Spatiotemporal Evolution of Mountainous Ecosystem Services in an Arid Region and Its Influencing Factors: A Case Study of the Tianshan Mountains in Xinjiang

Abstract

Mountainous ecosystems provide humans with multiple ecosystem services (ESs), but global changes and anthropogenic activities threaten the supply of such services in arid regions. To maintain regional ecological security and achieve ecosystem sustainability, it is quite essential to understand the spatiotemporal characteristics of mountainous ESs in arid regions and clarify the main driving factors of different ESs. Using the integrated valuation of ecosystem services and tradeoffs (InVEST) and revised universal soil-loss equation (RUSLE) models, we evaluated the ESs provided by the Tianshan Mountains of Xinjiang from 2000 to 2020. The research showed that: (i) over the 20 years in question, habitat quality and carbon storage remained relatively stable, while soil retention and water yield fluctuated significantly. (ii) All ES pairs exhibited synergies. Spatial synergy areas were concentrated in the northwestern and southwestern areas; spatial trade-off areas alternated with spatial synergy areas. (iii) Hotspots with at least two ESs covered 73% of the study region. Middle- and high-altitude areas were the main supply areas of ES. (iv) Land-use types were the dominant driving factor of habitat quality and carbon storage, while mean annual precipitation had the strongest explanatory power for water yield. Soil retention was mainly affected by mean annual temperature and the normalized difference vegetation index. Our findings could provide guidance for policymakers when developing measures for ecosystem conservation and management.

1. Introduction

Ecosystem services (ESs) refer to the diverse benefits that humans obtain directly or indirectly from ecosystems, including tangible material products and intangible services [1,2]. Such services are classified into supply, regulating, cultural, and support services required for maintaining other types of services that bridge the gap between human society and natural ecosystems and are crucial to socioeconomic sustainability and human well-being [3]. The status quo, trends, and spatial heterogeneity of ESs need to be accurately evaluated, in order to develop policies and management practices dealing with socioecological systems [4,5].

Mountainous areas cover 25% of the world’s total area, and although about 12% of the world’s population lives in mountainous areas, more than 50% of the world’s population directly or indirectly relies on mountain-derived resources [6]. By 2050, more than 24% of the world’s population in low-lying areas is expected to rely heavily on runoff contributions from mountainous areas [7]. These areas are also biodiversity hotspots [8], maintain global carbon and water cycles [9], and provide diverse ESs (e.g., water supply, soil conservation, climate regulation, and natural recreation) [10,11,12]. There is an emerging consensus that mountainous areas provide important and valuable ESs [13,14]. These services provide a livelihood to local residents and residents in adjacent low-lying areas [15] and form an important ecological security barrier for the region and even the country [16]. However, global changes may considerably alter the global supply of ESs, resulting in their decline, degradation, and even loss [3]. Owing to their heterogeneity and vulnerability, mountainous areas are more susceptible to changes in climate [17]. Moreover, human expansion into mountainous areas leads to the occupation of increasing amounts of ecological land (e.g., forest, grassland, wetland, and shrubland) [18], posing an increasingly severe threat to the supply of ESs [14]. Mountainous ecosystems in arid regions are more susceptible to climate changes and anthropogenic activities [19], presenting a great challenge for ecological conservation and sustainable management.

While studies have been conducted to assess the ecological conditions [20,21,22] and water resource changes [23,24] in arid regions in the context of climate change, there is little understanding of how climate change will affect ES provisioning in arid mountains, where ESs are a significant contributor toward the livelihood of local people and regional ecological security. In recent years, evaluating ESs in arid regions has received increasing attention, and a substantive body of research has mainly focused on natural reserves [25] and watersheds [26] with favorable ecological environments or densely populated urban agglomerations [27]. However, few efforts have been made to characterize the spatial heterogeneity of mountainous ESs in arid regions. Arid mountainous areas are complex ecosystems composed of diverse elements (e.g., mountains, waters, forests, lakes, farmlands, grasslands, sand, and ice) that exhibit evident vertical and longitudinal spatial heterogeneity [10]. An in-depth exploration of the temporal variation in ESs supply patterns in a biophysically heterogeneous environment would provide theoretical guidance for decision-makers to develop sustainability programs and conduct the differentiated management of mountainous ecosystems [11]. Thus, in comparison with research efforts found in the literature, our work has the following differences: firstly, this research assessed the heterogeneous characteristics of ESs across the mountains in arid regions; secondly, this research stresses the response of Ess to climate change.

The Tianshan Mountains region is a typical example of a large mountainous ecosystem in a temperate arid region [28]. Reputed as the “Water Tower of Central Asia”, the region contains the headstream of many rivers that flow into Central Asia [24] and has a significant impact on the water security of Central Asia as well as providing ecological security to both northwest China and Central Asia [21]. The ecological environment exhibits a high degree of complexity, spatial heterogeneity, and climate sensitivity, developing a complete vertical natural spectrum from a warm temperate and desert zone to an ice and permanent snow zone [28]. To promote the sustainable management of arid mountainous ecosystems and enhance their adaptability in coping with climate change, it is necessary to thoroughly study the spatiotemporal evolution patterns of mountainous ESs in arid regions, identify the important areas of ESs, and explore the constraints on the supply capacity of ESs. Understanding such aspects will facilitate the formulation of effective ES management policies.

Highlighting the Tianshan Mountains region as an example, we simulated four ESs, including habitat quality (HQ), soil retention (SR), water yield (WY), and carbon storage (CS), based on the integrated valuation of ecosystem services and tradeoffs (InVEST) and revised universal soil-loss equation (RUSLE) models. Accordingly, we explored the trade-offs and synergies among different ESs through correlation analysis, revealed the spatial pattern of trade-offs and synergies using spatial autocorrelation analysis, identified the importance level of ESs based on the number of hotspots, and analyzed the main driving factors of different ESs using the optimal parameter-based geographical detector (OPGD) model. The main objectives of this study were to: (1) quantify the spatiotemporal patterns of ESs from 2000 to 2020; (2) analyze the trade-offs and synergies among different ESs and spatial differences; (3) identify important areas of multiple ESs and their vertical zonality characteristics; and (4) reveal the contribution of natural–social factors to the spatial heterogeneity of ESs. The research framework is shown in Figure 1.

2. Materials and Methods

2.1. Study Area

Located in the central part of Xinjiang, the Tianshan Mountains region is 1852 km in length and covers a total area of about 23.53 × 10⁴ km², accounting for 14.13% of Xinjiang’s total area (Figure 2). The region spans eleven prefectures or cities, including Kizilsu Kirgiz Autonomous Prefecture, Kashi Prefecture, Akesu Prefecture, Yili Kazakh Autonomous Prefecture, Bortala Mongolian Autonomous Prefecture, Tacheng Prefecture, Bayingolin Mongol Autonomous Prefecture, Changji Hui Autonomous Prefecture, Urumqi City, Turpan City, and Hami City. The Tianshan Mountains are a major geographical boundary between southern and northern Xinjiang, with their southern and northern parts sandwiched by the Taklamakan Desert and Gurbantunggut Desert, respectively. This region has a highly complex topography characterized by a combination of mountains, valleys, and basins [21]. The region is dominated by a typical temperate continental climate, with large annual temperature differences and low annual precipitation. Its unique topography and climate have formed a typical “mountain–oasis–desert” ecosystem in the Tianshan Mountains. It is not only a prioritized area for biodiversity conservation in China but also one of the major global biodiversity hotspots [8].

2.2. Data Source

The data used included land-use data, a digital elevation model (DEM), the normalized difference vegetation index (NDVI), meteorological data, soil data, watershed boundaries, population density, and gross domestic product (GDP) density from 2000, 2005, 2010, 2015, and 2020 (

Table 1. Data source.

Data Set Name	Resolution	Data Source
Land-use data	30 m	Resources and Environmental Sciences Data Center, Chinese Academy of Sciences (https://www.resdc.cn/, accessed on 25 January 2022), including 6 primary types and 25 secondary types, with an overall evaluation accuracy greater than 80% [29]
DEM	90 m	Geospatial Data Cloud (http://www.gscloud.cn, accessed on 21 April 2022)
NDVI	30 m	National Ecosystem Science Data Center (http://www.nesdc.org.cn/, accessed on 25 January 2022)
Meteorological data	1000 m	National Tibetan Plateau Data Center (http://data.tpdc.ac.cn/, accessed on 10 May 2022), including monthly average temperature, monthly average precipitation, monthly potential evapotranspiration, etc.
Soil data	1000 m	Food and Agriculture Organization of the United Nations (https://www.fao.org, accessed on 21 April 2022)
Watershed boundary	-	National cryosphere Desert Data Center (http://www.ncdc.ac.cn, accessed on 21 April 2022)
Population data	100 m	WorldPop Datasets (https://www.worldpop.org/, accessed on 21 April 2022)
GDP data	1000 m	Resources and Environmental Sciences Data Center, Chinese Academy of Sciences (https://www.resdc.cn/, accessed on 21 April 2022)

Note: DEM is digital elevation model, NDVI is normalized difference vegetation index, and GDP is gross domestic product.

). Considering the inconsistent resolution of data from different sources, the resolution of all data was unified to 90 × 90 m to achieve good spatial consistency.

2.3. Ecosystem Service Evaluation

A series of ecological problems have emerged in the Tianshan Mountains, including grassland degradation, land desertification, and glacier retreat [24,29]. Thus, according to the classification of ESs proposed by the Millennium Ecosystem Assessment [3], along with current ecosystem characteristics and ecological pressures in the study area, four important ESs were investigated in this study, including habitat quality (HQ), soil retention (SR), water yield (WY), and carbon storage (CS). The widely distributed grasslands and forests provide important habitats for wildlife and have a profound effect on the global carbon cycle and the prevention of soil erosion. The glaciers provide the main source of water replenishment for rivers and play a crucial role in hydrological processes.

2.3.1. Habitat Quality

Habitat quality (HQ) is an important agent of biodiversity [30]. The HQ is assessed by the InVEST habitat-quality module, as follows:

$${\mathit{HQ}=\mathit{H}}_{\mathit{j}} \times [1 - (\frac{{\mathit{D}}_{\mathit{xj}}^{\mathrm{z}}}{{\mathit{D}}_{\mathit{xj}}^{\mathrm{z}}{+\mathit{k}}^{\mathrm{z}}}\left)\right]$$

(1)

where HQ is the habitat quality, ${\mathit{H}}_{\mathit{j}}$ is the habitat suitability of land-use type j, z and k are scaling parameters, and ${\mathit{D}}_{\mathit{xj}}$ is the total threat level in grid cell x with land-use type j.

2.3.2. Soil Retention

Soil retention (SR) refers to the difference between potential soil erosion and actual soil erosion. Annual SR is calculated using the RUSLE, as follows [31,32,33]:

$$\mathit{SR}=\mathit{R} \times \mathit{K} \times \mathit{L} \times \mathit{S} \times \left(1 – \mathrm{C} \times \mathrm{P}\right)$$

(2)

where $\mathit{SR}$ denotes the amount of soil retention (t·ha⁻¹·yr⁻¹); R, K, L, S, C, and P denote the rainfall erosivity factor (MJ·mm·(ha·yr)⁻¹), soil erodibility factor (t·ha·(MJ·ha·mm)⁻¹), slope length factor, slope factor, vegetation cover factor, and water and soil conservation measure factor, respectively.

2.3.3. Water Yield

The water yield (WY) can be evaluated based on the principle of water balance [16]. In this study, the WY of the study region was calculated using the annual water yield module in the InVEST model. The equation for WY is as follows:

$${\mathit{Y}}_{\mathit{ij}}=\left(1-\frac{{\mathit{AET}}_{\mathit{ij}}}{{\mathit{P}}_{\mathit{i}}}\right)·{\mathit{P}}_{\mathit{i}}$$

(3)

where ${\mathit{Y}}_{\mathit{ij}}$ denotes the WY (mm) of pixel i of land-use type j; ${\mathit{AET}}_{\mathit{ij}}$ denotes the real evapotranspiration (mm) of pixel i of land-use type j; ${\mathit{P}}_{\mathit{i}}$ denotes the annual precipitation (mm) of pixel i.

2.3.4. Carbon Storage

The carbon storage (CS) in ecosystems mainly depends on four basic carbon pools: aboveground biomass (${\mathit{C}}_{\mathit{above}}$), belowground biomass (${\mathit{C}}_{\mathit{below}}$), soil carbon storage (${\mathit{C}}_{\mathit{soil}}$), and dead organic matter (${\mathit{C}}_{\mathit{dead}}$) [5]; the formula for CS is as follows:

$${\mathit{C}}_{\mathit{i}}={\mathit{C}}_{\mathit{i}\mathit{-}\mathit{above}}+{\mathit{C}}_{\mathit{i}\mathit{-}\mathit{below}}+{\mathit{C}}_{\mathit{i}\mathit{-}\mathit{soil}}+{\mathit{C}}_{\mathit{i}\mathit{-}\mathit{dead}}$$

(4)

$${\mathit{C}}_{\mathit{total}}={\sum }_{\mathit{i}=1}^{\mathit{n}}{\mathit{C}}_{\mathit{i}}\times {\mathit{S}}_{\mathit{i}}$$

(5)

where ${\mathit{C}}_{\mathit{i}}$ denotes the total carbon density of land-use type i (t·ha⁻¹); ${\mathit{C}}_{\mathit{i}\mathit{-}\mathit{above}}$, ${\mathit{C}}_{\mathit{i}\mathit{-}\mathit{below}}$, ${\mathit{C}}_{\mathit{i}\mathit{-}\mathit{soil}},$ and ${\mathit{C}}_{\mathit{i}\mathit{-}\mathit{dead}}$ denote the carbon density of aboveground biomass, belowground biomass, soil carbon, and dead organic matter of land-use type i (t·ha⁻¹); ${\mathit{C}}_{\mathit{total}}$ denotes the total CS (t); ${\mathit{S}}_{\mathit{i}}$ denotes the area of land-use type i (ha⁻¹); and n denotes the number of land-use types.

To further characterize the spatial variation of ESs, we defined regions with an interannual variation of ESs greater than 5% as regions with increased ESs, while a variation of less than −5% denotes regions with decreased ESs. Areas with an interannual variation of between −5% and 5% have basically remained unchanged.

2.4. Trade-Offs and Synergies among ESs

To fully reveal the trade-offs and synergies among ESs, we analyzed the correlations among the four ESs from 2000 to 2020, using correlation analysis methods. We first created a uniform 1 × 1 km sampling grid in ArcGIS 10.6, using the fishnet tool. Then, we sampled the ESs in different years and, finally, conducted a correlation analysis. The equation for the trade-offs and synergies is as follows:

$${\mathit{r}}_{\mathit{xy}}=\frac{{\sum }_{\mathit{i}=1}^{\mathit{n}}\left[\left({\mathit{x}}_{\mathit{i}}- \overline{\mathit{x}}\right)\left({\mathit{y}}_{\mathit{i}}- \overline{\mathit{y}}\right)\right]}{\sqrt{{\sum }_{\mathit{i}=1}^{\mathit{n}}{\left({\mathit{x}}_{\mathit{i}}- \overline{\mathit{x}}\right)}^2}\sqrt{{\sum }_{\mathit{i}=1}^{\mathit{n}}{\left({\mathit{y}}_{\mathit{i}}- \overline{\mathit{y}}\right)}^2}}$$

(6)

where ${\mathit{x}}_{\mathit{i}}$ and ${\mathit{y}}_{\mathit{i}}$ denote two types of ESs; ${\mathit{r}}_{\mathit{xy}}$ denotes their correlation coefficients. According to the results of correlation analysis, trade-offs and synergies were classified into seven levels: highly significant trade-off (r < 0, 0.01 < p < 0.05), significant trade-off (r < 0, 0.05 < p < 0.1), trade-off (r < 0, p > 0.1), irrelevant (r = 0), highly significant synergy (r > 0, 0.01 < p < 0.05), significant synergy (r > 0, 0.05 < p < 0.1), and synergy (r > 0, p > 0.1) [34].

Using the spatial-weight module of the GeoDa software, a bivariate spatial autocorrelation analysis was performed on ESs at the pixel scale to measure the spatial trade-offs and synergies among the ESs. High–high aggregation and low–low aggregation indicate spatial synergies, while low–high aggregation and high–low aggregation indicate spatial trade-offs [30].

2.5. Ecosystem Service Hotspots

The areas where each ES exceeded its respective average at the grid scale were defined as hotspots of that ES [26]. A distribution map of the ES hotspots in different years was generated by overlying the hotspots of the four ESs in the corresponding years. On the basis of the number of ES hotspots per grid cell, ES hotspots were classified into five grades: (1) area of extreme importance, (2) area of high importance, (3) area of medium importance, (4) area of general importance, and (5) non-critical area (

Table 2. The importance grades of multiple ESs.

Classification	Criteria
Area of extreme importance	Four ecosystem services exceeding their respective averages
Area of high importance	Three ecosystem services exceeding their respective averages
Area of medium importance	Two ecosystem services exceeding their respective averages
Area of general importance	Only one ecosystem service exceeding its respective averages
Non-critical area	No individual ecosystem service exceeding its respective averages

).

2.6. Selection of Driving Factors

When disturbances from nature or humans exceed the critical threshold that an ecosystem can withstand, the ecosystem structure and function will change, thereby affecting ecosystem services. ESs are influenced by multiple factors (e.g., biological, climate, soil, topographic, and social factors). Based on previous studies [5,17,34,35], we selected nine potential natural–social influencing factors (

Table 3. Selection of potential influencing factors of ESs in the Tianshan Mountains.

Type	Influencing Factors	Unit	Code
topography	Elevation	m	ELE
	Slope gradient	°	SLO
climate	Mean annual temperature	°C	MAP
	Mean annual precipitation	mm	MAT
land-use type	Land-use type	–	LU
vegetation	Normalized difference vegetation index	–	NDVI
socio-economic force	Gross domestic product density	104 yuan/km2	GDP
	Population density	person/km2	PD
	Grazing intensity	–	GI

), grouped into topography (elevation and slope), climate (mean annual precipitation and mean annual temperature), land-use type (land-use type), vegetation (NDVI), socio-economic force (population density, GDP density, and grazing intensity). Specifically, elevation and slope are important indicators of mountainous landscape heterogeneity [5], while temperature and precipitation are key variables affecting vegetation growth and ecological processes in arid regions [21], land-use change affects the supply and interaction of ESs [11], and NDVI change is an important indicator of ecosystem status [20]. Land-use changes can lead to changes in the corresponding ecological processes, thus affecting ESs [34]. Usually, an increase in PD or GDP will impose pressure on the ecosystem [36], which may reduce ecological space, threaten biodiversity, and thereby hinder the supply of ESs [37,38]. Additionally, grazing is the main disturbance to mountainous ecosystems in arid regions, while overgrazing can cause grassland degradation and adversely affect ecosystem functions [39,40]. It is possible to use grazing intensity (GI) to characterize the disturbance level of grazing activities.

2.7. Optimal Parameter-Based Geographical Detector Model

A geographical detector model is an important tool for detecting and exploiting spatial heterogeneity [41] and is applicable in the field of ESs [34]. To overcome the spatial data discretization process based on professional experience rather than data-driven approaches and integrate spatial scale effect, spatial heterogeneity was quantitatively characterized using an optimal parameter-based geographical detector (OPGD) model. Compared with the geographical detector model, the OPGD model can provide flexible and comprehensive solutions through a suite of visualization tools to explore spatial factors, patterns, and heterogeneity more effectively [42]. In this study, we identified factors affecting the spatial heterogeneity of ESs through factor and interaction detection.

3. Results

3.1. The Spatiotemporal Evolution of Ecosystem Services

Figure 3 shows the temporal variation and spatial distribution of ESs from 2000 to 2020. Overall, HQ tended to decline slightly, from 0.5827 in 2000 to 0.5823 in 2020. High-value areas were mainly concentrated in the central Tianshan Mountains (e.g., the Yili Valley and Bogda Mountain), where forests and high-coverage grasslands are widely distributed, and habitat suitability is high. Low-value areas were mainly distributed in the southwestern Tianshan Mountains, a transitional zone between the Tarim Basin and the Tianshan Mountains, where deserts are the dominant vegetation type and the ecological environment is extremely fragile. HQ has not changed significantly in most areas (99.66%) of the Tianshan Mountains over the 20 years in question.

SR showed a fluctuating downward trend, with an annual mean SR of 14.52 × 10⁸ t over the 20 years in question. The maximum and minimum values of SR were 18.38 × 10⁸ t in 2015 and 10.30 × 10⁸ t in 2020, respectively. High-value areas were mainly distributed in the central and northern Tianshan Mountains, exhibiting a spatial distribution consistent with that of forests and grasslands. Low-value areas were mainly located in the western and eastern Tianshan Mountains, where vegetation coverage is low and soil erosion is severe. Over the two decades studied herein, SR did not change significantly, except for a significant decline in the central Tianshan Mountains. SR basically remained unchanged in 76.14% of the study region but declined in 22.00% of the area.

From 2000 to 2020, the annual average WY was 154.99 mm. Annual WY first increased and then decreased, specifically increasing from 142.90 mm in 2000 to 176.55 mm in 2015 and decreasing to 128.07 mm in 2020. High-value areas were distributed on both sides of the Yili Valley, where rainfall is abundant. Low-value areas were mainly located in the eastern and southwestern Tianshan Mountains, where the climate is arid and evapotranspiration is intense. In the twenty years studied herein, WY increased significantly in the northwestern Tianshan Mountains but tended to decrease in other parts of the study area. WY tended to increase in 15.65% of the study region but decreased in 62.69% of the area.

The total CS tended to decrease slightly, and the annual average CS was 16.26 × 10⁸ t from 2000 to 2020. The maximum and minimum values of total CS were 16.40 × 10⁸ t in 2005 and 16.21 × 10⁸ t in 2020, respectively. High-value areas were mainly concentrated in the mountainous forest zone and grassland belt, whereas low-value areas were mainly concentrated in the southern desert and snow belts. Over the two decades in question, CS basically remained unchanged in most parts (99.87%) of the study region.

From 2000 to 2020, ESs in the study region exhibited significant spatial heterogeneity. The spatial distribution of HQ and CS basically remained stable, whereas that of SR and WY exhibited significant fluctuations.

3.2. Trade-Offs and Synergies of Ecosystem Services

The results of the correlation analysis (Figure 4) showed that 29 out of 30 ES pairs exhibited highly significant synergy (p < 0.01) from 2000 to 2020. The coefficient of correlation between HQ and CS was the largest and remained relatively stable. This is mainly because their spatial distribution patterns were basically the same, wherein densely vegetated areas not only provide suitable and diverse habitats but are also important carbon pools. The coefficient of correlation between WY and CS was the smallest, and they were no longer correlated by 2020. The HQ–SR and SR–CS correlations showed a fluctuating upward trend, whereas the HQ–WY and SR–WY correlations showed a fluctuating downward trend.

As shown in Figure 5, the Moran index of the annual means of all individual ESs exceeded 0.80 (p < 0.01) during 2000–2020, indicating that all ESs exhibited significant global spatial autocorrelation. This finding is consistent with the results of the correlation analysis. The HQ–CS pair exhibited highly significant synergies, and their Moran index and correlation coefficient were 0.68 and 0.79, respectively. This was followed by the WY–SR pair, the Moran index and correlation coefficient of which were 0.47 and 0.46, respectively. The WY–CS pair exhibited weak synergies, and their Moran index and correlation coefficient were both 0.02.

The results of the bivariate spatial autocorrelation analysis further revealed the spatial patterns of trade-offs and synergies between different ESs from 2000 to 2020. Synergies were dominant in the HQ–SR pair, covering 44.71% of the study region. High–high aggregation areas were mainly concentrated in the central Tianshan Mountains, whereas low–low aggregation areas were mainly distributed in the eastern and southern Tianshan Mountains. The 7.44 × 10⁴ km² spatial trade-off areas of the HQ–SR pair were concentrated in the eastern and western Tianshan Mountains and were widely scattered in the central Tianshan Mountains. The spatial synergy and trade-off areas of the HQ–WY pair covered 37.96% and 34.81% of the study region, respectively, and the northern and southern parts of the study area were dominated by synergies, while the eastern and central parts were dominated by trade-offs. In HQ-CS pairs, 75.63% of the regions showed a synergistic relationship in terms of spatial distribution, for which the high-high aggregation area is the largest, reaching 11.56 × 10⁴ km². The spatial synergy areas of the SR–WY pair accounted for 56.95% of the study region. Specifically, high–high aggregation areas (36.17%) were mainly distributed in the northern and central Tianshan Mountains, whereas low–low aggregation areas (63.83%) were mainly distributed in the eastern and southern Tianshan Mountains. The spatial synergy and trade-off areas of the SR–CS pair covered 47.94% and 32.74% of the study region, respectively, and were alternately distributed. The spatial synergy and trade-off areas of the WY–CS pair covered 40.34% and 40.33% of the study region, respectively, and exhibited evident spatial transition characteristics. Specifically, high–high aggregation areas were dominant in the northwestern Tianshan Mountains, while low–low aggregation areas were dominant in the southwestern Tianshan Mountains, and the high–low aggregation and low–high aggregation areas were distributed between them.

3.3. Spatiotemporal Variation in Ecosystem Service Hotspots

Figure 6 shows the spatial distribution of hotspots of multiple ESs from 2000 to 2020. The proportions of different grades of ES hotspots remained relatively stable in the twenty years under study. Specifically, the non-critical area, area of general importance, area of medium importance, area of high importance, and area of extreme importance accounted for 16%, 10%, 32%, 21%, and 20% of the entire study region, respectively. The area of extreme importance was mainly distributed in the northern Tianshan Mountains and on both sides of the Yili Valley, while the area of high importance was mainly distributed in the central Tianshan Mountains. It is noteworthy that the area of high importance also existed in the northwestern Tianshan Mountains in 2005, 2010, and 2020, indicating that ESs in this area had improved in these years. Specifically, WY in these regions increased significantly in 2005, 2010, and 2020. An increase in precipitation and a decrease in evapotranspiration can significantly increase WY. The area of medium importance was mainly distributed in the eastern and northwestern Tianshan Mountains; the area of general importance was mainly distributed in high-altitude mountainous areas, while the non-critical area was mainly distributed in the southern Tianshan Mountains. In the study region, the spatial distribution pattern of ES hotspots is consistent with that of the landscape, exhibiting both the vertical zonality and the mountain–oasis–desert environmental gradient of the area.

To further investigate the topographic gradient characteristics of ESs, we calculated the proportions of ES hotspots along the altitudinal gradient in 2020 (Figure 7). The results showed that the non-critical area was mainly distributed in low-altitude areas (more specifically, areas with an altitude of less than 2000 m), with the largest proportion of areas (44.00%) being at an altitude of 1400 to 1600 m. The areas of general importance were concentrated at altitudes of more than 3600 m, where the ground is covered with permanent glaciers and snow, providing limited ESs. The areas of medium importance were mainly located at altitudes of less than 2400 m. The areas of high importance and extreme importance were mainly concentrated at altitudes of 2400 to 3600 m. The region is the main distribution area for forests and grasslands, accounting for 51% and 50% of their total area, respectively. Thus, this area provides a variety of ESs, such as biodiversity, carbon storage, water retention, and soil retention. Overall, middle- and high-altitude arid mountainous areas provide diverse ESs and play a crucial role in maintaining the ecological security of arid mountainous areas.

3.4. Mechanism Underlying the Spatial Variation of Ecosystem Services

3.4.1. Comparison of the Influencing Factors between Different Ecosystem Services

The OPGD model was used to explore the individual effects of the driving factors of ESs; the results are shown in Figure 8. The dominant factors influencing HQ were LU, NDVI, and MAT, with average q-values of 0.4708, 0.4443, and 0.4018, respectively. Over the two decades studied herein, the determining power of LU, GDP, and GI increased significantly by 0.3459, 0.0330, and 0.0247, respectively. SR was mainly influenced by MAT, NDVI, ELE, and MAP, with average coefficient values of 0.3087, 0.3069, 0.2915, and 0.2859, respectively. From 2000 to 2020, the contribution of LU increased by 0.0523, whereas the contribution of GDP decreased by 0.0543. MAP demonstrated the greatest explanatory power of WY, with a q-value of 0.9014. In addition, MAT and ELE showed better explanatory power on WY, and their q-value were 0.5756 and 0.4457, respectively. The contribution of LU and GDP decreased significantly by 0.1267 and 0.0936 during the 20 years in question, respectively. CS was predominantly influenced by LU, NDVI, and MAT, with average q-values of 0.5168, 0.4249, and 0.4059, respectively. In the twenty-year period studied, the contribution of LU and GDP increased significantly by 0.3092 and 0.0516, respectively, whereas the contribution of MAP decreased by 0.0624.

3.4.2. Interactive Effect of Driving Factors

The effects of various factors on ESs are often not independent; therefore, the interaction effects of different drivers were further tested, and the 2020 results were selected for specific analysis. Figure 9 shows the interactive effect of natural and anthropogenic factors on ESs. There were 36 pairs of interactive effects, which manifested as a nonlinear or bilateral enhancement for HQ, SR, WY, and CS. This indicated that the interaction of any two factors contributed more significantly to the spatial variation in ESs than a single factor; in other words, the effects of natural and social factors on ESs can promote each other.

For HQ (Figure 9a), 16 pairs of drivers showed nonlinear enhancement, with the largest interaction effect seen between MAP and ELE (0.3983); 20 pairs of drivers showed a bivariate enhancement, with the largest interaction effect between LU and NDVI (0.8105). For SR (Figure 9b), this was nonlinearly enhanced by 14 pairs of driving factors, among which NDVI and SLO produced the most significant interactive effect (0.6061). SR was bilaterally enhanced by 22 pairs of driving factors, among which ELE and NDVI produced the most significant interactive effect (0.5246). For WY (Figure 9c), there were 15 and 21 pairs of driving factors that showed nonlinear enhancement and bivariate enhancement, respectively. The interaction effect of LU and MAP and of MAT and MAP was the largest, reaching 0.9367 and 0.9382, respectively. For CS (Figure 9d), 23 pairs of drivers showed a bivariate enhancement, and the rest showed a nonlinear enhancement, among which LU and NDVI produced the most significant interactive effect (0.7950).

4. Discussion

4.1. Validation of Ecosystem Service Results in the Tianshan Mountains

The InVEST and RUSLE models have gradually emerged as basic tools in evaluating ESs, and they have proven to provide scientific and reasonable results [5,13]. Therefore, these models were used in this study to evaluate the mountainous ESs in an arid region. The results provide basic data for analyzing trade-offs and synergies and for identifying ES hotspots, and the quality of such data will affect the reliability of the results of subsequent related analyses. Thus, precision validation of each ES is fundamental to ensure the quality and accuracy of ES assessment results. In this study, the results of ES evaluation were compared with measured data and the related research results.

We found that areas with high HQ were mainly distributed in mountainous areas, while areas with low HQ were concentrated in desert areas and were typically represented by the Tarim Basin and Junggar Basin, on the southern and northern slopes of the Tianshan Mountains. The previous study also verifies our results [27,43].

The annual average SR of the study region was 53.73 t/km², 59.67 t/km², 39.51 t/km², 63.5 t/km², and 35.89 t/km² in 2000, 2005, 2010, 2015, and 2020, respectively. Using the RUSLE model, Zhang et al. [44] calculated the SR in the Three-North Shelter Forest areas of China, finding that the annual average SR of the Tianshan Mountains was 35.57 t/km², 30.02 t/km², 40.82 t/km², 35.52 t/km², and 28.69 t/km² in 2000, 2005, 2010, 2015, and 2020, respectively. Their results are basically consistent with the results of our study. As is consistent with previous findings [45,46], our study found that SR was highest in mountainous areas but was lowest in the desert and Gobi areas in the northern and southern Tianshan Mountains.

The water resources data are counted by administrative districts in the Xinjiang Water Resources Bulletin, and the area covered is not consistent with the study area. Therefore, we recalculated the administrative districts involved in the Tianshan Mountains region (including Urumqi City, Karamay City, Turpan Prefecture, Hami Prefecture, Changji Hui Autonomous Prefecture, Bortala Mongolian Autonomous Prefecture, Akesu Prefecture, and Yili Kazakh Autonomous Prefecture) using the same parameters and verified them with data published in the Xinjiang Water Resources Bulletin (2015). The published total water resources of the region were 33.277 billion m³, while the total water production estimated in our study was 33.249 billion m³. We further analyzed the simulated total water production of each administrative district and the actual total water resources of these administrative districts and found a significant linear correlation between them (r = 0.97, p < 0.01); the regression fitting results (regression coefficient = 0.735; R² = 0.94) were basically consistent with the WY evaluation results in this study.

The key to validating CS results is the accuracy of carbon density. By integrating the relevant experimental and literature data, Xu et al. [47] found that the measured aboveground, belowground, and soil carbon density of grasslands in the Tianshan Mountains ranged from 0.14 to 1.99 t·ha⁻¹, 1.74 to 7.09 t·ha⁻¹, and 24 to 253.5 t·ha⁻¹, with average values of 0.68 t·ha⁻¹, 3.70 t·ha⁻¹, and 126.69 t·ha⁻¹, respectively. The corrected aboveground and belowground carbon densities of grassland were close to the average measured values, whereas the corrected soil carbon density was significantly different from the average measured values. Furthermore, the average vegetation coverage of sampling sites was higher than 0.5, indicating that high-coverage grasslands were dominant at the sampling sites, whereas sampling sites with medium- or low-coverage grasslands were relatively few. Therefore, the corrected soil carbon density was significantly different from the measured carbon density. Based on data from China’s first and second national soil surveys, Zhang et al. [48] estimated the soil carbon density in western China and found that the average carbon density of oasis farmlands, forests, meadow grasslands, and desert grasslands was 79.5 t·ha⁻¹, 158.8 t·ha⁻¹, 123.6 t·ha⁻¹, and 63.9 t·ha⁻¹, respectively. Compared with the corrected soil carbon density in the present study, the estimated soil carbon density of forests and grasslands was significantly different, with relative errors of 32.10% and 19.42%, respectively, whereas the estimated soil carbon density of other land use types was slightly different. Overall, the CS values of this study are basically consistent with the available measured data.

In summary, the results of HQ, SR, WY, and CS, assessed by the InVEST and RUSLE models, are generally consistent with previous studies in the same arid zone and the Tianshan Mountains. Evidently, the simulation results of ESs in arid regions are highly reliable and can reflect the actual ES variation in the study region. The differences in related values may be associated with the spatiotemporal scales selected in the research area and model parameter settings.

4.2. Implications and Suggestions for Ecosystem Management

Owing to their rugged topography, low temperature, steep slopes, and geographical remoteness [49], mountainous areas are highly susceptible to climate change [20], especially in arid and semiarid areas, which is confirmed by our findings. It is becoming increasingly obvious that climate change is one of the greatest threats contributing to land use and vegetation coverage change today and, possibly, in the future [21,50]. The ecosystems of the Tianshan Mountains are extremely vulnerable and highly sensitive to climate change, especially regarding water scarcity [21]. As shown in Figure 10a,b, the Tianshan Mountains have experienced considerable increases in temperature and intensified fluctuations in precipitation in the twenty years under study. The increasing temperature leads to increased evaporation, which threatens plant growth in arid regions. Despite the increase in precipitation, the climate in most parts of Tianshan Mountains is extremely arid, with the evaporation rate usually higher than the actual precipitation. Drought-induced water deficit suppresses vegetation growth, and the negative ecological effects of climate warming are emerging [51]. Therefore, MAT has significant effects on biodiversity, soil conservation, water yield, and carbon storage. Precipitation is also an important factor affecting SR and WY. The increase in precipitation is beneficial to improving hydrological conditions, regulating the water cycle, and improving the overall level of WY [38]. With the transformation of hydrothermal conditions, the vegetation types and growth status of terrestrial ecosystems will change correspondingly, which will affect the capacity of soil retention. Meanwhile, abnormal and fluctuant precipitation has brought on the risk of soil erosion [52]. Therefore, in the future, there is a need for mitigation and adaptation strategies in arid mountainous ecosystems that can enhance the supply capacity of ESs to address climate change and realize sustainability goals.

As is consistent with previous studies [53,54], our results show that anthropogenic disturbance to ESs is slight; the reason for this may be as follows. Firstly, most of the study area belongs to the key ecological function area of Xinjiang; thus, most areas were not allowed to be used for economic activities. Secondly, economic growth could also allow more funds to be invested in ecological management, partially neutralizing the negative effects brought about by economic development [55]. Thirdly, the establishment of protected areas and priority areas for biodiversity conservation has effectively protected the ecosystem and biodiversity of the Tianshan Mountains. Nevertheless, the impact of grazing on multiple ESs should be given due attention. Large-area grasslands in the Tianshan Mountains are used as pasturelands, so grazing constitutes the primary anthropogenic disturbance to these ecosystems. Moderate grazing can promote plant growth under water-stress conditions [39], whereas overgrazing will lead to the degradation of grasslands and shrubs [22,40], posing a great challenge to the restoration of mountainous ecosystems. At this point, grassland grazing patterns should be improved to maintain a reasonable grazing capacity and reduce the impact of grazing on the ESs provided by mountainous grasslands.

Mountainous ESs are significantly spatially heterogeneous along the different environmental gradients [5,10,13]. In this study, we found that high-altitude mountainous areas can provide multiple important ESs. Therefore, adaptive management interventions are needed for forest and grassland ecosystems at middle and high altitudes, as well as glacial ecosystems at high altitudes, these being key areas and priorities for ecosystem conservation and restoration. This will be of great significance to regional ES enhancement, ecological security, and sustainable development, and will contribute to the sustainable management of arid zone mountainous ecosystems.

4.3. Limitations and Future Work

Despite its merits and contributions in evaluating the ESs of the Tianshan Mountains, this study has a few limitations. First, this study assessed only four types of major ESs, while other ESs, such as livestock production and natural recreation, were ignored. This was mainly because of the lack of corresponding statistical data and the absence of detailed data on the township level. Second, the parameters adopted in the evaluation models were mainly selected based on relevant studies in similar regions. In future studies, we need to localize and adjust the model parameters used for ES estimation through data collection and observation in the field to improve the accuracy of the model assessment results. Third, the impact of climate change on the ecosystems of the Tianshan Mountains is noticeable, and the simulation of ESs under different climate change scenarios needs to be enhanced in the future. Finally, this study analyzed the vertical zonal differences in ESs only in terms of the altitudinal gradient. Since the Tianshan Mountains stretch across a wide east–west area and are usually divided into East, Middle, and West Tianshan, the factors affecting the pattern of ESs may also differ; therefore, the longitudinal zonal differences in ESs need to be considered in the future.

5. Conclusions

Using multi-source data and multiple models, we quantitatively evaluated the ESs in the Tianshan Mountains of Xinjiang and analyzed the characteristics of their spatiotemporal evolution. Through correlation analysis and bivariate spatial autocorrelation analysis, we then analyzed the trade-offs and synergies of multiple ESs. ES hotspots were identified using a spatial overlay analysis. Using the OPGD model, we finally ascertained the natural–social factors contributing to the spatial heterogeneity of ESs. The results showed that ESs in Tianshan exhibited significant spatial heterogeneity, with higher ES supply capacity in densely vegetated mountainous areas in the western and central parts of the study area, but low capacity in sparsely vegetated areas in the southwestern and eastern regions. There were 29 ES pairs exhibiting highly significant trade-off values. The areas of spatial synergy were mainly distributed in the northwestern and southwestern Tianshan Mountains, whereas the areas of spatial trade-off were scattered in other areas. ESs in the study area were dominated by areas of medium importance, accounting for 32% of the study area, and were mainly distributed in areas with an altitude of less than 2400 m. The areas of high importance and extreme importance covered 21% and 20% of the study area, respectively, and were concentrated in areas between 2400 and 3600 m above sea level. LU is the main driving factor of HQ and CS; MAT and NDVI had a strong explanatory power for SR, and MAP was the dominant factor in determining the spatial distribution of WY. Therefore, policymakers should develop appropriate management measures based on the factors affecting different ESs to achieve ecological conservation and the sustainable development of mountainous areas in arid regions.