GEOID-Flood: A Large-Scale Multi-Modal Benchmark Dataset for Flood Segmentation

Researchers introduced GEOID-Flood, a new large-scale benchmark for flood segmentation built from Copernicus Emergency Management Service activations.

The dataset contains more than 14,000 tiles from 219 flood events in 65 countries, with pre- and post-event Sentinel-1, pre-event Sentinel-2, DEM data, and validated labels.

It was used to compare foundation models and conventional encoders across single-image, multi-temporal, and multi-modal settings.

Results show that multi-modal fusion and fine-tuning improve transfer to unseen flood events, making GEOID-Flood a stronger testbed for geospatial flood mapping models.