Municipal wastewater effluent characterization and ...

Science of the Total Environment 625 (2018) 1198–1207

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Municipal wastewater effluent characterization and variability analysis in view of an ozone dose control strategy during tertiary treatment: The status in Belgium Michael Chys a,⁎, Kristof Demeestere b, Ingmar Nopens c, Wim T.M. Audenaert a,c,1, Stijn W.H. Van Hulle a a b c

LIWET, Department of Industrial Biological Sciences, Ghent University Campus Kortrijk, Graaf Karel de Goedelaan 5, B-8500 Kortrijk, Belgium EnVOC, Department of Sustainable Organic Chemistry and Technology, Ghent University, Coupure Links 653, B-9000 Ghent, Belgium BIOMATH, Department of Mathematical Modelling, Statistics and Bioinformatics, Ghent University, Coupure Links 653, B-9000 Ghent, Belgium

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Powerful surrogates for water quality are revealed in view of tertiary ozonation. • Potential for a soft-sensor is established to measure alkalinity levels online. • Spectral measurements show site or event dependent deviations of treatment plants. • Variable water quality is a challenge for efficient control of ozone dose.

a r t i c l e

i n f o

Article history: Received 11 October 2017 Received in revised form 4 January 2018 Accepted 4 January 2018 Available online xxxx Editor: Yolanda Picó Keywords: Water resource recovery facilities Secondary municipal wastewater effluent Trace organic contaminants Water quality Principle component analysis

a b s t r a c t Ozonation is known for removing trace organic contaminants (TrOCs) from secondary wastewater effluent. However, its implementation and overall efficiency on a broad scale depends on effluent characteristics, which can differ both in time as well as between different treatment plants (nowadays referred to as water resource recovery facilities (WRRFs)). Therefore, water quality was assessed over time at 15 different Belgian sampling locations to increase the understanding of effluent variability in view of online control of the tertiary ozonation step. Conventional and surrogate parameters as well as those specifically related to tertiary ozonation (e.g. instantaneous ozone demand) were assessed. Little differences between the different locations were found for spectral measurements (e.g. UVA254 or fluorescence). The small amount of observed outliers was clearly site or event dependent. A lower variability (for spectral measurements) is advantageous in simplifying the development and application of a generic control framework based on these spectral measurements. In addition, also variations in TrOC concentration levels seemed to be small, as the concentration of most individual compounds resided within one order of magnitude over multiple sampling events at two different WRRFs. The combination of this low variability in TrOC levels in the effluent before ozonation with a control strategy using a TrOC removal efficiency set-point, allows to indicatively assess absolute TrOC levels after ozonation. In contrast, significant variations between different plants (especially smaller sized plants) were observed and could be related to the conventional water quality parameters alkalinity (correlated with the electrical conductivity) and pH which are both known to have an influence on the ozonation process. This confirms that a differential dosing control

⁎ Corresponding author. E-mail addresses: [email protected], [email protected] (M. Chys). 1 Present affiliation: AM-TEAM, Advanced Modelling for Process Optimisation, Hulstbaan 63, B-9112 Sint-Niklaas, Belgium.

https://doi.org/10.1016/j.scitotenv.2018.01.032 0048-9697/© 2018 Elsevier B.V. All rights reserved.

M. Chys et al. / Science of the Total Environment 625 (2018) 1198–1207

1199

strategy (i.e. accounting for the matrix reactivity) should be applied instead of one solely based on the (organic) effluent load before ozonation. © 2018 Elsevier B.V. All rights reserved.

1. Introduction The presence of TrOCs in municipal wastewater effluents poses an environmental burden on receiving water bodies as residual pharmaceuticals, personal care products, hormones etc. may trigger unwanted ecological effects (Gerrity and Snyder, 2011; Kidd et al., 2007). Tertiary treatment technologies as an additional step to remove these contaminants before discharge are currently under development. One promising technology is ozonation (Margot et al., 2013). Although ozonation has been successfully tested to remove TrOCs, the implementation of this technology and its overall efficiency on a broad scale might depend on effluent characteristics, which can differ both in time and between different water resource recovery facilities (WRRFs). Indeed, the composition of effluent depends on the source of the wastewater and involves both organic and inorganic substances. The effluent organic matter (EfOM) is composed of recalcitrant natural organic matter (NOM) (humic and fulvic acids), synthetic organic (micro)pollutants created or used during domestic applications, and soluble microbial products such as extracellular polymeric substances produced during biological wastewater treatment (Guo et al., 2011). Similar to organic compounds, inorganic compounds are influenced by the preceding secondary (and biological) treatment and originate from several chemical compounds, of which some contain nitrogen and phosphorous. Influents of WRRFs are known to be highly influenced by a diurnal pattern but also by fluctuating weather conditions (e.g. large rain events dilute the wastewater). Although these received wastewaters are intensively treated in conventional WRRFs and buffering occurs (especially in larger plants), the effluent quality still depends on the influent and its variability. Therefore, it is important to know the composition of secondary effluent to understand the possible reactions and interactions between organic and inorganic compounds present in the water matrix during subsequent ozonation (Shon et al., 2006). Tertiary ozonation and its reactions are influenced by the presence of impurities. The main parameters affecting the stability of ozone are chemical and physical water characteristics such as temperature, pH, EfOM content (and composition), alkalinity scavengers (e.g. carbonate and bicarbonate), and also the presence of highly ozone reactive inorganic constituents (e.g. NO− 2 -N). Indeed, alkalinity scavengers compete with other constituents for the available ozone given their high reactivity (kO3 ≈ 106–108 M−1 s−1) (Buxton et al., 1988). Other parameters such as the water turbidity have similar effects on the oxidation efficiency. The major sink of oxidants such as ozone is considered to be the presence of EfOM. Ozone is dosed to remove targeted TrOCs but these are only present in minor concentrations (ng L−1 to μg L−1 levels) compared to organic matter (mg L−1 levels). The EfOM concentration levels are mostly expressed by parameters such as COD or DOC, but also by more generic spectral measurements such as UV absorbance at 254 nm (UVA254) or total fluorescence response (Bahr et al., 2007; Gerrity et al., 2012; Wert et al., 2009). Moreover, not only the level but also the composition of EfOM plays a major role (Nöthe et al., 2009). For example, the instantaneous ozone demand (IOD) gives an idea about the level of highly ozone reactive species. This influences ongoing reactions, as e.g. Buffle and Von Gunten (2006) could observe high transient HO• concentrations related to a first initial phase with fast ozone reactions. Different types of organic matter can be associated with specific regions within a fluorescence 3D Excitation-Emission matrix (EEM). These different types show distinct behavior and properties (Chen et al., 2003; Sgroi et al., 2017a). Spectral measurements have

been put forward as online measurements to control the ozone dose for TrOC removal purposes (Bahr et al., 2007; Chys et al., 2017; Gerrity et al., 2012; Lee et al., 2013; Wert et al., 2009). This kind of measurement acts as a surrogate for the complete (organic) matrix and is related to the abatement of TrOCs. As main input parameters for a generic applicable ozone dose control framework, the variability of these characteristics both in time as well as between WRRFs is of high importance for a successful implementation and TrOC removal control. The current study aims at investigating the variability of both conventional physical-chemical water quality parameters and others, such as spectral measurements or ozone specific parameters (e.g. IOD), between different WRRF locations and over time. Different methods exist to determine the treatability of (secondary) wastewaters by ozonation and have proven to be useful (Schindler Wildhaber et al., 2015). For some parameters such as IOD or even TrOCs, (long-term) variability data among different plants is only scarcely available at large scale. This manuscript investigates effluent characteristics before ozonation and sheds light on the range of water quality parameters in municipal WRRF effluents. As a case study, 15 different locations in Belgium are selected, however results can be transposed to other countries. In addition, in view of implementing an effective control framework of a tertiary ozonation process, (i) correlations between effluent characteristics influencing the ozonation process and online measurements as potential control input parameters are investigated, (ii) characteristic differences between different WRRFs are revealed on a statistical founded basis by using principal component analysis, and (iii) an advice is provided on the usability of spectral measurements in a generic control framework. 2. Materials and methods 2.1. Experimental set-up and sampling procedure Secondary effluent was collected from 13 WRRFs located in different regions in Belgium with an I.E. (Inhabitant Equivalent) ranging from 500 to 116,100 (and operated by Aquafin NV or Ipalle). These WRRFs, and their specific effluents, were selected based on their location, receiving water (i.e. influent), and configuration (i.e. size and type of (biological) wastewater treatment). A detailed and schematic overview of the selected WRRFs is respectively given in Table 1 and Fig. S1, containing specific information on the treatment train preceding each sampling location. Two WRRFs contained a tertiary treatment - i.e. a constructed wetland (CW) and a sand filtration (SF) - and samples were taken before and after this tertiary treatment, bringing the total number of sampling locations to 15. Samples for all 15 locations were mainly obtained during three sampling campaigns spread over approximately 10 months (i.e. March '15, April '15 and January '16). Additional samples were taken at locations 1, 2, 5, 9, 13 and 14 in June '16 and were also included in this study. During each campaign, the time between sampling of the different locations was kept as short as possible to achieve comparable weather conditions, as good as practically feasible. As the number of samples for each individual location was limited in this approach, the results are mainly of value to provide a wide and general overview of the potential range of water quality characteristics. To provide more insights on a year's scale, the WRRFs of Aartselaar (location 11) and Harelbeke (location 15) were sampled more intensively to also include long time variations. Location 11 was sampled during 3 consecutive months – July to September '15 – with 17 additional

1200


Table 1 Overview of selected WRRFs, type of receiving wastewater (influent) and the associated sampling locations after secondary (or tertiary) treatment. Nr.

WRRF location

I.E.

Type of receiving wastewater (influent)

1

Kruiseke

500

100% municipal wastewater

2

(Secondary or tertiary) treatment technology

Size categorya

Biodisk + settler

Cat. 3

Biodisk + settler + constructed wetland

3

Mont-Saint-Aubert

600


Aerated and polishing lagoon

4

Oedeghnien

730


Biodisk + settler

5

Heule

14,000

±100% municipal wastewaterb

CASc

6

Péruwelz

14,000

Mainly municipal wastewater

CASc

7

Ath

22,000


CASc

8

Ingelmunster

35,000

Municipal and industrial wastewater (from food industry = good biological degradation)

CASc

9

Comines-Warneton

40,000

Mainly municipal wastewater (50% from Belgium, 50% from France)

CASc

10

Froyennes

50,000


CASc

11

Aartselaar

54,000

Municipal and industrial wastewater (from food industry = good biological degradation)

CASc

12

Menen

66,000

Mainly municipal wastewater (70% from Belgium, 30% from France)

CASc

13

Waregem

80,000

Municipal and industrial wastewater (from textile industry = difficult to degrade biologically)

CASc

14 15 a b c

Harelbeke

116,100

Mainly municipal wastewater (low degree of industrial wastewater)

Cat. 2

Cat. 1

CASc + sand filtration CASc

Size category determined by the inhabitant equivalent of each plant: large (I.E. N 80,000, cat. 1), medium (I.E. = 24,000–80,000, cat. 2) and small (I.E. b 24,000, cat. 3) sized plants. Receiving wastewater can be highly diluted e.g. because of a large share of receiving rain water. CAS = conventional activated sludge.

samples in October to December '14 and February '15 and location 15 during a little less than 1.5 years. During one year, a total of 273 samples (both grab and composite, see Table S1) were collected and analyzed for several physical-chemical characteristics (e.g. alkalinity, COD, pH, NO− 2 -N, etc.), UV–Visible spectra and fluorescence. Sampling of secondary effluents is less sensitive to the applied sampling strategy (e.g. compared to WRRF influents) due to the buffering of the complete treatment train (Aymerich et al., 2017). The combined set of grab and composite samples ensured representative samples retaining the variability of the effluent (Ort et al., 2010; Ort and Gujer, 2006). Samples were stored at 4 °C or analyzed immediately when possible. Samples for TrOC analysis were taken during a period of 2 years at the Harelbeke plant (7 samples) and during 2 months at the Aartselaar plant (47 samples). This resulted in an interesting dataset regarding the variability of TrOC concentrations over longer time periods, since this has only been scarcely investigated so far. Moreover, on a larger scale, very limited TrOC data for WRRF effluents in Belgium are available in the open literature. The two plants were selected for practical reasons (e.g. easily accessible, long term availability of automatic sampling devices, experience during previous campaigns, etc.) and their similarity with many other plants in Belgium. Samples were taken during different weather conditions (e.g. rain and dry weather), especially for the Aartselaar plant which contains three months of extensive and continuous sampling during long dry periods, first flush events and rainy conditions.

2.2. Analytical methods TrOCs were analyzed using a validated SPE-UHPLC-HRMS method, making use of a benchtop Q-Exactive™ Orbitrap mass spectrometer (Thermo-Fisher Scientific, USA). Details about the analytical procedure were provided by Vergeynst et al. (2017) and SI-Text 1. − + Nitrite (NO− 2 -N), nitrate (NO3 -N), ammonium (NH4 -N) and COD were determined spectrophotometrically using Hach-Lange cuvettes

and a DR2800 spectrophotometer (Hach-Lange, Belgium). Alkalinity (as mg L− 1 CaCO3) was determined according to standard methods (Eaton et al., 2005). Turbidity was measured with a portable Hi 98,703 (Hanna Instruments, USA). Conductivity (EC) and pH were registered by a multi-meter (HQ30D or HQ40D, Hach Belgium). The instantaneous ozone demand (IOD), defining the rapid ozone consumption within 5 s after dosing, was determined based on Hoigné and Bader (1994) and Roustan et al. (1998). Specifics related to the IOD determination are given in SI-Text 2.

2.3. Spectral measurements UV–Visible (UV-VIS) absorption data were obtained using a Shimadzu UV-1601 or UV-1800 spectrophotometer. Spectra were taken between 200 and 800 nm with 0.5 nm increments, using 1 cm quartz-cuvettes. The notation of the absorption coefficient (m− 1) is given as UVAi or colori with i related to the wavelength, e.g. 254 nm (UVA254) or 436 nm (color436). A few specific wavelengths (i.e. 210, 220, 254, 310 and 436 nm) were selected for further use based on their relevance and their previous usage in literature (Blaen et al., − 2016). Nitrogen species such as NO− 3 -N and NO2 -N are known to absorb below 250 nm and show maximum absorbance values around 220 and 210 nm, respectively, although specific determination of both might vary in the literature (Huebsch et al., 2015). The organic matrix is systematically represented by the absorbance at 254 nm (aromatic and double-bond moieties) (Dickenson et al., 2009; Wittmer et al., 2015), 310 nm (often used as a general representation of the (natural) organic matter) (Audenaert et al., 2011) and 436 nm (color) (Nanaboina and Korshin, 2010; Nöthe et al., 2009). Fluorescence EEMs were obtained using 1 cm quartz-cuvettes and a Shimadzu RF-5301 Fluorimeter. Dilution was performed when UVA254 N 0.3 cm−1 or when fluorescence measurements exceeded the maximal measurable intensity (Li et al., 2014; Murphy et al., 2011; Ohno, 2002). The dilution ratio was mathematically accounted for after correction of the raw EEMs, and before determination of the


fluorescence indices. Fluorescence intensities were measured at excitation wavelengths of 220–450 nm in 5 nm increments, and emission wavelengths of 280–600 nm in 1 nm increments. Excitation and emission slit widths were set at 5 nm, and a response time of 0.25 s was utilized. Raman scans of demineralized water were obtained at an excitation wavelength of 350 nm over an emission wavelength range of 365–450 nm in 0.2 nm increments, for the calculation of the Raman peak. The area of the Raman peak was used to normalize the fluorescence intensity of all spectra, finally expressed as RU (Raman units) (Murphy et al., 2013; Shutova et al., 2014). The 3D EEM was further used to derive several variables such as the total fluorescence (TF), determined by integration of the volume under the whole EEM. Different regions within the EEMs have been associated with different types of organic matter and properties, as defined by Chen et al. (2003) and modified by other authors, e.g. Sgroi et al. (2017a). Five different regions were used representing aromatic proteins, tyrosine-like substances (I1), aromatic proteins, tryptophan-like substances (I2), fulviclike and humic-like substances (I3), microbial byproducts, proteins, tryptophan-like substances and biopolymers (I4), and humic-like substances (I5). These different regions are typical EEM peaks present in wastewater matrices (Sgroi et al., 2017a). Details on the range of excitation and emission wavelengths, associated with these regions, are given in Table 2, and further information is available in detailed literature (and references therein) (Chen et al., 2003; Sgroi et al., 2017a). Directly measured EEM intensities (i.e. peak-picking method) might be representative for different individual fluorescing components, and result in an independent dataset. This in contrast to the more complex PARAFACapproach (Carstea et al., 2016; Li et al., 2013). Despite the fact that each fluorophore has its own affinity towards ozone (or its reaction products), directly measured intensities at defined peaks remain independent of a priori assumptions. The latter are for example made by earlier developed PARAFAC models that are applied to a new dataset. Although the peak-picking method is advantageous for the current research – providing more insights on effluent variability in view of tertiary ozonation – further investigations or developments towards a reliable control framework might need more advanced data processing to isolate e.g. identical reacting fluorophores. For example, Chys et al. (2017) showed a clear difference of ozone affinity among such components.

2.4. Data analysis Principal Component Analysis (PCA) was used as a pattern recognition method, aiming to reduce the large number of variables into a smaller number of principal components (PCs) or representative variables (Van Hulle and Ciocci, 2012). To reduce the contribution of variables with minor significance and increase the ease of interpretation, varimax normalized rotation was carried out (Brereton, 2003). The extracted number of PCs was defined by using the Kaiser criterion (Kaiser, 1960), retaining only PCs with Eigenvalues greater than unity and analyzing the scree plot (see Section 3.3). SPSS statistics 24 was used to perform all statistical analysis (www.spss.com). Table 2 Fluorescence regions and their excitation-emission wavelength boundaries (Chen et al., 2003). The fluorescence intensity used for further analyses was determined at the peak wavelength, similar to Sgroi et al. (2017a). Variable

I1 I2 I3 I4 I5

Excitation wavelengths (nm)

Emission wavelengths (nm)

Region boundary

Peak wavelength

Region boundary

Peak wavelength

220–250 220–250 220–300 250–300 250–580

225 225 245 275 345

250–320 320–390 390–580 250–390 300–450

290 340 440 345 440

1201

3. Results and discussion 3.1. Effluent water quality 3.1.1. Physical-chemical water characteristics An overview of all effluent characteristics determining the overall physical-chemical water quality is given in Fig. 1, showing distinct variations for most parameters. Variation among all measurements is observed especially for alkalinity (between 84 and 384 mg CaCO3 L−1), −1 NO− ), EC (between 396 3 -N (between 1.13 and 19.0 mg N L −1 and 1347 μS cm ), COD (between 5.8 and 84.6 mg O2 L− 1), IOD (between 2.9 and 16.0 mg O3 L−1), and a few of the spectral measurements (e.g. UVA210, UVA220, UVA254 and I2) with for example UVA210 and UVA254 ranging between 67 and 395 m−1 and 9–64 m−1 respectively. These can all (except for NO− 3 -N) be considered to influence ongoing reactions during tertiary ozonation (Von Gunten, 2003). Given the large number of measurements (maximum of 273) obtained for each parameter in a time period of a little under 1.5 years and at 15 different sampling locations, this logically results in varying observations. Effluent (and its quality) is indeed susceptible to operational deviations (e.g. malfunctions), variations in receiving influent of the WRRF due to e.g. weather conditions, diurnal patterns, etc. but also the location might play an important role as some plants are receiving specific industrial wastewater. Focusing on the combined samples for all 15 locations (without the additional taken samples for location 11 and 15), alkalinity (Fig. 2a) and IOD (Fig. 2c) exhibit no clear specific dependence on the sampling locations (in contrast to COD, UVA254 and TF, Fig. 2b,d,e) and are highly variable within one plant. For locations 1 and 2 (both originating from the WRRF of Kruiseke), however, significant lower alkalinity values are noticed. This is presumably related to the very small size of the WRRF, and hence its higher susceptibility to fluctuations (i.e. weather conditions, flow or constituents) in the receiving municipal wastewater. The large variations of IOD, up to a factor of 5 difference between the minimal and maximal measured values, have a large impact on a potential tertiary ozonation step. IOD is representative for the amount of ozone susceptible moieties, and low (high) values are indicative for a decreased (increased) need of unselective oxidant species to react with the other, ozone recalcitrant moieties. Additionally, the initial phase of fast ozone reactions is related to very high observed transient HO• concentrations, playing an essential role during the oxidation processes (Buffle and Von Gunten, 2006). Logically, variations in the IOD are an indication of the HO• formation potential, influencing consecutive reactions with less ozone reactive moieties. A few remarkable outlier measurements are noticed for turbidity and fluorescence (TF, I2, I4). Short-term large increases of turbidity are not illogical considering the possibility of flushing out more suspended particles (e.g. sludge) from the settler succeeding biological treatment. The specific data-points from Fig. 1 (high values of turbidity) can be allocated to sampling location 1 and 11, respectively originating from (i) the smallest WRRF before passing through a constructed wetland (CW) and (ii) one of the largest plants sampled at a time with increased flow (i.e. first flush), as a presumable consequence of abnormalities in the settler. Outliers for turbidity can thus be attributed to specific events, whereas the high variability in fluorescence response (specifically for TF, I2 and I4) is clearly plant specific. Fig. 2e shows, for each sampling location individually, the TF for samples taken at similar moments. The response at locations 13 and 14 (both WRRF of Waregem, respectively before and after a tertiary sand filtration) is clearly remarkable with a difference up to 3 to 13 times resembling respectively the minimum and maximum values of the other plants. Although these locations (i.e. 13 and 14) show values comparable to other locations concerning characteristics such as alkalinity (see Fig. 2a), pH (little variation in general, see Fig. 1), turbidity, etc., relatively high values are also noticed for COD and UVA254 (Fig. 2b–d). The increased TF is mainly caused by I2 and I4, both remarkably higher compared to I1, I3 and I5, and representing

1202


Fig. 1. Water quality characteristics of all effluent samples presenting (a) conventional physical-chemical characteristics and (b) spectral measurements, analyzed for all 15 sampling locations. The whiskers of the boxplots indicate the 10th and 90th percentile of the data distribution, while the white dots are the outliers considering all measured data.

regions within the EEMs that contain tryptophan-like substances (in contrast to I1, I3 and I5). The increased response for I2 and I4, and the presumable increased presence of tryptophan-like substances, is clearly specific for these locations. Both UVA254 (Fig. 2d) and COD (Fig. 2b) show a similar pattern of which the larger values observed at locations 1, 13 and 14 can be associated with the previous mentioned abnormalities (location 1) and the increased presence of tryptophan-like substances and substances having a significant degree of aromaticity (associated with the increased values of e.g. I1 and I2, representing aromatic proteins and tryptophanlike substances, in location 13 and 14). Locations 13 and 14 are located at a conventional WRRF receiving a share of industrial textile wastewater mixed with the municipal wastewater, unique among the selected WRRFs (Table 1). Textile wastewater is known to contain a significant amount of dyes which retain some absorption (Von Sonntag and von Gunten, 2012). The latter are traditionally difficult to be removed during secondary (biological) wastewater treatment. Additionally, previous research indicates highly increased intensity values during fluorescence measurements associated with this kind of substances (Wu et al., 2016). Dyes consisting of aromatic, heterocyclic and nitrogen containing moieties such as azobenzene or triazines typically show a low reactivity towards ozone and consequently a very low yield for O3 consumption (Chu and Ma, 2000; Von

Sonntag and von Gunten, 2012; Wu et al., 2016). Despite the higher values of general EfOM parameters such as UVA254 (Fig. 2d), TF (Fig. 2e) and COD (Fig. 2b), no remarkable increase of IOD (Fig. 2c) is noticed. This supports the hypothesis of an increased amount of ozone recalcitrant moieties. As such, the presence of these rather plant-specific substances in the effluent poses a challenge for consecutive ozonation. 3.1.2. TrOC concentrations 54 samples were analyzed for 40 TrOCs (all pharmaceuticals) of which 22 TrOCs are retained for discussion and displayed in Fig. 3. Some basic characteristics of these TrOCs, in view of tertiary ozonation, are summarized in Table S2. Overall, TrOCs are detected in a range from 1 ng L−1 to N2 μg L− 1. Although samples are taken at two different WRRFs (respectively nmax = 47 and 7 for location 11 and 15) and almost over a 2 years period, rather small variations are noticed for the individual compounds. Fig. 3 shows that, for 21 of the 22 TrOCs and especially for those having an average concentration of 10 ng L−1 or above, the measured concentrations are within one order of magnitude. A large variation (3 orders of magnitude) is only noticed for sulfamethazine (i.e. an antibiotic used to treat infections as e.g. bronchitis) which presence might depend on specific periods of increased illness among the population. Diclofenac is consistently detected at the highest concentrations (on average 928 ng L−1; variation over a factor of 5.3).


1203

Fig. 3. Concentration levels (ng L−1) of 22 TrOCs, quantified in minimum 20% of all samples. The whiskers of the boxplots indicate the 10th and 90th percentile of the data distribution, while the black dots are the outliers considering all measured data. The quantification frequency is indicated on the right indicating the total number of samples in which the TrOC was measured above the detection limit (nNDL, see Vergeynst et al., 2017).

Fig. 2. Selected physical-chemical water quality characteristics of effluent samples taken during three sampling campaigns of which all samples within one campaign were taken within similar time intervals. The characteristics are depicted as (a) alkalinity, (b) COD, (c) IOD, (d) UVA254 and (e) TF. WRRF sampling location numbers refer to Table 1. Grey bars indicate the maximal and minimal measured values.

Other TrOCs quantified at high concentrations such as carbamazepine and venlafaxine show even smaller differences between the minimum and maximum values, i.e. within a factor of respectively 3.5 and 4.1. This indicates a more or less constant discharge of TrOCs in the receiving water bodies, which might be advantageous when applying ozone control strategies based on the relative decrease of TrOC levels (ΔTrOC, expressed in %) in the effluent. Almost stable TrOC levels in the effluent before ozonation combined with a properly working control strategy implying a ΔTrOC-set-point, allows a ballpark estimation of the absolute TrOC levels to be expected after ozonation. On the other hand, the low variability of TrOC concentration levels makes it difficult and even reasonless to determine reliable correlations with other parameters (e.g. fluorescence indices or other water quality parameters). Such correlations (Sgroi et al., 2017a, b) have been developed for samples collected at different stages of a WRRF or at different locations in wastewater-impacted rivers, with TrOC concentrations varying within several orders of magnitude. More (variable) data would be needed to confirm such relationships among the data obtained in the current study. 3.2. Correlation between water quality characteristics Potential correlations between different physical-chemical water characteristics are determined according to the correlation matrix

displayed in Table 3. All characteristics, and their correlations, can be roughly summarized in three main groups - keeping the goal of tertiary ozonation in mind - which are (i) mainly influenced by the secondary biological treatment such as the nitrogen containing species (NO− 2 -N, + NO− 3 -N, NH4 -N) and organic loads, (ii) indicative for the load in the effluent such as EC or alkalinity (some depending partially on the biological treatment), and (iii) representing the organic matrix as a whole (e.g. TF, COD) or as specific (groups of) moieties such as UV-VIS or fluorescence spectral measurements. These three groups are introduced to facilitate the discussion. − + Remaining concentrations of NO− 2 -N, NO3 -N and NH4 -N after biological treatment have a good correlation with each other, up to the 0.01 significance level. Most WRRF sampling locations were preceded with a biological treatment (i.e. activated sludge) in several different configurations (see Table 1). Nitrification is transforming NH+ 4 -N − species into NO− 2 -N and NO3 -N. Denitrification further transforms − NO− 3 -N to N2 with NO2 -N being an intermediate product of this complete reaction (Henze et al., 2008). It is therefore logical that the occurrence of one of these constituents correlates with the removal (or production) of another. The presence of NO− 2 -N can for example be a sign of ineffective biological treatment (nitrification-denitrification) as it is aimed to transform all nitrogen to N2 gas (Tchobanoglous et al., 2003). Furthermore, also spectral measurements such as UVA210 and UVA220 show an excellent correlation (Pearson coefficients above 0.9 at the 0.01 significance level) with NO− 3 -N which is part of the nitrogen-containing species. The UV absorbance at 210 or 220 nm has been recognized as an indicator for nitrogen-containing species such − as NO− 2 -N or NO3 -N, respectively, as both show clear absorbance maxima around these wavelengths (Blaen et al., 2016; Huebsch et al., 2015). However, specific determination of both can be influenced by one another as well as by other constituents absorbing around these wavelengths (e.g. a large variety of organic moieties such as humic acids). This is indicated by the correlation (Pearson coefficient 0.20–0.27 with NO− 2 -N) with UVA254, UVA310 and color436 at the 0.01 confidence level, and probably contributes to the low correspondence of NO− 2 -N

K

L

M

UVA220

UVA254

UVA310

P

Q

R

S

T

I1

I2

I3

I4

I5

N

J

UVA210

O

I

IOD

TF

H

NH4+-N

color436

F

G

NO3--N

E

COD

NO2--N

C

D

Turbidity

B

pH

EC

A

Alkalinity

1

−0.21**

−0.09

−0.11

–0.23**

0.68**

0.09

0.02

−0.07

−0.12

−0.1

−0.15*

0.07

0.16*

−0.19*

0.21**

−0.03

−0.02

0.05

0.09

−0.01

0.05

0.01

0.27**

−0.62**

−0.18**

0.26**

−0.55**

0.05

0.05

0.26**

0.80**

−0.06

0.03

0.23**

−0.74**

0.02

–

1

−0.05

B

A

0.35**

0.35**

0.25**

0.33**

0.08

0.32**

0

0.14*

0.15*

−0.29**

−0.37**

0.25**

−0.45*

−0.16

−0.1

0.07

−0.07

1

–

–

C

0.07

0.07

−0.03

0.03

0.12*

0.05

0.78**

0.65**

0.48**

−0.1

−0.13*

0.28**

0.33

−0.63**

0.20**

0.64**

1

–

–

–

D

0.39**

0.42**

0.43**

0.42**

0.38**

0.49**

0.75**

0.79**

0.75**

−0.02

0.01

0.59**

0.53**

−0.77**

0.37**

1

-

–

–

–

E

−0.13

−0.04

−0.1

−0.05

0.06

−0.09

0.27**

0.20**

0.22**

−0.11

−0.01

0.24**

0.64**

−0.50*

1

-

-

–

–

–

F

−0.38*

−0.77**

−0.1

0.48**

0.22

0.13

−0.64**

−0.49**

−0.54**

0.98**

0.94**

−0.84**

−0.42*

1

-

-

-

–

–

–

G

0.02

0.59**

0.56**

−0.14

0.25

0.3

0.74**

0.68**

0.75**

−0.32

−0.12

0.39*

1

-

-

-

-

–

–

–

H

0.46**

0.33**

0.48**

0.30**

0.12

0.40**

0.39**

0.59**

0.67**

−0.43**

−0.33**

1

-

-

-

-

-

–

–

–

I

−0.20**

−0.19**

−0.09

−0.1

0.15*

-0.09

−0.13*

−0.12*

−0.15*

0.91**

1

-

-

-

-

-

-

–

–

–

J

−0.22**

−0.19**

−0.13*

−0.09

0.18**

−0.1

−0.15*

−0.17**

−0.23**

1

-

-

-

-

-

-

-

–

–

–

K

0.67**

0.42**

0.71**

0.39**

0.21**

0.58**

0.68**

0.95**

1

-

-

-

-

-

-

-

-

–

–

–

L

0.57**

0.37**

0.54**

0.35**

0.24**

0.49**

0.84**

1

-

-

-

-

-

-

-

-

-

–

–

–

M

0.32**

0.20**

0.18**

0.17**

0.25**

0.24**

1

-

-

-

-

-

-

-

-

-

-

–

–

–

N

0.42**

0.94**

0.82**

0.94**

0.16**

1

-

-

-

-

-

-

-

-

-

-

-

–

–

–

O

0.20**

0

0.01

0.08

1

-

-

-

-

-

-

-

-

-

-

-

-

–

–

–

P

0.17**

0.98**

0.63**

1

-

-

-

-

-

-

-

-

-

-

-

-

-

–

–

–

Q

0.69**

0.65**

1

-

-

-

-

-

-

-

-

-

-

-

-

-

-

–

–

–

R

0.21**

1

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

–

–

–

S

1

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

-

–

–

–

T

Table 3 Correlation matrix of the physical-chemical effluent characteristics (given as letters A to T) with indication of the Pearson coefficient. One or two asterisks (*) are marking the significant correlations at the 0.05 and 0.01 level (2-tailed), respectively − (also indicated as light or dark grey shaded). The strongest correlations (Pearson coefficient ≥ |0.75| and b|0.75| but ≥|0.50|) are respectively indicated in green and yellow (all at the 0.01 confidence level except NO− 2 -N vs NO3 -N)·The physical−1 − + chemical variables are: alkalinity (A, mg CaCO3 L−1), pH (B), EC (C, μS cm−1), turbidity (D, NTU), COD (E, mg O2 L−1), NO− ), IOD (I, mg O3 L−1), UVA210, UVA220, UVA254, UVA310 and color436 (J-N, m−1), 2 -N, NO3 -N and NH4 -N (F-H, mg N L TF and I1 − 5 (O-T, RU).

1204 M. Chys et al. / Science of the Total Environment 625 (2018) 1198–1207


and UVA210 (or UVA220) in Table 3. All NO− 2 -N measurements (n = 198) were below 1 mg N L−1 (indicating a good removal during biological −1 treatment), in contrast to NO− ) and other UV 3 -N (up to 19.0 mg N L absorbing constituents that might be present (e.g. organics, see Fig. 1). To ensure a reliable and trustworthy correlation between nitrogencontaining species and UV-VIS measurements, the need for multivariate data analysis (e.g. principle component analysis or partial least square regression) is recognized (Huebsch et al., 2015). Furthermore, also alkalinity is partially depending on the biological treatment and some correlation can be observed with a few previously discussed characteristics (e.g. NO− 3 -N, UVA210, UVA220). The presence of CO2− (and HCO− 3 3 ) acts as an important scavenger for hydroxyl radicals (HO•) with reaction rates in the order of 106–108 M−1 s−1 and, therefore, increasing the ozone demand of the wastewater (Buxton et al., 1988). Mainly occurring as a HO• scavenger, little effect can be expected on the IOD (direct ozone reactions) as exemplified by their low correlation (Pearson coefficient 0.23). Water quality characteristics being indicative for the effluent load, such as alkalinity and EC, exhibit a good correlation with each other (Pearson coefficient of 0.68, 0.01 level, n = 147). Both characteristics are determined by the presence of ions in the water. Alkalinity is largely affected by the biological treatment process and roughly refers to the bases that can be converted to uncharged species by acidification. It should be mentioned, however, that EC is (especially more than alkalinity) representative for all ions present and not as a sole representation for the alkalinity levels. Dilution of an effluent affects both parameters. An additional benefit of such a correlation is that EC is easily measured online, while alkalinity determination requires sampling and offline analyses in the lab. Therefore, such a correlation poses a potential indirect online follow-up of the alkalinity, known to influence ozone reactions (Von Gunten, 2003). Nevertheless, further investigations are necessary to develop this kind of soft sensor as there is currently no indication of plant independence, hence possessing great enhancement potential. A large portion of the observed correlations are related to the bulk organics present in the effluent in one way or the other. COD as a surrogate for the entire (chemically oxidizing) organic matrix shows good correlations with IOD and all spectral measurements (all significant at the 0.01 level), with the exception of UVA210 and UVA220 (both previously related to NO− 3 -N). More specifically, UVA254 (as also UVA310 and color436) show the best correlation with COD (Pearson coefficient 0.75 or above). This is in line with previous research which has led to the availability of sensors enabling the online monitoring of COD (Blaen et al., 2016). IOD, as a representation of the highly ozone reactive species, is also correlated with COD, however, to a slightly lesser extent (Pearson coefficient 0.59). The IOD is thus clearly influenced by the preceding biological treatment (hence its correlation with e.g. NO− 3 -N, Pearson coefficient of 0.84) and the organic matrix, although no perfect correlation exists. This means that next to the amount of organics present, their composition (more specifically the portion of highly reactive moieties) and the presence of other reactive constituents (e.g. NO− 2 -N) are influential. Furthermore, the IOD is correlated with UVA254 (Pearson coefficient 0.67) to a similar extent. The latter is known to be greatly influenced by (electron rich) aromatic and saturated moieties absorbing at this wavelength (Westerhoff et al., 1999). The fluorescence derived parameters represent the organic matrix, hence their good correlations with others such as UVA254, COD, etc. (Table 3). Nevertheless, while no clear differences are noticeable between the UV-VIS derived parameters (with the exception of UVA210 and UVA220), I1 clearly gives different relationships compared to TF and I2–5. In addition to previous observations, I2 and I4 are strongly correlated with each other as both are representative for tryptophanlike substances. Similarly, also I3 and I5 show some correlative behavior as both (partially) represent humic acid-like substances, and both have the highest correlations to IOD and UVA254 (Pearson coefficient 0.46–0.71) compared to I1,I2 and I4 (Pearson coefficient ≤ 0.42 for all,

1205

see Table 3). I1, representing aromatic proteins and tyrosine-like substances, is in contrast to other derived parameters clearly not correlated with the IOD. This suggests that this surrogate measurement is an indicator for the amount of recalcitrant moieties within the EfOM. It is hypothesized that this can be useful to follow-up ozone recalcitrant organics, which is of high interest in order to achieve insight into the overall reaction process. Although aromatic moieties (e.g. susceptible to ozone) are showing intensities in these regions, other lower reactive species might also be present at these wavelengths. Further investigation (e.g. kinetic degradation experiments) is required to confirm these findings. To focus more into detail on specific moieties, e.g. PARAFAC can be used to deconvolute the EEMs. Interest in this technique has recently increased substantially and it has been used for different applications by several authors (Chys et al., 2017; Li et al., 2014; Murphy et al., 2011, 2013; Oloibiri et al., 2017; Shutova et al., 2014). 3.3. Principle component analysis in view of the control of a tertiary ozonation step The resulting rotated principle component (PC) matrix is presented in Table S3. It is decided to use these four PCs, with each an Eigenvalue significantly larger than 1 (= Kaiser criterion), for further analysis (Kaiser, 1960). This decision is supported by a total explained variance of 74.9% and the fact that the scree plot contains a clear inflection point when retaining these four components. The Kaiser-MeyerOlkom (KMO) criterion for sampling adequacy (KMO = 0.67 N 0.5) is fulfilled, and also the Barlett's test of sphericity (χ2(153) ≈ 3017.5; p = 0.00 b 0.05) indicates sufficient correlation between items to perform PCA (Smeti et al., 2009). All variables shown in Table 3 are included in this analysis when N100 measurements could be attributed. This excludes the limited data available for NO− 3 -N (n = 34) and NH+ 4 -N (n = 28). The first and second PC are respectively explaining 25.0% and 23.7% of the total variance and contain most of the information on measurements related to the organic matrix (Table S3). PC1 describes most information from COD, IOD, turbidity and UV-VIS measurements, such as UVA254 and UVA310. PC2 retains most information obtained from fluorescence derived measurements with the clear exception of I1, previously defined to have a distinct behavior. The third and fourth PCs are more related to conventional physical-chemical water characteristics. PC3 describes most information of I1 (representing potential ozone refractory organics), UVA210 and UVA220 (related with inorganic nitrogen-containing species after biological treatment). PC4 is clearly a representation of the basic water quality mainly containing information of pH, EC, alkalinity and NO− 2 -N. In Fig. 4, the score plots of the four PCs is given comparing PC1 with PC2 (Fig. 4a), and PC3 with PC4 (Fig. 4b). Other combinations of PCs in score plots does not give any additional (visual) information and are therefore not shown. The PCA analysis provides more insight into the variability of, and potential coherence between, different plants. PC1 and PC2 are in large quantity situated around the zero point, indicating no large variations within these PCs. Nevertheless, some outlier values are noticed which can be attributed to (i) some disturbance - by e.g. particles or due to analytical anomalies - in the measurements (PC1, from e.g. sampling locations 1 and 11), and (ii) the increased fluorescence response noticed in sampling locations 13 and 14 (PC2). It is noteworthy that both of these observations have been mentioned previously, when describing the individual characteristics of the different sampling locations. Both PC1 and PC2 are mainly describing variations in the spectral measurements. A low variability of spectral measurements - and a general lack of plant dependent behavior - as input to an ozone dose control framework reduces the chance for errors and will simplify the development of a generic framework. Nevertheless, some outlier values are present which still need to be accounted for and require extensive validation of a surrogate-based correlation model during long-term operation and for different WRRFs.

1206


Fig. 4. Score plots of the PCA analysis with all physical-chemical water characteristics giving (a) PC1 vs. PC2 and (b) PC3 vs PC4. The data is color coded based on WRRF size (I.E.) ranging from dark blue (smallest = 500 I.E.) to yellow (largest = 110,000 I.E.) (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

PC3 and PC4 – being characteristic of basic and general water quality parameters – are exhibiting a much higher degree of variation. If the plant size is accounted for, most variation in measurements is attributed to decreasing WRRF size. The larger plants exhibit much less variability, presumably a consequence of the large buffering capacity of such a plant. The smaller a WRRF, the more direct influence of abnormalities in the process operation or in variations of the receiving influent is noticed. Variations of PC3 can be because of anomalies in the biological treatment and directly influence the ozonation process when e.g. an increase of NO− 2 -N occurs. On the other hand, the presence of I1, hypothesized to describe low ozone reactive substances, increases the need for indirect ozone reactions and therefore an (increased) generation of HO•. PC4, containing e.g. information related to alkalinity and pH, has its direct influence on the ozonation process. Whereas alkalinity acts more as an inhibitor of the ozonation process, an elevated (decreased) pH leads to an elevated (decreased) production of HO•. This greatly influences the chain reactions or autocatalytic decomposition of ozone, according to known radical chain reactions pathways (Beltran, 2004; Von Gunten, 2003). 4. Conclusion Fifteen different locations (on 13 WRRFs in Belgium) were selected to collect (secondary) effluent samples over a maximum period of almost 2 years. months. The main goal was to assess the water quality variability in view of (controlling) a tertiary ozonation step. The use of PCA resulted in an overall insight of variations between WRRFs. Spectral measurements such as UV-VIS or fluorescence are put forward as input for an online ozone dose control framework in the literature. These measurements showed limited variations among the different plants, with the low number of observed outliers being clearly site or event dependent. In terms of applicability, the observed low variability among different municipal WRRF plants facilitates the development and implementation of a generically applicable ozone dose control framework based on these spectral measurements. Models validated for a representative subset of municipal wastewater effluent can be easily applied on other similar WRRF plants. In addition, also variations in TrOC concentration levels seemed to be small, given that the presence of most individual compounds remained within the same order of magnitude over multiple sampling events at two different WRRFs. This poses an advantage when applying ozone dose control strategies based on the relative decrease of TrOC levels since absolute TrOC levels to be expected after ozonation can be estimated. On the other hand, significant variations among the different plants (especially smaller sized plants) could be noticed related to the basic

water quality, exemplified by e.g. alkalinity and pH which are known to have an influence on the ozonation process. Based on an extensive correlation analysis, the observed correlation between e.g. alkalinity (i.e. known to scavenge HO•) and electrical conductivity (i.e. online measurable) opens the perspective of including these variations in a future online applicable control framework. Similar observations were made for less ozone reactive substances (tyrosine-like compounds) based on fluorescence measurements (I1), hypothetically defined by their complete lack of correlation with the IOD. Further (kinetic) investigation is however needed to confirm these findings. The low correspondence between most basic water quality parameters (e.g. alkalinity, IOD, etc.) and the spectral measurements confirms the need for a feedback control approach for tertiary ozonation systems, as a feedforward prediction of the matrix composition (and/or reactivity) seems not to be applicable. Based on the insights obtained in this work, showing the potential for online ozone dose control using spectral measurements, further research on (i) diurnal or seasonal variability of the effluent water quality characteristics within a single plant and (ii) the use of more advanced data processing to isolate e.g. identical reaction fluorophores will be helpful to better understand the robustness and to improve reliable ozone dose control frameworks.

Acknowledgements The financial support (AUGE/11/016) from the Hercules Foundation of the Flemish Government is acknowledged for the UHPLC-QExactive™ mass spectrometry equipment. Ghent University is acknowledged for the PhD grant of Michael Chys, and the Special Research Fund (Ghent University) for funding the automated SPE equipment (01B07512). The authors further like to thank all staff and students helping during the different sampling campaigns and analysis, but also Aquafin NV and Ipalle for giving access to their respective treatment plants. This project was initiated within the LED H2O project which belongs to the LED network (www.lednetwerk.be) and is financially supported by the Flemish Knowledge Center Water (Vlakwa vzw). Appendix A. Supplementary data Detailed information on experimental procedures (TrOC analysis, IOD determination), WRRFs design and collected samples, properties of quantified TrOCs, and the Principle Component Analysis is given as additional, supplementary material. Supplementary data associated with this article can be found in the online version, at https://doi.org/ 10.1016/j.scitotenv.2018.01.032.


References Audenaert, W.T.M., Vermeersch, Y., Van Hulle, S.W.H., Dejans, P., Dumoulin, A., Nopens, I., 2011. Application of a mechanistic UV/hydrogen peroxide model at full-scale: sensitivity analysis, calibration and performance evaluation. Chem. Eng. J. 171:113–126. https://doi.org/10.1016/j.cej.2011.03.071. Aymerich, I., Acuña, V., Ort, C., Rodríguez-Roda, I., Corominas, L., 2017. Fate of organic microcontaminants in wastewater treatment and river systems: an uncertainty assessment in view of sampling strategy, and compound consumption rate and degradability. Water Res. https://doi.org/10.1016/j.watres.2017.08.011 (in press). Bahr, C., Schumacher, J., Ernst, M., Luck, F., Heinzmann, B., Jekel, M., 2007. SUVA as control parameter for the effective ozonation of organic pollutants in secondary effluent. Water Sci. Technol. 55:267–274. https://doi.org/10.2166/wst.2007.418. Beltran, F.J., 2004. Ozone Reaction Kinetics for Water and Wastewater Systems. Blaen, P.J., Khamis, K., Lloyd, C.E.M., Bradley, C., Hannah, D., Krause, S., 2016. Real-time monitoring of nutrients and dissolved organic matter in rivers: capturing event dynamics, technological opportunities and future directions. Sci. Total Environ. 569: 647–660. https://doi.org/10.1016/j.scitotenv.2016.06.116. Brereton, R.G., 2003. Chemometrics: Data Analysis for the Laboratory and Chemical Plant. Wiley, UK https://doi.org/10.1007/s10809-005-0223-6. Buffle, M.O., Von Gunten, U., 2006. Phenols and amine induced HO• generation during the initial phase of natural water ozonation. Environ. Sci. Technol. 40:3057–3063. https://doi.org/10.1021/es052020c. Buxton, G.V., Greenstock, C.L., Helman, W.P., Ross, A.B., 1988. Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals in aqueous solution. J. Phys. Chem. Ref. Data 17:513. https://doi.org/10.1063/1.555805. Carstea, E.M., Bridgeman, J., Baker, A., Reynolds, D.M., 2016. Fluorescence spectroscopy for wastewater monitoring: a review. Water Res. 95:205–219. https://doi.org/10.1016/j. watres.2016.03.021. Chen, W., Westerhoff, P., Leenheer, J.A., Booksh, K., 2003. Fluorescence excitation– emission matrix regional integration to quantify spectra for dissolved organic matter. Environ. Sci. Technol. 37:5701–5710. https://doi.org/10.1021/es034354c. Chu, W., Ma, C.W., 2000. Quantitative prediction of direct and indirect dye ozonation kinetics. Water Res. 34:3153–3160. https://doi.org/10.1016/S0043-1354(00)00043-9. Chys, M., Audenaert, W.T.M., Deniere, E., Mortier, S.T.F.C., Van Langenhove, H., Nopens, I., Demeestere, K., Van Hulle, S.W.H., 2017. Surrogate-based correlation models in view of real-time control of ozonation of secondary treated municipal wastewater - model development and validation. Environ. Sci. Technol. 51:14233–14243. https://doi.org/ 10.21/acs.est.7b04905. Dickenson, E.R.V.V., Drewes, J.E.J.E., Sedlak, D.L., Wert, E.C., Snyder, S.A., 2009. Applying surrogates and indicators to assess removal efficiency of trace organic chemicals during chemical oxidation of wastewaters. Environ. Sci. Technol. 43:6242–6247. https:// doi.org/10.1021/es803696y. Eaton, A.D., Clesceri, L.S., Rice, E.W., Greenberg, A.E., 2005. Standard Methods for the Examination of Water and Wastewater. 21st edition. APHA American Public Health Association, Washington, DC. Gerrity, D., Gamage, S., Jones, D., Korshin, G.V., Lee, Y., Pisarenko, A., Trenholm, R.A., von Gunten, U., Wert, E.C., Snyder, S.A., 2012. Development of surrogate correlation models to predict trace organic contaminant oxidation and microbial inactivation during ozonation. Water Res. 46:6257–6272. https://doi.org/10.1016/j. watres.2012.08.037. Gerrity, D., Snyder, S., 2011. Review of ozone for water reuse applications: toxicity, regulations, and trace organic contaminant oxidation. Ozone Sci. Eng. 33:253–266. https://doi.org/10.1080/01919512.2011.578038. Guo, J., Peng, Y., Guo, J., Ma, J., Wang, W., Wang, B., 2011. Dissolved organic matter in biologically treated sewage effluent (BTSE): characteristics and comparison. Desalination 278:365–372. https://doi.org/10.1016/j.desal.2011.05.057. Henze, M., van Loosdrecht, M.C.M., Ekama, G.A., Brdjanovic, D., 2008. Biological Wastewater Treatment: Principle, Modelling and Design. IWA Publishing, London. Hoigné, J., Bader, H., 1994. Characterization of water quality criteria for ozonation processes. Part II: lifetime of added ozone. Ozone Sci. Eng. 16, 121–134. Huebsch, M., Grimmeisen, F., Zemann, M., Fenton, O., Richards, K.G., Jordan, P., Sawarieh, A., Blum, P., Goldscheider, N., 2015. Technical note: field experiences using UV/VIS sensors for high-resolution monitoring of nitrate in groundwater. Hydrol. Earth Syst. Sci. 19:1589–1598. https://doi.org/10.5194/hess-19-1589-2015. Kaiser, H.F., 1960. The application of electronic computers to factor analysis. Educ. Psychol. Meas. 20, 141–151. Kidd, K.A., Blanchfield, P.J., Mills, K.H., Palace, V.P., Evans, R.E., Lazorchak, J.M., Flick, R.W., 2007. Collapse of a fish population after exposure to a synthetic estrogen. Proc. Natl. Acad. Sci. 104:8897–8901. https://doi.org/10.1073/pnas.0609568104. Lee, Y., Gerrity, D., Lee, M., Bogeat, A.E., Salhi, E., Gamage, S., Trenholm, R.A., Wert, E.C., Snyder, S.A., von Gunten, U., 2013. Prediction of micropollutant elimination during ozonation of municipal wastewater effluents: use of kinetic and water specific information. Environ. Sci. Technol. 47:5872–5881. https://doi.org/10.1021/es400781r. Li, W.-T., Chen, S.-Y., Xu, Z.-X., Li, Y., Shuang, C.-D., Li, A.-M., 2014. Characterization of dissolved organic matter in municipal wastewater using fluorescence PARAFAC analysis and chromatography multi-excitation/emission scan: a comparative study. Environ. Sci. Technol. 48:2603–2609. https://doi.org/10.1021/es404624q. Li, W.T., Xu, Z.X., Li, A.M., Wu, W., Zhou, Q., Wang, J.N., 2013. HPLC/HPSEC-FLD with multiexcitation/emission scan for EEM interpretation and dissolved organic matter analysis. Water Res. 47:1246–1256. https://doi.org/10.1016/j.watres.2012.11.040. Margot, J., Kienle, C., Magnet, A., Weil, M., Rossi, L., de Alencastro, L.F., Abegglen, C., Thonney, D., Chèvre, N., Schärer, M., Barry, D.A., 2013. Treatment of micropollutants

1207

in municipal wastewater: ozone or powdered activated carbon? Sci. Total Environ. 461–462:480–498. https://doi.org/10.1016/j.scitotenv.2013.05.034. Murphy, K.R., Hambly, A., Singh, S., Henderson, R.K., Baker, A., Stuetz, R., Khan, S.J., 2011. Organic matter fluorescence in municipal water recycling schemes: toward a unified PARAFAC model. Environ. Sci. Technol. 45:2909–2916. https://doi.org/10.1021/ es103015e. Murphy, K.R., Stedmon, C.A., Graeber, D., Bro, R., 2013. Fluorescence spectroscopy and multi-way techniques. PARAFAC. Anal. Methods 5:6557–6566. https://doi.org/ 10.1039/c3ay41160e. Nanaboina, V., Korshin, G.V., 2010. Evolution of absorbance spectra of ozonated wastewater and its relationship with the degradation of trace-level organic species. Environ. Sci. Technol. 44:6130–6137. https://doi.org/10.1021/es1005175. Nöthe, T., Fahlenkamp, H., Von Sonntag, C., 2009. Ozonation of wastewater: rate of ozone consumption and hydroxyl radical yield. Environ. Sci. Technol. 43:5990–5995. https://doi.org/10.1021/es900825f. Ohno, T., 2002. Fluorescence inner-filtering correction for determining the humification index of dissolved organic matter. Environ. Sci. Technol. 36:742–746. https://doi. org/10.1021/es0155276. Oloibiri, V., De Coninck, S., Chys, M., Demeestere, K., Van Hulle, S.W.H., 2017. Characterisation of landfill leachate by EEM-PARAFAC-SOM during physical-chemical treatment by coagulation-flocculation, activated carbon adsorption and ion exchange. Chemosphere 186. https://doi.org/10.1016/j.chemosphere.2017.08.035. Ort, C., Gujer, W., 2006. Sampling for representative micropollutant loads in sewer systems. Water Sci. Technol. 54, 169–176. Ort, C., Lawrence, M.G., Rieckermann, J., Joss, A., 2010. Sampling for pharmaceuticals and personal care products (PPCPs) and illicit drugs in wastewater systems: are your conclusions valid? A critical review. Environ. Sci. Technol. 44:6024–6035. https://doi.org/ 10.1021/es100779n. Roustan, M., Debellefontaine, H., Do-Quang, Z., Duguet, J.-P., 1998. Development of a method for the determination of ozone demand of a water. Ozone Sci. Eng. 20: 513–520. https://doi.org/10.1080/01919519809480338. Schindler Wildhaber, Y., Mestankova, H., Schärer, M., Schirmer, K., Salhi, E., von Gunten, U., 2015. Novel test procedure to evaluate the treatability of wastewater with ozone. Water Res. 75:324–335. https://doi.org/10.1016/j.watres.2015.02.030. Sgroi, M., Roccaro, P., Korshin, G.V., Greco, V., Sciuto, S., Anumol, T., Snyder, S.A., Vagliasindi, F.G.A., 2017a. Use of fluorescence EEM to monitor the removal of emerging contaminants in full scale wastewater treatment plants. J. Hazard. Mater. 323: 367–376. https://doi.org/10.1016/j.jhazmat.2016.05.035. Sgroi, M., Roccaro, P., Korshin, G.V., Vagliasindi, F.G.A., 2017b. Monitoring the behavior of emerging contaminants in wastewater-impacted rivers based on the use of fluorescence excitation emission matrixes (EEM). Environ. Sci. Technol. 51:4306–4316. https://doi.org/10.1021/acs.est.6b05785. Shon, H.K., Vigneswaran, S., Snyder, S.A., 2006. Effluent organic matter (EfOM) in wastewater: constituents, effects, and treatment. Crit. Rev. Environ. Sci. Technol. 36: 327–374. https://doi.org/10.1080/10643380600580011. Shutova, Y., Baker, A., Bridgeman, J., Henderson, R.K., 2014. Spectroscopic characterisation of dissolved organic matter changes in drinking water treatment: from PARAFAC analysis to online monitoring wavelengths. Water Res. 54:159–169. https://doi.org/ 10.1016/j.watres.2014.01.053. Smeti, E.M., Thanasoulias, N.C., Lytras, E.S., Tzoumerkas, P.C., Golfinopoulos, S.K., 2009. Treated water quality assurance and description of distribution networks by multivariate chemometrics. Water Res. 43:4676–4684. https://doi.org/10.1016/j. watres.2009.07.023. Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2003. Wastewater Engineering: Treatment and Reuse. McGraw-Hill, Boston. Van Hulle, S.W.H., Ciocci, M.C., 2012. Statistical evaluation and comparison of the chemical quality of bottled water and Flemish tap water. Desalin. Water Treat. 40:183–193. https://doi.org/10.1080/19443994.2012.671166. Vergeynst, L., K'oreje, K., De Wispelaere, P., Harinck, L., Van Langenhove, H., Demeestere, K., 2017. Statistical procedures for the determination of linearity, detection limits and measurement uncertainty: a deeper look into SPE-LC-Orbitrap mass spectrometry of pharmaceuticals in wastewater. J. Hazard. Mater. 323:2–10. https://doi.org/10.1016/j. jhazmat.2016.05.077. Von Gunten, U., 2003. Ozonation of drinking water: part I. Oxidation kinetics and product formation. Water Res. 37:1443–1467. https://doi.org/10.1016/S0043-1354(02)00457-8. Von Sonntag, C., von Gunten, U., 2012. Chemistry of Ozone in Water and Wastewater Treatment. IWA Publishing, London. Wert, E.C., Rosario-Ortiz, F.L., Snyder, S.A., 2009. Using ultraviolet absorbance and color to assess pharmaceutical oxidation during ozonation of wastewater. Environ. Sci. Technol. 43:4858–4863. https://doi.org/10.1021/es803524a. Westerhoff, P., Aiken, G., Amy, G., Debroux, J., 1999. Relationships between the structure of natural organic matter and its reactivity towards molecular ozone and hydroxyl radicals. Water Res. 33:2265–2276. https://doi.org/10.1016/S0043-1354(98)00447-3. Wittmer, A., Heisele, A., McArdell, C.S., Böhler, M., Longree, P., Siegrist, H., 2015. Decreased UV absorbance as an indicator of micropollutant removal efficiency in wastewater treated with ozone. Water Sci. Technol. 71:980–985. https://doi.org/10.2166/ wst.2015.053. Wu, J., Ma, L., Chen, Y., Cheng, Y., Liu, Y., Zha, X., 2016. Catalytic ozonation of organic pollutants from bio-treated dyeing and finishing wastewater using recycled waste iron shavings as a catalyst: removal and pathways. Water Res. 92:140–148. https://doi. org/10.1016/j.watres.2016.01.053.