Observed conditions, qualified forecasts, and the gaps between them.
Median reporting lag 0.3 h — half of these fires were last reported more recently than that, half longer ago.
Measured
170 fires
Site weather, fuel dryness and the ground inside the fire's own perimeter all went into the estimate.
Partial
5 fires
Some site data was available and the rest of the model ran on neutral values, which widens the range rather than shifting it.
Modelled
0 fires
Size and complexity only. These estimates are the weakest on the page and are labelled as such in the table.
Reporting lag
0.3 h
An estimate cannot be fresher than the acreage it was computed from. NIFC publishes in batches, so this is the floor under everything here.
Multiple words all have to match. Press / to jump to the search box, Esc to clear. Click any column heading to sort.
Type rates the incident's complexity; Management Type is the team assigned to run it. They differ on about a third of fires, and roughly half report neither.
Est. days left — median days remaining, with the 25th–75th percentile beneath. Hover for the drivers. measured — site weather and fuel dryness partial — some site data modeled — size and complexity only
| Fire↕ | ST↕ | Acres↕ | Contained↕ | Type↕ | Management Type↕ | Est. days left↕ | Personnel↕ | Age↕ | ERC↕ | Slope↕ | Agency↕ |
|---|
Sources
Four public feeds, pulled three times a day.
Fire perimeters and incident records from NIFC WFIGS, weather from NWS api.weather.gov, fire danger from the national ERC series, and terrain from USGS 3DEP. No third-party requests leave this page.
Method
A survival model, not a countdown.
Days remaining is the median of a lognormal fitted to historical closures by complexity, then adjusted for containment, size, observed growth, fuel, dryness and forecast rain. The range beneath each estimate is the middle half of that distribution.
Limits
It under-predicts, and we publish by how much.
On fires that have since ended, the model's days-remaining estimate has run short, and it is worst on the longest fires. The record page carries the measured error and the share of outcomes that fell inside the published range.
Not here
No rates, no margins, no crews.
This build is assembled from a database holding none of them, so there is nothing commercial to leak rather than a filter that might be forgotten.