Defensible Catastrophe Modeling Methodology
Delphi-1.0 replaces backward-looking static catastrophe tables with high-resolution satellite telemetry, multi-frequency physics, and Tweedie-Poisson severity distributions.
Empirical Telemetry & Feature Architecture
NASA FIRMS & VIIRS Thermal Telemetry
375m active fire detection bands calibrated against historical CAL FIRE burn perimeters from 2000–2026. Provides live spatial proximity scoring to active incident fronts.
ECMWF ERA5 Climate & Fuel Moisture Reanalysis
Hourly reanalysis of surface temperature, relative humidity, and 10m wind gust vectors transformed into 100-hour and 1000-hour dead fuel moisture (FMC) curves.
USGS 3DEP High-Resolution LiDAR Topography
Digital elevation models resolving slope gradients, slope aspect (sun exposure), and topographic channelling effects that accelerate convective flame front velocities.
CDI Safer from Wildfires Structural Attributes
Granular parcel-level building characteristics including roof construction material, enclosed eaves, dual-paned windows, and 0–100ft defensible space clearances.
Compound Poisson-Tweedie Loss Estimation
Traditional linear models fail on heavy-tailed catastrophe risks with zero-inflated loss claims. Delphi-1.0 combines a Poisson occurrence rate λ(x) with a generalized Gamma severity distribution Γ(μ, σ) to calculate exact Expected Loss Cost (ELC):
Calibrated across 26 years of historical out-of-sample commercial claims to minimize underwriting loss variance while maximizing policy capture in defensible brush buffer zones.