ARTICLE pubs.acs.org/est
Evaluation of Contaminant Removal of Reverse Osmosis and Advanced Oxidation in Full-Scale Operation by Combining Passive Sampling with Chemical Analysis and Bioanalytical Tools Beate I. Escher,*,† Michael Lawrence,‡ Miroslava Macova,† Jochen F. Mueller,† Yvan Poussade,§ Cedric Robillot,|| Annalie Roux,|| and Wolfgang Gernjak‡ †
)
The University of Queensland, National Research Centre for Environmental Toxicology (Entox), The University of Queensland, 39 Kessels Road, Brisbane, Queensland 4108, Australia ‡ Advanced Water Management Centre (AWMC), The University of Queensland, St. Lucia, Queensland 4172, Australia § Veolia Water Australia, Brisbane, Queensland 4000, Australia Queensland Manufactured Water Authority (WaterSecure), Brisbane, Queensland 4000, Australia
bS Supporting Information ABSTRACT: Advanced water treatment of secondary treated effluent requires stringent quality control to achieve a water quality suitable for augmenting drinking water supplies. The removal of micropollutants such as pesticides, industrial chemicals, endocrine disrupting chemicals (EDC), pharmaceuticals, and personal care products (PPCP) is paramount. As the concentrations of individual contaminants are typically low, frequent analytical screening is both laborious and costly. We propose and validate an approach for continuous monitoring by applying passive sampling with Empore disks in vessels that were designed to slow down the water flow, and thus uptake kinetics, and ensure that the uptake is only marginally dependent on the chemicals’ physicochemical properties over a relatively narrow molecular size range. This design not only assured integrative sampling over 27 days for a broad range of chemicals but also permitted the use of a suite of bioanalytical tools as sum parameters, representative of mixtures of chemicals with a common mode of toxic action. Bioassays proved to be more sensitive than chemical analysis to assess the removal of organic micropollutants by reverse osmosis, followed by UV/H2O2 treatment, as many individual compounds fell below the quantification limit of chemical analysis, yet still contributed to the observed mixture toxicity. Nonetheless in several cases, the responses in the bioassays were also below their quantification limits and therefore only three bioassays were evaluated here, representing nonspecific toxicity and two specific end points for estrogenicity and photosynthesis inhibition. Chemical analytical techniques were able to quantify 32 pesticides, 62 PCPPs, and 12 EDCs in reverse osmosis concentrate. However, these chemicals could explain only 1% of the nonspecific toxicity in the Microtox assay in the reverse osmosis concentrate and 0.0025% in the treated water. Likewise only 1% of the estrogenic effect in the E-SCREEN could be explained by the quantified EDCs after reverse osmosis. In comparison, >50% of the estrogenic effect can typically be explained in sewage. Herbicidal activity could be fully explained by chemical analysis as the sampling period coincided with an illegal discharge and two herbicides dominated the mixture effect. The mass balance of the reverse osmosis process matched theoretical expectations for both chemical analysis and bioanalytical tools. Overall the investigated treatment train removed >97% estrogenicity, >99% herbicidal activity, and >96% baseline toxicity, confirming the suitability of the treatment train for polishing water for indirect potable reuse. The product water was indistinguishable from local tap water in all three bioassays. This study demonstrates the suitability and robustness of passive sampling linked with bioanalytical tools for semicontinuous monitoring of advanced water treatment with respect to micropollutant removal.
’ INTRODUCTION The presence or absence of organic micropollutants is a major factor in the acceptance of recycled water for indirect potable reuse. In addition to stringent guidelines1 that specify the maximum concentrations for a vast suite of chemicals in recycled water used to augment drinking water supplies, there exists a need for further investigations on unknown or emerging hazards. r 2011 American Chemical Society
Yet, despite the regulatory impetus, more chemical monitoring is unlikely to be the solution as online monitoring of trace levels of Received: December 8, 2010 Accepted: May 13, 2011 Revised: May 11, 2011 Published: May 25, 2011 5387
dx.doi.org/10.1021/es201153k | Environ. Sci. Technol. 2011, 45, 5387–5394
Environmental Science & Technology individual organic parameters is currently not possible, and the number of parameters that can be monitored in a cost-effective manner can never be more than a subset of all possible contaminants. Bioanalytical tools, such as cell based and low complexity bioassays, complement chemical analysis by providing potencyscaled information on groups of micropollutants with a common mode of toxic action. Another advantage of bioassays is that they are quasi sum parameters, and unlike individual chemicals, the effect of the mixture might still be detectable at very low levels despite individual chemicals falling below their quantification limit. Bioanalytical tools have been employed since the 1970s for water quality assessment and for assessing the treatment efficacy and efficiency of wastewater treatment, initially focusing on genotoxicity,2 cytotoxicity,3 and, in the late 1990s, endocrine disrupting effects.4 More recently, the formation of disinfection byproduct during tertiary treatment or drinking water preparation using chlorination or alternative disinfection steps, where halogenated byproduct are formed, have been investigated with bioassays, including the Microtox assay for nonspecific toxicity and various genotoxicity assays.57 Tertiary treatment that combines flocculation, oxidation by ozone followed by biological treatment has been shown to reduce specific toxicity, such as estrogenicity, below the limit of quantification, and reduce nonspecific toxicity substantially.812 Nonetheless, micropollutant levels were still high enough to elicit distinct responses in the bioassays, and it was possible to evaluate the treatment efficiencies of these process steps. Cao et al.13 utilized bioassays to compare ozonation, ultrafiltration, and reverse osmosis and found reverse osmosis to be the most efficient technology to remove genotoxic effects, acute toxicity, and effects caused by binding to the retinoic acid receptor. However, to our knowledge, there is no study that has evaluated the effects of further polishing steps post reverse osmosis (RO) or evaluated a full-scale advanced water treatment plant based on membrane technology. This lack of data can easily be rationalized because effects after RO are mostly below quantification limits, as we have confirmed in our laboratory14 with a test battery composed of six bioassays; the same test battery has also been successfully applied for assessing other tertiary treatment processes.15 The test battery was comprised of bioassays for nonspecific toxicity, photosynthesis inhibition, estrogenicity, neurotoxicity (exemplified by acetylcholinesterase inhibition), genotoxicity, and activation of the arylhydrocarbon receptor. However, as the three latter end points were below the quantification limit in previous studies, in the present work we used the Microtox assay as a measure of nonspecific toxicity,16 the E-SCREEN to assess estrogenicity,17 and the combined algae test to detect the direct inhibition of photosystem II via chlorophyll fluorescence and to assess nonspecific effects via growth rate measurements.18 Representative samples can be collected using passive samplers resulting in a time-averaged concentration over a set period of time.19 Passive sampling has been typically applied in environmental systems (streams receiving effluent, rivers, groundwater, etc.), yet very few studies have employed passive sampling on (waste)water treatment plants20 and fewer have combined it with bioanalytical tools.21 An important consideration for the combination of passive sampling and bioassays is to ensure that the sampling kinetics is relatively uniform for a wide range of chemicals, meaning that the chemicals of interest have similar physicochemical properties and/or the uptake is not
ARTICLE
substantially affected by their properties. The later applies if the flow rate is sufficiently low that sampling kinetics are controlled by the stagnant water layer on the sampler surface. Stagnant flow results in conditions where a wide range of chemicals follow relatively similar uptake kinetics, and uptake is time-integrative over several weeks. This feature can be exploited for applications in continuous monitoring where low sampling rates are acceptable and where it is important to have no gaps in the sampling. The goal of this study was to develop, for the first time, a monitoring system for advanced water treatment plants (AWTP) that combines passive samplers with bioanalytical tools. This study describes two separate deployments such that the potential of combining passive sampling, chemical analysis, and bioassays can be demonstrated. The first deployment directly compared the concentrations of chemicals in composite water samples with the sampled amounts of chemicals in the passive samplers. This allowed the calculation of an apparent sampling rate. If the apparent sampling rate is fairly independent of the hydrophobicity of the individual chemicals constituting the mixtures, bioanalytical tools can be used in a meaningful way, and their results converted to aqueous concentrations. We further evaluated how much of the observed biological effect could be explained by the concentrations of the chemicals identified during chemical analysis. In the second deployment, we evaluated the robustness and reproducibility of passive sampling in combination with bioassays for the assessment of treatment efficiencies.
’ MATERIALS AND METHODS Sampling Site. Eight custom-made marine grade (418) stainless steel vessels were permanently installed as bypasses at representative locations within the treatment train at the Bundamba AWTP, QLD, Australia (see the Supporting Information (SI), Figure SI-1A to C). Vessels were installed immediately before the reverse osmosis (RO feed, ROF), after recombination of all reverse osmosis train outputs (RO permeate, ROP), following advanced oxidation (AOP), and after storage in the output of the treated water tank (Final). In addition, the reverse osmosis concentrate (ROC) was sampled. For control, passive samplers were also deployed in local tap water and Milli-Q water. Passive Sampling. Eight membrane protected styrene divinylbenzene-reverse phase sulfonated Empore extraction disks (ED) in Teflon housings were deployed in each vessel for 9d (ROF and ROC) or 27d (all other locations) using the sampling scheme depicted in the SI, Figure SI-1D. Preliminary experiments were performed to ensure linearity of uptake kinetics during 28d (unpublished results). Details of the experiments are described in the SI, Text SI-1. Composite Water Samples. Nine-day composite water samples were collected continuously using a Watson Marlow 520UN pump at a rate of 1.2 mL/min22 into a conditioned, acetonewashed 20 L glass bottle contained in a refrigerated sampler. Water samples were extracted with solid phase extraction using 60 mg Waters Oasis HLB as described previously for chemical analysis23 and for bioassays.15 Chemical Analysis. 106 Chemicals listed in Table SI-1 in the SI were quantified by Queensland Health Forensic and Scientific Services (QHFSS) using liquid chromatography coupled with tandem mass spectrometry (LC/MS-MS) as described in refs 23 and 24. 5388
dx.doi.org/10.1021/es201153k |Environ. Sci. Technol. 2011, 45, 5387–5394
Environmental Science & Technology
ARTICLE
Table 1. Comparison between the Toxic Equivalent Concentrations TEQw Predicted from Chemical Concentrations and Those Experimentally Determined in the 9d Composite Water Samples (ROC1, ROC2, ROC3)a TEQw
ROC1
ROC2
first 9d composite sample
ROC3
second 9d composite sample
third 9d composite sample
Microtox baseline-TEQchem (mg/L) baseline-TEQbio (mg/L)
0.43 16.98 ( 0.44
0.16 11.95 ( 1.12
0.13 8.01 ( 2.38
% effect explained by target analytes
2.54%
1.34%
1.65%
EEQchem (ng/L)
0.0003
0.027
0.030
EEQbio (ng/L)
3.12 ( 0.47
2.87 ( 0.23
2.89 ( 0.30
% effect explained by target analytes
0.01%
1.17%
1.25%
DEQchem (μg/L) DEQbio (μg/L)
82.55 69.81 ( 4.79
10.83 6.62 ( 0.80
2.13 1.54 ( 0.10
% effect explained by target analytes
118%
163%
138%
E-SCREEN
Photosynthesis inhibition
a
Toxic equivalent concentrations TEQchem in ROC were calculated from the concentrations of chemicals detected with chemical analysis (Table SI-1) and modeled relative potencies (RP, Tables SI-1 and SI-2). The experimentally determined toxic equivalent concentrations TEQbio for the Microtox assay, the E-SCREEN, and the combined algae test (2 h photosynthesis inhibition end point) were derived from measured EC50 values.
Bioanalytical Tools. The Microtox assay and the E-SCREEN were performed as described in ref 15, and the combined algae test with Pseudokirchneriella subcapitata was described in ref 18. Data Evaluation. Effects in all bioassays were expressed as EC50, the concentration required to cause 50% of the maximum effect, and the EC50 values were converted to toxic equivalent concentrations in the bioassay (TEQbio) with eq 1
TEQ bio ¼
EC50 ðreference compoundÞ EC50 ðsampleÞ
according to a procedure described in ref 25. Most of these input parameters have been previously collected for the analytes quantified in this study and were reported by Hawker et al.26 The resulting logDlipw(pH 7) are listed in Table SI-1. Relative potencies of each chemical (RPi) were calculated by dividing the EC50 of the reference chemical by the EC50 of chemical i (eq 418)
ð1Þ
For the Microtox assay, baseline-TEQbio represents the concentration of a virtual baseline toxicant that would elicit the same effect as the mixture of chemicals in the sample.18 In the E-SCREEN the reference compound is 17β-estradiol, hence the TEQ are referred to estradiol equivalent concentrations (EEQ). In the combined algae test, the end point of 2 h inhibition of photosynthesis is represented by the reference compound diuron, and effects are reported as diuron equivalent concentrations (DEQ). The end point of inhibition of algal growth rate is influenced both by specific inhibition of photosynthesis and by nonspecific baseline toxicity; therefore, both baseline-TEQ and DEQ could be calculated for this end point.18 In order to compare the TEQ from bioanalysis and chemical analysis, the EC50-values of individual chemicals were measured in the E-SCREEN test and for the herbicides in the combined algae test. As all compounds are contributing to baseline toxicity and we could not determine the EC50 values of 106 chemicals, the EC50 in Microtox (eq 2) and the growth end point in the combined algae test (eq 3) were predicted from Quantitative Structure Activity Relationships (QSAR) derived for these end points18 logð1=EC50 ðMÞÞ ¼ 0:84 logðDlipw ðpH 7ÞÞ þ 1:69
ð2Þ
logð1=EC50 ðMÞÞ ¼ 0:95 logðDlipw ðpH 7ÞÞ þ 1:16
ð3Þ
The physicochemical parameters required as input parameters in the QSARs, the liposome water distribution ratios logDlipw(pH 7), were calculated from the octanolwater partition coefficients (logKow) of the neutral species and acidity constants pKa
RPi ¼
EC50 ðreference chemicalÞ EC50 ðiÞ
ð4Þ
After calculating RPi for each chemical i, the contributions to the overall mixture toxicity of each compound were added, lending to calculated TEQchem (eq 527) TEQ chem ¼
n
n
∑i-1 TEQ i ¼ ∑i-1 RPi 3 Ci
ð5Þ
Sampling Rates. The TEQ in the EDs (TEQED) in units of μg/ED/d quantified with the bioassays were converted to aqueous TEQ (TEQw) with eq 6 using the average sampling rate Rs (L/ED/d) derived in this study (see the SI, Text SI-2)
TEQ W ¼
TEQ ED RS 3 t
ð6Þ
’ RESULTS AND DISCUSSION Direct Comparison of Chemical Analysis and Bioanalytical Tools for Composite Samples of the Reverse Osmosis Concentrate. In the first sampling campaign, 106 chemicals (32
pesticides, 62 PPCPs, and 12 EDCs) were quantified in three consecutive 9d composite water samples of the ROC. The ratios between the measured concentrations of the three sampling periods for a given chemical were on average 1.3 (max. 5, four outliers). Given the low temporal variability, concentrations were reported as mean values in the SI, Table SI-1. The outliers were peaks of 335 μg/L atrazine and 390 μg/L simazine and their 5389
dx.doi.org/10.1021/es201153k |Environ. Sci. Technol. 2011, 45, 5387–5394
Environmental Science & Technology
ARTICLE
Table 2. Comparison between the Toxic Equivalent Concentration in the Passive Samplers TEQED between Chemical and Biological Analysis in the First Sampling Campaigna TEQED
ROC
ROF
ROP
AOP
blank
baseline-TEQchem (mg/ED/d) baseline-TEQbio (mg/ED/d)
0.0029 0.32 ( 0.12
0.0003 0.20 ( 0.05
0.000006 0.13 ( 0.04
0.000002 0.08 ( 0.02
0.01 ( 0.005
% effect explained by analytes
0.91%
0.15%
0.0047%
0.0025%
Microtox
E-SCREEN EEQchem (pg/ED/d)
0.392
0.007
0.002
0.003
0.003
EEQbio (pg/ED/d)
35.5 ( 37.2
6.4 ( 5.9
0.23 ( 0.20
