Creek Fire Snow Persistence GeoAI: analysis-ready panels, diagnostics, and supporting rasters for the 2020 Creek Fire

Analysis-ready geospatial products and diagnostics for a multisource GeoAI study of post-fire snow persistence after the 2020 Creek Fire (Sierra Nevada, California). From Wildfire Severity to Snow Persistence: A Multisource GeoAI Study of the 2020 Creek Fire Authors: Parastoo Farajpoor and Mohammadreza Narimani (University of California, Davis). Contents CreekFire_Snow_Persistence_GeoAI_derived_data.zip — study boundary, matched BACI panel, analysis cells, manuscript tables, spatial-CV / SHAP / OOF diagnostics, final figures 01–09, replication scripts, manifests, and documentation. CreekFire_Snow_Persistence_GeoAI_supporting_rasters.zip — clipped HLS and MODIS 500 m persistence stacks, SDD-proxy rasters, Copernicus DEM AOI clip, and Sentinel-2 severity at 500 m. Locked results. Matched pairs n = 3,778; ML cell-years n = 22,665. Overall persistence BACI = +0.0017 (95% CI 0.0003–0.0032); high-severity BACI = +0.026. Spatial-CV XGBoost level R² = 0.811; anomaly R² = 0.046. SHAP elevation + winter temperature ≈ 65.7%. Grid 500 m, EPSG:32611. Companion code: https://github.com/MohammadrezaNarimaniUCDavis/CreekFire_Snow_Persistence_GeoAI Products support association and spatially honest screening; they are not operational snow forecasts. Source layers retain original licenses.