Data and Code Archive: Spatial Heterogeneity of Urban Thermal Sensitivity - JSTS Framework, Dhaka City Corporation, Bangladesh

This archive contains the complete data and code supporting the paper "Spatial Heterogeneity of Urban Thermal Sensitivity: A Jacobian-Based Comparison with TreeSHAP Attribution in Dhaka City Corporation, Bangladesh" submitted to Environmental Research Letters (IOP Publishing, Q1). The study introduces the Jacobian-based Spatial Thermal Sensitivity (JSTS) framework, which compares pixel-level MLP backpropagation gradient sensitivity with XGBoost TreeSHAP distributional attribution for urban land surface temperature driver identification across 338,688 pixels at 30 m resolution. Archive contents:- notebooks.zip: All 9 Jupyter analysis notebooks- gee_scripts.zip: Google Earth Engine Landsat 9 export script- data.zip: Processed CSVs, NumPy arrays, trained models, raw GeoTIFF exports- outputs.zip: Publication figures (TIFF) and output GeoTIFF maps- shapefiles.zip: Dhaka City Corporation 86-ward boundary Code repository: https://github.com/akprottay/jsts-dhaka-urban-heat