Simulation of nitrogen and phosphorus pollution in typical agricultural and forested basins as well as relevant reduction effect based on SWAT model

Non-point source nitrogen and phosphorus pollution is a critical threat to aquatic ecosystems and a potential risk to drinking water safety. To precisely control nitrogen and phosphorus pollution in the river basin, in this study, we identified key pollution areas of the river basin, analyzed the main characteristics of pollution sources as well as their contribution to the river basin pollution, and conducted simulation analysis on reduction measures based on the SWAT model. The results showed the following. (1) The simulation effect of the calibrated model was good, and sub-basins 3, 39 and 96 were the key source areas, the main sources of pollution were combined pollution from livestock and poultry breeding and planting industry. (2) Crops had the largest input and output for both nitrogen and phosphorus, the output of which was 6,137.8 t/a and 562.4 t/a respectively. The urban point sources had the highest output rates of nitrogen and phosphorus, 75.7% and 67.5% respectively. (3) With the optimal combination of reduction measures, nitrogen and phosphorus were reduced by 1,438.9 t and 85.3 t respectively, i.e., the reduction rates were the highest. The reduction effect for total nitrogen was better than that for total phosphorus.


INTRODUCTION
The protection of water resources is essential for local social and economic development. To strengthen the protection of water resources in river basins, industrial point source pollution in major river basins of China has been effectively controlled. However, non-point source pollution is still a serious problem, which has significantly compromised water quality, and made the governance of river basins more difficult (Chen et al. ; Wang et al. ). The nonpoint source nitrogen and phosphorus pollutants are mainly from agricultural and urban pollution sources, among which the urban pollution sources include living pollution sources and industrial pollution sources in urban hubs and scattered rural areas. Agricultural pollution sources may be divided into livestock and poultry breeding, agricultural planting, and aquaculture, from which the time, route, quantity and category for pollution are uncertain. However, if their spatial distribution can be reflected accurately, the managers of river basins may accurately locate the key areas for governance. Therefore, some hydrological models, such as the HSPF model, AGNPS model and SWAT model, have been widely applied. The SWAT model takes into account soil type, land type, climate, and vegetation type, and its simu- Nitrogen or phosphorus pollution is closely related to rainfall runoff process, soil type, vegetation type, agricultural production mode (fertilizer application, tillage mode, etc.) and agricultural management measures. At present, many river basins have implemented different management measures, which have caused a great difference in nitrogen and phosphorus load reduction. Generally, the managers take the optimal integrated management measures dependent on local conditions, e.g. the optimal management measures for the Mary River Basin summarized by Richards et al. (), by which the phosphorus load was reduced by 50%. Wojciechowska et al. () studied different management measures and policies in three small basins. In addition, for regions with serious soil and water loss, the prevention and control of soil and water loss and nutrient loss are often considered together in management measures for relevant basins, so as to make a comprehensive strategy (Niraula et al. ; Himanshu et al. ). In this study, a simulation of non-point source nitrogen and phosphorus pollution in the Yongzhou Basin of Xiangjiang River was carried out using local census data of pollution sources (from the second national census of pollution sources carried out recently) and farmland survey data (from the latest farmland quality survey report) based on the SWAT model, and three major goals for basin management and control were realized: (1) the key pollution areas of the river basin were identified; (2) the main characteristics of basin pollution sources and their contribution to the basin pollution were analyzed; (3) specific reduction measures were put forward, and the reduction effect was simulated.
This study may be taken as a reference for effective control of nitrogen or phosphorus pollution in typical agricultural and forested basins, and especially the analysis of the presented key source areas, the apportionment of the sources in the basin and nitrogen and phosphorus pollution control countermeasures may support local initiatives better to protect the source of Xiangjiang River.

Overview of the region
The study region is located on the upper reaches of the Xiangjiang River. Xiangjiang River is the largest river in Hunan Province, and a major tributary of the Yangtze River as well. It originates from Haiyang Mountain, Lingchuan County, Guangxi Zhuang Autonomous Region in the west, and Wild Dog Ridge, Lanshan County, Hunan Province in the east. Its total length is 948 km, and the area of the basin is 94,700 km 2 . As a basin with the most abundant water resources and the highest development and utilization rates, it belongs to the Dongting Lake system and is dominated by a subtropical monsoon climate, with long-term average annual precipitation of 1,200-1,900 mm, and annual average temperature at 17.6-18.6 C. Red soil and yellow soil are its main soil types, with perennial average sunshine between 101.5 and 113 kcal/cm 2 .

Planting conditions
In this study region, double-season rice, maize, and oilseed rape are the main crops. To simulate soil nitrogen and phosphorus loss with farmland runoff, it is assumed that the double-season rice is planted in paddy fields and singleseason maize in upland fields every year. The main fertilizers are nitrogen fertilizers and compound fertilizers. Urea with nitrogen content of 46% is the main nitrogen fertilizer while for the compound fertilizers, mainly a fertilizer containing nitrogen, phosphorus and potassium, 15% content for each, is adopted. For double-season rice and maize, there are base fertilizer and topdressing fertilizer for each season of rice. The application of fertilizers for the main crops is as shown in Table 1.

Database construction
The database for the SWAT model was divided into two categories: spatial database and attribute database. All spatial data were converted into the projection coordinate system: WGS_1984_UTM_Zone_49N; after that the data of meteorological, land use type, soil type and the daily load of urban point sources and livestock and poultry breeding pollution sources were used for simulation, likewise the hydrology and water quality data were used for model calibration and verification respectively. See Table 2 for the names, parameters and sources of the main data.

Model construction and land use distribution
In this study, ARCSWAT was used to perform DEM image recognition, calculate the catchment area of the basin, extract the river network and delimit the sub-basin. Based on the thresholds of land use type, soil type and slope division set as 10%, 10% and 15%, the total basin area was 18,633.6 km 2 , and the basin was divided into 115 subbasins and 818 HURs. There were eight types of land in the study region. The land use types that covered the largest areas were forest land, paddy field and dry land, of which the areas were 11,393.5 km 2 , 3,186.8 km 2 and 2,057.2 km 2 respectively, accounting for 89.3% of the total basin area of the study region together. The study region is a typical region dominated by agriculture and forestry, as shown in

Soil type distribution
The soil type map reflects the spatial distribution of soil in the study region. This study utilized HWSD data downloaded from the FAO website to extract a 1:1,000,000 soil type map of Yongzhou, effective numerical range: 11,333-11,927. Some attribute data can be found through the HWSD_DATE database, and the rest of the data can be calculated by SPAW software. Thus, the soil database can be established ( Figure 3).

MODEL CALIBRATION AND EVALUATION OF THE MODEL Model calibration
In this study, the SUFI method of SWAT-CUP software was used for sensitivity analysis of model parameters, in which 71 parameters related to runoff and nitrogen or phosphorus cycle were selected for iterative modelling with the Global Sensitivity Analysis method. The parameters with the highest sensitivity are as shown in Table 3.

Evaluation of the model
In the study, the coefficient of determination (R 2 ) and Nash-Sutcliffe efficiency coefficient (Ens) were selected as the indexes for evaluation of the model. The coefficient of determination (R 2 ) reflects the correlation between simulated value and actual value of the model. The larger R 2 , the better the consistency between the simulated value and the actual value. The Nash-Sutcliffe efficiency coefficient (Ens) represents the overall efficiency of model simulation. The closer the value to 1, the better the suitability of the model simulation. The formulas for calculation of R 2 and Ens are as follows:    Table 4 and Table 5. The maximum sub-basin inputs of total nitrogen and total phosphorus were 796.01 t and 88.49 t respectively, and the top three sub-basins for total    basin of Xiaoshui River were less polluted, the total nitrogen output intensity was less than 2.90 kg·ha À1 , and the total phosphorus output intensity was less than 0.28 kg·ha À1 .
In the north basin, sub-basin 35 at the upper reaches of Xiangjiang River, sub-basins 21 and 39 at the lower reaches of Xiangjiang River, tributaries 2 and 3 of Luhong River, and sub-basin 96 in the southern basin were seriously polluted, with total nitrogen output intensity greater than 11.09 kg· ha À1 , and total phosphorus output intensity greater than 1.11 kg·ha À1 . Obviously, for sub-basins 3 and 39 in the north, both inputs and outputs of total nitrogen and total phosphorus were very high while for sub-basin 96 in the south, the key source area for nitrogen and phosphorus treatment in this basin, the outputs of total nitrogen and total phosphorus were very high.  Note: Crop-planting emissions in each sub-basin ¼ emissions load for planting industry × farmland area/total area of sub-basin.

Analysis of output rate and contribution of nitrogen and phosphorus pollution
The pollution sources in the study region were divided into livestock and poultry breeding, crop cultivation and urban point sources. As shown in Table 6, for both total nitrogen input and total phosphorus input, crop cultivation ranked   Table 7, when the fertilizer reduction rate was set at 5%, the total nitrogen and total phosphorus of the basin were reduced by 288.4 t and 14.6 t respectively. Both reduction rates were greater than 20% for urban point sources and 10% for livestock and poultry breeding. When the fertilizer reduction rate was set at 20%, the total nitrogen and total phosphorus in the basin  and phosphorus output rates for pollution from urban point sources were the highest, 64.9% and 65.1% respectively, which indicated that chemical fertilizers are the most important source of non-point source pollution, and the application control of chemical fertilizers should also be combined with the control of livestock and poultry breeding and urban point source control.
4. In terms of source control, the reduction effect of chemical fertilizer application control was the best, and total nitrogen and total phosphorus could be reduced by 1,438.9 t and 85.3 t respectively, reduction rates: 7.5% and 3.9%. The effect for total nitrogen reduction was better than for total phosphorus reduction.
Nitrogen or phosphorus pollution is closely related to rainfall runoff process, soil type, vegetation type, agricultural production mode (fertilizer application, tillage mode, etc.) and agricultural management measures. Considering that terrace farming in typical agricultural and forestry basins is a common phenomenon, and soil erosion caused by terrace farming is closely related to water-body nitrogen and phosphorus pollution, it is therefore worthwhile to study the

DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.