Trihalomethanes formation in Iranian water supply systems: predicting and modeling

BDCM

bromodichloromethane

BFM

bromoform

CFM

chloroform

DBCM

dibromochloromethane

DBPs

disinfection by-products

DOC

dissolved organic carbon

MCL

maximum contaminant level

MDLs

method detection limits

NOM

natural organic matter

SUVA

specific ultraviolet absorbance

THMs

trihalomethanes

TOC

total organic carbon

TTHM

total trihalomethane

UVA

ultraviolet absorbance

WDS

water distribution system

WTP

water treatment plant

Chlorine is a widely used disinfectant because of its efficiency, cost-effectiveness, and prevention of water-borne disease transmission (Sadiq & Rodriguez 2004; Pavon et al. 2008; Siddique et al. 2014). Most surface waters contain natural and anthropogenic organic matters that are trihalomethanes (THMs) precursors (Panyapinyopol et al. 2005). Chloroform (CHCl₃), bromodichloromethane (CHCl₂Br), dibromochloromethane (CHClBr₂), and bromoform (CHBr₃) are the main THM species.

Recently, reproductive consequences of THMs in drinking waters such as spontaneous abortion, stillbirth, and intra-uterine growth reduction, and harmful effects on the liver, kidneys, bladder, and central nervous system have been reported (Legay et al. 2010; Richardson & Postigo 2012). Numerous studies have described the formation and relevant health risks of THM compounds (Sohn et al. 2004; Kumar et al. 2014). Toxicological information and guidelines related to THMs (mg/L) are shown in Table 1.

The formation of THMs during chlorination of drinking water needs to be monitored to ensure compliance with guidelines. A mathematical model that can simulate disinfection by-products (DBPs) is very useful in predicting THMs formation. Such models have been developed for various purposes, such as identification of the significant operational and water quality parameters controlling the formation of THMs, investigation of the kinetics of their formation, and THMs prediction as an alternative to field measurements (Sadiq & Rodriguez 2004). Modeling THMs requires the extension of experimental and mathematical relationships between THM levels in treated water and water quality and operational parameters if it is to be effective in THMs formation (Di Cristo et al. 2013; Al-Omari et al. 2014). The presence of THMs in chlorinated waters depends on the pH, water temperature, contact time between water and chlorine, type and dose of disinfectant, bromide ion concentration, and nature and concentration of natural organic matter (NOM) (Ristoiu et al. 2009; Fooladvand et al. 2011). Numerous models have been developed to predict THMs formation in drinking waters by recognizing the effective parameters (Chowdhury et al. 2009; Abdullah & Hussona 2013). These predictive models employ different water quality parameters, and individual models typically use three to eight parameters. However, the interaction parameters are typically not included in the modeling of THMs (Sohn et al. 2004).

The present study is the first attempt to evaluate THMs formation and to develop a predictive model based on multiple regression analysis using raw and treated water characteristics, operational conditions, and the properties of drinking water in two water distribution systems (WDS1 and WDS2) in Ahvaz, Iran. Ahvaz has a population of 1.2 million, and its drinking water is supplied by two water treatment plants (WTP1 and WTP2). The Karun River is the major source of raw water for WTP1 and WTP2. The deterioration of the quality of this river's water is a serious public concern; it receives tonnes of untreated municipal, industrial, and agricultural wastewaters annually. Recently, due to reduced precipitation in the region, the river's pollution has been concentrated, especially in terms of organic matter. Furthermore, the conventional treatment process used in local WTPs cannot guarantee organic matter removal. Consequently, the risk of THMs formation in the finished water is increasing.

In this research, we presented the profile of THMs and factors leading to their formation in WDS1 and WDS2. In addition, based on these results, we developed a predictive multiple regression model and conducted a sensitivity analysis on the model.

WTP1 and WTP2, which supply the water requirements of about 1 million inhabitants of Ahvaz, were monitored and the formation of THMs was modeled. Both plants use conventional treatment process trains and are mainly fed with the Karun River. These plants supply about 500,000 m³ of treated water to the city daily. The treatment trains consist of prechlorination, coagulation by the addition of poly aluminum chloride, flocculation, sedimentation, rapid sand filtration with a backwashing period of 8–12 hours, and final chlorination (typically up to 2.5 mg/L).

We designed the sampling period to cover both cold and warm seasons and so consider the impact of seasonal changes on the water quality. A 36-week sampling period was carried out from January to September 2011. Samples were taken from various locations to provide a comprehensive evaluation of the water quality in WDS1 and WDS2 (Table 2). Ninety water samples were collected to determine THM species. Thirty-six samples were taken from WDS1 and 54 from WDS2. Sampling in the distribution networks was performed biweekly at 10–12 a.m. Before collecting each sample, the faucet was opened for about 5 minutes to collect a representative sample and to ensure that each water sample came from the distribution network (Rodriguez et al. 2007).

For measurement of THMs, tap water was collected in 125-mL amber glass bottles with polypropylene screw caps and TFE-faced septa. Before sampling, we added a sodium thiosulfate solution (10%) to the bottles to eliminate residual chlorine and to prevent further formation of THMs until analysis. Special care was exercised during sampling to avoid any air bubbles in the sample. The samples for analyzes of dissolved organic carbon (DOC), UV absorbance at 254 nm (UV₂₅₄), bromide ion, and ammonia were collected in 1 L amber glass bottles. Before sampling, the bottles were washed with detergent, rinsed with deionized ultra-pure water, and placed in an oven at 400 °C for 1 hour. Once collected, the samples were stored in the dark at 4 °C and analyzed within 24 hours. The temperature, pH, and free residual chlorine of the water samples were measured in situ. The pH and free residual chlorine were determined by a digital Multi 1000 Cl/pH meter (Palintest Company) and the water temperature was measured with a digital thermometer.

Ammonia was analyzed according to the method suggested by the HACH Company, using a DR5000 spectrophotometer (HACH, Germany). Bromide concentration was measured using an ion chromatograph (2695 Alliance model, Waters) equipped with an electrochemical detector (2465 model, Waters) in the calibration range of 6–200 μg/L.

THM species analysis was carried out with the head space technique using a gas chromatograph (6890 model, Agilent) equipped with a μECD detector. A capillary fused silica DB-1701 column 30 m × 0.32 mm × 1.0 μm film thickness was used. Injections were made in splitless mode, with helium as carrier gas and nitrogen as makeup gas. Sample volumes of 10 mL were filled in 20 mL glass vials and spiked with 50 μL of the internal standard (10 mg/L p-bromofluorobenzene). The vials were immediately sealed with Teflon coated septa and aluminum crimp caps. After equilibration (10 minutes at 70 °C) a volume of 100 μL of the head space was injected splitless into a gas chromatograph. The temperature program was: initial 40 °C for 1 minute, 5 °C/min to 100 °C for 0 minute, 100 °C/min to 200 °C for 3 minutes. A linearity regression function was set up based on calibration measurement. There was good linearity in the detected range, and correlation coefficients (R²) were 0.9989, 0.9999, 0.9996, and 0.9992 for CHCl₃, CHCl₂Br, CHClBr₂, and CHBr₃, respectively. Water samples were spiked with a specific amount (1, 10, and 20 μg/L) of four THMs to determine the recovery. The average recoveries of the four THMs were in the range of 84–92% (n = 7) and relative standard deviations were 4.5–9.1% indicating acceptable precision and accuracy of the analytical procedure. The method detection limits (MDLs) were determined from the standard deviation of 1 μg/L spiked samples. The corresponding MDLs for CHCl₃, CHCl₂Br, CHClBr₂, and CHBr₃ were 0.15 μg/L, 0.11 μg/L, 0.12 μg/L, and 0.15 μg/L, respectively.

A multiple regression model was developed for analysis of THMs formation within these two distribution systems, using the forward selection procedure of SPSS 17. The dependent variable was THMs concentration in the distribution systems, and the filtered (i.e. after sand filters) and treated water (i.e. in the distribution systems) qualities and operational parameters from WTP1 and WTP2 were considered as the independent variables. The variables were tested for normality using the Kolmogorov–Smirnov test. We implemented several regression techniques, including linear, logarithmic, exponential, and quadratic in an attempt to formulate and adopt the equations exhibiting the highest coefficient of determination.

Average seasonal values of the filtered (i.e. after sand filters) and distributed water quality parameters and operational parameters of WTP1 and WTP2 are shown in Table 3. These parameters were analyzed as they are important in THMs formation. Apart from the ammonia concentration and Cl₂ dose, whose values were almost two times greater in WDS1 than in WDS2, the values of the remaining parameters were comparable in the two distribution systems. Importantly, it was observed that the values of some of the parameters measured in WDS1 and WDS2 increased as the water temperature increased due to seasonal change from about 17 °C in winter to 30 °C in summer. According to Table 3, the humic nature of NOM (as indicated by SUVA) in the distribution systems increased linearly as the water temperature increased. Bromide also showed a similar trend, and its concentration increased in both distribution systems from about 140 to 171 μg/L. Overall, the elevated concentrations of these parameters in summer resulted in the increased concentrations of THMs in the WDS1 and WDS2 distribution systems.

As shown in Table 4, we observed seasonal and spatial variations in the THMs concentrations. The total concentrations of THMs of Ahvaz's drinking water in winter, spring and summer were 20.5–86 μg/L, 18.92–66.06 μg/L, and 17.35–174.75 μg/L, respectively. In addition, the total THMs concentrations in WDS1 and WDS2 were 17.35–174.75 μg/L and 18.92–99.5 μg/L, respectively. As shown in Figure 1, the maximum seasonal average of the THMs in both distribution systems was related to summer temperature. For WDS1, the maximum value was 70.47 ± 22.3 μg/L and occurred at the end of the distribution system (i.e. Sp1-2), whereas the value for WDS2 was 89.22 ± 60.8 μg/L, measured at the midpoint of the distribution system (i.e. Sp2-2). The minimum average concentration was related to the finished water of the treatment plants, where the THMs concentrations were 25.97 ± 5.17 μg/L and 31.54 ± 8.63 μg/L in Sp1-1 and Sp2-1, respectively. The one-way ANOVA-Tukey test was used to compare the average concentrations of THMs between seasons as well as at the sampling points. Comparison of the average concentration of THMs between sampling points (Table 5) showed a significant difference between the primary point and endpoint of both WDS1 and WDS2 (p_value < 0.05). Thus, we observed an increase in the THMs levels as the distance from treatment plants increased. This spatial variation is related to the long residence time of water within the distribution systems (Rodriguez et al. 2007).

As shown in Figure 1 and Table 4, the brominated THM species were the predominant THM compounds in all of the sampling points of WDS1 and WDS2, accounting for about 80% of the total THMs concentration, whereas, chlorinated THM species accounted for only 20%. Prior studies also reported brominated compounds as the major THM species (Kampioti & Stephanou 2002; Duong et al. 2003).

We compared the THMs concentration in the various seasons to determine whether seasonal values differed significantly, and indeed statistical analysis (ANOVA) did reveal a difference (p_value < 0.05). However, this difference was only significant for the values measured in summer and spring (p_value < 0.05). The seasonal variations in THMs concentration are likely due to the seasonal difference of the water temperatures and bromide concentrations. As shown in Table 3, the maximum level of the water temperature and bromide concentrations in WDS1 and WDS2 were in summer. The seasonal differences in the water temperature can affect reaction rates between chlorine and precursors. Thus, high levels of reaction rates result in the formation of higher levels of THMs (Sohn et al. 2001). The experimental results from batch studies indicated that increased water temperatures in summer yielded a greater THM formation potential for the Karun River water. The summer values were 1.2–1.6 times higher than those of the spring and fall seasons (Fooladvand et al. 2011).

Unlike in colder seasons, summer's higher bromide concentration can be due to the elevated water temperature and higher evaporation rate in the water source (Toroz & Uyak 2005). We observed higher concentration of the bromide in the Karun River during summer than in spring and winter.

We compared the THMs concentrations obtained in this study with the standard values set by the regulatory agencies (i.e. USEPA and WHO). Overall, we observed that the THMs concentrations in WDS1 and WDS2 were less than the maximum contaminant level (MCL) values set by these agencies. However, the concentration of the THMs exceeded the MCL value at a number of sampling times (Figure 2(a) and 2(b)). As can be seen from Figure 2, almost all of the days when the THMs concentrations exceeded the MCL set values were in summer. This observation is consistent with studies reporting the higher likelihood of THMs formation in warmer seasons (Whitaker et al. 2003; Hassani et al. 2010).

A two-tailed Pearson test was employed to determine the correlation matrix for the THMs with regard to the water quality and operational parameters in WDS1 and WDS2 (Table 6). As shown in Table 6, we observed a positive and significant correlation (r = 0.413, p_value = 0.012) between THMs and water temperature in WDS1, while the Pearson correlation test for WDS2 indicated that the THMs concentration has a positive and significant correlation with several parameters, including temperature (r = 0.472, p_value < 0.01), DOC (r = 0.352, p_value < 0.01), UV₂₅₄ (r = 0.311, p_value = 0.022), Br⁻ (r = 0.485, p_value < 0.01), and chlorine dose (r = 0.48, p_value < 0.01). In addition, this test showed a negative correlation between the THMs and free residual chlorine (r = −0.521, p_value < 0.01) in WDS2. The remaining parameters (i.e. pH and ammonia) did not show a correlation with the formation of THMs due to slight variations.

The positive correlations of the THMs with temperature and with soluble organic matter (DOC, UV₂₅₄) indicate that the THMs formation increased as the water temperature and soluble organic matter in surface water increased. This observation agrees with the results reported by several researchers (Toroz & Uyak 2005; Ye et al. 2009). In addition, a negative and significant correlation observed in this study between the free residual chlorine and THMs concentration shows that the seasonal variations in water temperature influenced the reaction rates between chlorine and THMs precursors. This influence led to the decay of more chlorine and reduction of the free residual chlorine, and therefore a higher demand for chlorine dose (Toroz & Uyak 2005).

A positive and significant correlation was seen between the THMs and the Br⁻ concentration, indicating the elevated THMs concentration with increase of the Br⁻ concentration (Zha et al. 2014). In addition, the low UV₂₅₄ and SUVA values in the drinking water demonstrated that the hydrophilic fraction of NOM was greater than the hydrophobic fraction. Therefore, formation of the brominated THMs was expected due to the higher reaction tendency of NOM with Br⁻ than with chlorine (Liang & Singer 2003). In addition, in the presence of chlorine and bromine ions in water containing NOM, the haloform substitution efficiency (i.e. CHX₃ produced per mole X₂ consumed) for bromine is greater than that for chlorine because the bromine has higher electron density and lower bond strength than chlorine atoms (Westerhoff et al. 2004).

Various regression models were exercised to develop a structured model for THMs prediction in the distribution systems. The sample size (n), significance (P_value), and coefficient of correlation (R²) for each model are presented in Table 7. Based on the coefficients of determination, the best fit was obtained from the linear regression model for both WDS1 (R² = 0.47, P_value < 0.05) and WDS2 (R² = 0.54, P_value < 0.05). However, the coefficient of determination value of the THMs linear predictive model for WDS2 was greater than that for the WDS1 (Table 7). This observation holds true for overall comparison of the various regression models applied for WDS1 and WDS2, probably because the number of samples was greater for WDS2 than that for WDS1. Furthermore, the greater coefficient of determination for WDS2 can be attributed to more significant correlation values between THMs and independent variables in WDS2 than in WDS1. The coefficient of determination obtained in this study was comparable to those presented in the literature (Golfinopoulos & Arhonditsis 2002; Abdullah & Hussona 2013). Nikolaou et al. (2004) reported coefficients of determination ranging between 0.38 and 0.58, using field data for development of a predictive model for THMs formation. Although the model developed in this research agreed with prior studies on THMs predictive models, it appears that the parameters affecting THMs formation in water treatment systems require further investigation to increase the precision of these predictive models (Nikolaou et al. 2004).

The highest concentration of THMs, which was observed in summer, was 174.75 μg/L. A comparison of seasonal THM values in the two distribution systems revealed, respectively, THM concentrations in summer 1.6 and 1.4 times higher than the average values obtained in the spring and winter seasons. The results of this study showed that the minimum and maximum concentrations of the THMs in WDS1 were related to the finished water samples and water samples taken from the furthest points of the distribution system, respectively; whereas, the values in WDS2 were related to the finished water samples and water samples taken from the midpoint of the distribution system, respectively. In general, except for a few cases, THM concentrations in the Ahvaz drinking water were lower than the MCL values. The brominated THM species were the predominant THM compound in the Ahvaz drinking water, accounting for 80% of the total THM concentration. The bromide ion level in water is the key factor in controlling THMs bromination. Significant correlations were observed between THMs and the individual variables including temperature, Br⁻, free residual chlorine, DOC, UV₂₅₄ and chlorine dose for WDS2. For WDS1, significant correlation was only observed with temperature. Among the THMs formation predictive models used, the highest coefficient of determination was obtained with the linear regression model for both WDS1 and WDS2. The development of a precise model for prediction of THM formation was a challenging task due to limited data, lack of proper knowledge of THM formation mechanisms, and continuous changes in parameters affecting THM formation. Based on the results of this study, it was concluded that further investigations are required to develop a more precise predictive model for THM formation in drinking WDSs based on laboratory-controlled experimentations as well as field-scale investigations. This research was the first field-scale study on modeling THMs formation in the Ahvaz drinking WDSs. The overall results may help to identify operational strategies for water utility operators to use in minimizing THMs formation and improving decision-making for management of public health risks associated with THMs.

This paper is derived from thesis of Leila Atari. Financial support through the Vice-Chancellor for Research Affairs of Ahvaz Jundishapur University of Medical Sciences, grant No. U-89305, is gratefully acknowledged. The authors have declared no conflict of interest.