Geochemistry Source Attribution and Satellite-Prediction Boundaries for Heavy Metals in Dhaka's Five Rivers

This archive is the data companion to the analysis code at https://github.com/rakibhhridoy/five-rivers-heavy-metals, supporting the manuscript "Geochemistry Source Attribution and Satellite-Prediction Boundaries for Heavy Metals in Dhaka's Five Rivers" (Hasan, Rahman, Zubyer, Rupa, Arabi, and Jolly). Contents: - data/ — the curated tabular dataset used by the analysis pipeline: primary 17-station sediment and water chemistry, real Google AlphaEarth V1 satellite embeddings, hydrological-proximity features, ERA5/CHIRPS climate covariates, and the harmonized 111-row secondary literature dataset with its audit trail. - gis/ — the full geospatial tree (~3.4 GB uncompressed): computed spectral-index rasters (NDVI, EVI, SAVI, NDWI, MNDWI, AWEI, NDBI, UI, BUI, NDSI, LST), per-year land-use/land-cover rasters (2017-2025) for change detection, the SRTM digital elevation model used for D8 flow routing, brick-kiln/industrial/sampling point shapefiles, inverse-distance-weighting interpolation rasters, and the deep-learning model-search notebooks and trained model weights.