Next Article in Journal
PAFNet: A Parallel Attention Fusion Network for Water Body Extraction of Remote Sensing Images
Previous Article in Journal
Biased Aerosol Wet Deposition CAM5 Simulations: A Result of Misrepresented Convective-Stratiform Precipitation Partitioning When Benchmarked Against SPCAM
Previous Article in Special Issue
Assessing the Performance of Multiple Satellite-Based Evapotranspiration Models over Tropical Forests
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 18
Issue 1
10.3390/rs18010152
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Combining In Situ and Remote-Sensing Data to Assess the Spatial Pattern and Changes of Major Grassland Types in Xinjiang, China, Under Climate Change Scenarios

by
Jin Zhao
1,2,*,
Kaihui Li
1,3,
Qianying Shao
1,2,
Jie Bai
1,2,
Yanming Gong
1,3 and
Yanyan Liu
1,3
1
State Key Laboratory of Ecological Safety and Sustainable Development in Arid Lands, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Urumqi 830011, China
2
Xinjiang Key Laboratory of RS&GIS Application, Urumqi 830011, China
3
Bayanbulak Alpine Grassland Observation and Research Station of Xinjiang, Urumqi 830011, China
*
Author to whom correspondence should be addressed.
Remote Sens. 2026, 18(1), 152; https://doi.org/10.3390/rs18010152
Submission received: 22 September 2025 / Revised: 22 December 2025 / Accepted: 26 December 2025 / Published: 3 January 2026
(This article belongs to the Special Issue Remote Sensing Monitoring and Assessment of Forest, Grassland, Wetland and Urban Ecosystem)
Browse Figures
Versions Notes

Highlights

What are the main findings?
  • The total area of these nine grassland types showed a declining trend from 2041 to 2100. Among them, the areas of Alpine Meadow (AM), Alpine Steppe (AS), Temperate Steppe (TS), and Temperate Desert (TD) experienced the most significant decrease.
  • Grassland types in China are defined by precipitation, humidity, and community composition. This study showed that temperature and environmental variables play a crucial role and significantly contribute to the predicted distribution of grassland types.
What are the implications of the main findings?
  • The decline in grassland areas in high-altitude mountainous regions, such as AM, AS, and TS, indicates that climate change has a significant impact on grasslands in the mid-to high-altitude areas of Xinjiang.
  • The existing classification system for grassland types in China should take more fac-tors into account, such as temperature and other relevant parameters.

Abstract

Examining the long-term spatiotemporal distribution of grassland types and their transitions is crucial for better understanding regional and global changes. Most research in this field has examined the spatial distribution, temporal dynamics of grasslands, and their causes as a unified entity. This study predicted the distribution of nine major grassland types in Xinjiang under three climate change scenarios from 2041 to 2100 based on 1980s grassland maps, field data in 2023, and 28 factors. The total area of the nine grassland types showed a decreasing trend from 2041 to 2100. The lowland meadow (LM), temperate meadow steppe (TMS), temperate steppe desert (TSD), temperate desert steppe (TDS), and mountain meadow (MM) expanded, while significant declines occurred in alpine meadow (AM), alpine steppe (AS), temperate desert (TD), and temperate steppe (TS). Among cumulative contribution rate of the 28 factors examined in this study, NDVI, vegetation type, slope, elevation, soil_symbol, soil_ph, Bio1, Bio5, Bio8, Bio9, Bio10, Bio12, Bio13, Bio15, and Bio18 played important roles in most grassland types. LM, TD, and AS grassland were found to be more sensitive to E (environment), while AM, TDS, and TSD were more influenced by T (temperature). The distributions of MM and TMS are significantly influenced by the combined effects of all three categories of factors. For TS, the impacts of both temperature and environmental factors are substantial. These findings provided a robust foundation for conservation planning and the sustainable management of grassland ecosystems in temperate and alpine regions.

1. Introduction

Grassland makes up the largest proportion of China’s terrestrial ecosystems, accounting for approximately 27.8% of the country’s land area (https://www.forestry.gov.cn/search/622296, (accessed on 23 April 2025)). This ecosystem provides a wide range of ecosystem service functions that contribute to the prevention of soil erosion [1], groundwater recharge [1], the regulation and maintenance of biodiversity [2,3], food production [4], recreation, and tourism [5,6,7]. Despite their previously mentioned functions and significance, grasslands are currently undergoing changes due to climate variations and unsustainable human activities such as agriculture and excessive grazing [8,9]. Any alterations in the extent of grassland or their ability to provide ecosystem services will result in substantial societal consequences [10]. Consequently, elucidating the long-term spatial distribution and transition patterns of diverse grassland categories is critical for improving the understanding of environmental change at regional and global scales [11].
Most research in this field has examined the spatial distribution, temporal dynamics of grasslands, and their causes as a unified entity [3,10,12,13]. A minority of studies have also focused on the extraction and analysis of changes in grassland types by examining the diverse plant populations of grasslands and their combinations in different habitats based on satellite imagery and the phonological-topographic features of a few grassland types [11,14]. Some studies have also examined the biomass [15] and carbon sequestration capacity [16] caused by changes in grassland types. However, research on the comprehensive distribution patterns and spatial transitions of various grassland types remains limited. Furthermore, only the current and historical status of grassland types can be discerned based on satellite imagery, while predicting future conditions remains beyond current technical capabilities.
Climate change is a key factor altering the distribution of vegetation and grassland species [17,18]. Climatic factors have a significant impact on ecosystem water and heat dynamics, leading to significant alterations in the species composition, GPP and community structure of ecosystems [19], poses a serious hazard to the biodiversity of desert ecosystems [20]. On the other hand, climate change based on CMIP6 also strengthens vegetation resilience to compound dry-hot events through modifications in ecosystem patterns, thereby indirectly enhancing sandstorm prevention services [21,22,23].
Extensive research has demonstrated that the main factors influencing the status of grassland vegetation include precipitation, temperature, and additional climatic factors [24,25]. Moreover, geomorphological differences can alter the habitat conditions for vegetation both directly and indirectly, thereby influencing the growth and distribution of vegetation. It is, therefore, crucial to investigate the relationships between climatic variables, geomorphological landforms [26], and the distribution of different types of grassland, as well as their correlations with environmental variables. Grassland types can be predicted based on climate features and basic vegetation characteristics while considering the terrain, soil, and spectral characteristics. It is necessary to conduct grassland research over a long time series to further investigate the spatiotemporal characteristics of grassland, analyze current environmental stressors on grassland ecosystems, and project future trends of grassland ecological health [25].
Widely utilized species distribution modeling approaches include the DOMAIN model [27], boosted regression tree (BRT) [28], BIOCLIM [29,30], generalized linear model (GLM) [31], maximum entropy (MaxEnt) [32,33,34], random forest (RF) [35], MARS [36], k-nearest neighbor (K-NN) models [37,38], ENphylo and Environmental Niche Factor Analysis (ENFA) [39], and support vector machine(SVM) [40]. The strengths and limitations of the model are presented in Table 1. Among these models, the MaxEnt model demonstrates some of the best performance [41] due to its insensitivity to the absence of occurrence data and its ability to effectively avoid the risk of model overfitting by adjusting the regularization multiplier [42,43,44]. Overall, the MaxEnt plays an important role in wildlife conservation [45,46,47], biodiversity [48,49], prediction and quality zoning of potential suitable areas of crop [50,51], and species invasion [52].
The MaxEnt model has been applied to predict the distribution of individual or dominant grassland species [53,54], providing data and support for wildlife conservation and management. However, its application in predicting the distribution of relatively comprehensive grassland types has not yet been reported. Accurate prediction of grassland types at this comprehensive level would provide critical support for the development of differentiated grassland type resource management strategies and for promoting sustainable regional livestock production.
However, there are many classification criteria for the suitability zone of the MaxEnt model, which leads to significant differences in the size of the suitability zone. For example, using the natural breaks (Jenks) option in the ArcGIS reclassification tool, studies grouped potential distribution probability value ranges 0 to 100 into three levels: “high potential” (>0.7), “moderate potential” (0.36–0.7), and “least potential” <0.35 [55]; four levels: “High Suitability” (>0.75), “Moderate Suitability” (0.5–0.75), “Low Suitability” (LS) (0.25–0.5), and “Unsuitable” (025–0.25) [3]; as well as “high potential” (>0.6), “good potential” (0.4–0.6), “moderate potential” (0.2–0.4), “least potential” (<0.2), and no risk [56,57,58]. The area of the suitability zone differs with the natural breaks value. The area of the suitability zone varies with changes in the natural breaks value, making it difficult to determine the specific distribution and area of species, which poses challenges for management and policy formulation. To enhance the accuracy of the predictive models, the threshold values for each grassland class must be verified using a method that can identify the area closest to both the actual data and the predicted results. The comprehensive analysis of these threshold values is essential to ensure the reliability and precision of the predictions. This approach is employed in this study to minimize discrepancies between the observed and projected outcomes, thereby refining the accuracy of the predictions. This meticulous technique not only enhances the precision of predictions but also contributes to a more robust understanding of spatial patterns within different grassland classes.
The objectives of this study were as follows: (1) to determine the primary influencing factors of various grassland types in Xinjiang; (2) to establish the threshold value of the MaxEnt model for predicting the probability of different grassland types; and (3) to analyze the spatial and temporal changes of future grassland classes and further explore future challenges.

2. Materials and Methods

2.1. Study Area

The Xinjiang Uygur Autonomous Region (Xinjiang) is situated in an arid zone of Northwestern China, characterized by a distinctive basin-and-range topography. This landscape is defined by the Altai Mountains to the north, the Kunlun Mountains to the south, and the Tianshan Mountains transecting the center, which collectively enclose the expansive Tarim and Junggar basins. Climatically, Xinjiang exemplifies an arid-to-semiarid environment, with an average annual temperature ranging from 9 to 12 °C. Precipitation exhibits a strong north–south gradient, averaging about 150 mm per year [59]. Based on climatic characteristics (heat) and the basic characteristics of vegetation, and fully considering topography, soil, and economic factors, the grasslands are classified into 20 categories, building upon the original classification from National Inventory of Grassland Resources by the National Forestry and Grassland Administration of China. The Xinjiang grasslands were grouped into the following 11 types: alpine meadow (AM), alpine steppe (AS), alpine meadow–steppe (AMS), alpine desert (AD), lowland meadow (LM), mountain meadow (MM), temperate steppe (TS), temperate desert (TD), temperate desert–steppe (TDS), temperate steppe–desert (TSD), and temperate meadow–steppe (TMS). The grassland categories are shown in Table 2. In 2023, there were only 16 and 30 field sites for AD and AMS, respectively, making it unsuitable to use these numbers to predict the growth of these two types of grasses. The other nine grassland types were included in this study.

2.2. Field Sites

The variation in grassland types is a long-term process. The data comprised vector maps of grassland type in the 1980s, which were utilized to generate grassland type vector points to construct a predicted grassland type map for the 1980s. These data in the 1980s come from China’s first grassland resource survey, conducted over a 10-year period from 1979 to 1990. It is used as key data for MaxEnt modeling and threshold determination, representing grassland types from 1980 to 2010 (a 30-year period). The predicted grassland-type map was compared with the actual grassland-type map, and the probability threshold that best aligned with the actual area was determined for optimal prediction. The sampling data in 2023 represent grassland-type point data from the 30-year period between 2010 and 2040. There were 6013 field sites in the 1980s and 4385 in 2023 across nine grassland types (Figure 1). Based on this, predictions of grassland type changes for the next two 30-year periods are made (2041–2071, 2071–2100).

2.3. Factors

Survey points were integrated with shared socioeconomic pathway (SSP) climate data to forecast the future spatial distribution of grassland types. Climate factors were selected from bio1 to bio19 (Table 3). Terrain factors included slope, aspect, and elevation information obtained from the geospatial data cloud at http://www.gscloud.cn/, (accessed on 10 October 2023). Soil factors consisted of the soil pH value (Soil_ph) and soil type (soil_symbol) derived from global soil pH datasets, as well as China’s comprehensive map detailing various soil types. The latter is a detailed map compiled by the Nanjing Institute of Soil Science under the Chinese Academy of Sciences that encompasses an array of 72 distinct soil classes and 247 subclasses. There are 23 unique soil types in Xinjiang. The landform-type (Geomor) data used in this study pertained specifically to the Xinjiang region, and the river data measured the distance from river channels with an adjacent buffer zone of 1 km. The normalized difference vegetation index (NDVI) data employed in this study consisted of the maximum NDVI dataset for Xinjiang covering the period of 2000–2020. NDVI data processing were performed on the Google Earth Engine platform using Landsat 5/7/8 remote-sensing data. These data had a spatial resolution of 30 m. Cloud and shadow removal were performed prior to NDVI extraction utilizing linear interpolation and S-G smoothing methods. All data were resampled to 90 m resolution of the Digital Elevation Model (DEM). NDVI (30 m resolution) was resampled to 90 m using the average value method. Meteorological, soil, and other data were interpolated and resampled to 90 m based on measurements from ground stations and survey points.

2.4. Future Climate Factor Data

Future climate factor data for this study were derived from CHELSA CMIP6 scenario data, a high-resolution climate database for global land surface areas (https://chelsa-climate.org/, (accessed on 15 October 2023)). This dataset has a spatial resolution of 30 arc-seconds (approximately 1 km × 1 km) and includes bioclimatic variables projected for two future time periods: 2041–2070 and 2071–2100. The future climate scenarios utilized in this study are based on the SSP, which represent different socioeconomic development trajectories and their implications for greenhouse gas emissions and climate change impacts. SSP126 represents a sustainable development pathway with low greenhouse gas emissions, aiming for a radiative forcing of 2.6 W/m2 by 2100; SSP370 represents a middle to high-end emission scenario, leading to moderate levels of warming with a radiative forcing of 7.0 W/m2 by 2100; and SSP585 represents a high-emission scenario with extensive fossil fuel use, resulting in high greenhouse gas concentrations and a radiative forcing of 8.5 W/m2 by 2100. These climate scenarios provide essential data for assessing potential future impacts on ecological systems, considering varying levels of climate change and greenhouse gas emissions.

2.5. MaxEnt Model

The MaxEnt method is a sophisticated machine-learning technique based on the principle of maximum entropy that employs species occurrence data and environmental factors [60]. The calibration phase of the model is to determine which parameter combination best represents the phenomenon of interest by identifying the combination that exhibits the best fit with the data [61,62]. The statistical significance was determined by the partial receiver operating characteristic (ROC) curve. Meanwhile, the “Percent variable contributions” and “importance” routines were selected to evaluate the relative importance of environmental variables. If the correlation coefficient between two variables exceeded 0.8 in absolute value, the variable with the lower contribution rate was removed. In this study, 75% of the total data were employed for the calibration model, and the remainder was utilized to evaluate the model predictions [63]. This work selected logistic regression as the model output format. The maximum number of background points was set at 10,000, and 10 replicates and cross-validation were employed to ensure the stability of the model. The average results and average accuracy are calculated. The area under the curve (AUC) value obtained from the ROC curve was utilized to evaluate the prediction accuracy of the MaxEnt model. The AUC can serve as an independent measure of the overall model precision that is not influenced by a specific threshold. The AUC ranges between 0.5 and 1.0, with stronger model predictive ability at higher AUC values. After modeling the climate suitability for the distribution of nine grassland types for 2041–2070 and 2071–2100, the change trend of grassland type was calculated using ArcGIS software v3.6.

3. Results

3.1. Prediction Accuracy and Threshold

Each grassland type is converted into binary based on its threshold value (Table 4), rather than using a fixed threshold. The 9 grassland types are then merged to form the Xinjiang grassland map. When two or more values exceed the threshold at a given location, the grassland type with the highest threshold value is selected as the grassland type. Threhold of all grassland types greater than 0.5. The minimum for LM and TD was 0.5, while the maximum for TMS was 0.74. The AUC values for all nine grassland types achieved excellent results, with all of values showing above 0.85. Most grassland types surpassed 0.9, demonstrating outstanding performance.

3.2. Comparison of the Actual and Predicted Grassland Area in the 1980s

In Northern and Southern Xinjiang, there were up to seven grassland-type gradients from high to low altitude: AM/MM, AS/TMS, TS, TDS, TSD, TD, and LM. In the 1980s, among the nine grassland types, the TD type had the largest area, accounting for 37.3% of the total area. LM and TDS followed, representing 13.83% and 9.9% of the total area, respectively. The TMS made up the smallest area, contributing only 2.23% of the total area. Other grassland types occupied between 5% and 8% of the total grassland area.
The predicted area on the map for the 1980s was smaller than the actual grassland distribution (Figure 2). The predicted area for LM reached 97% of its actual area in the1980s, while the predicted area for TDS only reached 58% of its actual area in the 1980s (Figure 2). The predicted area proportions for other grassland types were similar to the actual distribution of grassland types. The TD type had the largest predicted area, followed by LM grassland, with AM grassland in third place. The predicted area for MM and TS was below 80% of actual area, while other grassland types exceeded 80%.

3.3. Importance of Various Factors

First, the importance and contribution of all influencing factors are calculated and ranked (Figure 3). Then, the correlation coefficients of the indicators are calculated. Factors with correlation coefficients greater than 0.8 are retained only if they have a larger contribution rate. The final simplified variable combinations for each grassland type were shown (Figure 4), with zero indicating unused variables. Additionally, the variables categorize into three main categories (Table 3): temperature (T), precipitation (P), and environmental factors (E), integrating the overall importance and contribution of the three categories and quantifying the influence of temperature, precipitation, and environmental factors for each grassland type (Table 5).
Based on the two heatmaps (Figure 4), the contribution and importance of various environmental drivers for each grassland type are shown. The importance ranking for LM identified vegetation type, slope, and distance to river channel as the three most significant predictors. Correspondingly, the top three predictors in terms of contribution rate were vegetation map, slope, and soil type. For AS, the same three factors exhibited the highest values for both importance and contribution rate: elevation, vegetation type, and Bio1. This consistency, combined with the finding that AS represents a cold-tolerant type, suggests that temperature is a primary determinant of its distribution. The variable importance for TD highlighted vegetation type, Bio12, and elevation. In the contribution rate analysis, vegetation type, Bio12, and soil pH were most influential, with the latter indicating a potential association with alkaline soil conditions. For AM, the three most important variables were Bio10, elevation, and NDVI. The contribution rate analysis ranked Bio10, Bio12, and elevation highest, aligning with the ecological preference of AM for cold and humid high-mountain climates. Considerable discrepancy was observed between importance and contribution rankings for TDS. Variable importance was highest for Bio1, elevation, and Bio9, while the top contributors were Bio1, soil type, and NDVI. In contrast, TSD showed complete agreement between the two metrics. Both importance and contribution rate ranked Bio1, Bio13, and Bio5 as the top three factors. The rankings for MM differed substantially between the two evaluation metrics. Variable importance was dominated by Bio15, elevation, and Bio19, whereas the highest contribution rates were attributed to NDVI, Bio10, and Bio12. For TMS, the most important variables were Bio18, NDVI, and Bio15, while NDVI, Bio18, and Bio8 contributed most significantly. Finally, for TS, variable importance was led by Bio8, Bio9, and NDVI. The contribution rate analysis identified Bio8, NDVI, and Bio12 as the most influential predictors.
Among cumulative contribution rate of the 28 factors examined in this study, NDVI, vegetation type, slope, elevation, soil_symbol, soil_ph, Bio1, Bio5, Bio8, Bio9, Bio10, Bio12, Bio13, Bio15, and Bio18 played important roles in most grassland types. By contrast, the impacts of Bio 14, Bio16, Bio17, aspect, and landform were relatively minor (Figure 4).
Importance and contribution varied with grassland type (Table 5). Some grassland types, such as LM, TD, and AS grassland, were found to be more sensitive to E, while others, such as AM, TDS, and TSD, were more influenced by T. The distributions of MM and TMS are significantly influenced by the combined effects of all three categories of factors, with no single factor exhibiting absolute dominance. For TS, the impacts of both temperature and environmental factors are substantial.

3.4. Future Changes of Grassland

An analysis of changes in grassland area was conducted for mid-century (2041–2070) and late-century (2071–2100) periods relative to a baseline under three Shared Socioeconomic Pathways (SSP126, SSP370, and SSP585). The results indicated that climate change will drive divergent responses among grassland types, characterized by an overall expansion of LM, TMS, TDS, TSD, and MM, contrasted with a substantial contraction of AM, AS, TS, and TD. The total grassland area declined by 16–19% across all scenarios by the end of the century (Table 6).
Based on the projected area change rates (Table 6), TMS and MM exhibited a consistent and significant increase in area. TMS demonstrated the most pronounced expansion, with area increases exceeding 130% throughout the 21st century under all scenarios. The growth was most notable during the mid-century period (up to 162% under SSP370). MM also showed a robust increasing trend, particularly under the SSP370 scenario in the late-century period, where its area increased by 60.1%. AM and AS were highly vulnerable to future climate conditions, showing severe area losses, with projected area reductions of approximately 70% across all periods and scenarios, indicating high sensitivity to warming. TS and TD also declined substantially. The contraction of TD, the largest grassland type, was a primary contributor to the net loss in total grassland area. These types exhibited relative resilience, with moderate increases or minimal changes. LM and TDS areas increased moderately (10–30%), showing greater adaptability. TSD was the most stable, with area changes generally ranging from −5% to +8%.
Across the 2041–2071 and 2071–2100 periods, consistent declines in total grassland area were projected under the SSP126, SSP370, and SSP585 scenarios, with specific grassland types showing an increase. The largest projected grassland area was obtained under the SSP370 scenario, while the smallest was obtained for the SSP585 scenario. The overall patterns of expansion and contraction were consistent between the low-emission (SSP126) and high-emission (SSP585) scenarios; greater variability was observed under SSP585. The medium-emission scenario (SSP370) occasionally resulted in the highest growth rates for some meadow types (e.g., MM) by the late-century, yet the negative impacts on alpine grasslands remained severe.
The predicted AS grasslands are primarily distributed in the mountainous regions surrounding the Tarim Basin in southern Xinjiang at an altitude range of 3200–4700 m (Figure 5). It typically forms a vertical belt distribution in the subalpine and alpine zones of major mountain ranges such as the Tianshan, and Kunlun Mountains. MM is mainly found in Northern Xinjiang, in the middle mountain belts of the Altai Mountains (1400–2100 m), the northern slopes of the Tianshan, and Western Tianshan (1700–2400 m). It is predicted that MM will decrease in the Altai Mountains and increase in the Tianshan mountain. The MM mainly consists of tall grasses and miscellaneous grasses, forming typical mountain meadow types by high canopy cover, tall grass layers, and high productivity. LM grasslands are found in the lakes, flat river channels (with slope of <4), and river terminals. LM shows slight expansion from the original scope, showing little change under different climate patterns (Figure 6). AM grasslands are predominantly located in Altai, the Tianshan Mountains, Southern Xinjiang, and Eastern Xinjiang. The AM grasslands in Southern Xinjiang are projected to decrease significantly, with reductions also observed in Altai and Western Mountains. TDS grasslands are transitional types between temperate grasslands and desert grasslands. In addition to drought-resistant perennial bunchgrasses, these grasslands also contain large amounts of desert semi-shrubs, saltbushes, and drought-resistant shrubs, which can comprise 30–50% of the vegetation. In Northern Xinjiang, they span from some plains at the foot of the mountains to the middle and lower mountain belts. In Southern Xinjiang, they extend into the middle and subalpine zones. The area of TDS grasslands has expanded in Northern Xinjiang, including the Altai and Eastern Xinjiang. In Southern Xinjiang, predictions showed TDS grasslands shifting toward lower altitudes in the Southern Tarim region. TDS grasslands showed little change, with a slight increase based on its original spatial distribution. Similarly, there was little change in the area of TSD, with a slight increase. TSD is primarily found in the plains of Northern Xinjiang. There is also some sporadic distribution in mountain areas, but the areas are small. In the future, TSD moves eastward, with an increase in the Eastern Tianshan and a decrease in the Southern Tianshan. TMS grasslands are primarily concentrated in Northern Xinjiang, with an increase in their distribution area. This predicted expansion is particularly evident in Altai, Ili, and is mainly derived from MM grassland, farmland, and temperate grasslands. TS is widely distributed in mountain areas and is the dominant grassland type in the region. In Northern Xinjiang, they form a belt in the middle and lower mountain belts of each mountain range, while in Southern Xinjiang, they occur in the middle and subalpine zones. The distribution altitude increases from north to south and from west to east. The TS area decreases in the Tianshan, with an increase in the western part of the Tianshan. TD is widely distributed and represents a temperate desert type of grassland in Xinjiang, divided into plain and mountain types. The plain desert occurs in the alluvial–accumulation plains at the foot of the mountains, in valleys, and in basins throughout Xinjiang. It is predicted that in the future, TD will decrease in the northern basin of Xinjiang and increase in Southern Xinjiang.

4. Discussion

4.1. Accuracy and Importance of Factors

4.1.1. Accuracy

The application of MaxEnt models in the prediction of single or multiple species has been extensively investigated, including in plants and animals. However, few studies have explored its use in grassland ecosystems. The accurate mapping of grassland type is seen as challenging in heterogenous landscapes. In this study, utilizing large occurrence data, the model demonstrated high accuracy, with predictive performance that ranged from good to excellent. The high performance confirmed the applicability of the model. Desert vegetation cover is sparse, and the boundaries of different desert types are difficult to delineate. As the sample size decreases, the uncertainty increases, requiring higher probabilities to obtain accurate predictions. Conversely, with larger sample sizes, even lower probabilities can lead to accurate distribution predictions.
Furthermore, MaxEnt assumes that the absence of a species in a given location is due to unsuitable environmental conditions, which may not always be the case. This assumption can lead to overfitting or biased predictions when suitable habitats are not sampled [64]. MaxEnt performs well with small numbers of occurrence points, which is often the case in ecological studies [65]. However, MaxEnt models are sensitive to overfitting, especially when too many features are used or when the sample size is small [66]. Recent advances, such as regularization, have attempted to mitigate this issue, but it remains a challenge. Therefore, grassland types with few sample points were excluded from the simulation in this study to ensure the accuracy of the results. Additionally, the removal of correlations between indicators was necessary to prevent overfitting. The model does not explicitly account for spatial autocorrelation in the presence data, which may inflate prediction accuracy in clustered data [67].

4.1.2. Various Factors

The high importance of vegetation indices indicates the potential of remote-sensing data in grassland type prediction. Utilizing remote-sensing data with a higher spatiotemporal resolution is likely to enhance and refine the results of related research. Sentinel 2 images with a higher spatiotemporal resolution may prove an asset in the near future, although the data time series is still relatively short for the evaluation of changes in vegetation. Further research could integrate additional vegetation indicators, such as the leaf area index (LAI), soil moisture, and chlorophyll fluorescence, which are more sensitive and offer higher precision. In this study, the accuracy for desert areas was relatively low, while the prediction accuracy for meadow and steppe types was high. This study only included river distance as a variable without incorporating groundwater-related indicators. Climatic factors, such as precipitation and temperature, make key contributions to grassland dynamics and their distribution across the Xinjiang region. This study did not include anthropogenic disturbance, which is among the key drivers of grassland type. Future research will consider how to achieve quantitative assessment of anthropogenic factors like population density, road density, residential density, grazing density, GDP, and land-use type.

4.2. Future Spatial Distribution and Changes of Grassland

The vertical gradients of grassland types varied with altitude, with a minimum of three-to-four grassland type gradients and a maximum of seven grassland type gradients. These gradients consisted of AM/MM, AS/TMS, TS, TDS, TSD, TD, and LM grasslands, with elevation from high to low. Although the gradient pattern was consistent, the starting and ending elevation ranges differed between Northern and Southern Xinjiang.
It was evident from the initial findings that the predicted extent of certain grassland types was underestimated, indicating potential spatial repetition or clustering phenomena within these ecological communities. In the Altai region, the results show that AM, MM, LM, and TM grasslands are decreasing, while the extent of TSD is diminishing and the elevation of TD regions is rising. In the Ili region, the findings indicate that AM and TS are being converted into TS grasslands. In Southern Xinjiang, AM, AS, and TS are undergoing degradation. Part of the reason for this finding is the limited number of sampling points for AM and AS, in addition to the restrictive assumptions and poor generalization ability of the MaxEnt model. However, grassland plant in Southern Xinjiang is impacted by droughts and overgrazing stress caused by pastoral activities. As a result, alterations in climate conditions can exert a key effect on regional vegetation cover. The current study simulated potential future trends of grassland cover change in Xinjiang over the period of 2041–2100 for two climate change scenarios, highlighting the driving influence of climate change on vegetation cover. The results suggest that most types of grassland cover in Xinjiang will display decreasing trends instead of increasing trends from 2041 to 2100 under the three climate change scenarios.
LM is widely distributed across the entire region, mainly in low-flat areas such as river floodplains, wide valleys, lakeshores, and basin depressions. It develops primarily in areas with abundant underground water and river overflow. Therefore, it is less affected by climate change. The AS is a cold-tolerant type within the grassland group. AM develops under the cold and humid climate conditions of high mountain areas. It is widely distributed in the snow and ice zones of the Tianshan, Altai. Therefore, altitude and temperature are very important factors for AS and AM. When the temperature rises in the future, the suitable habitat area decreases. TD is primarily composed of saltbush and semi-shrubs, with a decrease in the northern deserts and an increase in the Southern Tarim Basin. TD is widely distributed in arid regions, making prior knowledge of vegetation types important. Besides environmental factors, both moisture and temperature play roles. Among the individual moisture factors, annual precipitation has a significant impact. TDS play a significant role in the grassland composition, forming a desert grassland vegetation type in association with drought-resistant bunch grasses. Shrubs rely on groundwater and river channel recharge. Due to low precipitation, its impact on this type of vegetation is limited. TS is perennial drought-tolerant herb, and its growth is affected by temperature rise. MM and TMS are mainly distributed in the more humid climate areas of Northern Xinjiang, with warming and humidification promoting the increase in area.
The results suggest a potential shift in biome dominance, with cold-adapted alpine grasslands undergoing drastic declines, while some meadow and desert steppe types expand. This imbalance poses substantial risks to biodiversity and ecosystem function. Consequently, conservation and management strategies must be tailored to the specific vulnerabilities and responses of different grassland types.

5. Conclusions

This study predicted the distribution of nine major grassland types in Xinjiang under climate change scenarios from 2041 to 2100 based on 28 factors, including meteorological variables (temperature and precipitation), environmental factors (soil, altitude, slope, aspect, and landform), and vegetation factors (NDVI and vegetation type maps). The total area of the nine grassland types showed a decreasing trend from 2041 to 2100. The areas of LM, TMS, TDS, TSD, and MM grasslands exhibited slight increases, while AM, AS, TD, and TS experienced significant declines. In Northern and Southern Xinjiang, there were up to seven grassland-type gradients from high to low altitude: AM/MM, AS/TMS, TS, TDS, TSD, TD, and LM grasslands. Among cumulative contribution rate of the 28 factors examined in this study, NDVI, vegetation type, slope, elevation, soil_symbol, soil_ph, Bio1, Bio5, Bio8, Bio9, Bio10, Bio12, Bio13, Bio15, and Bio18 played important roles in most grassland types. By contrast, the impacts of Bio 14, Bio16, Bio17, aspect, and landform were relatively minor. Some grassland types, such as LM, TD, and AS grassland, were found to be more sensitive to E, while others, such as AM, TDS, and TSD, were more influenced by T. The distributions of MM and TMS are significantly influenced by the combined effects of all three categories of factors, with no single factor exhibiting absolute dominance. For TS, the impacts of both temperature and environmental factors are substantial.

Author Contributions

J.Z.: Writing—original draft, Visualization, Methodology. K.L.: Co-first author, Resources, Investigation, Conceptualization. Q.S.: Software, Formal analysis. J.B.: Data curation. Y.G.: Investigation. Y.L.: Investigation. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by National Key Research and Development Program of China (grant number 2023YFF0805603) and the Tianchi Talent Program of Xinjiang Province, China. Xinjiang “Tianshan Talents” program project (Grant No. 2023TSYCCX0086) the General program of National Nature Science foundation of China (No. 42071141).

Data Availability Statement

All data were created in this study can be provided by the corresponding author upon reasonable request.

Conflicts of Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References

  1. Lŭ, L.; Zhao, Y.; Chu, L.; Wang, Y.; Zhou, Q. Grassland coverage change and its humanity effect factors quantitative assessment in Zhejiang province, China, 1980–2018. Sci. Rep. 2022, 12, 18288. [Google Scholar] [CrossRef] [PubMed]
  2. Tong, L.; Liu, Y.; Wang, Q.; Zhang, Z.; Li, J.; Sun, Z.; Khalifa, M. Relative effects of climate variation and human activities on grassland dynamics in Africa from 2000 to 2015. Ecol. Inform. 2019, 53, 100979. [Google Scholar] [CrossRef]
  3. Lu, L.; Sun, Z.; Qimuge, E.; Ye, H.; Huang, W.; Nie, C.; Wang, K.; Zhou, Y. Using remote sensing data and species-environmental matching model to predict the potential distribution of grassland rodents in the northern china. Remote Sens. 2022, 14, 2168. [Google Scholar] [CrossRef]
  4. Bi, X.; Li, B.; Zhang, L.; Nan, B.; Zhang, X.; Yang, Z. Response of grassland productivity to climate change and anthropogenic activities in arid regions of Central Asia. PeerJ 2020, 8, e9797. [Google Scholar] [CrossRef]
  5. Bonari, G.; Fajmon, K.; Malenovský, I.; Zeleny, D.; Holusa, J.; Jongepierová, I.; Kocárek, P.; Konvicka, O.; Uricár, J.; Chytry, M. Management of semi-natural grasslands benefiting both plant and insect diversity: The importance of heterogeneity and tradition. Agric. Ecosyst. Environ. 2017, 246, 243–252. [Google Scholar] [CrossRef]
  6. Czarniecka-Wiera, M.; Szymura, T.H.; Kącki, Z. Understanding the importance of spatial scale in the patterns of grassland invasions. Sci. Total Environ. 2020, 727, 138669. [Google Scholar] [CrossRef]
  7. Zhang, N.; Li, Z.; Feng, Y.; Li, X.; Tang, J. Development and application of a vegetation degradation classification approach for the temperate grasslands of northern China. Ecol. Indic. 2023, 154, 110857. [Google Scholar] [CrossRef]
  8. Bocksberger, G.; Schnitzler, J.; Chatelain, C.; Daget, P.; Janssen, T.; Schmidt, M.; Thiombiano, A.; Zizka, G. Climate and the distribution of grasses in West Africa. J. Veg. Sci. 2016, 27, 306–317. [Google Scholar] [CrossRef]
  9. Zoungrana, B.; Ouedraogo, B.; Yanogo, I. Potential impact of future climate change on grassland cover in Burkina Faso. Environ. Sci. Pollut. Res. 2024, 31, 57229–57241. [Google Scholar] [CrossRef]
  10. Schils, R.L.; Bufe, C.; Rhymer, C.M.; Francksen, R.M.; Klaus, V.H.; Abdalla, M.; Milazzo, F.; Lellei-Kovacs, E.; Berge, H.; Bertora, C.; et al. Permanent grasslands in Europe: Land use change and intensification decrease their multifunctionality. Agric. Ecosyst. Environ. 2022, 330, 107891. [Google Scholar] [CrossRef]
  11. He, P.; Shi, Y.; Ding, H.; Yang, F. Classification and Transition of Grassland in Qinghai, China, from 1986 to 2020 with Landsat Archives on Google Earth Engine. Land 2023, 12, 1686. [Google Scholar] [CrossRef]
  12. Zhang, R.; Liang, T.; Guo, J.; Xie, H.; Feng, Q.; Aimaiti, Y. Grassland dynamics in response to climate change and human activities in Xinjiang from 2000 to 2014. Sci. Rep. 2018, 8, 2888. [Google Scholar] [CrossRef]
  13. Tian, L.; Tao, Y.; Fu, W.; Li, T.; Ren, F.; Li, M. Dynamic Simulation of Land Use/Cover Change and Assessment of Forest Ecosystem Carbon Storage under Climate Change Scenarios in Guangdong Province, China. Remote Sens. 2022, 14, 2330. [Google Scholar] [CrossRef]
  14. Xu, D.; Chen, B.; Shen, B.; Wang, X.; Yan, Y.; Xu, L.; Xin, X. The classification of grassland types based on object-based image analysis with multisource data. Rangel. Ecol. Manag. 2019, 72, 318–326. [Google Scholar] [CrossRef]
  15. Xu, G.; Zhang, H.; Chen, B.; Zhang, H.; Innes, J.L.; Wang, G.; Yan, J.; Zheng, Y.; Zhu, Z.; Myneni, R.B. Changes in vegetation growth dynamics and relations with climate over China’s landmass from 1982 to 2011. Remote Sens. 2014, 6, 3263–3283. [Google Scholar] [CrossRef]
  16. Li, H.Q.; Li, F.; Xiao, J.F.; Chen, J.Q.; Lin, K.J.; Bao, G.; Liu, A.J.; Wei, G. machine learning scheme for estimating fine-resolution grassland aboveground biomass over China with Sentinel-1/2 satellite images. Remote Sens. Environ. 2024, 311, 114317. [Google Scholar] [CrossRef]
  17. Zhang, M.; Chen, A.; Xing, X.Y.; Yang, D.; Wang, Z.C.; Yang, X.C. Changes in grassland types caused by climate change and anthropogenic activities have increased carbon storage in alpine grassland ecosystem. Glob. Planet. Chang. 2025, 250, 104803. [Google Scholar] [CrossRef]
  18. Boonman, C.C.; Huijbregts, M.A.; Benitez-Lopez, A.; Schipper, A.M.; Thuiller, W.; Santini, L. Trait-based projections of climate change effects on global biome distributions. Divers. Distrib. 2022, 28, 25–37. [Google Scholar] [CrossRef]
  19. Luo, Y.; Gong, Y. Diversity of desert shrub communities and its relationship with climatic factors in Xinjiang. Forests 2023, 14, 178. [Google Scholar] [CrossRef]
  20. Parra-Tabla, V.; Albor-Pinto, C.; Tun-Garrido, J.; Angulo-Perez, D.; Barajas, C.; Silveira, R.; Ortiz-Diaz, J.J.; Arceo-Gomez, G. Spatial patterns of species diversity in sand dune plant communities in Yucatan, Mexico: Importance of invasive species for species dominance patterns. Plant Ecol. Divers. 2018, 11, 157–172. [Google Scholar] [CrossRef]
  21. Meng, N.; Yang, Y.Z.; Zheng, H.; Li, R.N. Climate change indirectly enhances sandstorm prevention services by altering ecosystem patterns on the Qinghai-Tibet Plateau. J. Mt. Sci. 2021, 18, 1711–1724. [Google Scholar] [CrossRef]
  22. Yao, H.B.; Wen, Z.M.; Zhang, T.Y.; Li, X.Y.; Wang, L.L.; Li, Z.W. Spatiotemporal pattern of GPP of grassland ecosystem in Northern China based on CMIP6. Res. Soil. Water Conserv. 2024, 31, 266–274. [Google Scholar]
  23. Yao, T.; Wu, C.H.; Yeh, P.J.-F.; Hu, B.X.; Jiao, Y.F.; Li, Q.F.; Niu, J. Widespread global enhancement of vegetation resistance to compound dry-hot events due to anthropogenic climate change. Ecol. Indic. 2025, 178, 113880. [Google Scholar] [CrossRef]
  24. Liu, Y.; Zhang, Z.; Tong, L.; Khalifa, M.; Wang, Q.; Gang, C.; Wang, Z.; Li, J.; Sun, Z. Assessing the effects of climate variation and human activities on grassland degradation and restoration across the globe. Ecol. Indic. 2019, 106, 105504. [Google Scholar] [CrossRef]
  25. Chen, C.; Li, G.; Peng, J. Spatio-temporal characteristics of Xinjiang grassland NDVI and its response to climate change from 1981 to 2018. Acta Ecol. Sin. 2023, 43, 1537–1552. [Google Scholar] [CrossRef]
  26. Xu, H.; Cheng, W.; Wang, B.; Song, K.; Zhang, Y.; Wang, R.; Bao, A. Effects of Geomorphic Spatial Differentiation on Vegetation Distribution Based on Remote Sensing and Geomorphic Regionalization. Remote Sens. 2024, 16, 1062. [Google Scholar] [CrossRef]
  27. Carpenter, G.; Gillison, A.N.; Winter, J. DOMAIN: A flexible modelling procedure for mapping potential distributions of plants and animals. Biodivers. Conserv. 1993, 2, 667–680. [Google Scholar] [CrossRef]
  28. Elith, J.; Leathwick, J.R.; Hastie, T. A Working Guide to Boosted Regression Trees. J. Anim. Ecol. 2008, 77, 802–813. [Google Scholar] [CrossRef]
  29. Xie, C.; Liu, C.; Wang, H.; Liu, D.; Jim, C.Y. Distribution pattern of large old Ginkgobiloba in China under climate change scenarios. Ecol. Evol. 2024, 14, e11367. [Google Scholar] [CrossRef]
  30. Booth, T.H.; Nix, H.A.; Busby, J.R.; Hutchinson, M.F. BIOCLIM: The first species distribution modelling package, its early applications and relevance to most current MAXENT studies. Divers. Distrib. 2014, 20, 1–9. [Google Scholar] [CrossRef]
  31. Wainwright, B.J.; Leon, J.; Vilela, E.; Hickman, K.J.E.; Caldwell, J.; Aimone, B.; Bischoff, P.; Ohran, M.; Morelli, M.W.; Arlyza, I.S.; et al. Wallace’s line structures seagrass microbiota and is a potential barrier to the dispersal of marine bacteria. Environ. Microbiome 2024, 18, 23. [Google Scholar] [CrossRef] [PubMed]
  32. Phillips, S.J.; Anderson, R.P.; Dudík, M.; Schapire, R.E.; Blair, M.E. Opening the Black Box: An Open-Source Release of Maxent. Ecography 2017, 40, 887–893. [Google Scholar] [CrossRef]
  33. González-Miguéns, R.; Cano, E.; Guillén-Oterino, A.; Quesada, A.; Lahr, D.J.G.; Tenorio-Rodríguez, D.; de Salvador-Velasco, D.; Velázquez, D.; Carrasco-Braganza, M.I.; Patterson, R.T.; et al. A needle in a haystack: A new metabarcoding approach to survey diversity at the species level of Arcellinida (Amoebozoa: Tubulinea). Mol. Ecol. Resour. 2023, 23, 1034–1049. [Google Scholar] [CrossRef] [PubMed]
  34. Sillero, N.; Campos, J.C.; Arenas-Castro, S.; Barbosa, A.M. A curated list of R packages for ecological niche modelling. Ecol. Model 2023, 476, 110242. [Google Scholar] [CrossRef]
  35. Breiman, L. Random Forests. In Machine Learning; Schapire, R.E., Ed.; Kluwer Academic Publishers: Dordrecht, The Netherlands, 2001; Volume 45, pp. 5–32. [Google Scholar]
  36. Elith, J.; Leathwick, J. Predicting Species Distributions from Museum and Herbarium Records Using Multiresponse Models Fitted with Multivariate Adaptive Regression Splines. Divers. Distrib. 2007, 13, 265–275. [Google Scholar] [CrossRef]
  37. Akpoti, K.; Kabo-bah, A.T.; Dossou-Yovo, E.R.; Groen, T.A.; Zwart, S.J. Mapping suitability for rice production in inland valley landscapes in Benin and Togo using environmental niche modeling. Sci. Total Environ. 2020, 709, 136165. [Google Scholar] [CrossRef]
  38. Akpoti, K.; Dossou-Yovo, E.R.; Zwart, S.J.; Kiepe, P. The potential for expansion of irrigated rice under alternate wetting and drying in Burkina Faso. Agric. Water. Manag. 2021, 247, 106758. [Google Scholar] [CrossRef]
  39. Mondanaro, A.; DiFebbraro, M.; Castiglione, S.; Melchionna, M.; Serio, C.; Girardi, G.; Belfiore, A.M.; Raia, P. ENphylo: A new method to model the distribution of extremely rare species. Methods Ecol. Evol. 2023, 14, 911–922. [Google Scholar] [CrossRef]
  40. Drake, J.M.; Randin, C.; Guisan, A. Modelling ecological niches with support vector machines. J. Appl. Ecol. 2006, 43, 424–432. [Google Scholar] [CrossRef]
  41. Elango, A.; Raju, H.K.; Shriram, A.N. Predicting ixodid tick distribution in Tamil Nadu domestic mammals using ensemble species distribution models. Ecol. Process 2025, 14, 13. [Google Scholar] [CrossRef]
  42. Phillips, S.J.; Anderson, R.P.; Schapire, R.E. Maximum entropy modeling of species geographic distributions. Ecol. Model 2006, 190, 231–259. [Google Scholar] [CrossRef]
  43. Guga, S.; Xu, J.; Riao, D.; Li, K.W.; Han, A.; Zhang, J.Q. Combining MaxEnt model and landscape pattern theory for analyzing interdecadal variation of sugarcane climate suitability in Guangxi, China. Ecol. Indic. 2021, 131, 1872–7034. [Google Scholar] [CrossRef]
  44. He, Y.T.; Wang, G.; Ren, Y.; Gao, S.; McKirdy, S.J.; Chu, D. Maxent modelling combined with fuzzy logic provides new insights into Predicting the distribution of potato cyst nematodes with limited data. Comput. Electron. Agric. 2024, 222, 109035. [Google Scholar] [CrossRef]
  45. Pang, D.; Wan, Y.; Chen, Y.; Li, Y.D.; Wang, L.R.; Chen, G.T.; Pu, F.G.; Ding, J.; Li, J.Q.; Dai, Q.; et al. Climate change and conservation strategies for the Anhui musk deer: Habitat suitability and climate refuges in the Dabie Mountains. Landsc. Ecol. 2025, 40, 94. [Google Scholar] [CrossRef]
  46. Jiang, F.; Song, P.; Zhang, J.; Wang, D.; Li, R.; Liang, C.; Zhang, T. Assessment approach for conservation effectiveness and gaps for endangered species based on habitat suitability: A case study of alpine musk deer in western China. Ecol. Indic. 2025, 170, 113080. [Google Scholar] [CrossRef]
  47. Zhao, K.; Wang, N.; Xu, J.; Tian, S.; Zhang, Y. Habitat use and spatial distribution patterns of endangered pheasants on the southern slopes of the Himalayas. Glob. Ecol. Conserv. 2025, 57, e03414. [Google Scholar] [CrossRef]
  48. Liu, R.; Kong, H.; Wang, Q.; Li, Y. Identify priority protected areas for biodiversity conservation adapting to future climate and land cover changes. Ecol. Indic. 2025, 170, 113068. [Google Scholar] [CrossRef]
  49. Luo, X.; Yang, M.; Khan, M.S.; Huang, W.; Liu, S.; Liao, F.; Li, Y. Climate and topographic drivers of species and functional diversity in subtropical shrublands of Guangdong, China. Ecol. Indic. 2025, 176, 113641. [Google Scholar] [CrossRef]
  50. Wu, F.; Wang, Y.; Zheng, M.; Wang, J.; Pan, J.; Liu, L. Prediction and quality zoning of potentially suitable areas for Panax notoginseng cultivation using MaxEnt and random forest algorithms in Yunnan Province, China. Ind. Crop. Prod. 2025, 229, 120960. [Google Scholar] [CrossRef]
  51. Wang, C.; Shi, X.; Liu, J.; Zhao, J.; Bo, X.; Chen, F.; Chu, Q. Interdecadal variation of potato climate suitability in China. Agric. Ecosyst. Environ. 2021, 310, 107293. [Google Scholar] [CrossRef]
  52. Adhikari, P.; Lee, Y.H.; Adhikari, P.; Poudel, A.; Seo, C.; Lee, D.H.; Park, Y.S.; Hong, S.H. Global assessment of invasion risk: Ardisia elliptica, one of the most noxious tropical shrubs in the world. Ecol. Process 2025, 14, 55. [Google Scholar] [CrossRef]
  53. Jin, L.; Zhao, J.; Shao, Q.Y.; Ji, B.; Sui, X.; Gong, Y. Simulation of Potential Distribution Patterns of Common Plant Species in Xinjiang Grassland under Climate Change Scenarios. Acta Agrestia Sin. 2025, 33, 2973–2991. [Google Scholar]
  54. Li, M.; Wang, J.B.; Zhang, X.J.; Zhang, Y.; Wang, Z.R.; Yang, Y.S. Distribution of potential suitable areas of dominant species in alpine grasslands and alpine meadows in the Tibetan Plateau under future climate scenarios. Acta Ecol. Sin. 2024, 44, 10162–10177. [Google Scholar]
  55. Choudhury, M.; Deb, P.; Singha, H.; Chakdar, B.; Medhi, M. Predicting the probable distribution and threat of invasive Mimosa diplotricha Suavalle and Mikania micrantha Kunth in a protected tropical grassland. Ecol. Eng. 2016, 97, 23–31. [Google Scholar] [CrossRef]
  56. Yang, X.Q.; Kushwaha, S.P.S.; Saran, S.; Xuc, J.; Roy, P.S. Maxent modeling for predicting the potential distribution of medicinal plant, Justicia adhatoda L. inlesser Himalayan foothills. Ecol. Eng. 2013, 51, 83–87. [Google Scholar] [CrossRef]
  57. Gu, C.; Tu, Y.; Liu, L.; Wei, B.; Zhang, Y.; Yu, H.; Wang, X.; Yangjin, Z.; Zhang, B.; Cui, B. Predicting the potential global distribution of ageratina adenophora under current and future climate change scenarios. Ecol. Evol. 2021, 11, 12092–12113. [Google Scholar] [CrossRef]
  58. Zhang, L.; Liu, H.; Zhang, H.; Chen, Y.; Zhang, L.; Kudusi, K.; Taxmamat, D.; Zhang, Y. Potential distribution of three types of ephemeral plants under climate changes. Front. Plant Sci. 2022, 13, 1035684. [Google Scholar] [CrossRef]
  59. Yang, L.; Wei, W.; Wang, T.; Li, L. Temporal-Spatial Variations of Vegetation Cover and Surface Soil Moisture in the Growing Season across the Mountain-Oasis-Desert System in Xinjiang, China. Geocarto Int. 2022, 37, 3912–3940. [Google Scholar] [CrossRef]
  60. Jim’enez-Valverde, A.; Peterson, A.T.; Sober’on, J.; Overton, J.M.; Arag’on, P.; Lobo, J.M. Use of niche models in invasive species risk assessments. Biol. Invasions 2011, 13, 2785–2797. [Google Scholar] [CrossRef]
  61. Jayasinghe, S.L.; Kumar, L. Modeling the climate suitability of tea [Camellia sinensis (L.) O. Kuntze] in Sri Lanka in response to current and future climate change scenarios. Agr. For. Meteorol. 2019, 272–273, 102–117. [Google Scholar] [CrossRef]
  62. Cobos, M.E.; Peterson, A.T.; Barve, N.; Osorio-Olvera, L. Kuenm: An R package for detailed development of ecological niche models using Maxent. PeerJ 2019, 7, e6281. [Google Scholar] [CrossRef]
  63. Zhang, K.; Yao, L.; Meng, J.; Tao, J. Maxent modeling for predicting the potential geographical distribution of two peony species under climate change. Sci. Total Environ. 2018, 634, 1326–1334. [Google Scholar] [CrossRef]
  64. Warren, D.L.; Seifert, S.N. Ecological niche modeling in MaxEnt: The importance of model complexity. Ecol. Appl. 2011, 21, 1113–1121. [Google Scholar] [CrossRef]
  65. Elith, J.; Hsu, S.H.; Phillips, S.J. Spatial predictions of species distributions: An evaluation of different techniques for modeling the geographic distributions of species using presence-only data. Ecography 2019, 42, 380–392. [Google Scholar]
  66. Radosavljevic, A.; Anderson, R.P. Making better MAXENT models of species distributions: Complexity, overfitting, and evaluation. J. Biogeogr. 2014, 41, 629–643. [Google Scholar] [CrossRef]
  67. Naimi, B.; Araújo, M.B. Future challenges in ecological niche modeling: Evaluating new methods for species distribution models. Glob. Ecol. Biogeogr. 2021, 30, 1–9. [Google Scholar]
Remotesensing 18 00152 g001
Figure 1. Field sites of nine grassland types.
Figure 1. Field sites of nine grassland types.
Remotesensing 18 00152 g001
Remotesensing 18 00152 g002
Figure 2. Predicted and actual grassland map in the 1980s: (a) actual grassland map; (b) predicted grassland map.
Figure 2. Predicted and actual grassland map in the 1980s: (a) actual grassland map; (b) predicted grassland map.
Remotesensing 18 00152 g002
Remotesensing 18 00152 g003
Figure 3. Importance and contribution of all factors.
Figure 3. Importance and contribution of all factors.
Remotesensing 18 00152 g003
Remotesensing 18 00152 g004
Figure 4. Importance and contribution of decorrelated factors.
Figure 4. Importance and contribution of decorrelated factors.
Remotesensing 18 00152 g004
Remotesensing 18 00152 g005aRemotesensing 18 00152 g005bRemotesensing 18 00152 g005c
Figure 5. Distribution of grassland types in Xinjiang under climate change patterns.
Figure 5. Distribution of grassland types in Xinjiang under climate change patterns.
Remotesensing 18 00152 g005aRemotesensing 18 00152 g005bRemotesensing 18 00152 g005c
Remotesensing 18 00152 g006aRemotesensing 18 00152 g006bRemotesensing 18 00152 g006cRemotesensing 18 00152 g006d
Figure 6. Spatial change of grassland types under climate patterns.
Figure 6. Spatial change of grassland types under climate patterns.
Remotesensing 18 00152 g006aRemotesensing 18 00152 g006bRemotesensing 18 00152 g006cRemotesensing 18 00152 g006d
Table 1. Advantage and disadvantage of species distribution models.
Table 1. Advantage and disadvantage of species distribution models.
ModelAdvantagesDisadvantages
BRTHigh predictive accuracy; handles complex nonlinear relationships and interactions; robust to outliers Computationally intensive; slower training; complex parameter tuning; “black-box” nature
DOMAINSimple, robust; requires only presence data; suitable for rare species; derives simple niche characteristics Lower predictive accuracy; cannot handle categorical environmental variables; sensitive to sampling bias
BIOCLIMAlgorithm simple, easy to use Results are binary; sensitive to outliers; climate variables are treated equally; lower accuracy
GLMGood model interpretability; shows relationship between environmental factors and target via regression equation Cannot handle qualitative environmental factors; accuracy depends on sample size
MaxEntHigh prediction accuracy; performs well with relatively small sample sizes; can use only presence data Limited model adjustability;
RFHigh accuracy; handles nonlinear relationships; robust to outliers and noise; provides variable importance Requires large data volume for accuracy;
K-NNSimple implementation; strong performance; model-free classification Impractical to use a fixed k value for all test samples; time-consuming to determine optimal k for each sample
ENphyloFast and accurate for rare species (even <20 occurrences); good evaluation scores Specifically designed for rare species; performance with common species less explored
ENFADerives simple niche characteristics; can identify main limiting factors Lower accuracy; cannot handle categorical environmental variables
SVMHigh accuracy; simulates complex relationships Accuracy requires large data volume; poor transferability; high computational cost
Table 2. Definition of Grassland type.
Table 2. Definition of Grassland type.
Grassland TypeDefination
Humidity PrecipitationComposition
AS0.3–0.6200–350Cold-tolerant perennial, xerophytic bunchgrasses, or xerophytic small shrubs.
AM>= 1.0>400Cold-tolerant perennial mesophytic herbaceous plants, and may include mesophytic alpine shrubs.
MM>= 1.0>500Perennial mesophytic herbaceous plants.
TMS0.6–1.0350–500Perennial, drought-tolerant herbaceous plants dominant, mixed with a significant amount of mesophytic plants.
TD<0.1<100Super-xerophytic shrubs and semi-shrubs.
TSD0.10–0.13100–150Drought-tolerant semi-shrubs and desert shrub components dominant, containing a certain proportion of xerophytic herbaceous or semi-shrubby steppe components.
TS0.3–0.6250–400Perennial xerophytic herbaceous plants in semi-arid area.
TDS0.13–0.3150–300Perennial xerophytic bunchgrasses dominant, with a certain amounts of xerophytic and drought-resistant small shrubs and shrub desert species.
LMTemperate, subtropical, and tropical river floodplains, coastal mudflats, lake basin edges, inter-hill lowlands, valley floors, and alluvial fan edges with groundwater levels below 0.5 m.Perennial hygrophytic or mesophytic herbaceous plants.
Table 3. Model variables.
Table 3. Model variables.
Variable TypeVariable NameVariable DescriptionsUnit
T
(temperature)		Bio1Mean annual air temperature°C
Bio2Mean diurnal air temperature range°C
Bio3Isothermality°C
Bio4Temperature seasonality°C/100
Bio5Mean daily maximum air temperature of the warmest month°C
Bio6Mean daily minimum air temperature of the coldest month°C
Bio7Annual range of air temperature°C
Bio8Mean daily mean air temperatures of the wettest quarter°C
Bio9Mean daily mean air temperatures of the driest quarter°C
Bio10Mean daily mean air temperatures of the warmest quarter°C
Bio11Mean daily mean air temperatures of the coldest quarter°C
P
(precipitation)		Bio12Annual precipitation amountkg m−2month−1
Bio13Precipitation amount of the wettest monthkg m−2month−1
Bio14Precipitation amount of the driest monthkg m−2month−1
Bio15Precipitation seasonalitykg m−2
Bio16Mean monthly precipitation amount of the wettest quarterkg m−2month−1
Bio17Mean monthly precipitation amount of the driest quarterkg m−2month−1
Bio18Mean monthly precipitation amount of the warmest quarterkg m−2month−1
Bio19Mean monthly precipitation amount of the coldest quarterkg m−2month−1
E
(Environment)		ElevationTopographic elevationm
SlopeThe degree of steepness of the surface element°
AspectThe direction of the slope°
Geomorlandform type-
NDVIVegetation index-
RiverDistance from the riverkm
Soil_phSoil pH value-
Soil_symbolSoil symbol-
VegetationVegetation symbol-
Table 4. Prediction Accuracy and Threshold of grassland type.
Table 4. Prediction Accuracy and Threshold of grassland type.
Grassland TypeSite Number
(in the1980s)		AUC
(in the1980s)		ThreholdSite Number
(2023)		AUC
(2041–2100)
LM9690.9150.54870.93–0.939
AM7780.9560.662180.951–0.958
AS4030.9750.671460.97–0.976
MM5910.9520.664530.934–0.96
TMS2150.9340.742000.92–0.93
TSD3630.9090.614280.933–0.938
TS6650.9340.65820.944–0.951
TDS6370.8890.556680.903–0.922
TD13920.8540.512030.873–0.876
Total6013 4385
Table 5. Contribution and importance.
Table 5. Contribution and importance.
ContributionLMAMASMMTMSTSTSDTDTDS
T 256.61327.526.145.274.618.555.1
P 022.71.723.826.514.611.820.46.7
E 9820.785.348.747.440.213.661.138.3
ImportanceLMAMASMMTMSTSTSDTDTDS
T 10.285.411.33221.755.862.226.962.8
P 00.95.83538.59.921.919.14.4
E 89.713.882.93339.934.315.854.132.9
Table 6. Area of grassland types and change. km2.
Table 6. Area of grassland types and change. km2.
Grassland Type2041–2071
SSP126		Change Rate2041–2071
SSP370		Change Rate2041–2071
SSP585		Change Rate
LM88,765.4515.3691,324.4118.6986,993.2413.06
AM13,134.43−66.9912,539.25−68.4912,468.04−68.67
AS7596.53−74.888716.06−71.189528.82−68.49
MM29,250.2630.3629,324.7930.729,148.2629.91
TMS28,319.49148.0129,861.75161.5228,755.66151.83
TS13,596.02−60.5112,395.62−6413,119.34−61.9
TSD35,168.522.7335,582.53.9434,818.951.71
TDS39,183.3818.7739,719.8920.440,855.9623.84
TD146,627.75−21.84145,396.56−22.5145,259.36−22.57
Total401,641.82−17.29404,860.84−16.63400,947.62−17.44
Grassland Type2071–2100
SSP126		Change Rate2071–2100
SSP370		Change Rate2071–2100
SSP585		Change Rate
LM86,152.1111.9789,755.9116.6589,670.6516.54
AM12,363.8−68.9311,468.13−71.1811,886.33−70.13
AS8203.83−72.878262.18−72.688358.49−72.36
MM31,682.3241.235,916.3560.0729,231.8330.28
TMS26,499.92132.0826,499.92132.0828,000.21145.22
TS12,379.45−64.0410,911.65−68.3111,871.54−65.52
TSD32,462.78−5.1733,590.62−1.8837,059.888.26
TDS40,801.9923.6842,899.9830.0440,929.5724.06
TD140,736.54−24.98149,606.22−20.25141,123.88−24.77
Total391,282.75−19.43393,887.59−18.89398,132.37−18.02
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zhao, J.; Li, K.; Shao, Q.; Bai, J.; Gong, Y.; Liu, Y. Combining In Situ and Remote-Sensing Data to Assess the Spatial Pattern and Changes of Major Grassland Types in Xinjiang, China, Under Climate Change Scenarios. Remote Sens. 2026, 18, 152. https://doi.org/10.3390/rs18010152

AMA Style

Zhao J, Li K, Shao Q, Bai J, Gong Y, Liu Y. Combining In Situ and Remote-Sensing Data to Assess the Spatial Pattern and Changes of Major Grassland Types in Xinjiang, China, Under Climate Change Scenarios. Remote Sensing. 2026; 18(1):152. https://doi.org/10.3390/rs18010152

Chicago/Turabian Style

Zhao, Jin, Kaihui Li, Qianying Shao, Jie Bai, Yanming Gong, and Yanyan Liu. 2026. "Combining In Situ and Remote-Sensing Data to Assess the Spatial Pattern and Changes of Major Grassland Types in Xinjiang, China, Under Climate Change Scenarios" Remote Sensing 18, no. 1: 152. https://doi.org/10.3390/rs18010152

APA Style

Zhao, J., Li, K., Shao, Q., Bai, J., Gong, Y., & Liu, Y. (2026). Combining In Situ and Remote-Sensing Data to Assess the Spatial Pattern and Changes of Major Grassland Types in Xinjiang, China, Under Climate Change Scenarios. Remote Sensing, 18(1), 152. https://doi.org/10.3390/rs18010152

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop