The Runoff Evolution and the Differences Analysis of the Causes of Runoff Change in Different Regions: A Case of the Weihe River Basin, Northern China

Abstract

The runoff levels of the major hydrological stations in the Weihe river basin (WRB) have been found to present decreasing trends. However, the conspicuous spatial differences in the hydro-meteorological conditions have led to variations in the rainfall–runoff pattern in each of the sub-basin areas. The aims of this research study were to reveal the main factors contributing to the runoff changes in the different regions—and it has significance in the water resources rational allocation and protection in the different regions. Three statistical methods were used to analyze the law of precipitation and runoffs of five hydrological stations. The SWAT (Soil and Water Assessment Tool) model was used to reconstruct the runoff in the impact period. The effects of climate change and human activity on runoff were separated by comparing measured runoff and reconstructed runoff. The results show that the closer the proximity to the downstream hydrological station, the more the runoff decreased. In the tributaries and upstream hydrological stations (Zhuanhtou (ZT), Zhangjiashan (ZJS), and Linjiacun (LJC)), from 1970 to 2016, the dominant factor of the runoff reduction was determined to be climate change, and accounted for 148.2%, 98.9%, and 90.5%, respectively. In the hydrological stations of middle and lower reaches (Xianyang (XY) and Huaxian (HX)), the contributions of the climate change to the runoff reduction were 49.7% and 44.3%, respectively, and the impacts of human activity accounted for 50.3% and 55.7%. The impacts of human activity on the runoff reduction were slightly greater than that of the climate change. Due to the different leading factors affecting runoff change in the basin, in response to future climate change, for tributaries and upstream areas, land use should be rationally planned to achieve the optimal balance of water volume in each part of the basin, which is of great significance to the protection and utilization of water resources. As for the middle and downstream regions, reasonable planning should also be focused on the amount of water withdraw, water resource allocations, and water conservancy project construction. According to the factors affecting runoff, corresponding strategies are proposed for different regions, which have important research significance for the protection and sustainable development of watershed water resources.

1. Introduction

Climate change and human activity affect or change original water cycle processes to some extent [1,2]. A basin water cycle is an embodiment of the dynamic balance of the water quantity in a basin. The changes in a water cycle are most directly reflected by changes of runoff [3]. Due to the negative impacts of global climate changes and high-intensity human activity, the conducting of relevant studies has become a hot issue in the field of hydrology research [4]. Various methods have been implemented to evaluate the influences of climate change and human activity on the annual runoff patterns of river basins. For example, the elastic coefficient method [5], the cumulative slope method [6], the precipitation–runoff regression relation method [7], the Budyko conceptual model analysis method [8], the method of reconstructing natural runoff by using hydrological model [9,10,11], and so on.

China’s Weihe river basin (WRB) is located in the north–south climate transition zone and has been a relatively intense area of human activity since ancient times. It has been found that with negative impacts of global climate change and human activity, the runoff in the WRB has declined by 35% since the 1980s [4]. In previous studies, in order to reveal the possible causes of the runoff changes, researchers have conducted runoff attribution analyses from different perspectives. Three general categories were introduced in the aforementioned studies. The first involved the reconstructing the natural runoff scenarios or setting-up different scenarios through modeling techniques and then revealing the influences of different identified factors. For example, Chang et al. [4] used the Variable Infiltration Capacity (VIC) Macroscale Hydrologic Model to reconstruct natural runoff patterns during the perturbation period of human activities, and then conducted a quantitative attribution analysis of the long-term and interdecadal changes in the runoff patterns of the WRB. Li et al. [12] combined different simulation scenarios by using the SWAT model and then compared the results of the different scenarios to reveal the negative impacts of climate change, land usage, and other direct human activities on the watershed runoff. The second category involved the implementation of regression analyses methods for the precipitation runoff to reveal the contributing factors without reconstructing the natural runoff conditions. The results of the study conducted by Du et al. [7] indicated that the runoff patterns and climate change were highly consistent, and the regression relationship between the climatic factors and runoff was barely affected by human activity. It was found that the effects of the climatic factors and human activities on the runoff of the WRB could be directly isolated from the runoff according to the regression relationship without reconstruction of the natural runoff conditions. The third category included parameter variations or statistical analyses. For example, Jiang et al. [13], Jingjing et al. [14], and Wu et al. [15] adopted the Budyko-type equations of conceptual models to analyze the runoff changes in WRB from perspectives of the model parameters, climate change, and human activity, respectively. In another related study, based on a water balance principle, Zhao et al. [16] used the elastic coefficient and cumulative slope methods to analyze the factors contributing to the runoff changes in the WRB. The results of all of the aforementioned studies have promoted an understanding of the runoff evolution in the entire area.

It is well known that different climatic conditions and geomorphic types cause different mechanisms of rainfall–runoff in different regions, and the impacts of climate change and human activities on runoff are also known to be different. Previously, the majority of researchers have mainly been focused on the attribution analyses of the total runoff amount change over the entire basin. However, a few studies were conducted on the changes in the runoff patterns in different regions at small sub-basin scale. Take the WRB as an example, from south to north, the northern slopes of the Qinling Mountains, Guanzhong Plain, and Loess Plateau exist, which result in major differences between northern and southern climatic conditions, as well as variations in types of soil and geomorphological conditions [3]. As such, the types of rainfall–runoff characteristics in different regions have been observed to vary greatly [12,17].

Therefore, the entire WRB was divided into five regions (sub-basins) according to the water resources division of China. In this study, the runoff patterns from the hydrological stations were targeted in each region. Then, through a comparison of the impacts of climate change and human activity on the runoff, the possible influencing factors that may have caused the runoff change of the different hydrologic stations were identified. The results of this research study are of great significance to regional water resource rational allocation, planning, and protection.

2. Study Area and Data

2.1. Study Area

The Weihe River originates from Niaoshu Mountain in Dingxi, which is located in Gansu Province. It is the largest tributary of the Yellow River and flows through the provinces of Ningxia, Gansu, and Shaanxi. The total length of the Weihe River is approximately 818 km. The basin area measures approximately 134,800 km², and the WRB is located between 104°00′ E–110°20′ E and 33°50′ N–37°18′ N. There are two large tributaries of the Weihe River, namely the Jing River and the Beiluo River, which account for 33.7% and 20.0% of the total WRB area, respectively. The Yellow River basin is a first-level water resource region according to China’s water resource regions. The WRB is a second-level water resource region, which includes five three-level water resource regions as the research target sub-areas (namely, five three-level sub-basins), as shown in Figure 1.

The WRB was selected as the area for analysis of the reason of runoff change in different sub-basins, where there is a great difference in rainfall–runoff between the north and south. The WRB is also located in a transition zone of arid and humid regions and has the characteristics of a continental monsoon climate zone. The western section of the basin is high, and the eastern is low. Also, the soil and geomorphology types are relatively complex and variable [12]. Therefore, different natural and man-made regions have been formed in the northern, central, and southern areas of the basin. The northern section is located in an arid climate zone, with the vast loess plateau and a relatively small population. In this area, sparsely distributed vegetation conditions and loose soil exist, which easily cause soil erosion. The central part of the area mainly consists of the fertile Guanzhong Plain, which is characterized by a large population, developed industries, and agricultural activity. There are also urban agglomerations, such as Xi’an, Xianyang, and Weinan. The majority of the large- and medium-sized irrigated areas within the basin are located in the central region. The population has shown great strength in developing the available water resources. The southern section of the WRB includes the Qinling Mountains. The mean annual temperature ranges from 7.8 to 13.5 °C, and summer seasons are known to be hot and rainy. The precipitation ranges from 350 to 800 mm. Also, affected by topography and other factors, the precipitation distribution in the WRB has been observed to decrease from the southeast to the northwest, and the spatial differences are relatively large. The complex and diverse climatic characteristics and soil geomorphological types of the WRB have also led to major differences in the hydrological processes of the different sub-basins [17].

2.2. Data Sources

Runoff data: There are many hydrological stations in the WRB. In the current study, five hydrological stations were selected. These included Zhuangtou (ZT) station, which monitors the Beiluo River sub-basin above it; Zhangjiashan (ZJS) station, which monitors the Jing River sub-basin above it; Linjiacun (LJC) station, which monitors the sub-basin of the upper Weihe River; Xianyang (XY) station, which monitors the middle Weihe River sub-basin; and Huaxian (HX) station, which monitors the sub-basin of the lower Weihe River sub-basin. The time period examined in this study was from 1960 to 2016. Additionally, the data were selected by considering their representativeness, continuity, and availability. The temporal resolution of the selected runoff was monthly. All of the measured runoff data were obtained from the Hydrological Date Year Book of the People’s Republic of China. The distribution and location of hydrologic stations are shown in Figure 1.

Meteorological data: The data of 29 evenly distributed sites from 1956 to 2016 were selected by screening the weather stations in the WRB. The obtained data covered the entire research area. Figure 2 shows the distribution of each site. The temporal resolution was daily. The meteorological data included precipitation, maximum temperature, minimum temperature, wind speed, relative humidity, and sunshine hours. Meteorological data from 1960 to 2016 were mainly used to establish the hydrologic model and analyze the temporal distribution of precipitation in the WRB. The data from 1956 to 1959 was used to warm up the SWAT model and improve the initial simulation accuracy, which would not participate in the calculation of data analysis results. The meteorological data were obtained from the National Meteorological Information Center (http://cdc.cma.gov.cn).

Other data: The spatial raster data for the SWAT model included digital elevation data (DEM), soil data, and land use data. The scale and spatial resolution of the DEM and soil data were 1:1,000,000 and 1 km, respectively. The DEM and soil data were provided by the Cold and Arid Regions Sciences Data Center at Lanzhou (http://westdc.westgis.ac.cn). The soil data were obtained from the China Soil Map Based Harmonized World Soil Database (HWSD). The land use data with a 1 km spatial resolution were provided by the Data Center for Resources and Environmental Science of the Chinese Academy of Science (http://www.resdc.cn). Detailed information about the type of the soil and the land use is shown in Figure 3 and Figure 6.

3. Methodology

3.1. Methods Used for Determining the Evolution Laws of the Precipitation and Runoff

In this section, the methods of the tendencies and change point analysis for precipitation and runoff are detailed. Mann–Kendall trend and linear trend methods were used in the trend analysis, and two methods of Mann–Kendall change point testing and accumulated variances were used in the runoff change point analysis. The linear trend and accumulated variance methods are commonly used in these types of analysis processes. Detailed descriptions of the aforementioned methods can be found in other related studies [18,19,20,21].

3.1.1. The Mann–Kendall (M–K) Trend Test

The sequence is $x_1,x_2,x_3,\dots ,x_n$. Firstly, pick all the dual values $x_i$ and $x_j$, ($j>i$; $i=1,2,\dots ,n-1$; $j=i+1,i+2,\dots ,n$). $p$ is the number of times $x_i$, which is greater than $x_j$. The mathematical expected value of $p$ is calculated by Formula (1):

$$E(p)=\frac{1}{4}n(n-1)$$

(1)

Then, construct the statistics of the M–K iteration correlation test Formula (2):

$$U=\frac{\tau }{{[Var(\tau )]}^{1/2}}$$

(2)

In the formula, $\tau =\frac{4p}{n(n-1)}-1$, $Var(\tau )=\frac{2(2n+5)}{9n(n-1)}$, $n$ is the number of the sequence.

As the length of the sequence increases, the value of $U$ is quickly converged to a standard normal distribution.

$U_{\alpha /2}$ is the critical value of the standard normal distribution with a probability exceeding $\alpha /2$. When $\left|U\right|>U_{\alpha /2}$, the change trend of the sequence is significant. In this study, when the absolute value of $U$ is greater than or equal to 1.96, it shows that the Mann–Kendall trend test is significant at the 95% confidence level [19,20,21].

3.1.2. The Mann–Kendall (M–K) Abrupt Change Point Test

The Mann–Kendall change point test is a non-parametric statistical test. A non-parametric test is also called a non-distributed test. It has the advantage that it does not require a sequence to follow a certain distribution rule, and it is not disturbed by a small number of outliers. It is more suitable for type variables and sequential variables and is easier to calculate. The method is based on the sequence being independent and continuous. A brief description of the Mann–Kendall change point test method in this section is as follows:

For the sequence of $x_1,x_2,x_3,\dots ,x_n$ with sample sizes, construct a sequence of $S_k$ using Formula (3):

$$S_k={\displaystyle \sum _{i=1}^k{h}_i}(k=2,3,\dots ,n)$$

(3)

Here, $h_i=\{\begin{array}{c}+1\\ 0\end{array}\begin{array}{c}\\ \end{array}\begin{array}{c}{x}_i>x_j\\ x_i\le x_j\end{array}\begin{array}{c}\\ \end{array}\begin{array}{cc}(i=1,2,\dots ,n)& (j=1,2,\dots ,i)\end{array}$. Then, by assuming that the sequence distribution is randomly independent, the statistics are as follows:

$$U{F}_k=\frac{[S_k-E(S_k)]}{\sqrt{Var(S_k)}}(k=2,3,\cdots ,n)$$

(4)

In the formula, $U{F}_0=0$. $E(S_k)$ and $Var(S_k)$ are the mean value and variance of the count accumulation, respectively. The sequence $(x_1,x_2,x_3,\dots ,x_n)$ is independent and continuous. $E(S_k)$ and $Var(S_k)$ can be calculated by the Formulas (5) and (6), as shown:

$$E(S_k)=\frac{n(n-1)}{4}$$

(5)

$$Var(S_k)=\frac{n(n-1)(2n+5)}{72}$$

(6)

Here, $U{F}_k$ is based on the above formula in the sequential order of $x_1,x_2,\cdots ,x_n$. The method of calculating $U{F}_k^{\ast }$ is the same as that of $U{F}_k$, where the sequence is inverted to reverse sequence columns $(x_1,x_2,\cdots ,x_n)$. $U{B}_1=0$. $U{B}_k=-U{F}_k^{\ast }(k=n,n-1,\cdots ,1)$. If there is an intersection point between $UB$ and $UF$, the intersection point is the change point. When the sequence is calendar year data, the intersection point corresponds to a year of the change point.

3.2. Method Used for Separating the Effects on the Runoff Patterns in the River Basin of the Climate Change Conditions and Human Activities by Reconstructing the Runoff with the Hydrological Model

3.2.1. The Basic Theory for Separating the Effects of Climate Change and Human Activity on Runoff

Monitored runoff shows the influence of climate change (decreased precipitation and increased evaporation) and human activity (such as water withdrawal and change in the type of land use). Analysis of hydro-meteorological changes provides an understanding of changes in the hydrological system over the past 50 years, but identifying the causes of these changes is a challenge. To separate the influence of climate change and human activity, the runoff sequence is divided into two periods, the base period and the change period, using the runoff sequence change point and the rainfall–runoff relationship. Taking 1960–1969 as the base period and using the calibrated SWAT model to simulate the runoff from 1970 to 2016, while maintaining the land use unchanged, the detailed exposition is shown as Figure 4. The red line (R1) represents the measured runoff in the base period from 1960 to 1969. The blue dotted line (R0) is the simulated runoff by SWAT model in the change period, which is effected by climate change. The black curve (R2) is the measured runoff in the change period, which is influenced by climate change and human activity.

The total change of runoff affected by climate change and human activity is $\Delta R$, which is the difference between $R_1$ and $R_2$. The influence of climate change on runoff is $\Delta R_C$, which is the difference between $R_1$ and $R_0$. The $\Delta R_H$ is the influence of human activity and is the difference between $R_0$ and $R_2$ [4,22].

$$\Delta R=R_2-R_1=\Delta R_H+\Delta R_C$$

(7)

$$\Delta R_H=R_2-R_0$$

(8)

$$\Delta R_C=R_0-R_1$$

(9)

According to the total amount of the runoff change ($\Delta R$) during the change periods (compared with the base period), the contribution of climate change (${\eta }_H$) and human activity (${\eta }_C$) were calculated. The specific calculation formulas are as follows:

$${\eta }_H=-\frac{\Delta R_H}{\Delta R_H+\Delta R_C}=-\frac{\Delta R_H}{\Delta R}\times 100\%$$

(10)

$${\eta }_C=-\frac{\Delta R_C}{\Delta R_H+\Delta R_C}=-\frac{\Delta R_C}{\Delta R}\times 100\%$$

(11)

From the above, the impacts of climate change and human activity on runoff can be separated by Formulas (7)–(9). Then, the Equations (10) and (11) can be used to digitalize the percentage of the influence from climate change and human activity on the runoff change. The important step is to set up the SWAT model for reconstructing the runoff during the change period, that of the R₀.

3.2.2. Brief Introduction and Establishment of the SWAT (Soil and Water Assessment Tool) Model

The SWAT (Soil and Water Assessment Tool) model was developed in 1994 by the Agricultural Research Center of the United States Department of Agriculture (USDA), and it is a semi-distributed model based on physical process development [23,24]. The SWAT model adopts a daily step-size calculation and can output three types of step-size (daily, monthly, and yearly). It can be used to simulate and analyze the changing climate conditions, land usage, soil types, management measures, water, sediment, and non-point source pollution of a basin [10,22]. As hydrological simulation methods can take complex hydrological processes into consideration [25], and the hydrological variables and the evolution laws can be accurately expressed by this method, it is popular for the quantitative assessment of the impacts of climate change and human activities on runoff [26].

In the SWAT model, the whole watershed is first divided into several sub-basins and then further divided into hydrological response units (HRUs). These HRUs are defined as uniform spatial units with unique soil, land use, and slope characteristics, which can describe the spatial heterogeneity of land cover and soil types within each sub-basin. The HRU is the smallest computing unit in the SWAT model. Based on the water balance equation, the hydrological components (such as precipitation, actual evaporation, soil water content, surface runoff, entering the vadose zone from the soil profile, return flow, etc.) were simulated in each HRU. The surface runoff of each HRU was summed in each sub-basin and then the confluence to the outlet of the watershed by the stream system, calculated by the equation of modified Soil Conservation Service (SCS) curve. A detailed description and the calculating formulas are given in the technical report [27,28] of the SWAT model. Based on the guideline of the SWAT model and the characteristics of the WRB, we set up the SWAT model and established the spatial database. The main steps were as follows: Firstly, we generated the stream confluence system and divided the watershed (the WRB) into several sub-basins according to the terrain, DEM, and river channels. Secondly, we defined the raster of the land use data, the soil data, and the slope data in the WRB. Finally, the sub-basins were divided into multiple hydrological response units (HRUs) with unique land use, soil, and slopes in the WRB.

In the steps of setting up the model, based on the DEM data, the WRB was divided into 121 sub-basins by the SWAT model (Figure 5). According to the requirements of the land use type of SWAT model database, the collected raster of the land use data was reclassified into eight categories, which included cultivated land, forest land, grassland, and so on, as shown in Figure 6. The HWSD soil database is a USDA standard soil database which does not require soil particle size reclassification. Most soil parameters can be used to directly set up a SWAT model database; however, some of the soil parameters (such as the SOL_BD, SOL_AWC, SOL_K, and so on) were calculated by SPAW software [29], which could not be directly extracted from the HWSD database. Therefore, the soil database of the SWAT model was successfully established. According to the DEM and the “Technical specification for investigation of land use status” issued by the National Agricultural Zoning Committee in 1984, the WRB was divided into two distinct slopes, which were >2° and <2°, respectively. According to the land use raster, soil raster, and slope raster, the HRUs were generated. Then, in order to simplify the HRUs and improve the operational speed, the thresholds of HRUs were set as 10% for land use, 15% for soil, and 10% for slopes within a sub-basin according to the recommendation of Zhao et al. [30]. In total, 850 HRUs were generated in this study.

The SWAT model’s weather database contained meteorological data and a weather generator (WGEN). The meteorological data included daily precipitation, maximum temperature, minimum temperature, average wind speed, relative humidity, and solar radiation. The solar radiation data were calculated by the sunshine hours [31,32], and then the meteorological data were organized into a format that could be recognized by the SWAT model. The weather generator is a meteorological parameters database calculated from meteorological data. A detailed description of the meteorological parameters was introduced as a technical report [28]. The role of the weather generator (WGEN) is to generate climate data and fill in gaps.

The model was calibrated by the scale of the sub-regional (the five sub-basins were previously detailed in Figure 1). In this study, an SUFI-2 algorithm in the SWAT Calibration and Uncertainty Programs (SWAT-CUP) tool was used to perform a parameter sensitivity analysis, a parameter determination, and a model verification for runoff processes [33,34]. Then, according to the results of the parameter sensitivity analysis and the recommends of Zuo D et al. [3] and Chang et al. [4], 19 parameters (including CN2, GW_DELAY, SOL_Z, and so on), which had been determined to have the greatest influences on the runoff, were selected to calibrate the SWAT model [35]. Due to the fact that the basin area was relatively large, and the landform and soil types were complex and diverse, a method of “first tributary then main stream, first upstream, and then downstream” was adopted to calibrate the parameters of the model. The specific method was as follows: Firstly, the parameters of each tributary and upstream station were independently calibrated (ZT station, ZJS station, and the upstream LJC station), which output the optimal parameters of each station-controlled sub-basin. The optimal parameters were used to verify the simulation effects of single stations. If the simulation of runoff was acceptable, the optimal parameters of the upstream and tributaries were input into the SWAT model. Then, the parameters of the sub-basin between the LJC and XY stations of the main stream were calibrated. Similarly, the optimal parameters of each sub-basin were finally determined. In this process, we also considered the uncertainty factors of the model. The SUFI-2 algorithm was used to perform multiple iterative calculations within the range of model parameters, and then the final parameter range and a set of optimal parameters with better simulation results were obtained. The parameters and their ranges are shown in

Table 1. Initial iteration ranges and optimal values of the model parameters.

Parameter Name	Brief Description	Initial Range	ZT Station		ZJS Station		LJC Station		XY Station		HX Station
			Final Range	Optimal Value	Final Range	Optimal Value	Final Range	Optimal Value	Final Range	Optimal Value	Final Range	Optimal Value
v_TLAPS.sub	Temperature lapse rate (°C/km)	0~50	5.34~16.1	10.7	6.95~18.9	12.5	36.7~49.8	48.4	3.49~11.3	3.59	8.95~26.9	16.5
v_SLSUBBSN.hru	Average slope length (m)	10~150	10~34.5	16.3	15.5~35.6	15.9	70.8~99.8	88.2	28.8~69.3	45.6	85.3~139	93.8
v_CANMX.hru	Maximum canopy storage (mm H2O)	0~100	0~16.9	7.53	0~14.5	0.07	34~59.8	51.6	33.7~60.6	46.7	27.8~75.9	72
v_ESCO.hru	Soil evaporation compensation factor	0.01~1	0.48~0.63	0.55	0.78~1	0.91	0.68~0.89	0.88	0.74~0.93	0.83	0.01~0.39	0.19
v_EPCO.hru	Plant uptake compensation factor	0.01~1	0.15~0.33	0.29	0.01~0.18	0.08	0.38~0.55	0.51	0.01~0.17	0.06	0.53~0.85	0.61
r_CN2.mgt	SCS runoff curve number	−0.5~0.5	−0.39~−0.2	−0.29	−0.33~−0.15	−0.18	−0.02~0.29	0.12	−0.05~0.41	0.23	0~0.35	0.16
r_BIOMIX.mgt	Biological mixing efficient	−0.5~0.5	−0.23~−0.02	−0.02	0.24~0.41	0.32	−0.06~0.12	−0.01	−0.2~0.13	−0.03	0.12~0.47	0.32
r_SOL_Z().sol	Depth from soil surface to bottom of layer	−0.5~0.5	−0.45~−0.26	−0.36	0.11~0.37	0.32	−0.5~−0.34	−0.49	−0.5~−0.31	−0.48	−0.3~0.09	−0.17
r_SOL_AWC().sol	Available water capacity of the soil layer	−0.5~0.5	−0.5~−0.36	−0.49	−0.37~−0.15	−0.37	−0.22~−0.02	−0.09	0.23~0.47	0.28	−0.29~0.15	0.06
r_SOL_K().sol	Saturated hydraulic conductivity	−0.8~0.8	0.19~0.52	0.24	0.52~0.8	0.6	−0.16~0.11	−0.07	−0.11~0.21	0.02	−0.55~−0.05	−0.37
r_SOL_ALB().sol	Moist soil albedo	−0.5~0.5	−0.25~−0.09	−0.24	−0.43~−0.29	−0.37	−0.02~0.21	0.11	−0.12~0.07	0.03	−0.07~0.31	0.25
v_GW_DELAY.gw	Groundwater delay (days)	0~500	355~464	449.7	331~443	390.1	165~2461	215.6	0~108.4	7.16	39.3~299	183.1
v_ALPHA_BF.gw	Base flow alpha factor (days)	0~1	0.73~1	0.9	0.21~0.48	0.37	0.55~0.72	0.57	0.33~0.56	0.47	0.53~0.92	0.85
v_GW_REVAP.gw	Groundwater “revap” coefficient	0.02~0.2	0.16~0.19	0.17	0.16~0.19	0.18	0.03~0.06	0.05	0.09~0.13	0.1	0.06~0.14	0.09
v_CH_N2.rte	Manning’s “n” value for the main channel	0~0.3	0~0.06	0.01	0.02~0.09	0.08	0.09~0.16	0.14	0.21~0.27	0.22	0~0.08	0.03
v_RCHRG_DP.gw	Deep aquifer percolation fraction	0~1	0.2~0.45	0.39	0.13~0.37	0.33	0.63~0.85	0.84	0.48~0.76	0.68	0.42~0.79	0.67
v_GWQMN.gw	Threshold depth of the water in the shallow aquifer required for return flow to occur	0~5000	3220~4237	3474	73~1162	568.8	1730~2716	2117	3386~4463	3864	1440~3058	1727
v_REVAPMN.gw	Threshold depth of the water in the shallow aquifer for “revap” to occur (mm)	0~500	314~435	371.5	358~435	387.5	345~427	356.9	343~448	384.2	0~169	75.5
v_CH_K2.rte	Effective hydraulic conductivity in the main channel alluvium	0~150	121~150	144.9	19.9~40.0	29.7	111~138	127.3	22~66.2	48.2	29.1~80.5	65.2

NOTE: v_ means the existing parameter value is to be replaced by the given value; r_ means an existing parameter value is multiplied by (1+given value); the detailed description of the parameters was introduced in the SWAT model technical report [27,28].

.

4. Analysis of the Results

4.1. Evolution Laws of the Precipitation and Runoff

4.1.1. Analysis of the Evolution Law of the Precipitation

Due to the difference in surface elevation between the WRB (the difference between the highest and lowest is about 3000 m, as shown in Figure 1), we used the co-kriging interpolation method modified by elevation to process the precipitation data. The co-kriging interpolation method is an extended form of the common kriging method. It uses two or more variables, one of which is a pivot variable and the other is an auxiliary variable. The spatial autocorrelation of the main variables and the interactive correlation between the main and the auxiliary variables are combined for unbiased optimal estimation. A detailed description for this method is given by Hevesi et al. [36]. In addition to longitude and latitude, the spatial distribution of precipitation is sometimes affected by altitude [37]. In this study, we used longitude and latitude as the main variables and geographic elevation as the auxiliary variable. The process of calculating precipitation was as follows: Firstly, daily precipitation data of 29 sites from 1960 to 2016 were summarized into annual data. Then, the co-kriging interpolation method was used to calculate WRB precipitation year-by-year from 1960 to 2016. The annual precipitation sequence of the WRB is shown in Figure 7. The mean annual precipitation in the entire basin was 552.7 mm. From 1960 to 2016, the change rate of WRB’s precipitation was −0.962 mm/a. Through the analysis of the M–K trend test, the U value was determined to be −1.26 (|−1.26| < 1.96). On the change trend of multi-year, the decrease of the whole WRB’s precipitation showed an insignificant decreasing trend.

In order to analyze the spatial distribution of mean annual precipitation in the WRB, we averaged the annual precipitation from 1960 to 2016 at 29 stations. Then, we drew a spatial distribution map of WRB’s average precipitation by using the co-kriging interpolation method (in Figure 8). From a spatial perspective, it was observed that the precipitation was generally lower in the southeast and greater in the northwest (Figure 8). The maximum average annual precipitation in the southeast was determined to be 721.4 mm; the minimum in the northwest was only 405 mm. Therefore, the precipitation difference between the northwest and southeast regions was approximately 320 mm.

4.1.2. Analysis of the Evolution Law of the Runoff Evolution

Figure 9 shows the inter-annual variation process line of the measured runoff at each station in the study area. According to the M–K trend test (

Table 2. Detailed trend patterns of the runoff of each hydrological station in the WRB.

Hydrological Station	Annual Runoff (1 × 108 m3)	Change Ratio (1 × 108 m3/a)	Tendency	M–K Trend Test of the U Value	Critical Value of the U Value	Confidence Levels	Significance Test of Change
Zhuangtou	7.9	−0.09	Decreasing	−4.03	±1.96	95%	Significant
Zhangjiashan	15.4	−0.24	Decreasing	−4.78	±1.96	95%	Significant
Linjiacun	18.6	−0.46	Decreasing	−5.59	±1.96	95%	Significant
Xianyang	36.8	−0.73	Decreasing	−4.15	±1.96	95%	Significant
Huaxian	63.3	−0.93	Decreasing	−3.35	±1.96	95%	Significant

), the measured runoff of the five hydrological stations presented a significant decreasing trend. Meanwhile, the decreasing rates of each station were observed to be different. The annual runoff decreasing rate of the HX station was observed to be the largest (93 million m³/a). The annual runoff decreasing rate of the ZT station was the smallest (9 million m³/a), as shown in Table 2.

The variation in the measured runoff at the hydrological stations showed different characteristics at different periods. By taking the HX Station as an example, which had the largest runoff reduction in the lower reaches of the WRB, it can be seen that by comparing the runoff during the different time periods with the runoff from 1960 to 1969, the 1990s and 2000s displayed the largest reductions, reaching 54.5% and 53.3%, respectively. The reduction during the 1980s was observed to be minimal and only reached 17.7%. Regarding the decline of the overall attenuation of the runoff over the entire basin area, larger changes were observed between the 1970s to the 1980s and the 1980s to the 1990s.

Furthermore, the accumulated variance method and M–K change point testing method were used to identify the possible change points. It was found that the runoff change points of the HX and XY stations of the main stream occurred during the early 1970s and mid-1980s. The change points in the two tributary stations were observed to be concentrated in the 1990s, which indicates that an increased number of change points existed from the upstream to the downstream areas. The years of the possible change points of the runoff evolution in each station are detailed in

Table 3. Detailed change points of the runoff of each hydrological station.

Hydrological Stations	Possible Change Points Years
	Cumulative Anomaly Method	Mann–Kendall Change Point Test
Zhuangtou	1994	1998
Zhangjiashan	1997	1995
Linjiacun	1971; 1993	1990
Xianyang	1971; 1987; 1994	1976; 1984
Huaxian	1970; 1985; 1991	1977; 1986

.

4.1.3. Division of the Base period and Change Period for Attribution Analysis

Precipitation is known to be an important factor that influences runoff changes. Therefore, the changes of precipitation are bound to result in different hydrological processes [35]. In this study, in order to focalize the change points of the runoff in the WRB, reveal the possible causes, and select the base period of the attribution analysis, the cumulative characteristics of the precipitation and runoff of the WRB were analyzed in-depth. It can be seen from the cumulative precipitation curve (Figure 10) and cumulative rainfall–runoff curve (Figure 11) that the slope of the cumulative precipitation line has no obvious inflection point, but the precipitation–runoff cumulative curve displays two obvious turning points, which occurred in 1970 and 1991. Therefore, when combined with the change points and the law of the runoff evolution, it was found that the precipitation–runoff was stable without obvious change points in the 1960s.

Since human activity was mainly concentrated in the middle and lower reaches of the study area, which contained fewer tributaries and upper reaches prior to the 1990s, it was concluded that the influences of human activity on the tributaries and upstream runoff had gradually begun after the 1990s with an increased intensity of human activities. The rainfall–runoff relationship and the law of the runoff evolution illustrated the reasonableness of the runoff change points in the tributaries and upstream areas during the 1990s. In addition, based on the upstream–downstream precipitation–runoff relationship, it was determined that the period ranging from 1960 to 1969 (steady precipitation–runoff relationship) could be used as the base period for this study’s SWAT model set-up and the in-depth attribution analysis, and then reveal the causes of runoff change.

4.2. Attribution Analysis of the Runoff Evolution

4.2.1. SWAT Model Verification and Uncertainty Analysis

(1) SWAT Model Verification

According to the above analysis, the abrupt change points of runoff change in the WRB are around the years of 1970 and 1991. Before 1970, the annual runoff change is relatively stable (no abrupt change point) and relative in the wet period. According to the study of Kannan et al. [38], the period for SWAT model calibration verification should be stable and no abrupt change point. Therefore, from 1960 to 1965 was selected as the parameter calibration period and 1966–1969 as the model validation period. To evaluate the month-by-month simulation results of the SWAT model, the Nash–Suttcliffe (NS), the certainty coefficient (R²), and the percentage bias (PBIAS) were used to verify the simulation results of the model. The detailed computational formulas are as follows:

$$NS=1-\frac{{\displaystyle \sum _{i=1}^n{(Q_{o,i}-Q_{m,i})}^2}}{{\displaystyle \sum _{i=1}^n{(Q_{o,i}-\overline{Q_O})}^2}}$$

(12)

$$R^2=\frac{{\left[{\displaystyle \sum _{i=1}^n(Q_{o,i}-\overline{Q_o})(Q_{m,i}-\overline{Q_m})}\right]}^2}{{\displaystyle \sum _{i=1}^n{(Q_{o,i}-\overline{Q_o})}^2{\displaystyle \sum _{i=1}^n{(Q_{m,i}-\overline{Q_m})}^2}}}$$

(13)

$$PBIAS=\frac{{\displaystyle \sum _{i=1}^n(Q_{o,i}-Q_{m,i})}}{{\displaystyle \sum _{i=1}^n(Q_{o,i})}}$$

(14)

In the formulas, Q_o,i is the measured value, and Q_m,i is the simulated value by SWAT model. $\overline{Q_o}$ is the average value of the measured value, and $\overline{Q_m}$ is the average value of the simulated value by SWAT model.

According to the conclusion and recommendation of Moriasi D N [39], it was generally believed that when the NS and R² are greater than 0.5 and the PBIAS less than 25% in a runoff simulation, the results are acceptable. In the process of using a hydrological model, 0.5 is the kipping value. If the model evaluation index such as NS is less than 0.5, the model cannot well reflect the regional hydrological process, and there would be a large deviation in the process of simulation, which does not apply to the area. The detailed ranges of evaluation indicators were given in the relevant research [40]. The NS ranges are: NS ≤ 0.5 unsatisfactory; 0.5 < NS ≤ 0.65 satisfactory; 0.65 < NS ≤ 0.75 good; and 0.75 < NS ≤ 1.00 very good.

The simulation applicability analysis of the hydrological station is shown in

Table 4. Calibration and verification results of the SWAT model.

Stations	Calibration Period					Verification Period
	NS	R2	PBIAS	P-Factor	R-Factor	NS	R2	PBIAS	P-Factor	R-Factor
HX Station	0.88	0.85	1.92%	0.63	0.44	0.88	0.70	2.73%	0.85	0.71
XY Station	0.80	0.88	8.74%	0.58	0.54	0.79	0.78	6.65%	0.63	0.67
LJC Station	0.68	0.75	11.64%	0.56	0.60	0.64	0.54	18.54%	0.46	0.68
ZJS Station	0.80	0.66	10.87%	0.53	0.68	0.77	0.53	23.67%	0.85	0.95
ZT Station	0.72	0.62	3.77%	0.67	0.80	0.70	0.53	19.67%	0.69	0.93

. From the perspective of the hydrological station with the worst simulation effect, the evaluation indexes of NS and R² are above than 0.5, and the worst result of simulation is satisfactory. From the comprehensive perspective of the all hydrological stations in the WRB, the accuracy of the model is not only satisfactory, but most of the results of the NS and R² are above 0.65, which indicates that the simulation effect of most hydrological stations is good. The PBIAS of each station is less than 25%, and most stations are less than 15%, which meets the requirements. The SWAT model has good applicability in WRB and can well simulate the runoff of the different sub-basins. So it can be used to reconstruct the runoff and separate the effect of climate change and human activity on runoff change.

(2) Uncertainty Analysis

We selected the SUFI-2 algorithm as a method for parameter estimation and uncertainty analysis. The uncertainty of the input data, model structure, parameters, and measured data were considered in the algorithm, which reflected the results within the parameter range after the calibration. The uncertainty interval at the 95% confidence level (95% prediction uncertainty, 95PPU) after the parameter calibration contains most of the measured data. Through Latin hypercube sampling calculation, the total uncertainty of the simulation results was obtained by the cumulative distribution of the output results on the interval of 2.5% (L95PPU) and 97.5% (U95PPU). Among them, 97.5% (U95PPU) and 2.5% (L95PPU) were the upper and lower boundaries of the uncertainty interval (95% prediction uncertainty, 95PPU) [34,41,42].

Two indicators in the SUFI-2 algorithm were used to measure the results of the uncertainty analysis. P-factor is the percentage of the measured data contained in 95PPU interval. R-factor is the average width of the 95PPU interval divided by the standard deviation of the measured data [42,43]. In theory, the range of P-factor is 0–100%, and the range of R-factor is 0-∞. A P-factor close to 1 and an R-factor close to 0 is a simulation that is close to the measured data. According to Abbaspour et al. [42], on the basis of their experience, the value of an R-factor close to 1 would be satisfactory.

The result of the uncertainty analysis varies with the range of parameter values. A small parameter range can obtain narrower uncertainty intervals and can increase the confidence level of the simulation, but it reduces the sensitivity of the parameter variation and causes most of the observation data to fall outside the uncertainty interval. However, a large parameter range can reflect the influence of the parameters on the simulation results, but this results in a wider uncertainty interval, which reduces the confidence level of the simulation. The interval from 2.5% to 97.5% was selected as the 95% confidence interval (95PPU) of the SUFI-2 algorithm. In the process of analysis, the parameters ranges for the uncertainty analysis and the optimal value were detailed in Table 1. The simulation intervals for the uncertainty analysis of monthly runoffs are shown in Figure 12. Two indicators (P-factor and R-factor) indicating the uncertainty of the simulation are shown in Table 4. On the whole, the P-factor in the calibration period is less than in the verification period, but the R-factor in the verification period is closer to 1 than in the calibration period. From upstream to downstream, the simulation uncertainty of the downstream hydrologic station is lower than that of the upstream hydrologic station.

4.2.2. Analysis of the Impacts on the Runoff of Climate Change and Human Activity by Using the Hydrological Model to Reconstructing the Runoff

In the terms of the runoff evolution of each station in the WRB, it can be seen from columns 1 and 5 in

Table 5. Impacts of climate change and human activity on the runoff of each hydrological station in the WRB.

Station	Year	Measured (1 × 108 m3/a)	Simulated (1 × 108 m3/a)	Total Change (1 × 108 m3/a)	Climate Change		Human Activity
					Reduction (1 × 108 m3/a)	Ratio (%)	Reduction (1 × 108 m3/a)	Ratio (%)
ZT (Tributary)	1960–1969	10.10	(Base Period)
	1970–1979	8.34	6.35	−1.76	−3.75	−212.59%	+1.99	+112.59%
	1980–1989	9.20	7.09	−0.90	−3.01	−333.74%	+2.11	+233.74%
	1990–1999	7.49	3.48	−2.61	−6.62	−253.23%	+4.01	+153.23%
	2000–2009	6.01	6.36	−4.09	−3.74	−91.26%	−0.36	−8.74%
	2010–2016	5.81	8.34	−4.30	−1.76	−40.99%	−2.53	−59.01%
	1970–2016	7.47	6.20	−2.63	−3.90	−148.21%	+1.27	+48.21%
ZJS (Tributary)	1960–1969	21.66	(Base Period)
	1970–1979	17.42	13.88	−4.25	−7.78	−183.31%	+3.54	+83.31%
	1980–1989	17.09	14.77	−4.57	−6.89	−150.82%	+2.32	+50.82%
	1990–1999	14.00	11.52	−7.66	−10.15	−132.37%	+2.48	+32.37%
	2000–2009	9.97	13.5	−11.69	−8.16	−69.82%	−3.53	−30.18%
	2010–2016	10.95	18.38	−10.72	−3.28	−30.64%	−7.43	−69.36%
	1970–2016	14.07	14.16	−7.59	−7.51	−98.91%	−0.08	−1.09%
LJC (Upstream)	1960–1969	31.24	(Base Period)
	1970–1979	22.23	18.85	−9.01	−12.39	−137.57%	+3.38	+37.57%
	1980–1989	23.22	19.74	−8.02	−11.50	−143.39%	+3.48	+43.39%
	1990–1999	12.9	12.94	−18.35	−18.30	−99.76%	−0.04	−0.24%
	2000–2009	9.05	15.65	−22.19	−15.60	−70.28%	−6.6	−29.72%
	2010–2016	10.75	20.79	−20.49	−10.45	−51.00%	−10.04	−49.00%
	1970–2016	15.94	17.39	−15.30	−13.85	−90.54%	−1.45	−9.46%
XY (Middle stream)	1960–1969	61.96	(Base Period)
	1970–1979	36.67	47.86	−25.29	−14.1	−55.75%	−11.19	−44.25%
	1980–1989	45.46	52.36	−16.49	−9.60	−58.19%	−6.90	−41.81%
	1990–1999	22.47	36.13	−39.49	−25.83	−65.41%	−13.66	−34.59%
	2000–2009	22.80	45.35	−39.16	−16.61	−42.43%	−22.54	−57.57%
	2010–2016	29.26	54.69	−32.70	−7.27	−22.24%	−25.43	−77.76%
	1970–2016	31.47	46.8	−30.49	−15.16	−49.70%	−15.34	−50.30%
HX (Lower stream)	1960–1969	96.12	(Base Period)
	1970–1979	59.37	77.35	−36.75	−18.77	−51.08%	−17.98	−48.92%
	1980–1989	79.12	88.14	−17.00	−7.98	−46.95%	−9.02	−53.05%
	1990–1999	43.76	61.49	−52.36	−34.63	−66.14%	−17.73	−33.86%
	2000–2009	44.85	76.52	−51.27	−19.59	−38.22%	−31.67	−61.78%
	2010–2016	54.08	93.57	−42.04	−2.55	−6.06%	−39.5	−93.94%
	1970–2016	56.37	78.51	−39.75	−17.61	−44.30%	−22.14	−55.70%

NOTE: “+” means the amount or percentage of runoff h increased; “−” means the amount or percentage of the runoff reduced. In the calculation formula of 3.2.1, R₀, R₁, and R₂ correspond to the fourth column (simulated), the third column (measured) from 1960 to 1969, and the third column (measured) from 1970 to 2016 in the table, respectively.

that the runoff of the five hydrological stations showed decreasing tendencies. There was a decreasing runoff characteristic observed at different stations. When compared with the base period, the total runoff reductions of the main stream stations were greater than those of the tributary stations. Also, the downstream station displayed greater runoff reductions than the midstream and upstream areas during the period ranging from 1970 to 2016. Among these stations, the HX station had the largest reduction (3.975 billion m³), and the ZT station was found to have the smallest reduction (263 million m³). The separation of effects of climate change and human activity on runoff by SWAT model is intuitively shown in Figure 13.

(1) At different Hydrological Station Regional Scales

As for the tributaries and upper reaches (ZT, ZJS, and LJC stations) of the study area, compared with the base period from 1970 to 2016, the dominant factor affecting runoff change was climate change. The contributions of runoff reduction caused by climate change at the ZT, ZJS, and LJC stations were 148.2%, 98.9%, and 90.5%, respectively. About the middle and lower reaches (XY and HX stations) of the study area, especially the Guanzhong Plain located here, there were mainly irrigated and conurbation areas. When compared with the base period from 1960 to 1969, climate change and human activity presented negative effects on runoff during each of the time periods. The ratios of the contribution of climate change to runoff decrease at the XY and HX stations were 49.7% and 44.3%, respectively, and the ratios of the contribution of human activity were 50.3% and 55.70%, respectively. The contributions of climate change were slightly less, with a contribution ratio of less than 50%. The contributions of climate change were less than that of human activity.

It was observed that, due to the large irrigated areas and rapid development of urban agglomerations in the middle and lower reaches (especially following the reform and opening up in that region), the influences of human activity on runoff, such as the continuous increases in human water consumption (water consumption increased by 782 million m³ from 1986 to 2007 [44]) and the construction of water conservancy projects (such as dams and reservoirs), had become increasingly intense. Therefore, human activity had a greater influence on the regional runoff in the middle and lower reaches than on the runoff in the tributaries and upstream areas, and had become the dominant factor affecting the runoff patterns.

(2) At the Different Time Period Scale

For the middle and lower reaches (XY and HX stations), in all periods (1970s, 1980s, 1990s, 2000s, and 2010s), the effect of human activity on runoff was to reduce runoff, and the contribution rate of human activity to runoff reduction shows a generally increasing trend over time. Taking HX station as an example, the contribution rate of human activity to runoff reduction in each period was 48.92% (1970s), 53.05% (1980s), 33.86% (1990s), 61.78% (2000s), and 93.94% (2010s), respectively. As regards the tributaries and upper reaches (ZT, ZJS, and LJC stations), it was found that human activity increased the runoff before the 1990s and reduced the runoff after the 1990s. The contribution rates of human activity to runoff change in each period were 37.57% (1970s), 43.39% (1980s), −0.24% (1990s), −29.7% (2000s), and −49.00% (2010s), respectively, in LJC station.

It was ascertained that the changes in land use type had direct influences on the rainfall–runoff processes. Since the 1970s, China’s land usage has gone through major changes [45,46,47], which can be roughly divided into two stages. During the first stage (from the early 1970s to the middle and late 1980s), major efforts were made to develop agriculture and to reclaim land area, which were dominated by barren mountains and wasteland [48,49]. The second stage (from the 1990s to present) included a period of soil and water conservation treatments, which focused on ecological protection and returning farmland to forest land and grassland [49,50]. The land usage changes in the WRB were no exception. Therefore, based on the results of various related studies [51,52,53,54], the capacity to produce runoff in cultivated land areas is known to be larger than in grassland and forest land under equal climate conditions. The Loess source and hilly sections of the study area were controlled by three hydrological stations (ZT, ZJS, and LJC) at the tributary and upstream locations. The loose soil and sparsely distributed vegetation over those areas may have easily led to soil erosion and water loss conditions. However, after the 1990s, in order to control the regional soil and water losses and protect eco-environments, many engineering projects (such as retuning farmland to forest land and grassland, increased measures for soil and water conservation, and so on) were implemented, which contributed to reduced runoff. Therefore, these factors jointly formed the evolution characteristics of the runoff from 1960 to 2016, which was consistent with the fact that the land usage had been transformed during that period. Therefore, the effect of human activities on the control area of the three hydrological stations (ZT, ZJS, and LJC) was first to increase the runoff and then to reduce the runoff.

(3) The Dominant Factor Which Affected the Runoff Patterns

In the tributaries and upstream, the dominant factor affecting runoff change was climate change. Human activity had little effect on the change of runoff; however, with the passage of time, human activities first increased the runoff and then reduced the runoff. Human activity had a greater influence on the regional runoff in the middle and lower reaches than in the tributaries and upstream and became the dominant factor which affected the runoff patterns. In this study, from the viewpoint of the dominant factors of runoff change in the different regions, it was considered that reasonable planning for the land usage and soil and water conservation was important for the development and utilization of water resources in upstream and tributaries regions. Furthermore, reasonable planning should also be focused on the amount of water withdraw, water resource allocations, and water conservancy project construction in the middle and downstream regions.

5. Discussion and Conclusions

5.1. Discussion

Under the influence of climate change (especially global temperature rise) and high-intensity human activity, water resources in various regions of the world have also undergone differential changes. In this context, taking the WRB as an example, the evolution of runoff was analyzed. The runoff monitored by the five hydrological stations in the basin showed a significant reduction trend, which is closely related to the impact of climate change and human activity in the past 57 years, and the runoff of different hydrological station monitoring areas shows different decreasing laws. While the region is developing at a high speed, it is important and necessary to understand the causes of regional runoff changes and cope with future climate change. In the process of analyzing the causes of runoff change, the determination of runoff abrupt point has an important impact on the results. However, in the existing method of determining the abrupt change point, mathematical statistics analysis is the main method, which may cause some deviation by a single method. Therefore, we used the cumulative anomaly method, the M–K abrupt change point test, and the precipitation–runoff cumulative curve method to jointly determine the year of the abrupt change point. Furthermore, the error in the analysis results was reduced to ensure the true validity of the calculation. Finally, 1969 was determined as the year of the abrupt change point in the evolution of runoff in the past 57 years.

When analyzing the cause of runoff change, the SWAT hydrological model was mainly used to reconstruct the runoff without effect from human activity. According to the water balance principle, the SWAT model can better simulate the runoff by considering the hydrological components in the complex hydrological process [23,24]. However, the accuracy and uncertainty of the SWAT model have some influence on the analysis results of the causes of runoff change. In the process of using the SWAT model, according to the conclusion and recommendation of Kannan et al. [38], Moriasi [39], Abbaspour et al. [41,42], Leandro et al. [43], and so on, the model was strictly tested for parameter calibration and model verification, and the uncertainty of the model was also analyzed. The results show that the model can simulate runoff well and has less uncertainty. The SWAT model has good applicability in the WRB and can be used for analysis of the causes of runoff change.

Through the analysis of the causes of runoff change by using the SWAT model to reconstruct runoff and hydrological sequence analysis method, the results show that for the runoff detected by the HX and XY stations located in the lower reach of the river, the contribution of human activity to runoff change in 1970–2016 was greater than climate change. With the passage of time, the impact of human activity on runoff has been increasing, which is closely related to the rapid economic development, the strong water withdrawal activities, the construction of water conservancy projects, and so on, in the central Guanzhong Plain. According to relevant statistics, the water consumption increased by 782 million m³ from 1986 to 2007 [44]. In tributaries and upstream, the influence of climate change on runoff change was dominant. The impact of human activity on runoff is characterized by an increase in runoff and then a reduction in runoff, which is closely related to the policy of land use [45,46,47,48,49]. Under the policy of land reform and returning farmland to forest and grassland, the area of cultivated land increased first and then decreased. It is well known that the runoff produced under the same precipitation in cultivated land is greater than that in forest and grassland [51,52]. Therefore, the effect of human activity on runoff driven by the change in the underlying surface appears to decrease first and then increase. The otherness between different regions is related to geographical location, climatic conditions, land use types, soil types, social development status, and so on. At tributaries and upper reaches’ region, the topography is mainly hills and loess depressions. The loose soil causes soil erosion, and the social development is relatively backward with little population and small water consumption. The middle and lower reaches’ regions are plain with more developed agriculture, population, large and medium-sized urban agglomerations, and large water consumption.

It should be noted that in this study, the SWAT model—which was established in the base period—was used to reconstruct the natural runoff. In this way, the influences on runoff change of climate change and human activity could be effectively separated. This was based on the scenario analysis of non-interference between climate change and human activity. In fact, there was mutual influence between climate change and human activity. Human activity on CO₂ emissions will have an impact on climate change. Thus, Climate change and human activity interact and restrict each other to some degree. Therefore, the impacts of climate change and human activity on runoff need to be quantitatively identified more accurately, and the impacts of many factors need to be further studied.

In addition, based on the assumption that there was no human influence from 1960 to 1969, human activity began to influence the change of runoff after 1970. However, in fact, there was human activity from 1960 to 1969 (through the analysis of the evolution of the runoff and the change point), which had little or no obvious influence. The data of the period without human activity are not available, so the impact of human activities on runoff is not actual, but is compared with the 1960–1969 level.

5.2. Conclusions

In the current research study, statistical methods were applied to analyze the laws of the precipitation and runoff. A method of reconstructing runoff by the SWAT model was implemented for the purpose of quantitatively evaluating the influences of climate change and human activity on runoff and the corresponding variation from the different hydrological stations. Thus, it revealed the reasons for the decrease of runoff in different areas. The main results obtained in this study are as follows:

(1)

From 1960 to 2016, the annual precipitation showed an insignificant decreasing trend (−0.962 mm/a) with large variance in space (maximum and minimum differences were approximately 320 mm). Meanwhile, the runoff of each station displayed a significant decreasing trend with different reductions observed at the different hydrographic stations. At the same time, this study used the M–K change point test method, along with an accumulated variance method, and combined the precipitation–runoff accumulation curve and the actual changes in runoff in the WRB. The base period was determined to be the period from 1960 to 1969.

(2)

Climate change was the dominant factor for reducing the runoff of the tributaries and upstream areas (ZT, ZJS, and LJC stations), and the contribution ratio of the climate change to the runoff reduction reached 90%. Human activity had a slightly greater impact on the reduction of runoff, which was measured at middle and lower stations (XY and HX stations). The contribution ratios of human activity to runoff reduction were determined to be 50.30% and 55.70%, respectively, at the XY and HX stations.

(3)

Under the perspective of the influence of human activity on runoff, in response to future climate change, for upstream and tributaries areas, rational planning of land use is of great significance to the protection and utilization of water resources. As for the middle and downstream regions, reasonable planning should be focused on the amount of water withdraw, water resource allocations, and water conservancy project construction.

The distinction of the water withdrawal levels, surface vegetation types, and socio-economic developments in different regions have created differences in the dominant factors of runoff variation in different regions of the basin. Corresponding measures are taken to influence the dominant factors of runoff reduction in different regions, which has important research significance for water resource protection and sustainable development in the basin.