Quantification of the electron donating capacity and UV absorbance of dissolved organic matter during ozonation of secondary wastewater effluent by an assay and an automated analyzer

Ozonation of secondary wastewater treatment plant effluent for the abatement of organic micropollutants requires an accurate process control, which can be based on monitoring ozone-induced changes in dissolved organic matter (DOM). This study presents a novel automated analytical system for monitoring changes in the electron donating capacity (EDC) and UV absorbance of DOM during ozonation. In a first step, a quantitative photometric EDC assay was developed based on electron-transfer reactions from phenolic moieties in DOM to an added chemical oxidant, the radical cation of 2,2 ′ -azinobis(3-ethylbenzothiazoline-6-sulfonate) (ABTS ·+ ). The assay is highly sensitive (limit of quantification ∼0.5 mg DOC ·L −1 ) and EDC values of model DOM isolates determined by this assay were in good agreement with values determined previously by mediated electrochemical oxidation (slope = 1.01 ± 0.07, R 2 = 0.98). In a second step, the photometric EDC measurement method was transferred onto an automated fluidic system coupled to a photometer (EDC analyzer). The EDC analyzer was then used to monitor changes in EDC and UV absorbance of secondary wastewater effluent treated with ozone. While both parameters exhibited a dose-dependent decrease, a more pronounced decrease in EDC as compared to UV absorbance was observed at specific ozone doses up to 0.4 mg O 3 ·g DOC −1 . The concentration of 17 α-ethinylestradiol, a phenolic micropollutant with a high ozone reactivity, decreased proportionally to the EDC decrease. In contrast, abatement of less ozone-reactive micropollutants and bromate formation started only after a pronounced initial decrease in EDC. The on-line EDC analyzer presented herein will enable a comprehensive assessment of the combination of EDC and UV absorbance as control parameters for full-scale ozonation. © 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license. ( http://creativecommons.org/licenses/by/4.0/ )


a b s t r a c t
Ozonation of secondary wastewater treatment plant effluent for the abatement of organic micropollutants requires an accurate process control, which can be based on monitoring ozone-induced changes in dissolved organic matter (DOM). This study presents a novel automated analytical system for monitoring changes in the electron donating capacity (EDC) and UV absorbance of DOM during ozonation. In a first step, a quantitative photometric EDC assay was developed based on electron-transfer reactions from phenolic moieties in DOM to an added chemical oxidant, the radical cation of 2,2 -azinobis(3-ethylbenzothiazoline-6-sulfonate) (ABTS ·+ ). The assay is highly sensitive (limit of quantification ∼0.5 mg DOC ·L −1 ) and EDC values of model DOM isolates determined by this assay were in good agreement with values determined previously by mediated electrochemical oxidation (slope = 1.01 ± 0.07, R 2 = 0.98). In a second step, the photometric EDC measurement method was transferred onto an automated fluidic system coupled to a photometer (EDC analyzer). The EDC analyzer was then used to monitor changes in EDC and UV absorbance of secondary wastewater effluent treated with ozone. While both parameters exhibited a dose-dependent decrease, a more pronounced decrease in EDC as compared to UV absorbance was observed at specific ozone doses up to 0.4 mg O 3 ·g DOC −1 . The concentration of
E-mail address: vongunten@eawag.ch (U. von Gunten). 2010 ; Patel et al., 2019 ;Prasse et al., 2015 ;von Gunten, 2018 ). A number of comprehensive tests in pilot- ( Gerrity et al., 2011 ;Huber et al., 2005 ;Knopp et al., 2016 ;Snyder et al., 2006 ;Ternes et al., 2003 ) and full-scale treatment systems Hollender et al., 2009 ;Zimmermann et al., 2011 ) have demonstrated the efficacy and feasibility of ozonation for decreasing the load of organic micropollutants discharged into surface waters. Numerous wastewater treatment plants (WWTPs) in Europe are therefore upgrading their treatment trains to include full-scale ozonation systems ( VSA, 2020 ). However, a key challenge for the operation of these systems remains the real-time control of the ozone dosage to simultaneously meet the treatment objectives and water quality requirements.
Treatment objectives of ozonation are the abatement of selected indicator substances. For example, regulatory frameworks of auhttps://doi.org/10.1016/j.watres.2020.116235 0043-1354/© 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license. ( http://creativecommons.org/licenses/by/4.0/ ) thorities in Switzerland ( Federal Office for the Environment, 2018 ) and California ( California Environmental Protection Agency, 2018 ) define abatement targets for specific indicator substances. The treatment objectives specified by the Swiss Federal Office for the Environment demand an average abatement of 80% for a suite of indicator substances from raw wastewater to WWTP effluent, which requires the application of minimum specific ozone doses . At the same time, water quality requirements limit the upper value for ozone exposures (i.e., ∫ c O3 dt ) due to the formation of undesired oxidation by-products which are not degraded during biological post-treatment ( Hollender et al., 2009 ;Zimmermann et al., 2011 ). In particular, high ozone exposures in bromide-containing waters lead to the undesired formation of bromate, a regulated, potentially carcinogenic compound ( Soltermann et al., 2016 ;Soltermann et al., 2017 ;von Gunten and Hoigné, 1994 ). Furthermore, energy demand for ozone production also incentivizes lower ozone doses ( Katsoyiannis et al., 2011 ). To balance these treatment objectives and water quality requirements, an accurate, real-time determination of the optimal ozone dose based on robust control parameters is critical.
In drinking water treatment, oxidant exposures have been applied successfully as control parameters ( Kaiser et al., 2013 ). During ozonation, organic micropollutants may either be transformed by ozone or hydroxyl radicals which are transiently formed by reactions of ozone with the water matrix ( Buffle et al., 2006a ). The rate at which individual organic micropollutants are transformed is therefore a function of (i) the second-order rate constants of the reactions of the micropollutants with ozone and hydroxyl radical and (ii) the ozone and hydroxyl radical exposures (i.e., ∫ c O3 dt and ∫ c ·OH dt ) ( Buffle et al., 2006b ). While ozone and hydroxyl radical exposures can be determined on-line during drinking-water ozonation ( Kaiser et al., 2013 ), technical limitations of current analytical instrumentation prohibits quantifying the same parameters during ozonation of secondary WWTP effluents due to fast ozone depletion ( Buffle et al., 2006b ).
To circumvent this obstacle in wastewater treatment, surrogate parameters associated with readily observable changes in the properties of dissolved organic matter (DOM) have been suggested as control parameters for ozonation. In this context, ozone-induced transformations of chromophoric moieties in DOM ( Buffle et al., 2006a ;Wenk et al., 2013 ), have received a lot of attention. Ozoneinduced decreases in absorbance at 254 nm were found to correlate with the abatement of different classes of organic micropollutants ( Bahr et al., 2007 ;Nanaboina and Korshin, 2010 ;Wert et al., 2009 ;Wittmer et al., 2015 ) and viruses ( Wolf et al., 2019 ). Therefore, the relative decrease in UV absorbance allows to assess micropollutant abatement on-line and is considered a suitable control parameter for full-scale ozonation Park et al., 2017 ;Pisarenko et al., 2012 ;Stapf et al., 2016 ;Wittmer et al., 2015 ). Recently, empirical models were proposed that additionally incorporate changes in total DOM fluorescence to improve previous process control systems ( Chys et al., 2017 ;Gamage et al., 2013 ;Park et al., 2017 ). However, absorbance values measured at 254 nm primarily predict the content of aromatic carbon in DOM ( Weishaar et al., 2003 ) and thus only provide limited information on the ozone reactivity of DOM. Therefore, the application of an additional, complementary control parameter based on decreases in the content of ozone-reactive DOM moieties would provide a more robust process control system based on two independent parameters.
A promising complementary approach is the quantification of the electron donating capacity (EDC) of DOM, which provides a measure for the phenol content of DOM ( Aeschbacher et al., 2010 ). The EDC is operationally defined as the number of electrons transferred from electron-donating moieties in DOM to a chemical oxidant at a defined reduction potential ( E H ) and pH within a certain reaction time. Previous studies used the radical cation of 2,2 -azino-bis(3-ethylbenzothiazoline-6-sulphonate) (ABTS ·+ ) as chemical oxidant ( Aeschbacher et al., 2012 ;Chon et al., 2015 ;Walpen et al., 2016 ) because (i) ABTS ·+ is reversibly reduced in a pH-independent one-electron transfer and (ii) the standard reduction potential of the redox couple ABTS ·+ /ABTS ( E H 0 = + 0.68 V ( Scott et al., 1993 )) is sufficiently high to allow for the oxidation of phenolic moieties in DOM. Ozonation of DOM samples, including model DOM isolates, surface water and WWTP effluents, resulted in decreasing EDC values with increasing specific ozone doses ( Önnby et al., 2018b ;Wenk et al., 2013 ), consistent with the ozone-induced transformation of phenolic DOM moieties forming e.g., quinones and ring cleavage products via the Criegee mechanism ( Ramseier and von Gunten, 2009 ;Tentscher et al., 2018 ). Recently, the potential application of changes in EDC as control parameter for ozonation was demonstrated in laboratory experiments ( Chon et al., 2015 ). Wastewater effluent samples, which were exposed to increasing ozone doses exhibited pronounced, dosedependent decreases in EDC, which correlated with the abatement of selected organic micropollutants with different ozone reactivities. Taken together, these findings warrant further investigations on the applicability of EDC as control parameter for ozonation. However, these existing methods are not suited for such an application.
EDC values have previously been determined by mediated electrochemical oxidation (MEO) ( Aeschbacher et al., 2010 ). Subsequent development of analytical methods based on size-exclusion chromatography coupled to a post-column reaction with ABTS ·+ ( Chon et al., 2015 ;Önnby et al., 2018b ) and flow-injection analysis ( Önnby et al., 2018b ;Walpen et al., 2016 ) significantly enhanced the sensitivity for EDC quantification, a requirement for analyzing solutions with low DOM contents. While these analytical methods already introduced some degree of automation (i.e., automatic sample injection), they were not designed for a continual, long-term process monitoring. The primary reasons prohibiting the application of these methods for long-term monitoring are (i) the long delay times from sampling to obtaining EDC values, (ii) the requirement of advanced chromatographic or electrochemical instrumentation resulting in high investment costs and frequent maintenance, and (iii) high reagent consumption rates. Recently, a simple, field-deployable, photometric EDC method was presented ( Yuan et al., 2019 ). Even though the method was not automated, the study demonstrated the potential for on-site photometric EDC measurements of DOM in natural and engineered systems.
The objective of this study was to develop an automated EDC analyzer for monitoring relative changes in EDC and the UV absorbance of DOM during ozonation of secondary WWTP effluents. To this end, a photometric EDC assay was developed by adapting existing EDC methods and comprehensively validated. In a subsequent step, this method was implemented on an automated fluidic system, the EDC analyzer. In a third step, the performance of the EDC analyzer was assessed by measuring changes in EDC of a secondary WWTP effluent during ozonation and the resulting decreases in EDC were compared to changes in UV absorbance and the abatement of selected organic micropollutants. In a fourth and final step, the applicability of the combination of EDC and UV absorbance as control parameters for ozonation was evaluated.

Chemicals
A list of the chemicals used in this study (including information on their purity and supplier) is provided in Section S1.1 of the Supplementary Information (SI). All aqueous solutions were prepared in ultra-pure water with a resistivity of > 18.2 M ·cm (Barnstead NanoPure, Thermo Scientific, Switzerland).

Model DOM isolates
Six model DOM isolates were purchased from the International Humic Substances Society (St. Paul, Minnesota, USA): Nordic Lake Reference Fulvic Acid (NLFA, 1R105F), Pony Lake Reference Fulvic Acid (PLFA, 1R109F), Pahokee Peat Standard Humic Acid (PPHAS, 1S103H), Suwannee River II Standard Fulvic Acid (SRFA, 2S101F), Suwannee River II Standard Humic Acid (SRHA, 2S101H), and Suwanee River I Aquatic Natural Organic Matter (SRNOM, 1R101N). These model DOM isolates were selected because their EDC values (i) were determined previously by MEO ( Aeschbacher et al., 2012 ) and (ii) cover a wide range due to the different origins of the DOM samples. Stock solutions (c DOC ∼ = 50 mg DOC L −1 ) of these model DOM isolates were prepared by dissolving the isolates in water while adding sodium hydroxide solution (0.1 M) to maintain a pH of 6.5 ± 0.5.

Secondary WWTP effluent samples
Grab samples were collected at the municipal WWTP Werdhölzli in Zurich, Switzerland, where conventional secondary clarification is followed by ozonation and sand filtration for biological post-treatment (details on the treatment processes are provided in Section S1.2 and Fig. S1 , SI). The wastewater samples were collected from three different points in the treatment process: (i) the ozonation-reactor influent, (ii) the ozonation-reactor effluent, and (iii) the sand-filter effluent. To approximately collect the same wastewater parcel in all three locations, the sampling times for sample (ii) and (iii) accounted for the hydraulic retention time of 28 and 15 min in the ozonation reactor and sand filter, respectively (calculated based on reactor volumes and flow rates). Relevant ozonation process parameters for the week before, during, and after sampling are provided in Figs. S2 and S3 (SI). The grab samples were filtered within two hours of collection using prerinsed cellulose-nitrate membranes (0.45-μm pore size, Sartorius Stedim Biotech, Germany) to remove suspended particles potentially causing abrasion in the valve and syringe pump of the EDC analyzer (see Section 2.6.3 ). Concentrations of DOC, nitrite and ammonium as well as pH and alkalinity of these wastewater samples are provided in Table S1 (SI).

Ozonation of secondary WWTP effluent containing micropollutants
Increasing doses of ozone were added to the ozonation-reactor influent sample to (i) demonstrate the analytical performance of the EDC analyzer with a real, secondary WWTP effluent and (ii) to assess the predictive power of relative changes in EDC and UV for micropollutant abetment. For this purpose, three micropollutants (17 α-ethinylestradiol, bezafibrate, and atrazine) were each added to a separate aliquot of the filtered, secondary WWTP effluent. The starting concentration of each micropollutant was 1 ± 0.05 μM. While this concentration is higher than the concentrations typically observed for these compounds in secondary WWTP effluents, the relative abatement of these compounds was shown to be independent of the initial micropollutant concentration as long as the added micropollutants do not significantly affect the ozone and hydroxyl radical exposures ( Zimmermann et al., 2011 ). At this starting concentration, each micropollutant only contributed slightly to the scavenging of ozone ( < 3%) or hydroxyl radicals ( < 6%) by the sample matrix (see Section S1.5, SI, for the estimation of the scavenging contribution). These micropollutants were selected to span a wide range of reactivities towards ozone ( Table S2 , SI). A fourth aliquot was prepared to which no micropollutants were added. Subsequently, ozone stock solution was added to the four aliquots of the secondary WWTP effluent at increasing specific ozone doses (0-0.9 mg O3 ·mg DOC −1 ) at a temperature of 23 ± 1 °C under stirring. These ozonated samples were then stored at room temperature for > 24 h to guarantee full ozone depletion before analysis. EDC, UV absorbance, as well as bromide and bromate concentrations were quantified in the micropollutant-free aliquot to avoid potential interferences from the presence of added micropollutants. Accounting for the volume of the added micropollutantand ozone-containing solutions, the DOC concentration of this secondary WWTP effluent was diluted to 5.5 mg DOC ·L −1 for these laboratory ozonation experiments and was assumed to remain constant during ozonation ( von Sonntag and von Gunten, 2012 ).

Quantification of EDC values of DOM samples
To quantify the EDC of DOM sample solutions, we developed and validated a photometric EDC assay and subsequently implemented this assay on an automated fluidic system, the EDC analyzer. The reagent and buffer solutions as well as the detailed experimental procedure for the EDC assay and EDC analyzer are described in the following paragraphs.

Reagent and pH buffer solutions for the EDC assay and analyzer
For both, the EDC assay and EDC analyzer, an ABTS-and a chlorine-containing solution and a pH buffer solution were prepared weekly and stored at room temperature protected from light. The ABTS-containing solution was made by dissolving ABTS (1 mM final concentration) in dilute sulfuric acid (7.5 mM, pH = 2.0). The chlorine-containing solution (1 mM final concentration) was prepared by diluting a concentrated hypochlorite stock solution in water. Chlorine concentrations were determined photometrically from alkaline aliquots (pH 11) using the molar absorption coefficient of hypochlorite ε(292 nm) = 359 M −1 cm −1 ( Wang et al., 2012 ). The stability of the chlorine and ABTS stock solutions was confirmed photometrically (data not shown). The pH buffer solution contained 500 mM of phosphate and resulted in a pH of 7.0 in the final reaction mixture of the EDC assay and analyzer.

Experimental procedure of the EDC assay
The EDC assay was adapted from existing photometric methods for quantifying EDC values of DOM ( Walpen et al., 2016 ; The EDC assay consisted of four steps: First, an ABTS ·+ -containing reagent solution was prepared by adding a chlorine solution to an ABTS-containing solution, resulting in the oxidation of ABTS to ABTS ·+ . Second, the absorbance of the undiluted DOM sample was measured at 254 nm. Third, a phosphate buffer solution and an aliquot of the ABTS ·+ -reagent solution were mixed with the DOM sample solution in a disposable cuvette. Fourth, the absorbance at 728 nm was measured after a reaction time of 15 min. (b) On the EDC analyzer, the same analytical procedure is executed automatically by a valve and syringe-pump system coupled to a photometer. Different solutions were mixed by sequentially aspirating them via a selector valve into the syringe and subsequently circulating the resulting mixture to and from a mixing chamber. The ABTS ·+ -containing reagent solution was temporarily stored in one of the two mixing chambers. Absorbance values at 255 and 730 nm were quantified by dispensing the DOM sample or the reaction mixture from the syringe to a double-wavelength LED-photometer. Yuan et al., 2019 ). For the EDC assay presented in this study ( Fig. 1 a ), the ABTS ·+ -reagent solution was first prepared in bulk by adding 350 μL of chlorine solution (1 mM) per milliliter of ABTS solution (1 mM) which oxidized nominally 70% of the ABTS to ABTS ·+ (i.e., c ABT S ·+ / ( c ABT S ·+ + c ABTS ) = 70% ). Second, the UV absorbance of an undiluted DOM sample was measured at 254 nm as described above. Third, 2500 μL of a DOM sample (see Section 3.1.5 for the final DOC concentration) or DOM-free blank solution and 160 μL of pH buffer solution were added to a disposable cuvette with a 10-mm pathlength (4.5 mL, poly(methyl methacrylate), Brand, Germany). Subsequently, 540 μL of ABTS ·+reagent solution was added to the buffered sam ple solution. Fourth, the resulting absorbance at 728 nm was measured on a spectrophotometer after a reaction time of 15 min unless stated otherwise (see Section 3.1.2 for a discussion on the reaction time).

EDC analyzer
To automate liquid handling and absorbance measurements of the EDC assay, we developed a fluidic system coupled to a photometer ( Fig. 1 b ). The fluidic system consisted of a syringe pump (Level-3 Cadent 3, IMI Precision, Switzerland) equipped with a single 2.5-mL zero-dead-volume syringe and a 12-way rotary selection valve (IMI Precision, Switzerland) as well as two mixing chambers which were connected to the selection valve. Both mixing chambers were built from vertically aligned 2.5-mL pipette tips ( Amornthammarong et al., 2010 ). The detector was a doublewavelength LED photometer (Mikron 31, Runge, Germany) with a 10-mm pathlength and two LED-light source modules that emitted at wavelengths of 255 and 730 nm, respectively. The fixed wavelength of the LED modules led to a minor shift in the detection wavelength compared to the EDC assay and no correction was made for the molar absorption coefficient.
The EDC analyzer was programmed to replicate the four experimental steps of the EDC assay in a measurement cycle ( Fig. 1 b ), during which one DOM-free blank and two DOM-containing samples were analyzed. This allows to determine the relative change in EDC and UV absorbance during ozonation (i.e., analysis of a DOM sample from (i) the ozonation-reactor influent and (ii) from the ozonation-reactor effluent). A measurement cycle consisted of the following steps: First, the ABTS •+ -reagent solution was prepared by sequentially aspirating chlorine-and ABTS-containing solutions into the syringe. The resulting solution was then cycled five times from the syringe to the first mixing chamber and back. The mixing chamber was filled from the top and emptied from the bottom, resulting in an inversion of the syringe content, allowing for an efficient mixing ( Amornthammarong et al., 2010 ). After a total of five mixing cycles, the entire ABTS ·+ -reagent solution was dispensed again to the mixing chamber for temporary storage. Second, the DOM-free blank solution was aspirated into the syringe and a fraction of this blank solution was dispensed to the photometer to zero the absorbance signals at both wavelengths. Third, buffer solution and an aliquot of the ABTS ·+ -reagent solution stored in the first mixing chamber were additionally aspirated into the syringe which still contained DOM-free blank solution. The syringe content was then cycled to and from the second mixing chamber. Fourth, the resulting mixture was passed through the photometer to measure the absorbance at 730 nm after a reaction time of 15 min. Subsequently, the steps two to four were repeated for the two DOM-containing samples to complete one measurement cycle. However, instead of zeroing the absorbance signal during the second step, the absorbance of the DOM samples was measured at 255 nm.
The EDC analyzer was controlled with a computer through an RS-232 communication line. Serial commands and system status were automatically sent and received, respectively, using a script written in Python (Version 3.7, Van Rossum and Drake, 1995 ). An additional Python script was used for data acquisition from the photometer. Both Python scripts are available at https://github. com/walpen/edc _ analyzer .

Calculation of EDC from absorbance values
The EDC of a DOM sample (i.e., the number of electrons transferred from DOM to ABTS ·+ ) was calculated based on the reductive loss of ABTS ·+ in the reaction mixtures. We confirmed that ABTS ·+ was primarily lost through reduction to ABTS by DOM in oneelectron transfer reactions using SRFA as model DOM (see Section S1.6, SI). ABTS ·+ concentrations were quantified based on measured absorbances at 728 or 730 nm using the molar absorption coefficient ε(728 nm, pH 7) = 14 0 0 0 M −1 cm −1 ( Walpen et al., 2016 ). To calculate the EDC, the residual ABTS ·+ concentration in the reaction mixture containing DOM was compared to that of a reaction mixture containing only the DOM-free blank (i.e., ultrapure water) and normalized to the DOC concentration of the sample in the reaction mixture ( Eq. (1) 2) ).

Effect of pH on the oxidation of ABTS by chlorine
The rate of oxidation of ABTS to ABTS ·+ by chlorine decreases with increasing pH ( Pinkernell et al.,20 0 0 ). Because a fast, complete and reproducible preparation of ABTS ·+ is critical for both methods presented herein, the EDC assay and the EDC analyzer, the kinetics and the stoichiometry of the formation of ABTS ·+ was assessed photometrically in the pH range from 1 to 4. At a pH of 1.5 and 2.0, the reaction was completed within one minute ( Fig.   S4 , SI) and the formed ABTS ·+ subsequently remained stable for at least one hour ( Fig. S5 , SI). At pH values > 2.0, the reaction was slower and incomplete ( Fig. S4 , SI). At pH 1.0, the reaction was fast but incomplete, presumably due a side reaction involving Cl 2 ( Deborde and von Gunten, 2008 ). Therefore, ABTS ·+ -reagent solutions for all subsequent analyses were prepared by adding chlorine solution to an ABTS solution acidified with sulfuric acid resulting in a pH of 2.0 in the ABTS ·+ -reagent solution.

Reaction time of the EDC assay
To determine a suitable reaction time for the EDC assay, the evolution of the ABTS ·+ concentration in EDC assay reaction mixtures with SRFA as a model DOM was assessed. For this purpose, the same volume of ABTS ·+ reagent solution was added to a set of SRFA-containing solutions with increasing DOC concentrations and subsequently the resulting change in absorbance at 728 nm was measured over the course of 30 min. The absorption coefficient and hence the ABTS ·+ concentration of the DOM-free reaction mixture (i.e., c DOC = 0 mg DOC L −1 ) decreased slightly by 2.7% during the experiment ( Fig. 2 a ), consistent with the previously observed instability of ABTS ·+ at neutral pH conditions (e.g., Cano et al., 1998 ). Compared to this DOM-free reaction mixture, the decrease in ABTS ·+ concentration of the reaction mixture containing SRFA was more pronounced and bi-phasic. The ABTS ·+ concentration first decreased in an exponential and subsequently after approximately 15 min in a more linear fashion.
For every time point, the ABTS ·+ concentration correlated linearly with the DOC concentration (shown for t = 15 min, inset of Fig. 2 b ). According to eq. (2) , the absolute values of the slopes of these linear regression models represent the loss in ABTS ·+ concentration per DOC concentration and therefore the EDC. The EDC increased rapidly up to approximately 15 min and subsequently continued to increase at a lower rate ( Fig. 2 b ). However, no plateau was reached even for longer reaction times of up to 120 min (data not shown). This finding may be attributed to a presumably wide range of reactivities of electron-donating moieties in DOM towards ABTS ·+ . In addition, oxidation of phenolic moieties forms phenoxyl radicals. These radical species can form coupling products which may also donate electrons to ABTS ·+ ( Hotta et al., 2002, Hotta et al., 2001. As a consequence, a constant value may only be reached after much longer reaction times. Because of this lack of an obvious end point, an arbitrary time point of t = 15 min was selected for the quantification of EDC values for all subsequent analyses. This operationally defined reaction time of 15 min encompasses the rapid changes in ABTS ·+ concentration during the first phase mentioned above. In addition, EDC values obtained for a set of model DOM isolate using the EDC assay with a 15minute reaction time were in good quantitative agreement (see Section 3.1.6 ) to previously published EDC values quantified by MEO, where an electrochemical equilibrium was reached within approximately 60 min ( Aeschbacher et al., 2012 ).

Interferences from selected dissolved inorganic species
Potential interferences of four inorganic species at concentrations relevant for wastewater on the resulting ABTS ·+ concentration in the EDC assay were tested: manganese(II) (0-110 μg ·L -1 ), iodide (0-125 μg ·L -1 ), bromide (0-800 μg ·L -1 ), and nitrite (0-4.6 mg ·L -1 ; i.e., 0-1.4 mg-N ·L -1 ). Among these inorganic species and in the absence of DOM, only iodide affected the ABTS ·+ concentration ( Fig. S6 , SI). Iodide had an EDC value of 0.5 ± 0.2 mol -·mol iodide -1 after a reaction time of 15 min, suggesting a reduction of ABTS ·+ by iodide. However, for DOM samples with DOC concentrations expected in the context of ozonation of secondary WWTP effluents (i.e., 3-10 mg DOC ·L -1 ), the contribution of iodide to the measured total EDC would be negligible compared to the contribution of DOM . For example, for an SRFA solution of 3 mg DOC ·L -1 (EDC SRFA = 5.14 mmol e-·g C -1 ), a concentration of 125 μg ·L -1 iodide would increase the EDC of this SRFA solution by only 4%. In contrast, manganese(II) and nitrite did not react with ABTS ·+ in the absence of SRFA ( Fig. S6 , SI). However, in the presence of SRFA, increasing manganese(II) and nitrite concentrations resulted in increasing EDC values of SRFA ( Fig. S7 , SI). At the highest concentrations tested, i.e., at 110 μg ·L -1 manganese(II) and 4.6 mg ·L -1 nitrite, the measured EDC values of SRFA in the presence of these ions were elevated by 6% and 3%, respectively as compared to the EDC measured for SRFA without these ions. Iodide and bromide did not exhibit such effects. Taken together, these four investigated inorganic species will have a negligible effect on the EDC quantification if present below the tested concentrations.
We note that iron(II) associated with organic matter is oxidized in a one-electron transfer by ABTS ·+ and therefore contributes to the measured EDC by 1 mol e-·mol iron(II) ( Lau et al., 2015 ). In cases where iron(II) is expected to be present in significant amounts (e.g., DOM sampled from anoxic environments), the contribution of iron(II) to the EDC may be accounted for by separately quantifying the iron(II) concentrations in sample aliquots. In secondary WWTP effluents, iron(II) is not expected to reach significant concentrations due to preceding oxic treatments.

Effects of the reduction potential E H and initial ABTS ·+ concentration on the measured EDC
The effects of the apparent reduction potential ( E H ) and the initial ABTS ·+ concentration on the measured EDC were systematically assessed in a series of EDC assays using SRFA as a model DOM. The E H in the reaction mixtures is governed by the ABTS ·+ :ABTS concentration ratio according to the Nernst equation (e.g. Schwarzenbach et al., 2003 , chap. 14). To assess the dependency of the measured EDC on the initial E H , a series of EDC assays of SRFA with identical concentrations of ABTS ·+ , but increasing concentrations of ABTS were conducted. The calculated initial E H values of these solutions ranged from + 643 mV to + 730 mV (see Table S3 , SI). All reaction mixtures containing DOM-free blanks had similar residual ABTS ·+ concentrations after a reaction time of 15 min (78.2 ± 1.2 μM, n = 8, mean ± standard deviation; Fig. S8a , SI). However, the EDC of SRFA determined in these reaction mixtures increased from 3.15 mmol e − ·g C −1 at an initial E H of + 643 mV to 5.48 mmol e − ·g C −1 at an initial E H of + 730 mV ( Fig. S8b , SI). We attribute this effect to the broad distribution of apparent standard reduction potentials of phenolic or other electron-donating moieties present in DOM ( Aeschbacher et al., 2012 ) and the resulting increasing thermodynamic feasibility or faster kinetics of the oxidation of these moieties by ABTS ·+ with increasing E H . These findings are consistent with a previous study in which the EDC of DOM quantified by mediated electrochemical oxidation increased with increasing E H applied to the working electrode of the electrochemical analysis cell ( Aeschbacher et al., 2012 ).
Overall, these findings suggest that the initial E H and ABTS ·+ concentration of the reagent solution have strong effects on the reaction kinetics or thermodynamics of the reduction of ABTS ·+ by electron-donating moieties in DOM. To obtain reproducible EDC values, initial ABTS ·+ concentrations have to be controlled accurately and changes in E H over the course of the reaction should be minimized. Similar to a previous study ( Walpen et al., 2016 ), we selected an initial E H of + 0.71 V for the EDC assay because this E H is slightly above the standard reduction potential, E H 0 , of the redox couple ABTS ·+ :ABTS (i.e., E H 0 = + 0.68 V; Scott et al., 1993 ) thereby ensuring only small decreases in E H over the course of the reaction (i.e., decreases in the ABTS ·+ :ABTS concentration ratio lead to smallest E H decreases around the E H 0 ). A concentration of 87.5 μM was selected as initial ABTS ·+ concentration, which yielded an absorption coefficient of 1.23 cm −1 at 728 nm.

Working range of the EDC assay
To assess the DOC concentration range that can be applied in the EDC assay, the upper limit of the working range and the limit of quantification (LOQ) were determined for two model DOM isolates, NLFA and PLFA. These isolates were selected because their EDC values were previously found to be high and low, respectively, compared to EDC values of other isolates. As expected, increasing DOC concentrations of both NLFA and PLFA resulted in decreasing absorption coefficients at 728 nm and thus decreasing ABTS ·+ concentrations ( Fig. 3 a ). The measured decreases followed linear trends at low DOC concentrations but deviated from linearity at higher DOC concentrations towards higher ABTS ·+ concentrations. This deviation likely resulted from decreases in E H over the course of the reactions, which become more pronounced as the reacting DOC concentrations increased (see discussion above). To determine the upper limit of the linear range for NLFA and PLFA, we fitted linear regression models to the corresponding ABTS ·+ concentrations ( Fig. 3 a ) and sequentially removed data points with the highest DOC concentration until the R 2 -value of the model rose above a predefined threshold of 0.995. The upper limits of the linear ranges, i.e., the largest DOC concentrations of the models to first fulfill this condition (solid lines in Fig. 3 a ), were 3.1 and 7.8 mg DOC ·L −1 for NLFA or PLFA, respectively. The limit of quantification (LOQ = a Blank − 10 ·σ a , Blank ( Currie, 1999 )) corresponded to a DOC concentration of 0.11 and 0.44 mg DOC ·L −1 for NLFA and PLFA, respectively. LOQ values are similar to those determined for the recently published methods for EDC quantification based on size-exclusion chromatography and flow-injection analysis ( Önnby et al., 2018b ;Walpen et al., 2016 ). Both of these methods were previously used to quantify the change in EDC of real wastewater samples upon chemical oxidation with chlorine and ozone. Taken together, we recommend diluting DOM-sample solutions to DOC concentrations between 0.5 and 3.0 mg DOC ·L −1 in the final reaction mixtures for routine analysis with the EDC assay.

Comparison of the EDC assay and MEO
To compare results obtained by the EDC assay to previously published values determined by MEO ( Aeschbacher et al., 2012 ), six model DOM isolates were analyzed. For every isolate, a series of reaction mixtures containing increasing DOC concentrations ( n ≥ 3) was prepared and linear regression models were fitted to the residual ABTS ·+ concentrations versus DOC concentration ( Fig.  S10 , SI). Overall, the resulting EDC values ( Fig. 3 b ) closely matched the EDC values obtained for the same isolates determined by MEO (slope = 1.01 ± 0.07, intercept = −0.1 ± 0.4, R 2 = 0.98).

EDC quantification of a SRFA solution
An EDC analyzer was developed to automate liquid handling and absorbance measurements using a combined valve and syringe-pump system. Exemplary absorbance traces ( λ 1 = 255 nm; λ 2 = 730 nm) over the course of one measurement cycle for the analysis of SRFA are shown in Fig. S11 (SI). The UV absorption coefficient a 255nm of the tested SRFA solution was 0.181 ± 0.001 cm −1 ( n = 2). Normalizing this value by the DOC concentration yielded a specific UV absorbance (SUVA 255 ) of 4.63 ± 0.02 L ·(mg DOC ·m) −1 , which is in good agreement with a previously determined value of 4.4 L ·(mg DOC ·m) −1 ( Walpen et al., 2016 ). As expected from the EDC assay, the addition of the ABTS ·+ reagent to the SRFA solution resulted in a lower absorbance at 730 nm after a reaction time of 15 min compared to the DOM-free blank sample. This is caused by the electron transfer reactions from DOM to ABTS ·+ . The decrease in the absorption coefficient of 0.1686 ± 0.0 0 05 cm −1 ( n = 2) corresponded to a loss of 12.05 ± 0.04 μM ABTS ·+ . Normalizing this loss in ABTS ·+ concentration by the DOC concentration yielded an EDC value of 4.8 ± 0.2 mmol e − ·g C −1 which is only slightly smaller than the value obtained from the EDC assay (i.e., 5.1 ± 0.1 mmol e − ·g C −1 ). A possible explanation for this small deviation as well as a systematic comparison of this value to previously published values to validate the EDC analyzer is provided in Section 3.2.2 . The total duration of the measurement cycle was 74 min, during which the UV absorbance and the EDC values of two DOM samples were determined. We note that by optimizing the analysis sequence (e.g., by reducing the number of mixing cycles and rinsing steps, or by shortening of the reaction time), the analysis time for two samples may be shortened. However, since we considered the current analysis rate to be sufficient for detecting changes in sluggish systems such as WWTPs, we did not work towards shortening the duration of the measurement cycles.

Performance assessment of the EDC analyzer
The performance of the EDC analyzer was assessed by testing the repeatability of EDC and UV absorbance measurements of a single SRFA solution over a total of seven measurement cycles. The relative standard deviations of the EDC and UV absorption coefficients were both small with 1.9% ( n = 14, Fig. 4 a ) and 0.003% ( n = 14, Fig. S12a , SI), respectively.
Similar to the EDC assay, EDC values of six model DOM isolates were quantified using the EDC analyzer and compared to previously published EDC values of the same isolates obtained by MEO ( Aeschbacher et al., 2012 ). The EDC values obtained using the EDC analyzer correlated linearly with the EDC values obtained by MEO ( R 2 = 0.98, Fig. 4 b ) with a negative intercept of −0.4 ± 0.3 and a slope of 0.95 ± 0.06. These overall slightly lower EDC values obtained by the analyzer compared to those obtained by MEO and the EDC assay (see Fig. 3 b ) presumably reflect a slower mixing of the DOM sample and ABTS ·+ reagent solution in the automated system resulting in a shorter reaction time in the EDC analyzer compared to the EDC assay. In the analyzer, solutions are sequentially aspirated into the syringe and subsequently delivered to a mixing chamber. Therefore, the solutions did not mix immediately due to differences in their densities. In contrast, solutions in the EDC assay are added to a cuvette and immediately mixed manually. The slight offset in the EDC values between the measurements may be eliminated by increasing the reaction time on the EDC analyzer. However, since the subsequent discussion is based on relative changes in EDC during ozonation, small offsets in absolute values are not expected to affect the data interpretation. Table 1 Electron donating capacity (EDC) and absorption coefficient ( a 255nm ) of the three wastewater samples collected from (i) the ozonation-reactor influent, (ii) the ozonation-reactor effluent, and (iii) the post-ozonation sand-filter effluent of the Werdhölzli wastewater treatment plant in Zurich, Switzerland. The ozonation-reactor influent sample had a DOC concentration of 6.4 mg DOC ·L − 1 and an estimated specific ozone dose of 0.6 ± 0.06 mg O3 ·mg DOC -1 was applied in the ozonation reactor at the time of sampling.

Sampling Point
EDC a 255nm (mmol e-·g C −1 ) (cm −1 ) Ozonation-reactor influent 3.63 ± 0.08 0.115 ± 0.001 Ozonation-reactor effluent 1.39 ± 0.03 0.062 ±0.001 Sand-filter effluent 1.29 ± 0.03 0.060 ± 0.001 The accuracy of the UV measurements was tested by comparing the results obtained using the EDC analyzer to UV absorption coefficients obtained using a spectrophotometer ( Fig. S12b , SI). The linear regression model fitted to the data had a slope of 1.05 ± 0.01 and an intercept which was not significantly different from zero. This small systematic bias was considered negligible and is not expected to affect relative changes in the UV absorbance.

Characterization of in-and effluent of an ozonation reactor by the EDC analyzer
To demonstrate the performance of the EDC analyzer with samples containing real DOM, filtered samples collected from the WWTP Werdhölzli in Zurich, Switzerland, were analyzed. The EDC of DOM from the ozonation-reactor influent was 3.68 ± 0.08 mmol e-·g C -1 (see Table 1 ) and decreased by 62% to 1.39 ± 0.03 mmol e-·g C -1 after passing the ozonation reactor with an estimated specific ozone dose of 0.6 ± 0.06 mg O3 ·mg DOC -1 (based on the measured DOC concentration and plant process data, Fig.  S3 , SI). This pronounced decrease in EDC reflects the oxidation of electron-donating moieties in DOM (primarily phenols) and is in good agreement with previous findings ( Chon et al., 2015 ;Önnby et al., 2018b ;Wenk et al., 2013 ). The sand filter only had a minor effect on the EDC which decreased by an additional 3% to 1.29 ± 0.03 mmol e-·g C -1 . The UV absorption coefficient a 255nm of Relative changes in the (a) electron donating capacity ( EDC · EDC −1 0 ) and absorbance at 255 nm ( A 255 nm · A −1 255 nm, 0 ) of a secondary WWTP effluent exposed to increasing specific ozone doses compared to the untreated sample. (b) The relative decrease in UV absorbance ( A 255 nm · A −1 255 nm, 0 ) , y-axis) from panel a replotted as a function of the relative decrease in EDC ( EDC · EDC −1 0 , x-axis). Residual concentrations (c) of 17 α-ethinylestradiol (purple, filled circles), bezafibrate (orange, filled squares), and atrazine (green, filled diamonds) in the same wastewater relative to their concentrations in the untreated sample. (d) Increase in bromate concentration in the same samples. The wastewater sample was collected from the ozonation-reactor influent at the Werdhölzli wastewater treatment plant (Zurich, Switzerland) and had a pH of 7.9 before ozone addition. The addition of ozone stock solution diluted the DOC concentration and alkalinity to 5.5 mg DOC ·L −1 and 3.2 mM HCO − 3 , respectively. The solid lines in panels (a-d) represent local polynomial regression models as a guide for the eye. the ozonation-reactor influent sample was 0.115 ± 0.001 cm −1 and decreased by 44% to a value of 0.062 ± 0.001 cm −1 after passing the ozonation reactor and further decreased by 2% by passing the sand filter to 0.060 ± 0.001 cm −1 ( Table 1 ; for full UV spectra see Fig. S13 , SI). These findings clearly demonstrate that the high sensitivity of the EDC analyzer allows to detect the relative decrease in EDC and UV absorbance of real, secondary-treated wastewater DOM.

Effect on EDC and UV absorbance
Increasing specific ozone doses (0-0.9 mg O3 ·g C -1 ) were added to DOM collected from the ozonation-reactor influent (Werdhölzli WWTP, Zurich, Switzerland) and the EDC analyzer was subsequently used to monitor the relative changes in EDC and UV absorbance compared to the untreated sample. Both parameters decreased with increasing specific ozone doses ( Figs. 5 a ), consistent with the ozone-induced oxidation of electron-donating and chromophoric moieties in DOM ( Wenk et al., 2013 ). The relative EDC showed a pronounced decrease at low specific ozone doses ( < 0.35 mg O3 ·g C -1 , Fig. 5 a ). At a specific ozone dose of 0.35 mg O3 ·g C -1 , the relative EDC decreased to 40%. Further increases in the specific ozone doses up to 0.9 mg O3 ·g C -1 resulted only in smaller additional decreases in the relative EDC down to 31%. The relative UV absorbance decreased to 60% and 42% at specific ozone doses of 0.35 and 0.9 mg O3 ·g C -1 , respectively ( Fig. 5 a ). Overall, the relative decrease in EDC was more pronounced compared to that in UV absorbance at low specific ozone doses ( Fig. 5 b ). In contrast, the relative UV absorbance continued to decrease at higher doses and approximately matched the relative decrease in EDC at the highest applied specific ozone dose of 0.9 mg O3 ·g C -1 . This observation is consistent with previous findings and has been attributed to the oxidative transformation of phenolic moieties to chromophoric quinone moieties at low ozone doses and the subsequent oxidation of these quinones at higher ozone doses ( Chon et al., 2015 ;Önnby et al., 2018a ;Ramseier and von Gunten, 2009 ;Tentscher et al., 2018 ). Fig. 5 c shows the relative residual concentrations of the selected micropollutants (i.e., 17 α-ethinylestradiol, bezafibrate, and atrazine) as function of specific ozone doses (0-0.9 mg O3 ·g C -1 ). As expected, the relative residual concentrations decreased with increasing specific ozone doses as a result of ozone-induced oxidation ( Lee et al., 2014( Lee et al., , 2013. Overall, micropollutant abatement was proportional to the apparent second-order rate constants ( k O3 ) of the reaction of these micropollutants with ozone at pH 7.9 ( Table   S2 , SI). The relative residual concentration of 17 α-ethinylestradiol ( k O3 = 1.2 ·10 7 M −1 ·s −1 ) decreased to 9% at a specific ozone dose of 0.25 mg O3 ·g C -1 . In contrast, bezafibrate ( k O3 = 590 M −1 ·s −1 ) required a higher specific ozone dose of 0.6 mg O3 ·g C -1 to reach a relative residual concentration of 16% and the concentration of atrazine ( k O3 = 6.0 M −1 ·s −1 ) only decreased to 36% at the highest tested specific ozone dose of 0.9 mg O3 ·g C -1 . Since ozone and secondarily formed hydroxyl radicals are the predominant oxidants ( Buffle et al., 2006a ) and because the second-order rate constants for the reactions of these micropollutants with hydroxyl radicals vary less than an order of magnitude (9.8 ·10 9 -3.0 ·10 9 M −1 ·s −1 , Table S2 , SI), variations in abatement efficiencies are largely explained by the variations in k O3 of these micropollutants. Finally, bromate formation was observed at specific ozone doses > 0.4 mg O3 ·g C -1 and increased abruptly from 2.0 to 20.4 μg ·L -1 between 0.4 and 0.9 mg O3 ·g C -1 ( Fig. 5 d ), consistent with the oxidation of bromide present in the untreated ozonation-reactor influent (c Br − = 258 ± 3 μg ·L -1 ) by ozone and hydroxyl radical ( von Gunten and Hoigné, 1994 ).

EDC and UV absorbance as surrogate parameters for micropollutant abatement and bromate formation
To assess the suitability of the relative decrease in EDC and UV absorbance as control parameters for ozonation, the relative micropollutant concentrations were replotted against the relative EDC ( Fig. 6 a ) and relative UV absorbance ( Fig. 6 b ). The relative con-centration of 17 α-ethinylestradiol decreased approximately linearly with decreasing relative EDC and reached a relative concentration of 9% while the relative EDC decreased to 52%. In contrast, concentrations of the less ozone-reactive micropollutants bezafibrate and atrazine showed smaller decreases to only about > 78% for the same decrease in EDC. At higher specific ozone doses, bezafibrate and atrazine concentrations continued to decrease while additional decreases in EDC were small. In comparison, relative decreases in micropollutant concentrations were steeper and more linear when plotted against the relative decrease in UV absorbance similar to previous findings ( Chon et al., 2015 ). Bromate concentrations sharply increased to > 10 μg ·L -1 for relative decreases in EDC and UV absorbance down to 34% and 44%, respectively ( Fig. S14 , SI). Overall, these results demonstrate that the EDC analyzer, which was designed for continual, long-term process monitoring, achieves a similar analytical performance and predictive power for micropollutant abatement as a previously presented laboratory method based on size-exclusion chromatography ( Chon et al., 2015 ). Since the EDC of DOM was particularly sensitive to low ozone doses up to 0.4 mg O3 ·g C -1 , monitoring changes in EDC absorbance may be specifically suited as control parameter for treatment processes requiring low ozone doses applied for example in combined ozonation and adsorption systems.

Conclusions
We developed and comprehensively validated (i) a photometric assay and (ii) an automated analyzer for the quantification of changes in EDC and UV absorbance of DOM during ozonation of secondary WWTP effluents. For this particular application, the two methods offer the following advantages over previous analytical systems: • In the photometric EDC assay, the reagents, buffer and DOM sample solution are mixed directly in disposable cuvettes, which minimizes solution handling and residual ABTS ·+reagent concentrations are directly quantified using a photometer enabling fast and scalable EDC measurements. • The design of the fluidic system of the EDC analyzer allows to perform continual, on-line EDC and UV analyses at a frequency of approximately two samples per hour. Compared to exist-ing continuous-flow methods, this fluidic system uses lower reagent volumes and is fully automated. Furthermore, the use of only a single rotary valve and a single syringe pump in the EDC analyzer minimizes maintenance and costs. • Analysis of model DOM isolates obtained with the EDC assay and the EDC analyzer are highly repeatable and closely match EDC values determined using established laboratory analysis methods such as MEO. • The EDC analyzer will enable a continual monitoring of the relative decreases in EDC and UV absorbance during ozonation and potentially allow to predict micropollutant abatement.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.