What we have validated.

The one California result we publish, stated exactly.

In a California validation of 87 destructive-fire events, PRECEDE flagged 73.6% (64/87) versus 50.6% (44/87) for a standard fire-weather-index threshold at the same 6.27% false-alarm rate.

PRECEDE scores point-precise weather at each location while the index uses 0.25-degree reanalysis; the asymmetry is a property of the comparison.

This is a California result. It is one operating point, pre-committed before it was scored, on a fixed 87-fire gold set.

• The score is an index, not a calibrated probability.
• PRECEDE's approach has been evaluated most rigorously in California, where the historical fire record is richest. Any accuracy figures we publish are specific to that California evaluation and do not transfer to other states.
• The underlying fuel-type term uses LANDFIRE LF2022 labels that post-date part of the fire panel. A checked pre-era relabeling changes no catch and no miss at this operating point.
• The live score includes an active-fire proximity term that the validated version did not have. On a day with a fire nearby, the live score can be higher than the validated score would have been; it cannot be lower.
• The validated model includes a known defect in its terrain input: the aspect term is applied with the opposite slope-facing convention from the one its bands assume, and the slope input is scaled higher than intended. The 73.6% result includes both as built. The correction is scheduled after the 2026 season as a pre-registered change and the result will be re-run.

How we hold ourselves to it

• PRECEDE pre-registered its 2026 season-end validation before any outcome was known (frozen 2026-08-27).
• Every published number was independently recomputed from the raw data by an adversarial internal audit (completed 2026-08-28); zero numbers broke.

Full technical documentation available on request.