Ember · reads the flame
Fire PML from physics, not buffer rings
Ember clusters buildings into burn-together fire areas using construction-conditional separation thresholds anchored to NFPA 80A and FM Global DS 1-20 — computed on real building footprints, not fixed-radius circles. Schedule PML is the worst single fire, found exactly.
What it reads
Construction-conditional clustering
A frame building ignites its neighbor at ~50 ft. A fire-resistive one needs near contact. Ember's pairwise thresholds vary by both buildings' ISO class — the physics the buffer-ring shortcut ignores.
Real building footprints
Edge-to-edge distances measured on Overture Maps building polygons in an equal-area projection — not centroid-to-centroid guesses.
Standards-anchored thresholds
The separation matrix is calibrated to NFPA 80A radiant-flux ignition limits and FM Global DS 1-20 exposure charts. Defensible numbers, not tuned knobs.
Terms-aware net PML
Deductibles, limits, margin clauses, coinsurance and blanket programs apply per cluster — gross and net PML in one pass.
How it works
- 1
Resolve every building
Schedule locations geocode and snap to real Overture Maps building footprints — actual polygons, not points, with synthetic fallbacks flagged honestly.
- 2
Measure edge to edge
Footprints reproject to an equal-area CRS and pairwise edge-to-edge distances compute exactly — the measurement buffer-ring methods approximate away.
- 3
Apply the physics
Two buildings burn together when their separation is inside a threshold conditioned on both constructions: frame-to-frame ~50 ft, fire-resistive needs near contact. Calibrated to NFPA 80A radiant-flux limits and FM Global DS 1-20.
- 4
Find the worst fire
Connected components form burn-together clusters; the schedule PML is the largest cluster loss — gross, or net of deductibles, limits, and program terms.
See Ember read a real schedule.
The live demo runs the production engine — instant showcase scenarios, plus your own addresses if you have demo credentials.
Open the live demo