Characterizing the effects of stormwater runoff on dissolved organic matter in an urban river (Jiujiang, Jiangxi province, China) using spectral analysis

The effect of stormwater runoff on dissolved organic matter (DOM) in rivers is one of the central topics in water environment research. Jiujiang is one of the first cities established in the green development demonstration zone of the Yangtze River Economic Belt (Jiangxi Province, China). Three-dimensional excitation–emission matrix fluorescence with parallel factor analysis (3DEEM-PARAFAC) and ultraviolet–visible (UV–Vis) spectroscopy were used to explore the effects of runoff on organic matter in Shili River (Jiujiang, Jiangxi Province, China). The results show that the runoff led to an increase of some critical pollutants and DOM concentrations, especially in the middle reaches of the river. The concentration and relative molecular weight of DOM in water increased as a result of runoff. Three humic-like (C1–C3) and two protein-like (C4 and C5) components of DOM were identified using the PARAFAC model. The sources of the three humic-like components (C1, C2, C3) were consistent, unlike those of the protein-like component C4. Compared with the pre-rainfall period, the content of humus compounds flowing into the river through the early rainwater runoff was lower, which caused the relative content and proportion of humic substances little change and protein-like species increasing. The DOM mainly derived from autochthonous sources, and runoff had limited effect on its characteristics. Jiujiang is a key demonstration city for Yangtze River conservation. Rainwater runoff is one of the pollution sources of urban rivers, which leads to the deterioration of water quality and influences the distribution characteristics of DOM in water bodies. The PARAFAC components could adequately represent different indicators and sources of DOM in urban rivers, providing an important reference for urban river management.


Introduction
The effect of stormwater runoff on dissolved organic matter (DOM) in rivers is one of the central topics in water environment research. DOM plays a vital role in the water environment and consists of complex organic compounds, mainly including humic-, fulvic-, and amino acid-like substances, as well as carbohydrates (Lipczynska-Kochany 2018; Hawkes et al. 2019). DOM also contains functional groups such as amino, carboxyl, aldehyde, and hydroxyl (Wang and Chen 2018), which can reflect the composition of dissolved organic compounds in water and play a vital role in aquatic ecosystems (Wu et al. 2008;Chen et al. 2019b). Furthermore, DOM is considered a carrier of toxic substances, such as heavy metals and persistent organic pollutants (Macoustra et al. 2020;Katsoyiannis and Samara 2007). DOM sources are mainly divided into two categories (Zhang et al. 2021b): allochthonous inputs (mainly from terrestrial plant waste, rainwater runoff, and atmospheric rainwater (Derrien et al. 2019)) and autochthonous fractions, mostly originating from degradation and release of aquatic organisms, such as Responsible Editor: Philippe Garrigues microorganisms, algae, and plankton (Artifon et al. 2019). As a result of rapid urbanization, the allochthonous inputs have increasingly important effects on DOM variations in rivers, among which the impact of rainwater runoff on DOM is particularly strong .
Surface (especially initial rainwater) runoff contains multiple organic and inorganic pollutants. Common runoff pollutants include nutrient elements [e.g., suspended solids (SS), chemical oxygen demand (COD), nitrogen (N), and phosphorus (P)], heavy metals (such as Pb and Cu), and organic compounds (such as polycyclic aromatic hydrocarbons) (Sansalone and Cristina 2004;Zhu 2019). These pollutants can cause serious pollution of river water, resulting in the degradation of the water ecological environment. During rainfall events, the DOM in soil can be discharged into rivers or lakes through surface and underground runoff (Ward et al. 2017). The effects of rainwater runoff on DOM have attracted great attention from the scientific community. For example, Singh et al. (2014) explored seasonal variation of DOM in watershed sources, showing that the DOM in stormwater runoff contained relatively high DOC concentrations in summer, with a stronger humic character compared to spring and winter. In addition, Zhu et al. (2019) studied DOM variations in summer rainwater and indicated that humic substances absorbing in both the ultraviolet (UV) and visible (Vis) regions were the predominant component of DOM in rainwater; the DOM composition was influenced by both terrestrial inputs and autochthonous microorganisms. Moreover, Xenopoulos et al. (2021) investigated how human activities and climate changes alter the DOM composition in freshwater; they found that increased precipitation may shift the DOM composition toward allochthonous forms relatively rich in humus. Furthermore, Chen et al. (2019a) studied the effects of hurricanes on the sources and composition of DOM in ten watersheds across five southeastern US states by quantifying the dissolved organic carbon (DOC) output over a rainstorm period. They showed that the proportion of terrestrial humic-like DOM compounds increased over the high runoff period, with a corresponding decrease in the fraction of autochthonous protein-like DOM species. Therefore, it is essential to identify the sources and characteristics of DOM in rivers, as well as evaluating the effects of runoff on these properties.
Three-dimensional excitation-emission matrix (3DEEM) fluorescence spectroscopy is widely applied in the analysis of DOM characteristics and sources in water environments; this technique is characterized by high sensitivity and reliability, with no damage to samples (Sankoda et al. 2018). In addition, the combination of 3DEEM fluorescence spectroscopy with parallel factor analysis (PARAFAC) has been extensively used to study the DOM composition and sources of fluorescent components in marine (Yao et al. 2016), natural water (Zhang et al. 2021a), wetland (Jin et al. 2021), and reservoir (Li et al. 2021) environments. Using 3DEEM-PARAFAC, Niloy et al. (2021) investigated the seasonal variation and fluorescence characteristics of DOM in the Ganges River, India. Feng et al. (2016) used 3DEEM-PARAFAC to identify eight fluorescent components of DOM in the water of a eutrophic lake (Taihu) and then analyzed their distribution characteristics and possible sources. Furthermore, Lin et al. (2020) analyzed the composition and sources of chromophoric DOM (CDOM) in the Wangchuan River (Shaanxi Province, China) using UV-Vis absorption spectroscopy and 3DEEM-PARAFAC. Based on the same techniques, Ma and Li (2020) analyzed the spatiotemporal variation of DOM characteristics in river systems of the Three Gorges Reservoir (China), while He et al. (2016) analyzed the spectral characteristics and sources of DOM in landscape rivers during the flood season. In addition to its application in natural water environments, 3DEEM-PARAFAC is widely used in the field of wastewater treatment. For example, Maqbool et al. (2016) successfully tracked the fluorescent components of DOM in a membrane bioreactor using 3DEEM-PARAFAC. These studies demonstrate the reliability of this method for the characterization of DOM in water environments.
Jiujiang is among the first cities established in the green development demonstration zone of the Yangtze River Economic Belt (Jiangxi Province, China). Rainwater runoff is one of the pollution sources of urban rivers in this area, leading to the deterioration of water quality and influencing the distribution patterns of DOM in water. However, comparative studies of DOM variations in rivers inflow lake in Jiujiang under the action of runoff are scarce. Therefore, the present study aimed to determine the fluorescence spectral characteristics of DOM in the Shili River (Jiujiang) before and after rainwater runoff, based on the 3DEEM-PARAFAC technique. UV-Vis absorption spectra were used to investigate the distribution patterns of DOM and the associated influencing factors in river water before and after rainwater runoff. The results of this study could provide useful fundamental data for the management and protection of urban rivers and represent an important reference for the control of water pollution caused by stormwater runoff in the Yangtze River basin.

Study area
The Shili River is located in Jiujiang City, northern Jiangxi Province of southeastern China (Fig. 1). Being one of the main tributaries of Bali Lake, the Shili River originates from the northern slope of Lushan Mountain and is formed by the intersection of the Lianxi and Lianhua water systems. This river runs from south to north through the Lushan District and the Jiujiang Economic-Technological Development Zone, converges with the Lianxi River at Jiujiang Vocational and Technical College, then diverts west and flows into Bali Lake. The total length of the Shili River from the watershed to the estuary is 12.9 km, and the weighted average slope of the river channel is 29.34‰. A total of 14 sampling points were selected along the river course, covering the upper (#1), middle (#2-#8), and lower (#9-#14) reaches.

Sampling and parameters measuring
In situ monitoring and collection of water samples were conducted in the Shili River at two stages: before rainfall (April 23) and after rainfall (April 24, rainfall = 30.6 mm). A total of 28 samples were collected in 360 mL clean polyethylene bottles and stored in a 4 °C refrigerator in the dark. Selected indicators were determined within 48 h of collection.
Among the water quality indicators, COD was determined using the potassium dichromate method. Ammonia nitrogen (NH 3 -N), total nitrogen (TN), and total phosphorus (TP) were measured using the methods given in the Water and Waste Water Monitoring and Analytical Methods (State Environmental Protection Administration of China 2002). DOC was determined using a total organic carbon analyzer (TOC-L CPN; Shimadzu Corp., Kyoto, Japan).

DOM absorption analysis
UV-Vis absorption spectra of water samples were measured using a spectrophotometer (UV-1780 PC; Shimadzu Corp., Kyoto, Japan). Absorbances were measured over the range of 200-800 nm using a 1 cm quartz cuvette at a scanning interval of 1 nm, with Milli-Q ultrapure water as blank. Absorption coefficients were calculated using Eqs. (1) and (2) (Li et al. 2021): where λ is the wavelength (nm), a * (λ) and a(λ) are the absorption coefficients at λ without and with scattering correction (m −1 ), D(λ) is the absorbance at λ, and r is the optical path (m). Herein, the absorption coefficient at 355 nm, a (355), was selected to indicate the relative concentration of CDOM, with higher values reflecting higher CDOM concentrations (Yu et al. 2016). The ratio of the UV absorbances at 250 and 365 nm (E2/E3) was used to represent the relative molecular weight of CDOM: a higher ratio denoted a lower relative molecular weight (Li et al. 2014;Wang et al. 2019). The spectral slope ratio, S R , was used to identify the DOM sources; S R < 1 indicated primarily allochthonous DOM sources, i.e., organic matter inputs with high molecular weight and strong aromaticity; S R > 1 indicated predominantly autochthonous DOM inputs  (1) a * ( ) = 2.303D( )∕r (2) a( ) = a * ( ) − a * (700) • ∕700  . The ratio between absorption coefficient at 254 nm and DOC concentration (SUVA 254 ) was employed as indicator of humified DOM aromaticity, expressed in L/ (mg·m) units. The content of aromatic substances in DOM is proportional to the DOM aromaticity: the higher the SUVA 254 value, the greater the degree of organic matter humification ).

DOM fluorescence measurement
The 3DEEM fluorescence spectral characteristics of DOM samples were measured using a spectrofluorometer (RF-6000; Shimadzu). Before testing, samples with DOC concentrations > 8 mg/L were diluted with Milli-Q water to a DOC concentration of 8 mg/L to eliminate internal filtering effects. Milli-Q ultrapure water was used as blank to record fluorescence spectra at excitation wavelengths (Ex) of 200-400 nm in 5 nm intervals and emission wavelengths (Em) of 250-450 nm in 2 nm intervals, with a scanning rate of 2000 nm/min.
The fluorescence index (FI) (Fan et al. 2018) is the ratio of the integral fluorescence intensities at Em = 470 and 520 nm, with Ex = 370 nm; this index is often used to evaluate the degradation degree and sources of DOM. The humification index (HIX) (Han et al. 2021) is the ratio of the integral fluorescence intensities at Em = 435-480 and 300-345 nm, with Ex = 254 nm; this index measures the degree of humification, with higher values indicating a higher humification. The biological index (BIX) (Zhou et al. 2020) is the ratio of the fluorescence intensities at Em = 380 and 430 nm, with Ex = 310 nm; this index measures the relative contribution from autochthonous sources of DOM and the level of DOM bioavailability.

Statistical analyses
Origin 8.0 and Excel 2016 were used to analyze and process the data. Maps were drawn using the software ArcGIS version 10.2.
In order to meet the requirements of parallel factor analysis, three techniques were used for parallel treatment of samples and distilled water calibration samples. PARAFAC was based on a mathematical model that applied alternating least squares to 3D data analysis. This technique reduces the dimensions of multiple EEMs. In this paper, MATLAB 2018 software and DOMFluor toolkit were used to perform PARAFAC analysis on the fluorescence data of water samples to calculate the fluorescence intensity of each component and the total fluorescence intensity (Kowalczuk et al. 2010;Stedmon and Bro 2008).

General characteristics
The measured values of the four water quality indicators (COD, NH 3 -N, TN, and TP) in the river reaches are shown in Fig. 2. The indicator values obtained after rainfall exhibited marked changes compared to those measured before rainfall. In particular, the COD concentrations before and after runoff varied in the ranges of 4.52-30.10 mg/L (mean: 17.13 mg/L) and 5.01-67.73 mg/L (mean: 31.75 mg/L), respectively. The COD concentrations in river water increased after stormwater runoff, and the variation with respect to the concentration before runoff was especially pronounced in the middle river reaches. Before runoff, the mean COD concentration in the middle river reaches did not meet the standard limit of class III water quality (15.91 mg/L), defined according to China's Environmental Quality Standard for Surface Water (GB3838-2002); however, it exceeded the standard limit of class V water quality (45.15 mg/L) after runoff. In the upper and lower river reaches, the COD concentrations remained basically unchanged before and after runoff.
The NH 3 -N concentrations varied in the ranges of 0.22-2.95 mg/L (mean: 1.46 mg/L) before runoff and 0.16-2.60 mg/L (1.29 mg/L) after runoff. In the upper and lower river reaches, the mean NH 3 -N concentrations before runoff were higher than those after runoff; however, the inverse trend was observed in the middle river reaches, which met the criteria for class IV water quality. The TN concentrations before and after runoff fell in the ranges of 2.90-7.60 mg/L (mean: 4.61 mg/L) and 1.60-7.30 mg/L (mean: 4.66 mg/L), respectively. The mean TN concentrations after runoff were slightly higher than those before runoff, and all samples met the class V water quality standards. In the upper river reaches, the mean TN concentration was higher before runoff (4.80 mg/L) than after runoff (1.70 mg/L), while the opposite trend was observed in the middle and lower reaches; all values exceeded the class V water quality standard (except for the upper reaches after runoff). The range of TP concentrations before and after rainfall were 0.03-0.22 mg/L (mean: 0.12 mg/L) and 0.03-0.28 mg/L (mean: 0.14 mg/L), respectively. In the middle river reaches, the mean TP concentration after runoff was significantly higher than that before runoff.
In a previous study, Gu and Zhang (2013) analyzed the pollutant composition of surface runoff in the main urban area of Kunming (Yunnan Province, China) and found relatively high concentrations of pollutants in stormwater runoff during the initial precipitation stage, followed by rapid concentration decreases in the middle and later stages. In the present study, the upper reaches of the Shili River were surrounded by numerous trees and thus received less rainwater runoff. At the same time, there was no considerable variation in the pollutant concentrations in the upper river reaches before and after runoff, owing to the relatively high flow rate of river water. The middle reaches of the Shili River constituted the urban river channel. When this area was scoured by rainwater in the initial stage, surface source pollutants (TN, TP, organic matter, etc.) carried by rainwater were predominant. In the middle reaches of urbanization area, the vegetation coverage rate was less and the rainwater runoff flowed into the river channel more. Consequently, the concentrations of all pollutants in the middle river reaches showed a relative increase after the end of runoff. In the lower reaches of the Shili River, the pollutant concentrations after runoff did not show a marked decrease compared to those before runoff. Compared to the upper river reaches, the lower reaches were surrounded by lush vegetation, which reduced pollution from stormwater runoff. Moreover, both surface runoff pollutants and rainfall intensity decreased over time, leading to a reduction in the surface runoff discharged into river water. Li et al. (2017) analyzed the pollutant concentrations of runoff on road surfaces in Nanning (Guangxi Zhuang Autonomous Region, China). They showed that higher rainfall intensity resulted in a more evident scouring of pollutants in rainwater runoff; the scouring of pollutants by different intensities of rainfall exhibited a distinctive initial effect. Overall, the runoff caused heavy pollution in the middle reaches and light pollution in the upper and lower reaches of the Shili River.

DOM absorption characteristics
The DOC and a(355) values were used to indicate DOM and CDOM concentrations, respectively (Fig. 3). The DOM concentrations before and after runoff varied in the ranges of 1.08-2.67 mg/L (mean: 1.50 mg/L) and 1.01-3.03 mg/L (mean: 1.70 mg/L), respectively; the corresponding CDOM concentration ranges were 1.38-2.99 mg/L (mean: 2.21 mg/L) and 1.15-3.91 mg/L (mean: 2.61 mg/L), respectively. The mean DOM and CDOM concentrations after runoff were both higher than those before runoff. This result indicates that rainwater runoff contained substantial amounts of organic matter, and the DOM of Shili River was mainly influenced by terrestrial DOM inputs. This finding is consistent with the results reported for aquatic ecosystems in rivers and lakes, for which the DOM concentrations in water were observed to increase as a result of increased rainfall and runoff (Zhou et al. 2015;Niloy et al. 2021).
The SUVA 254 values before runoff ranged from 3.64 to 9.44 L/(mg·m) [mean: 7.06 L/(mg·m)], and those after runoff varied in the 5.64-12.17 L/(mg·m) range, with a mean of 7.46 L/(mg·m) (Fig. 4a). The higher SUVA 254 values after runoff indicated that aromatic substances entered river water during the stormwater runoff process. In terms of spatial distribution, the mean E2/E3 ratios before runoff changed in Fig. 2 Basic water quality indicators before and after stormwater runoff the following order: upper reaches (5.80) > middle reaches (5.77) > lower reaches (5.61), whereas the ratios measured after runoff varied as follows: lower reaches (5.80) > middle reaches (5.62) > upper reaches (5.40) (Fig. 4b). The E2/E3 ratio ranges were 5.10-6.40 (mean: 5.70) before runoff and 5.07-6.23 (mean: 5.68) after runoff. The E2/E3 ratios before runoff were slightly higher than those after runoff, and little variation in the molecular weight of DOM was observed before and after runoff. In addition to the spectra, the S R of DOM was used to identify some characteristic DOM components (Fig. 4c). The S R values of river water before runoff ranged from 0.85 to 1.25 (mean: 1.05), while those after runoff varied from 0.89 to 1.31, with a mean of 1.09. The minor difference in the S R values before and after runoff suggested little allochthonous pollution from organic pollution sources. Moreover, all S R values were slightly greater than 1, indicating that the DOM composition was dominated by biological inputs. It is mainly DOM formed by metabolic emissions of aquatic organisms and the transformation of biological remains (Fig. 4).
The excitation peak of component C1 (265/482 nm) corresponded to the standard peak A, indicating the presence of   PARAFAC model was placed on OpenFluor, and existing PARAFAC components in different ecosystems were referenced and compared. The results showed that the excitation and emission min similarity score was 0.95.
Statistical analysis of the data showed that the proportions of the five DOM components before and after runoff were relatively balanced (Fig. 6). Before runoff, C2 and C1 accounted for the largest and smallest proportion of the DOM components (23.69% and 15.86%, respectively), whereas the proportions of C3, C4, and C5 were 21.00%, 21.11%, and 18.34%, respectively. After runoff, C4 represented the highest fraction among DOM components (26.80%, also exceeding the proportion of protein-like C4 before runoff), followed by C2 (23.47%), while the proportion of C1 was still the lowest (14.33%). The C3 proportion after runoff was 3.82% lower than that before, whereas the C5 proportion showed no quantitative difference before and after runoff.
The total fluorescence intensity of DOM components represented the DOM content in water body. The total fluorescence intensity of DOM components before and after runoff in Shili River water body varied from 4952.19 to 15,867.99 a.u. and 4403.94 to 20,389.40 a.u., respectively. In general, the total fluorescence intensity before runoff in the upstream and downstream was higher than that after runoff, while the total fluorescence intensity before runoff in the middle reaches was lower than that after runoff, which was basically consistent with the changes of DOC in the upstream and middle reaches, and COD in the middle reaches.

DOM Source
To further investigate the DOM properties and sources, we used three widely applied fluorescence indices, which provide a better description of the DOM fluorescence characteristics of river water before and after runoff. Both FI and BIX showed slight variations before and after runoff, while marked changes were observed for the HIX values (Fig. 7). The FI thresholds for terrestrial and autochthonous sources are 1.4 and 1.9, respectively;  . 6 The EEM spectrum percentage of DOM before and after stormwater runoff Fig. 7 The total EEM fluorescence intensity of DOM before and after stormwater runoff namely, an FI closer to 1.4 indicates that the DOM in water mainly derives from terrestrial sources, whereas a FI closer to 1.9 denotes autochthonous sources (Chen and Li 2019c). In the present study, the FI values before runoff varied between 1.82 and 2.14 (mean: 1.96), whereas those after runoff ranged from 1.83 to 2.05, with a mean of 1.97. The small relative variation in the FI values after runoff suggested that microbial activity played a major role in DOM input before and after runoff. The contribution of terrestrial sources of DOM derived from runoff was limited, and autochthonous DOM was dominant. Compared with before runoff, the water bioavailability of Shili River increased after runoff, resulting in an increase in the proportion of newly born DOM. After runoff, the E2/E3 values decreased, the molecular weight of CDOM increased, and the endogenous level increased, which was consistent with this result. BIX values of 0.6-0.7 and 0.8-1.0 denote DOM with weak and strong autochthonous characteristics, respectively (Huguet et al. 2009;Ziegelgruber et al. 2013). In our study, the BIX values before runoff varied in the range of 0.85-1.06 (mean: 0.93), whereas those after runoff ranged from 0.83 to 1.09 (mean: 0.97), thus showing a relatively small increase. These results imply that aquatic organisms or bacteria were the primary factors driving variations in humic substances in river water after runoff. The autochthonous contribution to DOM was slightly larger after runoff.
A HIX value below 4 indicates DOM with a weak humification degree (Huguet et al. 2009) and terrestrial DOM inputs is small. Before runoff, the HIX values were between 2.40 and 3.07 (mean: 2.91), indicating a low DOM humification degree in river water. After runoff, the HIX values ranged between 2.08 and 3.34, with a mean of 2.54. Although relatively large variations were observed in the middle river reaches, all HIX values before and after runoff were below 4. These results reveal that little humic-like DOM flowed into the river through runoff, while DOM with low humic content derived from aquatic microorganisms was predominant. Runoff had a weaker scouring effect on terrestrially derived organic matter, leading to a predominance of microbially produced DOM components. Overall, the fluorescence indices FI and BIX in the Shili River water showed little variation before and after runoff, whereas HIX exhibited marked changes in the middle reaches of the river.

Relationship between DOM composition and environmental factors
Correlation between DOM fluorescence components (C1-C5), fluorescence parameters, and environmental factors in water samples from the Shili River water before and after runoff was analyzed, and Pearson's correlation coefficients are shown in Fig. 8. The water quality parameters COD, NH 3 -N, TN, and TP showed a significant positive correlation with the PARAFAC components C1-C5, suggesting that the latter could well reflect the water quality indicators. Moreover, weak correlations were found between water quality and DOM fluorescence indices. The correlation coefficients between C1 and C2, C2 and C3, as well as C1 and C3 were 0.99, 1.00, and 0.97, respectively, while that between C1 and C4 was 0.68. The humic-like component C1 showed very strong positive correlations with C2 and C3, but its correlation with the protein-like component C4 was slightly weaker. This result suggests that the sources of humic-like components were consistent in time and space, at variance with those of the protein-like component C4. Moreover, the BIX parameter was negatively correlated with the humic-like (C1-C3) and protein-like (C4-C5) components; the corresponding correlation coefficients were − 0.34, − 0.29, − 0.26, − 0.64, and − 0.55, respectively, indicating that the autochthonous contribution to DOM in the Shili River was probably low. No strong correlation was observed between BIX and HIX.
The results of this study show significant differences in DOM components and environmental factors, which suggests the presence of a specific relationship between DOM composition and water quality in different aquatic ecosystems ( Fig. 9) (Cui et al. 2016;Jin et al. 2022). In addition, components C1-C3 had the same source, which was significantly different from the source of C4. Therefore, improving the water quality of Shili River can improve the water quality of Shili River. It was suggested to increase the capital investment, ensure sewage treatment, strengthen the utilization of recycled water, strengthen the publicity, and improve the awareness of environmental protection.

Conclusions
(1) Runoff led to an increase in COD, NH 3 -N, TN, and TP concentrations in water from the Shili River; these effects were relatively more significant in the middle reaches of the river.
(2) Optical analysis indicated that the runoff resulted in an increased concentration and relative molecular weight of DOM in river water, with little variation in the degree of DOM humification. The DOM mainly derived from autochthonous sources, and runoff had limited effect on its characteristics. (3) Three humic-like (C1-C3) and two protein-like (C4 and C5) components of DOM were identified using the PARAFAC model. Correlation analysis revealed that the sources of the three humic-like components were consistent, unlike those of the protein-like component C4. The water quality parameters showed significant positive correlations with components C1-C5; thus, the PARAFAC components of DOM can well represent different indicators of water quality in urban rivers.