Fire-regime risk summaries
Purpose
ws4.fire.risk summarizes simulated fire events as a risk distribution. It reuses the
P187 scfm formulas but reports the behaviour of a process run rather than validating a
published toy case.
Inputs
The summary reads event-log-like records with year, event_type, area_ha, and
metadata.scenario_id. Only fire events with an area contribute. A fire event without
an area is an error.
Outputs
The summary reports:
annual burned area, fire count, escaped count, and sizes;
total, mean, median, 90th-percentile, and maximum annual burned area;
burn probability and annual burn rate;
fire return interval;
fire counts and escaped fires per year;
mean, median, 90th-percentile, and maximum fire size;
optional P187 target diagnostics when a target and historical sample are supplied.
A fire return interval is None when nothing burned.
Reference case
tests/test_fire_risk.py uses four years, 1000 ha, 10 ha cells, and fire sizes 5, 20,
100, 1, and 2 ha. It asserts total burned 128 ha, mean annual burned 32 ha, median 14 ha,
90th-percentile 77.5 ha, burn probability 0.75, FRI 31.25 years, mean fire size 25.6 ha,
median 5 ha, and 90th-percentile 68 ha.
Limits
Deterministic summary only; no fire-growth simulation or calibration.
Target diagnostics need an explicit historical size sample.
Separate from event-perimeter validation.
LandMine FRI/reburn invariants remain a follow-up.
Process ensemble integration:
ProcessEnsembleResult.fire_regime_summaries.Optimizer account bridge:
ws4.opt.fire_risk.fire_regime_account_values.Process risk report:
ws4.opt.fire_risk.evaluate_process_fire_risk.Report artifact:
ProcessFireRiskReport.write(values, aggregated, JSON).Train/validation:
evaluate_process_fire_risk_validationandProcessFireRiskValidation.write.Combined bundle:
write_risk_verification_bundle.Existing optimizer output:
bundle_existing_risk_verification.Explicit account comparison:
ws4.opt.account_compare.compare_accounts.Shared-schedule integration test:
tests/test_process_estate_shared_schedule.py.