FireWx-FM: Wildfire Forecasts for the Contiguous United States

Spatial wildfire forecasts from FireWx-FM for the contiguous United States (CONUS). This dataset preserves the fire-risk maps published on the Fire-Eye platform, with forecast times and supporting metadata for research, comparison and reproducible analysis. Data at a glance Coverage: contiguous United States; nominal 5 km grid. Forecasts: +12, +24, +36 and +48 hours; each map describes a 12-hour window. Format: GeoTIFF; NAD83 / Conus Albers (EPSG:5070); scores from 0 to 1. Snapshot: 2026-09-23; 171 forecast maps. Issue times: Jul 7, 2026, 12:00 UTC to Sep 23, 2026, 0:00 UTC. Available maps by lead: +12 h: 138, +24 h: 11, +36 h: 11, +48 h: 11. Earlier dates do not necessarily include all four lead times; the catalog lists the available maps. Download and use Start with README.md. Open forecast_catalog.csv to find forecast dates and lead times, then download the corresponding FireWx-FM_CONUS_<start-date>_to_<end-date>.zip. Each ZIP opens independently and contains GeoTIFF maps and its own catalog. Dates in ZIP names refer to forecast issue dates in UTC. GeoTIFFs can be opened in QGIS, ArcGIS or Python. The JSON catalog and checksum manifest support automated reuse. Interpreting the forecasts Higher scores indicate greater modeled active-fire activity within a grid cell and 12-hour window. They are not calibrated ignition probabilities or predictions of individual fire size or spread. Values preserve the published platform maps and may include temporal smoothing. Versions and timing Releases preserve the forecast revisions captured by the archive process. Issue time, recorded availability and first capture time are provided separately. For evaluating advance predictions, use availability and capture times as well as issue time; historical backfills are retrospective results. A release DOI dates the archive publication, not the original availability of each forecast. Citation and license Use the citation and version DOI shown on this page when reporting results from this release. The dataset is available under Creative Commons Attribution 4.0 (CC BY 4.0).