Water quality attribution and simulation of non-point source pollution load flux in the Hulan River basin

Surface water is the main source of irrigation and drinking water for rural communities by the Hulan River basin, an important grain-producing region in northeastern China. Understanding the spatial and temporal distribution of water quality and its driving forces is critical for sustainable development and the protection of water resources in the basin. Following sample collection and testing, the spatial distribution and driving forces of water quality were investigated using cluster analysis, hydrochemical feature partitioning, and Gibbs diagrams. The results demonstrated that the surface waters of the Hulan River Basin tend to be medium–weakly alkaline with a low degree of mineralization and water-rock interaction. Changes in topography and land use, confluence, application of pesticides and fertilizers, and the development of tourism were found to be important driving forces affecting the water quality of the basin. Non-point source pollution load fluxes of nitrogen (N) and phosphorus (P) were simulated using the Soil Water and Assessment Tool. The simulation demonstrated that the non-point source pollution loading is low upstream and increases downstream. The distributions of N and P loading varied throughout the basin. The findings of this study provide information regarding the spatial distribution of water quality in the region and present a scientific basis for future pollution control.

Sample collection and analysis. Based on existing survey data and according to the distribution of the water system and land use in the survey area, field sampling of the Hulan River Basin was conducted during June and October 2018. Sampling points were distributed in the upper, middle, and lower reaches of the river, both upstream and downstream of confluences and cities. The spatial distribution of sampling points is shown in Fig. 3.
Water samples were collected according to the "Technical Specifications for Surface Water and Sewage Monitoring" (HJ/T91-2002). An HQ40d Hach water quality monitor was used to test water temperature, total dissolved solids (TDS), conductivity, dissolved oxygen, and redox potential. Water was stored at 0-4 °C and total nitrogen (TN), total phosphorus (TP), chemical oxygen demand (CODCr), and ionic composition were measured within eight hours. TN was determined using alkaline potassium persulfate digestion UV spectrophotometry (GB11894-89), TP was determined using ammonium molybdate spectrophotometry (GB11893-89), CODCr was determined using the dichromate method (GB11914-89). Data analysis. The statistical analysis of water quality indicators was conducted using SPSS. A cluster analysis and principal component dimensionality reduction were used to determine the spatial difference and similarity of water quality. Water chemistry type was determined according to the Shukalev classification 30 , and pollutant sources were analyzed using the end element map 1 22 . The SWAT model is used for basin wide simulations of surface source pollution; water resources assessment and management; soil and water conservation; prediction of the influence of climate change; and land management measures on hydrology, sediment and nutrient production, and migration in complex watersheds. The SWAT model is divided into four modules; hydrological, soil erosion and sediment transport, nutrient transport, and plant growth and management. The nutrient transport module of the SWAT model simulates the migration and transformation of N and P nutrients. The migration and transformation of N, particularly NO 3 contained in runoff, lateral flow, and infiltration, are calculated by the volume of water and the average degree of aggregation. Effects of filtration are considered for underground infiltration and lateral runoff. Nitrogen can be divided into dissolved N and adsorbed N, where dissolved nitrogen is mainly nitrate N. Before calculating the total amount of nitrate N, it is necessary to calculate the concentration of nitrate N in mobile water, and then multiply the concentration by the amount of water to obtain the total amount of nitrate N. The calculation of free water nitrate N concentration is as follows: where p mobile is the concentration of nitrate N in free water (kg/mm), p ly is the amount of nitrate N in the soil (kg/hm 2 ), W mobile is the amount of free water in the soil (mm), θ e is porosity, and SAT is the soil saturated water content. Adsorbed N is mainly organic N and is determined using the model developed by McElroy et al. and modified by Williams and Hann 31 . The expression is: where ρ orgNsurf is the amount of organic N loss (kg/hm 2 ), ρ orgN is the concentration of organic N in the soil surface layer to a depth of 10 mm (kg/t), m is the amount of soil loss (t), A hru is the area of the hydrological response unit (hm 2 ), and ε N is the nitrogen enrichment coefficient (dimensionless). Phosphorus is also divided into dissolved P and adsorbed P. The migration of dissolved P in the soil is mainly achieved by diffusion. Since dissolved P is not very active, the surface layer of P in dissolved form is rarely removed from surface runoff. Dissolved P transported by surface runoff is calculated by: where P surf is dissolved P lost through surface runoff (kg/hm 2 ), P solution,surf is dissolved P in soil (kg/hm 2 ), ρ b is soil bulk density (mg/m 3 ), A hru is surface soil depth (mm), and k d,surf is the soil P partition coefficient (dimensionless). Adsorbed P is mainly divided into organic P and mineral P, which are usually adsorbed on soil particles and migrate with runoff. The calculation expression is: where m Psurf is the amount of organic P loss (kg/hm 2 ), ρ P is the concentration of organic P in the surface soil (kg/t), m is the amount of soil loss (t), A hru is the area of the hydrological response unit (hm 2 ), and ε P is the P enrichment factor (dimensionless). www.nature.com/scientificreports www.nature.com/scientificreports/

Results
Water quality characteristics. The cluster analysis and principal component dimensionality reduction analysis were used to classify rivers in Hulan basin based on the spatial distribution of pollutants. Average values of water quality indicators for the seven tributaries flowing into the main stream of the Hulan River Basin were clustered using squared Euclidean distance as the clustering index and results are shown in Fig. 4. Tributaries can be divided into two groups-A: the Anbang, Numin, Yijimi, Ougen, and Small Hulan Rivers, and B: the Keyin and Gemuke Rivers. Figure 5 shows that the water quality of the sampling points can be divided into two groups, with green areas belonging to the sampling points of the Keyin and Gemuke Rivers. Sampling points 39 and 41 downstream of the Keyin River are not in this grouping due to the influx of other tributaries, which impact water quality. The water quality samples in the blue region are relatively similar and represent the sampling points of the remaining tributaries, corresponding well to the results of the cluster analysis.
Based on the spatial distribution of water quality, we divide the basin into Group A (Ampang, Numin, Yijimi, Ougen, and Small Hulan Rivers), Group B (Keyin and Gemuke Rivers), and the main stream of the Hulan River.
Water chemistry determined using the Shukalev classification method are shown in the Piper three-line diagram (Fig. 6) and the water chemistry type zoning diagram (Fig. 7).  www.nature.com/scientificreports www.nature.com/scientificreports/ It can be seen from the piper diagram that the water chemistry type of the Hulan River main stream is mainly HCO3-Ca, while that of the tributary from upstream to downstream within the basin changes from HCO3•SO4-Ca to HCO3-Ca•Mg.
The TDS of Group A is 50~80 mg/L and water chemistry type is predominantly HCO 3 •Cl-Ca(HCO 3 -Ca); however, the water type of the Yijimi River is SO 4 •HCO 3 -Ca. The TDS of Group B is 91~298 mg/L, and the water chemistry types are HCO 3 -Ca and HCO 3 •Na-Ca. The average TDS value of the Hulan River main stream is  www.nature.com/scientificreports www.nature.com/scientificreports/ 55~95 mg/L, and water chemical types are HCO 3 •Cl-Ca and HCO 3 -Ca. Downstream of the Yijimi River injection, the water type is SO 4 •HCO 3 •Cl-Ca. The TDS of surface water in the basin is generally low, increasing gradually downstream except for in the Keyin River following the injection of the Numin River. The TDS of Group B is much higher as compared to Group A, while the Hulan River main stream has a TDS between these two groups. The ion types of the rivers in the basin are gradually enriched downstream.
The average concentration of the major ions in basin surface waters is shown in Table 1. It is apparent that ion concentrations in Group A are lower as compared to Group B, while the ion concentrations of the main stream are between these two groups. Ion concentrations generally increase downstream. Anion concentrations in Groups A and B are HCO 3 − > Cl − > SO 4 2− and HCO 3 − »SO 4 2− > Cl − , respectively, while all groups have cation concentrations Ca 2+ > K + + Na + > Mg 2 +. Overall, HCO 3− and Ca 2+ are the dominant components.
Comprehensive environmental indicators characterize the overall salinity of the water body, including total hardness and conductivity. Oxidation reduction potential (Eh), pH, dissolved oxygen (DO), and biochemical oxygen demand (COD) are shown in Table 2. Total hardness and conductivity generally show an increasing trend downstream and Group B values are much higher as compared to Group A and the mainstream. The pH values of surface waters are between 7.51 and 8.09 (medium-weak alkaline water) and Group B pH is slightly higher as compared to the other two regions. The pH of the main stream increases gradually downstream, while Group A and B pH decreases gradually. The Eh of the surface water in the basin is 200 ± 10 mv. According to the "Environmental Quality Standard for Surface Water" (GB3838-2002), most of the dissolved oxygen (DO) levels meet Class I and II water quality standards, and very few sites are Class III (the middle reaches of the Keyin and Hulan Rivers). The COD of most surface waters is in the IV and V standard range, while the middle reaches of the Gemuke and Hulan Rivers are nearly twice as high as the Class III standard.
Based on the China's drinking water hygiene standards (GB 5749-2006), according to the Class III water standard, the most polluting components in the basin are total N and P. Figure 8 shows that the water quality of total N is generally Class IV and V, while some samples are inferior to Class V. The total N content in Group B is higher as compared to Group A and the Hulan mainstream. The total P content in Group B is higher as compared to Group A and the Hulan mainstream.
Driving forces of water quality. Figure 9 shows Gibbs diagrams of the basin. It is apparent that samples generally fall in the rock weathering control area, indicating that the water chemistry of the basin is mainly controlled by rock weathering 4,32,33 . Figure 10 shows that ions are mainly composed of silicate mineral weathering products, followed by carbonate mineral weathering products, corresponding well with the geological features of the region.
According to 2015 land use remote sensing data (Figs. 11 and 12) combined with field survey results, the basin is mainly comprised of cultivated and forested land, accounting for more than 80% of the total land use. Forest

Region Location
Average ion content (mg/L) www.nature.com/scientificreports www.nature.com/scientificreports/ land is concentrated in the upper reaches of the basin and the middle reaches are mainly cultivated. Following the confluence of tributaries in the lower reaches, residential land intensifies.

SWAT model simulation.
Model configuration and validation. Many input parameters are required for the SWAT model 34 , including digital elevation models (DEM), land-use area, soil type, meteorological data, and hydrological data 35 as shown in Table 3.
Due to the number of parameters in the SWAT model, individual calibration of parameters is difficult. Therefore, the sensitivity analysis method is generally used to determine the sensitivity of model parameters. Those parameters that have a large influence on model simulation results are selected using the SWAT-CUP sensitivity analysis tool to reduce the workload during model calibration and verification as shown in Table 4 36 .
The hydrological cycle forms the basis of the hydrological model; however, rainfall and runoff are the driving forces of non-point source pollution. Therefore, the calibration and verification sequence of SWAT model parameters are runoff, sediment, and water quality. The model was calibrated spatially from the upper to lower reaches at Tieli, Sifang, and Qinjia stations.  www.nature.com/scientificreports www.nature.com/scientificreports/   www.nature.com/scientificreports www.nature.com/scientificreports/ The accuracy of the model simulation results can directly reflect the applicability of the model in a study area. Here, the relative error (PBIAS), the deterministic coefficient (R 2 ), and the Nash efficiency coefficient (NSE) were used to evaluate model simulation results 37   Where Q i obs is the runoff observation value, Q i sim is the runoff simulation value, Q obs is the mean observed value, and Q sim is the mean simulated value. When the relative error between simulated and measured runoff is within ±20%, NSE > 0.5, and R 2 > 0.6; and the relative error between simulation and measured sediment is within ±30%, NSE > 0.5, and R 2 > 0.6, the SWAT model is considered consistent with observations and can be used for simulation of the basin 38     www.nature.com/scientificreports www.nature.com/scientificreports/ We can see from the Tables 5-7, simulation results for the upper reaches (Sifang and Tieli stations) are better as compared to those of Qinjia station (lower reaches) because runoff from the upper reaches is abundant, presenting a natural river form. However, downstream water conservancy facilities, including reservoirs and river dams mean that downstream water supply is insufficient, leading to intermittent flow in many places. Hence, upstream simulation results are more accurate results due to the significant influence of human activities downstream. Notwithstanding, model simulation results are in general accord with SWAT model requirements and can be applied to the Hulan River Basin.
SWAT model results. Figures 13 and 14 show the simulated distribution of total N and P. It is apparent that SWAT model simulations of non-point source pollution loading in the upstream (downstream) is relatively low (high). In the Keyin and Numin River sub-basins, the non-point source pollution load of total N is relatively high. Conversely, the non-point source pollution load of total P is relatively high in the Ougen, Yijimi and Xiaohulan River sub-basins.

Discussion
The variation of water chemistry from upstream to downstream (Fig. 2) shows that the dissolution of magnesium minerals gradually increases, while the dissolution of carbonate rocks gradually decreases, indicating a difference in lithology of the tributary source rock or riverbed sediments from upstream to downstream within the Hulan river basin. Due to the self-cleaning function of nitrogen in the water, there is no obvious accumulation of total N in the middle and lower reaches. Because of the poor self-purification function of P in water, total P  Table 7. Simulation evaluation of total phosphorus. www.nature.com/scientificreports www.nature.com/scientificreports/ gradually accumulates downstream, and the water quality deteriorates from Class III to Class V or worse (Fig. 8).
In summary, the surface waters of the Hulan River Basin tend to be medium-weakly alkaline with a low degree of mineralization. HCO 3 -Ca and HCO 3 •Cl-Mg•Na (HCO 3 •Cl-Na•Ca) are the main chemical types, and the ion composition of each region changes regularly. In terms of drinking water safety, total N and P concentrations exceed the standard.
Stratigraphic rock (soil) minerals determine the source of groundwater chemical composition through water-rock interaction, which is the material basis of chemical components in surface water. The atmosphere is filled with CO 2 of different origins, forming a gas-liquid-mineral three-phase system, and chemical reactions of atmospheric precipitation with certain chemical components and soil minerals occur at the contact surface of gaseous CO 2 with water as follows: According to qualitative lithology analysis, the main rock minerals are carbonates and silicates. On the basis of proton (H + ) generation in the water and gas system, water-rock interaction occurs during phreatic water flow through the pores of the unconfined aquifer. The dissolution of carbonate and aluminosilicate minerals provides a source of Ca 2+ and Mg 2+ in the phreatic water, and the dissolution of rock salt provides a source of Na + , K + and Cl − as follows:    Topography controls the spatial distribution of individual ionic components. The water-rock reaction dominated by leaching occurs in the upstream where the hydraulic gradient is large. Under leaching processes, some HCO 3 − type substances in soil enter the river and migrate as runoff to form HCO 3 -Ca type water with a low TDS. The particle size downstream, where the hydraulic gradient is slow, becomes finer, resulting in enhanced evaporation and concentration; hence TDS is gradually increased. As individual tributaries continue to flow into the mainstream, the concentration of mainstream components will change accordingly. For example, the Hulan River main water chemistry type is HCO 3 •Cl-Ca; however, following the merging of the Yijimi River in the middle reaches, the water is of the type SO 4 •HCO 3 •Cl-Ca.
Land use (Fig. 9) and water quality data indicate that water quality in densely populated regions is poor. In particular, the COD content is high at sampling points 26 and 27 on the Gemuke River, 21 and 22 on the Keyin River, and the lower reaches of the Hulan River. Human activities and daily life produce large volumes of sewage which discharges into the water body, leading to a deterioration in water quality. Land use and land cover changes result in temporal and spatial variability in water cycling, quantity, and quality. With the increase in human activity in the Hulan River basin, land cover within the basin has changed from natural vegetation to cultivated land, resulting in higher levels of N and P due to the large-scale use of fertilizer, herbicides, and pesticides. Different types of cultivated land lead to different degrees of N and P pollution. For example, the dry land in Group B is planted with soybean and corn, while Group A is mainly dominated by paddy fields. Consequently, the water quality of Group B is significantly worse than Group A.
Due to the development of tourism and the lack of oversight, the flow of the Yijimi River is severely restricted by fallen trees and prefabricated panels, resulting in the serious deterioration of water quality. Furthermore, a dam was built for the municipal landscape in the Gemuke River within Qing'an County, reducing river flow and causing serious eutrophication and poor water quality downstream.
At present, the evolution and cause analysis of surface water and groundwater quality within coastal areas and lakes, the hydrological charactersitics of which are different from rivers, is also a research hotspot [40][41][42][43] . If future research on water quality can proceed from the scale of the hydrological cycle, such as inland cycles and ocean cycles, more progressive and satisfactory results will be achieved.

conclusion
The TDS of Group B is higher as compared to the Hulan River mainstream, which in turn is higher as compared to Group A. The ion concentrations of the rivers in the basin are gradually enriched downstream. Surface waters of the Hulan River basin display relatively low TDS, and are generally medium-weakly alkaline fresh water. The water chemistry type is dominated by HCO 3 -Ca and HCO 3 •Cl-Mg•Na (HCO 3 •Cl-Na•Ca), and the ion composition of each region changes regularly. In terms of drinking water quality, total N and P exceed water safety standards.
The water chemistry of the basin is mainly controlled by rock weathering. Water ions are mainly composed of silicate mineral weathering products, followed by carbonate mineral weathering products which corresponds well with the geology of the region. The upstream hydraulic gradient is large, and water-rock processes are dominated by leaching. Downstream, particles become finer and TDS is gradually increased under enhanced evaporation conditions and the decreased hydraulic gradient.
Increased human activity in the river basin has altered land cover from natural vegetation to cultivated land, resulting in water quality degradation. The content of N and P is generally high due to the large-scale use of fertilizer, herbicides, and pesticides. The degree of N and P pollution differs according to the type of cultivated land. The non-point source pollution load is relatively low upstream and increases downstream. In the Keyin and Numin River sub-basin, the non-point source pollution load of total N is relatively high. Conversely, in the Ougen, Yijimi, and Xiaohulan River sub-basin, the non-point source pollution load of total P is relatively high.

Data availability
The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.