Interaction of Climate Change and Anthropogenic Activity on the Spatiotemporal Changes of Surface Water Area in Horqin Sandy Land, China

Abstract

Surface water dynamics are sensitive to climate change and anthropogenic activity, and they exert important feedback to the above two processes. However, it is unclear how climate and human activity affect surface water variation, especially in semi-arid regions, such as Horqin Sandy Land (HQSL), a typical part of the fragile region for intensive interaction of climate and land use change in northern China. We investigated the changes of spatiotemporal distribution and the influence of climatic and anthropogenic factors on Surface Water Area (SWA) in HQSL. There are 5933 Landsat images used in this research, which were processed on the Google Earth Engine cloud platform to extract water bodies by vegetation index and water index method. The results revealed that the area and number of water bodies showed a significant decrease in HQSL from 1985 to 2020. Spatially, the SWA experienced different amplitudes of variation in the Animal Husbandry Dominated Region (AHDR) and in the Agriculture Dominated Region (ADR) during two periods; many water bodies even dried up and disappeared in HQSL. Hierarchical partitioning analysis showed that the SWA of both regions was primarily influenced by climatic factors during the pre-change period (1985–2000; the mutation occurred in 2000), and human activity has become more and more significantly important during the post-change period (2001–2020). Thus, it is predictable that SWA variation in the following decades will be influenced by the interaction of climate change and human activity, even more by the later in HQSL, and the social sectors have to improve their ability to adapt to climate change by modifying land use strategy and techniques toward the sustainable development of water resources.

1. Introduction

Water resources are an important foundation of human society [1,2], and provide critical ecosystem services, production development, and ecosystem functions maintenance, and promotes sustainable development [3,4,5]. However, global fresh water is only 2.5% of the total world water [6,7], and most cannot be used directly. Therefore, with climate change and rising water demand, surface water, including glaciers, lakes, rivers, and swamps, are becoming more crucial for human life and terrestrial ecosystems [8]. China is one of several countries with severe fresh water shortage [9], and the situation severely restricts regional economic development [10]. With intensified climate change and anthropogenic activity in recent decades, water resources are becoming more and more scarce, particularly in arid and semi-arid regions with fragile eco-environments.

Growing studies have used remote sensing methods to monitor surface water dynamics [11,12,13,14], which has many advantages, such as high resolution, long-term availability, and large-scale coverage [15,16,17]. Thus, various water extraction methods have been proposed based on satellite-based methods, such as the single band threshold method, the multiband spectral relation method, the water index method, and the threshold method. Among these, the water body index can quickly obtain the distribution range of water bodies through simple band calculation and threshold processing, such as Normalized Difference Water Index (NDWI) [18], Modified Normalized Difference Water Index (MNDWI) [19], and the Automated Water Extraction Index (AWEI) [20]. In this study, we use a combination of the water body index, MNDWI, Normalized Difference Vegetation Index (NDVI), and Enhanced Vegetation Index (EVI), which can eliminate land noise and decrease wetlands mixed with water bodies and vegetation to improve the accuracy of extracting water bodies [21]. In recent years, the Google Earth Engine (GEE) cloud platform has been widely applied to process massive geospatial data quickly and efficiently [22,23,24]. Based on the GEE cloud platform, numerous studies have concentrated on surface water dynamics in some hot spots or regions of China, such as the Huai River Basin [25], Yangtze River Basin [26], and Poyang Lake [27], where water resources are relatively abundant. However, the surface water bodies in fragile regions of northern China have experienced remarkable changes under the influence of climate change and anthropogenic activity, which also needs more attention. Several studies have explored the dynamics of SWA in some sensitive or water-shortage regions of China, such as the North China Plain [28,29] and the Inner Mongolian Plateau [30,31,32,33], but most of them only focused on lakes or reservoirs at individual scales rather than all types of surface water bodies for the whole region.

Horqin Sandy Land, the largest sandy land in China, is part of the fragile agro-pastoral ecotone, where cultivation and grazing are the dominant land use modes. As the population grows, the interaction of natural and human factors has caused significant water shortage due to consistent overuse, which has ultimately undermined the pivotal supporting capacity of water resource to the ecosystems in HQSL. Thus, it is necessary to quantify the spatiotemporal dynamics of surface water in order to understand their driving mechanism over a long-term period and provide in-depth theoretical and data supports to the locals, decision-makers, and researchers for feasible restoration and management strategies in HQSL, as well as provide a case study for similar regions around the world.

Therefore, the objectives of this study are underlined as follows: (1) Trace the annual dynamic process of SWA in HQSL from 1985 to 2020; (2) Achieve a more precise classification of surface water by the frequency of water inundation in time-series; (3) Quantify the contributions of climate and anthropogenic activity to SWA variation separately.

2. Materials and Methods

2.1. Study Area

Horqin Sandy Land is geographically located (41°41′N–46°02′N, 117°49′E–123°42′E) in northern China, covering almost 125,100 km² (Figure 1). The study area has a high elevation in the northwest and southwest, and a low elevation in the east and central area. It has a typical semi-arid continental monsoon climate, the mean annual temperature is 3–7 °C, and mean annual precipitation is 350–500 mm; approximately 70% of the precipitation is concentrated in summer. However, the annual potential evapotranspiration is 1500–2500 mm, approximately five-fold the annual precipitation [34]. The XiLiao River, Hongshan Reservoir, and Mengjiaduan Reservoir are some of the larger surface water bodies in HQSL. There are 14 counties with a combination of farming and grazing as a typical land use pattern in the study region [35,36], which is divided into Agriculturally Dominated Regions (ADR) and Animal Husbandry Dominated Regions (AHDR) for a better interpretation of SWA variation in HQSL. The classification was made according to results by Chen [37], who adopted the methods of Canonic Correspondence Analysis and Systematic Cluster Analysis to sort counties into different groups.

2.2. Landsat Data

This study used all available Landsat images at 30 m spatial resolution, covering HQSL on the GEE platform (https://earthengine.google.com, accessed on 20 November 2021) from 1985 to 2020. A total of 5933 images were used, including 2810 scene Landsat 5 Thematic Mapper (TM) images (1985–2011), 2290 scene Landsat 7 Enhanced Thematic Mapper (ETM+) images (2000–2020), and 833 scene Landsat 8 Operational Land (OLI) Images (2013–2020). The total observation counts in HQSL from 1985 to 2020 are presented in Figure 2.

2.3. Climate and Anthropogenic Data

Climate and anthropogenic activity data were collected and used to estimate the driving factors of surface water variation. Annual Temperature (ATM) and Annual Precipitation (AP) were provided by China National Meteorological Information Center (http://data.cma.cn/site, accessed on 1 January 2022). Evaporation (ET) was taken from the National Earth System Science Data Center at 1 km resolution of monthly potential evapotranspiration dataset in China (1990–2020) (http://loess.geodata.cn, accessed on 1 January 2022). Anthropogenic activities data were collected from the regional statistical yearbooks of Tongliao and Chifeng, including population, agriculture gross domestic production (GDP), livestock number, and effective irrigated area. For more information regarding the datasets used in this study, see

Table 1. Driving factors used in the study.

Data	Unit	Year	Source
Annual Temperature	°C	1985–2020	China Meteorological Data Service Center
Annual Precipitation	mm	1985–2020
Evaporation	0.1 mm	1985–2020	National Earth System Science Data Center
Population	person	1985–2020	Regional statistical yearbooks
Livestock number	Head	1985–2020
Effective irrigated area	Hm2	1985–2020
Agriculture GDP	Chinese Yuan (CNY)	1985–2020

.

2.4. Water Extent Extraction

We can identify surface water bodies by the relationships between water and vegetation indices; the method has been extensively applied to extract water bodies [38,39,40] because of the advantage of simple and accurate extraction [19,26]. The water and vegetation indices contain NDVI, EVI, and MNDWI, and each index can be calculated based on the following spectral bands and equations:

$$MNDWI=\frac{{\rho }_{Green}-{\rho }_{SWIR1}}{{\rho }_{Green}+{\rho }_{SWIR1}}$$

(1)

$$NDVI=\frac{{\rho }_{NIR}-{\rho }_{Red}}{{\rho }_{NIR}+{\rho }_{Red}}$$

(2)

$$EVI=2.5\times \frac{{\rho }_{NIR}-{\rho }_{Red}}{1.0+{\rho }_{NIR}+6.0{\rho }_{Red}+7.5{\rho }_{Blue}}$$

(3)

where ${\rho }_{Blue}$ is the surface reflectance values of the blue band (0.45–0.52 μm), ${\rho }_{Green}$ is the green band (0.52–0.60 μm), ${\rho }_{Red}$ is the red band (0.63–0.69 μm), ${\rho }_{NIR}$ is the near-infrared band (NIR) (0.77–0.90 μm), and ${\rho }_{SWIR1}$ is the short-wave infrared-1 band (1.55–1.75 μm), respectively, in the Landsat sensors.

We employed a criterion proposed by Zou et al. [41]. The criterion MNDWI > EVI or MNDWI > NDVI can identify the pixels where the water signal is stronger than the vegetation signal; additionally, EVI < 0.1 can remove vegetation pixels or mixed pixels of water. Thus, only pixels that meet the criteria (MNDWI > EVI or MNDWI > NDVI) and (EVI < 0.1) can be categorized as water pixels, while others are categorized as non-water pixels [30].

Water Inundation Frequency (WIF), percentage of per-pixel, is calculated by Equation (4):

$$WIF\left(t\right)=\frac{1}{N_t}{{\displaystyle \sum }}_{i=1}^{N_t}{W}_{t,i}\times 100$$

(4)

where WIF denotes water body inundation frequency for each pixel, t denotes year, N_t is the total number of Landsat pixels in the year, and W_t,i indicates whether or not the pixel is a water pixel, with one for water and zero for non-water.

In summary, the research process is displayed in Figure 3, showcasing a comprehensive flowchart of four parts. Firstly, water bodies were extracted-by water index method. Secondly, extracted water bodies were classified as seasonal and permanent based on inundation frequency (Equation (4)) and the specific threshold ranges discussed in Section 3.3.1. Thirdly, the space–time distribution of area and number of surface water bodies were analyzed. Finally, a quantitative analysis was conducted to determine the impact of driving factors on surface water in HQSL.

2.5. Changepoint Detection by BEAST

Rbeast is a Bayesian model that is derived for time series analysis and detecting any significant breakpoints within the series [42]; it can be employed to any time-series data when its precondition are met. The model presumes that an input time series can be separated into four distinct parts: a seasonal part modeled with a harmonic function, a background part modeled with a piecewise linear regression function, a number of possible change points for both parts, and a degree of random noise [43].

2.6. Quantitative Analysis of Driving Factors and SWA

Correlation analysis was carried out to discover the relationships between each independent variable, as well as the associations between the SWA and all of the explanatory variables. Using the rdacca.hp package in R, the contribution of driving factors to SWA was determined through the hierarchical partitioning method, which can compute the explanatory ratio of each set of variables, and can also solve the multicollinearity problem of explanatory variables [44]. Thus, correlation analysis and hierarchical partitioning methods were used to explore the influence of climatic and anthropogenic factors on the SWA in HQSL.

3. Results

3.1. Accuracy Assessment of Water Extraction

Water extraction accuracy in terms of the omission and commission errors of Landsat images was evaluated, and Sentinel-2 MSI images with 10 m resolution were used as the reference data. Results were summarized into a confusion matrix; the kappa coefficient (Kc) and Overall Accuracy (OA) were 0.93 and 96.78%, respectively (

Table 2. Accuracy of the extracted surface water in this study.

Sentinel-2 MSI (10 m)
Water Body Map (2020)	Water	No-Water	Total	User Accuracy (%)
water	836	25	861	97.10%
No-water	47	938	985	95.22%
Total	883	963	1846	Overall Accuracy = 96.78%
Producer Accuracy (%)	94.68%	97.40%		Kappa Coefficient = 0.93

), indicating that the accuracy of water extent extraction was credible.

3.2. Change Point of SWA

We used Bayesian models from the R package Rbeast to detect the change point of SWA during the period 1985–2020. The results are shown in Figure 4a; SWA showed an obvious decline trend from 1985 to 2020, and the probability distribution indicated that the change point year most likely occurred in 2000, with 70.8% probability. Hence, the SWA change period was split into the pre-change period (1985–2000) and post-change period (2001–2020), according to change point. Subsequently, boxplots were made for the maximum, seasonal, and permanent water area in ADR, AHDR, and HQSL for two periods in Figure 4b. The results indicated that all of the three types of SWA had distinct phase variations in the two periods. Overall, SWA was significantly larger in the previous period than in the later period.

3.3. Spatiotemporal Patterns of SWA in HQSL

We calculated SWA dynamics from 1985 to 2020 in HQSL, and separately analyzed the ADR and AHDR of SWA variations in HQSL for a better understanding of how the driving factors affect SWA variations.

3.3.1. Long Term Water Body Inundation Frequency

Waterbody occurrence, also known as inundation frequency, is typically represented by the dynamics of surface water, which shows the proportion of reliable observations when water is found at the surface [13,45]. Generally, water pixels with an inundation frequency (WIF) ≥ 75% are defined as permanent water, and those in the range of 25% ≤ WIF < 75% are classified as seasonal water; the sum of seasonal water and permanent water constitutes maximum water. Permanent water usually occurs at the deepest parts of water bodies and retains water throughout the year. While seasonal water, for instance, small ponds, creeks, and other shallow water [46], often distributes at the shallow edges of the permanent water, and only exists in the rainy seasons and dries up in the drought period. According to the cumulative WIF map of 1985–2020, the maximum surface water area is 2364.14 km², accounting for 1.89% of the total area of HQSL (Figure 5). We then mapped 0.25° latitude and longitude summaries of surface water area for HQSL (Figure 5b,c). As can be seen, surface water is mainly distributed in 43°N–44°N, and the areas of seasonal water are larger than those of permanent water. While surface water distributed more unevenly in longitude, the peaks indicate that there are reservoirs distribution.

According to the water existence period, thresholds of the annual WIF can provide detailed information regarding surface water variation. Thus, the water pixels were divided, with 25%, 50%, and 75% inundation frequency as the threshold. Surface water was distributed unevenly in HQSL. Approximately 30.6% of water pixels distributed in ADR and 69.39% in AHDR; among them, seasonal water and permanent water accounted for 73.7% and 24.2%, respectively, in HQSL (

Table 3. Spatial distribution of water in HQSL during 1985–2020.

Zone	Surface Water of HQSL (km2)
	Seasonal (km2)		Permanent (km2)	Total
	25 ≤ WIF < 50	50 ≤ WIF < 75	75 ≤ WIF < 100
ADR	41.31	92.26	162.06	295.63
AHDR	333.78	143.95	192.50	670.23
Total	495.84	236.21	233.81	965.86

). Seasonal water covered 54% and permanent water covered 46% in ADR, and 71.3% and 28.7% in AHDR, respectively, which strongly indicated that there was a greater reduction of water availability during dry periods due to the higher proportion of seasonal water.

3.3.2. Annual Variation of SWA

Further, in order to specifically and precisely elucidate the changing process of SWA in HQSL, the entire study period was divided into two parts according to the change point. Overall, the SWA showed a markedly decreasing trend, with a rate of 15.11 km²/yr in ADR, 31.3 km²/yr in AHDR, and 46.43 km²/yr in HQSL during the period 1985–2020 (Figure 6). The SWA variation trend in AHDR was better matched to that of HQSL than ADR (Figure 6b,c). The largest SWA of the three regions all occurred in the pre-change period and had a decreasing trend at a rate of 6.38 km²/yr in ADR, 19.29 km²/yr in AHDR, and 25.58 km²/yr in HQSL, respectively, and the decreasing rate in ADR was significantly less than that in AHDR. In the post-change period, SWA demonstrated a steadily decreased trend at a rate of 7.88 km²/yr in ADR, 4.9 km²/yr in AHDR, and 10 km²/yr in HQSL, respectively, and decreased less in AHDR than in ADR. On the whole, the SWA reached a peak at 2364.14 km² in 1995, and then abruptly declined and hit the lowest value at 394 km² in 2009.

Spatially, most of the water bodies are distributed in the eastern part of HQSL, with several lakes and reservoirs along the rivers (Figure 1). Because rivers are getting too narrow for their changes to be clearly identified, we mainly focused on the water body changes in several lakes and reservoirs, including semi-artificial reservoirs (Hongshan reservoir, Mengjiaduan reservoir, and Muruin Sum reservoir), as well as a natural lake (Naiman west lake), to demonstrate the SWA interannual variations in HQSL (Figure 7). The largest reservoir in HQSL is The Hongshan reservoir, inaugurated in 1958; its water area peaked in 1986 and has been gradually decreasing since then. Naiman West Lake disappeared after the year 2000 due to drought and increasing water use. Similarly, the water area of Mengjiaduan reservoir also reached a high level in 1986, and its two water bodies shrunk significantly after 2000; the east reservoir completely disappeared until 2020, while the west reservoir was restored by water diversion to improve the local eco-environment. However, even after restoration, the area of Mengjiaduan reservoir was only roughly half of its maximum capacity. The Muruin Sum reservoir also experienced shrinkage, dried up after 2008, and recovered because of water diversion in 2013; since then, it has completely dried up. The changes in the lake and reservoirs reflect the variation of SWA very well in HQSL. Furthermore, we can clearly identify from the Landsat-series images of each group that cultivated land has appeared and increased since 2000, which was closely related to the shrinkage of SWA.

3.4. Variation of Surface Water Number

Given that the largest number of water bodies was 361 in 1986, we selected 1986 as the start year, 2000 as the change point year with the number 187, and 2020 as the last year with number 114 to explore the change of surface water number from 1986 to 2020 (Figure 8). Figure 8a clearly shows that the number of different sizes of water bodies continues to decrease from 1986 to 2020. In particular, the number of small water bodies (SWA ≤ 1 km²) varied greatly and was highly consistent with the interannual variation of the total number, but the number of medium water bodies (1 km² < SWA < 10 km²) increased significantly due to the fragmentation or shrinkage of the larger ones; the number of large water bodies (SWA ≥ 10 km²) also showed reducing trend (Figure 8b). Compared with Figure 6c, the total number of water bodies showed a similar trend with the area; they had a high correlation coefficient (r = 0.96) (Figure 8c). According to

Table 4. Detailed information on changes in the area and number of SWA in HQSL during the periods 1986–2000 and 2000–2020.

	1986		2000		2020		Changes (1986–2000)				Changes (2000–2020)
Area Class	Num.	Area	Num.	Area	Num.	Area	ΔNum.	ΔArea	ΔNum.	ΔArea	ΔNum.	ΔArea	ΔNum.	ΔArea
		(km2)		(km2)		(km2)		(km2)	(%)	(%)		(km2)	(%)	(%)
ADR	94	372.44	59	220.10	8	32.33	35	152.34	37%	41%	51	187.77	86%	85%
<1 km2	48	31.83	27	18.81	4	3.13	21	13.02	44%	41%	23	15.67	85%	83%
1–10 km2	37	87.22	24	56.43	3	11.61	13	30.79	35%	35%	21	44.79	88%	79%
>10 km2	9	253.39	8	144.92	1	17.59	1	108.47	11%	43%	7	127.31	88%	88%
AHDR	267	696.18	128	288.65	106	205.21	139	407.53	52%	59%	22	83.39	17%	29%
<1 km2	120	82.03	59	41.17	50	34.16	61	40.86	51%	50%	9	6.94	15%	17%
1–10 km2	137	328.27	66	206.98	54	128.9	71	121.29	52%	37%	12	78	18%	38%
>10 km2	10	285.88	3	40.64	2	42.15	7	245.24	70%	86%	1	1.55	33%	4%
HQSL	361	1068.62	187	508.07	114	237.54	174	560.55	48%	52%	73	270.46	39%	53%
<1 km2	168	113.86	86	60.09	54	37.29	82	53.77	49%	47%	32	22.71	37%	38%
1–10 km2	174	415.49	90	263.01	57	140.51	84	152.48	48%	37%	33	122.49	37%	47%
>10 km2	19	539.27	11	185.23	3	59.74	8	354.04	42%	66%	8	125.26	73%	68%

, both regions experienced a comparable level variance in the number and area of surface water. AHDR experienced a slighter decline than ADR between 1986 and 2000, and from 2000 to 2020, the number and area of surface water decreased by 86% and 85% in ADR, while it decreased by 17% and 29% in AHDR, respectively. The result was in line with the conclusion from Figure 6; overall, the number and area of surface water in HQSL decreased continuously, and the larger ones fluctuated more obviously in 2000–2020 than in 1985–2000. The maximum number of water bodies was in 1986, the second largest was in 1995, the tertiary peak was in 1998, and it then decreased to the minimum in 2009.

3.5. Attribution Analyses of SWA in HQSL

Climatic factors may alter the hydrologic cycle by the variation of meteorology elements such as precipitation, evaporation, and temperature [47], and can change the amount of water recharge and regimen of the river [48,49]. The anthropogenic activity also changes the dynamics of the water cycle by human interference, such as dam construction and extraction of water for production and domestic use [50]. Therefore, in this study we selected the main consuming factors of surface water in HQSL, including population, agriculture GDP, effective irrigated area, and livestock number (Figure 9). Spearman correlation analysis indicated that SWA was significantly correlated with ET, Pop, and Ag-GDP during the pre-change period in ADR, but highly correlated with Pop, EIA, and Ag-GDP in the post-change period (Figure 9a,c). The hierarchical partitioning method further verified the independent impacts of driving factors on SWA (Figure 9b,d), and the result showed that climatic factors and human activity can explain 24.6% and 3.5%, respectively, in the pre-change period, while it can explain 12.4% and 66.58%, respectively, in the post-change period. However, the Spearman correlation result indicated that SWA was significantly related to AP and ET in the pre-change period, and correlated with Pop, EIA, Ag-GDP, and livestock number during the post-change period in AHDR (Figure 9e,g). Hierarchical partitioning analysis (Figure 9f,h) showed that climatic factors and human activity can explain 24.7% and 10.3%, respectively, in the pre-change period, and explain 13.17% and 46.27%, respectively, in the post-change period.

Through hierarchical partitioning analysis, it can be seen that, during the pre-change period in both regions, the most influential factor on SWA was AP, followed by ET and Ag-GDP. In the post-change period, SWA was significantly affected by Pop and EIA in ADR, while it was mostly affected by EIA and livestock number in AHDR. The SWA of both regions was greatly affected by climate in the pre-change period, but significantly influenced by human activities in the post-change period. As consumers play a growing role in SWA change in ADR and AHDR, either Pop or livestock number, and EIA consistently influence the change of SWA, particularly in the post-change period in both regions.

4. Discussion

4.1. Variations of SWA

According to the previous studies, more than 60% of the SWA has shrunk in Horqin [33,51], and this research finding is consistent with the fact that SWA showed a significantly decreased trend in HQSL during the period 1985–2020. The three peaks of SWA occurred in 1986, with 2218.68 km², 1995 with 2364.14 km², and 1998 with 2072.02 km², accordingly; precipitation reached 585.15 mm, 539.30 mm, and 677.96 mm in the three years, respectively, above the average of precipitation. It was reported that heavy rainfall that occurred in the region was caused by the strongest El Niño event on record [52,53].

During the period 2001–2010, the SWA shrunk remarkably because of drought and human water consumption [35]. La Niña-influenced annual precipitation decreases during the periods [53], and the semiarid region experienced actual evapotranspiration increases as a result of climate warming [54]. Meanwhile, the water demand for agricultural irrigation increases when the temperature rises, causing water withdrawal and a decrease [55]. In fact, the shortage limitation of SWA was influenced by the interaction between climate change and human water consumption. With the growing population and continuous drought, more and more land was cultivated to increase harvest in the drought years [35,56]; therefore, more water resource were required to meet the demand for production and livelihood. Many studies have verified that the West Liao River surface flow began to decline in 1980 and had nearly dried up by 2000 [53,57]. After 1999, most lakes and reservoirs also decreased significantly [58], and even approached their dead storage [59,60].

Since 2011, the variation of SWA has tended to be moderate in HQSL, owing to fluctuating increasing precipitation [25], emphasis on water resource protection by the local governments, recharging some key reservoirs through water diversion, and improving irrigation technology [61]. The SWA in this region would have been shrinking and would even have disappeared if there was no such protection. The local governments are making great efforts to restore the surface water resources and protect the eco-environment. Even though the reservoir’s desiccation situation has been mitigated temporally in recent years, the eco-environment is still very fragile in HQSL. As shown in Figure 9, the SWA was mainly influenced by anthropogenic activity in the post-change period, and its impact will be getting heavier, and could even be a “threat” to the development and ecological reservation at the local scale. Meanwhile, the higher occurrence of extreme climate events projected might make the situation worse in the future [30]. Therefore, it also requires more consideration of climate change scenarios, especially extreme climate events, and the negative interaction of climate change and overutilization of water resource by human activity.

4.2. Difference of SWA in Relation to Climate and Human Factors between ADR and AHDR

The study comparatively analyzed SWA change due to the different water consumptions of production and livelihoods between ADR and AHDR and found that there were significant differences in trends and influencing factors between the two regions. Compared with ADHR, ADR is mainly distributed in flat areas with more rivers and reservoirs (Figure 1b), where water is well-supplied; therefore, the permanent water bodies had a larger area than the seasonal water bodies, and the variation of SWA was more moderate. While the decreasing trend of SWA was more significant in AHDR, due to more influence from seasonal water change, especially in the years of abnormal precipitation change during the pre-change period, the two regions were all mainly influenced by the climate in this period.

However, with the growing population, irrigation demand, and livestock number, human activity had an increasing and significant impact on SWA in the post-change period in both regions (Figure 9). The change of SWA decreased at a rate of 7.8 km²/yr in ADR, more than that in AHDR, which had a rate of 4.9 km²/yr, due to rapidly increasing demand for irrigation in ADR. The proportion of cultivated land area in Horqin Sandy Land increased four-fold from 8.9% in 2000 to 40.3% in 2019 [62]. According to the 2020 TongLiao Water Resources Bulletin and the ChiFeng Water Resources Bulletin [63,64], the water resources of HQSL were mainly used for the first industry, including farmland irrigation, afforestation, and animal husbandry, accounting for 78.2%, 4.9%, and 3.5%, respectively, of the total water consumption. However, one thing that cannot be ignored here is the fact that, as surface water shrinks and water quality deteriorates, local inhabitants are forced to withdraw more groundwater for irrigation and feed livestock in HQSL, accelerating the reduction trends of the groundwater table [65], which also directly lead to SWA reduction.

It is clear that the effect of groundwater, a vital component of the hydrological cycle, is connected with surface water through leakage, evaporation, and lateral recharge [66], as well as through the impact of municipal and domestic water use on water storage in HQSL. The transformation of surface water and groundwater and groundwater funnel effects will be taken into consideration in upcoming research. Improving water use techniques and efficiency are essential for both ecological and economic well-being in the agro-pastoral ecotone in the future.

5. Conclusions

This study employed a water extraction approach by the combination of vegetation index and water index to quantify the spatial and temporal variation of SWA in HQSL during the period 1985–2020. The spatiotemporal change of area and number of surface water bodies showed a similar decreasing trend in HQSL. The SWA in this area has been shrinking, and even disappeared in recent years under the interaction effect of climate and human activity, though it has been restored by human intervention. SWA rebound is a test of positive human interference with water resources and also a warning that the limitation of human interference could not change the general trend of SWA reduction in HQSL. Area and number of surface water showed a slighter decline in AHDR than that in ADR between 1986 and 2000, but rapidly decreased by 86% and 85% in ADR, and 17% and 29% in AHDR, respectively, from 2000 to 2020. Climate change mainly affected SWA in the pre-change period, while human activity had a greater effect during the post-change period, and it will play a more and more important role in water resource variation. Our results provide a quantitative analysis of long-term data for an in-depth understanding of water resource dynamics in HQSL, and the urgent need to use water efficiently with a strict planning strategy of long-run water management practices, including HQSL and the agro-pastoral ecotone in China and similar ones around the world.