HISTORICAL EVENT INTELLIGENCE · WILDFIRE (UNITED STATES)
Wildfire Data
1,925 US fires since 2016 with burned acreage and mapped perimeters, enriched from federal interagency sources — plus containment timelines, SIT-209 impacts, and satellite detections. Download the free eight-fire sample.
Read this first: wildfire enrichment is United States only
The authoritative sources behind burned acreage and perimeters — NIFC/WFIGS, GeoMAC, IAFPH, FRAP, FPA-FOD — are US federal. Our global wildfire catalog holds 45,471 hazard records, and the large majority of them are outside the United States: Australia, Canada, Russia, Brazil and others. Those records carry hazard metadata but no burned-area enrichment. Everything else on this page describes the US subset.
What's in the dataset
Fires joined to authoritative federal records, not alert polygons with a fire label.
Final burned acreage
From the federal interagency spine, with per-record source attribution — not alert-polygon area
Mapped perimeters
Dated polygons as WKT; through 2019 these form a growth time-series, one per fire from 2020
Containment timeline
Discovery, containment percentage, and control dates where reported
SIT-209 impacts
Structures destroyed and threatened, estimated cost, personnel, injuries — on 1,087 fires
Satellite detections
NASA FIRMS daily detection counts and maximum fire radiative power (MODIS, VIIRS)
Cause & match provenance
Reported cause, plus how each hazard was joined to the authoritative fire record
Counts are as of September 2026 and grow as the season accrues. Of the 1,925, exactly 1,088 are matched to the authoritative fire record by interagency key; the remainder are matched by perimeter geometry or observations. Counts describe distinct fires, deduplicated by fire identity — the platform sometimes tracks one fire as several hazard records.
Two eras of perimeter data
A seam worth knowing before you build on it — so here it is on the product page, not buried in a footnote.
Perimeter growth time-series
Dozens of dated polygons per fire showing the footprint expanding. Silver Creek in 2018 runs from 5 acres on 19 July to just over 20,000 across 73 observations, each with acreage and the change from the previous one.
One perimeter per fire
GeoMAC, the source of multi-observation history, was decommissioned in 2020. Its WFIGS successor is currently ingested as a single current or final perimeter. Modern perimeter history exists upstream and restoring it is on our roadmap.
That is a change in the upstream source, not a quality difference between the fires. We would rather you learn it here than discover it in your own analysis.
Start with the free sample
Eight fires, seven states, 2017 to 2026 — deliberately spanning both perimeter eras.
| Fire | Where | Acres | Notable |
|---|---|---|---|
| Silver Creek | CO 2018 | 20,127 ac | 73 perimeter observations — 5 acres to 20,000 |
| Ferguson | CA 2018 | 96,901 ac | 76 observations, 75 SIT-209 reports |
| Dollar Ridge | UT 2018 | 70,059 ac | 438 structures destroyed |
| Diamond Creek | WA 2017 | 129,000 ac | 44 observations, 8,757 FIRMS detections |
| Park | CA 2024 | 429,603 ac | 709 structures, 27,014 detections |
| Wapiti | ID 2024 | 129,063 ac | 30,629 detections |
| Dragon Bravo | AZ 2025 | 145,504 ac | Coconino County, on the Grand Canyon North Rim — 113 structures |
| 0445 Crosswhite | OR 2026 | 292,161 ac | No SIT-209 yet — current-season lag, shown not hidden |
Delivered as authoritative NDJSON plus flattened CSVs — hazards.csv, perimeters.csv (250 rows, every one carrying geometry — the observation counts above plus one archival final perimeter per fire), sit209.csv (567 rows), firms_detections.csv (460 rows) — all joined on a stable hazard identifier, WGS 84, geometry as WKT.
One caution we put in writing
The mistake this data invites, and how to avoid it.
The area field on a hazard record is not burned area. It is the alert-polygon footprint, and it differs from the true fire perimeter by large, non-constant factors. Use final acreage or the perimeter geometry. Substituting the alert area would produce confidently wrong acreage for exactly the underwriting and loss users who can least afford it — which is why it is stated here, in the dataset README, and in the field documentation rather than left to be discovered.
Two more worth knowing: null means not observed or not reported, never zero — current-season fires often have no impact reports yet. And authoritative archival perimeters are published retrospectively, so a 2018 fire can legitimately carry a final polygon dated 2021.
Two ways to get it
The same wildfire records, served for history or for right now.
Historical datasets
US wildfire extracts scoped to your geography and date range, with acreage, perimeters, containment and impacts — by subscription or one-time purchase.
Historical Event Intelligence →Real-time wildfire monitoring
Active fires worldwide as they are detected — satellite heat detections and agency-tracked incidents, with exposure.
Wildfire Intelligence →Watching something right now? Live wildfire activity shows what the feed is tracking this hour. Related reading: why a satellite fire detection is not yet wildfire intelligence.
Wildfire Data — FAQ
Where can I download historical wildfire data?
A free eight-fire sample is available with no signup, covering 2017 to 2026 across seven states: Diamond Creek, Dollar Ridge, Ferguson, Silver Creek, Park, Wapiti, Dragon Bravo and 0445 Crosswhite. It includes final acreage, dated perimeter polygons, containment and discovery timelines, SIT-209 incident reports, and NASA FIRMS satellite detections as NDJSON plus flattened CSVs. Larger US wildfire extracts are licensed through Historical Event Intelligence.
How much wildfire data is actually enriched?
As of September 2026, 1,925 distinct US fires from 2016 onward carry both burned acreage and at least one mapped perimeter, and 1,087 of those also carry SIT-209 impact data — structures destroyed, estimated cost, and personnel. Put another way: of US hazard records since 2016, about 2,430 resolve to an identified fire, and 79% of those carry both burned acreage and a mapped perimeter. That is the US subset of a global wildfire catalog of 45,471 hazard records. Wildfire records outside the United States carry hazard metadata but no burned-area enrichment.
Why is wildfire enrichment United States only?
Because the authoritative sources are US federal: NIFC/WFIGS, GeoMAC, IAFPH, FRAP and FPA-FOD. There is no equivalent single interagency spine for Australia, Canada, Russia or Brazil, which together account for a large majority of wildfire records in the catalog. We would rather scope the claim honestly than imply global burned-area coverage we do not have.
Does the data include fire perimeter progression over time?
For fires through 2019, yes — dozens of dated perimeter polygons showing the footprint expanding. Silver Creek in 2018 runs from 5 acres to just over 20,000 across 73 observations. From 2020 onward, each fire carries exactly one perimeter. That is a source change, not a data-quality difference: multi-observation history came from GeoMAC, which was decommissioned in 2020, and its WFIGS successor is currently ingested as a single current or final perimeter per fire. Modern perimeter history exists upstream and restoring it is on our roadmap.
Can I use the area field as burned area?
No — and this is the single most important caution on this page. Where an area field appears on a hazard record, it is the alert-polygon footprint, not the fire perimeter. The two differ by large and non-constant factors. Use final acreage or the perimeter geometry for burned area, never the alert-polygon area. Substituting one for the other produces confidently wrong numbers for exactly the underwriting and loss users who can least afford them.