Hazards of risk: Using wildfire simulations to identify and plan for plausible extreme events

Background/Question/Methods

The wildfire problem is so pervasive across the western United States, that state, regional, and national managers face difficult decisions about where to allocate scarce resources to achieve maximum community risk reduction in preparation for future wildfires. Quantitative wildfire risk assessments are a common decision support tool, using wildfire simulations to characterize average annual risk and providing managers with insight into the relative likelihood and potential consequences of wildfires across a planning landscape. However, using annual average metrics may unintentionally downplay risk characterization in fire regimes characterized by infrequent wildfires where the annual likelihood of wildfire is miniscule. For instance, much of western Oregon and Washington in the Pacific Northwest (PNW) region of the United States has not experienced wildfire in well over a century, largely as a result of regional climate patterns, but in September 2020 multiple very large, synchronous wildfires burned approximately 700,000 hectares, destroyed nearly 3,000 structures and caused the evacuation of 90,000 people. We used a regional set of wildfire simulations based on contemporary climate and landscape conditions and Microsoft Building Footprint data to identify and characterize plausible extreme wildfires in western Oregon and Washington which would result in very high community structure exposure. Our aim was to determine whether or not the simulations included any plausible events comparable to or more extreme than the 2020 wildfires, the most destructive on record. In addition, we evaluated and characterized plausible extreme wildfires in communities of western Oregon and Washington which have not experienced direct effects of wildfire in the past half century.

Results/Conclusions

The highest exposure simulated wildfires ignited on private land, burned relatively close to communities and were 500% smaller on average compared to historical high exposure wildfires (1984-2020) which generally ignited in remote locations on public land. Simulations did not project any multi-fire events comparable to the 2020 fires in terms of area burned or magnitude of structure exposure. Simulations did illustrate plausible extreme events in 25% of western Oregon and Washington communities whereas less than 1% of communities experienced extreme fires historically. Simulated extreme fires did not precisely forecast all historical events, but nonetheless can appropriately be used to motivate public outreach and disaster planning in communities for which wildfire risk might otherwise seem an improbable concern. While our work assessed community exposure, similar approaches may be useful for evaluating risk to critical habitat, source water areas, or other resources adversely impacted by extreme fires.