Extreme Event Detection Dataset

This dataset supports multi-temporal and multi-modal change detection of extreme events. It includes Sentinel-1 and Sentinel-2 satellite imagery for global wildfire events between 2023 and 2024, providing multiple pre-event and a single post-event observation for each event, along with expert-derived ground truth change masks. Sentinel-2 optical imagery (Level-1C, top-of-atmosphere reflectance) includes all 13 spectral bands resampled to 10 m spatial resolution. Sentinel-1 synthetic aperture radar (SAR) imagery (Level-1 GRD) with VV and VH polarization is included to support multi-sensor change detection and to mitigate cloud cover and illumination challenges. For each event, 12 pre-event and one post-event optical and SAR images were collected. Data are divided into non-overlapping 256 × 256 pixel tiles, retaining only tiles containing affected areas. Ground truth masks were derived from delineation and grading products of the Copernicus Emergency Management Service (EMS). EMS vector data were rasterized and projected to the WGS ’84 coordinate reference system at 10 m resolution. Sentinel-1 and Sentinel-2 data were accessed via Sentinel Hub under the framework of the Copernicus Programme, which provides free, full, and open access under the Copernicus Data and Information Policy. © European Union, Copernicus Programme © European Union, Copernicus EMS