Threshold Effects and Synergistic Trade-Offs in Ecosystem Services: A Spatio-Temporal Study of Kashgar’s Arid Region

Abstract

The complex trade-offs and synergies among ecosystem services (ESs) in arid regions influence the stability and sustainable development of regional ecosystems. As a representative oasis–desert transition zone, the Kashgar region requires quantifying the key drivers and thresholds influencing ecosystem services, which is crucial for regional management. This study examines the spatio-temporal changes and interactions of five types of ES (grain production, water yield, soil retention, carbon storage, and habitat quality) and employs Restricted Cubic Splines to quantify the nonlinear changes and threshold effects of natural and social drivers. The results indicate the following: (1) During the period from 2000 to 2020, supply services (grain production) and regulatory services (water yield and soil retention) showed growth, while support services (carbon storage and habitat quality) declined slightly; (2) the synergistic effects of ecological services improved across the entire region, but trade-off effects emerged in certain local areas; and (3) the NDVI is the core natural factor driving the spatio-temporal differentiation of ESs. In 2020, when the NDVI exceeded 0.35, it had an adverse impact on habitat quality and carbon storage. Among social factors, water yield and habitat quality exhibit the highest threshold points with land use development intensity. An increase in land development intensity significantly impacts the trade-off and synergistic relationships among ESs, leading to local imbalances in ES resource supply and demand. These findings enhance our understanding of the nonlinear characteristics and potential mechanisms of ecosystems in arid regions, providing a scientific basis for ecosystem management in these areas.

1. Introduction

As global climate change intensifies and human economic activities expand, the stability and resilience of ecosystem services (ESs) face unprecedented challenges [1,2]. These conditions pose serious constraints on the ability of ecosystems to provide services and goods to humans which, in turn, challenges the realization of global sustainable development goals [3]. In light of ongoing crises and opportunities, the conservation and sustainable use of ecosystem services is seen as one of the core pathways to achieve global food security, water resources management, biodiversity conservation, and other key objectives [4]. Therefore, thorough insights into the pressures informing the spatial and temporal variability of ESs and their interactions, especially the threshold effects under the influence of multiple environmental and social pressures, are essential for maintaining a harmonious relationship between humans and nature [5,6].

Arid regions, defined by unique geographical and climatic conditions, resource scarcity, and environmental fragility, present more complex spatio-temporal changes in ESs [7]. Existing research indicates that relationships among ESs exhibit considerable spatio-temporal heterogeneity, appearing as synergies (the simultaneous increase or decrease in two services) or trade-offs (where gains in one service result in losses in another) [8,9]. While numerous analyses have concentrated on individual or limited sets of ES types, studies integrating the overall impact of multilevel interaction mechanisms on regional ecosystem services remain scarce [10]. In addition, the spatial and temporal variations in ESs often exhibit significant nonlinear characteristics [11]. Specifically, when influential factors surpass a critical juncture, the ecosystem experiences a state shift, signifying the presence of threshold effects [12]. These thresholds represent critical points at which ecosystems transition from one state to another [13]. The identification and measurement of threshold effects in ES drivers are common practices in scientific research and management [14].

While current research on ES impact mechanisms has comprehensively demonstrated the positive and negative effects of driving factors on ESs [15] and further verified the presence of threshold effects [16], important research gaps remain. First, previous studies have explored the trade-offs and synergies among ESs by utilizing correlation analysis, linear regression, or other basic statistical methods. These approaches often rely on linear assumptions [17]. However, interactions among ESs in actual ecosystems can exhibit thresholds, nonlinear inflection points, and complex dynamic responses [11,18]. Second, there is a relative lack of quantifying the response characteristics and critical thresholds of different ESs to external disturbances at the regional scale in arid zones [19,20]. The Kashgar region is part of the Tarim River basin and serves as a typical example of China’s arid zone ecosystems. This region also functions as a transition zone between oases and deserts [21]. The fragility and instability of Kashgar’s ecosystems render them highly sensitive to external disturbances. As climate change intensifies, ecological and environmental pressures continue to increase, with land desertification and vegetation degradation significantly weakening the region’s ecosystem regulatory capacity. Meanwhile, population growth further exacerbates land use intensity, and overdevelopment may lead to an imbalance between supply services and regulatory services, thereby weakening the ecosystem’s natural recovery capacity [22]. Therefore, there is an urgent need to systematically reveal the nonlinear relationships from a trade-off and synergy perspective, particularly the characteristics of threshold changes from local to global scales and their ranges of influence, to provide new solutions for regulating thresholds in vegetation ecological restoration and land use planning [23]. In this context, selecting appropriate research methods for studying the nonlinear characteristics and critical thresholds of ecosystem services is particularly critical. In terms of research methodology, existing studies often involve subjective categorizations of continuous variables, especially in determining the number of categories and node positions, which contributes to a loss of information [14,24]. The use of RCSs (Restricted Cubic Splines) has been proposed to address this methodological limitation. RCSs can smooth fitting curves while avoiding overfitting, thus effectively capturing nonlinear features and threshold effects and has been widely used in ES studies in several regions [25,26].

This study is based on remote sensing data and GIS technology, combining the supply, trade-offs, and synergies of ESs in the Kashgar region from 2000 to 2020. This work analyzes the overall trade-offs and synergies in the Kashgar region and uses RCS curves to analyze the nonlinear responses of different ES drivers and their critical thresholds. The specific objectives of this study are (1) to seek the spatial–temporal patterns of ecosystem services, (2) identify the trade-offs and synergies among ecosystem services, (3) consider the nonlinear relationships and threshold relations between ecosystem services and their driving factors, and (4) propose regional adaptive management strategies that will enhance the integration of social systems and ecological systems in the region. The findings of this study are important in explaining how arid ecosystem services interact with various driving factors and their implications for the optimized and scientifically based sustainable resource ecological restoration and development of this vulnerable region [27,28,29].

2. Materials and Methods

2.1. Study Area

The Kashgar region, located in the southwestern portion of Xinjiang in northwestern China (85°31′–91°04′ E and 45°00′–49°10′ N), encompasses an area of approximately 1.62 × 10⁵ km² (Figure 1). As the largest oasis cluster in southern Xinjiang, Kashgar typifies an arid environment and occupies a zone of integrated agricultural and pastoral land use [30]. With a population approaching 5 million concentrated in the oasis, significant pressure is placed on the area’s delicate ecological balance. The growth of artificial oases, coupled with imbalanced water and soil resources, has created significant human–land conflicts, leading to contractions of the oasis–desert transition zones and desertification expanding along the oasis edges [31,32]. The stability and sustainable supply of Kashgar’s ecosystem services are essential for maintaining the ecological security of this region and the wider Midwest [33]. However, due to regional vulnerability and increased human activities, ecological problems such as biodiversity reductions and soil erosion have become increasingly prominent, seriously affecting the ability to sustainably supply ecosystem services [34].

2.2. Data Collection and Processing

This study used basic geographic data, meteorological data, remote sensing data, socio-economic data, and statistical yearbook data for the period of 2000–2020. The spatial resolution of the remote sensing data was used as the benchmark, and the other raster data were resampled to 1 km. The coordinate system was standardized as “WGS_1984_UTM_Zone_53N”. Temporally, except for geomorphological and soil data, which remained unchanged for many years, all other data remained continuous in time. More details on the data are shown in

Table 1. Data sources for this study.

Data Type	Data Source/Processing	Spatial- Resolution
Land use/land cover	Resource and Environment Science and Data Center (http://www.resdc.cn/ (accessed on 13 August 2025))	30 m
Precipitation	National Earth System Science Data Center, National Science & Technology Infrastructure of China (http://www.geodata.cn (accessed on 13 August 2025))	1 km
Temperature	National Earth System Science Data Center, National Science & Technology Infrastructure of China (http://www.geodata.cn (accessed on 13 August 2025))	1 km
Evapotranspiration	National Earth System Science Data Center, National Science & Technology Infrastructure of China (http://www.geodata.cn (accessed on 13 August 2025))	1 km
Root depth, soil texture, and organic carbon content	China soil map based harmonized world soil database (HWSD) (v1.1)	30 arc-seconds
Carbon density	A dataset of carbon density in Chinese terrestrial ecosystems (2010s) (http://www.doi.org/10.11922/sciencedb.603 (accessed on 13 August 2025))	/
Digital elevation model (DEM)	Resource and Environment Science and Data Center (http://www.resdc.cn/ (accessed on 13 August 2025))	90 m
Normalized difference vegetation index (NDVI)	Resource and Environment Science and Data Center (http://www.resdc.cn/ (accessed on 13 August 2025))	1 km
Grain production	Xinjiang statistical Yearbook	/
Road network	https://www.openstreetmap.org (accessed on 13 August 2025)	/
Human footprint index	Socio-economic data and application center https://sedac.ciesin.columbia.edu (accessed on 13 August 2025)	1 km
GDP	Resource and Environment Science and Data Center (http://www.resdc.cn/ (accessed on 13 August 2025))	1 km
Population density	Resource and Environment Science and Data Center (http://www.resdc.cn/ (accessed on 13 August 2025))	1 km

.

2.3. Methods

2.3.1. Research Framework

An analytical approach is proposed for optimal regional ecosystem management based on an ES assessment and nonlinear threshold effect analysis framework (Figure 2). (1) Based on the InVEST model, the spatial and temporal distribution patterns of five ESs in 2000, 2010, and 2020 were analyzed and compared. (2) Using Spearman analysis and RMSE, the spatio-temporal values of ES interactions were calculated. We also analyzed the trade-offs and synergistic relationships of the region to construct a comprehensive ES index. (3) Through a redundancy analysis, social and natural factors were introduced as drivers to analyze the intrinsic driving spatio-temporal mechanisms of Kashgar. (4) We identified the nonlinear threshold effects of drivers on the ESs using the RCS method. In this way, we developed an ecological management strategy based on regional differentiation in Kashgar.

2.3.2. ES Measurement

Five ES categories were selected for this study based on variations in topography, climate, vegetation, and water systems, as well as the intensity of human activity disturbance in the Kashgar region [35,36]. We also explored the region’s ecological and environmental challenges. Water resources are essential for oasis preservation in arid regions, and soil retention underpins agricultural productivity. Habitat quality reflects the ecosystem’s ability to support biodiversity and offer a habitat, acting as a key measure of ecosystem health. Carbon sequestration capacity represents the ecosystem’s critical capacity to reduce climate change and achieve carbon neutrality. Agricultural production, a significant economic activity in the Kashgar region, directly influences regional food security. The five selected ESs were grain production (GP), water yield (WY), soil retention (SR), carbon stock (CS), and habitat quality (HQ). GP calculations utilized food production values for each county, transforming statistical data into spatial distributions (

Table 2. Assessment formulas and parameter significance for ESs.

Ecological Functions	Equation	Introduction of Spatialization Assessment Methods
WY	$W_x=\left(1-{\displaystyle \frac{A_x}{P_x}}\right)\times P_x$ ${\displaystyle \frac{A_x}{P_x}}={\displaystyle \frac{1+{\omega }_x+R_x}{1+{\omega }_x{R}_x+\left(\frac{1}{R_x}\right)}}$ ${\omega }_x=z\times {\displaystyle \frac{C_x}{P_x}}, R_x={\displaystyle \frac{k_x\times E_0}{P_x}}$ $C_x=Min\left(D_s,D_R\right)\times P_{AWCx}$	Where Wx is annual water yield (mm), and Pxis the average annual precipitation of grid cell x (mm). Ax is average annual actual evapotranspiration of grid cell x (mm). ωx is a non-physical parameter of natural climate–soil properties, and Z is the zhang coefficient, taking the value of 1.5, which is closest to the natural runoff in the Regional Water Resources Bulletin; E0 is the potential evapotranspiration (mm); k is the vegetation evapotranspiration coefficient; and after correcting the results, k was set to 0.65 for cropland, 0.7 for forest land, 0.75 for grassland, 1 for watersheds, 0.3 for unutilized land, and 0.03 for construction land. Cx is the amount of plant-available water per unit area (mm); DS is the soil depth (mm), and DR is the root depth (%); and PAwcx is the proportion of plant-available water per unit depth of soil, which was calculated using the HWSD Soil Database in combination with empirical formulas [38].
SR	$SR=R\times K\times LS\times C\times P$ $R={\alpha }_1{P}_J^{{\beta }_1}$ $$\begin{gathered} K=\left\{0.2+0.3exp\left[-0.0256S\left(1-{\displaystyle \frac{L}{100}}\right)\right]\right\}{\left({\displaystyle \frac{L}{A+L}}\right)}^{0.3} \\ \times \left[1-{\displaystyle \frac{0.25B}{B+\mathrm{e}\mathrm{x}\mathrm{p}(3.72-2.95B)}}\right] \\ \times \left[1-{\displaystyle \frac{0.7N_1}{N_1+exp\left(22.92S{N}_1-5.51\right)}}\right], N_1 \\ =1-{\displaystyle \frac{S}{100}} \end{gathered}$$	SR represents soil erosion volume (t/(hm2·a)), which characterizes the ability of human activities (vegetation restoration and engineering measures) to inhibit soil erosion. R is the rainfall erosive force (MJ mm/(hm2·h·a)), while α1 and β1 are model parameters, with values of 0.0534 and 1.6548, respectively [39]. K is the soil erodibility factor [t·hm2·h/(hm2·MJ·mm)], where S, L, and A represent the mass fractions of sand, silt, and clay, respectively, and B is the mass fraction of organic carbon = organic matter mass fraction/1.724. During the calculation process, the particle size content was multiplied by 100, and the calculated K value was divided by 7.59 to obtain the soil erodibility K value in the International System of Units [40]. The version of the InVEST model used in this study does not require the preparation of slope length and slope angle factors during runtime; the module automatically calculates LS based on the DEM data of the study area. C is the vegetation cover factor and management factor, and P is the soil and water conservation measures factor [41].
CS	$C_{total}=C_{above}+C_{below}+C_{soil }+C_{dead}$	Where Ctotal is the total carbon stock in grid cell x (t/ha); Cabove is surface carbon density (t/ha); Cbelow is below-ground biomass carbon density (t/ha); Csoil is soil organic matter carbon density (t/ha); and Cdead is the carbon density of dead organic matter (t/ha). The carbon parameters were determined based on existing studies [42].
HQ	$Q=H_j\left(1-{\displaystyle \frac{D_j^2}{D_i^2+k^2}}\right)$	Where Qj is the habitat quality of land use type j; Hj is the habitat suitability of land use type j; Dj represents the level of stress to which land use type j is subjected; and k is the half-saturation constant with a value of 0.5.
GP	${GP}_{ij}={\displaystyle \frac{{NDVI}_{ij}}{\sum {NDVI}_{ij}}}\times GP$	In the spatial allocation of grain production, NDVI is used as a proxy indicator for food productivity [43]. First, the NDVI weights of each grid cell were calculated, and the total grain production of the region was allocated to the farmland grid cells according to the NDVI weights.Where GPij is the grain production of the grid cell (i,j); GP is the total grain production of the region; NDVIij represents the NDVI value of grid cell (i,j); and $\sum {NDVI}_{ij}$ is the total sum of NDVI values for all arable land grid cells in the region.

). Spatially distributed data for WY, SR, CS, and HQ were generated utilizing biophysical models [37].

2.3.3. Interactions and Relationships Among ESs

To explore the relationships between pairs of ESs, correlation and statistical analyses were conducted. Since the basic data for this study have a resolution of 1 km, to avoid data redundancy and reduce analytical bias caused by high spatial autocorrelation, a fishing net with a grid size 3 km was established. The raster data of the ESs were used to create the points. Spearman’s rank correlation analysis, implemented in R (version 4.0.3), was employed to identify trade-offs and synergies between each ES pair [44]. Positive correlation coefficients indicate synergistic relationships, which means that the provision of one service increases in parallel with the provision of another service. Negative correlation coefficients indicate trade-offs, meaning that the provision of one service increases while the provision of another decreases. To compensate for the inability of traditional correlation analysis to clarify the spatial heterogeneity among ES relationships and visualize the spatio-temporal distribution of ES relationships, we applied bivariate spatial autocorrelation analysis with the GeoDa software (1.14.0) [45]. The local correlations among ESs were expressed using bivariate local spatial correlation index (LISA) clustering plots [46]. Four types of clustering, HH (High–High), LL (Low–Low), HL (High–Low), and LH (Low–High), were defined in the LISA plots as four types of trade-offs and synergisms with different degrees; i.e., HH—strong synergism, HL—weak synergism, LH—weak trade-offs, and LL—strong trade-offs [47].

Following the quantitative method [48], we determined the overall synergy and trade-off indices among multiple ES. With normalized ES indicators, the ES benefit index was calculated utilizing Equation (1). In addition, the root mean square error (RMSE) statistical parameter was utilized to quantify the degree of trade-offs among multiple ESs. Specifically, RMSE reflects the variable rates of unidirectional change across different objectives [49], thus indicating asymmetries in simultaneously changing ESs and broadening the concept of trade-offs. RMSE was calculated utilizing Equation (2):

$${ES}_{std}={\displaystyle \frac{{ES}_{obs}-{ES}_{Min}}{{ES}_{Max}-{ES}_{Min}}}$$

(1)

$$RSME=\sqrt{{\displaystyle \frac{1}{n-1}}·{\sum }_{i=1}^n{({ES}_i-\overline{ES})}^2}$$

(2)

where ES_std represents a standardized value describing the magnitude of benefits for each ES_obs, denoting the observed value for each ES; ES_Max and ES_Min, respectively, express the maximum and minimum observed values for each ES; and ES_i stands for the standardized value of ES_i, which is also the expected value of ES_i.

2.3.4. Threshold Effect Analysis

This study aimed to identify the primary natural and social drivers influencing the provision of ecosystem services [50,51,52]. Based on existing research, 11 typical potential driver variables were selected and redundant analysis (RDA) was further used to screen out the key factors significantly influencing the provision of ecosystem services, providing data support for the subsequent threshold analysis. Natural factors include elevation (ELE), slope (SLO), annual precipitation (PRE), annual average temperature (TEM), normalized difference vegetation index (NDVI), and soil moisture (SM). These variables serve as the background conditions for ecosystem service provision, reflecting the spatial heterogeneity of the regional environment and constituting an important natural foundation for ecosystem service provision. Socio-economic factors include gross domestic product (GDP), population density (POP), human footprint index (HFI), distance from railroads (RAIL), and land use intensity (LUI). These variables primarily reflect the direct interference from or impacts of human activities and land use on ecosystem service provision. Among these variables, land use intensity (LUI) is used to quantify the development level and intensity of land use, serving as a tool to analyze the specific impacts of land use on ecosystem service provision. Based on relevant research, the LUI grading indicators for unutilized land, ecological land (forests, grasslands, water), farmland, and construction land are set as 1, 2, 3, and 4, respectively [53]. The LUI is calculated by combining the area shares of each category in the region:

$$L={\sum }_{i=1}^n{R}_i\times {\displaystyle \frac{A_i}{A_t}}$$

where L is the land use intensity, n is the number of land cover types, R_i is the utilization intensity factor of the ith land cover type, A_i is the area of the ith land cover type, and A_t is the total area of all land cover types.

Restricted Cubic Splines (RCSs) are a piecewise polynomial regression method employed to model nonlinear associations between independent and dependent variables. Depending on the placement and quantity of knots, RCS applies spline functions to transform independent variables, illustrating the relationship between dependent and independent variables as continuous curves [54]. The restricted cubic spline function is expressed using Equation (3):

$$f\left(x\right)=\left({\beta }_1{x}_i+{\beta }_{2}f\left(Z_i\right)\right)={\beta }_{0}x+{\sum }_{k=1}^{K-2}{\beta }_k{x}_k^3$$

(3)

If β₀ = β₁ = ⋯ = β_K−2 = 0, there is no correlation between independent and dependent variables; if β₀ ≠ 0 and β₁ = ⋯ = β_K−2 = 0, there is a linear correlation; and if β₀ ≠ 0 and βi ≠ 0, i > 0, there is a nonlinear correlation.

In threshold analysis, RCS facilitates the identification of critical data knots through smooth curve fitting to indicate significant transitions in ecosystem service functions [55]. The number of knots in the spline function has a large impact on the fitting effect of RCS, which will directly determine the shape of the fitted curve. To account for the model’s degree of freedom and prevent overfitting as much as possible, we screened the RCS containing 3–5 knots based on the Akaike information criterion (AIC) [56]. The AIC is a standard that comprehensively considers model complexity and goodness of fit, with lower values indicating better models. During model selection, we compared the AIC values for different numbers of nodes and selected the model with the lowest AIC to ensure the best fit, with the selection results detailed in Tables S1–S3 in Supplementary Materials. The positions of the knots were initially searched within quantile ranges (e.g., 5–95%), followed by precise threshold localization using the method of maximizing the logLik function. By observing the trends of RCS curves for possible nonlinear relationships, the threshold was further explored using segmented linear regression, in which jumps occur at the positions of the knots, which can be equivalent to the threshold in a certain sense [55,57]. To evaluate threshold fitting uncertainty, we tested for potential nonlinearity by comparing models incorporating only linear terms against models containing both linear and cubic spline terms utilizing such likelihood ratio tests. The RCS regression was conducted utilizing the ggrcs package in R (4.3.3) [58].

3. Results

3.1. Spatial–Temporal Changes in ES

Between 2000 and 2020, the five ES categories exhibited varied temporal trends across the three time points (2000, 2010, and 2020). Provisioning services (GP) and regulating services (WY, SR) presented upward trends, while supporting services (CS, HQ) presented minor declines. GP experienced a significant increase (49.9%), with WY and SR services rising by 27.12% and 15.14%, respectively, whereas CS and HQ decreased by 3.1% and 2.6%, respectively (Figure 3).

Regarding spatial distribution, high-value areas for GP experienced a process of “expansion–adjustment.” Between 2000 and 2010, these areas extended from the northwestern and central regions to the northern and southern regions. From 2010 to 2020, this expansion continued northwestward, coupled with a simultaneous contraction in the central region. WY and SR exhibited a spatial gradient, with higher values in the south and lower values in the north. CS and HQ demonstrated a strong spatial correlation, with high-value areas primarily located in the west-central and southern regions, following a zonal distribution.

3.2. Trade-Offs and Synergies Among ESs

This study analyzed the trade-offs and synergies among five ES categories in the Kashgar region across three time points (2000, 2010, and 2020). The Spearman correlation coefficients (

Table 3. Correlations between paired ESs from 2000 to 2020 (p < 0.01).

		GP	SR	WY	CS	HQ
2000	GP	1.0	0.52 *	0.23 *	0.34 *	0.61 *
	SR	——	1.0	0.22 *	0.12 *	0.31 *
	WY	——	——	1.0	−0.09 *	0.13 *
	CS	——	——	——	1.0	0.31 *
	HQ	——	——	——	——	1.0
2010	GP	1.0	0.48 *	0.41 *	0.38 *	−0.12 *
	SR	——	1.0	0.5 *	0.59 *	0.35 *
	WY	——	——	1.0	0.35 *	0.3 *
	CS	——	——	——	1.0	0.31 *
	HQ	——	——	——	——	1.0
2020	GP	1.0	0.38 *	0.56 *	0.71 *	0.6 *
	SR	——	1.0	−0.03	0.2 *	0.35 *
	WY	——	——	1.0	0.49 *	0.32 *
	CS	——	——	——	1.0	0.74 *
	HQ	——	——	——	——	1.0

GP: grain production; SR: soil retention; WY: water yield; CS: carbon storage; HQ: habitat quality; *: p < 0.01.

) indicated ten interaction pairs among the five ESs, with most exhibiting significant positive correlations. However, the CS-WY, HQ-GP, and WY-SR pairs demonstrated significant negative correlations. Generally, these relationships appeared primarily as weak trade-offs and moderate synergies. The strength of synergistic relationships, while variable over time, remained broadly stable. The relationship between CS and WY transitioned from a weak trade-off (correlation coefficient −0.09) to a strong synergy (correlation coefficient 0.49), while the relationship between WY and SR changed from moderate synergy (correlation coefficient 0.22) to a weak trade-off (correlation coefficient −0.03).

The spatial distribution of trade-offs and synergies between ESs was analyzed using GeoDa (Figure 4), revealing the spatial variability among the ESs. In terms of spatial distribution, the oasis core area mainly showed strong synergies (High–High), while the desert edge area was dominated by low synergies (Low–Low). With the passage of time, the synergistic effect of the ESs was further expanded, and the Low–Low distribution in the desert edge region revealed a shrinking trend. The spatial distribution patterns of HQ–CS, HQ–WY, and CS–WY remained consistent throughout the study period, in contrast to the significant spatial variability exhibited between GP–CS and GP–SR.

Between 2000 and 2020, the overall ES synergy index in the Kashgar region presented a general increase (Figure 5a). Areas with high index values extended northward from the southern and central portions of the region, developing from a scattered distribution into a more contiguous pattern. By 2020, high values were mainly located in the western, central, and southern areas, while low values were clustered along the northeastern desert edges. The ES trade-off index instead indicated a pattern of localized growth. Areas with high trade-off values, which were sparsely distributed in 2000, expanded considerably into the central and southern periphery by 2010. By 2020, high trade-off values were largely concentrated in the southern periphery and northwestern portions of the region. Areas exhibiting low trade-off values progressively decreased, with changes primarily concentrated in the southwestern region (Figure 5b).

These results indicate that while regional ES synergy generally improved, trade-off relationships intensified in certain locations, especially in the southern area. The increasing conflicts among ecosystem services and greater spatial heterogeneity in this southern region suggest difficulties in effectively managing regional ESs.

3.3. Threshold Effects Analysis of Ecosystem Services

3.3.1. Identification of Driving Factors

To determine the principal drivers of ES change, this study utilized RDA to appraise the relationships between ESs and explanatory variables (Figure 6). RDA results indicated statistically significant associations between ESs and explanatory factors (2000: R² = 67.11%, p < 0.001; 2010: adjusted R² = 77.43%, p < 0.001; 2020: adjusted R² = 76.67%, p < 0.001). Bar charts illustrated the differential contributions of explanatory variables over time, indicating a significantly greater influence from natural factors compared to social factors on ESs in the Kashgar region.

In 2000, soil moisture (SM) had the highest contribution among natural factors (35.1%), while the human footprint index (HFI) had the highest contribution among social factors (9.4%). In 2010, NDVI became the most influential factor affecting ESs, with the contribution of precipitation (PRE) increasing and that of soil moisture (SM) gradually declining. By 2020, land use intensity (LUI) and transportation accessibility (RAIL) reached their highest levels of contribution. The influence of temperature (TEM) gradually decreased, while the effects of topographic factors (elevation and slope) fluctuated. The impacts of population density (POP) and per capita GDP peaked in 2010 but gradually diminished thereafter. Based on the RDA results, given that GDP’s explanatory power for ES changes consistently remained below 1%, subsequent analyses excluded the GDP variable to focus on the influence of key driving factors.

3.3.2. Threshold Characteristics of Natural Factors

(1)

Topographic Factors

The distribution of slope thresholds across ESs was largely concentrated between 15° and 30°, with WY presenting the highest threshold value (Figure 7a). Three unique patterns defined the responses of different ESs to changes in slope. GP and CS followed similar trends, increasing with slope after surpassing their respective thresholds. When the slope range was between 0 and 30°, WY increased rapidly from a minimum value to a maximum value and then gradually decreased. SR and HQ exhibited minimal responsiveness to changes in slope, indicating gradual response curves indicative of greater stability in the effects of slope on these services. The slope affected the overall effects of ESs, and the synergy and trade-off indices changed over time. The synergy index displayed a threshold effect (30.2°) only in 2000, while the trade-off index demonstrated a threshold effect (20.3°) in 2020.

Threshold relationships between elevation and various ESs exhibited unique gradient effects, with threshold points decreasing in the sequence of WY > CS > HQ. Each ES presented a unique response profile to changes in elevation. The RCS curve of WY versus elevation showed a “U” shape, with a threshold effect between 2117 and 2359 m. Both CS and HQ demonstrated inverted “U-shaped” relationships with elevation, indicating that mid-elevation zones are more favorable for the delivery of these services. SR increased monotonically with elevation, signifying a strong capacity for adaptation in high-altitude environments (Figure 7b). Considering trade-off and synergy relationships, the threshold between the synergy index (3641 m) and elevation was typically greater than that of the trade-off index (2045 m), suggesting that mid- to high-elevation zones may cultivate the synergistic delivery of ecosystem services.

(2)

Climatic Factors

Different ESs responded to temperature with varying threshold effects and trends. CS and HQ demonstrated similar threshold effects, declining as temperatures rose above their respective thresholds. WY exhibited a “U-shaped” relationship with temperature, with the threshold point reaching a maximum of 9.5 °C in 2010 (Figure 8a). The trade-off index showed a threshold effect when the temperature was in the range of 1.5–2.7 °C. The RCS curve of precipitation for ESs showed an opposite trend to that of temperature. The threshold point of ESs reached its highest in 2010, which might be related to the increase in precipitation during that year. For the time-series characteristics of the combined effects of precipitation on ESs (Figure 8b), the threshold effect appeared in both trade-off and synergistic indices only in 2010, while the trade-off index showed a downward trend after the precipitation exceeded 488 mm. The synergistic index instead showed a positive effect, highlighted by the precipitation exceeding 228 mm.

(3)

Vegetation Coverage–Soil Factors

The relationship between NDVI and ESs showed the following characteristics in different periods. The threshold points increased year by year, from 0.05 to 0.08 in 2000 to 0.30–0.35 in 2020. GP showed an increasing trend with an increase in NDVI, while WY transformed from a negative to positive correlation at the threshold and when NDVI exceeded 0.35, which adversely affected HQ and CS (Figure 9a). The differences between the threshold points of NDVI and the trade-off and synergy indices were relatively small (about 0.13–0.23). The threshold effect manifested as a gradual weakening of the trade-off index in the high NDVI region, while the synergy index showed an upward trend in 2020 as NDVI exceeded 0.23.

All four ESs, except for SR, exhibited nonlinear response curves with soil moisture (SM), characterized by U-shaped or V-shaped patterns. Among them, GP, CS, and HQ showed similar trends, with threshold points and RCS curve slopes exhibiting consistent changes, first increasing and then decreasing. WY consistently displayed threshold points at higher soil moisture levels across the three time points (e.g., 140 in 2000 and 120 in 2020), indicating its strong dependence on high soil moisture conditions. The synergy and trade-off indices showed notable symmetrical U-shaped patterns at all three time points. As SM increased, both the synergy and trade-off indices initially decreased and then increased, with greater uncertainty observed under extreme low or high soil moisture conditions (Figure 9b).

3.3.3. Threshold Characteristics of Social Factors

HFI demonstrated two patterns in its effects on individual ESs (Figure 10a): (1) a threshold response wherein WY increased, and HQ decreased after their respective thresholds, with these thresholds varying over time, and (2) a monotonic response where both GP and CS increased continuously alongside HFI, with GP presenting greater sensitivity. The threshold points of the trade-off index show fluctuating changes over time, while RCS curve shows an inverted U-shaped change. Notably, the effects of HFI on the synergy index were more moderate.

POP displayed significant threshold effects and temporal dynamics across different ES. For GP, thresholds decreased over time, suggesting a lower critical population density necessary for GP. WY initially exhibited increasing threshold effects with POP, which subsequently weakened. POP’s positive impact on soil retention (SR) peaked at 6.7 people/km², while this threshold effect gradually lessened over time (Figure 10b). The threshold effect of POP on the trade-off indices progressively weakened and eventually disappeared, with thresholds varying between 12.5 and 15 people/km². POP exerted a relatively weak influence on synergy indices, indicating negative effects in 2020.

Beyond 44 km, railways negatively affected CS and HQ while continuing to positively affect GP. Threshold patterns between railway distance and both trade-off and synergy indices indicated dynamic temporal behavior. In 2010, the threshold between railroad distance and the trade-off index declined significantly from 2000 and showed no threshold effect by 2020, with a monotonically decreasing relationship between the synergy index and railroad distance (Figure 11a). WY and LUI represent an inverted U-shape change, with a threshold point between 1.15 and 2.13. Additionally, we observed a trend of declining land use after exceeding the threshold point in the high-intensity land use area. HQ and CS reveal a similar trend of curve change, except for 2010, where the threshold point was between 2.7 and 3. After exceeding this range, the constraints of LUI on ESs strengthened. The RCS curve of 2010 shows a U-shape change, with the threshold point reaching its lowest (1.04). Then, the slope of the RCS curve for the trade-off index gradually flattens out, with the threshold point showing a significant decrease over time. The synergistic index increased significantly in the low- and medium-intensity regions but leveled off in the high-intensity region (Figure 11b).

4. Discussion

4.1. Spatial–Temporal Evolution Characteristics and Driving Mechanisms of ESs in the Kashgar Region

Our study identified significant spatio-temporal variations in different ESs across the Kashgar region from 2000 to 2020. These temporal patterns are strongly correlated with the drivers determined through RDA.

The significant growth observed in GP is likely closely related to advancements in agricultural technology and the expansion of cultivated land, which is consistent with the findings of Zhang [11]. The improvements in WY and SR indicate that the enhancement of vegetation cover in the Kashgar region has played a significant role in preventing soil erosion and conserving water resources [59]. The RDA findings suggest that natural factors exerted a stronger influence on ESs than social factors in the Kashgar region, with the NDVI representing primary factors. However, the declines in CS and HQ suggest that excessive land development and climate warming may pose threats to the stability of ecosystem services [21,59].

Spatially, WY and SR exhibited clear north–south gradients, with higher values concentrated in the south. This spatial heterogeneity is largely attributed to topographical and climatic factors [60]. The mountainous terrain and piedmont oases of southern Kashgar creates advantageous conditions for enhanced WY and SR, whereas the northern plains and deserts, marked by an arid climate and high soil erosion risk, exhibit comparatively weaker ecological functions [21]. CS and HQ shared similar spatial distribution patterns, reflecting their correlation (0.74 in 2020) and suggesting coupled relationships. High-value zones for both were primarily located in the west-central and southern areas, forming a belt-like distribution. High-value CS areas demonstrated notable spatial shifts, primarily clustered in central regions, while high-value HQ areas were largely confined to the southern region. This spatial arrangement corresponded with the distribution of the NDVI. The explanatory power of LUI among social factors for ESs reached its peak in 2020. This result indicates that human activities influence the spatial distribution of ESs by altering land use patterns [52]. This effect is particularly significant in the northern region where intensive human activity led to reduced overall ESs and unique patterns of human-induced ecological changes. In contrast, the southern region sustained higher ES levels due to its favorable environmental conditions and comparatively minimal human intervention.

The spatio-temporal development of ESs in the Kashgar region can be described by three key processes: (1) dual influences, where natural factors shape spatial patterns and human activities, particularly land use changes, that drive temporal shifts; (2) regional differentiations, with southern areas exhibiting high ES capacity due to favorable environmental conditions and northern areas experiencing lower ES levels due to anthropogenic pressures; and (3) interconnected effects, where interactions between natural and social factors produce unique spatio-temporal patterns.

4.2. Dynamic Evolution Mechanisms of Ecosystem Service Trade-Offs and Synergies

Our analysis indicated significant spatio-temporal heterogeneity in ES trade-offs and synergies in the Kashgar region. The correlation analysis indicated that from 2000 to 2020, the synergistic relationships between ESs in Kashgar region strengthened as a whole, and the trade-off relationship intensified locally. In particular, the relationship between CS and WY gradually changed from a weak trade-off (correlation coefficient −0.09 in 2000) to a strong synergistic relationship (correlation coefficient 0.49 in 2020). This positive trend is primarily attributed to the implementation of human intervention measures. As a key region for ecological engineering and policy implementation, Xinjiang has seen substantial investments from local governments in large-scale ecological restoration projects such as the “Three-North Shelterbelt Program,” desert greening initiatives, and the Taklamakan Desert control project [61]. Additionally, ecological water diversion and river channel management in the lower reaches of the Tarim River have enhanced water resource availability and improved ecological resilience [62]. However, these ecological projects have also introduced some potential challenges, particularly concerning water resource balance and ecosystem stability [63,64]. In contrast, the WY–SR relationship changed from moderate synergy (correlation coefficient 0.22 in 2000) to a weak trade-off relationship (correlation coefficient −0.03 in 2020), demonstrating the dynamic nature of ES interactions. GP’s synergistic relationships with other ESs generally strengthened, reaching a correlation coefficient of 0.71 with CS in 2020, suggesting improved alignment between agricultural production and ecological conservation [65].

Spatially, the overall ES synergy index generally increased. Areas with high synergy values, initially concentrated in southern regions in 2000, spread northward and westward. The analysis of trade-off indices indicated stronger overall ES synergies across the region, yet trade-offs were amplified in certain areas, notably in the south, indicating that these regions were subjected to a combination of higher anthropogenic intensity and pressure on natural resources [10]. This observation corresponds to the threshold analysis results. WY and LUI exhibited an inverted U-shaped relationship, where WY declined as LUI increased beyond the threshold point. Climatic factors shape ES relationships through nonlinear responses. Temperature and precipitation, for instance, exhibit threshold effects on ES relationships. When temperature surpasses certain thresholds, WY tends to increase, while HQ and CS tend to decrease. Changes in precipitation prompt nonlinear responses in ecosystem functions, improving WY and SR but potentially creating resource supply and demand imbalances in certain locations. These varying responses amplify trade-off relationships among ESs.

The mechanisms governing ES trade-offs and synergies in the Kashgar region, according to the preceding analysis, can be described by three key factors: (1) the threshold effect of land use intensity, whereby the increase in land development intensity leads to an imbalance in the supply and demand of local ES resources; (2) varying climatic impacts, i.e., the threshold effects tied to temperature result in divergent ES responses, while precipitation changes induce nonlinear responses in ecosystem functions; (3) the effects of human activities, wherein advancements in coordination between agricultural production and ecological conservation demonstrate the positive effects of sound management, although concentrated human activity in certain areas can exacerbate ES trade-offs, emphasizing the need for regional management strategies tailored to local conditions.

4.3. Threshold Effects of Driving Factors and Management Implications

Analysis of the RCS curves demonstrated nonlinear relationships between natural and social driving factors and ESs. Threshold effects were particularly significant for natural factors. The trend in the NDVI—the most influential driver, with increasing threshold points over time—suggests that improved vegetation cover is gradually becoming more supportive of ESs. In 2020, when NDVI surpassed 0.35, HQ and CS were negatively affected. This result indicates that high vegetation cover may exacerbate the conflict between ESs [66]. Based on the results of trade-offs and synergies in the region, moderate vegetation cover may optimize resource use efficiency, while excessive vegetation may exacerbate competition for resources and reduce the equilibrium of the system. The threshold effects of soil moisture on GP, CS, and HQ exhibited “U-shaped” or “inverted U-shaped” characteristics. This result indicates that the impact of soil moisture on ESs is dual in nature. Moderate moisture promotes ecological functioning, while extreme moisture conditions may increase uncertainty in service provision. The dependency of WY on high soil moisture is particularly significant and remains consistent with the characteristic reliance on oasis irrigation for supplying water resources in the Kashgar region [63].

Temperature and precipitation drive changes in ESs by altering resource allocation and ecological processes with threshold effects and dynamic responses, respectively [67]. The positive effects of temperature on CS, HQ, and GP weakened after the threshold due to increased resource depletion, while a U-shaped dependence on WY was observed through evapotranspiration. The trade-off index showed a threshold effect at temperatures between 1.5 and 2.7 °C. This result indicates that under low temperature conditions, certain ES types tend while others decrease. The effect of different precipitation amounts on CS, HQ, GP, and WY were also significantly differentiated. When precipitation was below 228 mm, limited water resources suppressed CS, HQ, and GP. However, when precipitation exceeded 488 mm, WY increased significantly, and the state of excess water supply mitigated trade-offs between ecosystem services. However, this increase may also have weakened the additional contributions to CS and HQ. According to the threshold features of NDVI, vegetation restoration should be moderate and avoid over-intensity. Meanwhile, considering the threshold effects of climatic factors, for the arid region in the north, water-saving irrigation techniques and drought-tolerant planting strategies should be prioritized. For the oasis region in the south, vegetation restoration and an optimal allocation of water resources should instead be combined to enhance the supply capacity of ESs.

Topographic factors significantly modulated the spatial patterns, trade-offs, and synergistic relationships of ESs. The distribution of thresholds between the slope and ESs was mainly concentrated between 15° and 30°. Here, the sensitivity of WY to slope was the highest, indicating that WY responded more strongly to changes in topography. This difference provided a theoretical basis for the optimal management of different ESs over the slope gradient. In addition, the temporal variability of the synergy and trade-off indices may be related to land use changes and the implementation of ecological conservation measures, especially the significant improvements in ecosystem trade-off relationships over the last 20 years, which is further evidence of the effectiveness of environmental management [68]. The threshold effect of elevation on ESs showed a gradient distribution. This differential response may originate from the effects of soil depth and vegetation growth conditions on ESs [69]. Meanwhile, considering the threshold effect of climatic factors, for the northern arid region, water-saving irrigation techniques and drought-tolerant planting strategies should be prioritized. However, for the oasis region, the integrated ES supply capacity should be enhanced combined with vegetation restoration and the optimal allocation of water resources [70].

In terms of social factors, the threshold effects of HFI and ESs reflect the bidirectional regulation mechanism of HFI on WY and HQ; i.e., moderate HFI can enhance WY but exacerbate the decline of HQ. Meanwhile, the inverted U-shaped curve of HFI on the trade-off index suggests that human activities nonlinearly regulate the trade-off and synergistic relationships among ESs at different intensities. The effect of POP on ESs further demonstrates significant temporal dynamics. The gradually decreasing threshold point of GP on POP may be related to the expansion of agricultural land and technological advances induced by increasing population density [65]. On the other hand, the positive effects of WY and SR on POP indicate the potential promotion of water supply and biodiversity conservation via moderate population density. However, excessively high POP can lead to increased competition for resources, weakening its positive effects.

The results showed that CS and HQ have negative effects when the distance to the railroad exceeds 44 km. It is, therefore, recommended that ecological buffer zones be established along the railroad and that vegetation restoration and soil protection projects be implemented to mitigate the potential negative impacts of the railroad on the quality of the habitat. At the same time, ecological corridor design can be combined to enhance the ecological connectivity of areas along the railroad. The inverted U-shaped response of WY to LUI and the significant enhancement of CS and HQ in the low- and medium-intensity areas indicate that moderate land use intensity has a positive effect on ecosystem services. For regions with high land use development intensity, it is recommended to improve the stability and resilience of regional ecosystems through green infrastructure (e.g., ecological wetlands and biological corridors) and ecological compensation mechanisms.

4.4. Limitations

In this study, the nonlinear threshold effects of natural and social factors were revealed using the RCS method. The results not only revealed the dynamic response mechanism of ecosystem services in Kashgar, but also provided a scientific basis for ecological management in arid zones. In particular, the hierarchical regulation strategy based on trade-offs and synergies proposed by this study provides new ideas for complex ecosystem management. Nonetheless, several limitations remain.

First, in the selection of ecosystem services, this study included only five types (GP WY, SR, CS, and HQ), neglecting other ecosystem functionalities. Specifically, bio-cultural and recreational services, while crucial for sustainable regional development, were excluded due to data constraints and model capabilities. Second, the selection of influential factors primarily relied on readily measurable indicators identified in prior research. Therefore, certain significant yet challenging-to-measure factors may have been omitted. Future analyses should integrate factors such as climate change, land use transformations, and policy adjustments to thoroughly appraise their effects on regional ecosystem services and their interconnections by introducing a long series of policy data combined with machine-learning models (e.g., Random Forest or XGBoost). Finally, with respect to the analysis of the spatial and temporal evolution of the threshold effect, the time span of this study (2000–2020) was relatively short and could not be updated to 2023 due to the lack of hydrological validation data. Simultaneously, although the RCS method effectively captures nonlinear features, a combination of this approach with other models should be explored in the future. Considering these limitations, future research avenues could include the following:

(1)

Improve the accuracy and breadth of the time-series analysis by utilizing newer satellite datasets (such as Sentinel) and ecological observation data from public platforms, further enhancing the timeliness and reliability of the research results.

(2)

Use proxy variables (such as geographic social media data and public behavior data) or participatory mapping methods to assess the spatial distribution of cultural and entertainment services and their ecological value, comprehensively revealing the multidimensional characteristics of ecosystem services.

(3)

Conduct more in-depth analyses into the effects of climate change and land use changes and analyze their implications for exploring regional ecological security zones.

(4)

Implement comprehensive monitoring systems and adaptive regulatory models to accurately target regional ecological concerns and cultivate the coordinated development of economic and ecological protection.

5. Conclusions

This study, based on measuring key ecosystem services and their trade-offs/synergies in the Kashgar region, developed a composite ecosystem service index and analyzed the key factors and critical thresholds influencing its changes. The principal findings are summarized below:

(1)

From 2000 to 2020, ESs in the Kashgar region exhibited significant spatio-temporal variations. GP experienced the largest growth (49.9%), followed by WY and SR (increasing by 27.12% and 15.14%, respectively). Conversely, CS and HQ declined slightly (by 3.1% and 2.6%, respectively). Spatially, WY and SR displayed a north–south gradient, with higher values in the south. High-value areas for CS and HQ were primarily located in the central-western and southern parts of the region. Natural factors dominated the formation of spatial patterns, while human activities significantly influenced temporal trends through changes in land use patterns.

(2)

Trade-offs and synergies among regional ecosystem services demonstrated significant spatio-temporal heterogeneity. In 2020, the strongest synergistic relationships were identified between CS and HQ (0.74) and between GP and CS (0.71). While the overall regional ecosystem service synergy index rose annually, trade-off relationships intensified in the southern region. Specifically, the relationship between CS and WY transitioned from a weak trade-off to strong synergy, indicating the positive impact of ecological restoration projects on improving synergies. However, the relationship between WY and SR changed from moderate synergy to a weak trade-off, reflecting the disruptive effects of increased land use intensity on the balance of water and soil resource supply.

(3)

The RDA showed that natural factors are the primary driving forces influenced ESs, while social factors significantly regulate the dynamic changes in ESs by altering resource distribution and management patterns. The RCS analysis indicated several critical thresholds. Exceeding an NDVI of 0.35 negatively affected HQ and CS, and slope thresholds for individual ESs generally fell between 15° and 30°. Excessive increases in land use intensity may exacerbate resource conflicts, whereas moderate development and optimized allocation could enhance the coordination of service provision.

(4)

Given the trade-offs, synergies, and threshold differences in ESs, regional adaptive strategies should be developed to optimize and coordinate ecological functions. In low-precipitation areas with annual rainfall below 228 mm, water-saving irrigation technologies and drought-tolerant plant cultivation should be promoted to enhance WY and SR capacities, thereby alleviating the pressures on ecosystem service provision caused by resource scarcity. In areas with high vegetation coverage (NDVI > 0.35), the scale of vegetation restoration should be appropriately controlled. Vegetation structure should, moreover, be optimized through natural restoration, thinning, and ecological corridor construction to avoid excessive vegetation density leading to intensified resource competition, thereby reducing the trade-off between CS and HQ. For areas with slopes less than 15°, the focus should be on optimizing resource utilization efficiency to reduce the negative impacts of agricultural activities on CS and HQ. Simultaneously, promoting ecological agriculture models and the resourceful utilization of agricultural waste could enhance soil quality and carbon storage, allowing one to establish a comprehensive regional adaptive management system that integrates agricultural production with ecological conservation.