Spatial assessment of ecological sensitivity for sustainable planning in Lijiang city using GIS and AHP

Introduction: In response to the increasingly prominent problem of ecological environmental vulnerability.

Method: This study comprehensively applies the Geographic Information System combined with Analytic Hierarchy Process to objectively quantify and comprehensively assess the ecological sensitivity of Lijiang City, Yunnan Province, in southwestern China. Six key evaluation indicators will be selected: elevation, slope, slope orientation, water buffer zones, fractional vegetation cover, and land use. The weights of these factors are determined through expert evaluation, and a sensitivity distribution map is generated using weighted overlay technology. This achieves a multidimensional quantitative assessment and spatial visualization of ecological sensitivity across the entire region of Lijiang City.

Result: The results indicate that ecologically sensitive regions in Lijiang City display significant spatial heterogeneity, with areas classified as extremely and highly sensitivity comprising 49.91% of the total area.

Method: Based on the assessment results, the study uses the Global and Anselin Local Moran’s I, and Getis-Ord General G to conduct spatial autocorrelation and hotspot analysis of the comprehensive Ecological Sensitivity Index. The two datas show high consistency, revealing a complex spatial pattern of ecological sensitivity in Lijiang City, characterized by large clusters and small anomalies. Furthermore, sensitivity analysis is incorporated to demonstrate the robustness of the results.

Result and discussion: Finally, the study proposes recommendations for the delineation of ecological protection redlines and zoning control strategies, providing a scientific basis for the sustainable development of the Lijiang World Heritage Site. It also establishes a technical framework for ecological sensitivity assessment applicable to plateau and mountainous areas, which is valuable for building ecological security patterns in similar regions.

1 Introduction

Against the backdrop of intensifying global climate change and increasing human activities, ecologically vulnerable areas are facing unprecedented pressure, necessitating scientific assessment of their ecosystems. Lijiang City, located in Yunnan Province, China, is renowned for its unique ecological environment and rich biodiversity. It is also a crucial component of southwest China’s ecological barrier and a region with a fragile and sensitive ecosystem. However, with the development of tourism and urbanization, Lijiang City is facing severe environmental challenges, such as water shortages, groundwater pollution, and wetland degradation (Dong et al., 2015). Therefore, as urbanization and tourism development accelerate, understanding the ecosystem sensitivity of the region becomes critically important for balancing economic growth with environmental protection.

Ecosystem sensitivity (ES) refers to the degree to which an ecosystem responds to external pressures, disturbances, or changes, such as soil erosion, sandstorms, heavy snowfall, or human activities. It encompasses both the sensitivity of the ecosystem to environmental changes and its capacity for recovery and adaptation (Store et al., 2015; Chen, 2023). Specifically, this metric is operationalized by assessing factors like biodiversity and ecosystem functionality to measure an ecosystem’s resilience (Ji et al., 2023; Li et al., 2021). In practice, ecological sensitivity assessment is a quantitative approach used to analyze regional ecological stability. It commonly employs indicators selected through subjective methods with the goal of optimizing ecosystem service functions and supporting sustainable management decisions (Yao et al., 2022). This concept plays a critical role in environmental management and conservation, aiding in the formulation of sustainable development strategies and measures. Accurately evaluating ecosystem sensitivity can inform land-use planning, help identify conservation priorities, and mitigate ecological risks. Therefore, after obtaining ES assessment results, it is essential to translate them into conventional land suitability evaluation outcomes (Shi et al., 2024; Dong et al., 2025).

The GIS-AHP method, a prominent approach within the Multi-Criteria Decision-Making in GIS (MCDM-GIS) framework, is well-established for integrating multiple criteria into spatial assessments (Saaty, 1980). Its coupling with GIS enables efficient processing of diverse data for rapid evaluation (Nyimbili and Erden, 2020; Khan et al., 2025). Consequently, GIS-AHP has become a common tool for ecological sensitivity (ES) assessment in fragile regions, valued for its systematic factor integration and spatial visualization, as evidenced by applications from multiple articles to evaluate the ecological sensitivity of sites (Yilmaz et al., 2020; Xu et al., 2023; Guan et al., 2023; Zhu M. X. et al., 2025; Li et al., 2025). However, a core limitation persists: the subjective uncertainty in AHP’s criterion weight assignment. While recent studies attempt to mitigate this—such as by integrating Entropy or Fuzzy AHP methods for more objective weighting (Nyimbili and Erden, 2020; Huang et al., 2023)—systematic validation of weight sensitivity remains a gap. Therefore, this study innovates upon the traditional GIS-AHP framework by incorporating Sensitivity Analysis (SA) specifically on criterion weights, adopting the one-at-a-time (OAT) method to rigorously verify model stability.

To another point, for the ES study of Lijiang City, there are also some limitations. Fan and Yang (2007) used the DRASTIC model to assess the groundwater vulnerability in the Lijiang Basin, utilizing seven indicators including groundwater depth and net aquifer recharge. With the support of GIS technology, the evaluation revealed relatively high groundwater vulnerability and significant pollution risks in the basin. Ma et al. (2011) focused on the Lashihai Wetland in Lijiang and developed an ecosystem health evaluation system based on the AHP, incorporating both natural and socio-economic factors. They classified wetland stability into six levels and found no “extremely stable” areas in the region. Ma et al. (2014) introduced the DPSIR (Driving Forces-Pressure-State-Impact-Response) model and GIS analysis, adopting 17 indicators to predict and evaluate the water resource conditions of the Yanggong River Basin for the near-term (2020) and mid-term (2030). The results indicated that the current state of the basin was “extremely vulnerable,” projected to be “highly vulnerable” in the near term, and expected to improve to “moderately vulnerable” by the mid-term, though significant pressures would remain. Hu et al. (2025) collected data from 2000, 2010, and 2020 and used GIS to study the evolution of ES in the Lijiang River, though it should be noted that this area differs conceptually from Lijiang City. Wang (2024) applied GIS to analyze data from 2000 to 2020 in Eryuan County, Dali Bai Autonomous Prefecture, Yunnan Province, examining spatiotemporal changes in ES, but this study did not cover Lijiang City. Therefore, research on the ES of Lijiang City remains limited. Existing studies suffer from several shortcomings: insufficient systematic selection of research factors, with some overemphasizing natural elements (e.g., groundwater studies) and others focusing on socio-economic impacts (e.g., tourism pressure), failing to fully integrate multi-dimensional natural and human factors; weak scalability, as findings from localized areas (e.g., rivers, wetlands) are difficult to apply to city-wide spatial planning decisions. To address these research gaps, this study explores the construction of an integrated AHP-GIS evaluation framework. By coupling multi-dimensional natural and geographical factors—including elevation, slope, aspect, water buffer zones, fractional vegetation cover, and land use—it aims to establish a more systematic ecological sensitivity evaluation model applicable at the city-wide scale of Lijiang.

In summary, current ecological sensitivity assessment research exhibits two key gaps. First, Limited Analytical Depth: Most GIS-AHP studies stop at producing static sensitivity maps, lacking statistical diagnosis (e.g., spatial autocorrelation) to verify whether patterns form significant “hot-spot” or “cold-spot” clusters. This limits their utility for precise decision-support. Second, Insufficient Methodological Rigor: The inherent subjectivity in AHP’s expert-based weight assignment is widely acknowledged, yet systematic sensitivity validation of these weights is rare. This oversight undermines the robustness, credibility, and cross-study comparability of the results.

This study conducts a multi-dimensional ecological sensitivity assessment for Lijiang City using a coupled GIS-AHP method. It aims to identify core sensitive areas, and provide a scientific basis for sustainable urban planning. The paper makes three key contributions: (1) It establishes the first city-wide ecological sensitivity evaluation model for Lijiang by integrating multiple ecological and geographical factors. (2) It introduces spatial autocorrelation analysis (Moran’s Index) to quantify and explain the spatial agglomeration patterns of sensitivity. (3) It innovatively applies criterion-layer sensitivity analysis to address the subjectivity and robustness issues of traditional AHP, thereby enhancing the result reliability. In summary, this work provides direct scientific support for delineating protection zones in Lijiang and offers a replicable methodological framework for assessing complex mountainous regions, promoting both theoretical and practical advancements in landscape ecology.

2 Methods

2.1 Study area

The study area is located in the north-western part of Yunnan Province (25°59′∼27°56′N, 99°23′∼101°31′E), at the transitional zone between the Yunnan-Guizhou Plateau and the Qinghai-Tibet Plateau (Figure 1). Lijiang City is characterized predominantly by complex and diverse highland mountainous terrain. The region exhibits dramatic elevation variations, ranging from 1,015 m in the Jinsha River valley to 5,596 m at the summit of Jade Dragon Snow Mountain (the lowest-latitude modern glacier in the Northern Hemisphere), forming a typical vertical climate zonation. Lijiang experiences a semi-humid, low-latitude plateau mountain monsoon climate with distinct wet and dry seasons. The area boasts a well-developed river system, with waterways such as the Yanggong River and Jinsha River crisscrossing the region. Lashihai Wetland, an internationally important wetland, serves as a critical stopover for migratory birds. Additionally, Lijiang is endowed with rich biodiversity and cultural resources, with a forest coverage rate of 61.2% (Yunnan Provincial Department of Natural Resources, 2025). However, Lijiang faces prominent ecological fragility, including dual pressures of water scarcity and water environmental pollution. Studies indicate that the Lijiang Basin has high groundwater vulnerability, severe pollution, and low water resource security, already endangering the aquatic ecological safety of this World Heritage site (Fan and Yang, 2007). The rapid development of tourism has further intensified the conflict between water supply and demand. Simultaneously, there is a growing contradiction between ecotourism development and lakeshore protection in the region (Ma, L.G. et al., 2011). These characteristics make Lijiang City an ideal case study area for researching ES in high-altitude mountainous regions.

Figure 1

Figure 1. Location.

2.2 Data sources

This study conducted an ES analysis using 2021 Digital Elevation Model (DEM) data and 2020 geographic data obtained from the National Geomatics Center of China. The DEM data, sourced from the Geospatial Data Cloud (accessible at https://www.gscloud.cn/search), consisted of GDEMV3 30M resolution digital elevation data. These data were imported into ArcGIS 10.8 for pre-processing to derive elevation, slope, slope orientation, and water buffer zones information for Lijiang City. Subsequently, NDVI data were extracted from remote sensing imagery to analyze vegetation coverage. The satellite imagery, acquired from the National Geomatics Center of China, was processed using ENVI 5.3 and ArcGIS 10.8, incorporating pre-processing steps such as calibration and correction. The Normalized Difference Vegetation Index (NDVI) is a widely used remote sensing technique for assessing vegetation status based on the reflectance of visible and near-infrared (NIR) light (Mukhtar et al., 2024). The specific operational procedures are as follows: NDVI data were sourced from the Geospatial Data Cloud, specifically utilizing the Landsat 8 OLI_TIRS satellite digital product. To obtain high-quality vegetation information for the growing season, images covering the study area from February to March 2021 with cloud cover below 15% were selected. Subsequent preprocessing was performed using ENVI software for radiometric calibration. The NDVI values were then extracted using the band calculator, and data within the 5–95% confidence interval were statistically analyzed, and the Fractional Vegetation Cover (FVC) was calculated and assigned a value ranging from 0 to 1. Finally, data of land use were obtained from the Global Geographic Information Resource Directory Service System (available at https://www.webmap.cn/main.do?method=index) and analyzed in ArcGIS. All datasets were standardized to the WGS 1984 UTM Zone 47N coordinate system and resampled to a consistent resolution of 30 m × 30 m to ensure uniformity and compatibility for analysis. All data were accessed on August 10, 2025.

2.3 Modelling ecological sensitivity (ES)

Owing to the inherent complexity of ecosystems, the selection of ES factors lacks a universally fixed standard and must be contextualized within specific ecological environments. A comprehensive, multi-factor methodology is commonly employed for the selection of ES indicators. Drawing upon an extensive review of the literature, this study synthesizes a conceptual framework centered on two distinct yet complementary dimensions: geographical factors and ecological factors, both of which critically influence ES. As demonstrated in numerous studies, areas characterized by high altitude, steep slopes, bare land, and proximity to river channels exhibit heightened susceptibility to erosion and soil loss, thereby representing regions of high ES (Zhao, 2022; Yu et al., 2020; Zhang and Zhang, 2018; Zhang et al., 2024; Zheng et al., 2019).

In summary, this study draws upon the aforementioned literature and based on data authenticity and availability, integrates the ecological characteristics and geographical features of the research area to construct an ecological sensitivity evaluation system comprising six indicators: elevation, slope, slope orientation, water buffer zones, fractional vegetation cover (FVC), and land use. The rationale for selecting these indicators is as follows:

Elevation is a crucial determinant of ecosystem fragility. Higher altitudes correlate with increased numbers of sensitive plant species and greater ecosystem vulnerability, leading to elevated regional ecological sensitivity (Özhancı and Yılmaz, 2018). Given Lijiang’s significant elevation variations and pronounced ecological gradients, this factor is indispensable for its sensitivity assessment.

Slope is a primary driver of soil erosion and geological hazards. Steeper slopes correspond to higher probabilities of natural disasters and greater ecological sensitivity, while flat areas are less affected (Weng, 2020). Its inclusion is essential for identifying sensitive zones in this erosion-prone plateau mountainous region.

Slope Orientation creates microhabitat variations by influencing sunlight exposure, which directly affects temperature, vegetation, and soil conditions. Slopes with longer sun exposure typically exhibit lower ecological sensitivity. In Lijiang’s complex terrain, considering aspect enables a more spatially refined assessment.

FVC, measurable via remote sensing indices like NDVI, quantifies surface vegetation and reflects ecological status. Higher vegetation cover generally indicates higher ecological sensitivity (Huang and Nan, 2017). For Lijiang, with its diverse vegetation, FVC directly indicates regional ecological quality and efficiently reveals sensitivity patterns.

Water Bodies are vital for biodiversity and ecosystem function. Proximity to water typically increases ecological sensitivity (Yang and Yang, 2022). Lijiang’s dense water network forms valuable but vulnerable ecological corridors. Delineating buffer zones helps identify and protect these sensitive areas from human interference.

Land Use reflects human impact on the natural environment. More intensively used land types often have higher environmental carrying capacity and lower ecological sensitivity (Zhu M. X. et al., 2025). Incorporating this factor captures the significant pressure from Lijiang’s tourism and urban expansion, providing practical relevance to the assessment.

Thus, through a synthesis of these theoretical and empirical considerations, three physical geographical factors (elevation, slope, and slope orientation) and three ecological elements (NDVI, land use and water buffer zones) were systematically selected. These six factors constitute the core indicators for assessing ES, as categorized and detailed in Table 1.

Table 1

Table 1. Grading criteria of ecological sensitivity evaluation factors. Due to Lijiang City's location on a plateau, with an elevation ranging from 960 m to 5518 m, areas below 1000 m were classified as non-sensitive in the selection of elevation criteria.

Following the normalization of each individual evaluation result, the Natural Breaks Method (Jenks Natural Breaks) was employed to classify each factor into five levels: insensitivity, slightly sensitivity, moderately sensitivity, highly sensitivity, and extremely sensitivity. These levels correspond to assigned values of 1–5, indicating the degree of importance. Within the GIS environment, the Reclassify tool was used to convert the original raster map of each factor into a sensitivity classification map with grades ranging from 1 to 5, thereby establishing a single-factor grading standard for ES in Lijiang City.

Among these factors, FVC was calculated using the NDVI-based pixel dichotomy model. The formula is as follows:

$\mathit{NDVI}=\frac{\left(\mathit{N}\mathit{I}\mathit{R}-\mathit{R}\mathit{e}\mathit{d}\right)}{\left(\mathit{N}\mathit{I}\mathit{R}+\mathit{R}\mathit{e}\mathit{d}\right)}$

$\mathit{FVC}=\frac{\left(\mathit{N}\mathit{D}\mathit{V}\mathit{I}-{\mathit{N}\mathit{D}\mathit{V}\mathit{I}}_{\mathit{min}}\right)}{\left({\mathit{N}\mathit{D}\mathit{V}\mathit{I}}_{\mathit{max}}-{\mathit{N}\mathit{D}\mathit{V}\mathit{I}}_{\mathit{min}}\right)},$

Herein, NIR and Red represent the reflectance in the near-infrared and red bands, respectively. and correspond to the NDVI values at the 5% and 95% percentiles within the study area, which are used to mitigate the impact of outliers. A lower FVC value indicates poorer vegetation coverage, higher ES, and is thus assigned a higher sensitivity score.

2.4 Weight indicator determination

AHP is an effective method for addressing complex decision-making problems and assists in identifying and weighting decision criteria. This approach has been validated in numerous studies (Hamadouche et al., 2014; Haara et al., 2017; Store et al., 2015). Accordingly, this study employed AHP to qualitatively rank the ES of the study area. AHP enables the derivation of precise weights by comparing decision criteria, making it more suitable for multi-factor site analysis than traditional methods that assign singular weights (Saaty, 2001).

In this research, a combination of online and offline questionnaires was administered to experts with doctoral degrees in fields such as landscape architecture, urban planning, and environmental ecology. The survey, titled “Analysis of ES Importance in Lijiang City” was designed using pairwise comparison techniques. Questions included, for example, “Which factor do you consider more important for the ecological environment of Lijiang City, slope or land use, and to what extent?” Based on the Saaty 1–9 scale method, factors within the same hierarchy level were compared in pairs (Table 2) to construct a hierarchical structure model. For each expert’s ratings, the n indicators to be analyzed were transformed into a judgment matrix A (Table 3).

Table 2

Table 2. AHP judgment matrix scale and its meaning.

Table 3

Table 3. Judgment Matrix Let A be the judgment matrix; i represents the row factor and j the column factor, denoting the importance of factor i relative to factor j (where i, j = 1, 2, 3, …, n). In the AHP data structure, the main diagonal (from top-left to bottom-right) must always be 1, indicating that each factor is of equal importance to itself. The elements in the upper-right and lower-left triangles are reciprocally symmetric. Numerical values represent the relative importance between indicators—the larger the number, the greater the relative importance.

Since multiple experts were involved, it was necessary to calculate the geometric mean for each row of elements in judgment matrix A:

${\mathrm{G}}_{\mathrm{i}}=\sqrt[\mathbf{n}]{{\displaystyle {\prod }_{\mathrm{j}=\mathrm{1}}^{\mathrm{n}}}\mathrm{\text{aij}}},\left(\mathrm{i}=1,2,3,\dots ,\mathrm{n}\right)$

The vector G = T was then normalized to obtain the weight vector W = T:

${\mathbf{W}}_{\mathbf{i}}$ = $\frac{\mathbf{\text{Gi}}}{{\sum }_{\mathbf{j}=\mathbf{1}}^{\mathbf{n}}\mathbf{\text{Gj}}}$ ,where represents the weight of the corresponding factor.

Finally, a consistency check was performed, the maximum eigenvalue ${\mathrm{\lambda }}_{\max }$ was computed:

${\mathbf{\lambda }}_{\mathbf{max}}$ = ${\displaystyle {\sum }_{\mathbf{i}=\mathbf{1}}^{\mathbf{n}}}\frac{\left(\mathbf{\text{AW}}\right)\mathbf{i}}{\mathbf{\text{nWi}}}$, where $\left(\text{AW}\right)\mathrm{i}$ denotes the ith element of the vector $\text{AW}$.

followed by the consistency index (CI):

$\text{CI}=\frac{{\mathbf{\lambda }}_{\mathbf{max}}-\mathbf{n}}{\mathbf{n}-\mathbf{1}},$

The Random Consistency Index (RI) was referenced from predefined tables, and the Consistency Ratio (CR) was calculated:

$\text{CR}=\frac{\mathrm{\text{CI}}}{\mathrm{\text{RI}}},$

A value of CR < 0.10 indicates that the consistency of the judgment matrix is acceptable.

2.5 GIS weighted overlay and comprehensive eensitivity calculation

A weighted overlay analysis was performed in GIS by integrating the standardized sensitivity classification maps of each factor with the weights derived from the AHP method. This analysis aimed to evaluate the comprehensive ES of Lijiang City. The assessment model is defined as follows:

${\mathrm{S}}_{\mathrm{j}}={\displaystyle \sum _{\mathrm{i}=\mathrm{1}}^{\mathrm{n}}}\text{WiFi},$

Where:

${\mathrm{S}}_{\mathrm{j}}$ represents the comprehensive score of ES, $\text{Wi}$ denotes the weight of the ith indicator, $\text{Fi}$ refers to the score of the ith indicator, N is the total number of indicators.

2.6 Sensitivity analysis

In any decision-making process, particularly within the framework of Spatial Multi-Criteria Decision Making (MCDM), conducting Sensitivity Analysis (SA) is essential due to uncertainties inherent in input data, criteria selection, and threshold definitions (Youssef et al., 2015; Munier et al., 2019). Therefore, sensitivity analysis is typically implemented after problem-solving to verify whether the original solution remains valid when these uncertain data change, thereby providing a crucial basis for precise decision-making (Alinezhad and Amini, 2011; Youssef, A. M., et al., 2015).

In this paper, the One-at-a-time (OAT) method (Daniel, C., 1973) is employed to estimate criteria weight sensitivity. This is achieved by changing one input factor at a time while keeping all other factors fixed, assessing the resultant effects on the outputs of the ecological sensitivity model for Lijiang City, and visualizing these effects in a GIS environment.

If the weight of the i attribute (criterion) changes from ${\mathbf{w}}_{\mathbf{i}}$ to ${\mathbf{w}}_{\mathbf{i}}^{\prime }$, then the weights of the other criteria ${\mathbf{w}}_{\mathbf{j}}$ are adjusted according to the following formula:

${\mathbf{w}}_{\mathbf{j}}^{\prime }=\frac{\left(\mathbf{1}-{\mathbf{w}}_{\mathbf{i}}^{\prime }\right)}{\left(\mathbf{1}-{\mathbf{w}}_{\mathbf{i}}\right)}\times {\mathbf{w}}_{\mathbf{j}}$

Where ${\mathbf{w}}_{\mathbf{j}}$ is the new weight value of another attribute (criterion) to be changed; ${\mathbf{w}}_{\mathbf{i}}$ and ${\mathbf{w}}_{\mathbf{j}}$ are the initial weight values of the criteria before conducting SA.

2.7 Spatial autocorrelation analysis

To scientifically validate the structural and agglomerative characteristics of the spatial distribution of ecological sensitivity, this study employed the Global Moran’s I for quantitative analysis. Proposed by Moran in 1950, this index is a classical statistic for measuring spatial autocorrelation. The Global Moran’s I was selected to quantify the spatial distribution characteristics of the ESI in Lijiang City. This method addresses the limitation of GIS spatial visualization in statistical rigor by providing an objective and quantifiable statistical evidence, thereby significantly enhancing the scientific validity and persuasiveness of the conclusions.

Furthermore, to compensate for the inability of the Global Moran’s I to reveal localized spatial correlation patterns—as it only reflects overall trends—this study additionally applied the Anselin Local Moran’s I, also known as LISA (Local Indicators of Spatial Association), and the Getis-Ord General G statistic. These two methods are complementary: Getis-Ord General G statistic focuses on further diagnosing the high-low clustering of the site. Whereas, LISA is designed to accurately identify the four types of spatial association—high-high (HH), low-low (LL), high-low (HL), and low-high (LH), thereby providing a comprehensive characterization of spatial clustering and heterogeneity. Through this multi-method validation strategy, this study aims to ensure that the identified ecologically sensitive core areas (e.g., hotspots) and anomalous conflict zones (e.g., HL areas) are more robust and reliable.

Overall, the spatial autocorrelation analysis in this study was conducted at three levels. First, the Global Moran’s I was employed to examine whether significant spatial autocorrelation existed in the ecological sensitivity overall. Subsequently, the Getis-Ord General G statistic was used to determine whether high-value or low-value clustering trends dominated across the entire study area. Finally, the Anselin Local Moran’s I (LISA) was applied to precisely identify local spatial association patterns, thereby revealing the detailed pattern of spatial heterogeneity in ecological sensitivity. By integrating global, local spatial autocorrelation and Getis-Ord General G statistic analyses, this study comprehensively and deeply characterizes the spatial dependency and heterogeneity of ecological sensitivity in Lijiang City at both macro and micro scales.

3 Results

3.1 Single-factor ES analysis

Through GIS data processing, the distribution maps of ES for various factors in Lijiang City were obtained (Figure 2). The ES of Lijiang City is predominantly characterized by extremely and highly sensitive areas. It was observed that, apart from the water buffer zones map and slope orientation map, which showed less obvious spatial distribution patterns of ES, the spatial distribution patterns of other single-factor sensitivities were significant. Lijiang City exhibits generally high terrain elevation. Extremely sensitive areas (>2500 m) highly coincide with the ridge lines of major mountain systems such as Jade Dragon Snow Mountain and Mianmian Mountain, displaying a distinct zonal distribution pattern and accounting for as much as 57.7% of the area. In contrast, insensitive areas (<1000 m) are minimally distributed along the southeastern boundary of the city. Moderately and slightly sensitive areas are primarily concentrated in the southeastern region, corresponding to the Part of the Jinsha River; In terms of slope, extremely and highly sensitivitive zones are distributed in a zonal distribution pattern across the Tiger Leaping Gorge, Yunling and Mianmian Mountains in the central and northwestern parts of Lijiang City, as well as along the canyon areas of the Jinsha River, accounting for 40.2% of the total area. The sensitivity distribution related to slope orientation appears as irregular patchy patterns, with ES on shady slopes significantly lower than that on sunny slopes.

Figure 2

Figure 2. Single-factor Maps. From left to right, they are elevation analysis and slope analysis, water buffer zones analysis, slope orientation analysis, fractional vegetation cover analysis, and land use analysis. It can be clearly seen from the legend that different colors represent varying degrees of ecological sensitivity.

Turning to fractional vegetation cover, extremely sensitive areas are predominantly concentrated in the Jade Dragon Snow Mountain Nature Reserve and the western forest regions like Yunling and Laojun Mountain, covering 26.3% of the area. Highly sensitive areas overlap with the dry-hot valleys along the Jinsha River, some cultivated lands, and bare areas in the northern part of the city, where ecosystems exhibit weak resistance to disturbances, accounting for 21.9% of the area. In terms of land use types, given Lijiang’s abundant natural resources and mountainous terrain, a more detailed sensitivity evaluation was conducted for different land use categories, compared to other cities, the criteria for being classified as highly sensitive or above may be more stringent. Extremely sensitive areas mainly include wetlands and permanent snow and ice, accounting for 0.1% of the area. Highly sensitive areas consist of forests and water bodies, which essentially cover most of Lijiang City, representing 57.1% of the area. Shrublands and grasslands were classified as moderately sensitive, while artificial surfaces and barelands, despite their low ecological function, showed minimal change after disturbance and thus were assigned slightly sensitive ratings, accounting for only 2% of the area. At the same time, Lijiang City is also rich in water bodies, and the 0–200 m buffer zones around water areas exhibit extremely high and high sensitivity, indicating the critical importance of protecting ecosystems adjacent to water systems.

From a multi-factor perspective, it can be observed that in the central and western regions of the study area, areas with moderately and highly sensitive elevation levels coincide with those exhibiting extremely and highly sensitive slope levels. This reinforces the notion that topographic factors are fundamental drivers shaping the macro-level pattern, which aligns with Fan, T.'s (2007) conclusion that groundwater vulnerability in the Lijiang Basin is strongly controlled by topography. Furthermore, the extremely sensitive areas in the west and central parts of the city overlap in elevation, land use, and fractional vegetation cover maps. This indicates that the role of topographic factors is not isolated, rather, through spatial coupling with land use and fractional vegetation cover factors, they form continuous ES zones, which also reveals the non-linear effects of multi-factor interactions in complex mountain systems. Another interesting finding is that the Chenghai area, located in the southern part of the central basin, is shown as a low or insensitive area in terms of ES factors such as elevation, slope, slope orientation, and fractional vegetation cover, however, in the land use map, it is classified as a highly sensitive area. Overall, it can be inferred that due to the unique geographical and ecological environment, the central and western parts of the study area are highly sensitivity, while the Chenghai area is moderately sensitive or below, indicating a certain regularity in the distribution of ES.

3.2 AHP analysis results

This study collected importance scores from eight experts with doctoral degrees in landscape architecture, urban-rural planning, and environmental ecology through a combination of online and offline questionnaires. The geometric mean of these ratings was calculated to derive judgment matrices, which subsequently underwent consistency tests. The results are presented in Tables 4–7.

Table 4

Table 4. AHP Data.

Table 5

Table 5. AHP Results As shown in the table above, a 6th-order judgment matrix was constructed for the six indicators: Elevation, Slope, Slope Orientation, Water Buffer zones, Fractional Vegetation Cover, and Land Use. Using the AHP, the resulting eigenvector was calculated as (0.502, 0.374, 0.173, 0.903, 1.816, 2.233). The corresponding weight values for these six factors are: 8.362%, 6.229%, 2.889%, 15.047%, 30.259%, and 37.215%. Furthermore, based on the eigenvector, the maximum eigenvalue ${\mathrm{\lambda }}_{\max }$ was computed as 6.317. This value was used to calculate the Consistency Index (CI) as 0.063, which is utilized in the subsequent consistency check.

Table 6

Table 6. Random consistency RI table Since a 6th-order judgment matrix was constructed in this study, the corresponding Random Consistency Index (RI) value, referenced from standard tables, is 1.260. The RI value is used in the following consistency ratio calculation.

Table 7

Table 7. Summary of consistency test results

In summary, the AHP results indicate that land use exerts the most significant influence on the ES of Lijiang City, followed closely by fractional vegetation cover, with minimal difference between the two. In contrast, slope orientation has the least impact, followed by slope. These findings provide critical weighting references for subsequent GIS-based overlay analysis.

3.3 Comprehensive ES analysis of Lijiang City

After determining the weights through the AHP method and performing weighted overlay, the comprehensive ES distribution map of Lijiang City was generated (Figure 3). Using the Natural Breaks method, the Comprehensive Sensitivity Index (ESI) was classified into five levels: extremely sensitive (3.7–4.63), highly sensitive (3.26–3.7), moderately sensitive (2.78–3.26), slightly sensitive (2.26–2.78), and insensitive (0–2.26), assigned values of 5 to 1, respectively, indicating their degree of importance (Table 8).

Figure 3

Figure 3. The comprehensive ES distribution map of Lijiang City.

Table 8

Table 8. Comprehensive evaluation analysis of ecological sensitivity The minimum value of the comprehensive evaluation index is 1.09 and the maximum value is 4.63.

Extremely sensitive areas are widely distributed and account for the largest proportion, primarily located in the western part of the region (including Laojun Mountain, Yunling, Jade Dragon Snow Mountain, the core area of Lashihai Wetland, the banks of the Jinsha River), the central area (such as the Mianmian Mountain Range), and parts of the southern region (Yongsheng County). Highly sensitive areas are concentrated in the peripheral zones of the extremely sensitive areas. Moderately and slightly sensitive areas are widely distributed in transitional zones between highly and slightly sensitive areas, encompassing mid-low mountain hills, as well as some cultivated land and gardens. Insensitive areas account for the smallest proportion and are mainly found in the southern part of the region, such as the surrounding of Chenghai Lake and Gucheng County, and the southeastern part, such as Huaping County. They are also relatively scattered in the central part of the region, including the ancient town of Lijiang and its built-up areas, some stable forest areas in the northern mountains, exposed bedrock areas in the dry-hot valley of the Jinsha River, and Chenghai Lake and its surrounding areas. Overall, the spatial distribution pattern of ES in Lijiang City is significantly influenced by topography and water systems. Extremely and highly sensitive areas are not isolated but form distinct ecologically sensitive zones along mountain ridges and water systems, constituting key ecological barriers and vulnerable regions in Lijiang City.

Therefore, special attention should be paid to Yulong Naxi Autonomous County in the western part of Lijiang City. Due to its abundant natural resources, most of this area comprises extremely sensitive clusters, representing a complex area integrating socio-economic core zones and crucial ecological areas. Meanwhile, the sensitivity of this area stems from the superimposed effect of inherent natural fragility and high-intensity human activities, making it a region where the conflict between ecological conservation and economic development is most acute. Therefore, it is necessitating extremely cautious assessment of potential disturbances from large-scale engineering projects, such as transportation and hydropower development.

3.4 Spatial simulation of sensitivity analysis results

This study synthesizes the results of the AHP and employs the One-at-a-time (OAT) method from Sensitivity Analysis (SA) to vary the criterion weights and evaluate their impact. Specifically, it focuses on altering the weights of the criteria identified as having the strongest influence, assesses the sensitivity of each criterion to these predefined weight changes, evaluates their relative impact on the stability of the overall rankings, and visualizes the resulting spatial variations in the assessment.

Based on the AHP results, FVC and Land use were found to exert the greatest influence on the ecological sensitivity of Lijiang City, with weights of 30.3% and 37.2% respectively (values precise to three decimal places). Consequently, a sensitivity analysis was conducted by adjusting the weight values of these two key criteria. Their weights were systematically varied across five new incremental intervals, each spanning 20%, ranging from 0% to 80% (as detailed in Tables 9, 10).

Table 9

Table 9. Senxitivity analysis of Landuse adjusted to new weight values ranging from 0% to 80%.

Table 10

Table 10. Sensitivity analysis of FVC adjusted to new weight values ranging from 0% to 80%.

Subsequently, the newly proportioned maps were reclassified into five suitability classes (Value 1–5) using the Natural Breaks method in ArcGIS. The simulated SA outputs for Land use and FVC under weight scenarios of 0%, 20%, 40%, 60%, and 80% were geographically visualized (as shown in Figures 4, 5). A visual assessment of the spatial variability in the generated maps under the defined input weights was conducted to test the sensitivity of the model outputs.

Figure 4

Figure 4. The simulated sensitivity analysis(SA) outputs were geographically visualized at cut-off weight values of 0%, 20%, 37.2%, 40% 60% and 80% for the Landuse criterion.

Figure 5

Figure 5. The simulated sensitivity analysis(SA) outputs were geographically visualized at cut-off weight values of 0%, 20%, 30.3%, 40% 60% and 80% for the FVC criterion.

For the Land use criterion, no significant observable changes appeared on the resulting suitability maps when its original weight value was decreased by approximately 17.2% and increased by 2.8%, respectively. This indicates that the model is insensitive to weight changes within a range of 20%–40%, but becomes more sensitive beyond this threshold. Therefore, it can be stated that the model weight stability for the Landuse criterion is within the 20%–40% range. Furthermore, when the new weight for the Landuse criterion exceeds 40%, the stability of the ranking preference order is also correlated with the model’s sensitivity. Similarly, for the FVC criterion, decreasing its original weight by approximately 10.3% and increasing it by 29.7% did not lead to significant visual changes on the suitability maps. This indicates that the model weight sensitivity for FVC is within the 20%–60% range, and the order of criterion rankings remains consistent even when the FVC weight exceeds 40%, demonstrating the model’s stability. The SA results further reveal that other criteria have a significant spatial impact on the suitability assessment decision outcomes, as the Land use and FVC criteria predominantly influence the results only within their specified model sensitivity weight ranges.

3.5 Spatial autocorrelation and hot spot analysis

To scientifically validate the structural and agglomerative characteristics of the spatial distribution of ES in Lijiang City, this study employed the Global Moran’s I index for analysis. Given the excessive volume of original raster data (over 2 million pixels), a systematic random sampling method via GIS was used to generate 100,000 random points as a representative sample for computation. This approach significantly improved efficiency while ensuring the reliability of the results.

The results show that the global Moran’s I for ES in Lijiang City is 0.47 (I > 0 indicates positive spatial autocorrelation), with a Z-score of 325.26 and a p-value of 0.000 (i.e., <0.001). This reveals a strong positive spatial autocorrelation and highly clustered pattern in the ES of Lijiang City. The value I = 0.47, significantly greater than zero, clearly demonstrates a stable and intense positive spatial autocorrelation in the distribution of ES. This means that areas with similar sensitivity levels tend to cluster spatially: highly sensitive areas are significantly adjacent to other highly sensitive areas (high-high clustering), while slightly sensitivity areas are significantly adjacent to other slightly sensitivity areas (low-low clustering). This value represents a “moderate correlation” level, which is considered a relatively strong spatial association in landscape ecology studies, indicating that the ES of the study area is governed by intense spatial processes.

Furthermore, the extremely high statistical significance (Z = 325.26, far exceeding the critical value of 2.58 required for a 99% confidence level; p = 0.000) implies that the probability of observing such a strong spatial clustering pattern (I = 0.47) under a completely random process is minuscule. This rejects the null hypothesis of “random distribution of ES in Lijiang City” with a very high confidence level (far >99.999%), proving that this spatial pattern is genuine and driven by intrinsic factors (e.g., topography, hydrology, human activities), rather than formed randomly. It can be concluded that the spatial distribution of ES in Lijiang City is not characterized by “weak clustering and local variation” but exhibits a distinct pattern of “overall high clustering with local coherence,” reflecting the strong spatial self-organization effects of natural and anthropogenic factors such as steep terrain, water system distribution, and land use.

In summary, there is strong and significant spatial dependency and heterogeneity in the ES of Lijiang City. Its spatial distribution is highly structured, showing a regular clustered pattern jointly influenced by physical geographical elements and human activities.

While the Global Moran’s I demonstrates the overall existence of spatial clustering in ES, it does not specify the exact locations of these clusters. To precisely identify the spatial hot spots and cold spots of ES in Lijiang City, this study further applied the Getis-Ord General G statistic and Local Indicators of Spatial Association (LISA) for analysis. The results of the Getis-Ord General G analysis indicate that the observed G-value is slightly greater than the expected G-value, meaning the actual degree of high-value clustering is 10% higher than expected under random conditions—a very substantial and significant difference (Table 11). The high Z-score of 246 and corresponding p-value of 0.000 reject the null hypothesis that “high and low values are randomly distributed” with absolute statistical confidence (far >99.999%), confirming that the observed high-value clustering pattern is not accidental but a real pattern resulting from intense spatial processes. Combining the Moran’s I (I = 0.47) and General G analysis results, it can be concluded that the spatial clustering pattern of ES in Lijiang City is dominated by “high-high clustering.” In other words, high ES areas do not exist in isolation but tend to be adjacent to each other, forming spatially coherent clusters or regions of high sensitivity.

Table 11

Table 11. LISA Analysis Result.

This finding has important ecological and management implications. It confirms the existence of large-scale, contiguous ecologically vulnerable zones in Lijiang City, where responses to disturbances may be amplified or trigger chain reactions due to the “aggregation effect.” In terms of ecological protection, high-value clustering areas should be regarded as absolute priorities for conservation and restoration. Management measures should not be limited to isolated points but should focus on entire high-sensitivity patches, implementing systematic, regional-scale protection strategies.

To further reveal the specific local spatial correlation patterns and heterogeneity of ES in Lijiang City, the LISA was conducted based on the above research. This analysis successfully identified four significant spatial correlation types: high-high (HH) clustering, low-low (LL) clustering, high-low (HL) outliers, and low-high (LH) outliers. These types collectively reveal the local details and anomalies under the global pattern, providing critical insights for precise ecological management (Figure 6).

Figure 6

Figure 6. Four significant spatial correlation types Map.

From the generated analytical maps, it can be observed that High-High Clusters (HH Clusters) serve as the core ecologically sensitive areas. These refer to regions with high ES surrounded by other areas of similarly high sensitivity, representing the most critical manifestation of positive spatial correlation. Such clusters exhibit extensive and continuously distributed spatial patterns, forming the main body of the highly sensitive ecological zones in central and western Lijiang City. They are primarily concentrated in the western regions, such as the areas along the Yunling Mountains, the Lashihai Wetland, Yongsheng County, as well as along the surroundings of Jade Dragon Snow Mountain. In the central region, they are mainly distributed in the Mianmian Mountains in the north, Yongsheng County, and surroundings of Chenghai Lake in the south. These HH clusters represent the core vulnerable zones of Lijiang’s ecosystem, acting as extremely high-risk areas for geological hazards such as soil erosion, landslides, and collapses. Simultaneously, they play a key role in water conservation and biodiversity protection. These areas should be designated as the core zones of the “ecological protection redline,” subjected to the strictest conservation measures. All forms of development and construction activities must be prohibited, and priority should be given to implementing ecological restoration projects such as geological hazard prevention and soil erosion control.

Low-Low Clusters (LL Clusters) represent ecologically stable areas. These refer to regions with low ES surrounded by other areas of similarly slightly sensitivity, representing another form of positive spatial correlation indicative of ecological stability. LL clusters also exhibit coherent and plate-like distribution patterns. They are primarily concentrated in the urban built-up areas of Lijiang City, where extensive impervious surfaces have fundamentally altered the natural ecological functions of these areas, resulting in lower “sensitivity” to external disturbances. Additionally, the high-altitude snow-covered areas and stable alpine meadows of Jade Dragon Snow Mountain experience minimal human disturbance, with stable ecosystem structures and strong resistance to interference. Therefore, natural LL areas (e.g., high-altitude regions) serve as crucial ecological stable bases and barriers for Lijiang City. For artificial cold spots like urban built-up areas, the focus of management should not be on increasing their sensitivity but on actively incorporating green infrastructure during urban renewal processes to enhance their ecological functions and reduce their negative on surrounding sensitive areas.

Similarly, High-Low Outliers (HL Outliers) represent the forefront of ecological erosion. These refer to areas of high ES surrounded by regions of slightly sensitivity, representing a typical form of negative spatial correlation known as “high-value outliers.” HL outliers are typically sporadically embedded on the outer edges of LL clusters or transitional zones. Examples include isolated woodland or grassland patches on steep slopes at the urban fringe, surrounded by construction land, or small, isolated bare areas or severely degraded zones within large stable forest areas. Thus, these ecological “islands” and erosion points represent the ongoing or impending ecological erosion and destruction. They may be remnants of natural patches eroded by surrounding human activities (e.g., urban expansion, tourism development) or starting points formed by internal natural degradation or sudden disturbances (e.g., wildfires, landslides). HL points are critical signals for ecological security warnings and priority targets for ecological restoration. These points require focused attention to prevent the expansion of degradation. Ecological engineering measures should be employed to connect them with surrounding source areas, preventing them from becoming “stepping stones” for further degradation.

Turning to Low-High Outliers (LH Outliers), they represent opportunities for ecological restoration. These refer to areas of low ES surrounded by regions of high sensitivity, representing another form of negative spatial correlation known as “low-value outliers.” LH outliers are typically embedded within HH clusters. Examples include small, relatively flat terraces or river valley terraces within continuous steep mountainous areas, or small reservoirs or wetland parks constructed within highly sensitive watersheds, where ecological functions have been effectively managed and enhanced. These areas hold irreplaceable value in maintaining regional ecological connectivity. For naturally formed LH points (e.g., terraces), strict protection should be enforced, prohibiting development to allow them to serve as nodes for species habitat and dispersal. For artificially restored LH points (e.g., reservoirs), their successful experiences should be summarized and used as demonstration bases for ecological restoration, exploring the possibility of replicating their success models in surrounding highly sensitive areas.

In summary, LISA analysis clearly reveals the complex spatial pattern of “large clusters with small outliers” in the ES of Lijiang City. HH and LL clusters form the main body of the spatial pattern, reflecting the structured, patchwork ecological baseline shaped by macro-topography, climate, and human activities. HL and LH outliers, on the other hand, reveal local variations and spatial interactions under the global pattern. This analysis not only validates the global statistical results but, more importantly, translates macro-level policy guidance into specific geographical spatial units and points. It provides a direct and scientific basis for achieving refined and differentiated ecological management through “one strategy for one area, one measure for one point.” The findings of this study offer clear practical guidance for the ecological management of Lijiang City and similar plateau heritage sites.

4 Discussion

This study constructed an ecological sensitivity assessment model for Lijiang City based on a GIS-AHP framework, providing a comprehensive evaluation of the city’s ecologically sensitive areas. Furthermore, it innovatively applied the One-at-a-time (OAT) method to conduct sensitivity analysis on the AHP criterion layer. This approach facilitated an in-depth investigation into the sensitivity of criterion weights to changes, thereby enhancing the robustness and reliability of the spatially visualized model results.

Ecological sensitivity refers to a regional ecosystem’s vulnerability and response intensity to external disturbances under specific conditions (Wang et al., 2024). To validate reasonableness of the assessment, results were compared with another AHP-based study of Lijiang (Miao et al., 2024), which used different indicator system and scale. Both studies show high agreement: Lijiang is overall highly sensitive, while the Chenghai Lake region, the Yanggong River to Gucheng District corridor, and Jinsha River areas are moderately sensitive. This spatial consistency confirms the good regional applicability and reliability of our model.

Methodologically, compared to other multi-criteria decision-making studies related to the ecological environment, recent research has increasingly emphasized improving model robustness through the integration of multi-source indicators and sensitivity testing. Examples include using fuzzy logic and sensitivity analysis within a DPSIR framework for water security (Yuan et al., 2025), and combining AHP, FAHP, and F-TOPSIS to examine weight perturbation effects (Mandal et al., 2025). These efforts focus on “reducing subjectivity” and “enhancing robustness.”

Building on this, our study extends the traditional GIS-AHP framework into a complete analytical chain for spatial pattern diagnosis and decision robustness verification. First, global and local spatial autocorrelation analysis reveals the clustering structure of sensitive areas, going beyond simple classification maps. Second, a weight perturbation method based on the One-at-a-Time (OAT) approach is applied to the two most influential indicators, testing model stability across five perturbation intervals. This combined methodology reduces uncertainty from weight subjectivity, yielding a more robust and credible assessment.

It is noteworthy that the evaluation results show that the main water bodies, such as the Jinsha River, Chenghai Lake, and Lashihai Lake, are rated as mildly or insensitive, while their surrounding areas exhibit significantly highly sensitivity. Moreover, the hotspot analysis map also indicates that the water areas themselves are classified as LL Clusters, yet many HL Outliers are distributed within them. This appears to deviate somewhat from the expected outcome. This apparent contradiction is explained through both theoretical and model-based analysis. Theoretically, the pattern aligns with the “Ecotone” theory (Clements, 1905). The land-water ecotone is inherently fragile and highly sensitive to disturbance, providing high ecosystem service value in limited space (Zhu Z. et al., 2025). Furthermore, empirical support comes from a landscape risk analysis in northwest Yunnan, which identifies areas around Chenghai and Lashihai Lake as high-risk priority zones (Wang et al., 2025), validating our model’s focus on water margins. Mechanistically, the result stems from the dominant weights of Land use and FVC in our AHP model. While water bodies score high in land use classification, they score low in FVC due to lack of vegetation. Other factors consistently rate water as insensitivity or slightly sensitivity. Consequently, the composite sensitivity score of surrounding vegetated land often exceeds that of open water, directly causing the observed pattern. In summary, our model accurately identifies the critical zone for aquatic health not as the open water, but as its surrounding ecotone. This highlights the model’s value in pinpointing key conservation spaces, shifting the focus from the water body to its influential peripheral zone for targeted management.

To further validate the practical value of this research, comparing our results with Lijiang’s 2025 Territorial Ecological Restoration Plan reveals strong alignment (Lijiang Gov 2025). The HH Clusters we identified (e.g., around Laojun Mt., Jade Dragon Snow Mt., Lashihai Wetland, Chenghai Lake) spatially match the Plan’s core “Two Zones, One Belt, Multiple Points and Corridors” restoration areas, validating the Plan’s macro-layout. More importantly, while the Plan defines broad zones (e.g., Jinsha River sector), our granular raster-scale analysis pinpoints priority intervention units within them, such as critical land-water ecotones. We conclude these ecotones are significantly more sensitive than open water, underscoring the need to protect and restore this sensitive interface. Overall, the research value extends beyond robust sensitivity analysis. It provides a detailed scientific basemap that can deepen spatial planning and inform targeted restoration decisions for Lijiang.

From the perspective of the model mechanism, the generation of this pattern stems from the dominant role of Land use and FVC within the AHP weight system. In the land use classification, water bodies are assigned a fixed highly sensitivity rating, while fractional vegetation cover within water bodies is categorized as insensitive. Other influencing factors consistently classify water bodies as insensitivity or slightly sensitivity. This ultimately results in the composite sensitivity score of the surrounding areas potentially surpassing that of the water bodies themselves after overlaying all six factors, which is the most direct reason for the low sensitivity of the water bodies.

To another point, it is acknowledged that Lijiang’s intricate terrain—characterized by significant urban elevation variations and human-modified landscapes—may host numerous small, ephemeral, or artificially altered water branches that are not fully resolved by the DEM-based model. The hydrological network generated from DEM data exhibits minor discrepancies when compared to the actual urban drainage patterns of Lijiang City. These differences are anticipated, given the inherent limitations of DEM in capturing the complexities of engineered urban environments. To enhance the reliability of the simulation, the generated water flow paths—including main rivers and their tributaries—were carefully validated and adjusted through visual comparison with high-resolution actual maps of Lijiang City. This step significantly improved the positional accuracy of primary hydrological features.

Nevertheless, the overall configuration of the simulated water system is considered representative and sufficiently accurate for the purposes of this study, which focuses on macro scale ecological sensitivity assessment rather than detailed hydraulic engineering. The fact that major flow directions and accumulation areas align well with real-world features supports the validity of using the corrected DEM output as a meaningful input in the sensitivity evaluation framework. Future studies may benefit from integrating higher-resolution LiDAR-derived DEM, drainage infrastructure vector data, and field-based hydrological observations to further minimize uncertainties in environmentally complex and highly urbanized settings.

In summary, this study not only systematically assesses the spatial differentiation characteristics of ES in Lijiang City but also deepens the understanding of the processes behind these patterns through spatial statistical methods. It provides an important scientific paradigm and decision-making support for the refined management of ecosystems in high-altitude mountainous World Heritage Sites like Lijiang. It is worth noting that some seemingly contradictory phenomena in the evaluation results, such as the Jinsha River, Chenghai, Lashihai and other major water bodies being located in mild or insensitive areas, while their surrounding areas are shown as highly or extremely sensitive areas, provide a key perspective for us to deeply understand the spatial laws of ecological sensitivity.

5 Conclusion

This study takes Lijiang City, Yunnan Province as the research object, and integrates GIS-AHP, and spatial statistical methods to construct a sensitivity evaluation model that combines geographical and ecological factors. It systematically reveals the spatial differentiation patterns and underlying driving forces, while proposing planning recommendations and ecological management strategies. The results show that extremely and highly sensitive areas account for 49.91% of the total area, primarily concentrated such as Chenghai Lake, Lashihai Wetland, Laojun Mountain, Jade Dragon Snow Mountain, Lashihai Wetland, the banks of the Jinsha River. These areas are absolute priorities for ecological protection and restoration. A key innovation of this study is implementing a One-at-a-Time (OAT) sensitivity analysis on the criterion layer. This quantitatively tests the robustness of model outputs against weight perturbations, directly addressing a common methodological critique and significantly enhancing the credibility of the results.

The study also demonstrates that the spatial distribution of ES is not random but exhibits strong spatial dependency and clustering. The Global Moran’s I (I = 0.47) and the extremely high Z-score (325.26) indicate that the spatial structure is highly statistically significant. Getis-Ord General G hotspot analysis accurately identified statistically significant hotspots (high-value clusters) and cold spots (low-value clusters), providing clear targets for differentiated management. LISA analysis revealed a complex local pattern of “large clusters with small anomalies.” High-high (HH) and low-low (LL) clusters form the main spatial body, reflecting the controlling role of natural conditions, while high-low (HL) and low-high (LH) outliers reveal the frontiers of conflict between human disturbance and natural processes, as well as potential opportunities for ecological restoration, offering important insights for precise management. Similarly, the study confirms that ES is driven by both natural and anthropogenic factors. Steep terrain and critical aquatic ecological functions form the natural foundation of high sensitivity, while human land use activities (e.g., construction land expansion) intensify or alter the spatial expression of sensitivity through superimposed effect with natural conditions, highlighting the complexity of human-land relationships in Lijiang as a World Heritage Site.

Although this study verifies the robustness of the results under the model’s weight settings through sensitivity analysis, an important methodological limitation remains: the static assessment framework based on single-year data. This limitation may affect the research conclusions in two main ways. First, the response of ecosystems to disturbances is dynamic. Inter-annual fluctuations in vegetation cover and climatic conditions (e.g., drought or flood years) cannot be captured in a static model, which may cause the assessment results to precisely reflect only the specific environmental context of that particular year. Second, land-use change in rapidly urbanizing areas is continuous. Static land-use data struggle to fully capture development trends after the assessment baseline year, which may lead to an underestimation of future ecological pressures around newly built-up areas. Therefore, while the conclusions of this study are solid and credible in revealing the overall spatial pattern of ecological sensitivity determined by stable geographical factors (such as topography and water systems), they are less suited for dynamic processes requiring high temporal resolution monitoring. Future research should integrate multi-temporal remote sensing data to construct dynamic assessment models, thereby enhancing the predictive capacity for the evolution of ecological risks.

This study, using Lijiang City as a case study, constructed a comprehensive ecological sensitivity assessment model integrating GIS-AHP and spatial statistics. It innovatively enhanced the robustness of the AHP weight system through criterion-layer sensitivity analysis using the One-at-a-time (OAT) method. The research found that ecologically highly sensitive areas are concentrated around major mountain ranges and water systems. More importantly, the model revealed that the sensitivity of the land-water ecotone at the edges of water bodies is significantly higher than that of open water. This phenomenon is well-explained by the “Ecotone” theory, shifting the focus of ecological conservation towards the more functionally critical riparian zones. The research findings highly complement the municipal ecological restoration plan, providing a directly applicable “scientific basemap” for identifying priority protection areas and implementing differentiated spatial governance. Although the static assessment based on a single year has limitations, the methodological framework established in this study provides a replicable example for conducting rigorous, in-depth, and practical ecological assessments in complex mountainous regions.

Data availability statement

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

Author contributions

TY: Conceptualization, Data curation, Formal Analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review and editing. QL: Data curation, Investigation, Supervision, Writing – review and editing. WL: Data curation, Investigation, Supervision, Writing – review and editing.

Funding

The author(s) declared that financial support was not received for this work and/or its publication.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References

Alinezhad, A., and Amini, A. (2011). Sensitivity analysis of TOPSIS technique: the results of change in the weight of one attribute on the final ranking of alternatives. J. Industrial Eng. 4 (1), 23–28. Available online at: https://www.sid.ir/paper/624197/en.

Google Scholar

Chen, W. J. (2023). Ecological sensitivity study of oasis supported by GIS: a case study of yili river valley. J. Bull. Surv. Mapp. 0 (5), 107–114. doi:10.13474/j.cnki.11-2246.2023.0145

CrossRef Full Text | Google Scholar

Clements, F. E. (1905). Research methods in ecology. Lincoln, Nebraska: University Publishing Company.

CrossRef Full Text | Google Scholar

Daniel, C. (1973). One-at-a-time plans. J. Am. Statistical Association 68 (342), 353–360. doi:10.1080/01621459.1973.10482433

CrossRef Full Text | Google Scholar

Dong, R. C., Siyuan, L., Yuan, Q., Zhinan, D., Chunming, L., Shuanning, Z., et al. (2015). Avoidance analysis in urban sustainable planning based on ecologically sensitiveareas in lijiang city, Yunnan province, China. J. Acta Ecol. Sinca 35 (7), 2234–2243. doi:10.5846/stxb201306051352

CrossRef Full Text | Google Scholar

Dong, J., Li, L., Yang, W., Liu, B., and Zhang, X. (2025). Construction and optimization of ecological security patterns in ecologically fragile areas: a case study of lanzhou city, China. J. Front. Environ. Sci. 13, 163998. doi:10.3389/fenvs.2025.1639986

CrossRef Full Text | Google Scholar

Fan, T., and Yang, S. Y. (2007). Groundwater vulnerability assessment in lijiang basin. J. Jilin Univ. Earth Sci. Ed. 37 (3), 1070–1075. doi:10.3969/j.issn.1671-5888.2007.03.022

CrossRef Full Text | Google Scholar

Guan, X. K., Wang, X., Zhang, J., and Dai, Z. (2023). Regulation and optimization of cultivated land in different ecological function areas under the guidance of food security goals-a case study of mengjin county, Henan Province, China. J. Front. Environ. Sci. 11, 1115640. doi:10.3389/fenvs.2023.1115640

CrossRef Full Text | Google Scholar

Haara, A., Store, R., and Leskinen, P. (2017). Analyzing uncertainties and estimating priorities of landscape sensitivity based on expert opinions. J. Landsc. Urban Plan. 163, 56–66. doi:10.1016/j.landurbplan.2017.03.002

CrossRef Full Text | Google Scholar

Hamadouche, M. A., Mederbal, K., Kouri, L., Regagba, Z., Fekir, Y., and Anteur, D. (2014). GIS based multicriteria analysis: an approach to select priority areas for preservation inthe ahaggar national park, Algeria. Arabian J. Geosciences 7, 419–434. doi:10.1007/s12517-012-0817-x

CrossRef Full Text | Google Scholar

Hu, J. L., Li, H., Wang, Y., and Luo, N. (2025). Spatiotemporal evolution characteristics and driving forces analysis ofecological sensitivity in the lijiang river scenic zone. J. Nat. Conservation 87, 126993. doi:10.1016/j.jnc.2025.126993

CrossRef Full Text | Google Scholar

Huang, S., and Nan, L. (2017). Urban ecological sensitivity evaluation of anshun, China. Int. J. Environ. Sci. Dev. 8 (9), 630–634. doi:10.18178/IJESD.2017.8.9.1029

CrossRef Full Text | Google Scholar

Huang, J., Zhang, Y., Zhang, J., Qi, J., Liu, P., and Liang, C. (2023). Eco-geological environment quality assessment based on FAHP-CV combination weighting. J. Sustain. 15 (14), 10830. doi:10.3390/su151410830

CrossRef Full Text | Google Scholar

Ji, L., Dong, J., Kienberger, T., Huang, J., Liu, F., Wang, L., et al. (2023). Quantitative assessment and development utilization modes of space resources in closed and abandoned mines. J. Energy Sources, Part A Recovery, Util. Environ. Eff. 45 (4), 10366–10380. doi:10.1080/15567036.2023.2246427

CrossRef Full Text | Google Scholar

Khan, N. A., Alzahrani, H., Bai, S., Hussain, M., Tayyab, M., Ullah, S., et al. (2025). Flood risk assessment in the swat River catchment through GIS-based multi-criteria decision analysis. J. Front. Environ. Sci. 13, 1567796. doi:10.3389/fenvs.2025.1567796

CrossRef Full Text | Google Scholar

Li, X. D., Chen, Z. T., Liu, Z., and Yang, H. (2021). Decision-making on reuse modes of abandoned coal mine industrial sites in beijing based on environment-economy-society matter-element models. J. Math. Problems Engineering, Mathematical Problems Eng. 2021 (1), 9941182. doi:10.1155/2021/9941182

CrossRef Full Text | Google Scholar

Li, C. X., Ren, T., and Li, Y. H. (2025). Ecological sensitivity analysis in typical loess Plateau gully region: a case study of qingyang, Gansu. J. Desert Res. 45 (06), 166–175. doi:10.7522/j.issn.1000-694X.2025.00201

CrossRef Full Text | Google Scholar

Ma, L. G., Cao, Y. R., and Li, X. T. (2011). Assessment of ecosystem health in lashihai Lake using AHP method. J. Geo-information Sci. 13 (2), 234–239. doi:10.3724/SP.J.1047.2011.00234

CrossRef Full Text | Google Scholar

Ma, X., Xu, J. C., and Su, Y. F. (2014). Vulnerability assessment of yanggong basin in lijiang. J. Environ. Sci. Surv. 33 (6), 60–66. doi:10.13623/j.cnki.hkdk.2014.06.015

CrossRef Full Text | Google Scholar

Mandal, B., Goswami, K. P., and Mondal, S. (2025). GIS-based suitability assessment of stone crushing site selection using AHP, Fuzzy-AHP, and Fuzzy-TOPSIS models: navigating towards sustainable environmental management in brahmani-dwarka interfluve. J. Environ. Sustain. Indic. 26, 100704. doi:10.1016/j.indic.2025.100704

CrossRef Full Text | Google Scholar

Miao, P., Li, C., Xia, B., Zhao, X., Wu, Y., Zhang, C., et al. (2024). Incorporating ecosystem service trade-offs and synergies with ecological sensitivity to delineate ecological functional zones: a case study in the sichuan-yunnan ecological buffer area, China. Land 13 (9), 1503. doi:10.3390/land13091503

CrossRef Full Text | Google Scholar

Mukhtar, M. A., Shangguan, D., Ding, Y., Anjum, M. N., Banerjee, A., Butt, A. Q., et al. (2024). Integrated flood risk assessment in hunza-nagar, Pakistan: unifying big climate data analytics and multicriteria decision-making with GIS. J. Front. Environ. Sci. 12, 1337081. doi:10.3389/fenvs.2024.1337081

CrossRef Full Text | Google Scholar

Munier, N., Hontoria, E., and Jiménez-Sáez, F. (2019). Strategic approach in multi-criteria decision making, 275. Cham, Switzerland: Springer International Publishing.

Google Scholar

Nyimbili, P. H., and Erden, T. (2020). A hybrid approach integrating entropy-AHP and GIS for suitability assessment of urban emergency facilities. ISPRS Int. J. Geo-Information 9 (7), 419. doi:10.3390/ijgi9070419

CrossRef Full Text | Google Scholar

Özhancı, E., and Yılmaz, H. (2018). Sensitivity analysis in landscape ecological planning; the sample of Bayburt. J. Agric. Fac. Bursa Uludag Univ. 32 (2), 77–98. Available online at: https://dergipark.org.tr/en/download/article-file/570121.

Google Scholar

Saaty, T. L. (1980). The analytic hierarchy process: planning, setting priorities, resource allocation. New York, NY, USA: McGraw-Hill Int. Book Co.

Google Scholar

Saaty, T. L. (2001). Decision making for leaders: the analytic hierarchy process for decision in a complex world. Pittsburgh, PA: University of Pittsburgh.

Google Scholar

Shi, Y., Fan, X., Ding, X., and Sun, M. (2024). An assessment of ecological sensitivity and landscape pattern in abandoned mining land. J. Sustain. 16 (3), 1105. doi:10.3390/su16031105

CrossRef Full Text | Google Scholar

Store, R., Karjalainen, E., Haara, A., Leskinen, P., and Nivala, V. (2015). Producing a sensitivity assessment method for visual forest landscapes. J. Landsc. Urban Plan. 144, 128–141. doi:10.1016/j.landurbplan.2015.06.009

CrossRef Full Text | Google Scholar

Wang, S. (2024). Landscape ecological risk evaluation and driving factors in the Lake basin of central Yunnan Plateau. Chin. J. Eco-Agriculture 32 (3), 391–404. doi:10.12357/cjea.20230412

CrossRef Full Text | Google Scholar

Wang, Y., Dong, Y. K., and Zhen, W. J. (2024). Analysis of spatiotemporal changes in ecological sensitivity in eryuan county from 2000 to 2020. J. HUBEI Agric. Sci. 63 (1), 11–17. doi:10.14088/j.cnki.issn0439-8114.2024.01.003

CrossRef Full Text | Google Scholar

Wang, S., Liu, F. L., Du, W. J., and Wang, Q. H. (2025). Spatial-temporal evolution of landscape ecological risk and driving forces in the Plateau Lake basin of northwest Yunnan. Environ. Sci. 46 (5), 3114–3126. doi:10.13227/j.hjkx.202405125

PubMed Abstract | CrossRef Full Text | Google Scholar

Weng, J. L., Wu, X., Lin, S., Lu, Y., and Xiao, M. (2020). Study on ecological sensitivity of rural landscape in meikou, nanping city. J. Southwest For. Univ. Nat. Sci. 40 (01), 153–159. doi:10.11929/j.swfu.201812005

CrossRef Full Text | Google Scholar

Xu, H., Zhang, Z., Yu, X., Li, T., and Chen, Z. (2023). The construction of an ecological security pattern based on the comprehensive evaluation of the importance of ecosystem service and ecological sensitivity: a case of yangxin county, Hubei Province. J. Front. Environ. Sci. 11. 1154166. doi:10.3389/fenvs.2023.1154166

CrossRef Full Text | Google Scholar

Yang, Y. Y., and Yang, C. J. (2022). Sensitivity analysis of ecological environment in dongchuan district based on GIS. J. Bull. Surv. Mapp. 6 (03), 7–12. doi:10.13474/j.cnki.11-2246.2022.0068

CrossRef Full Text | Google Scholar

Yao, Z., Jiang, C., Zong-Cheng, C., Shi-Yuan, Z., and Guo-Dong, Z. (2022). Construction of ecological security pattern based on ecological sensitivity assessment in jining city, China. Pol. J. Environ. Stud. 31 (6), 5383–5404. doi:10.15244/pjoes/151866

CrossRef Full Text | Google Scholar

Yilmaz, F. C., Zengin, M., and Tekin Cure, C. (2020). Determination of ecologically sensitive areas in Denizli province using geographic information systems (GIS) and analytical hierarchy process (AHP). J. Environ. Monit. Assess. 192 (9), 589. doi:10.1007/s10661-020-08514-9

PubMed Abstract | CrossRef Full Text | Google Scholar

Youssef, A. M., Pradhan, B., Pourghasemi, H. R., and Abdullahi, S. (2015). Landslide susceptibility assessment at wadi jawrah basin, Jizan region, Saudi Arabia using two bivariate models in GIS. Geosciences J. 19 (3), 449–469. doi:10.1007/s12303-014-0065-z

CrossRef Full Text | Google Scholar

Yu, J., Li, F., Wang, Y., Lin, Y., Peng, Z., and Cheng, K. (2020). Spatiotemporal evolution of tropical forest degradation and its impact on ecological sensitivity: a case study in jinghong, xishuangbanna, China. J. Sci. Total Environ. 727, 138678. doi:10.1016/j.scitotenv.2020.138678

PubMed Abstract | CrossRef Full Text | Google Scholar

Yuan, L., Zhou, Z., He, W., Wu, X., Degefu, D. M., Cheng, J., et al. (2025). A fuzzy logic approach within the DPSIR framework to address the inherent uncertainty and complexity of water security assessments. J. Ecol. Indic. 170, 112984. doi:10.1016/j.ecolind.2024.112984

CrossRef Full Text | Google Scholar

Zhang, Q. Q., and Zhang, T. Z. (2018). Land consolidation design based on an evaluation of ecological sensitivity. J. Sustain. 10 (10), 3736. doi:10.3390/su10103736

CrossRef Full Text | Google Scholar

Zhang, Q. Y., Wan, W., and Zhiqiang, Z. (2024). Identification of harbin ecological function degradation areas based on ecological importance assessment and ecological sensitivity. J. Sustain. 16 (16), 6763. doi:10.3390/su16166763

CrossRef Full Text | Google Scholar

Zhao, Z. Y. (2022). Comprehensive evaluation and spatio-temporal variations of ecological sensitivity on the Qinghai-Tibet Plateau based on spatial distance index. J. Acta Ecol. Sinica 42 (18), 7403–7416. doi:10.5846/stxb202110283041

CrossRef Full Text | Google Scholar

Zheng, Y., Lan, S., Chen, W. Y., Chen, X., Xu, X., Chen, Y., et al. (2019). Visual sensitivity versus ecological sensitivity: an application of GIS in urbanforest park planning. J. Urban For. and Urban Green. 41, 139–149. doi:10.1016/j.ufug.2019.03.010

CrossRef Full Text | Google Scholar

Zhu, M. X., Zhang, T., and Hu, H. H. (2025). Ecological sensitivity evaluation of longfeng wetland provincial nature reserve based on GIS. J. of Wetl. Sci. 23 (5), 878–887. doi:10.13248/j.cnki.wetlandsci.20240181

CrossRef Full Text | Google Scholar

Zhu, Z., Zhang, C., Lu, Y., Ye, J., Fang, G., Sun, C., et al. (2025). Ecological risk assessment and prediction of riparian zones in the Jiangsu section of the yangtze river from a spatiotemporal perspective. J. Ecol. Indic. 174, 113492. doi:10.1016/j.ecolind.2025.113492

CrossRef Full Text | Google Scholar