The relationship between disinfection by-products formation and characteristics of natural organic matter in raw water

Exploring the potential of machine learning to understand the occurrence and health risks of haloacetic acids in a drinking water distribution system

Determining the occurrence of disinfection byproducts (DBPs) in drinking water distribution system (DWDS) remains challenging. Predicting DBPs using readily available water quality parameters can help to understand DBPs associated risks and capture the complex interrelationships between water quality and DBP occurrence. In this study, we collected drinking water samples from a distribution network throughout a year and measured the related water quality parameters (WQPs) and haloacetic acids (HAAs). 12 machine learning (ML) algorithms were evaluated. Random Forest (RF) achieved the best performance (i.e., R2 of 0.78 and RMSE of 7.74) for predicting HAAs concentration. Instead of using cytotoxicity or genotoxicity separately as the surrogate for evaluating toxicity associated with HAAs, we created a health risk index (HRI) that was calculated as the sum of cytotoxicity and genotoxicity of HAAs following the widely used Tic-Tox approach. Similarly, ML models were developed to predict the HRI, and RF model was found to perform the best, obtaining R2 of 0.69 and RMSE of 0.38. To further explore advanced ML approaches, we developed 3 models using uncertainty-based active learning. Our findings revealed that Categorical Boosting Regression (CAT) model developed through active learning substantially outperformed other models, achieving R2 of 0.87 and 0.82 for predicting concentration and the HRI, respectively. Feature importance analysis with the CAT model revealed that temperature, ions (e.g., chloride and nitrate), and DOC concentration in the distribution network had a significant impact on the occurrence of HAAs. Meanwhile, chloride ion, pH, ORP, and free chlorine were found as the most important features for HRI prediction. This study demonstrates that ML has the potential in the prediction of HAA occurrence and toxicity. By identifying key WQPs impacting HAA occurrence and toxicity, this research offers valuable insights for targeted DBP mitigation strategies.

Total organic halogen (TOX) in drinking water: Occurrence, correlation analysis, and precursor removal during drinking water treatment

Total organic halogen (TOX) in drinking water provides a measurement of the overall organic halogenated disinfection by-products (DBPs) formed during disinfection. Yangtze River Delta is one of the regions with the highest population density, the fastest urbanization process, and the most severe water pollution in China. Collecting water samples from full-scale drinking water treatment plants (DWTPs) in this region, this study firstly surveyed TOX occurrence in drinking water. Besides, the correlation of TOX formation potential (TOXFP) and trihalomethane formation potential (THMFP) with general water quality parameters (e.g., dissolved organic carbon [DOC], UV254, and specific ultraviolet absorbance) and the removal efficiencies of TOX precursors by different water treatment processes were also investigated. TOX levels in DWTP effluents (i.e., finished water) ranged from 29 to 165 μg/L (median 67 μg/L), and those in simulated distribution system waters ranged from 101 to 276 μg/L (median 158 μg/L). There were generally higher linear regression coefficient values for raw water (R2 = 0.51-0.88) than for treated water (R2 = 0.33-0.64) in terms of the relationship between DBP formation potentials and general parameters. However, a relatively stronger correlation between THMFP and TOXFP was observed for treated water (R2 = 0.80, p < 0.001) than for raw water (R2 = 0.64, p < 0.001). The overall treatment efficiencies of investigated parameters in DWTPs generally followed the order of UV254 > DOC > TOX precursors > THM precursors. Notably, the overall removal rates of DOC and TOX precursors in summer (averaging 59 % and 54 %, respectively) were obviously higher than those in winter (averaging 39 % and 38 %, respectively), which was assumed to be related to the seasonal variation of bioactivity in sand filter. These results could expand the knowledge of TOX in drinking water, and provide valuable perspectives to water industry and DBP research.

Rainstorms drive export of aromatic and concurrent bio-labile organic matter to a large eutrophic lake and its major tributaries

Lakes are hotspots for global carbon cycling, yet few studies have explored how rainstorms alter the flux, composition, and bio-lability of dissolved organic matter (DOM) in inflowing rivers using high-frequency monitoring. We conducted extensive campaigns in the watershed of Lake Taihu and made daily observations for three years in its two largest inflowing tributaries, River Dapu and River Yincun. We found higher DOC, bio-labile DOC (BDOC), and specific UV absorbance (SUVA₂₅₄) levels in the northwestern inflowing regions compared with the remaining lake regions. DOC and BDOC increased during rainstorms in River Dapu, and DOC declined due to local dilution and BDOC increased during rainstorms in River Yincun. We found that rainstorms resulted in increased DOM absorbance a₃₅₀, SUVA₂₅₄, and humification index (HIX) and enhanced percentages of humic-like fluorescent components, %polycyclic condensed aromatic and %polyphenolic compounds as revealed from ultrahigh-resolution mass spectrometry (FT-ICR MS), while spectral slope (S_275-295) and the percentages of protein-like C1 and C3 declined during rainstorms compared with other seasons. This can be explained by a combined flushing of catchment soil organic matter and household effluents. The annual inflows of DOC and BDOC to Lake Taihu were 1.15 ± 0.18 × 10⁴ t C yr^-1 and 0.23 ± 0.06 × 10⁴ t C yr^-1 from River Dapu and 2.92 ± 0.42 × 10³ t C yr^-1 and 0.53 ± 0.07 × 10³ t C yr^-1 from River Yincun, respectively, and the fluxes of DOC and BDOC from both rivers increased during rainstorms. We found an elevated frequency of heavy rainfall and rainstorms in the lake watershed during the past six decades. We conclude that an elevated input of terrestrial organic-rich DOM with concurrent high aromaticity and high bio-lability from inflowing rivers is likely to occur in a future wetter climate.

Changes in the Dissolved Organic Matter Characteristics Released from Sediment According to Precipitation in the Namhan River with Weirs: A Laboratory Experiment

In this study, changes in the properties of dissolved organic matter (DOM) released from sediments into water layers were investigated. To analyze the spatial and temporal variation in dissolved organic carbon (DOC), sediment and bottom water samples were collected upstream of the Gangcheon, Yeoju, and Ipo weirs of the Namhan River during the rainy and non-rainy seasons. The initial DOC was correlated with precipitation (R² = 0.295, p = 0.034) and residence time (R² = 0.275, p = 0.040). The change in the bottom water DOC concentration resulted from the DOC released from the sediments, which may cause water quality issues in the bottom water. The fluorescence analysis revealed that the DOM contained higher levels of hydrophilic and low-molecular-weight (LMW) organic matter in the non-rainy season and higher levels of hydrophobic and high-molecular-weight (HMW) organic matter in the rainy season. Since the Namhan River is the main resource of drinking water for the Seoul metropolitan area, our results can help to optimize the drinking water treatment process by reflecting the DOM characteristics that vary with the seasons. Furthermore, the statistical analysis confirmed that the nutrient content of pore-water and sediment can be used to estimate the DOM release rate from the sediment to the water layer. The results of this study provide a better understanding of DOM movement in aquatic ecosystems and the influences of rainfall on the water quality of the surface waterbody.

Effect of the Recycling Process on Drinking Water Treatment: Evaluation Based on Fluorescence EEM Analysis Using the Peak-Picking Technique and Self-Organizing Map

The recycling process is applied in many water treatment plants (WTPs), although this process can lead to adverse effects. The effect of the recycling process on the characteristics of dissolved organic matter was evaluated based on a fluorescence excitation-emission matrix using the peak-picking technique and self-organizing map (SOM). In this study, an evaluation of two WTPs, one with and one without a recycling system, was carried out. Both WTPs show moderate efficiency during the coagulation–flocculation process in removing DOC, fulvic acid-like, humic acid-like, and tryptophan-like substances. The recycling process causes increased values of fulvic acid-like, humic acid-like, and tryptophan-like substances and specific ultraviolet absorbance (SUVA) after the filtration process of about 31.0%, 35.7%, 22.2%, and 6%, respectively. Meanwhile, the WTP without recycling showed a reduction in the level of fulvic acid-like, humic acid-like, and tryptophan-like substances and SUVA by 23.3%, 52.9%, 27.8%, and 21.1%, respectively. Moreover, SOM analysis based on the peak-picking technique can determine differences in sample clusters due to the recycling process.

Health risk and source assessment of semi-volatile phenols, p-chloroaniline and plasticizers in plastic packaged (sachet) drinking water

The presence of U.S. EPA priority organic contaminants in drinking water poses a dire health risk on consumers. Packaged drinking water such as plastic sachet drinking water has significantly gained market in both developed and developing countries, especially, its dominance in the Ghanaian market. The treatment process, packaging, and storage of the sachet drinking water contribute to the levels of genotoxic semi-volatile phenols, p-chloroaniline, and plasticizers contamination in the drinking water. The study thus sought to investigate the levels of semi-volatile phenols, p-chloroaniline, and plasticizer contaminants in sachet drinking water on the Ghanaian market and the associated health risk of exposure. The study also investigated the possible sources of the contaminants. A total of thirty (30) different brands of sachet water on the Ghanaian market were studied. The samples were extracted in replicates (n = 3) using Solid Phase Extraction (SPE) cartridges and further analysed with GC-MS (SIM mode). The source apportionment was conducted using absolute principal component analysis coupled with multiple, linear regression (APCA-MLR) and automatic linear regression (APCA-MALR) modelling. The mean total levels for the phenols, p-chloroaniline, and plasticizers were between 210.2 and 18,914.9, 11.2 and 18,871.0, and 21.2 and 69,834.1 ng/L respectively. The cumulative non-cancer risk (hazard quotient) and cancer risk upon exposure were computed to range between 2.1 × 10^-3 and 1.2 and 1.5 × 10^-7 and 1.3 × 10^-4 respectively. About 37% of the samples had elevated cancer risk (>10^-6) which may contribute to the existing incidence, cause for concern. The five sources found for the contaminants were apportioned as "environmental background (major)", "water treatment/disinfectant", "plastic/plasticizers", "storage and preservation", and "residual inter-conversion/degradation sources".

Rainstorm events shift the molecular composition and export of dissolved organic matter in a large drinking water reservoir in China: High frequency buoys and field observations

Rainstorm events can flush large amounts of terrestrial organic-rich material into lakes that are used for drinking water. To date, few studies have been carried out to investigate how rainstorm events change the molecular composition, bio-lability, and flux of upstream-imported dissolved organic matter (DOM), which can impact the odor and taste of drinking water as well as the efficiency of wastewater treatment. We undertook high-frequency buoy monitoring and point sample collection (n = 495), during high, moderate, and low inflow discharge, in Lake Qiandao, a key drinking water source for about 10 million people. Data from two online fluorescent DOM sensors deployed and field samples collected at the river site, Jiekou, and the lake site, Xiaojinshan, showed that rainstorm events increased the specific UV absorbance (SUVA₂₅₄), humification index (HIX), humic-like components (C1-C2), and FT-ICR MS derived condensed aromatic and polyphenolic compounds (p < 0.001) and decreased the spectral slope of DOM (S_275-295), spectral slope ratio (S_R), biological index (BIX), and highly bio-degradable peptide-like and aliphatic substances (p < 0.001). Our results suggest that rainstorm events enhanced the export to the lake of colored, hydrophobic, and aromatic DOM. Upstream-derived dissolved organic carbon (DOC) concentrations decreased (p < 0.001), while DOC bio-availability (BDOC) increased only slightly (p < 0.05) during rainstorm events. The loss rate of DOC in Lake Qiandao is 0.82 × 10⁴ t C yr^-1, of which 0.30 × 10⁴ t C yr^-1 is highly bio-labile, and higher occurrences of both ≥ 25 mm d ^- 1 and ≥ 50 mm d ^- 1 rainfall events are anticipated by linear fittings for this region in the future. The application of in situ fluorescence sensors provides an early warning of DOC surge incidents caused by rainstorm events and may be useful in advising drinking water treatment plant managers of changes in raw water DOM quality and treatability.

Predicting unregulated disinfection by-products in small water distribution networks: an empirical modelling framework

Disinfection is used to deactivate pathogens in drinking water. However, disinfectants react with natural organic matter present in water to form disinfection by-products (DBPs). While a few of these DBPs have been studied extensively and are regulated in many countries, new unregulated DBPs (UR-DBPs) have also recently been identified in drinking water. The UR-DBPs are considered to be more toxic than regulated DBPs (R-DBPs). To understand the occurrence of UR-DBPs in a water distribution network (WDN), this research presents an approach to predicting the behaviour of emerging UR-DBPs such as dichloroacetonitrile (DCAN), trichloropropanone (TCP), and trichloronitromethane (TCNM) in WDNs. Water quality data, generated by sampling and laboratory analysis of 12 small communities, was used to develop predictive models. A framework was also proposed alongside the predictive models to estimate the concentration of emerging UR-DBPs under limited water quality sampling information. Moreover, the relationship between emerging UR-DBP concentrations and their identified predictors was further observed and evaluated by developing contour profiles. DCAN and TCP predictive models have coefficient of determination (R²) > 85%, whereas for TCNM model, the R² was > 65%. Water quality parameters including water temperature, turbidity, conductivity, and dissolved organic carbon concentrations were identified as key predictors. Similarly, trichloroacetic acid and bromodichloromethane were identified as key predictors among DBP families, to predict the occurrence of emerging UR-DBPs. Developed models and relationships between the UR-DBPs and predictors can help water utilities and regulators to manage the occurrence of UR-DBPs in small WDNs.

Strengths of correlations with formation of chlorination disinfection byproducts: effects of predictor type and other factors

Measurements of the UV-Vis absorbance (Abs) and intensity of fluorescence emission (Fluor), as well as of concentrations of total or dissolved organic carbon (OC) in aqueous samples are commonly used to estimate the potential for disinfection byproducts (DBPs) formation during water chlorination. In this work, based on 574 linear associations collected from 70 experimental research papers published over the period of 1997-2019, the strengths of the correlations of Abs, Fluor, and OC with DBPs concentrations are compared. The correlations were expressed as approximately normally distributed Z-scores using Fisher variance-stabilizing transformation. The effects of specific prediction method, chlorination agent, water source, and DBPs type, with consideration of possible effects due to the presence of bromide, are examined against Z-scores by ANOVA, testing main effects and some variables interactions. The performed analysis is a first attempt to expose differences and patterns in correlation strengths associated with DBPs formation, based on systematically covered broad existing literature. Abs and OC concentration of water samples tend to demonstrate the strongest correlations with DBPs formation as compared with specific UV absorbance (SUVA) or intensity of fluorescence emission. Correlations of DBPs formation during chloramination demonstrated weaker strengths as compared with other chlorination agents, suggesting more caution in predicting DBPs concentrations, based on simple descriptors such as Abs, OC, and Fluor. In a series of different water types, the correlations with DBPs formation are expected to be enhanced, when wastewater is chlorinated. Non-fluorescent matter may be an important contributor to DBPs formation during water chlorination. When fluorescence intensity is considered as a predicting tool, choosing humic-like rather than proteinaceous fluorescence may enhance the strengths of the correlations with DBPs formation. Different performances of Abs, OC, and Fluor in correlating with DBPs formation may be beneficial for their concurrent use helping to optimize removal of different DBPs precursors.

THM and HAA formation from NOM in raw and treated surface waters

The disinfection by-product (DBP) formation potential (FP) of natural organic matter (NOM) in surface water sources has been studied with reference to the key water quality determinants (WQDs) of UV absorption (UV₂₅₄), colour, and dissolved organic carbon (DOC) concentration. The data set used encompassed raw and treated water sampled over a 30-month period from 30 water treatment works (WTWs) across Scotland, all employing conventional clarification. Both trihalomethane (THM) and haloacetic acid (HAA) FPs were considered. In addition to the standard bulk WQDs, the DOC content was fractionated and analysed for the hydrophobic (HPO) and hydrophilic (HPI) fractions. Results were quantified in terms of the yield (dDBPFP/dWQD) and the linear regression coefficient R² of the yield trend. The NOM in the raw waters was found to comprise 30-84% (average 66%) of the more reactive HPO material, with this proportion falling to 18-63% (average 50%) in the treated water. Results suggested UV₂₅₄ to be as good an indicator of DBPFP as DOC or HPO for the raw waters, with R² values ranging from 0.79 to 0.82 for THMs and from 0.71 to 0.73 for HAAs for these three determinants. For treated waters the corresponding values were significantly lower at 0.52-0.67 and 0.46-0.47 respectively, reflecting the lower HPO concentration and thus UV₂₅₄ absorption and commensurately reduced precision due to the limit of detection of the analytical instrument. It is concluded that fractionation offers little benefit in attempting to discern or predict chlorinated carbonaceous DBP yield for the waters across the geographical region studied. UV₂₅₄ offered an adequate estimate of DBPFP based on a mean yield of ∼2600 and ∼2800 μg per cm^-1 absorbance for THMFP for the raw and treated waters respectively and ∼3800 and2900 μg cm^-1 for HAAFP, albeit with reduced precision for the treated waters.

Water quality following extensive beetle-induced tree mortality: Interplay of aromatic carbon loading, disinfection byproducts, and hydrologic drivers

The recent bark beetle epidemic across western North America may impact water quality as a result of elevated organic carbon release and hydrologic shifts associated with extensive tree dieback. Analysis of quarterly municipal monitoring data from 2004 to 2014 with discretization of six water treatment facilities in the Rocky Mountains by extent of beetle impact revealed a significant increasing trend in total organic carbon (TOC) and total trihalomethane (TTHM) production within high (≳50% areal infestation) beetle-impacted watersheds while no or insignificant trends were found in watersheds with lower impact levels. Alarmingly, the TTHM concentration trend in the high impact sites exceeded regulatory maximum contaminant levels during the most recent two years of analysis (2013-14). To evaluate seasonal differences, explore the interplay of water quality and hydrologic processes, and eliminate variability associated with municipal reporting, these treatment facilities were targeted for more detailed surface water sampling and characterization. Surface water samples collected from high impact watersheds exhibited significantly higher TOC, aromatic signatures, and disinfection byproduct (DBP) formation potential than watersheds with lower infestation levels. Spectroscopic analyses of surface water samples indicated that these heightened DBP precursor levels are a function of both elevated TOC loading and increased aromatic character. This association was heightened during precipitation and runoff events in high impact sites, supporting the hypothesis that altered hydrologic flow paths resulting from tree mortality mobilize organic carbon and elevate DBP formation potential for several months after runoff ceases. The historical trends found here likely underestimate the full extent of TTHM shifts due to monitoring biases with the extended seasonal release of DBP precursors increasing the potential for human exposure. Collectively, our analysis suggests that while water quality impacts continue to rise nearly one decade after infestation, significant increases in TOC mobilization and DBP precursors are limited to watersheds that experience extensive tree mortality.

Characterization of dissolved organic matter for prediction of trihalomethane formation potential in surface and sub-surface waters

Dissolved organic matter (DOM) in surface waters used for drinking purposes can vary markedly in character dependent on their sources within catchments. The character of DOM further influences the formation of disinfection by products when precursor DOM present in drinking water reacts with chlorine during disinfection. Here we report the development of models that describe the formation potential of trihalomethanes (THMFP) dependent on the character of DOM in waters from discrete catchments with specific land-use and soil textures. DOM was characterized based on UV absorbance at 254 nm, apparent molecular weight and relative abundances of protein-like and humic-like compounds. DOM character and Br concentration (up to 0.5 mg/L) were used as variables in models (R(2)>0.93) of THMFP, which ranged from 19 to 649 μg/L. Chloroform concentration (12-594 μg/L) and relative abundance (27-99%) were first modeled (R(2)>0.85) and from these, the abundances of bromodichloromethane and chlorodibromomethane estimated using power and exponential functions, respectively (R(2)>0.98). From these, the abundance of bromoform is calculated. The proposed model may be used in risk assessment of catchment factors on formation of trihalomethanes in drinking water, in context of treatment efficiency for removal of organic matter.

Predicting total organic halide formation from drinking water chlorination using quantitative structure-property relationships

Chlorinating water which contains dissolved organic matter (DOM) produces disinfection byproducts, the majority of unknown structure. Hence, the total organic halide (TOX) measurement is used as a surrogate for toxic disinfection byproducts. This work derives a robust quantitative structure-property relationship (QSPR) for predicting the TOX formation potential of model compounds. Literature data for 49 compounds were used to train the QSPR in moles of chlorine per mole of compound (Cp) (mol-Cl/mol-Cp). The resulting QSPR has four descriptors, calibration [Formula: see text] of 0.72 and standard deviation of estimation of 0.43 mol-Cl/mol-Cp. Internal and external validation indicate that the QSPR has good predictive power and low bias (‰<‰1%). Applying this QSPR to predict TOX formation by DOM surrogates - tannic acid, two model fulvic acids and two agent-based model assemblages - gave a predicted TOX range of 136-184 µg-Cl/mg-C, consistent with experimental data for DOM, which ranged from 78 to 192 µg-Cl/mg-C. However, the limited structural variation in the training data may limit QSPR applicability; studies of more sulfur-containing compounds, heterocyclic compounds and high molecular weight compounds could lead to a more widely applicable QSPR.

An overview of the methods used in the characterisation of natural organic matter (NOM) in relation to drinking water treatment

Natural organic matter (NOM) is found in all surface, ground and soil waters. During recent decades, reports worldwide show a continuing increase in the color and NOM of the surface water, which has an adverse affect on drinking water purification. For several practical and hygienic reasons, the presence of NOM is undesirable in drinking water. Various technologies have been proposed for NOM removal with varying degrees of success. The properties and amount of NOM, however, can significantly affect the process efficiency. In order to improve and optimise these processes, the characterisation and quantification of NOM at different purification and treatment processes stages is important. It is also important to be able to understand and predict the reactivity of NOM or its fractions in different steps of the treatment. Methods used in the characterisation of NOM include resin adsorption, size exclusion chromatography (SEC), nuclear magnetic resonance (NMR) spectroscopy, and fluorescence spectroscopy. The amount of NOM in water has been predicted with parameters including UV-Vis, total organic carbon (TOC), and specific UV-absorbance (SUVA). Recently, methods by which NOM structures can be more precisely determined have been developed; pyrolysis gas chromatography-mass spectrometry (Py-GC-MS), multidimensional NMR techniques, and Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS). The present review focuses on the methods used for characterisation and quantification of NOM in relation to drinking water treatment.

Characterization of disinfection byproduct formation potential in 13 source waters in China

The formation potential of four trihalomethanes (THMFP) and seven haloacetic acids (HAA7FP) in 13 source waters taken from four major water basin areas in China was evaluated using the simulated distribution system (SDS) chlorination method. The specific ultraviolet absorbance (SUVA254: the ratio of UV254 to dissolved organic carbon (DOC)), which ranged between 0.9 and 5.0 L/(mg x m), showed that the organic compounds in different source waters exhibited different reactivities with chlorine. The HAA7FP of source waters ranged from 20 to 448 microg/L and the THMFP ranged from 29 to 259 microg/L. The HAA7FP concentrations were higher than the THMFP concentrations in all but one of the samples. Therefore, the risks of haloacetic acids (HAAs) should be of concern in some source waters. TCM (chloroform) and BDCM (bromodichloromethane) were the major THM constituents, while TCAA (trichloroacetic acid) and DCAA (dichloroacetic acid) were the major HAA species. Br-THM (brominated THM species) were much higher than Br-HAA (brominated HAA species), and the formation of Br-DBP (Br-THM and Br-HAA) should be of concern when the bromide concentration is over 100 microg/L.