Using Hydrochemistry, Multi-Isotope, and MixSIAR Model to Analyze Nitrate Sources of Groundwater: A Case Study of the Yongning River Banks

Highlights

• A novel hydrochemistry-isotope-MixSIAR framework was developed and applied.
• Hydrochemical-isotopic coupling identifies nitrification as key N transformation.
• MixSIAR reveals Industrial Wastewater vs. Soil Nitrogen source dominance.

Abstract

Groundwater nitrate (NO₃⁻) pollution, caused by anthropogenic activities, poses a global threat to water security. The mixing of multiple nitrate pollution sources and the associated biogeochemical reactions may create a complex chemical background, which renders traditional hydrochemical methods and single δ¹⁵N isotope analysis approaches limited in accurately identifying pollution sources and quantifying their contribution ratios. Accordingly, we adopted an integrated framework incorporating hydrochemistry, isotopes, and the MixSIAR model. Within this framework, results from different components mutually validate each other, helping to achieve more accurate source identification and contribution quantification. Results revealed severe nitrate contamination with striking spatial heterogeneity: concentrations were significantly higher in the eastern region (9.3–1890.7 mg·L⁻¹, Mean: 472.8 mg·L⁻¹) than in the western region (8.5–204.1 mg·L⁻¹, Mean: 52.0 mg·L⁻¹). Hydrochemical and δ¹⁸O-NO₃⁻ evidence identified nitrification as the dominant nitrogen transformation process. Critically, the MixSIAR model quantified drastically different source contributions between the two regions. In the eastern industrial zone, industrial wastewater was the predominant source (61.3%), followed by manure and sewage (18.5%). In contrast, in the western agricultural area, natural and agricultural sources dominated, with soil nitrogen contributing 43.9% and chemical fertilizer 31.7%. The findings pinpoint specific pollution drivers for each region, offering a robust scientific basis for formulating differentiated and effective nitrate pollution control strategies.

1. Introduction

Groundwater, as a critical freshwater resource for maintaining ecological balance and human survival, serves as an important drinking water source in over 400 cities and vast rural areas of China [1]. However, alongside intensive agriculture and industrialization, nitrate (NO₃⁻) has become a global groundwater pollutant due to its high solubility, chemical stability, and strong mobility [2,3,4]. The World Health Organization and the United States Environmental Protection Agency limit NO₃⁻-N in drinking water to 10 mg·L⁻¹ [5]. Excessive NO₃⁻ in drinking water can cause health hazards, including blue baby syndrome and cancers [6]. Given its long-term persistence in oxidizing groundwater environments [7], once pollution exceeds the environmental self-purification threshold, it can disrupt the ecological balance of groundwater and threaten drinking water safety [8,9]. Therefore, effective nitrate remediation is urgent, and accurately tracing its sources is essential for the orderly implementation of groundwater nitrate pollution control and remediation. Existing research indicates that nitrate sources in groundwater are complex and can be categorized into point sources and non-point sources [10]. The point sources primarily stem from livestock and poultry breeding, manure and sewage (M&S), and industrial wastewater (IW), whereas the non-point sources mainly include soil nitrogen (SN), chemical fertilizer (CF) and atmospheric deposition (AD) [2,11]. Aiming at the nitrate pollution issue caused by the synergistic impacts of multiple pollution sources, nitrate and ammonium released from different sources undergo nitrification and denitrification along distinct seepage paths. These biogeochemical reactions induce significant nitrogen and oxygen isotope fractionation during nitrogen migration and transformation, and the pollutants subsequently converge and undergo mixing processes. The mixed polluted groundwater partially loses its initial isotopic characteristics, posing fundamental challenges to qualitative identification and quantitative calculation [12,13]. Therefore, we adopted an integrated approach combining hydrochemical analysis, multi-isotope tracing, and a Bayesian mixing model to identify nitrate sources and quantify their respective contributions.

Traditional hydrochemical methods, relying on ion correlations, have limited discriminatory power in multi-source mixed pollution scenarios [14]. Researchers found that distinct δ¹⁵N-NO₃⁻ isotopic signatures exist among different nitrate sources, which has allowed the single nitrogen isotope to be used to distinguish nitrate derived from soil nitrogen, domestic sewage, and chemical fertilizers. However, overlaps in their isotopic signatures lead to inaccuracies in nitrate source apportionment [15,16,17]. To overcome these limitations, the dual-isotope technique of nitrate (δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻) has emerged. The incorporation of δ¹⁸O-NO₃⁻ significantly improves the differentiation of isotopic signatures among nitrate from different sources. By coupling fingerprint information from both nitrogen and oxygen dimensions, it significantly enhances source discrimination capabilities [10,18]. Further integration with water stable isotopes (δ²H-H₂O, δ¹⁸O-H₂O) and hydrochemical indicators can effectively reveal nitrogen transformation processes and constrain source end-members, forming the mainstream qualitative tracing framework [19,20,21,22,23]. However, translating qualitative identification into precise quantification of multi-source contribution proportions remains a methodological bottleneck in nitrate pollution research. Early quantitative models (e.g., IsoSource), primarily based on mass balance, struggled to handle complexities such as the wide isotopic distribution caused by isotope fractionation (e.g., nitrification and denitrification) [24,25,26]. Emerging Bayesian mixing models (MixSIAR), by incorporating a probabilistic framework, can simultaneously account for source signature variability, fractionation effects, and mixing uncertainty, providing a powerful tool for quantitative source apportionment in complex systems. Therefore, integrating hydrochemical analysis, multi-isotope tracing, and Bayesian mixing models to form a comprehensive “identification-transformation-quantification” analytical framework represents the emerging methodological direction in current nitrate source tracing research.

This study focused on the Yongning River riparian area in the Naxi District, Luzhou City, located in the Sichuan Basin of southwestern China. The area features both concentrated industrial zones and dispersed agricultural land. Surveys indicate severe nitrate contamination in groundwater, but clear understanding of dominant pollution sources and their contribution proportions is lacking. Accordingly, this study aims to: (1) employ a combination of hydrochemical and multi-isotope (δ¹⁵N-NO₃⁻, δ¹⁸O-NO₃⁻, δ²H-H₂O, δ¹⁸O-H₂O) approaches to clarify nitrate pollution sources and key nitrogen transformation processes in the study area; (2) apply the MixSIAR model to quantitatively estimate the contribution proportions of different sources (soil nitrogen, chemical fertilizers, manure and sewage, industrial wastewater, and atmospheric deposition) to groundwater nitrate. The research findings aim to provide precise scientific evidence for groundwater nitrate pollution prevention and control in the region.

2. Materials and Methods

2.1. Study Area

The study area is located in Naxi District, Luzhou City, on the southern edge of the Sichuan Basin (Figure 1). It experiences a subtropical humid monsoon climate, with an average annual temperature of approximately 12 °C and abundant precipitation (average annual rainfall of 909.4 mm), concentrated primarily in summer. The terrain slopes from higher elevations in the south to lower elevations in the north, with elevations ranging from 190 m to 305 m. The phreatic aquifer in the study area is dominated by the Jurassic Upper Shaximiao Formation, which is also widely distributed outside the study area. With an average thickness of approximately 40 m, the aquifer is composed of mudstone intercalated with sandstone, and the mudstones are generally enriched in calcium-bearing carbonates (calcite), while the groundwater depth ranges from 5 to 15 m. Groundwater occurrence types mainly consist of bedrock fissure water and porous water in unconsolidated sediments. The study area is divided into eastern and western parts by the Yongning River, and the Yangtze River flows through the upper part of the western section and the left side of the eastern section. Groundwater in the western part flows from southwest to northeast and discharges into the Yongning River and the Yangtze River, whereas groundwater in the eastern part flows from northeast to southwest and also drains into the two rivers. Industrial and chemical parks are concentrated in the eastern region, while the western part is predominantly characterized by agriculture and scattered residential areas, and Industrial wastewater infiltrates into the aquifer indirectly through the ground surface. Groundwater in the aquifers on both sides discharges into the Yongning River.

2.2. Sampling and Analyses

In this study, field sampling was conducted in June 2024, including the collection of groundwater samples and samples from potential nitrate sources. The 15 shallow groundwater samples were collected from agricultural irrigation wells and groundwater monitoring wells across the study area. These wells were selected to cover the main hydrogeological units and major land-use types, agricultural land, construction land, and wasteland, rendering them representative. Groundwater level variations induced by seasonal changes were relatively small. To ensure sample representativeness, each well was purged for a minimum of 15 min prior to sample collection. Samples for cations, anions and isotopes were filtered through a 0.22 μm membrane filter on the sampling day and then collected in 500 mL high-density polyethylene bottles rinsed three times with distilled water. Potential nitrate source samples included SN samples (n = 7), M&S samples (n = 2), AD samples (n = 4), and IW samples (n = 3). SN samples were collected from surface soil at a depth of 0–30 cm. At each sampling site, the five-point sampling method was adopted: samples collected from five different positions at the same site were mixed uniformly to ensure representativeness, and then the mixed samples were placed in sealed bags for preservation. M&S samples were collected from human excrement-related sources, specifically septic tanks and domestic sewage outlets in the study area. After collection, all M&S samples were filtered through a 0.22 μm membrane filter and then transferred into 50 mL amber glass bottles for storage to avoid light interference. AD samples were collected by atmospheric deposition sampling, using 47 mm quartz fiber filters to capture atmospheric deposition substances. After the sampling was completed, the quartz filter was folded with the collection side facing inward to prevent the collected substances from falling off or being contaminated, and then placed into a sealed bag. IW samples were collected from industrial wastewater storage tanks in industrial parks within the study area. Similar to M&S samples, IW samples were filtered through a 0.22 μm membrane filter immediately after collection, and then stored in 50 mL amber glass bottles. All collected samples were stored in a portable refrigerator at −4 °C during transportation and storage to maintain sample stability and avoid changes in δ¹⁵N and δ¹⁸O values. All these samples were used for the analysis of δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻ isotopic compositions. The specific distribution of sampling points is shown in Figure 1.

In situ measurements of dissolved oxygen (DO), pH, electrical conductivity (EC), and oxidation-reduction potential (ORP) were measured using a calibrated Portable Multi-Parameter Meter (HQ40d, Hach, Loveland, Colorado, USA). The concentrations of HCO₃⁻ and CO₃²⁻ were immediately measured via hydrochloric acid titration according to standard protocols. The major cations (Na⁺, K⁺, Ca²⁺, Mg²⁺) and major anions (Cl⁻, SO₄²⁻, NO₃⁻) in the samples were determined using Ion Chromatograph (930, Metrohm, Herisau, Switzerland). We first used standard solutions to establish standard curves for anions and cations at concentrations of 10, 30, 50, 70, and 100 mg·L⁻¹, and then, 50 mL of undiluted raw sample was injected into the ion chromatograph (IC) to determine the concentrations of anions and cations successively. Each sample was measured in triplicate, and the average value was adopted. For samples with concentrations exceeding the upper limit of the standard curves, they were diluted by 10 or 50 times to obtain accurate concentrations. The analytical precision for anions and cations was controlled within ±0.2 mg·L⁻¹.

The δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻ values were obtained by denitrifying bacteria method using Isotope Ratio Mass Spectrometer (MAT 253, Thermo, Bremen, Germany). In this method, denitrifying bacteria lacking N₂O reductase convert NO₃⁻ to N₂O; the resulting N₂O is automatically purified by a trace gas analysis system and analyzed using an isotope ratio mass spectrometer under a high-purity He atmosphere. The δ¹⁵N and δ¹⁸O of NO₃⁻ are compared with the atmospheric N₂ and Vienna Standard Mean Ocean water (VSMOW). The analytical uncertainty of δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻ values was ±0.5 ‰ [27]. The δ¹⁸O-H₂O and δ²H-H₂O were determined using Isotope Ratio Mass Spectrometer (MAT 253, Thermo, Bremen, Germany). The reference standards include VSMOW (δD: 0‰, δ¹⁸O: 0‰), AP-31 (δD: −149.7‰, δ¹⁸O: −18.8‰), and PW2015 (δD: −34.5‰, δ¹⁸O: −5.3‰, internal standard). High-purity He (>99.999%) was used as carrier gas at 80 mL/min; high-purity H₂ and CO served as reference gases. The pyrolysis furnace was set at 1400 °C, and column temperature at 85 °C. For IRMS: emission current 1.5 mA, acceleration voltage 3 kV, vacuum 1.4 × 10⁻⁶ mBar, injection volume 0.1 μL. Sample values were calibrated using linear regression of the first and last three standards. Intermediate standards were used to monitor instrument performance. The reported as values relative to Vienna Standard Mean Ocean water (VSMOW). The analytical precision was ±0.2‰ for δ¹⁸O-H₂O and ±1‰ for δ²H-H₂O.

The results of δ¹⁵N-NO₃⁻, δ¹⁸O-NO₃⁻, δ²H-H₂O, δ¹⁸O-H₂O values were calculated as follows [28]:

$$\delta (\mathrm{‰})=\left({\displaystyle \frac{R_{\mathrm{sample}}}{R_{\mathrm{standard}}}}-1\right)\times 1000$$

(1)

where R_sample and R_standard are the isotope ratios of the samples and standards (¹⁵N/¹⁴N, ¹⁸O/¹⁶O, ²H/¹H) respectively.

2.3. Bayesian Mixing Model

To quantitatively estimate the proportional contributions of various pollution sources, the MixSIAR model, which is based on Bayesian theory (operating in the R language environment), was employed [29]. The model can be represented as follows (Equation (2)):

$$X_{ij}={\sum }_{k=1}^k{P}_k(S_{jk}+C_{jk})+{\epsilon }_{ij}$$

(2)

$$S_{jk}~N({\mu }_{jk},{\omega }_{jk}^2)$$

$$C_{jk}~N({\lambda }_{jk},{\tau }_{jk}^2)$$

$${\epsilon }_{ij}~N(0,{\sigma }_j^2)$$

where X_ij is the isotope signature j of groundwater sample i, P_k is the contribution rate of pollution source k; S_jk is the isotope j of source k; C_jk is the fractionation coefficient of isotope j of pollution source k; ε_ij is a residual that represents other unquantified changes between individual samples, and is distributed with mean 0 and standard deviation of σ_j.

Inputs to the MixSIAR model comprised three elements. First, the mixture data consisted of the measured δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻ values from groundwater samples. Second, the source end-members were defined by the mean and standard deviation of these isotopes for five potential sources, among which AD, M&S, IW, and SN were derived from direct source sampling, and chemical fertilizers were derived from a review of local literature. Third, the fractionation factors were assigned. The hydrochemical and isotopic assessment (Section 3.2.1) indicated a strongly oxidizing environment with minimal denitrification. Accordingly, the fractionation coefficients for both nitrogen and oxygen isotopes were set to zero.

Non-informative priors were used in the model. The Markov Chain Monte Carlo (MCMC) run length was set to 200,000 iterations, with the first 20,000 iterations discarded as a burn-in period to ensure chain convergence. The results were presented as the posterior distribution means and 95% confidence intervals for each source’s contribution.

2.4. Statistical Analyses

Descriptive statistics and Pearson correlation analysis were performed using Origin 2025 software.

3. Results and Discussion

3.1. Hydrogeochemical Characteristics and Spatial Differentiation of Nitrate

The mean, range, median, and standard deviation (SD) of the physicochemical variables (DO, ORP, pH, EC) and major ions (Na⁺, K⁺, Ca²⁺, Mg²⁺, Cl⁻, SO₄²⁻, NO₃⁻, HCO₃⁻) of the groundwater samples are summarized in

Table 1. The statistical characteristics of the hydrochemical compositions of the study area.

Parameters		Eastern Region (n = 6)				Western Region (n = 9)
		Max	Min	Mean	SD	Max	Min	Mean	SD
pH	-	7.9	6.4	7.06	0.61	7.44	5.7	6.64	0.66
EC	μS·cm−1	3560	202.1	1329.18	1181.78	657	99.3	487.37	195.86
ORP	mV	194.9	12.1	134.03	65.56	240.3	58.7	148.78	53.62
DO	mg·L−1	7.18	2.68	5.04	1.82	7.04	1.57	4.52	2.22
Na+	mg·L−1	674	15.46	142.67	260.58	51.95	12.15	24.05	12.87
K+	mg·L−1	21.44	1.52	10.53	7.93	15.34	0.41	5.02	4.82
Ca2+	mg·L−1	208.43	20.2	103.47	66.52	120	16.6	56.19	33.53
Mg2+	mg·L−1	52.5	4.79	25.04	15.62	22.9	3.45	15.51	5.64
Cl−	mg·L−1	173.85	15.69	54.4	61.37	80.47	8.22	26.22	22.93
SO42−	mg·L−1	252.06	4.63	92.51	91.83	131.38	4.02	52.9	39.19
NO3−	mg·L−1	1890.74	9.25	472.86	746.28	204.09	8.5	52.02	60.72

. According to these results, pH across both regions exhibited weakly acidic to near-neutral conditions (pH 5.70–7.90), with notable spatial heterogeneity. The eastern region showed pH values of 6.40–7.90 (mean 7.06), while the western region exhibited slightly more acidic conditions (5.70–7.44, mean 6.64). Both DO and ORP demonstrated significant variability, with eastern region values ranging from 2.68 to 7.18 mg·L⁻¹ (mean 5.04 mg·L⁻¹) for DO and 12.10–194.00 mV (mean 134.03 mV) for ORP, compared to western region ranges of 1.57–7.04 mg·L⁻¹ (mean 4.52 mg·L⁻¹) for DO and 58.70–240.30 mV (mean 148.78 mV) for ORP collectively indicating strongly oxidizing conditions throughout the study area, with significantly higher EC values in the eastern region compared to the west.

Groundwater hydrogeochemistry in the study area exhibits distinct spatial patterns between the western and eastern regions. In the western region, HCO₃⁻ dominates as the primary anion, with the abundance order HCO₃⁻ > SO₄²⁻ > NO₃⁻ > Cl⁻, while Ca²⁺ is the predominant cation, following the sequence Ca²⁺ > Na⁺ > Mg²⁺ > K⁺. In contrast, the eastern region also features HCO₃⁻ as the main anion, but the anion abundance shifts to NO₃⁻ > HCO₃⁻ > SO₄²⁻ > Cl⁻, with cations ordered as Na⁺ > Ca²⁺ > Mg²⁺ > K⁺. Notably, NO₃⁻ exhibits the highest average concentration among anions in the eastern groundwater. This phenomenon may be attributed to the presence of nitrate production enterprises. High concentrations of NO₃⁻ exist in industrial wastewater from chemical production. Leakage of such wastewater gradually infiltrates into the underground aquifer through the surface, resulting in a significant increase in NO₃⁻ levels in wells near the production base. These findings underscore the combined influence of anthropogenic activities and natural geological processes on groundwater composition in the study area.

The hydrochemical characteristics of groundwater in the study area were analyzed using Piper trilinear diagrams (Figure 2). The majority of groundwater samples plot in the HCO₃-Ca·Mg zone of the Piper diagram, suggesting water-rock interactions dominated by carbonate mineral (particularly calcite) dissolution. A minor proportion of samples display Cl·SO₄-Ca·Mg type, potentially influenced by anthropogenic activities such as agricultural fertilization with ammonium sulfate compounds and the discharge of domestic wastewater [30]. Notably, one single sample from the Chemical Industry Park in the eastern region is HCO₃-Na type, with this anomalous composition likely attributable to leakage of sodium-containing raw materials or products during industrial chemical production process.

The study area exhibits pronounced spatial heterogeneity in groundwater NO₃⁻ contamination, with significantly higher concentrations observed in the eastern region (9.25–1890.74 mg·L⁻¹) compared to the western region (8.50–204.09 mg·L⁻¹). Notably, 46.7% of sampling wells surpassed the World Health Organization drinking water standards (>50 mg·L⁻¹), comprising 4 wells in the western region and 3 wells in the eastern region (Figure 3). The spatial distribution of NO₃⁻ pollution exhibits distinct clustering patterns, with peak concentrations in the eastern region primarily sourced from groundwater wells within the chemical industrial park. In contrast, the high nitrate concentrations in the western region are primarily located in agricultural activity areas. These findings indicate that there is a phenomenon of NO₃⁻ pollution within the study area, with particularly severe pollution in the eastern region, where industrial activities appear to be the dominant contributing factor compared to agricultural influences in the western region.

The spatial distribution of NO₃⁻ contamination in the study area’s groundwater exhibits strong correlations with land use patterns. The chemical industry park in the eastern region demonstrates the most severe NO₃⁻ pollution, with two sampling wells showing exceptionally high concentrations ranging from 724.21 to 1890.74 mg·L⁻¹. The western region presents a different contamination profile, where farmland groundwater consistently contains higher NO₃⁻ concentrations compared to residential areas, reflecting the dominant influence of agricultural activities. These findings demonstrate that industrial operations constitute the primary nitrate source in the eastern region, while agricultural practices drive contamination in the western region.

The hydrogeochemical correlation analysis reveals distinct nitrate contamination patterns. In the eastern region (Figure 4a), the strong positive correlations between Ca²⁺ and Mg²⁺ (R = 0.88) and between Ca²⁺ and HCO₃⁻ (R = 0.95) reveal the contribution of the long-term water-rock interactions to groundwater chemistry. The particularly strong association between NO₃⁻ and Ca²⁺, Mg²⁺ implies substantial influence from chemical fertilizer containing these cations [31]. Conversely, in the western region (Figure 4b), the remarkably high correlation between NO₃⁻ and Cl⁻ (R = 0.82) points to potential manure and sewage contributions to groundwater NO₃⁻ levels [32], as these waste sources typically contain both ions in comparable ratios. However, correlation alone is insufficient to verify pollution sources, which will be further analyzed in subsequent sections using isotopic evidence. The significant negative correlation between NO₃⁻ and pH (R = −0.70, R = −0.51) suggests an underlying biogeochemical mechanism. Acidic conditions reduce soil’s adsorption capacity, thereby enhancing nitrate mobility [33]. These differential correlation patterns highlight region-specific nitrate pollution mechanisms within the study area.

3.2. Nitrogen Transformation Processes and Identification of Nitrate Sources

3.2.1. Nitrogen Transformation Processes

Stable isotopes (δ²H and δ¹⁸O) serve as valuable tracers for identifying groundwater recharge sources. Groundwater samples in the study area exhibit a range of stable isotope values: δ²H from −58.0‰ to −22.9‰ and δ¹⁸O from −10.71‰ to −3.83‰, reflecting mixing of meteoric recharge and localized evaporation effects. Figure 5 shows the relationship between δ¹⁸O-H₂O and δ²H-H₂O in groundwater. The local meteoric water line (LMWL: δ²H-H₂O = 7.28 δ¹⁸O-H₂O + 6.56 R² = 0.95) falls near the Global Meteoric Water Line (GMWL: δ²H-H₂O = 8 δ¹⁸O-H₂O + 10), with deviations attributable to regional climatic influences on isotopic fractionation, including temperature, humidity, and precipitation variations. This correlation confirms atmospheric precipitation as the predominant recharge source for groundwater [34], providing critical information for subsequent NO₃⁻ source identification through isotopic signatures.

The dual-isotopic signatures of NO₃⁻ (δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻) serve as valuable tracers for NO₃⁻ source identification, yet we need careful consideration of biogeochemical processes that may modify these values. In shallow aquifer systems, microbial-mediated N transformation, particularly nitrification and denitrification, can substantially alter the original isotopic signatures of NO₃⁻ sources [12]. Therefore, a thorough understanding of nitrogen cycling, particularly nitrification and denitrification, is essential for accurate interpretation of NO₃⁻ isotopes [35].

Denitrification is a microbially mediated process that reduces NO₃⁻ to nitrogen gases (N₂ or N₂O), concurrently decreasing NO₃⁻ concentration while enriching the residual nitrate in δ¹⁵N [36]. This biogeochemical process can be identified through combined hydrochemical and isotopic analyses, where distinct negative linear correlations of both δ¹⁵N-NO₃⁻ and δ¹⁸O-NO₃⁻ with ln(NO₃⁻) typically indicate significant denitrification in groundwater systems. As shown in Figure 6a, while both regions exhibit negative δ¹⁵N-NO₃⁻ versus ln(NO₃⁻) relationships, these correlations are statistically weak. As can be seen from Figure 6b, the δ¹⁸O-NO₃⁻ demonstrates divergent regional trends, showing a weak positive correlation with ln(NO₃⁻) in the eastern region but a weak negative correlation in the western region, consistent with previous observations [37]. Notably, active denitrification typically produces characteristic δ¹⁵N-NO₃⁻ to δ¹⁸O-NO₃⁻ enrichment ratios ranging from 1:1.3 to 1:2.1 [38,39]. As can be seen from Figure 7a, three points in the western region plot within the denitrification zone, suggesting possible localized activity. However, comprehensive assessment requires supplementary hydrochemical indicators, particularly DO and ORP, which serve as robust denitrification proxies [31,40]. The measured DO concentrations (eastern mean 5.04 mg·L⁻¹; western mean 4.52 mg·L⁻¹) consistently exceed the 2 mg·L⁻¹ threshold for significant denitrification [41,42,43], while ORP values (12.10–240.3 mV) confirm prevailing oxidative conditions fundamentally incompatible with substantial denitrification (Figure 7b). Consequently, denitrification is not the dominant nitrogen transformation process in the study area, warranting the approximation of isotopic fractionation coefficients as zero in subsequent MixSIAR model [44,45,46].

Nitrification, the microbial-mediated oxidation of NH₄⁺ to NO₃⁻ under aerobic conditions, proceeds via two sequential steps: nitrosation (NH₄⁺→NO₂⁻) and nitrification (NO₂⁻→NO₃⁻) [47]. In the study area’s subtropical humid monsoon climate, optimal temperatures significantly enhance nitrification rates through multiple pathways. During this process, the oxygen composition of NO₃⁻ derives from two sources: approximately two-thirds originates from water molecules (H₂O) and one-third from atmospheric O₂ [48], and the empirical value of atmospheric O₂ is +23.5‰ [49]. This stoichiometric ratio represents a simplified model, as natural conditions introduce complexities including abiotic oxygen exchange and microbial respiration effects on δ¹⁸O values. These factors can alter oxygen incorporation, potentially increasing the proportion of aqueous-derived oxygen atoms in NO₃⁻ beyond the theoretical two-thirds. Nevertheless, the δ¹⁸O-NO₃⁻ values can be determined using the following theoretical expression (Equation (3)):

$${\delta }^{18}O-{{\rm N}O}_3^{-}={\displaystyle \frac{2}{3}}{\delta }^{18}O-H_{2}O+{\displaystyle \frac{1}{3}}{\delta }^{18}O-O_2$$

(3)

The values of δ¹⁸O-H₂O in groundwater range from −10.706‰ to −3.825‰, making the δ¹⁸O-NO₃⁻ theoretical values range between +0.626‰ and +5.283‰. However, the measured δ¹⁸O values may be up to 5‰ higher than the calculated theoretical values [50]. As shown in Figure 7c, most of the δ¹⁸O-NO₃⁻ values in the western region are within the nitrification range and far lower than the atmospheric oxygen line of NO₃⁻, while the δ¹⁸O-NO₃⁻ values in the eastern region are higher than the theoretical line, which can be attributed to the high input δ¹⁸O-NO₃⁻ values, microbial respiration, and evaporation can also lead to the enrichment of δ¹⁸O in groundwater [48].

3.2.2. Source Identification Based on Dual Isotopes and Ionic Ratios

Cl⁻ is stable in groundwater and can often be used as a stable tracer to identify different sources, such as Cl⁻ containing minerals from natural sources, M&S, CF and IW from anthropogenic sources [51]. The relationship between NO₃⁻/Cl⁻ ratio and Cl⁻ concentration shown in Figure 7d can be used to distinguish NO₃⁻ from M&S, SN and CF. The higher Cl⁻ and lower NO₃⁻/Cl⁻ values in the eastern region indicate that NO₃⁻ mainly comes from M&S [41,52], and NO₃⁻ in samples with NO₃⁻/Cl⁻ values far exceeding 1 may come from product leakage in the chemical industry park. In both regions, NO₃⁻/Cl⁻ ratios ranged from 1 to 10, indicating mixed source pollution, while the NO₃⁻/Cl⁻ ratios are less than 1 with low Cl⁻ indicates that NO₃⁻ mainly comes from SN [21]. Samples from both regions that are not within a specific range may be the result of mixed effects from multiple sources [53,54].

In addition, δ¹⁵N and δ¹⁸O values provide direct evidence for the identification of NO₃⁻ sources in groundwater. δ¹⁵N values in the study area range from 2.1‰ to 20.1‰, and δ¹⁸O values range from 1.1‰ to 14.8‰. According to previous studies and the measured data in the study area [50,51,55], the values of δ¹⁵N and δ¹⁸O from the main NO₃⁻ sources in the study area are shown in

Table 2. Dual isotopic signatures of nitrate sources.

Region	Source	δ15N (‰)		δ18O (‰)
		Mean	SD	Mean	SD
Eastern region	Soil Nitrogen	2.54	3.02	4.65	4.26
	Chemical Fertilizer	−0.37	0.22	2.92	1.69
	Manure and Sewage	16.3	4.28	5.01	3.07
	Industrial Wastewater	7.26	1.37	17.64	1.75
Western region	Atmospheric Deposition	−4.99	1.29	48.45	11.23
	Soil Nitrogen	3.54	2.51	4.44	4.85
	Chemical Fertilizer	−0.37	0.22	2.92	1.69
	Manure and Sewage	8.71	2.30	19.37	0.20

.

The distinct δ¹⁵N and δ¹⁸O ranges of different sources (Table 2) allow for qualitative identification of major nitrate sources using a dual-isotope plot. As can be seen from Figure 7a, the sample points mainly fall in the areas of SN, M&S, and CF, indicating their primary contribution to NO₃⁻ sources. Although the dual-isotopes show that some samples are near the M&S terminal, the final mixed signal δ¹⁸O-NO₃⁻ may be the result of the physical and chemical interaction between IW and local nitrification products, which makes the isotope scatters fall within the range of M&S. Few samples fall in the AD range, indicating its limited influence on the pollution.

3.3. Quantitative Analysis of Nitrate Sources Based on MixSIAR Model

Building upon the aforementioned qualitative insights, the MixSIAR model was applied to perform quantitative source apportionment separately for the eastern and western regions. The MixSIAR results revealed distinctly different pollution source structures between the two regions (Figure 8).

SN accounted for the highest proportion of nitrate in groundwater in the western region, with a contribution rate of 43.9%, followed by CF (31.7%), M&S (20.2%), and atmospheric deposition (4.1%). Notably, SN accounts for the largest proportion in agricultural areas. A possible reason is that the isotopic signatures of SN and CF overlap to some extent, making it impossible to accurately distinguish them when calculating their contribution rates. IW accounted for the highest proportion of nitrate in the groundwater in the eastern region, and its contribution rate is 61.3%, followed by M&S (18.5%), SN (10.9%) and CF (9.3%). The nitrate in the groundwater of the eastern region is mainly affected by anthropogenic activities, whereas in the western region, it is primarily controlled by agricultural practices. In the eastern region, NO₃⁻ chemical production enterprises are mainly distributed, which leads to serious NO₃⁻ pollution of groundwater. By integrating Cl⁻ data and spatial distribution of NO₃⁻, the MixSIAR model corrected the misjudgment caused by isotope overlap, and the findings ultimately point to industrial wastewater leakage as the principal source. This quantitative result aligns with the agricultural land use in the western region, where soil nitrogen and chemical fertilizer collectively constitute the dominant nitrate sources.

4. Conclusions

This study established an integrated framework coupling in situ hydrogeochemistry, multi-isotope tracers (δ¹⁵N, δ¹⁸O, δ²H), and MixSIAR model to identify the NO₃⁻ sources and quantify groundwater NO₃⁻ source apportionment. Groundwater NO₃⁻ pollution is severe and exhibits significant spatial heterogeneity, with extremely high concentrations in the eastern region primarily driven by industrial point sources (chemical industrial parks), while moderate to low levels of pollution in the western region are mainly associated with agricultural non-point sources (soil nitrogen and fertilizers). Hydrochemical and isotopic evidence indicates that nitrification is the dominant nitrogen transformation process in the study area, where the oxidizing environment suppresses denitrification and creates favorable conditions for stable isotope-based source tracing. Quantitative apportionment via the MixSIAR model reveals that soil nitrogen (43.9%) and chemical fertilizers (31.7%) are the primary contributors in the western region, whereas industrial wastewater (61.3%) overwhelmingly dominates in the eastern region.

Based on the above findings, the following management recommendations are proposed: In the eastern industrial zone, priority should be given to strengthening leak-proof supervision in industrial parks and enhancing wastewater treatment. In the western agricultural area, it is essential to promote soil testing and formulated fertilization, optimize irrigation management, and implement non-point source pollution prevention and control measures.

This study has limitations that suggest valuable future research directions. First, only one sampling campaign was conducted in this study, which fails to characterize the annual variation in nitrate. Future studies will increase sampling frequency and spatial coverage to better characterize the temporal variations and detailed spatial distributions of nitrate sources. Second, insufficient collection of samples from potential nitrate sources in the study area resulted in considerable overlap in the isotopic signatures of different nitrate end-members (e.g., soil nitrogen and chemical fertilizers), thereby introducing uncertainties into the estimated contribution proportions of each source. Finally, regarding the fractionation factors associated with nitrogen cycling and transformation required for the MixSIAR model, this study only adopted qualitative analysis. Therefore, future research should focus on the following aspects: first, long-term and large-scale groundwater and end-member sampling should be conducted in the study area to accurately characterize the spatiotemporal variations in nitrate pollution and precisely distinguish the isotopic signatures of different end-members; second, multiple isotopic tracers (e.g., B, Sr) should be adopted to improve source identification accuracy with consideration of isotope fractionation induced by biogeochemical nitrogen cycling; finally, accurate isotope fractionation factors should be determined via laboratory experiments to ensure the reliability of the calculated contribution proportions of different nitrate sources.