Temporal and spatial variations of disinfection by-products in South Taihu's drinking water, Zhejiang Province, China

Chemical disinfectants like free chlorine and chlorine dioxide are commonly applied in drinking water distribution systems due to their cost-effectiveness and efficiency in eliminating microorganisms, thereby enhancing water quality and public health (Benson et al. 2017). However, the formation of harmful disinfection by-products (DBPs), which are produced by the interaction of disinfectants with various inorganic and organic substances (e.g., bromide, humic substances, amino acids, and environmental pollutants) (Yang & Zhang 2016; Jiang et al. 2020) untreated water, has raised concerns (Korotta-Gamage & Sathasivan 2017). Numerous studies have documented the correlation between prolonged exposure to DBPs via ingestion, inhalation, and dermal pathways (Kumari et al. 2015; Wang et al. 2019; Ates et al. 2020) and their toxic synergistic impacts on genes, cells, endocrine systems, fetuses, and cancer in animals (Hrudey et al. 2015; Kolb et al. 2017).

Trihalomethanes (THMs) and haloacetic acids (HAAs) are among the most frequently detected classes of DBPs, originating from the chlorination process of drinking water supplies (Hao et al. 2017). Regulatory bodies such as the United States Environmental Protection Agency (USEPA), the European Community, and the World Health Organization (WHO) recommend respective THMs concentrations of 80 μg/L, 100 μg/L, and [TTHMs/WHO guideline value] ≤ 1 in drinking water (WHO 2004; USEPA 2010). In China, the prescribed concentrations for bromodichloromethane (BDCM), dibromochloromethane (DBCM), tribromomethane (TBM), dichloroacetic acid (DCAA), and trichloroacetic acid (TCAA) are 60, 100, 100, 50, 100 μg/L, respectively, and [TTHMs/respective guideline value] ≤ 1 (MOH 2006). Emerging DBPs such as haloacetonitriles (HANs), haloketones (HKs), halonitromethanes (HNMs), and iodine-DBPs have recently gained attention due to their higher cellular cytotoxicity and genotoxicity than traditional c-DBPs (Jeong et al. 2015; Mao et al. 2016). Additionally, factors including water sources, treatment processes, pipe materials, hydraulic conditions, and sediments on pipe walls contribute to the occurrence of DBPs in different drinking water systems (Abbas et al. 2014; Kali et al. 2021).

Huzhou City, located in the northern part of Zhejiang Province on China's eastern coast, spans an area of 5,818 km² and is home to approximately 2.67 million people. The region's mild temperature and subtropical monsoon climate result in four distinct seasons. Primary surface water sources include Taihu Lake, reservoirs, and the Tiaoxi river. Taihu Lake, one of the five largest freshwater lakes in China, lies downstream of the Tiaoxi river. Over recent decades, population growth and industrial development have led to substantial sewage discharge into rivers and lakes, causing significant ecological and watershed damage (Hong et al. 2015; Wu et al. 2020).

The objectives of this study are (i) to examine the spatial and temporal variations of DBPs in 10 full-scale drinking water systems that employ the chlorine sequential disinfection method in their drinking water treatment plants (DWTPs); (ii) to identify contributing factors (e.g., water quality, operational parameters) to these variations; and (iii) to explore the relationships between different DBP levels to predict future haloacetonitrile (HAL) occurrences.

The DWTPs primarily relied on surface water sources, and chlorine-based disinfectants were employed for water disinfection. The investigated finished or tap water samples were treated using traditional techniques, including coagulation, sedimentation, filtration, and chlorination. Depending on the raw water sources at the DWTPs, we divided the 10 DWTPs into 4 water distribution systems: 2 sourced from reservoir water mixed with river water, 3 from reservoir water systems, 4 from the Tiaoxi water system, and 1 from the Taihu Lake water system.

The collection of the 120 samples followed the GB/T 5750-2006 guidelines. Prior to sampling, we allowed the water at each sampling point to stand for 10–30 min. We then filled one 5 L plastic container, one 2.5 L plastic container, and one 1,000 mL amber glass sample bottle, ensuring the glass bottle was filled to capacity without air gaps. After sample collection, we labeled the respective bottles with pertinent information including the water plant's name, sample type (source water, treated water, distribution endpoint water), collection time, and sample number. All samples were maintained at low temperatures (4 °C) and stored in light-protected conditions before being transported to the Huzhou Municipal Center for Disease Control and Prevention (CDC) laboratory for testing within 4 h.

An Agilent 7890B capillary gas chromatograph and an electron capture detector were purchased from Agilent Corporation (USA). Ultrapure water was produced using a Milli-Q ultrapure grade purification system (Millipore Corporation, USA). Other chemicals were of analytical reagent grade and were obtained from Sinopharm Chemical Reagent Co. Ltd (China). Standard solutions of chloroform (TCM), dichlorobromo-methane (DCBM), chlorodibromo-methane (DBCM), tribromomethane (TBM), dichloroacetonitrile (DCAN), trichloroacetonitrile (TCAN), chlorobromo-acetonitrile (BCAN), dibromoacetonitrile (DBAN), 1,1-dichloroacetone (1,1-dichloro-2-propanone, DCP), 1,1,1-trichloroacetone (1,1,1-trichlo-2-propanone, TCP), and trichloronitromethane (TCNM) were purchased from Accustandard.

We commenced our experiment by utilizing a 1 L brown glass sampling bottle. Prior to sample collection, 1.0 g of ascorbic acid was added to the bottle to inhibit the chlorination reaction. Subsequently, we filled the bottle with water and sealed it.

For the analysis of trihalomethanes (THMs), haloacetonitriles (HANs), halogenated ketones (HKs), and trichloronitromethane (TCNM), we extracted a 10 mL water sample and directly placed it into a 20 mL headspace vial. We further added 3 g of sodium chloride, sealed it immediately, and then shook the vial thoroughly. The sample was then analyzed using a headspace sampler-gas chromatograph as per the established conditions. The Agilent GC ChemStation was responsible for data acquisition, analysis, and subsequent result generation. The process involved using an HP-5 chromatographic column (30 m × 0.32 mm × 1.0 μm) and high-purity nitrogen (99.999%) served as the carrier gas. The temperature programming started at an initial 50 °C, maintained for 5 min, then gradually increased to 100 °C at a rate of 3 °C/min. Other specific parameters were set as follows: inlet temperature at 200 °C, electron capture detector temperature at 300 °C, a flow rate of 1.0 mL/min, and a split ratio of 50:1. The make-up gas flow was set at 30 mL/min. Heating box temperature: 70 °C, loop temperature: 80 °C, transfer line temperature: 90 °C; vial equilibration time: 20 min, pressure equilibration time: 0.1 min, injection time: 1 min; air filling pressure: 15 psi, filling flow: 50 mL/min, shake seven times, and vent after extraction. Injection volume: 1,000 μL.

THMs, HANs, HKs, and TCNM were executed using the headspace salt-assisted gas chromatography-electron capture detector (GC-ECD) method. Our quality control procedure was crafted to mitigate any potential interference from the pre-treatment process on the target substance's determination. We incorporated a process blank experiment to certify the absence of such interference. Each batch of samples was subjected to a blank test using pure water instead of the actual test sample, and no disinfection by-products (DBPs) were detected. Post-verification, we confirmed the experimental process as contamination-free. To create a standard curve, we introduced a volume of mixed standard solution into the headspace bottle resulting in concentrations ranging from 1 to 50 μg/L. This demonstrated a robust linear relationship, with correlation coefficients exceeding 0.998. Utilizing a low-concentration sample spiked at 5 μg/L, we computed the method's detection limits at a signal-to-noise ratio three times. We attained a detection limit for TCAN as low as 0.005 μg/L, while the detection limit for DBAN, possessing the lowest sensitivity, could still reach up to 0.200 μg/L. With factory water as the substrate, the recovery rates and relative standard deviations (RSDs) at 5.0 μg/L spiking fluctuated between 70.62–118.53% and 2.10–8.62%, respectively.

In case of HAAs, we extracted 25 mL of the water sample to be tested directly into a 50 mL extraction bottle, added 2 mL of concentrated sulfuric acid, and mixed thoroughly. About 3 g of anhydrous copper sulfate was swiftly added, followed by approximately 10 g of anhydrous sodium sulfate, shook well. Then, 4.0 mL of methyl tert-butyl ether containing 300 μg/L internal standards (1,2-DBP) was added, shaken, and stood still for 5 min. We took 3.0 mL of the supernatant into another 16 mL derivatization bottle, added 1.0 mL of freshly prepared sulfuric acid-methanol solution, and derivatize it on a heating block at 50 °C for 120 min. Took out the derivatization bottle, added 4 mL of saturated sodium bicarbonate solution dropwise after cooling to room temperature, covered with a stopper, shook and kept deflating; finally, we took 1–1.5 mL of the supernatant into the extraction bottle, and added a small amount of anhydrous sodium sulfate, took 2 μL of supernatant for gas chromatography analysis, completed data acquisition and data analysis by AgilentGC chemical workstation and got the results directly. We used DB-5 ms chromatographic column (30 m × 0.32 mm × 0.25 μm), high-purity nitrogen (99.999%) as a carrier gas, program temperature: initial 50 °C for 7 min, increased to 70 °C at 4 °C/min, then 30 °C/min to 250 °C, and then held for 5 min. The inlet temperature was 200 °C, the electron capture detector temperature was 300 °C, the flow rate was 9.0 mL/min, and the split ratio was 3:1. The make-up gas flow was 30 mL/min.

To ensure quality and control for HAAs, we adopted process blank experiments to verify the absence of interference in the test process. This involved replacing the test samples with pure water in each batch for blank tests. The standard curve, developed from these results, displayed strong linearity, with correlation coefficients across all concentrations exceeding 0.998. The recovery rates for the seven haloacetic acids were as follows: monochloroacetic acid 104.14%, dichloroacetic acid 104.79%, trichloroacetic acid 120.68%, monobromoacetic acid 109.35%, tribromoacetic acid 69.49%, monochlorobromoacetic acid 108.37%, dibromochloroacetic acid 90.47%. For monobromoacetic acid and monobromodichloroacetic acid, due to the unavailability of standard reference materials, their spiked recovery rates were not determined.

Various water quality parameters were measured in a laboratory setting in accordance with the National Standard Methods (Standard Examination Methods for Drinking Water, GB/T 5,750, 2006, as specified by the Chinese Ministry of Health). These parameters include pH (measured with a pH meter), temperature (thermometer), residual chlorine (analyzed by the 3,3–5,5-tetramethylbenzidine colorimetry method), UVA254 (UV-visible spectrophotometry, Shimadzu UV-2450), Cl⁻, and Br⁻ (ion chromatography, Dionex ICS-1500), along with turbidity (assessed using a turbidimeter, HACH 2100N).

We computed fundamental descriptive statistics for each individual variable. To identify seasonal differences in water quality parameters, we employed nonparametric hypothesis testing. Results that yielded two-tailed p-values of less than 0.05 were considered statistically significant. To explore the correlation between disinfection by-products (DBPs) and water quality parameters, we utilized both Spearman correlation analysis and principal component analysis. All statistical computations were conducted using SPSS software (Version 24).

All collected data underwent a normality test and exhibited a skewed distribution. We utilized nonparametric analysis to perform comparisons between the datasets. Table 2 depicts the characteristics of water quality and operational parameters across different seasons in Huzhou City. Temperature, turbidity, and BOD all demonstrated significant seasonal variability (p < 0.05). The median turbidity level in drinking water exhibited a spring > winter > autumn > summer pattern, which might be attributed to rainfall carrying particulate matter into the source water. The warmer temperatures in summer and autumn compared to spring and winter could provide a more favorable environment for the growth of bacteria and other microorganisms, thereby contributing to an increase in COD by releasing organic matter into the source water (Fang et al. 2010; Ly et al. 2017). However, certain drinking water parameters, such as pH, Cl⁻, Br⁻, UVA₂₅₄, and residual chlorine, did not exhibit any seasonal changes, potentially indicating that water treatment technology was not season-dependent and that consistent methodology had been applied throughout the year.

As source water variation can influence water quality parameters, we analyzed these parameters across different source waters in Huzhou City, as detailed in Table 3. The analysis revealed that the pH of Taihu water is higher than that of other source waters. This could be linked to eutrophication in Taihu Lake, where algal blooms consume large amounts of CO₂ during photosynthesis, subsequently elevating the pH of Taihu water (Wu et al. 2018). Other water quality parameters remained unchanged across different source waters, suggesting that these parameters were largely consistent regardless of the water source.

Table 4 presents the concentrations of DBPs in the drinking water of Huzhou city. For the skewed distribution of the data results, we used median values to depict each compound's concentration. THMs and HAAs were the principal constituents of DBPs in all samples, accounting for 34 and 64% of all DBPs, respectively. Together, they represented approximately 98% of the total DBPs, which was higher than the levels found in the drinking water of Beijing city (Wei et al. 2010). The respective levels of THMs and HAAs in the Huzhou region were 14.5 and 27.4 μg/L, which were lower than those reported in Iran's Bushehr Province (Dobaradaran et al. 2020), China's Haihe River, and Nigeria (Benson et al. 2017), were higher than those found in Bohai Bay, China (Niu et al. 2017), and India's river water (Padhi et al. 2019). TCM was the dominant component of THMs, with a concentration of 11.35 μg/L. Among HAAs, TBAA and DBCAA had the highest mean levels, peaking at 60.60 and 82.57 μg/L, respectively. The concentrations of other DBPs such as HANs, HKs, and HNMs were considerably lower than those of THMs and HAAs. The elevated levels of THMs and HAAs, relative to other DBPs in Huzhou city, could be attributed to several factors: (1) The water in this study displayed higher levels of UVA₂₅₄ and COD, indicating an elevated DOC level which usually leads to a high level of DBPs (Zhou et al. 2019). Additionally, the disinfection duration in the waterworks could affect the generation of DBPs (Mompremier et al. 2018). (2) Nine out of 10 water treatment plants utilized sodium hypochlorite for disinfection, with only one using ozone. Chlorine-based disinfection tends to produce higher levels of DBPs in water than ozone disinfection (Zheng et al. 2017).

Table 5 demonstrates the relationship between each group of DBPs and specific water quality parameters. UV₂₅₄ exhibited a significant negative correlation (p < 0.05) with HKs, while COD displayed a notable positive correlation (p < 0.05) with HANs. However, no discernable relationship was observed between COD or UV₂₅₄ and other DBPs groups. Alterations in UV₂₅₄ and COD typically induce changes in the generation and concentration of DBPs (Zhou et al. 2019). The removal of organics might explain the absence of a clear correlation between UV254, COD, and DBPs (Hong et al. 2015). The variance in organic matter removal rates across water plants in Huzhou possibly influences the formation of DBPs. Additionally, while UV254 and COD serve as indicators of organic matter in the water (Leziart et al. 2019), they lack the capability to discern whether such matter constitutes precursor compounds of DBPs (Ates et al. 2020).

Our study revealed a positive influence of temperature on the formation of HAAs, THMs, and HANs. These findings diverge slightly from those of Hong et al. (2013) and Yang et al. (2007). Several studies have documented that rising temperatures can catalyze DBP formation, as a warmer environment augments reaction rates. Conversely, elevated temperatures may impede the formation of unstable DBPs (such as HAAs, HANs, HKs) by accelerating their degradation rate (Chowdhury et al. 2011; Stefán et al. 2019). Thus, variations in temperature ranges can elicit either positive or negative shifts in DBP formation.

Apart from HNMs, our study found no significant correlation between pH and most DBPs, aligning with prior research (Zhou et al. 2019). This lack of notable association could be attributed to the comparable geographical environment and economic conditions of Huzhou and Jinhua, where variations in pH were relatively insignificant. Consequently, most DBPs did not exhibit a pronounced relationship with pH.

Turbidity demonstrated a substantial negative correlation (p < 0.01) with THMs, HANs, HKs, and HNMs. Turbidity is caused by a heterogeneous mixture encompassing various particle types and materials, including particles, colloids, and dissolved substances (Sillanpää et al. 2018). Notably, when these particles require chlorine, they can interfere with the chlorination disinfection process (Leziart et al. 2019), leading to a consequent reduction in the generation of DBPs.

The Cl⁻ demonstrated a significant (p < 0.05) positive association with most DBPs (HAAs, THMs, HANs). Concurrently, residual chlorine was found to correlate positively with HAAs, THMs, and HNMs. As the concentration of Cl⁻ and residual chlorine increases, the oxidation of organics in water samples is heightened, leading to an amplified formation of HAAs and THMs. This increase is directly proportional to the chlorine dose and reaction time (Zhang et al. 2017). However, it is worth noting that Cl⁻ and residual chlorine showed no significant relationship with HKs. This lack of correlation could be influenced by variables, such as pH, contact time, and overall water quality.

A principal component analysis was employed to investigate the potential factors influencing DBPs formation. The results revealed that four components explained 60.9% of the total variance (Table 6). The first component, with an initial eigenvalue of 3.084, accounted for 23.7% of the total variance, with significant loadings on HANs (0.898), HKs (0.885), HNMs (0.678), and THMs (0.589). This suggests that the first component represents the similar formation characteristics of these DBPs. The second component, with an initial eigenvalue of 1.716, accounted for 13.2% of the total variance and was significantly influenced by temperature (0.835), COD (0.828), and HAAs (0.433). This component suggests the profound role of temperature and organic matter in HAAs formation, thus influencing DBPs formation (Kim et al. 2017). The third component, having an initial eigenvalue of 1.676, explained 12.9% of the total variance. It was influenced by residual chlorine (0.802), turbidity (−0.619), and HAAs (0.567). This indicates that turbidity may impact the disinfection effect, signified by the consumption of chlorine and influenced HAAs formation. The fourth component, with an initial eigenvalue of 1.440, accounted for 11.1% of the total variance and was linked to Cl⁻ (0.858), Br⁻ (−0.608), and pH (0.596), which are fundamental constituents of water. During drinking water treatment procedures, the presence of Br⁻ and Cl⁻ significantly affected DBPs formation, and this process was pH-dependent.

This study examined the occurrence of 5 DBPs and relevant water quality parameters at 10 DWTPs of Huzhou. The results demonstrated that THMs and HAAs were the predominant DBPs across all samples, lower than those reported in Bushehr Province, Iran, Haihe River, China, and Nigeria, but higher than observations in Bohai Bay, China, and Indian river water. Seasonal variations indicated a similar trend in HAAs, THMs, HANs, and HNMs with concentrations in summer or autumn > spring. When considering water source variations, the levels of HAAs and THMs in Taihu water surpassed other sources. In a spatial context, the aggregate DBPs and HAAs levels were the highest in Nanxun, followed by Anji, while lower levels were found in Deqing and Changxing. The DBPs levels tended to increase with proximity to Taihu Lake. Correlation analysis between water quality parameters and DBPs indicated that chloride (Cl⁻ , in mg/L), temperature, and residual chlorine positively influenced the formation of DBPs, while turbidity exerted a negative effect. Principal component analysis further revealed that the formation characteristics of HANs, HKs, HNMs, and THMs were similar. Temperature, COD, and residual chlorine played crucial roles in the occurrence of HAAs, which might be associated with the pollution in Taihu and local climatic conditions. In conclusion, factors such as the formation and relations of HANs, HKs, HNMs, and THMs, organic matter content, chlorine consumption, and water constitution significantly contribute to the generation of DBPs.

This study was supported by the Zhejiang Province Public Technology Application Research Project (2016C37022), Zhejiang Province Public Technology Application Research Project (2016C33218), Zhejiang Province Medical and Health Science and Technology Project (2018RC072), Huzhou Science and Technology Project (2018GYB03), Key Laboratory of Emergency detection for Public Health of Huzhou. Key Supporting Discipline of Huzhou Medical.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.