Why Wildfire Models Must Estimate Return Periods — and Why Most Still Don't
- 1. Where Fires Burn ≠ Where Risk Is
- 2. Where Assets Are Changes Everything
- 3. Burned Area ≠ Loss
- 4. Why Tail Risk Dominates Wildfire Losses
- 5. Climate, Development, and Feedback Loops
- 6. The Modeling Gap: What’s Missing Today
- 7. Why Return Periods Are the Missing Link
- 8. Toward Return-Period–Aware Wildfire Models
- Closing Thought
Wildfires are no longer just an ecological concern. They are a financial, insurance, and infrastructure risk—and increasingly, a tail-risk problem. Yet most wildfire models still answer the wrong question.
They tell us where fires burn, not how often catastrophic losses occur.
That gap matters.
Below, drawing on Slides 2–10 of the presentation, I explain why return-period–aware wildfire modeling is essential—and why its absence remains a critical blind spot in current research and practice.
1. Where Fires Burn ≠ Where Risk Is
Most wildfire maps focus on burned area.
Using county-level burned fractions (burned area ÷ county area), we see familiar patterns:
- Pacific Northwest
- Northern California mountains
- Central Idaho and interior Nevada
- Parts of Florida and the Interior South
These regions burn frequently and extensively. But many are rural with limited economic exposure.
If risk were proportional to burned area, these places would dominate wildfire losses.
They don’t.
2. Where Assets Are Changes Everything
When we map property exposure—using housing units multiplied by median home value—the geography flips:
- Coastal California
- Major metros
- Growing Wildland–Urban Interface (WUI) corridors
- Florida’s coastal counties
Economic assets cluster away from the places with the most fire, but exactly where losses explode when fire does occur.
3. Burned Area ≠ Loss
When we combine fire frequency with exposure—estimating annual property value at risk—a different set of hotspots emerges:
- Pacific Northwest WUI zones
- California coastal and metro regions
- Florida coastal counties
- Rocky Mountain WUI corridors
Meanwhile, places like interior Nevada or central Idaho—dominant in burned-area maps—fade once asset weighting is applied.
Big fires in empty places don’t bankrupt insurers. Medium fires in asset-dense places do.
4. Why Tail Risk Dominates Wildfire Losses
The world’s costliest wildfire events illustrate this clearly:
- The 2017–2018 California fires produced tens of billions in insured losses
- Recent Los Angeles–area fires generated some of the largest wildfire losses globally, with insured losses estimated by Swiss Re near $40B and total economic losses estimated by AccuWeather at $250–275B
These were not the largest fire years by area burned.
They were rare, high-impact events hitting dense, expensive landscapes.
This is why insurers, reinsurers, and capital markets care about return periods, not averages.
5. Climate, Development, and Feedback Loops
Losses scale nonlinearly because multiple forces reinforce each other:
- Climate and fuel aridity
- Rapid WUI development
- Property value inflation
- Suppression limits during extreme weather
- Building codes and defensible-space variability
These feedbacks create fat-tailed loss distributions—exactly the regime where return periods matter most.
6. The Modeling Gap: What’s Missing Today
Most wildfire models do one of the following:
- Predict burned area
- Predict ignition probability
- Simulate fire spread for individual events
What they don’t do well is connect:
- Probability of large fires
- Spatial spread under weather extremes
- Interaction with urban form
- Resulting economic loss
- Frequency of catastrophic outcomes
Without that chain, we cannot answer:
Is this a 1-in-20 year loss… or a 1-in-200 year loss?
That distinction is everything for insurance pricing, capital adequacy, and resilience planning.
7. Why Return Periods Are the Missing Link
Return periods allow us to:
- Price insurance and reinsurance realistically
- Stress-test portfolios under rare fire years
- Compare wildfire risk across regions consistently
- Evaluate mitigation investments (codes, buffers, suppression)
- Align wildfire risk with other natural hazards (floods, hurricanes)
Yet return-period estimation is still rare in wildfire research, largely because it requires integrated, probabilistic, end-to-end modeling—from ignition to loss.
8. Toward Return-Period–Aware Wildfire Models
The path forward requires combining:
- Large-fire probability models
- High-resolution fire spread models
- Urban exposure and vulnerability models
- Ensemble weather and climate drivers
- Explicit loss modeling
Only then can we move from “where fires happen” to “how often catastrophic losses occur.”