An Analysis of Runoff Variation in a Small Basin in the Loess Plateau: Identifying the Variation Causes and Implications for Sustainable Water Management

Abstract

Analyzing the characteristics and causes of runoff variation in a typical small basin is beneficial for ecological restoration in the Loess Plateau. This study employed a series of statistical methodologies to examine the characteristics of meteorological changes and underlying surface evolution in the Qishui River Basin (QRB). To differentiate the impacts of climate change and human activities on runoff variation, we applied the Choudhury–Yang formula and the Double Mass Curve (DMC) method. Subsequently, by incorporating future watershed protection strategies and various SSP scenarios, we utilized the Soil and Water Assessment Tool to simulate future runoff while employing the DMC to identify underlying causes of runoff variation. The results suggested that human activity has a slightly greater impact than climate change on reducing runoff during the historical period, with only a 1% difference. However, this will change in the future as human impact becomes increasingly significant. Human activities such as afforestation have dual effects, encompassing positive effects such as improving water quality and mitigating soil erosion, as well as negative consequences such as diminishing local water availability and exacerbating drought. Effective policies should be implemented, involving the use of appropriate tree species and planting methods, finding an appropriate value of forest area, monitoring and evaluation, etc., in order to ensure that the policies are aligned with the broader social, economic, and environmental goals of the QRB. These findings provide valuable guidance for policy-makers in developing management strategies for future environmental changes.

1. Introduction

The ecological environment in China’s Loess region is very fragile [1,2,3]. The scarcity of water resources, fragmented topography, deteriorated vegetation, loose soil structure, and aggravated soil erosion have exacerbated the conflict between water resources, ecological environment, and economic progress. This has severely hindered regional socio-economic advancement [4,5,6,7]. Under the increasingly intensified impact of human activities and global climate change, the hydrological conditions, water resource shortages, and ecological–environmental problems of the Loess tableland—the predominant form of the loess landform—have garnered significant attention [8,9]. Due to severe soil erosion and extensive gully development, tens of thousands of small basins have formed on the Loess Plateau [10]. These basins serve as relatively independent geomorphic, production, and administrative units where villages are located. The small watersheds significantly impact the local natural landscape and social development. Therefore, scientifically and efficiently utilizing the water resources in these small watersheds on the Loess Plateau holds great positive significance for restoring the local ecological environment [11]. Currently, there is ongoing experimental demonstration and extension work focusing on the comprehensive management of small basins in the Loess Plateau. In 1979, the Chinese government recognized the 3-North Shelter Forest Program as an important national economic construction project to enhance the ecological environment [12]. This program aims to establish large-scale artificial forestry ecosystems in the three northern regions of China, including the northwest, north, and northeast. As part of this project, the control of soil and water loss in small watersheds on the Loess Plateau is a key component. Consequently, examining the historical and future changes in runoff under different ecological conditions, at the level of small basins, holds significant practical value for the study of water resources, soil and water conservation, as well as ecological environment restoration in this region. Existing studies typically focus on the entire region or large-scale watersheds of the Loess Plateau [13,14,15,16,17]. However, small watersheds constitute the most fundamental hydrological unit of the Loess Plateau. Therefore, investigating the causes of runoff variation in a small basin is more conducive to unveiling the mechanism of runoff change in the Loess Plateau. Numerous studies have demonstrated that the variability in runoff is a result of the confluence of climate change and anthropogenic activities [18,19,20,21,22]. For instance, rising temperatures due to climate change can impact regional hydrological processes by increasing potential evapotranspiration and altering precipitation intensity and frequency [17,23,24]. Furthermore, human activities have increasingly significant impacts on runoff due to the development of society. These impacts can be categorized as direct (e.g., water use and reservoir construction) or indirect (e.g., changes in land use patterns) [22]. In recent years, human activities on the Loess Plateau mainly affect runoff indirectly [25,26,27]. The underlying surface, particularly the vegetation coverage, has undergone changes due to human activities in the Loess Plateau. Vegetation plays a crucial role in regulating evaporation, infiltration, and interception. Alterations in vegetation cover can significantly influence the amount and timing of runoff [17,28,29,30]. These changes have resulted in alterations in basin runoff.

In recent decades, the Loess Plateau region has experienced a series of changes in hydrometeorological elements [31,32]. Since the 1980s, some rivers’ runoff has decreased sharply [33]. Many tributaries and end water systems have dried up. Under the condition that the annual rainfall remains relatively stable, the groundwater level in the Loess Plateau and hilly areas has dropped significantly many times [34]. Even in wet years with ample precipitation, large-scale hydrological droughts are still common and cause severe ecological shocks. At present, most studies focus on large-scale watershed studies, and there are few studies on runoff attribution analysis in small watersheds [13,14,15,16,17]. Studying the mechanisms underlying the evolution of water cycle elements in the Loess small basin under the influence of climate change and human activities and revealing the patterns of evolution in the regional underlying surface and runoff is of practical significance for the rational utilization, planning, and management of regional water resources [35,36].

The quantitatively identifying runoff attribution methods mainly include the Choudhury–Yang formula, the hydrological model simulation method, the elastic coefficient method based on the Budyko assumption, and the Double Mass Curve (DMC) method [37,38,39,40,41,42,43,44]. In recent years, in addition to the DMC method, the Choudhury–Yang formula based on the basin hydrothermal coupling balance has been widely used in many Chinese basins due to its relatively simple expression and proven high accuracy [45,46]. Many studies have been carried out on the runoff change and attribution analysis in the Loess area [47,48,49,50,51,52]. However, there still needs to be more quantitatively clarifying the impact of climate and underlying surface change on the runoff variation of small basins. After considering future land use changes, how will runoff be affected under different climate modes? This question has significant reference value for policy-making regarding small watersheds in loess areas undergoing ecological restoration.

The Qishui River Basin (QRB), located in the Loess Plateau, is a typical small watershed in the loess region that experiences severe soil erosion. The precipitation, temperature and vegetation coverage in this area are at an average level, accounting for 94.74%, 105% and 93.56% of the respective average values of the Loess Plateau. Therefore, this area is a suitable representative of the characteristics of small watersheds in the loess region. Taking the QRB as the research area, we aim (1) to analyze the evolution characteristics of meteorological elements and underlying surfaces in the QRB; (2) to conduct the attribution analysis of runoff in the QRB under changing environment in the historical period; (3) to analyze the causes of runoff change under different shared socioeconomic pathway (SSP) scenarios in the future period; and (4) to present sustainable development strategies based on the QRB runoff and underlying surface conditions in the future.

2. Materials and Methods

2.1. Study Area

The QRB, a typical small watershed in the Loess Plateau, was selected to conduct the attribution analysis of its runoff change (Figure 1). The basin covers a total area of 773.5 sq. km and is known for its significant variation in sediment transport, as well as harsh soil and water loss. Soil erosion in the region is severe, and the natural environment is unstable. Additionally, the region is characterized by a continental monsoon climate with uneven spatial and temporal distribution of rainfall. Rainfall and heat occur in the same season, increasing the risk of drought. The area also faces prominent challenges related to the complex interplay between water resources, ecological environment, and economic development, leading to frequent hydrological and meteorological drought events. Selecting the QRB as the research watershed is thus consequential, as it can signify the characteristics of diminutive watersheds in the Loess Plateau region.

2.2. Available Data

2.2.1. Meteorological Data and Digital Elevation Model (DEM) Data

In this study, Yaoxian meteorological station was selected (Figure 1); its protracted series of meteorological data spanning from 1969 to 2018 were acquired from the China Meteorological Data Service Center (https://data.cma.cn/, accessed on 15 December 2022). We mainly used precipitation, temperature (maximum, minimum, and average), average wind speed, sunshine duration, and average relative humidity data at daily scale. The 90 m resolution Digital Elevation Model (DEM) data are derived from the Geospatial Data Cloud (http://www.gscloud.cn/, accessed on 15 December 2022).

2.2.2. Normalized Difference Vegetation Index (NDVI) Data

Chinese annual Normalized Difference Vegetation Index (NDVI) spatial distribution dataset was obtained from the Resource and Environment Science and Data Center [53]. The data have a spatial resolution of 1 km × 1 km and a temporal resolution of one year. This dataset effectively portrays the distribution and change in vegetation cover in different regions of China on both spatial and temporal scales, making it one of the widely used datasets for describing the dynamic changes in vegetation growth [54,55,56].

2.2.3. Observed Runoff Data

Yaoxian station is a hydrological station at the outlet of the QRB; its annual runoff data from 1961 to 2018 were obtained from the Yellow River Basin Hydrological Yearbook of China [57].

2.2.4. Coupled Model Intercomparison Project Phase 6 (CMIP6) Data

This research utilized daily temperature and precipitation data from seven models of the Coupled Model Intercomparison Project Phase 6 (CMIP6), with basic information presented in

Table 1. Seven modes from CMIP6 used in this study.

Model Name	Country	Organization	Spatial Resolution Lat. × Long.
BCC-CSM2-MR	China	BCC	1.125° × 1.1°
CNRM-CM6-1	France	CNRM	1.4° × 1.4°
CanESM5	Canada	CCCMA	2.8125° × 2.8°
INM-CM4-8	Russia	INM	2° × 1.5°
MRI-ESM2-0	Japan	MRI	1.125° × 1.1°
IPSL-CM6A-LR	France	IPSL	2.5° × 1.3°
UKESM1-0-LL	UK	UKESM	1.875° × 1.25°

. The data are derived from https://esgf-node.llnl.gov/search/cmip6/ (accessed on 15 December 2022). We utilized the Bayesian Model Averaging (BMA) method [58] to integrate the seven models, and after conducting 30 rounds of cross-validation, we determined that BMA-based data fusion was more accurate and reasonable than the original data. We also employed the BMA method to downsample the initial data to a spatial resolution of 25 km. Furthermore, we applied the Non-stationary Bias Correction technique to rectify errors in the fused model [59]. This study used the combined scenarios (SSPs-RCPs) data for different shared socioeconomic pathways (SSPs) and representative concentration pathways (RCPs) [60]. Specifically, four combined scenarios were used in this study: SSP1-RCP2.6 (SSP1-2.6), SSP2-RCP4.5 (SSP2-4.5), SSP3-RCP7.0 (SSP3-7.0), SSP5-RCP8.5 (SSP5-8.5). According to the Intergovernmental Panel on Climate Change (IPCC) hypothesis, the four different carbon concentration emission scenarios represent, in turn, the forced scenario of a gradual increase in carbon emission levels and a more considerable temperature increase: SSP1-2.6 (+2.6 W/m² imbalance; representing a low forcing sustainability pathway), SSP2-4.5 (+4.5 W/m²; medium forcing middle-of-the-road pathway), SSP3-7.0 (+7.0 W/m²; medium- to high-end forcing pathway), and SSP5-8.5 (+8.5 W/m²; high-end forcing pathway) [61,62,63].

2.2.5. Land Use Data and Harmonized World Soil Database (HWSD)

Land use data were derived from the Resource and Environment Science and Data Center (https://data.cma.cn/, accessed on 15 December 2022), with a spatial resolution of 30 m [64]. We selected two periods of land use data in 2000 and 2020. Harmonized World Soil Database (HWSD) was obtained from the Food and Agriculture Organization of the United Nations (https://www.fao.org/, accessed on 15 December 2022), which is a 30 arc-second raster database with over 15,000 different soil mapping units [65].

2.3. Methodology

Figure 2 depicts the framework adopted in this study to examine the hydrological evolution characteristics and causes of the runoff change. Firstly, we systematically analyzed the evolution laws of precipitation, temperature, and vegetation on the Loess Plateau from both temporal and spatial perspectives using statistical methods based on the observed meteorological and NDVI data. We conducted a comprehensive analysis of the trend, periodicity, change point, and their correlations using the related methods based on the observed data from 1969 to 2018. Secondly, we divided the annual runoff time series into baseline and variation periods based on the observed runoff evolution characteristics. Then, the widely used hydrological sensitivity methods based on the Choudhury–Yang formula and DMC were used to quantitatively distinguish the contribution of climate variability and human activities to the runoff change in the QRB. Finally, we simulated the future annual runoff time series in the QRB under four scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5) using the Soil and Water Assessment Tool (SWAT). Then, we considered the future runoff sequences as change periods, and considered the 1985–2018 as the baseline period. Subsequently, the causes of runoff change in the future were analyzed by the DMC method.

2.3.1. Trend and Change Point Analysis

The Mann–Kendall method [21,66,67], Sen’s slope estimation method [68], and linear regression [69] were adopted to analyze the variation trend of regional precipitation, temperature, and vegetation. Furthermore, the Pettitt test [70,71], sliding t-test [72], and the Mann–Kendall method were applied to identify the abrupt change point of the runoff series.

2.3.2. Runoff Variation and Attribution Analysis Methods

To identify the change point of annual runoff, we employed the Pettitt, sliding t-test, and the Mann–Kendall method. In general, the change point can be identified using any of these three methods. However, to determine the most accurate time for the change point, we chose to use all three methods and find the intersection of their results as the true change point. Additionally, based on the Penman–Monteith model [73], the annual potential evapotranspiration of the QRB was calculated from 1969 to 2018. To quantify the effects of climate and underlying surface change on watershed runoff, we used the Choudhury–Yang formula to calculate the elastic coefficient, with the DMC method used to verify the calculated results. Based on these calculations, we quantitatively distinguished the specific contribution rates of climate factors and underlying surface changes to the reduction in natural runoff in the basin during 1969–2018.

When the climate and vegetation cover of a basin are known, the hydrological and climatic characteristics of the long series are subject to the principle of water and energy balance [40]. Based on this theory, the basin’s hydrothermal coupling equilibrium equation appeared, later named by experts and scholars as the Choudhury–Yang formula [74]. The equation is expressed as follows:

$$E=\frac{P{E}_0}{{\left(P^n+E_0^n\right)}^{1/n}},$$

(1)

where P is the multi-year average precipitation; E₀ is the multi-year average potential evapotranspiration; E is the multi-year average actual evapotranspiration; n is a parameter reflecting the underlying surface characteristics of the basin, such as terrain, soil, and vegetation [38].

Climatic factors were classified as precipitation and potential evapotranspiration. The elasticity of runoff with respect to climate and underlying surface was defined as the degree of variation in watershed runoff resulting from changes in unit climatic factors and the underlying surface. That is, if the value of the elasticity coefficient is represented by m, then a 1% increment in the x factor (comprising precipitation, potential evapotranspiration and underlying surface parameter n) corresponds to an m% increase in runoff. According to the climate and underlying surface elastic coefficient, the runoff changes caused by precipitation, potential evapotranspiration, and underlying surface parameter n can be estimated, respectively. In Formula (1), P, E₀, and n are independent. Combined with the water balance relationship in the basin (precipitation equals the sum of evaporation and runoff), the change in annual runoff R can be expressed in the following total differential equation:

$$dR=\frac{\partial R}{\partial P}dP+\frac{\partial R}{\partial E_0}d{E}_0+\frac{\partial R}{\partial n}dn.$$

(2)

The precipitation elastic coefficient ${\epsilon }_P$, the potential evapotranspiration elastic coefficient ${\epsilon }_{E_0}$, and the underlying surface elastic coefficient ${\epsilon }_n$ are, respectively, expressed in Formulas (3)–(5):

$${\epsilon }_P=\frac{dR/R}{dP/P},$$

(3)

$${\epsilon }_{E_0}=\frac{dR/R}{d{E}_0/E_0},$$

(4)

$${\epsilon }_n=\frac{dR/R}{dn/n},$$

(5)

$$\mathrm{\varnothing }=\frac{E_0}{P}.$$

(6)

To calculate ${\epsilon }_P$, ${\epsilon }_{E_0}$, and ${\epsilon }_n$ by combining Equations (1)–(5) and substituting them into Equation (6), the following formula can be obtained:

$${\epsilon }_P=\frac{{\left(1+{\mathrm{\varnothing }}^n\right)}^{1/n+1}-{\mathrm{\varnothing }}^{n+1}}{\left(1+{\mathrm{\varnothing }}^n\right)\left[{\left(1+{\mathrm{\varnothing }}^n\right)}^{1/n}-\mathrm{\varnothing }\right]},$$

(7)

$${\epsilon }_{E_0}=\frac{1}{\left(1+{\mathrm{\varnothing }}^n\right)\left[1-{\left(1+{\mathrm{\varnothing }}^{-n}\right)}^{1/n}\right]},$$

(8)

$${\epsilon }_n=\frac{\ln \left(1+{\mathrm{\varnothing }}^n\right)+{\mathrm{\varnothing }}^n\ln \left(1+{\mathrm{\varnothing }}^n\right)}{n\left[\left(1+{\mathrm{\varnothing }}^n\right)-{\left(1+{\mathrm{\varnothing }}^n\right)}^{1/n+1}\right]}.$$

(9)

The study period was divided into two sub-periods according to the change point. The multi-year average runoff depth of Period 1 was denoted as R¹ (in mm), while the multi-year average runoff depth of Period 2 was denoted as R² (in mm). The difference can express the change from Period 1 to Period 2 in the average multi-year runoff depth of the two periods:

$$∆R=R^2-R^1.$$

(10)

The attribution of runoff change can be expressed as

$$∆R=∆R_c+∆R_l,$$

(11)

where $∆R_c$ is the runoff variation caused by climate change, mm; $∆R_l$ is the amount of runoff change caused by the underlying surface change, mm. $∆R_c$ is divided into runoff changes $∆R_P$ caused by precipitation and potential evapotranspiration changes $∆R_{E_0}$, mm, as shown in Equation (12).

$$∆R_c=∆R_P+∆R_{E_0}.$$

(12)

$∆R_P$, $∆R_{E_0}$, and $∆R_l$ are calculated by Equations (13)–(15):

$$∆R_P={\epsilon }_P\frac{R}{P}∆P,$$

(13)

$$∆R_{E_0}={\epsilon }_{E_0}\frac{R}{E_0}∆E_0,$$

(14)

$$∆R_l={\epsilon }_n\frac{R}{n}∆n.$$

(15)

The DMC method, initially used to check hydrological or meteorological data coherence, has become widely applied in assessing runoff response to climate variability and human activities due to its simplicity [69]. Briefly, X represent the reference variable and Y represent the test variable, with the observation period spanning N years. The observed values are denoted as X_i and Y_i, where i ranges from 1 to N. The X and Y variables are cumulatively computed to derive a novel year-by-year cumulative sequence [21]:

$$X_i^{\prime }=\sum _{i=1}^N{X}_i,$$

(16)

$$Y_i^{\prime }=\sum _{i=1}^N{Y}_i.$$

(17)

The relationship between the two variables is established by regression analysis in the Cartesian coordinate system. A change in the slope of the DMC indicates a breakpoint of the original gradient of the curve. Using this method and change-point analysis, runoff data were separated into the baseline period and the variation period. It was widely used to estimate the impact of human activities [21,75,76,77,78]. In this study, by substituting the rainfall accumulation data in the post-mutation period into the equation of the rainfall-runoff double accumulation curve prior to the mutation, we obtained a set of runoff data that differed from the original post-mutation runoff data. We attribute this difference to runoff changes resulting from human activities.

2.3.3. Simulation and Attribution Analysis Methods for Future Runoff under Different SSP Scenarios

We used Soil and Water Assessment Tool (SWAT) to simulate future runoff under different SSP scenarios, including SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5. The SWAT model was calibrated using monthly runoff data, with the calibration period from 2001 to 2009 and the validation period from 2010 to 2018. The calibration resulted in the Nash–Sutcliffe efficiency coefficients of 0.78 and 0.64 for the calibration and validation periods, respectively, demonstrating satisfactory model performance. Additionally, the Kling–Gupta efficiency coefficients achieved values of 0.74 and 0.63 for the calibration and validation periods, respectively, confirming the model’s accuracy and reliability. The 3-North Shelter Forest Program Overall Plan aims to increase forest coverage on the Loess Plateau to at least 25.6% and up to 39.63% by 2050. Given that the forest coverage rate of the QRB was 23.15% in 2020, it is possible to reach the upper threshold. When the land use data were input into SWAT, the set land area in the QRB accounted for 39.63% of the total land use area. We combined meteorological data under different SSP scenarios to obtain future runoff series values from 2019 to 2050. On this basis, we employed the DMC method to analyze the causes of runoff change under the four scenarios. Here, the method of finding abrupt change points is the same as in Section 2.3.2.

3. Results

3.1. Variation Law of Hydrological Regime and Underlying Surface

3.1.1. Temperature Trend

The Mann–Kendall method was used to analyze the trend and detect change points in the annual mean temperature series of the QRB. The temperature slope was found to be 0.0336, which is greater than 0 and indicates an upward trend. Furthermore, the value of |Z| was found to be 5.0426, greater than 1.96, a more than 95% confidence level. Therefore, the temperature rises significantly. Overall, the temperature in the QRB shows a significantly increasing trend from 1969 to 2018 at the significant level of 5%. Moreover, an abrupt change in temperature occurred in 1995 at the significant level of 5%. Figure 3 displays the distribution of annual mean temperature from 1969 to 2018 in the basin. The annual mean temperature in the basin ranges from 12.67 °C to 12.84 °C. Notably, the northeastern region of the basin exhibits relatively higher annual mean temperatures.

Figure 4 presents the spatial distribution results of linear trends of the lowest, average, and highest temperatures in the QRB. From 1969 to 2018, the temperature consistently increases throughout the region, ranging from 0.18 to 0.36 °C/10a, with the lowest temperature exhibiting the fastest increase. The mean temperature exhibits the lowest warming rate, growing by 0.18–0.24 °C, followed by the maximum temperature which has a growth rate of 0.21–0.24 °C. The minimum temperature experiences the most significant warming trend, with a growth rate of 0.24–0.36 °C.

3.1.2. Precipitation Trend

Figure 5 illustrates the distribution of the annual mean precipitation in the basin from 1969 to 2018. The annual mean precipitation in the basin ranges from 536.76 mm to 545.20 mm. It is worth noting that the northern region of the basin experiences relatively higher annual mean precipitation. Figure 6 suggests the results of the spatial distribution of linear precipitation trends on annual and seasonal scales in the QRB. The precipitation shows a decreasing trend on the whole, with a decreasing rate of −0.8 mm/10a to −0.6 mm/10a (Figure 6a). During spring, precipitation decreases in the whole region, especially in the southwest of the QRB (Figure 6b). The spatial distribution of the summer precipitation trend is the opposite of that of the spring. However, the overall trend is decreasing, and the rate of decrease is more significant in the northeast region (Figure 6c). Compared with the annual, spring, and summer values, fall precipitation demonstrates a lower rate of decrease (Figure 6d). The overall precipitation trend in winter is increasing (Figure 6e).

Table 2. Precipitation trends calculated by three methods (M-K, Sen’s, Linear regression).

Time Scale	Slope (M-K)	Z (M-K)	b (Sen’s)	b/mm/a (LR)	Significance	Trend
Annual	−0.72	−0.57	−0.72	−0.91	×	↓
Spring	−0.46	−1.17	−0.46	−0.37	×	↓
Summer	−0.32	−0.36	−0.34	−0.31	×	↓
Fall	−0.28	−0.50	−0.29	−0.34	×	↓
Winer	0.11	0.84	0.11	0.10	×	↑

“×” means not significant (significance level is 5%); “↑” represents an upward trend while “↓” indicates a downward trend.

presents the annual and seasonal precipitation trends as well as their significance ( significance level is 5%), calculated using the Mann–Kendall method (M-K), Sen’s slope estimation method (Sen’s), and linear regression method (LR). Spring, summer, and autumn precipitations show decreasing trends, whereas winter precipitation shows an increasing trend. However, none of the trends are significant. These findings are consistent with those presented in Figure 6.

3.1.3. Variation Law of Underlying Surface

Figure 7 illustrates the variation trend of NDVI in the QRB from 1982 to 2018. The variation range of NDVI was 0.34–0.75. Notably, NDVI exhibited a significant upward trend, with a rising rate of 0.128/10a, indicating that the vegetation in the QRB was continuously improving. The vegetation coverage in the QRB has been consistently increasing since the implementation of afforestation policy in 1999.

Figure 8 depicts the average distribution, slope, and slope significance of NDVI in the QRB. Vegetation growth in the north was better than that in the south (NDVI in the north was higher than that in the south), and the maximum value of NDVI reached 0.88, while the south distribution of NDVI in the center was between 0.45 and 0.65 (Figure 8a). Figure 8b displays the change in the annual NDVI of the basin. Based on statistics, the proportion of grids with a slope greater than 0 was 97.6%, indicating that NDVI has significantly improved overall, and the vegetation coverage rate of the basin has continuously increased. In addition, we identified the grids with the NDVI trend value that passed the 5% significance test (Figure 8c). NDVI in the entire region shows a highly significant rising trend, with only a tiny part of NDVI in the southwest region exhibiting a slightly significant declining trend. The overall vegetation in the QRB has been effectively improved, and the regional soil erosion control effect is significant.

3.2. Characteristics of Historical and Future Runoff Variation in the QRB

Figure 9 shows historical annual runoff series and the future annual runoff under four SSP scenarios. It can be seen that the highest runoff was observed in the SSP1-2.6 scenario, while the lowest runoff was observed in the SSP3-7.0 scenario. Compared with the runoff during 1985–2019, the future runoff increased under the SSP1-2.6 scenario and decreased under the other three scenarios. In Section 3.4, we look for the cause of the runoff change in the future four scenarios. Among the runoff data used here, the annual runoff data from 1985 to 2018 are the historical measured data, and the annual runoff data from 2019 to 2050 are the runoff data simulated by SWAT under four scenarios.

3.3. Relative Impact of Climate Change and Human Activity on Historical Runoff Change in the QRB

The Pettitt test, sliding t-test, and the Mann–Kendall method were employed to identify the abrupt change point of runoff data collected from 1969 to 2018 at Yaoxian Station. Given the significant level of 5%, the abrupt change point of annual runoff appeared in 1985. Therefore, 1969–1984 is designated as Period 1 (P₁) and 1985–2018 as Period 2 (P₂). The next step involved calculations using Formulas (1)–(15).

Table 3. Calculation results including the precipitation elastic coefficient ${\epsilon }_P$, the potential evapotranspiration elastic coefficient ${\epsilon }_{E_0}$, and the underlying surface elastic coefficient ${\epsilon }_n$ of different periods.

	n	P (mm)	E0 (mm)	R (mm)	${\mathit{\epsilon }}_{\mathit{P}}$	${\mathit{\epsilon }}_{{\mathit{E}}_0}$	${\mathit{\epsilon }}_{\mathit{n}}$
1969–2018	2.85	542.70	914.52	37.52	3.48	−2.48 *	−2.26 *
P1 1	2.71	575.58	947.92	46.84	3.32	−2.32 *	−2.12 *
P2 2	2.97	520.45	891.93	31.21	3.63	−2.63 *	−2.39 *
Δ 3	0.26	−55.12	−55.99	−15.63	/	/	/

¹ P₁ represents the base period of runoff change during historical period (1969–1984). ² P₂ represents the change period of runoff change during historical period (1985–2018). ³ “Δ” denotes the variation of n (underlying surface parameter), P (precipitation), E₀ (potential evapotranspiration) and R (runoff) between period P₁ (1969–1984) and period P₂ (1985–2018), indicating the magnitude of changes in these factors. * The symbol “−” represents a decrease in annual potential evapotranspiration and underlying surface parameter n of QRB leads to an increase in annual runoff.

shows that the ${\epsilon }_{E_0}$, ${\epsilon }_n,$ and ${\epsilon }_P$ of Period 1 are −2.32, −2.12, and 3.32, respectively, indicating that when the potential evapotranspiration, underlying surface parameter n and precipitation decrease by 1%, the annual runoff decreases by −2.32%, −2.12%, and 3.32%, respectively. The ${\epsilon }_{E_0}$, ${\epsilon }_n$, and ${\epsilon }_P$ of Period 2 are −2.63, −2.39, and 3.63, respectively, indicating that when the potential evapotranspiration E₀, underlying surface parameter n and precipitation P decrease by 1%, the annual runoff decreases by −2.63%, −2.39% and 3.63%, respectively. Notably, from Period 1 to Period 2, when the potential evapotranspiration, underlying surface parameter n, and precipitation change by 1%, the variation range of runoff increases.

According to Table 3 and

Table 4. Contribution rates of climate change and human activity to runoff change.

	Climate Change		Human Activity	Total
	ΔRP (mm)	$∆{\mathit{R}}_{{\mathit{E}}_0}$ (mm)	ΔRl (mm)	Total
Variation of runoff	−13.28 *	5.71	−7.79 *	−15.36 **
Contribution rate	86.46%	−37.16%	50.70%	100%
Rate	49.30%		50.70%	100%

* The symbol “−” means that changes in precipitation and underlying surface parameter n reduce runoff. ** The symbol “−” means that the runoff decreases from the base period P₁ (1969–1984) to the change period P₂ (1985–2018).

, it can be concluded that a decrease of 1% in annual precipitation, annual potential evapotranspiration, and underlying surface parameter n of the QRB leads to a reduction in annual runoff by 3.48%, −2.48%, and −2.26%, respectively. Furthermore, the contribution rate of precipitation reduction to runoff reduction is 86.46%. In contrast, the contribution rate of potential evapotranspiration reduction to the reduction of runoff is −37.16%. The contribution rate of the increase in underlying surface parameter n to the reduction of runoff is 50.70%.

To verify the accuracy of the Choudhury–Yang formula, we adopted the DMC method to distinguish the contribution of climate change and human activities to the reduction of runoff in the QRB. The precipitation and runoff depth were accumulated, respectively. The double accumulation curve of cumulative runoff and precipitation depth is shown in Figure 10. The linear relationship between the two variables is significant.

In order to determine the calculated runoff depth after an abrupt change, the cumulative rainfall during 1985–2018 was substituted into the regression equation of the base period. The influences and contribution rates of climate change and human activities on runoff in the QRB were subsequently obtained, respectively, as shown in

Table 5. Impacts of climate change and human activities on runoff changes in the QRB.

Time	Average Annual Precipitation (mm)	Runoff Depth and Its Variation (mm)				Climate Change		Human Activity
		Obs1 3	Obs2 4	Calculated Value 5	Total Change	Contribution (mm)	Rate	Contribution (mm)	Rate
P1 1	561.43	46	31.21	38.93	−15	−7.28	48.54%	−7.72	51.46%
P2 2	526.6

¹ P₁ represents the base period of runoff change during historical period (1969–1984). ² P₂ represents the change period of runoff change during historical period (1985–2018). ³ Obs₁ represents the mean value of observed runoff during P₁. ⁴ Obs₂ represents the mean value of observed runoff during P₂. ⁵ “Calculated value” represents the mean value of runoff during P₂ calculated by Formula (16).

. By differentiating the effects of climate change and human activities on runoff, it can be observed that the runoff depth decreased by 7.28 mm due to climate change, and the contribution rate was 48.54%. The decrease in runoff caused by human activities was 7.72 mm, and the contribution rate was 51.46%. The calculated results agree with those of the Choudhury–Yang formula, which calculated that the contribution rate of climate change and human activities was 49.30% and 50.70%, respectively. The results of the Choudhury–Yang formula are reliable and genuinely reflect the causes of runoff changes in the QRB.

3.4. Relative Impact of Climate Change and Human Activity on Future Runoff Change in the QRB

Figure 11c displays the future runoff simulated by SWAT under different SSP scenarios from 2019 to 2050. It is observed that the future runoff is higher in the SSP1-2.6 scenario and lower in the SSP3-7.0 scenario, while the other two scenarios are intermediate. To identify the abrupt change points of runoff under various scenarios from 1985 to 2050, we integrated the future runoff with the measured runoff, and the results are summarized in

Table 6. Impacts of climate change and human activities on future runoff changes in QRB under four SSP scenarios.

Scenario	Time	Mutation Point	Precipitation (mm)	Mean Temperature (°C)	Runoff Depth and Its Variation (mm)				Climate Change		Human Activity
					Obs1 3	Obs2 4	Calculated Value 5	Total Change	Contribution (mm)	Rate	Contribution (mm)	Rate
SSP1-2.6	P2-126 1	2026	535.70	12.66	28.87	40.02	36.90	11.15	8.03	72.00%	3.12	28.01%
	P3-126 2		641.66	12.48
SSP2-4.5	P2-245	2020	522.39	12.89	32.31	29.25	35.68	−3.06	3.38	−110.32%	−6.44	210.32%
	P3-245		570.69	12.42
SSP3-7.0	P2-370	2024	532.48	12.74	30.39	18.83	33.88	−11.56	3.49	−30.20%	−15.05	130.20%
	P3-370		551.78	12.17
SSP5-8.5	P2-585	2018	523.65	12.84	33.04	29.63	36.19	−3.41	3.15	−92.54%	−6.56	192.54%
	P3-585		582.81	13.13

¹ P_2-xxx (xxx = 126/245/370/585) represents the base period of runoff change in the future. The base periods for the four scenarios are 1985–2025, 1985–2019, 1985–2023 and 1985–2017, respectively. ² P_3-xxx represents the change period of runoff change in the future. The corresponding change periods for the four scenarios are 2026–2050, 2020–2050, 2024–2050 and 2018–2050, respectively. ³ Obs₁ represents the mean value of observed runoff during P_2-xxx. ⁴ Obs₂ represents the mean value of observed runoff during P_3-xxx. ⁵ “Calculated value” represents the mean value of runoff during P_3-xxx calculated by the formula of the rainfall-runoff double accumulation curve prior to the mutation.

. The abrupt points of runoff differ for different scenarios. Hence, we set 1985 to abrupt change year as the base period of runoff, abrupt change year to 2050 as the period of runoff change. Subsequently, we used the DMC method to estimate the contribution of climate change and human activities to runoff change. The contribution of climate change and human activities to runoff change under the four scenarios is presented in Table 6.

Table 6 indicates the variation trend of runoff under four scenarios in the future: runoff increases under SSP1-2.6, while it decreases under the other three scenarios. As shown in Table 6 and Figure 11a, the precipitation during the runoff change period under scenario SSP1-2.6 was the largest among the four scenarios. Precipitation in all four scenarios was higher than that in the base periods, but only the SSP1-2.6 scenario had an increase in runoff. The increasing runoff may be caused by the largest increase in precipitation under the SSP2.6 scenario. Figure 11b illustrates that the average temperature under SSP5-8.5 is higher than that under the other three scenarios.

Under SSP3-7.0, the largest variation in runoff is observed, even though the mean temperature is not significantly high. In the SSP5-8.5 scenario, the temperature during the runoff change period increases compared with the base period. In contrast, the temperature in the other three scenarios is lower than that in the base period. In the SSP1-2.6 scenario, the contribution rates of climate change and human activities to the future runoff increase are 72% and 28.01%, respectively. Under the SSP2-4.5 scenario, the contribution rates of climate change and human activities to future runoff reduction are −110.32% and 210.32%, respectively. Similarly, under the SSP3-7.0 and SSP5-8.5 scenarios, the contribution rates of climate change and human activities to future runoff reduction are −30.20% and 130.20%, respectively, and −92.54% and 192.54%, respectively. It can be found that the contribution rate of human activities to future runoff change is less than that of climate change in the SSP1-2.6 scenario. However, in the other three scenarios, the contribution rate of human activities to future runoff change is far greater than that of climate change. Therefore, although climate change in the QRB is likely to intensify in the future, human activities can still intervene and affect climate change on runoff.

4. Discussion

4.1. Challenges to Ensuring Water Security in the Watershed

The SSP2-4.5 scenario is projected to be the most likely scenario for human society to face in the future according to Hausfather’s study [79]. Compared to other scenarios, human activities are expected to contribute the most to the future reduction in the Qishui River runoff under SSP2-4.5. Based on our research results, both in the past and in the future, human activities are the main factors affecting the Qishui River runoff reduction (Table 5 and Table 6). For the QRB, afforestation is the most important human activity, which can help to stabilize soils and reduce erosion. This can lead to improved water quality, increased biodiversity, and increased carbon sequestration, which can have a positive impact on the environment and society. Additionally, increased vegetation reduces soil erosion, thus reducing the sediment content of runoff and further discouraging river bed silting.

However, an important reason for runoff attenuation is the increase in vegetation coverage, which can have some negative effects. The QRB is a region characterized by high soil and water loss, and a decrease in river runoff may result in a range of problems. The reduced river velocity caused by decreased flow can cause sediment from upstream to accumulate, leading to increased silting of the riverbed. During the dry season, the regulation function of the riverbank groundwater is gradually lost, which in turn affects the runoff of the river, making the runoff attenuated and even cutting off the flow in severe cases. Additionally, the reduction in river runoff directly affects river ecology in arid and water-deficient areas. During both the wet and dry seasons, the reduced flow in the river can make it difficult for some organisms to reach the hydraulic conditions required for growth and reproduction. Moreover, the decrease in runoff can lead to decreased water availability, which can affect agriculture, industry, and households that rely on water for various purposes. This can cause decreased food production, higher food prices, and increased competition for water resources [80]. Runoff reduction poses challenges to water security in the QRB.

Based on the above analysis, it is evident that a larger forest area does not necessarily always result in better outcomes for human society. If the afforestation policy is not properly planned and implemented, it can even lead to the loss of biodiversity, soil degradation, and reduced productivity.

4.2. Countermeasures to Ensure Water Resources Sustainability

China’s afforestation policy in the Loess Plateau has been effective in reducing soil erosion, and it will continue in the QRB in the future to prevent future problems such as sediment accumulation in the Qishui River caused by reduced runoff. However, when implementing afforestation policies, it is essential to carefully consider the potential benefits and risks and develop strategies to minimize the negative impacts while maximizing the positive ones. While implementing afforestation such as the 3-North Shelter Forest Program, China’s government has proposed many effective measures to strengthen water resources sustainability. For instance, tree species that are well adapted to the local conditions have been carefully selected, which can effectively stabilize the soil and prevent erosion [81]. Species such as Chinese pine, black locust, and elm have been found to be effective in reducing soil erosion and improving soil quality in the Loess Plateau [82]. China has also implemented ecological engineering measures in the Loess Plateau such as contour plowing and checking of dams to reduce soil erosion and improve water retention in the soil. These measures help to slow down runoff of the Qishui River and allow more water to infiltrate into the soil, thereby reducing the risk of sediment accumulation in river downstream [83]. Moreover, local government encouraged public participation in afforestation programs, including the involvement of local farmers and communities in tree planting and land management activities [84]. This approach helps to ensure that afforestation policies are aligned with local needs and priorities, and that they are sustainable in the long term.

In addition to measures that have already been implemented, three other promising options can be implemented. Firstly, an appropriate value of forest area needs to be determined. A common conclusion from previous studies is that deforestation increases annual runoff, while afforestation decreases it [85,86,87,88]. When the QRB has the right amount of forest, the degree of soil and water loss in the basin is small, and the river runoff is also suitable for social and economic development [89]. For the Chinese government, a way to balance the relationship between afforestation area and river flow has become an important issue for policy makers to consider. Additionally, it is necessary to monitor and evaluate the relationship between a reasonable threshold and existing vegetation coverage. If the coverage exceeds this threshold, it should be reduced; if there is a significant gap with this threshold, it should be raised. Hence, the Chinese government can establish monitoring and evaluation systems to assess the effectiveness of afforestation policies and identify areas for improvement. For example, remote sensing technologies are used to monitor the progress of reforestation and soil erosion reduction, and hydrological modeling is used to evaluate the impact of afforestation on water resources [90,91]. Moreover, runoff attenuation caused by human activities has significantly affected river ecological environment [92]. Therefore, in the development and utilization of lacquer river water resources, local government should ensure that enough ecological water is reserved in the river to maintain the good cycle of the Qishui River ecosystem [92].

4.3. Uncertainties and Limitations

Our study revealed that the precipitation decreased and the temperature increased in the QRB during 1969–2018, and the climate of the basin became warmer and drier in the past 50 years, which is consistent with the study of [80,93,94]. The change in runoff in the historical period is mainly caused by human activities, which is consistent with the study of [95].

Although our study is consistent with previous research, there are still some uncertainties and limitations that need to be acknowledged. Firstly, while previous studies have validated the good performance of the Choudhury–Yang formula [47,96], there are still uncertainties and limitations in its accuracy. One possible reason for this is the simplification of physical processes. When using the Choudhury–Yang formula, like many other scholars, we believe that P, E₀ and n are three independent variables, but in fact they are not [97]. The change in precipitation causes the change in evapotranspiration, and also has a certain impact on regional vegetation growth. Therefore, there are some errors in the process of calculating the contribution rate of each factor. A second potential explanation is that the underlying surface parameter n in this formula only considers simple factors such as terrain, soil, and vegetation, which cannot fully reflect all the impacts of human activities [49]. In addition, large spatial resolution of meteorological data, land use data and soil data also causes certain errors in the process of SWAT model construction. Thirdly, in the process of correcting CMIP6 data, we selected an average of seven models. Averaging multiple models produced results that fit the actual values better than data from individual models. In the future, we could consider using a larger number of climate models to improve the accuracy of data correction even further.

5. Conclusions

Loess Plateau has the most severe soil and water loss worldwide. The hydrological evolution driven by climate change and human activities is a crucial scientific problem in this area. Based on the analysis of the meteorological elements and the underlying surface in the QRB, a small watershed in the Loess Plateau, we study the causes of the historical and future runoff variation under the changing environment. Our results reveal that human activities contribute more to runoff change than climate change, both in the historical period (1969–2018) and in the future (2019–2050). Specifically in the historical period, the contribution rates of climate change and human activities to runoff change were 51.46% and 48.54%, respectively, which were almost equal, with a difference of only about 1%. In the future, with the development of social economy, this situation will change, since the influence of human activities is far greater than that of climate change. In the future period, the contribution rate of climate change and human activities is projected to be −110.32% and 210.32%, respectively. Furthermore, our findings suggest that the decreasing trend in runoff is mainly attributed to the implementation of the afforestation policy in the QRB.

The implementation of afforestation policies is a critical step towards mitigating soil and water loss in the QRB. However, it can also result in a reduction in surface runoff, which can have adverse effects. Therefore, attention should be paid to the sustainable use of water resources in the future. It is essential to optimize the local industrial structure to prevent over-consumption of energy. Based on the findings of this research, we propose three measures to ensure the sustainability of water resources in the QRB: (1) Balance the relationship between afforestation area and river flow, and find a suitable forest area for sustainable development of the QRB; (2) establish monitoring and evaluation systems to assess the effectiveness of afforestation policies and identify areas for improvement; (3) ensure that sufficient ecological water is reserved in the river to maintain the good cycle of the Qishui River ecosystem. In general, ensuring afforestation policies and water security simultaneously is proving to be a challenging task in the Loess Plateau region due to the increase in forested area coupled with a decrease in available water. Going forward, it is crucial to conduct comprehensive studies that explore all aspects of the relationship between water and afforestation. This will enable the development of integrated and explicit plans for the sustainable use of resources while ensuring water security in the QRB.