Wildfire risk maps aren’t capturing the threat of extreme events and seasons

May 11, 2021

Firefighters work to keep flames from spreading through an apartment complex in Paradise, Calif. (AP Photo/Noah Berger)

Researchers have developed an approach for predicting how vulnerable different areas are to wildfires like the deadly one that raged through Paradise, California, in 2018.

The scientists compared computer simulations with historical observations to explore how communities across the western U.S. could be exposed to disastrous fire seasons and published their findings April 13 in Science of the Total Environment.

"There are lots more Paradises out there in the West — lots and lots more just waiting to happen," said Alan Ager, a research forester with the U.S. Department of Agriculture Forest Service's Rocky Mountain Research Station and first author of the study. "By illustrating these extreme events to communities, maybe we can convince people and land managers that they're prone also and help pinpoint where this is most likely to happen."

Ager and his colleagues noticed that many organizations were creating community risk maps in western states based on the average values for frequency and intensity of fires as predicted by computer simulations.

"I began to question, what do these colors really mean in an era of changing climate, extreme meteorological events and potential huge losses?" Ager said. "We don't live in an average world anymore."

Extreme wildfires, he and his team noted in the study, "are increasingly becoming a reality in many parts of the globe."

They decided to focus on rare extreme events and fire seasons described in the wildfire simulation models. Usually, risk maps are based on the average values of such simulated fires. Additionally, until now, Ager said, no one had really compared simulated events with data from real fires that burned into developed areas.

He and his team examined 54 million fires simulated over 10,000 hypothetical fire seasons, as well as 25 million building locations, to explore the size and frequency of blazes in extreme fire seasons and how many buildings the fires could reach in 11 western states. The researchers then compared these results with observations of actual burned areas from 1990 to 2018.

In the simulations, Ager said, "Weather data was sampled from historical conditions and 'replayed' to create replicate fire seasons, each day having different weather."

The largest burned area simulated for this period was nearly 32,000 square miles during a year that would have caused 1,782 fires and impacted 76,210 buildings. Another disastrous simulated season burned about half of that area but would have exposed 495,843 buildings to wildfires.

One particularly damaging fire would have reached 106,378 buildings in California and burned 1,817 square miles. For comparison, the Camp Fire — which destroyed the town of Paradise and killed more than 80 people, making it the deadliest wildfire in the Golden State's history — affected 19,558 buildings. 

Overall, the simulations offered pretty accurate predictions for the area of land burned annually during the years the researchers examined. The models overestimated the number of buildings that were exposed to wildfires, although more accurate patterns might have emerged had the researchers examined a longer period, Ager said. 

However, the most extreme simulated fire seasons outstripped historical fire seasons by 278% in terms of burned area, and 1,255% in terms of buildings exposed.

"It's not saying this fire is coming to you next year; it's saying that there are plausible fire scenarios that are way worse than anything we've seen, even without changes in contemporary climate," Ager said. "These extreme events are off the scale, and they're not well represented in risk maps."

In studies of past events, researchers investigated what caused particular homes to be exposed to or consumed by the flames. To do this, Ager said, they used a mapping classification system that broke parcels of land into discrete categories based on the building density and amount of vegetation. However, Ager and his team found that the number of buildings exposed to the simulated fires peaked in certain spots along continuous gradients that ranged from unpopulated forests to urban areas with little burnable vegetation. 

"As you move along gradients from wildlands and national forest into developed areas, there's going to be this optimal mix that allows fires to burn into developed areas, and an optimal number of buildings, such that exposure is optimized," Ager said. 

The technique his team developed could be used to provide more-detailed risk assessments for communities in fire-prone areas.

"Every community has got its optimum location on the fire path into town, where you're going to lose the most buildings," Ager said. "We can look in more detail at building exposure around developed areas, using higher-resolution data and fire-simulation information to predict where along gradients the optimal level of exposure is."

He and his colleagues are working on a digital atlas that will depict how fires can move through individual communities around the nation. Users will be able to zoom in on a community and see the footprints of past fires and management actions that have since been taken, the present risk level and the extreme events that could plausibly strike that area in the future. 

The registry, which will soon be made public, is currently being used by the U.S. Forest Service to determine where to prioritize forest restoration and hazard fuel reduction efforts, Ager said.

The study, "Predicting Paradise: Modeling future wildfire disasters in the western US," published April 13 in Science of the Total Environment, was authored by Alan A. Ager, Michelle A. Day, Karen C. Short and Isaac Grenfell, USDA Forest Service; Fermin J. Alcasena, USDA Forest Service International Visitor Program and Oregon State University; and Cody R. Evers, Portland State University.

We use cookies to improve your experience on our site and to show you relevant advertising.