Catastrophe models

While statistical and physical models can provide a view of how the hazard from wildfire may change in the future, in order for the information to be most useful to insurers, investors, and other groups with financial interests, the hazard information needs to be converted to monetary loss. Catastrophe models are an appropriate tool for effecting that conversion. These models typically simulate the financial losses from natural hazards (catastrophes), accounting for uncertainty in the year-to-year variability. The models can be built to reflect risk for a particular future climate (e.g., SSP or RCP at a time period of interest) but usually they are built to represent the current climate risk because that is a key concern for the insurance industry.

For the Wildfire Climate Change Projections for the U.S., Verisk conditions the existing Verisk Wildfire Model for the United States¹, which reflects the current climate, to create the Climate Change Projections, which reflect future climates of interest.

The Verisk Wildfire Model for the United States is an event-based model designed to capture the risk of wildfire to property within the 13 westernmost conterminous states (WA, OR, CA, ID, NV, MT, UT, WY, AZ, CO, NM, TX, and OK). The model simulates the behavior of all wildfires greater than 100 acres in size and uses stochastic weather data and weather-wildfire relationships to model stochastic wildfire activity. This process allows the model to simulate correlated countrywide wildfire activity and drives the inter-annual variability in fire behavior. The model relies on historical fire data to classify fire behavior in different ecological regions of the United States, and then creates stochastic fires for each ecological region. To simulate fire growth during an event, the fire-spread component of the model considers available fuel, fuel characteristics, terrain, and modeled wind conditions and explicitly captures all the major methods of fire spread (surface fires, canopy fires, and ember generation/spotting). Special focus has been placed on modeling how wildfires spread through the wildland-urban interface, accounting for suppression efforts. The vulnerability component of the model uses flame length, calculated within the spread model, as an intensity variable. Six secondary modifiers are included in the model, including two modifiers specific to wildfire (defensible space and Firewise USA™ communities).

There are three catalog versions of the Verisk Wildfire Model for the United States, the 10,000-year, 50,000-year, and 100,000-year catalogs of simulated events. Each catalog reflects the annual wildfire activity for the current climate, and accurately captures regional distributions and seasonal cycles. Any one of the synthetic years of activity are equally probable. Each stochastic year is generated using a perturbation of historical years between 1979-2015. The advantage of this strategy is that spatial structures corresponding to known climate variability (e.g., the El Niño-Southern Oscillation) are preserved in the model. The model uses an array of damage functions as well as a loss module to compute loss based on modeled fire intensity and building characteristics.