Dataset of Spatiotemporal Dynamics and Driving Mechanisms of Carbon Storage in Southern Sichuan, China Based on Sentinel-2 Imagery and Explainable Machine Learning (2019–2030)

This dataset supports the study of "Spatiotemporal Dynamics and Driving Mechanisms of Carbon Storage in Southern Sichuan, China." It includes: (1) high-accuracy land use/land cover classification (LUCC) maps for 2019–2025 derived from 10 m Sentinel-2 imagery via Google Earth Engine; (2) InVEST carbon storage simulation results for 2019–2030; (3) intPLUS multi-scale land-use change scenario simulations under SSP1-1.9 and SSP5-8.5; and (4) Random Forest–SHAP driving mechanism analysis outputs.

The LUCC classification integrates spectral, vegetation index, and texture features, with the optimal Random Forest model selected from eight machine learning algorithms (OA = 0.9756, Kappa = 0.9715). Carbon storage dynamics were simulated using the InVEST Carbon module, and SHAP explainability analysis was applied to quantify the non-linear threshold effects of elevation, temperature, and precipitation on carbon fixation.