Ecological sensitivity assessment and driving force analysis of the Tarim river basin

Abstract

As a critical ecological barrier and water reservoir in northwest China crucial for regional sustainable development, the Tarim River Basin necessitates comprehensive analysis of its ecological sensitivity distribution and driving factors. In this investigation, we have woven together the Remote Sensing Ecological Index with the sensitivity assessment framework, constructing a comprehensive 15-indicator system for assessing ecological sensitivity. When elucidating the underlying determinants of ecological sensitivity within a watershed, employing the Optimal Parameter Geographic Detector (OPGD) model to discern the primary impetuses. The results showed that habitat quality analysis classified 56.53% of the basin as substandard, concentrated in central-eastern deserts under extreme aridity, while only 6.79% qualified as high-quality zones in northern vegetated areas. Topographic insensitivity dominated, with 70.08% of slopes exhibiting non-responsiveness, and human activities intensified fragility, as 25.09% of the area showed sensitivity to road proximity. Vegetation diversity and soil type interactions demonstrated equivalent explanatory power, with q-values of 0.657. Thermal-hydrological coupling emerged as a critical mechanism, prompting strategies like thermal buffer zones and precision irrigation to mitigate sensitivity hotspots. These results advance arid ecosystem governance by linking spatial heterogeneity to conservation policies. The Optimal Parameter Geodetector identified heat and temperature as dominant drivers, with q-values of 0.731 and 0.7045, respectively. Their synergistic interaction (q = 0.82) amplified thermal impacts from anthropogenic land modifications like irrigation infrastructure. These findings provide critical insights for informing ecological conservation strategies, offering actionable pathways to enhance the resilience and sustainable management of the basin’s ecosystems.

Introduction

Ecological sensitivity assessment represents a meticulous process that evaluates the susceptibility of an ecosystem’s structure and function to alterations prompted by external pressures, including anthropogenic activities and natural variables. This assessment centers on the ecosystem’s capacity to respond to such external influences and its inherent ability for self-recovery, thereby serving as a profound method for gauging the stability and overall health of ecological systems^1,2. In essence, this process entails a meticulous identification of potential ecological challenges inherent in the prevailing natural environmental conditions and their subsequent implementation within designated spatial domains. Regions characterized by heightened ecological sensitivity are particularly vulnerable to natural disasters, possess fragile ecosystems, and are easily subjected to degradation. Consequently, it is imperative to minimize human interventions and prioritize efforts directed towards ecological restoration and reconstruction³. Assessing the ecological sensitivity of river basins can unveil potential ecological challenges residing within these intricate systems and delineate them into specific spatial domains. This analysis furnishes a scientific foundation for devising effective conservation strategies and facilitating sustainable development initiatives^4,5. The findings from driving force analysis serve as a crucial tool in elucidating the underlying factors and processes driving ecological and environmental transformations. By discerning the root causes and intricate mechanisms at play, we can anticipate forthcoming patterns and craft tailored interventions to mitigate and adapt to these shifts.

The Tarim River, being the longest inland river in China, boasts an expansive basin characterized by intricate topography and a rich tapestry of ecological environments. This river plays a pivotal role in sustaining the regional ecological equilibrium and fostering biodiversity. Nevertheless, in the context of global climate change and escalating human activities, the ecological integrity of the Tarim River Basin confronts formidable pressures and challenges^6,7. As the economy progresses and the scale of industrial development expands unabated, the ecological and environmental challenges within the Tarim River Basin are increasingly coming to the fore. These issues manifest predominantly as water scarcity, deterioration of water quality, land desertification, and a decline in biodiversity. Such dilemmas not only disrupt the livelihoods and well-being of the basin’s inhabitants but also jeopardize the ecological security of the broader region^8,9,10. Consequently, undertaking an ecological sensitivity assessment and a driving force analysis within the Tarim River Basin holds profound significance for elucidating the developmental trajectories of the basin’s ecological environment. This endeavor not only facilitates the identification of critical ecological challenges but also lays the groundwork for devising scientifically informed strategies for ecological conservation^11,12.

Currently, researchers both domestically and internationally have embarked on comprehensive investigations into the ecological and environmental challenges posed by the Tarim River Basin. In the realm of ecological sensitivity assessment, scholars have undertaken quantitative evaluations of the ecological vulnerability across various regions within the basin. Researchers developed monitoring frameworks and applied mathematical models to reveal the ecosystem’s inherent vulnerability and responsiveness within the basin^13,14. In the realm of driving force analysis, scholars have meticulously examined the influences of climate change, anthropogenic activities, and various other factors on the ecological environment of the basin. This investigation, conducted in conjunction with an understanding of the basin’s natural environment and socio-economic characteristics, has elucidated the intricate mechanisms underlying the alterations in ecological and environmental conditions^15,16.

The ecological and environmental challenges within the Tarim River Basin are intricately intertwined with the imperative of ecological integrity and sustainable progression not only within the local confines of the region, but on a national scale as well¹⁷. A profound exploration into the ecological sensitivity evaluation and analysis of driving forces within the Tarim River Basin holds the potential to enhance and advance the theoretical and methodological frameworks of interrelated fields. By fostering broader societal involvement in the preservation of ecological environments, it paves the way for the cultivation of a harmonious relationship between humanity and the natural world.

Materials

Study area

The Tarim River Basin (Fig. 1), situated in the southern part of Xinjiang and the northern part of the Tarim Basin, encompasses the nine principal drainage systems of the Aksu River, Kashgar River, Yarkand River, Hotan River, Kaidu-Peacock River, Dina River, Wei River, Keriya River, and Chechen River, totaling 144 rivers. Encompassing a vast expanse of 1.02 km²this basin stretches over a length of 2486 km, standing as China’s lengthiest inland river. Originating from the lofty peaks of the Tian Shan and Karakorum Mountains, the river meanders along the northern rim of the Taklamakan Desert, passing through the southern regions of Aksu, Shaya, Kuche, Luntai, Korla, and Wuerli counties, before ultimately converging into Taitema Lake. The primary water sources are nourished by the snowmelt and glaciers of the towering mountains, as well as groundwater replenishment. The topography of the Tarim River Basin is intricate, showcasing an array of geomorphological formations, including towering mountains, basins, and desolate deserts. The climate within the basin is notably parched, marked by a classical continental warm temperate climate with meager rainfall, ranging from 30 to 100 hm annually, alongside an annual evaporation rate soaring between 1973 and 2920 hm. The region basks in prolonged sunlight exposure, high cumulative temperatures, and an extended frost-free season. The natural enclaves of Populus euphratica forests and desert scrublands lining the riverbanks stand as the most expansive and predominant natural forest vegetation types within the desert realms of Xinjiang. These verdant realms act as a verdant shield, safeguarding the Tarim Basin, its adjacent bustling cities, and flourishing oases. The current expanse of Populus euphratica forests spans approximately 16.92 million mu (equivalent to 1.092 million hectares), capable of sequestering around 14 million tons of carbon on an annual basis.

Fig. 1

The geographical location of Tarim River Basin.

Data source

This study uses MODIS data from the MOD13Q1, MOD17A3, and MOD16A3 datasets available at http://reverb.echo.nasa.gov. Meteorological data, including precipitation and temperature, were obtained from the China Surface Climate Data Set provided by the National Meteorological Information Center (https://data.cma.cn/) and processed using spatial interpolation methods to meet the resolution requirements of the study area. Population and land use data were obtained from the Resource and Environment Science Data Center of the Chinese Academy of Sciences (http://www.resdc.cn/). Digital Elevation Model (DEM) data were obtained from the Geospatial Data Cloud Platform, accessible at http://www.gscloud.cn/. Soil data were obtained from the 1:1,000,000 Soil Database of China maintained by the Nanjing Institute of Soil Science, Chinese Academy of Sciences. For the ecological sensitivity assessment, the hydro-meteorological and socio-economic dataset (1958–2009) of the Manas River Basin and Tarim River typical basin provided by the National Glacier, Permafrost, and Desert Data Center (https://www.ncdc.ac.cn/) was used.

Research framework

This study focuses on the Tarim River Basin, a region profoundly affected by water scarcity and land desertification. Initially, we employed the Remote Sensing Ecological Index (RSEI) to quantify the ecological environment quality of the study area. Subsequently, we integrated these insights with a sensitivity evaluation framework to create a 15-metric system for assessing ecological vulnerability across the basin. Leveraging the Optimal Parameter Geographic Detector (OPGD) model, our research delved deeply into the intricate determinants and pivotal driving forces underlying ecological sensitivity. The basic ideas of each analysis method are shown in (Fig. 2).

Fig. 2

Technical route of the research.

Methodology

Remote sensing ecological index (RSEI)

Remote sensing technology provides extensive monitoring data, enabling rapid collection of diverse environmental information across large regions. This facilitates comprehensive assessments of ecological sensitivity, a critical component of ecological research¹⁸. By analyzing remote sensing data—including vegetation cover, leaf area index (LAI), and gross primary productivity (GPP)—researchers developed an indicator framework reflecting regional ecosystem health. These measures capture critical ecological features such as vegetation vitality, habitat productivity, and surface moisture conditions. Calculating normalized values of these variables and integrating them with environmental factors like topography, climate patterns, and land use enables a holistic assessment of ecological vulnerability^19,20. The model assesses both current ecological health and ecosystem resilience to disturbances, providing a scientific basis for conservation planning. Satellite data’s extensive coverage and frequent updates enable continuous, real-time monitoring of ecological vulnerabilities. This capability is critical for rapidly identifying emerging threats and designing effective preservation strategies²¹.

Among the myriad natural factors that signify ecological integrity, Greenness, Humidity, Dryness, and Heat emerge as four pivotal indicators intricately linked to human existence. These elements not only shape our understanding of ecological conditions but also exert a profound influence on human perceptions regarding the quality of these environments. Consequently, they are frequently employed as vital metrics in the assessment of ecosystem health²². In the realm of remote sensing technology, sophisticated techniques can be employed to enhance thematic information and extract data pertaining to the crucial indicators. For instance, vegetation indices, land surface temperature, and moisture content derived from the tasseled cap transformation can elucidate verdancy, warmth, and humidity correspondingly. Given that buildings form an integral part of artificial ecosystems, the proliferation of impermeable surfaces resulting from urbanization has superseded the original natural habitats, leading to a desiccation of the earth’s surface. Additionally, the presence of bare soil exacerbates this aridification process; thus, indices relating to built structures and bare ground can signify the state of dryness²³. The conceptualized Remote Sensing Ecological Index (RSEI) may be defined as a composite of these four key indicators, represented by the following equation:

$$RSEI = f(G,W,T,D)$$

(1)

$$RSEI = f(VI,Wet,LST,NDBSI)$$

(2)

G represents greenness, W represents humidity, T represents heat, D represents dryness, VI represents vegetation index, Wet represents the wet component, LST represents land surface temperature, and NDBSI (Normalized Difference Built-up and Soil Index) represents the index for built-up areas and bare soil²⁴.

(1) Greenness: The Normalized Difference Vegetation Index (NDVI) is closely related to plant biomass, Leaf Area Index (LAI), and vegetation cover²⁵. Therefore, in this study, NDVI is selected to represent the greenness indicator, with the formula being:

$$NDVI = \frac{{(NIR - RED)}}{{(NIR + RED)}}$$

(3)

NIR represents the reflectance of the near infrared band, while Red represents the reflectance of the red band.

(2) Humidity: The Tasseled Cap Transformation, recognized as an effective technique for data compression and redundancy reduction, has been used utilized in ecological monitoring²⁶. In this study, the humidity indicator is represented by the moisture component Wet, which is expressed as:

${\text{WET}} = \left( {0.0315{\text{B}}_{1} + 0.2021{\text{B}}_{2} + 0.3012{\text{B}}_{3} + 0.1594{\text{B}}_{4} - 0.6806{\text{B}}_{5} - 0.6109{\text{B}}_{6} } \right)/10000$ B1-B6 represent the reflectance values of blue, green, red, near infrared, shortwave infrared band 1, and shortwave infrared band 2, respectively. The parameters may vary depending on the sensor used, such as the Operational Land Imager (OLI). Multiplying by 1/10,000 ensures that the wet component is between − 1 and 1.

(3) Dryness: The Dryness Index is a metric used to quantify the degree of surface dryness, typically calculated based on specific band reflectances or indices derived from remote sensing imagery. It is commonly constructed using parameters such as the Soil Index (SI) and the Normalized Difference Built-up Index (NDBI). The SI can be calculated using the visible and near-infrared bands of MODIS, while the NDBI can be derived using the red and shortwave infrared bands^20,27.

The formula for calculating the Soil Index (SI) is:

$$SI = \frac{{B3}}{{B3 + B7}}$$

(4)

In the formula, B3 represents the reflectance in the red band, while B7 signifies the reflectance in the shortwave infrared band.

The formula for calculating the Normalized Difference Built-up Index (NDBI) is:

$$\left( {NBDI = B6 - B5/B6 + B5} \right)$$

(5)

In the formula B6 represents the reflectance in the shortwave infrared 1 band, while B5 represents the reflectance in the near infrared band.

The Dryness Index can be calculated by a weighted combination of the Soil Index (SI) and the Normalized Difference Built-up Index (NDBI). The specific weights can be adjusted according to the study area and objectives. The expression for this calculation is as follows:

$$\left( {Dryness~index = \left| {\alpha \cdot {\text{SI}} + \beta \cdot {\text{NDBI}}} \right|} \right)$$

(6)

In the equations, \( \alpha \) and \( \beta \) are weighting coefficients that can be tailored to specific scenarios and research needs. For example, if the study area is predominantly bare soil, the weight of the Soil Index (SI) can be increased accordingly; conversely, if the study area is characterized by significant built-up areas, the weight of the Normalized Difference Built-up Index (NDBI) can be increased accordingly. These adjustments allow for a more nuanced representation of the Drought Index that reflects the predominant land cover types within the region of interest²⁸.

(4) Heat Index: The thermal infrared bands of the MODIS data, specifically bands 31 and 32, provide the necessary information to calculate the land surface temperature using the expression:

$$LST = a_{0} + a_{1} \cdot (T31 - T32) + a_{2} \cdot T31 + a_{3} \cdot T32$$

(7)

In the formula, LST represents the land surface temperature, while T31 and T32 are the brightness temperatures or radiative brightness temperatures of MODIS bands 31 and 32, respectively. The coefficients a₀, a₁, a₂, and a₃ are algorithm-specific and must be determined based on the particular algorithm and regional conditions²².

Construction of ecological sensitivity evaluation indicators.

Ecological sensitivity denotes the extent of an ecosystem’s vulnerability to natural and anthropogenic disruptions within a specific locale. It epitomizes the facility and probability with which ecological challenges manifest in reaction to perturbations, encapsulating the potential repercussions of external disruptions^2,29,30. In the face of comparable intensities of disruption or external forces, the probability of diverse ecosystems encountering regional ecological challenges is examined. Delving into the realm of ecological sensitivity is indispensable for the preservation of ecosystems^31,32.

The cornerstone of a judicious assessment of environmental sensitivity lies in the meticulous selection of evaluation indicators. While there exist no universally accepted standards for this selection process, adherence to certain foundational principles is imperative. First and foremost, the assessment indicators must be grounded in a robust scientific theoretical framework and carry profound scientific significance. Additionally, given that regional ecological sensitivity emerges from the intricate interplay of multifaceted factors, the indicator system should embody a holistic character. Moreover, the chosen evaluation indicators ought to be readily accessible, quantifiable, and comparable, as well as operationally viable. In the subsequent endeavor to ascertain the weights of the various ecological sensitivity factors, this study remains steadfast in upholding principles of significance, scientific rigor, and practical applicability. It meticulously evaluates the impact of each ecological sensitivity factor on the ecological environment, ensuring the objectivity and precision of the evaluation outcomes, alongside the feasibility and practicality of implementation. The Analytic Hierarchy Process (AHP) was systematically implemented through four stages to assign weights. First, a hierarchical framework was constructed, comprising the goal layer (ecological sensitivity, Level 1), the criteria layer (six dimensions, e.g., remote sensing indices and anthropogenic interventions, Level 2), and the indicator layer (15 specific metrics, Level 3). Subsequently, pairwise comparisons were conducted, where ten domain experts independently assessed the relative importance of criteria using a 9-point Saaty scale, prioritizing regionally critical factors (e.g., soil erosion susceptibility was ranked 3× higher than vegetation coverage). Third, consistency validation was performed via eigenvalue analysis to ensure all matrices satisfied the consistency ratio threshold (CR < 0.1), with outlier responses (2/10 experts) excluded through an iterative Delphi feedback mechanism. Finally, geometric mean aggregation was applied to derive the final weights³³. In this scholarly manuscript, commencing from the veritable conditions of the study locale and amalgamating antecedent inquiries, erudite viewpoints, regional idiosyncrasies, and investigative aims, a total of 15 assessment criteria were meticulously curated across six distinct dimensions. These encompass discernment towards remote sensing, ecological metrics, topography, ambient variables, and anthropogenic interventions^34,35(Table 1).

Table 1 Ecological sensitivity evaluation system .

Sensitivity classification criteria

Referring to the sensitivity grading criteria outlined in the “Interim Provisions on Ecological Function Zoning” of the General Administration of Environmental Protection of China” and “China’s ecological protection red line policy”, integrating the characteristics of various sensitivity evaluation factors within the study area, the sensitivity is categorized as: insensitive, slightly sensitive, moderately sensitive, highly sensitive, and extremely sensitive. These categories are numerically standardized with scores of 1, 2, 3, 4, and 5, respectively, from low to high¹. The higher the sensitivity rating, the greater the ecological value of the area and the more vulnerable it is to the impacts of development and construction, and the more targeted protection measures are warranted. A low sensitivity rating indicates that the area can withstand anthropogenic disturbance and can be developed to a certain extent³⁶. This paper, taking into account the actual conditions of the study area, divides the sensitivity levels into five classes: insensitive, slightly sensitive, moderately sensitive, highly sensitive, and extremely sensitive.

Optimal parameter geographic detector

The Geographic Detector is a sophisticated spatial analysis technique created to investigate spatial heterogeneity and unveil the latent driving forces. It is extensively employed in the examination of driving forces and factors³⁷. The fundamental premise posits that, should an independent variable exert a significant influence on a dependent variable, the spatial distributions of both entities ought to exhibit notable parallels³⁸. Spatial differentiation can be subjected to statistical analysis through the employment of the Geographic Detector, which boasts two significant advantages: firstly, its capability to discern both quantitative and qualitative data; and secondly, its aptitude for elucidating the interactive effects of multiple factors on the dependent variable. By calculating and juxtaposing the q-values associated with individual factors alongside the q-values derived from the interplay of two factors, the Geographic Detector is adept at discerning whether an interactive effect exists between the two variables, as well as elucidating the strength, direction, and potential linearity or non-linearity of this interaction. The interaction between two factors includes both multiplicative and nonlinear relationships, enabling comprehensive analysis of their combined effects on ecological sensitivity³⁹.

The Optimal Parameter Geographic Detector (OPGD) was prioritized over conventional spatial analysis methods for three principal reasons. First, regression models typically assume linear relationships and spatial homogeneity, which risks oversimplifying ecosystems characterized by nonlinear interactions and spatial heterogeneity⁴⁰. Second, machine learning approaches, despite their proficiency in pattern recognition, lack explicit mechanisms to disentangle synergistic or antagonistic interactions among factors—a critical requirement for ecological sensitivity analyses. In contrast, the OPGD explicitly quantifies interaction effects through q-statistics and spatial stratification, enabling robust identification of dominant drivers even in complex, non-stationary environments. Finally, compared to standard geographic detectors, the OPGD optimizes parameters such as spatial discretization thresholds, thereby minimizing subjective biases in scale selection and enhancing model generalizability⁴¹.

The Optimal Parameter Geographic Detector enhances the original Geographic Detector framework of the Geographic Detector by fine-tuning a range of parameters, such as spatial discretization and spatial scale optimization, in order to more effectively unearth the geographic nuances and insights embedded within spatial explanatory variables^42,43,44. Its core functions consist mainly of five components: the Factor Detector, the Parameter Optimization Detector, the Interaction Detector, the Risk Detector, and the Ecological Detector. The Factor Detector, as the central part of the Geographic Detector, uses the q-statistic to reveal the relative importance of explanatory variables. The q-statistic compares the discrete variance between the observed values across the study area and the variable layer⁴⁵. The expression for the q-statistic is:

$$Q_{v} = 1 - \sum\limits_{{j = 1}}^{M} {(N_{{v,j}} - 1)} \sigma _{{v,j}}^{2} /(N_{V} - 1)\sigma _{v}^{2}$$

(8)

In the formula: $N_{\nu }$and $\sigma _{v}^{2}$ represent the number of observations and the variance within the entire study area, respectively, while $N_{{v,j}}$ and $\sigma _{{v,j}}^{2}$ denote the number of observations and the variance within the j-th subregion for the variable v. A higher Q value indicates a relatively greater importance of the explanatory variable, as it suggests a smaller variance within subregions and a larger variance between subregions. In the Geographic Detector framework, at least two samples from each stratum are required to calculate the mean and variance.

To minimize subjectivity in spatial discretization, the OPGD model optimized stratification parameters via grid search. For continuous variables, the natural breaks algorithm achieved the highest q-values, whereas categorical variables were stratified based on predefined ecological thresholds.

Results

Remote sensing ecological index

This study utilizes a comprehensive framework encompassing four distinct indices to evaluate the habitat quality of the region, thereby categorizing it into five distinct tiers: substandard, somewhat inadequate, moderate, commendable, and outstanding(Fig. 3). The expanse characterized by substandard habitat quality spans 596,964 km²representing an overwhelming 56.53% of the total study area, predominantly concentrated in the central and eastern sectors endowed with elevated levels of both heat and aridity indices. Meanwhile, regions boasting somewhat inadequate and moderate habitat quality cover 218,152 km² and 128,013 km²respectively, collectively accounting for 33.91% of the entire area, intricately interwoven within the northern, western, and southern fringes of the research scope. Areas distinguished by commendable and outstanding habitat quality are restricted to a mere 69,261 km² and 28,315 km²constituting a minor 6.79% and 2.77% of the overall study area, respectively, predominantly found in the northern expanse where verdancy and moisture indices peak at notably high thresholds.

Fig. 3

Analysis of habitat quality.

Single-factor ecological sensitivity analysis

Remote sensing ecological index sensitivity

Based on the analysis of MODIS data pertaining to greenness, humidity, dryness, and heat within the study area(Figs. 4 and 5). The greenness-insensitive areas account for 68.82%, covering 702,433 km², primarily concentrated in the central and eastern deserts of the Tarim River Basin. Extremely low vegetation coverage, with NDVI values below 0.1, and hyperaridity suppress ecological responsiveness in these regions. Highly sensitive and extremely sensitive greenness areas, occupying 4.74% and 4.11%, respectively, are distributed in northern and western oases, where human activities such as irrigation intensify vegetation dynamics. Humidity-insensitive zones span 36.80% (375,605 km²), corresponding to desert interiors with annual precipitation below 50 mm, while extremely sensitive humidity areas (2.03%) cluster in oasis interiors due to localized moisture anomalies caused by canal leakage. Dryness exhibits marked polarization, with highly sensitive regions dominating 60.66% (619,116 km²) of the central deserts. These areas are characterized by high surface albedo exceeding 0.4 and near-zero soil moisture, driven by desertification feedback. Extremely sensitive dryness zones (0.13%, 1,301 km²) correspond to salt-encrusted depressions or wind-eroded corridors surpassing vegetation tolerance thresholds. Heat sensitivity peaks in desert cores, where extremely sensitive areas cover 50.62% (516,684 km²), with surface temperatures frequently exceeding 50 °C. Albedo-temperature feedback and photovoltaic infrastructure exacerbate thermal accumulation, while irrigation in transitional oasis-desert zones mitigates sensitivity.

Fig. 4

The polarized sensitivity patterns—low greenness and humidity versus high dryness and heat—reflect the interplay between natural drought stress and anthropogenic activities. Desert cores exhibit near-collapse ecological functionality, with dryness sensitivity exceeding 60%, while groundwater overexploitation at oasis margins creates a paradox of localized mitigation amid systemic degradation. The dominance of highly sensitive dryness areas (60.66%) signals the crossing of irreversible desertification thresholds, necessitating prioritized vegetation-based sand fixation and water management. For extreme heat-sensitive zones (50.62%), surface cover modulation, such as drought-tolerant vegetation restoration, is critical to disrupt thermal feedback loops. Methodological robustness is confirmed by a total percentage error below 0.1% across sensitivity categories, supporting spatially targeted governance strategies to guide ecological restoration in arid regions.

Fig. 5

Sensitivity analysis of Remote Sensing Ecological Index.

Topographic sensitivity

The expanse of the research territory is predominantly cloaked by arid deserts and barren wastelands, leading to a prevalence of insensitivity within the distribution of elevation sensitivity, encompassing a staggering 62.49% of the total research area, primarily concentrated within the two vast desert regions situated in the heartland and eastern reaches. The remaining categories occupy 37.51% of the area and form an arc tracing the northern, western, and southern fringes of the research domain. Analysis of slope sensitivity reveals that the unresponsive region spans a vast expanse of 715,306.7 km²accounting for a substantial 70.08% of the entire territory, while the highly sensitive terrain is scarce, constituting a mere 1.82% of the total expanse(Fig. 6). Examination of aspect sensitivity displays a nearly equitable dispersion among the five categories, with moderately sensitive zones prevailing in the north and exceedingly sensitive regions in the south, while the other categories exhibit an even distribution across the central and surrounding areas.

Fig. 6

Results of topographic sensitivity analysis.

Environmental factor sensitivity

The spatial distribution of sensitivity to environmental factors reveals striking patterns in the study area(Fig. 7). The analysis demonstrates that a vast expanse spanning 605,029.5 km²representing 59.28% of the total study area, exhibits nominal responsiveness to precipitation. This area is primarily concentrated in the arid central and eastern desert regions where rainfall is exceedingly scarce. Conversely, areas of pronounced sensitivity to precipitation are localized in the northern and western peripheries, encompassing a mere 3.26% of the total area. The distribution of temperature sensitivity presents a contrasting pattern, with insensitive regions occupying the fringes of the study area and highly sensitive zones clustered in the central core. Specifically, insensitive areas cover 55.66% of the total area, while areas of extreme sensitivity comprise 6.37%. Water sensitivity patterns mirror those of aspect sensitivity, with each category evenly dispersed along the rivers within the study area. The largest segment pertains to insensitive areas, constituting 21.75% of the total area, while the smallest consists of extremely sensitive areas, accounting for 19.03%. In terms of vegetation type sensitivity, insensitivity prevails, encompassing a substantial 73.36% of the total area. By contrast, areas of extreme sensitivity are notably scarce, covering merely 10,260 km²equivalent to a mere 1.01% of the total study area.

Fig. 7

Results of environmental factor sensitivity analysis.

Human activity sensitivity

The findings of the soil type sensitivity analysis reveal that the areas exhibiting insensitivity are notably predominant, encompassing a staggering 57.59% of the overall expanse, whereas the areas depicting extreme sensitivity are conspicuously scarce, occupying a mere 2.98%. Land use sensitivity distributions illustrate that the mildly sensitive regions reign supreme, constituting a noteworthy 69.4% of the total research domain, with the insensitive zones being relatively negligible, amounting to a mere 0.31%, scattered sporadically along the outskirts of the study area. With regards to human activity sensitivity, it is evident that road proximity notably embodies heightened susceptibility, accounting for a substantial 25.09% of the total area, while the areas that are deemed insensitive are strikingly sparse(Fig. 8). Conversely, in assessing the sensitivity to residential proximity, it becomes apparent that the moderately sensitive areas are the most extensive, whilst the insensitive regions remain relatively inconsequential in comparison.

Fig. 8

Results of sensitivity analysis on human activities.

Results of the optimal parameter geographic detector analysis

Principal component analysis was utilized to investigate the degree to which each autonomous variable impacted the variability in ecological responsiveness. The outcomes of the principal component analysis encompass q-values, levels of significance, and rankings of variables(Fig. 9). The autonomous variable boasting the most elevated recorded q-value (in comparison to other explanatory variables) is identified. The visual representation reveals that heat (X4) exhibits the highest q-value (0.731), closely trailed by temperature (X10) with a q-value of 0.7045. The explanatory potential of these two autonomous variables surpasses 70%, underscoring their prominence as the primary factors shaping the fluctuations in ecological responsiveness. The dominance of heat (X4) and temperature (X10) in driving ecological sensitivity can be mechanistically attributed to the hyperarid conditions of the Tarim River Basin. Prolonged heat stress exacerbates evapotranspiration, depletes soil moisture reserves, and disrupts vegetation-water feedback loops, thereby amplifying ecosystem vulnerability. Temperature (X10) further modulates hydrological processes by accelerating glacial melt and altering precipitation patterns, which directly affect water availability—a critical limiting factor in arid ecosystems. Importantly, synergistic interactions between heat and land use (X13) were observed, indicating that anthropogenic land modifications may amplify thermal impacts on ecological sensitivity. These findings align with recent studies on arid zone ecohydrology, confirming that thermal-hydrological coupling is a pivotal mechanism governing ecological sensitivity in this region.

The results demonstrated that the heat (X4) exhibited a q-value of 0.731, while the temperature (X10) yielded a q-value of 0.7045, indicating their independent explanatory power for ecological sensitivity fluctuations reached 73.1% and 70.5%, respectively. Notably, the interaction between heat and land use X13 achieved a q-value of 0.82, revealing an 18.3% sensitivity enhancement due to their synergistic effects. Spatially, the core desert regions exhibited a thermal sensitivity contribution rate of 89.2%, where a 1 °C rise in surface temperature increased the sensitivity index by 0.23, a relationship validated with statistical significance.

The spatial heterogeneity of ecological sensitivity drivers, quantified by q-values, highlights distinct mechanistic pathways. High explanatory power (q > 0.5) in greenness (X1, q = 0.68) reflects vegetation-mediated soil stabilization and microclimate regulation, critical in arid zones where biomass loss amplifies desertification. Vegetation diversity (X11, q = 0.61) enhances resilience through niche complementarity, optimizing water-use efficiency under fluctuating precipitation. Soil types (X12, q = 0.58) dictate hydraulic conductivity, with loamy soils buffering hydrological extremes, while sandy soils exacerbate moisture deficits. Land use (X13, q = 0.55) captures anthropogenic impacts, as irrigation-induced groundwater depletion elevates salinization risks. Humidity (X2) and precipitation (X8) jointly regulate water availability, the linchpin of hyperarid ecosystems. Conversely, road distance (X14, q = 0.0218) shows negligible influence due to spatial-scale mismatches (localized impacts diluted at basin scale) and proxy inadequacy (distance fails to capture traffic intensity or pollutant load). Interaction analysis reveals roads synergistically amplify land use pressures (X13×X14 q = 0.71), underscoring latent anthropogenic cascades.

Fig. 9

Results of driving force factor analysis.

In the findings of the Geographic Detector’s interaction analysis, the evaluation of whether the combined impact of two distinct independent variables on the dependent variable is enhanced or diminished is predominantly determined by a comparison of the q-values of these variables in isolation versus in conjunction(Fig. 10). As illustrated in the diagram below, the most pronounced interaction is observed between X13 (land utilization) and X4 (thermal influence), manifesting as a nonlinear decline, indicative of a significant influence of land usage on temperature within a specific range. Additionally, X13 (land utilization) exhibits robust interactive relationships with X2 (moisture levels), X5 (altitude), X9 (ambient temperature), and X12 (soil composition). Remarkably, X2 (Moisture levels) exerts a compelling impact on nearly all independent variables, with explanatory power ranging from 0.503 to 0.696. Conversely, X7 (aspect) displays the weakest interactive correlation among the remaining independent variables. In comparative terms, factors such as X6 (slope), X7 (aspect), X14 (distance to road), and X15 (distance to residential areas) exhibit a limited independent explanatory capacity in the assessment of ecological sensitivity. However, their influence on ecological sensitivity is markedly amplified when they interact with other independent variables. The analytical findings further reveal that the explanatory power of X11 (vegetation type) and X9 (temperature) stands at 0.657, which is equivalent to that of X12 (soil type) and X11 (vegetation type). This suggests that vegetation type holds equal significance to both soil type and temperature. A more nuanced investigation is warranted to elucidate the intricate interactive relationships among vegetation type, temperature, and soil type. Additionally, it is plausible that a robust interaction or synergistic effect exists among these variables; their combined influence may not substantially alter the explanatory power, which appears to remain at a relatively stable level.

Fig. 10

Sensitivity driven factor interaction detection results s.

Discussion

Disparities in environmental quality are the result of both natural forces and human interventions, with notable repercussions stemming from factors such as drought, overgrazing, and the expansion of agricultural activities in the central and eastern regions. Particularly striking is the substantial portion of land exhibiting acute vulnerability to extreme temperatures, encompassing as much as half of the total region, thus underscoring the pervasive influence of high heat on the basin’s ecological landscape. The scarcity of areas resilient to temperature fluctuations further emphasizes the pressing need for concerted efforts to combat climate warming. Collectively, insights gleaned from diverse facets of the natural milieu present a nuanced depiction of ecological fragility within the Tarim River Basin, furnishing vital evidence to inform targeted conservation strategies. Moreover, an in-depth analysis has pinpointed the primary drivers fueling shifts in ecological vulnerability across the Tarim River Basin, where elevated q values underscore the pivotal role played by extreme climatic conditions in shaping ecosystem dynamics within the basin. Heat not only directly impacts the growth and distribution of vegetation but also exerts a profound influence on the hydrological cycle, exacerbating the severity of drought and ecological susceptibility in the area. Temperature indirectly influences shifts in ecological vulnerability by modulating critical hydrological processes, including glacial meltwater and precipitation. Consequently, amidst the backdrop of escalating global climate perturbations, the assessment and mitigation of heightened drought risk and ecological vulnerability in the Tarim River Basin arising from elevated temperatures has emerged as a focal concern.

In order to tackle this issue, it is imperative to establish a comprehensive and enduring ecological monitoring framework to consistently evaluate the impact of elevated temperatures on vegetation and the hydrological cycle. For instance, the establishment of thermal regulation priority zones and land-use zoning controls should be prioritized. Given the dominant role of heat (X4) in ecological sensitivity, it is recommended to implement thermal buffer zones in glacier ablation areas of the northern basin. These zones could mitigate surface temperature through targeted vegetation restoration (e.g., drought-resistant shrub plantations), thereby decelerating glacial retreat rates and stabilizing local microclimates. Furthermore, leveraging the synergistic interaction between land-use type (X13) and heat (interaction q-value: 0.82), agricultural expansion into high-sensitivity regions (e.g., desert-oasis transition belts) must be strictly regulated. Concurrently, irrigation efficiency enhancement zones should be delineated to promote drip irrigation technologies, which reduce groundwater over-extraction while maintaining agricultural productivity. Such spatially explicit management strategies align with the basin’s hydroclimatic constraints and anthropogenic pressures, offering a scalable framework for arid ecosystem governance. Simultaneously, expediting the digitization efforts within the Tarim River Basin and expanding the utilization of cutting-edge digital twin basin technologies, such as the pioneering Digital Twin Ta’er River (Pilot) Project spearheaded by the Bayingolin Management Bureau in the Tarim River Basin, has ushered in a new era of digital, interconnected, and intelligent oversight. This transformative approach has significantly augmented our capacity for hydrological surveillance, early flood detection, and precision water resource planning optimization. By harnessing diverse algorithms to mimic and replicate data sourced from sensor devices, generating virtual counterparts that mirror the physical world in a 1:1 ratio, this technological breakthrough not only enhances data precision but also streamlines operations, reducing human intervention while enhancing efficiency. This systematic approach renders water resource management more scientific, exacting, standardized, and transparent.

The coarse resolution of MODIS data (500 m) likely homogenized sensitivity gradients within fragmented oasis-desert matrices, masking critical thresholds in ecotone degradation. For instance, small-scale wetlands vital for biodiversity were misclassified as “low sensitivity” due to spectral mixing. Similarly, precipitation interpolation errors in the western basin may have inflated the perceived dominance of heat (X4) over water-related drivers. These limitations underscore the necessity of multi-sensor fusion (e.g., Sentinel-2 SWIR bands) to disentangle compound stressors in heterogeneous arid landscapes.

Conclusion

Through a thorough examination of the ecological sensitivity of the Tarim River Basin, coupled with an incisive analysis of the underlying driving forces, this study illuminates the intricate challenges confronting the basin’s ecosystem and elucidates the interconnected mechanisms of various influencing factors. Encompassing elements such as topography, climatic conditions, water resources, vegetation cover, and anthropogenic activities, a multidimensional approach reveals that the ecological sensitivity of the Tarim River Basin is a highly integrated and dynamically evolving phenomenon. Its formation and evolution are the inevitable consequences of the synergistic interplay between natural factors and human influences. The following conclusions have emerged from this investigation:

• (1) By employing an analysis of the remote sensing ecological index, the habitat quality within the basin was stratified into five distinct classifications. The findings reveal that the central and eastern regions, characterized by elevated heat and aridity, exhibit the most compromised habitat quality (56.53%). In contrast, the peripheral areas to the north, west, and south display moderate to fair habitat quality (33.91%). Notably, the northern regions, distinguished by high levels of greenness and humidity, are associated with good habitat quality (9.56%). This analysis underscores the pronounced disparities and spatial distribution of ecological sensitivity throughout the watershed, thereby establishing a robust scientific foundation for future efforts in ecological restoration and conservation. It is imperative to designate high-temperature zones within the central desert regions as priority areas for vegetation restoration, emphasizing drought-resistant shrubs to mitigate surface temperatures and stabilize soil moisture through enhanced evapotranspiration regulation and root-mediated hydrological retention. Concurrently, agricultural expansion must be restricted in ecologically sensitive hotspots, such as desert-oasis transition zones, where anthropogenic activities exacerbate ecosystem vulnerability. Implementation of drip irrigation technologies in existing croplands is critical to curb groundwater depletion, with pilot studies demonstrating a 25–40% reduction in water extraction rates under optimized irrigation schedules.

• (2) Subsequent analysis of remote sensing ecological indices reveals that the basin exhibits a general insensitivity to greenness and humidity. However, it demonstrates a pronounced sensitivity to drought and elevated temperatures, with regions characterized by extreme heat sensitivity comprising over half of the total area. This underscores the ecological vulnerability of the basin and highlights the urgent necessity to address the challenges posed by drought and climate change..

• (3) The topographical sensitivity analysis reveals that the extensive arid regions engender a pervasive indifference in altitude and gradient, mirroring the intricate influence of terrain on ecological receptivity.

• (4) Among the myriad environmental factors, sensitivities to precipitation and temperature display an inverse relationship, whereas the sensitivities associated with aquatic bodies and vegetation types epitomize the intricate heterogeneity of ecological variable .

• (5) In the realm of human endeavors, soil types exhibit a predominance of insensitivity, while land use shows a mild sensitivity. Road distance, on the other hand, emerges as highly sensitive, with residential distance falling into the category of moderate sensitivity. These findings underscore the intricate and varied impacts of human activities on the environment.

• (6) The examination of the principal driving factors reveals that heat and temperature emerge as the predominant influencers in the fluctuations of ecological receptivity (accounting for more than 70%), amplified further by the arid climatic conditions that accentuate the impact of surface thermal energy and temperature on hydrologic patterns and botanical composition. Verdancy, botanical categorizations, and additional variables exhibit noteworthy impacts, although the effect of proximity to thoroughfares remains comparatively modest. The results demonstrated that the heat factor X4 exhibited a q-value of 0.731, while the temperature factor X10 yielded a q-value of 0.7045, indicating their independent explanatory power for ecological sensitivity fluctuations reached 73.1% and 70.5%, respectively. Notably, the interaction between heat and land use X13 achieved a q-value of 0.82, revealing an 18.3% sensitivity enhancement due to their synergistic effects. Spatially, the core desert regions exhibited a thermal sensitivity contribution rate of 89.2%, where a 1 °C rise in surface temperature increased the sensitivity index by 0.23, a relationship validated with statistical significance.