Simulation and attribution analysis of runoff–sediment in the Upper Basin of Fenhe River, China

With the progress of human civilization and the development of society, many countries in the world are facing serious water shortage (Abedin et al. 2019; Li et al. 2019a). The problem of water shortage is not only related to the change in runoff but also affected by the law of sediment change. Therefore, many studies have been carried out on the simulation and analysis of runoff and sediment (Fu et al. 2009; Shi et al. 2017; Li et al. 2021; Wang et al. 2021; Burgan 2022).

The universal soil loss equation (USLE), developed by the Agricultural Research Service of the United States Department of Agriculture (USDA), is one of the most widely used sediment transport formulas at present. In recent years, scholars from various countries have made many improvements to the equation, such as modified USLE (MUSLE) (Arekhi et al. 2012; Gwapedza et al. 2021), revised USLE (RUSLE) (Renard et al. 1991), and modified version of the USLE (USLE-M) (Di Stefano et al. 2017). Several studies based on USLE and its improved equations have been carried out (Ozcan et al. 2008; Kinnell 2014; Kinnell et al. 2018). USLE-M has been embedded in several hydrological models for integrated runoff–sediment simulations, such as the Soil and Water Assessment Tool (SWAT) model (Arnold 1992).

Simulating and evaluating runoff and sediment processes based on a hydrological model with the sediment simulation function is a commonly used method. Examples include the Water Erosion Prediction Project (WEPP) model developed by the USDA (Laflen et al. 2004; Xu et al. 2023), the Annualized Agricultural Non-Point Source Pollution Model (AnnAGNPS) developed by USDA to simulate catchment-scale hydrology and pollutant migration and transformation (Licciardello et al. 2007), and the MIKE 21 model based on hydrodynamics (Nguyen Thi Diem et al. 2019). The SWAT model is one of the most widely used hydrological models in the world. It can not only simulate runoff with high precision (Zhang et al. 2020; Mengistu et al. 2021) but also performs well on sediment simulation (Gebremicael et al. 2013; Kang et al. 2021).

Changing environments, including climate change and human activities, have significant impacts on surface processes, such as the hydrological cycle, water and energy balance, and ecological environments at various spatial and temporal scales (Piao et al. 2010). Studying the water cycle process in a changing environment is one of the current research hotspots in hydrology (Xia & Tan 2002). Climate change impacts the water cycle due to shifts in temperature, rainfall, evaporation, infiltration, and other processes. It leads to dramatic spatial and temporal changes in water resources, which further result in runoff and sediment changes (Naik & Jay 2011; Micheletti & Lane 2016). Human activities, such as the underlying surface caused by urbanization, water and soil conservation measures, and coal mining, are also important drivers of change in hydrological and sediment elements (Li et al. 2016; Guo et al. 2021).

Quantitative assessment on the impact of different factors on the change in runoff and sediment is of great significance for controlling soil erosion and managing water resources. The commonly used methods include quantitative analysis and model calculation. Quantitative analysis mainly analyzes the changing characteristics of runoff and sediment data in a time series and uses the base period and change period of the basin to distinguish different influencing factors or superimposes the contribution of non-influencing factors and measures (Qi et al. 2020). Model calculation uses Geographic Information System (GIS) to analyze and calculate the geomorphic characteristics of the study area, and calculate the variation law of sewage sediment under different conditions in combination with the hydrological model to remove different influencing factors (Choudhury & Sil 2010; Chang et al. 2023).

The Loess Plateau is the region with the most severe soil erosion in China, where issues related to runoff and sediment are particularly acute. In recent years, many scholars have conducted research on simulation and attribution analysis of runoff and sediment under changing environments. Tian & Li (2010) found that the construction of a soil-retaining dam has great potential to reduce soil erosion in the Loess Plateau. Li et al. (2019b) found that small and medium-sized soil-retaining dams have a strong ability to retain sediment, while large soil-retaining dams play an important role in sediment retention. However, up to this point, the majority of analyses regarding the impact of soil-retaining dams on changes in runoff and sediment have relied on mathematical statistical analysis. This article attempts to establish a high-precision runoff and sediment simulation model based on SWAT. In addition, it investigates the impact of large-scale construction of water conservation measures in the upper reaches of the Fenhe River during the 1990s, with particular focus on the impact of soil-retaining dams.

Statistical methods are used to analyze the extended runoff data series to delineate the historical baseline period from the comprehensive impact period. The SWAT model is constructed and the precision of runoff and sediment simulation is verified. Subsequently, to account for the effects of climate change, land use changes, and soil-retaining dams, a SWAT model for the comprehensive impact period is developed. This study innovatively incorporates soil-retaining dams modeled as reservoirs and ponds into the SWAT model to simulate runoff and sediment. This method improves the accuracy of runoff and sediment simulation in typical watersheds on the Loess Plateau. Various simulation scenarios are established to isolate and assess the individual effects of distinct factors on runoff and sediment variation. The research results can inform water resource management, protection, and governance in areas with soil erosion.

The region had suffered from severe soil erosion problems for a long time, resulting in serious land degradation on the Loess Plateau (Tian et al. 2016). The local conditions indicate that a unique and comprehensive control of soil and runoff loss was carried out in the upper basin of the Fenhe River. These included the construction of soil-retaining dams, which had a significant impact on the runoff–sediment process. A total of 91 main dams and 65 medium-sized dams were constructed, with a total design capacity of 91.6 million m³ and a designed silt design capacity of 61.3 million m³. The area controlled by the soil-retaining dams is 760.2 km², accounting for 14.4% of the total area of the basin.

The data used to establish the SWAT model in the study area are shown in Table S1. The main land use types include cultivated land, grassland, forest land, urban and rural residential land, and water area. According to the land use distribution maps in 1980, 2000, and 2020, the changes of each type of land use are shown in Figure S1. Grassland, cultivated land, and forest land account for more than 95% of the study area. From 1980 to 2000, the proportion of various types of land use showed little change. From 2000 to 2020, the change is mainly manifested in the slight decrease of grassland, forest land, and water area, which all increase to urban and rural residential land.

According to the book Shanxi Soil (Liu et al. 1992), there are seven main soil types in the upper basin of Fenhe River – brown soil, cinnamon soil, cultivated loessial soil, skeleton soil, mountain meadow soil, fluvo-aquic soil, and water area (Figure S2). The majority (76.21%) of the study basin consists of cinnamon soil and cultivated loessial soil, which are loose and soft, and therefore easily washed away and eroded.

First, the long-term runoff and sediment data of three hydrological stations are calculated, and the mean, median, maximum, minimum, and skewness of the long-term hydrological data are analyzed. The trend analysis of the time series of hydrological data is carried out using the Sen Slope and Mann–Kendall trend tests. Beforehand, T-test and Pettitt's test were used to analyze the mutability and judge the mutation points of the historical runoff and sediment series. The Mann–Kendall test is used to supplement the verification and determine the mutation of the hydrological sequence. The results are used for the demarcation of different simulation periods.

The model calibration uses the measured runoff and sediment data of three hydrological stations in the upper reaches of the Fenhe River. The three stations include the Jingle station that manages the main upstream flow, the Shangjingyou station that oversees the biggest tributary, and the Fenhe River Reservoir station that regulates the outflow of the entire basin. The control areas of the three stations are 2,799, 1,140, and 5,268 km², respectively. In addition, the distribution of these stations is relatively uniform. In general, although there are only three hydrological stations, they can provide a basic understanding of the overall runoff and sediment situation in the entire basin.

The computer program SWAT-CUP is used to calibrate model parameters. It includes a variety of calibration methods, and the SUFI2 algorithm is used here. First, the sensitivity of runoff parameters is analyzed and calibrated to a high simulation precision. Then, the model established with calibrated runoff parameters is further used to simulate the sediment transport processes, and the sensitivity of sediment parameters is analyzed and calibrated.

Three indexes (Equations (4)–(6)) were used to evaluate the accuracy of runoff and sediment simulation.

• (1) Nash–Sutcliffe Efficiency (NSE):

  

  (4)
• (2) Correlation coefficient (R²):

  

  (5)
• (3) Relative deviation (Re):

  

  (6)

Here, Q_m,i is the simulated value, Q_o,i is the observed value, is the average of the observed value, and is the average of the simulated value.

The accuracy of the specific model is judged according to Moriasi et al. (2007), as shown in Table S2. In addition, it is generally believed that when R² >0.6, the model is reliable (Yue 2020).

Compared to the 1980s and 1990s, the rules of runoff and sediment transport after 2000 in the study area have changed greatly. This is related to the implementation of a large number of water conservation measures in the basin during this period. Among these measures are soil-retaining dams, a form of water conservation and sediment reduction project, which has had a profound impact on the change in runoff and sediment in the upper reaches of the Fenhe River. According to the subbasin division, all the soil-retaining dams are generalized into two forms: reservoirs and ponds.

Large soil-retaining dams with a design capacity of more than 1 million m³ are added to the SWAT model in the form of reservoirs. However, because the ponds can be defined in each sub-watershed, the inflow water is produced from the subbasin, and the flow of other subbasins is not accepted. This feature is in line with the working principle of small and medium-sized soil-retaining dams in the study area. Therefore, small and medium-sized soil-retaining dams are integrated at each subbasin mainstream in the form of ponds.

The validity of this method used to integrate soil-retaining dams into the SWAT model is verified by comparing the simulation accuracy of runoff and sediment with or without the addition of soil-retaining dams. Specifically, this refers to comparing the simulation results of SD (the scenario which integrates soil-retaining dams into the SWAT model during the comprehensive impact period; details are in Table S5) to the values without soil-retaining dams.

Factors affecting the runoff and sediment processes in the study area mainly include climate change, land use, soil-retaining dams, and other human activities. This paper mainly focuses on the transformation that happened between 2008 and 2017, following the implementation of extensive water conservation measures. Different simulation scenarios (Table S5) are set to analyze the influence attribution of climate change and different human activities on runoff and sediment. The variation ratio of runoff and sediment to the simulated value of S0 is proposed to describe the influence of each factor on runoff and sediment. The impact of each influence factor on runoff and sediment can be captured by the differences between scenarios S1–S5 and S0. The above comparison results are used as a reference standard for whether the addition of soil-retaining dams improves the accuracy of the model.

Sen Slope and Mann–Kendall trend tests were used to calculate the runoff and sediment trends at three hydrological stations. The calculation results are shown in Table 1. In the three stations of the study area, the annual variation of runoff is consistent; it first decreases and then increases. The annual variation of sediment transport is consistent, with a decreasing trend over time, and the overall reduction is significant. The sediment is mostly concentrated in the flood season (May–October), and the sediment transport reaches its maximum in August. There is almost no sediment transport in the non-flood season (November–April).

The mutation test was carried out on the runoff and sediment data of the three hydrological stations, and the results are shown in Table 1. Comparing the runoff mutations of the three stations, the mutation time points of Fenhe River Reservoir Station and Shangjingyou Station are the same, both of which occurred in 1983, while the mutation point of the Jingle Station occurred in 2007. This may be because the Jingle Station is located at the source of the Fenhe River. In contrast, human activities have less impact, fewer engineering facilities, and greater protection, while the impact of human activities on the Shangjingyou Station and the Fenhe River Reservoir Station is more significant. The mutation time of the sediment sequence of the Jingle Station and the Shangjingyou Station is the same, both in 1997, which is consistent with the end of the first phase of soil and water conservation comprehensive management planning project in the upper reaches of Fenhe River in Shanxi Province.

After several parameters were simulated, the most sensitive 12 runoff parameters (Table S3) and 12 sediment parameters (Table S4) were selected for parameter calibration. The parameter values after calibration are also listed in Tables S3 and S4.

According to the evaluation index values of each hydrological station in the calibration and validation periods, the SWAT model has good accuracy in runoff and sediment simulation in the study area. The model can be used to further verify the method of adding soil-retaining dams in the comprehensive influence period and conduct attribution analysis.

The soil-retaining dams were then added into the model (i.e. scenario SD), according to the former proposed method, and the model parameters were recalibrated. It was found that the simulation results were effectively improved, and the simulation accuracy requirements of ‘Satisfactory’ or ‘Good’ could be met (Figure 5). The results demonstrate that the proposed method to add soil-retaining dams into SWAT is reliable and reasonable.

• (1) Climate change

The average annual rainfall between 1981 and 1990 (i.e. the earlier historical period for sediment simulation previously mentioned) was 445.79 mm and during 2008–2017 (i.e. the comprehensive influence period previously mentioned) was 567.62 mm. This demonstrates an increased rainfall of 27.33% in the period 2008–2017 compared to 1981–1990. This shows that the selected baseline period is a relatively dry period, while the comprehensive influence period is a relatively wet period in history, which partly led to the larger runoff in the comprehensive influence period.

The annual evaporation ranged from 961.03 to 4,069.71 mm, and the average annual value was 1,343.21 mm, showing a decreasing trend. The average annual evaporation during 1981–1990 was 1,118.25 mm, and the evaporation during 2008–2017 was slightly higher than that during 1981–1990, which was 1,156.08 mm, with an increase of 3.38%.

• (2) Changes of human activities

The situation of land use change from 1980 to 2020 is shown in Figure S1. Between 1980 and 2020, the area of grassland, cultivated land, forest land, and water area decreased generally, while the area of built-up land increased significantly. In the conversion of various land use types, there is an obvious process of returning farmland to grass, but it is also accompanied by the process of replacing grassland with cultivated land. Moreover, some areas near water bodies are identified as built-up land, which may be because coal mines are being mined near water bodies and are identified as mining land.

The local conditions suggest that, during the period between the earlier historical period and comprehensive influence period, a comprehensive control of soil and runoff loss was carried out in the upper basin of the Fenhe River. This period saw the creation of 327 km² basic farmland terraces, 1,245 km² of vegetation area, 103 new soil-retaining dams, and 190 new silting beach land construction projects. In addition, 18 afforestation programs of more than 10,000 acres and 6 economic forests of more than 1,000 acres were built, and remarkable results were achieved in reservoir sand containment. Three basic projects were carried out: first, the construction of soil-retaining dams and the treatment of desert beach; second, to create basic farmland, implement the conversion of farmland to forest, and improve agricultural production conditions; and third, to build green projects to improve the ecological environment. It follows that the main impact factors of human activities include land use change, soil-retaining dams, farmland terraces, and other measures. The former two play a major role. For this reason, this study analyzed the impact of land use change and the construction of soil-retaining dams, in addition to climate change, while the remaining influencing factors are summarized as other factors.

According to the simulation scenarios presented in Table S5, the input data of the runoff–sediment model were changed to obtain the simulation results, and the influence results of various factors were calculated, as shown in Table 2.

For β_S1, the change direction of runoff and sediment at each station is the same, and they all increased under the influence of climate change. Although the increase in temperature significantly increases evaporation, the increase in precipitation is more significant, making it the main reason for the increase of runoff. The increase in runoff will further intensify the sediment transport, so the impact of climate change on sediment is more significant than runoff.

For β_S2, under the influence of land use change, the runoff and sediment have a significant reduction range, except for the runoff of the Jingle Station. Although the three stations are in the upper reaches of the Fenhe River, due to the different locations of each hydrological station, the land use change has spatial and temporal differences, so the runoff impact results of different hydrological stations vary. Among the results, the construction of water conservation projects such as afforestation and grass planting has played a role in reducing soil erosion. Therefore, the sediment at both sites was significantly reduced, and the reduction of sediment was much greater than that of runoff.

For β_S3, the runoff and sediment of different stations in the upper reaches of the Fenhe River are obviously reduced by the soil-retaining dams. The runoff decreased by 24.86–35.20% while the sediment decreased by 78.93–83.77%. The effect of soil-retaining dams is remarkable, and it reduces sediment more than runoff.

For β_other, under the action of other factors, the runoff and sediment at different stations increased to a certain extent. The runoff at the Jingle Station increased the most (41.26%), followed by that of the Fenhe River Reservoir Station (39.21%).

The indicators of each influencing factor are normalized, as shown in Table 3. The sediment at different stations is affected by different factors to varying degrees. Under the influence of land use, the sediment at the Jingle Station decreased the most, up to −92.81%, followed by the influence of soil-retaining dams, which also reduced the sediment significantly, up to −86.05%. At the same time, under the joint impact of climate change and other factors, sediment in the Jingle Station has increased to different degrees, but in general, sediment still shows a significant reduction. The runoff at the different stations shows varying trends. The runoff at Shangjingyou Station showed a decreasing trend, which was more significantly affected by land use and soil-retaining dams, −99.56 and −160.98%, respectively. In contrast, the runoff at Jingle and Fenhe River Reservoir Stations showed an increasing trend. In the Jingle Station, the land use increased the runoff by 46.46%, while in the Fenhe River Reservoir Station, the land use reduced the runoff by 99.26%.

Referring to the literature related to this research, we found some differences between this study and others. Previous studies mainly used statistical methods to analyze the changes of runoff and sediment. Li et al. (2015) focused primarily on the effects of grass coverage on runoff and sediment dynamics. You et al. (2019) employed the Pair–Copula function to perform a quantitative analysis of runoff and sediment at various sites along the Weihe River. Further studies on runoff and sediment have demonstrated the significant impact of soil and water conservation measures on these dynamics. Ran et al. (2013) used mathematical statistics to analyze and study the sediment reduction effects of different soil-retaining dams in the Dali River basin of the Yellow River Basin. Ju et al. (2021), based on a long series of runoff and sediment data through mathematical analysis, discussed the impact of soil-retaining dams on the runoff and sediment in the flow area. Shao et al. (2021) analyzed the long series of water and sediment in the Jialing River basin by quantitative analysis. The study found that the main factors leading to the decrease of runoff and sediment in the basin were human activities based on soil and water conservation measures. To analyze the effects of climate change and human activities, this study develops a distributed hydrological model, enhancing the SWAT model by incorporating soil-retaining dams modeled as reservoirs and ponds. The results show that the simulation accuracy of the model is improved after the soil-retaining dam is added. The proposed method holds potential for broader application in river basins characterized by extensive water conservation measures, particularly those with a significant number of soil-retaining dams.

The uncertainty in hydrological model-based runoff simulation is commonly attributed to the models themselves (Singh et al. 2014; Fu et al. 2015) and the selection of relevant parameters (Fu et al. 2015; Jalowska & Yuan 2019). In this study, the accuracy of the constructed model was further improved by calibrating the parameters to reduce the uncertainty of the simulation. By constructing the historical base period model, the applicability of the SWAT model in the upper reaches of the Fenhe River is verified. To study the influence of soil-retaining dams on the change in runoff and sediment in the upper reaches of the Fenhe River, the large soil-retaining dams are added to the SWAT model in the form of reservoirs and the small and medium-sized soil-retaining dams in the form of ponds, so as to further improve the simulation accuracy of runoff and sediment by the SWAT model in the comprehensive influence period. On the basis of better simulation of runoff and sediment in the upper reaches of Fenhe River, attribution analysis was carried out.

First, the results indicate that human activities, including land use changes and soil-retaining dams, impact sediment processes by more than 75% and runoff processes by over 70%. Among them, the impact of other unconsidered human factors on Jingle runoff reached 43%, indicating that the impact of other factors cannot be ignored. The upper reaches of the Fenhe River are rich in mineral resources. With the large-scale exploitation of resources in recent years, the impact of mineral resources development and utilization on runoff and sediment has gradually expanded. These effects can be further studied and verified in future research. Second, the results show that the accuracy of runoff and sediment simulation in the upper reaches of the Fenhe River can be improved by adding soil-retaining dams in the form of reservoirs and ponds to the model. However, this method has only been verified in the upper reaches of the Fenhe River. It is anticipated that this model approach will be applied to additional river basins to generalize its applicability and refine its methodologies. Third, this study considers the impact of soil-retaining dams and afforestation on runoff and sediment in the upper reaches of the Fenhe River. However, other activities may also have a significant impact on runoff and sediment processes, which will require further research and analysis.

In this paper, the SWAT model is used to simulate the runoff and sediment processes of the upper reaches of the Fenhe River. NSE, R², and Re were used to evaluate the simulation accuracy of the model. Furthermore, through the simulation of different scenarios, the attributions of different influence factors are separated. The conclusion is as follows:

1. Parameters derived from baseline periods are inadequate for accurately reflecting the runoff and sediment processes under scenarios involving soil-retaining dams. The model was updated to reflect land use patterns of 2020 and climatic conditions from 2008 to 2017. Soil-retaining dams were then integrated into the SWAT model, modeled as either reservoirs or ponds, and the parameters were recalibrated to encompass the impacts of climate change and various human activities during the comprehensive impact period. During the comprehensive impact period, the initially ‘unsatisfactory’ simulation results with the original calibrated parameters were improved to ‘satisfactory’ after integrating soil-retaining dams and recalibrating the parameters. The proposed method of incorporating soil-retaining dams into SWAT has proven to be effective and can potentially be applied to similar water conservation basins.

2. The simulation results show that the impact of human activities on runoff and sediment is much greater than that of climate change in the Loess Plateau basin. The analysis shows that human activities influence runoff at each station by more than 70% and sediment by over 75%. The influence factors of land use and soil-retaining dams account for a large proportion of human activities. Under the impact of soil-retaining dams, the sediment and runoff of each hydrological station decreased. Under the influence of land use, runoff and sediment decreased in most areas. However, climate change and other human activities have increased runoff and sediment at each hydrological station.

3. Beyond the impacts of climate change, land use, and soil-retaining dams, runoff and sediment dynamics are also influenced by factors like coal mining, which require detailed investigation in future studies. While this research method has been validated only in a single basin, its broader applicability to other basins requires further investigation to confirm its effectiveness across different hydrological contexts.

This research was supported by the National Natural Science Foundation of China (52379018, U22A20613), the Natural Science Foundation Program of Shanxi Province, China (202103021223113), the Special Funds for Scientific and Technological Innovation Teams of Shanxi Province (202204051002027). The authors would like to thank the editors and reviewers for their valuable comments and suggestions.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.