When a wildfire breaks out, how do emergency management officials decide when residents should evacuate?
“Wildfire evacuation decision-making is challenging because the incident commanders need to take into account fire progression, population distribution and evacuation traffic,” explained Dapeng Li, an assistant professor in the Department of Geography at South Dakota State University.
Li’s research uses data-based models to help emergency managers make those decisions. Specifically, he uses fire spread modeling to predict how fast the fire will travel and employs traffic simulation model to estimate how much time the people need to evacuate to safe places.
“It’s the coupling of the two models that helps us develop a better understanding of the dynamics and mechanisms behind the evacuation process. It’s more complex than we thought,” said Li. The wildfire evacuation modeling research was part of Li’s doctoral work at the University of Utah. His work is published in a special issue on fire evacuation modeling in Fire Technology.
“My work is directly related to public safety—it’s about survival. However, very few scholars are doing this type of research in the United States due to its interdisciplinary nature,” Li explained. With the number of wildfires increasing and more people building homes in forested areas, deciding when residents must evacuate is an important aspect of emergency management planning.
In December 2017, the Legion Fire in the Black Hills closed Custer State Park and forced the evacuation of the nearby communities of Fairburn and Buffalo Gap. In August, the Vineyard Fire was brought under control shortly before portions of Hot Springs had to be evacuated.
Warning triggers in environmental hazards take into account the natural, human, and built environments and can be considered as an evacuation warning and timing mechanism, Li explained. To account for this, Li couples wildfire spread and traffic simulation models to set wildfire evacuation triggers.
The first step is to model the progression of the fire on the terrain, he explained. “Topography matters in terms of fire spread. Fire spreads faster up a slope.” Other inputs include the available fuel, moisture and wind data for the area. To predict fire spread, Li applies fire spread modeling. “Then we use a trigger model to create a buffer around the locations based on fire spread rates and the time needed for the safe evacuation of a community. When the fire reaches the buffer boundary, we know how much lead time people have.”
Li used Julian, a residential area east of San Diego, as a case study to test his coupled model. The area has three communities and three highways, with approximately 1,500 people living there. However, he noted, the integrated model can be applied to any geographic area in the United States because it is based on open source software and open data.
To run the traffic simulation model for evacuation time estimation, Li used house locations from the City of San Diego’s Geographic Information System Department and the vehicle-occupancy data from the American Community Survey. “We need to estimate how many cars will be in the road network during the evacuation,” he said. “If it is too congested, the evacuees could be trapped en route by the fire and die in their cars.”
Coupling fire spread and traffic simulation models helps answer questions concerning the timing of warnings during evacuation, Li explained. Through the coupled model, incident commanders are able to simultaneously examine the spatial patterns of fire perimeters and evacuation traffic for every minute in the evacuation simulation. This could help them improve their situational awareness and issue more effective protective action recommendations.
In the future, Li would like to further investigate how to incorporate more human evacuation behaviors into wildfire evacuation modeling and integrate the models into one software system. “This will make it more convenient for incident commanders and their technical support team to use,” he added.