Annual 30-meter Forest Cover Loss Maps for China (2000-2024) based on ECF-TST Framework

OverviewThis dataset contains annual, 30-meter resolution forest cover loss (FCL) maps for China spanning the years 2000 to 2024. The data was generated using a novel detection framework, the Temporal-Spectral-Texture features fusion with Eco-regional Constraints (ECF-TST), which incorporates independent XGBoost models trained for 35 distinct ecoregions within China. This approach is specifically designed to address the challenges of ecological heterogeneity across the country.MethodologyFramework: Ecoregion-Constrained Fusion of Temporal, Spectral, and Texture features (ECF-TST).Base Data: Landsat time-series imagery.Model: Separate XGBoost models were trained for each of the 35 ecoregions using a rigorously validated reference dataset of 9,202 samples.Validation: The overall accuracy of the FCL maps is 0.846 (±0.037), which significantly outperforms a global model without ecoregional constraints.Data DetailsSpatial Coverage: China.Temporal Coverage: 2000-01-01 to 2024-12-31 (Annual maps).Spatial Resolution: 30 meters.Projection: WGS 84File Format: GeoTIFFMap Values: 0 - No Loss, 2000-2024 - Loss Year]。Usage NotesThis high-accuracy data product is valuable for studies on carbon cycling, forest dynamics, conservation policy assessment, and understanding the drivers of forest change in China and other highly heterogeneous regions. The dataset also supports analysis of spatiotemporal patterns and climatic thresholds related to forest loss.Related WorkThese maps form the basis for the analysis presented in the associated manuscript: "Spatiotemporal Patterns and Climate Tipping Points of Forest Loss in China Revealed by 20 Years of Remote Sensing and Ecoregion Modeling" (Under Review).