Grassland ecosystems in arid regions increasingly experience resilience loss due to intensifying climatic variability. However, the limited interpretability of conventional machine learning models constrains our understanding of underlying ecological drivers. This study constructs an integrative framework that combines temporal autocorrelation (TAC) metrics with explainable machine learning, employing Random Forest and SHAP (SHapley Additive exPlanations) analysis. Time series of satellite-derived vegetation indices from MODIS (2001–2023), particularly the kernel Normalized Difference Vegetation Index (KNDVI), support the generation of TAC and its trend-based derivative δTAC. The framework assesses ecosystem resilience across seven representative grassland types in Xinjiang, capturing diverse responses to climate variability and vegetation dynamics. Results reveal pronounced spatial heterogeneity: resilience declines in radiation-stressed arid zones, while hydrothermally stable regions maintain stronger recovery capacity. Key drivers include temperature variability and vegetation dynamics, with divergent effects among grassland types. Meadow and Typical Steppe exhibit higher resilience under stable hydrothermal regimes, whereas desert and alpine systems show greater sensitivity to warming and climatic fluctuations. This framework enhances diagnostic transparency and ecological insight, offering a spatially explicit, data-driven tool for resilience monitoring. The findings support the formulation of targeted adaptation strategies and sustainable grassland management in response to ongoing climate change.

Potential evapotranspiration (PET) is a crucial variable for implementing adaptation measures to mitigate the potential impacts of climate change on water resources. In the context of global warming, PET is essential for predicting water resource supply and demand, guiding irrigation and water management decisions. However, there is limited understanding of the spatiotemporal changes in PET and its driving factors in the hyper-arid regions of Northwest China. In this study, the Hargreaves model was employed to estimate PET in the Hami region from 1991 to 2020. By combining relevant climate data and partial correlation analysis, we investigated the spatiotemporal distribution patterns of PET within the study area and analyzed the factors influencing these patterns. The results showed the following: (1) From 1991 to 2020, the overall PET in the Hami region demonstrated a tendency to rise. The interannual trend rates of PET for the full year, spring, summer, autumn, and winter were 0.933, 2.744, 0.906, 0.488, and −0.406 mm·a-1, respectively. Despite a decreasing trend in winter PET, the other seasonal PET values and the annual PET values exhibited an increasing trend. (2) The spatial distribution of both annual and seasonal PET showed significant regional heterogeneity, following a consistent pattern marked by lower values in the central part and higher values in the surrounding areas. The southern region tended to have relatively high PET, while the northwestern region experienced comparatively low PET. (3) Partial correlation analysis indicated significant differences in the impact of various climatic factors on PET. The maximum temperature emerged as the dominant factor influencing annual PET variation, while precipitation played a leading role in influencing autumn PET variation. This study underscores the influence of climate change on PET in the Hami region, contributing to an enhanced comprehension of PET variations.

The response mechanisms of vegetation dynamics to climate change in arid regions, particularly under extreme low-altitude and high-temperature environments, remain unclear. Focusing on China’s lowest and hottest Turpan-Hami Basin, this study investigates the spatiotemporal evolution of vegetation cover (using MODIS NDVI) and its response to temperature, precipitation, and potential evapotranspiration (PET) based on data from 2001 to 2020. Theil–Sen trend analysis, the Mann–Kendall test, and Pearson correlation were employed. Key findings include the following: (1) NDVI exhibited a significant increasing trend, with the largest rise in winter and peak values in summer. Spatially, high NDVI was concentrated in oasis and mountainous forest-grassland zones, while low values prevailed in desert Gobi regions; 34.2% of the area showed significant improvement, though localized degradation occurred. (2) Temperature showed no significant overall correlation with NDVI, except for strong positive correlations in limited high-altitude cold zones (2.9%). Precipitation had minimal influence (no correlation in 75.4% of the area), with localized positive responses in northwestern foothills linked to runoff. PET exhibited positive correlations (weak or strong) with NDVI across nearly half of the region (46.8%), predominantly in oasis-desert and piedmont transition zones. (3) Human activities, notably irrigation and shelterbelt projects, are key drivers of oasis vegetation restoration. Critically, the positive PET-NDVI correlation challenges the conventional paradigm viewing evapotranspiration solely as water stress. This study elucidates the compound responses of vegetation dynamics to climatic and anthropogenic factors in a low-altitude arid region, providing a scientific basis for ecological restoration and water resource management optimization.

In this study, we investigate the impact of soil moisture at varying depths on the stability of Chinese ecosystems, with ecosystem stability assessed using the Enhanced Vegetation Index (EVI) and Gross Primary Productivity (GPP). A multi-perspective analysis is conducted across different climatic zones and land cover types. Sen’s Slope Estimation and the Mann–Kendall trend test, combined with linear regression and correlation analyses, are employed to analyze the long-term trends of EVI and GPP in different climatic zones and land cover types and to assess the effects of soil moisture changes on ecosystem stability. The research reveals the following findings: (1) On a national scale, both EVI and GPP exhibit positive growth trends, with more significant increases in humid areas and relatively slower growth in arid areas. In addition, EVI and GPP of different land cover types exhibit positive inter-annual variation trends, reflecting a gradual enhancement in ecosystem productivity. (2) Cluster analysis shows that EVI has strong spatial correlation, with a distribution pattern of low–low (L-L) clusters in the north and high–high (H-H) clusters in the south. L-H clusters are concentrated in the Huaihai, Southwest Rivers, and Pearl River basins, while H-L clusters are scattered along the eastern coast. The spatial correlation of GPP is mainly concentrated in the south and the northeast, with a distribution pattern of L-L in the northeast, L-H in the Yangtze River basin, and H-H in the south. H-L clusters are dispersed in the downstream area of the Yangtze River. Both EVI and GPP show a tendency for high-value aggregation in space, with high-value areas of EVI located in the south and low-value areas in the central and western regions. High-value areas of GPP are in the south, while low-value areas are in the northeast, particularly in the Yangtze River Delta. (3) The correlation between EVI, GPP, and soil moisture varies significantly across different climatic regions. Arid and semi-humid regions show significant correlations between specific soil moisture depths and EVI and GPP, while such correlations are not significant in humid regions. The EVI and GPP values of croplands and grasslands are significantly and negatively correlated with soil moisture at depths of 150–200 cm (SM4). Conversely, wetland GPP values increase significantly with increasing soil moisture. Other vegetation types do not show significant correlations with soil moisture. The results of this study provide an important basis for understanding the impact of climate change on ecosystem stability and offer scientific guidance for ecological protection and water resource management.