Resolving Nitrate Sources in Rivers Through Dual Isotope Analysis of δ¹⁵N and δ¹⁸O

Abstract

Nitrate (NO₃⁻) pollution in rivers within agricultural regions has become a global issue that cannot be ignored. Identifying the sources and transformation processes of NO₃⁻ is crucial for safeguarding water quality in agricultural catchment areas. This study traces the sources and transformation processes of NO₃⁻ in the Songhua River basin of Northeast China by analysing hydrochemical parameters and NO₃⁻ dual isotopes (δ¹⁵N and δ¹⁸O) in river water. It estimates the proportional contributions of NO₃⁻ sources using Bayesian modelling via the MixSIAR package (3.1.12) in the R programming language. (1) The relatively low NO₃⁻/Cl⁻ ratio and high chloride concentrations in the upstream section indicate that the primary sources of NO₃⁻ in this area are manure and sewage (M&S). (2) Dual isotope analysis of NO₃⁻ indicates that the primary sources of NO₃⁻ in the Songhua River basin are M&S, soil nitrogen (SN), and chemical fertilizers (CF). (3) Nitrification occurs throughout the entire watershed. (4) Model estimation results indicate that SN constitutes the primary source of NO₃⁻ throughout the entire watershed (48%), with no significant variation observed across the studied river sections. However, other major NO₃⁻ sources exhibit spatially significant differences, primarily manifested as follows: M&S constitute the primary upstream source of NO₃⁻ (39%), whilst downstream contributions are predominantly attributable to CF (20%). Intermediate regions experience combined impacts from both MS and CF sources.

1. Introduction

Since the last century, the rapid advancement of human society and economies has been accompanied by a mounting impact of anthropogenic activities—including urbanization, agricultural production, and industrialization—on the biogeochemical cycling of nitrogen [1]. A primary concern is that nitrate (NO₃⁻) has emerged as a widespread and increasingly severe pollutant in river systems globally [2].Elevated NO₃⁻ concentrations can trigger a suite of environmental issues, such as aquatic eutrophication, algal blooms, greenhouse gas emissions, and the overall deterioration of water quality [3]. Furthermore, high levels of NO₃⁻ in aquatic environments pose significant threats to human health, with documented associations with conditions including diabetes and methaemoglobinaemia [4]. These risks are particularly pronounced among vulnerable populations, such as infants, young children, and pregnant women.

The accumulation of NO₃⁻ stems from multiple sources, including agricultural fertilizer runoff, domestic sewage discharge, land use changes, and AD [5]. Consequently, NO₃⁻ pollution—characterized by excessive nitrogen loads in rivers and groundwater—has evolved into a critical environmental challenge with global ramifications for water security, ecosystem health, and public well-being. Variations in precipitation patterns, wastewater inputs, and agricultural practices across different river basins further alter nitrogen biogeochemical cycles. Against this backdrop, accurately identifying and quantifying the sources and transformation processes of NO₃⁻ in rivers is indispensable for implementing effective pollution management strategies, reducing nitrogen fluxes, and advancing environmental conservation efforts [6].

Rivers can receive NO₃⁻ from diverse sources, including AD (AD), CF, SN, M&S [7]. Meanwhile, NO₃⁻ undergoes complex migration and transformation processes within rivers, such as nitrification, denitrification, and assimilation [8,9]. To mitigate NO₃⁻ pollution in rivers, accurate identification of NO₃⁻ sources and quantification of their respective contributions are crucial. Early NO₃⁻ pollution tracing primarily relied on water chemistry methods, which involved collecting and analyzing basic water chemistry indicators (pH, EC, ORP, COD) as well as parameters like NO₃⁻, NO₂⁻ and NH₄⁺ [10,11]. Based on the conversion relationships between different nitrogen forms, combined with the analysis of various environmental parameters, the sources and characteristics of NO₃⁻ pollution in water bodies were examined. However, a river basin functions as an open material-cycle ecosystem, leading to unstable outcomes in water chemistry analysis [12]. Furthermore, the sources and transformations of NO₃⁻ in water are highly complex, and conventional analytical methods are generally unable to distinguish between different sources of NO₃⁻ pollution [13]. The development of stable isotope techniques provides an opportunity to address the aforementioned limitations. Since the 1970s, stable isotopes of NO₃⁻ (e.g., δ¹⁵N-NO₃⁻) have been used to differentiate between groundwater and riverine sources of NO₃⁻. In practical research, however, the complexity of environmental conditions inevitably causes a certain degree of overlap in the nitrogen isotope signatures of different sources [14]. Thus, using nitrogen isotopes alone cannot achieve precise source tracing. Consequently, an increasing number of researchers have adopted dual nitrogen and oxygen isotope techniques (δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻) to investigate the origins of NO₃⁻ in water bodies [5]. With ongoing technological advancements, δ¹⁵N and δ¹⁸O techniques have been widely applied in environmental pollution source tracing research, exhibiting particular advantages in the source apportionment of NO₃⁻ pollution in aquatic environments [15]. Recent studies that integrated the dual-isotope approach for NO₃⁻ with water chemistry parameters (e.g., NO₃⁻/Cl⁻ ratios and major ions) have proven effective in identifying both NO₃⁻ transformations and sources [16].

As one of the most important rivers in Northeast China, the water quality of the Songhua River is significantly affected by both natural factors and human activities. Over the past two decades, nitrogen pollution in the Songhua River basin has undergone substantial dynamic changes [17,18,19]. Shifting from the early industrial-dominated ammonia nitrogen pollution to a composite pollution pattern with NO₃⁻ as the primary pollutant. Between 2000 and 2010, ammonia nitrogen (NH₃-N) was the dominant form of pollutant, accounting for more than 60% of total inorganic nitrogen. The 2005 Jilin Petrochemical explosion led to combined contamination of nitrobenzene and ammonia nitrogen, causing the ammonia nitrogen concentration in the downstream reaches to peak at 8.2 mg/L⁻¹—four times the national standard [20]. After 2010, agricultural non-point source pollution became increasingly severe [21]. Accordingly, NO₃⁻ became the predominant pollutant form, with its proportion in total inorganic nitrogen increasing from 20% to 45%, while the proportion of ammonia nitrogen decreased to 40%. According to the Statistical Yearbook released by the Heilongjiang Provincial Bureau of Statistics, agricultural fertilizer application in the basin increased sharply (reaching 5.8 million tonnes in 2005, a 120% increase compared to 2000). Meanwhile, livestock farming has become increasingly industrialized, with nitrogen losses from manure exceeding 25%. In the Harbin section of the Songhua River, NO₃⁻ concentrations exceeded 15 mg/L⁻¹ in summer, which is 1.5 times higher than the drinking water standard limit set by the World Health Organization (WHO) [22]. Consequently, identifying NO₃⁻ sources and transformation processes at the Songhua River basin scale is crucial for controlling NO₃⁻ pollution.

2. Materials and Methods

2.1. Study Area

The Songhua River is one of China’s seven major rivers and the largest tributary of the Heilongjiang River within China’s borders. The Songhua River basin lies in northeastern China, situated between 41°42′ and 51°38′ north latitude and 119°52′ and 132°31′ east longitude (Figure 1). Spanning a total length of 1927 km, the basin extends 920 km from east to west, with a north–south width of 1070 km. Its catchment area spans approximately 556,800 square km, accounting for 30.2% of Heilongjiang Province’s total river basin area of 1.843 million square km. The basin has an average annual runoff volume of 76.2 billion cubic m and traverses three provinces and autonomous regions: Heilongjiang Province, Jilin Province, and the Inner Mongolia Autonomous Region.

The entire Songhua River basin may be divided into three distinct sections: the upper reaches of the Songhua River, the Nen River, and the main stem of the Songhua River. The river basin encom-passes both northern and southern headwaters. The southern headwater is the upper reaches of the Songhua River, originating from Tianchi Lake on Mount Changbai at an elevation of 2744 m. This section spans 958 km in length with a catchment area of 73,400 square km. The northern headwater is the Nen River, arising from the Greater Khingan Range at over 1000 m above sea level. It stretches 1397 km with a catchment area of 283,000 square km. The two rivers converge at the confluence near Sanchahe Town in Jilin Province; from this confluence, the main stem of the Songhua River extends 939 km before flowing into the Heilongjiang River at Tongjiang City.

The lower reaches of the Songhua River flow through the Sanjiang Plain, a core agricultural production area in Heilongjiang Province. Located in the eastern part of Heilongjiang, this plain spans 21 counties under the administration of five municipal cities within the province, with 52 state-owned farms operating across its territory. According to the 2024 Heilongjiang Statistical Yearbook, the total agricultural fertilizer application in the Sanjiang Plain reached 776,300 tons in 2023. Among this total, nitrogen CF accounted for 261,000 tons, making up approximately one-third of the overall fertilizer use. This reflects the region’s high reliance on CF for agricultural production. Meanwhile, leveraging its solid agricultural foundation, the Sanjiang Plain also boasts a well-developed livestock husbandry sector. In 2023, the region maintained a stock of around 900,000 large livestock, along with a combined total of over 2 million pigs and sheep, and some 4 million poultry. Such large-scale livestock farming generates substantial amounts of breeding waste, which poses another environmental concern if not properly treated. Collectively, the intensive agricultural cultivation and large-scale livestock farming in the Sanjiang Plain indicate significant human activity interference in the area. These activities are likely to increase the risk of agricultural non-point source pollution, thereby exerting potential adverse impacts on the water quality of the lower Songhua River.

The Songhua River Basin serves as the core region of China’s “Northeast Granary”, primarily producing rice, maize, and soybeans, with its annual grain output accounting for over 10% of the national total. After the Songhua River flows into the Heilongjiang River in Tongjiang City, the area enclosed by the Heilongjiang River, the lower reaches of the Wusuli River, and the Songhua River forms the renowned Sanjiang Plain. As the principal agricultural production zone within the Songhua River Basin, the Sanjiang Plain boasts abundant black soil resources. Within the entire watershed, the land use structure presents clear characteristics: agricultural land accounts for 47.01%, woodland 41.79%, grassland 2.33%, water bodies 1.74%, developed land 3.46%, and bare land 3.67%. Located in the northern temperate monsoon climate zone, the watershed features four distinct seasons, with warm and rainy summers, cold and dry winters, and significant annual temperature variations. The long-term average annual temperature ranges from 3 °C to 5 °C. July records the highest temperatures, with a daily average of 20–25 °C and historical peaks exceeding 40 °C; January witnesses the lowest temperatures, with a monthly average below −20 °C, and the lowest recorded temperature near Zhalantun on the Nenjiang River reached −42.6 °C. The long-term average annual precipitation is approximately 500 mm, with notable spatial differences: rainfall in the southeastern mountainous regions can reach 700–900 mm, while the arid western basin areas receive only 400 mm. The overall distribution pattern shows higher precipitation in hilly and mountainous areas than in plains, slightly higher in the south and centre, moderately higher in the east, and the lowest in the west and north. Temporally, precipitation is highly concentrated: rainfall during the flood season (June to September) accounts for 60–80% of the annual total, whereas winter precipitation (December to February) constitutes merely about 5% of the annual total.

2.2. Sample Collection

From 20 October to 5 November 2024, a total of 27 river water samples were collected from the main streams of three major rivers (the main stem of the Songhua River, the upper reaches of the Songhua River, and the Nenjiang River) as well as their principal tributaries (Figure 1). Specifically, this sampling campaign included 8 samples from the main stem of the Songhua River and 19 samples from its principal tributaries. It is noteworthy that there are several water conservancy projects in the upper reaches of the Songhua River, such as the Baishan Hydropower Station Dam and the Fengman Hydropower Station Dam. These dams exert a water storage effect on the upstream water bodies, which may potentially alter the concentration of pollutants (e.g., NO₃⁻) in the water. Owing to this interference, the water samples collected from the upper reaches lack reference value and were thus excluded from the subsequent analysis of this study. Based on the topographic features and channel characteristics of the Songhua River Basin, the study area was divided into three sections: upper, middle, and lower reaches. The upper reaches, extending from the estuary of the Sancha River to Harbin City, cover a total length of 240 km and are situated in a region with intensive human activities and a high level of industrialization. The middle reaches span from Harbin City to Jiamusi City, with a total length of 432 km. The lower reaches, stretching from Jiamusi City to Tongjiang City, cover 267 km in total and are located in an extensive agricultural cultivation zone. Prior to sampling, the polyethylene bottles were pre-treated by rinsing with 10% hydrochloric acid followed by thorough rinsing with deionized water. All water samples were filtered using 0.22-μm filter membranes. For the analysis of nutrients and isotopes, approximately 1 L of surface water (collected 2–3 m below the river surface) was pumped into pre-cleaned polyethylene bottles. These samples were then transported back to the laboratory immediately and stored in a freezer at −20 °C for subsequent analytical procedures.

2.3. Chemical Analysis

At the sampling site, T, pH, EC and dissolved oxygen (DO) were measured using a multi-parameter portable water quality analyzer (Professional Plus, Yellow Springs, OH, USA. Chloride ion (Cl⁻) concentrations were determined by ion chromatography [15].The chromatographic column is a Dionex ICS-90 from the United States. Water chemistry analyses were conducted at the Agricultural Non-point Source Pollution Laboratory, School of Water Resources and Civil Engineering, Northeast Agricultural University. Concentrations of nitrate nitrogen (N-NO₃⁻), ammonium nitrogen (N-NH₄⁺), total nitrogen (TN), and total phosphorus (TP) were analyzed with a continuous flow analyzer (AA3, SEAL Analytical, Hamburg, Germany). The permanganate index (CODMn) was measured spectrophotometrically using a Leizhi DGB-425 instrument at a single wavelength of 540 nm, with a detection range of 0.5~10.0 mg/L⁻¹. Chemical oxygen demand (COD) was determined following the national standard method HJ 828-2017 [23], which is based on the potassium dichromate method.

Nitrate isotope measurements were performed at the Isotope Testing Laboratory of the Beijing Academy of Agricultural Sciences. The δ¹⁵N and δ¹⁸O values of nitrate in water samples were analyzed using the denitrifier bacteria method coupled with a stable isotope analyzer (Delta V Plus, Thermo Fisher Scientific, Bremen, Germany) [24,25]. In this method, NO₃⁻ is first converted to N₂O by denitrifying bacteria (Pseudomonas aeruginosa, ATCC 13985, Manassas, VA, USA) lacking N₂O reductase activity. The resulting N₂O was then analyzed for δ¹⁵N and δ¹⁸O using a trace gas concentrator (Pre-con, Finnigan, Bremen, Germany) coupled with an isotope ratio mass spectrometer (IRMS, Delta V plus, Finnigan, Bremen, Germany). Stable isotope ratios are expressed as follows [26]:

$$\mathsf{\delta }{\mathrm{R}}_{\mathrm{sample}}\left(‰\right) =\left({\displaystyle \frac{{\mathrm{R}}_{\mathrm{sample}}-{\mathrm{R}}_{\mathrm{standard}}}{{\mathrm{R}}_{\mathrm{standard}}}}\right)\times 1000 (‰)$$

(1)

Herein, R_sample and R_standard denote the ratios of ¹⁵N/¹⁴N and ¹⁸O/¹⁶O for the sample and standard material, respectively. The reference value for δ¹⁵N is standard atmospheric nitrogen, whilst the value for δ¹⁸O is Vienna Standard Mean Ocean Water (VSMOW).

To ensure the accuracy and reliability of the analytical data, this study implemented systematic quality control and quality assurance protocols. Specific measures included the immediate filtration of water samples through membranes followed by dark, low-temperature storage. Field duplicate samples were collected and analyzed for 10% of the total sample set. The mean relative standard deviation for all water chemistry parameters (e.g., NO₃⁻, NH₄⁺, TN) was less than 5%, with method detection limits of 0.001 mg/L, 0.002 mg/L, and 0.02 mg/L, respectively. For the determination of nitrate nitrogen and oxygen isotopes, calibration was performed using international reference materials (USGS32, USGS34, USGS35). Based on duplicate analyses, the analytical precision was better than ±0.5‰ for δ¹⁵N and ±1‰ for δ¹⁸O. All data underwent rigorous cross-verification prior to entry and statistical analysis to confirm compliance with quality control criteria. In N₂O isotopic analysis, water samples were injected via an autosampler using the headspace method, with H₂O and CO₂ removed by two chemical traps (magnesium perchlorate and sodium hydroxide, Merck KGaA, Darmstadt, Germany). After cryogenic trapping and concentration, N₂O was separated at 35 °C using a capillary column (PoraPlot Q, 25 m, 0.32 mm id, 10 mm df, Agilent Technologies, Santa Clara, CA, USA) and subsequently analyzed by isotope ratio mass spectrometry (IRMS). For every batch of 12 samples, one each of the USGS32, USGS34, and USGS35 standards was inserted. The δ¹⁵N values were calibrated against USGS32 (δ¹⁵N = 180.0 ± 1.0‰) and USGS34 (δ¹⁵N = −1.8 ± 0.2‰), while δ¹⁸O values were calibrated using USGS34 (δ¹⁸O = −27.9 ± 0.6‰) and USGS35 (δ¹⁸O = 57.5 ± 0.6‰).

2.4. Bayesian Isotope Mixture Model

To quantitatively assess the contributions of nitrate sources in the Songhua River basin, this study applied a Bayesian stable isotope mixing model (MixSIAR). The model utilized δ¹⁵N and δ¹⁸O as dual tracers, CF, SN, M&S, and AD defined as the potential sources. Their isotopic signatures, referenced from published literature (Table 1), were incorporated into the model as “mean ± standard deviation” to represent the inherent variability of each source. The model was run with a non-informative prior and a continuous mixing structure. After 300,000 iterations, all parameters achieved Gelman-Rubin diagnostic values of less than 1.05, indicating model convergence. To evaluate the uncertainty of the source apportionment results, we calculated an uncertainty index (UI₉₀ = (C₉₅ − C₅)/0.9), which represents the normalized width of the 90% confidence interval for each source’s contribution rate [27,28]. In summary, the MixSIAR model can be represented as [29]:

$${\mathrm{X}}_{\mathrm{i}\mathrm{j}}={\displaystyle \sum _{\mathrm{k}=1}^{\mathrm{k}}{\mathrm{p}}_{\mathrm{k}}\left({\mathrm{s}}_{\mathrm{j}\mathrm{k}}+{\mathrm{c}}_{\mathrm{j}\mathrm{k}}\right)}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{j}}$$

(2)

$${\mathrm{S}}_{\mathrm{j}\mathrm{k}}~\mathrm{N}\left({\mathsf{\mu }}_{\mathrm{j}\mathrm{k}},{\mathsf{\omega }}_{\mathrm{j}\mathrm{k}}^2\right)$$

(3)

$${\mathrm{C}}_{\mathrm{j}\mathrm{k}}~\mathrm{N}\left({\mathsf{\lambda }}_{\mathrm{j}\mathrm{k}},{\mathsf{\tau }}_{\mathrm{j}\mathrm{k}}^2\right)$$

(4)

$${\mathsf{\epsilon }}_{\mathrm{i}\mathrm{j}}~\mathrm{N}\left(0,{\mathsf{\sigma }}_{\mathrm{j}}^2\right)$$

(5)

In the formula: X_ij denotes the isotope value j of mixture i, where i = 1, 2, 3, …, N and j = 1, 2, 3, …, j; S_jk denotes the isotope value j for source k (k = 1, 2, 3, …, k), which follows a normal distribution with mean μ_jk and standard deviation ω_jk; according to MixSIAR estimation, P_k represents the proportional contribution of source k; C_jk denotes the fractionation factor of isotope j from source k, which follows a normal distribution with mean λ_jk and standard deviation τ_jk; ε_ij represents the residual of isotope j in mixture i, obeying a normal distribution with mean zero and standard deviation σ_j.

2.5. Data Analysis

To assess spatial variations in water chemistry and isotopic composition across upstream, midstream, and downstream river sections, a one-way analysis of variance (ANOVA) was employed. This approach is suitable for evaluating the effect of a single categorical independent variable (river section) on multiple continuous dependent variables. The assumptions of independence (ensured by independent sampling), normality (verified by the Shapiro–Wilk test), and homogeneity of variances (confirmed by Levene’s test) were met prior to analysis. The following parameters were tested: pH, T, EC, DO, N-NO₃⁻, N-NH₄⁺, Cl⁻, TN, TP, COD, CODMn, NO₃⁻, NH₄⁺, δ¹⁵N-NO₃⁻, and δ¹⁸O-NO₃⁻. A result was considered statistically significant at p < 0.05. Here, the p-value represents the probability of observing the current data, or more extreme data, if the null hypothesis is true. When p < 0.05, the probability of such a difference is less than 5%. When p < 0.01, the probability is less than 1%. In other words, the smaller the p-value, the lower the probability that the observed difference is due to chance. A p-value below 0.05 suggests a correlation between the two variables, while a p-value below 0.01 provides stronger evidence of this. All analyses were performed using R version 4.2.0, with additional processing and visualization conducted in SPSS 27.0, Origin 2021, and ArcGIS 10.2.

3. Results

3.1. Spatial and Temporal Characteristics of Water Chemistry Parameters

Due to varying environmental conditions within the study area, several water quality parameters (pH, T, EC, DO, COD, CODMn, TP, TN, NO₃⁻, NH₄⁺) exhibited significant spatiotemporal variations during the investigation period (Figure 2). The recorded ranges for pH and T were 7.32~8.59 and 4.9~12.0 °C, respectively. A statistically significant difference (p < 0.01) in temperature was observed between the upstream and downstream sections, correlating with the latitudinal gradient across sampling sites. Electrical conductivity (EC) values spanned from 70.5 to 640.4 μS cm⁻¹, with a mean value of 274.19 ± 136.33 μS cm⁻¹. The EC in the upstream region was significantly higher than in the middle (p < 0.05) and lower reaches (p < 0.01). Relevant studies indicate that anions and cations in river water may lead to an increase in electrical conductivity (EC) [30]. The ranges for determining DO, COD, and CODMn in the samples were 4.96~15.13 mg/L⁻¹, 12.1~31.4 mg/L⁻¹ and 0.99~7.6 mg/L⁻¹ respectively, with average sample concentrations of 9.55 ± 0.37 mg/L⁻¹, 19.14 ± 4.55 mg/L⁻¹ and 3.39 ± 1.56 mg/L⁻¹. At 78% of the river sampling points, DO levels met the national Class I water quality standard (7.5 mg/L⁻¹) stipulated in the Chinese Surface Water Envi-ronmental Quality Standard (National Standard GB3838-2002) [31]. However, 25% and 29% of the sampling points for COD and CODMn, respectively, fell within the third-class water quality category of the national standard, indicating the presence of pollution within the Songhua River basin. Both parameters exhibited broadly similar spatial distribution patterns. Analysis of variance (ANOVA) results indicate that concentrations of COD (p < 0.05; p < 0.01) and CODMn (p < 0.01; p < 0.01) in the upstream section were significantly higher than those in the middle and downstream sections. This indicates that pollution levels in the upper reaches are higher than in the middle and lower reaches, a situation likely attributable to more active human activities in the upstream regions. In contrast, the DO, COD and CODMn levels in the Nenjiang River—the largest tributary of the Songhua River in the upstream region—all meet the national Class I standards [31]. Moreover, other water quality data from the Nenjiang River also demonstrate favourable results, which may be attributed to the forested land use type in its source region, the Greater Khingan Range, and the relatively low level of human activity in the area. The range of TP values was 0.019~0.168 mg/L⁻¹, with an average of 0.065 ± 0.035 mg/L⁻¹. Analysis results indicate no significant correlation exists within the watershed.

The measured concentrations of nitrate (NO₃⁻) and ammonium (NH₄⁺) across the watershed ranged from 5.16~82.89 mg/L⁻¹ and 0.24~4.53 mg/L⁻¹, respectively, with mean values of 14.1 ± 2.77 mg/L⁻¹ and 1.03 ± 0.88 mg/L⁻¹. Of the 27 water samples collected from the main stem and tributaries of the Songhua River basin, only those from the Mudanjiang River—the second-largest tributary of the Songhua—showed nitrate levels that substantially exceeded the national drinking water limit of 10 mg/L⁻¹. Notably, one sample from this tributary recorded an NO₃⁻ concentration as high as 82.89 mg L⁻¹, indicating serious nitrate pollution in this section of the river. NO₃⁻ concentrations in the main stem of the Songhua River increased downstream, with significantly higher concentrations observed in the lower reaches compared to the upper reaches (p < 0.05). However, no discernible correlation was observed among NO₃⁻ concentrations in the tributaries. In contrast, NH₄⁺ concentrations decreased along the river within the watershed, downstream concentrations were significantly higher than upstream (p < 0.05). This spatial distribution pattern may be due to stronger anthropogenic influences in the upstream areas and potential nitrification occurring within the watershed.

3.2. Spatial Distribution Characteristics of NO₃⁻ Stable Isotopes in the Songhua River

NO₃⁻ stable isotopes in the Songhua River basin exhibit distinct spatiotemporal distribution patterns (Figure 3). The concentration of δ¹⁵N-NO₃⁻exhibits a decreasing trend from upstream to downstream, the δ¹⁵N-NO₃⁻ values across the entire river basin ranged from +1.71‰ to +11.99‰ (mean value: +7.05 ± 3.13‰). With the upper, middle, and lower reaches exhibiting respective ranges of +3.56‰~+11.99‰ (mean value: +9.01 ± 2.88‰), +2.1‰~+10.19‰ (mean value: +6.4 ± 2.36‰), and 1.71‰~+10.01‰ (mean value: +4.96 ± 3.32‰), respectively. The ANOVA results indicate significant differences in δ¹⁵N-NO₃⁻ values across the watershed, with the upstream range being significantly higher than both the midstream (p < 0.05) and downstream (p < 0.01) sections. The δ¹⁸O-NO₃⁻ values within the watershed ranged from +1.2‰ to +13.81‰ (mean value: +7.45 ± 2.88‰). The δ¹⁸O-NO₃⁻ values for each river segment segmented within the watershed ranged from +1.2‰ to +13.81‰ (mean value: +5.5 ± 3.54‰), +3.81‰ to +10.33‰ (mean value: +8.45 ± 1.95‰), and +7.25‰ to +9.84‰ (mean value: +8.9 ± 0.89‰). The δ¹⁸O-NO₃⁻ values in the upstream section were significantly lower than those in the middle section (p < 0.01) and downstream section (p < 0.01).

Human activities have a relatively pronounced influence on the δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻ values, resulting in distinct clustering of the dual isotope values of NO₃⁻ [30]. Higher δ¹⁵N-NO₃⁻ values predominantly occur in the upper reaches of the watershed, whereas the middle and lower reaches, which are less affected by human activity, exhibit relatively lower values. There is a strong correlation between δ¹⁵N-NO₃⁻ values and areas of high urbanisation and dense human habitation, as δ¹⁵N-NO₃⁻ is primarily influenced by faecal matter and sewage. In contrast, relatively low δ¹⁸O-NO₃⁻ values are found in relatively densely populated upstream regions, while δ¹⁸O-NO₃⁻ values in the middle and lower reaches of agricultural areas and less densely populated regions are significantly higher. This is because high δ¹⁸O-NO₃⁻ values are primarily influenced by inorganic nitrogen fertilisers and AD.

4. Discussion

4.1. Nitrogen Dynamics Associated with Water Chemistry Parameters

The proportions of N-NO₃⁻ and N-NH₄⁺ within the total nitrogen (TN) pool indicate that these two forms constitute the predominant nitrogen species in both the mainstem and tributaries of the Songhua River. Combined, N-NO₃⁻ and N-NH₄⁺ accounted for 60.38% to 93.01% of TN, with a mean contribution of 80.31% ± 8.6%. Specifically, N-NO₃⁻ represented 40% to 77.75% of TN, averaging 61.9% ± 10.12%. Its proportion in the upstream section was significantly lower than in both the middle (p < 0.01) and downstream (p < 0.01) reaches. Furthermore, the N-NO₃⁻ concentration in the middle reaches was significantly lower than that in the lower reaches (p < 0.05). The proportion of N-NH₄⁺ within total nitrogen (TN) across the entire watershed ranged from 6.37% to 33.01%, averaging at 18.40% ± 7.15%. Interestingly, the spatial distribution pattern of N-NH₄⁺ relative to TN differed from that of N-NO₃⁻. The proportion of N-NH₄⁺ in TN decreased from upstream to downstream, with significantly higher proportions in the upstream section than in the middle and downstream sections (p < 0.01). Concurrently, a strong correlation was observed between the middle and downstream sections (p < 0.01). The range of TN values was between 1.87 and 91.02 mg/L⁻¹, with an average of 4.34 ± 4.55mg/L⁻¹. Correlation analysis revealed no significant association of TN across the entire watershed (p > 0.05). However, we observed that the TN concentration in the Mudanjiang River, the second largest tributary of the Songhua River, was 91.02 mg/L⁻¹, significantly exceeding the TN concentrations in other rivers. This may be an outlier in the sample. After removing the influence of this outlier, we reanalysed the samples. The results revealed significant differences in TN concentrations between the upstream and downstream sections. The spatial distribution pattern of TN concentrations exhibited characteristics similar to those of N-NH₄⁺, showing a gradual decrease from upstream to downstream. TN concentrations in the upstream section were significantly higher than in the middle and downstream sections (p < 0.01). Furthermore, a significant difference was observed between the middle and downstream sections (p < 0.05). As the subsequent discussion will demonstrate, nitrification occurs throughout the entire Songhua River basin. Therefore, the downward trend in TN concentrations along the river does not necessarily indicate denitrification within the basin, but may instead result from reduced nitrogen inputs and downstream dilution effects [32].

The high nitrate (NO₃⁻) values within the watershed are predominantly found in upstream areas with high levels of human activity. However, the nitrate concentrations in the two major tributaries, the Nenjiang and Lalin rivers, are significantly lower than the overall average nitrate concentration in the upstream region. This suggests that these two rivers are less polluted. The upper reaches of the Nenjiang and Lalin rivers originate from the Greater Khingan Range and the Changbai Mountains, respectively, traversing areas with relatively little human activity or industrial zones [33]. This may be a significant reason for the relatively low NO₃⁻ concentrations in both rivers. The subsequent section on the dual-isotope analysis of nitrate (NO₃⁻) further indicates that SN constitutes the primary source of pollution in the Nenjiang and Lalin Rivers, rather than anthropogenic factors such as M&S. In contrast, NO₃⁻ concentrations were generally higher at other sampling points in the upper reaches of these two tributaries. Specifically, the NO₃⁻ concentration in the upper reaches of the Songhua River (Sampling Point 2) was 13.09 mg/L⁻¹, compared to 5.17 mg/L⁻¹ in the Nenjiang River. At the Sancha River confluence (Sample Point 3), where the Songhua and Nenjiang rivers converge, the NO₃⁻ concentration was 10.43 mg/L⁻¹in the main channel. This exceeds the average NO₃⁻ concentration in the main channel, indicating that the upper reaches of the Songhua River and the Nenjiang tributaries exert a significant influence on NO₃⁻ concentrations within the main stem. Following the Sancha River confluence, NO₃⁻ concentrations in the first tributary of the Songhua River, the Bajia River, reached 13.34 mg/L⁻¹. This may be due to historical pollution in the Bajia River. Reports indicate that severe water pollution occurred in the Bajia River around 2017 due to various industrial discharges. Test results showed that the levels of nitrates, biochemical oxygen demand (BOD), fluoride and heavy metals such as cadmium in the river water exceeded national standards, resulting in the river being classified as Class V under national regulations. This caused significant disruption to the daily lives of local residents. High concentrations of nitrate (NO₃⁻) were predominantly found in the main stem of the Songhua River and its tributaries flowing through the Harbin metropolitan area. At sampling point 8, located on the main stem, NO₃⁻ concentrations reached 13.80 mg/L⁻¹, which is the highest level recorded in the upper reaches. This may be attributed to substantial discharges of domestic sewage and industrial effluent from urban areas. The pollution situation in the middle reaches of the Mudanjiang River is more severe, with nitrate concentrations reaching 82.89 mg/L⁻¹. This may be due to substantial sewage discharges from upstream cities and the transport of nutrients from farmland into the river via rainfall-driven surface runoff. The differing spatial characteristics exhibited by NO₃⁻ and NH₄⁺ may be due to increased NH₄⁺ concentrations in upstream waters, likely resulting from wastewater discharges originating from industrial zones and densely populated areas situated upstream. However, population density and industrial activity gradually diminish as one moves downstream along the river. Consequently, the discharge of sewage and faecal matter into the water decreases. Concurrently, localized denitrification processes may occur within the watershed, whereby some NH₄⁺ may be converted into ammonia gas and volatilised into the atmosphere, while the nitrification process converts NH₄⁺ into NO₃⁻ [30].These factors may collectively cause NO₃⁻ concentrations to increase progressively along the main river course, while NH₄⁺ levels gradually decrease.

4.2. NO₃⁻ Source Attribution with Isotopes Under Cl⁻ Concentration Constraints

In recent years, the molar ratio of NO₃⁻/Cl⁻ has successfully been employed as a key indicator for evaluating and identifying sources of NO₃⁻ in river water bodies [34,35,36]. Due to the biological and chemical inertness of chloride ions, it can serve as a stable tracer to trace sources of human activity, such as non-point source pollution from agriculture (NPS), SN or M&S. As it remains unaffected by physical, chemical and biological processes, its concentration only changes when it is mixed with other water sources [37]. Generally speaking, different sources of the nitrate ion exhibit varying nitrate-to-chloride ratios. For instance, M&S sources typically exhibit a low NO₃⁻/Cl⁻ ratio and high chloride concentrations, whereas CF sources are characterised by a high NO₃⁻/Cl⁻ ratio and low chloride concentrations [38]. Additionally, evaporite dissolution in river water can be identified in advance. As the study area is located on the Northeast Plain, where land use is predominantly agricultural with minimal evaporite distribution, Cl⁻ levels in this watershed are primarily influenced by anthropogenic factors [39]. Therefore, the contribution of rock salt (NaCl) dissolution is virtually negligible. The NO₃⁻/Cl⁻ ratio in the Songhua River basin varies considerably, ranging from 0.57 to 2.34. This indicates that nitrate (NO₃⁻) in the Songhua River may originate from multiple sources. Sampling points with a low NO₃⁻/Cl⁻ ratio and a high Cl⁻ concentration are predominantly located within upstream urban clusters. The results of the ANOVA analysis indicate a significant correlation between the upstream and both mid- and downstream sections in the NO₃⁻/Cl⁻ ratio (p < 0.01 and p < 0.05, respectively). The NO₃⁻/Cl⁻ ratio upstream ranges from 0.57 to 1.06, with 90% of NO₃⁻/Cl⁻ ratios being less than 1.0. Therefore, it is unreasonable to attribute this to CF. Concurrently, elevated chloride levels (averaging 15.02 mg/L⁻¹) were observed in the upstream river, indicating that the water is primarily contaminated by M&S. This may also reflect underdeveloped sewage systems in areas characterised by high population density and rapid industrial development, resulting in the direct discharge of urban wastewater into the river. Additionally, the NO₃⁻/Cl⁻ ratio in the upper reaches of the Nenjiang River (1.06) was markedly higher than the upstream average (0.72 ± 0.13) and the Cl⁻ concentration in the Nenjiang River was extremely low. This suggests that CF may contribute to the nitrate source in the Nenjiang River. The ratio of NO₃⁻ to Cl⁻ in the river water downstream remained relatively high (on average 1.14 ± 0.2), while Cl⁻ concentrations decreased sharply from Harbin City upstream to the tributaries and main channel downstream. These lower Cl⁻ concentrations downstream suggest that agricultural fertilisers may represent a significant source of NO₃⁻ input. After removing outliers from the middle reaches, we found that the average NO₃⁻/Cl⁻ ratio (0.99 ± 0.18) and the Cl⁻ concentration in this section were intermediate between the upstream and downstream averages. River water in the middle reaches of the Songhua River receives discharges of faecal matter and sewage from urban clusters upstream, as well as incorporating tributaries from various agricultural zones. This suggests that pollutants in the middle reaches may result from a combination of faecal matter, sewage and agricultural fertilisers.

In general, potential sources of nitrate (NO₃⁻) in rivers include sewage, agricultural runoff and CF, as well as AD [40,41]. NO₃⁻ isotope values in the Songhua River vary widely, indicating that NO₃⁻ pollution in the basin is influenced by multiple sources. According to previous research reports, the range of δ¹⁵N-NO₃⁻ values influenced by soil organic matter is +2.0 to +8.0‰, while the range for M&S is +5.0 to +25‰. The respective ranges for CF and AD are −6 to +6‰ and −10 to +5‰ [14,42]. As illustrated in Figure 3, the δ¹⁵N-NO₃⁻ values in the upper reaches of the Songhua River’s main stem and some tributaries are relatively high. With the exception of the Nenjiang and Lalin rivers, the major tributaries, all δ¹⁵N-NO₃⁻ values in the upper reaches fall within the range typical of M&S. Isotopic values at sampling points 3 and 8 reached +11.99 and +11.53‰, respectively, while values for other main and tributary streams ranged between +9.36 and +10.32‰. Previous studies indicate that elevated δ¹⁵N-NO₃⁻ values may be associated with M&S inputs [14]. This indicates that the primary sources of NO₃⁻ in the upstream river water are faecal matter and sewage discharges. The two major tributaries upstream, the Nen jiang and Lalin Rivers, exhibited distinct patterns. Their δ¹⁵N-NO₃⁻ values were significantly lower than those of other rivers in the upstream region (p < 0.01), at +3.56‰ and +4.02‰, respectively. Consequently, the primary sources of NO₃⁻ in the Nen jiang and Lalin Rivers are unlikely to be M&S. As illustrated in the figure, the δ¹⁵N-NO₃⁻ values for the Nen jiang and Lalin rivers fall within the overlapping region of SN, CF, and AD. This may reflect increased contributions to NO₃⁻ from all three sources. However, the lower δ¹⁵N-NO₃⁻ values should not be attributed to AD, as sampling occurred during the dry season when rainfall contributes minimally to NO₃⁻ levels. The relatively low δ¹⁵N-NO₃⁻ values in the Nen jiang and Lalin Rivers may be attributable to SN and CF being washed into the rivers via soil erosion and surface runoff. Indeed, soil erosion within agricultural catchments and the associated nutrient losses constitute the primary factors driving increased nutrient loading in river water. Furthermore, a series of agricultural management and land use strategies will accelerate soil erosion in rivers and streams.

The δ¹⁵N-NO₃⁻ values in the middle and lower reaches of the watershed are relatively concentrated, as can be seen in the figure. Notably, 76% of the data points in these sections fall within the SN range. This suggests that SN is a significant source of NO₃⁻ in agricultural watersheds during the dry season. In the middle reaches of the watershed, 27% of the values of δ¹⁵N-NO₃⁻ fell within the range attributable to CF and SN, while 45% of the sample points fell within the range attributable to SN and M&S. The Mulan River section lies within the overlapping range of all three sources. This pattern of δ¹⁵N-NO₃⁻ values in the middle reaches suggests that the sources of NO₃⁻ in this section are dominated by a combination of M&S, SN and CF. At the downstream sampling points, δ¹⁵N-NO₃⁻ values were extremely low across tributary sites. The Dulu River registered a value of merely +1.71‰, with characteristic values falling within the fertiliser range. Meanwhile, the Wutong and Anbang Rivers recorded values of +2.30‰ and +2.49‰, respectively. All δ¹⁵N-NO₃⁻ values from downstream tributaries fell within the ranges characteristic of SN and fertiliser application. This indicates that both sources significantly contribute to downstream NO₃⁻ concentrations. However, the δ¹⁵N-NO₃⁻ values in the downstream main channel differ from those in the tributaries, with the main stem tending to fall within the ranges associated with SN, M&S and their overlapping contributions. This suggests that M&S are significant sources of NO₃⁻ in the downstream main stem due to the river passing through Jiamusi City before entering the downstream section. Jiamusi is situated in the eastern part of Heilongjiang Province and is the second-largest city along the main course of the Songhua River within the province. According to the 2024 Statistical Yearbook released by the Heilongjiang Provincial Bureau of Statistics, Jiamusi City had a permanent resident population of 2.079 million, 66.7% of whom resided in urban areas. In 2023, the total wastewater discharge was 9.29 × 10⁷ tonnes, while the annual GDP was ¥90.93 billion, ranking sixth in the province. This suggests that urban influences significantly contribute to NO₃⁻ sources in the downstream main river channel. Concurrently, the aforementioned analysis demonstrates that δ¹⁵N-NO₃⁻ values correlate well with NO₃⁻/Cl⁻ ratios, thereby validating the feasibility of using NO₃⁻/Cl⁻ ratios to investigate NO₃⁻ sources.

The δ¹⁸O-NO₃⁻ values of nitrate from different sources usually have different oxygen isotope ratios [14,42]. It has therefore been employed to identify the sources of NO₃⁻ in river water bodies. The δ¹⁸O-NO₃⁻ values for CF range from +17‰ to +25‰. M&S exhibit the same range as SN, from −5‰ to +15‰. AD values reach a maximum of +25‰ to +75‰ [14]. When the δ¹⁸O-NO₃⁻ values of the Songhua River are compared with the aforementioned range, it can be seen that AD contributes almost nothing to NO₃⁻ concentrations in all river water bodies, which is consistent with previous research [43]. The results of the ANOVA analysis indicate that the δ¹⁸O-NO₃⁻ values at the upstream river sampling points are significantly lower than those at the middle and lower reaches (p < 0.05), with all but one of the upstream sampling points exhibiting values within the range of +1.20 to +7.51‰. According to the aforementioned reference range, the lower δ¹⁸O-NO₃⁻ values suggest that upstream sources of NO₃⁻ are associated with SN, manure and sewage. Of all the sampling points, the δ¹⁸O-NO₃⁻ value in the upper reaches of the Nen Jiang River was the highest recorded at +13.81‰. As AD contributes negligible amounts to NO₃⁻ concentrations, this indicates that fertilisers may account for an increasing proportion of NO₃⁻ sources within the Nen Jiang River. On average, the δ¹⁸O-NO₃⁻ values in the middle and lower reaches are higher than in the upper reaches, indicating that fertilisers contribute more significantly to NO₃⁻ sources in the middle and lower reaches than in the upper reaches. This aligns well with the characteristics of the δ¹⁵N-NO₃⁻ values analysed above and is corroborated by the subsequent MixSIAR analysis.

4.3. NO₃⁻ Transformation Processes in the Songhua River Basin

Nitrification and denitrification are key processes in the nitrogen cycle and play a crucial regulatory role in nitrogen loss and nitrous oxide emissions [44]. The isotopic composition of NO₃⁻ is altered during the conversion process. During nitrification, lighter isotopes are preferentially incorporated into NO₃⁻, often resulting in a decrease in δ¹⁵N-NO₃⁻ values [45]. The δ¹⁸O-NO₃⁻ values decrease by between −10 and +10%, while the denitrification process is accompanied by a reduction in NO₃⁻ concentration and the simultaneous generation of high NO₃⁻ di-isotope values. During the assimilation period, the ratio of δ¹⁵N-NO₃⁻ to δ¹⁸O-NO₃⁻ for residual NO₃⁻ approaches 1:1 [46,47]. During denitrification, the concentrations of residual δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻ increase, with the ratio of δ¹⁵N-NO₃⁻ to δ¹⁸O-NO₃⁻ in the remaining NO₃⁻ approaching 1.5:1 or even 2:1 [48]. Furthermore, lower levels of dissolved oxygen in rivers will hinder nitrification, and values below 4.0 mg/L are also considered detrimental to denitrification [49,50]. Therefore, dual isotope analysis of NO₃⁻ is employed to determine its transformation processes within the watershed [13]. In this study, δ¹⁸O-NO₃⁻ values at the sampling point 90 km upstream, as well as at all sampling points in the middle and lower reaches, fell within the range of −10 to +10‰. This indicates that nitrification processes occur within the Songhua River basin. Furthermore, DO values at all sampling points throughout the river basin exceeded 4.0 mg/L, suggesting an absence of denitrification in the Songhua River (Figure 4). As the NO₃⁻ concentration decreases due to the combined effects of denitrification and dilution, and as δ¹⁵N increases solely through denitrification, the relationship between the two is weakened or masked by dilution, potentially leading to an insignificant correlation. Therefore, we used the NO₃⁻/Cl⁻ ratio, the decline of which primarily reflects consumption by denitrification, to eliminate dilution effects. We investigated NO₃⁻ transformation in the Songhua River by examining the correlation between δ¹⁵N-NO₃⁻ and ln(NO₃⁻/Cl⁻) across different study areas. As can be seen in Figure 4, there is no significant correlation between the upstream and downstream sections of the Songhua River (p > 0.05). According to previous research reports, this ratio exhibits a negative correlation during denitrification. The model’s coefficient of determination (R²) is relatively low and the core explanatory variable is statistically insignificant, which is consistent with theoretical expectations. This indicates a degree of robustness in the results. To conduct a more rigorous assessment of the regulatory mechanisms governing the nitrification process, we analysed the water oxygen isotope composition (δ¹⁸O-H₂O). The δ¹⁸O-H₂O values range from −12.1‰ to −5.8‰. Theory indicates that the δ¹⁸O-NO₃⁻ value of nitrate produced through nitrification can be estimated using the formula: δ¹⁸O-NO₃⁻ = 2/3 δ¹⁸O-H₂O + 1/3 δ¹⁸Oair. This is because NO₃⁻ formed through nitrification derives two oxygen atoms from water and one from dissolved O₂. By comparing the δ¹⁸O-NO₃⁻ values measured in this study with the theoretical nitrification values calculated from the contemporaneous δ¹⁸O-H₂O values in the water bodies (Figure 5), we found that the measured δ¹⁸O-NO₃⁻ values at most sampling points were close to the theoretical values. This suggests that nitrification is the main source of nitrate in this region. However, some samples exhibited lower δ¹⁸O-NO₃⁻ values than the theoretical value. Research suggests that high concentrations of NH₄⁺ increase nitrification rates and utilise a greater proportion of oxygen atoms derived from H₂O. Concurrently, we observed that the exchange of water and oxygen alters the oxygen isotope signature of nitrate in the soil, resulting in lower δ¹⁸O-NO₃⁻ values in the water. Some samples exhibit elevated δ¹⁸O-NO₃⁻ values compared to theoretical values. This may be due to increased δ¹⁸O-H₂O values resulting from soil moisture evaporation or bacterial respiration causing oxygen to exhibit high δ¹⁸O values. In summary, the δ¹⁸O-NO₃⁻ values of river water in this study were significantly lower than those of precipitation and CF, and multiple lines of evidence (the range of δ¹⁸O-NO₃⁻, dissolved oxygen levels, δ¹⁸O-H₂O validation and the relationship between δ¹⁵N-NO₃⁻ and log(NO₃⁻/Cl⁻)) collectively and consistently indicate that nitrification is the dominant process in the Songhua River basin. This aligns with the conclusions of most previous studies, which found that denitrification primarily occurs in soils and groundwater and is uncommon in large river basins [51].

4.4. Allocation of Nitrogen Sources Under the MixSIAR Model

The MixSIAR model, implemented in R, was used to estimate the contribution of nitrate to the Songhua River basin. The ranges of characteristic values used in the model were primarily based on published literature from global and regional sources. These ranges have been extensively validated and applied in environmental isotope geochemistry, enabling the reliable differentiation of macroscopic characteristics of diverse origins (Table 1). The model outputs revealed four potential sources of NO₃⁻ pollution exhibiting distinct spatial characteristics (Figure 6). The mean probability estimate (MPE) showed that the contributions of M&S to NO₃⁻ levels in the upper, middle and lower reaches of the Songhua River were 39%, 31% and 25%, respectively. Correlation analysis showed that M&S’s contribution to NO₃⁻ levels in the upper reaches of the watershed was significantly higher than in the middle and lower reaches (p < 0.01). The highest value in the upper reaches was recorded at sampling point 8, where M&S accounted for 47% of the total. As previously mentioned, the upper reaches traverse urban areas characterised by relatively high levels of industrialisation and dense populations, and receive substantial discharges of M&S. Consequently, the M&S discharged into the upper reaches of the Songhua River constitute a significant source of NO₃⁻ within the basin. Furthermore, within the Nenjiang River—the largest tributary of the Songhua River—M&S account for only 26.3% of NO₃⁻ sources, whereas the proportion attributable to CF is significantly higher than in other upstream rivers at 23%. This may be attributed to the Nenjiang’s origin in the Greater Khingan Range, where the river predominantly traverses agricultural regions rather than major urban areas. This suggests that NO₃⁻ sources in the Nenjiang River are dominated by M&S, CF and SN.

Table 1. The MixSIAR model estimates the specific characteristic value range for nitrate pollution sources.

Sources	δ15N (Mean ± SD‰)	δ18O (Mean ± SD‰)	Reference
AD	−2.5 ± 7.5	50 ± 25	[52]
M&S	15 ± 10	5 ± 10	[42,52]
CF	0 ± 6	21 ± 4	[14,53]
SN	5 ± 3	5 ± 10	[54,55]

Table 1. The MixSIAR model estimates the specific characteristic value range for nitrate pollution sources.

Sources	δ15N (Mean ± SD‰)	δ18O (Mean ± SD‰)	Reference
AD	−2.5 ± 7.5	50 ± 25	[52]
M&S	15 ± 10	5 ± 10	[42,52]
CF	0 ± 6	21 ± 4	[14,53]
SN	5 ± 3	5 ± 10	[54,55]

The contribution rates of SN and CF in the upper, middle and lower reaches of the Songhua River were 46%, 50% and 49%, and 11%, 15% and 20%, respectively. SN did not exhibit significant variation across the entire watershed, with its average proportion remaining relatively consistent in all study areas. This indicates that SN is the primary source of NO₃⁻ in the watershed during the dry season, which is consistent with previous research. However, the contribution rate of CF exhibited significant spatial variation, with downstream contributions being markedly higher than upstream ones (p < 0.01). Among the three main tributaries downstream, CF accounted for over 20% of NO₃⁻ sources. This indicates that CF exerts a significant influence on NO₃⁻ levels in the lower reaches of the Songhua River during the dry season. Downstream, SN and CF contribute approximately 70% of the NO₃⁻ in river water, indicating that agricultural activities are the primary source of NO₃⁻ in this region. The downstream flow region is located in the easternmost part of China, in the Sanjiang Plain of Jiamusi City. This plain is the largest area of agricultural cultivation in Heilongjiang Province, where rice is the main crop. According to the 2024 statistical yearbook published by Heilongjiang Province, Jiamusi City’s cultivated area stood at 195,400 hectares among the main croplands in different regions. Rice cultivation covered 82,400 hectares within this, accounting for 42% of the total cultivated area and ranking first in the province. This figure exceeded the second-ranked region’s rice cultivation area by 48%. Due to the extensive rice cultivation in this agricultural region, CF and SN are washed into rivers via soil erosion and surface runoff during the rainy season’s rice planting period. This leads to an increased contribution from CF sources in downstream rivers, which could potentially exceed 20%. Therefore, future research should quantify NO₃⁻ leaching from surface runoff into river water.

Across the entire watershed, AD contributes least to NO₃⁻ levels, which is consistent with previous studies in other watersheds [56]. AD accounted for 3.6% of NO₃⁻ loads in the upstream region, 3.7% in the midstream region and 6.0% in the downstream region. In the upper reaches of the Nenjiang River, AD contributes 6.2% of NO₃⁻ loads, which is significantly higher than the overall average for this region. This may be because most AD undergoes microbial nitrification before entering the river. However, during heavy rainfall events in forested watersheds, the proportion attributable to AD may increase [53]. In terms of spatial variation, the contribution rate of AD in the downstream region is 1–2% higher than in the middle and upper reaches, indicating a significant atmospheric input to river water in this area.

4.5. Uncertainty Analysis

To determine the uncertainty in the MixSIAR model results, we employ the following formula to calculate the uncertainty index (UI₉₀).

$${UI}_{90}={\displaystyle \frac{(C_{95}-C_5)}{0.9}}$$

(6)

C₉₅ and C₅ denote the upper and lower bounds of the rapid growth segment (90% cumulative probability) within the posterior distribution curve. Throughout the study period, the contribution from AD remained relatively stable with minimal uncertainty. The UI₉₀ for AD was 0.16. The proportional contributions from M&S and CF exhibited moderate uncertainty, with UI₉₀ values of 0.27 and 0.31, respectively. SN demonstrated the greatest uncertainty, with a UI₉₀ value of 0.37. Uncertainty in quantitative nitrate estimates in rivers primarily stems from variations in nitrate sources entering the river and isotopic fractionation during migration. Due to the influence of rainfall processes, runoff processes within the catchment exhibit significant spatial and temporal variability. The interaction between rivers and groundwater also changes, thereby introducing uncertainties regarding the contributions of different nitrate sources. Therefore, the sampling frequency could be increased in future to allow more frequent monitoring and reduce the uncertainty arising from variations in nitrate sources.

5. Conclusions

5.1. Summary

This study uses stable isotopes of nitrate (¹⁵N and ¹⁸O) and the hydrochemical composition of river water to identify nitrate sources and transformation processes within the Songhua River basin. Bayesian isotope mixing models were used to estimate the contribution of multiple NO₃⁻ sources. Higher Cl⁻ concentrations and a higher NO₃⁻/Cl⁻ ratio in the upper reaches of the Songhua River indicate that M&S are the primary sources. Conversely, lower Cl⁻ concentrations and higher NO₃⁻/Cl⁻ ratios in the middle and lower reaches suggest that CF may contribute more significantly to NO₃⁻ levels downstream within the basin. This finding is consistent with results confirmed by dual isotope analysis. Higher dissolved oxygen concentrations indicate conditions that are unfavourable for denitrification in the river water. Furthermore, the ratio of δ¹⁵N-NO₃⁻ to ln(NO₃⁻/Cl⁻) shows no significant correlation, suggesting that denitrification is absent within the watershed. The MixSIAR model results indicate significant spatial variation in the contribution of NO₃⁻ sources. Within the Songhua River basin, SN accounts for half of the nitrate contribution, with no significant differences observed across the study areas. This suggests that SN is the main source of NO₃⁻ within the basin during the dry season. M&S exhibit distinct spatial characteristics compared to CF. Upstream, M&S account for 39% of the contribution, primarily due to concentrated industrial activities and manure-wastewater discharge. In contrast, downstream, the contribution ratio of CF increases significantly to 20%, primarily due to the region’s location within concentrated rice cultivation zones, which substantially influences the proportion of CF contributing to downstream rivers. The middle reaches receive domestic sewage from upstream cities and inflows from various tributaries, with NO₃⁻ contributions primarily originating from three sources: M&S, SN and CF. AD contributes the least to NO₃⁻ levels, with a maximum contribution rate of just 6%. Therefore, controlling the discharge of M&S in the upstream area, alongside improving agricultural management practices in the middle and lower reaches, is key to mitigating nitrate pollution in the Songhua River basin.

5.2. Research Limitations and Outlook

The traceability findings of this study are based on samples collected in October and November 2024. However, given the highly seasonal nature of precipitation, with the annual rainy season primarily concentrated between June and September, this may introduce a limitation. Firstly, the proportionate contribution of nitrate sources may vary seasonally. During the rainy season, nitrate is primarily sourced from agricultural fertiliser leaching driven by surface runoff. Our sampling period (October–November) falls after the rainy season and during a period of agricultural fallow. At this time, the proportion of nitrate contributed by point sources, such as manure and domestic sewage, or by slow-release sources, may increase relatively. Therefore, the source contribution ratios identified in this study more accurately reflect the dominant pollution sources within the catchment during the dry season or normal flow period. Secondly, isotopic signals may be disrupted by mixing and transformation processes. During the rainy season, heavy runoff events can cause the rapid mixing and dilution of nitrates from different sources. Intense biogeochemical processes, such as assimilation and nitrification, may also alter the initial isotopic values. In contrast, the flow during our sampling period was relatively stable, presenting a simpler hydrological scenario. This facilitated the identification of fundamental characteristics of the pollution sources within a ‘low-noise’ environment. However, it was unable to capture the isotopic dynamics under the high-intensity disturbances of the rainy season. In summary, this study provides insight into the composition of nitrate pollution sources in the Songhua River basin during the non-rainy season. However, caution must be exercised when generalising these findings to encompass the entire year, particularly when quantifying the peak contribution of agricultural fertilisers during the rainy season. Future research should incorporate systematic sampling throughout the entire hydrological year, particularly during the rainy season, to elucidate the seasonal transition patterns of pollutant source contributions and to quantify absolute loads more precisely from each source across the annual cycle.