Water resources are a key factor restricting human survival and social and economic development. The Miyun Reservoir, which is the only surface drinking water source in Beijing and the water storage reservoir of the South-to-North Water Diversion Project, plays a decisive role in ensuring water-use sustainability of the capital. This study focused on the Baihe River Basin, which is one of the important inflows of the Miyun Reservoir. The mathematical data on the climate and hydrological changes of the Baihe River Basin over the past 30 years were obtained, and the evolution law of the Baihe River runoff and its influencing factors was analyzed. Our analysis used the data obtained via multi-source meteorology to establish the Soil and Water Assessment Tool (SWAT) model of the Baihe River Basin. Sequential Uncertainty Fitting version 2 (SUFI-2) and Generalized Likelihood Uncertainty Estimation (GLUE) were used to simulate runoff from 1990 to 2017 on a monthly scale, along with sediment, total nitrogen (TN), and total phosphorus (TP) contents from 1990 to 2010. Then, the parameter uncertainty of the SWAT model and the applicability of the SWAT model in the Baihe River Basin were determined. The land-use transfer matrix showed that land-use changes are complex and the transformation forms are diverse. The simulation results showed that the transformation of land-use patterns in 1990 and 2010 had slight impacts on the hydrologic and water quality of the basin. By identifying the key source areas of non-point source pollution in the basin, the impacts of three management measures on the hydrological and water quality of the basin were simulated. The results showed that the reductions in flow rates, sediment content, TN, and TP (4.37, 31.93, 23.13, and 35.30%, respectively) obtained through terrace engineering were significantly better than those acquired via stubble mulch and contour planting. Additionally, this study uses the Sixth International Coupling Mode Comparison Program (CMIP6) climate change scenario and the BCC-CSM2-MR climate model coupled with the corrected SWAT model to predict future runoff, thereby providing references and suggestions for managing the Baihe River Basin and the Miyun Reservoir.

  • SUFI-2 and GLUE are used to simulate the runoff, sediment, TN, and TP.

  • The parameter uncertainty of the SWAT model and applicability is analyzed.

  • Land-use changes are complex and transformation forms are diverse.

  • By identifying key source areas of non-point source pollution, based on land use change, the impacts of four management measures are simulated.

  • The CMIP6 climate change scenario is used to predict future runoff.

Graphical Abstract

Graphical Abstract
Graphical Abstract

In the twenty-first century, water resources are a key factor restricting human survival and socio-economic development. The Miyun Reservoir is the largest reservoir in North China. Since 1960, it has been used to supply industrial, agricultural, and domestic water to Beijing, Tianjin, and Hebei. Hebei and Tianjin stopped receiving water in 1982 because of dwindling water supplies and increasing pollution (Hu et al. 2007). As the only surface drinking water source in Beijing and the water storage reservoir for the middle route of the South-to-North Water Transfer Project, the Miyun Reservoir plays a key role in ensuring water-use sustainability of the capital. The water quantity, quality, and changes in its upstream rivers have important and far-reaching impacts on the Miyun Reservoir and its management (Zhao et al. 2017). Studies have shown that non-point source pollution is the main cause of water quality deterioration in important lakes such as Taihu Lake and Chaohu Lake (Lei et al. 2019; Wang et al. 2019; Yan et al. 2020). In addition to nitrogen, phosphorus, and other pollution sources, sediments are also a common source of non-point source pollution in China (Wang et al. 2002; Xing et al. 2005; Zhang et al. 2007). Therefore, it is necessary to conduct targeted research on the inflow runoff of the Miyun Reservoir.

The distributed hydrological model successfully reflects the great impact of basin spatial heterogeneity on the water environment process, which was lacking in the early hydrological and water quality models; this greatly promotes the development of water quality models (Feng et al. 2018; Saila 2020; Athira & Sudheer 2021). Presently, the Mike System Hydraulic European (MIKE SHE), hydraulic simulation program FORTRAN (HSPF), and Soil and Water Assessment Tool (SWAT) are widely used as representative watershed comprehensive water quality models. The SWAT model is popular for non-point source pollution and redevelopment. Although the SWAT model shows high simulation accuracy in research worldwide, the actual situation in different regions is different. It is necessary to analyze the applicability of the SWAT model in different regions.

As an approximate mathematical description of the complex watershed water environment, the model cannot reproduce the watershed system process at 100%; therefore, there is great uncertainty in watershed hydrology and water quality simulations (Beven 1993). This is mainly due to the uncertainty of the model structure, model parameters, and input data (Moges et al. 2020). The model structure is incomplete in expressing the physicochemical properties and processes of the water environment, which lead to uncertainty; the errors caused by sampling and measurement (that is, the errors existing in the input data) will also affect the model parameters, resulting in model uncertainty. Although calibration and validation of the model can ensure the reliability of the simulation, there are many parameters in the comprehensive water quality model; the correlation and sensitivity of the parameters will affect the uncertainty of the parameter value, resulting in an increase in the uncertainty of the simulation (Lennart et al. 2019). The current technical methods cannot remove this uncertainty; therefore, it is necessary to adopt reasonable and effective evaluation methods to evaluate and analyze the uncertainty of model parameters and simulation results.

Studies have shown that land use, water and soil conservation projects, and water resource development and utilization in a basin will greatly affect its hydrological and water quality (Wang et al. 2021). With social development and the implementation of a policy converting farmland to forest, the land-use mode of the Baihe River Basin has changed greatly since 1990; thus, the cultivated land area and grassland area have decreased significantly, forest land area has gradually increased, vegetation coverage has become higher, the proportion of urban land has increased, and tourism has developed rapidly. Therefore, it is of great significance to study the impact of land-use changes on hydrology and water quality in the Baihe River Basin.

Global warming leads to frequent extreme weather and disasters, which significantly impact production and life. Increasing research on the impact of climate change on hydrology and water resources shows that humans pay attention to such issues (Ibrahim 2018; Bouras et al. 2019; Chen et al. 2020; He et al. 2020; Khamidov & Khamraev 2020; Wang et al. 2020). To date, the World Climate Research Program (WCRP) has held five international coupling mode comparison programs (CMIP), and the results of the Sixth International Coupling Mode Comparison Program (CMIP6) have been released. The applicability and prediction effect of the CMIP6 future scenario in different regions must be proven and tested by many studies, especially for runoff prediction (Gao et al. 2021; Song et al. 2021). China's National Climate Center has participated in the program for many years and has continuously updated the improved version of the BCC climate model. BCC-CSM2-MR is one of the three latest model versions of the National Climate Center's participation in the CMIP6 program (Xin et al. 2019; Huang 2020). Many scholars have achieved good results using BCC series climate models (Chen et al. 2011; Dong et al. 2013; Li et al. 2020). Under different emission scenarios and global climate models, the change in water resources in the same region may also have different trends. Therefore, much research is needed on the applicability of climate models.

The upstream water volume of the Miyun Reservoir has decreased sharply in recent years, and the regulation and storage functions of the reservoir are weak. The situation has eased since undertaking the ‘south water’ of the Middle Route Project of South-to-North Water Transfer; however, the management and treatment of upstream runoff cannot be relaxed. In order to better understand the hydrology and water quality of the Baihe River Basin and to provide suggestions for water resource management, this study applied the best management practices (BMPs) to this basin based on land-use change and used CMIP6 data for the first time to quantitatively project future meteorology and runoff. The main objectives of this study are to: (1) analyze climate and hydrological changes within the watershed; (2) simulate the hydrologic and water quality of the Baihe River Basin using the SWAT model; (3) analyze the impact of land-use change and BMPs on hydrologic and water quality, and (4) discuss the response of runoff to future climate change.

Study area

The Baihe River Basin is located in North China (40 °28′N–41 °30′N, 115 °26′E–116 °57′E) (Figure 1). The Baihe River is one of the two main inflow reaches of the Miyun Reservoir. It originates from Jiulongquan, Guyuan County, Hebei Province and converges with tributaries such as Chongli County, Fengning County, and Chicheng County in Hebei Province in Chicheng County, Hebei Province; finally, it flows into Beijing. Tributaries such as Huairou District and Yanqing District flow into the Miyun Reservoir through the Miyun District, with a total length of more than 250 km. The total drainage area is 8528.1 km2, accounting for approximately 60% of the catchment area upstream of the Miyun Reservoir (Xie et al. 2009). The mountain area in the basin is vast, with high mountains, steep slopes, vertical and horizontal gullies, high terrain in the northwest and low terrain in the southeast, and ground elevation between 181 and 288 m. The land-use modes were mainly forest land, cultivated land, and grassland. In recent years, the area under coniferous and broad-leaved mixed forests has gradually increased due to ecological construction policies; meanwhile, the area of grassland and cultivated land has decreased daily, and the proportion of urban land and industrial land has increased. The main crops are corn, wheat, rice, soybean, peanut, and sweet potato.

Figure 1

Map of the study area.

Figure 1

Map of the study area.

Close modal

The basin has a typical temperate continental monsoon climate, with an annual average temperature of 7.3–10.3 °C. Owing to landform influence, the temperature difference in the basin is large. The winter is cold and dry, northwest wind prevails, and summer is hot and rainy. Precipitation is mainly concentrated from June to September, accounting for approximately 75–80% of the annual precipitation. It is prone to meteorological disasters such as rainstorms and hail; the interannual distribution of precipitation is uneven. The annual average precipitation was 511.9 mm, and the precipitation in 1993 was the lowest (227.08 mm). The interannual difference between high and low runoff in the Baihe River Basin is significant.

The soil types in the Baihe River Basin are diverse and include cinnamon, brown, and gray cinnamon soils. Among them, cinnamon soil is dominant, accounting for up to 30% of the total soil content; it is mostly distributed in low mountain and hilly areas. Brown soil accounts for approximately 10% of the total content and is mostly distributed in the middle mountain forest area above 800 m. The vertical distribution of the soil was evident. There are great differences in the physical soil properties between the east and west of the basin. The soil in the east is predominantly a massive or flake structure, which is unsuitable for storing water and fertilizers. The west is a rocky mountainous area with thin soil that is vulnerable to wind and rain erosion.

At present, monitoring shows that the water body of the Miyun Reservoir can reach first-class water quality as the surface water source of domestic and drinking water, and the inflow water quality of the Baihe River Basin is basically stable and maintained in class II water bodies. The results show that the main sources of water pollution in the Baihe River Basin are upstream planting, rural agricultural waste, and enterprise sewage. The non-point source pollution is serious and can easily cause water and soil loss during the flood season and increase the concentration of pollutants in the river, while the water quality is good in the dry season. Therefore, it is necessary to supervise chemical fertilizer application and waste treatment.

In this study, the runoff data from the Zhangjiafen station (which is a hydrological observation station before the Baihe River) are stored, and the water quality monitoring data such as sediment, TN, and TP of the Baihe main dam were used for simulation research. Table 1 shows the data sources used in this study.

Table 1

Basic data sources used in the study

Data nameSpatial resolution(m)Time ResolutionTime-series lengthData sources
Runoff data Monthly 1990–2017 Zhangjiafen hydrological station 
Water quality data Monthly 1990–2010 Water quality monitoring data of Baihe main dam 
Meteorological data Daily 1990–2017 Data of China's national ground meteorological stations 
DEM 30 Geospatial data cloud platform of Chinese Academy of Sciences 
Land-use data 30 Data center of resources and environment science, Chinese Academy of Sciences 
Soil data 30 China cold and arid area data management center 
Data nameSpatial resolution(m)Time ResolutionTime-series lengthData sources
Runoff data Monthly 1990–2017 Zhangjiafen hydrological station 
Water quality data Monthly 1990–2010 Water quality monitoring data of Baihe main dam 
Meteorological data Daily 1990–2017 Data of China's national ground meteorological stations 
DEM 30 Geospatial data cloud platform of Chinese Academy of Sciences 
Land-use data 30 Data center of resources and environment science, Chinese Academy of Sciences 
Soil data 30 China cold and arid area data management center 

Method

SWAT model establishment

This study is based on a digital elevation model (DEM), referring to the surface information extracted from the DEM. The resolution of the land-use data is 30 × 30 m. After further matching and cutting with the study area and matching with the land-use data classification of the SWAT model, land use was reclassified according to the national secondary type classification table of land resources. After reclassification, the land-use types in the study areawere divided into eight categories: forest land, grassland, cultivated land, water area, urban land, residential land, industrial land, and unused land. Some data can be calculated to meet the needs of the SWAT model for soil data. Specific data include daily maximum and minimum temperatures, daily precipitation, daily relative humidity, daily solar radiation, and daily average wind speed. Among them, temperature and precipitation data are the necessary data, and all other data can be generated through the weather generator. In this study, daily observation data from Fengning, Huailai, Miyun, and Chengde stations from 1987 to 2017 were selected, as shown in Figure 1.

SUFI-2 method

The Sequential Uncertainty Fitting Version 2 (SUFI-2) algorithm is a sequential uncertainty fitting algorithm based on a Bayesian framework, which can represent the difference between simulated and observed values, and the uncertainty can be determined through their sequence and fitting process (Rokhsare et al. 2009). In this method, the uncertainties of the input process, model structure, parameters, and observed values are considered. The global search method is used for combinatorial optimization and uncertainty analysis, while the Latin hypercube sampling method is used to deal with large parameter groups (Michael et al. 2011; Li et al. 2017). SUFI-2 is a semi-automatic method that makes calibration more efficient, making it easier to perform the calibration process within an achievable time range. This has aroused more interest among researchers because if calibration is manual, adding a large number of parameters to the model can increase both the complexity of the process and the number of required calculations.

GLUE method

The Generalized Likelihood Uncertainty Estimation (GLUE) algorithm partially introduces the generalized likelihood uncertainty estimation method. Its significance is to allow a non-unique parameter set to exist in an over-parameterized model. This uncertainty analysis method is inspired by importance sampling and regional sensitivity analysis methods. Similarly, the GLUE method explains all factors that may lead to uncertainty. In the calculation process, a generalized likelihood measure was first generated. Many randomly selected parameter sets are evaluated and weighted by comparing their given thresholds and likelihood measures. The quantile of the cumulative distribution of the weighted behavior parameters is described as the uncertainty. In this study, Nash–Sutcliffe efficiency (NSE) was used to measure the model efficiency. The three main steps of the GLUE method are as follows:

Step 1: The ‘generalized likelihood measure’ L (θ) is defined. Then, a large number of parameter sets is randomly selected from the a priori distribution, and each parameter set is categorized as ‘behavior’ or ‘non-behavior’ by comparing the ‘likelihood measure’ with the given threshold.

Step 2: Each behavior parameter is given a ‘possibility weight’ according to the following:
(1)
where n is the number of behavior parameter sets.

Step 3: The prediction uncertainty is described as the prediction quantile from the realized cumulative distribution according to the weighted behavior parameter set. In this study, the NSE was selected as the model efficiency measurement.

Mann–Kendall mutation test

The Mann–Kendall mutation test (M–K test) is a nonparametric statistical test proposed and studied by Mann and Kendall to detect sequence trends. This method is easy to calculate, has no specific distribution requirements for samples, and is not affected by individual outliers. The method is widely used in non-normal distribution data, such as time-series change trends of temperature, precipitation, and runoff (Tang et al. 2012). In this method, the standard normal distribution UF of sequential time series and the standard normal distribution UB of reverse time series are calculated, plotted, and compared with the standard critical value (Zhang 2014a, 2014b). The specific principle is as follows:

There is a set of measured time-series data {xi|i = 1, 2,…, n}, and a new series is constructed using the following equation:
(2)
where dk is the value of the new series, and mi is the cumulative number of samples with xi > xj (ij ≥ 1).
The mean and variance of dk are defined as follows:
(3)
(4)
where E(dk) is the mean of dk, and var(dk) is the variance of dk.
Assuming random independence of time series, the statistic UFk is defined as follows:
(5)

Similarly, a statistical value UBk is calculated for the inverse time-series data {xi|i = n, n − 1,…, 1}. Given a significance level of a, the corresponding critical value t can be obtained. A UFk value greater than 0 indicates an upward trend, while a UFk value less than 0 indicates a downward trend. When UFk exceeds the critical line, the trend is evident. If the UFk and UBk curves intersect, and the intersection point falls between the critical lines, then the moment corresponding to the intersection point is the time that mutation begins.

Global climate model data and scenarios

In this study, BCC-CSM2-MR climate model data were selected to predict the future climate and runoff in the Baihe River Basin using 1987–2010 as the historical period and 2022–2064 as the future period. Although the BCC-CSM2-MR model can reconstruct actual precipitation in the Baihe River Basin during 1987–2010, a gap remains between the BCC-CSM2-MR model and the measured data. In this study, the daily bias correction (DBC) method (Chen et al. 2013; Feng et al. 2021) was used to correct the precipitation and temperature output in the historical period of the BCC-CSM2-MR climate model, and the correction coefficient was applied to the climate elements in the future period. After the DBC method correction, the relative errors of the average precipitation, maximum temperature, and minimum temperature of the BBC model in the historical period were reduced from 17.49, −35.99, and −104.52% to 2.97, −0.28, and −1.82%, respectively, compared with the measured data.

Climate scenarios, which are based on scientific assumptions about future states of the climate system, mainly consider factors other than the climate model itself and thus may impact climate change simulations. Differences in the factors and variable settings in different scenarios can greatly affect the simulation of climate elements. In this study, three scenarios, SSP1-2.6 (low emission scenario), SSP2-4.5 (medium emission scenario), and SSP5-8.5 (high emission scenario), were selected in the context of actual social development. SSP1-2.6 represents the combined effects of low social vulnerability, low mitigation stress, and low radiative forcing; SSP2-4.5 represents a combination of moderate social vulnerability and moderate radiative forcing, and SSP5-8.5 is the only shared socio-economic pathway that can achieve an anthropogenic radiative forcing of 8.5 W/m2 by 2100 (Zhang et al. 2019a).

BMP scenario setting

The Baihe River Basin is a typical non-point source pollution area. The SWAT model is currently a common method for simulating and evaluating management measures in non-point source pollution research. This study combines the SWAT model with the BMPs, which are a series of engineering and non-engineering measures widely used to control agricultural non-point source pollution (Boufala et al. 2021). In the SWAT model, BMP implementation can be simulated by changing the values of one or more specific parameters in the files, such as .mgt in the ‘sub-basins data’ module.

This article combines the actual situation in the Baihe River Basin, with measures taken in three types of management scenarios (Table 2). Research has shown that using protective farming measures can reduce pollutants and land disturbance (Xu et al. 2020). Therefore, this study chose two non-engineering methods (stubble mulching and contour planting), as well as one engineering measure (terrace farming). Land with a slope greater than 25° is unsuitable for terrace farming; therefore, this study set up terraces for cultivated land with slopes less than 25°.

Table 2

Measures and parameter setting of different scenarios

BMPsScenarioDescription of measuresParameter adjustment
Initial S0 None 
Non-engineering measures S1 Stubble mulching Add ‘Harvest Only’ to .mgt, Subtract 2 from the original CN value; USLE_P = 0.29, USLE_C = 0.7, OV_N = 0.3 
S2 Contour planting Subtract 3 from the original CN value; USLE_P = 0.5 at slope 0–5°; USLP_P = 0.7 at slope 5–25°; USLP_P = 1 at slope above 25° 
Engineering measures S3 Terrace farming (within 25°) Subtract 3 from the original CN value; USLE_P = 0.5, SLSUBBSN = 10 at slope 0–5°; USLE_P = 0.7, SLSUBBSN = 5 at slope 5–25° 
BMPsScenarioDescription of measuresParameter adjustment
Initial S0 None 
Non-engineering measures S1 Stubble mulching Add ‘Harvest Only’ to .mgt, Subtract 2 from the original CN value; USLE_P = 0.29, USLE_C = 0.7, OV_N = 0.3 
S2 Contour planting Subtract 3 from the original CN value; USLE_P = 0.5 at slope 0–5°; USLP_P = 0.7 at slope 5–25°; USLP_P = 1 at slope above 25° 
Engineering measures S3 Terrace farming (within 25°) Subtract 3 from the original CN value; USLE_P = 0.5, SLSUBBSN = 10 at slope 0–5°; USLE_P = 0.7, SLSUBBSN = 5 at slope 5–25° 

Evaluation index

The indicators for evaluating the degree of fit significantly affected how model performance was judged. In this study, the certainty coefficient (R2), NSE, root mean square error (RMSE), RMSE observations standard deviation ratio (RSR), and mean absolute error (MAE) were selected to evaluate the model simulation effect. The closer the R2 value is to 1, the better the fit of the simulated value of the model to the actual observed value. The prediction ability of the NSE evaluation model indicates the fitting effect between the simulated and observed values. The RSR approaching zero shows better performance. The MAE represents the average value of the absolute error between the predicted value and the observed value. The RMSE represents the deviation between the predicted value and the observed value, which is greatly affected by outliers. The different values of NSE and RSR can evaluate how satisfactory the model results are, with the closer NSE is to 1 and RSR is to 0, indicating that the model performs better (Moriasi et al. 2007; Burgan & Aksoy 2022). The calculation formula for the above evaluation indices is as follows:
(6)
(7)
(8)
(9)
(10)
where represents the measured value (m3/s); represents the simulated value (m3/s); represents the measured average value within the simulation time (m3/s); represents the simulation average value within the simulation time (m3/s), and n represents the total number of simulated or measured values.

Statistical analysis of watershed climate and hydrology

Interannual precipitation variation

The annual precipitation distribution in the Baihe River Basin from 1990 to 2017 is shown in Figure 2(a). In recent years, precipitation in the Baihe River Basin has mainly occurred between June and August. Precipitation in these three months accounts for 65.92% of the annual precipitation; peak precipitation occurs in July, accounting for nearly one-third of the annual precipitation (28.81%). The months with the least precipitation are December, January, and February, with the lowest precipitation occurring in January (1.9 mm or 0.37% of the annual precipitation).

Figure 2

(a) Annual distribution of precipitation in the Baihe River Basin. (b) Distribution map of annual precipitation in different years. (c) Annual precipitation variation and 5-year moving average.

Figure 2

(a) Annual distribution of precipitation in the Baihe River Basin. (b) Distribution map of annual precipitation in different years. (c) Annual precipitation variation and 5-year moving average.

Close modal

The 1990–2017 period can be divided into three periods: 1990–1999 (Stage I), 2000–2009 (Stage II), and 2010–2017 (Stage III). The precipitation distribution characteristics were analyzed for the three periods over time. Figure 2(b) shows that an evident change in precipitation distribution within a year represents a variation in the precipitation peak. The precipitation peak was the largest in the first stage and decreased by nearly 70 mm in the second stage. In the third stage, the precipitation peak decreased by approximately 30 mm compared to the first stage and the month in which the precipitation peak appeared also changed. The precipitation peak in the second stage was delayed from the multi-year averaged precipitation peak month (July) to August, which affected the occurrence of runoff.

The average annual precipitation of the Baihe River Basin over the past three decades from 1990 to 2017 was 510.2 mm; despite great differences in the interannual precipitation, no significant mutation trend was observed by the M–K trend test. As shown in Figure 2(c), the maximum annual precipitation in the Baihe River Basin was 645.6 mm in 1990; the minimum precipitation was 383.9 mm in 2009. The annual precipitation in the Baihe River Basin shows a linear decreasing trend. The interannual variation of annual precipitation was analyzed using the 5-year moving average method. It was found that the precipitation first showed a decreasing trend, then an increasing trend over the past 30 years, a decreasing trend from 1990 to 2001, and an increasing trend from 2002 to 2017.

Analysis of interannual runoff characteristics

The annual distribution associated with the inner diameter flow of the basin was uneven, which is closely related to the seasonal distribution of precipitation. Accordingly, this distribution showed ‘double peaks’. As shown in Figure 3(a), the temperature gradually rose in March, causing the snow and surface frozen soil to melt and supplement the surface runoff, which resulted in the first runoff peak. After April and May, vegetation began to grow, farmers began to cultivate, water consumption increased, and runoff decreased; July–September was the main flood season of the basin, with a runoff of 11–21.1 m3/s that accounted for 51.14% of the annual runoff.

Figure 3

(a) Monthly average runoff. (b) Runoff nonuniformity coefficient from 1990 to 2017. (c) Annual distribution curve of runoff in different years of the Baihe River Basin.

Figure 3

(a) Monthly average runoff. (b) Runoff nonuniformity coefficient from 1990 to 2017. (c) Annual distribution curve of runoff in different years of the Baihe River Basin.

Close modal

The calculation results associated with the annual distribution nonuniformity coefficient of runoff in the Baihe River Basin are shown in Figure 3(b). The figure shows inter-generational changes in the annual uneven coefficient of runoff distribution in the Baihe River Basin over the past 30 years, with obvious fluctuations runoff nonuniformity coefficient ranging from 0.2 to 1.9. The maximum and minimum values were observed in 1998 and 1999, respectively, and the uneven coefficient showed a decreasing trend over the past 28 years. The coefficient was more unstable before 1999 and changed significantly. From 2001 to 2009, the uneven coefficient distribution was more uniform. This shows that the annual runoff distribution prior to 1999 was different and uneven, the runoff difference between months decreased after 1999, and the distribution was more uniform than before.

The annual distribution changes of runoff in the Baihe River Basin were analyzed in three stages, as shown in Figure 3(c). The study area is an area with a high impact of human activities, and runoff is affected by human water withdrawal, reservoir construction, and land-use type changes. From the M–K trend analysis, we know that 1999 is the abrupt change point of runoff, and after 2000, the runoff volume in the study area is significantly reduced by human activities. The peak value of runoff in the second and third stages was significantly lower than in the first stage, only half of the peak value in the first stage, approximately 15 m3/s; Compared with the single peak in the first stage, there were obvious double peaks in the second stage, which appeared in March and August, respectively, while the double peaks in the third stage were distributed in April and July, the peak interval shortened, and the summer runoff peak occurred earlier than August in the first and second stages.

As shown in Figure 4(a), the linear correlation between annual precipitation and annual runoff from 1990 to 2017 in the Baihe River Basin was approximately 0.58. From the intra-year distribution in Figure 4(b), the change trend of annual runoff was basically the same as the precipitation trend, first decreasing and then increasing gradually, but the magnitude of the change was quite different. The consistent change trend of precipitation and runoff shows that in arid and semi-arid areas such as the Baihe River Basin, precipitation distribution within one year significantly impacted the runoff generation of the entire basin. Figure 4(c) shows that the runoff and precipitation changes were consistent in the Baihe River Basin, showing a decreasing trend first and then an increasing one. The lowest runoff value was observed in 2002.

Figure 4

(a) Linear correlation between precipitation and annual runoff. (b) Intra-year distribution curve of runoff in different years. (c) Runoff in the past 30 years and 5-year moving average.

Figure 4

(a) Linear correlation between precipitation and annual runoff. (b) Intra-year distribution curve of runoff in different years. (c) Runoff in the past 30 years and 5-year moving average.

Close modal

M–K mutation analysis

The M–K method was used to test and analyze the trends and mutations in annual runoff in the Baihe River Basin from 1990 to 2017. The significance level α was set to 0.05, while the corresponding critical value t was 1.96. The mutation test results for the runoff change are shown in Figure 5(a). The runoff of the Baihe River Basin increased and decreased from 1990 to 1999, with relatively large volatility. In 1990–1992, 1996, and 1998, the runoff showed an upward trend, while it showed a downward trend in 1993–1995 and 1997. Therefore, the positive sequence statistics (UF) and reverse sequence statistics (UB) from 1990 to 1999 had frequent intersections with no mutation significance. Since 1999, the runoff of the Baihe River Basin has shown an obvious decreasing trend, with no intersection between the positive and reverse sequence statistics and no sudden changes.

Figure 5

(a) Analysis of abrupt runoff changes over the past 30 years. (b) Precipitation–runoff double accumulation curve.

Figure 5

(a) Analysis of abrupt runoff changes over the past 30 years. (b) Precipitation–runoff double accumulation curve.

Close modal

Figure 5(b) shows the double accumulation curve of precipitation and runoff in the Baihe River Basin over the past 30 years. The figure clearly shows that the slope of the cumulative curve changed significantly in 1999. The slope was approximately 0.72 from 1990 to 1998 and 0.41 from 1999 to 2017, thereby decreasing by approximately 43.17%. This shows that human activity has significantly impacted runoff. Combining the M–K trend analysis and the double accumulation curve, 1999 was identified as the point of change in the Baihe River Basin runoff over the past 30 years.

The runoff and sediment, TN, and TP of the Baihe River Basin were placed on the same time scale (1990–2010) for mutation test analysis. Similarly, the significance level α was set to 0.05, and the corresponding critical value t was 1.96. The results are shown in Figure 6. The changes in runoff were consistent with the above analysis. There were frequent fluctuations from 1990 to 1999, yielding multiple intersections with no abrupt significance. The last intersection occurred in 1999, and sediment content increased from 1990 to 1992. The trend has been decreasing since 1993, and a significant mutation point appeared in 1999. TN showed an increasing trend from 1990 to 2000 (barring 1992) and has followed a decreasing trend since 2000, with a significant break point in 1999. TP showed an increasing trend from 1990 to 1999 with the exception of 1993 and showed a decreasing trend since 1999, with a significant break point in 1999. According to the changing trends and mutation analyses of the four variables, the changes in the Baihe River Basin from 1990 to 2010 can be summarized as follows: (1) 1999 was the significant mutation point; (2) although each variable fluctuated greatly from 1990 to 1999, the variation trend was not significant; (3) and all variables showed evident decreasing trends from 2000 to 2010.

Figure 6

Analysis of abrupt changes in runoff, sediment, TN, and TP in the Baihe River Basin from 1990 to 2010.

Figure 6

Analysis of abrupt changes in runoff, sediment, TN, and TP in the Baihe River Basin from 1990 to 2010.

Close modal

As shown in Figure 7, the sudden change in precipitation in 1999 greatly reduced the impact of precipitation on sediment transport; the slope decreased from 0.08192 to 0.00205, representing a reduction of approximately 97.5%. Runoff was reduced due to a decrease in precipitation, and the impact of runoff on sediment transport was also reduced. The slope was initially reduced from 0.11464 to 0.00437, representing a reduction of nearly 96.19%. The impact was major and significantly reduced the amount of runoff, sediment, and soil erosion after 2000.

Figure 7

(a) Precipitation–sediment transport double cumulative curve and (b) runoff–sediment transport double cumulative curve.

Figure 7

(a) Precipitation–sediment transport double cumulative curve and (b) runoff–sediment transport double cumulative curve.

Close modal

In addition, we also calculated the numerical characteristics of monthly runoff, sediment, TN, and TP before and after the mutation point, as shown in Table 3. The amount of sediment in 1990–1999 is much larger than that in 2000–2010; therefore, the standard deviation and variance of the former are also very large. For runoff and TN, the average values in 2000–2010 are also reduced by half compared to 1990–1999. On the contrary, the average value of TP in 2000–2010 is larger than that in 1990–1999. The Cs of the four hydrological water quality elements decreases significantly after the abrupt change point.

Table 3

Monthly scale statistical characteristics of the Baihe River Basin over 1990–1999 and 2000–2010 durations, respectively

1990–1999
2000–2010
AverageSDCvCsAverageSDCvCs
Runoff (m3/s) 11.60 19.45 378.15 4.56 6.37 6.97 48.64 3.35 
Sediment (103t) 31.80 132.60 7,582.09 6.38 0.79 2.80 7.81 4.26 
TN (106kg) 3.26 5.43 29.45 4.62 1.47 1.65 2.73 2.39 
TP (104kg) 0.48 1.03 1.06 4.79 1.81 2.15 4.63 2.99 
1990–1999
2000–2010
AverageSDCvCsAverageSDCvCs
Runoff (m3/s) 11.60 19.45 378.15 4.56 6.37 6.97 48.64 3.35 
Sediment (103t) 31.80 132.60 7,582.09 6.38 0.79 2.80 7.81 4.26 
TN (106kg) 3.26 5.43 29.45 4.62 1.47 1.65 2.73 2.39 
TP (104kg) 0.48 1.03 1.06 4.79 1.81 2.15 4.63 2.99 

SD, standard deviation; Cv, coefficient of variation; Cs, coefficient of skewness.

Basin hydrologic and water quality simulation

Analysis of hydrologic and water quality simulation results

The runoff data at the Zhangjiafen hydrological station was measured from 1990 to 2018. Because the actual observation data of China's surface meteorological station were only updated to May 2018, 1990–2017 was selected as the runoff simulation study period. The study period for sediments, TN, and TP spanned from 1990 to 2010. Based on the results of climate and hydrological analyses of the Baihe River Basin, 1999 was the mutation point for runoff, sediment, TN, and TP. Therefore, the simulation work was divided into two time periods, which adopted different parameter ranges for simulation: the 1990s (1990–1999) and the beginning of the twenty-first century (2000–2017, water quality simulation to 2010). For the 1990–1999 period, 1990–1995 was set as the calibration period and 1996–1999 was set as the validation period. For the 2000–2010 period, 2000–2007 was the calibration period and 2007–2010 was the validation period. The above variables were simulated on a monthly scale. Table 4 summarizes the results of the hydrology and water quality simulations.

Table 4

Summary of the hydrological and water quality simulation results

Simulation elementsEvaluation indexCalibration period in the 1990s
Validation period in the 1990s
Calibration period at the beginning of the twenty-first century
Validation period at the beginning of the twenty-first century
SUFI-2GLUESUFI-2GLUESUFI-2GLUESUFI-2GLUE
Runoff R2 0.84 0.84 0.92 0.93 0.60 0.63 0.59 0.63 
NSE 0.83 0.84 0.91 0.93 0.60 0.63 0.57 0.61 
RSR 0.41 0.40 0.41 0.41 0.67 0.66 0.70 0.69 
RMSE 6.06 5.98 10.04 10.11 3.71 3.57 5.02 4.75 
MAE 3.31 3.22 5.99 6.14 2.45 2.21 3.47 3.29 
Sediment R2 0.74 0.72 0.87 0.97 0.52 0.49 0.59 0.61 
NSE 0.73 0.72 0.85 0.96 0.52 0.48 0.55 0.53 
RSR 0.52 0.54 0.38 0.19 0.65 0.69 0.67 0.68 
RMSE 4.92 5.10 6.63 3.31 6.52 6.94 8.38 7.67 
MAE 2.25 2.18 2.78 1.28 1.90 1.98 1.93 1.97 
TN R2 0.61 0.60 0.84 0.84 0.58 0.63 0.65 0.64 
NSE 0.60 0.60 0.83 0.84 0.52 0.60 0.53 0.53 
RSR 0.63 0.63 0.62 0.60 0.71 0.69 0.69 0.69 
RMSE 3.31 3.39 2.32 2.49 10.78 10.0 12.5 12.47 
MAE 1.76 1.74 1.58 1.77 7.13 6.34 7.96 7.81 
TP R2 0.62 0.62 0.64 0.68 0.56 0.58 0.67 0.71 
NSE 0.60 0.61 0.62 0.66 0.50 0.51 0.52 0.53 
RSR 0.64 0.65 0.41 0.44 0.66 0.65 0.71 0.69 
RMSE 0.47 0.45 0.81 0.79 12.94 13.03 19.98 19.44 
MAE 0.26 0.23 0.47 0.45 7.65 7.31 11.6 11.38 
Simulation elementsEvaluation indexCalibration period in the 1990s
Validation period in the 1990s
Calibration period at the beginning of the twenty-first century
Validation period at the beginning of the twenty-first century
SUFI-2GLUESUFI-2GLUESUFI-2GLUESUFI-2GLUE
Runoff R2 0.84 0.84 0.92 0.93 0.60 0.63 0.59 0.63 
NSE 0.83 0.84 0.91 0.93 0.60 0.63 0.57 0.61 
RSR 0.41 0.40 0.41 0.41 0.67 0.66 0.70 0.69 
RMSE 6.06 5.98 10.04 10.11 3.71 3.57 5.02 4.75 
MAE 3.31 3.22 5.99 6.14 2.45 2.21 3.47 3.29 
Sediment R2 0.74 0.72 0.87 0.97 0.52 0.49 0.59 0.61 
NSE 0.73 0.72 0.85 0.96 0.52 0.48 0.55 0.53 
RSR 0.52 0.54 0.38 0.19 0.65 0.69 0.67 0.68 
RMSE 4.92 5.10 6.63 3.31 6.52 6.94 8.38 7.67 
MAE 2.25 2.18 2.78 1.28 1.90 1.98 1.93 1.97 
TN R2 0.61 0.60 0.84 0.84 0.58 0.63 0.65 0.64 
NSE 0.60 0.60 0.83 0.84 0.52 0.60 0.53 0.53 
RSR 0.63 0.63 0.62 0.60 0.71 0.69 0.69 0.69 
RMSE 3.31 3.39 2.32 2.49 10.78 10.0 12.5 12.47 
MAE 1.76 1.74 1.58 1.77 7.13 6.34 7.96 7.81 
TP R2 0.62 0.62 0.64 0.68 0.56 0.58 0.67 0.71 
NSE 0.60 0.61 0.62 0.66 0.50 0.51 0.52 0.53 
RSR 0.64 0.65 0.41 0.44 0.66 0.65 0.71 0.69 
RMSE 0.47 0.45 0.81 0.79 12.94 13.03 19.98 19.44 
MAE 0.26 0.23 0.47 0.45 7.65 7.31 11.6 11.38 

Table 4 shows that in the runoff simulation for the 1990s, the R2 and NSE of the SUFI-2 and GLUE simulation results reached 0.8 and above in both the calibration and validation periods, and the RSR was about 0.41. In the early twenty-first century, despite the effects of upstream reservoir management measures (Zhang 2014a, 2014b), R2 and NSE reached 0.57 and above, and all values of RSR were less than or equal to 0.70. For the sediment, TN and TP simulations, the indicators were also in a satisfactory condition, with better results in the 1990s than in the early twenty-first century. Overall, the results of the SWAT model for runoff and water quality simulations in the Baihe River Basin are credible. The MAE and RMSE of runoff in the 1990s were relatively large, because the precipitation peak in the 1990s was high, resulting in high runoff peak and frequency, which was difficult to simulate; in the early twenty-first century, the MAE and RMSE of TN and TP were relatively large, because the measured data in this period were relatively missing, and there were still peaks in the simulation, so the error was large, and the simulation results was unsatisfactory. The flow hydrograph comparison results of the measured and simulated runoff for each period are shown in Figure 8.

Figure 8

Comparison of simulated and observed runoff values in different periods. (a) Calibration (1990–1995), (b) validation (1996–1999), (c) calibration (2000–2007), and (d) validation (2008–2017).

Figure 8

Comparison of simulated and observed runoff values in different periods. (a) Calibration (1990–1995), (b) validation (1996–1999), (c) calibration (2000–2007), and (d) validation (2008–2017).

Close modal

For sediment simulation, the NSE and R2 of the two methods reached 0.7 in the 1990s and the validation period, indicating that the simulation effect was better. Because the effectiveness of the validation period was almost the same, the GLUE method performed better in the validation period, reaching above 0.9. In the early twenty-first century, owing to the impact of water transport from upstream reservoirs (lack of relevant data), R2 and NSE reached 0.5 and above. The monthly sequences of the measured and simulated values for sediment in each period are shown in Figure 9.

Figure 9

Comparison of simulated and observed sediment values in different periods. (a) Calibration (1990–1995), (b) validation (1996–1999), (c) calibration (2000–2007), and (d) validation (2008–2010).

Figure 9

Comparison of simulated and observed sediment values in different periods. (a) Calibration (1990–1995), (b) validation (1996–1999), (c) calibration (2000–2007), and (d) validation (2008–2010).

Close modal

Because of the influence of upstream reservoir management measures in the early twenty-first century, the simulation results of TN and TP were slightly worse than those of the 1990s; however, they still met the simulation requirements, reaching 0.5 and above. Figure 10 shows the fitting results of the measured and simulated values at each stage.

Figure 10

Comparison of simulated and observed TN and TP values in different periods. TN: (a1) calibration (1990–1995); (a2) validation (1996–1999); (b1) calibration (2000–2007); (b2) validation (2008–2010). TP: (c1) calibration (1990–1995); (c2) validation (1996–1999); (d1) calibration (2000–2007); (d2) validation (2008–2010).

Figure 10

Comparison of simulated and observed TN and TP values in different periods. TN: (a1) calibration (1990–1995); (a2) validation (1996–1999); (b1) calibration (2000–2007); (b2) validation (2008–2010). TP: (c1) calibration (1990–1995); (c2) validation (1996–1999); (d1) calibration (2000–2007); (d2) validation (2008–2010).

Close modal

Table 2 shows that the evaluation indexes of runoff and water quality simulation results for the 1990s were better than those for the 2000s, with higher R2 and NSE and lower errors. Comparing the simulation results of the two calibration methods (SUFI-2 and GLUE), the results of the SUFI-2 method were better than those of the GLUE method in the calibration period, while the GLUE method results were significantly better in the validation period. The results of the two methods had little difference overall in terms of the monthly process lines, and sometimes, the SUFI-2 results were closer to the measured values. Meanwhile, considering that SUFI-2 can be run only once based on the input optimal parameter values to obtain the optimal results (which GLUE cannot), the optimal parameter values obtained using SUFI-2 in the 1990s were substituted into the SWAT model to simulate runoff and water quality in the subsequent section on future runoff prediction and BMPs.

Uncertainty analysis of GLUE and SUFI-2 methods

Figure 11 shows correlation diagrams of the ‘behavior parameters’ of the SUFI-2 and GLUE methods for runoff, sediment, TN, and TP in the 1990s. For example, some NSEs showed downward trends with increased CN2 values, as well as significant upward trends with increased ALPHA_BNK values, entering a stable stage around 0.08. NSE showed a slight decline as the FILTERW value increased. Among other parameters, the scatter distribution of NSE changed with changing parameter values and was not significant. This indicates that, in the final parameter range corrected by GLUE, when the values of CN2 and FILTERW are low and the value of ALPHA_BNK is between 0.08 and 0.2, the model simulation will have lower uncertainty and a better simulation effect of TN. The above parameters have a stronger impact on the uncertainty of the simulation results of the GLUE method than others.

Figure 11

The correlation between ‘behavior parameters’ of simulated runoff, sediment, TN, and TP when using SUFI-2 and GLUE methods in the 1990s.

Figure 11

The correlation between ‘behavior parameters’ of simulated runoff, sediment, TN, and TP when using SUFI-2 and GLUE methods in the 1990s.

Close modal

Part of the NSE gradually gathered as the values of GW_DELAY, GWQMN, and CH_K2 increased; The NSE showed a tendency to diverge as GWSOLP increased; in the 0–0.3 range, NSE increased as the LAT_ORGP value increased; at 0.3 within the range of −1, NSE decreased as the LAT_ORGP value increased and showed a tendency to diverge. Compared with other parameters, the scatter distribution of NSE changes with the change in parameter values was not significant. This means that, in the final parameter range corrected by SUFI-2, when the GW_DELAY, GWQMN, and CH_K2 values were larger, the value of GWSOLP was smaller, and when the LAT_ORGP was approximately 0.3, the uncertainty of the model simulation was smaller, and the TP simulation effect was better. The above parameters had greater impacts on the uncertainty of the SUFI-2 simulation results than other parameters. Figure 11 shows the correlation of the SUFI-2 method ‘behavior parameters’ in the TP simulation.

Impact of land use on hydrology and water quality

Consulting Chinese government policy documents revealed that the water level of the Miyun Reservoir rose to nearly 155 m in 1994. To ensure the safety of residents, the State Council of China relocated the reservoir for the third time from 1995 to 1999, promoting the transformation of land use in the Baihe River Basin. At the same time, official implementation of the national water and soil conservation project in 2000 has promoted changes in the cultivated land, forest, and grassland of the basin.

ArcGIS was used to analyze the spatial distribution of land-use changes to obtain a land-use change matrix, and the results are shown in Table 5 and Figure 12. The overall land-use changes are severely scattered in space, distributed in scattered points throughout the basin, and show signs of changes along the river.

Table 5

Land-use change matrix in the Baihe River Basin (unit: km2)

AgriculturalForestGrasslandWaterUrbanRural areaIndustrial buildingUnused land
Agricultural 1,883.80 16.07 39.39 4.20 1.73 35.73 27.12 0.10 
Forest 14.22 4,039.35 28.64 1.72  1.69 1.69 0.01 
Grassland 47.37 92.01 2,394.31 1.78 0.15 3.63 18.51 0.16 
Water 15.56 1.13 2.57 60.21  0.60 0.88 0.00 
Urban 0.01    2.75    
Rural area 11.07 0.14 0.37 0.19 0.17 25.60 0.38  
Industrial building 0.00 0.06 0.05    1.46  
Unused land 0.58 0.00 2.07 0.00  0.10 0.22 10.35 
AgriculturalForestGrasslandWaterUrbanRural areaIndustrial buildingUnused land
Agricultural 1,883.80 16.07 39.39 4.20 1.73 35.73 27.12 0.10 
Forest 14.22 4,039.35 28.64 1.72  1.69 1.69 0.01 
Grassland 47.37 92.01 2,394.31 1.78 0.15 3.63 18.51 0.16 
Water 15.56 1.13 2.57 60.21  0.60 0.88 0.00 
Urban 0.01    2.75    
Rural area 11.07 0.14 0.37 0.19 0.17 25.60 0.38  
Industrial building 0.00 0.06 0.05    1.46  
Unused land 0.58 0.00 2.07 0.00  0.10 0.22 10.35 
Figure 12

Baihe River Basin land-use change matrix.

Figure 12

Baihe River Basin land-use change matrix.

Close modal

In 1990 and 2010, the main land-use types in the Baihe River Basin were forest, grassland, and arable land. The results show that the most drastic change was observed in grassland, of which about 47 km2 was converted to arable land to ensure the agricultural development and survival of the relocated migrants in the watershed; about 92 km2 was converted to forest land to support soil and water conservation efforts, and 18.5 km2 was converted to industrial construction land for economic development in the watershed. Additionally, a small portion of grassland was converted to rural, watershed, and town land. Affected by factors such as relocation, the conversion of forest and grassland to arable land reached 14.22 and 47.37 km2, respectively; the conversion of forest to grassland was 28.64 km2, and the conversion of grassland to forest was as high as 92 km2. Cultivated land (15.56 km2) accounts for the largest proportion of watershed transformation types and reflects the decreased water volume in the basin. The reduced arable land was mainly transformed into grassland, rural land, industrial land, and forest land.

To discuss the effects of land-use changes on runoff, sediment, and nitrogen and phosphorus load changes in the Baihe River Basin, land-use data for 1990 and 2000 were input into the calibrated SWAT model, while keeping soil, slope classification, and meteorological information unchanged. The simulation period was 1990 to 2010, with monthly simulation scales. Based on the simulation results, the cumulative changes in flow, sand transport, and TN and phosphorus were calculated from 1990 to 2010. The results are presented in Table 6 and Figure 13.

Table 6

Simulated total cumulative values and changes from 1990 to 2010 based on 1990 and 2010 land uses

19902010Amount of changeChange (%)
Runoff (1010m33.34 3.35 0.1 0.30 
Sediment (106 t) 7.18 7.2 0.02 0.28 
TN (107kg) 7.04 6.82 −0.22 −3.12 
TP (107kg) 1.62 1.59 −0.03 −1.85 
19902010Amount of changeChange (%)
Runoff (1010m33.34 3.35 0.1 0.30 
Sediment (106 t) 7.18 7.2 0.02 0.28 
TN (107kg) 7.04 6.82 −0.22 −3.12 
TP (107kg) 1.62 1.59 −0.03 −1.85 
Figure 13

Cumulative change in four variables under land-use change: (a) runoff, (b) sediment transport, (c) TN, and (d) TP.

Figure 13

Cumulative change in four variables under land-use change: (a) runoff, (b) sediment transport, (c) TN, and (d) TP.

Close modal

Without considering the influence of human water abstraction, although the land-use change in the basin is relatively complex, the flow in the basin has not changed significantly, only a slight increase, and the cumulative increase in 20 years is about 109 m3, about 0.30%. This phenomenon may be due to the fact that the transition of grassland and forest land in the watershed occupies a large proportion, and there is no significant difference in the impact of the two on flow. Moreover, the land-use changes in the watershed are particularly scattered. Although the total number of changes is large, the small land-use changes scattered to local areas have a weak impact on the flow in the watershed, so the changes in the flow in the watershed are not significant.

Figure 13(b) shows that despite complex land-use changes in the watershed, sediment transport did not change significantly. The percentage increase in cumulative sediment transport over the 20 years is basically the same as that of runoff, only 0.2%. Further analysis shows that the number of extreme precipitation events in the basin has decreased in recent years, and the number and value of runoff peaks have decreased accordingly, indicating that precipitation and runoff have a very important impact on sediment transport.

Studies have shown that the area of forest land is negatively correlated with the TN and TP loads, while the area of arable land is positively correlated with the TN and TP loads. Large amounts of pesticides and fertilizers are used in agriculture, and precipitation causes the residual nitrogen and phosphorus on the ground to be lost (Zhang et al. 2019b). The increase in forest area and the decrease in arable land area impact the total TN and TP in the watershed. During the past two decades, TN decreased by 3.12% and TP decreased by 1.85%. Although the percentage of decrease was not large, it still had a positive effect on the water quality of the basin. Research has shown that patchy scattered farmland increases the risk of nitrogen and phosphorus pollution (Mi & Fan 2013), and the area of arable land after moving and resettlement in the basin is small and centered, which may also account for the decrease in TN and TP.

Impact of management measures on hydrology and water quality

Identification of key source areas in the river basin

To gain insights into the spatial distribution characteristics of sediments, nitrogen and phosphorus loads in the watershed were determined based on the results of the Baihe River Basin SWAT model, soil erosion modulus, and nitrogen and phosphorus concentrations per unit area and used to identify the key source areas of soil erosion and nitrogen and phosphorus loads in the watershed. Based on the SWAT simulation results from 1990 to 1999 and 2000 to 2010, the key source areas in the basin were identified for each sub-watershed, and the results are as follows.

The multi-year averaged sediment transport of each sub-watershed output of the SWAT model, combined with the definition of sediment transport ratio (Li 2014) and calculation formula (Xie & Li 2012), calculates the soil erosion modulus of the watershed. According to the natural fissure classification method, the soil erosion intensity of the Baihe River Basin was classified among four categories: slight loss, moderate loss, heavy loss, and severe loss; the results are shown in Figure 14(a). The proportion of soil erosion area was as follows: slight loss, 39.05%; moderate loss, 22.03%; heavy loss, 20.16%; and severe loss, 18.77%. Micro-erosion and mild erosion were the main factors affecting losses; the degree of upstream erosion is heavier than the downstream erosion intensity. Among them, sub-watersheds 1, 4, 6, 7, 10, 12, 14, 15, 27, and 31 had severe erosion intensity.

Figure 14

(a) Grading of soil erosion intensity; (b) risk grade of TN loss; and (c) risk grade of TP loss.

Figure 14

(a) Grading of soil erosion intensity; (b) risk grade of TN loss; and (c) risk grade of TP loss.

Close modal

According to the multi-year averaged TN loss quality of each sub-basin determined using the SWAT model, the risk of TN loss in the Baihe River was divided into four categories using the natural cracking point method: slight, moderate, heavy, and severe losses (accounting for 32.63, 39.78, 25.26, and 2.32%, respectively, of losses); the results are shown in Figure 14(b). The overall risk of TN loss predominantly ranged from slight to moderate, and it was higher upstream than downstream. Among them, sub-basins 1–6, 8, 9, 12, 14, and 15 had higher risks of loss.

Similarly, the risk of TP loss in the Baihe River was categorized into four levels: slight, moderate, heavy, and severe losses (accounting for 51.46, 21.28, 9.66, and 17.59% of losses, respectively); the results are shown in Figure 14(c). The overall risk of TN loss predominantly ranged from slight to moderate loss, and it was higher upstream than downstream; sub-basins 1–6, 8–12, 14, and 24 had higher risk levels of loss. After comparing the level of soil erosion with the levels of TN and phosphorus loss, sub-basins 1, 4, 6, 8, 9, 10, 12, 14, and 15, which are mainly located in the upper reaches of the basin, can be considered the key source areas of non-point source (NPS) pollution in the Baihe River Basin.

Analysis and evaluation of the effectiveness of management measures

This section retains all data from the 1990s, except for the different set of management measures. The percentage reductions in annual mean flow, sediment, TN, and TP in key source areas (sub-basin scales) under different scenarios and management measures are compared with the initial scenario at that time, as shown in Table 7.

Table 7

Annual average flow, sediment, TN, TP load, and their changes under different scenarios

ScenarioCharacteristic value
Change rate (%)
Flow (104 m3)Sediment (104 t)TN (104 t)TP (t)FlowSedimentTNTP
Initial 41,659.97 60.89 2.38 281.73 
Stubble cover 40,749.10 45.48 2.07 200.56 −2.19 −25.31 −13.32 −28.81 
Contour planting 41,207.15 50.64 2.10 233.24 −1.09 −16.83 −11.88 −17.21 
Terraces 39,840.23 41.45 1.83 182.29 −4.37 −31.93 −23.13 −35.30 
ScenarioCharacteristic value
Change rate (%)
Flow (104 m3)Sediment (104 t)TN (104 t)TP (t)FlowSedimentTNTP
Initial 41,659.97 60.89 2.38 281.73 
Stubble cover 40,749.10 45.48 2.07 200.56 −2.19 −25.31 −13.32 −28.81 
Contour planting 41,207.15 50.64 2.10 233.24 −1.09 −16.83 −11.88 −17.21 
Terraces 39,840.23 41.45 1.83 182.29 −4.37 −31.93 −23.13 −35.30 

Among the non-engineering measures, the average annual flow, sediment, TN, and TP reduction percentages reached 2.19, 25.31, 13.32, and 28.81%, respectively, under the stubble cover scenario and 1.09, 16.83, 11.88, and 17.21%, respectively, under the contour planting scenario. Comparing the reduction effects of the two measures, both had very good water and soil conservation effects; however, the flow reduction period, sediment reduction period, TN, and TP were more affected by stubble mulch farming than by contour planting. This shows stubble cover farming had a better non-point source pollution control effect than contour planting. In engineering measures, the average annual flow period, sediment, TN, and TP reduction rates under the terraced project scenario reached 4.37, 31.93, 23.13, and 35.30%, respectively.

Comparing the three scenarios, engineering measures had significantly better reduction effects than those of non-engineering measures. The three scenarios can be ranked in descending order of effectiveness as terraced fields, stubble coverage, and contour planting.

Climate and runoff prediction under future climate change

To further study the impact of climate change-related factors on water resource changes in the Baihe River Basin, the BCC model data were coupled with a calibrated SWAT model to predict precipitation, maximum and minimum temperature changes, and runoff responses under three future climate change scenarios.

Future climate change analysis

Table 8 shows the performance of three future (2022–2064) climate change scenarios for the Baihe River Basin after the DBC bias correction method. As the emission scenarios increased, the annual average maximum temperature, annual average minimum temperature, and precipitation gradually increased as well. In the SSP1-2.6 low emission scenario, the annual average maximum temperature increased by approximately 0.6%, the annual average minimum temperature increased by approximately 1.9%, and precipitation increased by 16.7%. In the SSP2-4.5 medium emission scenario, the annual average maximum temperature increased by approximately 4.0%, the annual average minimum temperature increased by 5.6%, and precipitation increased by 20.5%. In the SSP5-8.5 high emission scenario, the annual average maximum temperature increased by approximately 8.7%, annual average minimum temperature increased by approximately 14.8%, and precipitation increased by 25.2%. Clearly, the annual average maximum temperature increase exceeded the annual minimum temperature, while the temperature and precipitation increases were gradual and proportionate to the low and high emission scenarios.

Table 8

Climate characteristic values and relative changes in the base and future periods

Climatic factorsPeriodEigenvaluesRelative changes (%)
Annual average maximum temperature (°C) Base period 17.3 
SSP1-2.6 17.4 0.6 
SSP2-4.5 18.0 4.0 
SSP5-8.5 18.8 8.7 
Annual average minimum temperature (°C) Base period 5.4 
SSP1-2.6 5.5 1.9 
SSP2-4.5 5.7 5.6 
SSP5-8.5 6.2 14.8 
Average annual precipitation (mm) Base period 498.6 
SSP1-2.6 581.8 16.7 
SSP2-4.5 600.6 20.5 
SSP5-8.5 624.4 25.2 
Climatic factorsPeriodEigenvaluesRelative changes (%)
Annual average maximum temperature (°C) Base period 17.3 
SSP1-2.6 17.4 0.6 
SSP2-4.5 18.0 4.0 
SSP5-8.5 18.8 8.7 
Annual average minimum temperature (°C) Base period 5.4 
SSP1-2.6 5.5 1.9 
SSP2-4.5 5.7 5.6 
SSP5-8.5 6.2 14.8 
Average annual precipitation (mm) Base period 498.6 
SSP1-2.6 581.8 16.7 
SSP2-4.5 600.6 20.5 
SSP5-8.5 624.4 25.2 

Figure 15 shows the comparison and analysis of the corrected monthly average minimum temperature with the base period. The overall minimum temperature changed less than the maximum temperature. The SSP1-2.6 low emission scenario changed slightly from the base period, with a relative change of −0.5 to 26.1%; the SSP2-4.5 medium emission scenario changed from −27.2 to 28.8% relative to the base period, and the SSP5-8.5 high emission scenario changed from −1.8 to 78.1% relative to the base period. The relatively high change percentages are due to the fact that the lowest temperature in March is approximately zero. Unlike the monthly average maximum temperature, the temperature change from January to June was higher than the temperature change from July to December, and the rise in temperature gradually increased from the low emission scenario to the high emission scenario.

Figure 15

Monthly averaged parameters under the corrected future scenario: (a) maximum temperature, (b) minimum temperature, and (c) precipitation.

Figure 15

Monthly averaged parameters under the corrected future scenario: (a) maximum temperature, (b) minimum temperature, and (c) precipitation.

Close modal

By comparing and analyzing the corrected monthly average precipitation with the base period, it was revealed that the SSP1-2.6 low emission scenario changed from −15.3 to 140.3% relative to the base period, the SSP2-4.5 medium emission scenario changed from −16.1 to 146.2% relative to the base period, and the SSP5-8.5 high emission scenario changed from −15.7 to 144% relative to the base period. Compared with the base period, the increase in September was the most significant, accounting for much of the increase in annual precipitation; the more consistent rule is that the precipitation change rate in winter and spring is relatively large, while the precipitation change period in summer is relatively low. The monthly average precipitation changes among the different scenarios did not show the same change patterns as the temperature. Except for May and October, the monthly precipitation in each of the three scenarios was higher than the base period. Precipitation began in the month of seasonal change, beginning in spring. The decrease in spring to summer and summer to autumn indicates that precipitation is significantly affected by seasonal changes. First, the precipitation variation in October in the SSP1-2.6 scenario was reduced the most (by 15.3%); second, the variation ranges of the three climate scenarios did not show the same magnitude as the temperature. In the increased emission scenario, the increased changes in SSP1-2.6 and SSP5-8.5 were basically the same, while the change in SSP5-8.5 was higher than that in the SSP1-2.6 scenario. Compared with SSP1-2.6, the monthly precipitation increase in the scenario SSP2-4.5 appeared unstable and showed no change pattern.

Runoff prediction under future climate scenarios

The calibrated SWAT model and subsurface conditions in the 1990s are used in this paper for future runoff prediction to facilitate comparison with the study in the previous part of the paper. The results of the runoff simulations are shown in Table 9. The runoff of the three future scenarios increased from the base period, which is consistent with the change trend of precipitation. SSP5-8.5 had the largest runoff change, with a relative change period of 61.3%; it was followed by SSP1-2.6, which had a relative change of 49.3%. The relative change period of SSP2-4.5 was lower than that of SSP1-2.6 (40.2%). This result is in agreement with the fact that the higher the precipitation, the higher the runoff volume. The precipitation–runoff correlation diagram in Figure 16 shows that the SSP5-8.5 runoff precipitation correlation coefficient R2 was the largest (approximately 0.79), indicating a good correlation. Similarly, the SSP1-2.6 runoff precipitation correlation coefficient R2 is approximately 0.74, which also shows a good correlation. For SSP2-4.5, however, the runoff precipitation correlation coefficient R2 was only 0.54, which is significantly lower than that of the other two scenarios. Therefore, it can be explained that SSP2-4.5 experienced more precipitation than SSP1-2.6, but the runoff increase was less than that of SSP1-2.6.

Table 9

Runoff and its relative changes in the base and future periods

Average annual runoff (m3/s)Amount of change (m3/s)Relative changes (%)
Base period 8.7   
SSP1-2.6 13.0 4.3 49.3 
SSP2-4.5 12.2 3.5 40.2 
SSP5-8.5 14.0 5.3 61.3 
Average annual runoff (m3/s)Amount of change (m3/s)Relative changes (%)
Base period 8.7   
SSP1-2.6 13.0 4.3 49.3 
SSP2-4.5 12.2 3.5 40.2 
SSP5-8.5 14.0 5.3 61.3 
Figure 16

Precipitation–runoff correlation in future scenarios.

Figure 16

Precipitation–runoff correlation in future scenarios.

Close modal

Figure 17(a) shows the annual change trends of runoff and precipitation under the three scenarios, which indicate highly consistent changes in runoff and precipitation. A comparison between the monthly average runoff processes of the three emission scenarios and the baseline period is shown in Figure 17(b). Since the runoff simulation for the future period is based on the calibrated SWAT model and the subsurface conditions in the 1990s, based on the simulation results in the previous section, the runoff obtained from the projection will also be very large when the precipitation in the future period is significantly higher than that in the base period. The main features of future runoff are as follows: first, the runoff peak appears in August–September, which is consistent with the increasing trend of precipitation and indicates that precipitation and runoff have a good correlation; second, the runoff exhibited double peaks throughout the year, which was consistent with the increase in precipitation. In the abovementioned scenario, the actual runoff statistics and analysis change trends were consistent, indicating credible simulation results; additionally, the runoff increased slightly in November, which was closely related to precipitation. The change trends of the SSP1-2.6 and SSP5-8.5 scenarios were highly consistent, whereas the change trend of SSP2-4.5 was different. Considering the low correlation between precipitation and runoff under SSP2-4.5, the phenomenon may be influenced by other factors and needs to be further explored.

Figure 17

(a) Annual change trend of runoff and precipitation in future scenarios. (b) Monthly average runoff and its changes in the base and future periods.

Figure 17

(a) Annual change trend of runoff and precipitation in future scenarios. (b) Monthly average runoff and its changes in the base and future periods.

Close modal

The future period from 2022 to 2064 was divided into three stages to observe and study the trend of runoff changes in the future, 2022–2035, 2036–2049, and 2050–2064. The results are listed in Table 10. Under the three scenarios, runoff decreased with time. Compared with the first stage runoff, SSP1-2.6 decreased by 15.2 and 17.3% in the second and third stages, respectively; SSP2-4.5 decreased by 3.1 and 10.7% in the second and third stages, respectively. In the second and third stages, SSP5-8.5 decreased by 23.1 and 27.3%, respectively. The reduction intensities of SSP1-2.6 and SSP5-8.5 were higher than that of SSP2-4.5.

Table 10

Annual average runoff values for the three stages in the future period and the relative changes

PeriodRunoff (m3/s)
Relative changes (%)
SSP1-2.6SSP2-4.5SSP5-8.5SSP1-2.6SSP2-4.5SSP5-8.5
The first stage 4.6 4.1 5.4    
The second stage 3.9 3.9 4.1 −15.2 −4.9 −24.1 
The third stage 3.8 3.6 3.9 −17.4 −12.2 −27.7 
PeriodRunoff (m3/s)
Relative changes (%)
SSP1-2.6SSP2-4.5SSP5-8.5SSP1-2.6SSP2-4.5SSP5-8.5
The first stage 4.6 4.1 5.4    
The second stage 3.9 3.9 4.1 −15.2 −4.9 −24.1 
The third stage 3.8 3.6 3.9 −17.4 −12.2 −27.7 

This study selected the BCC-CSM2-MR climate model under the CMIP6 program (developed by the National Climate Center) and selected three emission scenarios (SSP1-2.6, SSP2-4.5, and SSP5-8.5) corresponding to low, medium, and high emissions. The DBC bias correction method was used to perform the output data of the climate model by correcting, obtaining better correction results, and finally predicting the runoff in the future period (2022–2064). The results showed that compared with the base period, the temperature and precipitation under the three scenarios showed upward trends year by year, and the changes fluctuated within years. The precipitation and runoff trends of SSP1-2.6 and SSP5-8.5 were consistent in the three scenarios, and the peak runoff time of SSP2-4.5 was earlier than the above two scenarios. The future runoff and precipitation maintained a good correlation, and the runoff increased with precipitation. Although it showed an increasing trend, the runoff gradually decreased with time.

This study conducted a statistical analysis of the climate and hydrology of the Baihe River Basin over the past 30 years from 1990 to 2017. The M–K trend test did not find a sudden change in annual precipitation in the Baihe River Basin. The 5-year moving average revealed that the annual precipitation decreased first and then increased around 2001. The precipitation and runoff in the basin showed significant and consistent change trends. After the M–K trend test and the precipitation–runoff double accumulation curve, it was found that runoff, sediment, TN, and TP exhibited abrupt changes in 1999, and human activities had a greater impact on runoff changes; after 2000, the peak runoff changed from a single peak in the 1990s to a double peak, and there were inter-generational changes in the uneven coefficient of runoff during the year. Meanwhile, sediment amount transported in the basin has decreased greatly due to decreasing precipitation trends.

The SWAT model and the SUFI-2 and GLUE methods simulated the runoff, sediment, TN, and TP in the Baihe River Basin from 1990 to 1999 and 2000 to 2010 (runoff to 2017) on a monthly scale. The simulation effect met the simulation requirements; in the 1990s, R2 and NSE of runoff and sediment were both greater than 0.72, and the RSR was less than 0.54; while R2 and NSE values of TN and TP were both greater than 0.60, and the RSR was less than 0.65. The results show that the SWAT model has good applicability in the Baihe River Basin, and that the simulation effect at the beginning of the twenty-first century is slightly inferior to that of the 1990s; however, it still meets the simulation requirements, which are mainly affected by the upstream reservoir management measures. After the uncertainty analysis, it was found that the uncertainty of the SUFI-2 and GLUE methods in the Baihe River Basin simulation was mainly affected by parameter sensitivity and less affected by the correlation of the parameters.

This study used the calibrated SWAT model to simulate the impact of land-use changes on hydrology and water quality. The results show that the transformation of land-use patterns in 1990 and 2010 had slight impacts on the hydrology and water quality of the basin; through the cumulative curve, the cumulative runoff was increased by approximately 0.30%, sediment transport was increased by 0.28%, TN was reduced by 3.12%, and TP was reduced by 1.85%. There are two reasons why the changes in runoff and sediment in the basin are unlikely. First, the conversion of forest land and grassland accounts for the largest proportion of land-use changes in the basin, and the impact of the two on runoff and sediment is almost the same. The land-use change in the basin is particularly scattered. Although the total amount of change is large, the small amplitude changes scattered to the local area have a weak impact on the flow in the basin, so the changes in flow and sediment are not significant. The changes of TN and TP have decreased, although the decrease is not much, it still has a positive effect on improving the water quality of the watershed. It may be that the cultivated land in the basin has become more concentrated, and the effect of fertilization in cultivation has weakened. By identifying the key source areas of NPS pollution in the watershed, the effects of three management measures on runoff, sediment, TN, and TP were simulated, and the results showed that the three management measures had significant impacts. The most effective management measures for reducing periods of sediment, TN, and TP were terrace projects, stubble cover, and contour planting (ranked from high to low).

This study used the latest CMIP6 data and selected the BCC-CSM2-MR climate model developed by the Chinese National Climate Center for three scenarios (SSP1-2.6: low emission scenario, SSP2-4.5: medium emission scenario, and SSP5-8.5: high emission scenario) to predict the future climate and runoff of the Baihe River Basin. The results show that as compared with the base period, the temperature and precipitation under the three scenarios showed overall upward trends year by year, while changes fluctuated within the year. In each of the three scenarios, the runoff increased compared to the base period and the future runoff maintained a good correlation with precipitation, except for SSP2-4.5. Dividing the stages of the future period revealed that runoff is decreasing year by year.

This research was funded by the National Key R&D Program of China (2017YFC0406004) and the NSFC (41271004).

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

The authors declare there is no conflict.

Boufala
M.
,
El Hmaidi
A.
,
Essahlaoui
A.
,
Chadli
K.
,
Ouali
A. E.
&
Lahjouj
A.
2021
Assessment of the best management practices under a semi-arid basin using SWAT model (case of M’dez watershed, Morocco)
.
Modeling Earth Systems and Environment
8
,
713
731
.
Chen
H.
,
Shi
S.
&
Zhou
J.
2011
BCC climate model simulation evaluation of extreme climate events in China in the past 50 years
.
Chinese Journal of Atmospheric Sciences
34
(
05
),
513
528
.
Dong
M.
,
Wu
T.
&
Wang
Z.
2013
Simulation of precipitation and its variability in the 20th century with BCC_CSM1.0 model
.
Journal of Applied Meteorology
24
(
01
),
1
11
.
Feng
Z.
,
Sun
T.
&
Yang
Y.
2018
The progress of resources and environment carrying capacity: from single-factor carrying capacity research to comprehensive research
.
Journal of Resources and Ecology
9
(
2
),
125
134
.
Feng
Y.
,
Sun
F.
,
Liu
W.
,
Chen
J.
,
Wang
H.
,
Guo
Q.
,
Wang
Y.
,
Zhang
Q.
&
Sang
Y.
2021
Changes in compound hot and dry day and population exposure across china under climate change
.
International Journal of Climatology
42
(
5
),
2935
2949
.
Gao
H.
,
Feng
Z.
,
Zhang
T.
,
Wang
Y.
,
He
X.
,
Li
H.
,
Pan
X.
,
Ren
Z.
,
Chen
X.
,
Zhang
W.
&
Duan
Z.
2021
Assessing glacier retreat and its impact on water resources in a headwater of Yangtze River based on CMIP6 projections
.
Science of the Total Environment
765
,
142774
.
He
L.
,
Wang
Y.
&
Yin
F.
2020
Research on the multi-matching characteristics of water and soil resources and agricultural development in Central Asia under the influence of global climate change
.
Science China: Earth Sciences
50
(
09
),
1268
1279
.
Hu
Y.
,
Zhou
L.
&
Yan
M.
2007
Eco-economic system characteristics, asset base and functional benefit evaluation of Beijing Miyun Reservoir
.
Journal of Natural Resources
(
04
),
497
506
.
Huang
K.
2020
Research on the Best Management Measures for Non-Point Source Pollution in Danjiang Basin Based on SWAT Model
.
Xi'an University of Technology
.
Ibrahim
M. O.
2018
Global warming and its implications on meteorological and hydrological drought in the Southeastern Mediterranean
.
Environmental Processes
5
(
2
),
329
348
.
Khamidov
M. K.
&
Khamraev
M.
2020
Water-saving irrigation technologies for cotton in the conditions of global climate change and lack of water resources
.
IOP Conference Series: Materials Science and Engineering
883
(
1
),
012077
.
Lei
W.
,
Xu
Z.
&
Jixi
G.
2019
Effects of fertilizer types on nitrogen and phosphorous loss from rice-wheat rotation system in the Taihu Lake region of China
.
Agriculture, Ecosystems and Environment
285
,
106605
.
Lennart
B. E.
,
Torben
O. S.
&
Jens
C. R.
2019
Observational and predictive uncertainties for multiple variables in a spatially distributed hydrological model
.
Hydrological Processes
33
(
5
),
833
848
.
Li
M.
2014
The Impact of Land Use and Climate Change on Non-Point Source Nitrogen and Phosphorus Pollution in Miyun Reservoir Basin
.
Capital Normal University
.
Li
Q.
,
Zhang
J.
&
Gong
H.
2017
Comparison of multi-site uncertainty evaluation methods based on SWAT model
.
People’s Yellow River
39
(
01
),
24
29
.
Li
Y.
,
Zhu
R.
&
Liu
T.
2020
Analysis of the future temperature and precipitation trend of the Shule River Basin based on the BCC-CSM2-MR model
.
Plateau Meteorology
40
(
3
),
535
546
.
Mi
Y.
&
Fan
J.
2013
Effects of returning farmland to forest on sediment concentration and sediment transport in the middle Reaches of the Yellow River
.
Shaanxi Water Resources
(
04
),
127
129
.
Michael
S.
,
Christian
B.
&
Sérgio
K.
2011
Using precipitation data ensemble for uncertainty analysis in SWAT streamflow simulation
.
Journal of Hydrology
414
,
413
424
.
Moriasi
D. N.
,
Arnold
J. G.
,
Liew
M. W. V.
,
Bingner
R. L.
,
Harmel
R. D.
&
Veith
T. L.
2007
Model evaluation guidelines for systematic quantification of accuracy in watershed simulations
.
Transactions of the ASABE
50
,
885
900
.
Rokhsare
R.
,
Aazam
J.
&
Majid
A.
2009
Application of a SWAT model for estimating runoff and sediment in two Mountainous Basins in Central Iran
.
Water and Energy Abstracts
19
(
2
),
977
988
.
Saila
R.
2020
Pollution assessment of point and non-point sources of Buriganga River
.
Journal of Water Resources and Pollution Studies
5
(
2
),
26
30
.
Wang
X.
,
Wang
Y.
,
Cai
X.
&
Xu
Q.
2002
Research on the Status Quo of non-point source pollution in the Miyun Reservoir Watershed in Beijing
.
Environmental Science and Technology
(
04
),
1
3
.
Wang
Z.
,
Lin
C.
&
Xu
J.
2019
Study on the spatial difference and key influencing factors of non-point source particulate phosphorus load in Chaohu Lake Basin
.
Journal of Agro-Environmental Sciences
38
(
03
),
659
670
.
Wang
G.
,
Qiao
C.
&
Wang
J.
2020
Simulation of water balance in Lancang-Mekong River Basin under global change
.
Chinese Journal of Atmospheric Sciences
43
(
06
),
1010
1017
.
Wang
X.
,
Yin
L.
,
Ma
H.
,
Dai
Z.
&
Zhang
X.
2021
The influence of change of land-use type on hydrological cycle
.
IOP Conference Series: Earth and Environmental Science
671
(
1
),
012041
.
Xie
W.
&
Li
T.
2012
Research progress of sediment transport ratio in river basin
.
Journal of Peking University (Natural Science Edition)
48
(
04
),
676
685
.
Xie
H.
,
Wu
B.
&
Zhang
X.
2009
Analysis on the time-history evolution characteristics of water quality in the Baihe River Basin
.
Journal of Anhui Agricultural Sciences
37
(
24
),
11672
11674
.
Xin
X.
,
Wu
T.
&
Zhang
J.
2019
Introduction to the BCC model and its CMIP6 experiment
.
Progress in Climate Change Research
15
(
05
),
533
539
.
Xing
K.
,
Guo
H.
,
Sun
Y.
,
He
B.
&
Huang
Y.
2005
Simulation of non-point source pollution in the watershed: a case study of the Dianchi Lake Basin
.
Geographical Research
(
04
),
549
558
.
Xu
Z.
,
Yu
D.
&
Tao
G.
2020
Study on the pollution load of adsorbed nitrogen from soil erosion in Biliu River Basin based on grid unit
.
Marine Environmental Science
39
(
01
),
138
144
.
Yan
C.
,
Du
Z.
&
Yao
B.
2020
Spatial pattern analysis of nitrogen and phosphorus loss in the Baoxiang River Basin of Dianchi Lake
.
Environmental Science Research
33
(
12
),
2695
2704
.
Zhang
Y.
2014a
Attribution Analysis on the Change of Runoff in the Upstream of Miyun Reservoir During Last 30 Years
.
Master Thesis
,
Tsinghua University
,
Beijing
,
China
.
Zhang
P.
2014b
Analysis and Research on Hydrological Characteristics of the Upper Heihe River
.
Lanzhou University
.
Zhang
Y.
,
Zheng
F.
&
Wu
M.
2007
Research progress in agricultural non-point source pollution caused by soil erosion
.
Advances in Water Science
(
01
),
123
132
.
Zhang
L.
,
Chen
X.
&
Xin
X.
2019a
Short commentary on CMIP6 Scenario Model Intercomparison Project (ScenarioMIP)
.
Climate Change Research
15
(
5
),
519
525
.
Zhang
M.
,
Li
L.
,
Zhao
W.
,
Xu
J.
&
Zhao
W.
2019b
Spatial heterogeneity and cause analysis of water quality in the upper streams of Miyun Reservoir
.
Acta Scientiae Circumstantiae
39
(
06
),
1852
1859
.
Zhao
G.
,
Xu
Z.
&
Dong
Q.
2017
The impact of different management measures on water quantity and quality changes in the Miyun Reservoir basin
.
South-to-North Water Diversion and Water Science and Technology
15
(
02
),
80
88
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY-NC-ND 4.0), which permits copying and redistribution for non-commercial purposes with no derivatives, provided the original work is properly cited (http://creativecommons.org/licenses/by-nc-nd/4.0/)