Training/validation reference polygons and classified disturbance rasters for multi-temporal satellite monitoring of the Tasiast gold mine, Mauritania (2006–2024)

This dataset supports the paper "Satellite Monitoring of Open-Pit Mining Disturbance in Hyperarid Mauritania Using Random Forest." It contains the manually digitised training and validation reference polygons used to train per-epoch Random Forest classifiers (2012, 2017, 2024) and the manually digitised 2006 baseline disturbance footprint, together with the resulting per-epoch classified disturbance rasters. Polygons are provided as GeoJSON; rasters as GeoTIFFs in UTM Zone 28N (EPSG:32628). Source imagery (Landsat Collection 2, Sentinel-2 L2A) is publicly available via Google Earth Engine and is not redistributed here.