CoolFace
Datasetpublic

justuskarlsson/FireComp

FireComp: Next-Day Fire Spread Global benchmark for next-day wildfire spread prediction from 375 m VIIRS active-fire detections, with ERA5 weather, GFS forecasts and Alpha Earth terrain embeddings. 256×256 patches, 9 regions, 2017–2025. Code, documentation, loaders and paper: https://github.com/justuskarlsson/FireComp Contents Path Description next_day_v3/ Main dataset: 8 HDF5 shards (dataset_*.h5, zstd) + per-shard metadata (dataset_*.json)… See the full description on the dataset page: https://huggingface.co/datasets/justuskarlsson/FireComp.

sourceHugging Facecc-by-4.0updated 17h agoView on Hugging Face
0likes63downloads
Dataset Card

FireComp: Next-Day Fire Spread

Global benchmark for next-day wildfire spread prediction from 375 m VIIRS active-fire detections, with ERA5 weather, GFS forecasts and Alpha Earth terrain embeddings. 256×256 patches, 9 regions, 2017–2025.

Code, documentation, loaders and paper: https://github.com/justuskarlsson/FireComp

Contents

PathDescription
next_day_v3/Main dataset: 8 HDF5 shards (dataset_*.h5, zstd) + per-shard metadata (dataset_*.json), samples.json, stats.json (normalization)
next_day_v3/splits/{train,val,test}.jsonlFlat sample index per split (stratified temporal 60/15/25 within each region) — what the dataset viewer shows
next_day_v3_case_study/Small case-study subset used for paper figures
regions/Region definitions (wildfire_regions.json / .tif)
fire_areas.npzPer-fire spatial area lookup (fire-size stratification)

Usage

Clone/download into data/ of the GitHub repo and follow its README. Reading the shards requires h5py and hdf5plugin (zstd filter).