Impact of Human Activities on Hydrological Drought Evolution in the Xilin River Basin

Abstract

The impact of human activities on the hydrological cycle makes hydrological drought no longer a natural disaster in a strict sense, and influences the stationarity of the hydrologic process. In this context, assessment methods that consider nonstationary conditions are more reasonable in the study of hydrological drought. In this study, we used the SWAT (Soil and Water Assessment Tool) model to reconstruct the historical hydrological conditions during the period affected by human activities (1998–2019) of the Xilin River Basin. After calculating the standardized runoff index (SRI) at multiple time scales, we compared the drought characteristics of the basin under natural conditions and under the influence of human activities. The results show that human activities were the main reason for the significant decrease of runoff in the basin (an obvious change-point for runoff series is identified in 1998), which accounted for 68%. Compared with natural conditions, human activities delayed the occurrence of short-term drought in the basin and changed its seasonal distribution characteristics, resulting in an increase in the frequency of severe and extreme droughts in autumn; the corresponding drought frequency increased by 15% and 60%, respectively. Moreover, human activities have also prolonged drought duration, increased drought intensity, and increased the uncertainty of drought in the basin. The proposed method is demonstrated to be efficient in quantifying the effects of human activities on hydrological drought, and the findings of this study provide a scientific basis for water resource management, drought early warning, and forecasting under a changing environment.

1. Introduction

Climate change and human activities are two major factors affecting the hydrological cycle and driving the evolution of water resources. They are also important driving forces for the evolution of regional drought [1,2,3]. With the changing environment in recent decades, drought has shown a trend of widespread and frequent occurrence. The resulting water shortage, food crisis, environmental deterioration, and other problems, have seriously threatened human survival and sustainable economic development [4,5,6]. Currently, drought is classified into four categories: meteorological drought, socioeconomic drought, hydrological and agricultural drought [7]. Hydrological drought is affected by precipitation, infiltration, and runoff, involving the transformation of surface water, soil water and groundwater, and is related to the hydrological cycle and water balance [8]. The processes of its formation and completion are relatively slow, and the mechanism is more complicated than the other three categories of drought [9,10]. With the profound impact of high-intensity human activities on the hydrological cycle, hydrological drought is no longer a natural disaster in a strict sense, but rather a natural-anthropogenic composite disaster [11,12]. To accurately predict the occurrence and development of the hydrological drought and effectively manage it, the formation and development of drought cannot be considered solely from the natural perspective. The role of human activities in the formation and development of hydrological drought needs to be fully considered [13,14,15].

Human activities alter hydraulic connections in the water cycle by changing the underlying surface conditions of the basin, resulting in changes in surface runoff, consumption, discharge conditions, river regulation, and storage functions. These can further affect the occurrence and development of the hydrological drought [16,17,18,19,20]. Scientific evaluation and identification of the hydrological drought events and determination of the main driving factors in their evolution have become the focus of international drought research [21,22,23,24,25]. Based on the threshold method and standardized index method, Rangecroft et al. [26] evaluated the characteristics of drought before and after the completion of the San Juan Dam in northern Chile. They reported that reservoir regulation reduced the frequency, duration, and intensity of drought events but had no significant effect on multi-year drought. By applying the VIC (Variable Infiltration Capacity) model to analyze the attribution of runoff attenuation in the Weihe River Basin, Ren et al. [27] suggested that both climate change and human activities were important factors affecting the evolution of drought and flood in the basin. In particular, human activities were the dominant factor in the occurrence of short-term drought. Su et al. [28] used the SWAT model to reconstruct the historical hydrology of the Daling River Basin under natural conditions. Based on a multi-time scale standardized runoff index, that study compared the evolution of the basin’s drought under changing environments and concluded that human activities have intensified the duration and intensity of seasonal drought and increased the uncertainty of drought events in the basin. By discussing the occurrence and evolution of hydrological drought in the basin from the perspective of water cycle, we can further improve the accuracy and objectivity of drought evaluation.

The Xilin River Basin, a typical grassland inland river basin, was selected as the study area. In recent years, the rapid development of regional animal husbandry has doubled the total number of livestock in the study area, and the amount of livestock has far exceeded the carrying capacity of the natural grassland. Meanwhile, with the depletion of energy resources in central and eastern China, China’s coal power development strategy has gradually moved westward. The Xilingol League region (where the study area is located) is rich in energy resources, and is gradually becoming a main source of coal and a national energy strategic base supporting the long-term socioeconomic development of China. Intensive human activities have accelerated soil erosion and aggravated drought in the basin during the past two decades [29,30,31]. In this context, the SWAT model was used to reconstruct the natural runoff of the basin, and on this basis the temporal characteristics of drought and the influence of environmental change on drought evolution were examined. This study should provide a scientific basis for water resources management, drought early warning, and forecasting under a changing environment.

2. Materials and Methods

2.1. Study Area

The Xilin River Basin located in the Inner Mongolia Plateau (43°24’–44°39’ N, 115°25’–117°15’ E), surrounded by the Greater Khingan Mountains and the Yinshan Mountains covers a drainage area of 10,786 km². It is difficult for warm and humid air flow to enter; thus the climate is characterized by strong evaporation and insufficient precipitation [32,33], the average annual precipitation is about 300 mm, and the evaporation is 1900 mm (measured by an evaporating dish with a diameter of 20 cm). The total length of the main river is 175 km, and the altitude of the basin is 900–1650 m. The terrain slopes gradually from southeast to northwest (Figure 1). The landform has obvious stratification and zoning. The south is a three-stage basalt platform with many small volcanic cones scattered across the landscape. The other areas are occupied by low mountains, hills, and grasslands. More than 90% of the vegetation in the basin is natural pasture, which transitions from upstream meadow characterized by Leymus chinensis and Stipa baicalensis to downstream steppe characterized by Stipa grandis and Stipa krylovii. It is a typical representative temperate grassland basin. Under the dual influences of climate change and human activities, the grassland vegetation has been severely degraded [34]. As a result, loss of the hydrological function of the basin is ongoing and it is extremely sensitive to meteorological disasters (such as drought) [35].

2.2. Data

We collected observation data from one hydrological station, 15 rainfall stations, and one national meteorological station in the upper Xilin River Basin over the past 45 years (1975–2019). The hydrological data included the monthly observed runoff depth at the Xilin River Hydrological Station collected from Inner Mongolia Monitoring Center of Hydrology and Water Resources. The meteorological data included daily precipitation, temperature, wind speed, and relative humidity, collected from Inner Mongolia Weather Bureau.

To construct the SWAT model of the basin we also needed land surface parameters such as a high-resolution digital elevation model (DEM), as well as soil and land use data. The DEM data, which had a spatial resolution of 30 m × 30 m, was obtained from the Geospatial Data Cloud Platform (http://www.gscloud.cn (accessed on 12 June 2021)). Soil data was obtained from the 1:1,000,000 soil data provided by the Food and Agriculture Organization of the United Nations (FAO) (https://www.fao.org/soils-portal/soil-survey/soil-maps-and-databases/harmonized-world-soil-database-v12/en (accessed on 15 June 2021)). The land use data were obtained from the National Geomatics Center of China’s global 30 m land cover data (http://www.globallandcover.com (accessed on 15 June 2021)).

2.3. Methods

2.3.1. Reconstruction of Natural Runoff Based on SWAT Model

The SWAT model [36], developed by the Agricultural Research Center of the United States Department of Agriculture (USDA), was selected to reconstruct the water cycle in the basin under natural conditions. The SWAT model is a semi-distributed hydrological model with physical mechanisms that can simulate different hydrological processes in the basin by using the spatial data information provided by GIS and multiple spatial data sets [37]. Briefly, the SWAT model divides the basin into several sub-basins, then divides them into one or more hydrological response units (HRUs) according to the soil type, land use and slope of the sub-basins. It uses a conceptual model on each unit to calculate the net rainfall, then performs runoff calculation, and finally obtains the outlet section flow. The threshold values of land use, soil type and slope selected in this study were 5, 5, and 10% respectively. Based on this, the study area was divided into seven sub-basins and 214 HRUs. The model has been successfully applied to simulate the water cycle over the past few decades and generate a series of historically consistent long-term water and heat flux data [38,39,40].

We used the daily rainfall, daily maximum and minimum temperatures, wind speed, and relative humidity from 16 rainfall stations as meteorological input for the SWAT model, and the monthly runoff of 23 years (1975–1997) was used for the model construction and parameter calibration; the period 1975–1991 was used as the model calibration period, and the period 1992–1997 was used as the model validation period. The Nash-Sutcliffe efficiency coefficient (NSE), determination coefficient (R²), and relative bias coefficient (PBIAS) were used to evaluate the fitting effect of the model [41,42]. The specific information of the three evaluation indicators is shown in

Table 1. Expression of NSE, R² and PBIAS.

	Expression	Range	Satisfactory
NSE	$NSE=1-\frac{{\displaystyle \sum _{i=1}^N{(Q_{obs}-Q_{sim})}^2}}{{\displaystyle \sum _{i=1}^N{(Q_{obs}-{\overline{Q}}_{obs})}^2}}$	−∞–1	0.5 ≤ NSE ≤ 0.65
R2	$R^2=\frac{{[{\displaystyle \sum _{i=1}^N(Q_{obs}-{\overline{Q}}_{obs})}(Q_{sim}-{\overline{Q}}_{sim})]}^2}{{\displaystyle \sum _{i=1}^N{(Q_{obs}-{\overline{Q}}_{obs})}^2{\displaystyle \sum _{i=1}^N{(Q_{sim}-{\overline{Q}}_{sim})}^2}}}$	0–1	R2 ≥ 0.5
PBIAS	$PBIAS=\frac{{\displaystyle \sum _{i=1}^N(Q_{sim}-Q_{obs})}}{{\displaystyle \sum _{i=1}^N(Q_{obs})}}\times 100\%$	−∞–+∞	−25% ≤ PBIAS ≤ 25%

.

It was necessary to identify the impact period of human activities before reconstructing the natural runoff series. The Mann-Kendall trend test [43,44], Pettitt test [45], and precipitation-runoff double cumulative curve method [46] were used to analyze variability in the hydrometeorological elements in the study area. The precipitation-runoff double cumulative curve method is represented by a straight line, and the change in the gradient of the curve may infer that the characteristics of precipitation or runoff have changed [25]. This was utilized for the auxiliary detection of the changepoint of the runoff series in this study. According to the change point identification results, the study period was divided into two periods, the natural period (before the change point) and the impact period (after the change point). The observed hydrometeorological data for the natural period were used to calibrate the SWAT model, and model parameters obtained reflect the runoff producing conditions under the background of natural conditions of the basin. Then, the meteorological forcing data for the impact period were used as input data for the calibrated model to reconstruct natural runoff during the same period. We considered the simulated runoff during the impact period to be the natural runoff, which is only affected by climate factors.

2.3.2. Attribution Analysis of Runoff Change

Based on the SWAT model, we estimated the impact of climate change and human activities on runoff in the Xilin River Basin with the following assumptions: (1) climate change and human activities are the main factors causing changes in runoff, and (2) the two are independent of each other (meaning we ignored any climate feedbacks and the interaction between human society and the climate system (see the evaluation framework in Figure 2). On this basis, we combined the natural runoff series reconstructed by the model and the measured runoff series to conduct the attribution analysis of runoff changes. This provided a foundation for quantitatively assessing the impact of human activities on drought in the basin. The total amount of runoff change in the basin can be expressed as:

$$\mathrm{\Delta }{Q}_{tot}=\mathrm{\Delta }{Q}_{cv}+\mathrm{\Delta }{Q}_{ha}={\overline{Q}}_{2,obs}-{\overline{Q}}_{1,obs}$$

(1)

where $\mathrm{\Delta }{Q}_{cv}$ and $\mathrm{\Delta }{Q}_{ha}$ can be expressed as:

$$\mathrm{\Delta }{Q}_{cv}={\overline{Q}}_{2,sim}-{\overline{Q}}_{1,sim}$$

(2)

$$\mathrm{\Delta }{Q}_{ha}=\mathrm{\Delta }{Q}_{tot}-\mathrm{\Delta }{Q}_{cv}=({\overline{Q}}_{2,obs}-{\overline{Q}}_{1,obs})-({\overline{Q}}_{2,sim}-{\overline{Q}}_{1,sim})$$

(3)

Therefore, the impact of climate change and human activities on runoff can be calculated as:

$$I_{cv}=\frac{\mathrm{\Delta }{Q}_{cv}}{\left|\mathrm{\Delta }{Q}_{tot}\right|}\times 100\%$$

(4)

$$I_{ha}=\frac{\mathrm{\Delta }{Q}_{ha}}{\left|\mathrm{\Delta }{Q}_{tot}\right|}\times 100\%$$

(5)

where $\mathrm{\Delta }{Q}_{tot}$ and $\left|\mathrm{\Delta }{Q}_{tot}\right|$ are the total runoff change and its absolute value, respectively, $\mathrm{\Delta }{Q}_{cv}$ and $\mathrm{\Delta }{Q}_{ha}$ are the runoff changes caused by climate change and human activities, respectively, ${\overline{Q}}_{1,obs}$ and ${\overline{Q}}_{2,obs}$ are the measured annual average runoffs in the base period and change period, respectively, ${\overline{Q}}_{1,sim}$ and ${\overline{Q}}_{2,sim}$ are the simulated annual average runoffs in the base period and change period, respectively, and $I_{cv}$ and I_ha are the contribution rates of climate change and human activities to runoff change, respectively.

2.3.3. Identification of the Hydrological Drought Process Based on the Run Theory

(1) Drought index

Currently, the standardized runoff index (SRI) is widely used for hydrological drought identification [47]. The calculation is similar to the standardized precipitation index (SPI). After determining the suitable probability distribution of runoff in a specific time, the normal standardized runoff index can be calculated. The specific method is shown in the following formula [25]:

$$SRI=\left\{\begin{array}{l}-(t-\frac{C_0+C_{1}t+C_2{t}^2}{1+d_{1}t+d_2{t}^2+d_3{t}^3}), t=\sqrt{\ln (\frac{1}{{[F(x)]}^2})},0<F(x)\le 0.5\\ t-\frac{C_0+C_{1}t+C_2{t}^2}{1+d_{1}t+d_2{t}^2+d_3{t}^3}, t=\sqrt{\ln (\frac{1}{{[1-F(x)]}^2})},0.5<F(x)<1\end{array}\right.$$

(6)

where F(x) is a cumulative distribution function, and the coefficients are c₀ = 2.515517, c₁ = 0.802853, c₂ = 0.010328, d₁ = 0.432788, d₂ = 0.189269, d₃ = 0.001308.

This study focused on the change in drought characteristics for a period when the basin was affected by human activities. Thus, we used reconstructed runoff and observed runoff in the periods to calculate the corresponding SRI series, respectively. According to the division of SRI drought levels [25] in

Table 2. Drought grades classified with standardized runoff index.

Grade	Category	SRI
1	No drought	−0.5 < SRI
2	Mild drought	−1.0 < SRI ≤ −0.5
3	Moderate drought	−1.5 < SRI ≤ −1.0
4	Severe drought	−2.0 < SRI ≤ −1.5
5	Extreme drought	SRI ≤ −2.0

, we can compare and analyze the hydrological drought characteristics from the perspective of drought severity.

(2) Identification of the hydrological drought events

This study applied the run theory to define drought events and their characteristic variables (e.g., duration and intensity) [48,49]. The identification of drought events involved three steps. The first step was preliminary recognition. Three thresholds, R₀, R₁, and R₂ were set. When SRI < R₂ in a certain month, the month was preliminarily identified as a drought month (Figure 3b,d). The second step was the removal of non-drought periods. If the initially identified drought process lasted for only a month and SRI > R₂ (Figure 3a), the month was treated as a non-drought month and eliminated. If the drought lasted for more than one month, or the drought lasted for a month but its SRI < R₂, it was regarded as a drought event (Figure 3c–e). The third step involved combined drought. If two adjacent drought processes were separated by a month, and the interval month SRI < R₀ (Figure 3c), then the adjacent drought processes were merged into one drought event. The drought duration was the duration from the beginning to the end of the event. The drought intensity was the sum of the drought intensity of each month. Otherwise, it was considered an independent drought event (Figure 3d,e).

3. Results

3.1. Analysis of the Characteristics of Hydrometeorological Elements in the Basin

The evolution of the runoff and precipitation time series in the Xilin River Basin from 1975 to 2019 were analyzed using the M-K trend test and Pettitt mutation test (

Table 3. Results of Mann-Kendall trend test and Pettitt test for the annual precipitation and runoff series over the Xilin River Basin in the period of 1975–2019.

Element	Mann-Kendall Test		Pettitt Test
	Z	Trend	p-Value	Step Change Year
Runoff	−4.14	↓↓	4.33 × 10−6	1998
Precipitation	−0.17	↓	0.67	—

Note: ‘↓’ Indicates a downward trend; ‘↓↓’ Indicates a significant downward trend which significant at the 0.05 level; ‘—’ Indicates no significant abrupt point.

). According to Table 3, the evolution of the runoff and precipitation series in the study area experienced a downward trend in which the runoff change passed the significance test of α = 0.05, showing a significant downward trend. In addition, the runoff showed a significant deviation in 1998, while the precipitation time did not change abruptly.

The double accumulation curve of annual precipitation and annual runoff in the Xilin River Basin (Figure 4) further demonstrates that the annual precipitation in the study area has remained relatively constant while the runoff changed from 1975–1997 to 1998–2019. The runoff decreased significantly, resulting in a significant drop in the runoff coefficient. The above results further indicate that under the premise of no significant change in meteorological elements in the Xilin River Basin, the runoff changed drastically in 1998.

The rapid development of the animal husbandry and coal industry in the study area has promoted local socioeconomic development since the 21st century. According to statistics, the secondary and tertiary industries in the study area have increased by more than ten times over the past two decades [50]. Combined with the historical characteristics of human activities in the Xilin River Basin, 1975–1997 was selected as the natural period when hydrological processes were less disturbed by environmental changes. The period 1998–2019 was selected as the human activities impact period, during which human activities had a significant impact on hydrological processes.

3.2. Attribution Analysis of Runoff Attenuation

The measured monthly runoff data from 1975 to 1997 were used to calibrate the main parameters of SWAT model (

Table 4. Model parameters fitted value during calibration period.

Parameter	Definition	Fitted Value
CN2	SCS runoff curve number for moisture condition II	0.13
ESCO	Soil evaporation compensation factor	0.60
SOL_AWC	Available water capacity of soil	0.25
SOL_K	Saturated permeability coefficient of soil	0.10
SOL_BD	Volume weight of soil	0.10
ALPHA_BNK	Baseflow alpha factor for bank storage	0.13
CH_K2	Effective hydraulic conductivity of main channel	99.80

) automatically (by using the SWAT-CUP) during the natural period, in which 1975–1991 was taken as the model calibration period, and 1992–1997 was taken as the model validation period. By comparing the measured and simulated runoffs (Figure 5), it can be seen that the changes in simulated and measured runoffs were basically consistent during the natural period of the basin. Combined with the model simulation accuracy evaluation parameters (

Table 5. Performance of flow simulation for the Xilin River Basin using the SWAT model.

Period	p-Factor	R-Factor	R2	NSE	PBIAS	RMSE
Calibration (1975–1991)	0.62	0.36	0.77	0.75	15.79	0.23
Validation (1992–1997)	0.51	0.42	0.68	0.73	19.72	0.25

), it can be seen that NSE, R² and PBIAS in the model calibration period were 0.75, 0.77 and 15.79, respectively, and NSE, R² and PBIAS in the model verification period were 0.73, 0.68 and 19.72, respectively. Therefore, the calibrated SWAT model can effectively describe the water cycle process in the basin and can be further applied to reconstruct the natural runoff in the basin.

The runoff series of the impact period was reconstructed using meteorological data for the impact period and the model parameters of the natural period. By this method the runoff series for 1998–2019 (the impact period) under natural conditions was obtained (Figure 5). The results show similar changes in the measured and simulated runoffs during the impact period. However, the measured runoff was significantly less than the simulated runoff. The attribution analysis was combined with the corresponding evaluation framework of runoff under a changing environment, and the contribution rates of climate change and human activities to the runoff reduction in the Xilin River runoff were calculated (

Table 6. Effects of climate change and human activities on runoff reduction over the Xilin River Basin in the period of 1998–2019.

Basin	${\Delta }{\mathit{Q}}_{\mathit{t}\mathit{o}\mathit{t}}/\mathbf{mm}$	${\Delta }{\mathit{Q}}_{\mathit{c}\mathit{v}}/\mathbf{mm}$	${\Delta }{\mathit{Q}}_{\mathit{h}\mathit{a}}/\mathbf{mm}$	${\mathit{I}}_{\mathit{c}\mathit{v}}/\%$	${\mathit{I}}_{\mathit{h}\mathit{a}}/\%$
Xilin River Basin	−1.01	−0.32	−0.69	32	68

). The results suggest that measured runoff in the impact period was 1.01 mm lower than the simulated runoff, and that human activities were the main reason for the runoff reduction in the Xilin River Basin during the impact period. The runoff reduction was 0.69 mm. The contribution rate to the total runoff reduction was 68%.

3.3. Analysis of Hydrological Drought Evolution over Different Time Scales

By calculating the SRI over different time scales, the evolution of the hydrological drought over different time scales in the basin can be described. By comparing the SRI values calculated from measured and model reconstructed values, Figure 6 shows some impact of human activities on the drought over different time scales (1, 3, and 12 months).

The time series comparison results of the SRI₁ and SRI₃ (Figure 6a,b) shows that in the same period, the occurrence time of the hydrological drought based on the observed values lagged behind, which, based on the simulation values, indicates that the effects of human activities delayed the onset of short-term drought to some extent. In addition, the short time scale drought levels in the two cases were also quite different. The overall variation trends of the long-term drought (SRI₁₂) based on the measured and reconstructed values were similar. The drought and flood conditions reflected by the SRI in the same period were not significantly different from each other, but the drought levels were different (Figure 6c). Thus, human activities had a significant impact on short-term droughts, changing both the onset and magnitude of short-term droughts. For long-term droughts, human activities changed only the drought levels.

3.4. Analysis of Drought Frequency at Different Time Scales

To further analyze the impact of human activities on drought frequency in the basin, we calculated drought frequencies of different levels and months based on hydrological drought indices at different time scales. According to the frequency distribution of drought at different levels based on SRI₁ (Figure 7a₁), the hydrological drought distribution in the Xilin River Basin was dominated by mild drought under natural conditions. Under these conditions the drought frequency reached about 60%. Under the influence of human activities, the frequency of moderate drought was less than 40%. Human activities have increased the frequency of moderate, severe, and extreme droughts on a monthly scale in the basin. The effect was strongest for extreme drought, whose frequency was more than twice that of extreme drought under natural conditions. In the case of SRI₃, mild and severe droughts dominated under natural conditions, with drought frequencies of 52% and 11%, respectively (Figure 7b₁). Human activities have also caused the frequencies of moderate and extreme droughts to increase in the basin. In terms of long-term drought, mild drought still dominated under natural conditions (Figure 7c₁). However, its frequency was only 3% lower than the frequency under the impact of human activities. Human activities had little effect on the frequencies of moderate, severe, and extreme droughts in the basin.

From the monthly distribution of drought frequency, we can observe the significance of human activities on altering the seasonal distribution of the short-term hydrological drought in the basin under natural conditions (Figure 7a₂,b₂). Summer and autumn were the seasons with a high incidence of the hydrological drought in the Xilin River Basin, particularly the period from summer to autumn (August-September). Once the flood season had passed, the precipitation in the basin decreased rapidly, and the runoff decreased sharply, allowing the autumn drought to occur easily. Under natural conditions, the autumn drought in the basin was dominated by mild and moderate droughts. However, human activities have led to the occurrence of severe and extreme droughts in autumn, and the corresponding drought frequency increased by 15% and 60%, respectively (Figure 7b₂). According to the monthly distribution of drought frequency corresponding to SRI₁₂ (Figure 7c₂), human activities caused extreme drought in August and November. On an annual scale, human activities had little effect on the level and frequency of the long-term hydrological drought.

3.5. Analysis of Drought Events Characteristics

To further examine the impact of human activities on the characteristics of the hydrological drought in the basin, we used the run theory to extract drought events and their characteristic variables based on the drought index SRI at different time scales, and analyzed the differences in drought duration and drought intensity under the influence of natural conditions and human activities in the Xilin River Basin (Figure 8). Figure 8a shows that for the short-term droughts SRI₁ and SRI₃, their average durations under natural conditions were 1.8 and 3.3 months, respectively. Under the impact of human activities, their average durations were 2.2 and 4.3 months, respectively. Human activities also caused the longest short-term drought to extend for another 3–4 months. The duration of long-term drought in the basin under natural conditions was around 6–12 months. Under the impact of human activities, the duration of this type of drought was approximately 7–14 months, and the average duration was 1.2 months longer than that under natural conditions. It can be seen from Figure 8b that the average values of the short-term drought intensity under natural conditions were 1.2 and 2.6. The average values of drought intensity based on SRI₁ and SRI₃ increased by 1.0 and 1.4, respectively, due to human activities. Likewise, human activities increased the drought intensity of long-term drought by 1.0. In general, human activities have enhanced drought duration and intensity at different time scales, i.e., prolonging drought duration and increasing drought intensity. From the statistical distribution of drought characteristics, human activities have increased the uncertainty of hydrological drought characteristics and expanded the distribution range of drought duration and drought intensity.

4. Discussion

Previous studies on grassland inland river basins in semi-arid regions mostly focused on the impacts of environmental change on precipitation or runoff [51,52], while there have been few studies on the impact of human activities on hydrological drought. In this study, we found the precipitation and runoff in the study area has trended downward over the past 45 years, which is consistent with the findings of many previous studies [29,31,53]. Some studies have shown that in recent decades, a high latitude circulation anomaly has increased the air pressure and weakened the east Asian monsoon, which in turn has reduced the northward transport of water vapor [54,55]. This is one of the main reasons for the decrease in precipitation in the northern China.

Since the 21st century, the development of animal husbandry and the coal industry has driven the rapid development of the economy in the study area. With the growth of GDP, the underlying surface conditions (e.g., land use and land cover) of the basin have also changed. Although climate change and human activities have combined to cause a decrease of runoff in the study area, human activities have played a dominant role; this conclusion is similar to that of Gao et al. [56].

Human activities have a significant impact on short-term hydrological drought, especially in summer and fall. During this period, the study area is in the peak production and peak water use period, and intensive human activities lead to an extension of drought duration and increase of drought intensity [27,28]. The impact of human activities on long-term hydrological drought is not obvious. This phenomenon is determined by the occurrence mechanism of hydrological drought. As we know, meteorological drought is caused by a reduction of precipitation, and the resulting hydrological cycle anomaly may lead to the occurrence of hydrological drought at longer time scales [57]. Since climate change is a long-term process, the propagation process from meteorological drought to hydrological drought is mostly affected by climate change at longer time scales [58].

Quantifying the impact of human activities on hydrological drought was carried out on the basis of the hydrological model simulation in this study, and accurate simulation of the hydrological process directly affected the results of hydrological drought assessment. Due to the limitations of the observation stations, the uncertainty of the model is relatively high; therefore, intensive observations are necessary to obtain more data for calibrating the model. In addition, we can compare the simulation results of multiple hydrological models and use optimization method to reduce uncertainties. It should be pointed out that this study conducted only a generalized evaluation of the impact of various potential human activities (e.g., water withdrawal and changes in underlying surface conditions) on drought. There are still some deficiencies in the fine evaluation of the driving effect of various activities on drought. In the future, the contribution of different human activities to drought should be more accurately and precisely distinguished based on water use and water conservation project monitoring data from human activities.

5. Conclusions

The SWAT model was applied to the Xilin River Basin to reconstruct the historical hydrological series under natural conditions. The multi-time scale standardized runoff index (SRI) was used to compare and evaluate the characteristics of the hydrological drought in the basin under natural conditions and the influence of human activities. This study revealed the impact of human activities on the hydrological drought at different time scales. The main conclusions are as follows:

(1) The runoff and precipitation series in the Xilin River Basin showed a downward trend, and the change in the runoff demonstrated a significant downward trend. The mutation test and the double accumulation curve of precipitation and runoff indicated that runoff in the Xilin River Basin showed a significant change in 1998.

(2) The SWAT model was used to reconstruct the natural runoff of the basin during a period of human influence. We found that human activities caused a significant reduction in runoff, contributing 68% of the reduction in basin runoff.

(3) Using SRI as the drought index, we evaluated the characteristics of the hydrological drought over multiple time scales. The results suggest that human activities had an obvious effect on the short-term drought and certain delaying effect on the onset of drought during the same period. In the case of long-term drought, human activities changed the drought level only slightly.

(4) Analysis of the frequency and characteristic variables of the hydrological drought showed that human activities have increased the frequencies of severe and extreme droughts in the basin over a short time scale. At the same time, human activities have significantly modified the seasonal distribution characteristics of short-term drought under natural conditions, resulting in severe and extreme drought events in autumn. In addition, human activities have increased drought duration and drought intensity at different time scales (i.e., prolonged drought duration and increased drought intensity). Human activities have also increased the uncertainty of the hydrological drought characteristics and expanded the distribution range of drought duration and drought intensity.