A Study on the Spatiotemporal Heterogeneity and Driving Forces of Ecological Resilience in the Economic Belt on the Northern Slope of the Tianshan Mountains

Abstract

The assessment of ecological resilience in arid regions is crucial for understanding and mitigating the impacts of climate change and human activities, ensuring the sustainable management of these vulnerable ecosystems. Taking the Economic Belt on the Northern Slope of the Tianshan Mountains (EBNSTM) as the research area, a multi-dimensional evaluation model coupling vulnerability, health, and connectivity was used to explore the spatiotemporal variation and driving forces of ecological resilience. Firstly, a sub-item evaluation of ecological resilience was conducted from three aspects, including ecological vulnerability evaluation based on the CRITIC and AHP models, ecological health evaluation based on the InVEST model, and landscape connectivity evaluation based on the MSPA method. Then, the sequence polygon method was utilized to conduct a comprehensive multi-dimensional assessment of ecological resilience based on the aforementioned three evaluation results. Finally, the geographical detector model was utilized to identify the driving factors behind the spatial heterogeneity of ecological resilience. The results show the following: (1) From 2000 to 2020, the overall ecological resilience showed an upward trend and significant spatial heterogeneity. The overall distribution pattern exhibited a spatial feature of south higher, north lower, where the southern region displayed a clear high-high clustering characteristic, exerting a positive and radiating influence on surrounding areas. (2) The main driving factors of the spatial heterogeneity are DEM, precipitation, NPP, GDP, and PM2.5. And among different factors, the dual-factor enhancement effect is greater than the nonlinear enhancement of a single factor. (3) Human activities are important influencing factor, and the impact of urban expansion and economic growth on ecological resilience is becoming increasingly significant. Therefore, in the process of economic development, full consideration should be given to the self-repairing and adaptive capabilities of the ecosystem.

1. Introduction

Amid escalating global environmental changes and the growing frequency of human activities, ecological issues have increasingly become critical concerns for ecological security [1]. Climate change, in particular, has emerged as one of the most urgent global challenges, including extreme weather events, shifts in climate patterns, and widespread environmental degradation profoundly impacting the functionality and stability of ecosystems [2]. In 1973, the concept of resilience was initially introduced to ecology by the ecologist Holling. He explained resilience as the stable state of an ecosystem, and the central idea revolves around the ecosystem’s capacity to strive towards maintaining its original equilibrium in the midst of diverse impacts and disturbances [3]. In 2015, Boschma gradually shifted their research emphasis to resilience pertaining to the long-term dynamic progression of systems, with a growing focus on its application within the ecological environment. Assessing ecological resilience requires considering multiple levels, including society, the ecosystem, and the economy [4].

At present, ecological resilience, as a focal point of ecological research, has garnered considerable attention and made certain progress. The theoretical system and methods of ecological resilience research are also relatively mature. Aliakbart employed the loss-benefit approach to assess ecological resilience [5]. Wu evaluated ecological network resilience based on complex network theory [6]. In addition, there are various methods such as RSEI [7], entropy weight method [8], E-PSR modeling [9], and resilience assessment from a landscape perspective [10]. But there is still no unified research method or an evaluation index for ecological resilience at home and abroad, and there is a lack of a multi-dimensional evaluation system. From the perspective of evaluation indicators, there has been a gradual shift from single to multi-dimensional. For example, Zhang et al. established a coastal composite ecological resilience evaluation system [10], and Liu constructed a multi-dimensional urban resilience evaluation index system encompassing four dimensions: economy, infrastructure, society, and ecology [11]. Vanessa developed a decision support model for national resilience evaluation in SES, and Lee constructed an ecological resilience framework grounded in the resistance adaptation recovery theory framework [12,13]. Multi-dimensional indicator evaluation has made certain progress. However, comprehensive research on multi-dimensional evaluation methods is still insufficient, and there is a lack of exploration of the dual-factor interaction driving the effects of ecological resilience [14,15].

EBNSTM is one of the most populous and urbanized regions in Xinjiang, having advanced social and economic development. As a solid ecological barrier in the arid region of northwest China, due to its arid characteristics, it also faces challenges such as a fragile ecological environment, limited environmental carrying capacity, and weak environmental resilience, making it highly vulnerable to interference and impacts from external environmental factors [16]. For resource-based cities such as Changji, it is crucial to scientifically evaluate their transformation capabilities and development potential [17]. With the rapid development of industry and economy and the acceleration of urbanization, the fragile ecological environment has been further impacted, which undoubtedly has caused considerable obstacles to sustainable development under the “Belt and Road” initiative [18]. Studying its ecological resilience from a multi-dimensional perspective and scientifically evaluating the spatiotemporal variations and underlying factors of ecological resilience holds significant importance for ecological safeguarding and promoting long-term development within the area.

2. Materials and Methods

2.1. Study Area

EBNSTM is the most developed economic area in Xinjiang, situated between 84°46′E to 88°58′ E and 42°45′ N to 46°8′ N (Figure 1). It occupies 5.7% of the region’s total land but houses 23.3% of the population. The area produces 83% of Xinjiang’s heavy industry and 62% of its light industry, contributing to over 40% of the region’s GDP annually. As an important part of the “Silk Road Economic Belt”, the formation and development of the EBNSTM are closely related to the natural environment and geographical characteristics of the Tianshan Mountains. With its unique geographical and energy advantages, it has shown enormous development potential. However, the climate is arid, having relatively scarce yearly rainfall and notable seasonal differences. Therefore, it also confronts many challenges, including water scarcity, ecological fragility, and low resource and environmental carrying capacity [1].

2.2. Conceptual Framework

Synthesizing previous studies, we recognize that when ecosystems confront external climate changes and human activities, they are perturbed and respond accordingly. As depicted in Figure 2, the ecosystem’s state shifts from A to B upon disturbance and then begins to recover, transitioning from B to C over time. For regions with high resilience, the ecosystem may even undergo further renewal, achieving a new stable state, D.

Given the varying degrees of resilience among different ecosystems, their abilities to respond to disturbances, recover, and even renew manifest in distinct rates. According to the self-adaptive organization theory (Figure 3), when an ecosystem is subjected to external disturbances, it undergoes four stages. Specifically, an ecosystem is always in a certain state of stability, with varying levels of resistance to disturbances. When disturbances accumulate to a certain threshold, the system experiences a collapse and release, disrupting its steady state. However, over time, different ecosystems exhibit varying degrees of resilience and may return to their initial state or even undergo reorganization and update to achieve a new, higher level of stability. In this study, we select vulnerability, connectivity, and health as proxies for an ecosystem’s capacity for response, recovery, and self-renewal, respectively. These factors are intricately intertwined and mutually influential, yet they all play pivotal roles in the ecological environment and system. Consequently, we have constructed a novel resilience assessment framework aimed at comprehensively and systematically reflecting the state of ecosystems, providing a reference for regional development. Firstly, a sub-item evaluation of ecological resilience was conducted from three aspects, including ecological vulnerability evaluation based on the CRITIC (criteria importance through intercriteria correlation) and AHP (analytic hierarchy process) models, ecological health evaluation based on the InVEST (integrated valuation of ecosystem services and tradeoffs) model, and landscape connectivity evaluation based on the MSPA (morphological spatial pattern analysis) method.

2.3. Data Resources and Process

To build an effective evaluation system and better carry out the ecological resilience evaluation of the EBNSTM, relevant data have been collected and organized. Based on the basic situation of data collection, the final time scale determined in this study is four years: 2005, 2010, 2015, and 2020. The data mainly include terrain data, meteorological data, vegetation coverage data, social economic data, etc.

Table 1. Source, resolution, and format of data.

Data	Spatial Resolution	Time Resolution	Format	Sources
Vector boundary	N/A	N/A	Vector	National Tibetan Plateau Data Center (http://data.tpdc.ac.cn (accessed on 17 September 2023))
Meteorological station	N/A	N/A	Vector	Resources and Environmental Science Data Center (http://www.resdc.cn (accessed on 5 October 2023))
Precipitation	N/A	D	Grid
Wind speed	N/A	D	Grid
Land use	1km	Y	Grid
DEM	1km	N/A	Grid
NDVI (Normalized difference vegetation index), NPP (Net Primary Production)	1 km	N/A	Grid	Geographic Data Sharing Infrastructure, global resources data cloud (http://gis5g.com (accessed on 6 October 2023))
Real GDP (Gross Domestic Product) Industrial structure	N/A	Y	Excel	China Statistical Yearbook (https://www.stats.gov.cn (accessed on 6 October 2023))
PM2.5 (Particulate Matter 2.5)	1 km	D	Grid	https://zenodo.org/record/6398971 (accessed on 6 October 2023)
AOD (Aerosol Optical Depth)	1 km	N/A	Grid	http://modis.gsfc.nasa.gov (accessed on 6 October 2023)
Population density	1 km	N/A	Grid	World POP (https://www.worldpop.org (accessed on 6 October 2023))

shows the details.

2.4. Methods

2.4.1. Ecological Vulnerability Evaluation Based on CRITIC and AHP

Based on relevant research, this article selected 14 natural and socioeconomic indicators and employed a blend of subjective and objective assessment techniques (

Table 2. Ecological vulnerability evaluation indicator system based on CRITIC and AHP methods.

Criterion Layer	Indicator Layer	Attribution	CRITIC Weight	AHP Weight	Complex Weight
Terrain	Terrain index	−	0.033	5.452	2.743
	Degree of relief	−	6.228	5.452	5.840
Landscape	Landscape disturbance index	+	1.717	10.904	6.310
Meteorology	Temperature	−	1.372	4.673	3.023
	Precipitation	−	7.921	4.673	6.297
	Wind speed	+	1.636	1.558	1.597
Surface	NDVI	−	2.866	10.16	6.513
	NPP	−	12.56	20.32	16.440
Social economy	Industrial structure	+	6.339	2.990	4.665
	Population density	+	25.804	12.390	19.097
	Night lights	+	19.158	8.767	13.963
	Real GDP	+	10.797	6.344	8.571
Air	AOD	+	1.708	3.369	2.539
	PM2.5	+	1.86	2.938	2.399

The “+” represents a positive factor, and the “−” sign represents a negative factor.

). Comprehensive weight was calculated by adding the same weight using CRITIC and AHP methods [19]. AHP effectively captures decision-makers’ preferences, making it suitable for situations with clear, measurable preferences [20]. The CRITIC method, on the other hand, minimizes biases by relying on historical data and statistical analysis, offering an objective approach [21]. Combining the two allows for the refinement of AHP’s weights, leveraging both subjective preferences and improved accuracy.

The various influencing factors were overlaid and analyzed in ArcGIS 10.8 to obtain the vulnerability evaluation results.

$$EVI={\displaystyle \sum C_i{W}_j}$$

(1)

Among them, $C_i$ is the weight of each indicator, and $W_j$ is the standardized indicator, which is calculated using a grid calculator in ArcGIS 10.8.

TI is the terrain index, and the larger the terrain index, the weaker the ecosystem’s resistance to external interference and the higher its ecological sensitivity [22].

$$TI=\lg \left[\left({\displaystyle \frac{E}{\overline{E}}}+1\right)\times \left({\displaystyle \frac{S}{\overline{S}}}+1\right)\right]$$

(2)

Among them, $E$ stands for elevation, and $S$ represents slope; $\overline{E}$ and $\overline{S}$ signify average elevation and average slope, respectively.

$$E_i=aP+bS+cD$$

(3)

Landscape disturbance index $E_i$ is calculated using Fragstats 4.2 software utilizing land use data as a basis. The moving window method is used to calculate PD (fragmentation), SPLIT (separation), and DIVISION (dimensionality), respectively. The formula is as follows. Among them, a, b, and c are 0.5, 0.3, and 0.2, respectively [23].

$$X=\left(X_i-X_{max}\right)/\left(X_{max}-X_{min}\right)$$

(4)

$$X^{\prime }=\left(X_{max}-X_i\right)/\left(X_{max}-X_{min}\right)$$

(5)

To eliminate the differences between dimensions and attributes, all indicators are standardized. First, the data are resampled, then processed based on positive and negative indicators, and finally standardized [24]; $X$ is a positive indicator, and $X^{\prime }$ is a negative indicator.

2.4.2. Ecological Health Evaluation (EHI) Based on InVEST Model

Drawing on the VORS (vigor-organization-resilience-services) framework proposed by Peng et al. in evaluating urban ecosystem health, it is included in the assessment of ecosystem service capacity [25,26].

$$EHI=\sqrt{PH\times ESV}$$

(6)

$$PH=\sqrt[3]{V\times O\times R}$$

(7)

Ecosystem services (ESV) serve the ecosystem and evaluate three factors: water production (WY), carbon storage (CS), and soil and water conservation (SC). Physical health (PH) signifies the physical well-being of the ecological system, with V, O, and R representing vitality, organization, and resilience, respectively. V uses NDVI data, and O depends on the interactions between its components. Employing diverse metrics such as landscape heterogeneity (LH), landscape connectivity (LC), and landscape morphology (LS) [10], each factor is calculated separately in Fragstats 4.2 software.

$$O=0.4\times LH+0.4\times LC+0.2\times LS$$

(8)

R (Resilience) can be calculated by aggregating the multipliers of various land-use multiplied by their respective area weights (

Table 3. Resilience value of various land-use types.

Land-Use	Cropland	Forest	Grassland	Water	Impervious	Unused
R	0.3	0.6	0.8	0.7	0.2	0.4

) [22].

2.4.3. Landscape Connectivity Evaluation Based on MSPA Method

Morphological spatial pattern analysis (MSPA) serves as a method for image processing that utilizes principles of graphics to perform morphological transformations on images through operations such as erosion, expansion, closure, and opening. Many research results and practices have proven that the MSPA method is suitable for extracting functional connectivity evaluations of ecological networks [27,28]. Land use types including forests, grasslands, and water bodies within study area are used as foreground data, while other types are set as background data. The foreground is reclassified as 2, and the background is 1. MSPA is used to classify land use types, sort the importance of these patches, and determine their importance levels [29].

We use the Conefor 2.6 plugin to process land use types in ArcGIS 10.8, use Conefor 2.6 software to construct corridor scheme costs under different connectivity thresholds, and calculate the dPC indices of core patches with ecological source potential within the research scope [30].

2.4.4. Sequential Polygon Method

The polygon method mainly includes two types: sequence polygons and fully arranged polygons. The sequential polygon method starts from a fixed point and extends multiple line segments outward to form a polygon. These line segments represent specific indicator items, and their length reflects the specific numerical figures of the corresponding metric items. During the calculation process, each triangle composed of adjacent line segments that are in common points is calculated one by one using an ordered arrangement, and then the entire polygon’s area is obtained. This area value is the composite index [31]. This article selects three indicators, namely vulnerability inverse normalization (VI), ecosystem health normalization (HI), and connectivity normalization (CI), to construct a triangle. A is 60°.

$$R\mathrm{s}={\displaystyle \frac{1}{2}}\sin A\left(VI\cdot CI+CI\cdot HI+HI\cdot VI\right)$$

(9)

2.4.5. Geodetector

Geodetector serves as a spatial statistical analysis instrument that deeply explores the factors affecting regional heterogeneity, including differentiation and factor detectors, interaction detectors, risk area detectors, as well as ecological detectors [32].

$$\mathrm{q}=1-{\displaystyle \frac{{\displaystyle {\sum }_{h=1}^L{N}_h{\sigma }_h^2}}{N{\sigma }^2}}=1-{\displaystyle \frac{SSW}{SST}}$$

(10)

$$SSW={\displaystyle {\sum }_{h=1}^L{N}_h}{\sigma }_h^2,SST=N{\sigma }^2$$

(11)

where $h$ represents the count of layers pertaining to the dependent or independent variable, $L$ is the total number of layers, $N$ denotes the quantity of units, $N_h$ denotes the number of units in layer $h$, ${\sigma }_h^2$ denotes variation in the dependent variable value in layer $h$, ${\sigma }^2$ denotes the variation in the dependent variable’s value across the entire region and the sum of variance within the $SSW$ layer, and $SST$ stands for the overall dispersion within the entire region.

3. Results

3.1. Ecological Vulnerability Assessment

Utilizing the natural breakpoint approach tailored to the specific conditions of the study area, the ecological vulnerability grade was reclassified into five distinct categories: slightly (0–0.2), mildly (0.2–0.3), moderately (0.3–0.4), severely (0.4–0.5), and extremely vulnerable (>0.5).

In Figure 4 and Figure 5 in terms of time, the ecological vulnerability has shown an overall increasing trend, with the slightly vulnerable area decreasing first and then increasing. The mild vulnerability level has significantly decreased. Its area proportion dropped from 34.36% in 2005 to 11.76% in 2020, while the medium and above vulnerability levels have shown a clear upward trend. In 2005, mild vulnerability was the main focus, and from 2005 to 2020, the extent of severe vulnerability gradually expanded, showing a relatively consistent trend with the overall economic development level and urbanization process of the study area.

In terms of space, the ecological vulnerability of the central alluvial plain area is significantly higher than that of sparsely populated areas such as mountains and deserts on both sides of the north and south. Spatial heterogeneity is evident, aligning closely with the distribution of the population.

3.2. Ecological Health Evaluation

3.2.1. Evaluation Results of ESV

This paper is founded on using the InVEST model to obtain precipitation data, ETP (evapotranspiration) data, land use data, etc. within the designated research region, and we process them in ArcGIS 10.8 to obtain the results of water production, carbon sequestration, and soil and water conservation modules.

From Figure 6, it can be seen that over time, the water yield exhibits an inverted U-shaped trend, first increasing and then decreasing. The distribution is influenced by significant interannual variations in rainfall and evapotranspiration. Spatially, the southern Tianshan Mountains are located on the windward slope, receiving relatively heavy precipitation, resulting in an overall spatial pattern of south higher, north lower. In addition, the carbon storage values in 2005, 2010, 2015, and 2020 were 4.81 × 10⁹ t, 4.84 × 10⁹ t, 4.82 × 10⁹ t, and 4.78 × 10⁹ t, respectively, exhibiting an M-shaped inter-annual variation trend of first declining, then rising, and finally decreasing again. Overall, there was an increase of 2.68 × 10⁷ t. Spatially, high-value areas are mainly distributed in the mountainous regions of central and southern areas with high vegetation coverage, while low-value areas are mainly located in the northern desert regions. The highest value is 112,105 t/km², indicating significant regional differences, and the areas are showing an overall trend of concentrating towards the central and southern regions. Overall, soil conservation still displays an increasing trend. Spatially, it aligns with the water production situation, showing a trend of south higher, north lower. The soil conservation in the lower-altitude areas of the Tianshan Mountains is relatively high, while the soil retention in the southern mountainous areas is relatively low due to wind erosion, water erosion, and glacial frozen soil, which is closely associated with factors like terrain slope. According to the distribution pattern, the results of ESV of EBNSTM are closely related to their natural factors. The results of the two ecosystem services, soil conservation and water yield, are closely related to the topographic features of the region, specifically reflecting the distribution trend of the DEM. In contrast, carbon sequestration is primarily influenced by vegetation cover and thus is closely related to the LUCC.

3.2.2. Comprehensive Evaluation of ESV and EHI

This article combines ecosystem health assessment with ecosystem service assessment, as shown in the figure. The ecosystem health assessment outcomes are categorized into five grades: poor (<0.15), bad (0.15–0.3), average (0.3–0.5), good (0.5–0.7), and excellent (0.7–1).

As evident from the Figure 7, from 2005 to 2020, the health range above the general level accounted for 2.92%, 3.92%, 5.03%, and 4.44% of the total area, respectively, with a relatively low proportion, which implies that the overall ecological health status of the region is below the moderate standard. However, due to positive vegetation cover changes caused by climatic variations and anthropogenic actions in recent years, the overall health level shows an upward trend. In terms of space, there is a clear zoning feature, with two parallel health zones on the north and south sides appearing in the layout, the land use type in the central region and other regions being dominated by grassland, and the overall health status being at a normal level. The northern and southern mountainous areas are mostly unused land, and the health status is at a lower level, while Karamay, an important industrial town, and Urumqi, the demographic and economic center, exist at a highly developed level of ecological health due to strong human intervention, indicating that the ecological health of this region is also affected by human activities.

3.3. Landscape Connectivity Evaluation

Based on previous research and in combination with the actual situation of EBNSTM [28,33], we reclassify DPC into less important (<0.07), less important (0.07–0.2), generally important (0.2–0.8), important (0.8–2.5), and extremely important (2.5–5.3).

As can be seen from the Figure 8, the land use types in the southern region are dominated by grassland and forest, with a large number of patches that are relatively concentrated. In contrast, the northern region is mostly unused land, with fewer important patches that are unevenly distributed. In terms of time, the overall connectivity level has shown a significant downward trend, exhibiting a clear decrease in connectivity. This is primarily due to the significant increase in the area of construction land caused by human activities, which has destroyed the original forests, grasslands, and other ecological systems, leading to an aggravation of the fragility of the ecological environment and a substantial reduction in the connectivity of the region.

3.4. Spatiotemporal Heterogeneity Distribution of Composite Ecological Resilience

From Figure 9 and Figure 10, it can be seen that by using a grid calculator for overlay analysis in ArcGIS 10.8, a comprehensive resilience index was obtained. From the figure, it can be seen that over time, multi-dimensional ecological resilience has shown a significant upward tendency. Among them, the areas having high resilience index and high resilience index have significantly increased, transforming from 13,577.88 km² and 698.29 km² in 2005 to 19,052.14 km² and 3017.3 km² in 20 years; the area of moderately resilient areas has decreased, with a significant increase from 2015 to 2020, indicating a clear and stable upward trend in the overall ecology of the region. Marked spatial heterogeneity is evident, displaying a gradual decline from south to north, with higher values prevailing in the southern regions and lower ones in the north. The areas exhibiting high ecological resilience are primarily clustered in the southern and southeastern regions, strongly correlating with the superior quality of life and economic development. As per the Sankey diagram of the ecological resilience level transfer matrix, the proportion of areas with lower ecological resilience being transferred out is the highest, at 37.62%, 39.12%, 39.71%, and 33.04%, respectively. Other types show a clear trend of transferring in.

To explore the overall spatial attributes of ecological resilience, the assessment outcomes pertaining to it were used as observation variables, and Global Moran’s I was employed for exploration. The results showed that the significance level p was greater than 0.01, indicating that there was no significant spatial clustering. However, due to the unclear regional situation, Local Moran’s I was introduced for local in-depth exploration. The findings indicated that all years were successfully proven statistically significant at the 0.05 level. As shown in the Figure 11, the southern region showed obvious high–high clustering characteristics, and Moran’s index revealed a positive value, signifying a positive radiate influence of regions with high ecological resilience on their surrounding areas. The low–low clustering areas are principally found in the northern region and passed the local Moran’s index test, indicating that the region with high ecological resilience has a positive radiate driving influence on the neighboring area. That region is located in a low level of ecological resilience gathering area. Overall, there is a significant spatial agglomeration phenomenon in regional ecological resilience, and regional disparities are quite evident, while at the county level, it exhibits spatial heterogeneity from east to west.

3.5. Analysis of Driving Factors for Spatial Heterogeneity of Composite Ecological Resilience

Multiple factors contribute to the spatial differentiation of ecological resilience. This article selects 14 indicators, including DEM (X1), temperature (X2-Tem), precipitation (X3-Pre), wind speed (X4-WS), LUCC (X5), NDVI (X6), NPP (X7), per capita GDP (X8), population density (X9-POP), night light (X10-NL), AOD (X11), PM2.5 (X12), landscape disturbance index (X13-LDI), and industrial structure (X14-IS), to explore the driving factors. The ranking of q values is shown in

Table 4. Ranking of the q-values and its average ranking of the impact of individual driving factors on ecological resilience from 2005 to 2020.

Ranking	2005	2010	2015	2020	Average
X1	1	1	1	1	1
X2	9	10	9	8	7
X3	2	2	2	3	2
X4	10	11	8	10	8
X5	3	9	10	9	6
X6	11	8	6	6	6
X7	4	3	3	2	3
X8	6	6	4	5	4
X9	13	12	14	11	9
X10	14	14	12	14	10
X11	7	5	7	4	5
X12	5	4	5	7	4
X13	12	13	13	12	9
X14	8	7	11	13	8

.

From the q-value graph, the following can be seen: (1) The factor rankings are generally consistent within the evaluation period. Among them, the explanatory power of the DEM factor is the most significant, making the greatest contribution to spatial heterogeneity and serving as the primary driving factor. In addition, factors such as precipitation and NPP also make important contributions. (2) From the perspective of annual average explanatory power, factors such as DEM, NPP, precipitation, per capita GDP, and PM2.5 are among the top. (3) However, due to the fact that this region generally belongs to a low-density population area, with a significant polarization where only a very small number of people are relatively concentrated, while most areas belong to low-value zones with relatively small differences, the explanatory power of factors such as nighttime lights and population density is relatively low. Furthermore, their p-values are greater than 0.05, failing to pass the significance test. This indicates that the sample data are not significantly different from the original hypothesis, thus indicating that these factors have extremely weak or no significant impact on spatial heterogeneity. (4) However, as an indicator of human activities, LUCC changes also reflect the intensity of human activities, indicating that human activities are an important influencing factor. Therefore, the region is a product of both natural factors and human activities.

From the perspective of interaction (Figure 12), there is an overall trend of decline followed by an increase and then a further decline, which is mainly manifested as bivariate enhancement and nonlinear enhancement. Dominant factors such as DEM and precipitation exhibit bivariate enhancement in the interactions among various factors. Factors like PM2.5 and nighttime lights have strong nonlinear interactions with other factors, indicating that the interaction’s explanatory power between these two factors and other factors on heterogeneity exhibits a complex, nonlinear relationship. Factors like NPP and precipitation show bivariate enhancement, indicating that the combined effect of these two factors on the results is greater than the sum of their individual effects. In other words, the combined effect of NPP, precipitation, and other factors on the explanatory power of spatial variation is increasingly significant. The interaction of various factors is significant.

4. Discussion

Combining the results of this study with previous research on land use change (LUCC) and its impact on the ecological environment, we can discuss these findings and their implications from multiple perspectives. Firstly, land use change is considered a primary means by which humans alter the natural environment, particularly in arid regions where the interaction between land use patterns and natural conditions significantly influences ecological spatial changes [34]. This aligns with the findings of this study [35], which suggest that the changes in ecological resilience in Xinjiang are not only influenced by natural environmental factors but are also closely related to human activities. Specifically, the ecological resilience on the northern slope of the Tianshan Mountains is driven by multiple factors, including terrain, vegetation, climate conditions, pollution levels, and human socioeconomic activities. These findings correlate with the evolution of land use patterns and the impact of human activities on the ecological environment. Furthermore, the results emphasize human activities as a crucial factor affecting ecological resilience, which is consistent with prior studies indicating that land use changes in regions with intense human activity have a significant impact on the ecosystem [36]. In Xinjiang, particularly in the EBNSTM region, the pressure on the ecological environment is increasing due to infrastructure development such as highways, railways, transmission towers, and oil pipelines [37]. This further highlights the necessity of fully considering human activities when assessing ecological resilience, rather than relying solely on natural environmental factors. Thus, in regions affected by resilience loss due to natural environmental changes, natural restoration should be prioritized with limited human interference [38]. In reverse, for areas impacted by human-induced environmental damage, ecological restoration measures should be implemented. Additionally, the impact of land use changes on the ecological environment is not driven by a single factor but by a combination of factors. Our study found that compound factors play a critical role in shaping ecological resilience, which aligns with the concept of bivariate enhancement and nonlinear enhancement effects discussed in earlier studies. For instance, in regions with highly uneven population distribution like Xinjiang, the interaction of socioeconomic factors such as population and GDP with natural environmental factors contributes to the spatial heterogeneity of ecological resilience [39]. Therefore, assessing regional ecological resilience requires considering the interaction of different factors, particularly in the context of increasing human activity’s impact on the environment. Future research should focus on further exploring the relationship between land use changes and ecological resilience, particularly in large-scale infrastructure projects such as energy corridors, and assess their specific impact on ecological resilience. Future studies should also prioritize more refined data collection methods and long-term monitoring to address the limitations of the current study’s data. In particular, for arid and semi-arid regions like Xinjiang, future research should integrate various types of land use changes with the ecosystem’s self-repair and adaptation capabilities to inform more effective ecological protection strategies and sustainable development plans. This study can, to some extent, contribute to the effective utilization and improvement of the area’s ecological environment quality. However, due to the lack of grid data related to the proportion of the secondary industry during the data collection process, the use of point data interpolated with the inverse distance weighting method cannot fully and accurately reflect the local conditions. Additionally, the research years of 2005, 2010, 2015, and 2020 leave a gap in predictive research for the future. Future research should further explore the dynamic relationship between land use changes and the ecological environment, especially considering the impact of global climate change on arid regions. As Xinjiang continues to expand in energy development and infrastructure construction, assessing the specific effects of these changes on ecological resilience will be an important research direction. Overall, the findings of this study align with prior research and underscore the central role of human activities in ecological resilience. When evaluating and formulating ecological protection measures, it is crucial to consider the interaction between natural and human factors, particularly in regions undergoing accelerated development.

5. Conclusions

Due to the special geographic location and ecological traits of the EBNSTM, exploring ecological resilience in this region is of significant importance. We conducted an inclusive evaluation of the ecological resilience of EBNSTM through a multi-dimensional evaluation method. The results indicate that human activities are a crucial element affecting ecological resilience. Thus, it is necessary to fully consider the impact of human activities when understanding and assessing ecological resilience. This not only helps us obtain a clearer understanding of the current status and future development trends of the ecosystem but also aids us in developing more effective ecological protection strategies and sustainable development plans.

• The ecological resilience within EBNSTM exhibits significant spatial and temporal heterogeneity. From 2000 to 2020, it showed a remarkable upward trend overall, with a spatial pattern of south higher, north lower. The southern region displays a distinct high-high clustering feature, exerting a positive and forward-driving effect on the surrounding areas.
• The main driving factors of spatial heterogeneity in ecological resilience on the northern slope of the Tianshan Mountains are DEM, NPP, precipitation, PM2.5, and per capita GDP, indicating that its ecological resilience is closely related to the local natural environment, economic development, and human activities.
• Human activities are an important influencing factor, and the impact of the natural environment on ecological resilience is becoming increasingly significant. Among different factors, the bivariate enhancement effect is greater than the nonlinear enhancement of a single factor, demonstrating the importance of compound factors in shaping ecological resilience. Therefore, in the process of economic development, full consideration should be given to the self-repair and adaptation capabilities of the ecosystem.