# Wildfire — United States, 2017–2026 (sample dataset)

Eight enriched US wildfire incidents from the DisasterAWARE Enterprise historical hazard catalog, produced by the DataFeeds wildfire enrichment pipeline. Each record is a complete hazard object joined to authoritative federal interagency fire data: final burned acreage, dated fire perimeters, containment and discovery timelines, SIT-209 incident-management reports, and satellite fire detections. This sample is published to show exactly what the enrichment contains — including where its coverage changes over time.

**Events:** 8 wildfires across 7 states (AZ, CA, CO, ID, OR, UT, WA) · **Date range:** 2017–2026 · **Coordinate reference:** WGS 84 (EPSG:4326) · **Geometry encoding:** WKT

| Fire | State | Year | Final acres | Perimeter observations | SIT-209 reports | FIRMS detections | Structures destroyed |
|---|---|---|---:|---:|---:|---:|---:|
| Diamond Creek | WA | 2017 | 129,000 | 44 | 60 | 8,757 | 3 |
| Dollar Ridge | UT | 2018 | 70,059 | 45 | 61 | 6,880 | 438 |
| Ferguson | CA | 2018 | 96,901 | 76 | 75 | 15,912 | 11 |
| Silver Creek | CO | 2018 | 20,127 | 73 | 91 | 3,256 | 1 |
| Park | CA | 2024 | 429,603 | 1 | 107 | 27,014 | 709 |
| Wapiti | ID | 2024 | 129,063 | 1 | 90 | 30,629 | 19 |
| Dragon Bravo | AZ | 2025 | 145,504 | 1 | 83 | 33,646 | 113 |
| 0445 Crosswhite | OR | 2026 | 292,161 | 1 | — | 38,921 | — |

## Read this first: two eras of perimeter data

The four 2017–2018 fires carry a **perimeter growth time-series** — dozens of dated perimeter polygons showing the fire's footprint expanding over its lifetime. Silver Creek, for example, goes from 5 acres on 19 July 2018 to just over 20,000. The four 2024–2026 fires carry **exactly one perimeter each**.

That is not a quality difference between the fires; it is a change in the upstream source. Multi-observation perimeter history came from **GeoMAC**, which was decommissioned in 2020. Its successor, the NIFC **WFIGS** service, is the source from 2020 onward, and this pipeline currently ingests a single current/final perimeter per fire from it. Modern perimeter history exists upstream and restoring it is a known item on our roadmap; until that lands, **no fire after 2019 in this catalog has perimeter progression**. We would rather you learn that here than discover it in your own analysis.

## Known limits — please read before building on this

- **United States only.** The enrichment sources (NIFC/WFIGS, GeoMAC, IAFPH, FRAP, FPA-FOD) are US federal. Wildfire records elsewhere in the catalog — Australia, Canada, Russia, Brazil and others, which together are the large majority of wildfire records — carry hazard metadata but **no burned-area enrichment**.
- **`area_km2` is NOT burned area.** Where an `area_km2` field appears on a hazard record anywhere in this catalog, it is the **alert-polygon footprint**, not the fire perimeter. The two differ by large and non-constant factors. Use `final_acres`, or the perimeter geometry, for burned area. Never substitute `area_km2`.
- **Acreage and perimeters are present on roughly half of enriched US fires**, not all. This sample deliberately shows fires where the join succeeded and the geometry is complete.
- **Impact fields (structures, cost, personnel) come from SIT-209** incident-management reports and are sparse for the current season — the 2026 fire here has none yet, which is normal for recent incidents and is shown rather than hidden.
- **One fire can have more than one hazard record.** The platform sometimes tracks a single fire as several hazards; 440 fires in the catalog do, one of them as 17 records. This sample is deduplicated to one record per fire. When counting, say "hazard records", not "fires".

## Files

- **`wildfire-us-2016-2026.ndjson`** — authoritative, full fidelity. One complete enhanced hazard object per line (8 lines), in the same V2 shape the historical hazards API serves: core hazard fields plus the `fireAssessment` and `fireDetails` enrichment blocks. Every record validates clean against the published hazard contract. Everything in the CSVs derives from this file.
- **`wildfire-us-2016-2026-csv-bundle.zip`** — flattened relational views, joined on `hazard_uuid`:
  - `hazards.csv` (8 rows) — one row per fire: identity, location, dates, final acreage, cause, containment, match provenance, and summary impact/detection counts.
  - `perimeters.csv` (250 rows) — one row per perimeter observation: timestamp, source (GEOMAC / WFIGS_PERIM / IAFPH / FRAP), kind (`snapshot`, `current`, `final`, `final_perimeter`), acreage, acreage change from the previous observation, and the polygon as WKT. **This is where perimeter growth lives.** Every row in this bundle carries geometry.
  - `sit209.csv` (567 rows) — one row per SIT-209 report: acreage, containment %, personnel, structures destroyed and threatened, injuries, fatalities, estimated cost, and incident-management narrative remarks.
  - `firms_detections.csv` (460 rows) — one row per fire per day of satellite detections: detection count and maximum fire radiative power (MW), with contributing sensors and search radius.
  - `README.md` — this file.

## Field notes

- `hazard_uuid` is the stable identifier and the join key across every file.
- Timestamps in the NDJSON (`startedAt`, `endedAt`) are epoch seconds, UTC. CSV timestamps are ISO-8601 as reported by the source.
- `acres_change` in `perimeters.csv` is computed within this bundle as the difference from the previous observation of the same fire; null on the first observation and on final-perimeter rows.
- **Final-perimeter timestamps can postdate the fire by months or years.** Authoritative archival perimeters (IAFPH, FRAP) are published retrospectively; a 2018 fire may carry a final perimeter dated 2021. This is the publication date of the authoritative polygon, not fire activity.
- 45 polygons in this bundle were repaired for OGC validity (simplification had left degenerate rings that break some GIS tooling). Geometry is otherwise exactly as enriched.
- `match_status` describes how the hazard was joined to the fire record: `CONFIRMED` (authoritative key match), `PERIMETER_ONLY`, `OBSERVATIONS_ONLY`, `UNMATCHED`. All fires in this sample are `CONFIRMED`.
- **Null means not observed or not reported. It does not mean zero.**

## Provenance

Enrichment source: a fire identity spine built from WFIGS, GeoMAC, IAFPH, FRAP and FPA-FOD, joined to DisasterAWARE hazard records. Satellite detections from NASA FIRMS (MODIS, VIIRS S-NPP, VIIRS JPSS-1). Incident reports from SIT-209. Records were generated by the DataFeeds wildfire enrichment run of 2026-07-31 and converted to the V2 hazard shape.

Questions, or a coverage need this sample doesn't answer: get in touch — we would rather scope against your actual use than have you infer it from eight fires.