The warning came a year early
Walking through Paradise, California, in late November 2018, you wouldn’t recognise the once peaceful, suburban town. Nestled among the pines, its relatively modest house prices had attracted families and retirees alike. Where homes once stood, now lay piles of charred rubble. Roughly 90% of the town was levelled, and at least 85 lives were lost in the Camp Fire.
The immediate causes of the fire were well understood. Strong winds, exceptionally dry fuels and an ignition source allowed it to spread rapidly through the town. But why had those fuels become so combustible, and what did that reveal about wildfire risk?
Wildfire is unusual because the conditions that determine its severity develop over months, sometimes years, before a fire even ignites. Weather influences how much vegetation grows, how thoroughly it dries and how readily it can burn. The hazard is therefore shaped not only by today's weather, but by the landscape's accumulated response to previous seasons.
When insurance stopped following the risk
The insurance market responded in a way that few homeowners expected: with a gradual withdrawal of insurance cover. Between 2019 and 2023, the number of homeowner policies in extreme fire risk areas of California dropped by almost 20%, whilst low fire risk zones saw a slight increase in policy numbers.[1] Over the same time period, the number of FAIR plan policies (the state's insurer of last resort) in very high-risk wildfire ZIP codes more than tripled. If wildfire risk had clearly increased, why didn't insurers simply charge more for it? The answer lay in California's regulatory framework. Premiums were tied to historical losses rather than reflecting current levels of risk. Without the ability to increase rates to reflect the changing climate, insurance rates were artificially suppressed, prompting some to withdraw from the market altogether.
The next test: Los Angeles, 2025
Then came the next test. This time it wasn’t the forests of Northern California, but the hills above one of the wealthiest metropolitan areas in the country. The 2025 Palisades and Eaton fires became the costliest wildfire event in insurance history. The fires arrived just days after the first wildfire catastrophe model was submitted under California’s new ratemaking framework, a milestone that would allow insurers to assess wildfire risk using simulated future scenarios instead of historical losses alone. This marked an important development in how wildfire risk could be priced, but the Los Angeles fires also demonstrated that even forward-looking models must continue to evolve as the hazard itself changes.
Why wildfire remembers
The LA wildfires themselves were also another important reminder of why static, backward-looking approaches have struggled, and why wildfire differs from many other perils. To understand why Los Angeles burned in January 2025, you have to look back over the preceding two years. Two unusually wet winters had encouraged extensive vegetation growth, which was then drained of its moisture by California’s hottest summer on record. Autumn rains failed to materialise and severe drought conditions set in. By the time hurricane-force Santa Ana winds arrived in January, the landscape was loaded and primed to burn.
Climate scientists call this "hydroclimate whiplash". As Willis Research Network wildfire researcher Dr. Matthew Jones explains, it begins with "the accumulation of excessive stocks of vegetation through wet periods" that are then followed by "very dry conditions and extreme droughts". The "real crack of the whip" comes when the atmosphere becomes exceptionally dry, rapidly drawing moisture from vegetation until it becomes extremely flammable.
Reading the landscape in real time
This presents a challenge for catastrophe models. A model that captures fuel conditions at a single point in time is inevitably missing part of the story, because the landscape itself is constantly changing. Any model built around a static snapshot is gradually drifting away from reality as vegetation grows, dries and recovers between fire seasons.
20 years of the Willis Research Network
Related content
This is why wildfire modelling is increasingly moving towards dynamic approaches that track how landscapes evolve over time. One example is work underway between the Willis Research Network and Pinepeak which explores the impact of dynamically changing fuel loads on wildfire spread. By continuously ingesting updated vegetation and fuel moisture data, the model can reflect the risk landscape as it evolves across seasons, rather than locking in a static picture that ages from the moment it is produced.
As dynamic, memory-aware modelling starts to gain traction, the brokers and carriers who embrace it — and can sit with the discomfort of year-to-year changes in risk — will be the ones best placed to navigate whatever the next fire season brings. The landscape has always been changing, but our models are only now beginning to change with it.
Key takeaways
- The Camp Fire exposed the limitations of pricing tomorrow's wildfire risk using yesterday's losses.
- Wildfire risk evolves continuously as landscapes respond to changing weather and climate. Fires over the last decade showed that even forward-looking models need to account for the long-term memory of the landscape.
- Dynamic catastrophe models that account for changing fuel conditions represent the next step in wildfire risk assessment.
Footnote
Contacts
One example is work underway between the Willis Research Network and Pinepeak which explores the impact of dynamically changing fuel loads on wildfire spread.
