dataset · Global

ai4g-flood-dataset

Hugging Face dataset ai-for-good-lab/ai4g-flood-dataset

Claim this listing

Represent ai4g-flood-dataset? Send a claim for review so we can discuss listing updates. Claims do not automatically grant ownership or verification.

Data offered docs-verified: global flood detections (20 m) from Sentinel-1 SAR · flood extent GeoTIFFs with 80 m and 240 m buffers · flood recurrence maps (months with flooding detected) · per-detection auxiliary data: soil moisture, elevation/slope, temperature, land cover
Coverage docs-verified: Global · October 2014 – September 2024
Resolution docs-verified: 20 m detections, grouped into 3° × 3° tiles
Formats docs-verified: Parquet · GeoTIFF
Access docs-verified: Open access under MIT licence; no authentication listed

Use cases llm-inferred: Historical flood-prone area screening for FRA context using 10-year recurrence maps · Post-event flood extent comparison against validation data using unfiltered single-event parquet · Flood exposure screening for international project sites where UK datasets do not cover the area
Audience llm-inferred: flood consultants · hydrologists · GIS analysts · planners

Constraints docs-verified: Detection quality reduced in urban areas and areas with dense vegetation cover · Certain soil types can still produce false positives · Recurrence files may be missing for areas with no flooding detected after the exclusion mask; use the 80 m or 240 m exclusion mask instead · Hugging Face dataset viewer unavailable for this split (file too large) · Aggregations over several months or more require the recommended false-positive filters

Verified 2026-10-15 via HF API: exists; 17446 downloads/30d, 17 likes.

Metadata observed 14/10/2026 · research sweep: manual-2026-10-15 (Hugging Face datasets)