Water samples were collected from Chao Phraya River (CP) (Bangkok, Thailand) in December 2023. Suspended solids were removed using a 0.7 μm glass-fiber filter (GF/F, Whatman), and UV-Vis absorbance (200–600 nm) was measured with a 1-cm quartz cell and a DR6000 spectrophotometer (Hach, USA). The differential UV spectrum at 272 nm (ΔA₂₇₂) was calculated as the difference between the UV absorbance of the sample before and after chlorination [12]. Dissolved organic carbon (DOC) and total dissolved nitrogen (TDN) concentrations were analyzed using a multi N/C 2100 analyzer (Analytik Jena). Alkalinity was determined by titration method (APHA 2320B). Bromide concentrations were measured by colorimetric method using a DR6000 spectrophotometer following the APHA 4500 phenol red colorimetric method. Characteristics of the raw water sample are shown in Table S1 of the supplementary materials.

The water samples were then fractionated into HPI and HPO fractions using an ENV bond elute® cartridge [20]. Fluorescence EEM spectra were measured using an FS5 spectrofluorometer (Edinburgh, UK) using excitation wavelengths from 200 to 500 nm in 5 nm increments, with emissions recorded from 250 to 550 nm in 1 nm increments. EEM data were blank-subtracted and normalized to the area under the water Raman curve, following the procedure in a previous study [21]. The correction matrix for the inner-filtering effect was created from the UV-vis spectra [22]. For the chlorinated samples, sodium sulfite (Na₂SO₃) was selected as a chlorine-quenching agent, as its effect on the optical properties of the samples was minimal [23]. The fluorescence EEM results were reported as fluorescence intensity in Raman units (RU) using the peak-picking technique. Three distinct peaks were observed, corresponding to specific types of DOM.

Peak A (ex 250 nm/em 400 nm) was previously identified as fulvic-like DOM [13]. Peak C (ex 350 nm/em 450 nm) corresponds to humic-like DOM [13,24]. Peak T (ex 275 nm/em 340 nm) is associated with tryptophan protein-like DOM [13,24]. These characteristic peaks provide insight into the composition and types of organic matter present in the water samples analyzed before and after chlorination. The peak locations and characteristics of DOM are summarized in Table S2 of the supplementary materials.

Chlorination experiments were performed using two distinct protocols. Standard Formation Potential (SFP) tests were conducted according to Standard Method 5710 [25] to assess the chlorine reactivity of DOM under a higher chlorine dose for 1–7 days. Additionally, Uniform Formation Conditions (UFC) tests, which employ lower chlorine doses and shorter incubation periods, were performed to simulate DBP formation within distribution systems [26]. THMs and HAAs were extracted and analyzed following U.S. EPA methods 551.1 and 552.2, respectively, using a gas chromatography electron capture detector (GC-ECD, Shimadzu Nexis GC-2030, Japan) and Rtx®-5 column with ascorbic acid serving as the chlorine-quenching agent. The analyzed THMs comprised TCM, DBCM, BDCM, and TBM. The analyzed HAAs comprised MCAA, DCAA, TCAA, MBAA, and DBAA. All chlorination experiments, fluorescence EEM measurements, and DBP analyses were duplicated and the average values were reported.

For the characteristics of the raw water, the DOC concentration is 5.06 mg/L, higher than the 4.2 mg/L previously reported for the Chao Phraya River [6]. Seasonal variation may have an effect, as a previous study also reported that DOC concentrations in water samples collected between September and December were higher than those collected in January and February [27]. With a UVA₂₅₄ value of 0.162 cm⁻¹, the DOM in the sample exhibits a SUVA value of 3.20 L/mg-m, which is relatively high compared to other surface water sources in Thailand. For example, the Phong River showed SUVA values between 1.66 and 2.81 L/mg-m [21], whereas the Chi River had an average of 2.37 ± 0.21 L/mg-m [5]. This SUVA of 3.20 L/mg-m aligns closely with values observed in prior studies, indicating high DOM aromaticity [6]. The high SUVA values suggest that coagulation can be an effective treatment for DOM removal. The characteristics of the raw water sampled in December 2023 are summarized in Table S1 in the supplementary materials.

Fig. 1 illustrates the total chlorine demand under a high-chlorine dosage (25 mg-Cl/L) of raw water over a 7-day period. Specific chlorine demand (mg-Cl/mg-C), defined as the total chlorine demand normalized to the DOC concentration of the water sample, characterizes the chlorine reactivity of DOM [28]. On the first day of chlorination, the raw water showed a chlorine demand of 10.4 mg-Cl/L, resulting in a specific chlorine demand of 2.06 mg-Cl/mg-C. Both total and specific chlorine demand increased steadily from days 2 to 4, then increased gradually after day 4. The highest values were observed on day 7, with total and specific chlorine demands reaching 23.6 mg-Cl/L and 4.46 mg-Cl/mg-C, respectively. These results indicate that chlorine demand increases with chlorination time, with DOM oxidation occurring most rapidly between days 2 and 4, consistent with findings in a previous study [5]. Notably, the chlorine demand on day 1 accounted for 44.1% of the total chlorine demand by day 7.

Fig. 2a shows a set of UV absorbance spectra for Chao Phraya River DOM subjected to chlorination as the reaction time increased from 1–7 days, while the corresponding differential UV spectra are shown in Fig. 2b. For this experiment, a chlorine dose of 25.0 mg/L as Cl₂ was added to a water sample containing 5.06 mg/L DOC, and the reaction was stopped by the addition of Na₂SO₃ at chlorination times ranging from 1 to 7 days. A clear trend of decreasing UV absorbance with increasing reaction time can be observed in Fig. 2a, but the conventional spectra lack identifiable peaks. In contrast, the corresponding differential UV spectra have shoulder peaks that can be seen around 272 nm, and the magnitude of the ΔA increased with increasing chlorination time, substantially between 2 and 3 days of chlorination (Fig. 2b). The increase in ΔA₂₇₂ over time indicates progressive oxidation of UV-absorbing aromatic moieties in the DOM. Since chlorine preferentially targets aromatic structures during disinfection, the growth of ΔA₂₇₂ suggests the breakdown or transformation of these moieties through oxidative or electrophilic reactions [29].

The preferential reaction between chlorine and UV-absorbing functional groups in DOM makes ΔA₂₇₂ particularly effective as a sensitive and specific probe for monitoring chlorination reactions [11]. As chlorine preferentially reacts with the aromatic fraction of DOM [30], the observed maximum in the differential UV spectrum around 264–270 nm likely characterizes aromatic structures that underwent oxidation, addition, or electrophilic substitution reactions with chlorine [31]. These structural changes are known precursors to the formation of carbonaceous DBPs, such as THMs and HAAs [32,33] and could also be related to fluorescence loss.

The qualitative results of the fluorescence EEM spectra for the water sample before and after chlorination are presented in Fig. 3. The fluorescence intensities at peaks A, C, and T correspond to fulvic-like, humic-like, and protein-like substances in the DOM pool, respectively [24]. The raw water exhibited a predominance of fulvic-like DOM, with peak A showing an intensity of 1.16 RU, while peaks C and T had intensities of 059 RU and 0.06 RU, respectively (Fig. 3a). This indicates that the DOM in the raw water is primarily composed of humic substances. Following chlorination, the fluorescence intensities at these three peaks showed a partial decrease after 1 day and a significant reduction after 7 days of chlorination (Fig. 3b and 3c).

After 1 day of chlorination, the fluorescence intensities observed at peaks A, C, and T decreased by 39.5%, 49.5%, and 32.4%, respectively (Fig. 4). The decline in fluorescence intensities continued gradually as chlorination progressed. By day 7, the fluorescence losses at peaks A, C, and T (ΔEEM_A, ΔEEM_C, and ΔEEM_T) had reached 59.1%, 68.2%, and 48.0.0%, respectively. This loss of fluorescence, calculated as the difference between initial fluorescence intensity at a specific peak before and after chlorination, is indicative of changes in DOM characteristics, suggesting that the aromatic fractions of DOM were oxidized [5,17]. The greater ΔEEM_C can also be related to the higher chlorine reactivity of humic-like DOM compared to the fulvic-like DOM (peak A) and protein-like DOM (peak T) [5]. This oxidation occurred rapidly during the first 1 to 4 days of chlorination, stabilizing after day 4, which reflects the behavior of chlorine demand illustrated in Fig. 1 [34].

Fig. 5a and 5b illustrate the formation of four THMs and five HAAs over 1–7 days of chlorination. The concentration of THMs increased gradually throughout the 7-day chlorination period, with 55.6% of the total THMs formed after the first day. After 7 days, the highest concentration of THMs reached 1,041.1 μg/L. TCM was the predominant THM species, with concentrations ranging from 467.6 to 927.39 μg/L, comprising 80.8–89.1% of the total THMs (Fig. 5a). TCM was previously reported as the predominant THM species for a water source with a low bromide concentration [21], while BDCM and DBCM were the predominant THM species for a water source with a high bromide concentration [5]. The specific THM-FP was defined as the ratio of the total THM-FP to the DOC of the water sample. The specific THM-FP after 7 days of chlorination was 205.75 μg/mg-C.

The concentrations of the five HAAs formed during the chlorination of the raw water were lower than those of the four THMs (Fig. 5b). The higher formation potential of THMs compared to HAAs following the addition of chlorine to the water samples is consistent with the typical characteristics of organic matter found in surface water, which tends to react with chlorine to yield a greater quantity of THMs than HAAs. This observation is in agreement with the previous findings [35], which reported that, relative to the total organic halide content (TOX), THMs accounted for 20% and HAAs for only 10% of the DBPs formed in chlorinated drinking water. Among the HAAs, MCAA and MBAA were formed in the highest concentrations, comprising 25.4–28.8% and 69.6–72.7%, respectively. Notably, DCAA and DBAA were not detected in this study. The specific HAA-FP after 7 days of chlorination of the raw water was 113.95 μg/mg-C, much greater than 58.34 μg/mg-C previously reported in 2006 [7]. The concentrations of THMs and HAAs correlated well with the observed decrease in fluorescence intensity following chlorination (Fig. 4). Specifically, the reduction in fluorescence intensity, attributed to the oxidation of aromatic organic matter, resulted in the formation of by-products in the form of THMs and HAAs.

To better understand the formation potential (FP) of individual DBP species, it is important to track changes in the optical properties of DOM resulting from chlorination. Therefore, linear correlations were sought between the formation potential of the four THMs (THM-FP) and five HAAs (HAA-FP) over 1–7 days of high-dose chlorination (25 mg-Cl/L) and various chlorine reactivity surrogates, including ΔA₂₇₂, ΔEEM_A, ΔEEM_C, and ΔEEM_T (Fig. 6a–6c). All correlations in Fig. 6 are statistically significant (p-value < 0.05). ΔA₂₇₂ showed strong linear correlations with the formation of total THMs (R² = 0.80, p-value < 0.05) and total HAAs (R² = 0.60), including contributions from both chlorinated and brominated species, similar to previous studies [11,12]. According to the relationship depicted in Fig. 6a, the formation of DBPs could be quantitatively evaluated using the following Eq. (1)[36]:

where m is the slope between either THM-FP or HAA-FP and the decrease of A₂₇₂ due to the chlorination of DOM. The variable C₀ represents the y-coordinate where the line intersects the y-axis. As shown in Fig. 6a, the slope (m) for the relationship between ΔA₂₇₂ and THM-FP is greater than the slope for HAA-FP. This suggests that more THMs were formed per unit of absorbance decrease, compared to HAAs, during the chlorination of DOM. The presence of bromide in the raw water (0.24 mg/L) promoted the formation of THMs over HAAs. Previous studies have reported that increasing bromide concentration causes an increase in THM-FP but a decrease in HAA-FP [37].

For fluorescence surrogate parameters, the strongest correlation was found between THM-FP and ΔEEM_C (R² = 0.96), indicating that humic-like DOM peak C contributes significantly to THM precursors (Fig. 6b). The ΔEEM_A and Δ EEM_T also significantly correlated with THM-FP but to a smaller extent (R² = 0.82 and 0.88, respectively). Many studies reported that THM-FP correlated strongly with fluorescence intensities of the humic-like component in the raw water before chlorination [18,38,39]. The decrease in fluorescence intensity of the humic-like substances peak C strongly correlated with THM-FP for raw and coagulated water [5]. As ΔEEM_C increases over time, it implies that prolonged chlorination enhances the oxidative degradation of humic-like DOM. Thus, the oxidation reactions not only oxidize aromatic DOM but also generate reactive intermediates, which subsequently lead to increased THM formation through chlorine substitution and addition reactions.

The ΔA₂₇₂ and ΔEEM_C reflect distinct but complementary aspects of DOM transformation during chlorination. The ΔA₂₇₂ primarily captures changes in aromatic and conjugated structures that absorb UV light [29], while ΔEEM_C specifically tracks the degradation of humic-like substances [5]. These optical changes occur simultaneously with the formation of DBPs as chlorine reacts with DOM functional groups. The stronger correlation between ΔEEM_C and THM-FP (R² = 0.96) compared to ΔA₂₇₂ and THM-FP (R² = 0.80) suggests that humic-like components represented by peak C are particularly effective precursors for THM formation in Chao Phraya River DOM. This difference indicates that while aromatic content generally contributes to DBP formation, specific humic-like fluorophores may contain molecular structures that are preferentially reactive toward THM formation pathways.

HAA-FP, on the other hand, was strongly correlated to ΔEEM_A and ΔEEM_T (R² = 0.79 and 0.72, respectively), while the correlation with ΔEEM_C was notably weaker (R² = 0.58) (Fig. 6c). These results suggest that the oxidation of fulvic-like DOM, represented by peak A, is a more significant contributor to HAA-FP under the tested conditions. This contrast underscores the distinct reactivity of different DOM fractions with chlorine and their role in producing specific groups of DBPs. The relationships between fluorescence loss at specific peaks and DBP formation suggest that the optical properties of DOM can be related to structural properties and are important indicators of its susceptibility to chlorine attack. The fluorescence loss at peaks A and C is a more reliable predictor of carbonaceous DBP formation compared to ΔA₂₇₂, with distinct trends evident for THMs and HAAs. A different trend was reported in a previous study that ΔA₂₇₂ was the most accurate surrogate parameter for THM-FP in a range of raw and coagulated waters [39]. This difference could be attributed to variations in water characteristics, DOM composition, and chlorination methods between the studies. In the current study, fluorescence loss at specific EEM peaks more directly reflects the oxidative transformation of distinct DOM fractions as a function of time, which may be more representative of the chlorine-reactive sites responsible for DBP formation in this specific water matrix. The oxidation of both humic-like and fulvic-like DOM, as measured by ΔEEM_C and ΔEEM_A, appears to be a stronger indicator of DBP formation compared to ΔA₂₇₂ in the present study.

The raw water had a DOC concentration of 5.06 mg/L, with the DOM-HPI and DOM-HPO fractions of DOM contributing 2.38 mg/L (49.4%) and 2.44 mg/L (50.6%), respectively. The DOC recovery using the fraction technique was at 95.3%. Fig. S1 in the supplementary materials presents the fluorescence EEM spectra for the raw water, DOM-HPI, and DOM-HPO fractions. A notable difference in fluorescence spectra was observed between the DOM-HPI and DOM-HPO fractions. Specifically, the fluorescent intensities at peak T (protein-like DOM) were more pronounced in the DOM-HPI fraction compared to those in the DOM-HPO fraction. Additionally, the fulvic-like DOM (peak C) was predominantly present in the DOM-HPO sample [40].

The quantitative comparison of fluorescent intensities at peaks A, C, and T for raw water, DOM-HPI, and DOM-HPO fractions, both before and after chlorination, is illustrated in Fig. 7a and 7b. In the raw water sample, the DOM-HPI fraction exhibited fluorescent intensities of 0.51, 0.23, and 0.33 RU at peaks A, C, and T, respectively, whereas the DOM-HPO fraction showed intensities of 0.58, 0.30, and 0.29 RU at the same peaks (Fig. 7a). This difference in fluorescence characteristics between the HPI and HPO DOM fractions is consistent with previous studies on the compositional differences between these two DOM fractions [16]. The percentage of humic substances (%humic) was determined by calculating the ratio of the sum of intensities at peaks A and C to the sum of intensities across all three peaks [13]. The %humic was 74.6% for the bulk raw water, 67.8% for the DOM-HPI fraction, and 77.0% for the DOM-HPO fraction. The higher %humic for the DOM-HPO fraction indicates that this fraction is enriched in humic-like organic matter compared to the DOM-HPI fraction.

A 1-day chlorination following the UFC procedure [26], in which a much lower chlorine dose was applied to represent chlorine residual in the distribution system, resulted in a certain degree of decline in fluorescent intensities at all three peaks. This finding is consistent with previous studies showing that chlorination can induce structural changes in DOM, leading to alterations in its fluorescence properties [5,17]. For the bulk raw water, the greatest fluorescence loss occurred at peak C (humic-like DOM, 13.1%). Humic-like substances are known to be more susceptible to chlorine-induced alterations than other DOM fractions [32]. For the DOM-HPI fraction, the greatest fluorescence loss also occurred at peak C (10.5%), followed by peak A (9.3%) and peak T (2.8%). The DOM-HPI fraction contains more hydrophilic and lower molecular weight compounds, which can react to chlorine oxidation [16].

A greater fluorescence loss was observed for the DOM-HPO fraction, 17.7% loss at peak C, 6.2% at peak T, and 7.0% at peak A. The DOM-HPO fraction is comprised of more hydrophobic and higher molecular weight compounds, which are generally more susceptible to reacting with chlorine [32]. These findings suggest that the chlorination process can differentially impact the fluorescence properties of various DOM fractions, with the more humic-like and hydrophobic components experiencing greater alterations. Understanding these changes in DOM fluorescence can provide insights into structural transformations and reactivity of DOM during chlorination processes.

After 1 day of chlorination, the raw water, DOM-HPI, and DOM-HPO fractions led to THM formation of 436.5, 203.7, and 253.5 μg/L, respectively (Fig. 8a). These concentrations are consistent with a previous study investigating DBP formation from different DOM fractions and reporting that the HPO fractions always yield higher HAA and THM formation potentials than their corresponding hydrophilic fractions [41]. The specific THM-FP values were 86.3, 85.6, and 103.9 μg/mg-C for the raw water, DOM-HPI, and DOM-HPO, respectively (Fig. 8b), indicating varying reactivity of different DOM components with chlorine [32]. The higher specific THM-FP values for DOM fractions compared to raw water may be attributed to differences in the DOC: Br ratio. Notably, brominated THMs (BDCM and DBCM) appeared in higher concentrations in the DOM-HPI fraction, while the chlorinated THMs (TCM) formed at a higher concentration in the DOM-HPO fraction (216.0 μg/L) than in the DOM-HPI fraction (151.0 μg/L). This observation can be explained by the lack of bromide retention in ENV bond elute® cartridges, resulting in a higher bromide concentration in the DOM-HPI fraction. These findings emphasize the complex interactions between DOM fractions and chlorine, influenced by DOM concentration, composition, and the presence of bromide.

HAA formation in the raw water, DOM-HPI, and DOM-HPO were measured at 227.5, 80.6, and 76.5 μg/L, corresponding to specific HAA-FP of 45.0, 33.9, and 31.4 μg/mg-C, respectively (Fig. 8a and 8b). These findings align with a previous study demonstrating that the hydrophobic acid fraction within the DOM pool exhibits the greatest potential for forming both THMs and HAAs. Comparing the two fractions, DOM-HPO exhibited a higher specific THM-FP, while DOM-HPI had a greater specific HAA-FP [42]. This differential behavior can be attributed to the distinct molecular characteristics and reactive sites present in each DOM fraction [32]. The DOM-HPO fraction typically contains more aromatic structures and higher molecular weight compounds, which are known precursors for THM formation [43]. In contrast, the hydrophilic fraction may contain specific functional groups that preferentially form HAAs during chlorination [7]. Overall, the THM-FP and HAA-FP can be related to fluorescence loss, suggesting a potential correlation between the structural characteristics of various DOM fractions and their DBP formation potential.