At KatRisk, we’re excited to introduce our US Wildfire model: a fully probabilistic, CONUS-wide model that pairs Technosylva’s physics-based fire-science engine with KatRisk’s catastrophe and financial modeling. The result is real-world fire behavior translated into insurance-ready analytics for underwriting and portfolio decisions.
Wildfire has outgrown the old assumptions
Wildfire is the fastest-growing peril in global catastrophe losses, with insured losses rising roughly 12% a year. The January 2025 Los Angeles wildfires, Palisades and Eaton, are the largest insured wildfire event ever recorded at an estimated $40 billion. The US now averages about 7 million acres burned per year, more than double the 1990s average.
Wildfire is no longer a “Western states” problem or a “fire season” problem. Major fires now occur across the country and in every month of the year, and urban areas are increasingly in the path of catastrophic fire. Approaches built on historical loss and static hazard zones struggle to answer the question that actually matters to insurers: not just where fires have happened, but where they can happen next.
Fire science meets catastrophe modeling
KatRisk’s US Wildfire is built on more than 30 years of wildfire science. Its fire-spread engine is powered by Technosylva, whose models are used by major US utilities and fire-fighting agencies for real-world mitigation and emergency response. KatRisk wraps that fire-behavior science in insurance-grade catastrophe event generation, financial modeling, and portfolio analytics.
That combination is the point. Operational fire science tells you how a fire actually moves across terrain, fuel, and weather. Catastrophe modeling turns that behavior into the frequency, severity, and loss outputs underwriters and portfolio teams need. Our model has also been validated against the 2025 Palisades and Eaton fires.
The model covers one primary peril, the fire footprint itself, and one following peril, smoke, with burn and smoke footprints simulated per event including off-perimeter exposure.
How the model works
KatRisk US Wildfire is organized around three connected stages.
Ignition and fire size. A 50,000-year stochastic catalog of fire ignition points and sizes, built from historical fire and meteorological data and focused on fires larger than 1,000 acres. The catalog captures spatial clustering, seasonality, and climate signals across the country.
Fire spread and smoke. The Technosylva fire-spread engine runs each ignition to its modeled size across real fuel and terrain, producing roughly 22 million simulated fire footprints across the catalog. Embers and spotting are modeled explicitly, reflecting how fire leaps ahead of the visible front, and urban conflagration is handled directly, since structure-to-structure spread drives the total-loss neighborhoods seen in recent California and Hawaii disasters.
Vulnerability and mitigation. Damage functions translate hazard into loss, conditioned primarily on rate of spread and conditional flame length and refined by structure characteristics such as construction, roof type, building age, and defensible space. The model also captures the impact of mitigation, supporting underwriting and risk-reduction decisions with transparent scoring.
Built for how fire behaves today
A few things that set the model apart:
- Nationwide CONUS coverage. Fully probabilistic stochastic event coverage across the country, not just the traditional Western states.
- An operational fire-science spread engine. The same class of physics-based fire-behavior modeling trusted by utilities and fire agencies, scaled for insurance.
- Climate-aware risk. Wildfire risk is linked to KatRisk’s stochastic precipitation and sea-surface-temperature modeling, so El Niño and La Niña signals propagate through to fire risk, alongside forward-looking climate scenarios.
- Real-time event response. Near real-time monitoring and analytics to support decision-making during active wildfire events.
Resolution and outputs
Fire spread is modeled on a 30m grid and mapped to a multi-resolution grid, as fine as 60m in populated areas and as coarse as 1km in low-population areas, with real-time events resolved up to hourly. The model produces the full set of analytics insurers expect, including average annual loss, exceedance probability curves, event and year loss tables, and return-period hazard and loss, at property, account, and portfolio scale.
The model is also built to be transparent and customizable. Clients can interrogate the underlying data and incorporate their own claims history.
Why it matters for insurers
For teams writing wildfire risk, the value is the move from static hazard zones to dynamic, behavior-based risk. That means a clearer property-level view of where catastrophic fire can occur, an understanding of the conditions that drive ignition and escalation, and a way to see how wildfire risk accumulates across a portfolio, grounded in field-tested, peer-reviewed fire science.
To learn more about KatRisk’s US Wildfire model, request a demo.
