dataset · Global

FloodSimBench

Hugging Face dataset chrimerss/FloodSimBench

Claim this listing

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

Data offered docs-verified: 1 m DEM (static elevation) · 1 m water depth GeoTIFF frames at 5-minute steps · Maximum depth maps for 6-hour events · 5-class flood severity segmentation labels · Rainfall intensities by return period (10, 25, 50, 100-yr)
Coverage docs-verified: 10 US cities (HOU, AUS, DAL, OKC, LA, SF, NYC, ATL, ORL, MIA), ~4 km x 4 km tile per city
Resolution docs-verified: 1 m spatial; 5-minute temporal (72 frames per 6-hour scenario)
Formats docs-verified: GeoTIFF
Access docs-verified: Open dataset, licence CC-BY-4.0; no authentication (ENTRY auth: null)

Use cases llm-inferred: Train and benchmark AI surrogate models that emulate 1 m urban flood depth against physical simulations · Compare ML flood-depth forecasts (recursive, 1 h history to t+12 steps) against physics-based outputs · Test emulator generalisation to unseen urban topographies using the city-based LA and SF test split
Audience llm-inferred: AI/ML researchers · flood modellers · hydrologists · drainage engineers · GIS analysts

Constraints llm-inferred: Covers 10 US cities only; ENTRY lists regions as Global, but no UK coverage is documented · Physics-based simulated outputs for design storms (1 h rain then 5 h recession), not observed flood extents · Very large download: 461 GB total · Hugging Face dataset viewer is currently erroring (RowsPostProcessingError)

Verified 2026-10-15 via HF API: exists; 641 downloads/30d, 1 likes.

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