firestorm-spread-data

Predicted fire spread for FIRESTORM — the leap from "where is the fire now" to "where is it going." Mirrors the open PyreCast (Pyregence Consortium) ELMFIRE fire-spread forecasts into a slim GeoJSON the single-file frontend reads.

What it gives FIRESTORM

For each active fire PyreCast models, a predicted spread extent with an uncertainty envelope — the 10th / 50th / 90th ensemble percentiles (core / likely / outer cone). Rendered as a MODELED overlay so an operator can see the projected footprint, not just current heat.

Source + licensing

data.pyrecast.org/fire_spread_forecast/<fire>/<run>/pyretec/landfire/<pct>/isochrones_*.shp (UTM per-fire zone; reprojected to WGS84 here).

Licensing — important: PyreCast Terms §VI explicitly permit "Public safety applications by emergency management, fire service organizations, and government agencies" free of charge. FIRESTORM (DOI) qualifies. Every rendered feature is attributed to PyreCast/Pyregence. Operator authorized the agency-use reading for display; written redistribution permission is being pursued in parallel.

Demo-data honesty: every polygon carries model=ELMFIRE + run timestamp + percentile so the frontend badges it MODELED / PREDICTED with an uncertainty cone — never ground truth. A forecast is decision-support; accuracy is bounded by inputs (wind-dominant).

Output — data/spread.json

{ "generated_at":"...", "model":"ELMFIRE", "percentiles":["10","50","90"],
  "fires":[ {"fire":"ca-saddle","label":"Saddle (CA)","run_at":"...","rings":{"10":[[ [lng,lat],... ]],"50":[...],"90":[...]}} ] }

Frontend reads raw.githubusercontent.com/Deasus/firestorm-spread-data/main/data/spread.json.

Run locally

pip install pyshp pyproj      # pure-Python, no GDAL
python fetch_spread.py

Walks runs newest-first per fire (the very newest is sometimes still computing/empty → falls back). Keeps the largest rings per percentile, coarsened, to keep the payload ~200KB. 20-min GHA cron.