Integrated Multi-Source Dataset for Grassland Resilience Assessment in Northwestern China (2001–2023)

This dataset provides a comprehensive collection of multi-source environmental and ecological data used for assessing grassland resilience and early-warning signals of abrupt decline events in northwestern China, particularly the Xinjiang region.<br>The dataset integrates meteorological, vegetation, land use, and grazing information at a unified 1 km spatial resolution and annual temporal scale, supporting analyses of climate–grazing–resilience interactions.<b>Contents</b>The dataset contains the following components:<b>Weather station data_Xinjiang_2000–2019</b> — Daily temperature and precipitation from national meteorological stations, used for local-scale calibration and trend analysis.<b>ERA5_Climate_Variables_Xinjiang_2000_2023</b> — Reanalysis-based gridded climate variables (temperature, precipitation, radiation, evapotranspiration).<b>NDVI_MOD13A2_2001_2023</b> — MODIS MOD13A2 product-derived NDVI data, 1 km resolution, preprocessed and smoothed.<b>FVC_1km_Xinjiang</b> — Fractional Vegetation Cover (FVC) derived from remote sensing, annual mean rasters at 1 km resolution.<b>Grazing_Intensity_China_2001_2023</b> — National grazing intensity dataset (SU ha⁻¹) from the National Ecosystem Science Data Center, used to quantify anthropogenic pressure.<b>Land_use_Xinjiang_2000–2020</b> — Chinese Land Use/Cover Change (CLCD) dataset for grassland pixel selection.<b>Classification_of_grassland_China_2020</b> — National grassland classification map (1 km) indicating dominant grassland types.<b>Temporal Coverage</b>Meteorological data: 2000–2023NDVI/FVC: 2000–2023Grazing intensity: 2001–2023Land use: 2000 / 2020Grassland classification: 2020<b>Processing and Usage</b>All layers were resampled to 1 km and converted to point format for spatial modeling.The data were used to derive resilience indicators such as <b>temporal autocorrelation (TAC)</b>, its long-term trend (δTAC), and the <b>Proximity-to-collapse indicator</b> based on XGBoost models.Grassland classification and land use maps served for <b>sample selection and ecological stratification</b>.ERA5 and station data supported <b>climate variable interpolation</b> and long-term trend analysis.NDVI and FVC time series were used to identify <b>abrupt decline (AD) events</b> and assess vegetation stability.Grazing intensity data quantified human disturbance levels and their coupling effects with climatic variability.