedure of Poinapen and Ekama (2010a). The notes 1e3 in Fig. 3 are described below. Filtered/soluble organics fraction (VFA, FBSO and USO; Table 6 and Note 1 in Fig. 3): The concentration of volatile fatty acids (VFAs), C2H4O2 for HAc and C2 H3 O 2 for Ac , was measured with 5-pH point titration method (Moosbrugger et al., 1992) and the influent wastewater pH. The unit conversion factor from mg HAc/L to mg COD/L is 1.067, which is also the ratio of COD-to-Mass ( fcv) for VFAs. The concentration of the un-biodegradable soluble organics (USO) in the influent was accepted to be the filtered COD of the final effluent, i.e. 30.7 mg COD/L (as shown in Tables 3 and 5). The concentration of total soluble/ filtered organics consisting of VFAs, USO and FBSO was determined from the filtered COD test on the influent. Therefore, the FBSO COD, TOC, OrgN and OrgP concentrations can be calculated from the difference, as shown in Table 3 and explained in Table 5. The mass ratios of the USO ( fc ¼ 0.532 mg C/mg, fn ¼ 1.074 mg N/mg (high fn due to high measured USON, Tables 1 and 3), fp ¼ 0.014 mg P/mg) and FBSO ( fc ¼ 0.626 mg C/mg, fn ¼ 0.038 mg N/mg, fp ¼ 0.003 mg P/mg) were determined using the assumed fcv of 1.420 mg COD/mg for both the USO and FBSO fractions and calculating from the filtered TOC, OrgN (TKN-FSA), and OrgP (TP-OP) concentrations in Tables 1 and 3. The detailed calculation steps are listed below Table 5. Particulate organics fraction (UPO and BPO; Table 6 and Note 2 in Fig. 3): The un-biodegradable particulate organics fraction ( fUPO) of the system influent was determined above to be 0.129.
486
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
With the COD concentration of the USO, FBSO, VFA and UPO known, the COD concentration of the BPO is found by difference, i.e. 217.5 mg COD/L giving a BPO fraction ( fBPO) of 0.505 (see Table 3). The VSS concentration of the influent was measured at 186 mg/L which needs to be disaggregated into UPO and BPO VSS concentrations. The VSS, TOC, OrgN and OrgP concentrations of the influent UPO were calculated from the solids TSS, VSS, TOC, OrgN, and OrgP mass balances analysis over the SANI pilot plant (Lu et al., 2011), which gave for the UPO VSS, TOC, OrgN and OrgP 37.8 mg VSS/L, 20.2 mg C/L, 8.3 mg N/L and 0.9 mg P/L (Table 3). Therefore, the mass fraction ratios ( fcv, fc, fn, fp) of UPO could be calculated as set out in steps 1e4 below Table 6. With the COD, VSS, TOC, OrgN and OrgP of the UPO known, the concentrations for the BPO were calculated by difference from the measured total particulate concentrations, i.e. BPO concentration ¼ Total PO concentrations e UPO concentrations. The mass ratios of the BPO were calculated from the BPO concentrations (as explained in steps 1e4 below Table 6). Finally, using the Eqs. (15a)e(15c), the elemental compositions of UPO and BPO were determined from the mass ratios, i.e. C1.00H1.54O0.22N0.35P0.017 and C1.00H1.10O0.44N0.09P0.005, respectively, in Fig. 3.
Table 5 e Calculation procedure (step 1 to 7 below Table) to determine the mass ratios of the soluble (<0.45 mm filtered) organics in the raw wastewater. Influent
COD
Concentration 34.5b Mass ratio 1.067b USO Concentration 30.7d Mass ratio 1.42f FBSO Concentration 92.7e Mass ratio 1.42f Total Concentration 157.9c VFA
Mass 32.2a e 22.7g e 65.3g e e
TOC
OrgN
12.9b 0.0b 0.400b 0.000b 11.5d 23.2d 0.507g 1.022g 40.9e 2.5e 0.626h 0.038h 65.3c 25.7c
OrgP 0.0b 0.000b 0.3d 0.013g 0.2e 0.003h 0.5c
a Calculation procedure:From 5-pH point titration method (Moosbrugger et al., 1992, mg HAc/L); b Obtained from composition of HAc (CH3OOH); c Measured wastewater influent membrane filtered (<0.45 mm) COD, TOC, TKN minus FSA and TP minus OP; d Measured membrane filtered final effluent (from BAR2) COD, TOC, TKN minus FSA and TP minus OP concentrations; e Calculated from Total e VFA e USO; f Assumed COD/mass ratio for FBSO and USO; g Obtained from COD divided by assumed COD mass ratio for USO and FBSO (1.42); h Obtained by divided by TOC, OrgN and OrgP by mass.
Table 6 e Calculation procedure to determine the mass ratios of the particulate organics (PO) compositions of the influent wastewater (UPO and BPO) and the SRUSB reactor VSS (UPO and biomass) from measured PO COD, VSS, TOC, OrgN and OrgP concentrations. Influent BPO UPO PO Total
COD Concentration Mass ratio Concentration Mass ratio Concentration
SRUSB Reactor VSS BPO Biomass UPO PO Total
Concentration Mass ratio Concentration Mass ratio Concentration Mass ratio Concentration Mass ratio
c
VSS
217.5 1.468d 55.6b 1.471d 273.1a
148.2 e 37.8b e 186a
COD
VSS
h
51.4 1.468e 638.9h 1.521k 4479.2h 1.471f 5169.5g 1.477d
i
c
35 e 420i e 3045i e 3500g e
TOC c
OrgN c
OrgP c
82.2 0.555d 20.2b 0.534d 102.4a
8.7 0.059d 8.3b 0.220d 17.0a
1.1 0.0074d 0.9b 0.024d 2.0a
TOC
OrgN
OrgP
j
19.4 0.555e 230.6j 0.549l 1626.0j 0.534h 1876.0g 0.536d
j
2.1 0.059e 45.5j 0.108l 669.9j 0.220h 717.5g 0.205d
0.3j 0.0074e 7.1j 0.017l 73.1j 0.024h 80.5g 0.023d
a Calculation procedure:Measured concentrations of the influent PO, i.e. difference between unfiltered and filtered concentrations; b Concentrations of the UPO calculated by VSS, TSS, TOC, OrgN and OrgP mass balance around the SANI pilot plant (Lu et al., 2011); c Difference between total (PO) and UPO concentrations; d Mass ratios obtained by dividing the measured COD, TOC, OrgN and OrgP by the VSS concentrations; e Mass ratios of BPO in influent and SRUSB VSS are same; f Mass ratios of UPO in influent and SRUSB VSS are same; g Measured concentrations of the SRUSB PO solids; accumulation of UPO solids in the base of the SRUSB reactor is the reason why the COD of the SRUSB solids predicted by the hydrolysis model (8445 mg COD/L in Table 3) is different to the measured concentration in this Table 5 (5169.5 mg COD/L). This difference was accounted for in the calculation of the particulate organics (PO) mass ratios; h Measured SRUSB COD subdivided into the same proportions as that predicted by the hydrolysis model (Table 3), i.e. 87% UPO, 12% biomass and 1% BPO; i VSS concentrations of the BPO and UPO in the SRUSB solids obtained by dividing the COD concentrations by the COD/VSS ( fCV) mass ratios; j TOC, OrgN and OrgP concentrations of the UPO and BPO obtained by multiplying their VSS concentrations by the fc, fn and fp mass ratios; k Biomass VSS, TOC, OrgN and OrgP concentrations in SRUSB solids obtained by difference between PO concentrations and BPO and UPO concentrations, e.g. Biomass VSS ¼ PO VSS minus BPO VSS minus UPO VSS concentrations; l Biomass mass ratios ( fcv, fc, fn, fp) calculated by dividing the COD, TOC, OrgN and OrgP concentrations by the VSS concentration.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
Concentration and composition of VSS in the SRUSB (Table 6 and Note 3 in Fig. 3): The same principle of disaggregation of measured influent particulate organic (PO) solids COD, VSS, TOC, OrgN and OrgP above was applied to characterize the PO solids (VSS) in the SRUSB. The calculation procedure for doing this is given by steps 5e11 below Table 6. The SRUSB PO consist of: (1) accumulated un-biodegradable particulate organics (UPO) with the same composition as the influent UPO (step 6), (2) residual biodegradable particulate organics (BPO) with the same composition as the influent BPO (step 5), and (3) biomass with an unknown composition, which is to be determined (endogenous residue concentration was assumed to be zero, fED ¼ 0). As mentioned above, the COD concentrations of the UPO, BPO and biomass solids in SRUSB was calculated by applying the method of iteration calculation with the anaerobic hydrolysis kinetic model between the anaerobic batch tests and SANI pilot plant system (Section 3.2). For this it was accepted that all the VFAs and FBSO are completely utilized (as shown in Fig. 3), and the USO is un-biodegradable and so exists in the liquid phase of sludge with the same composition as the influent USO. The procedure for this iteration calculation with the hydrolysis model were described above and in this way it was found that the residual (unutilized) biodegradable particulate COD (Sbp) in SRUSB PO is very low, only 1% (84 mg COD/l) of SRUSB COD (Fig. 3). The COD of the SRUSB PO (8445 mg COD/ l) therefore comprises mostly UPO COD from the influent (7360 mg COD/l, 87%) and acidogen biomass COD (1085 mg COD/l, 12%) with a very low residual BPO COD (84 mg COD/l, 1%) (Table 4). The measured COD concentration of the SRUSB PO (5169.5 mg COD/l) was lower than that calculated (8445 mg COD/l) due to accumulation of UPO in the base of the SRUSB reactor, as explained by Lu et al. (2011). So the measured COD of the SRUSB solids (5169.5 mg COD/l) was disaggregated into UPO COD, biomass COD and BPO COD in the same proportions as the calculated COD with the hydrolysis model, i.e. 87% (4479.2 mg COD/l), 12% (638.9 mg COD/l) and 1% (51.4 mg COD/l) respectively, as shown in steps 7 and 8 below Table 6. With the fcv, fc, fn and fp mass ratios of influent UPO and BPO known, the VSS, TOC, OrgN and OrgP concentrations of the UPO and BPO were calculated from the COD concentrations (steps 9 and 10). The VSS, TOC, OrgN and OrgP concentrations of the biomass were then calculated by difference between the COD, VSS, TOC, OrgN and OrgP concentrations of the measured SRUSB PO concentrations and the known UPO and BPO concentrations (step 11). The fcv, fc, fn and fp mass ratios of the biomass were then calculated by dividing the biomass COD, TOC, OrgN and OrgP concentrations by the biomass VSS concentration (step 12). The biomass fcv, fc, fn and fp mass ratios so obtained were 1.521 mg COD/mg VSS, 0.549 mg C/mg VSS, 0.108 mg N/mg VSS and 0.017 mg P/mg VSS respectively as shown in Table 6. Assuming x ¼ 1 and using the Eqs. (15a)e(15c), the biomass composition of C1.00H1.284O0.356N0.169P0.0147 was determined. The same method was also applied in the composition determination of Autotrophic Denitrifiers (ADN, C1.00H1.64O0.384N0.236P0.014) in the anoxic bioreactor (BAR1) and Autotrophic Nitrifiers (AN, C1.00H1.75O0.43N0.205P0.015) in the aerobic bioreactor (BAR2).
5.
Steady state model validation
5.1.
Steady state model application and validation
487
The developed and calibrated steady state model was applied to the SANI pilot plant operated with an influent flow rate of 10 m3/d and no primary treatment. The internal recirculation flow ratio of the SRUSB was maintained at 3Q (Q ¼ 10 m3/d) while the recirculation flow rate between the aerobic and anoxic reactors was set at 2.5Q, which supplied nitrate for the autotrophic denitrification in the anoxic reactor (see Fig. 2). The results predicted by the steady state model using the spread sheet in Microsoft Excel and the experimentally measured data from the SANI pilot plant are compared in Table 7. Overall the steady state model predictions correspond well with the measured data for all three reactors, SRUSB, BAR1 and BAR2. For the SRUSB, the COD and sulfate removals are a little higher than the measured data because the SRUSB is assumed to be a completely mixing reactor, however, there exist about 5% ( fb) hydraulic short-circuiting flow (bypass flow) and 35% ( fd) dead space in the SRUSB of the SANI pilot plant (Lu et al., 2011), which correspondingly causes the lower effluent COD than the measured data. The measured effluent hydrogen sulfide is lower than the prediction, which can be attributed to the following two points: a) the COD removal is lower than the prediction, and b) a little hydrogen sulfide was oxidized to elemental sulfur, which is indicated by the white elemental sulfur on the internal wall of effluent pipe of the SRUSB. The above analysis can also explain the COD mass balance of 81% for the SRUSB. For the anoxic reactor (BAR1), though the measured hydrogen sulfide removal is the same as the prediction, the nitrate removed by autotrophic denitrification with hydrogen sulfide as the electron donors is lower than the measured data. This means that not all the hydrogen sulfide removed in the anoxic reactor was utilized by autotrophic denitrification, i.e. part of the hydrogen sulfide was oxidized by the dissolved oxygen that was brought in by the recirculation flow from the aerobic reactor to the anoxic reactor (as shown in Fig. 2). This can also be observed from the ratio of DH2SeSto-DNO3eN in the BAR1, which is 2.5 g H2SeS/g NO3eN and higher than the stoichiometric values of 1.43 g H2SeS/g NO3eN calculated by Eqs. (8) and (9) for the SANI pilot plant system and 1.85 g H2SeS/g NO3eN reported by Driscoll and Bisogni (1978).The nitrate flux denitrified in BAR1 is 35 m3/ d (13.5e2.2) mg NO3eN/L ¼ 395.5 g NO3eN/d. This is 45% of the system influent TKN flux (10 m3/d 87.5 mg TKN-N/ L ¼ 875 g N/d). The N balance over the BAR1 denitrification reactor including the nitrate denitrified is 93%, which is less than 100% mainly because the unfiltered TKN concentration difference between the influent (39.2 mg TKN-N/L) and effluent (35.7 mg TKN-N/L) is 3.5 mg TKN-N/L. Including the N in the UPO accumulation in the bottom of the SRUSB (323.6 g COD/d O 23.6 0.22 ¼ 48.5 g N/d), the nitrate denitrified (395.5 g NO3eN/d) and the effluent unfiltered TKN (23.4 10 m3/d ¼ 234 g TKN-N/d) and nitrate (16.8 10 m3/ d ¼ 168 g NO3eN/d), the overall system N balance is 97%, which is very good. Since the BAR1 was packed with plastic
488
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
Table 7 e Comparison of experimentally measured (M) values with pilot-scale (P1) and lab-scale (P2) steady state models predicted data. Parameter
SRUSB a
Influent unfiltered COD (mgCOD/L) Influent VFA (mgCOD/L) Influent FBSO COD (mgCOD/L) Influent BPO COD (mgCOD/L) Influent sulfate (mgS/L) Influent hydrogen sulfide (mgS/L)b Influent TKN/FSA/NO3 (mgN/L) Influent TP/OP (mgP/L) Influent total alkalinity (mg/L as CaCO3)c Influent flow rate, Qi (m3/d) Sludge retention time, RS (d) NHRT/AHRT (h) E (sludge COD produced/COD utilized)
Comparison a
COD removal (mgCOD/L) Sulfate variation (mgS/L)d Sulfide variation (mgS/L)d Total alkalinity variation (mg/L as CaCO3)c Nitrate variation (mgN/L)d FSA removal (mgN/L) Effluent unfiltered COD (mgCOD/L)a Effluent sulfate (mgS/L) Effluent hydrogen sulfide (mgS/L)b Effluent TKN (mgN/L) Effluent FSA (mgN/L) Effluent NO3 (mgN/L) Effluent TP (mgP/L) Effluent total alkalinity (mg/L as CaCO3)c COD Mass balance (%)e S N P a b c d e f
BAR1
431 34.5 92.7 217.5 195.7 4.0 87.5/44.8/0 7.8/5.3 223.6 10 90 16.3/4.1 0.0264
BAR2
65.9 e e e 142.3 31.3 39.2/16.9/13.5 5.9/5.3 281.4 35 110 9.4/2.7 0.1804
55.0 e e e 169.4 3.0 35.7/16.4/2.2 5.7/5.3 247.7 35 42 9.4/2.7 0.0632
M
P1
P2
M
P1
P2
M
P1
P2
332.1 130.3 þ120.1 þ513.3 e e 98.8 65.4 124.1 85.3 45.4 0 7.6 736.9 86 95 103 104
344.7 168 þ168 þ536.9 e e 86.3 27.7 172 e e e 7.6 760.5 100 100 100 100
127.1 61.9 þ61.9 þ209.7 e e 303.8 133.8 65.9 e e e 7.7 443.3 100 100 100 100
e þ27.1 28.3 33.7 11.3 e 55.0 169.4 3.0 35.7 16.4 2.2 5.7 247.7 90 99 93f 97
e þ28.3 28.3 28.1 13.5 e e 170.6 3.0 e e 0 5.8 253.3 100 100 100 100
e þ28.3 28.3 28.1 13.5 e e 170.6 3.0 e e 0 5.8 253.3 100 100 100 100
e þ2.9 3.0 127.2 þ14.6 13.0 53.9 172.3 0 23.4 3.4 16.8 5.6 120.5 98 100 106 98
e þ3.0 3.0 119.2 þ16.4 16.4 e 172.4 0 19.4 0 18.6 5.7 128.5 100 100 100 100
e e e 109.8 þ16.4 16.4 e 169.4 3.0 19.4 0 18.6 5.7 137.9 100 100 100 100
COD without hydrogen sulfide COD; Total hydrogen sulfide ¼ H2S þ HS, which are dissolved in the water. Total alkalinity ¼ H2 CO3 alkalinity þ alkalinity H2S; The positive value means generation and the negative value means consumption; The C balance over the SRUSB in not included due to the sludge accumulation in the base of the reactor; and Includes the nitrate denitrified calculated from the difference between the BAR1 influent and effluent nitrate fluxes.
media (specific surface area of 115 m2/m3), it is reasonable that part of the particulate organics were retained by the filter and caused the 84% of COD balance. As shown in Fig. 5, the volumetric denitrification rate (KDN ¼ 4.2 mg NO3eN/L/h) in BAR1 and the volumetric nitrification rate (KN ¼ 4.8 mg FSA-N/L/h) in BAR2 can be calculated by dividing the DNO3eN (13.5e2.2 ¼ 11.3 mg NO3eN/L) and DFSA-N (16.4e3.4 ¼ 13.0 mg FSA-N/L) by the actual HRT (2.7 h), respectively. The nitrification rate (KN) of 4.8 mg FSA-N/L/h (or 4.8 24 ¼ 115 g FSA-N/m3/d) is close to the value of KDN of 4.2 mg NO3eN/L/h (or 4.2 24 ¼ 101 g NO3eN/m3/d) indicating that equal sized anoxic and aerobic reactors are appropriate. The media surface specific denitrification and nitrification rates in BAR1 and BAR2 respectively can be calculated by dividing the volumetric nitrate (101 g NO3eN/ m3/d) and ammonia (115 g FSA-N/m3/d) removal rates by the specific media surface (115 m2/m3), i.e. nitrate denitrification rate ¼ 101/115 ¼ 0.88 g NO3eN/(m2 media surface.d) and
Fig. 5 e Rates of denitrification and nitrification in the BAR1 and BAR2 respectively.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
Ammonia nitrification rate ¼ 115/115 ¼ 1.00 g FSA-N/(m2 media surface.d). The media specific autotrophic NDR in BAR1 was five times higher than that achieved in an up-flow fixed plastic media heterotrophically denitrifying bed reactor under the same NLR for normal (non saline) municipal wastewater treatment, i.e. 0.15 g NO3eN/m2 media surface/d (Moosavi et al., 2005). The media specific ANR in BAR2 is about 25% higher than that observed in down-flow draft ventilated rock filled nitrifying trickling filters (receiving no organics) for normal (non saline) municipal wastewater, i.e. 0.83 g FSA-N/m2 media surface/ d (Muller et al., 2006) and about five times of that in a fixed plastic media nitrifying bed reactor under the same ALR for normal (non saline) municipal wastewater treatment 0.21 g NH4eN/m2 media surface/d (Moosavi et al., 2005).
5.2. Comparison of lab-scale and pilot-scale steady state models To show the improvement in the pilot-scale steady state model for the SANI pilot system with real saline wastewater compared with the lab-scale steady state model, the latter was also applied to the SANI pilot plant with the same inputs (i.e. influent characteristics, flow rates, organics and biomass compositions, HRT and SRT etc) as pilot-scale steady state model (as shown in Table 7 column P2). Since about 217.5 mg COD/L of BPO in influent has to be hydrolyzed before utilization and the lab-scale steady state model did not include the anaerobic hydrolysis kinetics, it gave much lower COD removal, sulfate reduction and total alkalinity variation values (P2) than those of pilot-scale steady state model (P1), which also correspondingly caused a great deviation between experimentally measured (M) values and lab-scale steady state model predicted (P2) data (see Table 7). Furthermore, compared with the pilot-scale steady state model (P1), the labscale model (P2) cannot predict the changes of sulfate and hydrogen sulfide concentrations in the aerobic reactor and subsequently the alkalinity consumption due to the H2S and HS oxidation to SO2 4 by DO was not counted into the labscale model, as shown in Table 7. So it was necessary to develop a pilot-scale steady state model and characterize the influent organics and biomass compositions for the SANI pilot system treating with real saline wastewater.
5.3. Simplified stoichiometries for the SANI pilot system Based on the above model calibration, the simplified stoichiometries (with main reactants and products) of the SANI pilot system can be expressed as: Biological sulfate reduction in the SRUSB: 100g COD þ 150:2g SO2 4 þ 43:7g H2 O/53:2g H2 S þ 1:9g Sludge þ 190:9g HCO 3
ð16aÞ
Autotrophic denitrification in the anoxic bioreactor (BAR1): 100g NO 3 þ 5:9g HCO3 þ 35:92g H2 S/22:58g N2
þ 101:42g SO2 4 þ 2:15g Sludge Nitrification in the aerobic bioreactor (BAR2):
(16b)
489
100g NHþ 4 þ 7:33g CO2 þ 346:67g O2 /5:22g Sludge þ þ 344:44g NO 3 þ 11:11g H þ 98g H2 O
(16c)
According to Eqs. (16a)e(16c), the observed sludge yield coefficients (Yobs) for the SRUSB, BAR1 and BAR2 in the SANI pilot plant were determined to be only 0.02 kg VSS/kg COD removed, 0.10 kg VSS/kg NO3eN (or 0.06 kg VSS/kg S) and 0.07 kg VSS/kg NH4eN, respectively, which confirms the very low sludge production for the SANI system compared with conventional activated sludge systems, which explicitly explains that the causes and conditions of sludge minimization in the SANI pilot plant for real saline sewage treatment.
6.
Conclusions
A steady state model for the SANI pilot plant treating real screened saline sewage has been developed in this study. For the SRUSB reactor, this model comprises three sequential parts: 1) a COD-based anaerobic hydrolysis kinetic part, 2) an elements (C, H, O, N, P, S), COD and charge balanced stoichiometry part, and 3) a mixed weak acid/base chemistry part. In order to validate the steady state model, 1) the organics compositions of system influent and VSS in the SANI pilot plant at the steady state were characterized via the fcv, fc, fn, and fp mass ratios, 2) the hydrolysis kinetic parameters in the Monod equation (KADm ¼ 3.25 mg COD/(mg COD.d), KADs ¼ 557 mg COD/L and mADmax ¼ 0.37 d1 at 25 C) were determined through the anaerobic hydrolysis batch testing, and 3) the steady state model was applied for the SANI pilot plant with the design and operating parameters as the inputs. It was found that the steady state model predictions are correlated quite well with the experimentally measured data. This model clearly explains the causes and conditions for minimal sludge production and oxygen demand in the SANI pilot plant for saline sewage treatment compared with conventional activated sludge nitrogen removal systems.
references
Claus, G., Kutzner, H.J., 1985. Physiology and kinetics of autotrophic denitrification by Thiobacillus denitrificans. Applied Microbiology and Biotechnology 22 (4), 283e288. Driscoll, C.T., Bisogni, J.J., 1978. The use of sulfur and sulfide in packed bed reactors for autotrophic denitrification. Journal of the Water Pollution Control Federation 50 (3), 569e577. Ekama, G.A., 2009. Using biomass stoichiometry to build a plantwide mass balance based steady state WWTP model. Water Research 43 (8), 2101e2120. Ekama, G.A., Wentzel, M.C., 1999. Denitrification kinetics in biological N and P removal activated sludge systems treating municipal wastewater. Water Science and Technology 39 (6), 69e77. Gujer, W., Henze, M., Mino, T., van Loosdrecht, M.C.M., 1999. Activated sludge model No. 3. Water Science and Technology 39 (1), 183e193. Lau, G.N., Sharma, K.R., Chen, G.H., van Loosdrecht, M.C.M., 2006. Integration of sulfate reduction, autotrophic denitrification and nitrification to achieve low-cost excess sludge minimization for Hong Kong sewage. Water Science and Technology 53 (3), 227e235.
490
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 7 5 e4 9 0
Lu, H., Wang, J., Li, S., Chen, G.H., van Loosdrecht, M.C.M., Ekama, G.A., 2009. Steady-state model-based evaluation of sulfate reduction, autotrophic denitrification and nitrification integrated (SANI) process. Water Research 43 (14), 3613e3621. Lu, H., Wu, D., Tang, D.T.W., Chen, G.H., van Loosdrecht, M.C.M., Ekama, G.A., 2011. Pilot scale evaluation of SANI process for sludge minimization and greenhouse gas reduction in saline sewage treatment. Water Science and Technology 63 (10), 2149e2154. Moosavi, G.H., Naddafi, K., Mesdaghinia, A.R., Nabizadeh, R., 2005. Simultaneous organics and nutrients removal from municipal wastewater in an un-flow anaerobic/aerobic fixed bed reactor. Journal of Applied Science 5 (3), 503e507. Moosbrugger, R.E., Wentzel, M.C., EkamaMarais v, G.R., Marais v, G.R., 1992. Simple Titration Procedures to Determine H2CO3* Alkalinity and Short Chain Acid Concentrations in Aqueous Solutions Containing Known Concentrations of Ammonium, Phosphate and Sulphide Weak Acid/bases. Water Research Council (WRC), Cape Town, SA. Muller, A.W., Wentzel, M.C., Ekama, G.A., 2006. Estimation of nitrification capacity of rock media trickling filters in external nitrification BNR. Water S.A. 31 (5), 611e618. Nikitas, P., Pappa-Louisi, A., 2000. Non-linear least-squares fitting with microsoft excel solver and related routines in HPLC modeling of retention I. considerations of the problems of the method. Chromatographia 52 (7e8), 477e486. Oh, S.E., Bum, M.S., Yoo, Y.B., Zubair, A., Kim, I.S., 2003. Nitrate removal by simultaneous sulphur utilizing autotrophic and heterotrophic denitrification under different organics and
alkalinity conditions: batch experiments. Water Science and Technology 47 (1), 237e244. Poinapen, J., Ekama, G.A., 2010a. Biological sulphate reduction with primary sewage in an upflow anaerobic sludge bed reactor e Part 5: steady state model. Water S.A. 36 (3), 193e202. Poinapen, J., Ekama, G.A., 2010b. Biological sulphate reduction with primary sewage in an upflow anaerobic sludge bed reactor e Part 6: development of a kinetic model for BSR. Water S.A. 36 (3), 203e213. Riefler, R.G., Smets, B.F., 2003. Comparison of a type curve and a least-squared errors method to estimate biofilm kinetic parameters. Water Research 37 (13), 3279e3285. Ristow, N.E., So¨temann, S.W., Wentzel, M.C., Loewenthal, R.E., Ekama, G.A., 2006. The effects of hydraulic retention time and feed COD concentration on the rate of hydrolysis of primary sewage sludge under methanogenic conditions. Water Science and Technology 54 (5), 91e100. So¨temann, S.W., Ristow, N.E., Wentzel, M.C., Ekama, G.A., 2005. A steady state model for anaerobic digestion of sewage sludges. Water S.A. 31 (4), 511e527. Tsang, W.L., Wang, J., Lu, H., Li, S., Chen, G.H., van Loosdrecht, M.C.M., 2009. A novel sludge minimized biological nitrogen removal process for saline sewage treatment. Water Science and Technology 59 (10), 1893e1899. Wang, J., Lu, H., Chen, G.H., Lau, G.N., Tsang, W.L., van Loosdrecht, M.C.M., 2009. A novel sulfate reduction-autotrophic denitrification-nitrification integrated (SANI) process for saline wastewater treatment. Water Research 43 (9), 2363e2372. Water Supplies Department (WSD) of the Hong Kong SAR Government, 2010. Annual Report 2008/2009.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 9 1 e4 9 9
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Equilibrium and intra-particle diffusion of stabilized landfill leachate onto micro- and meso-porous activated carbon Shrawan K. Singh, Timothy G. Townsend*, David Mazyck, Treavor H. Boyer Department of Environmental Engineering Sciences, University of Florida, P.O. Box 116450, Gainesville, FL 32611-6450, USA
article info
abstract
Article history:
Stabilized landfill leachate has previously been treated with activated carbon (AC);
Received 4 April 2011
however, information on the selectivity of AC depending upon the pore size is minimal.
Received in revised form
Isotherm and kinetic experiments were conducted using three commercially available AC
21 October 2011
products, one micro-porous and two meso-porous. Equilibrium adsorption and intra-
Accepted 4 November 2011
particle diffusion of organic matter from stabilized leachate was studied. Isotherm
Available online 15 November 2011
experimental data were fitted to Langmuir, Freundlich, and RedlichePeterson isotherm models in non-linear forms. Of the three isotherm models, the RedlichePeterson model
Keywords:
provided the best fit to the experimental data and showed a similar organic matter
Landfill leachate
adsorption capacity (approximately 0.2 g total organic carbon (TOC) g1 AC) for both micro-
Leachate treatment
porous and meso-porous AC. The organic matter effective intra-particle diffusion coeffi-
Adsorption
cients (De) in both AC types were on the order of 1010 m2 s1 for AC particle sizes greater
Micro-porous activated carbon
than 0.5 mm. Meso-porous ACs showed slightly higher De compared to micro-porous AC.
Meso-porous activated carbon
Rapid small-scale tests showed a maximum of 80% TOC removal from leachate by each AC
Fluorescence
investigated. Fluorescence spectroscopy showed a preferential adsorption of fulvic-type organic matter with an increase in empty bed contact time by each AC. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Treating leachate appropriately and cost-effectively has been a challenge for landfill operators given the inherent complexity and changing characteristics of this wastewater. Leachate properties vary with landfill age, operation method, moisture availability, waste composition, and climate. A notable change in organic matter is observed with landfill age. Young leachate is characterized by readily biodegradable compounds with a high biochemical oxygen demand (BOD5), while mature leachate consists of molecules recalcitrant to biodegradation with a low BOD5. Biological treatment methods are effective for reducing organic matter concentrations in young landfill leachate (Borghi et al., 2003), but they are relatively ineffective for mature landfill leachate. Stabilized leachate, often characterized as having a ratio of BOD5 to
chemical oxygen demand (COD) of less than 0.1, demands physicochemical treatment techniques to remove organic matter. Several methods have been investigated for treating stabilized leachate containing recalcitrant organic matter, including coagulation and flocculation (Tatsi et al., 2003; Comstock et al., 2010), chemical oxidation (Rivas et al., 2004; Tizaoui et al., 2007), membrane-based technologies (Trebouet et al., 2001; Thorneby et al., 2003), and activated carbon (AC) adsorption (Kurniawan et al., 2006; Maranon et al., 2009). Removal of organic matter is an important objective during leachate treatment and these technologies mineralize or transform organic matter into a more biodegradable form (e.g., chemical oxidation, ozonation) or transfer them to a solid media (e.g., coagulation and flocculation, ion exchange, AC).
* Corresponding author. Tel.: þ1 352 392 0846; fax: þ1 352 392 3076. E-mail address:
[email protected] (T.G. Townsend). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.007
492
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 9 1 e4 9 9
Adsorption onto AC has been reported as an effective method for removing high molecular weight refractory organic matter from aqueous solutions (Halim et al., 2010), and it was investigated previously for leachate treatment. The literature review conducted by Foo and Hameed (2009) showed that AC is a potentially viable option for treating landfill leachate. However, interpreting data from the literature is complicated due to a wide range of experimental conditions, differences in how the data are reported, and the inherent differences in leachate quality. Generally, results from AC adsorption studies on leachate come from experiments where AC has been used in combination with other treatment methods (e.g., biological treatment processes, ozonation, coagulation) as presented in Table S1 in Supplementary Materials, which summarizes many pertinent studies on the use of AC for leachate treatment. Most of the studies reported in the literature use microporous AC (pore size < 2 nm) for leachate treatment (Gotvajn et al., 2009); however, an alternative AC product, mesoporous AC (pore size 2e50 nm), may prove more appropriate for stabilized leachate treatment. Stabilized leachate contains high molecular weight organic matter, which may cause pore blockage and reduce the adsorption capacity of micro-porous AC (Liu et al., 2010); however, the larger pore openings of meso-porous AC may allow the large organic molecules to diffuse within the pores more easily and limit the effect on adsorption capacity. The authors did not find literature on stabilized leachate treatment using meso-porous AC. This research was conducted to study the selectivity of micro- and meso-porous AC for stabilized leachate treatment based on the primary design factors of an AC treatment system. One micro-porous and two meso-porous ACs were selected to compare their adsorption isotherm profiles, the rate limiting adsorption processes, and the rate of organic matter diffusion. Rapid small-scale column tests (RSSCT) were conducted to determine the amount and type of organic matter removed using these three ACs for leachate treatment, providing information for the selection of AC pore sizes for leachate treatment.
Nalgene containers and kept at 4 C in the dark until used in the experiments. The three AC products used in this study were a Calgon Filtrasorb F-300 (Calgon Carbon Corporation, Pittsburgh, PA), a Norit HD-4000 (the Netherlands), and a Darco 12 40 (Norit Americas Inc., Texas). The Calgon F-300 ˚ with showed an average pore size diameter of 18 A a predominantly micro-porous structure. The Norit HD-4000 (meso-porous AC-1) and the Darco 12 40 (meso-porous AC2) were predominantly meso-porous ACs with pore size ˚ and 42 A ˚ , respectively. The general chardiameters of 32 A acteristics of the three ACs are listed in Table S2 in Supplementary Materials.
2.2.
Batch experiments
Batch experiments were conducted to develop adsorption isotherms and to determine the organic matter intra-particle diffusivity onto selected ACs. Isotherm data provide important information for designing and scaling up the AC adsorption system, whereas the intra-particle diffusivity data provide information needed to assess the suitability and effectiveness of the adsorption process. Prior to each experiment, leachate was filtered using a glass fiber filter of pore size 0.7 mm. This filtration did not measurably affect the leachate total organic carbon (TOC). Experiments were conducted in duplicate at a temperature of 23 0.5 C.
2.2.1.
2.2.2.
2.
Material and methods
2.1.
Landfill leachate and AC
Experiments were conducted using leachate generated from a lined and capped landfill in Florida, USA (Alachua County Southwest Landfill). Leachate samples were collected in
Isotherm experiment
Isotherm experiments were conducted using eight predetermined adsorbent doses ranging from 0 to 100 g L1. AC in amounts ranging from 0 to 10 g was added in separate, cleaned, pre-dried amber glass bottles and 100 mL of leachate was added per bottle. A rotary shaker mixed the bottles (wrapped with aluminum foil to avoid light exposure) at 150 rpm for approximately 240 h, achieving equilibrium. The samples were then filtered using glass fiber filters to restrict AC particles from entering into the samples to be analyzed. The samples were analyzed for TOC using a TekmarDohrmann TOC-VCPH/CPN analyzer.
Diffusivity experiment
The organic matter intra-particle diffusivity was determined by kinetic experiments using AC particles of different sizes. Subsets of each AC product were crushed and sieved through sieve numbers 20 (0.85 mm), 35 (0.50 mm), 50 (0.30 mm), and 200 (0.075 mm). The experiment used the particles retained on top of each sieve. All fractions were cleaned with deionized water and dried at 105 C for 24 h. One gram of AC for each
Table 1 e Design parameters of full-scale and rapid small-scale columns for Calgon F-300, Norit HD-4000, and Darco 12 3 40 AC Parameter
Grain size-US mesh Hydraulic loading rate (mm min1) Time to process (hours)
Calgon F-300
Norit HD-4000
Darco 12 40
Full-scale
RSSCT
Full-scale
RSSCT
Full-scale
RSSCT
8 30 17 to 205 28.8
35 50 4.6 4
10 30 17 to 205 28.8
35 50 4.6 4
12 40 17 to 205 28.8
35 50 4.6 4
493
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 9 1 e4 9 9
2.4.
Analytical methods
2.4.1.
Leachate organic matter characterization
The type of organic matter present in the samples was characterized using fluorescence spectroscopy (Hitachi F-2500 Fluorescence spectrophotometer). Fluorescence spectrometry provides detailed information about organic matter by generating a three-dimensional picture (excitation-emission matrix (EEM)) of fluorescence intensity as a function of excitation and emission wavelengths. The EEM was divided into five regions (Region-I to Region-V), based on the characteristics of each type of organic matter. Region-I and Region-II represent aromatic protein-like organic matter; Region-III and Region-V belong to fulvic- and humic-like organic matter, respectively; and Region-IV represents microbial-derived organic matter (Chen et al., 2003). The amount of each type of organic matter was quantified using the fluorescence regional integration (FRI) technique developed by Chen et al. (2003). The Supplementary Materials contain a detailed procedure of organic matter characterization and quantification.
2.4.2.
Adsorption equilibrium analysis
To effectively design an adsorption system, it is essential to develop the most appropriate mathematical description of adsorption isotherms. Several isotherm equations (e.g., Eqs. (1), (2), and (3) have been developed for the AC and organic matter adsorption process; the parameters of these equations express the surface properties and the affinity of adsorbent at
-1
Qe (mg TOC adsorbed g AC)
a
(1)
140 Experimental data Freundlich model Langmuir model Redlich Peterson model
120 100 80 60 40 20 0
0
20
40
60
80
100
120
Ce(TOC mg L-1)
b -1
Crittenden et al. (1986, 1991) presented the methodology for predicting full-scale granular activated carbon (GAC) column performance using RSSCTs that can generate a breakthrough profile in a very short time using smaller volumes of water compared to a full-scale or pilot-scale column. RSSCTs have been previously used to predict the efficiency of different types of water and GAC (Mackenzie et al., 2005; Ying et al., 2006); however, no study was found to predict the performance of GAC for stabilized leachate treatment using RSSCT. A description of designing the RSSCTs is presented in the Supplementary Materials. Quartz glass RSSCT columns (0.5 cm diameter 12 cm length) were fabricated with Teflon inserts to prevent the sorption of organic matter. The design parameters of RSSCTs for each AC are presented in Table 1. Experiments were conducted at 6.6, 10, 15, 30, and 60 min empty bed contact time (EBCT) of the full-scale column; corresponding EBCTs for the small-scale columns are presented in Table S3 in Supplementary Materials. A syringe pump (Harvard apparatus model 33) was used to supply a constant flow rate to the columns. Samples were collected every 15 min and analyzed for TOC to determine the organic matter amount removed at each EBCT. A fluorescence spectroscopy of the samples was conducted to determine the organic matter type removed at each EBCT.
Qe ¼ KF Ce1=n
Qe (mg TOC adsorbed g AC)
Rapid small-scale column tests
140 120
Experimental data Freundlich model Langmuir model Redlich Peterson model
100 80 60 40 20 0
0
20
40
60
80
100
120
80
100
120
Ce(TOC mg L-1)
c
140 120
-1
2.3.
a fixed temperature and pH. The data from the isotherm experiments were fitted into Freundlich isotherm (Eq. (1)), Langmuir isotherm (Eq. (2)), and RedlichePeterson isotherm (Eq. (3)) model equations to assess the ACs’ adsorption capacity (Freundlich, 1906; Langmuir, 1916; Redlich and Peterson, 1959).
Qe (mg TOC adsorbed g AC)
particle size was added to 30 mL of leachate in 40-mL glass vials. The glass vials were mixed on a rotary shaker at 150 rpm; samples were collected hourly, filtered, and analyzed for TOC.
Experimental data Freundlich model Langmuir model Redlich Peterson model
100 80 60 40 20 0
0
20
40
60 -1
Ce(TOC mg L )
Fig. 1 e Adsorption of leachate onto (a) Calgon F-300, (b) Norit HD-4000, and (c) Darco 12 x 40 AC. Data points and error bars are the average and standard deviation of the duplicate experiments, respectively.
494
Qe ¼
Qe ¼
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 9 1 e4 9 9
Qm bCe 1 þ bCe
(2)
KR Ce
(3)
1 þ aR Cbe R
where Qe is the amount of solute adsorbed at equilibrium (mg g1); Ce is the concentration of adsorbate in solution at 1 equilibrium (mg L1); KF ðmg g1 Þ ððmg L1 Þn Þ1 and n are adjustable parameters in the Freundlich equation; Qm (mg g1) and b (mg L1)1 are adjustable parameters in the Langmuir equation; and KR (L g1), aR (L mg1)1, and bR are adjustable parameters in the RedlichePeterson equation. Each model varies with respect to its assumption of the thermodynamic adsorption process and has its own set of advantages and disadvantages. The Freundlich isotherm (Eq. (1)) has been widely used and can be applied for heterogeneous surfaces along with multi-layer adsorption processes. The amount of adsorbate adsorbed onto adsorbent increases infinitely with an increase in concentration (Freundlich, 1906); however, it does not fit data very well for low concentration systems. The Langmuir adsorption isotherm (Eq. (2)) assumes the monolayer sorption process (Langmuir, 1916), which limits its applicability at high concentration systems. Redlich and Peterson (1959) developed a three-parameter isotherm equation (Eq. (3)) to incorporate the features of both the Freundlich and Langmuir isotherms. At low concentrations, the RedlichePeterson isotherm follows monolayer sorption, and at high concentrations its behavior approaches the Freundlich isotherm. The most common approach of analysis is to use the linear regression of the experimental data and the isotherm with a coefficient of determination (r2) closest to the one assumed to be the best fit isotherm equation. However, transforming a non-linear equation to a linear form may generate an error in analysis; hence, the non-linear analysis approach, as used by Ho et al. (2002), was followed for each isotherm. A detailed description of non-linear analysis is presented in the Supplementary Materials.
2.4.3.
Intra-particle diffusivity analysis
The sorption of organic matter onto AC is a complex phenomenon where properties of both adsorbate and adsorbent play an important role. The adsorption process can be controlled by following sequential steps: (1) the bulk solution transport, where the adsorbate diffuses from solution to the boundary layer of solution surrounding the AC particles; (2)
film diffusion, where adsorbate diffuses through the liquid film surrounding the AC particles; and (3) pore diffusion and adsorption, where adsorbent is transported to the pores of AC to available adsorption sites. One or more of these processes can be involved in the adsorption process; the slowest process controls the rate of adsorption. Weber and Morris (1963) developed a widely accepted kinetic-based model that represents the time dependent intra-particle diffusion of components and showed that the sorption process is diffusion controlled if the rate is dependent upon the rate at which adsorbate and adsorbent diffuse towards one another. The model equation is shown as follows: qt ¼ kid t1=2 þ C
(4) 1
where qt (mg g ) is the adsorbate loading on the solid phase at time t, kid (mg g1 min1/2) is the intra-particle diffusion rate constant, and C (mg g1) is the constant that is proportional to the thickness of boundary layer; the larger the value of C, the greater the boundary layer thickness (McKay et al., 1980). The straight line plot of qt and t1/2 represents the sorption process as diffusion controlled; however, if the data shows multiple linear plots, the sorption process is controlled by more than one process. To determine the type of organic matter diffusion into AC particles, the TOC data obtained from the diffusivity experiments were fitted into Eq. (4). Additionally, Boyd et al. (1947) generated a homogeneous particle diffusion model (HPDM), as shown in Eq. (5), for adsorbent-phase controlled diffusion (such as intra-particle diffusion) of adsorbate onto spherical particles to determine the rate of diffusion process: FðtÞ ¼ 1
2 2 6 XN 1 z p De t exp p2 z¼1 z2 r2
(5)
where F(t) is the fractional attainment of equilibrium at time t, De is the effective diffusion coefficient of adsorbate onto adsorbent (m2 s1), r is the radius of adsorbent particle assumed to be spherical (m), and z is an integer. F(t) values can be calculated as FðtÞ ¼ qt =qe , where qt and qe are adsorbate loading on the adsorbent at time t and when equilibrium is achieved respectively. Vermeulen’s approximation of the HPDM model fit the whole range 0 < F(t) < 1, for adsorption on spherical particles as shown in Eqs. (6) and (7) (Vermeulen, 1953):
Table 2 e Isotherm parameters obtained using non-linear method for organic matter absorption onto Calgon F-300, Norit HD-4000, and Darco 12 3 40 AC Isotherm Freundlich Langmuir RedlichePeterson
Constant 1 n1 1
KF ðmg g1 Þ ððmg L Þ 1/n QM (mg g1) b (mg L1)1 KR (L g1) aR (L mg1)1 bR
Þ
Calgon F-300
Norit HD-4000
Darco 12 40
0.04 1.63 64.6 5.9 x 103 0.71 578.8 1.89
0.05 1.52 35.3 7.6 x 103 0.56 917.8 2.03
0.04 1.59 82.8 4.8 x 103 0.63 647.2 1.88
495
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 9 1 e4 9 9
a
20
16 14 12 10
Particle size 0.85-1.0 mm Particle size 0.50-0.85 mm Particle size 0.30-0.50 mm Particle size 0.075-0.30 mm
8 6 4 0.5
1.0
1.5
2.0
2.5
3.0
0.4
0.2
16 14 12 10
Particle size 0.85-2.0 mm Particle size 0.50-0.85 mm Particle size 0.30-0.50 mm Particle size 0.075-0.30 mm
8 6
1
2
3
4
1.0
1.5
2.0
2.5
3.0
0.6
14 12
4 0.5
Particle size 0.85-1.7 mm Particle size 0.50-0.85 mm Particle size 0.30-0.50 mm Particle size 0.075-0.30 mm
0.2
0
1
2
3
4
1.5
2.0
2.5
3.0
3.5
Time1/2 (hr1/2) Fig. 2 e Weber and Morris intra-particle diffusion for removal of TOC using (a) Calgon F-300, (b) Norit HD-4000, and (c) Darco 12 x 40 AC. Data points were obtained using the average of duplicate experiments in the model equations.
2 1=2 p De t FðtÞ ¼ 1 exp r2
(6)
ln 1 F2 ðtÞ ¼ kt
(7)
5
6
7
8
1.0 Particle size 0.85-1.7 mm Particle size 0.50-0.85 mm Particle size 0.30-0.50 mm Particle size 0.075-0.30 mm
0.8
0.6
0.4
0.2
0.0 1.0
8
0.4
-1
Normalized TOC (C/C0), mg L
Qe (mg TOC adsorbed g-1 AC)
16
6
7
Operation time (hr)
c
18
8
6
Particle size 0.85-2.0 mm Particle size 0.50-0.85 mm Particle size 0.30-0.50 mm Particle size 0.075-0.30 mm
0.8
3.5
20
10
5
1.0
0.0 0.5
0
-1
18
Normaized TOC (C/C0), mg L
Qe (mg TOC adsorbed g-1 AC)
b
Time1/2 (hr1/2)
c
0.6
Operation time (hr)
20
4
Particle size 0.85-2.4 mm Particle size 0.50-0.85 mm Particle size 0.30-0.50 mm Particle size 0.075-0.30 mm
0.8
0.0
3.5
Time1/2 (hr1/2)
b
1.0
-1
18
Normalized TOC (C/C0), mg L
Qe (mg TOC adsorbed g-1 AC)
a
0
1
2
3
4
5
6
7
8
Operation time (hr) Fig. 3 e Adsorption kinetics of organic matter at different particle sizes (a) Calgon F-300, (b) Norit HD-4000, and (c) Darco 12 x 40 AC. Data points and error bars are the average and standard deviation of normalized TOC obtained from duplicate experiments, respectively.
where k ¼ p2De/r2 and the slope of plot eln[1 F2(t)] and t gives the value of effective diffusion coefficient (De). The TOC data of diffusivity experiments were also fitted into Eq. (7) to determine the De for each AC.
496
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 9 1 e4 9 9
3.
Results and discussion
3.1.
Leachate characterization
The composition of leachate was relatively constant over the research period and was characterized by a slightly alkaline pH, dark color, low BOD5, and high COD (BOD5:COD ¼ 0.02), consistent with characteristics of typical stabilized leachate. Relevant leachate physicochemical characteristics are presented in Table S4 in Supplementary Materials. The fluorescence EEM of leachate (Figure S1 in Supplementary Materials) showed peaks in Region-II, Region-III, and Region-V, representing the presence of protein, fulvic-, and humic-like organic matter in leachate. The FRI analysis of EEM showed the maximum concentration of humic-like organic matter in leachate.
3.2.
Adsorption equilibrium
The non-linear isotherm plots for three different ACs are presented in Fig. 1. Among the three isotherms, the Freundlich and RedlichePeterson isotherms provided a better fit to the experimental data, with the best fit from the RedlichePeterson isotherm. The Langmuir isotherm equation predicted a very low adsorption for higher concentrations of adsorbate at equilibrium and showed a poor fit to the experimental data, supporting others’ observations that the Langmuir isotherm model does not fit high concentration solutions due to possible violation of key Langmuir isotherm model assumptions: monolayer coverage, sites equivalence, and sites independence (Ho et al., 2002). The parameters values of each of the three isotherm equations generated from isotherm analysis for the three ACs are shown in Table 2. The Freundlich isotherm constant (KF) for each of the three ACs was in the range of 0.04e0.05. The values of constant n showed a stronger adsorption bond in all three ACs and a favorable adsorption process. The langmuir isotherm adsorption constant Qm showed low surface concentration of adsorbate; the values were in the range of 35.3e82.8 mg TOC g1 AC. These values represent monolayer concentrations that are generally not true for high concentration heterogeneous solutions, similar to what Rivas et al. (2006) observed. The Freundlich and RedlichePeterson isotherms predicted similar organic matter adsorption capacity for each AC. The Freundlich isotherm predicted an adsorption of 0.3 g TOC g1
AC by both meso-porous ACs and a slightly higher adsorption of 0.47 g TOC g1 AC by micro-porous AC. The RedlichePeterson Model predicted lower adsorption capacities than the Freundlich model, with 0.22, 0.18, and 0.19 g TOC g1 AC with micro-porous AC, meso-porous AC-1, and mesoporous AC-2, respectively. Very low adsorption capacities of 0.03, 0.05, and 0.06 g TOC g1 AC for micro-porous AC, mesoporous AC-1, and meso-porous AC-2, respectively were predicted by the Langmuir isotherm model. Although there is little data available on leachate adsorption onto AC, Xing et al. (2008) reported a maximum adsorption capacity of 0.2 g TOC g1 of coal-based, micro-porous AC, similar to the adsorption capacity predicted by the RedlichePeterson model. Fettig et al. (1996) observed a lower adsorption capacity of 0.06e0.08 g TOC g1 for a micro-porous AC. Data were not found for the leachate organic matter adsorption capacity of meso-porous AC. Micro-porous AC has a larger surface area compared to meso-porous AC, indicating that micro-porous AC would have a greater adsorption capacity than meso-porous AC. However, the presence of high molecular weight organic matter in leachate challenges the adsorption capacity of micro-porous AC because the smaller average pore size (versus mesoporous AC) may result in more pore blockage compared to meso-porous AC. Consequently, the adsorption capacity of micro-porous AC would be affected more compared to mesoporous AC (Li et al., 2003; Karanfil et al., 2006), which is supported by the results that show no difference in adsorption capacity between the micro- and meso-porous AC.
3.3.
Intra-particle diffusivity
Kinetic experiments were conducted to determine the ratelimiting adsorption process and the rate of organic matter adsorption onto micro- and meso-porous AC. Fig. 2 shows the kinetic experiment data fitted on the Weber-Morris model for different particle sizes of three ACs. Each AC showed two straight lines for the particle sizes greater than 0.3 mm. The first portion of the straight line represents the diffusion process controlled by external surfaces, and the second portion of the straight line shows the intra-particle diffusion. Extrapolation of the linear portion of the plots to the y-axis gives intercepts that provide the boundary layer thicknesses. The micro-porous AC showed a boundary layer thickness of 12.1 mg g1 for the particles greater than 0.84 mm, which increased up to 15.7 mg g1 for the smallest size particles (0.19 mm). Increasing boundary layer thickness with
Table 3 e Effective diffusion coefficient (De) of organic matter onto different particle sizes of Calgon F-300, Norit HD-4000, and Darco 12 x 40 AC AC particle size range (mean sizes) mm
>0.84 (0.84) 0.50e0.85 (0.68) 0.30e0.50 (0.40) 0.075e0.30 (0.19)
Calgon F-300 Diffusion coefficients (m2 s1) 145 109 533 151
1010 1010 1011 1011
Norit HD-4000 2
r F(t) vs (t) 0.95 0.97 0.97 0.99
Diffusion coefficients (m2 s1) 369 182 643 105
1010 1010 1011 1011
Darco 12 x 40 2
r F(t) vs (t) 0.93 0.93 0.97 0.70
Diffusion coefficients (m2 s1) 253 124 728 809
1010 1010 1011 1012
r2 F(t) vs (t) 0.99 0.96 0.95 0.86
497
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 9 1 e4 9 9
-1 -1 2
Fractional DOM removal derived
a
from FRI analysis (AU-nm (mg L ) ), V/V0
1010 m2 s1 for AC particles greater than 0.5 mm and the meso-porous AC-1 had the highest De among all three ACs in an order of meso-porous AC-1 > meso-porous AC-2 > microporous AC. Li et al. (2007) observed the De in the order of 1010 m2 s1 for adsorptive removal of aquatic natural organic matter by AC. The intra-particle diffusivity results show that the use of meso-porous AC for stabilized leachate treatment can provide faster organic matter removal compared to micro-porous AC. Additionally, the adsorption profile does not change with 1.6 Region-I Region-II Region-III Region-IV Region-V Total volume
1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0
0
10
20
30
40
50
60
70
EBCT long column (min)
100
-1 -1 2
Fractional DOM removal derived
b
from FRI analysis (AU-nm (mg L ) ), V/V0
decreasing AC particle size suggests that intra-particle diffusion was slower for smaller AC particles. The boundary layer thickness of meso-porous AC-1 ranged from 12.0 to 14.3 mg g1 for the aforementioned different particle sizes and 13.7e16.2 mg g1 for meso-porous AC-2. The straight line deviation from the origin represents the difference in diffusion rates in the initial and final stages of adsorption. The AC particle sizes <0.3 mm did not show a similar intra-particle diffusion pattern for micro- and meso-porous AC. While the micro-porous AC showed two straight lines, which represent macro- as well as micro-porous diffusion, the meso-porous ACs showed one straight line, which indicates micro-porous diffusion as the main diffusion process. The type of organic matter intra-particle diffusivity was also determined. Generally, the intra-particle diffusivity can be termed as ‘constant’ if the kinetic adsorption profile does not vary with the change in AC particle size. However, if the adsorption profile varies with the size of AC particles, the diffusivity is termed as ‘proportional’ (Crittenden et al., 1991). Fig. 3 shows the organic matter adsorption profile versus time in all three ACs for different particle sizes. The micro-porous AC showed that as the size of AC particles changed, the kinetic adsorption profile also changed, representing the proportional diffusivity of organic matter onto AC. However, the meso-porous ACs showed constant organic matter intraparticle diffusivity, possibly resulting from the larger pore diameter of meso-porous ACs that allow larger organic molecules to easily enter into the pores, whereas the smaller pore diameter of micro-porous AC does not readily allow entrance. No previous studies were found to compare these results. The rate of organic matter diffusion onto AC was determined using HPDM. The slope of ln½1 F2 ðtÞ with respect to time (t) gives the value of effective diffusion coefficient (De). The data satisfactorily fit in the entire range of diffusion for each size of AC particles as shown by r2 values of each plot in Table 3. The De values (Table 3) were in the order of
1.6 Region-I Region-II Region-III Region-IV Region-V Total volume
1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0
0
10
20
30
40
50
60
70
40
Calgon F 300 AC Norit HD 4000 AC Darco 12x40 AC
20
0
0
20
40
60
EBCT corresponding to the full-scale column (min) Fig. 4 e Effect of EBCT on the TOC removal of leachate using Calgon F-300, Norit HD-4000, and Darco 12 x 40 AC. Data points and error bars are the average and standard deviation of percentage TOC removal obtained from duplicate experiments, respectively. Initial TOC [ 670 mg LL1.
-1 -1 2
TOC removal (%)
60
Fractional DOM removal derived
c
80
from FRI analysis (AU-nm (mg L ) ), V/V0
EBCT long column (min) 1.6 Region-I Region-II Region-III Region-IV Region-V Total volume
1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0
0
10
20
30
40
50
60
70
EBCT long column (min)
Fig. 5 e Effect of EBCT on DOM of each EEM region under leachate treatment using (a) Calgon F-300, (b) Norit HD4000, and (c) Darco 12 x 40 AC. Volumes were derived from FRI analysis.
498
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 9 1 e4 9 9
changing particle sizes of meso-porous AC. Hence, mesoporous AC products can be selected based on availability without affecting the adsorption efficiency for leachate treatment.
3.4.
Column experiment
RSSCT experiments were conducted to determine the efficiency of AC columns for removal of organic matter at different EBCTs equivalent to EBCTs of the full-scale columns. As shown in Fig. 4, RSSCT results showed increasing TOC removal with increasing EBCT. However, an EBCT greater than 30 min showed an insignificant increase in TOC removal for all three ACs. Additionally, each AC tested showed similar TOC removal efficiency. A maximum of 80% of TOC was removed from all three ACs at an EBCT of 60 min. A few of the studies reported in Table S1 in Supplementary Materials found the maximum COD removal ranged from 50 to 90% for different leachate types. Fluorescence EEMs of the samples were obtained for the samples generated at each EBCT for all three ACs. A FRI analysis of EEM was conducted; Fig. 5 shows the organic matter fractional removal of each region versus the EBCT for all three ACs. Organic matter from each region was removed with an increasing amount of removal as the EBCT was increased. However, among all organic matter types, the highest removal was observed for the fulvic-like organic matter from each AC. Smaller organic matter of Region-I showed high adsorption at lower EBCT; however, Region-I organic matter desorbed at higher EBCT due to preferential adsorption of fulvic-like organic matter onto AC (Liu et al., 2010).
4.
Conclusions
This research investigated the effectiveness and selectivity of micro- and meso-porous ACs for the treatment of stabilized leachate. The adsorption capacity, rate of organic matter diffusion, and organic matter type removed by one microporous and two meso-porous ACs were studied using isotherms, kinetics, and RSSCTs, respectively. The following conclusions were drawn: Among the Freundlich, Langmuir, and RedlichePeterson models, the RedlichePeterson isotherm predicts the best adsorption capacity onto AC for stabilized leachate treatment. Micro- and meso-porous AC provide similar organic matter adsorption capacity for stabilized leachate treatment; however, AC particle size is an important parameter to consider while selecting GAC for stabilized leachate treatment. Meso-porous AC provides relatively faster adsorption of organic matter than Micro-porous AC. The RSSCT results showed a maximum of 80% TOC removal from each of the three ACs, and fulvic-like organic matter were adsorbed the most onto ACs. The selection of the type of AC for stabilized leachate treatment can depend on the economics and availability of
AC. Even though the AC showed very encouraging results in terms of TOC removal, the process would be best suited when combined with other treatment steps. The high organic matter concentrations present in stabilized leachate may exhaust the AC quickly, likely making the treatment process economically unfavorable.
Acknowledgments The authors thank Calgon Carbon Corporation and Norit for providing the activated carbons and Dr. Paul Chadik for his help and support with the laboratory space.
Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.11.007.
references
Borghi, A.D., Binaghi, L., Converti, A., Borghi, M.D., 2003. Combined treatment of leachate from sanitary landfill and municipal wastewater by activated sludge. Chemical and Biochemical Engineering Quarterly 17 (4), 277e283. Boyd, G.E., Myers, L.S., Adamson, A.W., 1947. The exchange adsorption of ions from aqueous solutions by organic zeolites. III. Performance of deep adsorbent beds under non aquarium conditions. Journal of American Chemical Society 69, 2849e2859. Chen, W., Westerhoff, P., Leenheer, J.A., Booksh, K., 2003. Fluorescence excitation-emission matrix regional integration to quantify spectra for dissolved organic matter. Environmental Science and Technology 37, 5701e5710. Comstock, S., Boyer, G., Graf, S., Townsend, T., 2010. Effect of landfill characteristics on leachate organic matter properties and coagulation treatability. Chemosphere 81, 976e983. Crittenden, J.C., Berrigan, J.K., Hand, D.W., 1986. Design of rapid small-scale adsorption tests for a constant surface diffusivity. Journal of Water Pollution Control Federation 58 (4), 312e319. Crittenden, J.C., Reddy, P.S., Arora, H., Trynoski, J., Hand, D.W., Perran, D.L., Summers, R.S., 1991. Predicting GAC performance with rapid small-scale column tests. Journal of American Water Works Association 83 (1), 77e87. Fettig, J., Stapel, S., Steinert, C., Gieger, M., 1996. Treatment of landfill leachate by preozonation and adsorption in activated carbon columns. Water Science and Technology 34 (9), 33e40. Foo, K.Y., Hameed, B.H., 2009. An overview of landfill leachate treatment via activated carbon adsorption process. Journal of Hazardous Materials 171, 54e60. Freundlich, H.M.F., 1906. Over the adsorption in solution. Journal of Physical Chemistry 57, 385e470. Gotvajn, A.Z., Tisler, T., Koncan, T., 2009. Comparison of different treatment strategies for industrial landfill leachate. Journal of Hazardous Materials 162, 1446e1456. Halim, A.A., Aziz, H.A., Johari, M.A.M., Ariffin, K.S., 2010. Comparison study of ammonia and COD adsorption on zeolite, activated carbon and composite materials in landfill leachate treatment. Desalination 262, 31e35. Ho, Y.S., Porter, J.F., McKay, G., 2002. Equilibrium isotherm studies for the sorption of divalent metal ions onto peat: Copper,
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 9 1 e4 9 9
nickel, and lead single component system. Water, Air, and Soil Pollution 141, 1e33. Karanfil, T., Dastgheib, S.A., Mauldin, D., 2006. Exploring molecular sieve capabilities of activated carbon fibers to reduce the impact of NOM preloading on trichloroethylene adsorption. Environmental Science and Technology 40, 1321e1327. Kurniawan, T.A., Lo, W.-H., Chan, G.Y.S., 2006. Degradation of recalcitrant compounds from stabilized landfill leachate using a combination of ozone-GAC adsorption treatment. Journal of Hazardous Materials B137, 443e455. Langmuir, I., 1916. The adsorption of gases on plane surface of glass, mica, and platinum. Journal of American Chemical Society 40, 1361e1368. Li, F., Yuasa, A., Katamine, Y., Tanaka, H., 2007. Breakthrough of natural organic matter from fixed bed adsorbers: investigations based on size-exclusion HPLC. Adsorption 13, 569e577. Li, Q., Marinas, B.J., Snoeyink, V.L., Campos, C., 2003. Threecomponent competitive adsorption model for flow-through PAC system. 1. Model development and verification with a PAC/membrane system. Environmental Science and Technology 37, 2997e3004. Liu, F., Xu, Z., Wan, H., Wan, Y., Zheng, S., Zhu, D., 2010. Enhanced adsorption of humic acids on ordered esoporous carbon as compared with microporous activated carbon. Environmental Chemistry (doi:10.1002/etc.450). Mackenzie, J.A., Tennant, M.F., Mazyck, D.W., 2005. Tailored GAC for the effective control of 2-Methylisoborneol. Journal of American Water Works Association 97 (6), 76e87. Maranon, E., Castrillon, L., Fernandez-Nava, Y., FernandezMendez, A., 2009. Tertiary treatment of landfill leachate by adsorption. Waste Management Research 27, 527e533. McKay, G., Otterburn, M.S., Sweeney, A.G., 1980. The removal of colour from effluent using various adsorbents e III silica: rate processes. Water Research 14, 15e20. Redlich, O., Peterson, D.L., 1959. A useful adsorption isotherm. Journal of Physical Chemistry 63, 1024.
499
Rivas, F.J., Beltran, F., Carvalho, F., Acedo, B., Gimeno, O., 2004. Stabilized leachate: sequential coagulation flocculation þ chemical oxidation process. Journal of Hazardous Materials B 116, 95e102. Rivas, F.J., Beltran, F., Gimeno, O., Frades, J., Carvalho, F., 2006. Adsorption of landfill leachates onto activated carbon equilibrium and kinetics. Journal of Hazardous Materials B 131, 170e178. Tatsi, A.A., Zouboulis, A.I., Matis, K.A., Samaras, P., 2003. Coagulation-flocculation pre-treatment of sanitary landfill leachates. Chemosphere 53, 737e744. Thorneby, L., Hogland, W., Stenis, J., Mathiasson, L., Somogyi, P., 2003. Design of a reverse osmosis plant for leachate treatment aiming for safe disposal. Waste Management Research 21, 424e435. Tizaoui, C., Bouselmi, L., Mansouri, L., Ghrabi, A., 2007. Landfill leachate treatment with ozone and ozone/hydrogen peroxide systems. Journal of Hazardous Material 140, 316e324. Trebouet, D., Schlumpf, J.P., Jaouen, P., Quemeneur, F., 2001. Stabilized landfill leachate treatment by combined physicochemical-nanofiltration processes. Water Research 35 (12), 2935e2942. Vermeulen, T., 1953. Theory of irreversible and constant-pattern solid diffusion. Industrial Engineering Chemistry 45, 1664e1670. Weber, W.J., Morris, J.C., 1963. Kinetics of adsorption onto carbon from solutions. Journal of Sanitary Engineering Division ASCE 89, 31e60. Xing, W., Ngo, H.H., Guo, W.S., Hagare, P., 2008. Physicochemical processes for landfill leachate treatment: experiments and mathematical modeling. Separation Science and Technology 43, 347e361. Ying, W., Zhang, W., Chang, Q., Jiang, W., Li, G., 2006. Improved methods for carbon adsorption studies for water and wastewater treatment. Environmental Progress 25 (2), 110e120.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 0 e5 0 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A novel two-step coprecipitation process using Fe(III) and Al(III) for the removal and immobilization of arsenate from acidic aqueous solution Yongfeng Jia a,*, Danni Zhang a, Rongrong Pan a, Liying Xu a, George P. Demopoulos b a
Key Laboratory of Pollution Ecology and Environmental Engineering, Institute of Applied Ecology, Chinese Academy of Sciences, Shenyang 110016, China b Department of Mining and Materials Engineering, McGill University, Montreal, QC H3A 2B2, Canada
article info
abstract
Article history:
Lime neutralization and coprecipitation of arsenate with iron is widely practiced for the
Received 7 August 2011
removal and immobilization of arsenic from mineral processing effluents. However, the
Received in revised form
stability of the generated iron-arsenate coprecipitate is still of concern. In this work, we
9 November 2011
developed a two-step coprecipitation process involving the use of iron and aluminum and
Accepted 15 November 2011
tested the stability of the resultant coprecipitates. The two-step FeeAseFe or FeeAseAl
Available online 25 November 2011
coprecipitation process involved an initial Fe/As ¼ 2 coprecipitation at pH4 to remove arsenic from water down to 0.25 mg/L, followed by introduction of iron or
Keywords:
aluminum (Fe/As ¼ 2, Al/As ¼ 1.5 or 2). The two-step coprecipitates showed higher stability
Two-step
than traditional Fe/As ¼ 4 coprecipitate under both oxic and anoxic conditions. Leaching
Coprecipitation
stability was enhanced when aluminum was applied in the second step. The use of
Arsenate
aluminum in the second step also inhibited microbial mediated arsenate reduction and
Removal
arsenic remobilization. The results suggest that the two-step coprecipitation process is
Immobilization
superior to conventional coprecipitation methods with respect to the stability of the generated arsenic-bearing solid waste. The use of Al in the second step is better than Fe to enhance the stability. This work may have important implications to the development of new technologies for efficient arsenic removal from hydrometallurgical solutions and safe disposal in both oxic and anoxic environment. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Arsenic is commonly cooccurring with base metals and precious metals as arsenic-bearing minerals, such as arsenopyrite and arsenic sulfides (Riveros et al., 2001). In hydrometallurgical operations, arsenic is released into mineral processing solutions and effluents during extraction of metals by oxidation and acid dissolution of the arsenic-containing minerals. Because of its well known toxicity, arsenic in the
industrial effluents must be removed and immobilized as a stable solid and disposed safely for the prevention of contamination to nearby surface and ground waters. Among the available technologies, lime neutralization accompanied by coprecipitation of arsenic with ferric iron is the industrial method of choice for the removal of arsenic from acidic mineral processing effluents (Harris, 2003; Twidwell et al., 2005). Arsenate can be efficiently removed from the solution to sub-ppm level, when excess ferric iron is
* Corresponding author. Tel./fax: þ86 24 8397 0503. E-mail address:
[email protected] (Y. Jia). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.045
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 0 e5 0 8
present (Harris, 2000). The generated arsenic-bearing solids are disposed in specially designed waste storage ponds (Langmuir et al., 1999). FeIIIeAsV coprecipitate can easily pass the permissible level of toxicity characteristic leaching procedure (TCLP) measurement and is usually regarded as one of the most stable arsenic-bearing solid wastes. However, its long-term stability is still of concern because it may be reductively solubilized in tailings pond (McCreadie et al., 2000). Influx of natural organic matter into the tailings pond provides nutrients to microbes, as a result, the oxygen in porewaters of the subsurface tailings will be consumed and anaerobic environment will be developed. Below the redox boundaries, both AsV and FeIII may be reduced biotically or abiotically. This may lead to arsenic remobilization from the coprecipitate due to reductive dissolution of iron (hydr)oxides and weaker affinity of AsIII to various minerals (Raven et al., 1998; Pedersen et al., 2006). Additionally, ferrihydrite, which accommodates arsenic in the coprecipitated solid, may transform to higher crystallinity forms (e.g. goethite) hence releasing the associated arsenic, although the recrystallization process may be retarded by the incorporated arsenate (Ford, 2002). Efforts have been made in recent years to develop new technologies for improving the stability of arsenic-bearing hydrometallurgical wastes, including new methods of precipitation and encapsulation of arsenic-containing solid wastes, most of them involving the use of AlIII. Encapsulation of scorodite with aluminum phosphate and aluminum gel enhanced the stability of scorodite against leaching or reductive dissolution (Leetmaa, 2008; Lagno et al., 2010). Coprecipitation of AlIIIeFeIIIeAsV measurably enhanced the effectiveness of arsenic removal from aqueous solutions compared to FeIIIeAsV coprecipitation (Robins et al., 2005). The coprecipitated AlIIIeFeIIIeAsV solids were studied with respect to their nature, mineralogy and reactivity, and AlIII was found to hinder the recrystallization of amorphous iron hydroxide (Violante et al., 2009). However, the stability of AlIIIeFeIIIeAsV coprecipitate against both TCLP leaching and reductive dissolution is unknown. In this study, a novel two-step coprecipitation process was tested, with the first step involving FeeAs coprecipitation to remove arsenate from solution, followed by second step involving ferric or aluminum hydroxide precipitation to improve the stability of the coprecipitate under oxic and anoxic conditions. The overall objective of this work is to develop a new technology for arsenic removal and immobilization from industrial effluents, which is superior to conventional coprecipitation methods with respect to the stability of the generated arsenic-bearing solid.
2.
Experimental section
2.1.
Two-step FeeAseFe and FeeAseAl precipitation
The AsV stock solution was prepared by dissolving Na3AsO4$12H2O in DI-water and the pH was adjusted to w1.5 with H2SO4. The FeIII and AlIII solutions were prepared by dissolving Fe2(SO4)3$5H2O and Al2(SO4)3$18H2O in DI-water, respectively.
501
The two-step FeeAseFe or FeeAseAl precipitation process included initial FeIIIeAsV coprecipitation and subsequent FeIII-(hydr)oxide or AlIII-(hydr)oxide deposition. The first step involved neutralization of an acidic FeIIIeAsV solution (Fe/As ¼ 2) to pH4 with 1 M NaOH or Ca(OH)2 and stabilization for 1 h. NaOH is usually used in laboratory while in industry slaked lime is used as base. The second step was conducted by addition of FeIII or AlIII solution from burette into the Fe/As ¼ 2 coprecipitate slurry at molar ratio of Fe/As ¼ 2, Al/As ¼ 2 or 1.5, while the mixture was mechanically stirred and the pH was maintained constant at 4. The mixture was then raised to pH8, transferred to conical flasks and equilibrated in a rotary shaker at 25 C. The volume of the slurry upon neutralization was 500 mL. The coprecipitation systems and the produced solids are designated as Fe2As1Fe2, Fe2As1Al2 or Fe2As1Al1.5 respectively. One-step Fe/As ¼ 4 coprecipitation test was also performed for comparison with the above-mentioned two-step coprecipitation process. In this process, an acidic Fe/As ¼ 4 solution was adjusted to pH4 with NaOH or Ca(OH)2, equilibrated for 1 h at pH4, then raised to pH8. This coprecipitation process and the produced coprecipitate are designated as Fe4As1. To prepare the coprecipitate samples for stability tests, the solids were separated after 24 h equilibration, washed with DIwater and freeze-dried. The reference materials for X-ray diffraction (XRD) analysis were also synthesized in lab. Two-line ferrihydrite sample was synthesized at room temperature using the procedure reported in the literature (Schwertmann and Cornell, 1991). Poorly crystalline ferric arsenate sample was synthesized according to reference (Jia et al., 2006). Amorphous aluminum hydroxide sample was synthesized by introducing AlIII solution into pH4 DI-water, equilibrating at pH4 for 1 h and raising to pH8. The solid was separated, rinsed with DI-water and freeze-dried.
2.2.
Leaching stability under oxic conditions
Two methods were used to evaluate leaching stability of the coprecipitates under oxic conditions. The first method: 1 g of solid sample was added into 20 mL of DI-water with the pH pre-adjusted to the target value (i.e. 3e9). The mixtures were allowed to equilibrate in a rotary shaker at 25 C for 72 h, with the pH maintained constant. The solutions were separated by filtration through 0.22-mm membrane filters and analyzed for the concentration of arsenic. Note that this method simulates TCLP leaching procedure. The second method: 1 g of solid was mixed with 20 mL of DI-water and the mixture was agitated at 25 C for 24 h. The solution was then replaced with 20 mL of fresh DI-water and equilibrated for another 24 h. This test was repeated 10 times with the pH controlled constant at 5 and 8. The concentration of arsenic in the leachate of each step was analyzed. Such leaching method was used to simulate the behavior of arsenic at disposal site (Krause and Ettel, 1989).
2.3.
Stability against sulfide reduction
The stability of the precipitates against reductive dissolution under anoxic conditions was tested by reaction with sulfide
502
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 0 e5 0 8
(Lagno et al., 2010). The precipitates were mixed with Na2S solution in 50 mL serum bottles with butyl rubber stoppers and a screw-cap lid and then purged with ultrapure N2. The bottles were agitated in the dark at 25 C in a rotary shaker with the pH of the slurries maintained at 7 0.2. At different time intervals, the serum bottles were removed from the shaker and transferred to an anaerobic glove chamber. The slurries were filtered through 0.22-mm filters and the solids were digested with 1 M HCl to analyze the contents of reduced As and Fe in solid. The concentrations of AsIII, AsV and FeII in both filtrates and digested solids were analyzed. The tests were run in duplicate and the average values were reported.
2.4.
Stability against microbial reduction
The coprecipitates were also subjected to stability evaluation under the impact of anaerobic microbes. The microbial consortia used as inocula were enriched from the contaminated soil of an abandoned smelter site (Zhang et al., 2008). Detailed microbial community analysis was reported previously by Zhang et al. (2008). Briefly, we analyzed the microbial diversity by using 16S rDNA-dependent molecular phylogeny and constructed a near-full-length 16S rDNA gene clone library. After analysis of 197 clones by using restriction fragment length polymorphism (RFLP) method, we obtained 72 operational taxonomic units (OTUs), accounting for 51% of the content for total clone number in six OTUs. Sequencing analysis was performed on six bacterial clones in these six OTUs and we found that these clones belong to the group Caloramator, Clostridium, and Bacillus. The coprecipitates were sterilized by autoclaving and mixed with lactate-amended minimal media at solid/liquid ratio of 1/100 in 50 mL serum bottles. The lactate-amended minimal medium was made according to literature (Campbell et al., 2006; Zhang et al., 2008) and the initial lactate concentration was 10 mmol/L. The biotic incubations were inoculated with 1% (v/v) of inoculums (w105 cells) in each culture serum bottle. The biotic trials and the abiotic controls were then incubated in the dark at 30 C in a rotary shaker. All trials were conducted in triplicate under strict anoxic conditions. At different time intervals, the serum bottles were removed from the incubator and transferred to an anaerobic glove chamber. The slurries were filtered through 0.22-mm filters to determine the concentration of aqueous arsenic and iron. The solids were dissolved with 1 M HCl to determine the concentration of produced As(III) and Fe(II) in solid.
2.5. Determination of arsenic, aluminum and iron concentration Arsenic concentration was analyzed by using hydride generation-atomic fluorescence spectrometry (see SI). Fe concentration was determined on an atomic absorption spectrometer. Fe(II) concentration was determined on a UVevis spectrophotometer after reaction with 1,10phenanthroline. The concentration of aluminum in solution was determined using ICP-AES.
2.6.
XRD analysis
The mineralogical characteristics of the coprecipitates was analyzed by using XRD. The powder XRD patterns were obtained on a Rigaku D/max 2000PC X-ray diffractometer equipped with a copper target, a crystal graphite monochromator and a scintillation detector. The equipment was run at 56 kV and 182 mA by step-scanning from 10 to 90 2q with increments of 0.02 2q.
2.7. X-ray absorption near edge structure (XANES) analysis The change of arsenic oxidation state after sulfidization was analyzed by synchrotron-based XANES. The arsenic K-edge XANES spectra were collected on the X-ray absorption fine structure (XAFS) station at National Synchrotron Radiation Laboratory (NSRL) of China. The NSRL storage ring was operated at 0.8 GeV and the ring currents of 300 mA. The monochromator was a fixed exit double crystal monochromator with Si(111) crystals. Data was acquired in transmission mode.
3.
Results and discussion
3.1.
Characterization of the coprecipitates
The SEM microimages of the FeeAs and FeeAseAl coprecipitates are shown in Fig. 1. There is no apparent difference in their appearance under 105 times magnification. Both materials appear to be aggregates of homogeneous nucleated nano-sized particles (<10 nm) (Demopoulos, 2009). The two-step FeeAseAl coprecipitates were characterized using XRD and their patterns are compared with those of reference materials in Fig. 2. The Fe/As ¼ 2 coprecipitate produced at pH4 (i.e. Fe2As1, the product of the first step coprecipitation) shows two XRD bands lying between the characteristic bands of ferrihydrite and poorly crystalline ferric arsenate. According to previous reports, the coprecipitated FeIIIeAsV solid at acidic pH usually consists of poorly crystalline ferric arsenate and arsenateadsorbed ferrihydrite (Jia et al., 2003; Chen et al., 2009). The two-step coprecipitate (Fe2As1Al2) exhibits similar XRD patterns to those of Fe2As1. No characteristic bands of either crystalline or amorphous aluminum hydroxides are observed. The precipitated aluminum hydroxide was probably amorphous in nature and its XRD diffraction bands may be obscured by those of poorly crystalline ferric arsenate and ferrihydrite. After aging for 300 days at room temperature, no crystalline materials evolved from Fe2As1Al2. The crystallinity of precipitated aluminum hydroxide during hydrolysis of AlIII ions strongly depends on OH/Al ratio and the rate of hydrolysis (Bottero and Bersillon, 1989; Gella, 2007). At OH/Al 3, crystalline aluminum hydroxide was obtained whereas at OH/ Al 2.8, the hydrolysis product was amorphous (Gella, 2007). The OH/Al molar ratio applied in this work was 2.7. This is probably the reason why no crystalline aluminum hydroxide was observed in the two-step FeeAseAl coprecipitates by using XRD. Different mineralogy was also observed by Masue et al. (2007) and Violante et al. (2006, 2009) in previous studies on precipitation or coprecipitation of aluminum hydroxide or
503
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 0 e5 0 8
Fig. 1 e SEM microimages of Fe2As1Al2 (top) and Fe4As1 (bottom). Fe4As1 e one-step coprecipitation: an acidic Fe/ As [ 4 solution was neutralized to pH 8. Fe2As1Al2 e twostep coprecipitation: an acidic Fe/As [ 2 solution was neutralized to pH 4, followed by addition of FeIII or AlIII, then the pH was raised to 8.
2008). The second step involving introducing the other half of the iron (or equimolar aluminum) into the Fe/As ¼ 2 coprecipitate slurry was applied to improve its stability. The stability of the coprecipitates in original solutions in terms of As concentration was monitored with time (Fig. 3). Arsenic was released back to solution from solid for all coprecipitation processes. The concentration of arsenic for the one-step Fe/As ¼ 4 coprecipitation system (Fe4As1) gradually reached a plateau of w32 mg/L after two months of equilibration. Two-step coprecipitation process apparently suppressed arsenic release to solution. Arsenic release was reduced by approximately one third if the Fe/As ¼ 4 coprecipitation was conducted in two-step mode (Fe2As1Fe2), i.e. neutralizing Fe/As ¼ 2 solution to pH4 (step 1) followed by addition of equimolar FeIII and raising pH to 8 (step 2). The use of AlIII in step 2 was superior to FeIII in suppressing arsenic remobilization. Only 3.2 mg/L of arsenic was released after one month of equilibration for Fe2As1Al2, thereafter As concentration remained at a steady level. This is only one tenth that of one-step Fe/As ¼ 4 coprecipitation (Fe4As1). The release of As back to solution after coprecipitation was attributed to the decomposition of ferric arsenate compounds that formed at acidic pH and/or desorption of adsorbed arsenate when the pH was raised from acidic to slightly alkaline region (Jia and Demopoulos, 2008), since both species are known to be more stable in mildly acidic media (Robins et al., 1988; Krause and Ettel, 1989; Raven et al., 1998; Nishimura and Umetsu, 2000; Jia and Demopoulos, 2005). Additionally, conversion of freshly precipitated very amorphous iron
amorphous Al(OH)3 ferrihydrite
aluminum-ferric hydroxides where different experimental conditions were applied.
poorly crystalline ferric arsenate
Stability of the coprecipitates in mother liquors
Conventional coprecipitation method involves using excess iron, usually FeIII/AsV 4, to ensure efficient arsenic removal and the stability of the generated arsenic-bearing solid (Langmuir et al., 1999; Harris, 2000; Moldovan et al., 2003). In this work, half of the iron (i.e. Fe/As ¼ 2) was applied to remove arsenic from solution. The preliminary results showed that Fe/As ¼ 2 coprecipitation at mildly acidic pH was sufficient to remove arsenate to sub-ppm level. Arsenic concentration after coprecipitation was the lowest at pH4 (0.25 mg/L) compared to that at pH5 (0.58 mg/L) and pH6 (0.79 mg/L) (see SI). Hence, the first-step of the coprecipitation process was conducted by neutralizing an acidic FeIIIeAsV solution (Fe/ As ¼ 2) to pH4 to ensure efficient arsenic removal from solution. Lower Fe/As molar ratio was applied in order to reduce the cost of operation and the volume of solid waste, which are also important issues in industrial effluent treatment and solid waste disposal. However, the Fe/As ¼ 2 coprecipitate is not stable in neutral to mildly alkaline media (Jia and Demopoulos,
Fe2As1Al2
Intensity
3.2.
-- 300 day Fe2As1Al2 -- 1 day Fe2As1
0
20
40
60
80
100
2θ Fig. 2 e XRD patterns of the fresh and aged two-step FeeAseAl coprecipitates (Fe2As1Al2, see Fig. 1 for the sample information in details) together with the first step product and reference materials.
504
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 0 e5 0 8
hydroxides to more ordered forms may also contribute to arsenic mobilization, although the recrystallization process is usually retarded by adsorbed arsenate anions (Ford, 2002; Das et al., 2011). Precipitation of ferric or aluminum hydroxide on the initially precipitated Fe/As ¼ 2 solid provides a barrier to inhibit arsenic release. Microencapsulation mechanism cannot explain the enhanced stability of the two-step coprecipitates. Amorphous ferric and aluminum hydroxides are not as dense as other crystalline materials used in microencapsulation, e.g. aluminum phosphate (Lagno et al., 2010), hence cannot provide complete physical insulation to porewaters. Instead, they may serve as adsorbents to take up the arsenic released from the Fe/As ¼ 2 solid. When the Fe/As ¼ 2 coprecipitation system was raised to pH8, arsenic was released up to w120 mg/L (Jia and Demopoulos, 2008). If this amount is assumed to be uptaken by the amorphous iron or aluminum hydroxide precipitated in step two, the Fe/As or Al/ As molar ratio of the secondary precipitate is estimated to be 24 for Fe2As1Fe2 and Fe2As1Al2, or 18 for Fe2As1Al1.5. This is much higher than the nominal molar ratio of Fe/As ¼ 4 coprecipitate, hence leading to enhanced stability of the two step coprecipitates. The use of aluminum in the second step of the two-step coprecipitation process is superior to the use of equimolar ferric iron for the purpose of suppressing arsenic release at pH8. This is probably due to the difference of sorption capacity between ferrihydrite and Al-(hydr)oxide, because amorphous aluminum oxide was reported to sorb greater amount of arsenate than amorphous iron oxide at pH 8 (Goldberg and Johnston, 2001).
3.3.
Leaching stability
The coprecipitates prepared using different methods showed apparently different leaching stabilities as determined by equilibration in water for 72 h at pH3-9 (Fig. 4). Two-step coprecipitation enhanced the stabilities of the arsenic-
bearing solids (Fe2As1Al2 vs. Fe4As1). For NaOH-neutralized systems, the solubility of two-step coprecipitate in terms of As concentration at pH6-9 was one order of magnitude lower than that of one-step coprecipitation. For Ca(OH)2 neutralized coprecipitates, the solubility in terms of As concentration at pH5e9 was reduced several times when two-step coprecipitation was applied. The concentration of arsenic in Ca(OH)2 neutralized Fe/As ¼ 4 system at pH8 was 0.8 mg/L, which is comparable to the value of w1 mg/L (Jia and Demopoulos, 2008) and <2 mg/L (Langmuir et al., 1999) in similar laboratory systems and the value of w1 mg/L in tailings pond where the iron-arsenate coprecipitate from uranium hydrometallurgical industry is disposed (Rowson, private communication). Two-step coprecipitation reduced the solubility of the coprecipitate to <0.2 mg/L As at pH8 (Fig. 4). One-step FeeAs coprecipitate (Fe4As1) showed higher stability than two-step FeeAseAl coprecipitate (Fe2As1Al2) at pH 3. Because aluminum hydroxide has higher solubility than ferric hydroxide at this pH (Cornell and Schwertmann, 1996; Lindsay and Walthall, 1996), more arsenic will be released from Fe2As1Al2 compared to Fe4As1 after leaching at pH 3. All coprecipitates showed lower solubility in terms of arsenic concentration at slightly acidic pH. This is consistent with previous reports on the solubilities of iron-arsenate coprecipitates and scorodite (Robins et al., 1988; Krause and Ettel, 1989; Nishimura and Umetsu, 2000; Jia and Demopoulos, 2008). The use of Ca(OH)2 as opposed to NaOH as base improved the stability of both one-step and two-step coprecipitates. This effect was attributed to CaIIeFeIIIeAsV association, AsVeCaII coadsorption, and/or positively charged hydroxide surface by calcium ions (Harris and Monette, 1988; Wilkie and Hering, 1996; Jia and Demopoulos, 2005; 2008). Leaching stability of the two-step coprecipitate was also assessed by repeatedly contacting the solid with fresh water, which simulates its behavior at disposal site (Krause and Ettel, 1989). Arsenic concentration in the leachates of the two-step coprecipitate Fe2As1Al2 neutralized using NaOH and Ca(OH)2 as base is compared with that of one-step coprecipitate Fe4As1
40 Fe4As1 (NaOH)
100
30
Fe2As1Fe2
20 Fe2As1Al1.5 10 Fe2As1Al2 0 0
20
40
60
80
100
120
Time (day)
Fig. 3 e As concentration in solution with time after coprecipitation using NaOH as base under oxic conditions. Fe4As1: one-step coprecipitation; Fe2As1Fe2, Fe2As1Al1.5, Fe2As1Al2: two-step coprecipitation. See Fig. 1 for experimental conditions.
Dissolved As concentration (mg/L)
Aqueous As concentration (mg/L)
Fe4As1 100
Fe4As1 (Ca(OH)2) Fe2As1Al2 (NaOH)
10
10
Fe2As1Al2 (Ca(OH)2)
1
1
0.1
0.1
0.01
3
4
5
6
7
8
9
0.01
pH
Fig. 4 e Leaching stabilities of one-step (Fe4As1) and twostep (Fe2As1Al2) coprecipitates after equilibration in water at different pH for 72 h under oxic conditions. The coprecipitates were produced using NaOH and Ca(OH)2 as base respectively.
505
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 0 e5 0 8
3.4.
Stability against reductive dissolution
Reductive dissolution is an important factor affecting the stability of disposed iron-arsenate coprecipitates in tailings pond due to anoxic nature in subsurface zone. A reducing environment was simulated by addition of sulfide and the subsequent stability of the one-step FeeAs coprecipitate (Fe4As1) and two-step FeeAseAl coprecipitate (Fe2As1Al2) was assessed by monitoring the release of arsenic during reaction (Fig. 6). The one-step FeeAs coprecipitate was destabilized after reaction with sulfide. The concentration of aqueous arsenic increased rapidly to w80 mg/L in 7 days, corresponding to approximately 80% of arsenic being released. The dissolved arsenic by sulfide was present predominantly as AsV, with only <5% present as AsIII. The two-step FeeAseAl coprecipitate showed much higher stability against reduction by sulfide. Arsenic was dissolved to w22 mg/L after 7 days of reaction, corresponding to 22% of arsenic being released, which was only one fourth that of one-step FeeAs coprecipitate. FeIII of the coprecipitates was reduced to FeII by sulfide. In Fe4As1, FeII in the solid phase reached w140 mg/L after 7 d of reaction while dissolved FeII was w35 mg/L. The produced FeII in Fe2As1Al2 after reaction with sulfide was w80 mg/L in solid phase and w20 mg/L in aqueous phase. During reduction reaction, the produced FeII may react with SII and FeS precipitation may occur. In addition, the produced FeII may be partly re-incorporated into ferrihydrite to form amorphous Fe3O4. Structural aluminum in ferrihydrite was found to decrease reincorporation of produced FeII into ferrihydrite during microbial mediated reduction (Masue-Slowey et al., 2011). In the present study, aluminum hydroxide was deposited onto preprecipitated Fe/As ¼ 2 solid rather than being incorporated into ferrihydrite structure. Hence, retention of the produced FeII by ferrihydrite may not be retarded by aluminum. Arsenic speciation in the solid phases after reaction of Fe4As1 and Fe2As1Al2 with sulfide was characterized by XANES
and the spectra were compared with those of reference compounds (Fig. 7). Arsenic K-edge XANES spectra of the sulfidized coprecipitates were dominated by arsenate features. There might be small amount of arsenite present in the solids but the features were obscured by those of arsenate. Neither arsenic sulfide nor arsenopyrite was identified from XANES spectra. The results of anaerobic microbial mediated reduction and dissolution of the coprecipitates are shown in Fig. 8. For the FeeAs coprecipitate (Fe4As1), arsenate was reduced to arsenite. The concentration of total AsIII in the system (both solid and aqueous phases) reached w125 mg/L after five weeks of incubation, accounting for 60% of AsV being reduced to AsIII. Most of the produced AsIII was retained by the solid phase. Large quantities of arsenic were released as a result of anaerobic microbial activities. The concentration of dissolved arsenic increased to 26 mg/L after two weeks of incubation. Almost all dissolved arsenic was present as AsIII. Ferric iron was also reduced during anaerobic incubation. The concentration of total FeII in the system (both solid and aqueous phases) reached w260 mg/L after three weeks, accounting for 40% of ferric iron being reduced to ferrous. Less than one third of FeII was released to solution with the concentration of aqueous FeII reaching w65 mg/L. Most of reduced iron was retained by solid, with FeIII/FeII molar ratio of w2.6. Microbial mediated FeIII and AsV reduction may lead to As dissociation from the host minerals and release into aqueous
0.4
Fe4As1 (NaOH) Fe4As1 (Ca(OH)2) Fe2As1Al2 (NaOH)
0.3
Fe2As1Al2 (Ca(OH)2)
0.2 pH5 Dissolved As concentration (mg/L)
(Fig. 5). All coprecipitates showed very low arsenic concentration in leachates (w0.05e0.1 mg/L) at pH5. However, the stabilities were different when these coprecipitates were subjected to repeated freshwater-leaching at pH8. The twostep FeeAseAl coprecipitates were more stable than onestep FeeAs coprecipitates. After five times of solvent replacement, the concentration of arsenic in the leachate was constant at w0.1 mg/L for Fe2As1Al2 (either Ca(OH)2 or NaOH as base), which was ten or two hundred times lower than that of Fe4As1 neutralized using Ca(OH)2 (w1 mg/L) or NaOH (w20 mg/L) as base respectively. There was a sharp increase of As concentration in the leachates of the second replacement of solution for NaOHneutralized coprecipitates, whereas Ca(OH)2-neutralized coprecipitates showed relatively stable concentration of leached As. Since slaked lime is usually used as base in industry to neutralize acidic FeeAs solution, arsenic leaching probably varies little at disposal site. As discussed above, the enhanced stability of the two-step coprecipitates was attributed to the barrier role of the Al(hydr)oxide precipitated in the second step. The released arsenic from the first step Fe/As ¼ 2 coprecipitate can be efficiently captured by the second step precipitate Al-(hydr)oxide.
0.1
0.0 200 150 100 50
3 pH8 2 1 0 0
2
4
6
8
10
Number of solution replacements
Fig. 5 e Leaching stabilities of the coprecipitates at pH 5 and 8 under oxic conditions with the leachate replaced with fresh water every 24 h of equilibration.
506
Dissolved As concentration (mg/L)
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 0 e5 0 8
100
100 Fe4As1
80
Fe2As1Al2
60
80 60
As(T) - red As(III) - red As(T) -ox
40
40
20
20
0
0 0
Fe concentration (mg/L)
As(T) - red As(III) - red As(T) -ox
150
1
2
3
4
Fe4As1
5
6
7
0
1
2
3
4
5
6
7
Fe2As1Al2
Solid Fe(II)
150 Solid Fe(II)
100
100
Aqueous Fe(II) 50 0
50
Aqueous Fe(II)
0
1
2
3
4
5
6
7
0
1
2
3
4
5
6
7
8
0
Time (day)
Fig. 6 e Comparison of the stabilities of one-step and two-step coprecipitates (Fe4As1 vs. Fe2As1Al2, NaOH as base) against reductive dissolution by reacting with sulfide under following conditions: pH 7; As in the system: 100 mg/L; S/As [ 20. As(T): total arsenic; As(III): Arsenite; Fe(II): ferrous iron; red: under reducing conditions; ox: under oxic conditions.
phase. This is one of proposed mechanisms for elevated concentration of As in groundwater. Reduction of FeIII may result in either dissolution of host Fe oxyhydroxides or conversion to low surface area oxide minerals hence releasing their associated As. Reduction of AsV to AsIII may also cause its release since AsIII is well known for its weaker affinity to various minerals compared to AsV. For the surface-bound AsV, its reduction may proceed before FeIII reduction, whereas the AsV incorporated in the structure of Fe oxyhydroxides is released first as a result of reductive dissolution of Fe, followed by aqueous reduction to AsIII (Postma et al., 2010). Reduction of FeIII may in most cases transform Fe oxyhydroxides to mixed-valence iron oxides, e.g. amorphous Fe3O4 and/or green rust due to re-incorporation of the produced FeII. Most of their associated As, even partially reduced to AsIII, can still be uptaken by the newly formed biogenic secondary Fe oxide minerals provided that Fe/As molar ratio is sufficiently high (e.g. Fe/As ¼ 100) (Islam et al., 2005; Coker et al., 2006). The Fe/As molar ratio applied in this work was 4, hence microbial mediated AsV and FeIII reduction caused significant arsenic release. The results of microbial mediated AsV and FeIII reduction of the two-step coprecipitate Fe2As1Al2 and subsequent As release are compared with those of one-step coprecipitate Fe4As1 (Fig. 8). Apparently, two-step coprecipitation process involving the use of aluminum lowered the extent of AsV reduction. The concentration of total AsIII in the system maximized at w45 mg/ L, which was only one third that of its one-step coprecipitation counterpart Fe4As1. Released arsenic from Fe2As1Al2 reached a plateau of w15 mg/L after two weeks of incubation. This was also notably lower than that of Fe4As1, indicating that anaerobic microbial mediated remobilization of arsenic from the
coprecipitate was partially suppressed when two-step coprecipitation process was applied. Arsenate reduction was found to accelerate arsenic release from the ferrihydrite with incorporated structural aluminum due to lower affinity aluminum
arsenopyrite realgar orpiment arsenite arsenate
-II
Fe4As1+S
-II
Fe2As1Al2+S
11860
11870
11880
11890
Photon energy (eV) Fig. 7 e Arsenic K-edge XANES spectra of Fe4As1 and Fe2As1Al2 after reaction with sulfide along with reference compounds. See Fig. 6 for reaction conditions.
507
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 0 e5 0 8
As concentration( mg/L)
100
100 Fe4As1
80
Solid As(III) Aqueous As(III) Aqueous As(T) Aqueous As(T), sterile
Solid As(III) Aqueous As(III) Aqueous As(T) Aqueous As(T), sterile
60 40 20
5
10
15
20
25
30
35
5
Fe4As1
200
10
15
20
25
30
35
Fe2As1Al2
60 40
100
0 250 200
Solid Fe(II)
150
Solid Fe(II)
Aqueous Fe(II)
Aqueous Fe(II)
150 100 50
50 0
80
20
0 250 Fe concentration (mg/L)
Fe2As1Al2
0
5
10
15
20
25
30
35
0
5
10
15
20
25
30
35
0 40
Time (day)
Fig. 8 e Microbial reduction and remobilization of arsenic of one-step and two-step coprecipitates (Fe4As1 vs. Fe2As1Al2, Ca(OH)2 as base) at pH 7 under anoxic conditions. As(T): total arsenic; As(III): Arsenite; Fe(II): ferrous iron. As concentration in the system: 220 mg/L.
(hydr)oxides for arsenite compared to ferric (hydr)oxides (Masue-Slowey et al., 2011). When two-step coprecipitation process was applied as in the case of the present study, arsenate reduction was inhibited due to secondary deposition of aluminum (hydr)oxide. Although Al-(hydr)oxide adsorbs less arsenite compared to arsenate, the overall effect of secondary deposition of Al-(hydr)oxide on Fe/As ¼ 2 coprecipitate leads to suppressed arsenic remobilization.
4.
Conclusions
A novel two-step coprecipitation process for arsenic removal and immobilization from aqueous solutions was developed. This process involved the first step to remove arsenate via Fe/As ¼ 2 coprecipitation followed by the second step to improve the stability via aluminum hydroxide precipitation. Arsenate can be effectively removed to sub-ppm level. The generated arsenic-bearing solid showed considerably higher stability in terms of arsenic release under both oxic and anoxic conditions compared to conventional FeeAs coprecipitation method. The findings are important for the removal and immobilization of arsenic from hydrometallurgical solutions in the form of high stability arsenicbearing solid.
Acknowledgments The authors thank Natural Science Foundation of China for the financial support (40925011).
references
Bottero, J.Y., Bersillon, J.L., 1989. Aluminum and iron (III) chemistry. In: Suffet, I.H., MacCarthy, P. (Eds.), Aquatic Humic Substances: Influence on Fate and Treatment of Pollutants. American Chemical Society, Washington, DC, pp. 425e442. Chen, N., Jiang, D.T., Cutler, J., Kotzer, T., Jia, Y.F., Demopoulos, G.P., Rowson, J.W., 2009. Structural characterization of poorly-crystalline scorodite, iron(III)arsenate co-precipitates and uranium mill neutralized raffinate solids using X-ray absorption fine structure spectroscopy. Geochimica et Cosmochimica Acta 73 (11), 3260e3276. Coker, V.S., Gault, A.G., Pearce, C.I., Van Der Laan, G., Telling, N.D., Charnock, J.M., Polya, D.A., Lloyd, J.R., 2006. XAS and XMCD evidence for species-dependent partitioning of arsenic during microbial reduction of ferrihydrite to magnetite. Environmental Science & Technology 40 (24), 7745e7750. Campbell, K.M., Malasarn, D., Saltikov, C.W., Newman, D.K., Hering, J.G., 2006. Simultaneous microbial reduction of iron(III) and arsenic (V) in suspensions of hydrous ferric oxide. Environ. Sci. Technol 40, 5950e5955. Cornell, R.M., Schwertmann, U. (Eds.), 1996. The Iron Oxides: Structure, Properties, Reactions, Occurrence and Uses. VCH Publishers, New York, p. 194. Das, S., Hendry, M.J., Essilfie-Dughan, J., 2011. Effects of adsorbed arsenate on the rate of transformation of 2-line ferrihydrite at pH 10. Environmental Science & Technology 45 (13), 5557e5563. Demopoulos, G.P., 2009. Aqueous precipitation and crystallization for the production of particulate solids with desired properties. Hydrometallurgy 96 (3), 199e214.
508
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 0 e5 0 8
Ford, R.G., 2002. Rates of hydrous ferric oxide crystallization and the influence on coprecipitated arsenate. Environmental Science & Technology 36 (11), 2459e2463. Gella, V., 2007. Precipitation if aluminum (oxy)hydroxides from concentrated chloride solutions by neutralization. M. Eng. thesis, McGill University. Montreal, Canada Goldberg, S., Johnston, C.T., 2001. Mechanisms of arsenic adsorption on amorphous oxides evaluated using macroscopic measurements, vibrational spectroscopy, and surface complexation modeling. Journal of Colloid and Interface Science 234 (1), 204e216. Harris, B., 2003. The removal of arsenic from process solutions: theory and industrial practice. In: Young, C., Alfantazi, A., Anderson, C., James, A., Dreisinger, D., Harris, B. (Eds.), Hydrometallurgy 2003-Proceedings of the International Symposium Honoring Professor Ian M. Ritchie, vol. 2. TMS, Warrendale, PA, pp. 1889e1902. Harris, G.B., 2000. The removal and stabilization of arsenic from aqueous process solutions: past, present and future. In: Young, C.A. (Ed.), Minor Elements 2000. SEM, Littleton, CO, pp. 3e20. Harris, G.B., Monette, S., 1988. The stability of arsenic-bearing residues. In: Reddy, R.G., Hendrix, J.L., Queneau, P.B. (Eds.), Arsenic Metallurgy: Fundamentals and Applications. TMS, Warrendale, PA, pp. 469e498. Islam, F.S., Pederick, R.L., Gault, A.G., Adams, L.K., Polya, D.A., Charnock, J.M., Lloyd, J.R., 2005. Interactions between the Fe(III)-reducing bacterium Geobacter Sulfurreducens and arsenate, and capture of the metalloid by biogenic Fe(II). Applied and Environmental Microbiology 71 (12), 8642e8648. Jia, Y.F., Demopoulos, G.P., Chen, N., Cutler, J.N., Jiang, D.T., 2003. Preparation, characterization and solubilities of adsorbed and co-precipitated iron(III)-arsenate solids. In: Young, C., Alfantazi, A., Anderson, C., James, A., Dreisinger, D., Harris, B. (Eds.), Hydrometallurgy 2003-Proceedings of the International Symposium Honoring Professor Ian M. Ritchie, vol. 2. TMS, Warrendale, PA, pp. 1923e1935. Jia, Y., Demopoulos, G.P., 2005. Adsorption of arsenate onto ferrihydrite from aqueous solutions: influence of media (sulfate vs nitrate), added gypsum, and pH alteration. Environmental Science & Technology 39 (24), 9523e9527. Jia, Y., Xu, L., Fang, Z., Demopoulos, G.P., 2006. Observation of surface precipitation of arsenate on ferrihydrite. Environmental Science & Technology 40 (10), 3248e3253. Jia, Y.F., Demopoulos, G.P., 2008. Coprecipitation of arsenate with iron(III) in aqueous sulfate media: effect of time, lime as base and co-ions on arsenic retention. Water Research 42 (3), 661e668. Krause, E., Ettel, V.A., 1989. Solubilities and stabilities of ferric arsenate compounds. Hydrometallurgy 22 (3), 311e337. Lagno, F., Rocha, S.D.F., Chryssoulis, S., Demopoulos, G.P., 2010. Scorodite encapsulation by controlled deposition of aluminum phosphate coatings. Journal of Hazardous Materials 181 (1e3), 526e534. Langmuir, D., Mahoney, J., MacDonald, A., Rowson, J., 1999. Predicting arsenic concentrations in the porewaters of buried uranium mill tailings. Geochimica et Cosmochimica Acta 63 (19e20), 3379e3394. Leetmaa, K., 2008. Scorodite stabilization with aluminum hydroxyl gels. Masteral dissertation, University of McGill at Montreal. Lindsay, W.L., Walthall, P.M., 1996. The solubility of aluminum in soils. In: Sposito, G. (Ed.), The Environmental Chemistry of Aluminum. Lewis Publishers, New York, p. 338. Masue, Y., Loeppert, R.H., Kramer, T.A., 2007. Arsenate and arsenite adsorption and desorption behavior on
coprecipitated aluminum: iron hydroxides. Environmental Science & Technology 41 (3), 837e842. Masue-Slowey, Y., Loeppert, R.H., Fendorf, S., 2011. Alteration of ferrihydrite reductive dissolution and transformation by adsorbed As and structural Al: implications for As retention. Geochimica et Cosmochimica Acta 75 (3), 870e886. McCreadie, H., Blowes, D.W., Ptacek, C.J., Jambor, J.L., 2000. Influence of reduction reactions and solid-phase composition on porewater concentrations of arsenic. Environmental Science & Technology 34 (15), 3159e3166. Moldovan, B.J., Jiang, D.T., Hendry, M.J., 2003. Mineralogical characterization of arsenic in uranium mine tailings precipitated from iron-rich hydrometallurgical solutions. Environmental Science & Technology 37 (5), 873e879. Nishimura, T., Umetsu, Y., 2000. Chemistry on elimination of arsenic, antimony and selenium from aqueous solutions with iron(III) species. In: Young, C.A. (Ed.), Minor Metals 2000. SME, Littleton, CO, pp. 105e112. Pedersen, H.D., Postma, D., Jakobsen, R., 2006. Release of arsenic associated with the reduction and transformation of iron oxides. Geochimica et Cosmochimica Acta 70 (16), 4116e4129. Postma, D., Jessen, S., Hue, N.T.M., Duc, M.T., Koch, C.B., Viet, P.H., Nhan, P.Q., Larsen, F., 2010. Mobilization of arsenic and iron from Red River floodplain sediments, Vietnam. Geochimica et Cosmochimica Acta 74 (12), 3367e3381. Raven, K.P., Jain, A., Loeppert, R.H., 1998. Arsenite and arsenate adsorption on ferrihydrite: kinetics, equilibrium, and adsorption envelopes. Environmental Science & Technology 32 (3), 344e349. Riveros, P.A., Dutrizac, J.E., Spencer, P., 2001. Arsenic disposal practices in the metallurgical industry. Canadian Metallurgical Quarterly 40 (4), 395e420. Robins, R.G., Singh, P., Das, R.P., 2005. Coprecipitation of arsenic with Fe(III), Al(III) and mixtures of both in a chloride system. In: Reddy, R.G., Ramachandran, V. (Eds.), Arsenic Metallurgy. TMS, Warrendale, PA, pp. 113e128. Robins, R.G., Huang, J.C.Y., Nishimura, T., Khoe, G.H., 1988. The adsorption of arsenate ion by ferric hydroxide. In: Reddy, R.G., Hendrix, J.L., Queneau, P.B. (Eds.), Arsenic Metallurgy: Fundamentals and Applications. TMS, Warrendale, PA, pp. 99e112. Rowson, J., Private Communication. AREVA Resources Inc., Saskatoon, Sask., Canada. Schwertmann, U., Cornell, R.M., 1991. Iron Oxides in the Laboratory Preparation and Characterization, second ed. Wiley-VCH, Weinheim, New York. Twidwell, L.G., Robins, R.G., Hohn, J.W., 2005. The Removal of arsenic from aqueous solution by coprecipitation with iron (III). In: Reddy, R.G., Ramachandran, V. (Eds.), Arsenic Metallurgy. TMS, Warrendale, PA, pp. 3e24. Violante, A., Pigna, M., Del Gaudio, S., Cozzolino, V., Banerjee, D., 2009. Coprecipitation of arsenate with metal oxides. 3. nature, mineralogy, and reactivity of iron(III)-aluminum precipitates. Environmental Science & Technology 43 (5), 1515e1521. Violante, A., Ricciardella, M., Del Gaudio, S., Pigna, M., 2006. Coprecipitation of arsenate with metal oxides: nature, mineralogy, and reactivity of aluminum precipitates. Environmental Science & Technology 40 (16), 4961e4967. Wilkie, J.A., Hering, J.G., 1996. Adsorption of arsenic onto hydrous ferric oxide: effects of adsorbate/adsorbent ratios and cooccurring solutes. Colloids and Surfaces A: Physicochemical and Engineering 107, 97e110. Zhang, X., Jia, Y., Wang, X., Xu, L., 2008. Phylogenetic analysis and arsenate reduction effect of the arsenic-reducing bacteria enriched from contaminated soils at an abandoned smelter site. Journal of Environmental Sciences 20 (12), 1501e1507.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 9 e5 2 0
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Characteristic analysis on temporal evolution of floc size and structure in low-shear flow Weipeng He, Jun Nan*, Haoyu Li, Shengnan Li Skate Key Laboratory of Urban Water Resource and Environment, School of Municipal and Environmental Engineering, Harbin Institute of Technology, Harbin 150090, PR China
article info
abstract
Article history:
A series of flocculation tests were performed to investigate the effect of low-shear rates
Received 2 July 2011
(G ¼ 3e16 s1) on flocculation of kaolin suspension by polyaluminum chloride (PACl), with
Received in revised form
the goal of understanding floc growth mechanisms. Results were reported in terms of floc
9 November 2011
average size (dp) and boundary fractal dimension (Dpf), derived from a non-intrusive optical
Accepted 11 November 2011
sampling and digital image analysis technique. As expected, the rate of floc aggregation
Available online 23 November 2011
increased with increasing G, resulting in faster changes in aggregate size and structure in
Keywords:
for dp at the same shear rates, possibly due to the self-similarity of fractal aggregates. An
Flocculation
interesting finding was that at G ¼ 3 s1, an obvious plateau was observed for the average-
Shear
size evolution at steady state; for shear rates of 6 and 7 s1, the flocs exhibited some
Floc breakage
decrease after reaching the peak of size, mainly as a result of floc settling at steady state;
Fractal dimension
and for G ¼ 11e16 s1, a decrease in floc size was possibly attributed to the irreversibility of
Image analysis
PACl-floc breakage. The process of floc growth was described using a fractal growth model,
the initial stage of flocculation. Nevertheless, steady state was attained faster for Dpf than
which defined flocculation as the result of the combined processes of aggregation and restructuring. The conceptual model could effectively characterize temporal changes in floc size and structure, and found that fragmentation followed by reformation seemed to be more effective in forming larger and more compact aggregates than the restructuring process due to erosion and reformation, which may provide useful insights for the design of flocculation reactors. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The removal of particles from a suspension by sedimentation following flocculation is widely applied in water and wastewater treatment. Chemical coagulant/flocculant addition brings about a change in the nature of small suspended particles, rendering them unstable, whereas flocculation encourages these destabilized particles to contact with each other, leading to the formation of highly porous and
irregularly shaped aggregates, known as flocs (Jarvis et al., 2005). The size and structure of flocs are considered fundamental to the efficient operation of water treatment. If properly controlled, flocculation can be used to produce flocs with desired size distribution, structure and settling rate (Soos et al., 2008). There are several phases of floc growth during flocculation. At the beginning, the process is governed by aggregation. During this initial time period, particles aggregate rapidly due
* Corresponding author. 1518 Room, School of Municipal and Environmental Engineering, The Second Campus of Harbin Institute of Technology, 73 Huanghe Road, Nangang District, Harbin, PR China. Tel.: þ86 451 86084169; fax: þ86 451 86283001. E-mail addresses:
[email protected] (W. He),
[email protected] (J. Nan),
[email protected] (H. Li), lishengnan_830309@ 163.com (S. Li). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.040
510
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 9 e5 2 0
to a high rate of inter-particle collisions. As flocs grow and become larger, further growth is restricted by fluid shear because breakage occurs (Yukselen and Gregory, 2004), and low particle collision rates of larger aggregates with other ones (Serra et al., 2008). After a certain time, aggregation and breakage balance each other, resulting in a steady-state floc size distribution (Spicer et al., 1996). The evolution of floc size during flocculation is accomplished by the change of floc structure, which can be made more compact by breakage or restructuring (Wang et al., 2011). Thill et al. (2001) proposed that aggregate restructuring is an important mechanism in explaining flocculation kinetics. The breakage of flocs has been classified in two modes, i.e., surface erosion, where single primary particles or their small aggregates are eroded off the parent floc, and large-scale fragmentation, where a floc is broken into pieces with similar sizes (Jarvis et al., 2005; Vassileva et al., 2007). Different modes will produce different size distributions after breakage, thus providing an effective method to determine whether erosion or fragmentation has occurred (Becker et al., 2009). Another main difference between the two modes is the energy input, which is low for erosion and high for fragmentation (Vassileva et al., 2007). Yeung and Pelton (1996) used micromechanical techniques to pull flocs apart, and found that breakage occurs at the weakest spot inside the aggregate. According to their results, compact flocs will break as a result of surface erosion, whereas loose ones tend to undergo largescale fragmentation. Factors affecting flocculation process include coagulant type and dosage, particle concentration, solution pH, mixing intensity and duration (Chakraborti et al., 2003; Colomer et al., 2005). An open issue refers to the effect of shear rate history on steady-state properties of flocs (Soos et al., 2008). When a precipitated/polymeric flocculant was used, it was found that floc size and structure strongly depended upon how the shear rate (G) was varied before reaching the steady state (Spicer and Pratsinis, 1996a). The behavior for shear rates above 20 s1 has been investigated in several previous studies. Biggs and Lant (2000), Selomulya et al. (2001), and Soos et al. (2008), have suggested that the larger the shear the smaller the average/median aggregate size under steady-state conditions. On the contrary, the range G < 20 s1 has received relatively little attention except in the work of Colomer et al. (2005), showing that aggregate size increases with increasing shear rates ranging from 0.7 to 27.4 s1. A recent study was conducted by Serra et al. (2008) to examine particle flocculation in three different devices (paddle mixer, oscillating grid and Couette) over a wide range of shear rates (G ¼ 4e102 s1). The results showed that each device has a shear-rate range within which there is a transition from aggregation dominated conditions, where little breakage takes place and further aggregation may be limited by low particle collision rates of larger aggregates with other ones (Brakalov, 1987), to breakage dominated conditions, where the probability of floc breakage increases rapidly and becomes important to further growth. Therefore, it is essential that much attention be paid to understanding the behavior of flocculation under low-shear rates (here, G < 20 s1). In the present study, a series of flocculation tests were carried out to investigate the effect of low-shear rates
(G ¼ 3e16 s1) on the temporal evolution of floc size and structure during flocculation in a stirred tank. The properties of flocs were characterized by average size and fractal dimension, derived from a non-intrusive optical sampling and digital image analysis technique. Of particular interest was how an increase in the shear rate affected flocculation behavior in low-shear flow. This topic has important implications for increasing understanding of floc aggregation and breakage mechanisms.
2.
Material and methods
2.1.
Suspension
Kaolin clay (Tianjin, China) suspension was used as testing water sample. The stock suspension was prepared similarly to that of Yukselen and Gregory (2004). The particles had an average size of about 3 mm, determined by a particle counter (2200PCX, HACH, USA). For flocculation tests, the stock solution was diluted in the tap water of Harbin, China, giving a final clay concentration of 100 mg/L. Harbin tap water has medium total hardness (ca. 160 mg/L as CaCO3) and alkalinity (ca. 115 mg/L as CaCO3) and a pH of around 7.8. To avoid the disturbance of divalent metal ions such as Ca2þ and Mg2þ in tap water, a small amount of humic acid (Shanghai, China) was added into the testing sample (Yukselen and Gregory, 2004). The final suspension containing 100 mg/L kaolin and 2 mg/L humic acid had a turbidity of about 96 NTU, determined by an online turbidimeter (700AQ, WTW, Germany).
2.2.
Coagulant
Polyaluminum chloride (PACl) was selected as the coagulant. Stock PACl solutions of 1% w/w were prepared by dissolving the reagent in deionized water (5 g dissolved in deionized water to 500 mL). To prevent aging effects, a fresh stock solution was prepared for a sequence of experiments or renewed every two weeks. The solution was kept in a refrigerator at 5 C and, for the flocculation tests, directly pipetted in the testing water without further dilution.
2.3.
Apparatus
A modified version of the standard flocculation-test procedure was used in this work (Fig. 1a). From the moment of coagulant addition, a non-intrusive optical sampling technique was used to obtain digital images of particles (Fig. 2), which were then analyzed to characterize floc size and structural properties, and calculations of the fractal dimension. The basic procedures were adapted from the work of Chakraborti et al. (2000), who studied aggregate characteristics produced after mixing suspensions with different coagulant doses. A major advantage of their method is that it requires no sample handling, so there is no concern for disturbing the floc characteristics during measurements. In addition, Chakraborti et al. (2003) also used this in-situ technique to evaluate temporal changes in the fractal dimension of aggregates formed during flocculation. A similar measurement technique
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 9 e5 2 0
511
Fig. 1 e Schematic diagrams of (a) the in-situ recognition technique and (b) photograph of the R1342-type impeller (IKA, Germany).
was used by Xiao et al. (2011) to investigate the flocculation dynamics for different coagulants in different model waters. In the present work, the in-situ recognition system (Fig. 1a) consisted of an automated stroboscopic lamp to illuminate suspended particles in the tank, a high-speed digital charged coupled device (CCD) video camera (SVS-VISTEK GmbH, Germany) to capture particle images and a process control and image processing software package (FMans 10, China) to determine floc geometrical parameters. The camera was placed on the opposite side of the tank from the lamp, so that backlit shadows of particles were produced. Here, the CCD camera has a sensor matrix consisting of 992(horizontal) 510(vertical) pixels, giving an interrogation window of about 5665 mm 2920 mm. This means the monitoring system had a resolution of around 5.7 mm for particle imaging in the present flocculation study. A personal computer served to control the camera and provided storage for particle images. The flocculating reactor used in this study was a rectangular stirred tank (homemade) with a bottom length D ¼ 280 mm and a liquid height H ¼ 230 mm, and filled with 18 L of testing water sample (see Section 2.1) as working fluid. For agitation, a R1342-type impeller (IKA, Germany) with a diameter d ¼ 50 mm (Fig. 1b) was used and the center of the impeller was positioned at C ¼ H/3 from the tank bottom (Fig. 1a). The impeller has a power number of Np ¼ 1.27 and a dimensionless pumping capacity of Nq ¼ 0.79. This mixing system was successfully used in some of our previous studies, e.g., in Nan et al. (2009).
2.4.
Procedure
Optimal PACl dosage was determined by carrying out a series of flocculation tests with incremental increases in the
coagulant dose (between 1 and 10 mg/L as Al). For each test, after a certain amount of PACl coagulant dosed in the reactor, the testing suspension (100 mg/L kaolin and 2 mg/L humic acid in Harbin tap water) was then mixed rapidly at 400 rpm for 30 s, followed by a slow stirring phase at 100 rpm for 20 min. After a 20-min sedimentation (without mixing), the turbidity of supernatant (at 5 cm below the liquid surface) was measured by an online turbidimeter (700AQ, WTW, Germany). The amount of coagulant giving the minimum turbidity is the optimal dosage (Duan and Gregory, 1998), which was then used for all other tests in this study. For dynamic tests, after allowing 1 min for steady-state turbidity (measured by online turbidimeter) to be established, PACl solution was dosed and the suspension was mixed at 400 rpm for 30 s. Then the stirring speed (N ) was reduced to values ranging from 50 to 140 rpm, producing shear rates of G ¼ 3e16 s1 (Table 1), for 39 min. The G value range was calculated in a similar way as described in the study of Spicer et al. (1996), where the effect of different impellers were examined. These G values were quite close to the range used by Colomer et al. (2005), who investigated particle aggregation and breakage in low-shear flow by monitoring the resulting aggregate size distribution. The Reynolds number for the impeller is Re ¼ Nd2/n (see Table 1 for values), where n is the kinematic viscosity of the water (here, n ¼ 1.146 106 m2 s1), producing values of Re larger than 1000, corresponding to turbulent mixing. Images were taken to examine aggregate geometry and size distributions at different moments (here, every second) during each test. The images were then analyzed to track changes in aggregate morphology for a given experiment, as well as differences between experiments resulting from varying mixing speed (N ). The analyses were reported in
512
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 9 e5 2 0
Fig. 2 e Examples of floc images captured in situ during a typical flocculation process.
terms of the fractal dimension and associated floc average size (see Section 2.5). Before processing a particular image, each image was treated using contrast enhancement in order to produce the clearest possible particle images prior to analysis. This step effectively filtered out random noise, as well as particles that were not well focused. More than 10 consecutive images within a minute were analyzed to produce floc morphological parameters. The total number of aggregates captured in each minute decreased from around 1000 to 150 during flocculation for each shear. This reflected the fact that small particles had aggregated into floc. All the tests were conducted at room temperature (22 1 C) and replicated 2e3 times.
2.5.
surface morphology of the aggregate in two-dimensional projection, its values ranges from Dpf ¼ 1 for the projected area of a sphere (a circle), to Dpf ¼ 2 for a line (e.g. a chain of particles) (Mandelbrot et al., 1984). As observed by several authors (Li and Ganczarczyk, 1989; Spicer et al., 1996; Xiao et al., 2011), typical values of Dpf at steady state for aggregates produced under turbulent conditions are in the range of 1.1e1.4. Moreover, the mass fractal dimension, Df, is also a measure of floc structure and varies from 1 for a line of particles to 3 for a sphere, but there is no straightforward relationship between Df and Dpf (Logan and Kilps, 1995). In this study, flocs in more than 10 images within a minute were used to calculate the value of Dpf at the corresponding moment.
Floc size and fractal dimension by image analysis
The size, dp, of a floc of irregular shape, can be calculated in terms of the equivalent diameter by dp ¼ ð4A=pÞ1=2
(1)
where A is the projected area of the floc. Then the average size and the size distribution of flocs in more than 10 images within a minute were used to reflect floc size at the corresponding moment. Flocs generated during flocculation have been shown to be fractal, implying that they are self-similar and scale invariant (Johnson et al., 1996). In addition to floc size, a boundary fractal dimension was used to characterize the fractal properties of flocs. For a 2D projected particle image, the fractal dimension, Dpf, defines how the projected area of the particles scales up with the length of the perimeter (Meakin, 1987; Spicer and Pratsinis, 1996b) according to AfP2=Dpf
(2)
where P is the perimeter of an aggregate. Since Dpf depicts the
3.
Results and discussion
3.1.
Coagulant optimization
Initial flocculation tests were conducted to determine the optimal coagulant dosage, as described earlier. Turbidity removal rate reached a peak value of about 90% at the PACl dosage of 2.2 mg/L as Al, which was selected as the optimal dosage and then used for all other flocculation tests in this study.
3.2.
Floc size and structure
3.2.1.
Effect of shear rate on the evolution of floc size
The in-situ recognition technique (Fig. 1a) for floc morphology is shown to be a powerful tool for performing real-time, in-situ particle imaging acquisition to determine floc size distribution during flocculation, expressed as temporal changes in floc average size at different shear rates (Fig. 3). The G values shown in Table 1 were all investigated, but in order to
pffiffiffiffiffiffiffiffi Table 1 e Hydrodynamic values during flocculation. The fourth line shows the Kolmogorov microscale ðh[ n=GÞ for each shear. For the sake of simplicity, integer rounded values of G are used throughout the text. N (rpm) G (s1) Re h (mm)
50 3.3 (3) 1818 586
60 4.4 (4) 2182 511
70 5.5 (6) 2545 455
80 6.8 (7) 2909 412
90 8.1 (8) 3272 377
100 9.4 (9) 3636 348
110 10.9 (11) 3999 324
120 12.4 (12) 4363 304
130 14.0 (14) 4727 286
140 15.6 (16) 5090 271
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 9 e5 2 0
513
Fig. 3 e Variation in floc average size with flocculation time under different shear rates (G). For each shear rate, the overall process was divided into two phases: transitional phase (TP) and steady state (SS).
illustrate a significant aspect of the results, only six typical shear rates were given here. The average-size evolution, derived from the in-situ recognition technique, clearly exhibited continuous floc growth during flocculation (Fig. 3), i.e., for each shear rate, floc average size rapidly increased at the beginning, and then further increase in the size was restricted, resulting in a steady-state floc size, usually assumed to represent a dynamic balance between shear-induced aggregation and breakage. This was in accordance with previous findings (Spicer and Pratsinis, 1996a; Soos et al., 2008). To clearly illustrate the size evolution, the overall process was divided into two phases, i.e., transitional phase (TP), where floc size increased to the first maximum value, and steady state (SS), where floc size did not significantly increase and fluctuated in a certain range, either narrow or broad. At each phase, the rate of floc growth (Rfloc) may be considered as the difference between the rate of aggregation and the rate of floc breakage (Rbr) (Jarvis et al., 2005): Rfloc ¼ aRcol Rbr
studies (Spicer et al., 1996; Yukselen and Gregory, 2004; Serra et al., 2008), the initial process of flocculation is mainly governed by floc aggregation; in this case Eq. (3) can be rewritten as Rfloc z aRcol. As the collision frequency is proportional to the shear rate (Jackson, 1995; Mietta et al., 2009), the rate of floc aggregation increased with G in the transitional phase. As expected, the size distribution of flocs at the end of TP was shifted toward larger sizes with increasing shear rates (Fig. 4), i.e., when the applied shear increased from 3 to 7 s1, the number fraction of flocs smaller than 50 mm, formed at the end of TP for each shear, decreased from around 70e40%, and for shear rates of G ¼ 11e16 s1, this fraction for each shear reduced to the value below 30%. Fig. 3 also showed that the evolution of floc average size at steady state (SS) for each shear rate was significantly
(3)
where Rcol is the rate of particle collision, a is the fraction of collisions which result in attachment (collision efficiency), and aRcol is used to represent the rate of aggregation. Obviously, when the two terms on the right-hand side of Eq. (3) are equal, the net rate of floc growth is zero and the floc size attains a limiting value. This is of great importance in understanding the evolution of floc size under low-shear conditions. The results from the curves shown in Fig. 3 indicated that in the transitional phase (TP), the higher the shear rate (G) the faster the growth in aggregate size, due to the increase of floc aggregation rate with increasing energy dissipation (Selomulya et al., 2001; Colomer et al., 2005). As stated in many
Fig. 4 e Size distributions of flocs formed at the end of the transitional phase (TP-end) under different shear rates (G).
514
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 9 e5 2 0
different. At G ¼ 3 s1, the steady-state floc size did not change greatly, and an obvious plateau was observed at steady state. For shear rates of 6 and 7 s1, the flocs reached a larger steady state size, which was expected as higher shear rates would promote more collisions. At G ¼ 11 s1, the flocs evolved to the largest steady-state size, and when the applied shear increased to 12 s1 or above, the steady-state size started to decrease. An interesting observation from Fig. 3 was that the flocs grown at 6 s1 or above (notably at 6, 11 and 16 s1) exhibited some decrease after reaching the peak of size. For example, at G ¼ 6 s1, the flocs seemed to reach their maximum size of 113 mm at 18 min of flocculation before sinking to their final equilibrium size of about 70 mm in the last 10 min of flocculation. A similar phenomenon of a decrease to a steady-state size was observed by other workers, such as Selomulya et al. (2001), and Wang et al. (2011). Furthermore, Serra et al. (2008) found that the size of the aggregates formed under steady state conditions was always maximized at shear rates of G ¼ 20e30 s1. However, the optimal shear rate (around 11 s1) in our study appeared to be lower than that proposed by Serra et al. (2008), possibly due to different flocculation frequencies produced by different mixing systems. These flocculation behaviors above could be due to the nature of flow generated by the employed impeller. As discussed in previous work (Coufort et al., 2005; Serra et al., 2008), the flow was uniform within the stirred tank, i.e., in the impeller zone the turbulent kinetic energy was highly dissipated, whereas outside this zone the turbulent kinetic energy was found to be smaller. An estimate of the frequency of exposure of the flocs to the high-shear impeller zone could be obtained from the circulation time (tc ¼ V/(NqNd3), where V is the volume of water, and Nq is the dimensionless impeller pumping capacity: here, Nq ¼ 0.79). When the applied shear increased, the tc became shorter (Fig. 5), indicating that the flocs were exposed to the high-shear impeller zone more frequently (Spicer et al., 1996), thus increasing the probability of floc breakage. At a really low-shear rate (e.g., 3 s1), it seemed that little breakage occurred at steady state, possibly due to a relatively low frequency of circulation through the high-shear impeller zone. This finding was supported by the evolution of floc size distributions during flocculation at
Fig. 5 e Circulation time (tc) as a function of the shear rate (G).
G ¼ 3 s1 (see Section 3.2.3). According to Eq. (3), further growth was mainly limited by low particle collision rates of larger aggregates with other ones (Serra et al., 2008). Mietta et al. (2009) stated that aggregation depends linearly on G, whereas breakage depends on G3/2, implying that there is a critical shear-rate value where breakage is expected to dominate over aggregation. With increasing shear rates, more sufficient collisions were produced to make larger particles, resulting in an increase in the rate of aggregation. However, floc breakage could increase as well due to more frequent visits of the particles in the high-shear impeller zone. As described in Eq. (3), the difference between the rates of floc aggregation and breakage determined the rate of floc growth. Fig. 3 indicated that when the applied shear increased from 6 to 11 s1, the rate of floc growth increased, reflected by an increase in the steady-state size; whereas at G ¼ 12 s1 or above, the relatively high shear could produce sufficient collisions to make larger particles, but the probability of floc breakage would greatly increase due to shorter circulation time (Fig. 5), leading to a decrease in the rate of floc growth, reflected by a reduction in the equilibrium size.
3.2.2.
Effect of shear rate on the evolution of floc structure
The high-quality floc images (Fig. 2) captured by the in-situ recognition technique also allow detailed analysis of the structural features of flocs formed under various shear rates. According to Eq. (2), the boundary fractal dimension (Dpf) of flocs could be approximated from the slope of logelog regression of a series of projected areas versus perimeters of the particles. The Dpf at steady state ranged from 1.13 to 1.24 (Fig. 6). This was similar with the range reported for the kaolin-HA flocs after 30 min of slow flocculation by Xiao et al. (2011), but somewhat lower than the values (from 1.1 to 1.4) reported for the aggregates of polystyrene spheres (Li and Ganczarczyk, 1989). All the curves shown in Fig. 6 appeared to be of similar form. In general, a higher Dpf normally indicates a more irregular and/or elongated shape and a rougher surface for the particles, whereas a lower Dpf suggests a more spherical shape and a smoother surface of the particles (Li and Ganczarczyk, 1989; Spicer and Pratsinis, 1996a). Similarly, the overall process of the fractal-dimension evolution was divided into two phases, transitional phase (TP) and steady state (SS). In the transitional phase for each shear, the Dpf increased to a significant extent (Fig. 6), indicating that micro-flocs, with a regular shape and smooth surface, were formed rapidly as aggregation dominated. With the growth of flocs, clusterecluster aggregation gradually began to play an important role in the development of flocs, resulting in more fractal aggregates with a more elongated shape and rougher surface. However, the production of more irregular structures would be impeded by floc breakage. Therefore, the Dpf reached a steady state. A similar phenomenon was observed by Stone and Krishnappan (2003), who proposed that small particle clusters (micro-flocs) were the formational units of larger flocs during flocculation and the stability of larger flocs was a function of the shear stress at steady state. Furthermore, similar to the evolution of floc size shown in Fig. 3, the change in aggregate structure was faster with increasing shear rates in the beginning of flocculation (Fig. 6).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 9 e5 2 0
515
Fig. 6 e Variation in the boundary fractal dimension of flocs with flocculation time under different shear rates (G). For each shear rate, the overall process was divided into two phases: transitional phase (TP) and steady state (SS).
Also, the structure of flocs at steady state appeared to be sensitive to the shear rate (Fig. 6). The loosest flocs, with the largest Dpf value of around 1.23, were formed at G ¼ 3 s1, indicating less restructured, loose aggregates. For the flocs formed at shear rates of 12 and 16 s1, the Dpf was slightly lower than that found for the flocs formed at 6 and 7 s1. These suggested that more compact structures were obtained after floc breakage in low-shear flow. This phenomenon is consistent with a number of other studies (Jarvis et al., 2005; Wang et al., 2011). As described previously, the higher the shear rate, the more frequent exposure of the flocs to the impeller zone. The increased mixing and circulation at higher shear rates further led to increased likelihood of floc breakage by shear (Spicer et al., 1996). As expected, the flocs would break up at their weakest points and rearrange into more stable structures. After breakage, the elongated flocs broke up into smaller pieces that were closer to spherical objects than the original flocs (Xiao et al., 2011). In addition, during the process of floc breakage and re-growth, the surface of large flocs might be destroyed, uncovering some inner pores hidden in these aggregates (Wang et al., 2011), and thus small particles or clusters had more opportunity to be incorporated into these larger flocs (see Section 3.2.3). This behavior would not only make floc structure more compact but also reduce the number of small particles.
3.2.3.
to 7 s1, the dp continued to increase, reaching the first maximum value of 97 20 mm, but the Dpf began to increase from the first minimum value of 1.17 0.01 (attained at G ¼ 6 s1) to 1.18 0.01. It could be seen from Fig. 8a that the size distributions of flocs changed slightly at steady state for G ¼ 3 s1, which supported the finding described in Section 3.2.1. At G ¼ 7 s1, the size distributions of flocs grown at steady state was shifted toward larger sizes from the end of the transitional phase (i.e., the beginning of SS) to 31 min of flocculation (Fig. 8b), indicating that floc aggregation dominated over breakage at steady state. Nevertheless, a significant decrease in the number of flocs larger than 200 mm was observed at the end of SS, possibly due to the settling of larger
Floc size and fractal dimension at steady state
Most previous studies have correlated floc size and structure with the shear rate, but the range that has typically been investigated was G > 20 s1 (Colomer et al., 2005). An objective of this work was to evaluate the effect of low-shear rates on floc size and structure at steady state. The floc average size (dp) increased and the boundary fractal dimension (Dpf) decreased with G up to 6 s1 (Fig. 7), indicating that flocs became larger and more compact. When the applied shear further increased
Fig. 7 e Floc average size and boundary fractal dimension at steady state under different shear rates (G). Error bars represent the fluctuation degree, characterized by the standard deviation, of the two parameters at steady state for each shear rate.
516
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 9 e5 2 0
Fig. 8 e Floc size distributions at four typical flocculation moments for six representative shear rates (G). The TP-end and SSend represent the points at the end of the transitional phase (TP) and the end of the steady state (SS), respectively.
flocs with open and porous structure at the end of flocculation (Johnson et al., 1996). As the G further increased from 7 to 9 s1, a decrease in dp and an increase in Dpf seemed to be observed (Fig. 7), but this trend did not occur in the study of Serra et al. (2008). One possible explanation for these slight changes is that some single particles or their small aggregates may be removed from the surface of loose flocs formed early in the process of flocculation, resulting in an increase in the small particle size ranges. This fact could be easily seen in Fig. 8c and d. These two sub-figures also demonstrated that at G ¼ 8 s1, erosionlike breakage mainly occurred among flocs within the intermediate size range (e.g., 100e200 mm), but at G ¼ 9 s1, the number fraction of flocs larger than 400 mm began to decrease after the end of SS, indicating that some larger aggregates
seemed to break up in the fragmentation-like way. Furthermore, it is likely that aggregates of intermediate size undergo restructuring, starting with an initial stretching phase (Becker et al., 2009), thus producing more elongated flocs with higher values of Dpf. When the G increased from 9 to 11 s1, the dp sharply increased from 81 7 to 153 24 mm, and the Dpf decreased from 1.21 0.01 to 1.17 0.01 (Fig. 7). This was possibly attributed to the occurrence of large-scale fragmentation, producing floc pieces with similar sizes (Fig. 8e), which were re-available to other small but compact clusters or particles suspended in flocculation system. If the formed pieces were filled up by those small particles, the floc size distribution would be shifted toward larger sizes. Also, as discussed above, more compact structure could be obtained. For the higher
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 9 e5 2 0
Fig. 9 e Flocculation time (ts) required to reach steady state for floc size and fractal dimension under different shear rates (G). The time lag between the steady states of the floc size and the fractal dimension is also given here. shear rates (G ¼ 12e16 s1), the dp began to decrease again but the Dpf slightly varied around 1.17 (Fig. 7). As shown in Fig. 8f, the number fraction of flocs larger than 150 mm decreased more sharply from the end of TP to the end of SS at G ¼ 16 s1 than that at G ¼ 11 s1, indicating that more significant largescale fragmentation occurred at steady state. As a result, a lower final equilibrium size was obtained after 26 min of flocculation (Fig. 3), possibly due to the irreversibility of PAClfloc breakage (Yukselen and Gregory, 2004).
517
Serra et al. (2008) suggested that there was a shear rate resulting in a transition from aggregation dominated conditions to breakage dominated conditions. Based on the results shown in Fig. 7, when breakage became significant, there was also a shear rate resulting in a transition from erosion dominated conditions to fragmentation dominated conditions. Interestingly, at this transition G (11 s1), the dp reached the second maximum value, and the Dpf was 1.16 0.01, close to the second minimum value attained at G ¼ 12 s1 (Fig. 7), indicating the formation of largest flocs with relatively compact structure. This suggested that large-scale fragmentation seemed to be a more effective mechanism in forming larger flocs with compact structure than surface erosion. The flocculation time (ts) required to reach steady state appeared to be different for dp and Dpf for each shear in lowshear flow (Fig. 9), i.e., steady state was attained faster for Dpf than for dp at the same shear rates, possibly due to the selfsimilarity of fractal aggregates (Logan and Kilps, 1995; Chakraborti et al., 2003). As expected, an increase in G caused a reduction in ts for both dp and Dpf.
3.3.
Floc growth mechanisms in low-shear flow
Floc strength is a particularly important operational parameter in solid/liquid separation processes for the efficient removal of aggregated particles. It is generally accepted that floc strength is directly related to floc structure, and increased floc compaction may increase floc strength due to an increase in the number of bonds holding the aggregate together (Yukselen and Gregory, 2004; Jarvis et al., 2005). In other words, floc strength is highly dependent on the floc formation process. In
Table 2 e Schematic of the structural evolution models of aggregates in shear flows (taken from Becker et al., 2009).
518
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 9 e5 2 0
Fig. 10 e Schematic of floc growth during low-shear flocculation. The evolutional process of floc structure was proposed on the basis of the work performed by Spicer et al. (1996).
order to improve the understanding of aggregate structure and its growth process, a number of fractal aggregation models, including the diffusion-limited aggregation (DLA) model and its several variations (reaction-limited, clustereparticle, clusterecluster, etc.), have been established (Meakin, 1987; Bushell et al., 2002; Kockar et al., 2010). These models usually consider aggregation as an irreversible process neglecting breakage and restructuring. This is a critical limitation, when compared with naturally occurring processes, where a restructuring of primary particles occurs within an aggregate due to breakage and reformation (Chakraborti et al., 2003). Becker et al. (2009) investigated the restructuring behavior of colloidal aggregates in shear flows, and suggested that small aggregates, represented by Aggregate I (being composed of 55 primary particles), rotate like a rigid body in the shear flow, whereas large aggregates, represented by Aggregate III (being composed of 1000 primary particles), break up. They also demonstrated that aggregates of intermediate size, represented by Aggregate II (being composed of 305 primary particles), undergo restructuring, starting with an initial stretching phase, followed by an aggregate compaction. Restructuring of the aggregates with different sizes was illustrated in Table 2. As discussed above, the aggregation, breakage and restructuring processes govern the development of floc size and structure (Fig. 10). Initially, destabilized particles are transferred into contact with each other by the applied shear, producing small clusters in the DLA limit. Further aggregation mainly occurs among those small clusters resembling the clusterecluster aggregation (CCA) model, producing larger flocs with more highly branched structure. It is easily understood that these aggregates are more susceptible to breakage by the applied shear, followed by reformation/restructuring. After a certain time, breakage and reformation can lead to stronger and more compact flocs (Fig. 10), with an associated lower boundary fractal dimension (Dpf). Such results have been reported by Spicer et al. (1996), Stone and Krishnappan (2003) and Xiao et al. (2011), for example. Flocculation behavior in low-shear flow is presented in Figs. 3, 6 and 7. At shear rates of G ¼ 3e7 s1, floc aggregation was dominated over breakage, and flocs aggregated in the
DLA/CCA limit, where the growth of flocs mainly depended on the collision frequency resulting from the effective shear rate. In this case floc size increased in proportion to the shear rate, but further growth was limited by low particle collision rates of larger particles with other particles. At G ¼ 3 s1, aggregation rate was minimized due to the lowest particle collision, producing the smallest flocs with loosest structure. It was possible that these flocs might rotate like Aggregate I (Table 2) without any structural change in low-shear flow. Therefore, at steady state no significant fluctuation was observed for floc size and structure (Figs. 3 and 6). As expected, flocs formed at higher shear rates (e.g., 4e7 s1) were larger due to increased particle collisions. In accordance with Becker et al. (2009), these flocs with intermediate size appeared to undergo a restructuring process resembling Aggregate II (Table 2), starting with an initial stretching phase. This behavior may be another reason for the production of larger flocs with more compact structure. When the shear increased to 8 s1 or above, floc breakage seemed to be significant (Figs. 3, 6 and 7). Thomas et al. (1999) suggested that when flocs are smaller than the Kolmogorov pffiffiffiffiffiffiffiffi microscale (h ¼ n=G, see Table 1 for values) they become prone to breakage by surface erosion, whereas above the microscale flocs are thought to be more exposed to breakage by fragmentation. As shown in Fig. 7, there was a decrease in dp and an increase in Dpf at steady state for shear rates of 8 and 9 s1. This phenomenon indicated that the CCA aggregates (Fig. 10) seemed to break up by erosion, resulting in an increase in particles of small size range (Fig. 8c and d). It also demonstrated that the removal of single particles from the parent aggregate surface made the aggregate more irregularly shaped. Obviously, the higher the shear, the more susceptible the flocs are to breakage (Wang et al., 2011). At G ¼ 11e16 s1, the CCA aggregates (Fig. 8) began to behave like Aggregate III (Table 2). Interestingly, this restructuring process seemed to be more effective in forming larger and more compact aggregates than the restructuring process due to erosion and reformation, as mentioned above. This finding may be helpful for the design of flocculation reactors to obtain the desired floc structure.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 9 e5 2 0
4.
Conclusions
The main conclusions of this study were listed as follows: 1. Floc size and structure were monitored using an in-situ recognition technique, and found to continuously evolve in low-shear conditions. The higher the shear rate, the faster the growth in aggregate size at the beginning of flocculation, due to increased rate of floc aggregation with increasing energy dissipation. 2. A decrease in floc size was observed after reaching a peak size under the examined shear rates. For shear rates of G ¼ 3e7 s1, flocs formed with open and porous structure, likely permitted great quantities of flow through them to attain large settling velocities, thus making some porous flocs at steady state settle out of suspension. At G ¼ 11e16 s1, where floc breakage became significant, the decreased final equilibrium size was possibly attributed to the irreversibility of PACl-floc breakage. 3. Floc structure appeared to evolve to a steady state according to the rate of shear, and more compact structures were obtained after floc breakage in low-shear flow due to the process of floc breakage and re-growth. 4. Flocculation time that was required to reach steady state for dp and Dpf decreased with increasing G, but steady state was attained faster for Dpf than for dp at the same shear rates, possibly due to the self-similarity of fractal aggregates. 5. The development of floc size and structure in low-shear flow was described using a fractal growth model. According to this model, it was found that fragmentation followed by reformation seemed to be more effective in forming larger and more compact aggregates than the restructuring process due to erosion and reformation. This finding may provide useful insights for the design of flocculation reactors in order to obtain the desired floc structure.
Acknowledgments This study has been made possible through the funding from the National Major Project of Water Pollution Control and Manage of Eleventh Five Years of China (2009ZX07424-005-01). Weipeng He’s research is also supported by New Academic Researcher Awards for Doctoral Candidate from Ministry of Education (MOE), PRC. Comments and suggestions from anonymous reviewers are greatly acknowledged.
references
Becker, V., Schlauch, E., Behr, M., Briesen, H., 2009. Restructuring of colloidal aggregates in shear flows and limitations of the free-draining approximation. Journal of Colloid and Interface Science 339 (2), 362e372. Biggs, C.A., Lant, P.A., 2000. Activated sludge flocculation: on-line determination of floc size and the effect of shear. Water Research 34 (9), 2542e2550.
519
Brakalov, L.B., 1987. A connection between the orthokinetic coagulation capture efficiency of aggregates and their maximum size. Chemical Engineering Science 42 (10), 2373e2383. Bushell, G.C., Yan, Y.D., Woodfield, D., Raper, J., Amal, R., 2002. On techniques for the measurement of the mass fractal dimension of aggregates. Advances in Colloid and Interface Science 95 (1), 1e50. Chakraborti, R.K., Atkinson, J.F., Benschoten, J.E.V., 2000. Characterization of alum floc by image analysis. Environmental Science & Technology 34 (18), 3969e3976. Chakraborti, R.K., Gardner, K.H., Atkinson, J.F., Benschoten, J.E.V., 2003. Changes in fractal dimension during aggregation. Water Research 37 (4), 873e883. Colomer, J., Peters, F., Marrase, C., 2005. Experimental analysis of coagulation of particles under low-shear flow. Water Research 39 (13), 2994e3000. Coufort, C., Bouyer, D., Line´, A., 2005. Flocculation related to local hydrodynamics in a Taylor-Couette reactor and in a jar. Chemical Engineering Science 60 (8e9), 2179e2192. Duan, J., Gregory, J., 1998. The influence of silicic acid on aluminium hydroxide precipitation and flocculation by aluminium salts. Journal of Inorganic Biochemistry 69 (3), 193e201. Jackson, G.A., 1995. Comparing observed changes in particle size spectra with those predicted using coagulation theory. Deep Sea Research 42 (1), 159e184. Jarvis, P., Jefferson, B., Gregory, J., Parsons, S., 2005. A review of floc strength and breakage. Water Research 39 (14), 3121e3137. Johnson, C.P., Li, X., Logan, B.E., 1996. Settling velocities of fractal aggregates. Environmental Science & Technology 30 (6), 1911e1918. Kockar, H., Bayirli, M., Alper, M., 2010. A new example of the diffusion-limited aggregation: NieCu film patterns. Applied Surface Science 256 (9), 2995e2999. Li, D.H., Ganczarczyk, J., 1989. Fractal geometry of particle aggregates generated in water and wastewater treatment processes. Environmental Science & Technology 23 (11), 1385e1389. Logan, B.E., Kilps, J.R., 1995. Fractal dimensions of aggregates formed in different fluid mechanical environments. Water Research 29 (2), 443e453. Mandelbrot, B.B., Passoja, D.E., Paullay, A.J., 1984. Fractal character of fracture surfaces of metals. Nature 308, 721e722. Meakin, P., 1987. Fractal aggregates. Advances in Colloid and Interface Science 28, 249e331. Mietta, F., Chassagne, C., Winterwerp, J.C., 2009. Shear-induced flocculation of a suspension of kaolinite as function of pH and salt concentration. Journal of Colloid and Interface Science 336 (1), 134e141. Nan, J., He, W., Song, X., Li, G., 2009. Impact of dynamic distribution of floc particles on flocculation effect. Journal of Environmental Sciences 21 (8), 1059e1065. Selomulya, C., Amal, R., Bushell, G., Waite, T.D., 2001. Evidence of shear rate dependence on restructuring and breakup of latex aggregates. Journal of Colloid and Interface Science 236 (1), 67e77. Serra, T., Colomer, J., Logan, B.E., 2008. Efficiency of different shear devices on flocculation. Water Research 42 (4e5), 1113e1121. Soos, M., Moussa, A.S., Ehrl, L., Sefcik, J., Wu, H., Morbidelli, M., 2008. Effect of shear rate on aggregate size and morphology investigated under turbulent conditions in stirred tank. Journal of Colloid and Interface Science 319 (2), 577e589. Spicer, P.T., Pratsinis, S.E., 1996a. Coagulation and fragmentation: universal steady-state particle-size distribution. AIChE Journal 42 (6), 1612e1620.
520
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 0 9 e5 2 0
Spicer, P.T., Pratsinis, S.E., 1996b. Shear-induced flocculation: the evolution of floc structure and the shape of the size distribution at steady state. Water Research 30 (5), 1049e1056. Spicer, P.T., Keller, W., Pratsinis, S.E., 1996. The effect of impeller type on floc size and structure during shear-induced flocculation. Journal of Colloid and Interface Science 184 (1), 112e122. Stone, M., Krishnappan, B.G., 2003. Floc morphology and size distributions of cohesive sediment in steady-state flow. Water Research 37 (11), 2739e2747. Thill, A., Moustier, S., Aziz, J., Wiesner, M.R., Bottero, J.Y., 2001. Flocs restructuring during aggregation: experimental evidence and numerical simulation. Journal of Colloid and Interface Science 243 (1), 171e182. Thomas, D.N., Judd, S.J., Fawcett, N., 1999. Flocculation modelling: a review. Water Research 33 (7), 1579e1592.
Vassileva, N.D., Ende, D., Mugele, F., Mellema, J., 2007. Fragmentation and erosion of two-dimensional aggregates in shear flow. Langmuir 23 (5), 2352e2361. Wang, D., Wu, R., Jiang, Y., Chow, C.W.K., 2011. Characterization of floc structure and strength: role of changing shear rates under various coagulation mechanisms. Colloids and Surfaces A: Physicochemical and Engineering Aspects 379 (1e3), 36e42. Xiao, F., Lam, K.M., Li, X.Y., Zhong, R.S., Zhang, X.H., 2011. PIV characterisation of flocculation dynamics and floc structure in water treatment. Colloids and Surfaces A: Physicochemical and Engineering Aspects 379 (1e3), 27e35. Yeung, A.K.C., Pelton, R., 1996. Micromechanics: a new approach to studying the strength and breakup of flocs. Journal of Colloid and Interface Science 184 (2), 579e585. Yukselen, M., Gregory, J., 2004. The reversibility of floc breakage. International Journal of Mineral Processing 73 (2e4), 251e259.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 2 1 e5 3 0
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A compartmental model to describe hydraulics in a full-scale waste stabilization pond Andres Alvarado a,b,*, Sreepriya Vedantam b, Peter Goethals c, Ingmar Nopens b a
DIUC-Direccion de Investigacion, Universidad de Cuenca, Av. 12 de Abril s/n, Cuenca, Ecuador BIOMATH, Department of Mathematical Modelling, Statistics and Bioinformatics, Ghent University, Coupure Links 653, B-9000 Gent, Belgium c Department Applied Ecology and Environmental Biology, Ghent University, J. Plateaustraat 22, B-9000 Gent, Belgium b
article info
abstract
Article history:
The advancement of experimental and computational resources has facilitated the use of
Received 14 May 2011
computational fluid dynamics (CFD) models as a predictive tool for mixing behaviour in
Received in revised form
full-scale waste stabilization pond systems. However, in view of combining hydraulic
11 November 2011
behaviour with a biokinetic process model, the computational load is still too high for
Accepted 12 November 2011
practical use. This contribution presents a method that uses a validated CFD model with
Available online 19 November 2011
tracer experiments as a platform for the development of a simpler compartmental model (CM) to describe the hydraulics in a full-scale maturation pond (7 ha) of a waste stabili-
Keywords:
zation ponds complex in Cuenca (Ecuador). 3D CFD models were validated with experi-
Waste stabilization ponds (WSP)
mental data from pulse tracer experiments, showing a sufficient agreement. Based on the
Tracer study
CFD model results, a number of compartments were selected considering the turbulence
Computation fluid dynamics (CFD)
characteristics of the flow, the residence time distribution (RTD) curves and the dominant
Compartmental model
velocity component at different pond locations. The arrangement of compartments based
Hydrodynamics
on the introduction of recirculation flow rate between adjacent compartments, which in
Mixing
turn is dependent on the turbulence diffusion coefficient, is illustrated. Simulated RTD’s
Modelling
from a systemic tanks-in-series (TIS) model and the developed CM were compared. The TIS was unable to capture the measured RTD, whereas the CM predicted convincingly the peaks and lags of the tracer experiment using only a minimal fraction of the computational demand of the CFD model. Finally, a biokinetic model was coupled to both approaches demonstrating the impact an insufficient hydraulic model can have on the outcome of a modelling exercise. TIS and CM showed drastic differences in the output loads implying that the CM approach is to be used when modelling the biological performance of the fullscale system. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Waste Stabilization Ponds (WSPs) are known for economical treatment of wastewater in places where vast stretches of land are available. However, there is still room for optimizing
these systems from a performance perspective in terms of effluent quality and energy consumption. In order to allow optimization, a better process understanding is needed, which can be achieved by using mathematical models. In large systems with sufficiently high residence time, the hydraulic
* Corresponding author. DIUC-Direccion de Investigacion, Universidad de Cuenca, Av. 12 de Abril s/n, Cuenca, Ecuador. E-mail addresses:
[email protected],
[email protected] (A. Alvarado),
[email protected] (S. Vedantam),
[email protected] (P. Goethals),
[email protected] (I. Nopens). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.038
522
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 2 1 e5 3 0
behaviour becomes less important compared to systems like Ucubamba WSP, the system studied here, with small residence time and where short-circuiting is clearly observed. Hence, the hydrodynamics of these systems should be modelled thoroughly in conjunction with biochemical process models. In order to approximate mixing behaviour in reactors (some point between plug flow and completely mixed flow) without using the advection-dispersion partial differential equation, two classical systemic approaches can be used, i.e. the dispersion model or the tanks-in-series (TIS) model. The use of dispersion model in systems with large backmixing is questionable (Levenspiel, 1999) and, moreover, it does not account for stagnant or dead zones (Polprasert and Bhattarai, 1985). The TIS model has the advantage of being simple and, hence, not very computationally demanding and has been widely applied when modelling e.g. activated sludge wastewater treatment systems. However, this simplicity is also its major limitation. TIS models describe fluid flow in only one direction and can only account for some backmixing by maintaining the liquid longer in the system, adjusting the backmixing rate (Le Moullec et al., 2010). Recirculation fluxes, which in some systems (like the WSPs under study here) become very important, cannot be properly represented with TIS. The major issue with systemic models is that, when combined with a biokinetic model, the flaws in the mixing model will be “calibrated” by using degrees of freedom of the biokinetic model which is bad modelling practice. Indeed, the latter will reduce the predictive power of the model and renders it useless in any subsequent optimization study. Since the introduction of the potential of Computational Fluid Dynamics (CFD) models to predict the flow patterns in ponds by Wood et al. (1995), the interest in CFD modelling for WSP’s has grown particularly for practical applications. In this sense, there are a number of studies describing the usefulness of CFD models as a tool for design improvement and hydraulic assessment (Karteris et al., 2005; Peterson et al., 2000; Shilton, 2000; Shilton and Mara, 2005; Sweeney et al., 2005; Vega et al., 2003; Wood et al., 1998). Nevertheless, Shilton et al. (2008) shed light upon the limited scientific contributions existing with regard to the proper validation of CFD models against experimental data in full-scale systems. It was concluded that the computational resources are the main constraint, especially if the ultimate objective of the modelling is the incorporation of biochemical reactions. An intermediate solution between TIS and full-fledged CFD are so-called compartmental models (CM). A CM consists of a number of compartments Ci of volume Vi in more than one dimension and in which a recirculation flow Qri from compartment Ciþ1 to Ci occurs, along with the forward flow. Since the first successful efforts reported to create a CM based on CFD turbulence analysis for mixed chemical reactors by Alexopoulos et al. (2002) and Alex et al. (2002), several contributions at pilot-scale have been reported, including approaches for mixing reactors by Rigopoulos and Jones (2003), Bezzo et al. (2004), Kougoulos et al. (2005), Guha et al. (2006), Vakili and Esfahany (2009); and also for continuous flow reactors by Gresch et al. (2009) and Le Moullec et al. (2010). In these previous efforts, the aggregation criteria for compartmental arrangement still appears to be dependent on
the physical interpretation of every system. However, a better agreement exists in relation to the estimation of recirculation flow rate between adjacent compartments from the turbulent diffusion coefficient. CMs were even proposed for activated sludge systems but were, at that time, subject to overparameterisation (De Clercq et al., 1999). However, the latter can be solved by using a validated CFD model to determine these parameters. In this study, a complete approach is detailed for a full-scale WSP system. The aim of this paper is threefold: i) analyze the flow characteristics in a WSP using a tracer study, ii) develop and validate a CFD model with data obtained from the tracer study, and iii) develop a methodology that uses the CFD model predictions to build a compartmental model. In a broader perspective, this compartmental model highlights the influence of strong recirculation patterns typically observed in pond systems, on the overall flow behaviour and the impact on system performance when adding a biokinetic model.
2.
Methods
2.1.
The waste stabilization pond system
The Ucubamba WSP system, the biggest wastewater facility in Ecuador, treats the domestic effluent of the Andean city of Cuenca e Ecuador (400,000 inhabitants), located at an altitude of 2560 m a.s.l. The WSP is in operation since 1999 by the Municipal Company ETAPA. The system is divided in two identical flow lines (Fig. 1); both lines comprise of an aerated lagoon (using mechanical floating aerators), a facultative lagoon and a maturation pond. The average discharge influent to the system is 1.2 m3 s1. The total surface of the WSP is 45 ha, and the hydraulic retention time is 12 days. The maturation pond, which was used for the development of the methodology in this work, has an area of 7 ha and a depth of 1.7 m. The inlet/outlet of the pond consists of a submerged pipe of 0.9 m diameter and an overflow structure of 10 m length, respectively. The bathymetries executed showed only a slight sludge layer growth in the maturation ponds (less than 1% over the total pond volume and 10 cm max. elevation in few locations) which was assumed negligible with respect to the pond hydraulics.
2.2.
Tracer study
A tracer study was conducted in Maturation pond 1 to monitor the residence time distribution (RTD). This data (Espinoza and Rengel, 2009) served as the validation data for the CFD model developed for the pond. 1.65 L (20% pure substance) of the dye tracer Rhodamine WT, considered non-biodegradable and non absorptive to solids (Yanez, 1993), was injected in the inlet and measured at the outlet (stimulus-response technique) using a fluorometer (Aquafluor, Turner Designs) which has a minimum detection limit of 0.4 ppb. The samples were collected using an automatic sampler. The sampling frequency was once per 15 min before the first peak appeared and was increased up to once per 60 min after that until the tracer recovery exceeded 90%, which was observed after 30 days. The samples were analyzed at a fixed temperature of
523
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 2 1 e5 3 0
Fig. 1 e Layout of the Ucubamba waste stabilization pond (WSP).
19 C and the influence of the algal biomass present in the pond was investigated by measuring the background fluorescence concentration during a couple of hours before starting the test to avoid unrepresentative hydraulic characteristic results as suggested by Valero and Mara (2009).
2.3.
CFD modelling
Due to the relative positions of the inlet and outlet (submerged pipe lying at bottom and overflow structure respectively) a 2D CFD model, which could save considerable computational time, was not possible for this pond system. Therefore, three dimensional CFD simulations were performed for maturation pond 1 (Fig. 1) by using the commercial software FLUENT 6.3 (ANSYS, USA). A finite volume method was used dividing the computational domain into 75,000 quadratic grid elements. In order to account for the flow disturbances due to the small inlet and outlet regions as compared to the full pond, a denser grid was used in these regions as well as the pond corners, where dead zones could likely develop. Increasing the grid density in the remainder of the pond did not affect the predicted flow field. The latter was based on grid dependency simulations. Hence, in order to have a feasible computational expense, this grid size of 75,000 elements was maintained for the simulations of flow as well as determination of the RTD. The three-dimensional mass and momentum equations in a Cartesian system are given by: vr þ V$ðr! vÞ ¼ 0 vt v ! g ðr v Þ þ V$ðr! v ! v Þ ¼ Vp þ V$ðsÞ þ r! vt T v þ V! v 2=3V$! v I wherein,s ¼ m½V!
(1)
(2)
and V$! v ¼
vvx vvy vvz þ þ vx vy vz
(3)
These could be considered realistic; assuming the concentration of suspended solids in the pond (typically of the order of 30 mg/l) does not significantly affect the fluid properties. The turbulence in the pond was modelled using the k-ε model. This is a two-equation model to account for transported variables representing the kinetic energy in turbulence and the scale of turbulence. This model has been shown previously in literature to be useful for free-shear layer flows with relatively small pressure gradients. Similar to the case of wall-bounded flows or internal flows, this model holds good for cases where the mean pressure gradients are not too adverse. The model equations are as follows: ru0i u0j vuj m vk mþ t r˛ sk vxj vxi
(5)
m vε ε ε2 mþ t c1ε rui u0j c2ε r sε vxj k k
(6)
v v v ðrkui Þ ¼ ðrkÞ þ vt vxi vxj v v v ðrεui Þ ¼ ðrεÞ þ vt vxi vxj
The influence of wind and temperature gradients was neglected. Moreover, in order to initiate the strategy for building the compartmental model, it was thought advisable to limit the computational expense. As will be shown in the results section, the flow fields obtained at various vertical heights yielded nearly the same pattern, being of the same magnitude. Hence, it was decided to neglect the vertical diffusion which also simplifies the compartmental model. For the simulation of the tracer test, it was assumed that the tracer species had identical properties as the fluid in the primary phase, i.e. water. The species transport model was used for the determination of the RTD at the outlet of the pond. The latter is obtained by plotting the tracer concentration versus time at the pond outlet (Vedantam et al., 2006). The tracer is injected at time t ¼ 0 for one time step giving a species mass fraction of 1 and then is reset to 0 for the second and following time steps.
The scalar transport equation for the transport of the Rhodamine dye is given by:
2.4.
vr4k v v4 rvi 4k Dk k ¼ 0 þ vt vxi vxi
The systemic tanks-in-series model (TIS) was built as the simplest approach to model a dispersion flow with the capability to be coupled with any kinetic model (Levenspiel, 1999). The number of compartments (n) was evaluated according to some statistical properties of the RTD obtained from the CFD model:
(4)
The fluid in the pond was assumed to be incompressible and exhibiting Newtonian fluid behaviour with a density of 998.2 kg m3 and dynamic viscosity of 1.003 E-3 kg m1 s1.
Tanks-in-series modelling (TIS)
524
s2 2
t
¼
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 2 1 e5 3 0
1 n
(7)
where, t represents mean residence time [d], and s2 the variance of the RTD [d2]. From the knowledge of the process, the TIS model was arranged with backmixing flow between adjacent tanks and recirculation flow from the last to the first tank, which allows adjusting the model with the back flow rate (Levenspiel, 1999).
2.5. model
Modelling of TIS and CM integrated with biokinetic
The hydraulic performance of the systemic model and the CM was assessed obtaining the RTD with a pulse of inert solution. For this purpose, the modelling and simulation software WEST (Mike by DHI, Denmark) was used. Moreover, these models were also assessed coupling the biokinetic model ASM1 (Henze et al., 2000) to compare the behaviour of some effluent species in the models under a real wastewater influent. For this purpose, the benchmarking influent composition of BSM1 (Copp, 2002) was used in conjunction with the flow rate measured at the WSP Ucubamba.
3.
Results and discussion
3.1.
CFD model simulation
For every time instance, the necessary inlet turbulent boundary conditions were calculated. The transient velocity profile that was used for the tracer simulations is shown in Fig. 3a. From the unsteady CFD simulation, it was observed that the transient velocity profile did not impact the overall velocity and turbulence profile of the pond (Fig. 3b), which could be expected given the influent flow rate and the size of the pond system. Fig. 3b is an important observation as it means that the compartmental model can be developed based on a single flow field and can subsequently be used for varying inflow conditions (see Section 3.2). In fact, the compartmental model that is derived in the next section is valid under different conditions of diurnal influent flow and no transient behaviour needs to be addressed in the CM procedure. This can be generalised for large systems where the whole flow pattern is not significantly influenced by the relative small changes in the inlet discharges. One should however be careful when the characteristics change (e.g. extreme peak behaviour). Moreover, the flow field in smaller systems might be more prone to changes in influent flow rate and in that case different compartmental models might be required. They can, however, be built following the same methodology illustrated in this paper.
3.2.
First, an unsteady simulation using a constant velocity inlet boundary was run until a stable solution for velocity and turbulence profiles was obtained in the whole pond domain. Fig. 2 depicts the velocity contours at a horizontal plane (0.1 m from the surface) of the pond (since the vertical diffusion is neglected) along with velocity vectors after 3 days of simulation time of an unsteady CFD simulation with a constant influent profile. The pattern described in the CFD concurs with the field observations in the tracer test. In Fig. 2, the highest velocities in the pond (in the range 0.2 < v < 1.12) are located in the uncoloured zone at the inlet. Subsequently, a transient inlet velocity profile, based on the inlet conditions of plant online measurements, was provided as inlet boundary condition and simulated for the transient tracer simulation.
Compartmental model development
The procedure of building the compartmental structure started with an unsteady simulation using a constant velocity inlet boundary until a stable velocity and turbulence profile solution were obtained in the whole pond domain (Fig. 2). Ideally, a compartmental model comprises of a certain number of fully mixed volumes, which can be interconnected by means of certain exchange fluxes. The extension over TIS is that there is more freedom to define compartments and it is not restricted to one single dimension. With regard to the CFD flow predictions, defining compartments in one single dimension do not account for recirculation patterns. The challenge now lies in determining the exchange fluxes, which depends upon the flow variations and the number of tanks that could be identified to have a very close approximation of the mixing behaviour of the
Fig. 2 e Contours of flow velocities magnitude and vectors in maturation pond 1 obtained after 3 days of flow simulation using a constant influent profile.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 2 1 e5 3 0
525
Fig. 3 e (a)Transient velocity profile from system records (b) outputs of the velocity profile in the pond at two different times in Maturation 1 pond.
entire pond. Considering that this act is mainly in order to introduce the biokinetics at a later stage into the integrated compartmental model, the decision of the number of fully mixed volumes had to be based upon the major flow directions in the pond, hereby keeping the number of compartments as low as possible. The following steps were needed to set up the compartmental model: (1) determine different zones, (2) determine volumes of different zones, (3) determination of number of compartments per zone and (4) determination of convective and exchange fluxes in and between zones. They will be detailed in what follows.
3.2.1.
Determination of different zones
The compartmental structure definition started with a visual inspection of the velocity profile in the pond (Fig. 2). From the CFD study, it has been observed that the velocity components play a major role in the movement of the tracer. Hence, in order to divide the pond into compartments, the dominant flow direction was primarily used. Based on this, the entire pond domain was divided into three different zones based on expert interpretation of how the dominant flow impacts the RTD at the outflow of the pond. These were defined as follows: (1) zone 1 was defined based on the dominant spatial dimension of flow connecting the pond inlet to the pond outlet, leading to the first peak in the RTD curve; (2) near the pond outlet the liquid moved away from the dominant spatial direction, with lower velocity magnitudes (which was classified as zone 2); and (3) another zone (zone 3) existed, in which even lower magnitudes of velocity persisted, which lead to a recirculation zone.
3.2.2.
Determination of volume of different zones
The volumes of the different zones were determined as follows. The volume of zone 1 (V1) was calculated as the relation of the mean discharge by the mean residence time evaluated in the first section of the RTD curve Q=t1 from time 0e0.2 d (Fig. 4) which describes the first and major peak observed. Volumes in zone 2 and zone 3 were assigned by visual inspection of the velocity vectors and iso-contours plots. Since zone 2 is connected by means of a recirculation
flow to zone 1, the cross-sectional area of this zone must be connectable to the first tank in zone 1. The same reasoning was used for zone 3.
3.2.3.
Determination of number of compartments per zone
In zone 1, the number of compartments is obtained by making an analogy of the plug flow with the axial dispersion model (Fig. 4), allowing determination of the number of tanks as a function of the Peclet number (Pe). Thus, 2ðn 1Þ ¼ Pe ¼
uavg L D
(8)
where, n represents the number of equivolume fully mixed tanks which are connected in series, along the major flow dimension; uavg [m s1] is the average flow velocity along the zone evaluated from CFD results; L [m] is the characteristic length of the zone and D [m2 s1] is the mass diffusion coefficient. It is to note that all tanks had the same volume. The RTD was found to be sensitive to the division of zone 1 into compartments, which was to be expected as this formed a single direct connection from pond inlet to outlet. Using this approach, the optimal number of tanks in zone 1 was found to be 13. In zone 2, based on the flow direction, and since no recirculation was observed from the CFD simulation, one large completely mixed tank was opted for (subdividing in more compartments did not impact the solution). The rest of zone 2 was split into smaller tanks in regions where the flow direction changed. This was required in order to maintain an ease of calculation of distance between the tanks and determination of flow areas, as these were needed for calculating the exchange fluxes with zone 3 (see 3.2.4). In recirculation zone (zone 3), the compartments were chosen based on the visual observation of recirculation zones in the flow field analysis of CFD simulation. A similar reasoning as for zone 2 was followed, resulting in 2 large completely mixed tanks (flowing in opposite directions) and some smaller zones needed for easy determination of exchange fluxes with zone 2 compartments. In order to simplify the structure of the compartmental model, the dead zones were not considered as such, as they will not have a major impact on the RTD. Upon using this approach, velocity
526
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 2 1 e5 3 0
Fig. 4 e Comparison of the RTD curves obtained in the tracer experiment with CFD simulation.
gradients between the boundaries of the compartments would remain low, leading to a simpler approach, with reasonable accuracy. Since the flow is considered to be invariant along the depth based on the turbulent parameters, this division into compartments has been done on a horizontal slice and then extended with the same area for each tank along the depth of the pond. The final structure of the compartmental model is shown in Fig. 5.
3.2.4.
Determination of convective and exchange fluxes
Now the main structure of the CM was defined, the exchange fluxes between the different compartments still need to be determined. This was done based on the velocity vectors, velocity contours and understanding of the flow directions. In zone 1 the flow direction is selected based on the dominant spatial dimension, from the inlet to the outlet, since these ponds are continuous flow systems (Gresch et al., 2009). In zone 2 and zone 3 the visual inspection of the velocity vectors clearly determined the main flow direction. The exchange fluxes (indicated by the red arrows in Fig. 5) are determined based on the turbulence characteristics of the flow (k, ε) and the constant of the turbulence model Cm given by the diffusion coefficient Dt (Guha et al., 2006; Le Moullec et al., 2010). Thus, the exchange flux Qr can be calculated based on the cross-sectional area (A) and the distance between the compartments Δx as
Fig. 5 e Compartmental model layout of Maturation 1 pond.
Qr ¼
Dt A Dx
(9)
Once the backflows Qr are defined, the convective fluxes are automatically calculated by means of mass balance in the entire compartmental configuration, using the conservation of mass. Note that the criteria used to determine the volume and number of compartments (Sections 3.2.2 and 3.2.3) allowed an easy evaluation of Dx, i.e. the distance between the centres of the corresponding tanks where recirculation is occurring. In a way, this methodology confines the choice of number of compartments to the computational expense that can be handled, since the next step in modelling the WSP involves inclusion of a biokinetic model in each of the compartments. In order to establish the proof of concept, the compartmental model in this study was developed using 25 compartments (Fig. 5).
3.3.
Tracer experiment vs. CFD
The tracer experiment was performed during the coldest and driest months of the year to prevent the influence of the thermal stratification in the ponds and to minimize the influence of the rainfall in the discharge variability. The tracer was initially mixed with pond water to equalize the tracer temperature with the pond water and then added as a pulse into the channel just preceding the inlet pipe. Due to the huge size of the pond, the tracer movement through the water body was visible only for the first 30 min but it was sufficient to visually observe a circular pattern of the tracer around the pond. This movement depicts the first initial and major peak in the RTD curve (short circuiting), which is typical for this type of hydraulic systems (Shilton et al., 2008). Fig. 4 shows a reasonable agreement between the RTD curves from the tracer study and the CFD model prediction. The CFD model was able to capture the magnitude and timing of the first peak reasonably well, but the subsequent peaks are less clear in the experimental results (although there seems to be one between 0.25 d and 0.5 d). The disagreement could be due to the simplifying assumptions in the CFD model and e.g. the meteorological conditions not included in the model. This influence however, is not significant analyzing the whole
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 2 1 e5 3 0
527
Fig. 6 e Comparison of the RTD of the CFD model and experimental data against: a) TIS models with backmixing rate R [ 90%, varying the number of tanks n and without recirculation flow (n [ 25), b) TIS-R (n [ 25), varying the backmixing rate R, and c) TIS-BM.
behaviour of the tracer response plot. It is assumed that the CFD model is sufficiently valid for the pond.
3.4.
Tanks-in-series analysis
Studying the properties of the whole RTD curve of the CFD model, and according to Eq. (6), the number of tanks (n) in TIS model results in n < 1. The latter suggests a completely mixed behaviour in the pond, mainly due to the recirculation pattern. However, to assess the performance of the TIS model for pond hydrodynamics, initially, the TIS model was built with n ¼ 25 to match the number of tanks used in CM (Fig. 5). Considering the fact that a multitude of alternatives of TIS models could be
built, three different configurations were adopted and evaluated: 1) Tanks-in-series without any backmixing or recirculation flows; 2) TIS with recirculation flow from the last to the first tank (TIS-R); and 3) TIS with backmixing flow between all two adjacent tanks in the configuration (TIS-BM). Moreover, in the last 2 configurations, the backmixing rate was varied in order to get the best fit of the RTD. Fig. 6 depicts a series of curves of different TIS models, resulting in null agreement with CFD model output despite of the n tanks used. In Fig. 6(a) the number of tanks (n) is varied while maintaining the recirculation rate at R ¼ 0.9. It also contains a profile resulting from a configuration with n ¼ 25 without any backmixing or recirculation flow. It is observed that a large value of n (>100) could
Fig. 7 e Comparison of RTD curves obtained in CM versus TIS and tracer experiment.
528
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 2 1 e5 3 0
eventually match the first peak magnitude, albeit with less precision in time. However, such a large value of n severely impacts the computational requirements if a biokinetic model is to be coupled. Fig. 6(b) illustrates a series of curves of TIS, using n ¼ 25 and varying flow recirculation rates. It can be observed that despite the increase in flow recirculation, the magnitude of the main peak is never predicted accurately. Fig. 6(c) shows results of the TIS configuration incorporating adjacent backmixing (TIS-BM). Also here, no accurate RTD predictions were obtained. Hence, from a mixing behaviour point of view, none of the TIS model configurations was able to describe the observed system behaviour. However, for comparison purposes in the following sections, the configuration of TIS-R with n ¼ 25 and recirculation rate R ¼ 0.9 was adopted as the best TIS configuration.
3.5.
Compartmental model vs. tanks-in-series
Fig. 7 shows a comparison of the RTD curves obtained in TIS and CM against the experimental tracer data and CFD simulation. Table 1 summarises the quantification of the goodness of fit in terms of Sum of Squared Errors (SSE) and Relative Absolute Errors (RAE) of every hydraulic model tested. The different tested TIS models, as discussed before, are unable to capture the measured RTD, whereas the CM fits the RTD of both experimental data and CFD simulation reasonably well. In a WSP system exhibiting a strong recirculation pattern, the description of the main peaks is definitively the main concern in the RTD analysis and a subsequent coupling of a biokinetic model. In this study, the CM has positively achieved this goal. Although the CM behaviour is not perfect along the whole curve, it represents a drastic improvement over the systemic approach, still using a fairly limited amount of compartments. A further fine-tuning of the backmixing flows in CM could certainly improve the fit. However, any further effort in this way should consider first the limitations and consequently the accuracy of the experimental data when dealing with a full-scale system. It should also be stressed that the objective of the modelling exercise is important in this regard. The goal of developing the compartmental model here is to have
Table 1 e Sum of Squared Errors (SSE) and Relative Absolute Errors (RAE) between the hydraulic models and the experimental data. CFD CM TIS TIS-R TIS-BM Relative Absolute Error (RAE) Sum of Squares Error (SSE)
0.73 2.50
0.74 2.80
1.39 4.64
0.94 4.03
1.39 4.64
a realistic, yet simple model (minimal number of tanks) for the subsequent integration of a full-fledged biokinetic model. The latter will be the scope of a future paper. If the objective would be to have a very accurate description of mixing behaviour, one can relax on the constraint of total number of tanks.
3.6. Coupled biological and hydrodynamic model analysis In order to illustrate the impact of an inadequate mixing model, both the TIS and CM models were coupled with ASM1 and a simple dynamic simulation (14 d) was performed. It should be noted that a full analysis of the results was not the objective here. Results of effluent ammonium and autotrophic biomass concentration are given in Fig. 8 (using default ASM1 parameter values). It can be observed that both models exhibit significantly different system behaviour. The TIS model predicts washout of autotrophic organisms whereas the CM model retains the autotrophic biomass in the system. This results in no ammonium removal in the TIS approach. The example depicted, although not realistic in terms of the actual biological processes occurring in a WSP system, is very illustrative for the right selection of the mixing model when modelling biochemical processes. Often, a simple, and frequently applied, solution of the above problem used by many modellers is to increase the autotrophic growth rate. This would indeed avoid washout of nitrifiers. However, this practice introduces an error to cure for a flaw in model structure (Gujer, 2011). In general, to use the degrees of freedom that biokinetic parameters offer to solve for flaws in model structure is not good modelling practice. This will severely
Fig. 8 e Comparison of the outputs concentrations of autotrophic organisms (bottom) and ammonia (top) in TIS (n [ 25, flow recirculation R [ 90%) and CM model approaches coupled with ASM1 model.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 2 1 e5 3 0
deteriorate the prediction power of the model. The proposed CM approach is a better alternative as it addresses the pond hydraulics in a more accurate way, still resulting in a model that has a reasonable computational load and does not require modifying kinetic parameters. This was also illustrated before using experimental observations for an activated sludge system (Gresch et al., 2011; Le Moullec et al., 2011).
4.
Conclusions
The conclusions can be summarized as follows: A CFD model is built and demonstrates its robustness and accuracy to describe the hydrodynamics of a full-scale WSP but at a high computational cost. The further incorporation of external factors like wind and thermodynamic stratification can certainly improve the model but the computational demand will also increase considerably. The incorporation of biokinetic models would be in that case practically infeasible. The tracer experiments in large pond systems, although being costly in terms of time and resources, are very valuable in understanding the hydrodynamic behaviour of such systems. Considering the assumptions made in the CFD modelling, the validation of the model against tracer data is highly recommended. It is demonstrated that different TIS model configurations are not able to describe the pond’s mixing behaviour nor through changing the number of tanks, nor introducing recirculation flows, backmixing flows nor manipulating the backmixing/recirculation rate. A procedure is presented to derive a compartmental model (CM) based on a detailed flow analysis obtained from the validated CFD model for the WSP. The CM is able to reasonably predict the pond’s mixing behaviour at a very low computational demand, which facilitates the further inclusion of a biokinetic model. The impact of the mixing model on the behaviour of a biokinetic model is briefly illustrated by integrating ASM1 for different mixing models. For the same pond system, the TIS model predicted washout of autotrophic biomass and, hence, loss of nitrification, whereas the CM did not predict washout. This illustrates the importance of a mixing model that is representative for the system and not a crude oversimplification. Curing TIS performance by changing biokinetic parameters is not good modelling practice and only results in a fitted model with low predictive power.
Acknowledgements This work has been supported by VLIR-UOS Interuniversity Flemish Cooperation at Universidad de Cuenca, Ecuador. The authors would like to express their gratitude to ETAPA (Municipal Enterprise for Telecommunications, Water Supply and Sanitation of Cuenca), particularly to Galo Durazno, Director of Ucubamba WSP.
529
Nomenclature
c1ε, c2ε r p v m x, y, z k ε sk,ε Dt Qr s2 t
model constants (1.44, 1.92 respectively) [] density [kg m3] static pressure [Pa] velocity [m s1] molecular viscosity [kg m1 s1] directional components [m] kinetic energy [m2 s2] turbulence dissipation rate [m2 s3] turbulent Prandtl numbers for k and ε, respectively turbulence diffusion coefficient [m2 s1] exchange flux [m3 s1] variance of RTD [s2] time [s]
Abbreviations ASM Activated Sludge Model CFD Computational Fluid Dynamics CM Compartmental Model TIS Tanks-in-series RTD Residence Time Distribution WSP Waste Stabilization Pond
references
Alex, J., Kolisch, G., Krause, K., 2002. Model structure identification for wastewater treatment simulation based on computational fluid dynamics. Water Science and Technology 45 (4e5), 325e334. Alexopoulos, A.H., Maggioris, D., Kiparissides, C., 2002. CFD analysis of turbulence non-homogeneity in mixing vessels a two-compartment model. Chemical Engineering Science 57 (10), 1735e1752. Bezzo, F., Macchietto, S., Pantelides, C.C., 2004. A general methodology for hybrid multizonal/CFD models - part I. Theoretical framework. Computers & Chemical Engineering 28 (4), 501e511. Copp, J.B. (Ed.), 2002. The COST Simulation Benchmark. Description and Simulator Manual, Office for Official Publications of the European Communities, Luxembourg. De Clercq, B., Coen, F., Vanderhaegen, B., Vanrolleghem, P.A., 1999. Calibrating simple models for mixing and flow propagation in waste water treatment plants. Water Science and Technology 39 (4), 61e69. Espinoza, E., Rengel, P., 2009. Evaluacion Hidraulica de la Planta de Tratamiento de Aguas Residuales de Ucubamba. Universidad de Cuenca, Cuenca, Ecuador. Gresch, M., Brugger, R., Meyer, A., Gujer, W., 2009. Compartmental models for continuous flow reactors derived from CFD simulations. Environmental Science & Technology 43 (7), 2381e2387. Gresch, M., Braun, D., Gujer, W., 2011. Using reactive tracers to detect flow field anomalies in water treatment reactors. Water Research 45 (5), 1984e1994. Guha, D., Dudukovic, M.P., Ramachandran, P.A., Mehta, S., Alvare, J., 2006. CFD-based compartmental modeling of single phase stirred-tank reactors. AIChE Journal 52 (5), 1836e1846. Gujer, W., 2011. Is modeling of biological wastewater treatment a mature technology? Water Science and Technology 63 (8), 1739e1743.
530
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 2 1 e5 3 0
Henze, M., Gujer, W., Mino, T., van Loosdrecht, M.C.M., 2000. Activated Sludge Models ASM1, ASM2, ASM2d and ASM3. IWA, London, UK. Karteris, A., Papadopoulos, A., Balafoutas, G., 2005. Modeling the temperature pattern of a covered anaerobic pond with computational fluid dynamics. Water Air and Soil Pollution 162 (1e4), 107e125. Kougoulos, E., Jones, A.G., Wood-Kaczmar, M., 2005. CFD modelling of mixing and heat transfer in batch cooling crystallizers - aiding the development of a hybrid predictive compartmental model. Chemical Engineering Research & Design 83 (A1), 30e39. Le Moullec, Y., Gentric, C., Potier, O., Leclerc, J.P., 2010. Comparison of systemic, compartmental and CFD modelling approaches: application to the simulation of a biological reactor of wastewater treatment. Chemical Engineering Science 65 (1), 343e350. Le Moullec, Y., Potier, O., Gentric, C., Leclerc, J.P., 2011. Activated sludge pilot plant: comparison between experimental and predicted concentration profiles using three different modelling approaches. Water Research 45 (10), 3085e3097. Levenspiel, O., 1999. Chemical Reaction Engineering. Wiley & Sons, NJ, USA. Peterson, E.L., Harris, J.A., Wadhwa, L.C., 2000. CFD modelling pond dynamic processes. Aquacultural Engineering 23 (1e3), 61e93. Polprasert, C., Bhattarai, K.K., 1985. Dispersion model for waste stabilization ponds. Journal of Environmental EngineeringAsce 111 (1), 45e59. Rigopoulos, S., Jones, A., 2003. A hybrid CFD - reaction engineering framework for multiphase reactor modelling: basic concept and application to bubble column reactors. Chemical Engineering Science 58 (14), 3077e3089. Shilton, A., 2000. Potential application of computational fluid dynamics to pond design. Water Science and Technology 42 (10e11), 327e334.
Shilton, A., Mara, D.D., 2005. CFD (computational fluid dynamics) modelling of baffles for optimizing tropical waste stabilization Cn pond systems. Water Science and Technology 51 (12), 103e106. Shilton, A., Kreegher, S., Grigg, N., 2008. Comparison of computation fluid dynamics simulation against tracer data from a scale model and full-sized waste stabilization pond. Journal of Environmental Engineering-Asce 134 (10), 845e850. Sweeney, D.G., Nixon, J.B., Cromar, N.J., Fallowfield, H.J., 2005. Profiling and modelling of thermal changes in a large waste stabilisation pond. Water Science and Technology 51 (12), 163e172. Vakili, M.H., Esfahany, M.N., 2009. CFD analysis of turbulence in a baffled stirred tank, a three-compartment model. Chemical Engineering Science 64 (2), 351e362. Valero, M.A.C., Mara, D., 2009. The influence of algal biomass on tracer experiments in maturation ponds. Desalination and Water Treatment-Science and Engineering 4 (1e3), 89e92. Vedantam, S., Joshi, J.B., Koganti, S.B., 2006. CFD simulation of RTD and mixing in the annular region of a Taylor-Couette contactor. Industrial & Engineering Chemistry Research 45 (18), 6360e6367. Vega, G.P., Pena, M.R., Ramirez, C., Mara, D.D., 2003. Application of CFD modelling to study the hydrodynamics of various anaerobic pond configurations. Water Science and Technology 48 (2), 163e171. Wood, M.G., Greenfield, P.F., Howes, T., Johns, M.R., Keller, J., 1995. Computational fluid dynamic modeling of waste-water ponds to improve design. Water Science and Technology 31 (12), 111e118. Wood, M.G., Howes, T., Keller, J., Johns, M.R., 1998. Two dimensional computational fluid dynamic models for waste stabilisation ponds. Water Research 32 (3), 958e963. Yanez, F., 1993. Lagunas de Estabilizacion - Teoria, Disen˜o, Evaluacion y Mantenimiento. ETAPA, Cuenca, Ecuador.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 1 e5 3 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Occurrence and fate of organophosphorus flame retardants and plasticizers in coastal and marine surface waters Ulla E. Bollmann a,b,1, Axel Mo¨ller a,*, Zhiyong Xie a, Ralf Ebinghaus a, Ju¨rgen W. Einax b a
Helmholtz-Zentrum Geesthacht, Institute of Coastal Research, Department for Environmental Chemistry, Max-Planck-Straße 1, 21502 Geesthacht, Germany b Friedrich-Schiller-University of Jena, Institute for Inorganic and Analytical Chemistry, Lessingstraße 8, 07743 Jena, Germany
article info
abstract
Article history:
This comprehensive study focused on the spatial and seasonal variations of organophos-
Received 15 September 2011
phorus flame retardants and plasticizers (OPs) in surface water from the estuary of the
Received in revised form
River Elbe and the German Bight (North Sea). 100 surface water samples were extracted by
8 November 2011
solid phase extraction (SPE) and analyzed by gas chromatographyemass spectrometry
Accepted 9 November 2011
(GCeMS) with regard to 16 different OPs. The dominating substance was found to be tris(1-
Available online 19 November 2011
chloro-2-propyl) phosphate (TCPP) (Elbe: 40e250 ng L1, German Bight: 3e28 ng L1).
Keywords:
phosphate (TBEP), and triphenylphosphine oxide (TPPO) were detected in concentrations
TCPP
up to 180 ng L1. Seasonal trends were detected for the longitudinal profile of the Elbe
TCEP
estuary. Besides the dilution of river water with North Sea water toward the mouth, leading
River
to decreasing concentrations at the four sampling cruises (March, May, August, and
Estuary
October, 2010), an additional depletion of non-halogenated OPs was observed in summer
North Sea
which is supposed to be due to biodegradation or photodegradation.
Furthermore, triethyl phosphate (TEP), tri-iso-butyl phosphate (TiBP), tris(2-butoxyethyl)
Seasonal variations
In addition, a comparison of all important tributaries of the German Bight (Elbe, Ems, and Weser) as well as the indirect tributaries Meuse, Rhine, and Scheldt was done by multivariate statistical methods. It could be shown that the contribution of non-halogenated alkylated OPs in the Rhine was higher than in all other tributaries. The riverine input of OPs into the North Sea via the investigated tributaries was estimated to be about 50 t yr1. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Flame retardants and plasticizers based on organophosphorus acid esters (OPs) are a widely used substance group. Varying alkyl- and aryl-ester groups, some of them halogenated, lead to a large variation of their physico-chemical properties e ranging from very polar and volatile (e.g. trimethyl phosphate) to non-polar and non-volatile (e.g. tri(ethylhexyl) phosphate) (Reemtsma et al., 2008; SRC PhysProp
Database, 2010). This makes them useful for various different applications from flame retardants in polyurethane foam (especially the halogenated OPs) and plasticizers in flexible PVC (primarily non-halogenated OPs) to some minor applications as additives in computer housings or hydraulic fluids. A good overview of the usage of different OPs is given by Marklund et al. (2003). The worldwide usage of OPs was 207,200 t in 2004 (EFRA, 2010) and especially due to the replacement of the banned
* Corresponding author. Tel.: þ49 4152 872353. E-mail addresses:
[email protected] (U.E. Bollmann),
[email protected] (A. Mo¨ller). 1 Present address: Aarhus University, Department for Environmental Science, Frederiksborgvej 399, 4000 Roskilde, Denmark. Tel.: þ45 871 58462. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.028
532
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 1 e5 3 8
polybrominated diphenylethers (PBDEs) by OPs an increasing trend can be noticed (Stapleton et al., 2009). Already in the late 1970s first studies on the occurrence and fate of OPs in the aquatic environment were published (Sheldon and Hites, 1978; Saeger et al., 1979). Due to the fact that the emissions into the environment are still increasing, the OPs can be classified as “re-emerging pollutants” (Reemtsma et al., 2008). Since OPs are not chemically bonded to the polymeric materials, they can undergo diffusion processes and be emitted into the environment. By the reason of the large variation in physicalechemical properties, OPs can be transported in different environmental media and have been detected in various environmental compartments. High concentrations up to a few mg m3 were detected in indoor environments (Tollba¨ck et al., 2010) whereas in outdoor air concentrations of a few ng m3 were measured (Saito et al., 2007; Mo¨ller et al., 2011). Surface water concentrations ranged between a few ng L1 up to a few hundred ng L1 (Andresen et al., 2004; Regnery and Pu¨ttmann, 2010). Especially the halogenated OPs are supposed to be highly persistent in the environment and only slightly biodegradable. Due to this they are insufficient degraded in sewage treatment plants and enter the aquatic environment (Andresen et al., 2004; Marklund et al., 2005). Several OPs are known to be toxic (e.g. neurotoxic or carcinogen) and additionally, the lipophilic OPs have the potential to bioaccumulate (Reemtsma et al., 2008). This is the first comprehensive study of organophosphorus acid triesters (OPs) in the estuaries of all important tributaries of the German Bight (North Sea), i.e. the rivers Elbe, Weser, and Ems, as well as the indirect tributaries Meuse, Rhine and Scheldt, and the German Bight (North Sea) itself. Within this sampling region, OPs were analyzed in order to determine possible seasonal trends, and to compare the substance patterns of the different tributaries using multivariate statistical methods. Finally, the riverine input of OPs into the German Bight (North Sea) was estimated. This study covered 16 different OPs; thereunder three chlorinated alkyl phosphates (tris(2-choroethyl) phosphate, TCEP; tris(1-chloro-2propyl) phosphate, TCPP; tris(dichloroisopropyl) phosphate, TDCPP), ten non-halogenated alkyl phosphates (trimethyl phosphate, TMP; triethyl phosphate, TEP; tri-n-propyl phosphate, TPrP; tri-iso-propyl phosphate, TiPrP; tri-n-butyl phosphate, TBP; tri-iso-butyl phosphate, TiBP; tripentyl phosphate, TPeP; trihexyl phosphate, THP, tris(2-ethylhexyl) phosphate, TEHP; tris(2-butoxyethyl) phosphate, TBEP), two aryl phosphates (triphenyl phosphate, TPhP; tricresyl phosphate, TCP), and also the synthetic intermediate triphenylphosphine oxide (TPPO). A list of the OPs analyzed in this study with their chemical structures, applications, and physico-chemical properties is given in Table S1 of the supplementary material.
2.
Material and methods
2.1.
Chemicals
Information on CAS-No., producers, and purities of the used OP standards are listed in Table S3 in the supplementary material. SERDOLITH PAD 2 and 3 (analytical grade) were purchased by Serva (Germany) and mineral water was from
Bismarck-spring (Germany). All solvents were of highest purity (picograde) and obtained from Promochem (Germany). Sodium sulfate (granular, anhydrous for organic trace analysis) and hydrochloric acid (suprapur, 30%) were purchased from Merck (Germany).
2.2.
Sampling area and sampling
The sampling area (Fig. 1) is located in the south eastern part of the North Sea, called the German Bight, with the estuaries of its tributaries. On the one hand the North Sea has the largest intertidal mudflats in the world and on the other hand it is highly anthropogenic influenced by fishery, shipping and recreational activities. The German Bight is heavily influenced by the Elbe plume and in lower quantities by the Weser and Ems. In addition the diluted plume of the Rhine-MeuseScheldt-Delta enters the German Bight due to an easterly coastline flow. 1 L surface water samples were taken from the shores of the rivers Elbe, Weser, Ems, Rhine, Meuse, and Scheldt in August 2010. In addition samples from the estuary of the Elbe were taken during four expeditions with the German research vessel R/V LUDWIG PRANDTL. They were performed in March, May, August, and October 2010. The samples from the German Bight were taken during three expeditions with the German research vessel R/V HEINCKE in March, July, and September 2010. River water samples were immediately filtered (glass fiber filter, Whatman GF/C) and acidified with concentrated HCl for conservation. All samples were stored at 4 C until further analytical procedure (storage time max. five months).
2.3.
Analytical procedure
The analytical process was performed in a Varipro cleanroom (class 10000, Daldrop þ Dr. Ing. Huber, Neckartalfingen, Germany). Throughout the whole analytical procedure no plastic lab ware was used in order to reduce blanks. 500 mL (estuary) and 1000 mL (North Sea) surface water were spiked with 30 ng surrogates (TMP-d9, TEP-d15, TPrPd21, TBP-d27, TPhP-d15), respectively, and extracted by solid phase extraction (SPE) with 3 g SERDOLITH PAD 3 (cleaned with acetone, Soxhlet, 4 h) as adsorption material. The SPEmaterial was conditioned by 5 mL acetone and 10 mL precleaned mineral water (cleaned by rinsing over 25 g SERDOLITH PAD 2) and washed with 5 mL pre-cleaned mineral water after sample loading. Before elution with 50 mL dichloromethane the cartridges were dried by centrifugation (3000 rpm, 5 min). The extracts were reduced by rotary evaporation to 5e10 mL. Afterward, the water was frozen out over night at 18 C and residual water was removed by rinsing over Na2SO4. During further reduction in a constant nitrogen flow to 150 mL the solvent was changed to n-hexane. Finally, 200 pg 13C-hexachlorobenzene was added as injection standard. Analysis was done by an Agilent 6890 gas chromatograph coupled to an Agilent 5973 mass spectrometer (GCeMS). The injection was done by a PTV injector in pulsed splitless mode. The initial temperature of the injector was held at 60 C for 0.1 min and then increased at 500 C min1 to 300 C. The injection volume was 1 mL and the helium carrier gas flow was
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 1 e5 3 8
533
Fig. 1 e Sampling area and sampling stations.
1.3 mL min1. The GC was equipped with an HP-5MS column (30 m 0.25 mm i.d. 0.25 mm film thickness, J&W Scientific) and the oven temperature was 40 C for 4 min, afterward increased by 5 C min1 to 170 C (5 min), 10 C min1 to 230 C (5 min), 5 C min1 to 250 C, then 10 C min1 to 300 C. The MS transfer line and the ion source (electron impact chemical ionization, EI) were held at 280 C and 230 C, respectively. The MS operated in selected ion monitoring (SIM) mode. Masses for the detection of OPs and IS are included in Tables S1 and S2 in the supplementary material, respectively.
2.4.
Quality assurance
Recovery tests (500 mL pre-cleaned mineral water spiked to 60 ng L1 of each OP) showed absolute recoveries between 30 (TMP) and 105% (TPhP). Surrogates were used to compensate losses during the extraction process, hence, all concentrations were corrected by the recovery of the corresponding surrogate (see Table S4 in supplementary material). Relative recoveries ranged between 82 (TPPO) and 118% (TiBP). In some tests sodium chloride (to adjust a salinity of 30 PSU) and hydrochloride acid (to adjust pH 1) were added to confirm the robustness of the method within the present sample conditions. The limits of detection (LODs), calculated at a signal-tonoise (S/N) ratio of 3, ranged between 0.1 and 3.9 ng L1 and limits of quantification (LOQs), calculated at a S/N ratio of 9,
ranged between 0.2 and 11.7 ng L1. Method blanks were detected for TiBP, TPhP, TEHP, and TPPO ranging from 0.2 to 1.8 ng L1. Based on this the LODs and LOQs are calculated for these four substances using blank standard deviation method. Recoveries (Table S4), LODs and LOQs (Table S5) for all analytes are given in the supplementary material.
2.5.
Data treatment
The chromatographic evaluation was done by use of Agilent MSD ChemStation. Further data treatment was performed with Microsoft Office Excel 2003 and Statistica 6.
3.
Results and discussion
3.1.
Riverine and marine concentrations
In the estuary of the River Elbe up to 11 out of 16 analyzed OPs were detected in the dissolved phase. Most abundant OP in all samples was TCPP with individual concentrations from 40 to 250 ng L1. Furthermore, TEP (10e180 ng L1), TiBP (10e50 ng L1), TPPO (10e40 ng L1), and TBEP (
534
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 1 e5 3 8
mouth (Elbe-km 724) the total OP concentration ranged between 80 and 150 ng L1 (Fig. 2 and Table S6 in supplementary material). The concentrations in the different tributaries are equivalent to those in the estuary of the Elbe. In the non-tidal influenced part of the river the total OP concentration ranged between 400 and 550 ng L1. Only in the River Scheldt total OP concentrations up to 1092 ng L1 were detected. Individual concentrations ranged from a few ng L1 up to 570 ng L1 (Table 1). The dominating OP in all tributaries was TCPP (44e570 ng L1). Otherwise some differences in substance patterns were noticeable and, therefore, analyzed with multivariate statistical methods (see Section 3.3). The concentrations in all investigated rivers in this study are comparable to those in former studies (Table 2). Comparing the results for the Elbe to the concentrations measured by Andresen et al. (2007) and by the ARGE Elbe (2000) the levels for all substances are nearly similar. Indeed, the concentration of TCEP was slightly higher in the 1990s (ARGE Elbe, 2000). This is assumedly due to the consumption decrease of TCEP since 1989 (WHO, 1998). Fries and Pu¨ttmann (2001) observed TCEP in a concentration in the Upper Rhine similar to the Delta concentration in the present study. In contrast, the concentrations of the non-halogenated OPs TBP and TBEP are ten times higher in the Upper Rhine which is probably caused by production sites located at the Rhine and can be a reason for the different substance pattern, which will
P Fig. 2 e Total measured OP concentration ð OPÞ and concentrations of selected OPs at the beginning of the estuary (a, Elbe-km 609) and the mouth of the river Elbe (b, Elbe-km 719) in 2010 in ng LL1.
be discussed later (Section 3.3). The high consumption decrease in TCEP since 1989 is also noticeable in the Rhine: Knepper et al. (1999) detected up to 500 ng L1 in 1994, whereas in the later studies it decreased to less than 25 ng L1. The total OP concentration in the German Bight (North Sea) ranged from 5 to 50 ng L1. Similar to the estuary of the River Elbe, the major OPs were TCPP (3e28 ng L1), TEP (0.7e7 ng L1), TiBP (0.5e5 ng L1), TBEP (
3.2.
Spatial and seasonal variations
To obtain a better comparison of the OP concentrations along the River Elbe regarding the four different sampling months, the OP load was calculated as the product of the OP concentration and the water discharge, estimated for the sampling stations from the daily mean discharge data at the water gauge measuring station in Neu-Darchau. Upstream of Hamburg the total OP loading ranged between 0.1 and 0.5 g s1. Besides two slightly higher loadings at Elbe-km 639 and 649 in October the loading is nearly constant until Elbekm 689 (near Brunsbu¨ttel) in March, May and October. With increasing salinity toward the mouth a decreasing trend in the OP loading was observed downstream from Elbe-km 689. In contrast, in August the total OP concentration decreased over the entire estuary. As shown in Fig. 4, also the substance pattern changed in August along the entire Elbe estuary while it was similar for the three other campaigns. Especially the contribution of non-halogenated alkylated OPs decreased toward the mouth in August. In March and May the nonhalogenated OPs predominated (50%) whereas in August and October the contribution of halogenated was higher (70%). Similar to the river Elbe, changes in the percentages of halogenated and non-halogenated OPs were also observed for the German Bight (North Sea). In March and September the ratio of halogenated to non-halogenated OPs was 1.4 0.4 and 1.9 0.5, respectively, whereas in July the amount of halogenated OPs was slightly higher by a medium factor of 2.2 0.8 (see Fig. 3 and Figure S1 in supplementary material). The decreasing concentrations in the Elbe estuary toward the open sea in March, May, and October were caused by mixing processes with sea water which is confirmed by high negative correlation with salinity. In August an additional depletion occurred affecting the non-halogenated OPs. In the literature different degradation pathways for OPs can be found: biodegradation (Saeger et al., 1979) predominantly under aerobic conditions (Fries and Pu¨ttmann, 2003), alkaline hydrolysis (Mabey and Mill, 1978), as well as photodegradation (Muir et al., 1989; Regnery and Pu¨ttmann, 2009). Especially the short chain alkylated OPs are highly volatile and evaporation might occur with increasing temperature. In addition sorption processes (WHO, 1990, 2000) can be expected as reasons for decreasing concentrations in the dissolved phase. Due to the fact that sorption processes would occur for those OPs with high n-octanolewater partition coefficients and not only for non-halogenated OPs these processes can be excluded as
535
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 1 e5 3 8
P P Table 1 e Total ð OPÞ and individual concentrations in ng LL1 and total loadings ð OPLloadÞ in t yrL1 in the analyzed tributaries of the North Sea in August, 2010. P P OP loadb River Station OP in TCPP in TCEP in TDCPP in TEP in TiBP in TBEP in alk OPa in TPhP in TPPO in 1 1 1 1 1 1 1 1 1 1 in t yr1 ng L ng L ng L ng L ng L ng L ng L ng L ng L ng L Elbe
E1 E2 E3
435 249 96.9
134 84.3 44.0
35.5 9.25 4.93
30.8 10.9 6.4
22.3 33.9 6.99
19.2 15.4 4.31
94.3 24.2
7.4 3.5 1.4
10.3
80.7 65.2 25.2
8.1 5.0 2.1
Weser
W1 W2
386 58.3
167 24.3
34.0 3.29
26.6 5.3
13.2 5.16
13.2
48.4
5.5 1.7
77.2 15.1
1.6 0.3
Ems
Em1 Em2
397 222
175 89.8
34.2 11.5
35.3 8.0
49.2 27.9
11.1 4.81
42.7 38.9
7.7 2.8
42.1 38.0
0.4 n.a.
Rhine
R1 R2 R3 R4 R5 R6
360 326 466 375 586 485
79.2 74.8 139 159 115 122
12.6 12.4 22.0 25.8 18.4 14.9
14.3 13.2 30.6 20.5 18.6 15.7
55.1 37.2 52.2 29.7 82.7 55.0
84.0 78.6 80.6 16.8 78.2 68.7
30.3 32.6 53.9 38.7 51.7 28.5
9.5 9.4 20.4 38.7 37.8 44.8
73.7 65.6 64.8 43.6 183 133
23 n.a. 4.9 1.0 23 10
79.6
n.a.
185 71.5
n.a. n.a.
Meuse
M1
542
196
38.4
37.3
48.6
20.7
103
15.3
3.6
Scheldt
S1 S2
1092 355
570 164
69.9 19.0
67.0 19.2
84.5 67.7
5.27 5.04
72.0
36.6 7.3
a reason for the decreasing concentrations in August. Concluded from the comparison to secondary parameters such as pH-value and temperature, alkaline hydrolysis can also be excluded. This means that biodegradation as well as photodegradation and evaporation are likely the mechanisms leading to decreasing concentrations in August and affecting mostly the non-halogenated OPs. Consequently in summer halogenated OPs predominate, whereas in winter time it is nearly equalized with non-halogenated OPs. To conclude which one of the processes might be the leading process causing the decreases in OP concentrations in the water phase further research is necessary. Former studies for seasonal variations of OP concentrations in urban and rural lake water and in precipitation by Regnery and Pu¨ttmann (2009) showed no seasonal dependence. This can be confirmed with the results of this study by comparing
single substance concentrations of the four different sampling months at the same sampling point of the Elbe estuary, e.g. TCPP or TiBP at Elbe-km 609. The variations of the single substance concentrations follow no seasonal trends. However, as mentioned above the longitudinal profile of the Elbe estuary shows obvious seasonal dependent variations, which are caused by the different rates of degradation of the different substances.
3.3.
Substance pattern
As mentioned in Section 3.1 the substance pattern in the different tributaries seemed to be different. For this reason two multivariate statistical methods (cluster analysis and factor analysis) were conducted to find structures in the set of data and to make them visible. To neglect the seasonal
Table 2 e Concentrations of selected OPs in different studies in river water in ng LL1. River Elbe
Rhine
Ruhr Streams Danube Tiber Streams
Location (Country)
TCPP
TCEP
TiBP
TBP
Hamburg e Cuxhaven (D) Magdeburg e Hamburg (D) Dresden/Hamburg Stade (D)
40e250 200
5e20 60
10e50
2e7.5
0.3e4
789/399 19
494/103 23
3.1
12e25 24 50e500
17e84
6e28 218
28e54 321
1e2
This study Fries and Pu¨ttmann, 2001 Knepper et al., 1999
<40 183 24e52 87e323
10e200
Andresen et al., 2004 Quednow and Pu¨ttmann, 2008 Martı´nez-Carballo et al., 2007 Bacaloni et al., 2007 Haggard et al., 2006
Delta (NL) Ru¨sselsheim (D) Colone (D) Spring e Mouth (D) Hesse (D) Vienna (A) Rome (I) Arkansas (USA)
90 75e160 30e150 20e200 502 33e43 54e117
13e130 118 13e23
98e137
276 20e110 82e114 31e560
TBEP
TPhP
Ref. This study ARGE Elbe, 2000 Fries and Pu¨ttmann, 2001 Andresen et al., 2007
536
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 1 e5 3 8
Fig. 3 e OP concentrations in the German Bight (North Sea) in ng LL1 in March, July, and September, 2010 (no column [ not analyzed).
variations in the substance pattern (see Section 3.2) the following analysis was only based on the sampling in August, 2010, in all tributaries. In addition the values were standardized (from each value the average was subtracted and divided by the standard deviation) to allow comparisons of many variables with different dimensions (Einax et al., 1997). By means of a cluster analysis groups, i.e. clusters, are formed on the basis of multivariate distances to visualize similarities in a high-dimensional set of data (for further information see Einax et al., 1997). The final output of a cluster analysis, in this case conducted as hierarchical agglomerative clustering according to WARD, is a dendrogram (Figure S2 in supplementary material). This shows that one sample of the Scheldt (S1) and the samples from the two largest effluents of the Rhine (Nieuwe Waterweg, R5, and Hollandsch Diep, R6) differed a lot from all the other samples and formed separate clusters. In a smaller impact also a separation between the other samples of the Rhine and the remaining samples of the other rivers could be seen. For the interpretation of the hidden relationships a factor analysis was conducted converting correlated variables to so-called factors (for further information see Einax et al., 1997). In this case three factors with eigenvalues above 1 were extracted, which explain together about 80% of the total variance. A scatterplot of factors 1 and 2 (Fig. 5) shows a separation of the Scheldt sample S1 from the other samples due to factor 1 and of the Rhine samples caused by factor 2. This reveals that the Rhine differed a lot from the other rivers due to a higher fraction of non-halogenated OPs as TMP, TiBP, and TiPrP, which have highest loadings in factor 2. The separation of the Scheldt sample S1 in the cluster
analysis is caused by the very high OP concentrations at this sampling site (see Table 1). This is assumed because TCPP, TDCPP, TCEP, and TBP, substances which are common for all investigated rivers have high loadings in factor 1 (Table S7 in supplementary material). The reasons for these differences in the substance patterns of the different rivers can only be assumed. As indicated in Section 3.1 there are some production sites of OPs located along the river Rhine which can cause these differences. The patterns of all the other river might be due to normal usage of OP-containing products and other diffuse sources. Indeed, this needs to be confirmed by further research. Concluded from the comparison of the substance patterns in the different tributaries the slightly higher amount of nonhalogenated OPs in samples from the East Friesian coast (N4, N9) compared to those from the North Friesian coast (N2, N3) (see Fig. 3) is ascribed to a coastline inflow of the Rhine into the German Bight. In former studies this coastline flow was also recognized for other organic pollutants (Mo¨ller et al., 2010).
3.4.
Riverine input into the North Sea
Based on the sampling campaign in August, 2010, and daily mean water discharge values, the total riverine input of OPs was estimated to be 50 t yr1. The highest amount was discharged by the Rhine-Meuse delta (82%). Smaller amounts entered the North Sea via the rivers Elbe (11%), Scheldt (5%), Weser, and Ems (1% each). Annually, about 13 t TCPP, 6.1 t TEP, 5.8 t TiBP, and 3.5 t TBEP were discharged into the North Sea via the investigated tributaries.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 1 e5 3 8
537
A comparison to other organic pollutants shows the relevance for analyzing OPs in the marine environment. For example the input of poly- and perfluorinated compounds (PCFs), another important group of persistent organic pollutants, was estimated to be about 800 kg yr1 via the Elbe and 6e20 t yr1 via the Rhine (Mo¨ller et al., 2010), whereas it was 5.5 and 42.5 t yr1 of OPs, respectively.
4.
Conclusions
This study shows a seasonal dependency of the OP substance pattern. In summer non-halogenated OPs are affected by degradation processes, assumed to be biodegradation and photodegradation. This leads to a predomination of halogenated OPs in summer whereas in winter it is nearly equalized with non-halogenated OPs. Moreover, it could be pointed out that the substance pattern of the River Rhine differs from that of all other tributaries of the German Bight (North Sea), that were investigated in this study. It could be seen that the contribution of non-halogenated OPs is much higher in this river, which is probably due to OP production sites along the Rhine. The riverine input of OPs into the German Bight (North Sea) could be estimated to be 50 t yr1. This very high input reveals that further research in fate and transportation behavior of OPs in the marine environment is necessary.
Acknowledgment
Fig. 4 e Distribution of the OP concentrations in the Elbe estuary in March (a) and August (b) 2010 (blue: nonhalogenated alkylated OPs, yellow-red: halogenated OPs).
In contrast, Mo¨ller et al. (2011) estimated the atmospheric input of OPs into the German Bight to be about 50e70 times lower than the riverine input. This means, that the main OP fraction in the marine environment enters the sea via the rivers.
The authors would like to thank Thorben Ammann (University of Hamburg, Germany), FGG Elbe, FGG Weser, WSA Meppen, and Helpdesk Water of Rijkswaterstraat Waterdienst (Netherlands) for providing secondary parameters of the rivers. In addition, the captains and crews of R/V HEINCKE and R/V LUDWIG PRANDTL should be thanked as well as the AlfredWegener-Institut (Bremerhaven, Germany) for the possibility of taking part in the cruises of R/V HEINCKE. Moreover the authors would like to thank Armando Caba (Helmholtz-Zentrum Geesthacht) for taking the North Sea samples.
Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.watres.2011.11.028.
references
Fig. 5 e Scatterplot of the scores of factor 1 and 2 (E: Elbe, Em: Ems, M: Meuse, R: Rhine, S: Scheldt, W: Weser); Loadings factor 1: TCPP (0.958), TDCPP (0.966), TCEP (0.963), TBP (0.795); Loadings factor 2: TMP (0.816), TiPrP (0.872), TiBP (0.832).
Andresen, J.A., Grundmann, A., Bester, K., 2004. Organophosphorus flame retardants and plasticisers in surface waters. Science of the Total Environment 332 (1e3), 155e166. Andresen, J.A., Muir, D., Ueno, D., Darling, C., Theobald, N., Bester, K., 2007. Emerging pollutants in the North Sea in comparison to Lake Ontario, Canada, data. Environmental Toxicology and Chemistry/SETAC 26, 1081e1089.
538
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 1 e5 3 8
Arbeitsgemeinschaft zur Reinhaltung der Elbe (ARGE Elbe), 2000. Selected Organic Trace Pollutants in the River Elbe and its Tributaries in 1994e1999. http://www.arge-elbe.de/tl_fgg_neu/ veroeffentlichungen.html (in German). Bacaloni, A., Cavaliere, C., Foglia, P., Nazzari, M., Samperi, R., Lagana`, A., 2007. Liquid chromatography/tandem mass spectrometry determination of organophosphorus flame retardants and plasticizers in drinking and surface waters. Rapid Communications in Mass Spectrometry 21, 1123e1130. Einax, J.W., Zwanziger, H.W., Geiss, S., 1997. Chemometrics in Environmental Analysis. WILEY-VCH, Weinheim. European Flame Retardants Association, 2010. Market Statistics. www.cefic-efra.com (accessed 25.05.10). Fries, E., Pu¨ttmann, W., 2001. Occurrence of organophosphate esters in surface water and ground water in Germany. Journal of Environmental Monitoring 3, 621e626. Fries, E., Pu¨ttmann, W., 2003. Monitoring of the three organophosphate esters TBP, TCEP and TBEP in river water and ground water (Oder, Germany). Journal of Environmental Monitoring 5, 346e352. Haggard, B.E., Galloway, J.M., Green, W.R., Meyer, M.T., 2006. Pharmaceuticals and other organic chemicals in selected north-central and northwestern Arkansas streams. Journal of Environmental Quality 35 (4), 1078e1087. Knepper, T.P., Sacher, F., Lange, F.T., Brauch, H.J., Karrenbrock, F., Roerden, O., Lindner, K., 1999. Detection of polar organic substances relevant for drinking water. Waste Management 19, 77e99. Mabey, W., Mill, T., 1978. Critical review of hydrolysis of organic compounds in water under environmental conditions. Journal of Physical and Chemical Reference Data 7, 383e415. Marklund, A., Andersson, B., Haglund, P., 2003. Screening of organophosphorus compounds and their distribution in various indoor environments. Chemosphere 53 (9), 1137e1146. Marklund, A., Andersson, B., Haglund, P., 2005. Organophosphorus flame retardants and plasticizers in Swedish sewage treatment plants. Environmental Science & Technology 39 (19), 7423e7429. Martı´nez-Carballo, E., Gonza´lez-Barreiro, C., Sitka, A., Scharf, S., Gans, O., 2007. Determination of selected organophosphate esters in the aquatic environment of Austria. Science of the Total Environment 388, 290e299. Muir, D.C.G., Yarechewskiab, A.L., Grift, N.P., 1989. Biodegradation of four triaryl/alkyl phosphate esters in sediment under various temperature and redox conditions. Toxicological & Environmental Chemistry 18, 269e286. Mo¨ller, A., Ahrens, L., Sturm, R., Westerveld, J., van der Wielen, F., Ebinghaus, R., de Voogt, P., 2010. Distribution and sources of
polyfluoroalkyl substances (PFAS) in the River Rhine watershed. Environmental Pollution 158 (10), 3243e3250. Mo¨ller, A., Xie, Z., Caba, A., Sturm, R., Ebinghaus, R., 2011. Organophosphorus flame retardants and plasticizers in the atmosphere of the North Sea. Environmental Pollution 159, 3660e3665. Quednow, K., Pu¨ttmann, W., 2008. Organophosphates and synthetic musk fragrances in freshwater streams in Hesse/ Germany. CLEAN e Soil, Air, Water 36 (1), 70e77. Reemtsma, T., Quintana, J.B., Rodil, R., Garı´a-Lope´z, M., Rodrı´guez, I., 2008. Organophosphorus flame retardants and plasticizers in water and air. I. Occurrence and fate. Trends in Analytical Chemistry 27 (9), 727e737. Regnery, J., Pu¨ttmann, W., 2009. Organophosphorus flame retardants and plasticizers in rain and snow from Middle Germany. CLEAN e Soil, Air, Water 37 (4e5), 334e342. Regnery, J., Pu¨ttmann, W., 2010. Occurrence and fate of organophosphorus flame retardants and plasticizers in urban and remote surface waters in Germany. Water Research 44, 4097e4104. Saeger, V., Hicks, O., Kaley, R., Michael, P.R., Mieure, J.P., Tucker, E.S., 1979. Environmental fate of selected phosphate esters. Environmental Science & Technology 13 (7), 840e844. Saito, I., Onuki, A., Seto, H., 2007. Indoor organophosphate and polybrominated flame retardants in Tokyo. Indoor Air 17, 28e36. Sheldon, L.S., Hites, R.A., 1978. Organic compounds in the Delaware River. Environmental Science & Technology 12 (10), 1188e1194. SRC PhysProp Database Demo, 2010. http://www.syrres.com/ what-we-do/databaseforms.aspx?id¼386 (accessed 28.12.10). Stapleton, H.M., Klosterhaus, S., Eagle, S., Fuh, J., Meeker, J.D., Blum, A., Webster, T.F., 2009. Detection of organophosphate flame retardants in furniture foam and U.S. house dust. Environmental Science & Technology 43 (19), 7490e7495. Tollba¨ck, J., Isetun, S., Colmsjo¨, A., Nilsson, U., 2010. Dynamic non-equilibrium SPME combined with GC, PICI, and ion trap MS for determination of organophosphate esters in air. Analytical and Bioanalytical Chemistry 396, 839e844. World Health Organization (WHO), 1990. Environmental Health Criteria 110: Tricresyl Phosphate. WHO, Geneva. World Health Organization (WHO), 1998. Environmental Health Criteria 209: Tris(chloropropyl) phosphate, Tris(2-chloroethyl) phosphate. WHO, Geneva. World Health Organisation (WHO) (2000) Environmental Health Criteria 218: Tris(2-butoxyethyl) phosphate, Tris(2-ethylhexyl) phosphate, and tetrakis(hydroxymethyl) phosphonium salts. Geneva.WHO
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 9 e5 4 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Modeling volatile organic sulfur compounds in mesophilic and thermophilic anaerobic digestion of methionine Weiwei Du*, Wayne Parker Dept. of Civil and Environmental Engineering, University of Waterloo, 200 University Ave. W., Waterloo, ON, Canada N2L 3G1
article info
abstract
Article history:
Processes involved in volatile organic sulfur compound (VOSC) generation and degradation
Received 5 April 2011
in mesophilic and thermophilic digestion of methionine were identified, kinetically studied
Received in revised form
and a mathematical model was established. MM was found to be the only VOSC directly
11 November 2011
generated from methionine degradation. MM was methylated to form DMS and both MM
Accepted 15 November 2011
and DMS were subsequently degraded to H2S. Mixed-second order kinetics were found to
Available online 25 November 2011
best fit the VOSC generation and conversion processes. The kinetic constants (average values) for MM generation and methylation and MM and DMS degradation were estimated
Keywords:
to be 0.0032, 0.0047, 0.027, and 0.013 l g1 h1 respectively at 35 C and 0.0069, 0.0012, 0.0083,
Methyl mercaptan
0.005 l g1 h1 respectively at 55 C. More rapid MM release and slower VOSC decline at
Dimethyl sulfide
thermophilic temperature implied that VOSC could be more problematic at thermophilic
Volatile sulfur compounds
temperatures as compared to mesophilic conditions. ª 2011 Elsevier Ltd. All rights reserved.
Mesophilic Thermophilic Anaerobic digestion
1.
Introduction
Increasing complaints of odor problems at wastewater treatment plants has aroused attention. Volatile organic sulfur compounds (VOSC) and H2S from anaerobic digestion have been identified as the most odorous compounds related to sewage treatment because of their very low odor thresholds and very negative hedonic values. Even a small amount of VOSC and H2S can contribute to odor pollution (Smet and Langenhove, 1998). In addition, VOSCs and H2S in biogas are reactive and corrosive to metal pipes, biogas storage tanks, and biogas utilization equipment such as biogas engines and turbine generators. When biogas is used for more efficient electricity generators such as fuel cells, it has to be cleaned-up and reformed. VOSCs and H2S can poison the catalysts that are used in both reforming and fuel cells. It has been reported that when the total sulfur concentration is above 10 ppmv, the
anode of solid oxide fuel cell systems will be deactivated. In addition, when the concentration of H2S was higher than 200 ppmv, activated carbon treatment could not effectively accomplish sulfur removal to reach a required purity of biogas (Wheeldon et al., 2007). Proteineous materials and sulfate have been reported as major sources of the volatile sulfur compounds that are generated during anaerobic digestion of municipal sludge. These sulfur containing compounds are converted to VOSC and H2S under anaerobic conditions through mechanisms such as biological reduction, hydrolysis, methylation, and metal catalyzed oxidization (Drotar et al., 1987; Kadota and Ishida, 1972; Lomans et al., 2002; Hullo et al., 2007; Sreekumar et al., 2009). However, VOSC concentrations in digesters are reduced by methanogens that mediate the degradation of VOSC (Finster et al., 1992; Chen et al., 2005; De Bok et al., 2006).
* Corresponding author. Tel.: þ1 519 729 0850. E-mail address:
[email protected] (W. Du). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.043
540
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 9 e5 4 6
In anaerobic sludge digestion, biochemical generation of H2S from sulfate reduction and pertinent kinetics has been well established (Visser, 1995; Kalyuzhnyi et al., 1998; Kalyuzhnyi and Fedorovich, 1998; Fedorovich et al., 2003) while VOSC generation and conversion is not well understood. Temperature has been reported to affect VOSC generation and release in both the start-up and steady state operation of anaerobic digesters (Iranpour et al., 2005), however, quantitative information on the effects of temperature on VOSC behavior is limited. To better predict VOSC behavior in anaerobic sludge digestion, the occurrence of the sulfur related reactions in digestion of municipal sludges needs to be confirmed, corresponding kinetic information needs to be quantified, and the effect of temperature should be assessed. The predominant amino acids in proteins that contain sulfur are methionine (C5H11NO2S, CH3eSe(CH2)2eCH(NH2)e COOH) and cysteine (C3H7NO2S, HSeCH2eCH(NH2)eCOOH) and both have been reported to be VOSC precursors in municipal sludge digestion. Methionine has been reported to be degraded through different pathways under different conditions to produce either methyl mercaptan (MM), dimethyl sulfide (DMS) or H2S. Cysteine normally is considered as an organic precursor of only H2S under anaerobic conditions (Kadota and Ishida, 1972; Derbali et al., 1998). While the pathways for VOSC generation have been documented, kinetic information on VOSC formation and conversion in the degradation of methionine is not available and hence prediction of the VOSC behavior in anaerobic sludge digestion is challenging. This paper describes a study of the kinetics of VOSC generation from methionine and their subsequent degradation in mesophilic and thermophilic anaerobic sludge digestion. The development of a mathematical model that describes the conversions is presented.
2.
Materials and methods
Two groups of batch tests were conducted in the present study. The first group of tests included anaerobic digestion of methionine with and without methanogen inhibition, using digested municipal wastewater treatment sludge as the inocula. In methionine digestion without methanogen inhibition, the types of VOSCs that are generated and subsequently transformed were identified. The VOSC responses in these tests were also subsequently employed for the purpose of model verification. The biodegradation of VOSCs was eliminated when methanogens were inhibited and hence VOSC accumulation in methionine digestion with inhibition of methanogens was employed for estimation of the kinetics of VOSC formation from methionine. The second group of batch tests involved incubations that were dosed with individual VOSCs. In these tests, the VOSCs that were observed to be formed from methionine were dosed into a digested sludge inocula and their degradation was monitored for development of kinetic rate models. Batch incubations were conducted in non-stirred, 328 ml serum bottles that were sealed with screw-caps which were equipped with bromobutyl rubber septa. The temperatures of the mesophilic and thermophilic incubations were 35 and 55 C, respectively. Digested sludges that were obtained from
pilot-scale mesophilic (35 C) and thermophilic (55 C) digesters were pre-incubated at the target temperatures for 48 h, and then used as inocula. The pilot-scale digesters were operated at an average hydraulic retention time of 22 days and fed with a co-thickened sludge that was the underflow of primary settler collected from Waterloo Wastewater Treatment Plant. The yearly average flow rate of wastewater influent, waste activated sludge (WAS), and primary underflow was 36,655, 753, and 282 m3/day respectively. The average total suspended solids concentration (TSS) of the WAS and cothickened sludge was about 0.45% and 3% (VSS/TSS w 70%). The values for VSS destruction across the pilot-scale mesophilic and thermophilic digesters were approximately 60% and 64% respectively. The total sulfur concentrations in the pilot-scale mesophilic and thermophilic digesters were between 20 and 36 mg S/g VSS while the organic sulfur concentrations in the pilot-scale digesters were between 9 and 20 mg S/g VSS. The total volume of the liquid phase was 200 ml with initial volatile suspended solids (VSS) varying between 1.8 and 9.3 g/l. VSS concentrations were employed as a surrogate for the concentration of biomass which mediated biological sulfur related reactions. The specific reaction rates for methionine and VOSC degradation at 35 C and DMDS degradation at 55 C were found to be relatively fast and hence dilute inocula with an initial VSS between 1.8 and 3.0 g/l was used. For most of the tests at 55 C, undiluted inocula with higher initial VSS levels, between 6.4 and 9.3 g/l, was used due to the slow degradation rates of MM and DMS. All the serum bottles were purged with pure nitrogen before incubation. In methionine-dosed tests a 1000 mM methionine stock solution was added to achieve concentrations of 2e5 mM (equivalent to 6.9e70.2 mg S/g VSS when normalized by VSS). When dilute inocula were used at 55 C, MM accumulated and minimal decay was observed within 5 days. Therefore, undiluted inocula were used for methionine incubations at 55 C, which led to a reduced normalized initial S concentration. In bottles where methanogens were inhibited, a 2-bromoethanesulfonic acid (BES) stock solution was dosed to a final concentration of 15 mM as the methanogen inhibitor. Methionine incubations, with and without methanogen inhibition, were carried out with 5 replicates respectively. The replicate bottles were gradually sacrificed for total Kjeldahl nitrogen and ammonia analysis to monitor methionine hydrolysis with time (data not presented in this paper). At the same time, the replicate bottles ensured replicate headspace samples for VOSC analysis. In the tests that were dosed with VOSCs which were detected in the methionine incubations (MM and DMS), 2e5 different initial VOSC dosages were employed. The initial VOSC dosages were in the range of 0.92e10.57 mg S/g VSS. The analytical grade VOSC was dosed into serum bottles directly by syringes. Duplicate bottles were prepared for each initial dosage. The volume of biogas produced in the serum bottles was measured by release into a manometer, until the pressure in the serum bottle was equal to atmosphere pressure. The volume of water in the manometer which was replaced by the released gas was then recorded. The frequency of biogas release was gradually reduced from once every day to once a week over the duration of incubation. Methane generation was used to monitor the methanogen activity. The methane and carbon dioxide content of biogas samples were analyzed
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 9 e5 4 6
by an SRI 310C gas chromatograph (GC) equipped with a thermal conductivity detector and a 60 1/800 OD Porapak column (80/100 mesh). The VOSC content of the biogas was analyzed by a GC equipped with a capillary column (Valcobond VB-1, 100% dimethyl-polysiloxane, 30 m 0.32 mm, 0.4 mm) and a pulsed flame photometric detector (PFPD). For methionine incubation tests, the gas phase of all the bottles was sampled 2 times every day for the first 3 days, then once every 1e3 days until the end of the incubation. For the VOSC dosed tests, the gas phase was sampled with a frequency from 1 to 4 times a day. Total suspended solids/volatile suspended solids (TSS/VSS) of the inocula were measured following Standard Methods 2540 D and E respectively (APHA, AWWA, and WEF, 1992).
3.
Results and discussion
To better reflect the sulfur transformations during methionine degradation and VOSC conversion, the masses of methionine and the VOSCs were represented by their S equivalent and normalized by VSS mass in the serum bottles. VOSC concentrations in the headspace were measured by the GC-PFPD. Concentrations of VOSC in the liquid phase were quantified based on their respective liquid-gas partitioning coefficient (Du, 2010) with an assumption of instantaneous equilibrium for VOSC distribution between the liquid and gas phases. The total VOSC mass was the sum of VOSC masses in the two phases.
3.1.
Methionine incubations
MM and DMS have been reported to be generated through methionine degradation in anoxic salt marsh sediments (Kiene and Visscher, 1987), in treated biosolids (Higgins et al., 2006), and in food fermentation (Bonnarme et al., 2000). In the present tests, VOSC concentrations were monitored over time in the headspaces of methionine incubations without methanogen inhibition. Consistent with the literature, MM and DMS were detected in the sludge digestions with dosed methionine. Their concentrations (average of duplicates) with time are presented in Fig. 1. The variability of measurements was less than 10% of the means and hence was not presented in the figure. From Fig. 1 it can be seen that at 35 C, with an initial methionine dosage of 70.2 mg S/g VSS, MM rapidly increased to
Fig. 1 e MM and DMS concentrations versus time in methionine incubation without methanogen inhibition.
541
2.5 mg S/g VSS after 25 h of incubation and then gradually declined to less than 0.1 mg S/g VSS after 300 h of incubation. The concentration of DMS increased with MM concentrations, peaked at 0.9 mg S/g VSS after 100 h of incubation, and decreased to less than 0.1 mg S/g VSS after 200 h of incubation. Similarly at 55 C, with an initial methionine concentration of 6.9 mg S/g VSS MM rapidly increased and maintained a high concentration of 1.2 mg S/g VSS for the first 20 h of incubation and then gradually declined to less than 0.01 mg S/g VSS after 78 h. The concentration of DMS increased with MM concentrations, reached a peak value of 0.5 mg S/g VSS after 54 h of incubation, and declined to below the detection limit after 78 h. Kiene and Visscher (1987) reported that DMS was generated from MM and that MM peak concentrations were 6e30 fold of the peak DMS concentrations during methionine degradation in sediments. The observations of the present study with sludge digestion were consistent with the previous study at both temperatures that were examined. Without methanogen inhibition MM generation started immediately and the MM concentrations reached peak values earlier than the DMS values. The MM peak concentrations were much higher than DMS peak concentrations at both temperatures. The results suggest that MM was the direct VOSC product of methionine degradation while DMS was generated from MM as shown in Fig. 2. The disappearance of MM and DMS will be subsequently demonstrated to be mediated by methanogens. The time-course of the VOSC results will be subsequently used to validate the models that have been developed to simulate VOSC formation and decay during methionine incubation. Fig. 3 presents the concentrations of MM that were observed when methanogens were inhibited in the methionine incubations. The initial VSS concentrations in the bottles were utilized to normalize the sulfur mass. At both temperatures there was a significant lag in MM generation from methionine at 35 C (Fig. 3), however, after about 120 h of incubation, the mass of MM increased rapidly. After 400 h of incubation at 35 C, the mass of MM accumulated to a level of 65.0 mg S/g VSS, which represented 92.5% of the sulfur available from the dosed methionine (original sulfur dose was 70.2 mg S/g VSS). After 340 h of incubation at 55 C, the MM concentration approached a constant level that was maintained until the end of the batch test. The generated MM accumulated to a level of 45.0 mg S/g VSS, which represented
Fig. 2 e Conceptual model of VOSC generation and degradation pathways in anaerobic digestion of methionine.
542
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 9 e5 4 6
It is typically assumed that the inhibitor BES only affects methanogens (Thauer, 1998). However, the lag in generation of ammonia and MM in the methionine incubations with BES addition suggested that BES might also inhibit methionine hydrolysis. The data suggest that the biomass was able to acclimate to the inhibitory effect after 120 h and methionine was totally hydrolyzed at the end of the tests.
3.2.
Fig. 3 e MM generation with time in methionine incubation with methanogens inhibited.
essentially all the sulfur available from the dosed methionine (original sulfur dose was 45.9 mg S/g VSS). Zinder and Brock (1978) reported that MM was the direct product of methionine degradation in anaerobic sediments. The present study conducted with municipal digestion sludge confirmed this conclusion. In the methionine incubations with methanogens inhibited, MM was the only VOSC which was detected. It was previously observed that at both temperatures MM and DMS were detected in anaerobic incubations of methionine without methanogen inhibition and both VOSCs eventually disappeared. In both mesophilic and thermophilic incubations with methanogen inhibition, MM was the dominant sulfur containing product generated. Its concentration rapidly accumulated with time after the acclimation period and then maintained at a high concentration level until the end of incubation. These results confirmed that the degradation of both MM and DMS is mediated by methanogens. Pianotti et al. (1986) reported that inhibition of methionine degradation occurred when an initial methionine concentration which was higher than 2 mM was employed in the incubation and the inhibition led to a postponed sulfur release. In the present study, with an initial methionine concentration of 2e5 mM, the MM generation from methionine was only delayed in the methanogen-inhibited incubations. There was no lag in the methionine incubation without methanogen inhibition at both mesophilic and thermophilic temperatures. Therefore, high initial methionine concentration (>2 mM) was not sufficient to explain the time lag which only occurred in incubations with methanogen inhibition. Baena et al. (1998) reported that methionine degradation depended on the removal of its fermentation products by methanogens. Accumulation of fermentation products of methionine could be inhibitory for continuous methionine degradation. In the present experiment, no ammonia accumulation was detected (ammonia detection limit, 0.3 mg/l with 95% confidence) during the lag period of the inhibited bottles (data not presented). With an assumption that ammonia was one of the fermentation products of methionine, slow ammonia generation suggested an inhibited fermentation. Therefore, the temporarily inhibited methionine degradation was not caused by the accumulation of fermentation products.
VOSC incubations
Batch tests that were dosed with individual MM and DMS were conducted to facilitate estimation of degradation kinetic rate constants for these substances. Different initial dosages were employed in these tests to assess the effect of the concentration of VOSCs on their degradation. The estimation of the kinetic coefficients of VOSC degradation based on the timecourse of VOSC variation will be discussed in a subsequent section. The effect of the temperature was assessed by comparing VOSC degradation rates at temperatures of 35 and 55 C. Fig. 4 presents total MM and DMS masses (normalized by VSS) present in serum bottles versus incubation time in the mesophilic batch tests with an initial MM mass of 3.11 mg S/g VSS and an initial DMS mass of 2.92 mg S/g VSS respectively. The responses observed in Fig. 4 were representative of the trends in VOSC degradation that were observed in all the mesophilic batch tests and representative of the trends in the thermophilic batch tests when the initial masses of MM and DMS were less than 2 mg S/g VSS and 3 mg S/g VSS respectively. The decline of the mass of VOSCs in the serum bottles was attributed to biodegradation that was mediated by methanogens. The role of methanogens in VOSC degradation was observed in the methionine incubation tests and has been reported in the literature (Lomans et al., 1999). At both mesophilic and thermophilic temperatures the mass of DMS declined at a considerably slower rate as compared to MM. For instance, at 35 C with an initial MM concentration of 3.11 mg S/g VSS about 55 h was required to achieve 95% reduction in the mass of MM while 120 h was required to obtain the same reduction of DMS when the initial concentration of DMS was 2.92 mg S/g VSS (Fig. 4). At 55 C with an initial concentration of 1.84 mg S/g VSS, 60 h was
Fig. 4 e MM and DMS concentrations versus time (in mesophilic incubations and when initial masses of MM and DMS were less than 2 and 3 mg S/g VSS respectively in themophilic incubations).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 9 e5 4 6
required to achieve a 95% reduction of MM while DMS required 175 h to achieve the same extent of degradation when an initial concentration of 2.13 mg S/g VSS was employed. In the thermophilic batch tests, the degradation of MM was inhibited when the initial normalized mass of MM in the serum bottles was higher than 2 mg S/g VSS. Fig. 5 presents an example of this type of response when the initial normalized mass was 4.26 mg S/g VSS. From this figure it can be seen that after 240 h of incubation the normalized mass had reduced by less than 5% of the initial value. Inhibition of methanogen activity has been previously reported under conditions with high concentrations of MM (Kiene et al., 1986; Van Leerdam et al., 2006). Elevated concentrations of DMS (3.96, 6.6 and 10.57 mg S/g VSS) also appeared to inhibit the degradation of this compound in the thermophilic incubations. However, in this case the decline in concentration of DMS exhibited a staged and slow decline. As indicated in Fig. 5 the mass of DMS declined approximately linearly for the first 70 h of incubation and then the concentrations became essentially constant. By the end of the incubation (170 h) less than 70% of the dosed DMS had degraded (Fig. 5). Inhibition of DMS degradation was not observed in tests with initial dosages less than 3 mg S/g VSS. When the initial DMS dosages were low (<3 mg/g VSS), DMS declined steadily (Fig. 4). The inhibited DMS degradation at high initial concentrations was consistent with the results of Kiene et al. (1986) where high DMS concentrations inhibited methane generation. Inhibition was only observed at thermophilic temperature in the present study. Considering that in mesophilic and thermophilic municipal sludge digestion, the typical accumulated MM and DMS mass was less than 2 and 3 mg S/g VSS respectively, the inhibition conditions were not included in modeling VOSC conversions in sludge digestion. In the MM dosed tests, there was a transient accumulation and subsequent degradation of DMS in the serum bottles. Fig. 6 presents the MM and DMS responses that were observed in the mesophilic incubation of MM at an initial dose of 1.30 mg S/g VSS and is representative of the responses observed for the other initial doses at 35 C. From Fig. 6 it can be seen that approximately 25 h after the addition of MM, DMS was detected and its concentration slowly increased to a peak at around 30 h, which corresponded with the time when MM was exhausted. The transient accumulated DMS subsequently degraded below the detection limit within another 20e25 h.
Fig. 5 e MM and DMS concentrations versus time in thermophilic incubations at elevated initial doses.
543
Fig. 6 e DMS and MM concentrations versus time in MMdosed mesophilic incubation.
A similar response was observed in the thermophilic incubation of MM. Fig. 7 shows DMS accumulation in the thermophilic degradation of MM with an initial dose of 0.92 mg S/g VSS and is representative of the responses that were observed when the initial dose was less than 2 mg S/g VSS. From Fig. 7 it can be seen that after 5 h of incubation DMS was detected and the DMS concentration slowly increased with the decrease of MM. In contrast to the mesophilic tests, the generated DMS did not disappear at 60 h of incubation when the test was terminated. The results suggest that methylation of MM to form DMS occurred in the incubations at both temperatures. The continuous accumulation of DMS in the thermophilic incubation reflects a slower degradation rate for DMS at thermophilic temperatures as compared to mesophilic temperatures. This behavior was consistent with that observed in the DMS dosed tests. However, methylation was not the only mechanism which resulted in the reduced mass of MM-S in the incubations. In all the MM dosed tests at 35 C, the peak masses of DMS-S were about 20% of the initial dosed MM-S masses. At 55 C the DMSS mass slowly increased with the decrease of MM-S. When more than 90% of the dosed MM-S had degraded, the amount of the S associated with the generated DMS was less than 13% of the initial dosed S in the MM. Hence, generation of DMS could not represent the total amount of MM which was removed. These results suggest that direct reduction of MM to generate H2S was likely a competitive reaction with methylation.
Fig. 7 e DMS and MM concentrations versus time in MMdosed thermophilic incubation.
544
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 9 e5 4 6
Table 1 e Model of VOSC generation and conversion in methionine digestion. Process
C_met
Degradation of methionine DMS degradation MM degradation MM methylation to DMS
3.3.
C_MM
1
C_H2S
Rate
1
1 1
kmet$C_met$VSS k_DMS$C_DMSliq$VSS k_MM$C_MMliq$VSS kmethy$C_MMliq$VSS
1 1 1
1
Modeling and kinetic parameter estimation
The pathways of generation and consumption of VOSCs in methionine incubation presented in Fig. 2 were confirmed by the methionine and VOSC responses in the dosed batch tests. The stoichiometric relations between the sulfur containing compounds are summarized in Table 1. Mixed-second order kinetics (Equation (1)), were found to best depict anaerobic biological methionine degradation, degradation of MM and DMS, as well as the methylation of MM. r¼
C_DMS
dc ¼ k$C$VSS dt
(1)
Where r is the consumption rate of the sulfur containing compound (mg l1 h1), C is the sulfur concentration (as methionine or VOSC) in the liquid phase (mg/l), t is time (h), k is the mixed-second order kinetic parameter (l g1 h1), and VSS is the volatile suspended solids which represented the biomass concentration (g/l). According to the mixed-second order kinetics, the rates of VOSC generation/conversion processes depend on the concentration of substrate as well as the biomass concentration. In the present study, the biomass concentration was assumed to be a constant value over the incubation and represented by the initial VSS. The actual VSS deduction rate was up to 20% in the methionine incubations without methanogen inhibition and was about 10% in the other batch incubations. The assumption of negligible decrease of VSS through the digestion may result in slightly conservative estimates of the kinetic constant values. The specific mixed-second order kinetic expressions for each conversion are summarized in Table 1. The MM and DMS degradation kinetic constants were estimated with the data collected from incubations that were not inhibited with high MM and DMS concentrations. Methionine concentrations (C_met) were not directly measured in the experiments conducted in this study. Assuming a direct conversion of methionine to MM (Table 1) the rate of methionine degradation was determined from the rate of MM generation. Therefore, kmet was estimated by fitting MM accumulation in the methionine incubations.
The estimated values of kmet at mesophilic and thermophilic temperatures are presented in Table 2. The value of kmet for thermophilic digestion was two times higher than that for mesophilic digestion, reflecting the data that demonstrated the release of MM from methionine was more rapid at 55 C. Fig. 8 presents the fit of the model to the decline of DMS at two different initial dosages in the thermophilic batch tests and is representative of the model fit for DMS degradation under both mesophilic and thermophilic conditions. From Fig. 8 it can be seen that the calibrated model was able to describe the decay of DMS over a range of DMS concentrations. The estimated rate coefficients for DMS degradation at the two temperatures are summarized in Table 2. In the MM dosed batch tests, MM was converted to generate both H2S and DMS and the generated DMS also decayed. Fig. 9 presents representative responses for the decline of MM and the formation and disappearance of DMS with an initial MM mass of 4.67 mg S/g VSS at 35 C. The responses presented in Fig. 9 are representative of those observed for MM degradation with different initial dosages in both the mesophilic and thermophilic batch tests. The value of k_DMS employed in these simulations was the value that was estimated from for the previously described DMS degradation tests. The values of k_MM and kmethy were estimated by fitting the MM and DMS time-course curves observed in the MM dosed tests (Table 2). The decay rate coefficients for each process (MM and DMS decay and MM methylation) were estimated for each initial VOSC dosage and were found to be not statistically different between the various doses (P ¼ 0.01). Comparing the average values of the coefficients estimated at 35 C and at 55 C (Table 2), it can be seen that the MM and DMS decay rate constants were about 3.4 times and 2.7 times higher at 35 C as compared to 55 C. Hence, the values of the decay rate coefficients reflect the trends in compound disappearance that were previously discussed. The results suggest that methanogens have a reduced capacity for MM and DMS degradation at 55 C as compared to that at 35 C. The MM methylation rate constant was also greater at 35 C as compared to its value at 55 C with the average rate
Table 2 e Methionine degradation and VOSC conversion coefficients. Mesophilic
1
1
kMet-S (l g h ) k_DMS (l g1 h1) k_MM (l g1 h1) kmethy (l g1 h1)
Thermophilic
Average value
Range (P ¼ 0.01)
Average value
Range (P ¼ 0.01)
0.0032 0.013 0.027 0.0047
0.0023e0.0040 0.010e0.016 0.010e0.047 0.0038e0.0063
0.0069 0.005 0.0083 0.0012
0.0053e0.0085 0.0040e0.0057 0.0053e0.013 0.0005e0.0041
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 9 e5 4 6
545
Fig. 8 e Observed and predicted DMS concentrations versus time in thermophilic incubations. Fig. 10 e Simulated and monitored VOSC concentrations with time in mesophilic methionine digestion.
Fig. 9 e Observed and predicted concentrations of MM and DMS variation in MM-dosed mesophilic incubation.
coefficient at 35 C about 3.9 times the value at 55 C. This value was consistent with the faster decline of MM that was observed at 35 C. Considering the more rapid release of MM from methionine at 55 C that was previously described, it could be speculated that VOSC persistence under thermophilic conditions would be longer and at higher concentrations than that under mesophilic conditions, when similar initial sulfur masses are present. Verification of the combined model presented in Table 1 was conducted by comparing model simulations with the
Table 3 e Initial conditions and kinetic parameter values utilized in the simulation of methionine digestion. Parameter values Mesophilic 1
1
k_met (l g h ) k_MM (l g1 h1) kmehty (l g1 h1) k_DMS (l g1 h1)
0.0034 0.069 0.006 0.02
Thermophilic 0.0065 0.017 0.003 0.004
Initial mass/concentration
Dosed methionine (mg S/l) VSS (g/l)
Mesophilic
Thermophilic
160 2.28
64 9.30
Fig. 11 e Simulated and monitored VOSC concentrations with time in thermophilic methionine digestion.
VOSC quantities that were previously described for methionine digestion without methanogen inhibition. The average parameter values presented in Table 2 were employed as the initial values for these simulations and were adjusted within the range presented in Table 2 to optimize the model simulations with the measured data. The initial methionine doses and VSS concentrations and the adjusted values of the kinetic constants employed in these simulations are listed in Table 3. The observed and predicted VOSC quantities for methionine digestion without methanogen inhibition are presented in Figs. 10 and 11 for mesophilic and thermophilic digestion respectively. From these figures it can be observed that the simulated values exhibited good agreement with the monitored data and hence validated the model structure and the values of the kinetic coefficients.
4.
Conclusions
The processes involved in VOSC generation from methionine and the degradation of the byproducts were quantitatively characterized in this study. At both mesophilic and thermophilic temperatures MM was the direct volatile sulfur containing product of anaerobic methionine degradation. The
546
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 9 e5 4 6
generated MM was demonstrated to be methylated to form DMS. Both MM and DMS were found to be degraded by the methanogenic population. At both temperatures, the degradation of DMS was slower than that of MM. Inhibition was observed in thermophilic VOSC degradation when initial masses of MM was above 2 mg S/g VSS and DMS was above 3 mg S/g VSS. Mixed-second order kinetics were able to best describe MM generation, MM methylation, and MM and DMS degradation in anaerobic methionine incubation when digested sludge was employed as the inocula. The kinetic constant for MM generation at 55 C was about 2.2 times that of the kinetic constant at 35 C, which suggested a more rapid MM release at thermophilic temperatures. The kinetic constants of MM methylation and MM and DMS degradation at 35 C were about 3.9, 3.4, and 2.7 times of those observed at 55 C, which reflected the faster disappearance of MM and DMS at mesophilic temperature. The model developed in the present study which combined four VOSC conversion processes was able to predict VOSC generation and subsequent degradation in methionine digestion at two different temperatures. Inhibition of VOSC degradation which might be caused by high initial VOSC masses was not included in the present model because in typical sludge digestion systems accumulated MM and DMS masses were found to be less than 2e3 mg S/g VSS and hence would not be inhibitory.
references
APHA, AWWA, WEF, 1992. Standard Method for the Examination of Water and Wastewater, eighteenth ed.. Baena, S., Fardeau, M.L., Labat, M., Ollivier, B., Thomas, P., Garcia, J.-L., Patel, B.K.C., 1998. Aminobacterium colombiense gen. nov. sp. nov., an ammonia acid-degrading anaerobe isolated from anaerobic sludge. Anaerobe 4, 241e250. Bonnarme, P., Psoni, L., Spinnler, H.E., 2000. Diversity of lmethionine catabolism in cheese-ripening bacteria. Applied and Environmental Microbiology 66, 5514e5517. Chen, Y., Higgins, M.J., Maas, N.A., Murthy, S.N., Toffey, W.E., Foster, D.J., 2005. Roles of methanogens on volatile organic sulfur compound production in anaerobically digested wastewater biosolids. Water Science and Technology 52, 67e72. De Bok, F.A.M., van Leerdam, R.C., Lomans, B.P., Smidt, H., Lens, P.N.L., Janssen, A.J.H., Stams, A.J.M., 2006. Degradation of methanethiol by methylotrophic methanogenic archaea in a lab-scale upflow anaerobic sludge-blanket reactor. Applied and Environmental Microbiology 72 (12), 7540e7547. Derbali, E., Makhlouf, J., Vezina, L.-P., 1998. Biosynthesis of sulfur volatile compounds in broccoli seedlings stored under anaerobic conditions. Postharvest Biology and Technology 13 (3), 191e204. Drotar, A., Burton Jr., G.A., Tavernier, J.E., Fall, R., 1987. Widespread occurrence of bacterial thiol methyltransferases and the biogenic emission of methylated sulfur gases. Applied and Environmental Microbiology 53 (7), 1626e1631. Du, W., 2010. Modeling volatile organic sulfur compounds in anaerobic digestion. Univeristy of Waterloo, Waterloo, Ontario, Canada. Ph.D Thesis. Fedorovich, V., Lens, P., Kalyuzhnyi, S., 2003. Extension of anaerobic digestion model no. 1 with processes of sulfate reduction. Applied Biochemistry and Biotechnology 109, 33e45. Finster, K., Tanimoto, Y., Bak, F., 1992. Fermentation of methanethiol and dimethylsulfide by a newly isolated methanogenic bacterium. Archives of Microbiology 157, 425e430.
Higgins, M.J., Chen, Y.-C., Yarosz, D.P., Murthy, S.N., Maas, N.A., Glindemann, D., Novak, J.T., 2006. Cycling of volatile organic sulfur compounds in anaerobically digested biosolids and its implications for odors. Water Environment Research 78 (3), 243e253. Hullo, M.-F., Auger, S., Soutourina, O., Barzu, O., Yvon, M., 2007. Conversion of methionine to cysteine in Bacillus subtilis and its regulation. Journal of Bacteriology 189 (1), 187e197. Iranpour, R., Cox, H.H., Fan, S., Abkian, V., Kearney, R.J., Haug, R.T., 2005. Short-term and long-term effects of increasing temperatures on the stability and the production of volatile sulfur compounds in full-scale thermophilic anaerobic digesters. Biotechnology & Bioengineering 91 (2), 199e212. Kadota, H., Ishida, Y., 1972. Production of volatile sulfur compounds by microorganisms. Annual Review of Microbiology 26, 127e138. Kalyuzhnyi, S., Fedorovich, V., Lens, P., Pol, L.H., Lettinga, G., 1998. Mathematical modeling as a tool to study population dynamic between sulfate reducing and methanogenic bacteria. Biodegradation 9, 187e199. Kalyuzhnyi, S., Fedorovich, V., 1998. Mathematical modeling of competition between sulfate reduction and methanogenesis in anaerobic reactors. Bioresource Technology 65 (3), 227e242. Kiene, R.P., Oremland, R.S., Catena, A., Miller, L.G., Capone, D.G., 1986. Metabolism of reduced methylated sulfur compounds in anaerobic sediments and by a pure culture of an estuarine methanogen. Applied and Environmental Microbiology 52 (5), 1037e1045. Kiene, R.P., Visscher, P.T., 1987. Production and fate of methylated sulfur compounds from methionine and dimethylsulfoniopropionate in anoxic salt marsh sediments. Applied and Environmental Microbiology 53 (10), 2426e2434. Lomans, B.P., Op den Camp, H.J.M., Pol, A., Van der Drift, C., Vogels, G.D., 1999. Role of methanogens and other bacteria in degradation of dimethyl sulfide and methanethiol in anoxic freshwater sediments. Applied and Environmental Microbiology 65 (5), 2116e2121. Lomans, B.P., Pol, A., Op den Camp, H.J.M., 2002. Microbial cycling of volatile organic sulfur compounds in anoxic environments. Water Science and Technology 45 (10), 55e60. Pianotti, R., Lachette, S., Dills, S., 1986. Desulfurization of cysteine and methionine by Fusobacterium nucleatum. Journal of Dental Research 65 (6), 913e917. Smet, E., Langenhove, H.V., 1998. Abatement of volatile organic sulfur compounds in odorous emissions from the bioindustry. Biodegradation 9, 273e284. Sreekumar, R., Al-Attabi, Z., Deeth, H.C., Turner, M.S., 2009. Volatile sulfur compounds produced by probiotic bacteria in the presence of cysteine or methionine. Letters in Applied Microbiology 48 (6), 777e782. Thauer, R.K., 1998. Biochemistry of methanogenesis: a tribute to Marjory Stephenson. Mcrobiology 144, 2377e2406. Van Leerdam, R.C., de Bok, F.A.M., Lomans, B.P., Stams, A.J.M., Lens, P.N.L., Janssen, A.J.H., 2006. Volatile organic sulfur compounds in anaerobic sludge and sediments: biodegradation and toxicity. 25 (12), 3101e3109. Visser, A. 1995. The Anaerobic Treatment of Sulfate Containing Wastewater. Ph. D. thesis, Wageningen Agriculture University, Netherlands. Wheeldon, I., Caners, C., Karan, K., Peppley, B., 2007. Utilization of biogas generated from Ontario wastewater treatment plants in solid oxide fuel cell systems: a process modeling study. International Journal of Green Energy 4 (2), 221e231. Zinder, S.H., Brock, T.D., 1978. Methane, carbon dioxide, and hydrogen sulfide production from the terminal methiol group of methionine by the anaerobic lake sediments. Applied and Environmental Microbiology 35 (2), 344e352.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 4 7 e5 4 8
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Comment
Comment on “Sorption of emerging trace organic compounds onto wastewater sludge solids” by J. Stevens-Garmon, J.E. Drewes, S.J. Khan, J.A. McDonald and E.R.V. Dickenson [Water Research 45 (2011) 3417e3426] Karl J. Ottmar* Ottmar Environmental Science and Engineering, 118 Tidal Drive, Newport News, VA 23606, USA
article info Article history: Received 28 June 2011 Received in revised form 7 November 2011 Accepted 10 November 2011 Available online 19 November 2011
The recently-published work by Stevens-Garmon et al. (2011) provides valuable new data along with several important insights pertaining to the transport of several emerging contaminants in wastewater systems. It is encouraging to see agreement with earlier research (Ottmar et al., 2010) along with the intriguing findings in regards to correlations between the distribution coefficients (Kd) and the n-octanol/water partitioning coefficients (Kow). A closer examination of the information present in the article and supplementary information does raise a couple of issues, though. The first concern is in regard to the potential for residual contamination by the target compound on the biosolids. As stated in the methods section, the biosolids were centrifuged and then re-suspended in ultra pure water. They were then mixed for 5 min and centrifuged for 15 min. This process was conducted a total of three times as part of washing the biosolids. The methods section then lists an additional
washing procedure used after lyophilization and oven drying, although, this step was based on reducing the total organic carbon (TOC) to 10 mg/L, and the washing times are not listed. A potential question arises when this 1-h (plus additional rinsing) time period is compared with some of the kinetic tests found in the supplementary information. For some of the compounds (atenolol, dilantin, and enalapril as examples), the time to reach equilibrium appears to be much longer than 1 h, whereas for some of the others (amitriptyline, clozapine, verapamil, omeprazole, etc.), the data is not available, raising the question about those compounds as well. If the rate of desorption is comparable to the rate of sorption, then the possibility exists that the biosolids might still be contaminated with these compounds. One option to eliminate some of this ambiguity in terms of how fast the compounds reach equilibrium would be to fit the data to a kinetic model (and then list those coefficients). Either
DOI of original article: 10.1016/j.watres.2011.03.056. * Tel.: þ1 434 242 7042. E-mail address:
[email protected]. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.11.032
548
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 4 7 e5 4 8
a simple kinetic model or the two-site equilibrium/kinetic model would work for this purpose. A second part of the question is due to the fact that the sorbed concentrations were determined by measuring the change in the aqueous phase, and those values could potentially be skewed by smaller decrease in aqueous concentration. While the authors’ procedure to wash the solids is probably sufficient to ensure that any residual contamination desorbed, it may be prudent to also analyze the solids in future studies as a way to reduce the uncertainty. The second concern is in regards to the sorption isotherms of the positively-charged compounds. In the discussion, other sorption mechanisms beyond partitioning were alluded to, although not necessarily evaluated. If a mechanism closer to adsorption were to be occurring for the positively-charged compounds, then other laboratory experiments in the future could be helpful to bear this out (isotherms at different temperatures, testing for competition, isotherms over a larger concentration range). While that level of analysis would require additional experimental work and was probably
beyond the scope of this study, another alternative exists that could provide further insight. This would entail evaluating the isotherm data using both the Freundlich model and the linear model. The methodology of Kinniburgh (1986) can then be used to compare the two models to see which would provide a better fit.
references
Kinniburgh, D.G., 1986. General purpose adsorption isotherms. Environmental Science and Technology 20, 895e904. Ottmar, K.J., Colosi, L.M., Smith, J.A., 2010. Sorption of statin pharmaceuticals to wastewater-treatment biosolids, terrestrial soils, and freshwater sediment. Journal of Environmental Engineering 136, 256e264. Stevens-Garmon, J., Drewes, J.E., Khan, S.J., McDonald, J.A., Dickenson, E.R.V., 2011. Sorption of emerging trace organic contaminants onto wastewater sludge solids. Water Research 45, 3417e3426.