Yang River is located in the upper Yongding River, which belongs to the Haihe River system, one of the seven largest rivers of China. Its headwater includes the Dongyang River from Inner Mongolia Autonomous Region and the Nanyang River from Shanxi Province, north China. These two headwater rivers merge in Hebei Province and finally drain into the Guanting Reservoir, an alternate drinking water source for Beijing Municipality, the capital of China (Dai et al., 2020).

The studied river is mainly located in Zhangjiakou city, Hebei Province, which is abundant in mineral resources and consequently has many industrial enterprises. It drains an area of 1.5 × 10⁴ km and has a length of about 262 km. The annual average temperature and precipitation are 6.9°C and 397.5 mm, respectively, with rainfall mainly concentrated between June to September (Dai et al., 2020; Yang et al., 2021). This basin has a temperate continental monsoon climate. It belongs to a typical Farming-Pastoral Ecotone, with cropland and grassland as the dominant land cover followed by forest and urban land (Figure 1, Supplementary Table S1). The main fertilizer applied include nitrogen fertilizer (compound fertilizer, ammonium and urea) and phosphate fertilizer (Wang et al., 2020). The average application of nitrogenous fertilizer and compound fertilizer was 4.26 × 10⁴ tons N and 6.10 × 10⁴ tons N in 2019 and 2020 in Zhangjiakou city, respectively (http://tjj.hebei.gov.cn/). The livestock animals in Zhangjiakou city are dominated by cattle and sheep and goats, with about 4.47×10⁵ and 1.67×10⁶ heads, respectively (http://tjj.hebei.gov.cn/).

River water samples were collected from 17 sampling sites (M1–M17) in the mainstream and five sites (T1–T5) in the tributaries of the Yang River (Figure 1). Three sampling campaigns were conducted in December 2019 July 2020, and April 2021 along the Yang River, corresponding to the dry, wet, and normal seasons. The seasons are divided based on precipitation considering the precipitation of April, July and December accounted for 6%, 24% and 0.5% of annual precipitation, respectively, during the period of 2019 and 2020 (http://tjj.hebei.gov.cn/). Four snow samples were collected in December in the study area. All the water samples were filtered through 0.7-μm glass fiber filters (Whatman GF/F) with pre-combusted for 3 h at 450°C.

The field measurement for river water dissolved oxygen (DO) via a portable multi-parameter meter (WTW Multi 3430 IDS, Germany). An automatic flow analyzer determined the total dissolved nitrogen (TDN) concentrations (mg/L for N) and different forms of DIN concentrations (SKALAR Sans Plus Systems). The detection limit of TDN, NO₂ ⁻-N, NO₃ ⁻-N and NH₄ ⁺-N is 0.02 mg/L, 5 μg/L, 0.01 mg/L and 0.01 mg/L, respectively. Dissolved organic nitrogen (DON) concentration (mg/L for N) was calculated by subtracting DIN from TDN. Chloride concentration was determined by Dionex ion chromatography (Dionex Corp. Sunnyvale, CA, United States) with a precision of ≤5%. A total organic carbon analyzer determined the dissolved organic carbon (DOC) (Aurora 1030W + 1088, OI Analytical, United States). Analytical errors were ≤1.5% for DOC according to triplicate sample measurements. The composition of DOM was analyzed in the wet season based on Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS). The detailed analysis method for DOC concentration and DOM composition can be found in our another study (Ge et al., 2022). The nitrogen and oxygen isotopes of nitrate were measured by a bacterial denitrifier method, which reduces NO₂ ⁻-N and NO₃ ⁻-N to N₂O via a special kind of denitrifying bacteria with a lack of N₂O reductase (Sigman et al., 2001; Casciotti et al., 2002). After purification, the nitrogen and oxygen isotopes of N₂O were determined by an isotope ratio mass spectrometer (Delta V, Thermo Fisher). The international standards (USGS-32, USGS-34, USGS-35, IAEA-NO3) were used for calibration of the dual isotopes of nitrate (Yue et al., 2020). The delta (δ) notation in parts per thousand (‰) is reported to express the isotopic compositions of nitrate relative to the international standards (VSMOW for δ¹⁸O, atmospheric N₂ for δ¹⁵N).

Eight soil samples (0–10 cm) were collected on the riverbank from cropland (2 samples from site M1, one from M9 and one from M15) and forest and grass land (1 sample from site M6 and three from M2) in July 2020. The nitrate in fresh soil were extracted with 2 M KCl solution in a 1:4 mass ratio (soil: solution) after 1 hour of shaking (Rock et al., 2011). The nitrate concentration of KCl solution was below the detection limit by combusting KCl at 450°C for 4 hours (Wang et al., 2018). The nitrate extracted from soil was determined for dual isotopes of nitrate using the above bacterial denitrifier method. Additionally, some air-dried soil samples were sieved to 100 mesh for measuring the isotope of particle nitrogen in soil using isotope ratio mass spectrometer (Delta V, Thermo Fisher).

Based on Landsat Thematic Mapper imagery (30 m resolution), land use types were classified and the percentage of different land-use types was calculated by ArcMap 10.2 (Supplementary Table S1). Five riparian buffer zones of 500 m, 1 km, 3 km, 5 km, and 8 km were extracted to evaluate the influence of land use on the nitrogen source and transformation using Spearman’s correlation coefficients due to the non-normally distribution of data set. After the correlation analysis between nitrogen-related parameters and the percentage of land use in each buffer zone, the significant correlations were observed in 3-km buffer zones with discussed in detail in the below section unless otherwise noted. Linear regression analysis was used for trend analysis among nitrogen-related variables. Kruskal–Wallis non-parametric test (K-W test) was used to test the seasonal and temporal differences in concentrations of different nitrogen species as well as dual isotopes of nitrate. Principal Component Analysis (PCA) can be used to reduce data dimensionality by converting large data sets into several principal components (PCs), which “represent a process influencing the data” (Matiatos, 2016). In this study, four pCs were retained when eigenvalues were greater than 1. The data of TDN concentrations were not included during PCA analysis considered the significant correlation between TDN concentrations and almost all the variables. Of note, two sampling sites (M2 and M14) were excluded from the above correlation and PCA analysis due to their scattered pattern, which might be related to other complicated factors besides land use. The contribution of nitrate sources was estimated by a Bayesian mixing model, which was implemented in a Stable Isotope Analysis in the R (SIAR) package. The details of set-up parameters can be found in previous studies (Parnell et al., 2010; Xia et al., 2017) and our published studies (Li et al., 2019; Yue et al., 2020). The isotopic values of nitrate end-members for the SIAR were listed in Supplementary Table S2. All of the statistical analyses were carried out in R 4.1.2.

Total dissolved nitrogen (TDN) ranged from 0.59 mg/L to 23.96 mg/L, with a significantly higher average value in the dry season (7.96 ± 5.15 mg/L) than that of the wet season (3.59 ± 2.41 mg/L, p < 0.05, Supplementary Figure S1). About 80% of samples exceed Class V (2 mg/L) based on Chinese quality standards for surface water (GB3838-2002). The high TDN concentrations occurred in sites with more urban distribution and the correlation between dissolved nitrogen and land use types will be presented below.

Among different species of TDN, NO₃ ⁻-N is the primary form, followed by DON, NH₄ ⁺-N and NO₂ ⁻-N. Most samples’ NO₃ ⁻ concentrations accounted for >50% of TDN. The NO₃ ⁻-N concentrations varied from 0.10 mg/L to 20.61 mg/L, with a significantly higher level in the dry season (6.42 ± 4.39 mg/L) than in two other seasons (Figure 2). A higher level of NO₃ ⁻-N was found in the lower stream with more cropland and urban land even if no significant difference was found between the upper and lower streams (Figure 2).

The seasonal pattern of NH₄ ⁺-N concentrations was similar to TDN and NO₃ ⁻-N, with a significantly higher level in the dry season (1.26 ± 1.80 mg/L) than in normal season (0.65 ± 1.97 mg/L) and wet season (0.17 ± 0.10 mg/L, Figure 2). According to the Chinese quality standard for surface water, several samples had NH₄ ⁺-N concentrations beyond the Class V (2 mg/L) guideline (GB3838-2002). In contrast, DON concentrations showed a different seasonal pattern, with a significantly higher level in the normal season (1.13 ± 0.62 mg/L) than that of the wet season (0.60 ± 0.36 mg/L) and dry season (0.18 ± 0.18 mg/L, Supplementary Figure S1). The peak concentrations of DON occurred in the site (M11) close to industrial areas and then kept an elevated level in the lower reach with a high proportion of cropland and urban land. Indeed, the downstream had significantly higher DON concentrations than that of upstream (Supplementary Figure S1). NO₂ ⁻-N was detected in most samples, but it was the lowest among the different nitrogen forms, with most sites having about 1% of TDN. However, several samples in the wet season had NO₂ ⁻-N concentrations accounting for >5% of TDN. The NO₂ ⁻-N concentrations showed a significantly spatial difference with a higher level in the downstream than that of upstream (Supplementary Figure S1).

The δ¹⁵N-NO₃ ⁻ values displayed seasonal changes with a significantly higher mean in the normal season (+15.8 ± 4.3‰) than that of the wet season (+12.1 ± 3.9‰, Figure 2C). However, the dry season did not show a higher δ¹⁵N-NO₃ ⁻ value (+12.7 ± 3.2‰) than other seasons when compared with a general pattern of global rivers, which had an about one‰ increase in δ¹⁵N-NO₃ ⁻ in dry seasons than different seasons (Matiatos et al., 2021). The δ¹⁵N-NO₃ ⁻ values exhibited apparent spatial variations, with a significantly higher level in the lower stream (+15.0 ± 3.9‰) than that of the upper stream (+12.1 ± 4.0‰). The average values of δ¹⁸O-NO₃ ⁻ followed the order of normal season > wet season > dry season, but they did not show significant seasonal differences among the three seasons (Figure 2D).

Figure 3 displays the correlation between land use types and the dissolved nitrogen species and dual isotopes of nitrate. A significant positive correlation was observed between the percentage of the urban area and annual mean concentrations of TDN, NO₃ ⁻-N, DON, Cl⁻ as well as δ¹⁵N-NO₃ ⁻ values (Figure 3A). As for the cropland, the percentage of cropland showed a positive correlation with annual mean concentrations of TDN, NO₃ ⁻-N and δ¹⁵N-NO₃ ⁻ values. In contrast, a significant negative correlation was found between the percentage of forest and grassland and annual mean concentrations of TDN, NO₃ ⁻-N, DON, Cl⁻ as well as δ¹⁵N-NO₃ ⁻ values. In the dry season, a positive correlation was observed between concentrations of NO₂ ⁻-N and NH₄ ⁺-N as well as NO₃ ⁻-N, and δ¹⁵N-NO₃ ⁻ values were positively correlated with NO₃ ⁻ concentrations and Cl⁻ concentrations (Figure 3B). In the normal season, δ¹⁵N-NO₃ ⁻ values were positively correlated with δ¹⁸O-NO₃ ⁻ values and DON concentrations, and NO₃ ⁻ concentrations were positively correlated with DON and Cl⁻ concentrations (Figure 3C). In the wet season, a positive correlation was also observed between concentrations of NO₂ ⁻ and NH₄ ⁺ as well as DO (Figure 3D), which was similar to the dry season.

North China is dominated by the service industry, commerce and manufacturing industry, which contribute significant amounts of nitrogenous compounds to rivers (Zhang et al., 2021). Thus, the average NO₃ ⁻-N concentration in Yang River (4.04 ± 3.42 mg/L) was comparable to the average level of North China (4.74 ± 9.24 mg/L, Zhang et al., 2021) but was much higher than that of South China (1.74 ± 0.5 mg/L, Zhang et al., 2021). Likewise, the average of δ¹⁵N-NO₃ ⁻ in Yang River (+13.6 ± 4.2‰) was similar to the average δ¹⁵N-NO₃ ⁻ in North China (+12.6‰) but was much higher than that of South China (+8.1‰, Zhang et al., 2021). The higher levels of nitrogen concentration and δ¹⁵N-NO₃ ⁻ were also reported in other rivers with more distribution of cropland and urban land (Mayer et al., 2002; Qin et al., 2018; Wong et al., 2018).

To examine the effect of land use type on nitrate pattern, the correlations were analyzed between the riverine dissolved nitrogen concentrations and different land-use shares. There is a significant positive correlation between urban ratios and annual mean concentrations of TDN, DON and NO₃ ⁻-N as well as δ¹⁵N-NO₃ ⁻ values (Figure 3A), which was also reported in other studies (Ohte et al., 2010; Taufiq et al., 2019). The increasing urban area would lead to more domestic and industrial wastewater release, causing elevated nitrogenous concentrations and δ¹⁵N-NO₃ ⁻ values. The significant positive correlation between the percentage of urban land and both the DON and DOC concentrations (Figure 3A) also reflected the influence of domestic wastewater on organic matter variations. The negative correlation between forest and grassland area ratios and concentrations of TDN, DON and NO₃ ⁻-N as well as δ¹⁵N-NO₃ ⁻ values (Figure 3A) indicated that the role of forest and grass in nitrogen removal via assimilation or adsorption (Mayer et al., 2002; Ohte et al., 2010; Taufiq et al., 2019). A significant positive correlation was found between the percentage of cropland and TDN concentrations, NO₃ ⁻-N concentrations and δ¹⁵N-NO₃ ⁻ values, indicative of the contribution of agricultural activities to nitrogen variations in the Yang River, which was also reported in other studies (Mayer et al., 2002; Qin et al., 2018; Wong et al., 2018). Overall, the land-use types play an important role in variations of nitrogenous concentrations and δ¹⁵N-NO₃ ⁻ values in the Yang River.

The main factors driving the nitrate concentrations were identified using PCA from nitrogen-related variables. The PCA identified four principal components (PC1, PC2, PC3 and PC4) with accounting for 85.8% of the total variance for the nitrogen-related variables in the Yang River (Figure 4). The PC1 explained 47.1% of the variance with the three major land use types included, which suggested that PC1 reflected the effects of land use on nitrate pollution. The PC1 had positive loadings for NO₃ ⁻-N, DON, DOC, Cl⁻, δ¹⁵N-NO₃ ⁻, urban ratios and cropland ratios, indicating the nitrate sources from urban and agricultural activities. However, the negative loading for forest and grass areas ratios in the PC1 reflected the role of forest and grassland in water purification as discussed in the above correlation analysis. The PC2 explained 15.5% of the variance, with positive loadings for NH₄ ⁺-N, NO₂ ⁻-N and cropland but a negative loading for Cl⁻. The PC2 did not represent the nitrate source from agricultural activities, otherwise the positive loadings would be observed for both cropland and Cl⁻ since high Cl⁻ concentrations were reported in cropland due to the application of manure or organic fertilizer (Liu et al., 2006). Therefore, the PC2 might suggest the transformation of nitrogen (nitrification process) in the cropland as indicated by the positive loadings for both NH₄ ⁺ and NO₂ ⁻, which acted as the reactant and intermediate products of nitrification, respectively. The PC3 explained 13.0% of the variance, with positive loadings for NO₂ ⁻-N, NH₄ ⁺-N and forest and grass while a negative loading for cropland and δ¹⁸O-NO₃ ⁻, which suggested the nitrification process in forest and grassland since the nitrification process is closely associated with variations in δ¹⁸O-NO₃ ⁻ values, NO₂ ⁻-N and NH₄ ⁺-N concentrations. The PC4 explained 10.3% of the variance, with positive loadings for δ¹⁵N-NO₃ ⁻, δ¹⁸O-NO₃ ⁻, DON, DOC and forest and grassland ratios while a negative loading for NO₃ ⁻-N, which suggested the assimilation process since a simultaneous increase occurred in δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ while a decrease in NO₃ ⁻-N concentrations as plants absorbed nitrate. The denitrification is excluded from the above transformation process, which would be discussed in detail in the next section.

Nitrogen transformations, such as nitrification, denitrification and assimilation, may alter the nitrogen concentrations and involve isotopic fractionation in the aquatic system. Thus, it is needed to ascertain the nitrogen transformations before identifying nitrate sources. Nitrification is the oxidation process of ammonium and consequently creates nitrate as products with NO₂ ⁻ as intermediate products under aerobic conditions. δ¹⁸O-NO₃ ⁻ can be used as a tracer of nitrification since one oxygen atom of nitrification-derived nitrate is derived from ambient oxygen molecules and the other two oxygen atoms from ambient water molecules (Andersson and Hooper, 1983). Thus, δ¹⁸O-NO₃ ⁻ of nitrification can be calculated according to δ¹⁸O-O₂ (23.5‰, Amberger and Schmidt, 1987) and δ¹⁸O of river water in the studied basin (−14.6‰ to −7.2‰, Kong et al., 2021). As a result, the theoretical δ¹⁸O-NO₃ ⁻ is estimated to range from −1.9‰ to +3.0‰. About 22% of river water samples fell into this estimated range, indicating that only a part of nitrate was derived from in-stream nitrification with mainly observed in dry and wet seasons based on the δ¹⁸O-NO₃ ⁻ values. However, most samples in the Yang River had δ¹⁸O-NO₃ ⁻ values higher than the estimated δ¹⁸O-NO₃ ⁻, while they fell into the range of microbial nitrification, characterized by a range of −10‰ to +10‰ (Kendall et al., 2007). This pattern reflected the in-soil nitrification during the movement of nitrate to river, which was similar to another study (Xuan et al., 2022). Moreover, the δ¹⁸O-NO₃ ⁻ of soil extracts (−4.6‰ to +8.8‰) covered the range of δ¹⁸O-NO₃ ⁻ for most samples of the river water in the studies area (Figure 5), which also verified the critical role of the in-soil nitrification process.

In addition to δ¹⁸O-NO₃ ⁻, δ¹⁵N-NO₃ ⁻ and other variables can also be used to trace the nitrification process. When ammonium is limited, δ¹⁵N-NO₃ ⁻ values tend to be close to soil organic nitrogen with depleted ¹⁵N/¹⁴N ratios (Kendall et al., 2007). In the Yang River, NH₄ ⁺-N concentrations were much lower in the normal season and wet season than that in the dry season (Figure 2B), which might result from rapid nitrification (Sanchez et al., 2017; Yue et al., 2020). As intermediate products of nitrification, the NO₂ ⁻ concentrations increased with the elevated NH₄ ⁺ concentrations in the wet season and dry season (Figures 3B,D), corroborating the occurrence of the in-stream nitrification process during these two seasons, which was in line with the above conclusion inferred by δ¹⁸O-NO₃ ⁻ variations. Furthermore, the decreasing δ¹⁵N-NO₃ ⁻ values with an increasing NO₃ ⁻/Cl⁻ ratios (Figure 6A) also confirmed the nitrification in the normal season (Jiang et al., 2021). Alternatively, the correlations between δ¹⁵N-NO₃ ⁻ and NO₃ ⁻/Cl⁻ reflected the mixing process between different nitrate sources, discussed in the next section.

During natural attenuation by denitrification, the residual nitrate is preferentially enriched in ¹⁵N and ¹⁸O, with a rough ratio of 1.3:1 to 2:1 (Kendall et al., 2007). In this study, a significant positive correlation was found between δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ in the normal season (Figure 6B), but its ratios (δ¹⁵N:δ¹⁸O = 3.3, the inverse of slope) fell outside the expected ratios of denitrification. Thus, the denitrification process of surface water is not plausible to occur based on the above variation trend of nitrate isotopes and high DO concentrations (10.26 ± 1.23 mg/L, Supplementary Figure S2). The lack of denitrification in the Yang River is consistent with the study on nitrate transformation of groundwater in the Yang River basin (Yang et al., 2021), which reported the oxic conditions and weak correlation of δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ in the groundwater of studied area. This is also in accordance with previous studies in the semiarid aquifer systems, where denitrification is inhibited due to the oxic conditions while nitrification is allowed to accumulate nitrate (Sanchez et al., 2017; Gutiérrez et al., 2018; Ma et al., 2021). Although there is a lack of denitrification process in the surface water of Yang River, denitrification process might occur in the sediment. However, the effect of benthic denitrification on overlying water could be negligible since the coupled nitrification–denitrification within sediments would be decreased under the high NO₃ ⁻ concentrations and O₂ levels in overlying water column (Rysgaard et al., 1994).

Assimilation might occur in the Yang River given the supersaturated state of DO by photosynthesis at some sampling sites. It is reported that assimilation causes a simultaneous increase of δ¹⁵N and δ¹⁸O in a 1:1 ratio during NO₃ ⁻ uptake by phytoplankton (Granger et al., 2004). However, the expected isotopic signal of assimilation was not observed as reflected by the scattered trend of δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ in the wet and dry season. Even if δ¹⁵N-NO₃ ⁻ was positively correlated with δ¹⁸O-NO₃ ⁻ in the normal season, the slope of δ¹⁸O:δ¹⁵N (0.38) was much lower than 1 (Figure 6B). The lack of isotopic signal of assimilation might be related to repression of NO₃ ⁻ uptake by elevated NH₄ ⁺ concentration (>1.5 μmol/L, Montoya et al., 1991). But the NO₃ ⁻ uptake cannot be completely ruled out since the δ¹⁵N-NO₃ ⁻ values increased with the decreasing NO₃ ⁻/Cl⁻ ratios (Figure 6A). Another reason for the weak signal of assimilation might be the concurrence of assimilation and nitrification in soil, where in-soil nitrification result in a much lower enrichment of δ¹⁸O-NO₃ ⁻ than δ¹⁵N-NO₃ ⁻ as reflected by the soil extracts (δ¹⁸O: δ¹⁵N = 0.37, Figure 6B).

According to the land use of the studied area (Figure 1), the potential nitrate sources included chemical fertilizer (CF), soil organic nitrogen (SON), manure and sewage. NO₃ ⁻/Cl⁻ ratios are reported to distinguish nitrate sources (Liu et al., 2006; Xia et al., 2017). Sewage is featured by relatively high NO₃ ⁻/Cl⁻ ratios and Cl⁻ concentrations, whereas manure is characterized by low NO₃ ⁻/Cl⁻ ratios and high Cl⁻ concentrations as animal manure has low NO₃ ⁻ and high Cl⁻ concentrations (Liu et al., 2006; Xia et al., 2017). Chemical fertilizer is marked by high NO₃ ⁻/Cl⁻ ratios and low Cl⁻ concentrations. As shown by the graph of NO₃ ⁻/Cl⁻ ratios vs Cl⁻ concentrations (Supplementary Figure S3), the Yang River had a mixing of CF, manure and sewage. Rainwater might also contribute nitrate to the river since some samples had low NO₃ ⁻/Cl⁻ ratios and low Cl⁻ concentrations. The negative correlation of NO₃ ⁻/Cl⁻ and δ¹⁵N-NO₃ ⁻ (Figure 6A) also reflected the mixture between two different nitrate sources, with one source having low δ¹⁵N-NO₃ ⁻ values and high NO₃ ⁻/Cl⁻ (like CF), and the other one having high δ¹⁵N-NO₃ ⁻ values and low NO₃ ⁻/Cl⁻ (like manure).

As plotted in Figure 5, most dual isotopes of nitrate fell into the range of manure, sewage and SON, indicating their dominant contribution to riverine nitrate in the Yang River. Seasonally, the significantly higher δ¹⁵N-NO₃ ⁻ values in the normal season were indicative of ¹⁵N-enriched origins, i.e., sewage and manure. Spatially, a significantly higher δ¹⁵N-NO₃ ⁻ in the lower reach suggested an increasing contribution of sewage and manure, in accordance with the increasing ratios of urban and cropland down the river. The positive correlation of urban (or cropland) ratios and NO₃ ⁻-N concentrations and δ¹⁵N-NO₃ ⁻ values (Figure 3A) also confirmed the influence of sewage and manure. Although one sample fell into the range of synthetic nitrate fertilizer, nitrate fertilizer was not commonly applied in China (<2%, Liu et al., 2013). Moreover, compound fertilizer, ammonium and urea are the main synthetic fertilizer in the study area. Thus, the influence of synthetic fertilizer should be taken into account when considering the wide distribution of cropland in the Yang River and the positive correlation of cropland area and concentrations of TDN and NO₃ ⁻-N as well as δ¹⁵N-NO₃ ⁻ values (Figure 3A). One soil extracts sample fell into the range of ammonium fertilizer, while other soil extracts samples fell into the range of SON and manure, which coincided with the application of animal waste (sheep and cattle) as base fertilizer before planting winter wheat in the study area. Under the semi-arid climate, Yang River is mainly recharged by rainfall during the wet season. Thus, atmospheric precipitation (AP) is a possible nitrate source for the Yang River.

In addition to physicochemical parameters and nitrate isotopes, DOM composition could also be used to trace nitrate source and transformation, given the coupling relationship of carbon and nitrogen in the aquatic environment. It has been reported that DOM molecular signatures are related to land use types with cropland streams enriched in aliphatic and lignin-like compounds (Spencer et al., 2019; Ge et al., 2022). Our another study on molecular signatures of DOM in the Yang River indicated that DOM was predominantly from allochthonous inputs in association with land use types (Ge et al., 2022). Generally, the relative abundance of lignin-like compounds ranged from 77.5% to 90.5%, with high abundance found in cropland dominated sites in the study area. The relative abundance of aliphatic compounds ranged from 1.9% to 14.6%, with high abundance found in urban area (Ge et al., 2022). Correspondingly, the relative abundances of aliphatic compounds were positively correlated with Cl⁻ concentrations and negatively correlated with NO₃ ⁻/Cl⁻ ratios (Figures 7A,B), which might reflect the mixing process of different nitrate sources, one was manure with high Cl⁻ concentrations and low NO₃ ⁻/Cl⁻ ratios and the other one was synthetic fertilizer with a contrast pattern (Liu et al., 2006). Livestock farming is relatively developed in the Yang River and animal manure could contribute to DOM and then be degraded to inorganic nitrogen via mineralization and nitrification. The relative abundances of lignin were negatively correlated with Cl⁻ concentrations and positively correlated with NO₃ ⁻/Cl⁻ ratios (Figures 7C,D), indicating the transport of synthetic fertilizer-derived nitrate along with lignin when considering the lignin as a refractory component and terrestrial tracer (Kirk and Farrell, 1987; Ni et al., 2020a; 2020b). To sum up, a suite of water chemical parameters, DOM composition and nitrate isotopes confirmed the inputs of manure, sewage, SON, CF and AP for riverine nitrate in the Yang River.

The SIAR model estimated the proportional contribution of the nitrate sources aforementioned. The nitrate sources’ mean probability estimate (MPE) was plotted in Figure 8. As shown by the mixing model outputs, sewage was the dominant nitrate source (36% ± 20%). Manure, SON and CF contribution were intermediate (26% ± 17%, 19% ± 14%, 15% ± 11%), and AP contributed the least (4% ± 5%). This result agreed with the estimation of the whole Haihe River basin with manure and sewage, SON, CF and AP accounting for 39%–97%, 0%–44%, 0%–16% and 0%–12%, respectively (Zhang et al., 2021).

Overall, the proportion of nitrate sources varied seasonally and temporally. Sewage and manure contributed the most in the normal season (41% ± 21%, 29% ± 19%) relative to the other two seasons (Figure 8). The pattern of higher sewage and manure contribution in the normal season (spring) was similar to another study, which indicated that sewage and manure was frozen and trapped in riverbanks in winter while it turned to melt and enter the river in spring (Jiang et al., 2021). For another reason, animal waste (sheep and cattle) was used in winter as base fertilizer in the study area, and this organic fertilizer could be nitrified in the following spring. Consequently, the nitrification-derived nitrate would be leached to river water during irrigation and rainfall events in spring. The elevated δ¹⁵N-NO₃ ⁻ values with increasing cropland area (Figure 3) confirmed the livestock manure as a kind of organic fertilizer during agricultural activities in the Yang River basin. Spatially, the higher proportion of sewage was found in the downstream of Yang River, which has more urban area than the upper stream (Figure 8).

Compared with the wide distribution of cropland, CF only contributed 15% of nitrate, which might be related to China’s action of zero growth in chemical fertilizer by 2020 (MARC, 2015) and natural attenuation such as in situ and riparian attenuation as well as nitrate infiltration into groundwater. Nitrate attenuation via denitrification can be excluded, as discussed before, in the semi-arid area. Thus, the nitrogen attenuation via biological assimilation could be one reason for the low contribution of CF relative to the extensive distribution of cropland when considering the negative correlation between nitrate concentrations and forest and grassland area (Figure 3A). Nitrate infiltration to groundwater could also be an important cause, as reflected by the lower riverine NO₃ ⁻-N concentrations in Yang River (4.04 mg/L) relative to groundwater with an average of 6.15 mg/L during our sampling periods (Yang et al., 2021). Although CF contributed less, the non-point sources (manure, SON and CF) accounted for 60% of nitrate. Thus, the non-point sources should be paid more attention to management of nitrogen pollution.

The contribution of AP was the lowest, which was comparable to other reports (Matiatos, 2016; Yi et al., 2020; Jiang et al., 2021). The AP contribution was slightly higher in the normal and wet seasons than in the dry season. However, the seasonal difference was not significant. A spatial variation occurred in the Yang River, with a higher proportion in the upper stream than in the lower stream. The upper stream has more forest and grassland cover distribution, which might increase input from atmospheric NO₃ ⁻ due to the interception of wet deposition by forest and grass.