Study on Runoff Simulation of the Source Region of the Yellow River and the Inland Arid Source Region Based on the Variable Infiltration Capacity Model

Abstract

Hydrological process simulation and rainfall–runoff analysis are important foundations for reasonably evaluating changes in water resources. In this paper, the VIC (Variable Infiltration Capacity) hydrological model was used to simulate runoff without observed data for exploring the applicability of the model in the Kequ, Dari, and Jimai river basins in the source region of the Yellow River, and the Balegen River basin in the inland arid source region. The results show that, from 2015 to 2018, the VIC model had a good simulation effect. The Nash efficiency coefficients (NSE) of the four basins were all above 0.7, and the NSE of the Dari River basin reached 0.93. The relative error (RE) of the three river basins was about 5%, on average, and the RE of the Balegen basin was 6.50%, indicating that the model has good applicability in the study area. Climate perturbation experiments were performed to quantitatively analyze the relationship between rainfall and runoff. The results show that, in the source area of the Yellow River, rainfall and runoff are roughly linearly related. However, in the inland arid source area, temperature has a slightly greater impact on runoff than rainfall.

1. Introduction

As an important natural resource, water is an important factor to control in the construction of the ecological environment. However, due to the impact of climate change and human activities, water shortages are becoming increasingly serious and water problems become an obstacle to the sustainable development of the social economy [1,2,3]. The unique geographical environment and climatic conditions in Northwestern China have a very important impact on water resources and climate change in China and even the world.

Runoff is one of the main sources of water resources and an important part of the water cycle [4]. It is especially important in inland arid regions [5,6,7]. Research on the relationship between rainfall and runoff can clearly analyze the increase and decrease in runoff under climate change, so as to plan and manage the sustainable development of water resources. The driving factors of runoff mainly include two aspects: climate change factors and human activity factors. Climate change factors mainly include changes in precipitation, temperature, relative humidity, and wind speed; human activity factors mainly include changes, through human activities, on the underlying surface of the watershed and human development and utilization of water resources. Moreover, human/anthropic activities produce changes in the surface, affecting runoff and water infiltration, etc. The population of Northwest China is sparse, and human activities have little impact on underlying surface elements. Among them, climate change in the source region of the Yellow River contributes 70% to runoff, while human activities contribute only 30% [8]. Therefore, climate change has become the main factor for runoff changes in the region. Rainfall, as the main factor of climate change, affects the change in watershed runoff and has a direct impact on water resources, ecological environment, and social and economic development [9].

Hydrological models play an important role in the study of hydrological laws. With the continuous development of science and technology, the hydrological model with information technology as its core can quantitatively study complex hydrological processes, such as rainfall, evaporation and runoff [10,11]. The VIC (Variable Infiltration Capacity) model is a distributed hydrological model jointly developed by researchers at the University of Washington, University of California, Berkeley, and Princeton University, based on the ideas of Wood and others. This model can simulate both energy balance and water balance in the water cycle, making up for the lack of traditional hydrological models describing energy processes [12]. As a distributed hydrological model, the VIC model has some significant characteristics; for example, it can be combined with DEM (Digital Elevation Model), considering the spatial heterogeneity of hydrological parameters and hydrological processes. Wei Jie et al. [12] studied the variation of precipitation, temperature, and runoff in the upstream of the Yellow River basin, based on the VIC model, and estimated the future runoff. The results show that the average annual precipitation and temperature will increase in the future. Zhu Yuelu et al. [13] used the VIC model and GPP (Gringorten plotting position) algorithm to construct a non-parametric multivariate comprehensive drought index to study the drought characteristics of the Yellow River basin on the annual and inter-annual scale. Guan Xiaoxiang et al. [14] used the VIC model, Xin'anjiang model, WBM model, and GR4J model to simulate the runoff process of the Yellow River basin and compared the simulation results. The results show that the VIC and Xin'anjiang models are superior to the WBM and GR4J models.

At present, most researchers in Northwestern China use hydrological models to simulate the Yellow River basin and its surrounding watersheds, and there are very few studies on runoff simulation in inland arid source regions. In this paper, we used the VIC model to simulate runoff in the Kequ, Dari, and Jimai river basins of the Yellow River source region, and the Balegen River basin in the inland arid source region. The applicability of the VIC model was analyzed and the relationship between rainfall and runoff in the four river basins was quantitatively compared. This model has an important impact on the rational study of water resource changes in Northwest China.

2. Study Area

This paper is supported by the State Key Program of National Science of China (2017YFC0403603), which aims to study water resources in the source region of the Yellow River and the inland arid source region, which is the basis for selecting the study area.

2.1. Source Area of the Yellow River

The Kequ, Dari, and Jimai rivers in the source area of the Yellow River are the first-level tributaries of the Yellow River. They are all located in the southern part of the Guoluo Tibetan Autonomous Prefecture at the junction of the Sichuan, Gansu, and Qinghai provinces in China (Figure 1). Their terrain is butterfly-shaped, with high northwesterly and low southeasterly winds. The Kequ River basin covers an area of approximately 2460 km² within a geographical range of 98°37′–99°04′ E and 33°14′–33°56′ N. The Dari River basin covers an area of approximately 3377 km² within a geographical range of 99°10′–99°53′ E and 33°03′–33°51′ N. The Jimai River basin covers an area of approximately 1858 km² within a geographical range of 99°37′–100°17′ E and 33°19′–33°43′ N.

These three adjacent basins have a very cold and semi-humid climate, with no clearly defined seasons—only cold and warm seasons—and no period that is completely frost-free. The cold season lasts 7–8 months, and there is heavy wind and snow, and even natural disasters. The warm season has a humid climate and lasts 4–5 months. The highest temperature is 23.2 °, the lowest temperature is −34 °C, and the annual average temperature is −0.5 °C. The temperature difference between day and night is 15–25 °C, and the average annual precipitation is 474–540.9 mm. The soil is mainly alpine meadow soil and alpine shrub meadow soil, with low vegetation diversity and low vegetation coverage.

2.2. Inland Arid Source Area

The Balegen River in the inland arid source area is located in the northeast of the Qaidam Basin in the Haixi Mongolian and Tibetan Autonomous Prefecture of Qinghai Province, China (Figure 2), with high northwesterly and low southeasterly winds. The Balegen River basin lies within a geographical range of 96°45′–97°09′ E and 37°20′–37°38′ N. The Balegen River originates from Zongwulong Mountain and has a total length of 77 km. A small reservoir—Huaitou Tala Reservoir—is built at the mouth of the gorge. The watershed area above the dam site is 1025 km², and the average annual flow rate is 1.12 m³/s.

The Balegen River basin belongs to a plateau continental climate, which is characterized by very low temperatures and a lack of oxygen, dry air, less rain and wind, and no distinction between the four seasons of the year. Due to its location on the Qinghai–Tibet Plateau, there is sufficient sunlight in the watershed, the soil is dry, the surface vegetation cover is lower, and the growth is slow.

3. Methods and Data

3.1. VIC Model

The VIC (Variable Infiltration Capacity) hydrological model is a distributed land surface hydrological model developed based on the ideas of Wood et al. [15]. This was improved by Liang Xu et al. [16,17,18,19] to form the VIC–3L model. The model divides the watershed into several grids, and the grids are independent of each other, fully taking into account the spatial heterogeneity of the surface vegetation types, soil water content, and precipitation of each grid [20,21]. The VIC model considers the physical exchange process between atmosphere–vegetation–soil, which can simultaneously calculate the energy and water balances. It has a combined runoff mechanism with infiltration excess runoff and saturation excess runoff [22]. Therefore, the VIC model is widely used in the simulation of runoff in different watersheds. The model itself does not have a confluence mode, so one can choose a suitable confluence calculation. Usually, the unit line method is used for the slope confluence in the grid, and the linear DeSaint–Venant equation is used to calculate the river network confluence [23,24,25,26].

In this study, the Kequ, Dari, and Jimai river basins in the source region of the Yellow River, and the Balegen River basin in the inland arid source region, were divided into 52,73,43, and 30 grids, respectively, at a spatial resolution of 0.083° × 0.083°, and were used for runoff simulation from 2015 to 2018.

3.2. Data

3.2.1. DEM Data

DEM data were obtained from the Geospatial Data Cloud (http://www.gscloud.cn/), which used 90 m resolution DEM data downloaded by SRTM (Shuttle Radar Topography Mission).

3.2.2. Vegetation Data

Vegetation data used was global 1 km land surface coverage data provided by the University of Maryland (

Table 1. The global 1 km land surface coverage data provided by the University of Maryland.

Number	Vegetation Classification	Albedo	Minimum Stomatal Impedance (s/m)
0	Water	/	/
1	Evergreen needleleaf forest	0.12	250
2	Evergreen broadleaf forest	0.12	250
3	Deciduous needleleaf forest	0.18	150
4	Deciduous broadleaf forest	0.18	150
5	Mixed forest	0.18	200
6	Woodland	0.18	200
7	Wooded grasslands	0.19	125
8	Closed shrublands	0.19	135
9	Open shrublands	0.19	135
10	Grasslands	0.20	120
11	Crop land	0.10	120
12	Bare ground	/	/
13	Urban and built-up	/	/

). These data divided the land cover types into 14 types [27,28], referring to the parameters of each model in the global land surface data assimilation system to extract the vegetation coverage data of each grid. Figure 3 and Figure 4 show the vegetation classification data in the study area.

3.2.3. Soil Data

The soil data used were the soil classification of the global 5 min soil data released by FAO (Food and Agriculture Organization of the United Nations) [29] (

Table 2. The soil classification of the global 5 min soil data released by the UN FAO (Food and Agriculture Organization).

Number	Soil Classification	Porosity (m3/m3)	Saturated Soil Water Potential (m)
0	No data/ocean	/	/
1	Sand	0.445	0.069
2	Loamy sand	0.434	0.036
3	Sandy loam	0.415	0.141
4	Silt loam	0.471	0.759
5	Silt	0.523	0.759
6	Loam	0.445	0.355
7	Sandy clay loam	0.404	0.135
8	Silty clay loam	0.486	0.617
9	Clay loam	0.467	0.263
10	Sandy clay	0.415	0.098
11	Silty clay	0.497	0.324
12	Clay	0.482	0.468
13	Salt flats	/	/
14	Inland water	/	/
15	Rock debris or desert detritus	/	/
16	Glaciers	/	/

). Figure 5 and Figure 6 show the soil classification data in the study area.

The soil parameters can be divided into two categories. The first category is related to the soil properties and does not need to be changed after being determined in the model, such as the saturated hydraulic conductivity of the soil Ks, and the saturated soil moisture content (porosity) θs [30]. There are seven soil parameters in the second category, including the depths of the upper, middle, and lower soil layers—d1, d2, d3—the infiltration curve parameter B, and the three parameters related to the base flow—Dm, Ds and Ws [16,17]—which need to be determined in the simulation; the determination of these parameters has a great impact on the runoff. Details on the parameter calibration and sensitivity analysis are shown in Tables S1 and S2 and Figure S1.

The soil parameter file includes the average elevation of each grid, as well as the annual average precipitation used to calculate radiation and relative humidity and the annual average temperature used to calculate the soil temperature [31]. When the VIC model is running, it first reads the parameters of each line of the soil input file to obtain the soil information in the corresponding grid, and then extracts the vegetation parameters of the grid with the same ID number from the vegetation file according to the ID number of the grid. Therefore, the uniqueness and consistency of each grid ID in the soil input file and vegetation input file must be guaranteed.

3.2.4. Meteorological Data

The meteorological forcing data describe the meteorological parameters of each grid, including precipitation, maximum and minimum temperature, and wind speed. This is another important part in addition to the soil parameters. The accuracy of meteorological forcing data largely determines the accuracy of the model simulation.

Because there were very few meteorological stations in the study area, ITPCAS (Institute of Tibetan Plateau Chinese Academy of Science) (spatial resolution, 0.1°; time resolution, 3 h) was adopted in the source area of the Yellow River [32], and the ground daily grid data set (spatial resolution, 0.5°; time resolution, 1 day) of China Meteorological Data Service Center (http://data.cma.cn/en) was adopted in the inland arid source area. According to the ID number of each grid, the corresponding meteorological forcing data were generated.

3.2.5. Control File

The control file plays a guiding role in the operation of the model. This determines the time step of the model operation, the start and end dates of the simulation period, the path of the soil file, vegetation file, and meteorological forcing data, etc. [33].

3.2.6. Actual Measured Runoff

This study simulated the runoff of the Yellow River source area and the inland arid source area from 2015 to 2018. Due to the geographical limitations and the harsh climate environment, there was a lack of hydrological station data in the study area. According to the research of the hydrological model in the area without actual measured data [11], and the study on similar watershed division with the parameter transplantation method [34],in the source region of the Yellow River we selected the measured runoff of the Jimai Hydrological Station (99°39′ E, 33°46′ N) in 2015. In the inland arid source area, we selected the inflow of Huaitou Tala Reservoir (96°45′ E, 37°22′ N) in the Balegen River basin from 2015 to 2018. Then, we used the rainfall–runoff model method [35,36,37] to restore it to natural runoff as measured data.

4. Model Results and Analysis

Figure 7 shows the comparison between the simulated runoff and the measured runoff of the study areas in the period 2015–2018. On the monthly scale, the simulated runoff of the model was generally consistent with the measured runoff.

We selected the following two objective functions to evaluate the runoff simulation effect:

The Nash efficiency coefficients (NSEs): The Nash efficiency coefficient is used to describe the degree of fitting between simulated flow and measured flow. It reflects the consistency of the curve. The closer to 1, the better the fitting effect.

$$NSE=1-\frac{\Sigma (Q_{i,s}-Q_{i,o}{)}^2}{\Sigma (Q_{i,o}-\overline{Q_o}{)}^2}$$

The relative error (RE): The relative error is used to reflect the deviation between the simulated flow and the measured flow. This is the difference between each piece of simulated data and the observed data. The closer to 0, the better the simulation effect.

$$RE=\frac{{\overline{Q}}_s-{\overline{Q}}_o}{{\overline{Q}}_o}$$

where i represents the serial number; $Q_{i,s}$ represents the simulated flow sequence; $Q_{i,o}$ represents the actual measured flow sequence; ${\overline{Q}}_o$ is the actual measured average runoff; and $\overline{Q_s}$ is the simulated average runoff.

Table 3. The values of NSE (Nash efficiency coefficients) and RE (relative error).

Study Areas		NSE	RE
The source region of the Yellow River	Kequ River basin	0.74	6.01
	Dari River basin	0.93	5.66
	Jimai River basin	0.75	4.59
The inland arid source region	Balegen River basin	0.71	6.50

represents the NSE and RE of the simulation results in the study area. From Figure 7 and Table 3, we can see the NSEs in the study area were all above 0.7, and the NSE in the Dari River basin reached 0.93. The average RE of the three river basins in the source region of the Yellow River was approximately 5%, and the RE in the inland arid source region was 6.50%. As seen in Figure 7d, occasionally, the simulated value was larger or smaller than the measured value. The underestimated magnitude of change in streamflow in the simulation may be because of the role of reservoir regulation. However, the overestimated magnitude may be because of the data. In the Balegen River basin, due to the complex geographical environment and climatic conditions, the accuracy of the measured data was uncertain. We cannot accurately restore the measured data to natural runoff, so the simulated value is overestimated. Due to the lack of measured data, the Kequ 7(a), Dari 7(b), and Jimai river basins 7(c) cannot be comprehensively analyzed by the simulation results. This issue should be improved in future research.

5. Runoff Analysis in Climate Perturbation Experiments

5.1. Assumptions of Climate Perturbation Experiments

Climate perturbation experiments can be used as a simple and effective method to analyze the impact of climate change on runoff. According to the report “Global Warming of 1.5 °C” [38], which was released by the IPCC (The Intergovernmental Panel on Climate Change) in 2018, this method indicates the feasibility of controlling the global temperature rise to 1.5 °C. Based on this report, the meteorological data in the study area were adjusted using the method of coupling perturbation experiments and hydrological models. We assumed 20 climate change scenarios to analyze the runoff response in the study area. Among them, the temperature increased by 0.5, 1.0, and 1.5 °C, based on the original temperature data, and the rainfall increased by −10%, −5%, 5%, and 10%. The other meteorological parameters were unchanged and the VIC model was run to simulate the runoff.

5.2. Analysis of Runoff Change

The changes in runoff under the climate perturbation experiments are shown below (Figure 8).

The four graphs in Figure 8 show the impact of changes in rainfall on runoff as the temperature increases by 0, 0.5, 1.0, and 1.5 °C, respectively. When the temperature increased, the runoff of the river basin decreased to varying degrees. The rainfall–runoff changes in the four basins shown at 8a–d were approximately linearly related. In the Balegen River basin of (c) and (d), when rainfall decreased by 10%, the amount of runoff change was relatively large. This may be because the Balegen River basin is in an inland arid area and belongs to a plateau continental climate. The air in the basin is dry and windy. The runoff in the basin is a form of infiltration excess runoff. When the temperature rises and the rainfall suddenly drops, the runoff suddenly decreases.

As shown in Figure 8b, when the temperature increased by 0.5 °C and the rainfall remained unchanged, there was a slight increase in the runoff of the Balegen River basin, which may be caused by the uncertainty of the model and the uncertainty of the meteorological data, or may be due to the reservoir.

In any case, the runoff of each of the three river basins in the source region of the Yellow River decreased when the temperature rose, showing a good linear relationship. In the Balegen River basin of the inland arid source region, a linear relationship was not obvious. When the temperature increased by 1.0 and 1.5 °C and the rainfall decreased by 5%, the runoff changed abruptly. This mutation indicated that, in the Balegen River basin of the inland arid source region, the increase in temperature led to a greater decrease in runoff than rainfall.

In summary, we understand that both rainfall and temperature affected the changes in runoff. However, in the inland arid source area with infiltration excess runoff as the main runoff method, the temperature had a slightly greater impact on runoff than rainfall. In the Kequ, Dari, and Jimai river basins in the source region of the Yellow River, rainfall and runoff were approximately linearly related.

6. Conclusions and Discussion

In this study, VIC hydrological models were constructed for the Kequ, Dari, and Jimai river basins in the Yellow River source region, and the Balegen River basin in the inland arid source region. The results demonstrate that the VIC hydrological model has good applicability.

The climate perturbation experiments in the study area were analyzed and compared to elucidate the relationship between rainfall and runoff in the source region of the Yellow River and the inland arid source region. In the inland arid source area, with infiltration excess runoff as the main runoff method, the temperature had a slightly greater impact on runoff than rainfall. In the Kequ, Dari, and Jimai river basins in the source region of the Yellow River, rainfall and runoff were approximately linearly related.

This research has sustainable development significance for the water resource management of the Yellow River basin in the future. First, compared to rainfall, temperature is a more important factor in the inland arid source region. We must pay attention to temperature changes and, when they are evident, we should take appropriate measures. In addition, the impact of water on vegetation is very large. There are few types of vegetation in the inland arid source areas and the amount of vegetation is small. In order to protect vegetation and prevent disasters, we should pay more attention to changes in runoff. Second, the runoff changes in the source area of the Yellow River directly affect the downstream runoff. We must understand them from the source so as to understand the downstream runoff changes. The source area of the Yellow River is an important ecological protection area. To understand the changes in water resources in the source area of the Yellow River is of great significance to water resource management and efficient water use. Moreover, it is important to protect ecological stability and maintain the sustainable development of the Yellow River basin.

There are some limitations associated with this study. The research area lacks accurate measured data due to factors such as geographic location and climatic conditions. This has a certain impact on the applicability analysis of the VIC model. The climate perturbation experiments are simple and intuitive, but they cannot accurately analyze runoff. These issues should be noted and improved in future research.