Data and code for : Unveiling the Dust-Cloud Synergy A Machine Learning Attribution and Modeling Study of Surface Radiation in the Hexi Corridor

This dataset and code support the paper "Unveiling the Dust-Cloud Synergy A Machine Learning Attribution and Modeling Study of Surface Radiation in the Hexi Corridor" (under review at PeerJ). It contains processed radiation and aerosol data for solar energy assessment in the Hexi Corridor, China. Data sources: - ERA5 (3-hourly radiation variables: SSRD, SSRDC, FDIR) - CAMS EAC4 (3-hourly aerosol and cloud variables: DUAOD550, OMAOD550, SUAOD550, SSAOD550, BCAOD550, HCC, MCC, LCC) - WRF simulations (version 4.4.2) driven by ERA5/CAMS Contents: - Zarr format: radiation_asia.zarr, radiation_asia_v2.zarr (adds FDIR·cosθ), predictors_asia.zarr, daymean_solarMask.zarr - XGBoost models and SHAP values for SSRD and FDIR·cosθ (JSON + Parquet) - WRF namelist files (namelist.input, namelist.wps) - WRF-interpolated station data for Dunhuang (52418) and Jiuquan (52533) - Hexi Corridor boundary shapefile and NetCDF mask - List of retained predictors after VIF analysis (sel_vars_aod_cloud.txt) Missing data (obtain separately): - Original ERA5/CAMS reanalysis: available from Copernicus CDS - Raw WRF output files: too large to include, can be regenerated using namelist files - Station observations: confidential, not included (statement provided)