A Quantitative Approach for Identifying Nitrogen Sources in Complex Yeongsan River Watershed, Republic of Korea, Based on Dual Nitrogen Isotope Ratios and Hydrological Model

Abstract

Effective management of nitrate loading in complex river systems requires quantitative estimation to trace different nitrogen sources. This study aims to validate an integrated framework using soluble nitrogen isotope ratios (δ¹⁵N–NH₄ and δ¹⁵N–NO₃) and hydrological modeling (hydrological simulation program SPARROW) of the main stream and tributaries in the Yeongsan River to determine anthropogenic nitrogen fluxes among different land-use types in the complex river watershed. The δ¹⁵N–NH₄ and δ¹⁵N–NO₃ isotopic compositions varied across different land-use types (4.9 to 15.5‰ for δ¹⁵N–NH₄ and −4.9 to 12.1‰ for δ¹⁵N–NO₃), reflecting the different sources of nitrogen in the watershed (soil N including synthetic fertilizer N, manure N, and sewage treatment plant effluent N). We compared the soluble nitrogen isotopic compositions (δ¹⁵N–NH₄ and δ¹⁵N–NO₃) of the river water with various nitrogen sources (soil N, manure N, and sewage N) to assess their contribution, revealing that N from sewage treatment plant effluent as a point source was dominant during the dry season and N from forest- and soil-derived non-point sources was dominant due to intensive rainfall during the wet season. The coefficient of determination (R²) between the measured pollution load and the predicted pollution load calculated by the SPARROW model was 0.95, indicating a high correlation. In addition, the EMMA-based nitrogen contributions compared to the SPARROW-based nitrogen fluxes were similar to each other, indicating that large amounts of forest- and soil-derived N may be transported to the Yeongsan River watershed as non-point sources, along with the effect of sewage treatment plant effluent N as a point source. This study provides valuable insights for the formulation of management policies to control nitrogen inputs from point and non-point sources across different land-use types for the restoration of water quality and aquatic ecosystems in complex river systems. Given the recent escalation in human activity near aquatic environments, this framework is effective in estimating the quantitative contribution of individual anthropogenic nitrogen sources transported along riverine systems.

1. Introduction

The increase in nitrogen sources (e.g., sewage treatment plants, domestic sewage, chemical fertilizers, and livestock manure) entering rivers has recently emerged as an important environmental issue from the perspectives of water quality and human health [1,2,3]. In particular, anthropogenic exogenous organic matter derived from point and non-point sources such as agricultural soil including chemical fertilizer, livestock manure, and wastewater treatment plant effluent contributes to the annual increase in total nitrogen inputs to tributary streams [4]. Thus, point and non-point source management is critical in a complex river watershed of mixed land-use type with multiple pollution sources.

Water quality management in rivers and lakes has largely centered on point sources from industrial facilities and urban areas, which can be largely eliminated through advances in wastewater treatment technology. However, in watersheds of mixed land use such as rural areas, various non-point sources such as agricultural land, livestock, and forests are scattered, and unlike point sources, it is difficult to determine pollution sources and runoff paths, making it difficult to improve water quality [5,6,7]. In addition, excess nutrients from non-point sources deposited into streams and rivers cause water quality degradation, for example, eutrophication and hypoxia, and are transported to bays and coastal areas through geomorphic and hydrologic transport, causing serious environmental problems [8,9]. For this reason, managing non-point sources and understanding the processes that influence nitrate concentrations are critical for sustainable water management and water quality conservation.

Efforts have been made in past decades to identify and quantify nitrogen pollution sources for effective water quality management [10,11,12]. Policies are being made to reduce nitrogen sources and to improve the water quality flowing from upstream and downstream through efficient management of point and non-point sources, such as through ecological restoration and reservoir water quality management. Nevertheless, in watersheds with complex land-use patterns, nitrogen transport processes are unclear, and the quantification of sources remains difficult [13,14]. However, approaches using stable nitrogen isotope ratios of ammonium and nitrate (δ¹⁵N–NH₄ and δ¹⁵N–NO₃) have garnered attention in tracking sources of nitrogen pollution [15,16,17]. The technique employing the double-use of nitrogen (δ¹⁵N–NO₃) and oxygen (δ¹⁸O–NO₃) in stable isotope analysis of nitrates using a bacterial denitrification method has been introduced [18]. Notably, dual-isotope approaches using nitrogen and oxygen isotope ratios of nitrate have been successful in identifying nitrogen sources [19,20,21]. Theoretically, nitrate (NO₃) formed through nitrification has a total of three oxygen atoms and two of them originate from water (nitrite, NO₂⁻) and the remaining one originates from nitric oxide (NO). Therefore, the stable isotope ratios of nitrogen and oxygen in nitrates are determined by the nitrogen pool characteristics and chemical isotope fractionation that exist at the time of production. The natural ratios of stable nitrogen and oxygen isotopes of nitrates have been used to identify various nitrogen pollution sources such as agricultural land, forest reservoirs, and lakes [17,20,21]. This is because nitrate (NO₃) is the dominant form of nitrogen entering surface waters, and each nitrogen source has a distinct stable isotope ratio. For example, nitrate derived from manure and wastewater (10–20‰) has a much heavier nitrogen isotope ratio than nitrate derived from synthetic fertilizers (from −3 to 3‰), atmospheric deposition (from −15 to 7‰), or soil (from −6 to 9‰) [22]. To build libraries that reflect these characteristics and estimate the relative contribution of each pollutant, quantitative approaches such as linear mixed models have recently been applied. The iso-source program provided by the U.S. Environmental Protection Agency has traditionally been used due to its simplicity and ease of use. An end-member mixing analysis (EMMA) based on the R model has recently been utilized to compensate for the shortcomings of the iso-source model [16,20,21]. However, when it comes to quantitative estimation of nitrate sources, the overprinting of specific nitrate end-members causes over- and under-estimation of nitrate sources from complex land-use types, leading to quantitative limitations.

Hydrologic models have been widely applied to estimate watershed-scale hydrologic conditions and a wide range of pollutant loads derived from point and non-point sources using simple statistical and mechanistic estimation tools [21,23]. In this regard, a variety of hydrologic models, such as the agricultural non-point pollution source model, the soil and water assessment tool, the Hydrological Simulation Program—Fortran, and spatially referenced regression on watershed attributes (SPARROW), have recently been used to predict substantial changes in nitrate fluxes under different land-use types in a wide range of river basins [24,25,26,27]. Among these models, SPARROW is used to estimate the flux of nutrients from point, non-point, and diffuse sources to tributary streams and watersheds in complex watershed environments of different sizes [28]. The model effectively analyzes watershed function by utilizing three main factors: pollutant runoff from various land uses, pollutant transfer from land to streams, and attenuation and loss of streams and lakes [25,28]. In terms of scalability, the SPARROW model is well suited for application to both small and large watersheds [28]. In particular, nitrate fluxes were calculated using a hydrologic model to estimate the total stream nitrate balance, which reflects the hydrologic and chemical characteristics of different land-use types [21,29]. However, the lack of monitoring data can affect the accurate estimation of various nitrate flows in complex watersheds in terms of the calibration and validation of hydrologic models [30]. Therefore, the simultaneous application of both techniques (isotopic contribution estimation and hydrologic model simulation) can compensate for their respective limitations and increase the accuracy of quantitative loading assessments of nitrate sources in complex aquatic systems.

This study aims to quantify the nitrogen contribution from various non-point sources (livestock, domestic, and soil) in a complex river watershed, the Yeongsan River. The strategy of this study to validate this approach is to (1) analyze the stable isotope ratios of representative pollutant sources (end-member: agricultural soil including N fertilizer, manure, and sewage), (2) estimate the contribution of nitrogen sources by fitting the multi-isotope composition ratios (δ¹⁵N–NH₄ and δ¹⁵N–NO₃) to the EMMA model, and (3) cross-validate the nitrogen fluxes calculated from the hydrological model considering the river network. This study intends to determine quantitative inputs of nitrate origin transported to riverine systems based on cross-validation and integration of these two independent methods.

2. Materials and Methods

2.1. Site Description

The Yeongsan River, one of the four major rivers of Korea, is located in the southwestern part of the Korean Peninsula (longitude: 126°26′12″–127°06′07″ east; latitude: 34°40′16″–35°29′01″ north) (Figure 1). The Yeongsan River has a watershed area of 3468 km² and a flowage of 129.5 km, making it the fourth largest in South Korea and accounting for approximately 29% of the total area of Jeollanam-do Province. The main tributaries of the Yeongsan River include the Gwangju, Jiseok, Hampyeong, and Gomakwon streams and the Hwangryong River as a main river. Along the river basin, there are four agricultural reservoirs, including Damyang, Jangsu, Gwangju, and Naju reservoirs, in the main stream and tributaries upstream of the river; two large reservoirs in the middle and lower reaches of the main stream; and an estuary dam at the river mouth. In the study area, one-third of the river basin area is cultivated, and more than half of the basin is covered by forest. Coniferous, deciduous, and mixed forests predominate (45% of land cover), with agricultural land (paddy fields and crop land) comprising 37% and urban areas 7.9% [31]. The Yeongsan River basin belongs to the East Asian monsoon climate, which is characterized by intensive and prolonged precipitation during the summer months, with approximately 61% of the annual average precipitation (1305 mm) occurring in the summer (June to August, rainy season) [32]. While the Yeongsan River has an average flow of approximately 2.0 × 10⁹ m³ yr⁻¹ and is mostly used for agriculture, the drinking water for the 1.9 million people living in the vicinity is sourced from outside the watershed, which means that the river is severely polluted by runoff from large sewage treatment plants and agricultural areas. The actual capacity of the Gwangju Sewage Treatment Plant, which is located in the middle of the Yeongsan River, is 600,000 m³/day, and the flow rate upstream of the Yeongsan River is sometimes less than this value [33]. Nitrogen pollution levels are high, with an average concentration of total nitrogen of 4 mg/L and an annual average maximum concentration of over 27 mg/L. Nitrogen fertilizer use was 3.8 × 10⁷ kg yr⁻¹, with urea and compound fertilizers being used the most [34].

2.2. Sample Collection

River water sampling was conducted at the main-stream sites (11 sites) and the tributary stream sites (11 sites) of the Yeongsan River (Figure 1), and the survey was divided into pre-monsoon (May, June, and July 2018), monsoon rainfall (August 2018), and post-monsoon (September and October 2018) periods, considering the weekly cumulative rainfall. Each water sample was collected sequentially as the water flow moved downstream. Field surveys were conducted in the center of the stream using collection equipment pre-cleaned with deionized water, and samples were stored in sterile collection bottles for transport to the laboratory.

End-members were obtained from pure pollution sources classified into three groups: the soil, manure, and sewage sources. The manure source samples were compost effluent, pig manure liquid, and livestock wastewater; the sewage source samples were sewage treatment plant effluent; and the soil source samples were chemical fertilizer, paddy and field runoff, and forest runoff, which were directly collected for measuring ratios of stable nitrogen isotopes (δ¹⁵N–NH₄ and δ¹⁵N–NO₃) used as end-member values.

2.3. Analysis of Water Quality Parameters

Water temperature, pH, DO, and electrical conductivity were measured in the field using a field measurement device (EXO1 multiparameter sonde, YSI, Yellow Springs, Ohio, USA), and general water quality parameters (TN, DTN, NH₃–N, NO₃–N, Cl⁻) were collected in the field, and samples were transported to the laboratory and stored at 4 °C. Subsequently, TN (continuous flow method, AACS-V, BLtec, Seoul, Republic of Korea), DTN (continuous flow method, AACS-V, BLtec, Seoul, Republic of Korea), NH₃–N (continuous flow method, Auto Analyzer3, BLtec, Seoul, Republic of Korea), NO₃–N (ion chromatography, Dionex integration, Thermo Fisher Scientific, Waltham, MA, USA), and Cl⁻ (ion chromatography, Dionex integration, Thermo Fisher Scientific, Waltham, MA, USA) were analyzed according to the Standard Methods for the Examination of Water and Wastewater.

2.4. Analysis of Stable Isotopes

The collected samples were immediately filtered through GF/F filter paper (25 mm, 0.65 µm) that had been burned at 45 °C for 24 h to remove organic matter, and the filtrate was used for stable nitrogen isotope ratio measurements for ammoniacal (NH₄–N) and nitrate nitrogen (NO₃–N) using the Kjeldahl distillation method [35]. The prepared filtrate was distilled with a Kjeldahl distillation apparatus with the addition of magnesium oxide (MgO) for approximately 40 min, and ammoniacal nitrogen (NH₄–N) was captured with 0.01 N sulfuric acid (H₂SO₄) solution, and the remaining distilled sample was redistilled with Devardal alloy to capture nitrate nitrogen (NO₃–N). The captured ammoniacal (NH₄–N) and nitrate nitrogen (NO₃–N) samples were adjusted to pH 3 with 0.01 N sodium hydroxide (NaOH) and freeze-dried at –80 °C to obtain a white powder of ammonium sulfate. The pretreated white powder was placed in a tin capsule and sealed to measure stable isotope ratios of carbon and nitrogen using a stable isotope mass spectrometer (IRMS, Isotope Ratio Mass Spectrometry, Isoprime Vision, Manchester, UK) coupled to an elemental analyzer (EA, Elemental Analyzer, Vario Isotope Cube, Langenselbold, Germany). High-purity oxygen greater than 99.995% as the combustion gas and ultra-high-purity helium gas greater than 99.999% as the carrier gas were flowed at a rate of 180 mL/min. Carbon dioxide (99.995%) and nitrogen (99.999%), which were used as standard gases, were injected into the stable isotope mass spectrometer at time-lag intervals with the analyzed samples, and their isotopic area ratios were used to calculate the δ¹⁵N value as follows [19]:

δ = [(R_sample/R_standard) − 1] × 1000

R = ¹⁵N/¹⁴N

(1)

where R denotes the stable isotope ratio, ¹⁵N/¹⁴N, R_sample denotes the sample, and R_standard denotes the respective stable nitrogen isotope ratio for the standard. N₂ (atmospheric air) was used as the reference standard for nitrogen, and the analytical precision was within 0.5‰. To evaluate the precision and accuracy of Kjeldahl processes, two reference materials (IAEA-NO-3 (potassium nitrate) and IAEA-N-1 (ammonium sulfate)) were analyzed repeatedly. The δ¹⁵N–NO₃ and δ¹⁵N–NH₄ values of IAEA-NO-3 and IAEA-N-1 were 4.7 ± 0.3‰ and 0.4 ± 0.1‰, respectively, which are within the recommended values of analytical uncertainties.

2.5. End-Member Mixing Analysis (EMMA)

The nitrogen source contribution rate was estimated using stable isotope analysis in R based on the Bayesian mixing model (Bayesian Mixing Model in R; MixSIAR, version 3.1.10) [36]. SIAR mixture models have been successfully applied to consumer food source partitioning or contaminant tracking using different stable isotope ratios [16,37,38]. To determine the relative contribution of each anthropogenic source (soil, manure, and sewage) in the Yeongsan River, this study used an end-member mixing analysis (EMMA) model based on the Bayesian mixing model, specifying the pollution source representative as the source and the water quality data as the mixture.

2.6. Hydrological Model

SPARROW is a watershed modeling technique that uses multivariate statistical and process-based approaches to estimate pollutant sources and transport load in river watersheds. The mathematical core of the SPARROW method is a relationship that expresses the in-stream T-N load (i.e., transport rate) of a pollutant at the end-downstream of a given reach as the sum of the monitored and unmonitored contributions to the load at the location of all upstream pollutant sources. In this study, the SPARROW model was parameterized using Bayesian inference [27], and the equation is organized as follows [27]:

$$L_i=\sum _{n=1}^N{S}_{n,i}$$

$$\begin{array}{c}{Y}_i\sim N\left({\mu }_i,{\sigma }_i^2\right)\\ {\mu }_i\sim N\left(Loa{d}_i,{\delta }^2\right)\\ Loa{d}_i={\displaystyle \sum _{n=1}^N}{\displaystyle \sum _{j=1}^{J_i}}{\beta }_n{S}_{n,j}{e}^{\left(-\alpha Z_i\right)}{H}_i^s\end{array}$$

In the aforementioned equations, Y_i is the observed T-N load and follows a normal distribution with mean μ_i and variance σ²_i with a non-informative prior distribution. Load_i denotes the T-N load, N is the T-N load at each observation point, and J_i denotes the number of stream networks in the Yeongsan River system. In the equation, β_nS_n,j denotes the pollutant source factor for each network, e^(−αZi) denotes the delivery factor, and H^s_i denotes the decay factor. The use of Bayesian inference increases the accuracy of the model while accommodating the uncertainty associated with the model parameters.

3. Results and Discussion

3.1. Water Quality during Monsoon Period

The total precipitation during the monsoon rainy season (August) in the study area amounted to 400 mm. Specifically, during the storm event period, the precipitation peaked at 397.1 mm. In contrast, the pre- and post-monsoon periods experienced significantly lower levels of precipitation at 85.4–222.4 mm (May–July) and 125.2–129.7 mm (September–October), respectively (Figure 2). The changes in TN, TDN, NO₃–N, NH₄–N, and Cl⁻ concentration showed significant differences between the wet and dry seasons, which were attributed to variability in precipitation and pollutant sources (p < 0.01). During the study period, the concentration of total nitrogen in the Yeongsan River watershed ranged from 0.5 to 11.7 mg/L, mainly in the form of dissolved nitrogen. NO₃–N accounted for the largest proportion of dissolved nitrogen, ranging from 51.2 to 71.2% (Figure 2). During the study period, NO₃–N averaged 1.54 mg/L (0.22–6.52 mg/L), with averages of 0.13 ± 5.52 mg/L and 0.32–7.23 mg/L in the dry season (May, June, July, September, and October) and wet season (August), respectively. The average concentrations for the two seasons were statistically significantly different (N = 123, p < 0.01). In particular, in the study area, nitrogen fertilizer compounds in the form of reduced urea and ammonium are mostly used and are present in high concentrations in the soil. Therefore, nitrogen sources containing nitrogen fertilizers from forest or agricultural soil flow into the Yeongsan River due to concentrated rainfall during the rainy season, leading to higher NO₃–N concentrations [31]. Spatially, NO₃–N concentrations were higher in tributary sections than in main-stream sections, with extremely high concentrations in the Gwangju tributary stream, where domestic sewage was concentrated. In the context of sewage treatment plant operations, DIN concentrations can be affected by the treatment process, including primary treatment for physical sedimentation, secondary treatment for filtration and biological treatment, and tertiary treatment for UV and chemical treatment [16,39]. The basic type of treatment process may have insufficient nitrogen removal when water flows and nitrogen concentrations are high, therefore higher nitrogen concentrations near a sewage treatment plant (as a point source) suggest the possibility that untreated input sewage may be discharged nearby.

During the study period, the NH₄–N concentration averaged 1.54 mg/L (0.22–6.52 mg/L), with average values of 0.13 ± 5.52 mg/L and 0.32–7.23 mg/L in the dry season and wet season, respectively. The average concentrations for the two seasons were statistically significantly different (N = 123, p < 0.01). This is because nitrogen sources from fertilizers and manure from agricultural and livestock activities concentrated around the Yeongsan River have a large impact [8,40]. Spatially, NO₃–N and NH₄–N concentrations were higher in tributary sections than in main-stream sections, with extremely high NH₄–N concentrations in the Gwangju tributary stream with many livestock facilities and extremely high NO₃–N concentrations in the Yeongsan tributary stream with many domestic sewage discharges. This spatial distribution suggests that the origin of nitrogen sources in the Yeongsan River varies depending on the land-use type (urban, agricultural, or forest) around the study area [20,21].

3.2. Identification of Nitrogen Sources and Contribution Using Dual Isotope Ratios

Stable isotope analyses in river water samples are shown in the figure below, divided into wet and dry seasons (Figure 3). The mean values of δ¹⁵N–NH₄ and δ¹⁵N–NO₃ at all sites investigated in the main stream and tributaries of the Yeongsan River watershed were 13.1 ± 2.8‰ and 4.9 ± 1.6‰, respectively, with ranges of 6.0~23.3‰ and 0.5~3.0‰; 14.3 ± 2.9‰, 5.4 ± 1.4‰; and 9.6~17.5‰, 3.3~8.3‰, respectively, for the main stream, and 11.7 ± 2.3‰, 4.4 ± 1.6‰; and 8.9~16.0‰, 2.8~8.0‰, respectively, for the tributary streams (Figure 3). Seasonally, δ¹⁵N–NH₄ and δ¹⁵N–NO₃ values ranged from 6.0~15.0‰ and from 0.5~6.4‰, respectively, in the wet season and 7.1~23.3‰ and 1.2~13.0‰, respectively, in the dry season, with higher values in the dry season than in the wet season, especially in the main stream when compared to the tributaries. In general, soil-derived nitrogen and applied chemical fertilizers enter rivers through soil erosion during rainfall, resulting in relatively low δ¹⁵N–NH₄ and δ¹⁵N–NO₃ values [41]. This implies that heavy rainfall in the rainy season has increased the impact of nitrogen from soil (soil N) and chemical fertilizer (NH₄⁺ in fertilizer) sources, as these two sources have lower δ¹⁵N–NH₄ and δ¹⁵N–NO₃ values compared to manure and sewage [16,30]. In the dry season, the relatively heavier δ¹⁵N–NH₄ and δ¹⁵N–NO₃ values are attributed to the extremely high volume of discharges from sewage treatment plants and manure treatment plants, which are point sources rather than non-point sources. Anthropogenic inputs from point sources (manure and sewage) play a large role in increasing NO₃⁻ concentrations during rainfall events. This is also supported by the relationship between the NO₃⁻/Cl⁻ ratio and Cl⁻ concentration found in the Yeongsan River.

Since chlorine does not undergo physical, chemical, or biological processes, it is a good indicator of sewage impact and dilution [42]. High Cl⁻ concentrations mainly come from manure and sewage (Figure 4). The relationship between the NO₃⁻ and Cl⁻ concentration in Yeongsan River shows that the influence of soil and chemical fertilizers increases during the rainy season, and the influence of manure and sewage becomes apparent during the dry season. These features are also consistent with seasonal changes in δ¹⁵N–NH₄ and δ¹⁵N–NO₃ values.

NH₄ and NO₃ ions entering the water environment can potentially be divided into soil, manure, and sewage sources [30]. In this study, three types of pollution sources (soil, manure, and sewage) were selected in the study area (Figure 3). The difference between the isotopic values of the three end-members (manure, soil, and sewage) is determined by the isotopic values of the samples collected from the source or the raw material itself, such as commercial fertilizer. In addition, uncertainty in the selection of potential sources can lead to errors in the calculation of accurate source-specific contributions using EMMA models [16,20,43]. The validity of the end-member values selected in this study was examined, and the δ¹⁵N–NH₄ and δ¹⁵N–NO₃ isotopic values between each end-member showed distinct, non-overlapping differences between each end-member: from 2.7% to −1.8% for the soil source, from 28.1% to 6.8% for the manure source, and from 12.5%, 15.0% for the sewage source (p < 0.001), which were in a range similar to the values reported in the literature [16,20]. The δ¹⁵N–NH₄ isotopic abundance of the manure source is relatively heavier than that of other sources, which is attributed to the inclusion of a prominent nitrogen source from livestock manure [8]. The sewage source has the largest δ¹⁵N–NO₃ isotope values, which are remarkably similar to the δ¹⁵N–NO₃ values of wastewater treatment plant and pig manure treatment plant effluents reported in a previous study [44]. The existing literature reported that δ¹⁵N–NO₃ values increased as sewage treatment plant and pig manure treatment plant discharge entered the river, allowing for the identification of the nitrogen pollution source [16,21]. Moreover, among the three end-members, the soil source has the lowest δ¹⁵N–NH₄ and δ¹⁵N–NO₃ values. This represents a similar range to those in the existing literature, which reports that soil sources containing inorganic fertilizers, such as those applied to agricultural fields, have low δ¹⁵N–NH₄ and δ¹⁵N–NO₃ values as they enter rivers [20,44].

In this study, the three available end-member values and seasonal averages of stable nitrogen isotope ratios (δ¹⁵N–NH₄ and δ¹⁵N–NO₃) measured in rivers were used to estimate the contribution of each pollutant source after input to the EMMA model. During the survey period, as for the contribution of nitrogen pollution sources in the main stream (upper, middle, and lower reaches) and tributaries of the Yeongsan River system, the sewage source accounted for 24 ± 0.7%, the manure source for 31 ± 1.8%, and the soil source for 45 ± 1.6% (Figure 5). In the upper reaches, the soil source contributed 60 ± 2.5%, higher than the manure source (19 ± 0.2%) and the sewage source (21 ± 0.8%). In the middle reaches, the sewage source (32 ± 3.5%), soil source (32 ± 2.9%), and manure source (36 ± 3.7%) all contributed similarly (Figure 5). Finally, in the lower reaches, the manure source (50 ± 4.2%) contributed relatively more than the sewage source (18 ± 2.2%) and the soil source (32 ± 1.3%) (Figure 5).

The upper region of the Yeongsan River is mainly composed of agricultural land, and the share of forests in the land-use type is 55.7%, which is much higher than that of paddy fields (16.5%) and fields (9.3%) [34], which explains the high contribution of the soil source in the upstream. The middle region of the Yeongsan River has the largest urban area with 22.0% (132.1 ha), and the Gwangju 1 Sewage Treatment Plant (average discharge of 600,000 tons/day) is located there, which explains the high contribution of the sewage source in the middle reaches [31]. In addition, the total nitrogen load of the M11 (Juksan) site in the lower reach of the Yeongsan River is 485 tons [34], and the annual nitrogen fertilizer application is calculated to total 2,727 tons [45]. In general, the runoff rate from nitrogen fertilizer application is approximately 7% [46], and based on this, the runoff from nitrogen fertilizer at point M11 of the Yeongsan River was estimated to be approximately 62 tons. Therefore, it is determined that the contribution of the manure source (50%) to the downstream area is relatively higher than that of the sewage source (18%) and the soil source (32%).

The results of the comparative evaluation divided into dry and wet seasons showed that in the dry season, the contribution of the sewage source accounted for 25%, the manure source for 32%, and the soil source for 43%, while in the wet season, that of the sewage source and manure source decreased by 9% and 3%, respectively, and the soil source increased by 14%. The relatively higher contribution of the sewage source in the dry season compared to the wet season is a direct result of sewage treatment plant effluent [16,20]. The study area has three large sewage treatment plants that discharge many sewage nitrogen sources into the Yeongsan River during the dry season. However, during the rainy season, the discharge is reduced due to the risk of flooding. In the dry season, nitrogen concentrations are considerably higher compared to the wet season, and the δ¹⁵N–NH₄ and δ¹⁵N–NO₃ isotope ratios are heavier than those in the wet season, reflecting the representativeness of sewage as a major pollutant of anthropogenic origin [41]. Most of the sewer systems in the basin are combined systems and experience severe combined sewer overflows during the rainy season. In addition, most sewage treatment plants are operated in conjunction with a manure treatment plant, and the utilization rate of the manure treatment plant is adjusted according to the total amount of sewage entering the sewage treatment plant. As a result, the utilization rate of manure treatment plants decreases during the rainy season, which may have contributed to a lower contribution of the manure source in the rainy season than in the dry season. Meanwhile, the contribution of the soil source increased in the rainy season. The higher contribution of the soil source in the rainy season is due to non-point source pollution from surrounding forests and agricultural lands entering the Yeongsan River. In general, contributions from soil and fertilizer N sources tend to increase during the rainy season when there are intense downpours [20], and human disturbance results in soil particles containing more nitrogen entering rivers [30,41]. Due to the large amount of agricultural land effluent in the study area and the high dosage of synthetic fertilizers, the stable isotope ratios δ¹⁵N–NH₄ and δ¹⁵N–NO₃ of soil N and fertilizer N often overlap. This is because the nitrogen source in chemical fertilizer added to soil is transformed and synthesized by physical and biological processes together with the nitrogen source in the soil [42]. Although it is not possible to distinguish between the two sources in this study, the proportion of the soil source is higher than in the dry season due to heavy rainfall during the rainy season, which causes nitrogen sources from non-point sources such as forests and agricultural land to enter the Yeongsan River. In addition, the overall land-use type of the Yeongsan River shows an extremely high proportion of forests and agricultural land [34], which explains the high contribution of the soil source in the wet season. This study hypothesizes that nitrogen sources can be more effectively distinguished if dual δ¹⁵N–NH₄ and δ¹⁵N–NO₃ axes, rather than a single-isotope approach, are applied in complex river watersheds, especially those with sewage treatment plants. The identified manure and sewage sources based on the δ¹⁵N–NO₃ and δ¹⁸O–NO₃ values resulted in overlapping of corresponding values between some sources, leading to lack of identification accuracy [17,47,48]. Additionally, the sensitivity of δ¹⁵N–NH₄ to NH₄⁺–N concentration is greater than that of δ¹⁵N–NO₃ to NO₃⁻–N concentration, even when the NH₄⁺–N concentration is low [16]. Therefore, the applicability of δ¹⁵N–NH₄ and δ¹⁵N–NO₃ axes should be evaluated in complex river watersheds including agricultural areas and intensive livestock farming where nitrogen is predominantly released into aquatic environments in the form of NH₄⁺–N. To the best of our knowledge, the δ¹⁵N–NH₄ and δ¹⁵N–NO₃ axes have not yet been applied to identify nitrogen sources in agricultural aquatic environments. This technique should be validated in areas where sewage treatment facility effluent is discharged into rivers or in complex watersheds that face similar problems [49,50,51].

Excluding the previously mentioned non-point sources, groundwater inflation and air deposition, excluding non-point sources, can lead to increasing amounts of contaminants such as ammonium and nitrate in aqueous water systems [52]. The nitrogen source in atmospheric deposition comes mainly from the combustion of fossil fuels, such as coal and oil. Atmospheric deposition may be in a dry form, such as droplets and particulates, or a wet form, such as fog, rain, and snow. Furthermore, groundwater is a potentially important source of nitrogen, but it is difficult to quantify due to the fact that limited or poorly understood data on them exist. It is important to determine the amount of nitrogen contributed by these undocumented sources within watersheds to target those that contribute most to nitrogen pollution problems. Therefore, we should consider atmospheric deposition and groundwater inflow to complex watersheds in future studies.

3.3. Nitrate Loads Based on the EMMA and Hydrological Model (SPARROW)

Temporal and spatial changes in NO₃–N sources can be influenced by water quality releases from upstream, which in turn affect nitrate levels in water bodies [53]. Therefore, the amount of discharge from each sub-basin can be an important factor in controlling quantitative nitrate exports. This is because changes in discharge can potentially be controlled by a variety of ambient parameters, such as precipitation, inflow rate, and morphological characteristics of the river [36,54]. The input data of SPARROW model in the Yeongsan River watershed was collected, and a database (DB) was established using the land-use area (agricultural land, urban and forest) by reach within the Yeongsan River watershed, the discharge load of waste treatment plants, and the time of travel of the Yeongsan River watershed (Figure 6).

To improve the reliability of the multivariate regression model between water quality and pollutant sources, the Bayesian mixing model was utilized to estimate the values of the parameters of the pollutant sources responsible for the T-N pollution load and to estimate the contribution rate. The coefficient of determination (R²) between the measured pollution load and the predicted pollution load calculated by the SPARROW model was 0.95, indicating a high correlation (Figure 6). The MSE, an error analysis used to assess the accuracy of the model, was calculated to be 0.187. The closer the coefficient of determination (R²) is to 1 and the closer the MSE is to 0, the higher the accuracy and the better the performance of the model [27]. Based on the above results, the model is judged to be reliable.

As a result of applying SPARROW regression model to evaluate the contribution rate, among the overall contribution rates of pollution sources, the contribution rate of pollution load from agricultural land was found to be the highest in the upper reaches at 66%; the middle reaches had a high load from domestic sewage and sewage treatment plants discharged from urban areas; and the lower reaches of the Yeongsan River had a high load of manure sources originating from livestock farms (Figure 7). In terms of land-use area, the Yeongsan River watershed consisted of 1675 km² of forest, 1017 km² of agricultural land, and 347 km² of urban area. Comparing the contribution rate of pollution load with the area of land use type, the contribution rate of pollution load was high at 66% due to the high area of soil and agricultural land in the Yeongsan River watershed, confirming the importance of managing pollution n sources in soil systems such as forests and agricultural land. The graphs below compare the contribution rate calculated by the EMMA model using stable isotope ratios and calculated by the SPARROW model using land use types (Figure 8). As a result of comparing the contribution rates calculated by the SPARROW and EMMA model, it showed a similar trend, suggesting that the SPARROW model can be used not only to consider the reliability analysis of stable isotope tracer methods but also to establish strategies for source management in watersheds. In recent years, a number of studies have compared the contribution of pollutants from different sources using stable isotope ratios with the contribution of flow using a watershed model to improve the reliability of results [16,21,55].

The spatio-temporal increase in nitrate loading in the Yeongsan River was similar to increases in other large-scale watershed systems, including increased agricultural source fluxes of nitrogen in the Oak Creek region [56], increased sewage runoff in the Dead Run watershed [13], anthropogenic source inputs in the Xijiang River region [30], and increased sewage treatment plant loading in the Geumho River [21]. Although anthropogenic activities in the Yeongsan River are small compared to other watersheds, except for sewage treatment plants, excessive nitrate runoff in space and time seems to play a decisive role in the deterioration of water quality in the river basin. The integrated approach of the EMMA technique using stable isotope ratios and the SPARROW regressing model suggests that anthropogenic nitrate inputs from non-point source pollution (soil, agriculture, manure, and sewage) should provide a substantially higher contribution to nitrate levels in complex river basins. Soil- and fertilizer-derived nitrogen sources are dominant in the study area in terms of space and time. Frequent use of synthetic fertilizers can enhance nitrate diffusion into the soil column, which can increase groundwater infiltration of fertilizer-derived nitrate [57,58]. Under these conditions, the signature of anthropogenic nitrate-derived isotope ratios can be replaced by the simultaneous occurrence of substantial nitrogen transformations between the soil layer and groundwater [59,60,61,62,63]. It should be noted that this process may underestimate the contribution of nitrogen sources from soil and fertilizers.

3.4. Implication for the Applicability of Nitrogen Source Techniques and Their Further Possible Development

The integrated approach of this study provides a reliable and practical tool for estimating the individual contributions of non-point sources associated with anthropogenic activities to effectively manage increased anthropogenic nitrate inputs. While the two methods (EMMA and SPARROW) each have advantages in tracking the origin and hydrologic behavior of nitrogen river systems, the exclusive applicability of each approach to quantitatively estimating nitrate export to complex river systems may be limited. To address these limitations, this study proposes a phased cross-checking approach with complementary methods. So far, there has been no single approach to effectively assessing specific nitrate source fluxes and loads in river basins around the world. Furthermore, it is challenging to directly compare nitrate loads derived from different land-use types in complex watershed. The integrated approach in this study essentially involves a comparison of EMMA and SPARROW. In complex river systems, a phased approach that combines the total nitrate load from different land-use types in SPARROW with the contribution of each nitrogen source based on isotopic characteristics is needed. The integration of the two approaches compensates for each method, providing new insights into tracking different nitrate sources and showing the quantitative contribution of source-specific nitrate loads by land-use type. In addition, any obvious factors associated with nitrate fluxes (e.g., groundwater diffusion rates) should be considered in future studies. In particular, seasonal sampling over a long period will help to capture spatial and temporal variability based on the proposed framework. Therefore, this approach can be applied to the accurate estimation of nitrate load with spatio-temporal distribution according to various land-use types and can be helpful in establishing practical government policies for sustainable river water quality management and conservation.

4. Conclusions

For the quantitative estimation and management of nitrate fluxes in a complex river basin, a nitrate tracer study of anthropogenic origin into the Yeongsan River was conducted using stable isotope ratios (δ¹⁵N–NH₄ and δ¹⁵N–NO₃) and a hydrological model approach (SPARROW). As a result of comparing the isotopic composition of ammonium and nitrate (δ¹⁵N–NH₄ and δ¹⁵N–NO₃) in the river water with various nitrate sources (livestock manure, forest soil N, agricultural synthetic fertilizers, and sewage treatment plant discharge) to assess their contribution to pollution, N from sewage treatment plant effluent was dominant during the dry season, and N from forest and soil sources was dominant during the wet season. The EMMA-based contribution rate results compared to the SPARROW-based nitrate fluxes were similar to each other, indicating that large amounts of forest- and soil-derived N may be transported to the Yeongsan River watershed as non-point sources, along with the effect of sewage treatment plant effluent as a point source. Therefore, the integrated approach of this study provides important evidence for a substantial increase in anthropogenic causes based on enhanced nitrogen balance and fluxes in the river. As a result, an integrated approach is useful and meaningful for estimating the quantitative contribution of natural and anthropogenic nitrogen transported from point and non-point sources.