Habitat Quality Assessment Based on Ecological Network Construction: A Case Study of Eremias multiocellata in Xinjiang, China

Abstract

Habitat fragmentation represents a significant threat to biodiversity, particularly the survival of wild species. Constructing and optimizing ecological networks are critical for promoting sustainable biodiversity, especially in the conservation of unmanaged wildlife. To address this, this study focused on designing and optimizing an ecological network tailored to the preservation of the Xinjiang desert lacertid lizard (Eremias multiocellata). This study integrated a dual-model approach, applying the InVEST model for habitat quality assessment and the MaxEnt model for suitable habitat prediction. An overlay analysis identified 15 core ecological source areas spanning 126,044 km², primarily located in the desert–grassland transition zones of the central and western study areas. A total of 34 ecological corridors were established utilizing the minimum cumulative resistance model, totaling 3764 km in length. These include 11 long corridors, 17 short corridors, and 6 potential corridors. Additionally, 100 strategic points were identified: 41 pinch points, 38 barrier points, and 21 stepping stones. This study identifies priority areas and obstacles affecting the ecological connectivity of the species’ habitats and highlights the importance of small habitat patches for long-term species dispersal and habitat expansion, providing more comprehensive guidance for sustainable development and species conservation. Furthermore, the methodology provides valuable insights into biodiversity conservation and the optimization of the natural habitat spatial layout in desert ecosystems, along with novel methods for managing and conserving other unmonitored animal species in various ecosystems.

1. Introduction

As modern society continues to develop and urbanization accelerates, an increasing amount of land is being utilized for human production and construction [1]. This has led to changes in landscape patterns and habitat fragmentation, posing a severe threat to biodiversity [2]. Currently, the rate of biodiversity decline exceeds that of any period in human history [3]. Habitat loss directly impacts species survival, while fragmentation splits once continuous populations into isolated groups, hindering gene flow, reducing genetic diversity, and increasing extinction risks [4,5]. Landscape connectivity between habitats determines the potential for species migration and dispersal. Extensive research has shown that accurately identifying habitats and constructing and optimizing ecological networks are critical for promoting sustainable biodiversity, especially for the conservation of unmanaged wildlife [6,7,8].

The theoretical framework for constructing ecological network spaces is centered on landscape ecology, integrating theories from island biogeography, disturbance ecology, and other disciplines to achieve habitat protection goals through spatial connectivity strategies [9]. Using linear ecological corridors as carriers, it addresses spatial fragmentation caused by landscape heterogeneity and promotes the cross-regional flow of biological elements through multidimensional connection channels [10]. Establishing a rational ecological network is of paramount importance for the safeguarding and sustained development of unmanaged wildlife. In landscape ecology, three common methods for optimizing ecological network connectivity are used: identifying ecological pinch points, barriers, and stepping stones. Ecological pinch points constitute crucial locations within ecological corridors, governing their overall connectivity [11], while barriers represent areas hindering connectivity. Managing these two types of areas can significantly enhance network connectivity [12]. Stepping stones, small patches added between core habitats, serve as temporary shelters during species migration, improving migration success and efficiency [13].

Habitats form the basis of ecological network construction, and assessing habitat quality for unmanaged wildlife is crucial for effective subsequent habitat management [14]. Previous studies have used models such as MaxEnt (Maximum Entropy) and InVEST (Integrated Valuation of Ecosystem Services and Trade-offs) to predict species distribution and habitat quality [15,16]. The InVEST model utilizes land use data to evaluate the impact of land use modifications on habitat quality. This methodology requires minimal data inputs while delivering high visualizability and objective assessments [17]. Its powerful spatial analysis capabilities have made it widely used in assessing habitat quality [18,19,20,21]. The MaxEnt model, based on the Maximum Entropy principle, predicts potential species distributions using species occurrence data and environmental variables. It is relatively insensitive to spatial errors in location data, requires minimal distribution data, and outperforms other distribution-based modeling methods, making it suitable for identifying potential habitats, predicting range shifts under climate change, and modeling the spread of invasive species [22,23,24]. In recent research, combining multiple models rather than relying solely on a single analysis model has emerged as a prevalent approach, which helps avoid potential impacts on the accuracy of the results due to the limitations of using a single model [25,26,27,28].

Climate change is disrupting precipitation and temperature stability in many regions worldwide [29]. Desert ecosystems, influenced by typical continental climates, are characterized by extreme temperature fluctuations and low precipitation. Their complex environments support unique flora and fauna, but long-term drought and overexploitation of water, soil, and vegetation resources pose significant conservation challenges [30]. Desert animals adapted to extreme conditions will face escalating threats in the future [31]. Thus, identifying core habitats and prioritizing conservation is essential for protecting biodiversity and ecosystem balance. However, systematic research on constructing ecological networks in desert ecosystems remains limited [32].

On the other hand, desert animals exhibit well-defined habitat ranges, distinctive behavioral characteristics, and preferred microhabitats [33,34]. Moreover, their habitats are significantly influenced by various factors, including climate, altitude, and vegetation types [35,36,37]. Consequently, desert animals have emerged as a focal subject of extensive investigation in wildlife conservation research. For example, Feizabadi et al. (2023) emphasized the need to designate new protected areas and corridors for two desert carnivores by assessing core habitats and connectivity while advocating for road restrictions to enhance connectivity [38]. Cypher et al. (2021) provided new insights for protecting Ammospermophilus nelsoni through habitat modeling and ecological attribute analysis [39]. Krishna et al. (2016) found that blackbuck habitat use dynamically changes with seasonal variations in resources and risks by examining resource quality habitat use relationships [40].

The sustainable development of wild species biodiversity within unmanaged wilderness areas necessitates the ecological assessment of their habitats alongside the establishment and enhancement of ecological network spaces [41]. Research establishing ecological network spaces for unmanaged wild animals, particularly desert species, remains understudied in China. Most existing studies primarily focus on optimizing the ecological networks of species within established nature reserves. These studies are concerned with issues such as how to adjust the boundaries of nature reserves and which species are better adapted to the ecological conditions of these reserves while giving relatively scant attention to the sustainable development of unregulated wild animals [42,43,44]. On the other hand, although numerous previous studies have delved into habitat connectivity, the emphasis has predominantly been on large mammals and key protected species [45,46,47]. Research identifying migratory ecological corridors and nodes connecting habitats of specific unmanaged wild animal species remains notably scarce. In particular, research on constructing ecological network spaces for unregulated desert animals remains an entirely unexplored territory.

Located in Xinjiang’s Tarim Basin, China, the Taklimakan Desert constitutes China’s largest desert and ranks as the world’s second-largest mobile desert, forming an exemplary desert ecosystem [48]. Eremias multiocellata is a reptilian species with a wide distribution across the Tarim Basin. It predominantly inhabits regions characterized by loose sand, dunes, or relatively densely vegetated lower sandy terrains [49]. In recent years, a combination of factors—excessive land reclamation, overgrazing, inadequate conservation attention, and lack of proper management—has caused a precipitous decline in its population. The information on E. multiocellata in Xinjiang on the GIBF (Global Invasive Species Database and Information System) website has remained unchanged for nearly three decades. Although E. multiocellata has not been classified as an endangered species, it has been included in China’s List of Terrestrial Wild Animals under National Protection of Important Ecological, Scientific, and Social Value [50]. In addition, E. multiocellata exhibits a relatively high population density within the Tarim Basin. Its distinct morphological features render it easily identifiable, and its complex habitat environment makes it an ideal model for research on constructing ecological network spaces for unmanaged animal species [51].

This work implemented the InVEST model to assess the habitat quality of E. multiocellata and utilized the MaxEnt model to estimate its possible appropriate habitats. By overlaying and analyzing the results of these two models, core ecological source areas were identified. Subsequently, the Minimum Cumulative Resistance (MCR) model was used to construct ecological corridors. Furthermore, based on the identification of strategic points—including pinch points, barriers, and stepping stones—the ecological network space of E. multiocellata was optimized.

2. Study Area and Data Sources

2.1. Overview of the Study Area

An overview of the study area is shown in Figure 1. Located in Southwestern Xinjiang, China, between the Tian Shan and Kunlun Mountain Ranges, the study area includes Aksu Prefecture, Kizilsu Kirgiz Autonomous Prefecture, Kashgar Prefecture, Hotan Prefecture, Alaer City, Tumushuke City, and Kunyu City. With a geographic range of 73°50′–84°59′ E longitude and 34°34–2°29′ N latitude, it covers approximately 567,800 km², accounting for 34% of Xinjiang’s total area. The research region exhibits a topography that is elevated in the south and diminished in the north, characterized by limited precipitation, averaging 76.22 mm per year. It possesses a temperate continental desert climate characterized by elevated evaporation rates. Most of the area consists of deserts and Gobi, making it suitable for desert animals [52]. In recent years, driven by urban and rural development and mineral resource extraction, the habitats of the desert animals in the study area have gradually become fragmented, facing challenges such as reduced suitable habitat area and declining habitat quality, threatening their long-term survival prospects [53,54].

2.2. Data Source

This study uses the following data: (1) Based on the possible distribution areas, habitat characteristics, and our field observations reported in the literature, we selected nine sampling sites around the Tarim Basin in 2019 (Wuqia, Minfeng, Yutian, Hetian, Yecheng, Yingjisha, Kashi, Tashikuergan, and Bachu; see Figure 1), and a total of 123 samples were collected. The desert interior area where E. multiocellata rarely distributes was also incorporated for a better understanding of the future construction of ecological networks. (2) Digital elevation model, slope, and aspect data were downloaded from the Geospatial Data Cloud Platform (http://www.gscloud.cn), all with a spatial resolution of 30 m. (3) Land-use and vegetation type data were sourced from the Resource and Environment Science and Data Platform of the Chinese Academy of Sciences (https://www.resdc.cn), also at a 30 m spatial resolution. (4) A total of 19 bioclimatic variables were collected from WorldClim (http://www.worldclim.org) with a spatial resolution of 2.5 arc-minutes (

Table A1. The 24 environmental factors.

Code	Variable Name	Unit
Bio1	Annual mean temperature	°C
Bio2	Mean diurnal range	°C
Bio3	Isothermality	%
Bio4	Temperature seasonality	°C
Bio5	Max temperature of warmest month	°C
Bio6	Min temperature of coldest month	°C
Bio7	Temperature annual range	°C
Bio8	Mean temperature of wettest quarter	℃
Bio9	Mean temperature of driest quarter	°C
Bio10	Mean temperature of warmest quarter	℃
Bio11	Mean temperature of coldest quarter	℃
Bio12	Annual precipitation	Mm
Bio13	Precipitation of wettest month	Mm
Bio14	Precipitation of driest month	Mm
Bio15	Precipitation seasonality	—
Bio16	Precipitation of wettest quarter	Mm
Bio17	Precipitation of driest quarter	Mm
Bio18	Precipitation of warmest quarter	Mm
Bio19	Precipitation of coldest quarter	Mm
Veg	Vegetation type	—
Dem	Digital elevation model	m
Aspect	Aspect	°
Slope	Slope	°
Land	Land use	—

Note: The 6 selected environmental factors are highlighted in bold.

). All imagery and data were unified into the WGS 1984 UTM Zone 45N coordinate system.

3. Research Methodology

The overall research workflow is shown in Figure 2. The core ecological source identification method proposed in this study integrates the InVEST model and the MaxEnt model. The initial phase employed the InVEST model to assess land-use data within the research area, with the objective of evaluating habitat quality and pinpointing places of high habitat quality. The second step utilized the MaxEnt model to predict suitable habitats for E. multiocellata, incorporating 24 environmental factors. The third step focused on constructing an ecological resistance surface; a comprehensive indicator system for the resistance surface was developed to assess the impacts of natural and anthropogenic interventions on the migration of E. multiocellata. The relative weights of resistance factors were assigned using the Analytic Hierarchy Process (AHP), leading to the final construction of a complete resistance surface. The fourth step involved identifying core ecological sources; overlay analysis was performed on high-habitat-quality areas identified by the InVEST model and suitable habitats predicted by the MaxEnt model, with the intersected areas designated as core ecological sources. The fifth step was the construction and optimization of the ecological network, utilizing the MCR model to establish ecological corridors and identifying critical locations within the ecological network based on circuit theory and landscape shape indices.

3.1. Construction of Core Ecological Source Areas

3.1.1. Habitat Quality Assessment Based on the InVEST Model

Habitat quality is a crucial indicator for measuring biodiversity and ecosystem services [55]. In Xinjiang, China, the assessment of habitat quality in desert ecosystems can serve as an effective decision-making tool for the conservation of desert-dwelling species [56]. Parameters, including resistance surface, threat sources, and sensitivity to threat sources in the study area, were used to calculate the habitat quality index to obtain the final assessment results. According to the InVEST User Guide, the basic formulas are as follows [57]:

$$Q_{xj}=H_j\left[1-\left({\displaystyle \frac{D_{xy}^Z}{D_{xy}^Z+k^z}}\right)\right],$$

(1)

$$D_{xj}={\sum }_{r=1}^R{\sum }_{r=1}^{Y_r}\left({\displaystyle \frac{W_r}{{\sum }_{r=1}^R{W}_r}}\right)r_y{i}_{rxy}{\beta }_x{S}_{jr},$$

(2)

In Equations (1) and (2), $Q_{xj}$ is habitat quality; $H_j$ is the habitat score assigned to each land use type (scored from 0 to 1, where 1 = highest habitat suitability and 0 = no habitat); Z = 2.5; and $k$ is the scale parameter (or constant). The constant $k$ is set to 0.5. The variable $r$ represents the threat level of land use types, where $r$ = 1,2, …n; $R$ represents an index of all modeled degradation sources; $y$ denotes the index of all grid cells on the $r$-th raster map; $Y_r$ represents the set of grid cells on the r raster map; $W_r$ is the weight parameter, and ${\beta }_x$ is the accessibility factor.

$$i_{rxy}=1-\left({\displaystyle \frac{d_{xy}}{d_{rmax}}}\right)\left(linear\right),$$

(3)

$$i_{rxy}=exp\left[-\left({\displaystyle \frac{2.99}{d_{rmax}}}\right)d_{xy}\right](exponential),$$

(4)

In Equations (3) and (4), $d_{xy}$ is the linear distance between grid cells $x$ and $y$, and $d_{rmax}$ is the maximum effective distance that threat $r$ can reach in space.

Cultivated land and construction land are highly susceptible to anthropogenic disturbances (e.g., farming, grazing, and mineral resource development). These activities not only affect the species composition and population size of E. multiocellata but also pose risks to their survival, reproduction, and migration [58]. Therefore, five habitat factors—dryland, paddy fields, urban land, rural settlements, and other construction land—were designated as threat sources in this study.

By referring to the InVEST model user manual and studies on similar areas in the arid regions of northwest China [59,60], we determined the maximum impact distance of the threat factors. After experts scored different threat factors, the weights of the threat factors were determined using the Analytic Hierarchy Process (AHP) [61,62] (

Table 1. Threatening factors and their maximum impact distance and weight.

Threatening Factor	Maximum Influence Distance (km)	Weight	Attenuation Type
Dryland	7	0.6	Linear
Water field	7	0.6	Linear
Other construction land	9	0.9	Exponential
Rural residential land	3	0.4	Exponential
Urban land	10	0.8	Exponential

). Based on the habitat characteristics and field survey results of E. multiocellata, the species was predominantly distributed in the marginal zones of deserts and grasslands [63,64]. Consequently, the sensitivity parameters for land with low vegetation coverage and sandy land were set relatively high, whereas those for water areas and land types with high vegetation coverage were set relatively low. After comprehensively considering the habitat traits of E. multiocellata, this study determined the sensitivity parameters of five threat factors, as well as the habitat sensitivity parameters to threats and habitat suitability parameters (

Table 2. Habitat suitability and sensitivity scales.

Land Use Type	Habitat Suitability	Dryland	Water Field	Other Construction Land	Rural Residential Land	Urban Land
Water field	0.2	0.3	0.3	0.3	0.4	0.3
Dryland	0.2	0.3	0.3	0.3	0.4	0.3
Forest land	0.6	0.4	0.4	0.5	0.3	0.3
High-coverage grassland	0.2	0.6	0.6	0.6	0.5	0.5
Moderate-coverage grasslands	0.5	0.4	0.4	0.4	0.3	0.3
Low-coverage grassland	0.9	0.3	0.3	0.3	0.2	0.3
Waters	0.2	0.6	0.6	0.6	0.5	0.7
Urban land	0	0	0	0	0	0
Rural residential land	0	0	0	0	0	0
Other construction land	0	0	0	0	0	0
Sandy soil	0.9	0.1	0.1	0.2	0.2	0.1
Gobi *	1	0.2	0.2	0.2	0.3	0.1
Salinate field	0.4	0.3	0.3	0.4	0.2	0.2
Marshland	0.3	0.3	0.3	0.5	0.4	0.5
Bare land	0.9	0.1	0.1	0.3	0.1	0.2
Bare rock texture	1	0.1	0.1	0.2	0.1	0.1

* In China’s multi-period land use remote sensing monitoring database, the definition of Gobi is land whose surface is dominated by gravel and whose vegetation coverage comprises less than 5%.

), with the half-saturation parameter set to the default value of 0.5 [65,66,67].

3.1.2. Suitable Habitat Prediction Based on MaxEnt Model

A total of 24 environmental variables were selected, comprising 19 climatic factors, as well as the digital elevation model, slope, aspect, land use type, and vegetation type. To prevent overfitting due to multicollinearity, we used ENMTools (R 4.3.1) to assess the correlations among 24 environmental variables. Simultaneously, we employed the Jackknife tool (Version 3.4.4) to evaluate the contribution of each environmental factor to the model’s predictive performance. Variables with low contributions and correlation coefficients (|r| ≥ 0.8) were excluded from the analysis [68]. Additionally, since default model parameters may induce overfitting and reduce predictive accuracy, we applied the ENMeval package (R 4.3.1) for model optimization [69]. All data were unified to the same resolution and coordinate system before being converted into ASCII format. A random subset comprising 75% of observed geographic distribution points was partitioned for model training, while the remaining 25% served as validation data. Bootstrap resampling was employed, with 10 independent runs conducted. The output format was set to logistic using default parameters. Each raster value in the output signifies habitat suitability for the species, ranging from 0 to 1, with higher values denoting greater suitability in that area.

The Receiver Operating Characteristic (ROC) curve, a well-established method for evaluating the accuracy of species geographic distribution models, was employed to test the model’s precision [70]. The area under the ROC curve was assessed to evaluate the model’s performance. To evaluate the accuracy of the MaxEnt model, an AUC value close to 1 indicates a high prediction probability. Models with an explanatory power of approximately 0.7 or higher are considered meaningful [71]. Response curve tests were performed to analyze the influence of individual environmental factors on overall habitat suitability, thereby assessing their relative importance within the model and informing the construction of the ecological resistance surface. Model accuracy was evaluated based on the area under the ROC curve.

3.1.3. Overlay Analysis

The InVEST model can effectively reflect the impact of land use changes on the habitat quality of species; however, for wild animals whose precise habitats are inaccessible, only considering this environmental factor is insufficient for accurate assessment. In such cases, incorporating the MaxEnt model to account for environmental factors beyond land use can enhance the comprehensiveness and accuracy of research results. The MaxEnt model has demonstrated considerable robustness in predicting species distributions [72], and the performance and robustness of the model were evaluated using receiver operating characteristic curves [73]. In this study, core ecological source areas were thus delineated through overlay analysis of the InVEST and MaxEnt model outputs. Using ArcGIS 10.8, the mean values of the two model results were calculated separately and used as reclassification thresholds [74]. Based on the InVEST-derived habitat quality results, areas exceeding the mean value were designated as high-quality habitats, whereas those falling below it were classified as low-quality habitats. Regarding the habitat suitability predictions from the MaxEnt model, areas above the mean were designated as suitable habitats, and those below the mean were designated as unsuitable habitats. High-quality habitat areas and suitable habitats were overlaid in the analysis. Studies have shown that small-area patches are prone to fragmentation and contribute less to biodiversity [75]. Following previous research [76], we excluded patches smaller than 20.5 km² and those that were scattered [77], with the remaining patches selected as core ecological patches. Adjacent core ecological patches were merged, resulting in a total of 15 habitat patches for E. multiocellata, with the smallest patch area being 89.71 km². To further quantify the accuracy of the model improvement, we standardized the boundaries of the results from both models and the superimposed core ecological sources, converted the points, transformed the format to csv, and performed five-fold cross-validation using Python 3.11.1.

3.2. Construction of Ecological Resistance Surface Based on Analytic Hierarchy Process

The Analytic Hierarchy Process (AHP) constitutes a systematic decision-making methodology. It breaks down intricate problems into a hierarchy of factors, combining qualitative and quantitative analyses to facilitate scientific decisions in multi-objective, multi-criteria contexts [78,79,80]. Based on the habitat characteristics of E. multiocellata and the field survey results, the resistance factors affecting the migration of E. multiocellata usually involve factors such as vegetation coverage, surface substrate, and sunlight abundance. The sparseness of vegetation coverage has a significant impact on the crawling speed of E. multiocellata; lower vegetation coverage is more conducive to the rapid crawling of E. multiocellata [63]. In addition, sparse vegetation habitats can help E. multiocellata effectively escape attacks from predators [64]. When the surface substrate where E. multiocellata is located is sandy, its movement performance is stronger [64]. According to our field observations and the existing literature data [81], altitude also has an impact on the distribution of E. multiocellata. E. multiocellata is concentrated in areas below 1500 m above sea level, and the probability of its occurrence decreases as altitude increases. We speculate that this situation may be caused by temperature changes, as E. multiocellata usually lives in warm places [82]. Studies have shown that E. multiocellata is only active from 10:00 to 13:00 and prefers to bask in the sun near caves [83]. This indicates that aspect also affects the distribution of E. multiocellata, because sunny and shady slopes receive different durations of sunlight exposure. Based on the above considerations, we finally selected 5 resistance factors—altitude, slope, aspect, land use type, and vegetation type—to construct the comprehensive ecological resistance surface. Using the AHP principle and based on previous studies [84,85], we assigned the weights of the resistance factors as 0.13, 0.20, 0.15, 0.04, and 0.48, respectively. Based on our previous field investigations and prior studies, the weight of vegetation was set to the highest value due to its habitat characteristics. Based on the influence of different environmental factors on the migration process of E. multiocellata, we assigned resistance values to the secondary classification of these factors to better align with the species’ migration behavior. Environmental factors that facilitate migration were assigned lower resistance values, while those that hinder migration were assigned higher resistance values (

Table A2. Resistance assignment system of five resistance factors.

Resistance Values	Land	Slope/(°)	Dem/(m)	Aspect/(°)	Veg
10	Desert	≤20	≤1500	≤45	Grass land
60	Forest land	>20~30	>1500~3500	>45~135	Desert
100	Dryland	>30~45	>3500~4000	>135~225	Shrub
150	Other construction land	>45	>4000	>225	Coniferous broad leaved forest
Weight	0.04	0.20	0.13	0.15	0.48

). The resistance values were categorized into four levels: 10, 60, 100, and 150 [86]. As the resistance value increased, the interference with the migration process of E. multiocellata became more pronounced [87]. Finally, the comprehensive ecological resistance surface was obtained via weighted summation of the 5 resistance factor rasters using the raster calculator [82,88].

3.3. Construction of Ecological Corridors Based on Minimum Cumulative Resistance Model

In ecological research, the MCR model is extensively employed owing to its straightforward data requirements and the intuitive, clear outcomes it provides for identifying corridors [89]. As a geospatial analysis technique, it primarily serves to identify connectivity channels between different ecological units [50]. By calculating the cumulative migration resistance values across grid cells, the model constructs an optimal ecological network connecting core habitat patches. This approach effectively mitigates the isolation effects of landscape fragmentation on species migration and enhances the functional integrity of regional ecosystems. The formula for the MCR model is as follows [90]:

$$MCR=f_{min}{\sum }_{j=n}^{i=m}\left(D_{ij}\times R_i\right),$$

(5)

In Equation (5), MCR is the Minimum Cumulative Resistance value; $f_{min}$ is defined as a positive function that monotonically increases with both the distance from all ecological sources and their ecological resistance values; $D_{ij}$ is the spatial distance from ecological source $j$ to spatial unit $i$; $R_i$ represents the ecological resistance value of spatial unit $i$; $m$ is the number of resistance surface grids; and $n$ is the number of ecological sources.

The ecological resistance surface and core ecological source layers were imported via the Linkage Mapper tool (Version 3.0) within ArcGIS 10.8. The following options were selected: Identify Adjacent Core Areas, Construct a Network of Core Areas, Calculate Cost-Weighted Distances and Least Cost Paths, and Truncate Corridors. Considering the study area’s size and the movement capacity of E. multiocellata, the cost-weighted distance threshold was set to 3 km [77].

3.4. Identification of Strategic Points in the Ecological Network

3.4.1. Identification of Ecological Pinch Points

Within the framework of circuit theory, increased current density indicates a higher potential for ecological process flow, making such areas “ecological pinch points” in ecological corridors—even small-scale disruptions here can significantly affect overall connectivity [91,92]. The Ecological Pinch Points Mapper module within ArcGIS 10.8 was employed to iteratively identify ecological pinch points in “All-to-One” mode, utilizing core ecological source areas and the comprehensive resistance surface with a configured cost-weighted distance threshold of 3 km. Classification via the natural break method designated the highest-ranked category as ecological pinch points.

3.4.2. Identification of Barriers

Inspired by the random walk of electrons in electrical circuits, circuit theory provides an effective tool for simulating species movement in heterogeneous landscapes [18]. The Barrier Mapper module in the Linkage Mapper toolkit (Version 3.0) was used with a minimum search radius of 300 m, a maximum search radius of 1500 m, and a step size of 300 m, employing the moving window method to identify regions where removing the area would significantly improve connectivity across the landscape resistance surface [93]. The results were classified using the natural break approach, with the highest-ranked category identified as barriers.

3.4.3. Identification of Stepping Stones

As temporary stopover sites for long-distance species migration, the quantity and quality of stepping stones affect migration time, frequency, and success rates [94]. According to landscape ecology theory, larger patch areas and low-coverage grasslands enhance biodiversity for E. multiocellata [64,95]. Given the contribution of potential corridors to species migration, we selected large low-coverage grassland patches intersecting with long-distance and potential corridors as stepping stones [96]. According to the definition of the landscape shape index, more complex ecological land patch shapes are associated with more frequent exchanges of material and energy with surrounding patches, thereby enhancing the ecological functional connectivity of the patch [97]. Studies have shown that lizards inhabiting desert environments require structurally complex habitats for thermoregulation, foraging, concealment, and other essential behaviors [98,99]. The patch shape index was used for screening, which measures the complexity of patch shapes by calculating the deviation from the shape of a regular geometric figure with the same area [100]. The formula is as follows [97]:

$$Shape={\displaystyle \frac{0.25p}{\sqrt{A}}},$$

(6)

In Equation (6), $p$ is the perimeter of the patch and $A$ is the area of the patch.

China’s low-coverage grassland patches were processed in ArcGIS 10.8 to calculate individual perimeters and areas, followed by patch shape index computation for each patch in Fragstats 4.2. Finally, patches ranking in the top 1% in terms of shape index and intersecting with long-distance and potential corridors were selected as stepping stones [97].

4. Results

4.1. Results of Habitat Quality Analysis

The results of the habitat quality analysis of E. multiocellata using the InVEST model are shown in Figure 3. The habitat quality ranged from 0 (low quality) to 1 (high quality), and the habitat quality index ranged from 0 to 0.89. By calculating the mean value of habitat quality in the distribution area, the mean value of the habitat quality index was determined to be 0.73, and the area of patches with values higher than the mean was 346,133 km², accounting for 60.9% of the study area. Meanwhile, the area of the patches with values lower than the mean was 221,086 km², accounting for 39.1% of the study area. Spatially, the areas with high habitat quality were distributed in block-like patterns in the eastern and central parts of the study area, while those with low habitat quality appeared as strip-shaped distributions.

4.2. Habitat Prediction Results

Through screening (Figure A1), we retained six key environmental factors for prediction (Table A1). The model achieved the minimum AICc value when RM = 2 and FC = LQHPT. Therefore, RM = 2 and FC = LQHPT were selected as the optimal model parameters for this study (Ziyang College of Environmental Science and Technology, Figure A3, Figure A4). The habitat prediction results for E. multiocellata using the MaxEnt model are shown in Figure 4. The AUC value was 0.956, indicating a strong model performance and suitability for subsequent experiments. The habitat suitability ranged from 0 to 1, with the suitability index ranging from 0 to 0.98 and an average of 0.29. Using ArcGIS 10.8, the suitable habitat areas above the threshold comprised 228,525 km², accounting for 40.4% of the study area, while the unsuitable habitat areas below the threshold comprised 337,191 km², accounting for 59.6%. The Jackknife test results indicated that the annual temperature difference, slope, and land type were the key variables influencing E. multiocellata’s distribution. The model’s predicted values range from 0 to 1, with the higher distribution probabilities represented by green on the map. The suitable habitats are primarily distributed along the edges of the Tarim Basin, Tianshan Mountains, and Kunlun Mountains at elevations lower than 3500 m above sea level. To validate the model’s results, we considered the distribution and habitat characteristics, food availability, and hiding places of E. multiocellata. The model’s results were consistent with these habitat characteristics and are suitable for the subsequent construction of core ecological sources.

4.3. Construction of Core Ecological Source Area

As shown by the superposition analysis, the ecological source area was 132,474 km², accounting for 23.42% of the total study area (Figure 5). By combining the neighboring source patches and eliminating the small patches that were too fragmented, a total of 15 habitat patches were generated and designated as core ecological source areas, with a total area of 126,044 km². In terms of spatial distribution, the core ecological source areas exhibit a high aggregation degree, primarily distributed between the Tian Shan and Kunlun Mountain Ranges as well as the western part of the Tarim Basin, and are distant from water bodies such as lakes and rivers (Figure 5). In terms of altitude, 90% of the core ecological source sites are distributed in areas lower than 3500 m above sea level, consistent with the results highlighted in Section 4.2. The MaxEnt result (red) covers 96,051 km² (16.98% of the study area), the InVEST result (yellow) covers 213,659 km² (37.77%), and the unsuitable area comprises 123,532 km² (21.83%). Additionally, the results of the five-fold cross-validation indicated that the average accuracy of the InVEST and MaxEnt models in the core source area was 0.858 and 0.884, respectively.

4.4. Construction of Ecological Resistance Surface

The ecological resistance surface was generated by five resistance factors: land use type, slope, aspect, vegetation, and elevation. Among the five factors, vegetation type had the greatest influence, while slope had the least influence. Overall, the integrated ecological resistance in the study area showed a gradual decrease and then an increase from the Tarim Basin to the periphery (Figure 6). The interior of the Tarim Basin formed a strip-shaped low-resistance zone across the study area. Most of the areas had medium-low resistance values, and the areas with higher ecological resistance values were mainly distributed in the high-elevation areas of the Kunlun and Tianshan Mountain Ranges, suggesting that the high-elevation and large-slope environments exert a more significant influence on the distribution of E. multiocellata and are not conducive to the migration of the species’ populations.

4.5. Ecological Corridor Construction

Thirty-four corridors were constructed using Linkage Mapper, including 11 long, 17 short, and 6 potential corridors (Figure 7). The longest corridor was 419.27 km, the shortest corridor was 0.89 km, and the aggregate length of the corridors was 3,764.858 km. Due to the aggregation and dispersion of the 15 ecological source sites, the ecological corridors were centrally located in the northeastern part of the study area, displaying a pattern of long in the east and short in the west. Among them, patch 12 had the largest number of connected corridors, serving as a transit station, and its ecological status should be prioritized in future research. Six long-distance corridors (C1, C2, C3, C7, C10, and C11) run through the Tarim Basin, establishing north–south connectivity and promoting connectivity between the Aksu, Tumushuke, and Hotan regions. Two long-distance corridors, C6 and C9, promote connectivity between Kashgar and the Kizilsu–Kirghiz Autonomous Prefecture. In contrast, corridors D1, D2, C3, C6, and C9 play important roles in connecting the periphery and the inner ecological sources, and corridors D4, D5, D7, D8, D10, D12, D14, and D16 promote the connectivity of the inner sources of E. multiocellata.

4.6. Identification of Strategic Points for Ecological Corridor Optimization

Based on the “pinch points–barriers–stepping stones” strategic point optimization method, a total of 41 pinch points, 38 barriers, and 21 stepping stones were identified (Figure 8). The majority of the critical locations were situated in the southwestern region where Kashgar and the Kizilsu–Kyrgyz Autonomous Prefecture meet, and the land type in this region is mainly grassland, with the Gobi and desert on the east and west sides, respectively, which are very favorable for E. multiocellata migration. More barriers were distributed in the corridors of D2, D12, C1, C9, and C10, but pinch points were lacking, which should be prioritized in subsequent management and maintenance. The stepping stones were mainly distributed in the Aksu region and the area around the promenade, especially in the northern region of the research area, which compensates for the high number of barriers that require restoration in the corresponding area.

5. Discussion

The habitat quality results from our InVEST model showed that areas with high habitat quality were primarily found in the Tarim Basin and remote places with high elevations and rivers, where land use types are mainly low-cover grasslands, deserts, and Gobi. The areas with a lower habitat quality were mostly high-altitude regions, construction sites, and water-dense areas, which is consistent with the distribution characteristics of E. multiocellata avoiding rivers and alpine areas [34]. A higher habitat quality is more favorable for the sustainable development of E. multiocellata biodiversity; therefore, these areas need to be actively protected and managed [101].

The overall mean value of the MaxEnt model was low because E. multiocellata mainly dispersed near the periphery of the Tarim Basin, resulting in a low mean value overall. This is consistent with the species’ habitat preferences [63,64] and verifies the accuracy of the model predictions. The coldest monthly minimum temperature constituted the primary environmental factor influencing E. multiocellata’s habitat distribution, and the Gobi and the sparse scrubland at the edge of deserts and semi-deserts were the most suitable habitats, with these areas accounting for 78.8% of the entire source area. The response curve showed that the species was more likely to be distributed in areas with an annual temperature range of approximately 34.7 °C and gentle slopes, especially in the sparse scrubland at the edge of the desert and the Gobi. Therefore, the MaxEnt model in this study not only predicted the suitable habitat of E. multiocellata but also indirectly verified the results of the previous study [102]. The implementation of China’s Three-North Project in Xinjiang has led to a continuous increase in forest and grass vegetation and a gradual decrease in sandy land in this region [103]. In addition, with the development of new urbanization and agricultural modernization, the area available for E. multiocellata protection is minimal, necessitating the strengthened conservation of its suitable habitats.

According to the results of the integrated resistance surface, since the Hotan River passes through the interior of the Tarim Basin from south to north, a strip-like low resistance area is formed within the basin as the Hotan River passes through the interior of the Tarim Basin from south to north. Most areas had low to medium resistance values, indicating that E. multiocellata populations in the study area face a relatively low ecological resistance during migration, which suggests that their connectivity can be enhanced in the future by optimizing their ecological network space at a relatively low cost.

The difference between the core ecological source sites and the InVEST results was larger because the habitat quality assessment of the Gobi was set higher in the input data to match the ecological characteristics of E. multiocellata. The InVEST model only considered the evaluation of habitat quality predicated on current land use alterations, and it is necessary to combine other models to further improve the habitat evaluation system. Previous studies have mostly employed the MaxEnt model to forecast the possible suitable habitats for species [104,105]. However, the inappropriate parameterization of the model [106] and the failure to account for the actual situation of LULC often lead to an excessive predictive output [107]. In our proposed integrated model, potential and highly suitable distributional habitats for species can be predicted more accurately. Compared with the standalone InVEST model, the accuracy rate increased by 56.7%; compared with the standalone MaxEnt model, it increased by 37.1%. The average accuracy from the five-fold cross-validation exceeds 0.85, significantly improving the accuracy of the core ecological source site identification.

The 100 strategic sites identified in this study can be used to further explore ways to reduce the habitat fragmentation of E. multiocellata and promote connectivity among the source sites. These results can provide a theoretical basis for the construction and protection of refuges. Previous studies have shown that during migration, refuges can maximize survival rates by reducing the energy and time lizards spend on thermoregulation and providing access to escape from predators [108,109]. The north–south span of habitat patches was the largest between the Hotan and Aksu regions in the eastern region of the research area. Especially in the Aksu region, the habitat patches were relatively small, spaced far apart from other patches, and had a dense distribution of barriers, accounting for 34.2%. To avoid the formation of ecological islands in this region, prioritizing corridor protection while enhancing connectivity between the source sites is essential [110]. In addition, the construction of stepping stones should be strengthened. The Taklamakan Desert divides the habitat between the two regions, but the north–south-oriented Hotan River runs through the desert, connecting the two habitats. As a seasonal river, the flood and dry periods of the Hotan River correspond to the emergence periods of the species, which facilitates its migration [111]. Therefore, priority should be given to strengthening and protecting stepping stone construction between the two regions in order to avoid the formation of isolated and segregated regional populations, which would affect gene exchange and genetic diversity among the E. multiocellata populations. In the western part of the study area, at the border of Kashgar and the Kizilsu–Kirghiz Autonomous Prefecture, where the land is dominated by a bare rocky texture, the ecological pinch points are most densely distributed, accounting for 39% of the total. This indicates that the survival environment of E. multiocellata in this region is superior to that in other regions and needs to be protected as a priority. Barriers on the two corridors passing through the area need to be further optimized to maintain and ensure the species’ sustainability.

In recent years, mineral resource development has intensified in the Kashgar region. Such anthropogenic impacts may cause damage to some of the pinch points; the biological bridge method, therefore, can be considered to mitigate the effects in this region. In Xinjiang, government authorities typically employ context-specific measures such as straw checkerboards made from reed stalks for sand fixation. This method has been empirically validated as the most effective approach for windbreak, sand stabilization, and water conservation [112], and it can effectively facilitate the restoration of desert steppes [113], thereby providing stopover points for the migration of E. multiocellata. With respect to machinery for laying straw checkerboards, Chinese scholars have developed a series of windbreak and sand-fixation straw checkerboard-laying robots, which have significantly reduced labor costs [114]. Barriers exhibit a high spatial density within the study area’s northern sector. This is influenced by topographic isolation on the one hand, and, on the other hand, by the overall west-to-east topographic gradient of Xinjiang, which causes rivers to converge and form the Tarim River Basin in the region’s northeast. Here, water source obstructions and the expansion of towns and settlements may have hindered the migration of E. multiocellata. Overall, future management priorities in the study area should remain centered on the Aksu region. While part of the Aksu region lies within the Taklamakan Desert, which provides a suitable habitat for E. multiocellata, mountain ranges and oases fragment this habitat, threatening it with isolation and disrupting genetic exchange among populations. Additionally, large-scale cotton cultivation has encroached on natural habitats; pesticides have contaminated the food chain; and urbanization, along with transportation networks, has further exacerbated habitat fragmentation. In recent years, the legal protection system in Xinjiang has been continuously improved. Regulations such as the “14th Five-Year Plan for the Protection and Development of Wild Fauna and Flora in Xinjiang Uygur Autonomous Region” have created a favorable ecological environment for the reproduction of wild animals. A robust compensation mechanism for damages caused by wild animals has been established: in cases where economic losses are sustained in crop cultivation or animal husbandry due to the ecological protection of wild animals, the relevant authorities will conduct investigations and verifications, after which an appropriate compensation will be granted to the affected villagers. During the planning of large-scale projects, core patches should be avoided, or ecologically compensatory industries can be developed to balance economic development and ecological protection. In the process of constructing corridors and addressing barrier points, local residents should be encouraged to participate in the maintenance and construction under the guidance of government departments, thereby obtaining ecological compensation funds. The Chinese government can implement zoning-based controls and strictly enforce the ecological protection red line to ensure the scientific conservation of E. multiocellata and their symbiotic species within ecological protection zones while strictly prohibiting natural resource exploitation activities. Furthermore, local authorities can utilize events such as World Wildlife Day and the Wild Animal Protection Publicity Month of the Autonomous Region to organize awareness-raising campaigns delivered by ecological experts targeting local farmers and herdsmen, thereby enhancing the public awareness of wildlife protection. Real-time monitoring and research are also necessary. By understanding the population dynamics, distribution, and factors threatening the species in a timely manner, we can develop more effective conservation strategies and improve network modeling.

The methodological framework and parameter settings proposed in this study can provide a theoretical basis for constructing ecological networks for other desert-dwelling reptiles. The habitat characteristics of E. multiocellata are similar to those of other desert species, such as Phrynocephalus axillaris. However, due to the more complex habitat environment of E. multiocellata, the predicted suitable habitat area is smaller than that of P. axillaris. Nonetheless, its complex habitat environment enables E. multiocellata to utilize a greater variety of land types as stepping stones during migration.

However, for non-desert ecosystems, due to the distinct habitat preferences of species therein, the model parameters require separate consideration and adjustment. For non-desert ecosystems, taking forest and wetland ecosystems as examples, both the InVEST model and the MaxEnt model have proven effective in assessing habitat quality and suitability for species within these ecosystems [115,116,117,118]. This substantiates the feasibility of the dual-model approach proposed in this study to enhance the accuracy of habitat prediction. However, specific parameters within the models need to be adjusted according to the habitat characteristics of the species for different ecosystems. For instance, the key environmental variables influencing species habitat suitability in wetland ecosystems differ significantly from those in desert ecosystems. Given that wetlands are characterized by abundant water resources, complex hydrological conditions, and specific vegetation communities, these environmental factors must be prioritized in the model construction.

6. Conclusions

In this study, E. multiocellata, an unmanaged wildlife species in the Xinjiang region of China, was used to construct an ecological cyberspace. The habitat quality and suitability for the species were assessed and predicted using the InVEST and MaxEnt models. Using superposition and reclassification, we extracted areas with values higher than the average model indices as core ecological source areas. Using the MCR model, we constructed ecological corridors, identified ecological pinch points and barriers (based on circuit theory), and determined stepping stones using the landscape shape index. This study also integrated two existing models for ecological source site construction and enhanced the accuracy and rationality of the species’ ecological source area identification by incorporating multi-level environmental factors. This approach applies to research on unmonitored fauna where obtaining accurate sample data is challenging. The pinch points–barriers–stepping stones optimization method provides a foundation for future conservation assessment of unmonitored wildlife. By applying this method, research can more intuitively identify priority areas for conservation and improve landscape connectivity between species’ habitats.

This study identifies priority areas and obstacles affecting the ecological connectivity of species’ habitats and highlights the importance of small habitat patches for supporting long-term species dispersal and habitat expansion, providing more comprehensive guidance for sustainable development and species conservation. Furthermore, it offers insights into biodiversity conservation and the optimization of natural habitat layouts in desert ecosystems, as well as new strategies for managing and conserving other unmonitored animal species in various ecosystems.