146 wildfire alerts.
Three of them were fires.
21 continuous days correlating NASA satellite hotspots against CAL FIRE's incident list. One fire surfaced 18.4 hours before it reached the public record. The 143 that were not fires are on this page too.
Where the heat is
Hours ahead of the public list
Lead is measured to the moment the incident appeared on the public feed — never to ignition. Baseline: 162 incidents, median 0.9 h, p90 14.1 h. Only Dutchflat clears p90. Timber is the strongest evidence here and it is not ours: CAL FIRE first published a 17:44Z start, then revised it back to 03:35Z four days later — their own record moving 14 hours to agree with a hotspot we had already flagged.
And the 143 that were not
One square per alert, in order. Conversion is 2.1%. Most were agricultural or prescribed burns, industrial heat, or satellite artefacts — which is what an instrument tuned to surface a candidate rather than miss one produces, and what it should say out loud.
- Scope21 days, one region, three matches. A signal, not a validated rate — and nothing here is an accuracy figure.
- Match ruleA named incident within 5 km not already listed. At 10 km the count rises to 22, dominated by one 94,105-acre fire absorbing twelve alerts at 8.5 km — measured and discarded as coincidence.
- PrecisionA VIIRS pixel is ~375 m. Marker positions carry that uncertainty and no more.