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 Jean-Claude Block Université H. Poincaré, Nancy I France David Dixon University of Melbourne Australia Hiroaki Furumai University of Tokyo Japan 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 Maria Reis Universidade Nova de Lisboa/FCT Portugal
Damien Batstone University of Queensland Australia Nicolas Bernet INRA France Guang-Hao Chen Hong Kong University of Science & Technology China Rob Eldridge Glen Waverley Australia 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 Peter Hillis AECOM Australia
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 Vayalam Venugopalan BARC Facilities India
Hong-Ying Hu Tsinghua University China
Hang-Shik Shin Korea Advanced Institute of Science and Technology Korea
Jae-Hong Kim Georgia Institute of Technology USA
Thomas Ternes Bundesanstalt für Gewässerkunde Germany
Günter Langergraber University of Natural Resources and Applied Life Sciences, Vienna (BOKU) Austria
Stefan Wuertz University of California, Davis USA
Shang-Lien Lo National Taiwan University Taiwan
Eduardo von Sperling Federal University of Minas Gerais Brazil Julie Zilles University of Illinois at Urbana-Champaign USA Anastasios Zouboulis Aristotle University of Thessaloniki Greece
Editorial Office
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 ) 1 e1 0
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Corrosion in drinking water pipes: The importance of green rusts a, Joanna Swietlik *, Urszula Raczyk-Stanisławiak a, Paweł Piszora b, Jacek Nawrocki a a b
, Poland Department of Water Treatment Technology, Faculty of Chemistry, Adam Mickiewicz University, ul. Drzymały 24, 60-613 Poznan , Poland Department of Chemistry of Materials, Faculty of Chemistry, Adam Mickiewicz University, Grunwaldzka 6, 60-780 Poznan
article info
abstract
Article history:
Complex crystallographic composition of the corrosion products is studied by diffraction
Received 29 April 2011
methods and results obtained after different pre-treatment of samples are compared. The
Received in revised form
green rusts are found to be much more abundant in corrosion scales than it has been
7 October 2011
assumed so far. The characteristic and crystallographic composition of corrosion scales
Accepted 9 October 2011
and deposits suspended in steady waters were analyzed by X-ray diffraction (XRD). The
Available online 25 October 2011
necessity of the examination of corrosion products in the wet conditions is indicated. The drying of the samples before analysis is shown to substantially change the crystallographic
Keywords:
phases originally present in corrosion products. On sample drying the unstable green rusts
Green rust
is converted into more stable phases such as goethite and lepidocrocite, while the content
Corrosion products
of magnetite and siderite decreases. Three types of green rusts in wet materials sampled
Drinking water
from tubercles are identified. Unexpectedly, in almost all corrosion scale samples signifi-
Distribution system
cant amounts of the least stable green rust in chloride form was detected. Analysis of
Steady water
corrosion products suspended in steady water, which remained between tubercles and
Tubercles
possibly in their interiors, revealed complex crystallographic composition of the sampled material. Goethite, lepidocrocite and magnetite as well as low amounts of siderite and quartz were present in all samples. Six different forms of green rusts were identified in the deposits separated from steady waters and the most abundant was carbonate green rust GR(CO2 3 )(I). ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Drinking water distribution systems in several countries have a large amount of iron and steel pipes that are subjected to corrosion causing economic, hydraulic and aesthetic effects, including water leaks, increasing pumping costs, and corrosion products build up. Estimated overall percentage of steel and iron tubes in working networks amounted 53% in Poland (Kwietniewski et al., 2011), 56,6% in USA (AWWA, 2004) and 67,2% in Italy (Veschetti et al., 2010). The highest share of iron
and steel pipes were noted in large cities, e.g. 91% in Warsaw and 93% in Innsbruck (Kwietniewski et al., 2011). Corrosion in steel or cast iron water distribution pipes, is not only responsible for the destruction of pipe material but also for deterioration of potable water quality due to unwanted chemical and biochemical reactions occurring in the distribution systems (McNeill and Edwards, 2001; Edwards, 2004; Hansen et al., 1996; Huck and Gangon, 2004; Ona-Nguema et al., 2002; Chaves, 2005). Corrosion of iron and steel pipes releases iron into distributed water that can re-precipitate
* Corresponding author. Tel.: þ48 618293430; fax: þ48 618293409. E-mail address:
[email protected] (J. Swietlik). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.006
2
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0
forming corrosion scales, also referred to as tubercles (Gerke et al., 2008). Corrosion scales are quite reactive species; they can actively modify physicochemical parameters of water in the distribution system not only by releasing Fe oxyhydroxides (red water) but also by reactions with e.g. chlorinated disinfection by-products (Chun et al., 2005), nitrates (Hansen et al., 1996) or with natural organic matter (NOM) (Nawrocki et al., 2010). Corrosion scales usually consist of a corroded floor, a porous core containing solid and often also fluid, a hard shell layer and surface layer, present on the top of shell-like layer (Baylis, 1926; Baylis, 1953; Sarin et al., 2004a,b; Gerke et al., 2008; Ray et al., 2010). The core of tubercles may be composed of a mixture of iron oxyhydroxides such as goethite and lepidocrocite, green rusts, magnetite, hematite, ferrous hydroxide, ferric hydroxide and siderite (Tuovinnen et al., 1980; Sontheimer et al., 1981; Sarin et al., 2001, 2004a,b). The hard shell layer is mainly composed of magnetite and goethite while the surface layer is more heterogeneous and can be composed of goethite, lepidocrocite, ferric hydroxide, silicates, phosphates and carbonates (Sontheimer et al., 1981; Sarin et al., 2001). The structures of corrosion scale have been examined by many methods: elemental analysis _ 1995; Lin et al., 2001; Gerke ´ z, (Rudnicka and Swiderska-Bro et al., 2008), diffraction (Lin et al., 2001; Sarin et al., 2003, 2004a; Tang et al., 2006; Gerke et al., 2008) or microscopic (Sarin et al., 2001; Tang et al., 2006; Gerke et al., 2008). Magnetite, lepidocrocite and goethite have been repeatedly found (Sarin et al., 2001, 2004a) as the main crystallographic forms of corrosion scales. Maghemite, siderite and green rusts (GR) have also been detected in tubercles, however their amounts differ in reports of different authors. The most interesting and unstable fraction of corrosion by-products is made by green rusts (GR). GRs are layered double hydroxides (LDH) composed of positively charged layers and charge balancing anions located in the interlayer region. GRs with brucite-type layers containing Fe(II) and Fe(III) can have sulphate, chloride, carbonate (Refait et al., 2003) as anions and also water filling the interlayer spaces. The crystal structures of GRs depend on the type of interlayer anion (McGill et al., 1976; Hansen, 1989; Evans and Slade, 2006). The anions compete to form GRs but carbonate is considered as forming the most stable green rust e GR(CO2 3 ) (Refait et al., 1997; Reffas et al., 2006). Green rust in sulphate form, the so called GRII, was reported already in 1976 as a part of corrosion scales in drinking water pipe (McGill et al., 1976). The basic research on corrosion has shown that microbial corrosion also leads to the formation of GR(SO2 4 ) or closely coexists with sulphate green rust (Ge´nin et al., 1991, 2002; Refait et al., 2006; Pineou et al., 2008). Microbially-influenced corrosion has been reviewed by Beech and Gaylarde (1999). Green rust in sulphate form is preferentially formed even at the cost of drastic decrease in sulphate concentration in a surrounding solution (Refait et al., 2003). In other words formation of GR(SO2 4 ) may act as a “sulphate pump”. It has been recently shown that the “steady water” surrounding the corrosion scales may sometimes contain fewer sulphates than the water delivered by the treatment plant (Nawrocki et al., 2010). GR(Cl) are considered as less stable and easily oxidizable than the other (carbonate or sulphate) forms of green rusts (Refait et al., 1997). Moreover, hydroxycarbonate green rust has been also shown to be the
main product of bioreduction of lepidocrocite (Ona-Nguema et al., 2002). Thus the question arises: why green rusts have been rather rarely reported as corrosion scales component? Usually the scales are dried before XRD analysis, no serious precautions are taken to avoid atmospheric oxygen, and this fact probably prevents detection of the phases actually existing in the corrosion scales. Details about preparation of the corrosion products for X-ray diffraction (XRD) measurements have been omitted in many papers (McGill et al., 1976; Badan et al., 1991). Analytical results are very similar when measurements have not been preceded by any special sample treatment, and, in most cases, magnetite, lepidocrocite and goethite are the only observed crystal phases (Badan et al., 1991; Frateur et al., 1999; Lin et al., 2001; Sarin et al., 2001, 2003, 2004a,b; Gismelseed et al., 2004). Similar results have been reported for the air dried samples of corrosion scales (Tang et al., 2006; Gerke et al., 2008). It indicates that observation of the intermediate and metastable corrosion products, such as green rust, requires a special sample treatment. The rapid reaction of GRs with atmospheric oxygen makes it difficult to identify these compounds in natural and engineered systems (Gismelseed et al., 2004; Lin et al., 2001; Zhang et al., 2008). Previously it had been expected that detection of GR phases could be possible by XRD simply after covering a flat sample with glycerol (Hansen, 1989; Zegeye et al., 2007a,b) or freeze-drying (Williams and Scherer, 2001). However, detailed research has shown a progressive sample decomposition during measurement, even for samples covered with glycerol (Mazeina et al., 2008). Unfortunately, also the freeze-drying procedure could alter the structure of corrosion products (Greffie´ et al., 2001). Moreover, sample drying, even in inert atmosphere, can result in product decomposition due to removal of the interlayer water. Therefore, even complex sample treatments before XRD analysis cannot guarantee stabilization of the rust sample. During presented study the corrosion products sampled from working distribution networks were carefully protected against contact with oxygen and drying, and transported to the laboratory directly after sampling. The storage of samples in inert atmosphere enabled to reveal of corrosion products’ complex crystallographic composition by diffraction methods. The comparison of wet and not oxidized deposits and suspensions with those usually analyzed and described in literature (i.e. oxidized and dried) show significant and very important differences in corrosion products structure. The main goal of this paper was to show how sample pretreatment can change the results of XRD analysis. Moreover microscopic images of the scales were taken to confirm the existence of various crystals in the tubercles.
2.
Experimental
2.1.
Sample collection
A large distribution system delivering drinking water produced in three different water treatment plants (WTPs) to about 1.6 million citizens was investigated. Table 1 presents full description of treatment trains in all WTPs as well as the number of pipe samples taken from the network.
3
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0
Table 1 e Water treatment plants description, chloride concentration in treated water and number of taken pipe samples. Water source
Water treatment train
Average Cl concentration in treated water [mg L1]
Number of pipe samples
WTP-1
Wisła River (bottom-infiltration)
130.8
9
WTP-2
skie Lake Zegrzyn
23.8
7
WTP-3
Wisła River (bottom-infiltration)
Coagulation with alum, sand filtration and slow sand filtration (with a layer of active carbon), final disinfection with Cl2/ClO2. Preozonation, alum coagulation, sand filtration, final disinfection with Cl2/ClO2. Aeration, sand filtration and final disinfection with Cl2/ClO2
120.0
3
Treatment plant
During the sampling procedure selected pipes were unearthed in a specially reinforced ditch, and next brushed and rinsed with water to remove soil from the surface. Then, the 30e70 cm fragments of 80e120 mm diameter cast iron pipes were cut off and horizontally removed from the network using supportive strips attached to the shoulders of excavator. In the next step, steady water e the black water surrounding and partly filling the tubercles in the pipes, containing suspension of corrosion products and usually retaining between corrosion scales on the bottom of the corroded pipe (Fig. 1) e was collected (Nawrocki et al., 2010). Steady water present on the bottom of the tubing (usually no more than 150 mL) (Nawrocki et al., 2010), was carefully drained to a glass bottle, transported to the laboratory and kept in a refrigerator at 4 C pending the analysis. Next, the deposits present in steady water samples were separated and analyzed in form of wet pulp. Simultaneously, the pipe segments were protected against drying and the contact with oxygen by placement in an argon saturated container (portable glovebox, Witko, Poland) and transported to the laboratory within 1e10 h. Afterwards, the samples of corrosion scales were taken e one part in wet state, in argon saturated 100 mL plastic containers, and the second part left in opened containers for drying at 25 C with oxygen access. The chemical compositions of corrosion by-products, both deposits and tubercles, were
Fig. 1 e Photograph of iron pipe sample (the steady water samples are taken from the hollows located between tubercles at the bottom of pipe fragment).
analyzed by ICP, and their morphology and structural composition were analyzed by SEM, TEM and XRD. To emphasize the complex composition of liquid containing loose deposits the steady water samples were analyzed by IC and ICP (Table 2).
2.2.
X-ray diffraction (XRD)
The rust deposits were taken from different fragments of pipes sampled from working distribution systems. The pipes with tubercular deposits were cut off and quickly transported to a laboratory. The wet samples were transferred to a mortar, homogenized as sampled (wet) and divided into two parts. One part was loaded as a dense slurry into a flat sample holder and covered with kapton foil; while the other part was dried in air. Then X-ray diffraction measurements were taken for both, dried and wet sample in the standard BraggeBrentano geometry on a Bruker D8 diffractometer using CuKa radiation. For the high-resolution synchrotron X-ray diffraction measurements, small quantities of rust-in-water suspension were ground with a pestle in an agate mortar, and introduced as slurry into 0.5 mm diameter glass capillaries. X-ray powder diffraction data were collected on a high-resolution X-ray powder diffractometer (beamline ID31 at ESRF) selecting Xrays of 0.41274(6) A wavelength from the white undulator source. Data were collected for half an hour, normalized against monitor counts and detector efficiencies, and rebound into steps of 2q ¼ 0.001 . The samples were introduced into glass capillaries to avoid degradation of the airsensitive phases, and the capillaries were rotated during measurement to minimize the effect of preferential orientation. Rietveld refinement was performed using the Topas package (Bruker AXS, 2003). The starting structure models were taken from the following papers: magnetite (Fe3O4) (Fleet, 1981); goethite (a-FeOOH) (Verdonck et al., 1982); lepidocrocite (g-FeOOH) (Ewing, 1935); green rust II with SO2 4 anion, denoted as GR2_SO4 (Fe6(OH)12(SO4)(H2O)8) (Simon et al., 2003); green rust I with Cl anion, denoted as GR1_Cl (Fe6(OH)10Cl(H2O)3) (Allmann and Donnay, 1969); siderite (FeCO3) (Graf, 1961); green rust I with CO2 3 anion, denoted as GR1_CO3 (Fe6(OH)12(CO3)(H2O)3) (Aissa et al., 2006); quartz (SiO2) (Glinnemann et al., 1992); akagene´ite (b-FeOOH) (Stahl et al., 2003); chukanovite (Fe2(CO3)(OH)2) (Pekov et al., 2007). For each crystal phase, the atomic displacement parameters
4
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 e1 0
Table 2 e The results characterizing the quality of steady waters. Parameter
Location 1
DOC [mg L1] pH Fe [mg L1] Mn [mg L1] Al [mg L1] Pb [mg L1] Zn [mg L1] K [mg L1] Na [mg L1] Mg [mg L1] Ca [mg L1] 1 NHþ 4 [mg L ] Ptot [mg L1] Alkalinity [mval L1] Cl [mg L1] F [mg L1] Br [mg L1] 1 NO 3 [mg L ] 1 SO2 4 [mg L ] Formate [mg L1] Acetate [mg L1] Oxalate [mg L1]
WTP-1 9.399 6.06 360 8.6 2.5 e 0.6 4.9 39 14 94 e 0.77 0.60 857.5 0.660 1.37 0.00 43.00 0.066 0.052 0.378
Location 2
Location 3
33.307 6.25 1120 8.0 e e e 4.4 40.3 16.1 170 e 0.036 e 1057 e 4.26 0.11 140.53 e e 1.519
3.828 5.84 960 3.5 0.06 e e 3.8 23.0 12.6 170 e 0.00 e 1808.2 0.470 8.10 0.00 267.99 0.323 1.656 0.129
Location 4 WTP-3 3.350 5.81 nm nm nm nm nm nm nm nm nm nm nm e 664.4 0.425 1.11 1.23 55.44 0.193 1.219 0.033
Location 5
Location 6
Location 7
3.100 5.91 nm nm nm nm nm nm nm nm nm nm nm e 847.8 0.421 2.07 1.28 90.84 0.295 0.949 0.111
WTP-2 5.348 6.85 0.4 1.4 e e 0.002 4.3 17 11.3 107 e e 0.30 159.2 0.174 0.17 0.00 132.69 0.068 0.034 0.030
10.503 6.26 250 48 13 0.05 0.3 5.4 16.7 27 240 30 1.7 e 1346.6 0.552 1.75 0.27 80.61 0.044 e 0.421
nm e Not measured due to small sample volume.
were constrained to be equal for all atoms because of the multiphase character of the sample. The peak shape profile parameters were well fitted with a pseudo-Voigt function for the LaB6 standard and as such fixed for all phases of the corrosion sample. Background was corrected using a Chebyshev polynomial function.
2.3.
DOC measurement
DOC was analyzed by means of a TOC 1030 system (I.O. Analytical, USA). Before analysis all samples were filtered through 0.45 mm filters (Fisherbrand, Fisher Scientific). The detection limit (MDL) was 0.01 mg C L1, while RSD of the method was 15. The measurements were carried out using a UVeVis-scanning spectrophotometer (UV-2401PC) from Shimadzu. The decomposition of Al polymers by the Ferron reagent was measured in 1 cm cuvettes at l ¼ 368 nm within a time span of 40e3600 s. The kinetic curves were analyzed according to the method of Scho¨nherr et al. (1983, 1987) by computer-aided calculation. The absorbance data of the time curves were converted into the logarithm of non-reacted Al. The concentrations of monomeric Al and polymeric Al were determined from the intersections of the tangents at t ¼ 0 of the initial and the final periods of the converted time curves and described in this text as Ala and Alpoly, respectively. The concentration of Alb was calculated from the difference between Alpoly and monomeric Ala. A comparison with 27Al NMR data showed that Alb measured by the Ferron method is equitable to Al13 (Scho¨nherr et al., 1983; Parker and Bertsch, 1992a). Alpoly described by Scho¨nherr et al. (1983, 1987) and Bertram et al. (1994) correspond to the Al30 polycations first described by Allouche et al. (2000) and Rowsell and Nazar (2000). The fractions of Al species in a solution determined via Ferron kinetic or 27Al NMR have been assessed to be equivalent (Bertram et al., 1996; Parker and Bertsch, 1992b; Changui et al., 1990).
2.4.
27
Al MAS NMR
A PAClAl30 solution with 15 mM Altot and 0.1 M KCl was prepared by dissolving 0.73 g Locron-S and 1.8 g KCl in 250 mL water at room temperature. Samples were obtained by removing 50 mL from PAClAl30 solution after 1, 3, 6, and 24 h. Subsequently, sample pH was increased to 7 with 0.01 M KOH, and the aggregated flocs were separated from solution by centrifugation at 27.7 103 G and freeze-dried for three days. Precipitated Al flocs were analyzed by solid-state 27Al multiquantum magic-angle spinning nuclear magnetic resonance (27Al MQMAS NMR). All 27Al NMR spectra were obtained on a Bruker Avance 400 WB spectrometer operated at 104.3 MHz with a spinning frequency of 12 kHz. For every sample a pulse
56
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2
of p/3 and a recycle delay of 1 s was used. The free-inductiondecays (FIDs) were Fourier transformed with a line broadening of 30 Hz. Line fitting was performed with Gaussian peaks using the Igor Pro software package. The species at around 7 ppm was assigned to octahedral aluminum (Aloct). The Al concentrations for 61 ppm and at 66e67.6 ppm signals were multiplied by 13 and 15, respectively, to obtain the concentrations for Al13 and Al30. The Al species represented by the peak at 35 ppm is referred to as Ali in this text.
2.5.
Titration experiments
Polyaluminum solutions of 100 mL with total aluminum concentrations of 15 mM and an ionic strength of 0.1 M KCl were titrated with 0.1 M KOH in a 150 mL water jacket vessel at 25 C under continuous magnetic stirring and argon atmosphere. All acidebase titration experiments were carried out by means of a computer-controlled titration device (736 GP Titrino, Metrohm). Titrations were started after 30 min aging and 10 min pH stability control. Aggregation of Al nanoparticles in the titrated PAClAl30 solution with [Altot] of 15 mM was observed using a laser beam (632.8 nm). Surface-specific charge (s) and protonation state (Z ) of the PACl solution were calculated according to Casey et al. (2005) from titration data under the assumption that most Al is present as Al30 with positive charges of þ18 at fully protonated state: s¼
ðZ þ Smax Þ Alsurf
vB Kw þ Hþ þ ðv0 þ vÞ ½H with Z ¼ ½Al30 tot
(4)
(5)
Z: Protonation number; Smax: total charge at fully protonated state; Alsurf: total number of surface Al atoms; v0: initial volume of Al30 in solution (mL); v: volume of the added base (mL); B: concentration of the added base (M); ½Al30 tot : total Al30 concentration as calculated for each titration step (mol/L). For arsenic removal experiments, 1 mL of the PACl solution was replaced by 1 mL of neutral arsenic stock solution (5 mM As) once the desired pH was held constant for 10 min, resulting in an arsenic concentration of 50 mM in the vessel. For kinetic titration experiments, the pH was kept constant for up to four hours and triplicates were taken after 5, 15, 30, 60, 120, 180 and 240 min. Samples were centrifuged at 3.22 103 G and filtered through 0.45 mm filters (Infochroma AG). Above pH 6, filters with 0.2 mm led to instant clogging. In order to use the same separation technique for all experiments it was decided to use 0.45 mm filters.
2.6.
Coprecipitation experiments
Coprecipitation experiments were conducted in 330 mL polyethylene terephthalate bottles with 97e98.9 mL aliquots of synthetic groundwater. A volume of 1 mL of 27, 133, and 347 mmol/L arsenite or arsenate solution was added to yield arsenic concentrations of 0.27, 1.33 and 3.47 mmol/L. The coprecipitation process was initiated by adding PAClAl30 solution from Locron-S stock solutions (5.1 mM, 51.2 mM, and 512.2 mM Altot) to yield Altot concentrations of
0.01e10.3 mM. The bottles were placed in a reclined position on a shaker with horizontal movement at 250 rpm for 30 min at room temperature, and the pH was allowed to equilibrate. After 4 h of settling, the pH was measured and the supernatant solution was filtered through 0.45 mm nylon filters and the filtrate was analyzed for As concentrations and cations. To investigate the arsenic density on Al nanocluster, experiments were carried out in 50 mL sterile polyropylene tubes with 42.5 mL synthetic water at pH 7.5 0.1. To each batch, 5 mL of As(III) or As(V) stock solutions (adjusted to pH 7.6) were added to achieve initial arsenic concentrations of 0.3e66.7 mmol/L. Finally, 2.5 mL of a Locron stock solution (5.1 mM Altot) was added to yield total Al concentrations of 255 mmol/L.
2.7.
Field tests
Arsenic removal has been tested on natural groundwater from reducing Pannonian aquifers in three artesian wells from the towns Avram Iancu, Ciumeghiu and Sepreus (R112, R113, and R164, notation after Rowland et al., 2011) in the counties of Bihor and Arad in SW-Romania ca. 100 km south of Oradea, and from three pumped wells in the Nea Triglia geothermal area of Chalkidiki prefecture, Northern Greece, identified as KL59, KL103, and Pilot. In Romania, artesian wells have been selected for high arsenic concentrations. 330 mL PET bottles were filled with 100 mL groundwater, and the PAClAl30 solution was added in two different concentrations to yield an As:Altot molar ratio of 0.15 and 0.015. The bottle was vigorously shaken for 5 min and left for settling in an upright position over night before the supernatant solution was filtered (0.45 mm). The samples were taken before and after treatment, immediately acidified by adding 2% 1 M HCl, and stored in the dark at 4 C until analysis by HG AFS. In the geothermal area of Nea Triglia, Greece, water was collected from each well in 4 L containers, which were rinsed three times with distilled water before use and filled up to 3.2 L. Locron-S was added in two different concentrations: 0.22 g/L ([Altot] ¼ 1.1 mM) and 0.62 g/L ([Altot] ¼ 3.2 mM). The waters were vigorously shaken three times and left to react for four hours. Samples were collected before and after treatment by filtering through 0.45 mm filters. For ICPeMS analysis 20 mL of sample was acidified with 1% HNO3 and stored at 4 C. For As(III) and As(tot) analysis 15 mL of water was acidified with 250 ml 1 M HNO3 and stored at 4 C in a brown PET bottle until analysis with HG AFS and ICPeMS.
3.
Results and discussion
3.1.
Characterization of PAClAl30
3.1.1.
Acidebase properties of PAClAl30
The chemical behavior of PACl in solution is crucial to understand the removal processes and the best conditions of coagulant application. The initial pH of the PAClAl30 solution is very similar to that of an Al30 solution with an Al30 concentration of 25 mM. Increasing the pH from 4.7 to 6.7 leads to a loss of 18 protons (Z ¼ 18, Fig. 1a). Aggregation of Al nanoparticles in a PAClAl30 solution with [Altot] of 15 mM was
57
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2
stable afterward (Fig. 2). The initial content of monomeric Al (Almono, Fig. 2) decreased within the entire time of the experiment (five days) from 7.8% to below 1%. However, Al13 increased its share from 13.7% to 25.4% (Fig. 2), partially by polymerization of monomeric Al, and partially from structural transformation of larger aggregates.
3.1.3.
Al floc analysis with
27
Al NMR
27
Fig. 1 e Titration curves of a 1 h altered PAClAl30 solution with total aluminum concentrations of 5.1 mM (corresponding to 392 mM Al13 or 170 mM Al30) in comparison with Al13 (data from Furrer et al., 1992) and Al30 (data from Casey et al., 2005). a) Z Value. The dashed lines indicate the point of complete deprotonation of Al13 and Al30, b) Surface-specific charge. The titration until pH 8 was carried out in 63 min. The pHPZC of the PAClAl30 solution is at 6.7 and corresponds well with the pHPZC of the 25 mM Al30 solution.
observed with a laser (l ¼ 632.8 nm) at pH 5.5 (Z-Value ¼ 4). The pH of the point of zero charge (pHPZC) of the PAClAl30 solution at pH 6.7 correlates with the pHPZC of a solution with lower Al13 and Al30 concentrations, and the obtained values are similar at a surface-specific charge smaller than 0.1 (Fig. 1b). The deprotonation of the main compound of the used PACl, Al30, is known to occur in a broad pH window and takes place at terminal water ligands (Casey et al., 2005). The presented results indicate that the PAClAl30 solution deprotonates over the same large pH window than Al30 (4.7e6.7). Strongest flocculation of the aluminum was found above pHPZC 6.7, where the electrostatic repulsion was smallest. The processes of aluminum coagulation, floc formation and the consequent efficiency of precipitation and liquidesolid separation are enhanced with increasing pH.
3.1.2.
Kinetic analysis with Ferron
Al13 and Al30 solutions are known to change their Al composition with time (Casey et al., 2001), and the formation of monomeric Al could lead to a reduction of As binding sites. The kinetic analysis of Al species in PAClAl30 solutions with the Ferron method showed that the content of Al30 slightly decreased from 78% to 74% within the first 48 h but remained
Solid-state Al NMR analysis of Al flocs from a PAClAl30 solution with 15 mM [Altot] showed one broad peak representing octahedral aluminum (Aloct) centered at ca. 7 ppm (Fig. A.1, supplemental material). Two contributions were distinguished for an asymmetrical peak representing Al(O)4: one at 61 ppm (Al13 ε-Keggin) and another at 66e67.6 ppm. The signal at 66e67.6 ppm falls between the reported chemical shifts for Al30 (68e72 ppm, Allouche et al., 2000) and the Al13 dKeggin (65.5 ppm). Comparison with precipitates of pure Al30 solutions showed the same chemical shift for Al30. It therefore corresponded to an Al30 polymer that was slightly deformed by aggregation. Al13 remained between 15% and 23% without notable trend (Fig. 2). The average of the Al30 share over 24 h was 48 3%, while octahedral aluminum not linked to Al13 and Al30 species varied from 27% to 40%. A small peak at 35 ppm representing 0.4e0.8% of total Al, referred to in this text as Ali (Fig. 2), may be attributed to an intermediate octahedral Al species that was formed during the transformation of Al30. Compared with the Al species distribution in the aqueous solution, about 20e30% of Al30 was transformed into
Fig. 2 e The share of Al species from [Altot] [ 15 mM in % in a freshly prepared aqueous PAClAl30 solution (empty symbols) and in formed Al flocs (filled symbols). Al speciation of the PAClAl30 solution was determined by the Ferron method. The indicated mole fractions Al30 (squares), Al13 (upward triangles) and Almono (empty circles) for the PACl solution are derived from the measured concentrations for Alpoly, Alb, and Ala after Scho¨nherr et al. (1983, 1987). Al flocs from a PAClAl30 solution with 0.1 M KCl were analyzed by solid-state 27Al MAS NMR. Aloct (filled circles) corresponds to the octahedral aluminum species not present in the structure of Al13 or Al30, Ali (downward triangles) is an intermediate, octahedral aluminum species.
58
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2
octahedrally coordinated Al clusters during aggregation. However, Al13 contents in the PAClAl30 solution were similar to Al13 contents in the precipitated Al flocs. Therefore, aggregation processes did not affect the Al13 content (Fig. 2). In total, two thirds of the total aluminum in the aggregates were represented by the Al nanoclusters Al13 and Al30.
3.2.
Arsenic removal with PAClAl30
The removal of arsenic from aqueous solutions with PAClAl30 includes two processes, which can occur simultaneously or subsequently: (i) the chemical interaction of As species with 2 the polynuclear Al. The As(V) species H2AsO 4 and HAsO4 and the As(III) species H3AsO3 form dissolved AleAs complexes already in the acidic pH range, where no aggregation and precipitation of polynuclear Al can be observed. (ii) The deprotonation of polynuclear Al complexes is dependent on pH: Al nanoclusters fully deprotonate and aggregate strongly at near-neutral or higher pH values. The actual removal of arsenic from solution results as a consequence of aggregation and precipitation of the nanoclusters that might include relevant amounts of bound arsenic. Since the As(V) species have a higher binding affinity for aluminum atoms than the As(III) species (Manning and Goldberg, 1997) the former are removed more efficiently than the latter.
3.2.1.
The effect of pH
The pH value is a key parameter controlling the deprotonation state and the aggregation of polynuclear Al species in PACl. Strong aggregation of Al clusters at pH > pHPZC leads to a pronounced removal of total aluminum from solution (Fig. 3). As(V) was completely removed at pH 7 and 8, whereas As(III) removal reached its maximum of 80% at pH 8 (Fig. 3). Although optimum removal pH of both species was achieved at pH 7e8, the interaction of As(III) and As(V) with Al nanoclusters is different. As(V) was removed in the same proportion than Al over the entire pH range within the margins of error (Fig. 3). Hence, As(V) is uniformly distributed over all Al surface binding sites, including those of dissolved Al species at pH 5, 6 and 6.5 (Fig. 3), and As(V) forms soluble complexes with Al nanoclusters. Hence, a covalent binding by ligand-exchange reactions is taking place independently of the charge of Al clusters. The simultaneous elimination of As(V) with Al was consistent with the precipitation of the As(V)-Al complexes. As(III) removal at all pH values was significantly lower than As(V) removal. This indicates a weaker affinity between As(III) and Al nanoclusters. As shown in Fig. 3, strong flocculation leads to enhanced As(III) removal at pH 7 and 8. This shows that Al aggregation is the most important process for the removal of both arsenic species with PAClAl30. Hence, the pH needs to be kept at the optimum for the removal of Al nanoclusters. Kinetic investigation over 4 h reaction time showed no strong variation of arsenic removal (Fig. 3). Therefore, the reaction of As with Al nanoclusters was completed after at most 5 min. Speciation measurements showed no oxidation of As(III). At optimal pH, 99% of Al was removed by precipitation of Al flocs and filtration. At the total added Al concentration of 405 mg/L this means that 4 0.6 mg/L remained in solution.
Fig. 3 e Removal of As(III), As(V) and Al in dependence of pH. As(V) was added as Na2HAsO4 and As(III) was added as NaAsO2 in the concentration of 45 mM with samples taken in triplicates after target pH was reached. Error bars indicate the standard deviation of measurements over 5, 15, 30, 60, 120, 180, and 240 h. Locron-S solutions with [Altot] [ 15 mM were altered for 30 min and had an ionic strength of 0.1 M KCl. Equilibration pH at t [ 0 was 4.8. For As(V), the five titrations with target pH 5, 6, 6.5, 7, and 8 required 0.53, 4.56, 8.67, 8.09, and 9.47 mL 0.1 M KOH and reached the target pH values after 3.9, 29.7, 55.9, 18, and 15.7 min, respectively. The five As(III) titrations with target pH 5, 6, 6.5, 7, and 8 required 0.39, 3.82, 8.09, 9.47, and 10.8 mL 0.1 M KOH and reached the target pH values after 2.7, 13.8, 52, 18, and 15 min. As(V) and As(III) removal is increasing with higher pH and increasing aggregation of the PACl. Arsenate is removed by 100% above pHPZC of PAClAl30 at 6.7 (0.22 As/coagulated Al), maximum arsenite removal of 80% is achieved at pH 8.
Hence, the maximum contaminant level for aluminum in drinking water of 0.2 mg/L suggested as indicator parameter by the EC Drinking Water Directive 98/83/EC was exceeded by a factor of 20. For practical application at small-scale the Al removal process therefore needs to be optimized by using smaller filters or centrifugation.
3.2.2.
The effect of coagulant dosage in water applications
Al30 is more acidic than Al13 due to the presence of acidic hH2O functional groups (Rustad, 2005). To understand how the application of PACl with high Al30 content affects the water chemistry and the arsenic removal efficiency in natural well water, we used simulated groundwater with the chemical composition of the Pannonian Basin (Table 1) at initial pH of 8 0.4. Independent from the initial arsenic concentration, 98.6e99.4% As(V) were removed with total aluminum concentration of 1e6 mM (Fig. 4b). Aluminum concentrations below 1 mM showed a scattered Al removal and consequently a scattered As removal. Above 6 mmol Al/L the pH dropped to 6.5 or lower. Titration experiments revealed that low pH reduced aggregation and precipitation of aluminum clusters due to charge repulsion (Fig. 3), and this was confirmed by Al removal data at high initial Al concentrations with low and
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2
Fig. 4 e Effect of increasing aluminum concentrations on a) equilibrium pH, b) arsenate removal with and without Si as 0.34 mM Na2SiO3, the insert zooms to low Al concentrations, and c) silica removal in synthetic water (composition see Table 1). [As(V)]initial [ 0.3 mmol/L, 1.3 mmol/L and 3.47 mmol/L. The pH decreases with increasing aluminum concentration from pH 8 at low Al concentrations to pH 6.8 at high Al concentrations. The dotted line in a) indicates the pHPZC of PAClAl30, above which Al aggregation and removal is at maximum.
high pH for reference (Fig. 4c): highest Al removal was achieved when the pH is above pHPZC, and under these conditions As(V) is removed by >90%. As(III) was removed by 35% at most (Fig. 5b), but it improved in general with increasing Locron-S addition due to the increase in available surface, as long as the pH is not too low to prevent the formation of big aggregates. As(III) removal in titration experiments is about twice as efficient as in batches with synthetic water where pH was at equilibrium (Fig. 5a). PAClAl30 solutions with high Al concentration were more acidic but also showed more coagulation and aggregation when brought to pH 7e8 than solutions with smaller Al
59
Fig. 5 e Effect of increasing aluminum concentrations on a) equilibrium pH, b) arsenite removal with and without Si as 0.34 mM Na2SiO3 and c) silica removal in synthetic water (composition see Table 1). [As(III)]initial [ 0.3 mmol/L, 1.3 mmol/L and 3.47 mmol/L. The pH decreased with increasing aluminum concentration from pH 8 at low Al concentrations to pH 6.8 at high Al concentrations. The two lines in b) show the average removal with and without Si. As(III) removal is enhanced by the presence of Si.
concentrations. Therefore, As(III) can be removed by up to 80% with high Al concentration at optimum pH conditions (Fig. 3). The adsorption of As(V) on Al was investigated by increasing initial arsenic concentrations to 66.7 mmol/L in a solution of 250 mM Altot in water of the same composition than used in batch experiments. The maximum adsorption density for As(V) was determined as 0.2 M/M Al. This is a 60% increase compared to alum coagulation (0.12 M As(V)/Al) reported by Edwards (1994). The As(III) adsorption density is lower by a factor of 100 (Fig. 6).
60
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2
smaller filter pores is a better alternative. Water treatment facilities in the US using Al2(SO4)3 salts reported similar remaining aluminum concentrations ranging from 0.01 to 2.7 mg/L, with an overall average of 0.16 mg/L after treatment (WHO, 1998).
3.3.
Fig. 6 e Removal of As(III) and As(V) from synthetic water (composition see Table 1) by PAClAl30 (250 mmol/L Altot) with increasing addition of arsenic concentrations (0.3e66.7 mmol/L) at pH 7.5 ± 0.1.
3.2.3.
The effect of silica
Activated silica is used as a coagulation aid for aluminum and ferric-based treatment. It forms negatively charged colloids that increase the size of aluminum hydroxides during coagulation. Water of the Pannonian Basin already contained 9.5 mg Si/L in average (Table 1). We conducted batch experiments with and without adding silica to synthetic water and found initial silica concentrations were reduced after water treatment by up to 50% with 3 mM Altot, while at the lowest and highest PAClAl30 addition no Si was removed (Figs. 4c and 5c). Hence, silica removal depended on the effectiveness of aluminum coagulation and separation. Fan et al. (2003) found an increasing As(V) removal by polyaluminum sulfate and aluminum sulfate with Si dosages of 4e10 mg Si/L. We have no evidence for As(V) removal being effected by the presence of Si, because As(V) was already removed by 98e100% without silica. However, As(III) removal ranged between 10% and 35% with silica and between 0% and 15% without silica (Fig. 5b), unless the pH was at or above 7. This suggests that the presence of Si is favorable to the removal of As(III), either by enhancing the coagulation of aluminum or by forming complexes with Al and As(III). Aluminum was removed by nearly 100% in batch experiments (Figs. 4c and 5c). However, the concentration of Al remaining in solution ranged between 0.03 and 0.8 mg/L with an average of 0.35 mg/L for waters containing no Si concentrations and 0e1.1 mg/L at Si concentrations of 9.5 mg/L with an average value that is slightly below the drinking water limit of 0.2 mg/L suggested by the WHO for drinking water (Fig. A.2, supplemental material). For high and low Al concentrations, the separation of Al from the aqueous phase is scattered and can be as low as 60% due to concentration and pH effects as discussed above (Fig. 4c). The removal of Al can be enhanced by better separation methods than used in this study, e.g. smaller filter pores and ultracentrifugation. Centrifugation is neither cost-effective nor practical for large-scale field operations. It was observed during this study that filters smaller than 0.45 mm clogged easily when filtering Al solutions above pH 6. Therefore, using several filter steps from 0.45 mm to
Field validation
In Romania, groundwater from the artesian wells R112, R113, and R164 has As concentrations in the range of 74e185 mg/L, mainly present as As(III). Samples R112 and R164 had similar arsenic concentrations (80 mg/L and 74.5 mg/L), while R113 groundwater contains 185 mg/L (Table A.1, supplemental material). The three wells generally differed in their water composition, but were within the average groundwater chemistry of the area (group 1, Rowland et al., 2011). The well R113 with the highest As concentration had the highest concentration of TOC (6.1 mg/L), and the groundwater with the lowest As concentration (R 164) had the lowest TOC concentration (0.49 mg/L). The enhanced release of arsenic due to competition with TOC on sorption sites has been proposed for Hungarian groundwater by Varsa´nyi and Kova´cs (2006). Fig. 7a shows that 33e38% of arsenic was removed from As(III)-rich water with an As-to-Al ratio of 0.14, and 12.6e24.1% were removed with an As-to-Al ratio of 0.014. In this study at three well sites, TOC concentrations of up to 6 mg/L seemed to have no effect on As(III) removal. H3AsO3 did not change its oxidation state during treatment (Table A.1, supplemental material). Arsenic removal was tested on water of the geothermal area of Chalkidiki from two open tanks KL59 and KL103 and from a freshly pumped well (Pilot) with initial arsenic concentrations of 193.2 mg/L, 2293 mg/L, and 168.8 mg/L, respectively, mainly present as As(V). The results showed that initial As(V) concentrations of 170 mg/L were reduced to below the drinking water guide value with 1 mM Altot, and are further decreased to less than 5 mg/L with 3.2 mM. High As(V) concentrations of up to 2300 mg/L can be lowered to 10.8 mg/L with the addition of 3.2 mM Altot (Fig. 7b), resulting in a molar ratio of 0.01 M As(V)/M Al. The As removal data from these field trials in two different areas confirm laboratory findings despite different water compositions. For all investigated wells, measured aluminum concentrations were below the detection limit of 0.1 mg/L after the settling of Al flocs and filtration (0.45 mm) of the supernatant solution. These data are in contrast with the results from batch experiments and a result of the reduction of boundary effects during up scaling. Initial silica content was in the range or one third above the concentrations used in lab experiments. For As(III)-rich Romanian groundwater Si was removed by ca. 50% after treatment, while in the case of As(V)rich Greek groundwater Si removal was up to 30% (Table A.1, supplemental material). Manganese concentration was reduced by 25e50%. Other metals present in the waters like copper and iron were eliminated substantially by 40% and 93e99%, respectively. Trace elements like uranium and vanadium showed a tendency to be reduced (Table A.1, supplemental material). The initial pH of the Romanian and Greek groundwater was within the optimal range for removal. During treatment with PAClAl30 the pH decreased by 0.2e1
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 5 3 e6 2
61
processes and removal potential. The protonation state of Al30 governs the formation of Al flocs, and consequently the removal of both arsenic species, to occur between pH 6.5 and 8. This is a common pH range for natural groundwater and therefore no pH adjustment has to be undertaken for precipitation to take place. Analysis of the PAClAl30 solution and the precipitate obtained from a commercial coagulant powder verified that polyaluminum clusters remain dominant during flocculation and precipitation. Highly concentrated PAClAl30 (>5 mM) solutions result in pH ranges that are not favorable for flocculation and precipitation () with intermittent vibration (tC: 60 min and a: 50%).
indispensable to ensure both an efficient and an economic operation.
3.3.
Operation flux and critical flux
3.3.1.
Effect of operation flux on membrane fouling
Flux is considered as one of the most important factors affecting membrane fouling in MBRs (Chang et al., 2002). Its appropriate selection is crucial to maintain the fouling rate at a satisfactory level during long-term operations. A series of filtrations was performed to investigate the effect of operational flux on filtration performance. The performances of both the vibrated filtration and the aerated filtration were compared, as shown in Fig. 7. The flux was varied between 14 and 30 L/m2 h, and the MMV system operated in an intermittent mode, with a tC of 4 min, a PV of 8 W and an a of 50%.
Fig. 6 e Effect of intermittent cycle duration: filtration with vibration cycle of (-) 120 min, (>) 24 min and (:) 4 min (a: 50%).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2
As expected, a faster rate of fouling was found at higher fluxes for both systems. The filtration resistance at the corresponding fluxes was always found to be significantly lower for the vibrating module. At J ¼ 30 L/m2 h, a sharp rise of filtration resistance was observed for both the aerated and the vibrated module. However, the MMV system experienced a much lower fouling at the flux ranges of 14e26 L/m2 h. These results suggest that the MMV system could ensure higher operational fluxes compared to conventional submerged MBR systems.
3.3.2.
Effect of vibration on critical flux
Membrane fouling, in general, is managed by operating the system below its CF (Le-Clech et al., 2006). The CF is broadly defined as the maximum flux value, at which (theoretically) no particle deposition on the membrane surface occurs. This critical value depends on several factors such as the feed condition, membrane properties, hydrodynamics and operation conditions (Wu et al., 2008). The enhanced shear upon vibration can facilitate the increase of CF. Its correlation was evaluated by measuring CF values at different PVs (Fig. 8). The experiment was carried out at a ¼ 100% and at a constant vibration frequency of 50 Hz. The PV was adjusted by varying the vibration amplitude. Fig. 8 indicates that the higher the PV, the higher the CF, in accordance with the literature (Genkin et al., 2006; Beier et al., 2006; Altaee et al., 2010). For example, a CF value of 46 L/m2 h was obtained at a PV of 15.4 W, which is about 3 times higher than the CF measured for the similar feed and membrane in a conventional aerated lab-scale MBR (Bilad et al., 2011a). The PV and CF were found to be proportional in the range of the studied PV. The intercept with the y-axis in Fig. 8 gives the CF for the non-vibrating operation. Since the MMV system is operated at a fixed frequency, PV is proportional here to the applied vibration amplitude, and that is in line with earlier findings for the VHFM system (Beier et al., 2006).
3.4. Long-term filtration, multiple membranes operation and energy consumption The applicability of the MMV system to control membrane fouling in short-term filtration duration has been proven in
69
the previous sections. However, a filtration test over an extended time frame is indispensable to provide more convincing results. In the following experiments, the activated sludge filtration was studied from two different aspects: (1) examining the long-term filtration resistance profile and (2) investigating the effect of multiple membranes in the MMV system to allow evaluation of the energy consumption.
3.4.1.
Long-term filtration
Since the membrane used is different from the one used in Sections 3.2.3 and 3.3, preliminary experiments were performed to select the optimum values for tC, a and PV. In order to better represent the full-scale operation, the filtration was now performed with a relaxation time included in the intermittent filtration. The choice of filtration cycle duration and ratio was also based on a preliminary experiment in which both of these parameters were varied. The results of the aforementioned preliminary experiments are provided as Supplementary material. For the long-term filtration, the filtration was operated in a 5 min cycle that consisted of 4.5 min of filtration and 0.5 min of relaxation. The experiment was performed in two sequential runs. Initially, 5 PEK membranes were run in parallel. Two membranes were operated in the aerated zone, and 3 membranes were operated with vibration. The distance between the membranes in the MMV system was about 5 mm. The operational parameters J, PV, tC and a for the MMV were set at 16 L/m2 h, 6.4 W, 5 min and 50%, respectively. The applied flux was selected as the flux generally applied for the particular membrane in full-scale applications, and the PV was obtained as the result of a preliminary test. PV was set to be low enough to reduce energy consumption, but high enough to provide an acceptable fouling control. Fig. 9 shows the profile of the filtration resistances during the long-term filtration. After seven days of operation, fouling was found to be more severe for all modules in the vibrated system compared to the ones in the aerated system. This is an obvious contradiction with all the previous results, but can be explained by the arrangement of the membranes. The strongest resistance increase was observed in the case of the vibrated membrane in the second position (i.e. situated
Fig. 7 e Effect of operational flux on filtration resistance in (a) aerated system and (b) MMV system.
70
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2
process, but also suggest the importance of adequate design and arrangement of the membranes in one module.
3.4.2.
Fig. 8 e CF determined at different PVs. The shaded area represents optional conditions at which filtration is enhanced due to the additional shear caused by the membrane vibration.
between the two others), suggesting an inappropriate distance between the membranes. Apparently, the membranes in the MMV system were situated so near to each other that the liquid between the membranes moved in-phase and almost became stagnant, moving together with the membrane. To confirm this hypothesis, the filtration was stopped, the second (middle) vibrated membrane was omitted from the reactor, and the filtration was continued. The remaining membranes were chemically cleaned prior to the filtration re-start. Fig. 9 clearly shows that the two vibrating membranes (now with a distance of 10 mm in between) performed better in terms of fouling than the aerated ones throughout the 15 extra days of operation, even though membrane 3 showed a jump on days 13e16 which cannot be explained. These results not only confirm the efficacy of the MMV system in a long-term filtration
Fig. 9 e The profile of filtration resistance during the longterm filtration. Membranes 1e3 were operated in the MMV system, and membranes 4e5 operated in a normal submerged aerated MBR.
Multiple membrane operation and energy consumption
The reduction of energy consumption associated with fouling control is currently one of the main objectives in MBR research. The energy consumption of submerged MBRs is several times higher than that of conventional activated sludge processes (Cornel et al., 2003), and it mainly comes from the energy associated with the coarse bubble aeration for fouling control (Gander et al., 2000). The use of MMV system might offer a promising alternative as a new approach to control fouling in the MBRs. In most shear-enhanced filtration systems, the energy consumption (ED, kWh/m3) is dominated by the energy that is consumed by the vibration engine. Therefore, the energy consumption associated with the MMV system was monitored during this particular test. However, since the ED is calculated based on the volume of permeate, the scale of the plant becomes very significant, mostly favoring large-scale applications. To evaluate the ED of the MMV system, the filtration with multiple membranes was conducted. The MMV system was loaded with up to 6 membranes, to check if there were any changes in filtration performance when the number of membranes in the module increased. Six filtration runs with activated sludge were performed with the MMV system, with each time a different numbers of membranes attached to the module. One additional filtration with six membranes in one module was also performed without vibration for comparison. The filtration parameters were set at J of 16 L/m2 h and PV of 6.4 W, similar to the values used in the long-term test. The profile of the resistance of filtration with different numbers of membranes on the MMV system is shown in Fig. 10. The filtration resistances are given as the average values and the deviations are represented by the shaded area. The results suggest that the lab-scale MMV system could
Fig. 10 e The profile of the filtration resistance for the filtration with different number of membranes mounted on the MMV system.
71
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2
Table 2 e Energy demand of lab-scale MBR and comparison with literature data. Reactor (membrane) Lab-scale MBR (KUBOTA flat sheet) Pilot-scale MBR (KUBOTA flat sheet) Full-scale MBR (Zenon hollow fibre) Full-scale MBR (Zenon hollow fibre) Full-scale MBR (hollow fibre)c
ED (kWh/m3)
J (L /m2 h1)
a
12.12 2.03b 6.06 4.88 0.64 < 0.60 1.07
16 19 25 23 26 20
A (m2)
Reference
0.016 0.096 16
Present study Fenu et al. (2010)
10,160 12,130 63,366
Fatone et al. (2007) Verrecht et al. (2010) Gil et al. (2010)
a Calculated from PV with one vibrating membrane in the reactor. b Calculated from PV with six vibrating modules in the reactor. c Theoretical setup for cost sensitivity analysis.
sustain with at least six membranes (each has 0.016 m2 effective area). The addition of up to six membranes did not significantly affect the filtration performance. This result is in line with the new generation of VSEP system, where increasing the membrane area does not significantly affect the ED (Jaffrin, 2008). The addition of more membranes to the MMV system was not feasible in the current setup, due to the limited space available inside the lab-scale reactor tank. The available information on energy consumption of full- or pilot-scale MBRs in scientific literature is scarce. Table 2 contains some related data of selected publications from the last 5 years and furthermore includes ED data from the labscale MMV system. The ED associated with the MMV system was calculated by using a J, PV and a of 16 L/m2 h, 6.4 W and 50%, respectively, similar to the ones used in the long-term test. Table 2 also confirms that the ED associated with coarse bubble aeration is strongly affected by plant scale. The ED of the pilot-scale MBR (Fenu et al., 2010) is almost 10 times the one of the full-scale ones. This should be considered when analyzing the lab-scale MMV system data. Calculating the ED using six membranes (2.03 kWh/m3) gives a 6 times smaller value than with one membrane (12.12 kWh/m3). Nevertheless, this value is much lower than the ED of a pilot-scale MBR that operates with a 160 time larger membrane area, suggesting a rather economic design of the system, despite being far from optimized yet. It is worth noting that the ED of the lab-scale MMV system is about 3.5 times higher than the ED of the best performing full-scale MBR listed in Table 2. However, direct comparison of these data is not entirely reliable, since the ED a lab-scale setup is not of economical scale and the feed and sludge characteristics of an MBR have a serious influence on the filtration performance. From these comparisons, it can thus be expected that an optimised MMV system (frequency, amplitude, vibration cycle, etc) may lead to a significant cost reduction for fouling control in MBRs. With even membrane moving in one direction and odd membrane in the other, more compact, but still efficient modules could possibly be realized.
conventional MBR processes in terms of realisable flux and fouling control. Significant improvement of CF was obtained due to the enhanced shear at the liquidemembrane interface. The filtration was found sensitive to several operation factors such as the vibration parameters (e.g., vibration power and cycles) and the applied flux. The long-term experiments confirmed the efficacy of the MMV system, but also suggested the importance of an appropriate membrane arrangement in the MBR in the module. The energy demand of vibration, resulting in the highest of all the MBR costs, was found practically constant when the number of modules mounted in the MMV system was increased from one up to six, while increasing treated water volumes 6-fold. The MMV-aided filtration, after process optimisation, is expected to lead to significant cost less membrane area to be installed when taking advantage of the higher CF and energy (especially when expending the number of membranes per module) reduction in (up-scaled) MBRs. This novel membrane fouling limitation method seems very promising in MBRs, but also for the currently progressing anaerobic MBRs where coarse air bubbling is not an option, and possibly also other fouling sensitive ultrafiltration and nanofiltration operations, like algae harvesting.
Acknowledgements The authors gratefully acknowledge the financial support provided by K.U. Leuven (CECAT excellence, GOA and IDO financings), by the Flemish Government (Methusalem funding) and by the Federal Government (IAP grant). Special thanks to Waterleau for providing the sludge seed and giving technical support at start-up of the lab-scale MBR; Wim Ruttens (Intellitech) for his very useful help in the system development; and Toray Membrane Europe for providing the A4 size membrane element samples.
Abbreviations and symbols 4.
Conclusions
Innovative magnetically induced membrane vibration proved very promising in a lab-scale MBR treating synthetic wastewater treatment. Results of both the filtration and the CF measurements showed clear advantages of this system over
A CF ED J JV
membrane surface area (m2) (normalised) critical flux (L/m2 h) energy demand (kWh/m3) permeate flux (L/m2 h) permeate flow velocity (m/s)
72
L MBR MLSS MMV NF PV PEK PVDFT RF RM RT t tC TMP SVI UF DV a h
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 6 3 e7 2
membrane permeability (L/m2 h Pa) membrane bioreactor mixed liquor suspended solid (g/L) magnetically induced membrane vibration nanofiltration vibration power (W) polyethylene (Kubota) polyvinylidene fluoride (Toray) fouling resistance (m1) intrinsic embrane resistance (m1) filtration resistance (m1) filtration time vibration cycle time (min) trans-membrane pressure (Pa) sludge volume index (mL/g) ultrafiltration permeate volume (L) intermittent vibration fraction (%) dynamic viscosity of permeate (Pa s)
Appendix A. Supplementary data Supplementary data associated with this article can be found in the online version at doi:10.1016/j.watres.2011.10.026
references
Altaee, A., Al-Rawajfeh, A.E., Baek, Y.J., 2010. Application of vibratory system to improve the critical flux in submerged hollow fiber MF process. Separation Science and Technology 45 (1), 28e34. American Public Health Association/American Water Works Association (APHA)/Water Environment Federation, 1992. Standard Methods for the Examination of Water and Wastewater, 18th ed. APHA, Washington DC, USA. Beier, S.P., 2008. Dynamic Microfiltration e Critical Flux and Macromollecular Transmission. Ph.D. dissertation, Technical University of Denmark. Beier, S.P., Jonsson, G., 2007. Separation of enzymes and yeast cells with a vibrating hollow fiber membrane module. Separation and Purification Technology 53 (1), 111e118. Beier, S.P., Jonsson, G., 2009. A vibrating membrane bioreactor (VMBR): macromolecular transmission e influence of extracellular polymeric substances. Chemical Engineering Science 64 (7), 1436e1444. Beier, S.P., Guerra, M., Garde, A., Jonsson, G., 2006. Dynamic microfiltration with a vibrating hollow fiber membrane module: filtration of yeast suspensions. Journal of Membrane Science 281 (1e2), 281e287. Bilad, M.R., Declerck, P., Piasecka, A., Vanysacker, L., Yan, X., Vankelecom, I.F.J., 2011a. Development and validation of a high-throughput membrane bioreactor (HT-MBR). Journal of Membrane Science. doi:10.1016/j.memsci.2011.05.052. Bilad, M.R., Declerck, P., Piasecka, A., Vanysacker, L., Yan, X., Vankelecom, I.F.J., 2011b. Treatment of molasses wastewater in a membrane bioreactor: influence of membrane pore size. Separation and Purification Technology 78 (2), 105e112. Bilad, M.R., Westbroek, P., Vankelecom, I.F.J., 2011c. Assessment and optimization of electrospun nanofiber-membranes in a membrane bioreactor (MBR). Journal of Membrane Science 380, 181e191.
Chang, I.-S., Le-Clech, P., Jefferson, B., Judd, S.J., 2002. Membrane fouling in membrane bioreactors for wastewater treatment. Journal of Environmental Engineering 128 (11), 1018e1029. Cornel, P., Wagner, M., Krause, S., 2003. Investigation of oxygen transfer rates in full scale membrane bioreactors. Water Science and Technology 47 (11), 313e319. Cui, Z.F., Chang, S., Fane, A.G., 2003. The use of gas bubbling to enhance membrane processes. Journal of Membrane Science 221, 1e35. Drews, A., 2010. Membrane fouling in membrane bioreactors e characterisation, contradictions, cause and cures. Journal of Membrane Science 363 (1e2), 1e28. Drews, A., Evenblij, H., Rosenberger, S., 2005. Potential and drawbacks of microbiologyemembrane interaction in membrane bioreactors. Environmental Progress 24 (4), 426e433. Fatone, F., Battistoni, P., Pavan, P., Cecchi, F., 2007. Operation and maintenance of full-scale municipal membrane biological reactors: A detailed overview on a case study. Industrial & Engineering Chemistry Research 46 (21), 6688e6695. Fenu, A., Roels, J., Wambecq, T., De Gussem, K., Thoeye, C., De Gueldre, G., Van De Steene, B., 2010. Energy audit of a full scale MBR system. Desalination 262 (1e3), 121e128. Gander, M., Jefferson, B., Judd, S., 2000. Aerobic MBRs for domestic wastewater treatment: a review with cost considerations. Separation and Purification Technology 18, 119e130. Genkin, G., Waite, T.D., Fane, A.G., Chang, S., 2006. The effect of vibration and coagulant addition on the filtration performance of submerged hollow fibre membranes. Journal of Membrane Science 281 (1-2), 726e734. Gil, J.A., Tu´a, L., Rueda, A., Montan˜o, B., Rodrı´guez, M., Prats, D., 2010. Monitoring and analysis of the energy cost of an MBR. Desalination 250 (3), 997e1001. Jaffrin, M.Y., 2008. Dynamic shear-enhanced membrane filtration: A review of rotating disks, rotating membranes and vibrating systems. Journal of Membrane Science 324 (1e2), 7e25. Kola, A., Ye, Y., Stuetz, R., Le-Clech, P., Chen, V., 2011 Transverse vibration as a fouling limitation strategy in membrane bioreactors. Oral presentation presented at ICOM 2011, Amsterdam, The Netherlands, July 23e29, 2011. Le-Clech, P., Jefferson, B., Chang, I.S., Judd, S.J., 2003. Critical flux determination by the flux-step method in a submerged membrane bioreactor. Journal of Membrane Science 227 (1e2), 81e93. Le-Clech, P., Chen, V., Fane, T.A.G., 2006. Fouling in membrane bioreactors used in wastewater treatment. Journal of Membrane Science 284 (1e2), 17e53. Low, S.C., Cheong, K.T., Lim, H.L., 2009. A vibration membrane bioreactor. Desalination and Water Treatment 5, 42e47. Meng, F., Chae, S.-R., Drews, A., Kraume, M., Shin, H.-S., Yang, F., 2009. Recent advances in membrane bioreactors (MBRs): membrane fouling and membrane material. Water Research 43 (6), 1489e1512. Rosenberger, S., Kraume, M., 2003. Filterability of activated sludge in membrane bioreactors. Desalination 151, 195e200. Verrecht, B., Maere, T., Nopens, I., Brepols, C., Judd, S., 2010. The cost of a large-scale hollow fibre MBR. Water Research 44 (18), 5274e5283. Wu, Z., Wang, Z., Huang, S., Mai, S., Yang, C., Wang, X., Zhou, Z., 2008. Effects of various factors on critical flux in submerged membrane bioreactors for municipal wastewater treatment. Separation and Purification Technology 62 (1), 56e63. Yan, X., Gerards, R., Vriens, L., Vankelecom, I.F.J., 2010. Hollow fiber membrane fouling and cleaning in a membrane bioreactor for molasses wastewater treatment. Desalination and Water Treatment 18, 192e197.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Effect of moderate pre-oxidation on the removal of Microcystis aeruginosa by KMnO4eFe(II) process: Significance of the in-situ formed Fe(III) Min Ma a,b, Ruiping Liu a, Huijuan Liu a, Jiuhui Qu a,* a b
Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China Graduate School of Chinese Academy of Sciences, Beijing 100039, China
article info
abstract
Article history:
This study developed a novel KMnO4eFe(II) process to remove the cells of Microcystis aer-
Received 12 May 2011
uginosa, and the mechanisms involved in have been investigated. At KMnO4 doses of
Received in revised form
0e10.0 mM, the KMnO4eFe(II) process showed 23.4e53.3% higher efficiency than the
21 September 2011
KMnO4eFe(III) process did. This was first attributed to the moderate pre-oxidation of M.
Accepted 15 October 2011
aeruginosa by KMnO4, achieved by dosing Fe(II) after a period of pre-oxidation, to cease the
Available online 25 October 2011
further release of intracellular organic matter (IOM) and the degradation of dissolved
Keywords:
process led to high levels and insufficient molecular weight of DOM, inhibiting the
Microcystis aeruginosa
subsequent Fe(III) coagulation. Additionally, Fe(II) contributed to lower levels of the in-situ
KMnO4eFe(II) process
formed MnO2, the reduction product of KMnO4 which adversely affected algae removal by
organic matter (DOM). The extensive exposure of M. aeruginosa to KMnO4 in KMnO4eFe(III)
Moderate pre-oxidation
Fe(III) coagulation. However, the in-situ formed Fe(III), which was derived from the oxida-
The in-situ formed Fe(III)
tion of Fe(II) by KMnO4, in-situ MnO2, and dissolved oxygen, dominated the remarkably high
MnO2
efficiency of KMnO4eFe(II) process with respect to the removal of M. aeruginosa. On one
Dissolved organic matter
hand, in-situ formed Fe(III) had more reactive surface area than pre-formed Fe(III). On the other hand, the continuous introduction of fresh Fe(III) coagulant showed higher efficiency than one-off dosage of coagulant to destabilize M. aeruginosa cells and to increase the flocs size. Moreover, the MnO2 precipitated on algae cell surfaces and contributed to the formation of in-situ formed Fe(III), which may act as bridges to enhance the removal of M. aeruginosa. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
The removal of algae is a well-concerned issue for drinking water treatment plants (DWTPs) that suffer from algae bloom in water sources. The conventional treatment processes, i.e., coagulation/flocculation, sedimentation, and filtration, cannot effectively remove algae cells. This is ascribed to the nature of algae cells, including the low
specific density, high motility, negatively-charged surface, and diverse morphology (Bernhardt and Clasen, 1991; Pieterse and Cloot, 1997; Ma and Liu, 2002). Consequently, the proliferation of algae often has adverse effects of the increased coagulant demand, the clogging and penetrating of filters, and more frequent backwashing (Schmidt et al., 1998; Plummer and Edzwald, 2001). Additionally, the dramatic growth of algae is also associated with other water quality
* Corresponding author. Tel.: þ86 10 62849160; fax: þ86 10 62923558. E-mail address:
[email protected] (J. Qu). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.022
74
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1
problems such as unpleasant taste and odor, toxins, and disinfection by-products (DBPs). There have been proposed several strategies to enhance the removal of algae. Generally, air flotation showed better performance than sedimentation with an overall improvement of 15e20% (Henderson et al., 2008a). The dissolved air flotation (DAF) contributed to the removal efficiency of above 94%, with the highest value of 99.8%, toward different algae species of cyanobacteria Microcystis aeruginosa, green C. vulgaris, and diatom Cyclotella. Ultrafiltration (UF) was another effective approach to remove M. aeruginosa with removal efficiencies of above 90% (Tan et al., 2008). Unfortunately, the process innovation from sedimentation to DAF and UF is usually hindered by the heavy investment, increased operating cost, and complicated operation. It is well preferred to achieve the enhanced removal of algae without large-scale reconstruction. The strategies such as increasing coagulant doses, applying coagulants with better efficiency, and introducing inorganic polymers to aid coagulation may be included. Pre-oxidation is another feasible strategy which has been popular all round the world. The addition of various oxidants [i.e., Cl2, O3, KMnO4, ClO2 and Fe(VI)] prior to coagulation has been shown to significantly improve the algae removal by coagulation or filtration. During alum coagulation at different doses, chlorine at 2 mg/L increased the removal of Chlamydomonas and Euglena gracilis by 85% and 95%, respectively (Steynberg et al., 1996). To achieve the same 50% removal of control algae cultures, the addition of ClO2 (1, 3 and 5 mg/L) and O3 (2.6, 4.6 and 8.1 mg/L) may reduce the alum doses by 13e75% and 41e58%, respectively (Sukenik et al., 1987). KMnO4 at 1.7 mg/L increased the algae removal from 70% to 100% when 40 mg/L of alum was dosed (Chen et al., 2009). Ferrate was also reported to significantly enhance the removal of algae and reduce the coagulant demand, even at very short pre-oxidation time (Ma and Liu, 2002). Among these oxidants, KMnO4 is the most convenient to use in case of urgent algae bloom. However, extensive pre-oxidation causes the lysis of algae cells and the release of intracellular organic matters (IOM), and this should be well considered before applying strong oxidants. In addition to the elevated risk of DBPs formation, the released IOM also inhibited coagulation when they were of high levels or insufficient molecular weight (MW) (Bernhardt et al., 1985). Therefore, an ideal pre-oxidation should be “moderate” to achieve the balance between the two goals of avoiding extensive pre-oxidation of algae and improving the removal of algae. To our best knowledge, rare studies have focused on this theme before. As a strong oxidant, the effects of KMnO4 on algae removal varied with its doses (Petrusevski et al., 1996). The appropriate doses benefited the removal of algae by coagulation whereas the excessive doses showed adverse impacts (Chen et al., 2009). Moreover, the overdose of KMnO4 leads to other problems such as the increase in the residual Mn, color, and turbidity. To avoid the aforementioned side effects of extensive pre-oxidation, a novel KMnO4eFe(II) process, i.e., the introduction of Fe(II) after the pre-oxidation of algae cells by KMnO4 for certain period, is brought forward in this study to achieve moderate oxidation of algae cells. Additionally, Fe(III)
will be freshly formed in this KMnO4eFe(II) process due to the oxidation of Fe(II) by KMnO4 and/or oxygen. This kind of Fe(III) formed in the original place of Fe(II) in aqueous solution is defined as the in-situ formed Fe(III), which is more effective than the pre-formed Fe(III) [e.g., Fe2(SO4)3] in removing contaminants such as arsenic (Guan et al., 2009) and phosphate (Lee et al., 2009). Furthermore, the continuous transformation of Fe(II) to Fe(III) may show different coagulation behaviors from the one-off addition of Fe(III). However, there is a lack of studies which have investigated the mechanisms involved in these processes. This study aimed to: (1) compare the removal efficiencies of M. aeruginosa by the KMnO4eFe(II) and KMnO4eFe(III) processes; (2) investigate the roles of different species, i.e., dissolved organic matter (DOM), in-situ formed MnO2, and insitu formed Fe(III), in the removal of algae by KMnO4eFe(II) process, and demonstrate the significance of the in-situ formed Fe(III); (3) propose the interfacial reactions involved in the removal of algae cells by KMnO4eFe(II) process. A laser particle size analyzer was employed to determine the dynamic growth and size distribution of algae flocs; to date, no studies have employed it for this application.
2.
Materials and methods
2.1.
Materials and reagents
An axenic strain of M. aeruginosa (No. FACHB-905) previously described by (Shen and Song, 2007) was used in this study, and was cultured in BG-11 (Rippka et al., 1979) medium. Full details of growth conditions are presented in the Supporting Information. All chemicals used were reagent-grade and all solutions were prepared with deionized water. Ferrous sulfate (FeSO4) or ferric sulfate [Fe2(SO4)3] solutions were prepared just before experiments. Freshly-formed MnO2 was prepared by mixing KMnO4 and MnCl2 at mol ratio of 2:3, immediately before being dosed to algae suspension; therefore, it could be considered as in-situ formed MnO2. Algogenic organic matter (AOM) was derived from algae cells at the exponential growth phase. Full details about AOM extraction procedure are described in the Supporting Information. Source water was collected from Miyun reservoir located in north China. And its qualities were as follows: turbidity 2.8 NTU, pH 7.7e7.9, temperature 23e25 C, DOC 2.60 mg/L, Ca 43.2 mg/L.
2.2.
Jar tests
M. aeruginosa cultures were harvested at the exponential growth phase and then spiked with source water to obtain the cell density of 1.0 106 cells/mL. The pH of this algae suspension was about 8.4. Jar tests were performed with 300 mL sample in 500 mL beaker and conducted on a programmable jar tester (MY3000-6, MeiYu, China). After the dosage of KMnO4 or MnO2, samples were rapidly mixed at 250 rpm for 5 min FeSO4 or Fe2(SO4)3 was dosed after the preoxidation. The coagulation process consisted of the mixing at 200 rpm for 2 min and 40 rpm for 15 min, consecutively. Then,
75
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1
2.3.
Analytical methods
2.3.1.
Determination of the levels of Kþ, Fe and Mn
An inductively coupled plasma optical emission spectrometer (OPTIMA 2000, PerkinElmer, USA) and/or an inductively coupled plasma mass spectrometer (5000a, Agilent, USA) was used to determine the concentrations of Kþ, Fe and Mn.
2.3.2.
Measurement of residual KMnO4 and UV254 of DOM
Residual KMnO4 and UV254 of DOM were measured with a UV/ vis spectrophotometer (U-3010, Hitachi Co., Japan) at 530 and 254 nm, respectively.
2.3.3.
located downstream of the laser particle size analyzer and run at a rate of 28 rpm d50 served as the representative floc size.
3.
Results and discussion
3.1. Removal of M. aeruginosa by KMnO4eFe(II) and KMnO4eFe(III) processes Fig. 1 compares the removal efficiency of M. aeruginosa between two processes of KMnO4eFe(II) and KMnO4eFe(III) with KMnO4 doses ranging from 0 to 10 mM KMnO4eFe(II) process exhibited significantly higher removal of algae compared to KMnO4eFe(III) process. Without KMnO4, the coagulation by Fe(III) and Fe(II) contributed to algae removal of 26.5% and 49.8%, respectively. KMnO4 at 1.7 mM increased the removal of algae to 36.5% for KMnO4eFe(III) process and to remarkably higher efficiency of 89.7% for KMnO4eFe(II) process. The elevated KMnO4 doses of 2.3e10.0 mM further enhanced the algae removal, which was 91.0e95.6% in KMnO4eFe(II) process and 45.2e62.6% in KMnO4eFe(III) process, respectively.
3.2. Dynamic growth of flocs in KMnO4eFe(II) and KMnO4eFe(III) processes The dynamic growth of flocs in KMnO4eFe(II) and KMnO4eFe(III) processes were compared at KMnO4 doses of 1.7 and 5.0 mM, and the obviously different trends in flocs growth were observed (Fig. 2). The KMnO4eFe(II) process contributed to smaller flocs than KMnO4eFe(III) process initially; however, there showed larger flocs after a critical contact time. Quantitatively, at 1.7 mM of KMnO4, the d50 values in KMnO4eFe(II) process was below 200 mm in the initial 720 s. After that, d50 values increased steadily to as high as 239 mm at 990 s. Comparatively, the d50 values in KMnO4eFe(III) process were 198 and 204 mm at contact time of 720 and 990 s, respectively. Similar trends were observed at 5.0 mM of KMnO4 except for lower intersecting contact time (330 s) and more significant difference between d50 at 990 s in
100
Algae removal (%)
samples were quiescently settled for 30 min. The equivalent molar ratio of KMnO4 to FeSO4 is 1:3; while the molar ratios of KMnO4 to FeSO4 in this study ranged from 120:1 to 20:1. After the settling, samples were siphoned 2 cm below the water surface and divided in two subsamples: the first sample was measured for residual algae by optical density at 680 nm (OD680) with a U-3100 spectrophotometer (Hitachi Co., Japan). The remaining sample was immediately subjected to filtration through a filter (0.45 mm, glass fiber) and then determined for concentrations of Kþ, and total Fe and Mn. In some tests, the level of residual Fe(II) was measured during the coagulation and settling. Details about residual Fe(II) measurement are presented in Supporting Information. For tests measuring the effects of KMnO4 on algae cells, at each pre-determined time, samples withdrawn from the bulk samples were filtered through a filter (0.45 mm, glass fiber) and divided into two subsamples. One subsample was immediately measured for residual KMnO4 and UV254. The other was for the analysis of Kþ concentration. Details about tests studying the effect of pre-oxidation time on MW of DOM in samples containing M. aeruginosa cells are shown in Supporting Information. All experiments were conducted in at least duplicate. Standard error was calculated by software SPSS 13.0. For the tests studying effects of dissolved organic matter (DOM) and the dosage strategy of coagulant on the algae coagulation, the floc growth was measured with a laser particle size analyzer. Jar tests were conducted with 900 mL sample in 1L beaker. For the tests studying the effect of AOM on algae removal, AOM stock solution was added to algae suspension before any treatment. The mixing procedure was: 250 rpm for 2 min and then 50 rpm for 15 min. For the tests of coagulant dosage strategy, certain volumes of Fe(III) were dosed consecutively at several pre-determined times, with the same total dose of Fe(III). The mixing procedure was: 250 rpm for 15 min and then 50 rpm for 5 min.
75
50
25
KMnO 4 -Fe(II)
Size measurement and structure determination of flocs
In some tests, the floc growth during the coagulation processes was studied with a laser particle size analyzer (Malvern Instruments, UK). Jar tests were conducted with 900 mL sample on a programmable jar tester. Algae suspension was drawn from the jar through a latex tube to the sample cell of the laser particle size analyzer; and then back to the jar by a peristaltic pump (BT00-300M, Longer, China). To avoid disturbing floc before the measurement, the pump was
KMnO 4 -Fe(III) 0 0
2
4
6
8
10
12
KMnO4 dose ( M) Fig. 1 e Comparison of the removal of M. aeruginosa at different KMnO4 doses between the KMnO4eFe(II) and KMnO4eFe(III) processes. Cell density: 1.0 3 106 cells/mL. Fe dose: 197.4 mM. Pre-oxidation time: 5 min.
76
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1
400
15 M
a
300 10 % Volume
Floc size d 50 ( m)
KMnO 4 dose: 1.7
200
100
Fe(II) 4 min Fe(II) 17 min Fe(III) 4 min Fe(III) 17 min
a 1.7
M KMnO4
b 5.0
M KMnO4
5
KMnO 4 -Fe(II) KMnO 4 -Fe(III) 0 15
4000
b
M
300 10 % Volume
Floc size d 50 ( m)
KMnO 4 dose: 5.0
200
100
0 0
300
600
900
1200
Time (s) Fig. 2 e Comparison of floc growth during the coagulation processes between the KMnO4eFe(II) and KMnO4eFe(III) processes. Cell density: 1.0 3 106 cells/mL. Fe dose: 197.4 mM. Pre-oxidation time: 5 min.
two processes (148 mm). This may be attributed to the more rapid formation of the in-situ Fe(III) at higher dose of KMnO4. To better understand the difference between KMnO4eFe(II) and KMnO4eFe(III) processes, it is essential to study the change of size distribution of flocs in coagulation. Flocs at 4 and 17 min served as the representative flocs in the floc growth and steady-state phases, respectively. As shown in Fig. 3, for both processes the volume percentage of large flocs increased as coagulation processed. In addition, for KMnO4eFe(II) process there was a substantial decrease in the volume percentage of small flocs whose sizes ranged between 10 and 50 mm (F10e50 mm). The difference between the volume percentage of F10e50 mm at 4 and 17 min was much less significant in the algae coagulation by KMnO4eFe(III).
3.3. Evaluation of different effects on M. aeruginosa removal by KMnO4eFe(II) process 3.3.1.
Contribution of KMnO4 oxidation
KMnO4 in the KMnO4eFe(II) process first served as an oxidant to inactivate algae cells [Eq. (1)]. Fig. 4a illustrates the kinetics of Kþ release at KMnO4 doses from 0.5 to 3.0 mg/L. The release of Kþ, which may indicate the integrity of cell membrane (Peterson et al., 1995), increased with elevated KMnO4 doses and prolonged contact time. Additionally, there was a rapid increase in the level of Kþ within the initial 5e10 min, which was followed by a more gradual increase thereafter. The exposure of M. aeruginosa to 3 mg/L of KMnO4 for 2 min
5
0 10
100
1000
Floc size ( m) Fig. 3 e Particle size distribution for M. aeruginosa flocs after the growth phase in the KMnO4eFe(II) and KMnO4eFe(III) processes. KMnO4 dose: (a) 1.7 mM and (b) 5.0 mM. Fe dose: 197.4 mM. Cell density: 1.0 3 106 cells/mL.
resulted in the Kþ release of as high as 59.6%. There showed consistent trends between Kþ release and KMnO4 decay, and the levels of KMnO4 decreased significantly in the initial 10 min and then decreased slightly (Fig. S1).
KMnO4 þ Cell / Cell* þ MnO2 þ released IOM
(1)
The release of IOM also occurred after the exposure of M. aeruginosa to KMnO4 [Eq. (1)], as indicated by the UV254 analysis (Fig. 4b). The increase in KMnO4 doses resulted in higher UV254 values and thus more significant release of IOM. Interestingly, the levels of IOM decreased to some extent with prolonged contact time, different from that of Kþ release. This is attributed to the adsorption activity of hydrous MnO2 toward DOM (Chen and Yeh, 2005) and to the degradation effect of KMnO4 [Eqs. (2) and (3)].
DOM þ MnO2 / DOM h MnO2
(2)
KMnO4 þ DOM / MnO2 þ DOM*
(3)
The zeta potential of M. aeruginosa suspension was consistently ranging from 2.0 to 4.0 mV before and after KMnO4 oxidation at 3.2e19.0 mM. This trend has also been
77
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1
100
120 KMnO4 (mg/L): 0 2.0
1.0
a
Algae removal (%)
75
60
+
K release (%)
90
0.5 3.0
Fe(II) Fe(III)
a
30
50
25
0
3.0
0
b
b
0.10
1.5 Residual Fe (%)
-1
(cm )
0.09
UV
254
0.08 0.07 0.06
0.3 0.2 0.1 0.0 197.3
0.05 0
10
20
30
Pre-oxidation time (min) Fig. 4 e Effect of KMnO4 pre-oxidation on the variation of (a) KD release and (b) UV254 of DOM. Cell density: 1.0 3 106 cells/mL. The concentrations of KD due to KMnO4 addition were subtracted accordingly. Error bars were not shown in (a) for the consideration of readability.
reported before (Chen and Yeh, 2005). The surface charge of Scenedesmus cells may be an indicator on the changes in the outside cell membrane (Plummer and Edzwald, 2002). It is inferred that KMnO4 altered the outside membrane of M. aeruginosa cells to a small extent.
394.7
Fig. 5 e Comparison of (a) algae removal efficiency and (b) Fe in solution after coagulation between Fe(II) and Fe(III) coagulation processes. Cell density: 1.0 3 106 cells/mL.
The effect of settling time on the removal of M. aeruginosa in Fe(III) and Fe(II) coagulation was also investigated (Fig. 6). Despite of the similar removal efficiencies of algae in two processes after a 30-min settling, cells settled more rapidly by Fe(II) coagulation when settling period was prolonged. After a 60-min settling, Fe(II) coagulation contributed to algae removal of 64.0%, which was as low as 36.0% by Fe(III) coagulation. After that, the algae removal by Fe(II) coagulation increased steadily to as high as 82.3% after 150-min settling,
Contribution of the in-situ formed Fe(III)
Levels of residual Fe were higher in KMnO4eFe(II) process as compared to those in KMnO4eFe(III) process (Fig. S2a). The less participation of Fe in algae removal in KMnO4eFe(II) process indicates the different coagulation behaviors of Fe(II) and Fe(III) toward algae cells. To exclusively investigate the contribution of the in-situ formed Fe(III), this study first compared the removal of M. aeruginosa by Fe(III) and Fe(II) coagulation without KMnO4 (Fig. 5a). The algae removal by Fe(II) coagulation was 10%e41% higher than that by Fe(III) coagulation at different Fe doses from 197.4 to 394.7 mM. However, the levels of residual Fe associated with Fe(II) coagulation was 5e8 times higher than those in Fe(III) coagulation, with more than 98.4% of Fe(II) insitu oxidized to Fe(III) by oxygen accordingly (Fig. 5b). Similar trends were also observed in coagulation with KMnO4 preoxidation (Fig. S2a). The discrepancy between the algae removal and the contents of Fe in the precipitates in those two processes indicates a surprisingly higher efficiency of Fe(II) coagulation toward algae removal.
100 Fe(II) Fe(III) 75 Algae removal (%)
3.3.2.
276.3
Coagulant dose (as Fe) ( M)
50
25
0 0
30
60
90
120
150
Settling time (min) Fig. 6 e Comparison of settling rates between Fe(II) and Fe(III) coagulation processes. Cell density: 1.0 3 106 cells/ mL. Fe dose: 197.4 mM.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1
and only 4.3% of improvement was achieved by Fe(III) coagulation with the settling time increasing from 30 to 150 min. The different coagulation and settling behaviors of flocs between Fe(II) and Fe(III) coagulation is in support of the better performance of the in-situ formed Fe(III) than the pre-formed Fe(III) with respect to the algae removal. In the KMnO4eFe(II) process, the residual KMnO4 and oxygen, as well as the in-situ formed MnO2 which will be discussed in detail later, showed good activities to oxidize Fe(II) to Fe(III) [Eqs. (4e6)]. Fe(II) acted as the source to continuously provide the in-situ Fe(III) (Fig. S3), which was an important contribution to the remarkably higher removal of algae in KMnO4eFe(II) process.
Fe(II) þ KMnO4 / in-situ Fe(III) þ MnO2þ
(4)
Fe(II) þ MnO2 / in-situ Fe(III) þ Mn2
(5)
Fe(II) þ O2 / in-situ Fe(III)
(6)
First, continuously introduction of the in-situ Fe(III) inhibited the adverse impacts of the coagulant aging. A timedependent mechanism inferred by Flynn et al. depicted a three-step hydrolysis process of Fe(III): 1) formation of low MW complexes, consisting of [Fe(OH)]2þ, [Fe(OH)2]þ and [Fe2(OH)2]4þ; 2) formation and aging of polynuclear polymers, such as [Fep(OH)r(H2O)s](3pr)þ or [FepOr(OH)s](3p2rs)þ; and 3) precipitation of Fe(III) oxides (Fe2O3) and hydroxides [Fe(OH)3 and FeO(OH)] (Flynn, 1984). Over these dissolutionprecipitation processes, the structure of iron oxides became more ordered by in-situ rearrangement (Jolivet et al., 2004). The aging of hydrolyzed Fe(III) also increased the particle size and thus decreased the reactive surface area that was accessible for reaction (Vikesland et al., 2007), reducing the reactivity of Fe(III) oxides as a whole. Due to the slower aging of coagulant as a whole, more active surface areas and higher reaction rates can be achieved by Fe(II) coagulation than by Fe(III) coagulation. Furthermore, hydrolysis intermediates of in-situ formed Fe(III) might immediately attach to and bind with the aged iron hydroxide precipitate to achieve positive surfaces. Thus, the continuous provision of in-situ formed Fe(III), rather than the one-off dosing of Fe(III), advanced the effective agglomerate of tiny particles and the subsequent settling thereafter. The presence of algae cells and DOM with large amount of negative functional groups (i.e., eOH, eCOOH) even complicated the roles of in-situ Fe(III) involved in. The hydrolytically formed iron hydroxo complexes and iron oxide hydroxides can form surface complexes with organic anionic polymers (OAP) which contains eOH and eCOOH groups (Bernhardt et al., 1985; Pivokonsky et al., 2006). The pre-formed Fe(III) promptly hydrolyzed to Fe(OH)3 precipitates soon after its being dosed. In contrast, the in-situ formed Fe(III) and its hydrolysis products may be more active to form these complexes than the pre-formed Fe(III) for the aforementioned reasons. And these interesting characteristics enhanced the interactions between algae cells and Fe(III) to improve algae removal. Although it is difficult to raise the exact reactions, the significance of the in-situ formed Fe(III) can be proposed.
Moreover, the heterogeneous oxidation of Fe(II) on the surfaces of MnO2 also played an important role. KMnO4 rapidly decayed and transformed to the in-situ MnO2 after its interaction with algae suspensions [Eqs. (1) and (2)], and MnO2 may precipitate on the surfaces of algae cells to some extent [Eq. (7)] (Chen and Yeh, 2005). Besides the possibly residual KMnO4 and oxygen, the in-situ MnO2 also contributed to the oxidation of Fe(II). This suggestion is supported by the much higher levels of residual Mn in KMnO4eFe(II) process than those in KMnO4eFe(III) process (Fig. S2b). The heterogeneous oxidation of Fe(II) to Fe(III) on the surfaces of tinny MnO2 particles may continuously provide the in-situ formed Fe(III) to destabilize MnO2 particles, algae cells, and the algae cells with precipitated MnO2 [Eqs. (8) and (9)]. Correspondingly, the role of in-situ formed MnO2 in the coagulation was altered, improving the removal of algae. And the role of in-situ formed MnO2 will be discussed in detail later.
Cell þ MnO2 / Cell h MnO2
(7)
Fe(II) þ DOM-MnO2 / in-situ Fe(III) þ DOM þ Mn2þ
(8)
Fe(II) þ Cell-MnO2 / in-situ Fe(III) þ Cell þ Mn2þ
(9)
To further prove the suggested advantage of continuous provision of in-situ Fe(III) over one-off dosage of pre-formed Fe(III), the dynamic growth of flocs under different dosingFe(III) strategies was investigated (Fig. 7).Despite of the same levels of total Fe, the consecutive dosing of Fe(III) contributed to the formation of larger flocs than the one-off dosing-Fe(III) strategy. After slow-mixing, the d50 value in Stategy-I [i.e., 197.4 mM Fe(III) at 0 min] was 95 mm, which increased to 112 mm in Stategy-IV [i.e., 131.6 mM Fe(III) at 0 min and 32.9 mM Fe(III) at 5 and 10 min]. The higher d50 values of 120 and 125 mm were
150 Strategy I Strategy II Strategy III Strategy IV
120 Floc size d 50 ( m)
78
90 60 30 Rapid mixing 0 0
400
800
Slow mixing 1200
Time (s) Fig. 7 e Floc growth in the coagulation with different dosage strategies of Fe(III). (I) 197.4 mM (0 min); (II) 131.6 mM (0 min) D 65.8 mM (5 min); (III) 65.8 mM (0 min) D 65.8 mM (5 min) D 65.8 mM (10 min); and (IV) 131.6 mM (0 min) D 32.9 mM (5 min) D 32.9 mM (10 min). Cell density: 1.0 3 106 cells/mL. Sample volume: 900 mL. Mixing procedure: 250 rpm, 15 min and then 50 rpm, 5 min.
79
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1
observed in Stategy-III [i.e., 131.6 mM Fe(III) at 0 min and 65.8 mM Fe(III) at 5 min] and Stategy-II [i.e., 65.8 mM Fe(III) at 0, 5, and 10 min], respectively.
Contribution of the in-situ formed MnO2
MnO2 may also affect the removal of algae. Fig. 8 compares the effect of KMnO4 and the in-situ formed MnO2 at the same Mn doses on the removal of algae by Fe(III) coagulation. MnO2 showed adverse influences on the algae removal. In the control experiment, Fe(III) coagulation alone removed 23.3% of algae cells. The presence of MnO2 at 0.83e6.67 mM decreased the algae removal to 19.7e21.3%. This might be ascribed to the repulsive forces between the negativelycharged algae cells and MnO2. The isoelectric point of MnO2 was reported to be in the range of 2.8 and 4.5 (Posselt and Anderson, 1968). At pH 8.3 in this study, the negativelycharged MnO2 inhibited the collision and aggregation between algae cells. In contrast, MnO2 at 0.83e6.67 mM improved the algae removal by 2.7e8.0% at pH 3.8 (Fig. S4), which may be attributed to the adsorption of positivelycharged MnO2 onto algae cells and thus the increase in the specific weight of cells (Chen and Yeh, 2005). Different from MnO2, KMnO4 at 0.63e6.67 mM enhanced the removal of algae by 2.0%e30.3%. These different behaviors indicate the significant role of KMnO4 oxidation on the removal of algae cells.
3.3.4.
Effect of DOM on the floc growth
KMnO4 pre-oxidation also contributed to the release of IOM and thus the elevated levels of DOM (Fig. 4b). The impacts of DOM on the dynamic growth of flocs during Fe(III) coagulation varied with its concentration. As shown in Fig. 9, DOM at low levels obviously benefited the formation of larger flocs. The d50 values at 960 s increased from 121 mm to 158 and 212 mm due to the addition of 0.5 and 1 mg/L of DOM. However, DOM at high concentrations of 2 and 3 mg/L negatively decreased d50 to lower values of 130 and 123 mm. There are some previous studies that show the effect of IOM on the coagulation. IOM mainly consists of protein, polysaccharides, and lipids (Henderson et al., 2008b), and these cyanobacterium-derived organics were reported to behave like anionic and non-ionic
60 MnO 2
Algae removal (%)
KMnO 4
AOM addition (mg/L)
Floc size d50 ( m )
3.3.3.
300 0 0.5 1.0 2.0 3.0
200
100
0 0
200
400
600
800
1000
Time (s) Fig. 9 e Effect of DOM on the floc growth during the coagulation process. Cell density: 1.0 3 106 cells/mL. Fe(III) dose: 197.4 mM. Sample volume: 900 mL. Mixing procedure: 250 rpm, 2 min and then 50 rpm, 15 min.
polyelectrolytes, depending on their concentrations and MW (Bernhardt and Clasen, 1991). At low DOM levels, the cyanobacterium-derived AOM species may serve as polymer bridges between algae cells and Fe(III) hydroxides to aid coagulation (Bernhardt et al., 1985). However, the released IOM from M. aeruginosa cells showed a high ratio of protein to DOM (Henderson et al., 2008b), and the eOH and eCOOH groups within cyanobacteria protein can form protein-coagulant complexes with coagulants (Pivokonsky et al., 2006). These interactions inhibited the cross linking and clustering of Fe-hydroxide polymers and increased the coagulant demand accordingly. Thus, DOM showed both positive and adverse effects on the removal of algae by Fe coagulation, depending on the levels of DOM.
3.4. Proposed mechanisms involved in the removal of M. aeruginosa by KMnO4eFe(II) process Fig. 10 illustrates the schematic diagram of the possible mechanisms of KMnO4eFe(II) process on algae removal. KMnO4 pre-oxidation inactivated cells of M. aeruginosa and disturbed their integrity to benefit the algae removal, as indicated by the release of Kþ and IOM. Meanwhile, the MnO2 from KMnO4 reduction may precipitate onto algae surfaces,
45 Mn(VII)
Without Mn
Mn(VII)
1) Inactivation
30
MnOOH Algae
MnOOH Algae
+
0.83
1.67
3.33
6.67
Mn dose ( M) Fig. 8 e Comparison of MnO2 and KMnO4 on the removal of M. aeruginosa by Fe(III) coagulation. Cell density: 1.0 3 106 cells/mL. Fe dose: 197.4 mM. Pre-oxidation time: 5 min.
K , DOM
O2
4) Redox In situ-Fe(III)
5) Hydrolysis
2) Adsorption Mn(II)
3) Release
15
Fe(II)
DOM
Algae
DOM with lower level and higher MW
Moderate pre-oxidation
MnO2
Algae
Mn(II) Algae
Fresh Fe(III) polymer spheres with more active surface areas Continuous formation of In situ-Fe(III)
Fig. 10 e Schematic diagram of the possible mechanisms of KMnO4eFe(II) process on algae removal.
80
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1
even though solution pH was as high as 8.3. It is assumed that the co-existing Ca2þ acted as cation bridges to enable the complexation and adsorption of MnO2 onto the surfaces of algae cells, mainly through the eOH and eCOOH groups on the surfaces of their outside membranes. SEM/EDX analysis demonstrated the involvement of Ca2þ in interactions between MnO2 and algae cells (data not shown). After coagulation with the KMnO4 pre-oxidation at 0.83e13.3 mM, the elemental Ca did exist on the surfaces of algae cells with the atomic ratios of Ca to Mn ranging from 1.5:1 to 4:1. The aggregation of MnO2 onto algae surfaces increased the cell density, accelerated the settling of algae, and enabled the heterogeneous oxidation of Fe(II) to in-situ Fe(III) on algae surfaces. On the other hand, the moderate oxidation of algae cells achieved in KMnO4eFe(II) process played important roles. The subsequent introduction of Fe(II) ceased the inactivation of algae cells by KMnO4 to inhibit further IOM release and DOM degradation, as indicated by the less release of Kþ in KMnO4eFe(II) than that in KMnO4eFe(III) process (Table S1). Additionally, the released IOM in KMnO4eFe(II) process was in lower levels with less fractions of small molecular weight (Fig. S5), and this also indicated beneficial effects of moderate oxidation. Furthermore, at KMnO4 doses of 1.7e10.0 mM, the residual Mn2þ concentrations in KMnO4eFe(II) process were 2e18 times higher than those in KMnO4eFe(III) process (Fig. S2b). Mn2þ can also adsorb onto MnO2 (Forrez et al., 2010) and act as cation bridges to improve algae aggregation [Eqs. (10e12)].
Mn2þ þ MnO2 / MnO2 h Mn2þ
(10)
MnO2 h Mn2þ þ Cell / MnO2 h Mn2þ h Cell
(11)
and Duan, 2001) and thus dominates the flocs growth. In the one-off dosing strategy, Fe(III) immediately precipitates and grows to limited size during rapid-mixing with strong shear forces. In contrast, the continuous transformation of Fe(II) provides fresh Fe(III) and increases the discrete number of clusterecluster bond(s) and the magnitude of cohesive force to form stronger and larger flocs. Additionally, in the presence of soluble Fe, the formation of compact flocs and the precipitation of in-situ formed Fe(III) is in progress simultaneously to benefit particles aggregation. This is far to be well illustrated and needs to be studied to give more clear view in the removal of algae by KMnO4eFe(II) process.
4.
Conclusions
KMnO4eFe(II) process showed better capability toward algae removal than KMnO4eFe(III) process. This is attributed to the combined effects of moderate pre-oxidation and the continuous formation of in-situ Fe(III) in KMnO4eFe(II) process.After a period of pre-oxidation, the positive impacts of KMnO4 had been achieved, which would not increased with further prolonged pre-oxidation time. Thus, the introduction of Fe(II) at this time interval would avoid the extensive oxidation of algae cell and the thus the significant release of IOM, achieving moderate pre-oxidation of algae cells and facilitating the coagulation. Besides, the simultaneously formed in-situ Fe(III) was more effective than the pre-formed Fe(III) with respect to algae removal. This is attributable to that 1) in-situ formed Fe(III) had more reactive surface area and 2) that the continuous introduction of fresh coagulant [i.e. in-situ formed Fe(III)] benefited the floc growth than the one-off dosage of preformed Fe(III).
Acknowledgments MnO2 h Mn
2þ
þ DOM / MnO2 h Mn
2þ
h DOM
(12)
Moreover, the continuous formation of fresh Fe(III) and its hydrolysis products dominated in the remarkable efficiency toward algae removal via aforementioned mechanisms. The fresh Fe(III) hydrolysis products may not only form Fe-DOM complexes, but also attach and bind to the negativelycharged surfaces of algae cells. Additionally, the heterogeneous oxidation of Fe(II) by the MnO2 on algae surfaces, and the in-progressing Fe(III) precipitation achieved strong binding of flocs. The continuous provision of fresh Fe(III) also favored the aggregation of destabilized tiny flocs (i.e., algae cells, Fehydroxide, MnO2, DOM, OAP-Fe complexes) to form larger flocs during flocculation. The size distribution analysis of flocs at 4 and 17 min indicates the shift to the right of the major peak (Fig. 3). This supports the entrapment and/or adsorption of tiny flocs onto the surface of large ones and the continuous floc growth during floc growth phase. The significant finding of the present work is that continuous introduction of fresh coagulates is able to promote the growth of floc. Due to the limited solubility of Fe(III), the amorphous Fe(III) hydroxide plays an essential role (Gregory
This work was supported by the National Basic Research Program of China (Grant 2007CB407301), the National Natural Science Foundation of China (Grant No. 51078345), the Funds for the Creative Research Groups of China (Grant No. 50921064) and CAS Major Projects of Knowledge Innovation Program (kzcxl-yw-06-02).
Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.10.022.
references
Bernhardt, H., Clasen, J., 1991. Flocculation of microorganisms. Journal of Water Supply: Research & Technology - AQUA 40 (2), 76e87. Bernhardt, H., Hoyer, O., Schell, H., Lusse, B., 1985. Reaction mechanisms involved in the influence of algogenic matter on flocculation. Z. Wasser Abwasser Forsch 18 (1), 18e30.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 7 3 e8 1
Chen, J.J., Yeh, H.H., 2005. The mechanisms of potassium permanganate on algae removal. Water Research 39 (18), 4420e4428. Chen, J.J., Yeh, H.H., Tseng, I.C., 2009. Effect of ozone and permanganate on algae coagulation removal - Pilot and bench scale tests. Chemosphere 74 (6), 840e846. Flynn, C.M.J., 1984. Hydrolysis of inorganic iron (III) salts. Chemical Reviews 84 (1), 31e41. Forrez, I., Carballa, M., Verbeken, K., Vanhaecke, L., Schlu¨sener, M., Ternes, T., Boon, N., Verstraete, W., 2010. Diclofenac oxidation by biogenic manganese oxides. Environmental Science & Technology 44 (9), 3449e3454. Gregory, J., Duan, J.M., 2001. Hydrolysing metal salts as coagulants. Pure and Applied Chemistry 73 (12), 2017e2026. Guan, X., Ma, J., Dong, H., Jiang, L., 2009. Removal of arsenic from water: effect of calcium ions on As(III) removal in the KMnO4eFe(II) process. Water Research 43, 5119e5128. Henderson, R., Parsons, S.A., Jefferson, B., 2008a. The impact of algal properties and pre-oxidation on solid-liquid separation of algae. Water Research 42 (8e9), 1827e1845. Henderson, R.K., Baker, A., Parsons, S.A., Jefferson, B., 2008b. Characterisation of algogenic organic matter extracted from cyanobacteria, green algae and diatoms. Water Research 42 (13), 3435e3445. Jolivet, J.P., Chane´ac, C., Tronc, E., 2004. Iron oxide chemistry. From molecular clusters to extended solid networks. Chemical Communications (5), 481e487. Lee, Y., Zimmermann, S.G., Kieu, A.T., Gunten, U.V., 2009. Ferrate (Fe(VI)) application for Municipal Wastewater treatment: a novel process for Simultaneous Micropollutant oxidation and phosphate removal. Environmental Science & Technology 43 (10), 3831e3838. Ma, J., Liu, W., 2002. Effectiveness and mechanism of potassium ferrate(VI) preoxidation for algae removal by coagulation. Water Research 36 (4), 871e878. Peterson, H.G., Hrudey, S.E., Cantin, I.A., Perley, T.R., Kenefick, S. L., 1995. Physiological toxicity, cell membrane damage and the release of dissolved organic carbon and geosmin by aphanizomenon flos-aquae after exposure to water treatment chemicals. Water Research 29 (6), 1515e1523. Petru sevski, B., van Breemen, A.N., Alaerts, G., 1996. Effect of permanganate pre-treatment and coagulation with dual coagulants on algae removal in direct filtration. Journal of
81
Water Supply: Research & Technology - AQUA 45 (5), 316e326. Pieterse, A.J.H., Cloot, A., 1997. Algal cells and coagulation, flocculation and sedimentation process. Water Science and Technology 36 (4), 111e118. Pivokonsky, M., Kloucek, O., Pivokonska, L., 2006. Evaluation of the production, composition and aluminum and iron complexation of algogenic organic matter. Water Research 40 (16), 3045e3052. Plummer, J.D., Edzwald, J.K., 2001. Effect of ozone on algae as precursors for trihalomethane and haloacetic acid production. Environmental Science & Technology 35 (18), 3661e3668. Plummer, J.D., Edzwald, J.K., 2002. Effects of chlorine and ozone on algal cell properties and removal of algae by coagulation. Journal of Water Supply: Research & Technology - AQUA 51 (6), 307e318. Posselt, H.S., Anderson, F.J., 1968. Cation sorption on colloidal hydrous manganese dioxide. Environmental Science & Technology 2 (12), 1087e1093. Rippka, R., Deruelles, J., Waterbury, J.B., Herdman, M., Stanier, R. Y., 1979. Generic assignments, strain histories and properties of pure cultures of cyanobacteria. Journal of General Microbiology 111, 1e61. Schmidt, W., Hambsch, B., Petzoldt, H., 1998. Classification of algogenic organic matter concerning its contribution to the bacterial regrowth potential and by-products formation. Water Science and Technology 37 (2), 91e96. Shen, H., Song, L.R., 2007. Comparative studies on physiological responses to phosphorus in two phenotypes of bloom-forming Microcystis. Hydrobiologia 592, 475e486. Steynberg, M.C., Pieterse, A.J.H., Geldenhuys, J.C., 1996. Improved coagulation and filtration as a result of morphological and behavioural changes due to pre-oxidation. Journal of Water Supply: Research & Technology - AQUA 45 (6), 292e298. Sukenik, A., Teltch, B., Wachs, A.W., Shelef, G., Nir, I., Levanon, D. , 1987. Effect of oxidants on microalgal flocculation. Water Research 21 (5), 533e539. Tan, S., Li, P., Chen, L., 2008. Test research for algae-removing by ultrafiltration. Journal of Water Resources & Water Engineering 19 (4), 73e75. Vikesland, P.J., Heathcock, A.M., Rebodos, R.L., Makus, K.E., 2007. Particle size and aggregation effects on magnetite reactivity toward carbon tetrachloride. Environmental Science & Technology 41 (15), 5277e5283.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Influence of temperature and salinity on Ostreopsis cf. ovata growth and evaluation of toxin content through HR LC-MS and biological assays Laura Pezzolesi a, Franca Guerrini a, Patrizia Ciminiello b, Carmela Dell’Aversano b, Emma Dello Iacovo b, Ernesto Fattorusso b, Martino Forino b, Luciana Tartaglione b, Rossella Pistocchi a,* a b
Centro Interdipartimentale di Ricerca per le Scienze Ambientali, Universita` di Bologna, Via S’Alberto 163, 48123 Ravenna, Italy Dipartimento di Chimica delle Sostanze Naturali, Universita` degli Studi di Napoli “Federico II”, Via D. Montesano 49, 80131 Napoli, Italy
article info
abstract
Article history:
In the Mediterranean Sea, blooms of Ostreopsis cf. ovata and Ostreopsis siamensis have
Received 9 March 2011
become increasingly frequent in the last decade and O. cf. ovata was found to produce
Received in revised form
palytoxin-like compounds (putative palytoxin, ovatoxin-a, -b, -c, -d and -e), a class of
4 October 2011
highly potent toxins. The environmental conditions seem to play a key role in influencing
Accepted 16 October 2011
the abundance of Ostreopsis spp. High cell densities are generally recorded in concomitance
Available online 25 October 2011
with relatively high temperature and salinity and low hydrodynamics conditions. In this
Keywords:
ovata isolate were investigated. The highest growth rates of the Adriatic strain were
study the effects of temperature and salinity on the growth and toxicity of an Adriatic O. cf. Ostreopsis cf. ovata
recorded for cultures grown at 20 C and at salinity values of 36 and 40, in accordance with
Temperature
natural bloom surveys. Toxicity was affected by growth conditions, with the highest toxin
Salinity
content on a per cell basis being measured at 25 C and salinity 32. However, the highest total toxin content on a per litre basis was recorded at 20 C and
Toxicity Ovatoxins Haemolysis assay
salinity 36, since under such conditions the growth yield was the highest. O. cf. ovata had lethal effects on Artemia nauplii and juvenile sea basses, and produced haemolysis of sheep erythrocytes. A comparison between haemolysis neutralization assay and HR LC-MS results showed a good correlation between haemolytic effect and total toxin content measured through HR LC-MS. Considering the increasing need for rapid and sensitive methods to detect palytoxin in natural samples, the haemolytic assay appears a useful method for preliminary quantification of the whole of palytoxin-like compounds in algal extracts. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Massive blooms of the benthic dinoflagellates Ostreopsis spp. are reported worldwide in many tropical and temperate regions (Faust et al., 1996; Vila et al., 2001; Aligizaki and
Nikolaidis, 2006; Chang et al., 2000; Ciminiello et al., 2008; Mangialajo et al., 2011; Rhodes et al., 2011). In the Mediterranean Sea, blooms of O. cf. ovata and Ostreopsis siamensis have been reported since the late ‘70s (Taylor, 1979; AbboudAbi Saab, 1989) but, in the last decade, they have become
* Corresponding author. Tel.: þ39 (0) 544 937376; fax: þ39 (0) 544 937411. E-mail address:
[email protected] (R. Pistocchi). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.029
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2
increasingly frequent and resulted in benthic biocenosis sufferings and human health problems. Ostreopsis spp., typically, proliferate in shallow and sheltered waters, with low hydrodynamism; they form a rustybrown coloured mucilaginous film, which covers reefs, rocks (Bottalico et al., 2002), and soft sediments (Vila et al., 2001) as well as seaweeds (Vila et al., 2001; Bottalico et al., 2002; Aligizaki and Nikolaidis, 2006; Totti et al., 2010), marine angiosperms, and invertebrates (Bianco et al., 2007; Totti et al., 2007). The presence of Ostreopsis spp. in coastal waters may pose a real threat to coastal food web and fishery (Aligizaki et al., 2008). Several marine organisms, in particular sea urchins, have lost their spines and died during blooms of O. cf. ovata or O. siamensis (Grane´li et al., 2008; Sansoni et al., 2003; Shears and Ross, 2009); however, the effects on marine organisms and on ecosystem dynamics are still unknown. Ostreopsis spp. are thought to produce palytoxin (or its analogues) (Taniyama et al., 2003), one of the most potential toxic marine compounds, which acts on the Naþ/Kþ pump converting it into an ionic channel and causing the subsequent depletion of the Kþ ions (Habermann, 1989). This hypothesis was later supported by identification of putative palytoxin as the causative toxin of human poisonings which occurred during O. siamensis blooms (Onuma et al., 1999) and, most importantly, by identification of some palytoxin-like compounds from various Ostreopsis spp. Particularly, ostreocin-D was isolated from O. siamensis and structurally elucidated by NMR (Usami et al., 1995; Ukena et al., 2001) while mascarenotoxins were identified, basing only on MS evidence, as palytoxin-like compounds from Ostreopsis mascarenensis (Lenoir et al., 2004). Putative palytoxin and ovatoxin-a were detected in field and cultured samples of O. cf. ovata, collected along the Ligurian coasts (Italy) (Ciminiello et al., 2006; 2008) as well as in O. cf. ovata cultures from the Adriatic and Tyrrhenian Sea (Guerrini et al., 2010) by liquid chromatographymass spectrometry (LC-MS). Recently, several new ovatoxins, namely ovatoxin-b, -c, -d, and -e, were also detected in an Adriatic O. cf. ovata culture through an in-depth high resolution (HR) LC-MS investigation (Ciminiello et al., 2010). Currently, O. cf. ovata blooms occur each year from June to late October at several sites on the Italian coastline, characterized by different environmental conditions, such as seawater temperature in the range 18e30 C and salinity in the range 30e39 (Pistocchi et al., 2011). However, O. cf. ovata has never been detected in the Northwestern Adriatic sea, at sites located close to the Po river delta, where low salinity values occur and a coast-offshore salinity gradient affecting microphytobenthos distribution in the northern Adriatic Sea was observed (Totti, 2003); this suggests that some environmental conditions play a key role in influencing O. cf. ovata growth and/or its geographical dispersal. Several authors indicated seawater temperature is an important factor affecting cell proliferation (Tognetto et al., 1995; Sansoni et al., 2003; Simoni et al., 2004; Aligizaki and Nikolaidis, 2006; Mangialajo et al., 2008). In most studies (as reviewed by Pistocchi et al., 2011), high temperature values (24e29 C) were associated with increases of Ostreopsis cell number in seawater; however, in the Adriatic (Totti et al., 2010) and Catalan seas (Vila et al., 2001) such positive
83
correlation has not been observed. Recently, the influence of temperature on O. cf. ovata growth and toxicity has been also reported by Grane´li et al. (2011) using a Tyrrhenian isolate from the Ligurian coast: the highest toxicity was found in cultures grown at 20 C, whereas the highest algal biomass was recorded at 30 C. In the present study, we report on in-depth investigation on the effect of some environmental parameters on the growth and toxicity of O. cf. ovata. An Adriatic O. cf. ovata isolate, whose growth and toxin profile had been previously characterized at 20 C and salinity 36 during the exponential and stationary phases (Guerrini et al., 2010), was used. Cultures were grown at different temperature (20, 25 and 30 C) and salinity values (26, 32, 36 and 40); HR LC-MS analyses were carried out to determine their toxin profile, including the recently found ovatoxins (Ciminiello et al., 2010), and to evaluate the total toxin amount released in the extracellular medium during the stationary growth phase. A further object of the present study was to compare the total toxin content of algal extracts measured by HR LC-MS with the results obtained through the haemolysis assay (Riobo´ et al., 2008), with the aim of gaining information on the accuracy of the haemolytic test, a rapid and very sensitive biological assay widely employed for palytoxins detection (Riobo´ et al., 2011). Finally, the toxicity of O. cf. ovata cultures on crustaceans and fish was also investigated using Artemia sp. assay and the ichthyotoxicity test with juvenile sea basses (Dicentrarchus labrax) (IRSA-CNR, 2003) to evaluate the potential O. cf. ovata impact on the other marine organisms.
2.
Materials and methods
2.1.
Cultures of Ostreopsis cf. ovata
O. cf. ovata was isolated using the capillary pipette method (Hoshaw and Rosowski, 1973) from water samples collected along the Adriatic coast of Italy (Marche region, Numana sampling site, strain OOAN0601) in October 2006, in proximity to the seaweeds Cystoseira sp. and Alcidium corallinum. After an initial growth in microplates, cells were cultured at 20 C under a 16:8 h L:D cycle from cool white lamp in natural seawater, at salinity 36, adding macronutrients at a five-fold diluted f/2 concentration (Guillard, 1975) and selenium. In order to evaluate the effect of environmental parameters on growth and toxicity of O. cf. ovata, temperature and salinity experiments were carried out. In the salinity experiment, cultures (at 20 C) were established at salinity 26, 32, 36 and 40 in a thermostatic room, maintaining light irradiance at 100e110 mmol m2 s1. Salinity levels 26, 32 and 36 were obtained by diluting seawater (salinity 38) with deionized water, while salinity 40 was obtained by evaporation of the seawater. In the temperature experiment, cultures (salinity 36) were established at 20, 25 and 30 C in water baths kept in the same thermostatic room, thus light irradiance slightly decreased to 90 mmol m2 s1. Phaeodactylum tricornutum (strain PTN0301 from the North Sea, Holland) was cultured using f/2 medium under the same conditions and used in the experiments either for comparisons or control.
84
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2
Both temperature and salinity experiments were carried out by using, for each condition, 2 series of batch cultures. One was used to evaluate the growth profile and the other the toxin content.
2.1.1.
Evaluation of growth profile
Since the evaluation of the growth profile of O. cf. ovata in batch cultures was complicated by the presence of mucous aggregates, the sampling method developed by Guerrini et al. (2010) was used for counting. For each temperature/salinity level, 15 Erlenmeyer flasks containing 200 ml of culture were grown in parallel; every other day, two out of the initial flasks were treated with HCl to a final concentration of 4 mM. Acid addition dissolved mucous aggregates and homogenous sampling could be performed. After counting, the two acidified flasks were discarded. Cell counts were made following Utermo¨hl method (Hasle, 1978) and specific growth rate (m, day1) was calculated using the following equation: m¼
ln N1 ln N0 t1 t0
where N0 and N1 are cell density values at time t0 and t1. Calculation of cell volume was made with the assumption of ellipsoid shape using the following equation (Sun and Liu, 2003):
V ¼ ðp=6Þ a b c where a ¼ dorsoventral diameter (length), b ¼ width, c ¼ mean anteriorposterior diameter (height).
2.1.2.
Evaluation of toxin content
For each temperature and salinity level, a set of four culturing flasks was set up. Due to limitations in the availability of the equipment, the salinity experiment was carried out in a thermostatic room using 1500 mL flasks, while the temperature experiment was carried out by placing the 800 mL flasks in water baths. Cell counting was carried out on one out of the four flasks as described above. Five replicate counts were collected from one of the four flasks for each treatment and used to determine the cell density and to express toxin content on a per cell basis. Culture collection was carried out during the late stationary growth phase by gravity filtration through GF/F Whatman (0.7 mm) filters at day 21st and 25th for the salinity and temperature experiment, respectively. Cell pellets and growth media for each temperature/salinity level were provided for chemical analysis.
2.2.
Chemical analysis
2.2.1.
Chemicals
All organic solvents were of distilled-in-glass grade (Carlo Erba, Milan, Italy). Water was distilled and passed through a MilliQ water purification system (Millipore Ltd., Bedford, MA, USA). Acetic acid (Laboratory grade) was purchased from Carlo Erba. Analytical standard of palytoxin was purchased from Wako Chemicals GmbH (Neuss, Germany).
2.2.2.
Extraction
Cell pellets and growth media for each temperature/salinity level were extracted separately. For each pellet sample 9 mL of a methanol/water (1:1, v/v) solution was added and the solution sonicated for 30 min in pulse mode, while cooling in ice bath. The mixture was centrifuged at 3000 g for 30 min, the supernatant was decanted and the pellet was washed twice with 9 mL of methanol/water (1:1, v/v). The extracts were combined and the volume was adjusted to 30 mL with extracting solvent. The obtained mixture was analyzed directly by HR LC-MS (5_ml injected). Each growth medium was extracted five times with an equal volume of butanol. The butanol layer was evaporated to dryness, dissolved in 5 mL of methanol/water (1:1, v/v) and analyzed directly by HR LC-MS (5 ml injected). Recovery percentage of the above extraction procedures were estimated to be 98% and 75% for the pellet and growth medium extracts, respectively (Ciminiello et al., 2006).
2.2.3. High resolution liquid chromatography-mass spectrometry (HR LC-MS) High resolution (HR) LC-MS experiments were carried out on an Agilent 1100 LC binary system (Palo Alto, CA, USA) coupled to a hybrid linear ion trap LTQ Orbitrap XL Fourier Transform MS (FTMS) equipped with an ESI ION MAX source (ThermoFisher, San Jose`, CA, USA). Chromatographic separation was accomplished by using a 3 mm gemini C18 (150 2.00 mm) column (Phenomenex, Torrance, CA, USA) maintained at room temperature and eluted at 0.2 mL min1 with water (eluent A) and 95% acetonitrile/water (eluent B), both containing 30 mM acetic acid. A slow gradient elution was used: 20e50% B over 20 min, 50e80% B over 10 min, 80e100% B in 1 min, and hold 5 min. This gradient system allowed a sufficient chromatographic separation of most palytoxin-like compounds (Table 1). HR full MS experiments (positive ions) were acquired in the range m/z 800e1400 at a resolving power of 15,000. The
Table 1 e Molecular formulae (M) of ovatoxins, elemental composition of their relevant A and B moieties and most abundant peaks of [M D 2HeH2O]2D and [M D 2H D K]3D ion clusters for each compound. Toxin Palytoxin Ovatoxin-a Ovatoxin-b Ovatoxin-c Ovatoxin-d Ovatoxin-e
Rt (min)
M
A moiety
B moiety
[M þ 2HeH2O]2þ
[M þ 2H þ K]3þ
10.78 11.45 11.28 10.90 11.07 11.07
C129H223N3O54 C129H223N3O52 C131H227N3O53 C131H227N3O54 C129H223N3O53 C129H223N3O53
C16H28N2O6 C16H28N2O6 C18H32N2O7 C18H32N2O7 C16H28N2O6 C16H28N2O7
C113H195NO48 C113H195NO46 C113H195NO46 C113H195NO47 C113H195NO47 C113H195NO46
1331.7436 1315.7498 1337.7623 1345.7584 1323.7456 1323.7456
906.8167 896.1572 910.8318 916.1628 901.4884 901.4884
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2
following source settings were used in all HR LC-MS experiments: a spray voltage of 4 kV, a capillary temperature of 290 C, a capillary voltage of 22 V, a sheath gas and an auxiliary gas flow of 35 and 1 (arbitrary units). The tube lens voltage was set at 110 V. Due to commercial availability of the only palytoxin standard, quantitative determination of putative palytoxin, ovatoxin-a, -b, -c, -d, and -e in the extracts was carried out by using a calibration curve (triplicate injection) of palytoxin standards at four levels of concentration (25, 12.5, 6.25, and 3.13 ng mL1) and assuming that their molar responses were similar to that of palytoxin. Extracted ion chromatograms (XIC) for palytoxin and each ovatoxins were obtained by selecting the most abundant ion peaks of both [Mþ2HeH2O]2þ and [Mþ2H þ K]3þ ion clusters (Table 1). A mass tolerance of 5 ppm was used.
2.3.
Toxicity assays
2.3.1.
Artemia sp. assay
The assay was carried out according to the short-term test of the IRSA-CNR (2003) method, consisting in a 24 h exposure of Artemia sp. to the potentially toxic sample. 10 nauplii were incubated in 1 mL of sample in a glass tube for 24 h. Firstly, aliquots of a culture grown at 20 C and salinity 36, containing five increasing concentrations of live cells, lysed cells, algal extracts and growth media, were tested in triplicate. Live cell aliquots were sampled during the stationary phase of the culture. Lysed cell aliquots were obtained by sonicating 10 mL of the culture for 3 min. Algal extracts were obtained as reported above and diluted (1:100 to 1:10,000) with seawater. A palytoxin stock solution (12.5 mg mL1) in methanol/water (1:1, v/v) was diluted with seawater and tested in the concentration range 500e10,000 pg mL1. Growth medium aliquots were obtained by filtering 50 mL of the culture through GF/F Whatman (0.7 mm) filters. The toxicity of O. cf. ovata cultures grown at different temperature/salinity conditions was evaluated by Artemia sp. assay, using only live cells. Five different concentration levels of each sample were obtained through dilution with seawater, and were tested in triplicate. The effects on Artemia sp. of sample exposure were evaluated after 24 h by counting the number of dead organisms. Seawater samples, methanol/ water (1:1, v/v) solution (diluted 1:100, v/v with seawater) and f/2 medium at the investigated salinity levels (diluted 1:5 with seawater) were used as control. EC50 values were calculated (see below Section 2.4).
2.3.2.
Haemolytic assay
Haemolytic assay was carried out following the procedure proposed by Bignami (1993) and modified by Riobo´ et al. (2008). The test is based on photometrical determination of haemoglobin released from sheep erythrocytes following exposure to haemolytic compounds. Sheep blood was kindly provided by the Department of Veterinary Public Health and Animal Pathology (University of Bologna). Erythrocytes were separated from plasma by centrifugation (400 g at 10 C for 10 min) and washed twice with a solution containing phosphate buffered saline (PBS) 0.01 M, pH 7.4, bovine serum albumin (BSA), calcium chloride (CaCl2 2H2O) 1 mM and boric
85
acid (H3BO3) 1 mM. Finally, the erythrocytes solution was diluted with PBS at a final concentration of 1.7 108 red cells mL1. According to the reported method (Riobo´ et al., 2008), two blood solutions, one added of ouabain (2.5 mM) and one ouabain-free, were prepared to a final concentration of 1.7 107 erythrocytes. 1 mL of each blood solution was mixed with 1 mL of the sample diluted in PBS (either pellet extract or palytoxin standard previously dissolved in methanol/water (1:1, v/v)) and incubated at 25 C for 20 h. After the incubation, samples were centrifuged at 400 g for 10 min and the supernatant absorptions were measured at 405 nm. Two replicates of algal extract at different concentration levels, control solutions for blanks (PBS buffer and methanol/water (1:1, v/v) in PBS) and total haemolysis sample were prepared for each experiment. Palytoxin standard at seven concentration levels (4e196 pg mL1) were used for generating the calibration curve. Stock solutions of the algal extracts and palytoxin standard used in the haemolytic assay were quantified by HR LC-MS. The haemolytic effects of the algal extracts were expressed either on cell basis (cell mL1) or on toxin content basis (pg mL1). EC50 values obtained by testing the palytoxin standard and the algal extracts were calculated (see below Section 2.4).
2.3.3.
Fish bioassay
Sea basses (D. labrax) employed in the assay were collected from the hatchery Valle Ca’ Zuliani (Pila di Porto Tolle, Rovigo, Italy). After the transfer, they were kept in a 60-70 L aquarium, aerated by a small dispenser (Hailea) and kept at room temperature and salinity 36 for one month. For the experiments, 2 L aerated tanks containing algal culture were used. Three juveniles (5.0 1.0 g) were put into each tank, kept at 20 C, during a 16:8 h lightedark period and observed for 4 days. Two replicates of O. cf. ovata culture grown for 4e6 days at 20 C and salinity 36 were tested at three concentration levels. An equal volume of P. tricornutum culture was used as control. Fish were considered dead when gill opercular movements ceased.
2.4.
Data analysis
The 50% effect concentration (EC50) of each sample for the Artemia sp. and haemolytic assays was estimated by fitting the experimental concentration-response curves to a logistic model: y¼
a x b 1þ EC50
Where: y ¼ endpoint value; x ¼ substance concentration; a ¼ expected endpoint value in absence of toxic effect; b ¼ slope parameter. The parameters of the equation, including the EC50, were estimated using the non-linear regression procedures implemented in Statistica (Statsoft, Tulsa, OK, USA). An independent estimate of EC50 was obtained for each of the experiments. Differences in cell biovolume, EC50 value, and toxins concentration among the samples were tested by using the multivariate analysis-of-variance (ANOVA) test, using
86
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2
Statistica (StatSoft, Tulsa, OK, USA) software. Whenever a significant difference for the main effect was observed (P < 0.05), a NewmaneKeuls test was also performed.
3.
Results
Batch cultures of an Adriatic strain of O. cf. ovata, collected along the Marche coasts of Italy (Numana sampling site) in October 2006, were established in order to evaluate the effect of salinity and temperature on algal growth and toxin profile. Particularly, in the temperature experiment, cultures were set at 20, 25 and 30 C by maintaining salinity at 36 and light irradiance at 90 mmol m2 s1, while in the salinity experiment cultures were established at salinity 26, 32, 36 and 40, by maintaining temperature at 20 C and light irradiance at 100e110 mmol m2 s1.
3.1.
Growth pattern and cell volume
The growth profiles of O. cf. ovata cultures exposed to different salinity and temperature values were analyzed by measuring the cell density every 2e3 days from the beginning of the exponential phase to the end of the stationary phase (Fig. 1A and B). Under the different growth conditions, O. cf. ovata growth rates in the range 0.34e0.49 day1 were observed. For the temperature experiment, during the first 5 days cells grew better at 25 C; at the end of the exponential phase the maximum growth rate of 0.49 day1 was recorded at 20 C, followed by 0.43 and 0.34 day1 at 25 and 30 C, respectively. For the salinity experiment (carried out at 20 C) growth rate was not significantly affected by the salt concentration (0.43e0.47 day1) (ANOVA, P > 0.05). In the stationary phase, the maximum density was 13,000e16,000 cell mL1 at 20 C and intermediate salinities
(32 and 36), while the cell yield dropped to 7500 cell mL1 both at salinity 26 (temperature 20 C; Fig. 1A) and temperature 30 C (salinity 36; Fig. 1B). In the course of the experiments, we noticed that the culture volume played a key role on the final cell yield: decreasing cell densities were obtained as culture volumes increased from 200 mL to 800 mL up to 1500 mL. Another aspect we considered in the salinity and temperature experiments was the cell biometric measurement. It is to be noted that O. cf. ovata cells appeared highly different both in size and in shape, within each cell culture; therefore, a statistically significant cell number (n 50) was used for estimating the mean biovolumes. In both salinity and temperature experiments, a significant difference was observed between cell volumes measured in the exponential and stationary phases (ANOVA, P < 0.001). For the salinity experiment, the highest difference among biovolumes was observed in the exponential phase where a mean value of 22,000 mm3 was reached at the lowest salinity (26) and resulted significantly higher (Post-hoc SNK test, P < 0.001) than those observed at 36 and 40 (14,000 and 13,000 mm3, respectively). An intermediate biovolume mean value was observed at salinity 32 (17,000 mm3). In the stationary phase, cells were more homogenous in size, with cell volumes in the range 28,000e30,000 mm3; however, the value reported at salinity 32 (22,400 mm3) resulted significantly lower than those observed at the other salinity levels (Posthoc SNK test, P < 0.001). For the temperature experiment, in both growth phases cell volumes decreased as temperature increased, with a maximum biovolume of 22,000 mm3 being reached at 20 C (stationary phase), which was significantly higher (Post-hoc SNK test, P < 0.001) than biovolumes measured at 25 and 30 C (16,000 and 15,000 mm3, respectively).
-1
Abundance (cell mL )
3.2. 100000
26 psu 32 psu 36 psu 40 psu
A
10000 1000
100 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
-1
Abundance (cell mL )
Time (day)
B
100000
20°C 25°C 30°C
10000
1000
100 0
2
4
6
8 10 12 14 16 18 20 22 24 26
Time (day) Fig. 1 e Growth pattern of O. cf. ovata cultures exposed to different salinity (A) and temperature (B) conditions.
Determination of toxin content by HR LC-MS
Cell pellets and growth media of O. cf. ovata cultures grown at the different temperature and salinity values were collected during the late stationary growth phase. Samples were extracted separately, and the crude extracts were used to evaluate the toxin profile. HR LC-MS experiments were acquired in full MS mode by using an LC method which allowed chromatographic separation of the major components of the toxin profile. The spectra were acquired in the mass range m/z 800e1400 where each palytoxin-like compound (Table 1) produces bi-charged ions due to [M þ H þ K]2þ, [M þ H þ Na]2þ, and [Mþ2H]2þ, tri-charged ions due to [Mþ2H þ K]3þ and [Mþ2H þ Na]3þ, and a number of ions due to multiple water losses from the [Mþ2H]2þ and [Mþ3H]3þ ions. The presence of putative palytoxin and of all the ovatoxins (ovatoxin-a, -b, -c, -d, and -e) recently identified in O. cf. ovata (Ciminiello et al., 2008; 2010) was highlighted in all the analyzed samples. Significant differences were observed in the total toxin content of different algal extracts (ANOVA, P < 0.001), whereas the relative abundance of individual toxins were quite similar: ovatoxin-a was by far the major component of the toxin profile (47e56% of the total toxin content; Post-hoc SNK test,
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2
Table 2 e Total toxin content (putative palytoxin, ovatoxin-a, -b, -c, -d, and -e) of O. cf. ovata culture pellet and medium extracts, measured by HR LC-MS in both salinity and temperature experiment. Data are expressed as mg per Litre of culture (mg LL1). Cell density (cell LL1) and extracellular release (%) are also reported. Total toxin content (ug L1) Cell L
1
Extracellular release (%)
Pellet
Medium
Total
3,450,333 4,646,333 4,281,333 5,619,000
57 95 76 68
17 14 12 11
74 109 88 80
23 13 14 14
Temperature 9,869,587 20 C 5,581,677 25 C 4,493,377 30 C
155 129 81
25 25 30
180 154 111
14 16 27
Salinity 26 32 36 40
P < 0.001), followed by ovatoxin-b (24e27%), ovatoxin-d and -e (15e18%), ovatoxin-c (4e8%) and putative palytoxin (0.5e3%) on the basis of their decreasing relative abundance. Total toxin content of pellet and medium extracts expressed as mg L1 culture in both salinity and temperature
A
palytoxin ovatoxin-a ovatoxin-b ovatoxin-c ovatoxin-d,-e tot
20
-1
Toxin content (pg cell )
25
10 5
26
32
36 pellet
40
26
32 36 medium
40
Salinity (psu) 25
-1
experiments are reported in Table 2. Toxin contents were significantly higher in the cell pellet than in the corresponding culture medium (Post-hoc SNK test, P < 0.001), resulting in relatively low release percentages (13e16%) in most of the growth conditions applied; however, the release increased up to 23 and 27% under the most unfavourable growth conditions, namely salinity 26 (temperature 20 C) and temperature 30 C (salinity 36), respectively. Total and individual toxin contents on a per cell basis (pg cell1), are reported in Fig. 2A and B for salinity and temperature experiments, respectively. Small differences in total toxin content were observed between the experiments that should have provided similar results, namely the cultures grown at temperature 20 C and salinity 36. Such differences could be due to the slightly different growth conditions, among which the difference in light intensity and in culture volume could have played a major role. In the salinity experiment, total toxin content reached the highest value in the culture grown at 32 (20 pg cell1) and the lowest at 40 (12 pg cell1). For the temperature experiment, O. cf. ovata grown at 25 C was found to have a total toxin content of 23 pg cell1, while cultures grown at 20 and 30 C produced 16 and 18 pg cell1, respectively. This last finding apparently is not in agreement with the maximum concentration (mg L1) observed at 20 C (salinity 36), which was indeed affected by the high cell yield of the culture (Table 2). Particularly, culture grown at 20 C showed a cell density almost two-fold higher than the others.
3.3.
Haemolytic assay
15
0
Toxin content (pg cell )
87
B
palytoxin ovatoxin-a ovatoxin-b ovatoxin-c ovatoxin-d,-e tot
20 15 10 5 0 20
25 pellet
30
20
25 medium
30
Temperature (°C)
Fig. 2 e Total and individual toxin contents of putative palytoxin, ovatoxin-a, -b, -c, -d, and -e of O. cf. ovata cultures grown under different salinity (A) and temperature (B) conditions. HR LC-MS measurements (pg cellL1) were carried out for both pellet and medium extracts at the end of stationary growth phase.
All the O. cf. ovata culture extracts investigated in the present study were tested by haemolytic assay and the results, expressed as haemolysis percentage versus cell number present in 1 mL of assay solution (cell mL1), are reported in Fig. 3. All showed a strong delayed haemolysis of sheep erythrocytes, which was specifically inhibited by ouabain, even at concentrations corresponding to very low cell densities; however, a percentage of not-specific haemolysis was left over even in the presence of ouabain as shown in Fig. 3 (dotted lines). The haemolytic activity of cultures grown at different salinity levels (Fig. 3A) followed a pattern similar to that measured by HR LC-MS; particularly the highest haemolysis (83%) was observed for the culture grown at salinity 32 (total toxin content ¼ 20 pg cell1) followed by cultures grown at salinity 26 (haemolysis 79%, total toxin content ¼ 16 pg cell1), 36 (haemolysis 76%, total toxin content ¼ 18 pg cell1), and 40 (haemolysis 74%, total toxin content ¼ 12 pg cell1). For the temperature experiment (Fig. 3B) cells grown at 20 C reported the lowest haemolytic activity (haemolysis 76%, total toxin content ¼ 16 pg cell1) in agreement with HR LC-MS results, while an 82% haemolytic effect was observed for both cultures grown at 25 and 30 C, despite the slightly different toxin content of 23 and 18 pg cell1, respectively. All the above data were consistent with HR LC-MS results expressed as pg cell1. A comparison of the results of the haemolytic assay with the quantitative results achieved by HR LC-MS could provide useful information about the haemolytic activity of ovatoxins
Haemolysis (%)
88
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2
100 90 80 70 60 50 40 30 20 10 0 1
activity of the overall ovatoxins is quite similar to that of palytoxin standard. Values obtained for the palytoxin standard were interpolated using a non-linear estimation curve, described by the reported equation (f1, Fig. 4). The resulting EC50 values for palytoxin standard and algal extracts were not significantly different (ANOVA, P > 0.05), being 22 2 and 25 8 pg mL1, respectively.
26 psu 26 psu+OUA 32 psu 32 psu+OUA 36 psu 36 psu+OUA 40 psu 40 psu+OUA
A
2
3
4
5
6
7
8
Haemolysis (%)
100 90 80 70 60 50 40 30 20 10 0
B
0
1
20°C 20°C+OUA 25°C 25°C+OUA 30°C 30°C+OUA
2
3
4
5
6
7
8
9
Artemia sp. assay
3.4.
-1
Concentration (cell mL )
10
-1
Concentration (cell mL ) Fig. 3 e Haemolytic activity of O. cf. ovata extracts grown at different salinity (A) and temperature (B) conditions on sheep erythrocytes in absence (solid lines) and in presence (dashed lines) of ouabain (OUA). Data are expressed as haemolysis percentage (%) versus cell number mLL1 assay (cell mLL1).
in comparison with that of palytoxin. The haemolytic activity of the algal extracts from salinity and temperature experiments were also expressed as haemolysis percentage versus concentration of pg total toxin contained in 1 mL assay solution (pg mL1) as measured by HR LC-MS. These data are compared in Fig. 4 with those obtained for the haemolytic activity of palytoxin standard tested at seven different concentrations. This clearly suggests that the haemolytic
Artemia sp. assays were carried out using both live and lysed cells of O. cf. ovata cultures as well as the algal extract and the growth medium of a culture grown at 20 C and salinity 36. O. cf. ovata live cells induced rapid and high mortality of Artemia sp. nauplii, even at low cell concentrations. From the EC50 values of all the samples calculated at 24 h (Table 3), cell toxicity appeared relevant and significantly different (ANOVA, P < 0.001): the growth medium resulted significantly less toxic than the live cells (Post-hoc SNK test, P < 0.001), with an EC50 value of 720 cell mL1 versus 8 cell mL1, respectively. This result confirmed the presence of small amounts of toxins released in the growth medium. The lysed cells induced a similar mortality as the algal extract, as evidenced by the comparable EC50 values (Post-hoc SNK test, P > 0.05). EC50 values of all the tested O. cf. ovata samples were expressed also as pg of toxins per mL assay (pg mL1) (Table 3), basing on the total toxin contents measured by HR LC-MS. A palytoxin standard at five levels of concentrations (500e10,000 pg mL1) was also tested; it presented an EC50 value of 4606 pg mL1, which was significantly higher (Posthoc SNK test, P < 0.001) than that of the algal extract (1146 pg mL1). Administration of live O. cf. ovata cells, grown at different salinity and temperature conditions, to Artemia sp. resulted in no significant differences among the measured EC50 values (ANOVA, P > 0.05). For the temperature experiment the lowest EC50 value was measured for live cells grown at 25 C (EC50 ¼ 6 2 cell mL1) compared to those grown at 20 and palytoxin standard algal extracts palytoxin standard+OUA algal extracts+OUA
100
Haemolysis (%)
90 80
f1
70 60 50 40 30 20 10 0 0
25
50
75
100
125
150
175
200
-1
Concentration (pg mL )
Function (f1) PLTX standard
y = (85.5819)-(85.5819)/(1+(x/(21.473))^(log(9)/(log((254.169)/(21.473)))))
Fig. 4 e Haemolytic activity of O. cf. ovata extracts and palytoxin standard on sheep erythrocytes in absence and in presence of ouabain (OUA). Data are expressed as haemolysis percentage (%) versus concentration of palytoxin equivalent per mL of assay (pg mLL1). Equation f1: non-linear estimation curve obtained for the palytoxin standard.
89
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2
et al., 2011). Since very few laboratory studies on the effects of environmental parameters on growth and toxicity of Ostreopsis isolates have been reported (Grane´li et al., 2011; Ashton et al., 2003; Morton et al., 1992), we carried out a detailed study on an Adriatic strain of O. cf. ovata grown at different temperature and salinity conditions.
Table 3 e The 50% mortality on Artemia nauplii (EC50) is expressed both as cells of O. cf. ovata per mL (cell mLL1) and as total toxin content per mL (pg mLL1) basing on the HR LC-MS quantification. Values are reported for O. cf. ovata live and lysed cells, extract, and growth medium. EC50 value obtained for palytoxin standard is reported as pg mLL1. Each value is the mean of three replicates ± standard error. EC50 (cell mL1) 85 96 6 80 7 720 54 e
O. cf. ovata live cells O. cf. ovata lysed cells O. cf. ovata extract O. cf. ovata medium Palytoxin standard
4.1.
EC50 (pg mL1) 115 1376 1146 1822 4606
The Adriatic O. cf. ovata strain was tolerant to salinity variation in the range 26e40. Very similar growth rates and yields were observed within the tested salinity range, with the lowest growth yield being recorded at salinity 26. This was in good agreement with field measurements performed during Mediterranean O. cf. ovata blooms (Totti et al., 2010; Monti et al., 2007) as well as with results of a survey of epiphytic dinoflagellates along the Hawaiian coast, (Parsons and Preskitt, 2007) in which O. cf. ovata was the only dinoflagellate to be negatively correlated with salinity. In the temperature experiment, the analyzed Adriatic O. cf. ovata strain reached the highest growth yield at 20 C, whereas the lowest yield was recorded at 30 C. Our results are in good agreement with field surveys in the Adriatic Sea, where O. cf. ovata proliferation occurs from the end of August to October, when water temperature is about 20e22 C (Totti et al., 2010; Monti et al., 2007). On the contrary, Grane´li et al. (2011) indicated, for a Tyrrhenian O. cf. ovata strain, that high water temperatures (26e30 C) increased both growth rate and yield; this is consistent with the field surveys reporting O. cf. ovata blooms in the Tyrrhenian Sea in the middle of the summer. Our results and those observed by Grane´li et al. (2011) suggest that Adriatic and Tyrrhenian strains are differently affected by temperature. For the morphometric characters, in both salinity and temperature experiments, a certain cell size variability was observed; however, the cell volumes reported under the different growth conditions did not show a specific pattern, particularly in the stationary growth phase. A marked variability in the biovolumes of O. cf. ovata cells from the same culture was observed (Guerrini et al., 2010), and is in agreement with field observations (Aligizaki and Nikolaidis, 2006; Bianco et al., 2007).
72 86 272 137 781
30 C (EC50 ¼ 11 3 and 14 1 cell mL1, respectively). For the salinity experiment the lowest EC50 value was obtained at salinity 32 (EC50 ¼ 9 2 cell mL1), while cells grown at the other salinities reported values of 17 3, 24 8, and 17 4 cell mL1 for salinity 26, 36 and 40, respectively. This appears in good agreement with total toxin contents (pg cell1) measured by HR LC-MS (Fig. 4).
3.5.
Fish bioassay
Table 4 shows the results of the ichthyotoxic assay performed with different concentrations of O. cf. ovata live cells. Sea bass mortality occurred after 1 day of exposure, only at the highest O. cf. ovata cell density (2367 cells mL1). Before dying, loss of balance and fish floating was observed. After 45 h from the beginning of the assay, even fish exposed to lower algal concentrations (1138 and 425 cells mL1) began to die and they were all dead after 72 h. Fish exposed to the control diatom P. tricornutum (789, 100 cells mL1) survived and behaved normally till the end of the experiment (96 h).
4.
Growth and cell size pattern
Discussion
Several field surveys have indicated that environmental conditions play a major role in determining Ostreopsis spp. proliferation (reviewed by Mangialajo et al., 2011 and Pistocchi
Table 4 e Toxicity of different concentrations of O. cf. ovata cells on fish (Dicentrarchus labrax). Phaeodactylum tricornutum was used as control and was tested at the reported cell density. Results are reported as n dead organisms/n tested organisms, and time is expressed as hours (h). Time (h)
P. tricornutum 1
789 100 cell mL
0 28 30 31 45 51 52 72 96
O. cf. ovata
O. cf. ovata 1
2 367 cell mL
O. cf. ovata 1
1 138 cell mL
425 cell mL1
Tank 1
Tank 2
Tank 1
Tank 2
Tank 1
Tank 2
Tank 1
Tank 2
0/3 0/3 0/3 0/3 0/3 0/3 0/3 0/3 0/3
0/3 0/3 0/3 0/3 0/3 0/3 0/3 0/3 0/3
0/3 1/3 2/3 2/3 3/3
0/3 1/3 2/3 3/3
0/3 0/3 0/3 0/3 3/3
0/3 0/3 0/3 0/3 1/3 1/3 1/3 3/3
0/3 0/3 0/3 0/3 2/3 3/3
0/3 0/3 0/3 0/3 0/3 0/3 1/3 3/3
90
4.2.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2
Toxin profile
Putative palytoxin and all the ovatoxins so far known (Ciminiello et al., 2010) were detected in O. cf. ovata extracts. In the cultures grown under different conditions the relative abundance of individual toxins was similar, with ovatoxina and putative palytoxin being the major and the minor component of the toxin profile, respectively. The highest total toxin content on a per cell basis (pg cell1) was recorded in cultures grown at 25 C, while the highest total toxin concentration on a per litre basis was recorded at 20 C, namely under conditions that induced the highest growth yield. A reverse correlation between growth and toxin production has been reported also by Grane´li et al. (2011), as found also for other dinoflagellates (Etheridge and Roesler, 2005; Errera et al., 2010). As for the salinity experiment the highest total toxin content (pg cell1) was measured at salinity 32, while it decreased at lower and higher salinity values. However, no clear correlation between growth and toxin content was observed in the salinity experiment. The extracellular release increased as the temperature increased, with the maximum 27% value being observed at 30 C, the most unfavourable growth condition in the temperature experiment. This suggests that high temperatures favour cell lysis, leading to toxins being released in the growth medium. Similarly, in the salinity experiment, the highest release was also measured at the most unfavourable growth condition (26 C). Comparable results were obtained for Protoceratium reticulatum (Guerrini et al., 2007) and this could represent a response of the cells to the osmotic stress.
4.3. data
Haemolysis results in comparison with HR LC-MS
Palytoxin converts Naþ/Kþ pump into a non-selective cation channel, causing cell lysis; ouabain and other cardiac glycosides are used as indicators for the site of action since these compounds are specific ligands for the Naþ/Kþ-ATPase. The haemolytic assay proposed by Riobo´ et al. (2008) is a rapid and sensitive method to determine palytoxin content. In our study, it was successfully applied to the analyses of O. cf. ovata extracts in order to gain information about the haemolytic activity of ovatoxins. The haemolytic assay resulted highly reproducible even among separate set of experiments and using different blood samples. The haemolytic activity was tested by using O. cf. ovata extracts obtained from cultures set up at different growth conditions. The obtained data showed a good correlation between haemolysis percentage and the total toxin content measured through HR LC-MS. Although comparison of LC-MS and haemolysis assay results has already been done (Rhodes et al., 2010), in this work a detailed and quantitative cross check between biological assay and chemical analysis was applied to palytoxin-like compounds for the first time. Useful information was obtained from haemolytic tests after pretreatment with ouabain. They showed that ovatoxins behave similarly to palytoxin, suggesting a common mechanism of action, which involves a binding to the Naþ/Kþ pump. The haemolytic activity of all the O. cf. ovata extracts was found to be very similar to that of palytoxin, as confirmed also
by the similar EC50 values. These data suggested that ovatoxins, which represent the major components of the O. cf. ovata extracts (99.5e97%), have a similar haemolytic effect as palytoxin standard. It has to be noted that, in our analyses, the total activity of ovatoxins was measured and it has still to be ascertained whether individual components of the ovatoxin profile present different haemolytic activity. This will be possible when each ovatoxin will be isolated as a pure compound and used to evaluate its haemolytic activity. So far, the haemolytic assay appears to be a good method for preliminary quantification of the whole of palytoxin-like compounds in algal extracts: equation (f1, Fig. 4) obtained from the haemolysis curve, indicating the total haemolysis, can be a powerful tool to evaluate total toxin concentration of algal extracts, especially in laboratories where LC-MS is not available. However, some drawbacks of this assay are represented by the interference of other possibly co-occurring haemolytic compounds and its inability to define toxin profile.
4.4.
Toxicity for crustacean and fish
The toxicological assays revealed a marked toxicity of compounds produced by O. cf. ovata on Artemia nauplii and juvenile sea basses. For Artemia sp., the assay performed with the live O. cf. ovata cells reported mortality of nauplii even at very low cell densities and the EC50 value was significantly lower than those obtained for O. cf. ovata lysed cells, algal extract, and growth medium (Table 3). The difference in EC50 values of O. cf. ovata live cells versus both O. cf. ovata lysed cells and algal extract can be related to a different toxin uptake by the Artemia sp. nauplii: live cells were actually ingested by nauplii whereas either lysed cells or algal extract were taken up through filtration. Thus, this latter mechanism of toxin uptake seems to be less powerful than ingestion. This suggests that herbivorous fish, feed on seaweeds where the benthic dinoflagellates proliferates, is the most vulnerable to O. cf. ovata toxicity. The high EC50 value of the O. cf. ovata growth medium also deserves consideration. This can be related to the low toxin extracellular release emerging by HR LC-MS data (Table 2). Despite the apparently low toxicity of O. cf. ovata growth medium on Artemia nauplii, a long lasting bloom could be anyway hazardous to marine crustaceans, particularly considering that cell lyses and toxin extracellular release increase at the end of the stationary phase reached at the end of the bloom. Unlike the haemolytic assay, the Artemia sp. assay was not able to detect differences in the toxin contents of O. cf. ovata cultures grown at different salinity and temperature conditions. This could be due to the extreme sensitivity of Artemia sp. nauplii to O. cf. ovata live cells (EC50 values ranging from 6 to 24 cell mL1), which has not been observed for any other harmful algae so far (Pezzolesi et al., 2010). Thus, Artemia sp. assay is not able to catch relatively small differences among different samples and, therefore, it cannot be used for quantitative purposes. In the ichthyotoxic assay, sea basses exposed to O. cf. ovata live cells died within a few days despite they are known not to feed on microalgal cells. This mortality could be attributed to an haemolytic effect of palytoxin-like compounds on the gills,
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2
where Naþ/Kþ ATPase activity is high in the juvenile stage of sea basses (Varsamos et al., 2004). However, we cannot exclude an effect due to accidental ingestion of algal cells, which were contained in the surrounding water at high density.
Acknowledgements This research was supported by MURST PRIN, Rome, Italy. We thank Prof. Poglayen of the Department of Veterinary Public Health and Animal Pathology (University of Bologna) for kindly proving us the sheep blood, and the hatchery of Valle Ca’ Zuliani (Pila di Porto Tolle, Rovigo, Italy) for the juvenile sea basses. We thank Dr. Andrea Pasteris for the advice on the statistical analysis. We are grateful to Dr. Beth Strain for English revision of the manuscript.
references
Abboud-Abi Saab, M., 1989. Les dinoflagelle´s des eaux coˆtie`res libanaises-espe`ces rares ou nouvelles du phytoplancton marin. Lebanese Science Bulletin 5, 5e16. Aligizaki, K., Nikolaidis, G., 2006. The presence of the potentially toxic genera Ostreopsis and Coolia (Dinophyceae) in the North Aegean Sea, Greece. Harmful Algae 5, 717e730. Aligizaki, K., Katikou, P., Nikolaidis, G., Panoub, A., 2008. First episode of shellfish contamination by palytoxin-like compounds from Ostreopsis species (Aegean Sea, Greece). Toxicon 51, 418e427. Ashton, M., Tosteson, T., Tosteson, C., 2003. The effect of elevated temperature on the toxicity of the laboratory cultured dinoflagellate Ostreopsis lenticularis (Dinophyceae). Revista de Biologı´a Tropical 51, 1e6. Bianco, I., Sangiorgi, V., Penna, A., Guerrini, F., Pistocchi, R., Zaottini, E., Congestri, R., 2007. Ostreopsis ovata in benthic aggregates along the Latium Coast (middle Tyrrhenian Sea). International Symposium on Algal Toxins, SITOX, p. 29. Bignami, G.S., 1993. A rapid and sensitive hemolysis neutralization assay for palytoxin. Toxicon 31, 817e820. Bottalico, A., Micella, P., Feliciti, G.P., 2002. Fioritura di Ostreopsis sp. (Dinophyta) nel porto di Otranto. In: Gruppo di Lavoro per l’Algologia. Societa` Botanica Italiana, Chioggia 8e9 November 2002. Chang, F.H., Shimizu, Y., Hay, B., Stewart, R., Mackay, G., Tasker, R., 2000. Three recently recorded Ostreopsis spp. (Dinophyceae) in New Zealand: temporal and regional distribution in the upper North Island from 1995 to 1997. New Zealand Journal of Marine & Freshwater Research 34, 29e39. Ciminiello, P., Dell’Aversano, C., Fattorusso, E., Forino, M., Magno, G.S., Tartaglione, L., Grillo, C., Melchiorre, N., 2006. The Genoa 2005 outbreak. Determination of putative palytoxin in Mediterranean Ostreopsis ovata by a new liquid chromatography tandem mass spectrometry method. Analytical Chemistry 78, 6153e6159. Ciminiello, P., Dell’Aversano, C., Fattorusso, E., Forino, M., Tartaglione, L., Grillo, C., Melchiorre, N., 2008. Putative palytoxin and its new analogue, ovatoxin-a, in Ostreopsis ovata collected along the Ligurian coasts during the 2006 toxic outbreak. Journal of the American Society for Mass Spectrometry 19, 111e120. Ciminiello, P., Dell’Aversano, C., Dello Iacovo, E., Fattorusso, E., Forino, M., Grauso, L., Tartaglione, L., Guerrini, F., Pistocchi, R., 2010. Complex palytoxin-like profile of Ostreopsis ovata.
91
Identification of four new ovatoxins by high-resolution liquid chromatography/mass spectrometry. Rapid Communications in Mass Spectrometry 24, 2735e2744. Errera, R.M., Bourdelais, A., Drennan, M.A., Dodd, E.B., Henrichs, D.W., Campbell, L., 2010. Variation in brevetoxin and brevenal content among clonal cultures of Karenia brevis may influence bloom toxicity. Toxicon 55, 195e203. Etheridge, S.M., Roesler, C.S., 2005. Effects of temperature, irradiance, and salinity on photosynthesis, growth rates, total toxicity, and toxin composition for Alexandrium fundyense isolates from the Gulf of Maine and Bay of Fundy. Deep-Sea Research II 52, 2491e2500. Faust, M.A., Morton, S.L., Quod, J.P., 1996. Further SEM study of the marine dinoflagellates: the genus Ostreopsis (Dinophyceae). Journal of Phycology 32, 1053e1065. Grane´li, E., Vidyarathna, N.K., Funari, E., Cumaranatunga, R., 2008. Climate change and benthic Dinoflagellates - the Ostreopsis ovata case. In: Proceedings of the 13th International Conference on Harmful Algae, p. 42. Hong Kong, 3e7 November, 2008. Grane´li, E., Vidyarathna, N.K., Funari, E., Cumaranatunga, P.R.T., Scenati, R., 2011. Can increases in temperature stimulate blooms of the toxic benthic dinoflagellate Ostreopsis ovata? Harmful Algae 10, 165e172. Guerrini, F., Ciminiello, P., Dell’Aversano, C., Tartaglione, L., Fattorusso, E., Boni, L., Pistocchi, R., 2007. Influence of temperature, salinity and nutrient limitation on yessotoxin production and release by the dinoflagellate Protoceratium reticulatum in batch-cultures. Harmful Algae 6, 707e717. Guerrini, F., Pezzolesi, L., Feller, A., Riccardi, M., Ciminiello, P., Dell’Aversano, C., Tartaglione, L., Dello Iacovo, E., Fattorusso, E., Forino, M., Pistocchi, R., 2010. Comparative growth and toxin profile of cultured Ostreopsis ovata from the Tyrrhenian and Adriatic Seas. Toxicon 55, 211e220. Guillard, R.R.L., 1975. Culture of phytoplankton for feeding marine invertebrates. In: Smith, W.L., Chanley, M.H. (Eds.), Culture of Marine Invertebrates Animals. Plenum Press, New York, pp. 26e60. Habermann, E., 1989. Palytoxin acts through Naþ, Kþ-ATPase. Toxicon 27, 1171e1187. Hasle, G.R., 1978. The inverted microscope method. In: Sournia, A. (Ed.), Phytoplankton Manual. Monographs on Oceanographic Methodology, vol. 6. UNESCO, Paris, pp. 88e96. Hoshaw, R.W., Rosowski, J.R., 1973. Methods for microscopic algae. In: Stein, J.R. (Ed.), Handbook of Phycological Methods. Cambridge University Press, New York, pp. 53e67. IRSA-CNR, 2003. Metodi analitici per le acque, APAT Manuali e Linee Guida 29/2003. In: 8060-Metodo di valutazione della tossicita` acuta con Artemia sp., vol. 3 1043e1050. Lenoir, S., Ten-Hage, L., Turquet, J., Quod, J.P., Bernard, C., Hennion, M.C., 2004. First evidence of palytoxin analogues from an Ostreopsis mascarenensis (Dinophyceae) benthic bloom in Southwestern Indian Ocean. Journal of Phycology 40, 1042e1051. Mangialajo, L., Bertolotto, R., Cattaneo-Vietti, R., Chiantore, M., Grillo, C., Lemee, R., Melchiorre, N., Moretto, P., Povero, P., Ruggirei, N., 2008. The toxic benthic dinoflagellate Ostreopsis ovata: quantification of proliferation along the coastline of Genoa, Italy. Marine Pollution Bulletin 56, 1209e1214. Mangialajo, L., Ganzin, N., Accoroni, S., Asnaghi, V., Blanfune´, A., Cabrini, M., Cattaneo-Vietti, R., Chavanon, F., Chiantore, M., Cohu, S., Costa, E., Fornasaro, D., Grossel, H., MarcoMiralles, F., Maso´, M., Ren˜e´, A., Rossi, A.M., Montserrat, S., Thibaut, T., Totti, C., Vila, M., Leme´e, R., 2011. Trends in Ostreopsis proliferation along the Northern Mediterranean coasts. Toxicon 57, 408e420. Monti, M., Minocci, M., Beran, A., Ivesa, L., 2007. First record of Ostreopsis cfr. ovata on macroalgae in the Northern Adriatic Sea. Marine Pollution Bulletin 54, 598e601.
92
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 8 2 e9 2
Morton, S.L., Norris, D.R., Bomber, J.W., 1992. Effect of temperature, salinity and light-intensity on the growth and seasonality of toxic dinoflagellates associated with ciguatera. Journal of Experimental Marine Biology and Ecology 157, 79e90. Onuma, Y., Satake, M., Ukena, T., Roux, J., Chanteau, S., Rasolofonirina, N., Ratsimaloto, M., Naoki, H., Yasumoto, T., 1999. Identification of putative palytoxin as the cause of clupeotoxism. Toxicon 37, 55e65. Parsons, M.L., Preskitt, L.B., 2007. A survey of epiphytic dinoflagellates from the coastal waters of the island of Hawai‘i. Harmful Algae 6, 658e669. Pezzolesi, L., Cucchiari, E., Guerrini, F., Pasteris, A., Totti, C., Pistocchi, R., 2010. Toxicity evaluation of Fibrocapsa japonica from the Northern Adriatic Sea through a chemical and toxicological approach. Harmful Algae 8, 504e514. Pistocchi, R., Pezzolesi, L., Guerrini, F., Vanucci, S., Dell’Aversano, C., Fattorusso, E., 2011. A review on the effects of environmental conditions on growth and toxin production of Ostreopsis ovata. Toxicon 57, 421e428. Riobo´, P., Paz, B., Franco, J.M., Va´zquez, J.A., Murado, M.A., 2008. Proposal for a simple and sensitive haemolytic assay for palytoxin: toxicological dynamics, kinetics, ouabain inhibition and thermal stability. Harmful Algae 7, 415e429. Riobo´, P., Franco, J.M., 2011. Palytoxins: biological and chemical determination. Toxicon 57, 368e375. Rhodes, L., Smith, K., Munday, R., Briggs, L., Selwood, A., 2010. Ostreopsis ovata isolated from Rarotonga, Cook Islands, 14th International Conference of Harmful Algae, Hersonissos-Crete (Greece), 1e5 November. Conference abstract book, p. 229. Rhodes, L., 2011. World-wide occurrence of the toxic dinoflagellate genus Ostreopsis Schmidt. Toxicon 57, 400e407. Sansoni, G., Borghini, B., Camici, G., Casotti, M., Righini, P., Fustighi, C., 2003. Fioriture algali di Ostreopsis ovata (Gonyaulacales: Dinophyceae): un problema emergente. Biol. Amb 17, 17e23. Shears, N.T., Ross, P.M., 2009. Blooms of benthic dinoflagellates of the genus Ostreopsis; an increasing and ecologically important phenomenon on temperate reefs in New Zealand and worldwide. Harmful Algae 8, 916e925. Simoni, F., Gori, L., Di Paolo, C., Lepri, L., Mancino, A., 2004. Fioriture d’Ostreopsis ovata, Coolia monotis, Prorocentrum lima nelle macroalghe del mar Tirreno settentrionale, Mediterraneo (seconda fase di studio). Biol. Ital. 10, 68e73.
Sun, J., Liu, D., 2003. Geometric models for calculating cell biovolume and surface area for phytoplankton. Journal of Plankton Research 25, 1331e1346. Taniyama, S., Osamu, A., Masamitsu, T., Sachio, N., Tomohiro, T., Yahia, M., Tamao, N., 2003. Ostreopsis sp., a possible origin of palytoxin (PLTX) in parrotfish Scarus ovifrons. Toxicon 42, 29e33. Taylor, F.J.R., 1979. A description of the benthic dinoflagellate associated with maitotoxin and ciguatoxin, including observations on Hawaiian material. In: Taylor, D.L., Seliger, H. H. (Eds.), Toxic Dinoflagellate Blooms. Elsevier/North-Holland, New York, pp. 71e76. Tognetto, L., Bellato, S., Moro, I., Andreoli, C., 1995. Occurrence of Ostreopsis ovata (Dinophyceae) in the Tyrrhenian Sea during summer 1994. Botanica Marina 38, 291e295. Totti, C., 2003. Influence of the plume of the river Po on the distribution of subtidal microphytobenthos in the Northern Adriatic Sea. Botanica Marina 46, 161e178. Totti, C., Cucchiari, E., Romagnoli, T., Penna, A., 2007. Bloom of Ostreopsis ovata on the Conero riviera (NW Adriatic Sea). Harmful Algae News 33, 12e13. Totti, C., Accoroni, S., Cerino, F., Cucchiari, E., Romagnoli, T., 2010. Ostreopsis ovata bloom along the Conero Riviera (northern Adriatic Sea): relationships with environmental conditions and substrata. Harmful Algae 9, 233e239. Usami, M., Satake, M., Ishida, S., Inoue, A., Kan, Y., Yasumoto, T., 1995. Palytoxin analogs from the dinoflagellate Ostreopsis siamensis. Journal of the American Chemical Society 117, 5389e5390. Ukena, T., Satake, M., Usami, M., Oshima, Y., Naoki, H., Fujita, T., Kan, Y., Yasumoto, T., 2001. Structure elucidation of ostreocin D, a palytoxin analog isolated from the dinoflagellate Ostreopsis siamensis. Bioscience, Biotechnology, and Biochemistry 65, 2585e2588. Varsamos, S., Wendelaar Bonga, S.E., Charmantier, G., Flik, G., 2004. Drinking and Naþ/Kþ ATPase activity during early development of European sea bass, Dicentrarchus labrax: Ontogeny and short-term regulation following acute salinity changes. Journal of Experimental Marine Biology and Ecology 311, 189e200. Vila, M., Garce´s, E., Maso`, M., 2001. Potentially toxic epiphytic dinoflagellate assemblages on macroalgae in the NW Mediterranean. Aquatic Microbial Ecology 26, 51e60.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Recycled water: Potential health risks from volatile organic compounds and use of 1,4-dichlorobenzene as treatment performance indicator Clemencia Rodriguez a,b,*, Kathryn Linge c,1, Palenque Blair d,2, Francesco Busetti c,3, Brian Devine a,4, Paul Van Buynder a,e,5, Philip Weinstein f,6, Angus Cook a,7 a
School of Population Health, Faculty of Medicine, Dentistry and Health Sciences, The University of Western Australia, 35 Stirling Hwy, (M431) Crawley 6009 Western Australia, Australia b Department of Health, Government of Western Australia, Grace Vaughan House 227 Stubbs Terrace, Shenton Park, 6008 Western Australia, Australia c Curtin Water Quality Research Centre, Department of Chemistry, Curtin University, Kent Street, Bentley 6102 Western Australia, Australia d Water Corporation of Western Australia, 629 Newcastle Street Leederville, 6007 Western Australia, Australia e Fraser Health Authority, C200, 9801 King George Boulevard Surrey, BC V3T 5E5, Canada f University of South Australia, City West Campus, GPO Box 2471 Adelaide, 5001 South Australia, Australia
article info
abstract
Article history:
Characterisation of the concentrations and potential health risks of chemicals in recycled
Received 25 May 2011
water is important if this source of water is to be safely used to supplement drinking water
Received in revised form
sources. This research was conducted to: (i) determine the concentration of volatile organic
19 September 2011
compounds (VOCs) in secondary treated effluent (STE) and, post-reverse osmosis (RO)
Accepted 16 October 2011
treatment and to; (ii) assess the health risk associated with VOCs for indirect potable reuse
Available online 25 October 2011
(IPR). Samples were examined pre and post-RO in one full-scale and one pilot plant in Perth, Western Australia. Risk quotients (RQ) were estimated by expressing the maximum
Keywords:
and median concentration as a function of the health value. Of 61 VOCs analysed over
Water recycling
a period of three years, twenty one (21) were detected in STE, with 1,4-dichlorobenzene
Water quality
(94%); tetrachloroethene (88%); carbon disulfide (81%) and; chloromethane (58%) most
Organic pollutants
commonly detected. Median concentrations for these compounds in STE ranged from
Indirect potable reuse
0.81 mg/L for 1,4-dichlorobenzene to 0.02 mg/L for carbon disulphide. After RO, twenty six
Volatile organic compounds
(26) VOCs were detected, of which 1,4-dichlorobenzene (89%); acrylonitrile (83%) chloro-
Reverse osmosis
methane (63%) and carbon disulfide (40%) were the more frequently detected. RQ(max) were all below health values in the STE and after RO. Median removal efficiency for RO was variable, ranging from 77% (dichlorodifluoromethane) to 91.2% (tetrachloroethene).
* Corresponding author. Department of Health, Government of Western Australia, Grace Vaughan House 227 Stubbs Terrace, Shenton Park, 6008 Western Australia, Australia. Tel.: þ8 93884812; fax: þ8 93884910. E-mail addresses:
[email protected] (C. Rodriguez),
[email protected] (K. Linge),
[email protected] (P. Blair),
[email protected] (F. Busetti),
[email protected] (B. Devine), Paul.VanBuynder@fraserhealth. ca (P. Van Buynder),
[email protected] (P. Weinstein),
[email protected] (A. Cook). 1 Tel.: þ8 92667534; fax: þ8 92663547. 2 Tel.: þ8 94203328; fax: þ8 94203195. 3 Tel.: þ8 92663273; fax: þ8 92663547. 4 Tel.: þ8 64888667; fax: þ8 64881188. 5 Tel.: þ1 604 587 7621; fax: þ 1 604 587 7625. 6 Tel.: þ61 8 830 25129; fax: þ61 8 830 20828. 7 Tel.: þ8 64887804; fax: þ8 64881188. 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.032
94
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
The results indicate that despite the detection of VOCs in STE and after RO, their human health impact in IPR is negligible due to the low concentrations detected. The results indicate that 1,4-dichlorobenzene is a potential treatment chemical indicator for assessment of VOCs in IPR using RO treatment. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Volatile organic compounds (VOCs) are organic chemicals that have a relatively low boiling point (250 C measured at a standard atmospheric pressure of 101.3 kPa) and high vapour pressure relative to their water solubility. This class of chemicals therefore easily volatilize from water to air at room temperatures and enter the atmosphere upon contact with an airewater interface. Substances that are included in the VOC category are: aliphatic hydrocarbons (e.g. hexane), aromatic hydrocarbons (e.g. benzene, toluene and the xylenes), halogenated hydrocarbons (e.g. tetrachloroethene) and oxygenated compounds (e.g. acetone and similar ketones). VOCs are widely used and comprise an important group of environmental contaminants. They are produced in large volumes and are associated with numerous products and applications, including household cleaners, fuel additives, and commercial and industrial solvents. VOCs dissolve many other substances and are used as cleaning and liquefying agents in fuels, degreasers, adhesives, solvents, polishes, cosmetics, refrigerants, drugs, and dry cleaning solutions (Zogorski et al., 2006). VOCs may be emitted from fabrics, carpets, fibreboard, plastic products, glues, solvents, household cleaners, printed material, methylated spirits, paints and paint products (such as thinners or varnishes), disinfectants, cosmetics, degreasing products, and fuels. They are hence discharged to wastewater treatment plants (WWTP) from a large number of sources including commercial enterprises, industries, and residential households. VOCs have been detected in many water types, including secondary treated effluent (STE). Aliphatic hydrocarbons, aromatic hydrocarbons, halogenated volatiles and dimethyl disulfide account for approximately 70% of all VOCs detected in municipal STE (Koe and Shen, 1997). Although VOCs concentrations in raw wastewater may range from 1 to 150 mg/ L, atmospheric emissions during treatment generally lead to significantly lower dissolved concentrations in STE (Atasoy et al., 2004, Battistoni et al., 2007). Adsorption and biodegradation can reduce the concentration of VOCs in WWTPs. VOC release to the atmosphere during collection and in particular during aeration treatment is considered the most important method of removal of VOCs from STE (Fatone et al., 2011). The aeration that occurs during wastewater treatment and during many sludge treatment processes can achieve more than 90% removal of the VOCs concentration in raw wastewater (NRC, 1996). For example, Wu et al. (2002) reported a 96% decrease in total VOCs in a WWTP during exposure to the atmosphere via air stripping (Wu et al., 2002). In some circumstances, VOCs may also be found in public drinking water supplies as a result of spills, discharges, atmospheric deposition or leaching from contaminated soils.
Tetrachloroethylene, trichloroethene, 1,1-dichloroethene and benzene are examples of VOCs that are occasionally detected (Williams et al., 2002). Industrial discharges may lead to the release of VOCs into groundwater (along with gasoline oxygenates). For example, eighteen (18) of eighty-eight (88) VOCs were detected in twenty eight (28) wells sampled in the San Diego GroundWater Ambient Monitoring and Assessment study (Wright et al., 2005). Groundwater contamination with non-aqueous phase liquids, such as chlorinated solvents and petrol hydrocarbons, may pose a health risk if used as a drinking water source as they can be difficult to remove by treatment (Patterson et al., 1993). The presence of VOCs in drinking water is of concern because some of these compounds have adverse health effects, including potential carcinogenesis, and because they can change the taste and odour of drinking water. The health impact of VOCs varies greatly from those that are highly toxic, to those with no known health effect. As with other organic chemicals, the extent and nature of the health effect will depend on the level of exposure and length of exposure. Some VOCs may adversely affect the liver, kidneys, spleen, and stomach, as well as the nervous, circulatory, reproductive, immune, cardiovascular, and respiratory systems. Some VOCs may affect cognitive abilities, balance, and coordination. At high levels of exposure, VOCs can cause central nervous system depression (Boyes et al., 2000; Brouwer et al., 2005; Herpin et al., 2009) and can be irritating upon contact with the skin, to mucous membranes of the eyes or to the mucous membranes if inhaled (Toccalino et al., 2006; WHO, 2011). Acute symptoms after exposure to some VOCs (mainly from inhalation) include headaches, dizziness, visual disorders, and memory impairment. The chronic health effects to the general public from ingestion of VOCs at low concentrations in drinking water are less well understood but health values are well above offensive taste/odour thresholds and contain significant safety margins.
2.
Regulatory framework
VOCs comprise almost half of the 129 priority pollutants designated by the USEPA for limitation or prevention of introduction to water (USEPA, 1994), and some are also included in the European Commission priority pollutant lists. The EU adopted decision No 2455/2001/EC, which established a list of 33 priority substances in the field of water policy, include the following VOCs: benzene, C10e13-chloroalkanes, 1,2-dichloroethane, dichloromethane, hexachlorobutadiene, trichlorobenzenes and trichloromethane (European Commission, 1998). In Australia, the National Industrial Chemical Notification and Assessment Scheme (NICNAS) is the federal agency
95
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
assessing the human health risk of industrial chemicals including VOCs introduced into the country. The assessment reports include hazard, exposure assessments and risk characterisation for occupational health and safety, environment and public health. The reports are available on the web and are classified as (i) new chemicals; (ii) priority existing chemicals; and (iii) other assessments (http://www.nicnas.gov.au/ Publications/CAR.asp). National guidelines for sewerage systems e in particular those pertaining to (i) Effluent Management and (ii) Acceptance of Trade Waste (Industrial Waste) e provide a framework for sewerage authorities responsible for the management, monitoring, disposal and implementation of trade waste management programs. Regulations exist for levels of VOCs in various contexts, such as water systems (drinking water, sewage discharges and stormwater disposal), occupational settings and air emissions. Australian Federal and State regulations often limit the quantity of VOCs that are emitted from sources such as industrial facilities, WWTPs and landfills. In Western Australia, the Department of Environment and Conservation regulates the level of contaminants allowable in wastewater streams under the Environmental Protection Act 1986 and may prescribe specific license conditions for VOCs, depending on the type of industry (NWC, 2011). Examples of compounds that have to be reduced or removed from industrial facilities before discharge to sewers include benzene, trichlorethene, vinyl chloride and xylenes. Steam or air stripping, carbon adsorption and solvent extraction are all methods used for removing VOCs from wastewater before secondary treatment for compliance with regulatory requirements. Regulatory agencies and institutions set drinking water values for VOCs based on toxicological and epidemiological assessments. For the majority of VOCs that are assumed to be non-carcinogenic, it is hypothesised that there is a threshold dose below which no adverse health effects will occur. Consequently, drinking water guidelines are calculated based on tolerable daily intake (TDI) values. The TDI values are derived from toxicological studies conducted in animals and epidemiological data when available. Some VOCs are known or suspected carcinogens. For those VOCs classed as carcinogenic (without a threshold dose), guideline values are derived using mathematical models that combine toxicological data with the concept of acceptable levels of risk for lifetime consumption. The WHO guideline values are derived using an acceptable level of risk of one in a hundred thousand excess cancers attributable to a particular VOC consumption at the guideline concentration (WHO, 2011). For the Australian
Drinking Water Guidelines (ADWG), the acceptable level of risk is mainly based on one in a million excess cancers (NHMRC, 2011). In many cases the toxicological data used is the same but the assumptions used to calculate the health value varies. For example, the Canadian standard for 1,4dichlorobenzene is 5 mg/L (FPT Committee on Drinking Water, 2010) while the regulatory value for: USEPA is 75 mg/L (USEPA, 2011); WHO is 300 mg/L (WHO, 2011) and for Australia is 40 mg/L (NHMRC, 2011). The derivation of the health value ¼ 300 mg/L in the WHO guidelines is based on non cancer effects using the lowest observed adverse effect level (LOAEL) of 150 mg/kg for kidney effects in a two-year rat study with an uncertainty factor of 1000. In contrast the Canadian guidelines classified 1,4-dichlorobenzene as Group II e probably carcinogenic to humans based on a National Toxicology Program report and the calculation is based on the slope of a doseeresponse data with linear extrapolation to zero. The estimated ranges of concentrations corresponding to a lifetime risk of one in a hundred thousand are used. The work presented in this paper is part of a larger project investigating the effectiveness of microfiltration/reverse osmosis (MF/RO) to treat STE for indirect potable reuse (IPR), a key water conservation strategy for Western Australia (DOHWA, 2009). The objectives of these study were to (i) evaluate the range and concentration of volatile organic compounds (VOCs) in secondary treated effluent (STE) and post-reverse osmosis (RO) treatment for 61 VOCs; (ii) assess the health risk associated with VOCs for indirect potable reuse (IPR) with post-RO water and, (iii) determine the efficacy of RO to remove VOCs. This study provides the most extensive analysis of VOCs in treated wastewater in Australia published to date.
3.
Methods
3.1.
Sample sites
Six (6) sampling events were conducted from November 2006 to June 2008 (Table 1), with an overall total of 32 sampling days. Typically a single sampling event consisted of between 4 and 6 sampling trips over a week, with sampling focused on STE or and post-RO water. However, on a number of occasions, sampling of post-MF water was also undertaken. During membrane treatment, wastewater undergoes chloramination before MF to prevent RO membrane fouling and samples of post-MF water were analysed to determine the
Table 1 e Measurement of VOCs by event and location. Event 1 2 3 4 5 6 Total
Month
#days
Year
Groundwater
Secondary treated effluent
Post microfiltration
Post-reverse osmosis
Total
November May/June September January April June
4 6 6 6 5 5 32
2006 2007 2007 2008 2008 2008
0 108 0 114 0 0 222
159 432 336 342 392 362 2023
53 162 106 0 56 108 485
158 162 336 342 336 262 1596
370 864 778 798 784 732 4326
96
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
effect of chloramination during the MF/RO process. Grab STE samples represent the three major WWTPs in Perth and were taken directly from Beenyup WWTP and Subiaco WWTP and at the influent stream of the Kwinana Water Reclamation Plant (KWRP) for Woodman Point WWTP. Samples post-MF and post-RO were collected from two advanced treatment plants at KWRP and Beenyup Pilot Plant (BPP) in order to characterise water quality through the membrane treatment process. Details of each have been previously published (DOHWA, 2009) but briefly, KWRP treats secondary treated wastewater from Woodman Point WWTP by MF/RO to produce approximately 16 ML/day of general process water for neighbouring industrial facilities, reducing Perth’s total demand for drinking water by about 2%. The BPP treats a small volume of STE (approximately 100 kL/day) from Beenyup WWTP by MF/RO and is the first stage of a larger project investigating indirect potable reuse of Beenyup STE. The BPP was commissioned after the May/June 2007 sampling event (sampling event 2 in Table 1). Both plants are owned by the Water Corporation of Western Australia. Woodman Point WWTP receives wastewater primarily from residential areas, but also receives about 6% of wastewater from industrial facilities (DOHWA, 2009), while Beenyup WWTP has a sewer catchment that is mainly residential in nature. Grab samples were also collected from groundwater (GW) during sampling events 2 and 4. The groundwater in this context corresponds to the raw drinking water source. Standard protocols were used to ensure adequate sample preparation, preservation and transportation to the laboratory. Laboratory blanks, trip and field blanks were also analysed and constituted about one third of the samples analysed.
3.2.
Analysis of VOCs
The selection of VOCs was based on their risk profiles and factors. The following criteria were used to guide analytes inclusion in the target list: (i) the VOC is currently or has been registered for use in Australia; (ii) there is a high likelihood of the VOC being detected in wastewater based on known chemical and physical properties; (iii) the VOC has previously been detected in natural waters or wastewater; (iv) there were public perceptions that the chemical may pose a possible public health hazard; and (v) the VOC is listed in the ADWG (2004) (NHMRC, 2004) or other international regulatory agencies as regulated or as part of the USEPA contaminant candidate list. All VOCs except acrolein, acrylonitrile and methyl tertiary butyl ether (MTBE) were measured by purge and trap GCeMS. Acrolein, acrylonitrile, MTBE were extracted and preconcentrated by headspace solid-phase microextraction (SPME) before GCeMS analysis. Quantification was performed by mass spectrometry (MS) with electron ionisation (EI), with peak identification and calculation of recovery was aided by inclusion of surrogate standards. Limits of detection were determined for every analytical run and were calculated using the standard deviation of replicate analyses of a standard solution of appropriate concentration (typically 0.05e0.1 mg/L). Standard deviations were then converted to 95% confidence intervals using the student’s t-test.
Relative standard uncertainties were calculated using an uncertainty budget that incorporated precision, calibration standard preparation, sample volume, and linear regression of the calibration curve. Sample homogeneity was considered a negligible source of uncertainty. Acrolein and acrylonitrile were not validated to the same extent as other VOCs because they were only analysed during sampling event 3. The VOCs analysed, standard relative uncertainty at 0.5 mg/L, and average limits of detection (LODs) are presented in Tables 2 and 3.
3.3.
Data analysis
Unlike other chemical classes of compounds (e.g. dioxins), there is no common toxicological mechanism for VOCs, and therefore the potential human health risk was evaluated for individual compounds. Risk quotients (RQ) were estimated by expressing the maximum and median concentration in STE as a function of the health value for detected VOCs. For VOCs without detections in STE, RQs were calculated as the ratio between the LOD and the health value as a worst case scenario. A three tiered screening health risk assessment approach was used for the derivation of the health values. The basis for the tool has been discussed and applied in previous publications (Rodriguez et al., 2007a, 2007b). Under this system, VOCs were allocated to “tier 1 (regulated contaminants)”; “tier 2 (unregulated contaminants with toxicity information)” or “tier 3 Threshold of Toxicological Concern (unregulated contaminants with no toxicity information)”. For VOCs in tier 1, the order or priority for setting the health benchmark values was the ADWG (2011) (NHMRC, 2011), the ADWG (2004) (NHMRC, 2004), WHO guidelines including the 2nd addendum to the 3rd edition published in 2006 (WHO, 2011), the Drinking water standards and health advisories from the USEPA (USEPA, 2011) and the California Drinking Water Notification Levels and Response Levels (CDPH, 2010), based on the methodology previously described (Rodriguez et al., 2007b). Data were analysed in Stata version 10 (Stata Corp, 2007). Comparison of median concentrations was performed using non-parametric tests. For median comparison between KWRP and BPP the ManneWhitney test was used, while the Kruskal Wallis X2 test was used for comparison of median values of three or more groups. Results are reported at a significance level of 5% ( p < 0.05).
4.
Results
A total of 61 VOCs were analysed in at least one sampling event. A total of 4326 measurements were included in the statistical analysis for VOCs after excluding QA/QC samples. The distribution of sampling by event and location is presented in Table 1.
4.1.
Secondary treated effluent (STE)
Twenty one (21) VOCs (34% of the total) were detected in STE (Table 2). The most frequently detected VOC was 1,4-dichlorobenzene (93.9% of STE samples), followed by
Table 2 e VOCs detected in secondary treated effluent, post-microfiltration and/or post-reverse osmosis water and corresponding RQs. STE samples Parameter
CASR No Mean LOD SRU (%) Tier Health (mg/L) (0.5 mg/L) value (mg/L) 0.099 0.041 0.018 0.026 0.017 0.030 0.056 0.032 0.030 0.046 0.244 0.045 0.033 0.066 0.272 0.091 0.078 1.471 0.031 0.118 0.109 0.029 0.087 0.052 0.042 0.099 0.027 0.033 0.080 0.083 0.017 0.047 0.136 0.096
31.2 20.8 16.4 17.7 8.4 13.2 20.9 15.9 ND 30.4 20.7 53.4 23.2 20.1 21 32.6 37 32 41.7 27.9 30.2 17 48 53.7 18.2 40.2 52.6 13.2 31.7 59 20.1 26.7 37.4 50.4
1 2 1 1 1 1 2 1 1 1 2 2 3 2 2 1 1 1 2 1 1 1 1 2 1 1 1 1 2 2 1 3 1 1
0.057
34.6
3
5 330 40 1500 3 60 600 40 6 1 10 700 0.7 0.7 1000 4 300 40 100 50 800 20 600 260 600 600 30 300 260 330 100 7 30 150 7
Source
n
USEPA 2011 CDPH 2010 WHO 2011 ADWG 2011 ADWG 2011 ADWG 2011 USEPA 2011 ADWG 2011 USEPA 2011 ADWG 2011 USEPA 2011 IRIS 1990 TTC TTC USEPA 2011 ADWG 2011 ADWG 2011 USEPA 2011 USEPA 2011 ADWG 2011 ADWG 2011 WHO 2011 ADWG 2011 CDPH 2010 ADWG 2011 ADWG 2011 ADWG 2011 ADWG 2011 CDPH 2010 CDPH 2010 ADWG 2011 TTC ADWG 2011 CDPH 2008
29 28 29 29 29 29 29 29 6 29 29 15 29 29 24 8
3.5 3.6 10.3 10.3 10.3 34.5 10.3 93.1 50.0 27.6 3.5 80.0 24.1 62.1 4.2 12.5
0.02 0.0001 0.0005 0.00003 0.007 0.0009 0.00003 0.02 0.003 0.08 0.002 0.00002 0.04 0.004 0.0002 0.02
0.04 0.0002 0.005 0.0001 0.02 0.002 0.001 0.08 0.01 0.1 0.02 0.0006 0.8 0.02 0.0005 0.05
26
3.9
0.03
0.13
29 28 29
86.2 14.3 48.3
0.009 0.0002 0.003
0.2 0.0003 0.04
28
17.9
0.00005
0.0001
TTC
RQ (median)
RQ (max)
n
Post-RO samples
% of RQ RQ n % of RQ RQ detection (median) (max) detection (median) (max)
9 9 9
66.7 11.1 55.6
0.0007 0.0005 0.00003
9 9 9
55.6 11.1 100.0
0.002 0.0002 0.03
9 9 3 9 9 7
22.2 11.1 100.0 33.3 88.9 14.3
9
0.009 0.004 0.003
26
7.7
0.0001
0.0002
0.08 0.02 0.0003 0.09 0.006 0.0005
27 27 0.003 27 0.001 27 0.07 27 6 0.08 27 0.02 27 0.001 15 0.8 27 0.03 27 0.0005 22
3.7 3.7 3.7 11.1 88.9 83.3 29.6 7.4 40.0 11.1 63.0 4.6
0.000007 0.003 0.0004 0.00003 0.005 0.01 0.08 0.002 0.00002 0.03 0.003 0.0002
0.0002 0.02 0.0009 0.0002 0.02 0.02 0.1 0.02 0.0002 0.09 0.01 0.0005
66.7
0.0005
0.004
26 24 27 27 26 27 26 26 26 26 27 27 27
9 9 9 9 9 9 9 9
22.2 88.9 88.9 55.6 66.7 22.2 88.9 66.7
0.0002 0.05 0.0004 0.02 0.0003 0.0003 0.0005 0.0004
0.02 0.2 0.008 0.03 0.004 0.003 0.004 0.006
7.7 4.2 7.4 14.8 26.9 7.4 7.7 3.9 15.4 15.4 3.7 11.1 3.7
0.0002 0.03 0.0001 0.003 0.0002 0.001 0.0002 0.0002 0.0001 0.0002 0.0005 0.0001 0.0004
0.0005 0.04 0.002 0.006 0.002 0.003 0.0003 0.0003 0.0001 0.0004 0.004 0.0002 0.0005
9 9 8 9 9
66.7 11.1 25 44.4 11.1
0.0004 0.0001 0.01 0.01 0.001
0.002 0.0003 0.3 0.04 0.002
9
22.2
0.02
0.02
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
1,1,2-trichloroethane 79-00-5 1,2,4-trimethylbenzene 95-63-6 1,2-dichloropropane 78-87-5 1,2-dichlorobenzene 95-50-1 1,2-dichloroethane 107-06-2 1,2-dichloroethene, cis 156-59-2 1,3-dichlorobenzene 541-73-1 1,4-dichlorobenzene 106-46-7 Acrylonitrile 107-13-1 Benzene 71-43-2 Bromomethane 74-83-9 Carbon disulfide 75-15-0 Chloroethane 75-00-3 Chloromethane 74-87-3 Dichlorodifluoromethane 75-71-8 Dichloromethane 75-09-2 Ethylbenzene 100-41-4 MTBE 1634-04-4 Naphthalene 91-20-3 Tetrachloroethene 127-18-4 Toluene 108-88-3 Trichloroethene 79-01-6 m-xylene 108-38-3 n-butylbenzene 104-51-8 o-xylene 95-47-6 p-xylene 106-42- 3 1,2,3-trichlorobenzene 87-61-6 Chlorobenzene 108-90-7 tert butylbenzene 98-06-6 1,3,5-trimethylbenzene 108-67-8 1,3-dichloropropene 542-75-6 2-propyltoluene 28729-54-6 Styrene 100-42-5 Trichlorofluoromethane 75-69-4 (Freon 11) p-Isopropyltoluene 99-87-6
% of detection
Post-MF samples
LOD, limit of detection; CASR No, registry number for each chemical assigned by the Chemical Abstracts Service, a division of the American Chemical Society; SRU, Standard Relative Uncertainty; n, total number of samples; STE, secondary treated effluent; all values are expressed in mg/L; RQ, risk quotient; ND: not determined; IRIS, Integrated Risk Information System of the USEPA (IRIS, 1990 #64); TTC: Threshold of Toxicological Concern; USEPA Drinking Water Standards and Health Advisories (USEPA, 2011); CDPH: California Department of Public Health (CDPH, 2008, 2010); Australian Drinking Water Guidelines(ADWG) (NHMRC, 2011); TTC: Threshold of Toxicological Concern.
97
98
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
Table 3 e VOCs without detections in any of the samples and corresponding “worst-case scenario” RQs. Parameter 1,1,1,2-tetrachloroethane 1,1,1-trichloroethane 1,1,2,2-tetrachloroethane 1,1,2-trichloro-1,2, 2-trifluoroethane (Freon 113) 1,1-dichloroethane 1,1-dichloroethene 1,1-dichloropropene 1,2,3-trichloropropane 1,2,4-trichlorobenzene 1,2-dibromomethane 1,2-dibromo-3-chloropropane 1,2-dichloroethene, trans 1,2-dichloropropene 1,3-dichloropropane 2,2-dichloropropane 2-chlorotoluene (ortho) 4-chlorotoluene (para) Acrolein Bromobenzene Carbon tetrachloride Ethylene Dibromide (1,2-dibromoethane) Hexachlorobutadiene Isopropylbenzene Vinyl Chloride n-propylbenzene sec-butylbenzene
CASR No
Mean LOD (mg/L)
630-20-6 71-55-6 79-34-5 76-13-1
0.076 0.037 0.022 0.029
75-34-3 75-35-4 563-58-6 96-18-4 120-82-1 8003-07-4 96-12-8 156-60-5 563-54-2 142-28-9 594-207 95-49-8 106-43-4 107-02-8 108-86-1 56-23-5 106-93-4 87-68-3 98-82-8 75-01-4 103-65-1 135-98-8
SRU (%) (0.5 mg/L)
n
Tier
27.8 14.3 12.2 28.2
73 73 73 24
2 1 1 1
0.065 0.056 0.039 0.057 0.024 0.100 0.048 0.041 0.032 0.077 0.225 0.174 0.260 0.300 0.137 0.052
12.3 20.5 23.7 21.7 55.3 ND 31.3 21.5 19.1 23.8 31.9 41.3 83.6 ND 38.3 24.3
73 73 73 73 73 7 73 73 66 73 73 73 73 12 73 73
1 1 3 2 1 3 1 1 3 3 3 2 2 2 2 1
0.059 0.188 0.156 0.074 0.205 0.025
19.1 46.1 43.8 22.8 64.2 ND
66 73 73 73 73 7
1 1 2 1 3 2
Health value (mg/L)
Source
RQ
USEPA 2011 USEPA 2011 CDPH 2008 CDPH 2008
0.08 0.0002 0.02 0.00002
5 30 0.7 100 30 0.7 1 50 0.7 0.7 0.7 140 140 3.5 70 3
CDPH 2008 ADWG 2011 TTC USEPA 2011 ADWG 2011 TTC WHO 2011 WHO 2011 TTC TTC TCC CDPH 2010 CDPH 2010 OCS 2011 USEPA 2011 ADWG 2011
0.01 0.002 0.06 0.0006 0.0008 0.1 0.05 0.0008 0.05 0.1 0.3 0.001 0.002 0.09 0.002 0.02
0.4 0.7 770 0.3 7 260
WHO 2011 ADWG 2011 CDPH 2010 ADWG 2011 TTC CDPH 2010
0.15 0.3 0.0002 0.3 0.03 0.0001
1 200 1 1200
LOD: limit of detection; CASR No: registry number for each chemical, assigned by the Chemical Abstracts Service, a division of the American Chemical Society; SRU: Standard Relative Uncertainty; n: total number of samples; STE: secondary treated effluent; all values are expressed in mg/L; RQ: risk quotient; USEPA Drinking Water Standards and Health Advisories (USEPA, 2011); CDPH: California Department of Public Health (CDPH, 2008, 2010); Australian Drinking Water Guidelines (NHMRC, 2011); TTC: Threshold of Toxicological Concern; OCS: Office of Chemical Safety, Australian Government (Office of Chemical Safety, 2011); ND: not determined.
tetrachloroethene (87.9%), carbon disulfide (81.2%) and chloromethane (57.6%). Of the 21 VOCs detected, fourteen (14; 67%) were detected in less than 30% of the samples analysed, indicating that the presence of the compounds in STE is not consistent. Median concentrations for these compounds were dominated by non-detects, reported as their correspondent LOD. Seven (7) VOCs (i.e. 1,4-dichlorobenzene, cis-1,2-dichloroethene, carbon disulfide, chloromethane, tetrachloroethene, acrylonitrile and trichloroethene) had a percentage of detections greater than 30% across all samples (Table 2). For these compounds, median concentrations ranged from 0.81 mg/L for 1,4-dichlorobenzene to 0.02 mg/L for carbon disulphide. Comparison for the 7 VOCs with percentage detections greater than 30% showed that median concentrations were generally higher for influent STE from KWRP compared to BPP (ManneWhitney p ¼ 0.0001). For example, the median concentration of chloromethane at KWRP STE influent was double the median concentration at BBP STE influent (KWRP median ¼ 0.12 mg/L; BPP median ¼ 0.06 mg/L). The median concentration of tetrachloroethene in the STE entering KWRP was 5.4 times higher than the corresponding at BPP (KWRP median ¼ 2.2 mg/L; BPP median ¼ 0.41 mg/L). For 4 VOCs the concentrations in KWRP STE influent were statistically higher than BPP STE influent (Fig. 1), i.e. cis-1,2-dichloroethene
(ManneWhitney test, p ¼ 0.001), chloromethane ( p ¼ 0.004), tetrachloroethene ( p ¼ 0.02) and trichloroethene ( p ¼ 0.001). For 1,4-dichlorobenzene and acrylonitrile, the median concentrations were slightly higher at BPP STE influent but the differences were not statistically significant. Samples from Subiaco WWTP were not included in the comparison as fewer samples were taken at this location compared to KWRP and BPP. Seasonal comparison of median VOC concentrations is presented in Fig. 2. Again comparison is only made for compounds with percentage detections greater than 30% (i.e. 1,4-dichlorobenzene, cis-1,2-dichloroethene, carbon disulfide, chloromethane, tetrachloroethene and trichloroethene), except for acrylonitrile which was not included because it was only analysed during sampling event 3 (n ¼ 6, Table 2). Overall median VOC concentrations were higher in spring (0.125 mg/L) and winter (0.12 mg/L) than in summer (0.025 mg/L) and autumn (0.022 mg/L). These differences were statistically significant (Kruskal Wallis X2 p ¼ 0.0001). The VOCs: 1,4-dichlorobenzene, tetrachloroethene, and trichloroethene all had highest median concentrations in spring, whereas the highest median concentrations for cis-1,2dichloroethene and chloromethane were equal in spring and winter. The highest median concentration for carbon disulphide was recorded in winter, although it should be noted that it was not analysed in spring sampling events (Table 1).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
99
Fig. 1 e Median VOCs concentration in STE by plant in mg/L. *VOCs with statistically significant differences in concentrations between plants.
4.2.
Post-MF water
Twenty seven (27) VOCs were detected in post-MF samples, a higher number than in STE samples (21). Sixteen (16) of the 27 VOCs in the post-MF samples were also detected in STE (Table 2). Of the 27 VOCs detected in post-MF water, 18 were detected at KWRP only (Fig. 5) and 1 (i.e. 1,3-dichloropropene) was detected at BPP only. Eight (8) VOCs were detected at both locations: 1,2-dichlorobenzene, 1,4-dichlorobenzene, benzene, carbon disulfide, chloromethane, tetrachloroethene, toluene and o-xylene.
detections) followed by acrylonitrile (83.3%), chloromethane (62.9%) and carbon disulfide (40.0%), with respective median concentrations of 0.19 mg/L, 0.13 mg/L, 0.09 mg/L, and 0.02 mg/L. Five VOCs (i.e. ethyl benzene, naphthalene, n-butylbenzene, m-xylene and p-xylene) were detected in post-RO water and post-MF water but not in STE. Three VOCs were only detected in post-RO water (i.e. 1,2,3-trichlorobenzene, chlorobenzene and tert butylbenzene), although all with a percentage of detection below 11%. The percentage detection for all VOCs detected in post-RO water but not in STE was below 10% except for p-xylene (15.4%) (Table 2).
4.3.
4.4.
Post-RO water
A total of 26 VOCs (43%) were detected in post-RO water and 18 of these VOCs were also detected in STE (Table 2). The most commonly detected VOC was 1,4-dichlorobenzene (89.9%
Groundwater
Three (3) VOCs (i.e. 1,4-dichlorobenzene, benzene and toluene) were detected in groundwater. 1,4-dichlorobenzene (0.005 mg/L) was detected once in a sample taken from bore
Fig. 2 e Median VOCs concentration in STE by season in mg/L. * VOCs with statistically significant differences among seasons.
100
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
VOCs IARC Classification 2, 3%
4, 7%
1: Carcinogenic to humans 10, 16% 27, 44%
2A: Probably carcinogenic to humans 2B: Possibly carcinogenic to humans 3: Unclassifiable as to carcinogenicity in humans NE: Not evaluated
18, 30%
Fig. 3 e Distribution of the VOCs analysed according to the IARC cancer classification.
4.5.
Screening health risk assessment
Of 61 VOCs analysed, 34 (55.7%) were classified in tier 1, 17 (27.9%) were classified in tier 2 and the remaining 10 (16.4%) were classified in tier 3. The list of VOCs analysed, the corresponding tier, health value and calculated RQs are presented in Tables 2 and 3. The VOCs were classified according to the IARC cancer classification and the USEPA cancer classification. Almost half of the VOCs analysed had not been evaluated (27, 44.6%). Of the 34 VOCs evaluated: two (2) are classified as carcinogenic to humans
1, 1 1, ,2-T 2, 4- rich tri lo 1, me roe 2- th D ylb tha 1, ichl en ne 4- or ze di op n c e 1, hlo rop 1, 2-d ro an 2b di ich en e ch lo ze r 1, lor oe ne 3- oe th di th an 1, chlo en e e 2r di ob , c ch en is lo ro zen be e Ac nze ry n lo e ni t r Br B i om en le C om zen ar bo eth e n a C dis ne hl D u ic hl Ch oro lfide e or od loro tha iflu m ne e D oro tha ic ne m hl or eth om an et e Te ha tra ne ch lo MT ro B et E he Tr n ic To e hl l u or oe ene th e o- ne xy le ne
-200
Efficiency (%) 0 -100
100
line A during sampling event 4 (January 2008). A replicate sample from bore line A and a sample from bore line B taken the same day were below the LOD (0.003 mg/L). Benzene was also detected above LOD (0.04 mg/L) in all groundwater samples taken during sampling event 4. Two replicate samples from bore line A were 0.08 mg/L and 0.13 mg/L, while the concentration in the single bore line B sample was 0.1 mg/L. Toluene (0.54 mg/L) was detected once in a sample from bore line B during sampling event 2. The toluene concentration of bore line A sample taken on the same day was below LOD (0.13 mg/L).
Fig. 4 e Reverse osmosis removal efficiency of VOCs detected in secondary treated effluent.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
101
Fig. 5 e Median concentrations of VOCs in paired secondary treated effluent, post-MF water and post-RO water samples for KWRP and BPP.
(benzene and vinyl chloride); four (4) are classified as probably carcinogenic to humans (tetrachloroethene, trichloroethene, ethylene dibromide and 1,2,3-trichloropropane); ten (10) are classified as possible carcinogenic to humans (group 2B) and; eighteen (18) are unclassifiable as to carcinogenicity in humans (group 3) Fig. 3. No significant differences were observed in the IARC cancer classification distribution of the detected VOCs. Similarly, of the 61 VOCs analysed, 16 (26.2%) had not been evaluated by the USEPA. Two (2) VOCs are classified
as human carcinogens, seven (7) are classified as probable human carcinogens; seven (7) are classified as possible human carcinogens, and sixteen (16) are not classifiable as to human carcinogens. The RQ for the VOCs detected in any STE, post-MF or postRO sample are presented in Table 2. In STE samples, both RQ(max) and RQ(median) were always below 1. Most calculated RQ(max) were between one and three orders of magnitude below 1, whereas all RQ(median) were between one and
102
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
four orders of magnitude below 1. However, RQ(max) was only slightly below 1 for chloroethane (RQ ¼ 0.8); tetrachloroethene (RQ ¼ 0.2); MTBE (RQ ¼ 0.13) and; benzene (RQ ¼ 0.1). RQs for VOCs detected in post-MF water but not in STE or post-RO water (Table 2) were all below 1 with the highest RQ(max) ¼ 0.3 for 2-propyltoluene. In post-RO water, both RQ(max) and RQ(median) were again consistently below 1. The highest RQ(max) was for benzene (0.14), with all other values between one and three orders of magnitude below 1. For RQ(median), all values were between one and four orders of magnitude below 1. The results indicate that chemical concentrations measured in post-RO water are not of human health concern. A total of 26 VOCs were not detected in any STE, post-MF or post-RO samples (Table 3). RQs for undetected VOCs were calculated using the average LOD as a worst case scenario. For twenty (20) of the undetected VOCs the calculated RQs were between one and four orders of magnitude below 1. For the other 6 undetected VOCs, RQs were slightly higher but all remained below 1 ranging from 0.11 (1,3-Dichloropropane) to 0.32 (2,2-dichloropropane). The human health risk from these VOCs is therefore estimated to be negligible. This screening health risk assessment is based on VOCs concentrations in water for human consumption. However, for a comprehensive risk assessment, it is necessary to consider other routes of exposure, including inhalation and dermal uptake. In the Perth context, VOC concentrations in post-RO water are very low and any potential human risk from inhalation, dermal contact and ingestion will be further minimised by the retention time of the injected recycled water into the confined aquifer. Consequently, no human health risk is anticipated at the VOCs concentrations detected in the postRO water. With respect to other environmental sources, it is likely that diet accounts for some VOC exposure (FlemingJones and Smith, 2003), with inhalation accounting for a larger portion of human intake through VOCs emitted by cigarette smoke, vehicles, household products and industrial pollution. Consequently, risks to human health from VOCs in recycled water for IPR are likely to be negligible with negligible impact on public health compared to other sources of exposure.
5.
Discussion
5.1.
VOC concentrations in STE
A total of 21 VOCs were detected in STE, of which 14 (67%) were detected in less than 30% of the samples analysed. The low concentration and inconsistent occurrence of VOCs in STE in our study may indicate that (i) adequate industrial waste acceptance criteria are in place to limit or prohibit discharge of substances from commercial or industrial premises in Perth, and/or; (ii) WWTPs are able to effectively remove VOCs from raw wastewater. The concentration of VOCs detected in STE depends on the relative importance of removal processes during the wastewater treatment, such as adsorption onto sludge, chemical transformation, volatilization, and biodegradation. The results are consistent with previous studies showing low concentrations of VOCs in STE attributable to
significant reductions during municipal wastewater treatment (more than 90% removal) (Wu et al., 2002). Significant differences in the median concentrations of VOCs were found between the STE influent of KWRP and BPP. This may be related to the differences in level of contamination of the wastewater in the catchments and geographical variability in industrial activities. The higher VOCs concentrations seen in STE samples at KWRP compared to BPP may be related to the fact that KWRP is on the site of an oil refinery. This finding reinforces the importance of wastewater characterisation for projects considering IPR, given the different nature of industry, trade waste agreements/regulations, sewer arrangements and WWTP process in place. Our results are consistent with other studies indicating that the VOCs detected in a WWTP are closely related to the industrial activities in the catchment. For example Cheng et al. (2008) found that the more common VOCs detected in STE were as follows: acetone, acrylonitrile, methylene chloride, and chloroform for the petrochemical districts; acetone, chloroform, and toluene for the science-based districts; and chlorinated and aromatic hydrocarbons for the multiple industrial districts (Cheng et al., 2008). In contrast, Fatone et al. (2011) found that BTEX compounds (excluding benzene) to be the most commonly detected VOCs in five municipal WWTPs, assumed to result from vehicle emissions. Seasonal differences in some VOC concentrations were also observed, with higher concentrations observed in winter and spring compared to summer and autumn. VOCs are more likely to be stable and detectable in cold water because warm temperatures can cause VOCs to volatilize (Metz et al., 2007) and to more readily undergo degradation by the activated sludge process (Martı´nez et al., 2006). This seasonality was consistent when duplicate seasonal sampling was undertaken in summer and winter. The results also correspond with findings from air pollution studies that report higher concentrations of VOCs in air during summer (Millet et al., 2005).
5.2.
VOC concentrations in groundwater
Three VOCs (i.e. 1,4-dichlorobenzene, benzene and toluene) were detected in groundwater, just above the LOD. Toluene and benzene may indicate potential VOC contamination of groundwater, as previously reported in association with landfills and leaking underground petrol storage tanks (Zogorski et al., 2006). Given the limited number of samples and low concentrations detected, further investigation is required to confirm the presence of these compounds in Perth’s groundwater. VOCs have been frequently detected in shallow ground-water beneath urban areas (up to 90% of samples) (Hamilton et al., 2004). Samples taken during this study were a mixture of groundwater from shallow aquifers and deep, confined aquifers. In general, deep aquifers are less vulnerable than shallow aquifers to anthropogenic contaminants that originate on or near the land surface. In other locations, VOC contamination has been observed in public wells which draw on proportionately large volumes of groundwater situated below developed areas (Zogorski et al., 2006). However, in Western Australia, all public drinking water bores are protected by catchment protection reserves.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
5.3. The effect of MF/RO treatment on VOC concentration In both the BPP and KWRP, wastewater undergoes chloramination before MF to prevent RO membrane fouling. Samples of post-MF water were therefore analysed to determine the effect of chloramination during the MF/RO process. Paired wastewater, post-MF and post-RO samples were taken on 6 occasions at KWRP (1 in sampling event 1, 3 in sampling event 2, 1 in sampling event 3 and 1 in sampling event 6) and on 3 occasions at BPP (sampling events 3, 4 and 6). At KWRP, there were 21 analytes for which the highest median concentration was measured in a post-MF sample: chloromethane, trichlorofluoromethane (freon 11), carbon disulphide, cis-1,2-dichloroethene, trichloroethene, benzene, toluene, ethyl benzene, o-xylene, m-xylene, p-xylene, styrene, 1,3,5-trimethylbenzene, tert butylbenzene, 1,2,4trimethylbenzene 1,3-dichlorobenzene, 2-propyltoluene, pisopropyl toluene, 1,2-dichlorobenzene, n-butyl benzene, and naphthalene. Although the percentage detections for some of these analytes were low, 11 were present in more than 50% of post-MF samples: chloromethane (100%), cis-1,2dichloroethene (83%), toluene (83%), ethyl benzene (100%), oxylene (100%), m-xylene (100%), p-xylene (100%), styrene (67%), 1,3,5-trimethylbenzene (83%), 1,2,4-trimethylbenzene (100%), and naphthalene (83%). During this research project, chloramination was found to increase the concentration of some disinfection by-products (DBPs) during MF/RO treatment, particularly the halomethanes (DOHWA, 2009, Linge et al. in prep). Chloromethane is considered a disinfection by-product (Krasner et al., 2006) and may have also been formed by chloramination. However, cis-1,2-dichloroethene is a chlorinated solvent and would not be expected to be formed by disinfection. All of the other nine VOCs frequently detected in post-MF samples are aromatic compounds associated with gasoline or diesel exhaust (Elbir et al., 2007; Watson et al., 2001) or with oil refinery emissions (Chen et al., 2006; Scheff and Wadden, 1993). The KWRP MF/RO plant is located on the site of an oil refinery that produces petrol, diesel, liquefied petroleum gas, aviation gasoline, jet fuel and bitumen and it is likely that trace concentrations of associated compounds would be found in water samples from KWRP. The low concentrations measured in post-RO water and field blanks compared to post-MF water suggests that this contamination did not occur during sampling, but most likely occurred during the MF treatment where the water is exposed to the atmosphere for about 25 min. A number of the tanks also have air vents that may enable some exposure to the atmosphere both before MF and after RO of up to an hour, although the vents are less likely to be a significant source of exposure compared with the open tanks. At BPP, there were only 4 VOCs for which the highest median concentration was measured in a post-MF sample: i.e. carbon disulphide (66%), toluene (100%), 1,3-dichloropropene (33%), and 1,2-dichlorobenzene (66%). While they are not typically measured as disinfection by-products, dichlorobenzenes are produced from the chlorination of benzene (IARC, 1999) and therefore it is possible that they may form, petroleum-based contamination. However the source of these VOCs is not obvious at this time.
5.4.
103
Treatment performance
Overall treatment efficiency was calculated as a proportion of removal, comparing STE and post-RO samples that were matched for plant and date. For those parameters reported below LOD after RO, the efficiency was calculated assuming a concentration equal to half the LOD. Very high variability in the removal of VOCs was observed, as illustrated in Fig. 4. For ten (10) of the twenty one (21) VOCs detected in STE, the median removal efficiency was above 70%. The median removal efficiency ranged from 77.0% for dichlorodifluoromethane to 91.2% for tetrachloroethene. For 17 samples, corresponding to 6 VOCs (1,4-dichlorobenzene ¼ 1 sample, benzene ¼ 4, carbon disulfide ¼ 3, chloromethane ¼ 7, dichlorodifluoromethane ¼ 1, o-xylene ¼ 1), the concentrations in post-RO samples were higher than their paired STE samples. For 10 (59%) of these paired samples, the concentration in post-RO water was not statistically different from the STE: the difference was within the uncertainty of the analytical method and therefore calculation of removal efficiency using these data is inconclusive. Differences were seen (outside of the limits of uncertainty) for carbon disulphide (1 sample), chloromethane (4 samples), dichlorodifluoromethane (1 sample), and o-xylene (1 sample). As discussed in Section 4.3, elevated concentrations of carbon disulphide and o-xylene occurred in post-MF samples, due to atmospheric exposure to oil refinery emissions during MF treatment at KWRP. While post-RO water did not have similar levels of atmospheric exposure, the volatile nature of the compounds studied mean it is possible that there were individual occasions where trace contamination from atmospheric exposure occurred. Therefore the increased concentrations seen in a few post-RO samples may be a result of similar atmospheric exposure. Chloromethane was also elevated in post-MF samples. Reverse osmosis is the only barrier in the treatment process for chemical removal of VOCs and it is apparent that calculation of RO chemical removal by comparing STE and post-RO samples can be confounded by potential impacts from chloramination and atmospheric contamination, as described above. Thus the treatment efficiency of RO alone was determined using paired post-MF and post-RO samples by plant (Fig. 6). Calculations confirmed that the degree of RO treatment efficiency was higher than overall recycling plant treatment efficiency for all VOCs detected in STE with the exception of chloroethane, tetrachloroethene, toluene and dichlorodifluoromethane. Furthermore, by using post-MF data, treatment efficiency could be calculated for 10 additional VOCs, which were measured in post-MF samples but were not present in STE. As shown in Fig. 6, the majority of positive VOCs in post-MF water occurs at KWRP (n ¼ 26) compared to BPP (n ¼ 9). Variability in RO treatment efficiency (calculated using post-MF and post-RO data) as represented by standard deviation, fell slightly but remained relatively high, although this may relate to the smaller number of paired samples available for analysis. The rejection of VOCs in a range of different RO membranes has been reported as highly variable. For some VOCs (1,1,1-trichloroethane, carbon tetrachloride, p-, m- and o-xylenes, tetrachloroethylene, 1,2-dichlorobenzene),
104
60 40
1, 2, 4t 1, rim 2- et D h 1 ic ylb 1, ,2- hlo en 2- di ro ze 1, dic chlo pro ne 3, hl ro p 5- or b an tr oe e e 1, ime the nze 3- th n n 1, Dic ylb e, e 3- h e ci d lo n s 1, ich ro zen 4- lo pr e di ro op ch be e lo n n 2- ro ze e p r be n e op nz yl en to Br B lue e om e ne C o nz ar m e bo e ne t C n d han hl is e C or ul hl oe fid or t e o h Et me ane hy th l a N ben ne ap z ht en Te ha e tra ch S len lo tyr e ro en Tr et e ic T he hl ric or h To ne of lo lu lu ro e or et ne om he e ne n- m tha b u - ne ty xyl lb en pen e Is op o ze ro -x ne py yle lto n lu e p- en xy e le ne
0
20
Efficiency (%)
80
100
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
Beenyup Pilot Plant
Kwinana Water Reclamation Plant
Fig. 6 e RO removal efficiency of VOCs detected in post-MF wastewater by plant.
rejection has been reported as higher than 90% (Agenson et al., 2003), but for other VOCs rejection has been reported as much lower (e.g. 6e54% for benzene). Using paired post-MF and post-RO samples, 56% of analytes in this study had rejection efficiency greater than 80% e including 1,4-dichlorobenzene (87%) , while 37% had rejection efficiency greater than 90%. Lower median rejection was associated with benzene (56%), bromomethane (48%) and styrene (63%). It has been found that rejection of VOCs is influenced by solute size, molecular charge, branching of functional groups and Kow (Agenson et al., 2003). VOCs with poorer rejection usually have smaller molecular width and length, and lower Kow. VOCs are not highly hydrophobic (logKow < 3), and Kow has been found to influence rejection, which suggests that there is some degree of interaction between the solute and the membrane.
5.5.
Identification of treatment performance Indicator
The use of chemical indicators in recycled water has been proposed for occurrence monitoring and for assessing treatment process performance assessment (Benotti et al., 2009; Dickenson et al., 2011; Drewes et al., 2008). A treatment performance indicator is an individual chemical occurring at quantifiable level that represents a family of trace constituents with certain physicochemical and biodegradable characteristics that are relevant to fate and transport during treatment, in this case RO. A treatment performance indicator should provide a conservative assessment of removal of represented parameters, and should be sensitive to minor changes in RO treatment performance.
The criteria for selection of a Treatment Performance Indicator are: Quantifiable using an established and preferably accredited analytical method; Frequently detected in feed water, preferably 100% detections; Present in feed water at significant concentrations; generally greater than five times LOD; The selection of a chemical indicator for VOCs is challenging given the diversity of compounds in this chemical group. However, it is neither practical nor feasible to assess for all potential VOCs present in recycled water during routing monitoring. In our study, 1,4-dichlorobenzene was identified as a potential treatment performance indicator for assessment of VOCs in IPR using RO treatment (Blair et al., 2010). 1,4-dichlorobenzene was detected in almost 94% of the STE samples and with a median concentration that was over 30 times the average method LOD, therefore fulfilling the criteria for selection of a Treatment Performance Indicator. While the median treatment efficiency by MF/RO treatment was relatively high (87%), the low method LOD meant that 1,4-dichlorobenzene was still detected in almost 89% of the post-RO water samples, ensuring accurate estimates of treatment performance could be calculated. The high frequency of detection of 1,4-dichlorobenzene in STE is attributed to its long history of domestic use in toilet products, moth repellents, and mildew control agents (Aronson et al., 2007; NICNAS, 2000). Dichlorobenzene has also been
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
reported as one of the less biodegradable VOCs in wastewater treatment (Koe and Shen, 1997). As a small (molecular weight ¼ 147 g/mol) and uncharged molecule, the RO rejection of 1,4-dichlorobenzene is mostly attributed to its relatively high logKow (3.4) compared to other VOCs. The relevance of 1,4-dichlorobenzene as a chemical indicator of VOCs will be further validated during the Groundwater Replenishment Trial (GWRT) (DOHWA, 2009). During the three year trial, 8 ML/day of STE from the Beenyup WWTP will be treated by MF/RO and ultra violet disinfection before injection into a confined aquifer. If successful, this approach could significantly reduce the analytical cost and complexity of monitoring water treatment systems performance for removal of VOCs.
6.
Conclusions
The screening health risk assessment indicates that the individual VOCs measured in recycled water have a low potential to affect humans from long-term consumption after RO treatment. Detection of VOCs in STE can occur as a result of the widespread use of these compounds. However, the impact on potable supplies through augmentation with recycled water treated by MF/RO is likely to be negligible at the concentrations observed in Perth. Calculated MF/RO treatment removal was variable, with some concentrations in post-RO water higher than in the STE. For some VOCs, this may be due to uncertainty in the analytical method. However, for others it is attributed to industrial contamination during the MF/RO process, or formation during chloramination. For most VOCs, RO treatment efficiency was higher than overall MF/RO treatment efficiency, however more analysis of VOCs before and after RO treatment is recommended to better characterise the RO treatment variability. Management of risks in IPR is dependent on advanced treatment technologies and comprehensive risk management approaches to ensure compliance with drinking water guidelines. Frequent monitoring of a treatment performance indicator such as 1,4-dichlorobenzene (in conjunction with the continuous online monitoring of critical control points) during the MF/RO treatment process is likely to be sufficient to ensure adequate removal of VOCs. If successful, this approach has the potential for significantly reducing the analytical costs and complexity of monitoring water treatment systems performance for removal of VOCs, and other chemical groups.
Acknowledgements This research was conducted as part of the Premiers Collaborative Research Project (PCRP) a collaborative effort between Department of Health, Department of Water, Department of Environment & Conservation, Water Corporation of Western Australia, The University of Western Australia, Curtin University of Technology, Chemistry Centre WA and the National Measurement Institute (NMI). State
105
government funding was provided by the Office of Science & Innovation.
references
Agenson, K., Oh, J.-I., Urase, T., 2003. Retention of a wide variety of organic pollutants by different nanofiltration/reverse osmosis membranes: controlling parameters of process. Journal of Membrane Science 225 (1e2), 91e103. Aronson, D., Bosch, S., Gray, D.A., Howard, P., Guiney, P., 2007. A comparative human health risk assessment of pdichlorobenzene-based toilet rimblock products versus fragrance/surfactant-based alternatives. Journal of Toxicology and Environmental Health Part B: Critical Reviews 10 (7), 467e526. Atasoy, E., D, T., K, S., 2004. The estimation of NMVOC emissions from an urban-scale wastewater treatment plant. Water Research 38 (14e15), 3265e3274. Battistoni, P., Cola, E., Fatone, F., Bolzonella, D., Eusebi, A.L., 2007. Micropollutants removal and operating strategies in ultrafiltration membrane systems for municipal wastewater treatment. Industrial & Engineering Chemistry Research 46 (21), 6716e6723. Benotti, M., Trenholm, R., Vanderford, B., Holady, J., Stanford, B., Snyder, S., 2009. Pharmaceuticals and endocrine disrupting compounds in U.S. drinking water. Environmental Science & Technology 43 (3), 597e603. Blair, P., Linge, K., Rodriguez, C., Busetti, F., Drewes, J., Turner, N., 2010. Development of the System to Manage Chemical Risk During Indirect Potable Reuse. Proceedings Water Reuse & Desalination Water Reuse Association, Sydney. Boyes, W.K., Bushnell, P.J., Crofton, K.M., Evans, M., Simmons, J.E., 2000. Neurotoxic and pharmacokinetic responses to trichloroethylene as a function of exposure scenario. Environmental Health Perspectives 108 (Suppl 2), 317e322. Brouwer, D.H., de Pater, N.A., Zomer, C., Lurvink, M.W., van Hemmen, J.J., 2005. An experimental study to investigate the feasibility to classify paints according to neurotoxicological risks: occupational air requirement (OAR) and indoor use of alkyd paints. Annals of Occupational Hygiene 49 (5), 443e451. CDPH, 2008. Maximum Contaminant Levels and Regulatory Dates for Drinking Water U.S. EPA vs California, California Department of Public Health, California. CDPH, 2010. Drinking Water Notification Levels and Response Levels. California Department of Public Health, California. Chen, C.-L., Shu, C.-M., Fang, H.-Y., 2006. Location and characterization of emission sources for airborne volatile organic compounds inside a refinery in Taiwan. Environmental Monitoring and Assessment 120, 487e498. Cheng, W.-H., Hsu, S.-K., Chou, M.-S., 2008. Volatile organic compound emissions from wastewater treatment plants in Taiwan: legal regulations and costs of control. Journal of Environmental Management 88 (4), 1485e1494. Dickenson, E.R.V., Snyder, S.A., Sedlak, D.L., Drewes, J.E., 2011. Indicator compounds for assessment of wastewater effluent contributions to flow and water quality. Water Research 45 (3), 1199e1212. DOHWA, 2009. Premier’s Collaborative Research Program (20052008). Characterising Treated Wastewater for Drinking Purposes Following Reverse Osmosis Treatment. Department of Health, Western Australia (DOHWA), Perth, WA. Drewes, J.E., Sedlak, D., Snyder, S., Dickenson, E., 2008. Development of Indicators and Surrogates for Chemical Contaminant Removal During Wastewater Treatment and Reclamation. Water Reuse Foundation, Alexandria, VA.
106
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 9 3 e1 0 6
Elbir, T., Cetin, B., Cetin, E., Bayram, A., Odabasi, M., 2007. Characterization of volatile organic compounds (VOCs) and their sources in the air of Izmir, Turkey. Environmental Monitoring and Assessment 133 (1e3), 149e160. European Commission, 1998. Guidelines of water intended for human consumption. Official Journal of the European Communities. Fatone, F., Di Fabio, S., Bolzonella, D., Cecchi, F., 2011. Fate of aromatic hydrocarbons in Italian municipal wastewater systems: an overview of wastewater treatment using conventional activated-sludge processes (CASP) and membrane bioreactors (MBRs). Water Research 45 (1), 93e104. Fleming-Jones, M.E., Smith, R., 2003. Volatile organic compounds in foods: a five year study. Journal of Agricultural and Food Chemistry 51, 8120e8127. FPT Committee on Drinking Water, 2010. Guidelines for Canadian Drinking Water Quality. Federal-Provincial-Territorial Committee on Drinking Water, Ontario. Hamilton, P.A., Miller, T.L., Myers, D.N., 2004. Water Quality in the Nation’s Streams and AquiferseOverview of Selected Findings, 1991e2001. U.S.Geologycal Survey, Reston, Va, p. 20. Herpin, G., Gargouri, I., Gauchard, G.C., Nisse, C., Khadhraoui, M., Elleuch, B., Zmirou-Navier, D., Perrin, P.P., 2009. Effect of chronic and subchronic organic solvents exposure on balance control of workers in plant manufacturing adhesive materials. Neurotoxicity Research 15 (2), 179e186. IARC, 1999. Some Chemicals that Cause Tumours of the Kidney or Urinary Bladder in Rodents and Some Other Substances, p. 674. International Agency for Research on Cancer, Lyon, France. IRIS 1,1,2,2-Tetrachloroethane; CASRN 79-34-5; 09/30/2010 USEPA, Integrated Risk Information System. IRIS Carbon disulfide; CASRN 75-15-0; 09/30/2010 USEPA, Integrated Risk Information System. Koe, L.C.C., Shen, W., 1997. High resolution GC e MS analysis of VOCs in wastewater and sludge. Environmental Monitoring and Assessment 44 (1), 549e561. Krasner, S.W., Weinberg, H.S., Richardson, S.D., Pastor, S.J., Chinn, R., Sclimenti, M.J., Onstad, G.D., Thruston Jr., A.D., 2006. Occurrence of a new generation of disinfection byproducts. Environmental Science & Technology 40 (23), 7175e7185. Linge, K.L., Blythe, J.W., Busetti, F., Blair, P. and Heitz, A. (in prep) Formation of halogenated disinfection by-products during wastewater recycling employing microfiltration and reverse osmosis. Water Research. Martı´nez, S.A., Rodrı´guez, M.G., Morales, M.A., 2006. Stability analysis of an activated sludge bioreactor at a petrochemical plant at different temperatures. International Journal of Chemical Reactor Engineering 3. Metz, P.A., Delzer, G.C., Berndt, M.P., Crandall, C.A., Toccalino, P. L., 2007. Anthropogenic Organic Compounds in Ground Water and Finished Water of Community Water Systems in the Northern Tampa Bay Area, Florida, 2002e04, p. 48Report 2006e5267. U.S. Geological Survey Scientific Investigations, Reston, Virginia. Millet, D.B., Donahue, N.M., Pandis, S.N., Polidori, A., Stanier, C.O., Turpin, B.J., Goldstein, A.H., 2005. Atmospheric volatile organic compound measurements during the Pittsburgh Air Quality Study: results, interpretation, and quantification of primary and secondary contributions. Journal of Geophysical Research 110 D07S07. NHMRC, 2004. Australian Drinking Water Guidelines, p. 615. National Health and Medical Research Council and Natural Resource Management Ministerial Council, Artarmon, NSW.
NHMRC, 2011. Australian Drinking Water Guidelines. National Health and Medical Research Council and Natural Resource Management Ministerial Council, Canberra ACT. NICNAS, 2000. In: Australia, C.o. (Ed.), Summary: Priority existing chemical assessment reports. PEC No. 13. paradichlorobenzene. National Industrial Chemicals Notification and Assessment Scheme, Australian Goverment, Canberra ACT. NRC, 1996. Use of Reclaimed Water and Sludge in Food Crop Production. National Academic Press, Washington. NWC, 2011. Environmental Health Management. National Water Commission, Australian Goverment, Canberra ACT. Office of Chemical Safety, 2011. Acceptable Daily Intakes for Agricultural and Veterinary Chemicals. Office of Health Protection, Department of Health and Ageing, Canberra. Patterson, B.M., Pribac, F., Barber, C., Davis, G.B., Gibbs, R., 1993. Biodegradation and retardation of PCE and BTEX compounds in aquifer material from Western Australia using large-scale columns. Journal of Contaminant Hydrology 14 (3e4), 261e278. Rodriguez, C., Cook, A., Van Buynder, P., Devine, B., Weinstein, P., 2007a. Screening health risk assessment of micropollutants for indirect potable reuse schemes: a three-tiered approach. Water Science and Technology 56 (11), 35e42. Rodriguez, C., Weinstein, P., Cook, A., Devine, B., Buynder, P.V., 2007b. A proposed approach for the assessment of chemicals in indirect potable reuse schemes. Journal of Toxicology and Environmental Health, Part A: Current Issues 70 (19), 1654e1663. Scheff, P.A., Wadden, R.A., 1993. Receptor modeling of volatile organic compounds. 1. Emission inventory and validation. Environmental Science & Technology 27 (4), 617e625. Stata Corp, 2007. In: C. (Ed.), Stata Statistical Software Station. TX: StataCorp LP, Texas. Toccalino, P.L., Rowe, B.L., Norman, J.E., 2006. Volatile Organic Compounds in the Nation’s Drinking-water Supply Wells e what Findings May Mean to Human Health. U.S. Geological Survey, Sacramento, California, p. 4. USEPA, 1994. Emerging Technology Report Cross-flow Pervaporation System for Removal of VOCs from Contaminated Wastewater, p. 129. Risk Reduction Engineering Laboratory, Office of Research and Development, U.S. EPA, Cincinnati, Ohio. USEPA, 2011. 2011 Edition of the Drinking Water Standards and Health Advisories. Office of Water U.S. Environmental Protection Agency, Washington, DC. Watson, J.G., Chow, J.C., Fujita, E.M., 2001. Review of volatile organic compound source apportionment by chemical mass balance. Atmospheric Environment 35 (9), 1567e1584. WHO, 2011. Guidelines for Drinking-water Quality. World Health Organization, Geneva, Switzerland. Williams, P., Benton, L., Warmerdam, J., Sheehans, P., 2002. Comparative risk analysis of six volatile organic compounds in California drinking water. Environmental Science & Technology 36 (22), 4721e4728. Wright, M., Belitz, K., Burton, C., 2005. California GAMA Program: Ground-water Quality Data in the San Diego Drainages Hydrogeologic Province. U.S. Geological Survey, Sacramento, California, p. 102. Wu, C., Lu, J., Lo, J., 2002. Analysis of volatile organic compounds in wastewater during various stages of treatment for hightech industries. Chromatographia 56 (1), 91e98. Zogorski, J., Carter, J.M., Ivahnenko, T., Lapham, W.W., Moran, M.J., Rowe, B.L., Squillace, P.J., Toccalino, P.L., 2006. Report on Volatile Organic Compounds in the Nation’s Ground Water and Drinking-water Supply Wells, p. 112. U.S. Department of the Interior, U.S. Geological Survey, Reston, Virginia.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Evaluation of the flocculation performance of carboxymethyl chitosan-graft-polyacrylamide, a novel amphoteric chemically bonded composite flocculant Zhen Yang a, Bo Yuan a, Xin Huang a, Junyu Zhou a, Jun Cai a, Hu Yang a,*, Aimin Li b, Rongshi Cheng a a
Key Laboratory for Mesoscopic Chemistry of MOE, Department of Polymer Science & Technology, School of Chemistry & Chemical Engineering, Nanjing University, Nanjing 210093, PR China b State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing 210093, PR China
article info
abstract
Article history:
In the present work, a novel amphoteric chemically bonded composite flocculant (car-
Received 14 July 2011
boxymethyl chitosan-graft-polyacrylamide, denoted as CMC-g-PAM) was successfully
Received in revised form
prepared and used to flocculate the kaolin suspension. The flocculation performance of
22 September 2011
CMC-g-PAM in acidic, neutral, and alkaline conditions was systematically evaluated by
Accepted 16 October 2011
light scattering in combination with fractal theory, as well as by traditional turbidity and
Available online 25 October 2011
zeta potential measurements. Based on the experimental facts from in situ size and fractal dimension measurements, different flocculation mechanisms play key roles at various pH levels, resulting in substantially varied flocculation kinetic processes under three pH
Keywords: Amphoteric
chemically
bonded
conditions. In acidic condition, patching was the main mechanism involved in the opposite
composite flocculant
zeta potential between CMC-g-PAM and the kaolin suspension. A flat configuration was
Flocculation kinetics
favored when the polymeric flocculant was adsorbed onto the particle surface, leading to
Flocculation mechanism
a slower initial floc growth rate but larger and denser flocs. Bridging was the dominant
Fractal dimension
mechanism in neutral and alkaline conditions. A faster initial rate of bridging resulted in
Light scattering
smaller and more open floc structures. A rearrangement process in neutral pH subsequently led to more compact flocs, whereas no restructuration of flocs occurred in alkaline conditions because of the electrostatic repulsion of the same negative charges on the flocculant and particles. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Flocculation is an important industrial process for solideliquid separation during the primary purification of wastewater. Attention has been increasingly paid to natural polymer flocculants owing to their biodegradability, wide availability, and environment-friendliness (Renault et al., 2009). Chitosan, poly-b(1 / 4)-2-amino-2-deoxy-D-glucose, obtained from the deacetylation of chitin (the second most abundant natural polymer), is
one of the most outstanding candidates among these flocculants (Guibal et al., 2006; Renault et al., 2009). This polymer presents an abundance of free amino groups along its backbone chain that are positively charged in acidic media and show prominent flocculation performance. However, the poor water solubility (Wang et al., 2009) of chitosan at neutral or higher pH remains a limitation in its practical application. Given that this polymer has high tailorability for an abundant number of functional groups on its
* Corresponding author. Tel.: þ86 25 83686350; fax: þ86 25 83317761. E-mail address:
[email protected] (H. Yang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.024
108
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4
backbone, chemically modified chitosan materials have been manufactured to overcome this limitation. Among these materials, amphoteric chitosan (Yang et al., 2011a, b), containing both cations and anions on the backbone chain, can easily convert chitosan into a water-soluble form over wide pH ranges. Considering that chitosan itself is a cationic polyelectrolyte, the simplest amphoteric chitosan is carboxyalkyl chitosan (Chen and Park, 2003). However, at high pH, carboxyalkylated chitosan also suffers from a drawback: enhanced electrostatic repulsion against negatively charged particles that commonly exist in natural turbid water. Grafting provides an effective method (Sand et al., 2010; Wang et al., 2008) for overcoming this limitation. Synthetic polymeric flocculants, such as polyacrylamide (PAM), have already been proven to be useful in water purification, although they exhibit defects such as poor shear resistance and nonbiodegradability (Biswal and Singh, 2004; Sen and Pal, 2009). Therefore, amphoteric chemically bonded composite flocculants of natural and synthetic polymers are expected to combine the advantages of both variants (Biswal and Singh, 2004; Sen et al., 2009) through graft copolymerization. On the one hand, flexible synthetic polymer chains on comb-like flocculants (Sen et al., 2009) have better approachability (Tripathy et al., 2009) than carboxyalkylated chitosan to suspended particles in water. On the other hand, compared with synthetic polymers, new tailor-made chitosan-based materials can be shear stable and controllably biodegradable (Biswal and Singh, 2004). Generally, for the aforementioned amphoteric flocculants, different pH levels result in various changes in the densities and morphologies of the polymer chains (Yang et al., 2011b). Hence, similar to the fact that various Al(III) species of aluminum-containing flocculants at various pH levels lead to different flocculation mechanisms (Lin et al., 2008), amphoteric flocculants at various pH levels also undergo diverse flocculation mechanisms. Thus, a good understanding of these mechanisms is important for industrial applications to optimize the flocculation process (Yu et al., 2010). Traditional approaches for studying flocculation mechanisms are usually indirect (Zhou and Franks, 2006), and use finalresult-based methods (Yu et al., 2006), such as turbidity, sedimentation rate, or zeta potential (ZP). In these methods, only the overall parameters are obtained, and no direct information about floc properties is measured. This is despite the fact that floc properties are crucial physical properties that should be studied to obtain a better understanding of the flocculation mechanism as well as control the separation process (Fabrizi et al., 2010; Wei et al., 2009). In past decades, several microscopic approaches (Thomas et al., 1999) to model the floc properties and flocculation process have been formulated based on Smoluchowski’s collision theory (von Smoluchowski, 1917). However, considering that the principal assumptions of this theory cannot be satisfied in a real flocculation process, these approaches present a number of shortcomings (Thomas et al., 1999). Therefore, an alternative way is to focus on macroscopic measurements. One such approach is the fractal theory, which considers that irregular and porous flocs have been found to be geometrically fractal during flocculation (Mandelbrot, 1983). In the fractal concept, the most important and powerful quantitative parameter is the fractal dimension (Zhou and Franks, 2006), which indicates the space-
filling capacity (Thomas et al., 1999), that is, the compactness, of the floc. Larger fractal dimensions signify more compact flocs, which are usually preferred in most situations in water treatment to yield lower sludge volumes and easier sedimentation. Fractal dimensions can be expressed depending on the space dimension. The two commonly used fractal dimensions are two-(D2) and three-dimensional (i.e., mass fractal dimension, or DF) presentations. D2 is defined by the power law relationship between the projected area (A) and the characteristic length (l ), which is usually measured by image analysis (IA) (Liao et al., 2006). DF is the power law relationship between mass (m) and l, which can be obtained by light scattering (LS). Compared with IA, LS is more time-saving, simpler, and more accurate. LS tests can also easily supply average size and size distribution, as well as DF, in a continuous manner, if the equipment is pre-programmed for continuous data acquisition (Rasteiro et al., 2008). Therefore, the flocculation kinetics (i.e., the complicated stages of a flocculation process), which has been recognized as more deeply related to flocculation mechanism (Yu et al., 2006; Zhou and Franks, 2006), can be evaluated together in a single experiment. In the present study, a novel amphoteric chemically bonded composite flocculant, carboxymethyl chitosan-graftpolyacrylamide (CMC-g-PAM), is prepared. Given that the new composite flocculant has better solubility over a wide pH range than other flocculants, the flocculation performance of CMC-g-PAM is investigated using turbidity and ZP methods at various pH levels. Considering that few studies on the application of fractal theory on biodegradable natural polymerbased flocculants have been reported, IA and LS methods are introduced to evaluate the flocculation performance of CMCg-PAM. Using online data acquisition of LS, the flocculation kinetics and mechanisms at various pH levels are studied in detail, and various stages during the flocculation process are finally proposed.
2.
Materials and methods
2.1.
Materials
Chitosan, with 85.2% degree of deacetylation and 8.34 105 g/mol viscosity-average molecular weight as calculated from the intrinsic viscosity (Wang et al., 1991), was purchased from Shangdong Aokang Biological Co., Ltd. Monochloroacetic acid (Zibo Lushuo Economic Trade Co., Ltd.), acrylamide (Nanjing Chemical Reagent Co., Ltd.), and ceric ammonium nitrate (CAN) (Sinopharm Chemical Reagent Co., Ltd.) were used without further purification. Kaolin, with a mean particle size of 7.4 mm, was purchased from Sinopharm Chemical Reagent Co., Ltd. Polyaluminium chloride (PAC) and PAM (weight-average molecular weight: 1.20 107 g/mol) were purchased from Guangzhou Yurun Chemical Technology Co., Ltd. All other chemicals were purchased from Nanjing Chemical Reagent Co., Ltd.
2.2.
Preparation of CMC-g-PAM
The preparation route of CMC-g-PAM is shown in Fig. S1 of the supporting information. CMC was synthesized using the same method as described in the authors’ previous work (Yang et al.,
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4
2011b), and the substitution degree of the carboxymethyl groups was 48.3%, as calculated from 1H nuclear magnetic resonance (1H NMR) spectrum. Graft copolymerization for the preparation of CMC-g-PAM was performed as follows: A specific amount of solid CMC was dissolved in 400 mL of 1 wt% hydrochloric acid. After stirring under N2 for 30 min, a certain amount of CAN as the initiator was added into the CMC solution. The solution was then kept for 5 min for pretreatment of the CMC by the initiator, followed by the dropwise addition of an acrylamide monomer aqueous solution into the bottle for 20 min. The pretreatment of CMC by the initiator could efficiently suppress the formation of the PAM homopolymer (Sonmez et al., 2002). After 3 h of reaction under N2, copolymerization was stopped, and the samples were precipitated in acetone. The solid product was filtered, washed, extracted using acetone as the solvent in a Soxhlet apparatus for 72 h to remove impurities in the sample (Ceresa, 1973; Sonmez et al., 2002; da Silva et al., 2007; Ali and Singh, 2009) and then finally vacuum dried at room temperature. The final product was CMC-g-PAM, with a weight-average molecular weight of 3.10 106 g/mol, as calculated from its 1H NMR spectrum.
2.3.
Characterization
Fourier transform infrared (FTIR) spectra were recorded using a Bruker Model IFS 66/S FTIR spectrometer. The interval of tested wave numbers was 650e4000 cm1. 1H NMR spectra were recorded on a Bruker AVANCE Model DRX-500 spectrometer, operating at 500 MHz, in a mixed solvent composed of CF3COOD and D2O with a mass ratio of 1:1. ZP was measured using a Malvern Model Nano-Z Zetasizer. Hydrodynamic radius (Rh), a parameter commonly used to characterize the size of the polymer chains in solution, was determined by a Brookheaven Model BI200SM dynamic light scattering apparatus.
2.4.
Flocculation experiment
2.4.1.
Flocculation procedure
Floc size measurement
Floc sizes were analyzed by LS using a Malvern Mastersizer 2000 system. Jar tests were conducted in the same process as described in Section 2.4.1. The suspension was measured by continuous recycling of water flowing through the sample cell of the instrument. The apparatus was settled as described in the work of Bushell and Amal (2000). Floc size was defined by volume-weighted mean diameter (Dv), calculated and provided by the data processing software, P 4 d Dv ¼ P 3 (2) d where d is the diameter of the equivalent sphere of each single floc, according to the Mie scattering theory.
2.4.3.
Determination of fractal dimension
D2 was measured by IA and determined from the slope of the logelog plot of A and l (Chakraborti et al., 2000, 2003) AflD2
(3)
In the present study, the largest projection length was taken as l. Afterward, the flocs were carefully withdrawn from the jar and placed into a glass dish with water. An optical microscope (model: Leica DMLP) equipped with a digital camera (Victor Company of Japan, Ltd.) was used to take photos. A and l were derived from the photos using Imagepro Plus 6.0 software. DF was measured by LS. The theory of the LS technique has been reported in the literature (Rasteiro et al., 2008; Yu et al., 2006; Zhou and Franks, 2006) and was used in the present study. In this technique, two parameters, light intensity (I ) and scatter vector (Q), were measured. Q is given by the following equation (Schaefer et al., 1984): Q¼
4pnsinðq=2Þ l
(4)
where n, q, and l are the refractive indices of the medium, scattered angle, and wavelength of radiation, respectively. DF was then determined from the negative slope of logelog plot of I and Q (Jarvis et al., 2005),
Kaolin (1 g) was added into 1 L of distilled water. After 3 min of ultrasonic stirring at 200 rpm, the suspension was used as synthetic water. The pH levels were modified by adding HCl or NaOH solution. The stock solutions of the flocculants were always freshly prepared in distilled water before each flocculation test. Jar tests were conducted using 1-L jars and a six-place programmed paddle mixer model of TA6 (Wuhan Hengling Tech. Co. Ltd.) at room temperature. The detailed procedure consisted of an initial period of rapid mixing for 5 min at 200 rpm, followed by 15 min of slow mixing at 50 rpm, and finally settling for 40 min. After this procedure, samples collected from the supernate were analyzed for turbidity and ZP. Turbidity was measured by the Turbidity Indicator model of ATZ-A22 (Wuxi Guangming Instrument Factory). The residual turbidity percentage (%) is expressed as Residual turbidity percentage ð%Þ ¼ ðTtreated =Traw Þ 100%
2.4.2.
109
(1)
where Traw and Ttreated are the turbidities of raw and treated water, respectively.
IfQ DF
(5)
3.
Results and discussion
3.1.
Characterization of CMC-g-PAM
The FTIR and 1H NMR spectra of chitosan, CMC, and CMC-gPAM are shown in Figs. S2 and S3 of the supporting information, respectively. Aside from the characteristic FTIR peaks of chitosan in Fig. S2(i), the new peak at 1584 cm1 for COO (Chen and Park, 2003) in Fig. S2(ii) indicates that CMC is successfully prepared. In Fig. S2(iii), the new strong band around 1651 cm1 is assigned to amide-I (CeO stretching) on the PAM chain (Yuan et al., 2010). The bands of COO on the CMC backbone and amide-II (NeH bending) on the PAM chain overlap with each other and result in a peak at 1601 cm1 (Biswal and Singh, 2004; Yuan et al., 2010). The new peak at approximately 1447 cm1 is due to CeN stretching in the graft copolymer (da Silva et al., 2007; Yuan et al., 2010). Since the
110
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4
homopolymer of PAM is removed during the preparation process as mentioned above, the detected PAM chain should be chemically bonded to the chain backbone of CMC. The 1H NMR spectra (Fig. S3) also confirm the final structure of the amphoteric flocculant product, where PAM is successfully grafted onto the CMC. Aside from the signal of protons from chitosan in Fig. S3(i), the intense resonance at 3.93 ppm in Fig. S3(ii) is attributed to the protons on the carboxymethyl groups (Chen and Park, 2003). The new strong peaks at around 2.10 and 1.50 ppm in Fig. S3(iii) are the resonances of methylene and methine protons in the PAM chain (Yuan et al., 2010), respectively. According to the 1H NMR spectrum (Fig. S3(iii)), the grafting ratio (G) of CMC-g-PAM is also calculated to be around 212%, where G is defined as (Yuan et al., 2010) Gð%Þ ¼ ðWPAM =WCMC Þ 100%
flocculation performance of CMC-g-PAM is systematically studied at different dosages and various pH levels. First, the turbidity measurement, a commonly used method, is adopted to evaluate the optimal dosages of the flocculant (Fig. 2). From the turbidityedosage profiles, at pH 4, 7, and 11, the optimal dosages are 2, 5, and 16 mg/L, respectively. With increasing pH, the best residual turbidity percentage corresponding to the optimal dosage also increases, and the flocculation window becomes narrower.
(6)
where WPAM and WCMC are the weights of the PAM branches and the CMC backbone, respectively. As previously mentioned, CMC-based flocculants have enhanced solubility, which is beneficial for a wider application range. Therefore, the solubility of CMC-g-PAM in aqueous solution with various pH levels is tested according to the method of Chen and Park (2003) and shown in Table S1 of the supporting information. As expected, CMC-g-PAM reveals a remarkable improvement in solubility. It is soluble over the wide pH range from 1 to 13. The pH dependence of the ZP of kaolin and CMC-g-PAM were also measured (Fig. 1). The results demonstrate that kaolin particles are negatively charged over the whole pH range, whereas CMC-g-PAM has an isoelectric point at around pH 5 and shows a positive ZP below the isoelectric point. In addition, the ZP of CMC-g-PAM is close to zero at pH levels ranging from 5 to 7 because of nonionic PAM branch chains (Song et al., 2009).
3.2. Evaluation of the optimal dosages of the flocculant at various pH levels A variety of external parameters can affect the flocculation efficiency of flocculants, of which dosage and pH are the two most important ones (Rojas-Reyna et al., 2010). Hence, the
Zeta potential (mV)
0
-10
-20
Kaolin CMC-g-PAM
-30
2
4
6
8
10
12
pH Fig. 1 e ZPepH profiles of the kaolin suspension (---) and ). CMC-g-PAM solution (
Fig. 2 e Residual turbidity percentage (-) and ZP ( ) of the supernate as a function of CMC-g-PAM dosage at various pH: 4 (a), 7 (b), and 11 (c). Insert: The chain structures of CMC-g-PAM in solution at various pH.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4
Poorer flocculation efficiency with increasing pH has also been reported for a non-grafted chitosan-based amphoteric flocculant (Yang et al., 2011b). Since the conformation of the flocculant molecule in solution has great effects on its flocculation performance, the pH dependence of Rh is measured and shown in Fig. S4 of the supporting information. At lower pH values, the positive charges of the flocculant lead to high Rh because of the effects of intra-chain electrostatic repulsion (Radeva, 2001). Both the extensive conformation and positive charges of the flocculant are beneficial for flocculation efficiency and result in lower optimal dosage requirements. At pH 7, the decrease in Rh indicates that the polymer chains collapse owing to the loss of the net charge based on Fig. 1, which reduces the number of active flocculation sites and leads to a higher optimal dosage. At a higher pH level, the negative charges on the CMC-g-PAM increase, therefore, the polymer chain with a high Rh again becomes extensive. However, negative charges also result in greater electrostatic repulsion between the flocculants and the suspended particles, finally leading to poorer flocculation performance. Compared with non-grafted amphoteric flocculants (Yang et al., 2011b), CMC-g-PAM exhibits a more extended flocculation window and less sensitivity to dosage because of the better approachability of the grafted PAM chain to the suspended particles. Although CMC-g-PAM shows the poorest performance at pH 11, the best residual turbidity percentage can also reach a value of less than 5%. The flocculation performance of several conventional flocculants, such as alum, PAC, and PAM, are investigated at various pH levels and shown in Fig. S5 of the supporting information for comparison. Based on Fig. S5, the optimal dosages and the corresponding residual turbidity percentages of different flocculants are summarized in Table 1. Based on Table 1 and Fig. S5, the three conventional flocculants also show effective turbidity removal effects similar to CMC-gPAM. However, after further comparison, CMC-g-PAM consistently reveals the best flocculation performance amongst all four flocculants, and shows both the lowest optimal dosage and highest turbidity removal efficiency at pH 4 and 7. While PAM has the lowest optimal dosage at pH 11, its residual turbidity is much higher than that of the new flocculant. These flocculation behaviors demonstrate that CMC-gPAM is applicable for primary water purification, whether under acidic, neutral, or alkaline conditions.
Table 1 e The optimal dosages and corresponding residual turbidity percentages of various flocculants. Flocculants pH ¼ 4
CMC-g-PAM Alum PAC PAM a
Dosageoptimal (mg/L) RTlowest b (%) pH ¼ 7 Dosageoptimal (mg/L) RTlowest (%) pH ¼ 11 Dosageoptimal (mg/L) RTlowest (%)
2 0.86 5 2.74 16 4.65
8 4.70 12 4.56 25 4.63
4 3.19 8 3.73 20 4.71
2 2.28 6 5.08 12 9.54
a Dosageoptimal is the optimal dosage where the lowest residual turbidity percentage occurs. b RTlowest is the residual turbidity percentage at the optimal dosage.
111
The ZP of the supernates after flocculation is also provided along with the turbidity in Fig. 2. Under acidic conditions (Fig. 2(a)), ZP increases with increasing dosage owing to the opposite charge between the flocculant and the kaolin particles (Fig. 1). The net charge is nearly equal to zero at the optimal dosage. However, the flocculation mechanism cannot be explained simply by charge neutralization. This argument is based on the following: (i) No significant restabilization of suspended particles occurs at excess dosages (Yang et al., 2011b); (ii) ZP does not reach an extraordinarily high positive value, but shows a plateau of almost zero when excess flocculant is added; (iii) floc structures and flocculation kinetics are distinctively different from those that obey charge neutralization mechanisms (Rasteiro et al., 2008; Yu et al., 2006), which will be studied in the following section. Therefore, multimechanisms, including electrostatic patching and/or bridging, are probably involved in the flocculation process of CMC-g-PAM under acidic conditions. Under neutral or alkaline conditions (Fig. 2(b) and (c)), ZP also increases with increasing dosage. Although there is no opposite charge on the flocculant relative to kaolin based on Fig. 1, the increased ZP is due to the adsorption of a layer of polymer chains on the particle surface; these cause the shear plane to shift farther from the particle surface and the increase in the magnitude of ZP. Moreover, regardless of the pH condition, ZP cannot be used to evaluate the optimal dosage of CMC-g-PAM (Fig. 2). This result again demonstrates the limitation of the ZP method (Rasteiro et al., 2008; Yu et al., 2006). This method can only be used to predict the optimal dosage of flocculants when simple charge neutralization is the predominant flocculation mechanism. In addition to the traditional methods, as an alternative, Blanco et al. (1996) developed another method that optimizes the flocculation process based on monitoring of the mean floc size. Thus, in the present work, the dependence of floc size on the flocculant dosage is measured, and other crucial floc structure properties, such as fractal dimensions, are analyzed. The results are all illustrated in Fig. 3. In Fig. 3(a), larger flocs are obtained at the optimal dosage under each pH, which confirms the feasibility of this method for predicting the optimal dosage. The floc sizes grow until the dosages reach the optimal one, after which aggregation no longer takes place. The floc size even decreases at excess dosages because of steric stabilization and electrostatic repulsion (Rasteiro et al., 2008; Yu et al., 2006). Extraordinarily large floc sizes are obtained under acidic conditions. This result further denies the simple charge neutralization mechanism, given that the floc produced by charge neutralization is always smaller than that by bridging (Zhou and Franks, 2006). The microscopic photos shown in Fig. S6 of the supporting information also agree well with the results. In Fig. 3(b), DF values are calculated by linear fitting the logelog plot of I dependence on Q. Examples of the linear fitting are shown in Fig. S7 of the supporting information. For each pH investigated, the largest DF (i.e., the densest floc), as well as floc size, always occurs at the optimal dosages. However, when the flocculant is in excess, steric stabilization and electrostatic repulsion results in looser floc structures.
112
a
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4
300
pH=4 pH=7 pH=11
Floc size (µm)
250 200 150 100
particles. In Fig. 3(c), the D2 dependence on the flocculant dosage is in accordance with the results of DF, where D2 is calculated from Figs. S6 and S8 of the supporting information. However, considering that both DF and D2 are calculated through linear fitting, the data correlated for DF (correlation coefficient R2 > 0.99) are higher than those for D2 (R2 in the range of 0.80e0.97), which is attributed to the higher accuracy of the LS method than that of IA.
3.3. Monitoring of the flocculation process: flocculation kinetics
50 0
0
5
10
15
20
Dosage (mg/L)
b
2.4
pH4 pH7 pH11
2.2 2.0
DF
1.8 1.6 1.4 1.2 1.0 0
5
10
15
20
Dosage (mg/L)
c
2.0 1.8
D2
1.6
1.4
pH=4 pH=7 pH=11
1.2
1.0
0
5
10
15
20
Dosage (mg/L) Fig. 3 e The floc size (a), DF (b), and D2 (c) as functions of CMC-g-PAM dosage at various pH: 4 ( ), 7 (-), and 11 ( ).
Two points should be noted here: (i) DF at pH 4 is always larger than that in the two other pH levels, implying that patching mechanism is dominant under acidic conditions according to the literature (Rasteiro et al., 2008), and (ii) the extremely loose floc structure obtained under alkaline conditions is mainly due to the repulsive forces between negative charges of the adsorbed polymer and the kaolin
In the aforementioned discussion, results obtained from LS are consistent with the traditional turbidity measurement. From the variations in size, as well as D2 and DF as functions of flocculant dosage (Fig. 3), different flocculation mechanisms at various pH levels are primarily derived. Previous studies (Yu et al., 2006; Zhou and Franks, 2006) have reported that flocculation kinetics is highly related to the flocculation mechanism. In addition, flocculation kinetics is very complicated, and various stages, such as particlepolymer mixing, attachment of polymer onto the surface particles, rearrangement of adsorbed polymer chain, collisions of destabilized particles, breakage of floc due to shearing, and so on, occur continuously or simultaneously in the process. As mentioned in Section 1, LS can provide realtime floc structural information. Hence, from the aspects of size and fractal dimensions of flocs, flocculation processes at various pH levels are monitored in situ in the present work. The floc size and DF dependence on time of CMC-g-PAM at each optimal dosage under pH levels of 4, 7, and 11 are shown in Fig. 4, where DF is calculated based on Fig. S9 of the supporting information. In Fig. 4, the following results are obtained: (i) The floc in the acidic solution has the largest final size, whereas it has the slowest initial growth rate, and (ii) the size and DF of flocs become nearly constant after a brief period of time under both neutral and alkaline conditions, whereas those under acidic conditions continue growing through multiple steps over the measured time range. To explain these phenomena, feasible stages in the flocculation processes at pH levels of 4, 7, and 11 are proposed as follows. At pH 4, given that the amino groups on CMC backbone are protonated, the polymer chains are cationic and show extensive morphology with a large Rh of 171 nm in solution because of intra-chain electrostatic repulsion (Radeva, 2001) as shown in Fig. S4. There is a strong attraction between the flocculant and kaolin particles for the opposite charges. Therefore, at the initial stage of flocculation, a positively charged flocculant with large Rh tends to be adsorbed onto the particle in a flat configuration (Yu et al., 2006; Zhou and Franks, 2006). In this type of adsorption, the size of the kaolin particles coated by the polymer is almost unchanged (Rasteiro et al., 2008; Yu et al., 2006), leading to a slow floc growth rate. However, the coverage on the surface of the suspended particles by the polymeric flocculant is not full and uniform. Negative charges on bare parts of the surface of one particle easily attach to the excess net residual positive charges of the polymer coating of another particle (Mihai and Dragan, 2009), summarized as the patching mechanism. Through patching, different primary flocs aggregate to form
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4
a
300
Floc size (µm)
Under the alkaline condition, CMC-g-PAM again achieves a large Rh of 221 nm in solution. Owing to the better approachability of the PAM branches, bridging is the main flocculation mechanism. Therefore, the initial floc growth rate is also rapid. However, for aggregates, electrostatic repulsion between polymeric chains and suspended particles prevents particles from further entering into the interior of the floc. Thus, no rearrangement of floc occurs, causing the floc to reach the earliest equilibrium and achieve the most open structure (Fig. 4).
pH=4 dosage=2 mg/L pH=7 dosage=5 mg/L pH=11 dosage=16 mg/L
250
113
200 150 100 50 0
4. 0
5
10
15
20
Time (min)
b
2.4 2.2 2.0
DF
1.8 1.6 1.4 pH=4 dosage=2 mg/L pH=7 dosage=5 mg/L pH=11 dosage=16 mg/L
1.2 0
5
10
15
Conclusion
20
Time (min) Fig. 4 e Time dependence of floc size (a) and DF (b) at the optimal dosage of CMC-g-PAM at various pH: 4 ( ), 7 (-) and 11 ( ).
larger flocs, resulting in a continuously increased floc size. Rearrangements of the polymeric flocculant and original particles in the floc (Rasteiro et al., 2008; Yu et al., 2006; Zhou and Franks, 2006) always occur, accompanied by newly formed larger flocs, thereby inducing increases in DF and size. Despite the dominant mechanism of patching, bridging due to the better approachability of PAM chains to suspended particles (Sen et al., 2009) may also have valid effects and contribute to the formation of large flocs. At pH 7, when the ZP of CMC-g-PAM is close to zero, the bridging flocculation mechanism plays the predominant role. At the beginning of the flocculation process, different segments (PAM branches or the main backbone) of the same polymer chain can bind more than one particle very rapidly (Rojas-Reyna et al., 2010), that is, link multiple particles together. Thus, the initial floc growth rate is high. However, the primary aggregates are unstable and can further rearrange. The flocculant polymer chains with an originally small Rh of 125 nm transform into more extended and flatter configurations that stick onto the surface of suspended particles (Rasteiro et al., 2008; Yu et al., 2006), resulting in more compact but slightly smaller-sized flocs.
A novel amphoteric chemically bonded composite flocculant, CMC-g-PAM, is designed and successfully prepared. The new flocculant demonstrates improved solubility. Flocculation tests at various pH levels prove the applicability of the designed flocculant over a wide range of pH for primary water treatment. The optimal dosages of the flocculant obtained from turbidity, size, and fractal dimension test are consistent; ZP is not applicable under a more complicated flocculation mechanism. By measuring the floc size and fractal dimension, different flocculation mechanisms are also found at various pH levels. Patching is predominant under acidic conditions and causes larger and denser flocs, whereas bridging plays a key role under neutral and alkaline conditions, producing smaller and looser flocs. Moreover, three diverse kinetic processes of flocculation at various pH levels are proposed, according to the online monitoring results of LS. CMC-g-PAM favors adsorption onto the surface of suspended particles under a platform at acidic conditions, resulting in a slow initial growth rate but continuous floc growth. Under neutral or alkaline conditions, the flocculant prefers bridging particles together, causing a rapid initial rate and requiring a shorter period of time to reach equilibrium. In addition, under alkaline conditions, no further rearrangement of polymeric flocculant and particles in flocs occurs because of electrostatic repulsion.
Funding Supported by the Natural Science Foundation of China (Grant No. 51073077, 50633030, 50938004 and 50825802) and the Fundamental Research Funds for the Central Universities (Grant No. 1105020504 and 1116020510).
Appendix. Supplementary material Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.10.024.
references
Ali, S.K.A., Singh, R.P., 2009. An investigation of the flocculation characteristics of polyacrylamide-grafted chitosan. Journal of Applied Polymer Science 114 (4), 2410e2414.
114
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 0 7 e1 1 4
Biswal, D.R., Singh, R.P., 2004. Characterisation of carboxymethyl cellulose and polyacrylamide graft copolymer. Carbohydrate Polymers 57 (4), 379e387. Blanco, A., Negro, C., Hooimeijer, A., Tijero, J., 1996. Polymer optimization in paper mills by means of a particle size analyser: an alternative to zeta potential measurements. Appita Journal 49 (2), 113e116. Bushell, G., Amal, R., 2000. Measurement of fractal aggregates of polydisperse particles using small-angle light scattering. Journal of Colloid and Interface Science 221 (2), 186e194. Ceresa, R.J., 1973. Block and Graft Copolymerization. Wiley, New York. Chakraborti, R.K., Atkinson, J.F., Van Benschoten, J.E., 2000. Characterization of alum floc by image analysis. Environmental Science & Technology 34 (18), 3969e3976. Chakraborti, R.K., Gardner, K.H., Atkinson, J.F., Van Benschoten, J. E., 2003. Changes in fractal dimension during aggregation. Water Research 37 (4), 873e883. Chen, X.G., Park, H.J., 2003. Chemical characteristics of Ocarboxymethyl chitosans related to the preparation conditions. Carbohydrate Polymers 53 (4), 355e359. da Silva, D.A., de Paula, R.C.M., Feitosa, J.P.A., 2007. Graft copolymerisation of acrylamide onto cashew gum. European Polymer Journal 43 (6), 2620e2629. Fabrizi, L., Jefferson, B., Parsons, S.A., Wetherill, A., Jarvis, P., 2010. The role of polymer in improving floc strength for filtration. Environmental Science & Technology 44 (16), 6443e6449. Guibal, E., Van Vooren, M., Dempsey, B.A., Roussy, J., 2006. A review of the use of chitosan for the removal of particulate and dissolved contaminants. Separation Science and Technology 41 (11), 2487e2514. Jarvis, P., Jefferson, B., Parsons, S.A., 2005. Breakage, regrowth, and fractal mature of natural organic matter flocs. Environmental Science & Technology 39 (7), 2307e2314. Liao, J.Y.H., Selomulya, C., Bushell, G., Bickert, G., Amal, R., 2006. On different approaches to estimate the mass fractal dimension of coal aggregates. Particle & Particle Systems Characterization 22 (5), 299e309. Lin, J.L., Huang, C.P., Chin, C.J.M., Pan, J.R., 2008. Coagulation dynamics of fractal flocs induced by enmeshment and electrostatic patch mechanisms. Water Research 42 (17), 4457e4466. Mandelbrot, B.B., 1983. The Fractal Geometry of Nature. Freeman, New York. Mihai, M., Dragan, E.S., 2009. Chitosan based nonstoichiometric polyelectrolyte complexes as specialized flocculants. Colloids and Surfaces A e Physicochemical and Engineering Aspects 346 (1e3), 39e46. Radeva, T., 2001. Physical Chemistry of Polyelectrolytes. Marcel Dekker Inc., New York. Rasteiro, M.G., Garcia, F.A.P., Ferreira, P., Blanco, A., Negro, C., Antunes, E., 2008. The use of LDS as a tool to evaluate flocculation mechanisms. Chemical Engineering and Processing 47 (8), 1329e1338. Renault, F., Sancey, B., Badot, P.M., Crini, G., 2009. Chitosan for coagulation/flocculation processes e an eco-friendly approach. European Polymer Journal 45 (5), 1337e1348. Rojas-Reyna, R., Schwarz, S., Heinrich, G., Petzold, G., Schutze, S., Bohrisch, J., 2010. Flocculation efficiency of modified water soluble chitosan versus commonly used commercial polyelectrolytes. Carbohydrate Polymers 81 (2), 317e322. Sand, A., Yadav, M., Behari, K., 2010. Preparation and characterization of modified sodium carboxymethyl cellulose via free radical graft copolymerization of vinyl sulfonic acid in aqueous media. Carbohydrate Polymers 81 (1), 97e103.
Schaefer, D.W., Martin, J.E., Wiltzius, P., Cannell, D.S., 1984. Fractal geometry of colloidal aggregates. Physical Review Letters 52 (26), 2371e2374. Sen, G., Pal, S., 2009. Polyacrylamide grafted carboxymethyl tamarind (CMT-g-PAM): development and application of a novel polymeric flocculant. Macromolecular Symposia 277, 100e111. Sen, G., Kumar, R., Ghosh, S., Pal, S., 2009. A novel polymeric flocculant based on polyacrylamide grafted carboxymethylstarch. Carbohydrate Polymers 77 (4), 822e831. Song, H., Wu, D., Zhang, R.Q., Qiao, L.Y., Zhang, S.H., Lin, S., Ye, J., 2009. Synthesis and application of amphoteric starch graft polymer. Carbohydrate Polymers 78 (2), 253e257. Sonmez, H.B., Senkal, B.F., Bicak, N., 2002. Poly(acrylamide) grafts on spherical bead polymers for extremely selective removal of mercuric ions from aqueous solutions. Journal of Polymer Science Part A e Polymer Chemistry 40 (17), 3068e3078. Thomas, D.N., Judd, S.J., Fawcett, N., 1999. Flocculation modelling: a review. Water Research 33 (7), 1579e1592. Tripathy, J., Mishra, D.K., Yadav, M., Sand, A., Behari, K., 2009. Modification of kappa-Carrageenan by graft copolymerization of methacrylic acid: synthesis and applications. Journal of Applied Polymer Science 114 (6), 3896e3905. von Smoluchowski, M., 1917. Experiments on a mathematical theory of kinetic coagulation of coloid solutions. Zeitschrift fu¨r Physikalische Chemie e Sto¨chiometrie und Verwandtschaftslehre 92 (2), 129e168. Wang, J.P., Chen, Y.Z., Zhang, S.J., Yu, H.Q., 2008. A chitosanbased flocculant prepared with gamma-irradiation-induced grafting. Bioresource Technology 99 (9), 3397e3402. Wang, J.P., Chen, Y.Z., Yuan, S.J., Sheng, G.P., Yu, H.Q., 2009. Synthesis and characterization of a novel cationic chitosanbased flocculant with a high water-solubility for pulp mill wastewater treatment. Water Research 43 (20), 5267e5275. Wang, W., Qin, W., Bo, S.Q., 1991. Influence of the degree of deacetylation of chitosan on its mark-houwink equation parameters. Makromolekulare Chemie-Rapid Communications 12 (9), 559e561. Wei, J.C., Gao, B.Y., Yue, Q.Y., Wang, Y., Li, W.W., Zhu, X.B., 2009. Comparison of coagulation behavior and floc structure characteristic of different polyferric-cationic polymer dualcoagulants in humic acid solution. Water Research 43 (3), 724e732. Yang, H., Lu, Y.B., Cheng, R.S., 2011a. In: Davis, S.P. (Ed.), Chitosan: Manufacture, Properties, and Usage. Nova Science Publishers, New York. Yang, Z., Shang, Y.B., Lu, Y.B., Chen, Y.C., Huang, X., Chen, A.M., Jiang, Y.X., Gu, W., Qian, X.Z., Yang, H., Cheng, R.S., 2011b. The flocculation properties of biodegradable amphoteric chitosanbased flocculants. Chemical Engineering Journal 172 (1), 287e295. Yu, J.F., Wang, D.S., Ge, X.P., Yan, M.Q., Yang, M., 2006. Flocculation of kaolin particles by two typical polyelectrolytes: a comparative study on the kinetics and floc structures. Colloids and Surfaces A e Physicochemical and Engineering Aspects 290 (1e3), 288e294. Yu, W.Z., Gregory, J., Campos, L., 2010. Breakage and regrowth of Al-humic flocs e effect of additional coagulant dosage. Environmental Science & Technology 44 (16), 6371e6376. Yuan, B., Shang, Y.B., Lu, Y.B., Qin, Z.Q., Jiang, Y.X., Chen, A.M., Qian, X.Z., Wang, G.Z., Yang, H., Cheng, R.S., 2010. The flocculating properties of chitosan-graft-polyacrylamide flocculants (I) e effect of the grafting ratio. Journal of Applied Polymer Science 117 (4), 1876e1882. Zhou, Y., Franks, G.V., 2006. Flocculation mechanism induced by cationic polymers investigated by light scattering. Langmuir 22 (16), 6775e6786.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
A model for predicting resuspension of Escherichia coli from streambed sediments Pramod K. Pandey a, Michelle L. Soupir a, Chris R. Rehmann b,* a b
Dept. of Agricultural and Biosystems Engineering, Iowa State University, Ames, IA 50011, USA Dept. of Civil, Construction, and Environmental Engineering, Iowa State University, Ames, IA 50011, USA
article info
abstract
Article history:
To improve the modeling of water quality in watersheds, a model is developed to predict
Received 8 February 2011
resuspension of Escherichia coli from sediment beds in streams. The resuspension rate is
Received in revised form
expressed as the product of the concentration of E. coli attached to sediment particles and
14 October 2011
an erosion rate adapted from work on sediment transport. The model uses parameter
Accepted 16 October 2011
values mostly taken from previous work, and it accounts for properties of the flow through
Available online 23 October 2011
the bottom shear stress and properties of the sediment through the critical shear stresses for cohesive and non-cohesive sediment. Predictions were compared to resuspension rates
Keywords:
inferred from a steady mass balance applied to measurements at sixteen locations in
Resuspension
a watershed. The model’s predictions matched the inferred rates well, especially when the
E. coli
diameter of particles to which E. coli attach was allowed to depend on the bottom shear
Sediment
stress. The model’s sensitivity to the parameters depends on the contributions of particle
Microbial transport
packing and binding effects of clay to the critical shear stress. For the current data set, the
Watershed modeling
uncertainty in the predictions is controlled by the concentration of E. coli attached to sediment particles and the slope used to estimate the bottom shear stress. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Pathogens impair 480,000 km of rivers and shorelines and 2 million ha of lakes in the U.S., and the cost to implement total maximum daily load (TMDL) plans is estimated as $0.9 to $4.3 billion per year (USEPA, 2010). To predict the risk of bacteria to public health and allocate load reductions fairly, models that include accurate representations of the key processes of fate and transport are required (Fries et al., 2008). For example, the high concentrations of bacteria in suspended sediment and bed sediment suggest that the understanding of interactions between pathogens and sediment must be improved (Droppo et al., 2009). Sediment disturbance can account for the majority of total bacterial contamination (Nagels et al., 2002), and a one-dimensional
model applied to the field measurements of Jamieson et al. (2005) showed that including interactions with the sediment improved the predictions of Escherichia coli concentrations in the stream (Rehmann and Soupir, 2009). However, models that U.S. regulatory agencies use to determine pollutant load reductions usually do not include resuspension of bacteria as a source. Even when resuspension is included in models, how to predict it is uncertain. Many researchers have either specified the resuspension rate (e.g., Petersen et al., 2009) or expressed it mainly as a function of flow (Wilkinson et al., 1995; Tian et al., 2002; Collins and Rutherford, 2004). Kim et al. (2010) added a model of resuspension of E. coli to the Soil and Water Assessment Tool (SWAT); resuspension was estimated using a simplified version of Bagnold’s stream power
* Corresponding author. Tel.: þ1 515 294 1203; fax: þ1 515 294 8216. E-mail address:
[email protected] (C.R. Rehmann). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.019
116
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6
Nomenclature a a1 b b1 C1 C2 c3 c5 Ca d E E0 E0a fa g H2 kn2 N n na ns
coefficient for the effects of particle packing on the critical shear stress sc [L2] coefficient in the alternative model of the resuspension rate (Eq. (14)) [L13b1 Tb1 1 ] coefficient for the effects of particle packing on the critical shear stress sc [M1 L3] exponent in the alternative model of the resuspension rate (Eq. 14) [-] concentration of E. coli in the water column [#L3] concentration of E. coli in the sediment [#L3] prg(s 1)/6, coefficient for the effect of clay on the critical stress sc [M L2 T2] coefficient for the effect of clay on the critical stress sc [M L1 T2] concentration of E. coli attached to sediment in the bed [#L3] diameter of sediment particles to which E. coli attach [L] erosion rate for sediment [L T1] erosion rate at the threshold of erosion [L T1] coefficient in the predicted resuspension rate [L T1] fraction of E. coli in the water column that are attached to sediment [-] acceleration of gravity [L T2] depth of the sediment containing E. coli [L] net growth rate in the sediment [T1] number of parameters [-] Manning roughness coefficient [-] exponent in the predicted resuspension rate [-] exponent in the erosion rate for sediment [-]
equation, which has been criticized for not including the effect of grain size on sediment transport (Ferguson, 2005). Hipsey et al. (2008) accounted for properties of the sediment by including a critical shear stress computed from the Shields criterion, but although the Shields criterion holds for noncohesive sediment, its validity for cohesive sediment is questionable (Mehta and Lee, 1993). Sanders et al. (2005) assumed resuspension to be proportional to the shear stress, while Bai and Lung (2005) expressed resuspension as a nonlinear function of the difference between the shear stress and a critical shear stress. As Rehmann and Soupir (2009) noted, resuspension of microorganisms from a sediment bed depends in general on properties of the flow and sediment (e.g., Lick, 2009, ch. 3), the type of microorganism (Hipsey et al., 2008), and the presence of biofilms (e.g., Droppo et al., 2001). To predict E. coli resuspension reliably, theory for transport of cohesive sediment must be considered because most bacteria attach to cohesive particles (Black et al., 2002). For non-cohesive sediments such as sands, which typically have particle diameters greater than 62 mm, the main forces to consider are the dislodging tendency of the fluid shear stress and the submerged weight of a particle. For cohesive sediment, however, inter-grain forces complicate the predictions. Clay and very fine silt ( scn, E ¼ E0
n sb scn s ; sc scn
(1)
where E0 ¼ 106 m/s is the erosion rate at the threshold of erosion (Lick, 2009). Lick (2009) found the exponent ns to be approximately equal to 2 for small and intermediate particles, while others expressed the erosion rate as linearly proportional (i.e., ns ¼ 1) to the difference between the bottom stress and a critical stress (Amos et al., 1996). A compilation of data shows that the critical shear stress for non-cohesive sediment depends on the particle diameter d: scn ¼ 414d;
(2) 2
where scn is in N/m and d is in m (Lick, 2009). For cohesive sediment, the packing of the particles, which is quantified by the bulk density rb, and extra binding forces caused by clay must be considered. Combining the work of Roberts et al. (1998) and Lick et al. (2004), Lick (2009) proposed aebrb c5 ; sc ¼ scn 1 þ 2 þ d c3 d
(3)
where a and b are coefficients that Lick (2009) specified as 8.5 1016 m2 and 9.07 cm3/g, respectively. The coefficient c3 is given by p c3 ¼ rgðs 1Þ; 6
(4)
where r is the density of water, s is the specific gravity of the sediment particle, and g is the acceleration of gravity. The coefficient c5 depends on the clay fraction; for 2% bentonite added to quartz particles, c5 ¼ 7 N/m2 (Lick et al., 2004). Lick (2009) proposed Eq. (1) as a uniformly valid formulation for erosion. For large particles and no clay fraction, the bulk density does not affect the critical shear stress (Fig. 1), and Eq. (1) follows a form that applies to fine-grained, coarsegrained, cohesive sediments, and non-cohesive sediment (Lick, 2009). As the particle size decreases, effects of cohesion dominate (i.e., sc [ scn), and Eq. (1) reduces to a form similar to that of Roberts et al. (1998) for cohesive sediment. When the binding effects of clay provide the main resistance to particle motion, the critical shear stress depends only weakly on the
Fig. 1 e Dependence of critical shear stress on particle diameter. The dotted line is the critical stress for noncohesive sediment. The dashed line is the critical stress for cohesive sediment with a bulk density rb of 1.26 g/cm3 and no effects of clay (c5 [ 0 N/m2), and the solid line is the critical stress for cohesive sediment with rb [ 1.26 g/cm3 and c5 [ 21 N/m2.
particle diameter (Fig. 1) because both the mobilizing force and the resisting force depend on the surface area of the particle. The erosion rate in Eq. (1) can be adapted to predict the rate of resuspension of E. coli. Bacteria attach to and bioflocculate around solid particles (Black et al., 2002) and deposit to the bottom sediments; attached fractions for streams ranges from 55% during storms (Characklis et al., 2005; Krometis et al., 2007) to between 80 and 100% (Auer and Niehaus, 1993; Hipsey et al., 2008). When sediment is resuspended, an influx of E. coli from the streambed results (Whitman et al., 2006). Therefore, we predict the E. coli resuspension rate Ra by multiplying the erosion rate by the concentration Ca of E. coli attached to sediment in the bed: Ra ¼ Ca E0a
n sb scn a : sc scn
(5)
The coefficient and exponent are changed to E0a and na to allow for possible differences from E0 and ns in Eq. (1). The resuspension rate is nonzero only when the bottom shear stress exceeds the critical shear stress for non-cohesive sediment. We expect Eq. (5) to be useful in predicting resuspension rates because it accounts for effects of the flow and sediment, as well as the concentration of E. coli in the streambed.
3.
Methods
We applied the model to the Squaw Creek watershed to predict the resuspension of E. coli attached to stream bottom sediments into the water column. The model was evaluated using data collected from sixteen sites in the watershed. At
118
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6
each site, flow geometry was measured, and the concentrations of E. coli in streambed sediment and the overlying water column were determined. The resuspension rates predicted with Eq. (5) were compared to values inferred from the onedimensional model of Rehmann and Soupir (2009). A sensitivity analysis was conducted to assess the influence of certain parameters on model output, and the parameters controlling the uncertainty in the resuspension rate were identified. Table 1 lists the parameters used in computing the predicted and inferred resuspension rates, and it indicates whether the parameters were measured, estimated, taken from previous work, or calibrated.
3.1.
Study area
Squaw Creek passes through four counties of central Iowa, U.S.A. before discharging into the South Skunk River near Ames (Fig. 2). The total area of the Squaw Creek watershed, as defined by the 10-digit hydrologic unit code, is 59,327 ha, and the average basin slope is 2%. Soils consist of loamy Wisconsin glacial till and clayey lacustrine deposits, including loam, silty clay, clay loam, and silty clay loam (Iowa NRCS, 2011); about 87% of the soil is fine, and another 8% is sandy. The study area has a humid climate with an average yearly rainfall of 865.4 mm and average annual high and low temperature of 15.6 and 3.3 C, respectively. The stream network of Squaw Creek watershed was generated using 30 m digital elevation maps from the U.S. Geological Survey’s Earth Resources Observation and Science Center and the geographic information systems software ArcGIS 9 (ArcMap version 9.3.1) to identify the tributaries and main stream. Land cover was determined with a 2002 map for Iowa obtained from the Natural Resources Geographic Information System library, a repository developed by the Iowa Department of Natural Resources. About 74% of the watershed was under
agricultural management (corn 41%, soybean 33%, and row crops 0.4%), 17% of the watershed was under grassland (ungrazed grass 11%, grazed grass 2.5%, CRP grass 1.7%, and alfalfa 1.8%), and 2.7% was deciduous forest. Additionally, 5.4% of the watershed land cover was road, residential, and commercial and 0.3% was water and wetlands. The watershed has 20 listed confined feeding operation units, and hogs are the major livestock.
3.2.
Measurements
Data were collected from the sampling locations to predict E. coli resuspension rates. Water temperature and cross section geometry were measured at sixteen locations on 17 July 2009. The mean air temperature during the sampling was 18.4 C, and although the sky was mostly overcast, precipitation was zero. The mean discharge for the day was reported to be 3.6 m3/ s at the U.S. Geological Survey gaging station 05470500 on Squaw Creek in Ames, which is at the same cross section as our sampling location 16 (Fig. 2); the discharge varied by less than 2% during the sampling. The bulk density of the streambed sediments, expressed as weight per unit volume, was determined from wet and dry weight (dried in the oven at approximately 75 C for 2 days) of sediments (Roberts et al., 1998). The Manning roughness coefficient n was taken to be 0.036 using information for natural streams in Chow (1959, pp. 112e123). Water samples were collected from the center of the stream by lowering a Horizontal Polycarbonate Water Bottle Sampler (2.2 L, Forestry Suppliers Inc., Mississippi, U.S) from a bridge into the center of the stream at all the locations. Sediment samples were collected from the top 2e3 cm of the streambed using a Shallow Water Bottom Dredge Sampler (15 cm 15 cm opening, Forestry Suppliers Inc., Mississippi, U.S) at the same location as water samples. Three replicates of water and sediment samples were used in microbial analyses. Immediately after collection, samples were stored at 4 C and
Table 1 e Parameters used to compute the predicted and inferred resuspension rates. The second column indicates whether the parameter was used in the predicted rate (P), inferred rate (I), or both (B). The fourth column lists the uncertainty, expressed as a percentage of the parameter’s value, assumed in the analysis in Section 4. Parameter C1 ¼ conc. of E. coli in water (CFU/100 ml) C2 ¼ conc. of E. coli in sed. (CFU/100 ml) R ¼ hydraulic radius (m) A ¼ cross sectional area (m2) rb ¼ bulk density of the sediment (g/cm3) T ¼ temperature ( C) Q ¼ discharge (m3/s) E0a ¼ coefficient (m/s) a ¼ coefficient for bulk density effect (m2) b ¼ coefficient for bulk density effect (cm3/g) c3 ¼ coefficient for clay effect (N/m3) H2 ¼ depth of sediment containing E. coli (m) n ¼ Manning roughness coefficient S ¼ slope (m/m) fa ¼ attached fraction na ¼ exponent c5 ¼ coefficient for clay effect (N/m2) d ¼ particle diameter (mm)
Rates
Value
Uncty (%)
Source
I B P P P B P P P P P I P P B P P B
2.25 102 5.47 103 1.63 103 3.43 104 0.10e0.76 0.5e3.5 1.26 17.0e24.6 3.6 1 106 8.5 1016 9.07 8.46 103 0.02 0.036 2.5 104 1.0 1.0 21 1.0; 0.5e3.5
15 15 10 10 5 5 5 0 0 0 0 50 15 20 15 10 10 50
Measured Measured Measured Measured Measured Measured Measured at station 16 Lick (2009) Lick (2009) Lick (2009) Lick (2009), computed with s ¼ 2.65 Estimated from sediment sampler Estimated from Chow (1959) Estimated with Manning’s eq. at station 16 Estimated using range in Hipsey et al. (2008) Estimated from Amos et al. (1996) Calibrated/estimated from Lick (2009) Calibrated using ranges in Oliver et al. (2007) by fitting d ¼ asb with a ¼ 1.9 mm/Pa
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6
119
Fig. 2 e Squaw Creek watershed and sampling locations (1e16). Discharge was measured at the U.S. Geological Survey gaging station near station 16. The land cover is also shown.
analyzed within 24 h. The E. coli attached to particles were detached by stirring the mixture of sediment and purified water (ratio 1:1) for 15 min at approximately 200 rpm using a magnetic stir bar. The resulting solution was used to enumerate E. coli in the sediment. The values of C1 and C2, the E. coli concentrations in the water and the sediment, respectively, were determined by membrane filtration techniques (APHA, 1999) on modified mTEC agar (EPA, method 1603).
3.3. rates
Calculation of predicted and inferred resuspension
Resuspension rates were predicted with Eq. (5). All E. coli were assumed to be attached to sediment grains; that is, the attached fraction fa ¼ 1 and Ca ¼ C2. This choice is consistent with the assumptions and work reviewed in Hipsey et al. (2008), which showed attachment between 80 and 100%. The
bottom shear stress was computed from a force balance for steady, uniform flow: sb ¼ rgRS;
(6)
where r is the water density, g is the acceleration of gravity, and R is the hydraulic radius. The slope S was estimated from Manning’s equation to be 2.5 104. Values of the coefficients a and b from Lick (2009) were used, and the coefficient E0a was assumed to be equal to E0 given by Lick (2009). The coefficient c5 was calibrated, and the exponent na was taken to be 1, as suggested by Amos et al. (1996), who found the erosion rate to be linearly proportional to the difference between the shear stress and a critical shear stress. The critical shear stresses scn and sc require an estimate of the diameter d of the particles to which the E. coli attach. A constant value of the diameter and a diameter that is linearly proportional to the bottom shear stress were
120
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6
used. The merits of these approaches are discussed in Section 4.1. To evaluate the predictions, resuspension rates were inferred from the mass balance model of Rehmann and Soupir (2009). Considering settling, resuspension, and net growth, they determined that a steady-state mass balance for the sediment yields C2 fa ws ¼ ; C1 vr kn2 H2
Rai ¼ vr C2 ¼ fa ws C1 þ kn2 H2 C2 ¼ fa ws C1 ð1 þ FÞ;
(8)
where F ¼ kn2 H2 C2 =fa ws C1 indicates the relative importance of settling and net growth in the mass balance for E. coli in the sediment. For example, when F [ 1, settling is unimportant, and net growth balances resuspension. Once the exponent na was chosen, the parameters to determine or calibrate were the coefficient c5 and the diameter of the particles to which E. coli attach. The optimal values of the parameters for the predictions using Eq. (5) were chosen by minimizing s, the sum of the squares of the differences between the logarithms of the inferred and predicted resuspension rates: X
log10 Ra log10 Rai
DRa Ra
2 ¼
2 X 2 N N X 1 vRa Dyi Syi Dyi ¼ ; yi Ra vyi i¼1 i¼1
(12)
where N is the number of parameters and Dyi is the uncertainty in yi.
(7)
where ws is the settling velocity, vr is the resuspension velocity, kn2 is the net growth rate in the sediment, and H2 is the depth of sediment containing E. coli, which is estimated to be about 2 cm for our experiments. The settling velocity ws was estimated with Stokes’s law. The net growth rate is the difference between the growth rate and the natural mortality rate, which were computed as functions of water temperature using the relations in Hipsey et al. (2008). The resuspension velocity at each sampling location was computed from Eq. (7), and the inferred resuspension rate was computed as
s¼
individual parameters, which were assumed to be independent, with the formula of Taylor (1997, p. 75):
2
4.
Results and discussion
4.1. rates
Concentrations, critical stresses, and resuspension
E. coli concentrations were large. The concentration C1 of E. coli in the water column ranged from 225 to 5467 CFU/100 ml with a mean of 789 CFU/100 ml and standard deviation of 1255 CFU/ 100 ml (Fig. 3). All but one of the concentrations exceeded the U.S. water quality standards (USEPA, 2001), which state that the geometric mean of at least five samples during a 30-day period must not exceed 126 CFU/100 ml and that a singlesample must not exceed 235 CFU/100 ml. The concentration C2 of E.coli in the sediment ranged from 1633 to 34,257 CFU/100 ml with a mean of 13,597 CFU/100 ml and standard deviation of 10,463 CFU/100 ml. Concentrations in the sediment were 2e90 times (mean ¼ 30, s.d. ¼ 29) higher than concentrations in the water column. Previous studies reported the ratio C2/C1 to be 10e10,000 (Buckley et al., 1998; Davies and Bavor, 2000; Bai and Lung, 2005). Resuspension rates inferred using Eq. (8) ranged from 11 to 167 CFU/m2s, depending on whether the diameter d of the particles to which E. coli attach was set to a constant value (Fig. 4a) or allowed to vary with hydraulic conditions (Fig. 4b). The inferred rates are smaller than those of
(9)
Values of s were computed with the resuspension rates expressed in CFU/m2s. A quantitative measure of predictive skill (Willmott, 1981) was computed to assess the agreement between predicted and inferred E.coli resuspension rates: P jRa Rai j2 ; Skill ¼ 1 P ðjRa Rai j þ jRa Rai jÞ2
(10)
where the overbar denotes an average over all sampling locations. A skill of 1 indicates perfect agreement between predicted and inferred resuspension rates, while a skill of zero indicates poor performance. To understand the dependence of the resuspension rate on the parameters and help in applying the model, the sensitivity and uncertainty were computed. The relative sensitivity of the resuspension rate to each parameter yi was computed with Syi ¼
yi vRa Ra vyi
(11)
(Haan, 2002). The relative uncertainty in the resuspension rate was computed by propagating the uncertainties of the
Fig. 3 e Concentrations of E. coli in the water column and sediment. Squares denote measurements from the main channel, and circles denote measurements from the tributaries. The solid vertical line is set at the USEPA’s single-sample standard for E. coli, and the dotted lines are contours of C2/C1.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6
a
b
clay, the critical stress used for the predictions in Fig. 4 did not depend strongly on the particle diameter (Fig. 1). Estimates of critical stress in other cases vary widely because of the characteristics of the sediment, the presence of biofilms, the depositional history of the bed, and the approach used to define the critical stress. For field measurements in a stream in which 32% of the sediment was finer than 75 mm, Jamieson et al. (2005) computed critical shear stresses of 1.5e1.7 N/m2 using the Manning’s roughness coefficient and the discharge at which E. coli NAR first appeared in discrete samples during a storm. El Ganaoui et al. (2004) analyzed sediment samples from a field site and differentiated between the fluff layer, a surface layer of fine sized particles and organics with a mean particle diameter of 10e20 mm and critical shear stresses of 0.025e0.05 N/m2, and a layer with coarser particles, which had critical shear stresses that were 10e20 times larger. Droppo et al. (2001) conducted laboratory experiments on kaolinite clay with a mean diameter of 5 mm and contaminated sediment from a field site that had particle sizes less than 63 mm. The critical stress increased from 0.024 N/m2 to 0.325 N/m2 when a biofilm was allowed to grow, and critical stresses of 0.100e0.135 N/m2 for beds deposited under shear exceeded the stresses of 0.047e0.054 N/m2 for beds deposited under quiescent conditions. The critical stress estimated for Squaw Creek had a magnitude representative of natural sediment beds with biofilms and a realistic deposition history.
4.2.
Fig. 4 e Comparison between predicted and inferred resuspension rates. The solid line indicates perfect agreement, and the dashed lines indicate difference by a factor of 2. Squares denote measurements from the main channel, and circles denote measurements from the tributaries: (a) Constant value of the particle diameter: d [ 1.0 mm, s [ 1.04, and skill [ 0.82. (b) Particle diameter linearly related to bottom shear stress: d [ asb with a [ 1.9 mm/Pa, s [ 0.40, and skill [ 0.85.
Jamieson et al. (2005), who measured resuspension rates of 8200e15,000 CFU/m2s in a stream during storms. One cause of the discrepancy is that Jamieson et al. (2005) artificially seeded the bed with E. coli NAR; concentrations of E. coli in the sediment corresponding to the three storms highlighted by Jamieson et al. (2005) were between 1.2 105 and 5.5 105 CFU/100 ml, or about 3e300 times larger than in our experiments. In fact, the resuspension velocities vr ¼ Rai/ C2 in our experiments (4 107e1 106 m/s) were only 2e25 times smaller than the values (2 106e1 105 m/s) Jamieson et al. (2005) observed. The critical shear stress for cohesive sediment was about 1.1 N/m2 at all sampling stations. Because of the effects of
121
Predicting resuspension
Using values of parameters from previous work and a constant value of the particle size (Table 1), the model predicted thirteen of the resuspension rates within a factor of 2 and all within a factor of 5 (Fig. 4a). The model predicted the resuspension rates from the main channel and tributaries about equally well. As noted in Section 2, most of the parameters in Table 1 were either measured or taken from Lick (2009). The coefficient c5 was set to 21 N/m2; this value is three times that used by Lick et al. (2004) for quartz particles with 2% bentonite. Because grain size analyses showed that the sediment samples consisted of between 1 and 7% clay (i.e., particle sizes < 8 mm), a larger value of c5 is reasonable. The diameter d of particles to which E. coli attach must be specified to compute both the predicted and inferred resuspension rates. A single value of 1 mm used for all sites yielded s ¼ 1.04 and a skill of 0.82. Attachment of E. coli to small particles is consistent with previous findings. For example, Oliver et al. (2007) observed that 65% of E. coli attached to particles smaller than 2 mm. Because of the uncertainty in the diameter of particles to which E. coli attach, the diameter was allowed to depend on the bottom shear stress as d ¼ asb, where a is a coefficient. The optimal value of a of 1.9 mm/Pa yielded diameters between 0.5 and 3.5 mm, which fall within previously observed ranges (Oliver et al., 2007), and it changed the range of inferred resuspension rates because of the dependence of the settling velocity on particle diameter. The model predicted all of the within a factor of 2 (Fig. 4b), and it yielded s ¼ 0.40 and a skill of 0.85. Again, the model predicted the resuspension rates from the main channel and tributaries about equally well.
122
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6
Table 2 e Relative sensitivity of the predicted resuspension rate to the various parameters. As noted in the text, the parameters fb [aexpðbrb Þ=d2 and fc [ c5/c3d represent the contributions of bulk density and clay content, respectively, to the critical shear stress sc. Parameter
Relative sensitivity
Parameter
Relative sensitivity
Hydraulic radius R
na
sb sb scn
Exponent na
s scn na ln b sc scn
Slope S
na
sb sb scn
Bulk density rb
na
brb fb fb þ fc br f na b b fb þ fc
Concentration Ca
1
Coefficient b
Coefficient E0a
1
Coefficient a
na
Diameter d
na
Coefficient c5
na
fc fb þ fc
This version of the model involves no more calibration parameters, and the relationship between the diameter and shear stress appeals to physical intuition. Once the exponent na and coefficient c5 were specified using information from Amos et al. (1996) and Lick (2009), the only parameter to adjust in the first application (Fig. 4a) was the diameter d. The second application (Fig. 4b) also had only one parameter to adjust, the coefficient a. The assumed relationship d ¼ asb implies that as the bottom shear stress increases, larger particles can be resuspended. Also, the coefficient a can be related to the Shields parameter, which is used to determine conditions under which non-cohesive sediment will start moving: j¼
sb a ¼ : rgðs 1Þd rgðs 1Þ
(13)
With a ¼ 1.9 mm/Pa Eq. (13) yields a Shields parameter of about 33. This value is much larger than the critical Shields parameter for initiation of motion of 2 mm quartz particles (Cao et al., 2006). The larger value we obtained is reasonable because it deals with suspended sediment instead of initiation of motion and because the Shields criterion does not account for the cohesive effects involved in the transport of small particles.
4.3.
fb fb þ fc fb scn fb þ fc sb scn
sediment (sb [ scn) and effects of clay control the critical shear stress for cohesive sediment (fc [ fb). For similar reasons, the sensitivity to the bulk density, particle diameter, and coefficients a and b in Eq. (3) is smaller. When the effects of bulk density outweigh those of the clay content (fb [ fc)d for example, for a soil with greater bulk density or smaller clay content (i.e., reduced c5), most of the sensitivities change little, but the resuspension rate becomes most sensitive to the bulk density and the coefficient b because of the exponential dependence on both. Although the predicted resuspension rate is not sensitive to the particle diameter, the inferred resuspension rate can be. With a settling velocity computed with Stokes’s law, the inferred rate is always twice as sensitive to the diameter as it is to the concentration of E. coli in the water column and the attached fraction (Table 3). When settling is more important than net growth in the mass balance for E. coli in the sediment (F < 1) as at station 14, the inferred resuspension rate is most
Sensitivity and uncertainty
Calculating the sensitivity can help in determining the parameters for other situations. The relative sensitivity can be computed analytically (Table 2); the parameters fb ¼ aexpðbrb Þ=d2 and fc ¼ c5/c3d represent the contributions of bulk density and clay content, respectively, to the critical shear stress defined in Eq. (3). Because the resuspension rate is linearly proportional to both the coefficient E0a and the concentration Ca of attached E. coli in the sediment, the relative sensitivity to those parameters is always 1. All other sensitivities depend on the parameter values (Fig. 5). For the parameter set used in Fig. 4b, the resuspension rate is most sensitive to the slope, hydraulic radius (which for these measurements is approximately equal to the water depth), the concentration of attached E. coli, and the coefficients E0a and c5. The magnitude of these sensitivities is approximately equal to the exponent na, or 1, because the bottom shear stress is much greater than the critical shear stress for non-cohesive
Fig. 5 e Absolute value of the relative sensitivity of the predicted resuspension rate to the parameters listed in Table 2. Sensitivity is computed for station 13. Black bars are computed for the parameter set used to compute the rates in Fig. 4b. Gray bars use the base parameter set with c5 [ 2.5 N/m2, and white bars use the base parameter set with rb [ 1.45 g/cm3.
123
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6
Table 3 e Relative sensitivity of the inferred resuspension rate to the various parameters. As noted in the text, the parameter F[kn2 H2 C2 =fa ws C1 measures the relative importance of settling and net growth in the mass balance for E. coli in the sediment. The net growth rate kn2 and its derivative with respect to temperature are taken from Hipsey et al. (2008). Parameter Diameter d Conc. C1 in water column Attached fraction fa
Relative sensitivity
Parameter
2 1þF 1 1þF 1 1þF
sensitive to the diameter and less sensitive to the concentration of E. coli in the sediment, the depth of sediment containing E. coli, and the water temperature T (Fig. 6). For larger values of F, when net growth is more important that settling, the inferred rate is most sensitive to temperature except for temperatures corresponding to the peak in the net growth rate (i.e., vkn2/vT ¼ 0). The growth and decay relations in Hipsey et al. (2008) suggest that sensitivity to temperature will be small around temperatures of 22.6 C. The differences between these two cases is illustrated by the sensitivities for stations 6 and 11 (Fig. 6). Uncertainty is 30% in the predicted resuspension rate and 47e77% in the inferred resuspension rate. For the individual uncertainties in the parameters listed in Table 1, about 75% of the uncertainty in the predicted rate comes from the slope and the concentration Ca of E. coli attached to sediment. Another 20% comes from the coefficient c5 and the hydraulic radius, which is approximately equal to the water depth in most of our cases, and the remaining uncertainty comes from the exponent na. Efforts to reduce uncertainty in the predictions should involve better estimates of Ca and either measuring the slope more accurately or measuring the bottom shear stress with another method, such as one based on velocity measurements at a cross section (e.g., Kim et al., 2000).
Conc. C2 in sediment Depth H2 with E. coli Temperature T
Relative sensitivity F 1þF F 1þF F T vkn2 1 þ F kn2 vT
The main contributions to the uncertainty in the inferred resuspension rate depend on F. When net growth is more important than settling (large F), the depth of sediment containing E. coli controls the uncertainty, and when settling is more important than net growth (small F), the particle diameter controls the uncertainty. In the former case, the uncertainty can be reduced by measuring E. coli concentrations at different depths in the sediment; such measurements would also allow the assumption of uniform concentration to be assessed and revised. In the latter case, the uncertainty occurs because of the dependence on particle diameter through the settling velocity. Reducing the uncertainty in the settling velocitydand thus the resuspension ratedis difficult for several reasons. As Rehmann and Soupir (2009) reviewed in detail, flocculation, which can control the deposition of cohesive sediment (Droppo, 2001), can cause Stokes’s law to overestimate the settling velocity (Burban et al., 1990). Even without flocculation, settling velocities in a flowing stream fall below those from Stokes’s law far from the bed and exceed them near the bed (Cuthbertson and Ervine, 2007). Further uncertainty is introduced by the range of particle sizes present in stream sediment and tendency of E. coli to attach to particles of different sizes (Oliver et al., 2007).
4.4.
Model assessment
The key advantages of our model are that its parameters are related to observable physical quantities and that it accounts for the properties of the flow, sediment, and organisms. Alternative models for computing resuspension rates include those that assume a constant resuspension velocity vr (Chapra, 1997) and those that relate resuspension to the discharge Q using a formula of the form Ra ¼ a1 Ca Q b1 ;
Fig. 6 e Absolute value of the relative sensitivity of the inferred resuspension rate to the parameters listed in Table 3: Black bars are computed for station 14 (F [ 0.4, T [ 17.7 C), gray bars are computed for station 6 (F [ 43, T [ 19.0 C), and white bars are computed for station 11 (F [ 16, T [ 22.7 C).
(14)
where a1 and b1 are coefficients. Examples of models like (14) include those of Wu et al. (2009), who computed the concentration of resuspended organisms, and Collins and Rutherford (2004), who computed the number of E. coli resuspended per unit time. Predictions with a constant resuspension velocity of 5.2 107 m/s and Eq. (14) with a1 ¼ 8 107 and b1 ¼ 0.29 (with discharge expressed in m3/s) also provide good fits to the inferred resuspension rates in Fig. 4b (Table 4). However, choosing the parameters in these two models is difficult in situations without measured or inferred resuspension rates to be used for calibration. For example, to specify the resuspension rate in their model, Petersen et al. (2009) used the average of the resuspension rates reported
124
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6
Table 4 e Comparison of methods of predicting resuspension rates. The last two models were evaluated with the dataset computed with variable particle diameter. The last two columns show the number of predictions within factors of 2 and 5 of the measured values. Model
Eq. (5), constant d Eq. (5), constant j Constant vr Eq. (14)
s
1.04 0.40 0.45 0.27
Skill
0.82 0.85 0.98 0.95
Number within a factor of. 2
5
13 16 14 15
16 16 16 16
by Jamieson et al. (2005). As noted in Section 4.1, that rate is much higher than the inferred rates from our study. The ranges of resuspension velocity are closer, but even with the smallest value of vr from Jamieson et al. (2005)dwhich is four times larger than the optimal value, the predictions using constant resuspension velocity are worse than all four cases in Table 4. The coefficients a1 and b1 in (14) are even harder to specify: Collins and Rutherford (2004) did not report their values of the coefficients. Wu et al. (2009) related the concentration of resuspended organisms to Q4.5; this exponent is much larger than b1, and the strong dependence on flow was not reflected in our measurements of E. coli concentrations in the water column. In contrast, the parameters in Eq. (5) can be measured, observed, or estimated from previous work; the most challenging parameters to specify are the exponent na, which was taken from the work of Amos et al. (1996); the coefficient c5, which was estimated from the clay fraction and the results in Lick (2009); and the particle diameter, which was discussed in detail in Section 4.1. The ability of Eq. (5) to account for sediment properties gives it wider applicability than Eq. (14). For the data in Fig. 4b, the bottom shear stress is much larger than the critical shear stress for non-cohesive sediment, and the binding effects of clay make the critical shear stress for cohesive sediment depend only weakly on the particle diameter (Fig. 1). With sc approximately constant, the resuspension rate from (5) is proportional to Ca snba , and if the bottom shear stress can be expressed as a function of the discharge raised to some power, then Eq. (14) should work well. However, in streams with sediment that has a larger bulk density or a smaller clay fraction, the critical shear stress for cohesive sediment will not be constant, and predictions with Eq. (14) will not be as successful. The proposed formula (5) for predicting resuspension rates can in principle be applied in unsteady flow. In contrast, a model with specified resuspension velocity would be difficult to apply because the velocity would have to vary in time. The ability to use (5) to predict resuspension in unsteady flows is important because resuspension typically is largest during the rising limb of storm hydrographs (Jamieson et al., 2005). To apply Eq. (5), estimates of the shear stress would need to be obtained by modifying the force balance in Eq. (6) by considering effects of unsteadiness and nonuniformity or showing that they are negligible, as in Jamieson et al. (2005). Our study also demonstrates the challenge of estimating resuspension from field measurements. The inferred
resuspension rate from Eq. (8) was computed from a steadystate mass balance in Eq. (7). An analysis similar to that of Rehmann and Soupir (2009) shows that the flow in Squaw Creek was approximately steady: The time scale of unsteadinessdestimated as Q/(dQ/dt) using discharge measured at the U.S. Geological Survey’s gaging station at our station 16dwas about 11.5 h. This time scale is about 6 times larger than the time scale for settling (C2H2/C1faws), 20 times larger than the time scale for net growth (k1 n2 ), and 30 times larger than the time scale for resuspension (H2/vr). Therefore, the mass balance in Eq. (7) should hold approximately. Still, as discussed in Section 4.3, resuspension rates inferred with Eq. (8) are uncertain because they require estimates of the settling velocity and depth of sediment containing E. coli, and the various processes contributing to growth and decay of E. coli (Hipsey et al., 2008) are difficult to quantify in the field. Future work involves incorporating the resuspension rate in Eq. (5) in watershed-scale models such as SWAT. Such models provide discharge and channel geometry, from which shear stresses can be estimated. Spatial variations in quantities such as sediment properties can pose a challenge, especially in cases in which the resuspension rate is sensitive to the bulk density. However, because our model shows that variations in sediment properties become less important when the binding effects of clay control the critical shear stress sc, in those casesdas in the case of the Squaw Creek watersheddthe model should be easier to apply. Also, our use of the model for Squaw Creek, as well as future comparisons with the performance of other watershed-scale simulations including resuspension (Collins and Rutherford, 2004; Wu et al., 2009; Kim et al., 2010), will guide users in selecting the model’s parameters. The resulting model should help in creating plans to improve water quality in areas affected by E. coli contamination.
5.
Conclusions
We predicted resuspension of E. coli from sediment beds in streams by expressing the resuspension rate as the product of the concentration of E. coli attached to sediment particles and an erosion rate adapted from work on sediment transport. The model accounts for properties of the flow through the bottom shear stress and properties of the sediment through the critical shear stresses for cohesive and non-cohesive sediment. To evaluate the model’s predictive skill, its predictions were compared to resuspension rates inferred from a steady mass balance applied to measurements at sixteen locations in a watershed. Sensitivity and uncertainty were computed to determine the parameters that affect the predictions most strongly and to identify ways to improve the model. The main conclusions of this study are as follows: 1. The model performed well using parameter values mostly taken from previous work. The coefficient representing the binding effects of clay was increased from a previously reported value because of the higher clay content in the sediment in our study. The application of the model in which the particle diameter was linearly proportional to the bottom shear stress (i.e., constant Shields parameter)
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6
performed better than an application with constant particle diameter, while maintaining the same number of model coefficients. 2. Although two simpler models also performed well, the proposed model can be applied more easily in situations without measured or inferred resuspension rates because its parameters are related to observable physical quantities and it accounts for properties of the flow, sediment, and organisms. Furthermore, its ability to be applied in unsteady flow is important because resuspension is often largest during the rising limb of a storm hydrograph. 3. When the binding effects of clay control the critical shear stress, the predicted resuspension rate is more sensitive to properties of the flow, and when the bulk density controls the critical shear stress, the predicted resuspension rate is more sensitive to properties of the sediment. For the current data set, the uncertainty in the predictions would be reduced by reducing uncertainty in the concentration of attached E. coli and the slope used to compute the bottom shear stress.
Acknowledgments The authors thank the U.S. Environmental Protection Agency Region 7 (contract no. X7-97703701-1) for generous support of this work and Kendal Agee, Andrew Paxson, Charles Velasquez, and Ray Sims for assistance with sample collection and analysis. The first two authors acknowledge support from the National Science Foundation under grant 0967845; any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.
references
American Public Health Association (APHA), 1999. Standard Methods for the Examination of Water and Wastewater. AWWA, Water Environment Federation. Amos, C.L., Sutherland, T.F., Zevenhuizen, J., 1996. The stability of sublittoral, fine-grained sediments in a subarctic estuary. Sedimentology 43 (1), 1e19. Auer, M.T., Niehaus, S.L., 1993. Modeling fecal coliform bacteria. I. Field and laboratory determination of loss kinetics. Water Research 27 (4), 693e701. Bai, S., Lung, W.-S., 2005. Modeling sediment impact on the transport of fecal bacteria. Water Research 39 (20), 5232e5240. Black, K.S., Tolhurst, T.J., Paterson, D.M., Hagerthey, S.E., 2002. Working with natural cohesive sediments. Journal of Hydraulic Engineering 128 (1), 2e8. Buckley, R., Clough, E., Warnken, W., Wild, C., 1998. Coliform bacteria in streambed sediments in a subtropical rainforest conservation reserve. Water Research 32 (6), 1852e1856. Burban, P.Y., Xu, Y.J., McNeil, J., Lick, W., 1990. Settling speeds of flocs in freshwater and seawater. Journal of Geophysical Research 95 (C10), 18213e18220. Cao, Z.X., Pender, G., Meng, J., 2006. Explicit formulation of the Shields diagram for incipient motion of sediment. Journal of Hydraulic Engineering 132 (10), 1097e1099.
125
Chapra, S.C., 1997. Surface Water-Quality Modeling. McGraw-Hill, New York, NY, p. 510. ISBN 0-07-011364-5. Characklis, G.W., Dilts, M.J., Simmons III, O.D., Likirdopulos, C.A., Krometis, L.-A.H., Sobsey, M.D., 2005. Microbial partitioning to settleable particles in stormwater. Water Research 39 (9), 1773e1782. Chow, V.T., 1959. Open-channel Hydraulics. McGraw-Hill, New York, NY, p. 112. ISBN 07-010776-9. Collins, R., Rutherford, K., 2004. Modelling bacterial water quality in streams draining pastoral land. Water Research 38 (3), 700e712. Cuthbertson, A.J.S., Ervine, D.A., 2007. Experimental study of fine sand particle settling in turbulent open channel flows over rough porous beds. Journal of Hydraulic Engineering 133 (8), 905e916. Davies, C.M., Bavor, H.J., 2000. The fate of stormwater-associated bacteria in constructed wetland and water pollution control pond systems. Journal of Applied Microbiology 89 (2), 349e360. Droppo, I.G., 2001. Rethinking what constitutes suspended sediment. Hydrological Processes 15 (9), 1551e1564. Droppo, I.G., Lau, Y.L., Mitchell, C., 2001. The effect of depositional history on contaminated bed sediment stability. Science of the Total Environment 266 (1e3), 7e13. Droppo, I.G., Liss, S.N., Williams, D., Nelson, T., Jaskot, C., Trapp, B., 2009. Dynamic existence of waterborne pathogens within river sediment compartments: implications for water quality regulatory affairs. Environmental Science & Technology 43 (6), 1737e1743. El Ganaoui, O., Schaaff, E., Boyer, P., Amielh, M., Anselmet, F., Grenz, C., 2004. The deposition and erosion of cohesive sediments determined by a multi-class model. Estuarine Coastal and Shelf Science 60 (3), 457e475. Ferguson, R.I., 2005. Estimating critical stream power for bedload transport calculations in gravel-bed rivers. Geomorphology 70 (1e2), 33e41. Fries, J.S., Characklis, G.W., Noble, R.T., 2008. Sediment-water exchange of Vibrio sp. and fecal indicator bacteria: Implications for persistence and transport in the Neuse River Estuary, North Carolina, USA. Water Research 42 (4e5), 941e950. Haan, C.T., 2002. Statistical Methods in Hydrology. Iowa State University Press, USA, p. 391. ISBN 978e0813815107. Hipsey, M.R., Antenucci, J.P., Brookes, J.D., 2008. A generic, process-based model of microbial pollution in aquatic systems. Water Resources Research 44 (7), W07408. doi:10.1029/02007wr006395. Iowa Natural Resources Conservation Service, 2011. Information about Soils. http://www.ia.nrcs.usda.gov/soils.html Accessed 15.06.11. Jamieson, R.C., Joy, D.M., Lee, H., Kostaschuk, R., Gordon, R.J., 2005. Resuspension of sediment-associated Escherichia coli in a natural stream. Journal of Environmental Quality 34 (2), 581e589. Kim, J.W., Pachepsky, Y.A., Shelton, D.R., Coppock, C., 2010. Effect of streambed bacteria release on E. coli concentrations: monitoring and modeling with the modified SWAT. Ecological Modelling 221 (12), 1592e1604. Kim, S.C., Friedrichs, C.T., Maa, J.P.Y., Wright, L.D., 2000. Estimating bottom stress in tidal boundary layer from acoustic Doppler velocimeter data. Journal of Hydraulic Engineering 126 (6), 399e406. Krishnappan, B.G., 2007. Recent advances in basic and applied research in cohesive sediment transport in aquatic systems. Canadian Journal of Civil Engineering 34 (6), 731e743. Krometis, L.A.H., Characklis, G.W., Simmons, O.D., Dilts, M.J., Likirdopulos, C.A., Sobsey, M.D., 2007. Intra-storm variability in microbial partitioning and microbial loading rates. Water Research 41 (2), 506e516.
126
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 5 e1 2 6
Lick, W., 2009. Sediment and Contaminant Transport in Surface Waters. CRC Press, pp. 81e93. ISBN 978-1-4200-5987-8. Lick, W., Jin, L.J., Gailani, J., 2004. Initiation of movement of quartz particles. Journal of Hydraulic Engineering 130 (8), 755e761. Mehta, A., Lee, S.C., 1993. Problems in linking the threshold condition for the transport of cohesionless and cohesive sediment transport. Journal of Coastal Research 10 (1), 170e177. Mehta, A.J., 1989. On estuarine cohesive sediment suspension behavior. Journal of Geophysical Research-Oceans 94 (C10), 14303e14314. Nagels, J.W., Davies-Colley, R.J., Donnison, A.M., Muirhead, R.W., 2002. Faecal contamination over flood events in a pastoral agricultural stream in New Zealand. Water Science and Technology 45 (12), 45e52. Oliver, D.M., Clegg, C.D., Heathwaite, A.L., Haygarth, P.M., 2007. Preferential attachment of Escherichia coli to different particle size fractions of an agricultural grassland soil. Water Air and Soil Pollution 185 (1e4), 369e375. Partheniades, E., 1965. Erosion and deposition of cohesive soils. Journal of the Hydraulics Division-American Society of Civil Engineers 91, 105e139. Paterson, D.M., 1989. Short-term changes in the erodibility of intertidal cohesive sediments related to the migratory behavior of epipelic diatoms. Limnology and Oceanography 34 (1), 223e234. Petersen, C.M., Rifai, H.S., Stein, R., 2009. Bacteria load estimator spreadsheet tool for modeling spatial Escherichia coli loads to an urban bayou. Journal of Environmental Engineering 135 (4), 203e217. Rehmann, C.R., Soupir, M.L., 2009. Importance of interactions between the water column and the sediment for microbial concentrations in streams. Water Research 43 (18), 4579e4589. Roberts, J., Jepsen, R., Gotthard, D., Lick, W., 1998. Effects of particle size and bulk density on erosion of quartz particles. Journal of Hydraulic Engineering 124 (12), 1261e1267. Sanders, B.F., Arega, F., Sutula, M., 2005. Modeling the dryweather tidal cycling of fecal indicator bacteria in surface waters of an intertidal wetland. Water Research 39 (14), 3394e3408.
Stone, M., Krishnappan, B.G., Emelko, M.B., 2008. The effect of bed age and shear stress on the particle morphology of eroded cohesive river sediment in an annular flume. Water Research 42 (15), 4179e4187. Sutherland, T.F., Amos, C.L., Grant, J., 1998. The effect of buoyant biofilms on the erodibility of sublittoral sediments of a temperate microtidal estuary. Limnology and Oceanography 43 (2), 225e235. Taylor, J.R., 1997. An Introduction to Error Analysis: The Study of Uncertainties in Physical Measurements. University Science Books, Sausalito, CA, p. 75. ISBN 0-93570242-3. Tian, Y.Q., Gong, P., Radke, J.D., Scarborough, J., 2002. Spatial and temporal modeling of microbial contaminants on grazing farmlands. Journal of Environmental Quality 31 (3), 860e869. U.S. Environmental Protection Agency, 2001. Protocol for Developing Pathogen TMDLs. EPA 841-R-00e002, Office of Water (4503F) Washington, D.C, 6-4. U.S. Environmental Protection Agency, 2010. WATERS (Watershed Assessment, Tracking & Environmental ResultS) Washington, D.C. van Rijn, L.C., 2007. Unified view of sediment transport by currents and waves. I: initiation of motion, bed roughness, and bed-load transport. Journal of Hydraulic Engineering 133 (6), 649e667. Whitman, R.L., Nevers, M.B., Byappanahalli, M.N., 2006. Examination of the watershed-wide distribution of Escherichia coli along southern Lake Michigan: an integrated approach. Applied and Environmental Microbiology 72 (11), 7301e7310. Wilkinson, J., Jenkins, A., Wyer, M., Kay, D., 1995. Modelling faecal coliform dynamics in streams and rivers. Water Research 29 (3), 847e855. Willmott, C.J., 1981. On the validation of models. Physical Geography 2, 184e194. Wu, J.Y., Rees, P., Storrer, S., Alderisio, K., Dorner, S., 2009. Fate and transport modeling of potential pathogens: the contribution from sediments. Journal of the American Water Resources Association 45 (1), 35e44.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Novel ferromagnetic nanoparticle composited PACls and their coagulation characteristics M. Zhang, F. Xiao, X.Z. Xu, D.S. Wang* State Key Laboratory of Environmental Aquatic Chemistry, Research Center for Eco-Environmental Sciences, CAS, POB 2871, Beijing 100085, China
article info
abstract
Article history:
Effects of magnetic nanoparticles on inorganic coagulants and their coagulation perfor-
Received 14 February 2011
mances were studied in the present work. The Fe3O4eSiO2 core-shell particle (FSCSP) and
Received in revised form
superfine iron (SI), were compounded with polyaluminium chloride of basicity 2.0 (PACl2.0),
9 September 2011
providing magnetic PACl2.0s (MPACl2.0s). The physiochemical properties of ferromagnetic
Accepted 16 October 2011
nanoparticles were investigated using transmission electron microscopy (TEM), the BET
Available online 25 October 2011
method and a zeta potentiometric analyzer. The Al species distributions of the MPACl2.0s
Keywords:
coagulation performances. Floc properties were assessed by use of the electromotive
Magnetic composited PACl
microscope (EM) and small angle laser light scattering (SALLS). The results showed that
Magnetic coagulation
modified layers of nanoparticles mitigated agglomeration. FSCSP had a larger specific area
Ferromagnetic nanoparticles
and pore volume than SI. The addition of ferromagnetic nanoparticles obviously increased
and PACl2.0 were examined by liquid
27
Al NMR. Jar tests were employed to evaluate the
Coagulation characteristics
the content of Alun. MPACl2.0s performed better than PACl2.0 in turbidity removal and DOC
Floc property
removal when dosed less than 0.06 mmol/L as Al. Generally, PACl2.0 þ FSCSP (50 mg/L) performed best. Large, loose and weak flocs were produced by MPACl2.0s, which were preferred for the magnetic powder recycling. A plausible structure, Al species-nanoparticles cluster, contributing to the unique properties of MPACl2.0 flocs, was proposed. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Magnetic coagulation, the combination of magnetic matericals and coagulants, has been tested for its effectiveness. The presence of ferromagnetic materials and magnetic fields makes magnetic collection and separation possible, leading to higher efficiency, larger handling capacity, easier manipulation and comparatively lower energy consumption (Ambashta and Sillanpa¨a¨, 2010). Hence, investigations have been mainly focused on several aspects as follows: (I) Seeding of magnetic particles in the treating water to produce magnetic flocs that can be rapidaly collected in a magnetic separator (Li et al., 2010); (II) Integration of magnetic ion exchange (MIEX)
hybrid systems (Singer and Bilyk, 2002) to improve treatment efficiency (Zhang et al., 2007; Korbutowicz et al., 2008); (III) Development and improvement of the magnetic separation system/instrument (Svoboda and Fujita, 2003); (IV) Development of magnetic coagulants by compounding magnetic materials with traditional coagulants. Although magnetic seeding coagulation and MIEX resin adsorption can greatly improve removal efficiency, the expense of the magnetic powder and MIEX resin have restricted their practical application. Nanosorbents, nanocatalysts and so on have been evaluated (Savage and Diallo, 2005). In aqueous systems, iron oxide particles are hydrated, and FeeOH groups can completely cover their surface.
* Corresponding author. Tel./fax: þ86 10 62849138. E-mail addresses:
[email protected] (M. Zhang),
[email protected] (F. Xiao),
[email protected] (X.Z. Xu),
[email protected] (D.S. Wang). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.025
128
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5
Hydrous iron oxides have an amphoteric character. The FeeOH sites on the surface can react with Hþ or OH ions from dissolved acids or bases. A positive (FeeOH2þ) or negative (FeeO) charges can be formed on the surface by protolytic reactions depending on the pH of the electrolyte solution (Ille´s and Tomba´cz, 2006). Magnetic nanoparticles exhibit a finitesize effect and/or high surface-to-volume ratio, resulting in a higher adsorption capacity. Additionally, the magnetic flocculants can be recovered by an external magnetic field. Therefore, there is high interest in the application of magnetic nanomaterials to water treatment (Shen et al., 2009). Composite magnetic coagulants are considered to have advantages over traditional coagulants since their magnetic components may optimize coagulation behaviors and facilitate the magnetic separation following the coagulation process. Jiang et al. (2010) successfully synthesized magnetic polyferric chloride (MPFC) by combining Fe3O4 nanoparticles with PFC. The MPFC, a physical mixture, gave a synergistic improvement in Microcystis aeruginosa removal efficiency compared to only PFC. Coagulated flocs showed a higher settling velocity by wrapping up Fe3O4. Liu et al. (2006) developed a special flocculant to control freshwater cyanobacterial blooms. In their study, hydrochloride acid was firstly mixed with flyash in a proper ratio for modification, and the mixture and the magnetic powder were then put into the algae contaminated water; the aquatic hazardous substances (algal toxins) were thereby effectively adsorbed and removed. Polyaluminum chloride (PACl) has been widely used and claimed to be superior to other traditional coagulants since it had the advantages of less alkalinity consumption and wider temperature and pH tolerance (Odegaard et al., 1990; Van Benschoten and Edzwald, 1990). The hydrolysis products of PACl comprise a series of Al species, especially Al13 polycation, [AlO4Al12(OH)24(H2O)12]7þ, which has been widely accepted as the critical species in particle aggregation by strong charge neutralization (Johansson, 1960). In addition, PACl can be temporarily refractory to hydrolysis before adsorption onto particle surfaces (Hu et al., 2006). Thus, combining ferromagnetic nanoparticles with PACl should produce favorable results. However, almost no research has addressed this to date, especially the magnetic coagulation mechanism and the characterizations of the formed flocs. In the present work, ferromagnetic nanoparticles were mixed with PACl2.0 to develop a novel composite magnetic coagulant MAPCl2.0. The developed coagulant was characterized in terms of Al species distributions by a liquid 27Al NMR method. In addition, jar tests were conducted to evaluate the MPACl2.0’s performance by examining turbidity and DOC removal. The formation, breakage and regrowth of MPACl2.0 flocs were subsequently investigated. Finally, a model was proposed to elucidate the coagulation mechanisms of MPACl2.0s with special emphasis on the role of the ferromagnetic nanoparticles.
2.
Material and methods
2.1.
Chemicals
FeCl3$6H2O, FeCl2$4H2O, NaHCO3, tetraethyl orthosilicate (TEOS) and absolute ethanol were obtained from the
National Medicines Corporation Ltd. of China. NH3$H2O, NaOH and hydrochloric acid were provided by Beijing Chemical Plant. AlCl3$6H2O， NaNO3 and Kaolin were obtained from the Xilong Chemical Corporation Ltd. of Shantou, the Jinke Fine Chemical Institute of Tianjin, and the Dongxu Chemical Plant of Beijing, China, respectively. Humic acid was produced by SigmaeAldrich Corporation Ltd. of USA. All reagents were of analytical grade without further purification except for being specified, and deionized water was used in preparing all solutions.
2.2.
Preparation of the ferromagnetic nanoparticles
Fe3O4eSiO2 core-shell particles (FSCSP) were prepared in the laboratory. Firstly, Fe3O4 particles were synthesized by coprecipitation (Hong et al., 2009): FeCl3 and FeCl2 with a molar ratio of 1.7 were prepared in an N2 atmosphere. Excess ammonia aqueous solution was then quickly added into the Fe3þ/Fe2þ solution with vigorous stirring until no precipitates could be seen in the solution, and then another 1 h’s stirring was conducted. Finally, precipitates were collected and washed for several times with deionized water and ethanol until the pH value decreased to 7. A SiO2 layer was built up by directly introducing Fe3O4 particles into the primary silica particles by the Sto¨ber process (Lu et al., 2008): a certain amount of ethanol, deionized water, aqueous ammonia and TEOS were added in a three neck flask in a 40 C water bath. Fe3O4 particles were added into the flask at different time under mechanical stirring. The final product was washed, dried at 50 C and conserved in the dryer. The other ferromagnetic nanoparticle used in this study is the superfine iron (SI), which was provided by the Research Center for Nano Technology of the Chinese Iron and Steel Institute. Its physiochemical properties may be found in the Master’s thesis of Chunfeng Hou (2009).
2.3.
Synthesis of PACl2.0 and of MPACl2.0s
The PACl2.0 was prepared by slow base titration of 0.5 mol/L AlCl3 solution with 0.2 mol/L NaHCO3 solution under rapid stirring to achieve the target molar OH/Al ratio (basicity value) of 2.0 (Tang, 2006). MPACl2.0s were synthesized by blending PACl2.0 with ferromagnetic nanoparticles on a shaking table at room temperature (25 C) for 2 h with a speed of 250 rpm. The final composited MPACl2.0s were aged for 24 h, and conserved at 4 C. FSCSPs, two loading concentrations of 50 and 100 mg/L, were respectively added into PACl2.0 solutions. Hence the formed MPACl2.0s can be described as PACl2.0 þ FSCSP (50 mg/L) and PACl2.0 þ FSCSP (100 mg/L) respectively. Loading concentrations were determined by means of apparent dispersibility. It may be noted that FSCSP aggregated more easily than the SI. Therefore, in order to make the ferromagnetic nanoparticles fully distributed at the surface of PACl2.0, only 25 mg/L was chosen as the loading concentration of the SI (PACl2.0 þ SI (25 mg/L)).
129
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5
2.4. Characterization of the ferromagnetic nanoparticles, PACl2.0, and MPACl2.0s The particle size and morphology of FSCSP and SI were determined by transmission electron microscopy (TEM) (S-570, Hitachi, Japan). The surface charges of the two nanoparticles were analyzed with a zeta potential meter (Zetasizer, Malvern, UK) by adjusting the pH of their suspensions from 3 to 11 with 1.0 and 0.1 mol/L NaOH and HCl solutions. Moreover, zeta potentials of MPACl2.0s and PACl2.0 were also measured. The specific surface area and pore size distribution were measured by a Brunauer, Emmett, and Teller (BET) surface area analyzer (ASAP-2000, Micromeritics, US) for nitrogen adsorption. The BET method was carried out under relatively high vacuum and measured primarily the external area of the particles and aggregates. Al species of MPACl2.0s and PACl2.0 were characterized by 500 MHz 27Al NMR, and instrumental settings and experimental conditions were addressed elsewhere (Xu et al., 2003). The internal standard was 0.05 mol/L NaAlO2 with its chemical shift at 80 ppm downfield. The signals near 0 and 62.5 ppm represent mononuclear Al (Alm) and Al13, respectively. Alun (the amount of undetectable species) was obtained by: Alun ¼ AlT Alm Al13
2.5.
(1)
Preparation of synthetic water
A synthetic water was prepared by dissolving the HA and kaolin stock suspensions with 1 mmol/L NaNO3 and 0.8 mmol/L NaHCO3 so that the ionic strength and alkalinity of the solution could be kept at 1.0 mmol/L and 80 mg/L respectively. The pH was adjusted to 7.00 0.02 using 0.1 mol/L, 0.01 mol/L NaOH or/ and 0.1 mol/L, 0.01 mol/L HCl. The synthetic water had average measured zeta potential, turbidity, UV254 and DOC values of 21.4 mV, 75.6 NTU, 0.254 cm1 and 2.318 mg/L, respectively. All measurements were obtained at room temperature.
2.6.
Coagulation jar tests
range of 0.02e0.20 mmol/L as Al. Before dosing, the three kinds of MPACl2.0s were shaken well in order to disperse the composited systems as uniformly as possible. The jar-test procedure consisted of a 30 s premix (250 rpm), a 1 min rapid mix (200 rpm), a 15 min slow mix (30 rpm) and a 20 min settling period. A small amount of sample was taken immediately to measure the zeta potential (Malvern, Zetasizer 2000, UK) after a 30 s rapid mix. After settling, the supernatants were analyzed by a TOC analyzer (TOC-Vcph, Shimadzu, Japan) and UV254 (UVeVis 8500, Hitachi, Japan) after filtration through a 0.45 mm membrane. The turbidity was measured by a turbidimeter (Hach 2100P Turbidimeter, USA).
2.7.
Floc characterization
Flocs were immediately collected from beakers after the slow stirring phase of the jar test with great care to avoid unnecessary “breakage”. Image observation was done under a low resolution of 1 mm (10 objective) using an electromotive microscope (Axioskop 2 mot plus, Carl Zeiss Co., Germany). To investigate and compare the growth, breakage and regrowth of flocs produced by MPACl2.0s and PACl2.0, experiments similar to those of Zhu et al. (2009) were carried out. Briefly, coagulation tests were performed under the optimal dosage of 0.08 mmol/L as Al. After the slow stirring phase, flocs were suddenly exposed to a 1 min strong stirring of 400 rpm for breakage, and then regrowth was undertaken at 40 rpm for 15 min. Dynamic floc size was monitored in the whole procedure using SALLS (Mastersizer 2000, Malvern, UK). Size measurements were taken every 40 s. The inter-particle bonds that hold aggregate flocs together are considered as the cohesive strength of the flocs. A size ratio method is used here with an index (strength factor) to express the strength of particle flocs, i.e., Strength Factor ¼
Jar tests were performed on a programmable jar test apparatus (Daiyuan Jar Test Instruments, China) to investigate the coagulation performance of MPACl2.0s within the dosage
d2 100% d1
(2)
where d1 and d2 are the mean sizes of the flocs before and after the shear breakage, respectively. A higher value of the
Fig. 1 e TEM images of two ferromagnetic nanoparticles: (a) FSCSP, (b) SI.
130
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5
d3 d2 100% Recovery Factor ¼ d3 d2
Results and discussion
3.1.
Properties of ferromagnetic nanomaterials
3.1.1.
Size and morphologies of FSCSP and SI
TEM images of FSCSP and SI are shown in Fig. 1. In Fig. 1(a), FSCSPs can be easily observed in the spherical shape with an average size of 20 nm. The non-uniform-sized nanospheres form submicron aggregates with a visible 10 nm shell. The agglomeration is due to the magnetic dipoleedipole attraction and the large specific surface area of the magnetic particles (Liu et al., 2004), in spite of fierce stirring during the silica coating process. Fig. 1(b) confirms the core-shell structure of SI. The passivated layer is Fe3O4 according to the product description. The TEM results reveal that the sizes of the ferromagnetic nanoparticles are of nanoscale before agglomeration. It has been reported that Fe3O4 is superparamagnetic if its diameter is around 16 nm (Hong, 2008). Unlike ordinary ferromagnetic particles, superparamagnetic materials have no hysteresis effect e once the external magnetic field is cut off, the remnant magnetism will disappear. This characteristic is favored the magnetic separation process since the magnetic sludge or flocs may be easily scraped when the magnetic field is off (Ngomsik et al., 2005).
3.1.2.
Specific surface area and pore volume
Nanoparticles have been proven to have relatively large specific surface areas, leading to greater interface reaction rates (Li et al., 2006). The specific surface areas and pore diameters were 78.76 m2/g and 153.24 A for FSCSP and 22.51 m2/g and 63.48 A for SI respectively. They were determined by BET surface measurement. Apparently, compared with the commercially available SI, the lab-prepared FSCSP has a larger surface area with more sites to react with Al species or water pollutants.
3.1.3.
FSCSP SI
20
10
0 2
3
4
5
6
7
8
9
10
11
12
pH value -10
(3)
where d3 is the mean size of the particle flocs after reflocculation at the original shear rate. A higher Recovery Factor suggests a greater flocculation and regrowth capability of the flocs after the shear breakage.
3.
30
Zeta Potential/mV
strength factor indicates a higher ability of the flocs to resist breakage when exposed to an elevated fluid shear. When the shear intensity was reduced after the breakage phase, re-flocculation of the particles could take place. A reversibility factor is used here to measure the reflocculation potential of the particles when the shear is returned to its original level. A modified size ratio approach may be applied to calculate the reversibility (Recovery Factor) by
-20
-30
Fig. 2 e Variation of the zeta potentials of the ferromagnetic nanoparticles with pH.
point (IEP) of FSCSP is at pH 5.2 while that of SI is at pH 6.6. It is worth noting that magnetite can produce charges in the protonation and deprotonation reactions of ^FeeOH surface sites as expected for an amorphous solid (Hajdu´ et al., 2009). Therefore, both particles can be positively charged in composited coagulants since the pH value of PACl2.0 is around 4.02, lower than IEPs of SI and FSCSP (Table 1). Zeta potentials of the MPACl2.0s apparently rise with the addition of magnetic nanoparticles, and the pH values of PACl2.0 and MPACl2.0s are very close around 4.0 Among the three coagulants, PACl2.0 þ FSCSP (100 mg/L) has a highest zeta potential of 21.5 mV PACl2.0 þ FSCSP (50 mg/L) has a lowest zeta potential. The zeta potential of PACl2.0 þ SI (25 mg/L) is 20.5 mV.
3.2.
Al species distributions of MPACl2.0s and PACl2.0
All the coagulants were aged for 1 week. The liquid 27Al NMR method was used to characterize Al species distributions in MPACl2.0s and the original PACl2.0. The results are shown in Fig. 3. The Alm, Al13 and Alun contents of PACl2.0 were 12.8%, 63.0% and 24.2%, respectively. It should be noted that the Alm and Al13 contents decline whereas the Alun content increases after compounding with ferromagnetic nanoparticles. Such change is more obvious for the two PACl2.0 þ FSCSPs. When the loading concentration is 50 mg/ L, Alm and Al13 contents decrease to 9.9% and 54.0%, respectively. For 100 mg/L, they further decrease to 9.3% and 52.5%, respectively.
Table 1 e Characteristics of electric charges for PACl2.0 and MPACl2.0s. Types of coagulants
pH
Zeta potential (mV)
PACl2.0 PACl2.0 þ FSCSP (50 mg/L) PACl2.0 þ FSCSP (100 mg/L) PACl2.0 þ SI (25 mg/L)
4.02 3.99 3.96 3.95
3.6 15.0 21.5 20.5
Surface electric charge
The electric charge characteristics of both nanoparticles are shown in Fig. 2. Zeta potentials of FSCSP and SI decrease with increase in pH values, changing from þ18.7 mV to þ26.0 mV down to 13.0 mV and 27.7 mV, respectively. The isoelectric
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5
3.3. Coagulation performances of MPACl2.0s and PACl2.0 with different dosages
100
Al species distributions/%
131
90 80 70 Alm
60
Al13
50
Alun
40 30 20 10 0 PACl2.0
PACl2.0+ FSCSP (50 mg/L)
PACl2.0+ FSCSP (100 mg/L)
PACl2.0+ SI (25 mg/L)
Fig. 3 e Comparison of Al species distributions.
All MPACl2.0s are in the form of solideliquid mixure. The high surface energy, strong adsorption and magnetic attraction of nanoparticles are apt to play a significant role in their combination with PACl2.0. The high density of reactive surface sites and the great intrinsic reactivity of ferromagnetic nanoparticles will enhance the reactivity of the nanoscale particles (Li et al., 2006), and allow Al species to adsorb on their surface. In the process of compositing and aging the MPACl2.0, it seems that Alm and Al13 may firstly aggregate into [Al13]n or [Alm]n on the surface ofthe nanoparticles, then transform to a sol or amorphous solid phase, and finally turn to other higher order aggregates. As a result, Al species-nanoparticle clusters form (shown in Fig. 7(a)), being identified as Alun. They are not only large in size and easy to precipitate but also have a crucial positive effect on bridging, enmeshment and sorption flocculation capacity. MPACl2.0s with different nanoparticles and loadings lead to different Al species distributions. According to the BET results, FSCSP has a larger surface area and pore volume, offering more adsorption sites for Al species. Therefore, the content of Al species-nanoparticle clusters in PACl2.0 þ FSCSP, in the presence of Alun, will rise. As for MPACl2.0 with SI, since SI was more positively charged in PACl2.0 (Table 1), the electrostatic repulsion between Al species and SI particles may hinder their approach and aggregation, resulting in an inconspicuous increase of Alun.
The HA and Kaolin particles removal efficiencies of MPACl2.0s and PACl2.0 and the zeta potential of the coagulated suspensions were comparatively investigated, as shown in Fig. 4. As anticipated, the addition of the MPACl2.0s and PACl2.0 effectively reduced the surface charge of the particles in the synthetic water, resulting in particle destabilization (Fig. 4). As the dosage increased, the particle zeta potentials approached zero. Further increase in flocculant dose caused a certain extent of charge reversal of the particles. The isoelectric dosage (IED) of PACl2.0 þ FSCSP (50 mg/L) is close to 0.0 mmo/L as Al, slightly lower than that of the others which were higher than 0.10 mmol/L as Al. The whole dosage range can be divided into three parts according to the change of zeta potentials and turbidity removals: (I) No significant flocculation from 30 to 10 mV. Dosages of coagulants are too low for MPACl2.0s and PACl2.0 to effectively flocculate/coagulate with water pollutants. (II) Obvious flocculation from 10 to 0 mV. As the dosage increases, big flocs are formed and zeta potentials are close to the IEP. Optimal turbidity removals are achieved in this range. (III) Restability from 0 to 10 mV. In the high dosage range, the coagulation performance is ineffective. The jar-test flocculation and sedimentation results for turbidity and DOC removals from the synthetic waters are in general agreement with the zeta potential analysis. For all four flocculants, it is obvious that the optimal dosage range for MPACl2.0s and PACl2.0 lies in Part II, and the optimal dosage is 0.08 mmol/L as Al. At the optimal dose, 90% of DOC and more than 90% of turbidity can be removed. Especially for PACl2.0 þ FSCSP (50 mg/L), the DOC and turbidity removal efficiencies can reach 92% and 98%, respectively. In comparison, PACl showed a slightly worse water treatment performance. The results indicated that the nanoscale magnetic particles did play a significant role in DOC and particle removal. Firstly, ferromagnetic nanoparticles could affect the hydrolyzed Al species by increasing Al species-nanoparticle clusters, which may increase effective collision rates. Secondly, the addition of ferromagnetic nanoparticles would enhance the adsorption capability due to their large specific surface areas and magnetic dipoleedipole attraction, resulting in a higher removal rate of water pollutants (Li et al., 2006). Therefore,
Fig. 4 e Variation of turbidity and DOC removal with coagulant dosages.
132
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5
Fig. 5 e Representative images of flocs formed with (a) PACl2.0; (b) PACl2.0 D FSCSP (50 mg/L); (c) PACl2.0 D FSCSP (100 mg/L) and (d) PACl2.0 D SI (25 mg/L).
3.4.
Floc properties
3.4.1.
Morphology of flocs
Microscopic examination in this study (Fig. 5) suggested the formation of large and fractal-like flocs, resulting in the removal of turbidity and DOC in the suspension. Representative images of flocs coagulated by PACl2.0, PACl2.0 þ FSCSP (50 mg/L), PACl2.0 þ FSCSP (100 mg/L), and SI (PACl2.0 þ SI (25 mg/L) are shown in Fig. 5. It may be noted the black dots were observed in the flocs formed by MPACl, revealing that nanoscale magnetic particles were successfully composited with the PACl. This can also imply the formation of Al speciesnanoparticle clusters. Those clusters influence not only the kinetics of aggregation, but also the structure and resulting fractal dimension of the aggregates.
3.4.2.
Floc formation, breakage and regrowth
The coagulation kinetics of PACl2.0 and MPACl2.0s are demonstrated in Fig. 6. It shows the variation in the average floc size during the hydrodynamic sequencing. In the jar tests, all the four coagulants were dosed of 0.08 mmol/L as Al. After the slow mix phase, the coagulation for four suspensions achieved a steady-state with the mean sizes denoted as d1 in Table 2. Then floc size was immediately reduced following an increase in shear. 1 min later, it can be assumed the flocs reached the size d2. A reversibility phenomenon in terms of
floc size was observed, as mentioned in previous studies (Yukselen and Gregory, 2002; Zhu et al., 2009). When the shear was reduced to its initial value, the broken particles could collide with each other again to form the larger ones. However, it may be noticed that the formed flocs could not regrow to anywhere near their previous size. This may be attributed to the different coagulation mechanisms. Flocs formed by charge neutralization should give total recoverability. Thus, the irreversible breakage of the flocs was considered as evidence that the flocs formed in these systems were not dominated by pure charge neutralization mechanisms and were therefore held together by chemical rather than physical bonds, such as the combination of entrapment
200
PACl2.0 PACl2.0+FSCSP (50 mg/L) PACl2.0+FSCSP (100 mg/L) PACl2.0+SI (25 mg/L)
160
Floc size/um
adsorption, enmeshment and sweep would have simultaneously taken place by adding MPACl2.0. That is the primary reason why the performance of MPACl2.0 coagulation is superior to that of PACl2.0 coagulation.
120
80
40
0 0
200
400
600
800
1000
1200
1400
1600
1800
2000
Time/s
Fig. 6 e Comparison of the floc formation kinetics curves.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5
Table 2 e Strength and recovery factors of MPACl2.0s and PACl2.0. Suspensions
d1 (mm)
d2 (mm)
d3 (mm)
Strength factor (%)
Recovery factor (%)
PACl2.0 PACl2.0 þ FSCSP (50 mg/L) PACl2.0 þ FSCSP (100 mg/L) PACl2.0 þ SI (25 mg/L)
136.5 165.4
43.6 42.9
73.9 51.3
31.9 25.9
32.6 6.9
139.0
46.7
77.9
33.7
33.8
173.3
46.4
61.5
26.8
11.9
bridging and complexation with coagulant metal hydrolysis species (Yukselen and Gregory, 2004). Table 2 summarizes the values of the strength factor and recovery factor for the suspensions with different coagulants. PACl2.0 þ SI (25 mg/L) and PACl2.0 þ FSCSP (50 mg/L) are less able to resist the shearing condition, and their strength factor values are 26.8% and 25.9%, respectively, lower than those of 33.7% for PACl2.0 þ FSCSP (100 mg/L) and 31.9% for PACl2.0. Meanwhile, recovery factors of PACl2.0 þ SI (25 mg/L) and PACl2.0 þ FSCSP (50 mg/L) are only as low as 6.9% and 11.9%. Compared with PACl2.0, the small addition of ferromagnetic nanoparticles slightly reduced the floc strength but severely weakened the recovery ability of the flocs. This can be explained according to the variation of the Al species shown in Fig. 3. The floc strength is dependent upon the interparticle bonds between the components of the aggregates. As reported by Wang et al. (2009), the Alm species can complex with HA to form large and strong flocs while the Al13 species react with HA to form small and unstable flocs. Those two HA-
133
Al flocs can be joined together by adsorption and bridging due to larger polymer and solid-phase Al(OH)3, resulting in larger flocs. In our study, the increase of Alun in MPACl2.0s, containing the probably existing Al species-nanoparticle clusters, may have helped to form larger flocs than PACl2.0. The decrease of Alm in MAPCl2.0s and the possible loose conformation of Al species-nanoparticle clusters may lead to weaker flocs. Moreover, the results also indicate that flocs formed by MPACl2.0s show lower strength and reflocculating ability after breakage. As required for the magnetic particle recycling, the smash of magnetic flocs/sludge is a prerequisite, after which the separation of magnetic particles from the flocs/sludge is easier to achieve. Therefore, the incompact flocs, induced by PACl2.0 þ FSCSP (50 mg/L) and PACl2.0 þ SI (25 mg/L), are favored for practical purposes because of the low energy consumption for flocs/sludge breakage.
3.5.
Mechanisms of nanoscale magnetic coagulants
The results of this work indicate that the ferromagnetic nanoparticles could play a significant role in coagulation. A plausible MPACl2.0 coagulation mechanism has been described schematically in Fig. 7. It can be noted that Al species-nanoparticle clusters will be formed when ferromagnetic nanoparticles are added into the PACl (as shown in Fig. 7(a)). The formed clusters would increase the proportion of Alun, which has a great potential to enhance charge neutralization, enmeshment and adsorption when aggregated with the pollutants. The flocs produced by those clusters exhibited different characteristics as shown in Fig. 4.
Fig. 7 e Coagulation mechanism of MPACl2.0: (a) formation of Al species-nanoparticle clusters in MPACl2.0, (b) coagulating process of MPACl2.0.
134
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5
As illustrated as Fig. 7(b), the flocs formed by PACl2.0 þ FSCSP (50 mg/L) and PACl2.0 þ SI (25 mg/L) are larger but weaker. As stated previously, ferromagnetic nanoparticles exhibited a preference to form Al species-nanoparticle clusters. These clusters enhance the bridge and adsorption effects, leading to a larger floc size. Moreover, the recovery factors of the flocs show marked differences. The recovery factors of PACl2.0 þ FSCSP (50 mg/L) and PACl2.0 þ SI (25 mg/L) are very low (6.9% and 11.9% respectively). This implies that, in a suitable range of ferromagnetic nanoparticle addition, the connections between the clusters and pollutants are unique, being based on chemical bonds rather than physical ones. Flocs formed by PACl2.0 þ FSCSP (100 mg/L) showed a different situation. They were stronger and had a comparable size with those formed by PACl. The difference is in the concentration of ferromagnetic nanoparticles. Apparently, PACl2.0 þ FSCSP (100 mg/L) had the highest concentration. Aggregation of nanoparticles happened easily, and thus Al species-nanoparticle clusters were reduced. Therefore, the bridge and adsorption effects of Alun were limited to some degree, and the floc size did not obviously increase. Additionally, the formed flocs were considerably porous and fractal. The excess nanoparticles could penetrate into flocs. The embodiment of those nanoparticles would increase the strength of the flocs as indicated in Table 2.
4.
Conclusions
Three novel ferromagnetic nanoparticle composites PACl2.0s (MPACl2.0s) were synthesized by compounding FSCSP and SI respectively with PACl2.0. They are described as PACl2.0 þ FSCSP (50 mg/L), PACl2.0 þ FSCSP (100 mg/L) and PACl2.0 þ SI (25 mg/L). A series of characterizations in terms of physiochemical properties illustrated that the ferromagnetic nanoparticles used in the study were all core-shell particles with slight agglomeration and positive charge PACl2.0. FSCSP had a larger surface area and pore volume than SI did. All MPACl2.0s were in the form of solideliquid mixing. Liquid 27Al NMR measurements indicated that, compared with PACl2.0, Alun content increased whereas Al13 and Alm declined with increase of ferromagnetic nanoparticle loadings. Al speciesnanoparticle clusters, attributed to Alun, are suggested to form. Coagulation jar tests revealed that the MPACl2.0s performed better than PACl2.0 in both turbidity removal, and DOC removal. Among the three MPACl2.0s, the coagulation behavior of PACl2.0 þ FSCSP (50 mg/L) was the best. The improved coagulation efficiency of MPACl2.0s may be considered to be due to the co-effect of Al species, ferromagnetic nanoparticles and the possible Al species-nanoparticle clusters. Investigation of floc properties revealed that the key factor that influenced floc growth, re-flocculation, size and strength was the loading of ferromagnetic nanoparticles. Moderate nanoparticle loadings induced large but weak flocs that were hard to rearrange. Such flocs are favored for magnetic particle recovery. The formation of Al species-nanoparticle clusters might have greatly contributed to the unique Al species distributions, the coagulation performance and the floc properties of
MPACl2.0. The additions of FSCSP and SI appeared to enhance the adsorption, enmeshment, and sweep abilities of MPACl2.0s, and to further strengthen the coagulated flocs by entering into the floc pores during flocculation. When the ferromagnetic nanoparticle loadings are controlled within a certain range, the connections between clusters and water pollutants were presumed to be chemical bonds since they had a considerably low recovery factors.
Acknowledgments This research was funded by the Natural Science Foundation of China under 50921064, 51025830 and 51008293. The authors are very grateful to technical support from the Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences.
references
Ambashta, R.D., Sillanpa¨a¨, M., 2010. Water purification using magnetic assistance: a review. Journal of Hazardous Materials 180, 38e49. Hajdu´, A., Ille´s, E., Tomba´cz, E., Borba´th, I., 2009. Surface charging, polyanionic coating and colloid stability of magnetite nanoparticles. Colloids and Surfaces A: Physicochemical and Engineering Aspects 347, 104e108. Hong, R.Y., 2008. Preparation and Application of Magnetic Nanoparticles and Fluid, first ed. Chemical Industry Press, China. 11e12. Hong, R.Y., Li, J.H., Zhang, S.Z., Li, H.Z., Zheng, Y., 2009. Preparation and characterization of silica-coated Fe3O4 nanoparticles used as precursor of ferrofluids. Applied Surface Science 255, 3485e3492. Hou, C. F., 2009. The characterization of nanoscale-iron and synthetic resins and the test of their activity. MD thesis, Water Pollution Control Group of Municipal Engineering, Northern JiaoTong University, China. Hu, C.Z., Liu, H.J., Qu, J.H., Wang, D.S., Ru, J., 2006. Coagulation behavior of aluminum salts in eutrophic water: significance of Al13 species and pH control. Environmental Science and Technology 40, 325e331. Ille´s, E., Tomba´cz, E., 2006. The effect of humic acid adsorption on pH-dependent surface charging and aggregation of magnetite nanoparticles. Journal of Colloid and Interface Science 295, 115e123. Jiang, C., Wang, R., Ma, W., 2010. The effect of magnetic nanoparticles on Microcystis aeruginosa removal by a composite coagulant. Colloids and Surfaces A: Physicochemical and Engineering Aspects 369, 260e267. Johansson, G., 1960. On the crystal structures of some basic aluminum salts. Acta Chemica Scandinavica 14, 771e773. Korbutowicz, M.K., Nowak, K.M., Winnicki, T., 2008. Water treatment using MIEXDOC/ultrafiltration process. Desalination 221, 338e344. Li, L., Pan, M.H., Brown, R.C., Leeuwen, J.V., 2006. Synthesis, properties, and environmental applications of nanoscale ironbased materials: a review. Critical Reviews in Environmental Science and Technology 36, 405e431. Li, Y.R., Wang, J., Zhao, Y., Luan, Z.K., 2010. Research on magnetic seeding flocculation for arsenic removal by superconducting magnetic separation. Separation and Purification Technology 73, 264e270.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 2 7 e1 3 5
Liu, D., Cai, D.Q., Wang, X.Q., Yu, Z.L., 2006. Removal of algal blooms in freshwater by magnetic polymer. China Environmental Science 26 (6), 677e679. Liu, Z.L., Wang, H.B., Lu, Q.H., Du, G.H., Peng, L., Du, Y.Q., Zhang, S.M., Yao, K.L., 2004. Synthesis and characterization of ultrafine well-dispersed magnetic nanoparticles. Journal of Magnetism and Magnetic Materials 283, 258e262. Lu, Z.Y., Dai, J., Song, X.N., Wang, G., Yang, W.S., 2008. Facile synthesis of Fe3O4/SiO2 composite nanoparticles from primary silica particles. Colloids and Surfaces A: Physicochemical and Engineering Aspects 317, 450e456. Ngomsik, A.F., Bee, A., Draye, M., Cote, G., Cabuil, V., 2005. Magnetic nano- and microparticles for metal removal and environmental applications: a review. Comptes Rendus Chimie 8, 963e970. Odegaard, H., Fettig, J., Ratnaweera, H.C., 1990. Coagulation with prepolymerized metal salts. In: Hahn, H.H., Klute, R. (Eds.), Chemical Water and Wastewater Treatment. Springer-Verlag, New York, pp. 189e220. Savage, N., Diallo, M.S., 2005. Nanomaterials and water purification: opportunities and challenges. Journal of Nanoparticle Research 7, 331e342. Shen, Y.F., Tang, J., Nie, Z.H., Wang, Y.D., Ren, Y., Zuo, L., 2009. Preparation and application of magnetic Fe3O4 nanoparticles for wastewater purification. Separation and Purification Technology 68, 312e319. Singer, P.C., Bilyk, K., 2002. Enhanced coagulation using a magnetic ion exchange resin. Water Research 36, 4009e4022.
135
Svoboda, J., Fujita, T., 2003. Recent developments in magnetic methods of material separation. Minerals Engineering 16, 785e792. Tang, H.X. (Ed.), 2006. Inorganic Polymer Flocculation Theory and Flocculants, first ed. China Architecture and Building Press, China, pp. 11e12. Van Benschoten, J.E., Edzwald, J.K., 1990. Chemical aspects of coagulation using aluminum salts - II. Coagulation of fulvic acid using alum and polyaluminum chloride. Water Research 24, 1527e1536. Wang, Y., Gao, B.Y., Xu, X.M., Xu, W.Y., Xu, G.Y., 2009. Characterization of floc size, strength and structure in various aluminum coagulants treatment. Journal of Colloid and Interface Science 332, 354e359. Xu, Y., Wang, D.S., Liu, H., Lu, Y.Q., Tang, H.X., 2003. Optimization of the separation and purification of Al13. Colloids and Surfaces A: Physicochemical and Engineering Aspects 231, 1e9. Yukselen, M.A., Gregory, J., 2002. Breakage and re-formation of alum flocs. Environmental Engineering Science 19 (4), 229e236. Yukselen, M.A., Gregory, J., 2004. The reversibility of floc breakage. International Journal of Mineral Processing 73, 251e259. Zhang, R., Vigneswaran, S., Ngo, H., Nguyen, H., 2007. A submerged membrane hybrid system coupled with magnetic ion exchange (MIEX) and flocculation in wastewater treatment. Desalination 216, 325e333. Zhu, Z., Li, T., Lu, J.J., Wang, D.S., Yao, C.H., 2009. Characterization of kaolin flocs formed by polyacrylamide as flocculation aids. International Journal of Mineral Processing 91, 94e99.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Integration of anammox into the aerobic granular sludge process for main stream wastewater treatment at ambient temperatures M.-K.H. Winkler, R. Kleerebezem, M.C.M. van Loosdrecht* Delft University of Technology, Department of Biotechnology, Julianalaan 67, 2628 BC Delft, The Netherlands
article info
abstract
Article history:
Anaerobic ammonium oxidation, nitrification and removal of COD was studied at ambient
Received 25 May 2011
temperature (18 C 3) in an anoxic/aerobic granular sludge reactor during 390 days. The
Received in revised form
reactor was operated in a sequencing fed batch mode and was fed with acetate and
10 October 2011
ammonium containing medium with a COD/N ratio of 0.5 [g COD/gN]. During influent
Accepted 17 October 2011
addition, the medium was mixed with recycled effluent which contained nitrate in order to
Available online 6 November 2011
allow acetate oxidation and nitrate reduction by anammox bacteria. In the remainder of the operational cycle the reactor was aerated and controlled at a dissolved oxygen
Keywords:
concentration of 1.5 mg O2/l in order to establish simultaneous nitritation and Anammox.
Anammox
Fluorescent in-situ hybridization (FISH) revealed that the dominant Anammox bacterial
Granular sludge
population shifted toward Candidatus “Brocadia fulgida” which is known to be capable of
AOB
organotrophic nitrate reduction. The reactor achieved stable volumetric removal rates of
Acetate
900 [g N2eN/m3/day] and 600 [g COD/m3/day]. During the total experimental period
Ambient temperature
Anammox bacteria remained dominant and the sludge production was 5 fold lower than what was expected by heterotrophic growth suggesting that consumed acetate was not used by heterotrophs. These observations show that Anammox bacteria can effectively compete for COD at ambient temperatures and can remove effectively nitrate with a limited amount of acetate. This study indicates a potential successful route toward application of Anammox in granular sludge reactors on municipal wastewater with a limited amount of COD. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Anaerobic ammonium oxidizing bacteria (Anammox) are capable of autotrophic ammonium oxidation with nitrite as electron acceptor (Strous et al., 1999). After the discovery of Anammox by Mulder in 1985, Anammox bacteria have successfully been implemented in full scale wastewater treatment systems to treat ammonium rich wastewater cost effectively (Abma et al., 2010; Sliekers et al., 2003; van der Star
et al., 2007; Wett, 2007). Currently Anammox is applied at mesophilic temperatures and on wastewater containing high concentrations of ammonium. In order to supply Anammox with nitrite for the oxidation of ammonium two different systems are proposed. Nitrite can be either produced in a separated tank by partial nitrification and in turn be fed into an non-aerated Anammox reactor (Sharon-Anammox) or produced in an oxygen limited one stage system (CANON) (Sliekers et al., 2003; van Dongen et al., 2001). In the latter
* Corresponding author. E-mail address:
[email protected] (M.C.M. van Loosdrecht). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.034
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4
system ammonium oxidizing bacteria (AOB) and Anammox grow together in one granule in which AOBs are located on the outer oxygen penetrated shell, where they oxidize ammonium to nitrite. Anammox can grow in the oxygen shielded inner core were ammonium and nitrite are available (Hao et al., 2005). Currently the nitritation/anammox processes are applied predominantly for treatment of sludge digester rejection water and effluent from industrial anaerobic wastewater treatment that both do not contain any or only limited amounts of organic carbon. If Anammox can be applied at lower temperatures and nitrogen concentrations, its application potential could be extended to municipal sewage treatment (Jetten et al., 1997). In order to implement Anammox in sewage treatment, pre-removal of COD is normally required. Heterotrophic growth results in a decrease of the SRT and a very high SRT is essential for successful cultivation of Anammox at ambient temperatures. Given the low yield and growth rate of Anammox bacteria, heterotrophic growth should be minimized, to maintain a high fraction of Anammox bacteria in the sludge. Nitrate produced either by Anammox or nitrite oxidizing bacteria would have to be removed by nitrate reduction processes. Pre-removal of organic carbon can be established after the A-stage in an A/B process, after physio-chemical pretreatment, or after anaerobic digestion (Jetten et al., 1997; Joss et al., 2009; Kartal et al., 2010; Wett, 2007). Simultaneous partial nitrification, anammox and denitrification has been reported for treating wastewater with an approximate COD/N ratio of 0.5 [g COD/gN] at temperatures of 30e36 C under constantly aerated conditions (Chen et al., 2009; Lan et al., 2011; Xu et al., 2010). In these studies the COD was removed by regular heterotrophic bacteria. Recently, it was reported that certain Anammox species have the capacity to oxidize volatile fatty acids with nitrate as electron acceptor, while forming ammonium with nitrite as intermediate (Gueven et al., 2005; Kartal et al., 2007a, 2007b). Anammox does not incorporate the fatty acids into biomass, but completely converts them into CO2 (Kartal et al., 2007a). Why Anammox remains growing autotrophically while oxidizing acetate is not well understood but the low biomass yield associated with autotrophic growth is beneficial for wastewater treatment since sludge production is minimized. When COD oxidation with nitrate can be catalyzed by Anammox, nitrate produced by Anammox bacteria (or nitrite oxidizing bacteria) can be removed resulting in a lower nitrogen effluent concentration. Previous research in anoxic reactors showed that heterotrophs will win the competition for nitrate if the COD/N ratio exceeds 1 (Gueven et al., 2005). If oxygen and acetate are available at the same time Anammox bacteria do not only need to compete with NOBs for nitrite but also general heterotrophs will get an advantage over Anammox since Anammox is inhibited by oxygen (Strous et al., 1997). A better strategy would be therefore to proceed a nitritation/anammox period with an anoxic COD oxidation period. This allows treating wastewater with easy degradable soluble compounds such as acetate to promote the oxidation of acetate with nitrate by Anammox bacteria. In this study we aimed to explore the possibility to convert COD (in the form of acetate) and ammonium at a COD/N ratio of 0.5 by Anammox and AOB in a granular sludge reactor at ambient temperature
137
(18 C 3). Hereto we operated a granular sludge SBR. During one fourth of the cycle time the ammonium and acetate were mixed with reactor effluent containing nitrate from the previous cycle. In this way nitrate reduction can be catalyzed anoxically by Anammox with acetate. During the remaining three-fourth of the cycle the reactor was aerated for ammonium removal by nitritation and Anammox.
2.
Material and methods
2.1.
Long term operation
A lab-scale anoxic/aerobic bubble column reactor with a total volume of 2.9 L was run for 390 days at ambient temperature (18 3 C) in sequencing fed batch mode. The reactor was inoculated with granular Anammox sludge from the Rotterdam Dokhaven Anammox reactor. The reactor was operated in two phases (Fig. 3). In phase I (day 0e260), the reactor was fed with medium containg ammonium and nitrite (115 mg NO2eN/l (NaNO2), 190 mg NH4eN/l (NH4Cl)) and low concentrations of COD (25e90 mg COD/l (C2H3OONa)) to establish
Fig. 1 e Shows the cycle operation in experimental set-up of 1) mixed anaerobic fed batch period with nitrate from aerobic period and ammonium and COD in the influent (60 min) 2) aerobic period (172 min) 3) settling period (3 min) 4) discharge period (5 min).
138
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4
Fig. 2 e Nitrogen compounds and COD (in form of acetate) during one cycle of operation in A) phase I and B) phase II. Graph displays nitrate (-), nitrite (C) ammonium (:) total nitrogen (3) and acetate (A) profiles over time. Note that opposed to normal cycle operation all media was fed at once into reactor in order to follow trends over time during anoxic period.
a stable system for combined nitritation and Anammox and organic carbon removal. During phase I the COD/N ratio was kept at 0.1 until day 170 after which it was raised gradually by decreasing nitrite and increasing COD in the influent until a ratio of 0.4 (day 260). In phase II the medium consisted of a nitrite free and acetate rich medium (100 mg COD/l) reaching a COD/N ratio of 0.5 (phase II; day 260e390). The mineral medium consisted of 0.2 mM MgSO4∙7H2O, 0.2 mM KCl, 2 mM NaHCO3, 0.2 mM K2HPO4, 0.1 mM KH2PO4. ‘Visniac and Santer’ solution was used to provide trace elements (Visniak and Santer, 1957). The pH was maintained at 7.2 0.2 during the aerobic period by dosage of hydrochloric acid and sodium hydroxide. The dissolved oxygen (DO) concentration was controlled at 1.5 mg O2/l. The DO was set by recirculating the off-gas and blending with fresh air. In this way the DO could be regulated while maintaing a constant superficial air velocity of 2 l/min (Mosquera-Corral et al., 2005). Samples were taken on a weekly basis and analyzed for N-compounds
by the use of standard test kits (Hach-Lange). Sludge bed height remained constant over time (1 cm). Biomass production was monitored over time by catching effluent from one cycle and determining dry weight and ash content.
2.2.
Cycle operation
The reactor was operated in a sequencing fed batch mode and the different periods are displayed in Fig. 1. During the mixed anoxic feeding period (60 min) nitrate produced during the previous cycle was mixed with medium containing acetate and ammonium. After the feeding period an aerated period for partial nitrification was introduced lasting 172 min, followed by a settling period (3 min), and an effluent withdrawal period (5 min). During the effluent removal period half of the reactor liquid volume (1.5 l) was discharged and half (1.4 l) remained in the system.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4
139
Fig. 3 e A) Volumetric conversion rates of ammonium (A) and nitrite (:) as well as production of nitrate (3) B) biomass production rate (C) and the COD/N ratio (-) over time.
The reactor performance during one cycle of operation was analyzed for phase I (nitrite in the feed) and phase II (no nitrite in the feed) (Fig. 2). Samples were taken every 10e20 min to measure ammonium, nitrite and nitrate by means of flow injection analysis (Quick Chem8500, Lachat instruments). Acetate was measured by using a High Performance Liquid Chromatography (HPLC).
2.3.
Microscopic characterization of granules
Granules were taken for microscopic analysis in order to assess their morphology and microbiological composition. Slicing and FISH was accomplished by the method proposed by (Winkler et al., 2011b) (Fig. 4) to see the spatial distribution of bacteria as a function of depth within the granule. FISH was performed for determination of general Anammox bacteria (Cy3) general bacteria Eub (Cy5) and Candidatus “Brocadia
fulgida” (Fluos). For AOB a mix of two probes was prepared and labeled with (Cy5). Probe sequences are listed in Table 1.
2.4.
Biomass yields
The estimated community composition was based on produced biomass per consumed acetate (Yx/HAC) and ammonium ðYX=NHþ4 Þ, respectively (Table 2). It was assumed that all acetate which was fed into the reactor (100 mg COD/l) would be metabolized by either Anammox or heterotrophic bacteria. In case of Anammox being the only active bacteria half of the consumed ammonium was supposed to be used for partial nitrification (AOB) and the other half by anaerobic ammonium oxidation. Since Anammox does not incorporate acetate into biomass no growth on acetate for Anammox was assumed. For the conversion from COD to VSS a factor of 1.4 was used (Scherer et al., 1983).
140
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4
Fig. 4 e Microbial images from phase I (a,c) and II (b,d) a) FISH on sliced granules with general anammox bacteria (red), Candidatus “Brocadia fulgida” (green) and general AOB (blue) b) FISH on sliced granules with general anammox bacteria (red), Candidatus “Brocadia fulgida” (green) and general Eub (blue) c, d) Light microscopic images of granules in phase I (c) and II (d). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
3.
Results
3.1.
Long term operation
The reactor was operated in two phases over a time period of 390 days. The volumetric conversion rates are displayed in Fig. 3 and are based on the difference between the total soluble nitrogen compounds in the influent and effluent of the reactor. In phase I when the reactor influent contained nitrite (115 mg NO2eN/l) and small amounts of acetate (25 mg COD/l)
the nitrogen removal rate reached 1200 [g N2eN/m3/day] and gradually decreased upon decreasing the nitrite and increasing the COD concentration in the medium (Fig. 3A). Phase I started with a low COD/N ratio of 0.1 and reached a ratio of 0.5 in phase II. In phase II no nitrite was supplied and the influent COD concentration was kept constant at 100 mg COD/l reaching an average volumetric N conversion rate of 900 [g N2eN/m3/day] as well as a COD removal rate of 600 [g COD/m3/day] (Fig. 3A). The measured nitrate in the effluent during phase I was 71 16 [mg NO3/m3/day] (Fig. 3A day phase I (day0e260)), and decreased when acetate was increased
Table 1 e Oligonucleotide probes and primers target microorganisms, and references used in this study. Probes Amx 368 Bfu613 EUB 338 EUB 338 III NSO190 NSO1225
Sequence (from ‘5 to ‘3)
Specificity
Reference
CCTTTCGGGCATTGCGAA GGATGCCGTTCTTCCGTTAAGCGG GCTGCCTCCCGTAGGAGT GCTGCCACCCGTAGGTGT CGATCCCCTGCTTTTCTCC CGCCATTGTATTACGTGTGA
All Anammox bacteria Candidatus Brocadia fulgida Most bacteria Verrucomicrobiales All AOB All AOB
(Schmid et al., 2003) (Kartal et al., 2008) (Amann et al., 1990) (Daims et al., 1999) (Mobarry et al., 1997) (Mobarry et al., 1997)
Probes for Anammox fulgida were tagged with the fluorescent dye Fluos (green) general Anammox with Cy3 (red) and Eubs as well as AOBs with Cy5 (blue). For analysis probes of one target group were mixed.
141
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4
Table 2 e Biomass yields for anammox, AOB and heterotrophic bacteria and the corresponding biomass concentration and relative community composition according to consumed substrate (100 mg COD/l; 150 mg NH4eN/l) in reactor during one day of operation in phase II.
Anammox AOB Heterotrophs
YCx/Hac
YCx=NHþ4
g VSS/m3/day
e e 0.4
0.07 0.13 e
30 63 336
Community composition [%] 32a 68a e
7b 15b 78b
Reference for yields (van der Star, 2008) (Blackburne et al., 2007) (Beun et al., 2001)
a Calculations based on assumption that all acetate was converted by anammox. b Calculations based on assumption that all acetate was converted by heterotrophic bacteria.
down to 40 5 [mg NO3/m3/day] (Fig. 3A day phase II (day 260e380)). The biomass production in phase I was 61 14 g VSS/m3/day when a COD/N ratio of 0.1 was applied and 126 49 g VSS/m3/day when the COD/N ratio was elevated until 0.5. During phase II when COD/N ratio was kept constant at 0.5 the biomass production decreased from initial values of approximately 150 g VSS/m3/day to values around 55 g VSS/ m3/day.
3.2.
Reactor performance during one cycle of operation
A cycle measurement was conducted and changes in nitrogen concentrations during one cycle of operation during phase I (influent COD/N ratio 0.1) and phase II (influent COD/N ratio 0.5) are depicted in Fig. 2. Note that for a cycle measurement influent was fed at once into the reactor at t ¼ 0 min to be able to measure concentration profiles over time. During normal operation the mixed feeding period lasted 60 min. In phase I, when nitrite was present during anoxic feeding (0e60 min) nitrite and ammonium decreased while nitrate was formed as is can be expected to common Anammox stoichiometry. During the aerobic period (60e232 min) ammonium was oxidized by AOB and Anammox. Nitrate formation was due to Anammox activity as well the oxidation of nitrite by a small proportion of NOB, this gave a nitrate accumulation of 42 mg NO3eN/l nitrate at the end of the cycle in phase I. In the phase II, acetate removal occurred simultaneously with nitrate removal until complete depletion of electron acceptor and donor (0e60 min) (Fig. 2B). During the aeration period (60e232 min) reactor performance in phase II (Fig. 2B) was similar to the aerated period observed in phase I (Fig. 2A) with the difference that nitrate accumulation was lowered from average values of 42 (phase I) to average values 20 mg NO3eN/l (phase II). Ammonium measurements in the liquid are not completely indicative for the conversions since significant ammonium adsorption occurred (Bassin et al., 2011). Upon feeding ammonium is adsorbed to the granular sludge, while during conversion it gradually desorbs again (maintaining an adsorption equilibrium) during conversion.
3.3.
Microscopic analyses of granular sludge samples
During the reactor operation granules of red granulated biomass originating from the full scale Anammox reactor system in Rotterdam (Fig. 4c) developed over time into lightreddish granules (Fig. 4d) indicating a change in community composition. During phase I the Eubacterial microbial population mainly consisted of AOB and Anammox (data not
shown) whereas no Candidatus “Brocadia fulgida” was detected (Fig. 4a). Candidatus “Brocadia fulgida” (Fig. 4b) accumulated in the sludge once the nitrite was omitted from the feed during phase II, this organism was neither detected in the seed sludge nor during phase I (Fig. 4a). During phase I typical nitrite-anammox architecture of granular sludge was observed (Vlaeminck et al., 2010), with nitrifiers (here EUB) forming an outside coating of the Anammox granule. During phase II this structure essentially remained but now also a small fraction of heterotrophs got intermixed with the Anammox population (Fig. 4b).
3.4.
Biomass yields
In order to investigate the conversion of acetate in the reactor we evaluated the biomass production and microbial community composition on calculated biomass production based on two conversion routes. If only Anammox bacteria and AOB would dominate the system and all acetate and ammonium would be consumed by Anammox bacteria and AOB a sludge productivity of approximately 93 mg VSS/m3/day is expected and hence a system strongly dominated by Anammox bacteria (32%) and AOB (68%). If all acetate was assumed to be oxidized by heterotrophs and all ammonium assumed to be converted by partial nitrification in combination with anaerobic ammonium oxidation the sludge production was calculated to be 429 g VSS/m3/day. If acetate was fully converted by heterotrophs it can be expected that the community was dominated by heterotrophic bacteria (78%), with only 7% of Anammox bacteria and 15% AOB in the community (Table 2).
4.
Discussion
This research showed a possible new application of Anammox for wastewater treatment containing COD and ammonium with a COD/N ratio up to 0.5. It was shown that despite feeding the reactor with acetate, Anammox activity could be maintained and acetate oxidation was combined with anaerobic ammonium oxidation while keeping sludge production low. Removal rates were similar to those reported in system based on one-reactor nitration Anammox processes (Abma et al., 2010; Siegrist et al., 2008; Wett, 2007). Earlier research has shown that Anammox bacteria are capable of using organic acids like propionate and acetate (Kartal et al., 2007a, 2008). These authors demonstrated that Anammox bacteria did not incorporate acetate in biomass leading to low sludge production. Their studies were conducted in flocculent sludge
142
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4
reactors at a higher temperature and with constant feeding under non-aerated conditions. Our study here showed that Anammox bacteria could outcompete normal heterotrophic denitrifying bacteria for acetate at ambient temperatures. We moreover combined the operation of a SBR similar to aerobic granular sludge systems (Winkler et al., 2011a) with the capability of Anammox to use acetate as a second electron donor for nitrate and nitrite reduction besides ammonium. Excess nitrate from the aerated period was used as electron acceptor to oxidize acetate present in the influent leading to significantly lower nitrate in the effluent in phase II compared to phase I (Figs. 2 and 3). FISH pictures on sliced granules showed a higher amount of Anammox bacteria as expected from calculations assuming that all acetate is consumed by normal heterotrophic bacteria (Table 2). According to this calculation the microbial population would consist of 78% heterotrophs and only a minor fraction of Anammox bacteria (7%) with a total expected biomass production of 429 g VSS/m3/day (Table 2). This is 5 fold higher than what is expected for the pure autotrophic Anammox based acetate oxidation (93 g VSS/m3/day). The measured biomass production remained low during the last 100 days of phase 2 with values of 65 14 g VSS/m3/day which is close to calculated value of 95 g VSS/m3/day which assumed acetate oxidation by Anammox bacteria only (Table 2). FISH images showed a significantly higher fraction of Anammox bacteria than what could be expected if heterotrophic denitrification would have outcompeted Anammox bacteria for actetate (Table 2, Fig. 4b). In addition, the color of the biomass remained red after raising the COD concentration to 600 g COD/m3/day in phase II (Fig. 4cd). A clear shift of the Anammox population toward Candidatus “Brocadia fulgida”, known for its capability to use acetate (Kartal et al., 2008), was detected by FISH (Fig. 4a, b). Earlier studies on Anammox bacteria using granular sludge have shown an increased growth of nitrifiers in smaller granules. This is due to the fact that smaller granules (or flocks) have a larger aerobic volume fraction than larger granules thus favoring the growth of aerobic bacteria (Volcke et al., 2010; Winkler et al., 2011b). Moreover, slow growing organisms (here Anammox bacteria) are expected to grow in the inner part of a biofilm whereas faster growing organisms (here AOB or general heterotrophs) are pushed toward the rim and in turn out of the biofilm (Picireanu et al., 2004). In the granular sludge process flocks and small granules are easier washed-out. This gives a smaller SRT for the population that dominates these smaller granules and flocks (mainly the aerobic organisms and not Anammox bacteria) which is similar to studies as conducted in a nitrifying e denitrifying biofilm airlift suspension reactor (Van Benthum et al., 1997). Visual observation of microbial community composition within a granule confirmed that Anammox bacteria are located in the middle of the granule and a smaller fraction of AOB and heterotrophs are located on the outer shell of the granule (Fig. 4a, b). Therefore also due to erosion the latter organisms will have a shorter retention time. This will lead to a higher SRT for granules and hence to an enrichment of Anammox bacteria in the sludge. Measurements on removal capacities showed a total volumetric nitrogen removal of 900 [g N2eN/m3/day] when the COD/N ratio was 0.5 [mg/mg].
Nitrate reduction to ammonium coupled to acetate oxidation proceeds theoretically in a 4.6 mg COD/mg N-ratio which is below the measured ratio (7 0.7) in the system indicating that also a fraction of COD was converted by traditional heterotrophic COD oxidation. Likely the combination of COD removal by anammox bacteria and a selective washout of heterotrophic bacteria led to the high accumulation of anammox bacteria in the studied system. Previous studies in which Anammox bacteria were exposed to organic acids reported successful operation at low COD/N ratios around 0.5 (Chen et al., 2009; Lan et al., 2011; Xu et al., 2010), an increase in heterotrophic growth at COD/N ratio around 1 (Udert et al., 2008) and a loss in Anammox activity at a constant COD/N ratio above 1 (Gueven et al., 2005). A wastewater stream containing high loads of COD seems to be unsuitable for Anammox bacteria although this likely depends on the actual ammonium load and hence the COD/N ratio. Here we showed an option to increase removal capacity of a one stage nitration Anammox process at ambient temperatures by reducing the excess nitrate from the aeration period by mixing it with the influent containing acetate. Herewith the growth of heterotrophic bacteria can be minimized because excess nitrate can be reduced under nonaerated conditions via the organotrophic pathway of Anammox bacteria. In a continuously aerated reactor heterotrophs would have the availability of a strong electron acceptor and donor (oxygen and acetate), while Anammox bacteria are inhibited by the oxygen (Strous et al., 1997), which would hence give heterotrophs an extra advantage over Anammox bacteria to oxidize acetate. Current research suggests the usage of Anammox based treatment systems after either an A/B process, after physiochemical pretreatment, or anaerobic digestion (Jetten et al., 1997; Joss et al., 2009; Kartal et al., 2010; Wett, 2007). If after an A-stage or a pretreatment some soluble COD is left and ammonium levels are high then the here presented treatment strategy could be applied. This forms an option to use Anammox in the main stream while obtaining a stable nitrogen removal process at ambient temperatures and low COD/N ratios. The approach chosen here is close to the operational conditions of an aerobic granular sludge system (De Bruin et al., 2004) allowing potential future combination of this novel high rate technology and Anammox processes. The advantage of Anammox bacteria being able to remove COD as well as nitrate makes the implementation of the Anammox technology indeed easier. Nitrate is produced in the regular Anammox conversion due to the coupling of CO2 reduction for biomass synthesis to oxidation of nitrite to nitrate. Moreover at lower temperatures it might be more difficult to prevent nitrite oxidizing bacteria to grow in the system. This could lead to too high nitrate concentrations in the effluent. Integrating anoxic periods in the sequencing fed batch cycles of aerobic granular sludge reactors would indeed give the option for nitrate removal by organotrophic Anammox bacteria. In the current perspective it is not easy to speculate on potential effluent nitrate levels within Anammox systems but we believe it will not be problematic to reach similar levels of around 60e75% of total nitrogen removal in a municipal treatment plant. The organic composition of wastewater does not only contain acetate and despite the fact that Anammox is
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4
reported to use other C sources such as propionate (Gueven et al., 2005; Kartal et al., 2007b) it remains unclear how competitive Anammox can be in an environment containing a variety of different carbon sources and how sensitive Anammox bacteria are to fluctuations in COD/N ratios.
5.
Conclusions
Here we used the acetate oxidizing capacity of Anammox bacteria at ambient temperature conditions and low COD/N ratios in a nitritation/anammox granular sludge system. Nitrate produced by Anammox bacteria or due to the presence of a fraction of nitrite oxidizing bacteria was reduced significantly. Sludge production remained low suggesting that Anammox successfully competed with general heterotrophs for acetate. This is a first step toward the application of Anammox bacteria in the main stream of municipal wastewater treatment processes.
Acknowledgements This study is partly funded by DHV and STOWA in the framework of the Dutch national Nereda research programme.
references
Abma, W.R., Driessen, W., Haarhuis, R., van Loosdrecht, M.C.M., 2010. Upgrading of sewage treatment plant by sustainable and cost-effective separate treatment of industrial wastewater. Water Science and Technology 61, 1715e1722. Amann, R.I., Krumholz, L., Stahl, D.A., 1990. Fluorescentoligonucleotide probing of whole cells for determinative phylogenetic, and environmental-studies in microbiology. Journal of Bacteriology 172, 762e770. Bassin, J.P., Pronk, M., Kraan, R., Kleerebezem, R., v. L., M.C.M., 2011. Ammonium adsorption in aerobic granular sludge, activated sludge and anammox granules. Water Research 45, 5257e5265. Beun, J.J., Heijnen, J.J., van Loosdrecht, M.C.M., 2001. N-removal in a granular sludge sequencing batch airlift reactor. Biotechnology and Bioengineering 75, 82e92. Blackburne, R., Vadivelu, V.M., Yuan, Z.G., Keller, J., 2007. Determination of growth rate and yield of nitrifying bacteria by measuring carbon dioxide uptake rate. Water Environment Research 79, 2437e2445. Chen, H.H., Liu, S.T., Yang, F.L., Xue, Y., Wang, T., 2009. The development of simultaneous partial nitrification, ANAMMOX and denitrification (SNAD) process in a single reactor for nitrogen removal. Bioresource Technology 100, 1548e1554. Daims, H., Bruhl, A., Amann, R., Schleifer, K.H., Wagner, M., 1999. The domain-specific probe EUB338 is insufficient for the detection of all bacteria: development and evaluation of a more comprehensive probe set. Systematic and Applied Microbiology 22, 434e444. De Bruin, L.M.M., De Kreuk, M.K., van der Roest, H.F.R., Van Loosdrecht, M.C.M., Uijterlinde, C., 2004. Aerobic granular sludge technology, alternative for activated sludge technology? Water Science and Technology 49, 1e9.
143
Gueven, D., Dapena, A., Kartal, B., Schmid, M.C., Maas, B., Van de Pas-Schoonen, K., Sozen, S., Mendez, R., Op den Camp, H.J.M., Jetten, M.S.M., Strous, M., Schmidt, I., 2005. Propionate oxidation by and methanol inhibition of anaerobic ammonium-oxidizing bacteria. Applied and Environmental Microbiology 71, 1066e1071. Hao, X.D., Cao, X.Q., Picioreanu, C., van Loosdrecht, M.C.M., 2005. Model-based evaluation of oxygen consumption in a partial nitrification-Anammox biofilm process. Water Science and Technology 52, 155e160. Jetten, M.S.M., Horn, S.J., v. Loosdrecht, M.M., 1997. Towards a more sustainable municipal wastewater treatment system. Water Science and Technology 35, 171e180. Joss, A., Salzberger, D., Eugster, J., Knig, R., Rottermann, K., Burger, S., Fabijan, P., Leumann, S., Mohn, J., Siegrist, H., 2009. Full-scale nitrogen removal from digester liquid with partial nitritation and anammox in one SBR. Environmental Science and Technology 43, 5301e5306. Kartal, B., Kuenen, J.G., van Loosdrecht, M.C.M., 2010. Sewage treatment with anammox. Science 328, 702e703. Kartal, B., Kuypers, M.M.M., Lavik, G., Schalk, J., Op den Camp, H.J.M., Jetten, M.S.M., Strous, M., 2007a. Anammox bacteria disguised as denitrifiers: nitrate reduction to dinitrogen gas via nitrite and ammonium. Environmental Microbiology 9, 635e642. Kartal, B., Rattray, J., van Niftrik, L.A., van de Vossenberg, J., Schmid, M.C., Webb, R.I., Schouten, S., Fuerst, J.A., Damste, J.S., Jetten, M.S.M., Strous, M., 2007b. Candidatus "Anammoxoglobus propionicus" a new propionate oxidizing species of anaerobic ammonium oxidizing bacteria. Systematic and Applied Microbiology 30, 39e49. Kartal, B., van Niftrik, L., Rattray, J., de Vossenberg, J., Schmid, M.C., Damste, J.S.S., Jetten, M.S.M., Strous, M., 2008. Candidatus ‘Brocadia fulgida’: an autofluorescent anaerobic ammonium oxidizing bacterium. Fems Microbiology Ecology 63, 46e55. Lan, C.J., Kumar, M., Wang, C.C., Lin, J.G., 2011. Development of simultaneous partial nitrification, anammox and denitrification (SNAD) process in a sequential batch reactor. Bioresource Technology 102, 5514e5519. Mobarry, B.K., Wagner, M., Urbain, V., Rittmann, B.E., Stahl, D.A., 1997. Phylogenetic probes for analyzing abundance and spatial organization of nitrifying bacteria (vol 62, pg 2157, 1996). Applied and Environmental Microbiology 63, 815. Mosquera-Corral, A., de Kreuk, M.K., Heijnen, J.J., van Loosdrecht, M.C.M., 2005. Effects of oxygen concentration on N-removal in an aerobic granular sludge reactor. Water Research 39, 2676e2686. Picioreanu, C., Kreft, J.U., van Loosdrecht, M.C.M., 2004. Particlebased multidimensional multispecies Biofilm model. Applied and Environmental Microbiology 70, 3024e3040. Scherer, P., Lippert, H., Wolff, G., 1983. Composition of the major elements and trace elements of 10 methanogenic bacteria determined by inductively coupled plasma emission spectrometry. Biological Trace Element Research 5, 149e163. Schmid, M., Walsh, K., Webb, R., Rijpstra, W.I.C., van de PasSchoonen, K., Verbruggen, M.J., Hill, T., Moffett, B., Fuerst, J., Schouten, S., Damste, J.S.S., Harris, J., Shaw, P., Jetten, M., Strous, M., 2003. Candidatus "Scalindua brodae", sp nov., Candidatus "Scalindua wagneri", sp nov., two new species of anaerobic ammonium oxidizing bacteria. Systematic and Applied Microbiology 26, 529e538. Siegrist, H., Salzgeber, D., Eugster, J., Joss, A., 2008. Anammox brings WWTP closer to energy autarky due to increased biogas production and reduced aeration energy for N-removal. Water Science and Technology 57, 383e388.
144
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 3 6 e1 4 4
Sliekers, A.O., Third, K.A., Abma, W., Kuenen, J.G., Jetten, M.S.M., 2003. CANON and Anammox in a gas-lift reactor. FEMS Microbiology Letters 218, 339e344. Strous, M., van Gerven, E., Kuenen, J.G., Jetten, M., 1997. Effects of aerobic and microaerobic conditions on anaerobic ammonium-oxidizing (anammox) sludge. Applied and Environmental Microbiology 63, 2446e2448. Strous, M., Kuenen, J.G., Jetten, M.S.M., 1999. Key physiology of anaerobic ammonium oxidation. Applied and Environmental Microbiology 65, 3248e3250. Udert, K.M., Kind, E., Teunissen, M., Jenni, S., Larsen, T.A., 2008. Effect of heterotrophic growth on nitritation/anammox in a single sequencing batch reactor. Water Science and Technology 58, 277e284. Van Benthum, W.A.J., Van Loosdrecht, M.C.M., Heijnen, J.J., 1997. Process design for nitrogen removal using nitrifying biofilm and denitrifying suspended growth in a biofilm airlift suspension reactor. Water Science and Technology 36, 119e128. van der Star, W., 2008. Growth and Metabolism of Anammox Bacteria: PhD thesis, Technical Univertity Delft. van der Star, W.R.L., Abma, W.R., Blommers, D., Mulder, J.W., Tokutomi, T., Strous, M., Picioreanu, C., Van Loosdrecht, M.C.M., 2007. Startup of reactors for anoxic ammonium oxidation: experiences from the first full-scale anammox reactor in Rotterdam. Water Research 41, 4149e4163. van Dongen, U., Jetten, M.S.M., Van Loosdrecht, M.C.M., 2001. The SHARON-Anammox process for treatment of ammonium rich wastewater. Water Science and Technology 44, 153e160.
Visniak, W., Santer, M., 1957. The thiobacilli. Bacteriological Reviews 26, 168e175. Vlaeminck, S.E., Terada, A., Smets, B.F., De Clippeleir, H., Schaubroeck, T., Bolca, S., Demeestere, L., Mast, J., Boon, N., Carballa, M., Verstraete, W., 2010. Aggregate size and architecture determine microbial activity balance for onestage partial nitritation and anammox. Applied and Environmental Microbiology 76, 900e909. Volcke, E.I.P., Picioreanu, C., De Baets, B., van Loosdrecht, M.C.M., 2010. Effect of granule size on autotrophic nitrogen removal in a granular sludge reactor. Environental Technology 31, 1271e1280. Wett, B., 2007. Development and implementation of a robust deammonification process. Water Science and Technology 56, 81e88. Winkler, M.K.H., Bassin, J.P., Kleerebezem, R., van Loosdrecht, M.C.M., van den Brand, T.P.H., 2011a. Selective sludge removal in a segregated aerobic granular biomass system as a strategy to control PAO-GAO competition at high temperatures. Water Research 45, 3291e3299. Winkler, M.K.H., Kleerebezem, R., Kuenen, J.G., Yang, J., van Loosdrecht, M.C.M., 2011b. Segregation of biomass in cyclic anaerobic/aerobic granular sludge allows the enrichment of anaerobic ammonium oxidizing bacteria at low temperatures. Environmental Science and Technology. Xu, Z.-Y., Zeng, G.-M., Yang, Z.-H., Xiao, Y., Cao, M., Sun, H.-S., Ji, L.-L., Chen, Y., 2010. Biological treatment of landfill leachate with the integration of partial nitrification, anaerobic ammonium oxidation and heterotrophic denitrification. Bioresource Technology 101, 79e86.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Aerobic degradation of sulfanilic acid using activated sludge Gang Chen a,b, Ka Yu Cheng a,*, Maneesha P. Ginige a, Anna H. Kaksonen a a b
CSIRO Land and Water, CSIRO, Floreat, WA 6014, Australia College of Environmental Science and Engineering, Donghua University, Shanghai 201620, China
article info
abstract
Article history:
This paper evaluates the aerobic degradation of sulfanilic acid (SA) by an acclimatized
Received 9 June 2011
activated sludge. The sludge was enriched for over three months with SA (>500 mg/L) as
Received in revised form
the sole carbon and energy source and dissolved oxygen (DO, >5 mg/L) as the primary
15 September 2011
electron acceptor. Effects of aeration rate (0e1.74 L/min), DO concentration (0e7 mg/L) and
Accepted 18 October 2011
initial SA concentration (104e1085 mg/L) on SA biodegradation were quantified. A modified
Available online 28 October 2011
Haldane substrate inhibition model was used to obtain kinetic parameters of SA biodegradation and oxygen uptake rate (OUR). Positive linear correlations were obtained between
Keywords:
OUR and SA degradation rate (R2 0.91). Over time, the culture consumed more oxygen per
Azo dye
SA degraded, signifying a gradual improvement in SA mineralization (mass ratio of O2: SA
Sulfonated aromatic amines
at day 30, 60 and 120 were 0.44, 0.51 and 0.78, respectively). The concomitant release of
Haldane kinetics
near stoichiometric quantity of sulphate (3.2 mmol SO2 4 released from 3.3 mmol SA) and
Oxygen uptake rate
the high chemical oxygen demand (COD) removal efficacy (97.1%) indicated that the
Wastewater treatment
enriched microbial consortia could drive the overall SA oxidation close to a complete mineralization. In contrast to other pure-culture systems, the ammonium released from the SA oxidation was predominately converted into nitrate, revealing the presence of ammonium-oxidizing bacteria (AOB) in the mixed culture. No apparent inhibitory effect of SA on the nitrification was noted. This work also indicates that aerobic SA biodegradation could be monitored by real-time DO measurement. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
Wastewaters generated from textile factories often contain sulfonated aromatic amines, which are primarily originated from the reductive cleavage of sulfonated azo dyes (10e50% of the applied dyes remain in the wastewater) (O’Neill et al., 1999). Sulfanilic acid (4-aminobenzenesulfonic acid, SA) is one of the most representative sulfonated aromatic amines (Perei et al., 2001). Due to environmental and health concerns, SA-contaminated wastewaters need to be treated prior to its discharge into the environment (Chung and Cerniglia, 1992; Oh et al., 1997; Topac et al., 2009). Currently, aerobic biodegradation is considered as the most effective and
environmentally benign approach to treat SA-contaminated wastewaters (Tan et al., 2005). However, the negatively charged sulfonyl group of SA molecule is known to suppress biodegradation by most heterotrophic microbial communities due to the low permeability of SA through bacterial membranes (Hwang et al., 1989). An acclimatization period is often required to enrich an efficient SA-degrading microbial community in the treatment process (Tan and Field, 2005; Tan et al., 2005). Earlier research has investigated the aerobic degradation of SA by using microbial enrichments. For example, the pioneering works by Feigel and Knackmuss (1988, 1993) have shown that a defined co-culture of Hydrogenophaga palleroni S1
* Corresponding author. Tel.: þ61 8 9333 6158; fax: þ61 8 9336211. E-mail addresses:
[email protected],
[email protected] (K.Y. Cheng). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.043
146
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1
and Agrobacterium radiobacter S2 could effectively degrade SA under aerobic conditions. Some other SA-degrading pure strains such as Pseudomonas paucimobilis (Perei et al., 2001), Agrobacterium sp. PNS-1 (Singh et al., 2004), Pannonibacter sp.W1 (Wang et al., 2009), Ralstonia sp. PBA and Hydrogenophaga sp. PBC (Gan et al., 2011) have also been reported. Arguably, these studies have only limited practical implication as they mainly focused on the use of either pure or co-cultures. To practically treat SA-contaminated wastewaters, mixed microbial cultures such as activated sludge could be of significance. Activated sludge obtained from municipal wastewater treatment plants has been demonstrated as an effective mixed culture inoculum for starting up SA-contaminated wastewater treatment processes. For instances, Tan et al. (2005) reported that SA could be aerobically degraded using activated sludge previously contaminated with a mixture of sulfonated aromatic compounds. In a separate study, efficient acclimatization of a SA-degrading mixed culture was achieved even using activated sludge that has not been previously exposed to SA (Carvalho et al., 2008). Nonetheless, the relationship between oxygen consumption and SA degradation in these aerobic mixed culture processes has not been properly defined. Since oxygen is the primary electron acceptor for these microbial processes, understanding the relationship between the kinetics of SA degradation and oxygen consumption would facilitate process optimization of aerobic SA wastewater treatment. Therefore, the objectives of this study were (i) to enrich an aerobic SA-degrading mixed culture using activated sludge that has not been previously exposed to SA as microbial inoculum, and (ii) to evaluate the kinetics of SA degradation and oxygen consumption of the enriched culture. The relationship between oxygen consumption and SA degradation was elucidated, and the feasibility of using realtime dissolved oxygen (DO) measurement as a strategy to obtain SA degradation kinetics during the treatment process was explored. To our knowledge, these issues have not been previously addressed particularly in a mixed culture environment. This work would shed light on our fundamental understanding of biological removal of sulfonated aromatic amines in wastewaters.
2.
Materials and methods
2.1.
Chemicals
Sulfanilic acid (SA) (4-aminobenzenesulfonic acid, CAS number 121-57-3) was purchased from SigmaeAldrich (Australia). Some selected physicalechemical properties of SA are: density 1.485 g/mL (25 C), water solubility in 10 g/L (20 C) and Henry’s constant 8.89 1013 m3/mol (25 C). A working stock solution of SA (5 g/L) was prepared by dissolving the SA in deionized water. All chemicals used in this study were of analytical grade.
2.2. Bacterial inoculum and SA containing synthetic wastewater Activated sludge was obtained from a domestic municipal wastewater treatment plant in Perth, Australia, and was
stored at 4 C prior to use. Unless specified otherwise, the basal medium used in this work had a composition of (mg/L): NH4Cl 125, NaHCO3 125, MgSO4$7H2O 51, CaCl2$2H2O 300, FeSO4$7H2O 6.25, and 1.25 mL L1 of trace element solution, which contained (g/L): ethylenediamine tetraacetic acid (EDTA) 15, ZnSO4$7H2O 0.43, CoCl2$6H2O 0.24, MnCl2$4H2O 0.99, CuSO4$5H2O 0.25, NaMoSO4$2H2O 0.22, NiCl2$6H2O 0.19, NaSeO4$10H2O 0.21, H3BO4 0.014, and NaWO4$2H2O 0.05 (Cheng et al., 2010). In some experiments where the effects of the background ammonium and sulphate were tested, NH4Cl and FeSO4$7H2O were omitted in the medium (i.e. Fig. 6). Defined SA concentration in the synthetic wastewater was prepared by adding a known volume of the SA stock solution to the basal medium. pH was adjusted to 7.0 by using phosphate buffer (KH2PO4 and K2HPO4, 15e30 mM).
2.3.
Reactor configuration and general operation
A 2-L glass continuously-stirred tank bioreactor was used in this study to aerobically acclimatize the SA-degrading culture. The working volume of the culture medium was 1.5 L. An adjustable aeration pump was used to supply oxygen to the suspended culture medium at an air flow rate that was varied from 0 to 1.74 L/min. The suspension liquor was continuously stirred by using an overhanging turbine impeller stirrer to maximize mass transfer. A DO sensor and process monitor (TPS Pty. Ltd., Australia) was used to measure the DO concentration in the suspension liquor. The DO data was periodically recorded into an excel spreadsheet using a LabVIEW computer program. The reactor was operated in batch mode at room temperature (25 2 C).
2.4.
Experimental procedures
2.4.1.
Reactor start-up
The enrichment process was initiated by mixing the activated sludge (10%, v/v) with the medium to obtain an initial mixed liquor suspended solids (MLSS) concentration of 2000 mg/L. The initial SA concentration in the medium was 500 mg/L during the first two weeks of operation (weeks 1 and 2). During this period, medium renewal was performed every 2 days. For medium renewal, the aeration pump and the overhanging mixer were switched off to allow a complete sludge settlement (ca. 30 min) and only the supernatant was decanted. From week 2 to 8 (i.e. two months after the initial start-up), the initial SA concentration was gradually increased to about 1000 mg/L and medium renewal was performed daily. No sludge wastage was performed during the initial two months to prevent washout of microorganisms relevant to the process. After the initial two months until the end of the acclimatization (i.e. 5 months), 75 mL of the mixed liquor was wasted daily to maintain a steady range of a MLSS of 5000e6000 mg/L (sludge age ¼ 20 days). During the entire acclimatization process, DO concentration in the reactor was maintained at over 5 mg O2/L to ensure sufficient supply of DO.
2.4.2. SA degradation of the acclimatized culture at different initial SA concentrations and different aeration rates SA degradation kinetics of the acclimatized sludge was quantified after the initial 2 months of acclimatization. SA
147
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1
degradation profiles were obtained with different initial SA concentrations. The relationship between SA degradation rate and initial SA concentration was established using a modified Haldane inhibition model as described by Maeda et al. (2005) (Equation (1)). rSA ¼
ds 1 1 mmax SX km SX ¼ ¼ mX ¼ dt Y Y KS þ S þ S2 =KI KS þ S þ S2 =KI
(1)
where rSA (mg/L h) is the SA degradation rate, ds/dt is the change in SA concentration over time, S (mg/L) is the SA concentration, m (1/h) is the specific microbial growth rate, km (1/h) is the product of the maximum specific microbial growth rate (mmax) and the inverse of the biomass yield on the substrate (Y, mg/mg), KS (mg/L) is the saturation constant, KI (mg/L) is the substrate inhibition constant, and X is the biomass concentration (MLSS, mg/L). Since in this study the constants mmax and Y were not measured, the constant km ¼ mmax/Y was determined instead (Maeda et al., 2005). The MLSS concentration was maintained at 5800 mg/L and the aeration rate was maintained at 1.74 L/min during this period of study. To evaluate the effects of different aeration rates (0e1.74 L/ min) on SA degradation rate, a series of batch experiments were conducted with an initial SA concentrations in all runs ranged from 310 to 340 mg/L, and the MLSS concentration was maintained at 5600 mg/L. Changes in SA concentration over time at different aeration rates were compared, and used to derive the specific SA oxidation rates which were then plotted against (i) aeration rate and (ii) average DO concentration in the mixed liquor.
2.4.3. Correlation between oxygen uptake rate and SA degradation rate At day 30, 60 and 120, experiments were conducted to obtain the correlation between SA degradation rate and the oxygen uptake rate (OUR) of the enriched culture. The 2-L bioreactor was operated with an initial SA concentration of 500e1000 mg/ L, MLSS of 5800 mg/L and aeration rate of 1.74 L/min. Over the course of each kinetic experiment, multiple mixed liquor subsamples (50 mL) were periodically taken from the bioreactor and immediately transferred into a separate, 50-mL vessel equipped with a DO probe for the determination of OUR using a dynamic method as described by Garcia-Ochoa et al. (2010). Once transferred in the vessel, the mixed liquor was continuously mixed with a small magnetic stirrer bar and was purged with air until DO reached >7 mg O2/L. Thereafter, the air supply was terminated and the decline in DO was recorded to calculate the OUR. The specific OUR values were obtained by normalizing the OUR value with the MLSS concentration and plotted against the corresponding specific SA degradation rate to obtain a linear correlation. The ratio of mg O2 consumed per mg SA degraded was obtained from the slope. The correlation between oxygen uptake rate (OUR) and initial SA concentration was also established by using the modified Haldane model equation. The specific OUR ðqO2 Þ, is calculated as below (Equation (2)): qO2 ¼
1 YX=O2
m þ m0 ¼
mmax k0 þ m0 ¼ þ m0 2 KS þ S þ S2 =KI YX=O2 KS þ S þ S =KI
where YX=O2 (mg/mg) is the yield of biomass on the oxygen, KS (mg/L) is the saturation constant, KI (mg/L) is the substrate inhibition constant, S (mg/L) is the SA concentration, k0 (1/h) is the product of the maximum specific growth rate of the microorganism (mmax) and the inverse of the yield of biomass on the oxygen ðYX=O2 Þ and m0 is the maintenance constant. Since in this study, the constants mmax and YX=O2 were not measured separately, the constant k0 ¼ mmax =YX=O2 was determined instead.
2.4.4. Real-time OUR determination in a continuously aerated SA-degrading culture The feasibility of using real-time DO measurement for obtaining in-situ SA-dependent OUR data was explored with the continuously aerated culture (MLSS ¼ 5800 mg/L; constant aeration rate of 0.31 L/min). Four aliquots of SA solution (final SA concentration ¼ 50e300 mg/L) were sequentially added to trigger oxygen consumption activity of the culture. Samples (2 mL) were periodically collected to establish concentration profiles of SA, ammonium (NHþ 4 eN), nitrite (NO2 eN), nitrate 2 (NO3 eN) and sulphate (SO4 ). Real-time OUR (OURRT) was computed from the DO profile according to Equation (3) (Garcia-Ochoa et al., 2010). OURRT ¼ KLa ðCs CÞ
dC dt
(3)
where KLa is oxygen mass transfer coefficient (1/h) of the reactor under the constant aeration and stirring rates (KLa ¼ 17.6 h1); dC/dt is the rate of DO change; Cs is saturation concentration of DO (mg O2/L); C is the real-time DO concentration (mg O2/L). The amount of oxygen consumed per each SA addition was obtained by integrating the peak areas in the OURRT profile.
2.5.
Chemical analysis
Mixed liquor samples taken from the bioreactor were immediately filtered using a 0.2 mm sterile filter (0.8/0.2 mm Supor Membrane, PALL Life Sciences) to remove biomass. SA concentration of the filtrate was quantified by using a UVspectrophotometer (Cary 50, Varian, USA) at a maximal SA specific adsorption wavelength of 250 nm (determined experimentally in this study by scanning the adsorption at various wavelengths). The samples were diluted (if required) to obtain a SA concentration between 0 and 25 mg/L, which resulted in a linear standard curve (R2 > 0.999) with absorbance. MLSS were quantified according to a standard method (APHA, 1995). Chemical oxygen demand (COD) was measured using a closed reflux dichromate COD method (HACH Method þ 8000, HACH Ltd). SO2 4 , NH4 eN, NO3 eN and NO2 eN were quantified by using Ion Chromatography (ICS-3000, DIONEX).
3.
Results and discussion
3.1. SA biodegradation kinetics of the aerobically enriched activated sludge
1
(2)
After approximately two months of acclimatization, a kinetic experiment was conducted to quantify the SA degradation
148
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1
2004). This result suggests that the SA degradation behaviour of the acclimatized culture is similar to these pure and coculture studies.
a SA (mg L-1)
1000 800 600
3.2.
400
b
0
Specific SA degradation rate (mg/ g MLSS ·h)
200
5
5
Time (h) 20
10
4
25
30
Model Simulation Experimental Value
3 2 1 0
600 400 800 SA (mg L-1)
200
1000
Fig. 1 e Effect of initial SA concentration on SA degradation by enriched activated sludge culture. (a) SA concentration profiles at different initial SA concentrations; (b) relationship between specific SA degradation rate and the initial SA concentrations (R2 [ 0.97). Notes: all experiments were conducted with the same MLSS concentration of 5800 mg/L and aeration rate of 1.74 L/min.
capacity of the activated sludge at different initial SA concentrations (100e1100 mg/L) (Fig. 1). As expected, a longer time was required for the degradation of a higher initial SA concentration (Fig. 1a). To better quantify the relationship between the specific SA degradation rate and the initial SA concentration, a modified Haldane model was used to fit the obtained data (Equation (1)). The model fits the experimental data well with km, KS and KI determined as 1.04 106 h1, 98.7 mg/L and 3424 mg/L, respectively (R2 ¼ 0.97) (Fig. 1b). Both the KS and KI values for the mixed culture are similar to those reported for other highly acclimatized SA-degrading pure and co-cultures (Gan et al., 2011; Wang et al., 2009; Singh et al.,
a
Aeration facilitates mass transfer of oxygen from the atmosphere into the mixed liquor where the bacteria can degrade the SA with DO as electron acceptor. To account for any possible abiotic loss of SA via aeration (e.g. volatilization), an abiotic test was conducted and the result indicated that the loss was negligible (data not shown). No apparent SA removal was noted with the SA enriched culture when aeration was not provided (data not shown). This agrees with the general consensus that SA can only be biodegraded under aerobic conditions (Pereira et al., 2011; Kuhn and Suflita, 1989; RazoFlores et al., 1996). With the exposure to aerobic conditions a significant increase in SA removal was observed (Fig. 2a and b). At aeration rate of 0.13 L/min (which resulted in an average apparent DO concentration of 1.3 mg O2/L), SA degradation rate was 20.3 mg/L h. When the aeration rate was further increased to 0.31 L/min (average DO w 3 mg O2/L), the culture reached a maximal SA degradation rate of 43.5 mg/L h. A further increase in aeration rate (also resulting in a DO increase) could no longer increase the SA degradation rate. This suggests that beyond a DO of w3 mg O2/L, factor(s) other than oxygen contributes towards further increases in SA degradation rates. Hence, for a cost effective treatment of SA containing wastewaters using activated sludge, aeration should only be provided to maintain a DO of 3 mg O2/L.
3.3. Oxygen consumption by the enriched SA-degrading culture The kinetics of oxygen consumption by the enriched SAdegrading culture was evaluated to establish the relationship between specific OUR and initial SA concentration. Increase in SA concentration from 0 to 200 mg/L resulted in a linear increase in the specific OUR, and a maximum specific OUR (6 mg O2/g MLSS h) was obtained with an initial SA concentration of about 180 mg/L. Further increase in SA concentration did not increase the oxygen demand of the culture.
b
40
(mg L-1·h-1)
SA oxidation rate
50
Effect of aeration rate on SA degradation
30 20 10 0
0.5 1 1.5 Aeration rate (L min-1)
0
2 4 6 -1 DO (mg O2 L )
Fig. 2 e Effect of aeration rate on SA degradation by enriched activated sludge culture. (a) SA degradation rates at different aeration rates; (b) SA degradation rates at different DO concentrations. Notes: all experiments were conducted with the same MLSS concentration of 5800 mg/L, and SA concentrations were maintained between 310 and 340 mg/L.
149
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1
Specific OUR (mg O2 / g MLSS ·h)
6
4 Model Simulation Experimental Data
2
0
200
400 600 800 1000 SA (mg L-1)
Fig. 3 e The relationship between specific oxygen uptake rate (OUR) by the enriched activated sludge culture and initial SA concentrations (MLSS [ 7000 mg/L, day 130).
120) (Fig. 4). This indicates that the mixed culture had become more acclimatized for a more complete mineralization of SA. Since the microbial metabolism of SA degradation consists of both catabolism (i.e. oxidation of SA and its metabolites) and anabolism (cell growth), of the theoretical 1.29 mg O2/ mg SA the culture appeared to have spent only 60.5% for the complete SA degradation (Fig. 4, day 120). The y intercepts in Fig. 4 suggest that the background oxygen uptake rate due to both the endogenous respiration and the background nitrification by the mixed culture to be 1.31 0.36 mg O2/g MLSS h. Although the described OUR-based method has an advantage of its simplicity and could give a clear trend of an over-time improvement of the microbial activity in mineralizing SA, it only indirectly indicates the degree of mineralization of SA by the mixed culture.
3.4. Close to complete SA mineralization by the acclimatized mixed culture
3.3.1. Gradual increase in oxygen demand indicates enhanced SA mineralization The amount of oxygen consumed by the mixed microbial culture to oxidize a definite amount of SA can be obtained from the slope of a linear regression between oxygen uptake rate (OUR) and SA degradation rate (i.e. oxygen consumed (mg) per SA degraded (mg)). In principle, the maximal value of this slope is denoted by the theoretical COD value of SA (i.e. 1.29 mg O2/mg SA, without nitrification). In this study, the SA degradation rate increased over the 3 months of acclimatization (a 1.8-fold increase in the oxygen demand from day 30 to
To clearly verify the degree of SA mineralization by the mixed microbial culture, a detailed mass balance experiment was conducted taking the COD and sulphate concentrations into account over the course of SA degradation (Fig. 5). The result clearly indicated that the enriched microbial consortia could drive the overall SA oxidation close to a complete mineralization. The evidences are: 1. A high COD removal efficacy of 97.1% (note that the theoretical COD values of SA are close to the experimental COD values), indicating that the degradation intermediates of SA had also been oxidized by the mixed culture. 2. A near stoichiometric quantity of sulphate (1 mole of SO2 4 released per mole of SA oxidized) was produced in proportion to the amount of SA degraded (i.e. 3.2 mmol released from 3.3 mmol SA), corroborating with SO2 4 a similar observation as reported by Tan et al. (2005). The result also reveals that in contrast to other pureculture studies where SA degradation coincided with the
SA COD Actual COD Theoretical SO42NH4+-N 300 NO2--N NO3--N
150
Day 30
(1.29 mg O2 / mg SA) Theoretical Curve
Day 60 Day 120 100
50
y = 0.78x + 6.93 R2 = 0.98
SA / COD (mg/L)
Oxygen uptake rate (mg O2/L· h)
800
600
200
400
100
200
NH4+ / NO2- / NO3- / SO42Concentration (mg/L)
Instead, a gradual decline in the specific OUR was obtained, indicating substrate inhibition of the microbial culture by the SA (Fig. 3). The relationship was also obtained by using a modified Haldane model (Equation (2)), for which a satisfactory fit was obtained with the measured data (Fig. 3). By using a non-linear regression procedure, the values of k0 and m0 were found to be 0.00714 and 0.000846 L/h with a correlation factor of 0.91. A similar relationship between specific OUR and initial substrate concentration was reported with activated sludge that used o-cresol (a toxic aromatic organic pollutant) as the sole source of carbon and energy instead of SA (Maeda et al., 2005).
y = 0.51x + 9.91 (R2 = 0.91) y = 0.44x + 5.94 (R2 = 0.95) 0
0 25 50 75 100 125 SA degradedation rate (mg/L· h)
Fig. 4 e The relationship between oxygen consumed and SA degraded by the enriched activated sludge culture over time (MLSS [ 5800 mg/L, aeration rate [ 1.74 L/min).
2
4 Time (h)
6
0
Fig. 5 e Degradation of SA by the enriched activated sludge culture and release of sulphate, nitrite and nitrate over time (MLSS [ 5800 mg/L, aeration rate [ 1.74 L/min, day 130).
150
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1
Peak 4
O2 Consumed (mg)
300
200
y = 0.901x - 11.91 R² = 0.998 Peak 3
100 Peak 1 0
Peak 2 200 300 100 SA Degraded (mg)
Fig. 7 e The relationship between oxygen consumed and SA degraded by the enriched activated sludge culture obtained from the SA-spiking experiment in Fig. 6. The amounts of oxygen consumed were calculated from the oxygen consumption rate data in Fig. 6b.
Fig. 6 e (a) Effect of SA additions on dissolved oxygen (DO) concentration and concentrations of ammonium, sulphate, nitrite and nitrate in the SA limited culture (MLSS [ 5800 mg/L, aeration rate [ 0.31 L/min; arrows indicate additions of SA; Dotted line indicates saturation concentration of DO (7.7 mg O2/L)); (b) real-time oxygen uptake rate profile in the SA-degrading culture.
accumulation of both ammonium and sulphate in the medium (Wang et al., 2009; Singh et al., 2006), the ammonium released from the SA oxidation by the enriched mixed culture was predominately converted into nitrate (Fig. 5). This suggests that the culture contained an active population of ammonium-oxidizing bacteria (AOB). It is worth mentioning that as opposed to other studies where non-sulfonated aromatic amines such as aniline was found to have a much severe inhibitory effect on AOB (i.e. nitrification) even at a magnitude lower concentration (5 mg/L, 125 times lower), the presence of SA appeared to have no inhibitory effect on the AOB as the release of nitrate was in phase with the decline in SA, indicating that SA oxidation proceeded simultaneously with nitrification (Kumar et al., 1984; Than et al., 2002). Although exploring the impact of AOB on SA degradation is beyond the scope of this paper, further studies are currently underway to address this fundamental issue.
DO in a continuously aerated culture was investigated (Fig. 6a), and the changes in OUR during the relevant period of exposure is shown in Fig. 6b. OUR profile was closely linked to the SA concentration profile (Fig. 7). In the absence of SA (0 h), close to saturation concentration of DO (7.7 mg O2/L) was recorded (Fig. 6a). Incrementally dosing the SA into the culture (as denoted by the solid arrows) resulted in an immediate increase in the OUR (Fig. 6b). The more SA added (as denoted by the increased SA concentration), the higher the OUR and the larger the OUR peak area (Peaks 1e4, Fig. 6b). Integrating these OUR peaks with time (i.e. the peak area) gave the amount of oxygen that was consumed by the culture to degrade the spiked SA (Fig. 7). It is noteworthy that although Figs. 4 and 7 report a similar linear regression curve for oxygen consumption and SA degradation (0.44e0.78 and 0.90 mg O2/mg SA, respectively), the experimental approaches taken were completely different. Overall, these results suggest that for the treatment of wastewater primarily contaminated with SA, online DO measurement might be used to continuously monitor the contaminant removal process (i.e. SA biodegradation). It should be noted that the anaerobic treatment of a complex azo dye containing wastewater would generally lead to a complex mixture of sulfonated (or non-sulfonated) aromatic amines residues, which would collectively contribute to the COD content in the final effluent (O’Neill et al., 1999). As we have only tested SA as the model sulfonated aromatic amine, the effectiveness of the proposed DO-based monitoring approach for practical treatment of anaerobically treated azo dye effluent would certainly warrant further studies.
4. 3.5. In-situ DO monitoring as a real-time indicator of SA biodegradation Online DO monitoring is widely used in activated sludge wastewater treatment processes to reveal oxygen demand and to control aeration. To explore the possibility of using DO as a real-time indicator for SA degradation using the acclimatized activated sludge, the impact of SA concentration on
Conclusions
Due to the complexities of using pure or co-cultures to treat wastewaters in general, mixed microbial consortia are favoured even to treat SA-contaminated wastewaters. The novelty of this work is that it describes for a first time that SA biodegradation could be monitored through online measurement of oxygen consumption in a mixed culture system (here activated sludge).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 4 5 e1 5 1
Based on the results, we conclude the following: An activated sludge that has not been previously exposed to SA could be acclimatised under aerobic condition to oxidize SA close to a complete mineralization. The enriched activated sludge exhibited kinetics of both SA degradation and oxygen consumption that could be described by Haldane substrate inhibition model. In contrast to other pure or co-culture, the ammonium released from SA was almost completely converted into nitrate, indicating a co-enrichment of ammonium-oxidizing bacteria (AOB). SA had no apparent inhibitory effect on the AOB activity (nitrification). To what extent the AOB would affect the SA degradation is beyond the scope of this study and hence warrants further research. Oxygen consumption was directly proportional to SA degradation, and hence the continuous DO measurement enabled a real-time monitoring of SA degradation by the mixed culture.
Acknowledgements This work was funded by the CSIRO Water for a Healthy Country Flagship. CG was supported by a scholarship from the China Scholarship Council. We thank Dr. Trevor Bastow and Ms. Yasuko Geste (CSIRO Land and Water) for assistance in the ion chromatography. We are also grateful to Drs. Bradley Patterson and Carlos Descourvie´res (CSIRO Land and Water) for their valuable comments on this work.
references
APHA (American Public Health Association), 1995. In: Eaton, A., Clesceri, L., Greenberg, A. (Eds.), Standard Methods for the Examination of Water and Wastewater, 19th ed., pp. 5-15e517. Washington, DC. Carvalho, M.C., Pereira, C., Goncalves, I.C., Pinheiro, H.M., Santos, A.R., Lopes, A., Ferra, M.I., 2008. Assessment of the biodegradability of a monosulfonated azo dye and aromatic amines. Int. Biodeterior. Biodegrad. 62, 96e103. Cheng, K.Y., Ho, G., Cord-Ruwisch, R., 2010. Anodophilic biofilm catalyses cathodic oxygen reduction. Environ. Sci. Technol. 44, 518e525. Chung, K.T., Cerniglia, C.E., 1992. Mutagenicity of azo dyes: structureeactivity relationships. Mutat. Res. 277, 201e220. Feigel, B.J., Knackmuss, H.J., 1988. Bacterial catabolism of sulfanilic acid via catecho-4-sulfonic. FEMS Microbiol. Lett. 55, 113e118. Feigel, B.J., Knackmuss, H.J., 1993. Syntrophic interactions during degradation of 4-aminobenzenesulfonic acid by a two species bacterial culture. Arch. Microbiol. 159, 124e130.
151
Gan, H.M., Shahir, S., Ibrahim, Z., Yahya, A., 2011. Biodegradation of 4-aminobenzenesulfonate by Ralstonia sp. PBA and Hydrogenophaga sp. PBC isolated from textile wastewater treatment plant. Chemosphere 82 (4), 507e513. Garcia-Ochoa, F., Gomez, E., Santos, V.E., Merchuk, J.C., 2010. Oxygen uptake rate in microbial process: an overview. Biochem. Eng. J. 49, 289e307. Hwang, S.Y., Berges, D.A., Taggart, J.J., Gilvarg, C., 1989. Portage transport of sulfanilamide and sulfanilic acid. J. Med. Chem. 32, 694e698. Kuhn, E.P., Suflita, J.M., 1989. Anaerobic biodegradation of nitrogen-substituted and sulfonated benzene aquifer contaminants. Hazard. Waste Hazard. Mater. 6, 121e134. Kumar, N.J., Krishnamoorthi, K.P., Swaminthan, T., 1984. Studies on nitrification of aniline with acclimated activated sludge. Biotechnol. Bioeng. 26, 197e202. Maeda, M., Itoh, A., Kawase, Y., 2005. Kinetics for aerobic biological treatment of o-cresol containing wastewaters in a slurry bioreactor: biodegradation by utilizing waste activated sludge. Biochem. Eng. J. 22, 97e103. Oh, S.W., Kang, M.N., Cho, C.W., Lee, M.W., 1997. Detection of carcinogenic amines from dyestuffs or dyed substrate. Dyes Pigm. 33, 119e135. O’Neill, C., Hawkes, F.R., Lourenco, N.D., Pinheiro, H.M., Delee, W., 1999. Colour in textile effluents-sources, measurements, discharge, contents and simulation: a review. J. Chem. Technol. Biotechnol. 74, 1009e1018. Perei, K., Rakhely, G., Kiss, I., Polyak, B., Kovacs, K.L., 2001. Biodegradation of sulfanilic acid by Pseudomonas paucimobilis. Appl. Microbiol. Biotechnol. 55, 101e107. Pereira, R., Pereira, L., van der Zee, F.R., Alves, M.M., 2011. Fate of aniline and sulfanilic acid in UASB bioreactors under denitrifying conditions. Water Res. 45, 191e200. Razo-Flores, E., Donlon, B., Field, J., Lettinga, G., 1996. Biodegradability of N-substituted aromatics and alkylphenols under methanogenic conditions using granular sludge. Water Sci. Technol. 33, 47e57. Singh, P., Birkeland, N.K., Iyengar, L., Gurunath, R., 2006. Mineralization of 4-aminobenzenesulfonate (4-ABS) by Agrobacterium sp. strain PNS-1. Biodegradation 17, 495e520. Singh, P., Mishra, L.C., Iyengar, L., 2004. Biodegradation of 4aminobenzensulfonate by a newly isolated bacterial strain PNS-1. World J. Microbiol. Biotechnol. 20, 845e849. Tan, N.C.G., van Leeuwen, A., van Voorthuinzen, E.M., Slenders, P., Prenafeta-Boldu, F.X., Temmink, H., Lettinga, G., Field, J.A., 2005. Fate and biodegradability of sulfonated aromatic amines. Biodegradation 16, 527e537. Tan, N.C.G., Field, J.A., 2005. Biodegradation of sulfonated aromatic compounds. In: Environmental Technologies to Treat Sulfur Pollution Principles and Engineering. IWA Publishing, London, pp. 377e392. Than, K., Gheewala, S.H., Annachhatre, A.P., 2002. Modeling of nitrification inhibition with aniline in suspended-growth process. Water Environ. Res. 6, 531e540. Topac, F.Q., Dindar, E., Ucaroglu, S., Baskaya, H.S., 2009. Effect of a sulfonated azo dye and sulfanilic acid on nitrogen transformation. J. Hazard. Mater. 170, 1006e1013. Wang, Y.Q., Zhang, J.S., Zhou, J.T., Zhang, Z.P., 2009. Biodegradation of 4-aminobenzenesulfonate by a novel Pannonibacter sp. W1 isolated from activated sludge. J. Hazard. Mater. 169, 1163e1167.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
TiO2 and Fe (III) photocatalytic ozonation processes of a mixture of emergent contaminants of water Eva M. Rodrı´guez, Guadalupe Ferna´ndez, Pedro M. Alvarez, Fernando J. Beltra´n* Departamento de Ingenierı´a Quı´mica y Quı´mica Fı´sica, Universidad de Extremadura, Avenida de Elvas S/N, 06006 Badajoz, Spain
article info
abstract
Article history:
A mixture of three emergent contaminants: testosterone (TST), bisphenol A (BPA) and
Received 29 July 2011
acetaminophen (AAP) has been treated with different photocatalytic oxidation systems.
Received in revised form
Homogeneous catalysts as Fe(III) alone or complexed with oxalate or citrate ions, hetero-
14 October 2011
geneous catalysts as titania, and oxidants such as hydrogen peroxide and/or ozone have
Accepted 18 October 2011
been used to constitute the oxidation systems. For the radiation type, black light lamps
Available online 28 October 2011
mainly emitting at 365 nm have been used. The effects of pH (3 and 6.5) have been investigated due to the importance of this variable both in ozone and Fe(III) systems.
Keywords:
Removal of initial compounds and mineralization (total organic carbon: TOC) were fol-
Titania
lowed among other parameters. For the initial compounds removal ozonation alone, in
Iron
many cases, allows the highest elimination rates, regardless of the presence or absence of
Ozone
UVA light and catalyst. For mineralization, however, ozone photocatalytic processes
Testosterone
clearly leads to the highest oxidation rates. ª 2011 Elsevier Ltd. All rights reserved.
Bisphenol A Acetominophen Photocatalytic oxidation Water treatment
1.
Introduction
Nowadays it is well established that advanced oxidation processes are recommended technologies for the removal of the so called emergent pollutants of the water. These compounds are mainly those from pharmaceutical origin (antibiotics, analgesics, etc) or dedicated to personal care but also they belong to other families such as plasticizers, pesticides, phenols, etc (Kolpin et al., 2002; Daughton and Ternes, 2000). Many of these compounds show a potential disrupting character for the endocrine system of living beings (Escher et al., 2011) and are frequently found in influents and, also, effluents of wastewater treatment plants that usually do not apply advanced chemical oxidation processes (Ternes, 1998; Nelson et al., 2011).
Photocatalytic oxidation processes (POP) are well known advanced oxidation where hydroxyl radicals are formed from the synergic effects of radiation, a catalyst and an oxidant. There are two types of POP depending on the nature of the catalyst. The homogeneous POP where metal ions as catalysts are used such as in the photo-Fenton process (Zepp et al., 1992) and heterogeneous POP where metal oxides play the role of semiconductors such as titania (Legrini et al., 1993; Bhatkhande et al., 2001). Another feature of POP is the nature of the oxidant used. In homogeneous POP hydrogen peroxide is a very used oxidant while in heterogeneous POP oxygen is the classical oxidant. The main role of these oxidants is rather different depending on the POP type. Thus, hydrogen peroxide in homogeneous POP directly acts to produce hydroxyl radicals by reacting with the catalyst, for
* Corresponding author. Tel.: þ34 924289387; fax: þ34 924289385. E-mail address:
[email protected] (F.J. Beltra´n). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.038
153
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
example Fe(II) in the Fenton processes (Fenton, 1876). In heterogeneous POP the oxidant, mainly oxygen, acts to capture electrons arriving to the conduction band of the semiconductor to minimize electron-hole recombination (Turchi and Ollis, 1990). Radiation type is also another variable in these processes with UVC radiation as the classical one especially in heterogeneous POP (Matthews and McEvoy, 1992). However new radiation sources are investigated to have POP implying radiation closer to the visible spectrum of light. In this sense, black light lamps that emit radiation in the 350e390 nm range with a maximum at 365 nm are of interest since titania semiconductor can be excited with radiation energy of up to 387 nm (Bhatkhande et al., 2001). Another oxidant of high interest in POP is ozone that is recently investigated in a process called photocatalytic ozonation. Due to its higher oxidizing character and reactivity ozone can improve the formation of hydroxyl radicals in a POP throughout several mechanisms (Agustina et al., 2005). In this work, some POP (homogeneous and heterogeneous) have been investigated to remove a mixture of emergent water pollutants and the subsequent remained organic carbon (TOC). Testosterone (TST), a steroid hormone from the androgen group which is prescribed at sexual functional disorders, vascular disorders and for therapy of tumors including cancer tumors, acetominophen (AAP), very used to reduce pain and fever but can cause serious liver and gastrointestinal side effects, and bisphenol A (BPA), used primarily to make polycarbonate plastic and epoxy resins, have been chosen as model compounds because of their frequent presence in water with different organic content (Kim et al., 2007; Stackelberg et al., 2007). Although these pollutants are present in real water at concentrations below some tenths of mg L1, in this work, however, concentrations of about 1e3 mg L1 have been applied to assure accurate measurements of concentrations, follow the TOC and check possible synergic effects between oxidants, light and catalysts to make predictions about possible mechanisms of photocatalytic ozonation reactions. In any case and as result of oxidation processes applied, concentrations of compounds studied in the order of hundreds of mg L1 (that is, just one order of magnitude higher than in real water) were followed. Also, some estimation about the importance of oxidation ways (direct ozonation and hydroxyl radical oxidation) is also made for mg L1 concentrations of pollutants.
2.
Experimental
2.1.
Products, experimental set-up and procedure
BPA, TST, AAP (see Fig. 1 for molecular structures) and citric acid were obtained from Aldrich (Spain), oxalic and perchloric acids from Merck (Spain) and powdered P25 TiO2 was directly obtained from the manufacturer, Degussa AG (Germany). Other chemicals used were at least reagent grade and used as received. Two 15 W black light lamps (HQ Power Lamp15TBL) emitting mainly 365 nm radiation were used. Photocatalytic oxidation experiments were carried out in a 4 L cylindrical borosilicate glass reactor that was provided with magnetic agitation, air feeding system and devices for
CH3 HO
OH CH3
BPA NH O
HO
AAP OH
O
TST Fig. 1 e Moleular structures of pharmaceuticals studied.
temperature and pH measurements. Two black lamps were installed on opposing walls outside the reactor and the overall system was placed in a closed box to avoid the disturbing effect of direct sunlight irradiation. To start the photocatalytic experiments, 3 L of a buffered BPA, AAP and TST aqueous solution, about 105 M/each (1.5e3 mg L1/each), at pH 3 or pH 6.5 (perchloric/perchlorate ionic strength 0.03 M) were fed to the reactor, air bubbling was set at 40 L h1 and the lamps were turned on. About 20 min later, time enough to allow the lamps work steadily, a known volume of catalyst or free radical promoting agent was added: aqueous Fe(III) solution, carboxylic acid, hydrogen peroxide and/or TiO2. More details of this photoreactor and the procedure followed to complete photocatalytic experiments can be seen in a previous work (Rodrı´guez et al., 2009a). Ozonation and photocatalytic ozonation experiments were performed in the same reactor supplied in these cases with inlets for gas feeding (through a diffuser), sampling, catalyst addition and temperature measurement and outlet for the non absorbing gas (see Figure 1SI of supplementary information). A Sander Labor ozone generator was used. Gas flow rate was always 36 L h1. In photocatalytic ozonation experiments, after the time needed for the lamp emission to reach stationary conditions has elapsed, catalysts were added and the ozone-oxygen gas was fed. In all cases, at regular time intervals, samples were withdrawn from the reactor to be analyzed. Experimental conditions applied were as follows: CBPA0 ¼ CTST0 ¼ CAAP0 ¼ 105 M (1.5e3 mg L1); CTP0 as BPA ¼ 1.6 105 M (3.6 mg L1); TiO2 dose: 0.1 g L1; CFe(III) 5 M (2.8 mg L11) (as ferric perchlorate); 0 ¼ 5x10 4 CH2O20 ¼ 5 10 M (17 mg L1); COxalic0 ¼ 4 104 M (36 mg L1);
154
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
Ccitric0 ¼ 4 104 M (77 mg L1); T ¼ 22e25 C; CO3g0 ¼ 2.5 104 M (12 mg L1) (except in some cases with 8.3x105 M (4 mg L1).
Chemical analysis
TST, AAP and BPA were analyzed by high-performance liquid chromatography with an HPLC-UV (Agilent 1100) system in a 15 cm long, 0.4 cm i.d. Kromasil C18 column with acetonitrile-water with 0.1% phosphoric acid as mobile phase (15/85 (v/v) for BPA and AAP, and 50/50(v/v) for TST), with a flow rate of 1 mL min1. Detection was made at 280, 220 and 244 nm for BPA, TST and AAP, respectively. Total iron concentration was determined by the ferrozine method (Stookey, 1970), Fe(II) concentration by the phenantroline method (Zuo, 1995), total polyphenol content (TP) by the Folin Ciocalteau method (Singleton and Rossi, 1965) and hydrogen peroxide by the method of Eisenberg (1943), following for all the determinations the procedure detailed in a previous work (Rodrı´guez et al., 2011). Bader and Hoigne´ (1981) method, based on the decoloration of the 5,5,7 indigotrisulphonate, was used to measure the dissolved ozone concentration. Ozone in the gas phase was monitored by means of an Anseros Ozomat ozone analyzer, the analysis based on the absorbance at 254 nm. The intensity of UVA radiation coming from the black light lamps into the aqueous solution was obtained by ferrioxalate actinometry (Hatchard and Parker, 1956). A photon flow of 7.32 107 E s1 was determined with both lamps simultaneously working. Total organic carbon was measured with a TOC-VSCH Shimadzu analyzer. Samples for HPLC and TOC analyses were mixed with Na2S2O3 to ensure the destruction of any oxidizing agent remaining in solution. In ozonation systems, remaining dissolved ozone in samples was eliminated by helium stripping except in cases for analyzing the dissolved ozone concentration. When needed samples were filtered before analysis (MachereyeNagel PET 0.45 mm filters).
3.
3.1.
3.1.1.
Experiments at pH 3
Fig. 2 presents, as example, the evolution of TST remaining dimensionless concentration with time for experiments carried out with ozone alone, ozone/Fe(III) in the dark and POP studied here: O2/Fe/III/UVA, O3/Fe(III)/UVA, O2/Fe(III)/Oxalate/ UVA, O3/Fe(III)/oxalate/UVA, O2/Fe(III)/H2O2/UVA, O3/Fe(III)/ H2O2/UVA and O3/H2O2/UVA. A blank experiment of ozone alone (not shown) did not lead to any difference from that of ozone/UVA radiation thus confirming the absence of ozone photolysis when radiations in the 350e390 nm range are applied. It can be seen from Fig. 2 that Fe(III) photocatalytic oxidation leads to partial removal of TST, about 63%, in 2 h reaction. Approximately, 40% removal was also reached for BPA and AAP at these conditions (see Figures 2SI and 3SI in the supplementary part). These results are due to hydroxyl free radical reactions, the free radicals coming from the photolysis of Fe(III) that at the pH of work is forming the species Fe(OH)2þ (Faust and Hoigne´, 1990). When oxalate is added, the
1.0
0.8
Results and discussion
Water POP can be classified as homogeneous or heterogeneous processes according to the liquid or solid phase nature of the catalyst used. For homogeneous and heterogeneous POP studied here, the systems O2/Fe/III/UVA and O2/TiO2/UVA can be considered as the basic ones around which any other POP involving changes of oxidant and catalyst type can be compared. Homogeneous and heterogeneous POP are studied here in this order changing the type of catalyst and/or oxidant (oxygen, hydrogen peroxide and ozone). However, for comparative reasons a number of blank experiments were also made involving ozone alone or combined with a catalyst. POP studied have been investigated at two pH values: 3 and 6.5. Although water at pH 3 could represent an unrealistic situation, POP at acidic conditions were investigated since these are the optimum ones for Fenton-type processes. Furthermore, efficiency improvement of advanced oxidation systems can be observed when hydroxyl radical scavengers such as carbonate/bicarbonate ions are removed from water. This step could be particularly convenient when a Fenton POP is immediately after applied. In a following step, addition of some
Homogeneous POP
Combinations of Fe(III) or complexed Fe(III) (with oxalate or citrate) with hydrogen peroxide or ozone were the homogeneous POP investigated.
0.6
CTST/CTST0
2.2.
alkali to raise the pH could not necessarily be an expensive step given the possible benefits obtained in the Fenton oxidation process. In any case, regardless of pH, UVA radiation experiments were firstly carried out with no photolysis results. This was expected since the compounds tested do not absorb radiation in the 350e390 nm interval (black light lamp emission range). Also, for heterogeneous experiments, TST, AAP and BPA did not show any appreciable adsorption on TiO2.
0.4
0.2
0.0 0
20
40
60
80
100
120
t, min
Fig. 2 e Evolution of dimensionless remaining concentration of testosterone with time during homogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. Symbols and systems: , O3/UVA; C Fe(III)/UVA; B O3/Fe(III)/UVA; > O3/Fe(III)/UVA*; : Fe(III)/Oxal/UVA; 6 O3/ Fe(III)/Oxal/UVA; ; Photo-Fenton; 7 O3/Photo-Fenton; ✯ O3/H2O2/UVA. See Section 2.1 for experimental conditions except when noted. *Ozone gas concentration: 4 mg LL1.
155
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
ferrioxalate complex is formed. This complex photolytically decomposes faster than Fe(OH)2þ producing higher concentration of hydroxyl free radicals (Hatchard and Parker, 1956; Safarzadeh-Amiri et al., 1997). Then, it is not surprising the increase of oxidation rate observed with this system compared to the previous one. Also, for the photo-Fenton system, addition of hydrogen peroxide to the Fe(III) photocatalytic system, the reaction rate also increases compared to the latter one and yields approximately similar conversions of the compounds studied. Nonetheless, comparison of this system with the H2O2-free photocatalytic oxidation is a rather difficult task since the oxidation rate is dependent on the H2O2 concentration effect that has not been investigated. Regarding the presence of ozone, it has to be noted that ozone alone, ozone/ UVA and ozone/Fe(III) processes lead to similar experimental results (only O3/UVA results are shown) since black light does not decomposes ozone and ozone-Fe(III) is an inert system regarding hydroxyl radical formation from this catalystoxidant combination. Very different results were obtained, however, when Fe(III) and black light were simultaneously applied. As observed from Fig. 2 TST conversions achieved with the ozone-free systems are much lower than those obtained when ozone processes were applied (similar results were observed in the case of BPA and AAP, see Figures 2SI and 3SI). For example, for the ozone alone process, after 20 min complete disappearance of the organics studied was noticed. This result was also similar to those of other ozone processes, regardless of the type of accompanying agent (UVA light, Fe(III), ferrioxalate, etc). As indicated above, intermediate conversion results, as far as the organic removal rates are concerned, were obtained when iron was accompanied with oxalic acid to form ferrioxalate complex or hydrogen peroxide. Thus, in between 60 and 80 min reaction, TST totally disappeared while 93 and 80e90% BPA and AAP removal, respectively, were observed in these POP at the same reaction times. In these systems hydroxyl radical oxidation is the only way of compound removal. Here, the photo-Fenton process results in the higher oxidation rates likely due to the continuous formation of radicals through the redox system: pH3
2þ
hn
FeðIIIÞ þ H2 O!FeðOHÞ ! hnFeðIIÞ þ HO
(1)
FeðIIÞ þ H2 O2 !FeðIIIÞ þ OH þ HO
(2)
Thus, the synergic effect of this system is evident.
The fact that differences between ozone processes are negligible is likely due to the importance of the direct ozone reactions compared to the contribution of hydroxyl radicals. The kinetic regime of these ozone reactions can be established from their corresponding Hatta values determined with Equation (3): Ha ¼
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi kD DO3 CM kL
(3)
where kD, kL, DO3 are the rate constant of the direct reaction between ozone and compound M, the individual mass transfer coefficient and ozone diffusivity in water, respectively, while CM is the concentration of the organic compound. Values of these rate constants can be obtained from literature (Andreozzi et al., 2003; Deborde et al., 2005). Table 1 shows the Ha results for the ozone reactions studied. As can be seen Ha is always lower than 0.3 which means the ozone reactions are slow. At these conditions and according to film theory (Beltra´n, 2004) dissolved ozone diffuses throughout the water film close to the gaseliquid interface and reaches the bulk water where it can react with target compounds (such as TST, BPA and AAP) and/or decompose, through different ways, in hydroxyl radicals. Thus, compounds can also be simultaneously consumed through reactions with hydroxyl radicals. If ozone target compounds direct reactions would have been fast (Ha>3) ozone would only be consumed through these reactions in the proximity of the gaseliquid interface. In this situation dissolved ozone would not reach the bulk water where ozone decomposition reactions to yield free radicals take place (Beltra´n, 2004). The fact that ozone can reach the bulk water, then, explains why in the different ozone processes the organic reaction rates are not exactly the same but differences, although low, are also observed (Beltra´n, 2004). These differences increase with the decrease of the direct reaction rate constant, which supports previous comments. In the case of TST, the direct ozone reaction rate constant is not known but Barron et al. (2006) and Brose´us et al. (2009) have reported for the ozone-progesterone reaction values of 480 30 M1 s1 in the 2e8 pH range and 601 9 M1 s1 at pH 8.1, respectively. Since both molecules, testosterone and progesterone, have similar molecular structure and nucleophilic points where ozone can attack, specifically their double carbon bond, it is expected a similar value for the rate constant of the ozone-TST reaction. The rate
Table 1 e Values of the rate constant of the direct reaction between ozone and compounds studied and corresponding Hatta numbers at different pH and reaction time.a Reaction
pH 3 kD, M1 s1
O3 þ BPA O3 þ TST O3 þ AAP a b c d
1.71 10 4.80 102 1.84 103
c d
0.298 0.051 0.097
t ¼ 5 min 0.196 0.037 0.044
pH 6.5 kD, M1 s1
Hatta values t ¼ 0 min
4 b
Reaction
t ¼ 10 min 0.083 0.027 0.025
O3 þ BPA O3 þ TST O3 þ AAP
6 b
1.08 10 4.80 102 1.37 106
c d
Hatta values t ¼ 0 min
t ¼ 5 min
t ¼ 10 min
2.296 0.051 2.613
1.241 0.034 1.388
0.459 0.02
Results correspond to ozonation alone experiments (similar results are obtained from data of other ozone processes). Deborde et al. (2005). Andreozzi et al., 2003. Value of the ozone-progesterone reaction according to Barron et al. (2006).
e
156
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
FeðIIÞ þ O3 /FeO2þ þ O2 FeO2þ /FeðIIIÞ þ HO þ HO
(4)
2FeO2þ /2FeðIIIÞ þ HO þ HO 2
(5) (6)
In reaction (4), Fe(II) formed from Fe(III) photolysis shown in reaction (1) reacts with ozone to yields FeO2þ that eventually gives hydroxyl radicals and hydrogen peroxide in the ionic form through reactions (5) and (6) (Logager et al., 1992). Also, Fe(III) will be again photoreduced through reaction (1) to yield more hydroxyl radicals. At pH acid, the ionic form of hydrogen peroxide will pass to its molecular form H2O2 that can react with Fe(II) in another Fenton process to yield more hydroxyl radicals. In addition, formation of hydrogen peroxide from direct ozone reactions gives rise to another photo-Fenton process. Other possible reactions such as the ozonehydrogen peroxide reaction to yield free radicals are negligible at pH 3 (Staehelin and Hoigne´, 1982). Then, development of these reactions confirm the synergic effect between ozonation and photocatalytic oxidation. In the cases studied here both O3/Fe(III)/UVA and O3/photo-Fenton systems show clearly this synergism. However, the results also show that this synergic effect is highly dependent on the relative importance between the rates of ozone-organic direct reactions and ozone decomposition reactions in hydroxyl radicals. Given the molecular structure of organics studied, polyphenol compounds are possible first intermediates of oxidation (Decoret et al., 1984). Then, total concentrations of these compounds were also followed with time as shown in Fig. 3. As it can be seen, results are significantly different depending on the presence or absence of ozone. Thus, in ozone-free oxidation, polyphenol compound concentration clearly increases with time and, in some cases, after reaching
2.4x10
2.0x10
1.6x10
CTP as BPA, M
constant value is much lower, however, than the ones corresponding to the ozone-BPA and ozone-AAP reactions. Because of TST-ozone reaction is the slowest one among those studied here, TST removal rate slightly increases when ozone and Fe(III)/UVA or ozone/photo-Fenton processes are simultaneously applied compared to the single ozonation process. This is undoubtedly due to the contribution of the reaction with hydroxyl radicals coming from the different mechanism in these POP. However, in the case of BPA and AAP differences among ozone processes in organic removal rates are even much lower, especially, in the case of BPA that result to be negligible. Since BPA and AAP are dissociating organic compounds the apparent rate constants of their direct reactions with ozone are pH dependent (Hoigne´ and Bader, 1983). At pH 3 these values are 17140 and 1840 M1 s1 for BPA and AAP, respectively (Andreozzi et al., 2003; Deborde et al., 2005). These results justify the higher reactivity of these compounds with ozone compared to TST. In Fig. 2 the effect of ozone concentration on the ozone/ UVA/Fe(III) process can also be observed at two concentration levels: 2.5x104 and 8.3x105 M (12 and 4 mg L1). It is clearly seen that the increasing concentration of ozone accelerates the organic removal rate which supports the action of ozone direct reaction but also the possible participation of ozone to increase hydroxyl radical formation through reaction (4)e(6):
1.2x10
8.0x10
4.0x10
0.0 0
20
40
60
80
100
120
t, min
Fig. 3 e Evolution of total polyphenol concentration measured as BPA with time during homogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. See Fig. 2 for symbols and experimental conditions.
a maximum value, it decreases. In these latter cases, it is clear that a more intensive oxidation process is being applied. Thus, the simple Fe(III) photocatalytic process does not lead to any decrease of these compounds (after 2 h, formation of polyphenols still continues) while Fe(III)/Oxalate/UVA system seems to be a stronger oxidizing process since, in this case, polyphenols reach a maximum concentration after 60 min and then slowly decrease. The most important process among ozone-free systems is the photo-Fenton system. Here, maximum polyphenol concentration is reached in 30 min and then the concentration slows down to 8 106 M as BPA (1.83 mg L1) in 2 h reaction. When ozone is applied, regardless of the ozonation system, polyphenol concentration diminishes since the beginning of the process to completely disappear in just 80 min reaction. These results are also expected because of the direct reactions between ozone and phenols are of the order of 103e104 M1 s1 at acid pH (Hoigne´ and Bader, 1983). The importance of ozone gas concentration applied is also seen in Fig. 3 since ozonation with 8.3 105 M (4 mg L1) retards complete polyphenol conversion up to 2 h reaction. Since Fe(III) is applied to photolytically decompose and yield Fe(II) and free radicals, the system becomes a redox one especially when ozone or hydrogen peroxide is also present. Then, it is interesting to follow the Fe(II) concentration with time. This can be seen in Fig. 4. As observed, in ozone-free systems, Fe(II) concentration increases with time specially in Fe(III) and Fe(III)/oxalate photocatalytic processes where Fe(II) is accumulated in water. When hydrogen peroxide is also present, there is a photo-Fenton process, Fe(II) concentration slowly increases and keeps a low stationary value after 60 min. This value 5 106 M (0.28 mg L1), in any case, is much lower than the one Fe(II) reaches after 2 h with the other two ozone-free systems (about 4 105 M (2.2 mg L1)). For the same reasons, in ozone systems, the trend Fe(II) concentration follows is similar to the one observed for the photo-Fenton process.
157
5x10
1.0
4x10
0.8
3x10
0.6
CTOC/CTOC0
CFe(II), M
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
2x10
1x10
0.4
0.2
0
0.0 0
20
40
60
80
100
120
0
20
40
t, min
Fig. 4 e Evolution of Fe(II) concentration with time during homogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. See Fig. 2 for symbols and experimental conditions.
The reactivity of ozone oxidation systems can also be deduced from the evolution of the dissolved ozone concentration (details are shown in Figure 4SI in supplementary data). Thus, the O3/Fe(III)/UVA system was observed as the most reactive one regarding ozone consumption since the concentration of ozone is close to the detection limit of the method applied (about 50 mg L1) for more than 1 h. Then, concentration starts to increase. Notice that polyphenol concentration starts to be negligible after 60 min reaction time (see Fig. 3). In the rest of ozone systems, dissolved ozone concentration follows a similar trend, low or null concentration during the first minutes and then an increase of concentration (for more details see Figure 4SI). These trends are likely due to the contribution of direct ozone reactions during the beginning of the process where BPA, AAP and TST were present and then to the decomposition reactions of ozone in free radicals that start to be important after fast ozone reacting compounds have disappeared. For example, in the ozonation alone system, the concentration of dissolved ozone was negligible during the first 20 min (notice that BPA, AAP and TST needed about 20 min for total conversion). After this time, the concentration of ozone started to rapidly increase to reach a stationary value after 50 min reaction (the time needed to remove polyphenols) (See also Figure 4SI for more details). Therefore, the low values of dissolved ozone concentration observed with the ozone photocatalytic systems are undoubtedly due to the higher ozone consumption in the mechanism of free radical reactions and Fe(II)ozone reaction. Finally, in Fig. 5 the evolution of total organic carbon (TOC) with time corresponding to the experiments commented is shown. Regarding TOC, Fe(III) photocatalytic oxidation gives no mineralization at all while ozonation alone only leads to 15% TOC removal after 2 h reaction. When some agent is added to this system (i.e. hydrogen peroxide or oxalate to have a photo-Fenton or Fe(III)/oxalate/UVA systems, respectively) or ozone combined with other agent is applied, TOC clearly
60
80
100
120
t, min
Fig. 5 e Evolution of dimensionless remaining TOC with time during homogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. See Fig. 2 for symbols and experimental conditions.
diminishes though with different rates depending on the oxidation system applied. The most efficient system is the O3/ Fe(III)/UVA oxidation that allows about 80% TOC conversion in 2 h. Also, the ferrioxalate systems lead to similar TOC conversions but in these cases TOC contribution from oxalate have to be considered. Given the fact that after 20e40 min dissolved ozone is already appreciable in water and that main ozone reacting compounds have disappeared (polyphenols) TOC consumption are undoubtedly due to hydroxyl free radical reactions. TOC results show in Fig. 5, then, clearly confirm a high synergic effect between ozonation and photocatalytic oxidation (Fe(III)/UVA) since none of these processes individually applied allow TOC conversions higher than 15% after 2 h reaction.
3.1.2.
Experiments at pH 6.5
A series of experiments of homogeneous POP were also made at pH 6.5. Since Fe(III) is unstable at this pH, experiments were carried out in the presence of oxalate or citrate to form the corresponding ferricarboxylate complexes. These complexes photolytically decompose to yield free radicals (Faust and Zepp, 1993; Zhang, 2000; Rodrı´guez et al., 2009a). In fact, ferrioxalate was already used in the experiments at pH 3 for comparative reasons. At pH 6.5, ferricitrate complex is also used because of its higher stability compared to ferrioxalate at ´ lvarez et al., 2010; these conditions (Nansheng et al., 1998; A Rodrı´guez et al., 2009b). Also, as at pH 3, some ozonation experiments were carried out. Thus, the following oxidation systems were studied at pH 6.5: Simple ozonation, O3/UVA, O3/Fe(III)/oxalate/UVA, O2/Fe(III)/ O2/Fe(III)/Oxalate/UVA, Citrate/UVA, O3/Fe(III)/Citrate/UVA and O3/H2O2/UVA. The evolution of BPA remaining dimensionless concentrations with time corresponding to the experiments at pH 6.5 is shown in Fig. 6 as an example (information of the changes observed in the concentrations of the other two model
158
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
1.0
0.8
CBPA/CBPA0
0.6
0.4
0.2
0.0 0
20
40
60
80
100
120
t, min
Fig. 6 e Evolution of dimensionless remaining concentration of BPA with time during homogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 6.5. Symbols and systems: , O3/UVA; C Fe(III)/Oxal/UVA; B O3/Fe(III)/Oxal/UVA; : Fe(III)/Cit/UVA; 6 O3/Fe(III)/Cit/UVA; ✯ O3/H2O2/UVA. See Section 2.1 for experimental conditions.
compounds, TST and AAP, is shown in Figures 5SI and 6SI of supplementary data, respectively). As can be seen from these figures, Ferrioxalate/UVA system only allows compound conversions between 20 and 25% in 2 h reaction. In fact, the reaction is stopped after about 20 min likely due to the unstability of the complex at pH 6.5. On the contrary, the higher stability of ferricitrate complex allows its application to reach between 90 and 95% compound conversion in 2 h. However, the best POP are those involving ozone. Thus, ozone processes lead to total conversion of initial compounds in between 15 and 20 min with reaction rates faster than those observed for the same compounds in experiments at pH 3 (see Figures 2, 2SI and 3SI for details). In this case, differences between ozone processes are even lower than those observed at pH 3 including the case of TST. The main reason of the increase of oxidation rates can be attributed to different mechanisms depending on the compound tested. Thus, BPA and AAP react with ozone at pH 6.5 faster than at pH 3 since both are dissociating compounds. For example, the rate constants of the direct reactions between ozone and the nondissociating and first dissociating forms of BPA are, according to Deborde et al. (2005), 1.68x104 and 1.06 109 M1 s1, respectively, the pKa being 9.6. With this information the apparent second order rate constant of the ozone-BPA reaction at pH 6.5 can easily be determined as 1.08 106 M1 s1. In the case of AAP (pKa ¼ 9.36), according to data of Andreozzi et al. (2003), similar calculations lead to a value of 1.37 106 M1 s1 at pH 6.5. In the case of TST, a nondissociating compound in water, the rate constant remains as at pH 3. Then, Hatta numbers of the ozone direct reactions are calculated with Equation (3) to yield values between 2.6 and 0.5 for ozone-BPA and ozone-AAP reactions and lower than 0.3 for the ozone-TST reaction during the first 10 min of
reaction (see Table 1). Then, according to what is commented in Section 3.3.1 it is reasonable to admit that both BPA and AAP are exclusively consumed through their direct reactions with ozone during this initial period (Beltra´n, 2004). Regarding TST, however, the ozone direct reaction is slow and competes with the ozone decomposition reactions to yield free radicals as at pH 3. Furthermore, the main route of reaction rate for TST is likely the hydroxyl radical oxidation since it is well known that ozone decomposition in free radicals is triggered with the increasing pH if direct reactions are not fast (Beltra´n, 2004) as it is this case. For example, one possible route is the reaction between ozone and the ionic form of hydrogen peroxide that at pH 6.5, even at very low concentrations, decomposes ozone at a high rate (the rate constant of the ozone-hydroperoxide ion reaction is 2.8 106 M1 s1 as Staehelin and Hoigne´ (1982) reported). This also explains the high rates observed with the O3/H2O2/UVA system but also in any of ozone involving oxidation since the ozone reactions with the organic present yields hydrogen peroxide through cycloaddition reactions to the aromatic ring (Mvula and von Sonntag, 2003). Although concentration of H2O2 was not followed in ozone experiments, but in the O3/H2O2/UVA system, it is well known the formation of hydrogen peroxide in ozonation processes (Mvula and von Sonntag, 2003; Leitzke and von Sonntag, 2009). In fact, in ozone processes, hydrogen peroxide formed in direct ozone reactions can lead to hydroxyl radical formation by reacting with ozone: M þ O3 /P þ H2 O2 pK¼9:3
(7)
þ H2 O2 %HO 2 þH
(8)
HO 2 þ O3 /HO2 þ O3
(9)
pK¼4:8
þ HO2 %O 2 þH
(10)
O 2 þ O3 /O3 þ O2
(11)
þ O 3 þ H /HO þ O2
(12)
Also, hydroxyl radicals are generated in a modified photoFenton process formed with ozone, Fe(III) and UVA through reactions (1), (7) and (2). The higher reactivity of ozone systems has also been checked by following the total polyphenol concentration with time (not shown). At pH 6.5, there is no variation of polyphenol concentration when Fe(III)/oxalate/UVA process is applied which corroborates the low organic conversion achieved with this system. With the Fe(III)/citrate/UVA system an increase of total polyphenol concentration is observed during the first 100 min of reaction to eventually slowly diminish. In the case of ozone systems, regardless of the type of ozonation process applied, total polyphenol content continuously decrease with time to become practically negligible after 2 h reaction. Also, the absence of dissolved ozone during the first minutes of reaction confirmed the development of ozone fast reactions in water (Beltra´n, 2004). Measurements of Fe(II) concentration in water also confirm the reactivity of ozone systems and the redox process formed since the
159
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
concentrations remain below 105 M (0.55 mg L1) while in the ferricitrate UVA system Fe (II) concentration were twice higher and remained constant after approximately 20 min of reaction. No appreciable formation of Fe(II) was noticed in the ferrioxalate/UVA system due to the absence of oxidation reactions. Finally, in Fig. 7 the variation of remaining dimensionless TOC with time corresponding to POP experiments at pH 6.5 is shown. As it can be seen, after 2 h reaction, less than 10% TOC conversion is observed with ozonation alone (or O3/UVA) and ferricitrate/UVA systems while no variation at all is seen with the ferrioxalate/UVA system which is in agreement with the results previously shown at pH 6.5 for the individual pharmaceuticals. The other three ozone advanced oxidation systems, however, lead to about 60% TOC conversion in 2 h although at different reaction rates. The fastest process is the combination between ozone and Fe(III)/oxalate/UVA that allows 50% TOC conversion to be reached in 90 min. However, this oxidation system is nearly inhibited after 2 h of reaction. On the contrary, the O3/H2O2/UVA and, specially, the O3/ Fe(III)/citrate/UVA systems present, after 2 h, faster oxidation rates and the processes do not seem to become inhibited. For a practical case, however, systems containing carboxylic acids are not recommended due to their contribution to increase TOC of the water. For example, in the O3/Fe(III)/citrate/UVA experiments applied here, citric acid was added well in excess of Fe(III) (see experimental part for concentrations) to assure total complexation of Fe(III). As a result, the organic carbon of water from citrate represented about 85% of measured TOC at the start of the experiment. Since the start of the reaction, a decrease of the concentration of total citrate (sum of free and complexed) in water was observed so that after about 70 min citrate had completely reacted with hydroxyl radicals (ozone does not react directly with citrate). At this reaction time, total dissolved iron concentration was about half the initially measured at the start of the experiment which means that some other iron complexes were formed since Fe(II)
concentration was very low. After about 2 h reaction, total dissolved iron was not detected in water. Also, at 70 min reaction only about 30% TOC was removed which means that some intermediates formed from citrate oxidation were in water. In any case, hydroxyl free radical oxidation continued until total iron complexes conversion was achieved in 120 min. Apart from the practical point of view, there is a clear synergism in the O3/Fe(III)/citrate/UVA system to remove TOC from water if results are compared to those obtained with ozonation or Fe(III)/citrate/UVA systems.
3.2.
Heterogeneous POP
In a second part of this work, heterogeneous catalytic photooxidation processes were studied. The experiments were also carried out at pH 3 and 6.5 for comparative reasons with the homogeneous ones. Now, TiO2 was used as solid catalyst or semiconductor since its good efficiency in photocatalytic oxidation processes has long been extensively tested (Legrini et al., 1993). Therefore, the O2/TiO2/UVA system can be considered as the basic one in the heterogeneous POP studied here.
3.2.1.
Experiments at pH 3
At pH 3 the following oxidation systems were investigated: O2/ TiO2/UVA, O3/TiO2/UVA, O2/TiO2/Fe(III)/UVA and O3/TiO2/ Fe(III)/UVA. As an example, the changes of organic remaining dimensionless concentration with time in the heterogeneous POP studied are shown in Fig. 8 for the case of BPA. As it can be seen, TiO2 photocatalytic oxidation is not as an efficient process as homogeneous Fe(III) photocatalytic oxidation since compound conversions achieved in 2 h are between 30 and 40% much lower (For information on AAP and TST changes see Figures 7SI and 8SI of supplementary part). Addition of Fe(III) to the heterogeneous TiO2 photocatalytic system allows 1.0
1.0
0.8
0.8
CBPA/CBPA0
CTOC/CTOC0
0.6
0.6
0.4
0.4
0.2
0.2
0.0 0
20
40
60
80
100
120
t, min 0.0 0
20
40
60
80
100
120
t, min
Fig. 7 e Evolution of dimensionless remaining TOC with time during homogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 6.5. See Fig. 6 for symbols and experimental conditions.
Fig. 8 e Evolution of dimensionless remaining concentration of BPA with time during heterogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. Symbols and systems: , O3/UVA; C TiO2/UVA; B O3/TiO2/UVA; : TiO2/Fe(III)/UVA; 6 O3/TiO2/Fe(III)/UVA. See Section 2.1 for experimental conditions.
160
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
a significant increase of the conversion that, at these conditions, reaches values between 85 and 97% which are already of the order of those achieved with homogeneous ferrioxalate photocatalytic oxidation. The increase observed is due to the sum of the contributions of both photocatalytic systems to generate hydroxyl free radicals although some synergic effect should not be disregarded. Thus, it is known that TiO2 photocatalytic oxidation generates small amounts of hydrogen peroxide from possible recombination of superoxide ion radicals formed when oxygen captures electrons in the conduction band (Wang et al., 2002): þ
hn
TiO2 !h þ e
(13)
þ
h þ OH /HO
(14)
O2 þ e /O 2
(15)
2O 2 þ H2 O/H2 O2 þ 2O2
(16)
Then, hydrogen peroxide in the presence of Fe(II) formed from Fe(III) photolysis and UVA radiation forms a new photoFenton process that can enhance the formation of hydroxyl radicals. However, by comparing Figs. 9 and 4 it is deduced that formation rate of Fe(II) is higher in the TiO2 photocatalytic process than in the Fe(III)/UVA process which would imply the action of Fe(III) to capture electrons of the conduction band to avoid electron-hole recombination (Quici et al., 2007; Mesta´nkova´ et al., 2005; Rodrı´guez et al., 2009a): FeðIIIÞ þ e /FeðIIÞ
(17)
Then, it is unlike that a photo-Fenton process develops in a great extension. In any case, ozone involving systems are the most efficient to simultaneously eliminate the organics BPA, TST and AAP as observed in Fig. 8 for the BPA case (more information is given in Figures 7SI and 8SI for the AAP and TST cases, respectively). In fact, oxidation rates are even something higher than those
obtained in the homogeneous ozone oxidation and photocatalytic oxidation at pH 3 and similar to those obtained in ozone systems at pH 6.5 as shown in Fig. 6 (for more information see also Figures 5SI and 6SI of supplementary data). However, on the contrary to what is observed during homogeneous ozonation processes at pH 3, where there was not practically differences between ozone systems (see especially Fig. 8 for BPA), the heterogeneous O3/TiO2/Fe(III)/UVA system presents the highest reaction rates which confirms the synergic effect between processes since in the presence of ozone-hydrogen peroxide is clearly formed as explained before and a photo-Fenton process is established. The highest reactivity of this system can also be deduced from data of dissolved ozone concentration shown in Fig. 10. In Fig. 10 it can be seen that heterogeneous O3/TiO2/Fe(III)/UVA system gives rise to very low concentrations of ozone during the first minutes of reaction (first 20 min, just the time needed to complete disappearance of initial compounds) and then an increase to reach stationary values, that depend on the ozone process applied. Thus, after the first 20 min during ozonation alone a stationary ozone concentration of about 4x105 M (1.9 mg L1) is reached in less than 40 min while in the case of photocatalytic ozonation, O3/TiO2/UVA, the concentration of dissolved ozone stays nearly constant at about 3x106 M (0.14 mg L1) for the whole reaction period and, finally, during O3/TiO2/Fe(III)/UVA oxidation ozone starts to accumulate in water from 43 min reaction to continuously increase to reach 3.5x105 M (1.7 mg L1) after 2 h. These results, however, cannot be interpreted by only considering the initial organic conversions but all other possible reactions developed as commented later. The evolution of Fe(II) concentration with time (see Fig. 9) supports the reactivity of the O3/TiO2/Fe(III)/UVA system and the possible development of a photo-Fenton process. Thus, Fe(II) concentration rapidly increases during the first 5 min to reach 105 M (0.56 mg L1) and then it slowly decreases during the next 40 min. Finally, it increases again to reach a stationary value something higher than 105 M (0.56 mg L1).
-5
5x10
5x10
4x10
4x10
3x10
3x10
-5
CO3d, M
CFe(II), M
-5
2x10
-5
2x10
-5
1x10
1x10
0
0 0
20
40
60
80
100
120
t, min
Fig. 9 e Evolution of Fe(II) concentration with time during heterogeneous photocatalytic oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. See Fig. 8 for symbols and experimental conditions.
0
20
40
60
80
100
120
t, min
Fig. 10 e Evolution of dissolved ozone concentration with time during heterogeneous ozonation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. See Fig. 8 for symbols and experimental conditions.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
It has to note that during the first minutes hydrogen peroxide concentration is expected to also increase due to ozone direct reactions. This was also observed in this work where hydrogen peroxide reached a maximum value, usually after 15e20 min and then it decreased as previously reported for similar ozone systems (Beltra´n et al., 2010). This trend seems to corroborate the development of the photo-Fenton process up to 45 reaction minutes. On the other hand, during TiO2/ Fe(III)/UVA system, Fe(II) accumulates in water just since the start of oxidation to reach, after 30 min, a stationary concentration of about 5 105 M (2.8 mg L1). In other words, all added Fe(III) is in the form of Fe(II) since the start of oxidation. Then, the absence of Fe(II) consumption suggests that in this oxidizing system photo-Fenton processes do not develop which also confirm the lower reactivity of the TiO2/Fe(III)/UVA system compared to the O3/TiO2/Fe(III)/UVA system as, also, shown later for the case of TOC reduction. Regarding the evolution of total polyphenols, compounds of high reactivity with ozone, it was observed (not shown) high degradation rates similar for the ozone systems applied (total conversion was reached in 60 min) and null or negligible decrease concentration in the ozone-free systems which supports the lower efficiency already shown to remove the initial organics and TOC (as commented below). Finally, as observed in Fig. 11, the evolution of TOC with time during the processes investigated not only confirm the higher efficiency of ozone combined systems but the results of Fe(II) and dissolved ozone concentrations shown in Figs. 9 and 10, respectively. Thus, from Fig. 11 it is seen that ozone-free photocatalytic oxidation and ozonation alone hardly allow TOC conversions lower than 10% in 2 h reaction. Results of ozone-free oxidation also support the low reactivity of polyphenols with these systems and, specifically the rapid formation of Fe(II) in the TiO2/Fe(III)/UVA system. During ozonation alone TOC is reduced about 10% during the first 30 min to reach a stationary value which explains why the dissolved ozone concentration starts to increase at about the same reaction time (see Fig. 10) to also reach a stationary value
1.0
CTOC/CTOC0
0.8
0.6
0.4
0.2
0.0 0
20
40
60
80
100
120
t, min
Fig. 11 e Evolution of dimensionless remaining TOC with time during heterogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 3. See Fig. 8 for symbols and experimental conditions.
161
from 40 min reaction suggesting the inhibition of oxidation. In addition, the absence of Fe(III) in this system (O3/TiO2/UVA) makes it not possible any photo-Fenton process development. In the case of heterogeneous photocatalytic ozonation, TOC degradation continuously takes place for the whole reaction period to reach a conversion of more than 90% after 2 h. This means that the process was not inhibited and explains the low concentration of dissolved ozone for the 2 h reaction. In the system, however, heterogeneous O3/TiO2/Fe(III)/UVA although TOC removal rates are the highest among oxidation systems applied during the first 40e50 min, the process is finally stopped so that maximum TOC conversion achieved, in 2 h, was practically the same as after 50 min, about 82%. This process inhibition is also deduced from data of dissolved ozone concentration shown in Fig. 10. Thus, in this figure it is also seen negligible ozone concentrations measured during the first 43 min, the time period where reactions did develop, but a sudden increase of ozone concentration to reach, after 2 h, a value close to that measured for the ozonation alone process. This undoubtedly means that some sort of inhibition of the process occurs after the first 40 min. A priori, two situations explain the inhibition observed in Fig. 11: a) a possible reduction of TiO2 activity for the fixation of Fe(III) atoms on the catalyst surface or b) the development of a photo-Fenton process during the first 40e50 min when TOC reduction was appreciable. The first explanation, however, is not possible since total dissolved Fe concentration kept practically constant for the whole reaction period, so that iron was not deposited on the TiO2 surface. The second explanation matches the results of the changes observed in dissolved ozone, hydrogen peroxide and Fe(II) concentrations with time as indicated above. Thus, it is evident the synergic effect of the O3/TiO2/Fe(III)/UVA system.
3.2.2.
Experiments at pH 6.5
Finally, a series of experiments of heterogeneous POPs were also carried out a pH 6.5. Now, to avoid Fe(III) precipitation, oxalic acid or citric acid was also added to form ferricarboxylates complexes, known for their photolytic activity (Faust and Zepp, 1993; Zhang, 2000; Rodrı´guez et al., 2009a). Fig. 12 shows, as example, the evolution with time of the remaining dimensionless concentrations of AAP in heterogeneous POPs. As it can be observed, typical TiO2 photocatalytic oxidation leads to slightly better conversion results than at pH 3 but organic conversions remain between 50 and 60% after 2 h reaction (more information is given in Figures 9SI and 10SI for the BPA and TST cases, respectively). When Fe(III) and oxalic acid are added conversions even diminish to about 20% likely due to the unstable ferrioxalate complex at this pH value something also observed in the corresponding homogeneous system (see Fig. 6). However, when citric acid was charged instead of oxalic acid, the higher stability of the ferricitrate complex allows organic conversions to increase to 75% for AAP, 87% for BPA and 90% for TST after 2 h. Nonetheless, when ozonation results are considered, once more, total conversion of compounds studied is reached in less than 20 min. In fact, with the heterogeneous ozone POP applied total conversion of BPA, AAP and TST can be achieved in just 5 min for the first two compounds and 10 min for the third (in the O3/TiO2/Ferrioxalate/UVA system). Contrary to
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
1.0
1.0
0.8
0.8
0.6
0.6
CTOC/CTOC0
CAAP/CAAP0
162
0.4
0.2
0.4
0.2
0.0
0.0
0
20
40
60
80
100
120
0
20
t, min
the results of ozonation processes in homogeneous (pH 3 and 6.5) and heterogeneous (pH 3) systems, differences of the time needed to reach total conversion are higher for different heterogeneous ozone POP at pH 6.5 as observed, for example, in Fig. 12 for AAP. Thus, for ozonation alone between 20 and 25 min are necessary for total conversions of BPA and AAP and TST, respectively. Thus, it is evident that in heterogeneous ozone POP removal of compounds are not only due to ozone direct reactions but also to possible surface and hydroxyl radical reactions. For example, photo-Fenton processes are likely to develop as also shown in experiments at pH 3. Results obtained with total polyphenol concentration (not shown) indicate high reaction rates in ozone processes faster than those achieved at pH 3 (total polyphenol disappearance is now observed after 40 min reaction). On the contrary, ozone-free heterogeneous POP show again a null or even an increase of polyphenol concentration (case of systems with ferricarboxylates) or a slight decrease (case of classical TiO2 photocatalytic oxidation). In the case of TOC evolution (as seen in Fig. 13) only combined heterogeneous ozone POP allow significant decrease of this parameter. Regarding the O3/TiO2/ UVA process, TOC conversion results at pH 6.5 are lower than those at pH 3 with 70% and 90% TOC reductions, respectively, after 2 h reaction. The highest efficiency of this system at acid conditions could likely be due to the reactions (18) and (12): O3ðadÞ þ e
/O 3
60
80
100
120
t, min
Fig. 12 e Evolution of dimensionless remaining concentration of AAP with time during heterogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 6.5. Symbols and systems: , O3/UVA; C TiO2/UVA; B O3/TiO2/ UVA; : TiO2/Fe(III)/Oxal/UVA; 6 O3/TiO2/Fe(III)/Oxal/UVA; ; TiO2/Fe(III)/Cit/UVA; 7 O3/TiO2/Fe(III)/Cit/UVA. See Section 2.1 for experimental conditions.
40
(18)
Regarding the O3/TiO2/Ferrioxalate/UVA system no inhibition is observed as a difference of what is seen at pH 3 for the O3/TiO2/Fe(III)/UVA system. In any case, when carboxylates are present an important fraction of TOC reductions achieved are likely due to the removal of these compounds which would also agree the decrease observed in total iron
Fig. 13 e Evolution of dimensionless remaining TOC with time during heterogeneous oxidation processes of a mixture of testosterone, bisphenol A and acetominophen in water at pH 6.5. See Fig. 12 for symbols and experimental conditions.
concentration with time as indicated below. Then, TOC is not a good parameter to follow the efficiency of carboxylate POP systems. For systems with ozone and ferricarboxylates a decrease of total dissolved iron was observed since the start of reaction in the case of ferrioxalate and after about 30 min for the case of ferricitrate, which can be attributed to the lower stability of the former at pH 6.5 (Abrahamson et al., 1994; Nansheng et al., ´ lvarez et al., 2010). On the other hand, Fe(II) stayed at 1998; A negligible levels during TiO2 photooxidation in the presence of ferrioxalate while about 2 106 M (0.11 mg L1) concentrations were achieved in the presence of ferricitrate systems, regardless of the presence of ozone in both cases. This confirms the higher reactivity of ferricitrate system likely due to the higher stability of the complex compared to that of ferrioxalate. Concentrations of dissolved ozone follow typical trends of ozonation systems, that is, very low concentration during the first 20 min reactions, as a consequence of fast direct reactions of ozone with initial compounds and, likely, first intermediates (measured as polyphenols) and an increase to reach a stationary concentration. The lowest dissolved stationary concentration was observed in the O3/TiO2/UVA system which suggests this system presented the highest reactivity.
3.3.
Kinetic aspects
Although there are basically one or two main kinetic contributions to the organic removal rate in the POP studied: the hydroxyl radical reaction and/or the direct ozone reaction, POP are complex systems involving different reaction mechanisms (direct reactions of ozone, reactions with hydroxyl radicals generated through different ways, possible direct oxidation with positive holes on the TiO2 surface, photoFenton processes, etc). Thus, the kinetic study has been limited to some aspects related to the direct ozonation
163
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
contribution in some of the ozone processes and the estimation of the hydroxyl radical concentration to remove the initial compounds and TOC.
3.3.1.
Kinetics of initial organics removal
In ozone-free systems, the organics at any time are removed through free radical oxidation so that a mass balance of any organic compound in water in the semibatch photocatalytic reactor used is as follows: dCM ¼ kHO CHO CM dt
(19)
where kHO, CHO and CM are the rate constant of the hydroxyl radical-M reaction and the concentrations of hydroxyl radical and compound M, respectively. From experimental values of the concentration and accumulation rate of M, left side of equation (19), determined in this work and the rate constant kHO obtained from literature data the concentration of hydroxyl radicals can be estimated. Since the three compounds studied were simultaneously treated it is expected that values of CHO determined from equation (19) be similar when applied to reaction rate and concentration data of BPA, AAP or TST. Literature only gives values of kHO for BPA and AAP reactions with hydroxyl radicals (Bisby and Tabassum, 1988; Gozmen et al., 2003) so that from data of these compounds CHO was easily found by applying equation (19) (See Table 2 for results). In the case of TST, the rate constant of the hydroxyl radical reaction and the corresponding CHO were determined by a trial and error method as follows. First, a value of kHO was assumed to get CHO from equation (19) at different conditions and oxidation process and then this value was compared to those already determined from data of BPA and AAP for the same oxidation process. The best value of kHO was finally the one minimizing the sum of the squares differences between CHO from TST data and the assumed kHO and the average values of CHO from data of BPA and AAP directly determined from Equation (19) at different reaction time and oxidation process. The optimum kHO value was found to be 1.5 1010 M1 s1. Table 2 also shows the results of CHO from TST data. As it can be seen, in most of cases, at a given time, pH and oxidizing system, similar results of CHO were obtained from Equation (19) for the three initial compounds. This was expected since the three
compounds were simultaneously treated. Also, it is seen that CHO values are nearly constant with time with variations lower than one order of magnitude. At pH 3, the highest concentration of hydroxyl radicals are produced in photoFenton and ferrioxalate/UVA and TiO2/Fe(III)/UVA systems for homogeneous and heterogeneous processes, respectively, while, at pH 6.5, the best processes for generating hydroxyl radicals are the ferricitrate/UVA systems with or without TiO2 for homogeneous and heterogeneous processes, respectively. These results coincide with those commented about the changes in the concentration of compounds with time. In ozonation processes, there is another contribution to the organic reaction rate: the direct ozone-M reaction, so that Equation (19) is now: dCM ¼ kD CO3 CM þ kHO CHO CM dt
(20)
In these cases, however, the estimation of hydroxyl radical concentration from Equation (20) presents a high uncertainty due to the low values measured for the dissolved ozone concentration. Nonetheless, in ozone systems an important objective is to know the relative effect of the direct and hydroxyl radical reactions. Here, the contribution of the direct ozone reaction to the removal of the organics studied, first term of the right side of Equation (20), has been estimated in a few cases, also because of the low values of the concentration of ozone (about 106 M (0.05 mg L1)) and possible interferences in determining the ozone concentration with some ozone systems (i.e. those where a Fenton reaction develops). This contribution, DR%, is given by Equation (21): DR% ¼
kD CO3 CM 100 dCM =dt
(21)
Thus, only for experiments at pH 3 in ozonation alone and photocatalytic ozonation in the presence of Fe(III), contribution of the direct reaction was estimated after 10 min reaction (for lower times dissolved ozone concentration was very close to the detection limit of the method). Both in ozonation alone and ozone/Fe(III)/UVA systems, DR%, in the case of BPA and AAP, was always something higher than 100% while in the case of TST was between 80 and 60%. The first result clearly suggests BPA and AAP were exclusively removed by ozone direct reaction (the values of DR% obtained higher than 100%
Table 2 e Estimated hydroxyl radical concentration from organic removal in ozone-free UVA photocatalytic processes (CHO 3 1014, M).a System
pH 3 t ¼ 5 min
System
t ¼ 30 min
t ¼ 60 min
t ¼ 5 min
BPA TST AAP BPA TST AAP BPA TST AAP Fe(III) TiO2 Ferrioxalate Photo-Fenton TiO2/Fe(III)
1.65 0.84 5.08 4.65 1.89
1.68 0.63 5.8 4.64 2.15
1.31 0.77 2.99 4.26 1.96
0.99 1.26 0.68 0.61 6.73 13.7 6.72 6.51 2.26 2.49
0.89 0.76 0.57 0.54 5.23 9.00 6.31 10.2 2.19 2.79
0.97 0.53 e 7.93 3.16
0.70 0.48 4.42 8.59 2.36
pH 6.5 t ¼ 30 min
t ¼ 60 min
BPA TST AAP BPA TST AAP BPA TST AAP Ferrioxalate Ferricitrate TiO2 TiO2/Ferrioxalate TiO2/Ferricitrate
1.10 3.03 0.82 0.85 2.46
0.82 2.49 0.58 0.58 1.87
0.90 2.48 0.83 0.75 1.81
0.30 3.41 1.05 0.30 2.43
0.20 2.51 0.64 0.23 1.81
0.25 2.64 0.90 0.25 1.80
0.17 3.97 1.28 0.20 2.46
0.11 2.76 0.69 0.16 1.77
0.15 2.98 0.95 0.16 1.89
a All systems with UVA black light radiation. Results from equation (19). Values of kHO for BPA and AAP are 1.1 1010 and 9.8 109 M1 s1, respectively (Bisby and Tabassum, 1988; Gozmen et al., 2003). The value of kHO for TST-OH reaction, 1.5 1010 M1 s1 was determined in this work as detailed in Section 3.3.1.
164
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
Table 3 e Estimated hydroxyl radical concentration from TOC results in different UVA POP (CHO 3 1014, M).a System
O3/Fe(III) O3/TiO2 O3/H2O2 O3/Fe(III)/TiO2 Photo-Fenton O3/Photo-Fenton Ferrioxalate O3/ferrioxalate
pH 3
System
t ¼ 15 min
t ¼ 60 min
t ¼ 120 min
4.14 5.14 1.30 7.74 3.04 4.62 5.78 5.45
5.71 6.65 1.34 3.97 7.04 3.08 2.06 8.61
3.60 2.28 3.83
O3/TiO2 O3/H2O2 O3/Ferrioxalate O3/TiO2/Ferrioxalate Ferricitrate O3/ferricitrate TiO2/ferricitrate O3/TiO2/ferricitrate
e 4.69 2.11 1.04 1.75
pH 6.5 t ¼ 15 min
t ¼ 60 min
t ¼ 120 min
2.71 1.52 3.91 3.57 0 1.31 0.29 1.46
2.91 1.71 2.04 4.31 2.14 2.73 0.25 3.12
2.91 2.25 0.58 5.54 0.32 1.93 0.22 4.17
a All systems with UVA black light radiation. Results from Equation (19) with TOC instead of CM. kHO ¼ 5 109 M1 s1.
can be attributed to some deviations of the reported data on kD from the actual ones). In the case of TST, contribution of free radicals is evident since values of DR% agree the results commented while discussing the organic concentration evolution (see Fig. 2).
3.3.2.
Kinetics of TOC oxidation
For any POP studied, once the initial compounds are eliminated hydroxyl radical oxidation is the main way of removal. Thus, if TOC is considered as lumped parameter of the organic concentration in water, the hydroxyl radical concentration can be estimated with equation (19), taken TOC as CM. In this case, for kHO a value of 5 109 M1 s1 was applied. This value is the average between 1010 and 5 107 M1 s1 values of rate constant for first reacting products, such as BPA, and typical end products of ozonation reactions, such as oxalic acid, respectively (Buxton et al., 1988). Thus, in Table 3 calculated values of CHO estimated from the modified Equation (19) are given. As it can be seen, in most of cases the order of magnitude was the same as in Table 2 for initial compounds (1014 M) with ozone processes showing the higher efficiency in producing hydroxyl radicals.
3.3.3. Kinetic considerations for low concentrations of pollutants In a practical case, real water contains concentrations of pollutants such as BPA, AAP and TST, as high as tenths of mg L1 so that it is important to make estimations of the contributions of ozone direct and hydroxyl radical reactions in this case. Since no experimental data was available Equation (19) cannot be applied but DR% can be determined from theoretical Equation (22): kD CO3 100 DR% ¼ kD CO3 þ kHO CHO
(22)
According to Equation (22) DR% is not dependent on pollutant concentration but on the concentration of hydroxyl radicals, dissolved ozone and rate constants of ozone direct and hydroxyl radical reactions. Also, from data of Table 2 or 3, it is deduced that concentration of hydroxyl radicals does not depend either on pollutant concentration but on the experimental conditions applied for ozone and catalyst concentrations and intensity of light, depending on the oxidation process. This particularly holds when dealing with pollutant concentrations of mg L1 level since any possible initiating,
promoting or scavenging effect of pollutants on hydroxyl radical formation would be negligible (Staehelin and Hoigne´, 1985). From Table 3 values of hydroxyl radical concentration between 1014 and 9 1014 M can then be used in equation (22). On the other hand, in water containing BPA, AAP and TST at mg L1 level, concentrations of dissolved ozone should be higher than those observed in this work just once concentrations of BPA, AAP and TST were below and close to the detection limit of the analytical method used here (100 mg L1). These low concentrations were reached at 15e25 min reaction times depending on the ozonation system as seen in Figs. 2, 6 and 8 and 12 (see also Figures 2SI, 3SI, 5SI, etc, of supplementary information). Thus, concentrations of ozone between 106 (0.05 mg L1) and 2 105 M (0.96 mg L1) can be taken as seen for example in Fig. 10. Then, with data of dissolved ozone and hydroxyl radical concentrations Equation (22) was used to estimate DR% for low pollutant concentrations. Application of Equation (22) leads to DR% values close to 100% in the case of BPA and AAP for all cases. Regarding TST, only when dissolved ozone concentration is 5 106 M (0.25 mg L1) and hydroxyl radical concentration is 1015 M, DR% is also close to 100%. In the rest of cases, contribution of hydroxyl radical oxidation varies from 70 to 20%. As a consequence, for removal of mg L1 concentrations of BPA, AAP and TST from water, photocatalytic ozonation systems are mainly recommended in the case of TST.
4.
Conclusions
Major conclusions reached in this study are: For homogeneous POP: The basic Fe(III) photocatalytic oxidation only allows partial removal of organics during 2 h reaction. Addition of oxalic acid or citric acid, necessary at pH 6.5 to avoid Fe(III) precipitation, or hydrogen peroxide (photo-Fenton process) can lead to total removal but more than 60 min are needed. On the contrary, ozonation processes lead to total removal of organics in less than 20 min. Ozonation alone and Fe(III)/UVA leads to very low TOC removal. At pH 3 photo-Fenton, ozone/Fe(III)/UVA and ozone/ferrioxalate/UVA systems lead to the highest TOC
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
removal in 2 h (about 80%). TOC reductions are higher in POP at pH 3, especially in Fe based photocatalytic processes. Total polyphenols are only completely removed at a significant rate in ozonation systems. In ozone-free POP, total polyphenol concentration remains constant or even increase except in the photo-Fenton process where some decrease in polyphenol concentration is also observed. A synergic effect between ozone and photocatalytic oxidation is clearly noticed, especially for TOC removal. For heterogeneous POP At pH 6.5 citric acid is better than oxalic acid to remove the organic compounds in reactions initiated by the corresponding ferricarboxylate photolysis. Ozonation processes allow again total conversion of organic in less than 15 min. Regarding TOC, basic photocatalytic ozonation (O3/TiO2/ UVA) and also O3/Fe(III)/carboxylic acid/UVA systems leads to the highest TOC reductions (about 90% at pH 3 and 80% at pH 6.5). However, TOC reduction is not a good parameter to measure the efficiency of carboxylic acid based Fe(III) photocatalytic oxidation because the TOC contributing fraction of added carboxylic acids. As in homogeneous systems, feeding ozone to a basic photocatalytic oxidation yields significant increases of TOC removal due to the synergic effects between both processes. About mechanism and kinetic aspects: Regardless of any type of ozonation process and pH, initial compounds, at 1e3 mg L1 to some hundreds mg L1, are mainly removed through direct ozone reactions, especially for the cases of BPA and AAP. Then, for the removal of initial organics no advanced oxidation is needed. In ozone-free systems, hydroxyl radical reactions are the way of oxidation. Concentration of hydroxyl radicals are of the order of 1014 M, regardless of the type of process. However, the highest values, at pH 3, correspond to photo-Fenton and ferrioxalate UVA and TiO2/Fe(III)/UVA systems for homogeneous and heterogeneous processes, respectively, while, at pH 6.5, the best processes for generating hydroxyl radicals are ferricitrate/UVA systems with or without TiO2 for homogeneous and heterogeneous processes, respectively. For TOC removal, however, advanced oxidation processes are needed, especially, ozone combined processes. In some ozone systems a modified photo-Fenton mechanism develops. Ozone photocatalytic processes show the higher efficiency in producing hydroxyl radicals. For concentrations of pollutants in the mg L1 level, photocatalytic ozonation is recommended for TST removal since ozonation alone would allow total removal of BPA and AAP.
Acknowledgments This work has been supported by the CICYT of Spain and the European Region Development Funds of the European Commission (Project CTQ2009/13459/C05/05).
165
Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.10.038.
references
Abrahamson, H.B., Rezvani, A.B., Brushmiller, J.G., 1994. Photochemical and spectroscopic studies of complexes of iron(III) with citric acid and other carboxylic acids. Inorganica Chimica Acta 226, 117e127. Agustina, T.E., Ang, H.M., Vareek, V.K., 2005. A review of synergistic effect of photocatalysis and ozonation on wastewater treatment. Journal of Photochemistry and Photobiology 6, 264e273. ´ lvarez, P.M., Rodrı´guez, E.M., Ferna´ndez, G., Beltra´n, F.J., 2010. A Degradation of bisphenol A in water by Fe(III)/UVA and Fe(III)/ polycarboxylate/UVA photocatalysis. Water Science and Technology 61.11, 2717e2722. Andreozzi, R., Caprio, V., Marotta, V., Vogna, D., 2003. Paracetamol oxidation from aqueous solutions by means of ozonation and H2O2/UV system. Water Research 37, 993e1004. Bader, H., Hoigne´, J., 1981. Determination of ozone in water by the indigo method. Water Research 15, 449e456. Barron, E., Deborde, M., Rabouan, S., Mazellier, P., Legube, B., 2006. Kinetic and mechanistic investigations of progesterone reaction with ozone. Water Research 40, 2181e2189. Beltra´n, F.J., Aguinaco, A., Garcı´a-Araya, J.F., 2010. Kinetic modelling of TOC removal in the photocatalytic ozonation of diclofenac aqueous solutions. Applied Catalysis B: Environmental 100, 289e298. Beltra´n, F.J., 2004. Ozone Reaction Kinetics for Water and Wastewater Systems. Lewis Publ., Boca Raton, Florida. Bhatkhande, D.S., Pagarkar, V.G., Beenackers, A.C.M., 2001. Photocatalytic degradation for environmental applicationsa review. Journal of Chemical Technology and Biotechnology 77, 102e116. Bisby, R.H., Tabassum, N., 1988. Properties of the radicals formed by one-electron oxidation of acetaminophen - a pulse radiolysis study. Biochemical Pharmacology 37, 2731e2738. Brose´us, R., Vincent, S., Aboulfadl, K., Daneshvar, A., Sauve´, S., Barbeau, B., Pre´vost, M., 2009. Ozone oxidation of pharmaceuticals, endocrine disruptors and pesticides during drinking water treatment. Water Research 43, 4707e4717. Buxton, G.V., Greenstock, C.L., Helman, W.P., Ross, A.B., 1988. Critical review of data constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals (OH/O) in aqueous solution. Journal of Physical and Chemical Reference Data 17, 513e886. Daughton, C.H., Ternes, T.A., 2000. Special report: pharmaceuticals and personal care products in the environment: agents of subtle change? Environmental Health Perspectives 108, 598e602. Deborde, M., Rabouan, S.D., Duguet, J.P., Legube, B., 2005. Kinetics of aqueous ozone-induced oxidation of some endocrine disruptors. Environmental Science and Technology 39, 6086e6092. Decoret, C., Royer, J., Legube, B., Dore´, M., 1984. Experimental and theoretical studies of the mechanism of the initial attack of ozone on some aromatics in aqueous medium. Environmental Technology Letters 5, 207e218. Eisenberg, G.M., 1943. Colorimetric determination of hydrogen peroxide. Industrial and Engineering Chemistry 15, 327e328. Escher, B.I., Baumgartner, R., Koller, M., Treller, K., Lienert, J., McCardell, C.S., 2011. Environmental toxicology and risk
166
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 5 2 e1 6 6
assessment of pharmaceuticals from hospital wastewater. Water Research 45, 75e92. Faust, B.C., Hoigne´, J., 1990. Photolysis of Fe(III)-hydroxy complexes as sources of OH radicals in clouds, fog and rain. Atmospheric Environment 24A, 79e89. Faust, B.C., Zepp, R.G., 1993. Photochemistry of aqueus iron(III) polycarboxylate complexes- roles in the chemistry of atmospheric and surface water. Environmental Science and Technology 27, 2517e2522. Fenton, H.J.H., 1876. On a new reaction of tartaric acid. Chemical News 33, 190. Gozmen, B., Oturan, M.A., Oturan, N., Erbatur, O., 2003. Indirect electrochemical treatment of bisphenol A in water via electrochemically generated Fenton’s reagent. Environmental Science and Technology 37, 3716e3723. Hatchard, C.G., Parker, C.A., 1956. A new sensitive chemical actinometer. II. Potassiumferrioxalate as a standard actinometer. Proceedings of the Royal Society London Service A 235, 518e536. Hoigne´, J., Bader, H., 1983. Rate constants of the reactions of ozone with organic and inorganic compounds. II. Dissociating organic compounds. Water Research 17, 185e194. Kim, S.D., Cho, J., Kim, I.S., Vanderford, B.J., Snyder, S.A., 2007. Occurrence and removal of pharmaceuticals and endocrine disruptors in South Korean surface, drinking and waste waters. Water Research 41, 1013e1021. 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 U. S. streams, 1999e2000: a national reconnaissance. Environmental Science and Technology 36, 1202e1211. Legrini, O., Oliveros, E., Braun, A.M., 1993. Photochemical processes for water treatment. Chemical Reviews 93, 671e698. Leitzke, A., von Sonntag, C., 2009. Ozonolysis of unsaturated acids in aqueous solution: acrylic, methacrylic, maleic, fumaric and muconic acids. Ozone Science & Engineering 31, 301e308. Logager, T., Holcman, J., Sehested, K., Pedersen, T., 1992. Oxidation of ferrous ions by ozone in acidic solutions. Inorganic Chemistry 31, 3523e3529. Matthews, R.W., McEvoy, S.R., 1992. A comparison of 254 nm and 350 nm excitation of TiO2 in simple photocatalytic reactors. Journal of Photochemistry and Photobiology A Chemistry 66, 355e366. Mesta´nkova´, H., Mailhot, G., Jirkovsky´, J., Kry´sa, J., Bolte, M., 2005. Mechanistic approach of the combined (iron-TiO2) photocatalytic system for the degradation of pollutants in aqueous solution: an attempt of rationalisation. Applied Catalysis B: Environmetal 57, 257e265. Mvula, E., von Sonntag, C., 2003. Ozonolysis of phenols in aqueous solutions. Organic and Biomolecular Chemistry 1, 1749e1756. Nansheng, D., Feng, W., Fan, L., Mei, X., 1998. Ferric citrateinduced photodegradation of dyes in aqueous solutions. Chemosphere 36, 3101e3112. Nelson, E.D., Do, H., Lewis, R.S., Carr, S.A., 2011. Diurnal variability of pharmaceutical, personal care product, estrogen and alkylphenol concentrations in effluent from a tertiary wastewater treatment facility. Environmental Science & Technology 45, 1228e1234.
Quici, N., Morgada, G.E., Gettar, R.T., Bolte, M., Litter, M.I., 2007. Photocatalytic degradation of citric acid under different conditions: TiO2 heterogeneous photocatalysis against homogeneous photolytic processes promoted by Fe(III) and H2O2. Applied Catalysis B: Environmental 71, 117e124. Rodrı´guez, E.M., Nu´n˜ez, B., Ferna´ndez, G., Beltra´n, F.J., 2009a. Effects of some carboxylic acids on the Fe(III)/UVA photocatalytic oxidation of muconic acid in water. Applied Catalysis B: Environmental 89, 214e222. ´ lvarez, P.M., Rodrı´guez, E.M., Ferna´ndez, G., Ledesma, B., A Beltra´n, F.J., 2009b. Photocatalytic degradation of organics in water in the presence of iron oxides: influence of carboxylic acids. Applied Catalysis B: Environmental 92, 240e249. ´ lvarez, P.M., Herna´ndez, R., Rodrı´guez, E.M., Ferna´ndez, G., A Beltra´n, F.J., 2011. Photocatalytic degradation of organics in water in the presence of iron oxides: effects of pH and light source. Applied Catalysis B: Environmental 102, 572e583. Safarzadeh-Amiri, A., Bolton, J., Cater, S., 1997. Ferrioxalatemediated photodegradation of organic pollutants in contaminated water. Water Research 31, 787e798. Singleton, V.L., Rossi, J.A., 1965. Colorimetry of total phenolic with phosphomolybdic-phosphotungstic acid reagents. American Journal of Enology and Viticulture 16, 144e158. Stackelberg, P.E., Gibs, J., Furlong, E.T., Meyer, M.T., Zaugg, S.D., Lippincott, R.L., 2007. Efficiency of conventional drinkigwater-treatment processes in removal of pharmaceuticals and other organic compounds. Science of the Total Environment 377, 255e272. Staehelin, S., Hoigne´, J., 1982. Decomposition of ozone in water: rate of initiation by hydroxyde ions and hydrogen peroxide. Environmental Science & Technology 16, 676e681. Staehelin, S., Hoigne´, J., 1985. Decomposition of ozone in water in the presence of organic solutes acting as promoters and inhibitors of radical chain reactions. Environmental Science & Technology 19, 1206e1212. Stookey, L.L., 1970. Ferrozine- a new spectrophotometric reagent for iron. Analytical Chemistry 42, 779e781. Ternes, T.A., 1998. Occurrence of drugs in German sewage treatment plants and rivers. Chemosphere 32, 3245e3260. Turchi, C.S., Ollis, D.F., 1990. Photocatalytic degradation of organic water contaminants: mechanisms involving hydroxyl radical attack. Journal of Catalysis 122, 178e192. Wang, S., Shirahisi, F., Nakano, K., 2002. A synergistic effect of photocatalysis and ozonation on depcomposition of formic acid in an aqueous solution. Chemical Engineering Journal 87, 262e271. Zepp, R.G., Faust, B.C., Hoigne´, J., 1992. Hydroxyl radical formation in aqueous reactions (pH 3-8) of iron (II) with hydrogen peroxide: the photo-Fenton reaction. Environmental Science & Technology 26, 313e319. Zhang, Y., 2000. Light and iron(III)-induced oxidation of chromium (III) in the presence of organic acids and manganese (II) in simulated atmospheric water. Atmospheric Environment 34, 1633e1640. Zuo, Y., 1995. Kinetics of photochemical/chemical cycling of iron coupled with organic substances in cloud and fog droplets. Journal of Hazardous Materials 59, 3123e3130.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Boron bioremoval by a newly isolated Chlorella sp. and its stimulation by growth stimulators Burcu Ertit Tas¸tan, Ergin Duygu, Go¨nu¨l Do¨nmez* Department of Biology, Faculty of Science, Ankara University, 06100 Bes¸evler, Ankara, Turkey
article info
abstract
Article history:
It has been well documented that excess concentrations of boron (B) causes toxic effects on
Received 3 August 2011
many of the environmental systems. Although Chlorella sp. has been studied to remove
Received in revised form
pollutants from water, its capacity to remove B has not been investigated yet. Boron
11 October 2011
removal levels of newly isolated Chlorella sp. were investigated in BG 11 media with
Accepted 19 October 2011
stimulators as triacontanol (TRIA) and/or sodium bicarbonate (NaHCO3) and without them,
Available online 28 October 2011
to test if they could increase the removal efficiency by increasing biomass. The assays were performed to determine the effect of different medial compositions, B concentrations, pH
Keywords:
and biomass concentrations onto removal efficiency. Boron removal was investigated at
Microalgae
5e10 mg/L range at pH 8 in different medial compositions and maximum removal yield
Boron
was found as 32.95% at 5.45 mg/L B in media with TRIA and NaHCO3. The effect of different
Bioremoval
pH values on the maximum removal yield was investigated at pH 5e9, and the optimum pH
Growth stimulators
was found again 8. The interactive effect of biomass concentration and B removal yield was
Wastewater treatment
also investigated at 0.386e1.061 g wet weight/L biomass. The highest removal yield was found as 38.03% at the highest biomass range. This study highlights the importance of using new isolate Chlorella sp. as a new biomaterial for B removal process of waters containing B. ª 2011 Elsevier Ltd. All rights reserved.
1.
Introduction
As known very well, boron (B) is one of the essential trace elements (Waggott, 1969) and its biological role has still been studied by a number of researchers (Lee et al., 2009; Sheng et al., 2009). In spite of its biologically important role in metabolism at its low concentrations (Frick, 1985; Wojcik et al., 2008; Lee et al., 2009), excess amounts of B is harmful and causes toxic effects (Davis et al., 2002; Gunes et al., 2006; Del-Campo Marı´n and Oron, 2007; Sasmaz and Obek, 2009). Boron is mainly released to the environment by discharged industrial wastewaters (Coughlin, 1998). Manufacturing facilities of heat resistant materials (Morioka et al., 2007), storage and distribution of solar energy systems (Abu-Hamed et al.,
2007), catalysts (Xu et al., 2009), ceramics (Christogerou et al., 2009) and glass (Crawford et al., 2007) can be presented as main examples of important B sources. On the other hand, Turkey has the largest B reserves (60%) in the world (Okay et al., 1985) and therefore, toxicity of B is come into more prominence. As Cervilla et al. (2009) attracted attention to the fact that B toxicity has become important especially in areas close to the Mediterranean Sea, where intensive agriculture has been developed. They evidenced that excess level of B in cultivated soils lead to B toxicity which caused inhibition of nitrate reduction and consequence increase in ammonium assimilation in tomato plants, accompanied with the loss of leaf biomass and disorders in organic nitrogen metabolism. Nable et al. (1997) in fact, previously stated that B-rich soils were
* Corresponding author. Tel.: þ90 312 212 67 20; fax: þ90 312 223 23 95. E-mail address:
[email protected] (G. Do¨nmez). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.10.045
168
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5
causing B toxicity in the field where crop yields decreased. They also mentioned that various anthropogenic sources of excess B might increase amount of B in soil to the toxic levels for plants, such as wastes from surface mining, fly ash, and industrial chemicals, and the most important source was irrigation water. As a result once the B level is higher than required it is needed removing excess of B in order to decrease the effects of B toxicity. According to the literature, some different methods have been tried to reduce B pollution in environment, including adsorptioneflocculation (Chong et al., 2009; Kavak, 2009), elec¨ ztu¨rk tro coagulation (Yılmaz et al., 2008), reverse osmosis (O et al., 2008), precipitation (Itakura et al., 2005), ion-exchange (Okay et al., 1985), use of B-selective resins (Simonnot et al., 2000) and some biological materials, such as duckweeds (Sasmaz and Obek, 2009; Del-Campo Marı´n and Oron, 2007). On the other hand, a number of studies have focused on the use of microalgae in removal of several pollutants from the culture media or wastewaters (de-Bashan and Bashan, 2010; Karacakaya et al., 2009; El-Sheekh et al., 2005). Chlorella sp. is known as one of the most useful microalgae for different purposes. This genus has been investigated in numerous studies considering its high growth rates under effect of different conditions (Lee et al., 2002; Valderrama et al., 2002; Sung et al., 1999) and it was shown that Chlorella sp. could remove pollutants with a high capacity and in an efficient way compared to many other aquatic organisms (Ruangsomboon and Wongrat, 2006; Gonza´lez et al., 1997; Hanagata et al., 1992). Its capacity to remove B has not been investigated previously; therefore this study aims to fill this gap in literature. In B removal studies, investigators did not add any growth stimulators to the culture media. However, it might be possible to enhance the growth of microorganism by enrichment of media with adding them. Another purpose of the present study was to investigate the effects of some growth stimulators effecting B removal process and to see whether they would enhance the growth of Chlorella sp. and increase the efficiency of bioremoval process. One of the growth stimulator used current study is Triacontanol (TRIA) and the other one is sodium bicarbonate (NaHCO3). TRIA, a long chain 30-carbon primary alcohol, (C30H61OH) is a well known plant hormone and growth regulator (Ries and Houtz, 1983). Stimulatory effects of TRIA on the photosynthesis, growth and net bioproductivity of some green algae and cyanobacteria species have also been reported (Karacakaya et al., 2009; Houtz et al., 1985a, 1985b). On the other hand, carbon is fixed by microalgae and is produced biomass and energy. Wang et al. (2008) reviewed that microalgae can fix carbon dioxide (CO2) from soluble carbonate forms such as NaHCO3 and use it into photosynthesis. It was shown that inexpensive carbon source of NaHCO3 was affected to growth and chlorophyll concentrations of microalgae by the way dissolving into the culture media and replacing the atmospheric CO2 and it was a product resulting from CO2 capture with alkali method and was a suitable C source for the growth of Chlorella vulgaris (Wang et al., 2010). It may be feasible to increase the growth of Chlorella sp. by adding TRIA and NaHCO3 into media and the efficiency of B removal processes can be increased. The aim of this study was to examine the hypothesis that (i) to determine the best B removal conditions in detail by changing experimental parameters, (ii) to examine if the
biomass production and B removal capacity of Chlorella sp. could be increased by adding TRIA and NaHCO3 into medium at different parameters, and (iii) to check if there was a potential offer by Chlorella sp. as an effective and eco-friendly biomaterial for developing B removal procedure. According to our best knowledge this is the first report about B removal by a new isolated microalgae Chlorella sp., with proposed target.
2.
Materials and methods
2.1.
Isolation and culture conditions
The microalgal culture was isolated from water supply located in Sorgun, Yozgat, Turkey. Samples were spread on the Petri plates containing BG 11 medium (Rippka, 1988) with 50.000 penicillin and were incubated at 25 2 C under continuous illumination (cool-white fluorescent, 48 mmol/m2s (2400 l)). The pH of the growth medium was adjusted to 8 by adding diluted (0.01 M) and concentrated (1 M) sulfuric acid or sodium hydroxide solutions. Cells from microcolonies on these plates were isolated by micromanipulation. The microalgal cells were purified at aseptic conditions by streaking the cells repeatedly on the BG 11 medium agar plate. At the final step, the purified microalgal cells were transferred to liquid media. In order to validate the axecinity, these liquid cultures were also tested for bacterial contamination by plating on bacteriological media. A series of batch culture experiments in unshaken flasks illuminated by cool-white fluorescent lamps were carried out at 48 mmol/m2s (2400 lx) light intensity. The microalgal cultures were transferred into 100 mL BG 11 medium in 250 mL Erlenmeyer flasks and incubated at 30 C under continuous illumination for 20 days. For the experiments, to give an initial concentration of about 0.1 g/L dry weight biomass inoculated in culture media.
2.2.
PCR and sequencing
Whole cells from an exponentially growing culture of the isolate were used for 18S rRNA gene amplification. 18S rRNA region was amplified with primers forward F1: 50 WACCTGGTTGATCCTGCCAGT-30 and reverse R1798: 50 GATCCTTCYGCAGGTTCACCTAC-30 are used as described by Luo et al. (2006). PCR reaction is carried out in 50 mL reaction and reaction mix included 0.2 mM of each primer, 0.2 mM of each dNTP, 1.5 mM of MgCl2 and 30 ng of template DNA. Super-HotTaq Taq DNA polymerase (Bioron GmbH, Germany) is used in the amplification. Amplification by PCR technique was carried out by an initial denaturation at 95 C for 10 min, followed by 35 cycles of denaturation at 95 C for 45 s, annealing at 60 C for 45 s, and elongation at 72 C for 45 s, with a final extension of 72 C for 10 min. Four primers are used to sequence the amplified product: 2 forward primers, F371: 50 AGGGTTCGATTCCGGAG-30 and F1132:50 -GAAACTTAAAKGAATTG-30 and 2 reverse primers, R584: 50 -GWATTACCGCGGCKGCTG-30 and R1283: 50 -CGGCCATGCACCACC-30 . BigDye 3.1 (Applied Biosystems Inc, USA) chemistry and ABI 3130 genetic analyzer (Applied Biosystems Inc, USA) is used in sequencing.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5
2.3.
TRIA, NaHCO3 and boron solutions
Stock TRIA (96%, w/v; Aldrich) solution was prepared by dissolving 0.5 g of the chemical in chloroform. Sodium bicarbonate solution (Merck) was prepared by dissolving 17.2 g/L of the NaHCO3 in distilled water. Stock solution of B was prepared by dilution of boric acid (H3BO3) (Carlo Erba) to a final concentration of 10 g/L of B. Appropriate volumes of the stock solutions were added to the BG 11 media.
2.4. Effects of different media compositions on boron removal To determine the effects of the different media compositions on B removal, the microalgae were cultivated in media containing increasing concentrations of B (5, 7.5 and 10 mg/L) in; (1) BG 11 control medium without any contents; (2) BG 11 medium with 1 mg/L TRIA; (3) BG 11 medium with 34 mg/L NaHCO3 and (4) BG 11 medium with 1 mg/L TRIA and 34 mg/L
169
NaHCO3 solutions (Tas¸tan et al., unpublished results) at pH 8 for 20 days of incubation period. For the experiments, 100 ml BG 11 media was inoculated with 0.1 g/L dry weight biomass.
2.5.
Effect of initial pH on boron removal
In order to highlight the effect of pH on B removal process by Chlorella sp., the experiments were also conducted at pH 5, 6, 7, 8 and 9 at 10 mg/L B concentration in BG 11 media with 1 mg/L TRIA and 34 mg/L NaHCO3 for 15 days incubation period. For the experiments, 100 mL BG 11 media was inoculated with 0.1 g/L dry weight biomass.
2.6. Effects of the biomass concentration on boron removal The effect of microalgal biomass concentrations on B removal was also examined at four different biomass concentrations by the fresh wet weight method. The experiments were performed
Fig. 1 e Comparison of the effect of different media compositions on removal yield of different boron concentrations by Chlorella sp. during the incubation period. (TRIA concentration, 1 mg/L; NaHCO3 concentration, 34 mg/L; pH 8; T 25 ± 2 C; illumination, 48 mmol/m2s (2400 l3)).
170
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5
in media with 1 mg/L TRIA and 34 mg/L NaHCO3 at pH 8 for 20 days of incubation period. Four different fresh wet weights of Chlorella sp. biomass were added to each treatment to give an initial biomass concentration of 0.386, 0.450, 0.655 and 1.061 g/L.
2.7.
Analytical methods
During the incubation period, 3 mL samples were taken at 5, 10, 15 and 20 days from each of the flasks. The B concentration was determined by measuring the absorbance at 585 nm with a Shimadzu UV 2001 model spectrophotometer by using carmine as the complexing reagent (Adams, 1990). The percentage removal of B and qm (the maximum specific B uptake) was calculated from equations used before by other researches (Tas¸tan et al., 2010). In the study, qm represents the maximum amount of B removal per unit dry weight of microalgal cells (mg/g), X maximum dried cell mass (g/L), and C0 the initial concentration of the B (mg/L), respectively. Cell growth of Chlorella sp. was determined by measuring optic density, maximum dried cell mass and specific growth rate parameters for any set of growth conditions. Optic density was measured at 600 nm with the spectrophotometer. The maximum dried cell mass was saved by the measurement of the pellets, which were dried at 80 C for overnight (Nu¨ve FN 400 model sterilizator) after centrifugation step (3421 g ¼ 5000 rpm for 100 ). Specific growth rate (m) was calculated
according to the equation m ¼ (ln X2 ln X1)/(t2t1), X2 and X1: dry cell weight concentrations (g/L) at time t2 and t1, respectively (Ip and Chen, 2005). The chlorophyll (a þ b) concentrations were also determined for chlorophyll a at 646.6 nm and chlorophyll b at 663.6 nm and calculated with the method developed by Porra et al. (1989). The chlorophyll concentrations were expressed in mg of chlorophyll per milliliter. All of the experiments were performed in triplicate.
3.
Results and discussion
Pairwise distance analysis carried out by using Mega4 software (Tamura et al., 2004, 2007) revealed that our species is in a close relationship to microalgal species Chlorella zofingiensis at % 97.09, Chlorella ellipsoidea at % 95.76, Chlorella sorokiniana at % 95.62, C. vulgaris at % 95.55 and Chlorella lobophora at % 95.33 identity, for 18S rRNA. As a result of these findings new isolate was identified as Chlorella sp.
3.1. Effect of different media compositions and boron concentrations on removal It is important to detect an interactive effect between media contents and B removal in order to optimize the removal
Fig. 2 e Interactive effect of boron removal yield and optic density (OD600) of Chlorella sp. during the incubation period. (TRIA concentration, 1 mg/L; NaHCO3 concentration, 34 mg/L; pH 8; T 25 ± 2 C; illumination, 48 mmol/m2s (2400 l3)).
171
8.37 27.56 1.74 3.923 0.21 7.08 30.57 1.94 2.580 0.11 5.45 32.95 1.39 1.836 0.15 8.47 26.06 1.85 2.524 0.30 6.28 27.39 1.73 1.834 0.03 Co (mg/L) Removal (%) qm (mg/g)
5.03 23.36 1.31 1.773 0.29
7.05 20.66 0.55 1.694 0.08
7.32 19.88 1.33 1.564 0.02
5.66 24.13 0.83 1.481 0.29
6.64 28.53 0.88 2.147 0.03
6.98 24.24 0.93 2.172 0.13
5.55 29.31 0.28 2.153 0.19
BG 11 þ TRIA þ NaHCO3 BG 11 þ NaHCO3 BG 11 þ TRIA BG 11 without any content
process. Previous studies showed that the initial concentration of B was an important parameter in B bioremoval efficiency that was measured in the culture media (Del-Campo Marı´n and Oron, 2007). Therefore, in the first series of the experiments this parameter was tested at three different B concentrations (5, 7.5 and 10 mg/L), which were designed also for measurement of the effect of different media compositions. In our previous study, we tested microalgal biomass amount was more remarkable in media with TRIA and/or NaHCO3 when compared with biomass amount in media without them (Tas¸tan et al., unpublished results). The aim was to test if TRIA and/or NaHCO3 could increase the removal efficiency by increasing biomass. Therefore, four different BG 11 media compositions were investigated for this purpose: (i) BG 11 without any content (control samples), (ii) BG 11 þ TRIA, (iii) BG 11 þ NaHCO3, (iv) BG 11 þ TRIA þ NaHCO3. The results presented in Fig. 1. show that the percentage of B removal decreased when initial B concentrations were increased from 5 to 10 mg/L. As presented in the Fig. 1aed, all of the removal yields showed significant increases only after 10 days of the incubation period. The removal capacity of Chlorella sp. in BG 11 media without any content is of significance important to comparison its efficiency in media with presence of TRIA and/or NaHCO3. As it can be seen clearly in Fig. 1a, Chlorella sp. removed 13.04% B at 5.03 mg/L concentration within 10 days of incubation period in media without any content. The maximum B removal was observed at the same concentration, i.e., 23.36% within 20 days. The removal capacity of Chlorella sp. was also not affected significantly by TRIA plus in media (Fig. 1b). For example, the highest B removal yield was 24.13% at 5.66 mg/L B in the media with TRIA. These results can be interpreted as a sign of ineffectiveness of TRIA when applied alone in low B concentrations. Boron removal capacity of Chlorella sp. in media with TRIA was higher than yields found in media without TRIA, at higher B concentrations. For example, the removal yield was 19.88% at 7.32 mg/ L B concentration in control media and it increased to 24.24% at 6.98 mg/L B concentration, when TRIA added to media. There are some of the other studies in the literature, describing removal of industrial pollutants in an effective way by including TRIA hormone in BG 11 culture media by freshwater cyanobacterium Synechocystis sp. (Karacakaya et al., 2009). The removing yield was about 22.0% at 58.5 mg/L Remazol Blue dye in media without TRIA and increased to about 25.0% in media with 10 mg/L TRIA. They have explained that increasing in dye removal to be due to TRIA growth hormone. Another observation was the considerable increase in B removal capacity of Chlorella sp. when NaHCO3 added into media. The maximum B removal yield was 29.31% at 5.55 mg/L B concentration within 20 days of incubation period. The removal capacity of Chlorella sp. in media with NaHCO3 was also higher than media with TRIA at higher B concentrations (Fig. 1c). On the other hand, another result, which is more important than above all, was obtained by adding of TRIA and NaHCO3 into culture media together where, enhanced the growth and removal capacity of Chlorella sp. As it can be seen clearly in Fig. 1d, TRIA stimulated all of the parameters
Table 1 e Comparison of the removal yields and qm values at different boron concentrations in different media compositions by Chlorella sp. (Co, boron concentrations; qm, the maximum specific boron uptake; TRIA concentration, 1 mg/L; NaHCO3 concentration, 34 mg/L; incubation period, 20 days; pH 8; T, 25 ± 2 C; illumination, 48 mmol/m2s (2400 l3)).
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5
172
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5
Table 2 e Effect of pH on boron removal and maximum specific boron uptake (qm) of Chlorella sp. (Boron concentration, 10 mg/L; TRIA concentration, 1 mg/L; NaHCO3 concentration, 34 mg/L; incubation period, 15 days; T, 25 ± 2 C; illumination, 48 mmol/m2s (2400 l3)). pH Removal (%) qm (mg/g)
5
6
7
8
9
13.64 0.50 2.429 0.188
15.24 1.33 2.640 0.281
18.71 0.78 2.702 0.215
23.06 2.78 3.268 0.196
13.50 0.43 1.592 0.074
measured in the presence of NaHCO3. Oxidation of the HCO3 to CO2 and its assimilation was increased the primary production in terms of dry weight. Therefore, microorganism tolerated increasing B concentrations easier than all other media contents. Increasing B concentrations up to 8.37 mg/L did not affect the efficiency in terms of yield, and the maximum B removal yield was 32.95% at 5.45 mg/L and was 27.56% at 8.37 mg/L B concentrations within 20 days of incubation period. For these results, further experiments were performed in media with TRIA and NaHCO3. The findings reported by Del-Campo Marı´n and Oron (2007), who used duckweed Lemna gibba in their study, bioremoval of B by L. gibba decreased when its initial concentrations in the media were increased, and the most efficient removal measured at below 2 mg/L B concentrations. They added that there was no removal at about 10 mg/L B concentration even on 12th day of cultivation. In the present study Chlorella sp. showed significant removal efficiency at 10 mg/L B concentration at the four different media compositions tested at 10th days of incubation period. These data demonstrate Chlorella sp. is a good bioaccumulator when compared other aquatic microorganisms. Fig. 2. gives more information about microalgal growth, in order to confirm the stimulation effect of TRIA and/or NaHCO3 and the relationship between algal growth and B removal. As seen in Fig. 2a, Chlorella sp. showed its minimum growth at 5 mg/L B concentration in media without any content (OD600 1.692 at 20th days). The growth of microorganism was increased by adding TRIA and/or NaHCO3 in media. The OD600 was 1.398, 1.484 and 1.693 at 0 mg/L B concentration at 15th days in media with TRIA, media with NaHCO3 and media with TRIA and/or NaHCO3, respectively. When TRIA and NaHCO3 added to media, indicating the presence of stimulators, Chlorella sp. was less affected at 5 mg/L B concentrations due to high growth rates. The comparisons of the maximum amount of B removal per unit dry weight of microalgal cells (qm) is shown in Table 1. In general, the maximum specific B removal values increased with increasing B concentrations up to a certain level due to stimulators. At the lowest B concentration (5 mg/L), the maximum B amount per unit dry weight of Chlorella sp. was 1.773, 1.481, 2.153 and 1.836 mg/g in media without any contents, media with TRIA, media with NaHCO3 and media with TRIA and NaHCO3, respectively. Comparison of qm values for B removal in media with TRIA and/or NaHCO3 and without them showed that there were slight difference, indicating the presence of hormonal stimulation. When B concentration increased to 7.32 mg/L, qm also decreased to 1.564 mg/g in media without any content. The qm values in media with TRIA were found higher than media without TRIA at higher B concentrations due to higher removal yields. In media with
TRIA and NaHCO3, qm increased nearly 2 times (3.923 mg/g) and considerably higher than other media experiments at 8.37 mg/L B concentration. The reason why the qm values in media with TRIA and NaHCO3 were higher than in media without them was due to the much higher growth rates and higher removal yields of Chlorella sp. The media with TRIA and NaHCO3 was the selected media for further experiments because of the results of above mentioned.
3.2.
Effect of initial pH on boron removal
To find a suitable pH for effective B removal by Chlorella sp., trials were performed at media with five different initial pH values (5e9) including nearly 10 mg/L B, and with TRIA and NaHCO3. The effect of pH on B removal after incubation for 15 days is exhibited in Table 2. As shown in the table, B removal was 13.64%, 15.24%, and 18.71% at pH 5, 6 and 7 respectively. The maximum specific B uptakes were 2.429, 2.640 and 2.702 mg/g at the same pH values, respectively. It was observed that the removal yield of B and maximum specific B uptake increased with an increase in pH values up to pH 8. Chlorella sp. removed B with the highest yield of 23.06% and maximum specific B uptake was 3.268 mg/g at pH 8. Further increase in pH to 9 resulted in decrease percentage of removal yield (13.50%) and maximum specific B uptake (1.592 mg/g). As these experiments demonstrated that Chlorella sp. showed its highest B removal yields at pH 8, the following experiments were conducted at this pH value. Some researchers also marked basic pH values for efficient B removal by the macroalgae Caulerpa racemosa var. cylindracea (Bursali et al., 2009) and by the electro coagulation in a batch reactor (Yılmaz et al., 2008).
Fig. 3 e Effect of different biomass concentrations on boron removal by Chlorella sp. at the end of 20 days. (Boron concentration, 10 mg/L; TRIA concentration, 1 mg/L; NaHCO3 concentration, 34 mg/L; pH 8; T 25 ± 2 C; illumination, 48 mmol/m2s (2400 l3)).
173
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5
Table 3 e Results of the removal yields, qm values, chl (aþb) (mg/mL) and m (1/d) at different biomass concentrations (g wet weight/L) by Chlorella sp. (Co, boron concentrations; qm, the maximum specific boron uptake; m, specific growth rate; TRIA concentration, 1 mg/L; NaHCO3 concentration, 34 mg/L; incubation period, 20 days; pH 8; T, 25 ± 2 C; illumination, 48 mmol/ m2s (2400 l3)). Biomass concentration (g wet weight/L) 0.386 0.450 0.655 1.061
3.3.
Co (mg/L) 8.30 8.37 8.93 9.19
qm (mg/g)
Removal (%) 19.59 27.56 35.52 38.03
1.23 1.74 2.36 3.83
Effect of biomass concentrations on boron removal
It was necessary to also determine the effect of biomass concentration for effective B removal. To find a suitable biomass concentration of Chlorella sp., experiments were performed at 10 mg/L B concentrations, in selected optimum media (BG 11 þ TRIA þ NaHCO3) at selected optimum pH level (8), with four different initial biomass concentrations (0.386e1.061 g wet weight/L). It is evident from Fig. 3. that B removal was really linked to the amount of biomass concentrations. The B removal yield was 4.48% at 0.386 g wet weight/L biomass concentration at 5th days of incubation period, and increased to 19.59% at the end of 20 days. When biomass concentration was increased from 0.450 to 0.655 g wet weight/L, the B removal yield was increased from 27.56 to 35.52%. Although 35.52% of 8.93 mg/L B was removed at 0.655 g wet weight/L biomass within 20 days, same removal yield (35.08%) was achieved at 9.19 mg/L B concentration within 15 days by increasing biomass concentration to 1.061 g wet weight/L. It also took only 5 days to remove 9.19 mg/L B with a yield of 18.11% at highest biomass concentration, instead of 20 days to remove 8.30 mg/L B with a yield of 19.59% at lowest biomass concentrations. The highest B removal yield was 38.03% at 1.061 g wet weight/L biomass concentration within 20 days. It is clearly seen in Fig. 3 that the relationship between these variables was linear. This linearity can be interpreted as the indication of the relationship between higher biomass concentrations and higher tolerance capability to B by Chlorella sp. The maximum amount of B removal per unit dry weight of microalgal cells (qm), chlorophyll (a þ b) concentrations and specific growth rate (m) at increasing biomass concentrations after 20 days of incubation period has also been determined and presented in Table 3. The effect of biomass concentrations on qm showed that the microalgal uptake capacity and removal yield of B at 0.450 g wet weight/L biomass concentration was higher than 0. 386 g wet weight/L. The qm was 3.268 mg/g at 8.30 mg/L B concentration at 0.386 g wet weight/L biomass concentration and increased to 4.204 mg/g when the initial biomass concentration increased up to 0.655 g wet weight/L. Although B removal yield was highest (38.03%) at the highest biomass concentration (1.061 g wet weight/L), qm value was lowest (3.174 mg/g) related to the high amount of biomass. When biomass had high amount, the removal of B per 1 g of the dry weight of the microalgae would have low value. The data given in Table 3 also shows the changes in chlorophyll (a þ b) concentrations under effect of a known B
3.268 3.923 4.204 3.174
Chl (a þ b) (mg/mL)
0.66 0.21 0.73 0.13
0.382 0.720 0.925 1.349
0.040 0.034 0.143 0.087
m (1/d) 0.088 0.0046 0.106 0.0023 0.095 0.0092 0.094 0.0035
concentration. The chlorophyll (a þ b) concentrations increased from 0.382 to 1.349 mg/mL in parallel to the increasing biomass concentrations. There was also a little difference between the specific growth rates; lowest m was 0.088 1/d at 0.386 g wet weight/L biomass concentration. As expected, when biomass had fewer amounts at a known concentration of B, specific growth rate would have low. On the other hand the maximum specific growth rate was 0.106 1/d at 0.450 g wet weight/L biomass concentration and it decreased to 0.095 1/d at 0.655 g wet weight/L biomass concentration. The specific growth rates of biomass concentrations between 0.655 and 1.061 g wet weight/L were very similar. But, when compared all of the biomass concentrations, and take into consideration of standard declinations, the initial biomass concentration of 0.655 g wet weight/L was the most suitable one. This biomass concentration had higher B removal yield and specific growth rate with a highest qm value. The reason why highest biomass concentration was not the best one due to the already in place much higher biomass of microalgae in media would not have an excellent growth rate.
4.
Conclusions
The present study describes a remarkable topic that is very popular and hard to study around the world. Boron removal in aqueous solutions by using efficient and economical methods is still a challenging problem. Our effort is contribution to the studies finding a sufficiently economical and efficient method which can be taken as a promising alternative technique offering an applicable solution to the problem. Compared to current physical or chemical B removal techniques, microalgal B removal process can be supported as an easily and is more environment friendly method. The ultimate aim of this study was to investigate if Chlorella sp. could serve as a biomaterial for B removal from water and its potential could be increased by the stimulatory effect of some growth stimulating agents. Increasing biomass concentrations in the presence of TRIA hormone and NaHCO3 at pH 8 resulted more prominent effects on B removal process in our experiments. The maximum B removal yield measured was 38.03% at 9.19 mg/L B concentration in the present study which was the highest yield that was obtained with the use of aquatic organisms, to the best of our knowledge. The results showed that Chlorella sp. can be used as a potential bioaccumulator for B removal process up to high B concentrations. Reports on the use of biomaterial for B removal are limited. Some investigators in fact, have used Chlorella sp. as
174
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5
a biomaterial for removal of some pollutants (Lim et al., 2010; Valderrama et al., 2002), but unfortunately it has not been used in such studies for B bioremoval. The present study not is only the first report that offered Chlorella sp. as an effective biomaterial, but also detected the most efficient way on B removal process, comprehensively.
Acknowledgments Financial support was gratefully acknowledged by the Scien_ ¨ BITAKtific and Technological Research Council of Turkey (TU _ BIDEB). The authors also wish to express their gratitude to reviewers for their valuable comments.
references
Abu-Hamed, T., Karni, J., Epstein, M., 2007. The use of boron for thermochemical storage and distribution of solar energy. Solar Energy 81, 93e101. Adams, V.D., 1990. Water and Wastewater Examination Manual, vol. 247. CRC Press LLC, Florida. Bursali, E.A., Cavas, L., Seki, Y., Bozkurt, S.S., Yurdakoc, M., 2009. Sorption of boron by invasive marine seaweed: Caulerpa racemosa var. cylindracea. Chemical Engineering Journal 150, 385e390. Cervilla, L.M., Blasco, B., Rı´os, J.J., Rosales, M.A., RubioWilhelmi, M.M., Sa´nchez-Rodrı´guez, E., Romero, L., Ruiz, J.M., 2009. Response of nitrogen metabolism to boron toxicity in tomato plants. Plant Biology 11, 671e677. doi:10.1111/j.14388677.2008.00167.x. Chong, M.F., Lee, K.P., Chieng, H.J., Ramli, I.I.S.B., 2009. Removal of boron from ceramic industry wastewater by adsorptione flocculation mechanism using palm oil mill boiler (POMB) bottom ash and polymer. Water Research 43, 3326e3334. Christogerou, A., Kavas, T., Pontikes, Y., Koyas, S., Tabak, Y., Angelopoulos, G.N., 2009. Use of boron wastes in the production of heavy clay ceramics. Ceramics International 35, 447e452. Coughlin, J.R., 1998. Sources of human exposure: overview of water supplies as sources of boron. Biological Trace Element Research 66, 87e100. Crawford, C.L., Marra, J.C., Bibler, N.E., 2007. Glass fabrication and product consistency testing of lanthanide borosilicate glass for plutonium disposition. Journal of Alloys and Compounds 444e445, 569e579. Davis, S.M., Drake, K.D., Maier, K.J., 2002. Toxicity of boron to the duckweed, Spirodella polyrrhiza. Chemosphere 48, 615e620. de-Bashan, L.E., Bashan, Y., 2010. Immobilized microalgae for removing pollutants: review of practical aspects. Bioresource Technology 101, 1611e1627. Del-Campo Marı´n, C.M., Oron, G., 2007. Boron removal by the duckweed Lemna gibba: a potential method for the remediation of boron-polluted waters. Water Research 41, 4579e4584. El-Sheekh, M.M., El-Shouny, W.A., Osman, M.E.H., El-Gammal, E. W.E., 2005. Growth and heavy metals removal efficiency of Nostoc muscorum and Anabaena subcylindrica in sewage and industrial wastewater effluents. Environmental Toxicology and Pharmacology 19, 357e365. Frick, H., 1985. Boron tolerance and accumulation in the duckweed, Lemna minor. Journal of Plant Nutrition 8, 1123e1129.
Gonza´lez, L.E., Can˜izares, R.O., Baena, S., 1997. Efficiency of ammonia and phosphorus removal from a Colombian agroindustrial wastewater by the microalgae Chlorella vulgaris and Scenedesmus dimorphus. Bioresource Technology 60, 259e262. Gunes, A., Soylemezoglu, G., Inal, A., Bagci, E.G., Coban, S., Sahin, O., 2006. Antioxidant and stomatal responses of grapevine (Vitis vinifera L.) to boron toxicity. Scientia Horticulturae 110, 279e284. Hanagata, N., Takeuchi, T., Fukuju, Y., Barnes, D.J., Karube, I., 1992. Tolerance of microalgae to high CO2 and high temperature. Phytochemistry 31, 3345e3348. Houtz, R.L., Ries, S.K., Tolbert, N.E., 1985a. Effect of triacontanol on Chlamydomonas’. Plant Physiology 79, 357e364. Houtz, R.L., Ries, S.K., Tolbert, N.E., 1985b. Effect of triacontanol on Chlamydomonas: II. Specific activity of ribulosebisphosphate carboxylase/oxygenase, ribulose-bisphosphate concentration, and characteristics of photorespiration. Plant Physiology 79, 365e370. Ip, P.-F., Chen, F., 2005. Production of astaxanthin by the green microalga Chlorella zofingiensis in the dark. Process Biochemistry 40, 733e738. Itakura, T., Sasai, R., Itoh, H., 2005. Precipitation recovery of boron from wastewater by hydrothermal mineralization. Water Research 39, 2543e2548. Karacakaya, P., Kılıc¸, N.K., Duygu, E., Do¨nmez, G., 2009. Stimulation of reactive dye removal by cyanobacteria in media containing triacontanol hormone. Journal of Hazardous Materials 172, 1635e1639. Kavak, D., 2009. Removal of boron from aqueous solutions by batch adsorption on calcined alunite using experimental design. Journal of Hazardous Materials 163, 308e314. Lee, J.-S., Kim, D.-K., Lee, J.-P., Park, S.-C., Koh, J.-H., Cho, H.-S., Kim S-, W., 2002. Effects of SO2 and NO on growth of Chlorella sp. KR-1. Bioresource Technology 82, 1e4. Lee, S.H., Kim, W.S., Han, T.H., 2009. Effects of post-harvest foliar boron and calcium applications on subsequent season’s pollen germination and pollen tube growth of pear (Pyrus pyrifolia). Scientia Horticulturae 122, 77e82. Lim, S.-L., Chu, W.-L., Phang, S.-M., 2010. Use of Chlorella vulgaris for bioremediation of textile wastewater. Bioresource Technology 101, 7314e7322. Luo, W., Pflugmacher, S., Pro¨schold, T., Walz, N., Krienitz, L., 2006. Genotype versus phenotype variability in Chlorella and Micractinium (Chlorophyta, Trebouxiophyceae). Protist 157, 315e333. Morioka, A., Sakurai, S., Okuno, K., Sato, S., Verzirov, Y., Kaminaga, A., Nishitani, T., Tamai, H., Kudo, Y., Yoshida, S., Matsukawa, M., 2007. Development of 300 C heat resistant boron-loaded resin for neutron shielding. Journal of Nuclear Materials 367-370, 1085e1089. Nable, R.O., Ban˜uelos, G.S., Paull, J.G., 1997. Boron toxicity. Plant and Soil 193, 181e198. Okay, O., Gu¨c¸lu¨, H., Soner, E., Balkas¸, T., 1985. Boron pollution in the Simav River, Turkey and various methods of boron removal. Water Research 19, 857e862. ¨ ztu¨rk, N., Kavak, D., Ko¨se, T.E., 2008. Boron removal from O aqueous solution by reverse osmosis. Desalination 223, 1e9. Porra, R.J., Thompson, W.A., Kreidemann, P.E., 1989. Determination of accurate extinction coefficients and simultaneous equations for assaying chlorophylls a and b extracted with four different solvents: verification of the concentration of chlorophyll standards by atomic absorption spectroscopy. Biochimica et Biophysica Acta 975, 384e394. Ries, S.K., Houtz, R., 1983. Triacontanol as a plant growth regulator. Horticultural Science 18, 654e662. Rippka, R., 1988. Recognition and identification of cyanobacteria. Methods in Enzymology 167, 28e67. Ruangsomboon, S., Wongrat, L., 2006. Bioaccumulation of cadmium in an experimental aquatic food chain involving
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 6 7 e1 7 5
phytoplankton (Chlorella vulgaris), zooplankton (Moina macrocopa), and the predatory catfish Clarias macrocephalus C. gariepinus. Aquatic Toxicology 78, 15e20. Sasmaz, A., Obek, E., 2009. The accumulation of arsenic, uranium, and boron in Lemna gibba L. exposed to secondary effluents. Ecological Engineering 35, 1564e1567. Sheng, O., Song, S., Peng, S., Deng, X., 2009. The effects of low boron on growth, gas exchange, boron concentration and distribution of ‘Newhall’ navel orange (Citrus sinensis Osb.) plants grafted on two rootstocks. Scientia Horticulturae 121, 278e283. Simonnot, M.O., Castel, C., Nicolaı¨, M., Rosin, C., Sardin, M., Jauffret, H., 2000. Boron removal from drinking water with a boron selective resin: is the treatment really selective? Water Research 34, 109e116. Sung, K.-D., Lee, J.-S., Shin, C.-S., Park, S.-C., Choi, M.-J., 1999. CO2 fixation by Chlorella sp. KR-1 and its cultural characteristics. Bioresource Technology 68, 269e273. Tamura, K., Nei, M., Kumar, S., 2004. Prospects for inferring very large phylogenies by using the neighbor-joining method. Proceedings of the National Academy of Sciences (USA) 101, 11030e11035. Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: molecular evolutionary genetics analysis (MEGA) software version 4.0. Molecular Biology and Evolution 24, 1596e1599. rul, S., Do¨nmez, G., 2010. Effective bioremoval Tas¸tan, B.E., Ertug of reactive dye and heavy metals by Aspergillus versicolor. Bioresource Technology 101, 870e876.
175
Valderrama, L.T., Del Campo, C.M., Rodriguez, C.M., de- Bashan, L.E., Bashan, Y., 2002. Treatment of recalcitrant wastewater from ethanol and citric acid production using the microalga Chlorella vulgaris and the macrophyte Lemna minuscula. Water Research 36, 4185e4192. Waggott, A., 1969. An investigation of the potential problem of increasing boron concentrations in rivers and water courses. Water Research 3, 749e765. Wang, B., Li, Y., Wu, N., Lan, C.Q., 2008. CO2 bio-mitigation using microalgae. Applied Microbiology and Biotechnology 79, 707e718. Wang, H.-M., Pan, J.-L., Chen, C.-Y., Chiu, C.-C., Yang, M.-H., Chang, H.-W., Chang, J.-S., 2010. Identification of anti-lung cancer extract from Chlorella vulgaris C-C by antioxidant property using supercritical carbon dioxide extraction. Process Biochemistry 45, 1865e1872. Wojcik, P., Wojcik, M., Klamkowski, K., 2008. Response of apple trees to boron fertilization under conditions of low soil boron availability. Scientia Horticulturae 116, 58e64. Xu, J., Chen, L., Tan, K.F., Borgna, A., Saeys, M., 2009. Effect of boron on the stability of Ni catalysts during steam methane reforming. Journal of Catalysis 261, 158e165. lu, R., Kocakerim, M.M., Kocadag istan, E., Yılmaz, A.E., Boncukcuog 2008. An empirical model for kinetics of boron removal from boron-containing wastewaters by the electrocoagulation method in a batch reactor. Desalination 230, 288e297.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6
Available online at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Survival dynamics of fecal bacteria in ponds in agricultural watersheds of the Piedmont and Coastal Plain of Georgia Michael B. Jenkins a,*, Dinku M. Endale a, Dwight S. Fisher a, M. Paige Adams b, Richard Lowrance c, G. Larry Newton d, George Vellidis b a
USDA-ARS J. Phil Campbell, Sr., Natural Resource Conservation Center, Watkinsville, GA 30677, USA Department of Biological and Agricultural Engineering, University of Georgia, Tifton, GA 31793, USA c USDA-ARS Southeast Watershed Research Laboratory, Tifton, GA, USA d Department of Animal Science, University of Georgia, Tifton, GA 31793, USA b
article info
abstract
Article history:
Animal agriculture in watersheds produces manure bacteria that may contaminate surface
Received 14 July 2011
waters and put public health at risk. We measured fecal indicator bacteria (commensal
Received in revised form
Escherichia coli and fecal enterococci) and manure pathogens (Salmonella and E. coli 0157:H7),
18 October 2011
and physicalechemical parameters in pond inflow, within pond, pond outflow, and pond
Accepted 21 October 2011
sediments in three ponds in agricultural watersheds. Bishop Pond with perennial inflow
Available online 2 November 2011
and outflow is located in the Piedmont, and Ponds A and C with ephemeral inflow and outflow in the Coastal Plain of Georgia. Bromide and chloride tracer experiments at Bishop
Keywords:
Pond reflected a residence time much greater than that estimated by two models, and
E. coli 0157:H7
indicated that complete mixing within Bishop Pond was never obtained. The long resi-
Fecal indicator bacteria
dence time meant that fecal bacteria were exposed to solar UV-radiation and microbial
Natural disinfection
predation. At Bishop Pond outflow concentrations of fecal indicator bacteria were signifi-
Ponds
cantly less than inflow concentrations; such was not observed at Ponds A and C. Both
Salmonella
Salmonella and E. coli 0157:H7 were measured when concomitant concentrations of
Watersheds
commensal E. coli were below the criterion for surface water impairment indicating problems with the effectiveness of indicator organisms. Bishop Pond improved down stream water quality; whereas, Ponds A and C with ephemeral inflow and outflow and possibly greater nutrient concentrations within the two ponds appeared to be less effective in improving down stream water quality. Published by Elsevier Ltd.
1.
Introduction
Watersheds with dairies, beef cattle, swine, and poultry operations together with wildlife and watersheds with crops receiving manure as a soil amendment are potential nonpoint sources of fecal bacteria and zoonotic pathogens such as Salmonella and Escherichia coli 0157:H7 (Ferguson et al., 2003). Salmonella is known to cause gastroenteritis and is a leading
cause of food related deaths (Mead et al., 1999). Several outbreaks of haemorrhagic colitis and haemolytic uremic syndrome have been caused by E. coli 0157:H7 infections traced to surface water (Olsen et al., 2002), groundwater (Hrudey et al., 2003), and from fresh spinach (Jay et al., 2007). Riparian filter strips have been developed and tested as measures to attenuate the movement of manure-borne microorganisms into surface waters (Coyne et al., 1995;
* Corresponding author. Tel.: þ1 706 769 5631x228; fax: þ1 706 769 8962. E-mail address:
[email protected] (M.B. Jenkins). 0043-1354/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.watres.2011.10.049
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6
Entry et al., 2000). Contamination, nevertheless, occurs as evidenced by the enumeration of Salmonella and E. coli 0157:H7 under base flow conditions in surface waters of an agricultural watershed when concentrations of fecal indicator bacteria, commensal E. coli, and fecal enterococci were below impairment criteria (Jenkins et al., 2008; 2009). Given that an infective dose of Salmonella can be as low as 100 cells (Bitten, 1984), and the infective dose of E. coli 0157:H7 can range between 10 and 100 cells (Jones, 1999), dilute concentrations of these two bacterial pathogens in recreational waters may pose a risk to public health. Analogous to waste stabilization ponds associated with domestic waste water treatment facilities (Davies-Colley et al., 2000), and wet ponds designed to treat urban runoff (Hathaway et al., 2009), ponds in agricultural watersheds may provide mechanisms for inactivating fecal bacteria. Natural mechanisms of disinfection in impoundments in agricultural watersheds would be similar, such as solar radiation (DaviesColley et al., 2000) and predation (Barcina et al., 1997). Jenkins et al. (2011) demonstrated that both insolation and predation were factors in die-off of commensal E. coli and fecal entercocci in an impoundment in a first-order watershed containing beef cattle, wildlife, and cropland fertilized with poultry litter. In a study of paired watersheds with significant agricultural land use and in which one watershed had a greater percentage of ponds than the other, Lowrance et al. (2007) showed a significant reduction in sediment and nutrient loads (NH4eN, NO3eN, total N, inorganic N, ortho P, Cl, and total suspended solids) from the watershed with the greater area of ponds. With the expansion of the interface between urban and agricultural land, the need for abatement of nonpoint sources of fecal pollution is an immediate concern. Fisher et al. (2000) reported that placement of grazing cattle upstream from a pond in the landscape was an effective means of reducing the down stream loads of fecal indicator bacteria. Their observations suggested that ponds situated in watersheds with stream inflows and outflows attenuate the load of fecal indicator bacteria and presumably pathogenic bacteria associated with animal agriculture. Comparison of the efficacy of ponds in pollution abatement is difficult because of site-specific differences in pond size, shape, storm flow characteristics, quality of inflow, and hydraulic loading rate (Thackston et al., 1987; Van Buren et al., 1996; Persson and Wittgren, 2003). Water and contaminant residence times and pond bathymetry are essential hydrologic information for determining inputeoutput relationships within small ponds. Hydrology and hydraulics, representing temporal distribution of the inflows and flow patterns that develop in ponds during events, respectively, are two primary factors influencing residence time (Walker, 1998). The theoretical residence time (nominal retention time) is computed as the ratio of impoundment volume to the discharge rate. The actual travel time distribution is, however, complicated by varying inflow rate, evaporation, seepage, wind, changes in storage, thermal stratification, and bypass flow through the impoundment. Sedimentation and biological factors also affect nutrient abatement. Small impoundments typically are built to fit across an existing creek or stream thus limiting ability for a hydraulically optimal design (Walker, 1998).
177
Most residence time determinations and resulting models are made from an analysis based on assumptions such as steady-state, plug flow, complete mixing, single inlet and outlet, and homogeneous systems, borrowed from principles used to design chemical reactors (Kadlec, 1994; Walker, 1998; Nauman, 2008). Many investigators (Nix, 1985; Kadlec, 1994; Werner and Kadlec, 1996; Walker, 1998; Persson and Wittgren, 2003) have pointed out that steady-state analyses and designs based on these assumptions are not suitable for ponds. Despite such complications the usual method employed to measure residence time is a tracer test whereby a quantity of a conservative tracer is injected at the inlet and concentration is monitored as a function of time at the outlet, until background levels are achieved again; thus, a retention time distribution (RTD) curve is developed. A number of numerical descriptors are then derived from the RTD to which hydraulic efficiency as proxy for pond performance can be correlated. Using such approach on shallow basins 60 to 600,000 m3 in size, Thackston et al. (1987) developed a model (Eq. (1)) that they recommended to estimate the hydraulic efficiency of similar basins in the absence of site-specific data needed to develop a more rigorous model. Tm =T ¼ 0:84 1 eð0:59ðL=WÞÞ
(1)
Where T is the theoretical volumetric resident time computed as V/Q, with V taken as pond volume and Q is flow rate; Tm mean residence time; L is length and W is width. Thackston et al. (1987) referred to the parameter Tm/T as “hydraulic efficiency.” The model neglects wind and depth effects. Konyha et al. (1995) noted that the definition of T implicitly assumes total mixing, and, therefore, gives only an approximation of the residence time where total mixing does not occur. Our objective was to undertake a systematic study of impoundments in watersheds in two physiographic regions of the USA (Piedmont and Coastal Plain) in which animal agriculture occurs, and determine the reduction on the load of fecal indicator bacteria, Salmonella and E. coli 0157:H7. This paper will focus on results of testing the hypothesis that these ponds under baseflow conditions reduce the concentrations and fluxes of fecal indicator bacteria. The relation of fecal indicator bacteria to dynamics of microbial communities and nutrient characteristics were examined. Because exposure time to natural mechanisms of inactivation such as solar UVradiation and microbial predation (Davies et al., 1995; SchulzFademrecht et al., 2008; Jenkins et al., 2011) are important for the elimination of fecal bacteria, retention time was determined at the Piedmont pond to better understand the processes associated with fecal bacterial inactivation.
2.
Materials and methods
2.1.
Study sites
One pond located in the Piedmont and two ponds in the Coastal Plain of Georgia were used in this study. A man-made impoundment, Bishop Pond, located in the Southern
178
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6
Piedmont, has been described in detail (Jenkins et al., 2008). Briefly, it is approximately 1.6 ha, contains approximately 2.4(10)10 L, and is located in a 100 ha first-order watershed consisting of grazed pastures, a cropped field that is amended with poultry litter annually, and a wooded riparian zone with wildlife and from which cattle are excluded. A perennial firstorder stream fed by a series of springs flows into and out of Bishop Pond which captures base and storm flow from approximately 60% of the 100-ha watershed. It has an approximate mean length of 235 m and mean width of 70.5 m (a w3.3 length to width ratio recommended as a minimum ratio for good pond performance e Van Buren et al. (1996)). The deepest part is about 4-m from permanent pool level and occupies an approximately 35 80 m2 area close to the outlet. The bed level then gradually rises toward the edges where it is about 0.4 m from permanent pool level. The pond holds approximately 24(10)10 L at pool level. Both Ponds A and C are man-made impoundments located in a sub-watershed of the Little River known as the University of Georgia Animal and Dairy Science Farm watershed (ADS watershed). Livestock in the 240 ha ADS watershed include 100 head of beef cows and calves. These are pastured in the watershed year round; there is a 250-cow free stall dairy and dairy heifer and dry cow feeding and grazing area. Pond A is approximately 0.7 ha, holds approximately 4.4(10)6 L and is at the edge of a pasture and is accessible to grazing cattle. Pond C is approximately 2.0 ha and holds approximately 2.5(10)6 L. It captures water draining from liquid manure application areas (spray fields), a restored riparian wetland, and paddocks where replacement dairy heifers are fed with limited grazing. Pond inflow and outflow at Bishop Pond were measured with standard structures (Brakensiek et al., 1979) including a 120 V-notch weir at inflow and a 0.46-m H-flume at the outflow. Sensors and data logging equipment were programmed to store average 5-min flow rates. These 5-min flow data were then integrated over time to produce flows at appropriate time intervals. Both Ponds A and C are outfitted with H-flumes and refrigerated ISCO samplers (Teledyne Isco, Inc., Lincoln, NE) to measure and sample outflow.
2.2.
Tracer studies
Two tracer studies were conducted using sodium bromide (NaBr) in summer 2009 and sodium chloride (NaCl) in early spring 2010 to determine residence time at Bishop Pond in the Piedmont. The greater expense of sodium bromide could not be justified and it was replaced with sodium chloride for the second tracer test. Commensal E. coli was applied as a microbial tracer at the spring 2010 tracer experiment.
2.2.1.
Bromide (Br)
In July 2009, 95.0 kg of NaBr equivalent to 73.8 kg of Br- was mixed with 200 L of inflow water in a polyethylene drum and drained through a rubber hose to the discharge point at approximately 0.11 L/s. The release occurred from 10:30 am to 11:00 am. The stream inflow rate at the time of spiking was 1.27 L/s. The spike point was approximately 5-m upstream of where the stream enters the pond. A Sigma 900 Max portable sampler (American Sigma, Loveland CO) was used to collect outflow samples in 24 glass bottles. Initially the sampler was
programmed to collect 150 mL samples at the hour and half past the hour into one 300 mL glass sampling bottle (1-hr sampling). This protocol was used from the initiation of the tracer study (July 14, 10:00 am) until August 6, 9:30 am. The program was then modified to collect 75 mL samples at the hour and half past the hour to composite over each 2-hr interval (2-hr sampling) from August 6, 10:00 am to September 4, 1:30 pm. Then, until 10:00 am on October 2, 75 mL samples were collected every hour and composited over each 4-hr interval (4-hr sampling). The sampler bottle rack containing 24 bottles was transported to the laboratory once all bottles were full. Duplicate subsamples from each bottle were poured into 125 mL specimen cups and kept at room temperature. A Br ion selective electrode sensor (Model WQBR Sensor; Nexsens Technology Inc. Beavercreek, OH) was initially used to measure Br- concentration in the samples. However, the probe was unstable and unreliable and its use was discontinued. Instead, duplicate 20-mL subsamples were taken from the specimen cup samples at 3 to 4 subsamples per day, time-spaced approximately equally, and sent to the USEPA lab in Athens, GA for analysis of Br on a Metrohm Ion Chromatograph (Metrohm IC Systems AG, Herisau, Switzerland). The typical CV between replicates was 2% or less. A total of 194 samples were analyzed for Br.
2.2.2.
Chloride (Cl)
In March 2010, 386 kg of NaCl equivalent to 234 kg of Cl was mixed with 3780 L of inflow water in a polyethylene drum and drained through a rubber hose to the same discharge point as the Br at approximately 0.42 L/s. The release occurred from 11:00 am to 13:30 pm. The stream inflow rate at the time of spiking was 3.68 L/s. The sampler program was similar to that used for bromide sampling except that only the 1-hr (8:00 March 4, to 9:00 March 29, 2010) and 4-hr composite samples (10:00 March 29 to 13:00 April 27, 2010) were used. A total of 481 bottle samples were collected. Subsamples were transferred to 125 mL specimen cups and kept at room temperature. From the specimen cups a total of 245 subsamples in 20-mL viles, made up of approximately 12 samples/day from spike day to March 15, 2010, 6 samples/day to March 29, 2010, and 1 sample/day thereafter through to April 27, 2010, were sent to the University of Georgia’s Soil and Water Analyses Laboratory (Athens-GA) for determination of Cl on an ion chromatography instrument with precision similar to that of the bromide assay.
2.2.3.
E. coli
A wild-type commensal E. coli strain isolated from fresh calf feces (Jenkins et al., 2003) was grown in five 100 mL volumes of Brain Heart Infusion (BHI) broth. Cells were harvested by centrifugation, washed in phosphate buffered saline (PBS), and each volume resuspended in 100 mL PBS. The total of 500 mL of commensal E. coli was poured and mixed into 100 L of pond water from the inflow site. From five subsamples of inoculated pond water E. coli concentrations were measured as previously described (Jenkins et al., 2008). Immediately after the introduction of the chloride tracer, the E. coli (at a load of 1.6 (10)13 cells) were placed in the pond inflow over approximately10 min. The discharge flux of E. coli from Bishop Pond was measured with an automated system that compiled
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6
179
total outflow volume through the flume in 5-min intervals. The total flow for this period was then computed from the 5min flow records. The 5-min flow amount multiplied by the corresponding concentration, assuming linear interpolation of concentration between sampling times, produced the flux for that period. This was repeated for all samplings. Cumulative addition of these fluxes produced the total flux of E. coli through the flume over 664 h (w28 d) of monitoring after which E. coli concentrations were at baseflow concentrations or not detectable.
(Hatch Company, Loveland, CO) and colorimetric techniques (Clesceri et al., 1998). Total suspended solids (TSS) were determined by standard methods (Clesceri et al., 1998). Concentrations of chlorophyll a were determined on subsamples following EPA Method 445.0 (Arar and Collins, 1997 - www.epa.gov/microbes/m450_0pdf e accessed 3/2/11) with fluorescence analysis on a Turner-Designs TD-700 Fluorometer (Turner Designs, Sunnyvale, CA).
2.3.
Analysis for fecal indicator bacteria E. coli and fecal enterococci, and total direct microbial counts have been described (Jenkins et al., 2008). Analyses for Salmonella and E. coli 0157:H7 have been described (Jenkins et al., 2008, 2009; 2011).
Sampling scheme
Inflow and outflow samples were taken at all three ponds. At stations along a transect across each pond both surface (5 cm) and deep (50 cm) samples were taken. Deep samples were taken at the extinction depth of solar UV-radiation. In addition, at the Bishop Pond site, two sites up stream from the inflow site were sampled. The samples at these sites were analyzed for commensal E. coli, fecal enterococci, and pond water chemical parameters (NH4eN, NO3eN, total N, inorganic N, ortho P, Cl, and total suspended solids). For pathogen analysis at the Bishop Pond site, for logistical considerations, water samples were taken at the inflow, and outflow sites, and one within pond surface sample. In addition, Salmonella and E. coli 0157:H7 were sampled at separate months during 2006, 2007 and part of 2008. At Bishop Pond Salmonella was sampled starting in August 2006 and followed by sampling in August and December 2006, February, March, April, May, July, October, and December 2007, February, March, April, May June, July, and August 2008. E. coli 0157:H7 was sampled starting in July 2006, followed by sampling in September, and November 2006, February, May, June, August, September, and November 2007, January, March, April, May June July, and August 2008. At Ponds A and C pathogen analysis for both Salmonella and E. coli 0157:H7 was undertaken after the protocols for each pathogen were established, and because of the many months of no inflow or outflow, within pond samples only were taken. Total direct microbial counts were undertaken at Bishop Pond; because of technical and logistic difficulties, total direct microbial counts were not performed on samples from Ponds A and C. At each within pond sampling site sediment samples were obtained with a small benthic dredge (Mighty Grab Dredge, Ben Meadows, Inc., Janesville, WI) and analyzed for commensal E. coli and fecal enterococci. Samples were taken monthly beginning February 2006 and ending August 2008. Because in- and outflows for Ponds A and C were ephemeral and non-existent for several months, we report results on fecal indicator bacteria and nutrients from those months for which inflow and outflow samples were taken.
2.4.
2.5.
2.6.
Microbiological analysis
Data analysis
Significant differences at P 0.05 between inflow, outflow, and within pond sites of concentrations of microorganisms (natural log-transformed) and nutrients were analyzed with Proc Mixed of SAS (Version 9.1; SAS Institute, Cary, NC) and the program’s repeated-measures option and treating time as the repeated measure (Littell et al., 1996). Means of bacterial concentrations were back transformed and presented as log10 MPN mL1 or L1 for the fecal indicator bacteria and pathogens, respectively. Linear relations between microorganisms and chemical characteristics were analyzed with the regression procedure Proc Reg of SAS (version 9.1).
3.
Results
3.1.
Bishop Pond retention time experiments
3.1.1.
Theoretical and Thackston model-based residence time
Base flow variability of the stream flow feeding Bishop Pond meant that the theoretical and Thackston model-based
Chemical analyses
Total N and P analyses were performed on a Technicon Autoanalyzer II (SEAL Analytical, Mequon, WI) using standard colorimetric techniques (Clesceri et al., 1998). Subsamples were filtered through prewashed Whatman 934AH glass fiber filters and analyzed for NO3eN, NH4eN, dissolved reactive P (ortho-P), and Cl on a Lachet 8000 Flow Injection Analyzer
Fig. 1 e Theoretical residence time (RT) and mean RT based on Thackston’s model as a function of flow rate at Bishop Pond.
180
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6
residence times also varied (Fig. 1). Both show an exponential drop with increasing flow rate. Mean flow rate during the JulyeSeptember 2009 bromide tracer test was 8.7 L s1 and was influenced by storm flows during the latter half of September when a peak of 85 L s1 was observed. For most of July and August base flow was at 3 L s1 or less. The theoretical and Thackston model-based residence times for a 3 L s1 flow are approximately 92 and 67 days, respectively. Mean flow rate during the chloride tracer test in MarcheApril 2010 was approximately 15 L s1 and was influenced by nine storm flows with peaks reaching 70 L s1. Base flow varied from 9 to 13 L s1. The theoretical and Thackston model-based residence times for a 15 L s1 flow are approximately 18 and 13 days, respectively.
3.1.2.
3.1.3.
Chloride tracer test
The outflow Cl concentration for the first 41 days was steady around 8 mg L1, and then decreased during the next 13 days (Fig. 3A). The corresponding 5-min average discharge rates fluctuated over this sampling period (Fig 3b). Base flow stayed around 10 L s1 while 11 storm flows occurred with peak flows ranging from 45 to 80 L s1. Chloride concentration had not decreased to background levels (w4.85 mg L1) by day 54 after spiking. Therefore, a non-linear regression was used to extend estimates of concentration to background levels and determination of Cl mass balance (Fig. 3A). Based on this model the 234 kg of chloride would have exited the pond in 69 days (w1664 h). The outflow Cl concentration (Fig. 3A) showed much scatter for the first 25 days after spiking (w600 h) indicating a spatially non-uniform mixing.
Bromide tracer test
Bromide concentration peaked at 1.41 mg L1 on day 8 then went down uniformly to 0.93 mg L1 by day 64 (Fig. 2A). Because of storms in September and increased discharge (Fig. 2B) Br concentrations dropped to 0.25 mg L1 by day 70 and then to background levels of 0.2 mg L1 by day 74. Cumulative Br mass exiting the pond amounted to 2.0 kg by day 8, 16.9 kg by day 64 (0.266 kg day1 for day 8e64), and 35.8 kg by day 70 (3.16 kg day1 for day 64e70). Total Br mass accounted for in the 76 days after spike when concentration returned to background levels was only 39.5 kg (54.3%) of the 72.8 kg added to the pond.
3.1.4.
Fig. 2 e Bromide tracer experiment: A. Bromide concentrations; B. Total daily discharge.
Fig. 3 e Chloride tracer experiment: A. Chloride concentrations; B. Total daily discharge.
E. coli tracer test
The time between samplings at the outflow after spiking with chloride and E. coli at the inflow ranged from 4 to 48 h and had a mean of 20.4 h. Assuming linear interpolation between sampling times, the flux of E. coli leaving the pond was estimated using 5-min mean flow rates and concentrations. To account for total flux of E. coli exiting Bishop Pond at each sampling time the 5-min fluxes were cumulatively summed. Two sources of E. coli were considered: the E. coli spike and the continuous input from the creek during this experiment. The initial E. coli concentration in the creek water (before spiking)
181
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6
the highest concentrations of total P and chlorophyll. In contrast to the continuous inflow and outflow fluxes at Bishop Pond, inflow and outflow fluxes of Ponds A and C were ephemeral and retention times for them could not be determined. The concentrations of fecal indicator bacteria at Bishop Pond were significantly greater in the upstream and inflow sites than within pond and pond outflow. No differences were observed between surface (5 cm) and deep (50 cm) samples within Bishop Pond. The total direct microbial counts were, contrary to the fecal indicator bacteria, significantly greater within the pond than upstream and pond inflow (Fig. 5). Concentrations of Salmonella within pond and pond outflow were significantly less than inflow concentrations. In contrast, differences between inflow, within pond and outflow concentrations of E. coli 0157:H7 were not observed (Fig. 6). Of the 17 sampling times, Salmonella was detected in pond inflow, within pond, and pond outflow nine, five, and three times, respectively. Of those times of detecting Salmonella, concentrations of commensal E. coli were below the impairment level of 126 cfu 100 mL1 (USEPA, 1986) one time in pond inflow, one time within pond, and three times in pond outflow. Similarly out of 17 sampling times, E. coli 0157:H7 was detected in pond inflow, within pond, and pond outflow nine, eight and six times, respectively. Of those times of detecting E. coli 0157:H7, concentrations of commensal E. coli were below impairment levels three times in pond inflow, eight times within pond, and five times in pond outflfow. In- and outflow fluxes of commensal E. coli, fecal enterococci, and the Pond’s microbial community paralleled their respective concentration patterns (Fig. 7). In the sediments of Bishop Pond concentrations of commensal E. coli were significantly less than concentrations of fecal enterococci (Fig. 8). Concentration of both E. coli and fecal enterococci correlated with pond water nitrate and ammonia concentrations (Table 2). Concentrations of the two pathogens, Salmonella (ranging between below the detection limit of 0.001 and 120 MPN L1) and E. coli 0157:H7 (ranging between below detection limit and 946 MPN L1), were several log10 orders of magnitude less than either of the fecal indicator bacteria (Fig. 6). No differences in concentrations of commensal E. coli and fecal enterococci were observed between pond inflow, within pond and pond outflow of Pond A. In contrast to Pond A, concentrations of commensal E. coli and fecal enterococci within Pond C increased compared to inflow concentrations (Fig. 9). Concentrations of E. coli and fecal enterococci were ten-times greater in Ponds A and C than in Bishop Pond. Differences between shallow and deep samples were not
Fig. 4 e Cumulative mass of E. coli cells in Bishop Pond outflow and hypothetical mass of E. coli remaining in Bishop Pond based on a T99-value of 20 days (Jenkins et al., 2011).
was 200 MPN mL1; this concentration we assumed to continue throughout the period of the experiment. Assuming these two sources, total input after the spike was 1.601(10)13 MPN of E. coli. Total flux of E. coli measured at the outflow 27 days after the spiking event was 3.69(10)10 cells which represented 0.23% of the total influx of E. coli. A comparison of the accumulative load of spiked E. coli exiting Bishop Pond with the load of E. coli remaining in the pond based on a hypothetical 20 days for E. coli to reach 99% inactivation (T99 ¼ 20 days) (Jenkins et al., 2011) indicated a convergence of loads exiting with that remaining in the pond (Fig. 4). Based on the die-off model (T99 ¼ 20 days), at 27 days less than 1% of the spiked load of E. coli would be remaining in the pond. The convergence of the model data and measured accumulative load at the outflow, thus, indicates die-off of E. coli.
3.2. Pond characteristics, fecal indicator bacteria, and pathogens Differences in nutrient and chemical concentrations between Bishop Pond and Ponds A and C were evident (Table 1). With the exception of NO3, Bishop Pond had the lowest concentrations of NH4, total N, ortho P, total P, TSS, and chlorophyll compared to Ponds A and C. Pond C had the highest concentrations of NO3, NH4, total N, and ortho P; whereas, Pond A had
Table 1 e Mean nutrient and chemical concentrations within Bishop Pond (BP), and ponds A and C. The mean of each within pond sampling site is treated as a replicate per pond (n [ 8). Different letters by means of each chemical indicates a difference at P < 0.05. Pond
NO3
NH4
Total N
Ortho P mg L
Bishop Pond A Pond C
0.264b 0.167a 1.712c
0.158a 0.079a 1.418b
0.897a 3.042b 4.315c
Total P
TSS
1
Chlorophyll mg L1
0.009a 0.111b 0.279c
0.051a 0.743c 0.501b
7.57a 118.10c 24.17b
42.4a 1379.3c 120.6b
182
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6
observed at either Pond A or Pond C (Fig. 9). Differences in within pond mean concentrations of Salmonella (ranging between detection limit and 112.3 MPN 100 mL1) were observed between ponds; whereas, no differences in concentrations of E. coli 0157:H7 (ranging between limit of detection and 7.3 MPN 100 mL1), commensal E. coli, and fecal enterococci were observed between Ponds A and C (Fig. 10). Each time Salmonella or E. coli 0157:H7 was detected in Pond A, the concentrations of commensal E. coli were greater than the impairment level. In contrast, at two sampling times at Pond C concentrations of commensal E. coli were below the impairment level when Salmonella was detected. E. coli 0157:H7
Fig. 5 e Mean concentrations (n [ 31) of commensal E. coli (Ec), and fecal enterococci (FE) at log10 MPN 100 mLL1, and total direct microbial counts (TC) at log10 cells mLL1 for each sampling site of stream inflow, within pond, and outflow at Bishop Pond. Different letters above means of each bacterial type and community indicates differences within each community at P £ 0.05.
Fig. 6 e Mean concentrations of Salmonella (n [ 17), E. coli 0157:H7 (Ec0157H7) (n [ 17) at MPN LL1, and commensal E. coli, and fecal enterococci (n [ 31) at log10 MPN 100 mLL1 associated with pond inflow, within pond, and pond outflow at Bishop Pond. Different letters over means between each of the bacterial communities indicate differences at P £ 0.05.
Fig. 7 e Flux rates of commensal E. coli, fecal enterococci, and total direct microbial counts for 32 months.
w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 7 6 e1 8 6
Fig. 8 e Mean sediment concentrations of commensal E. coli and fecal enterococci at Bishop Pond (n [ 30), and ponds A (n [ 24) and C (n [ 27). Means with different letters indicates a difference between organisms and ponds at P £ 0.05.
183
observed was 1.4 mg 1 (Fig. 2A) indicating very little mixing of Br in the pond. The September storms increased the Br flushing rate almost 12-times. In addition, only 53.5% of the added bromide exited the pond after 74 days indicating a resident time greater than the 92 and 67 days based on the theoretical or Thackston model, respectively. Since there were no storms soon after spiking, early flushing of the Br did not occur. Nearly half the added Br stayed trapped in the pond. Substances entering the pond under base flow would be retained for an extended period of time. A ‘well-mixed-vessel’ criterion for determining residence time does not appear to hold for Bishop Pond. Like the Br data, the Cl data indicated that a ‘well-mixedvessel’ criterion for determining residence time did not appear to hold for Bishop Pond. Assuming complete mixing in 24(10)10 L of pond water Cl concentration in the pond would have been approximately 9.79 mg L1 above background Cl concentration which was at 4.85 mg L1. But only 3.1 mg L1 above background was accounted for. This Cl imbalance suggests that the regression model used to extend the estimated concentration was likely invalid. The results of both
Table 2 e Coefficients of determinations (R2) from linear regression analysis of commensal E. coli (Ec), fecal enterococci (FE) against Bishop Pond water chemical parameters: nitrate (Nitr), ammonia (Amm). (N [ 327). Organisms
Variable
R2
Pr > F
Nitr Amm Nitr Amm
0.3096 0.3497 0.2550 0.2738
