Computer Models That Consider Human Behavior Could Prevent Chaos During Evacuations

The devastation wreaked by wildfires in the greater Los Angeles area earlier this year underscored the crucial need for urban areas to prepare to evacuate large numbers of people. Delays in evacuating — or the inability to evacuate safely — led to devastating outcomes, endangering not only those attempting to escape but also the first responders and emergency managers handling the crisis. 

These delays are often the consequence of insufficient preparation or confusion about what people need to do to evacuate safely. A research team at Boise State University’s Hazards and Climate Resilience Institute is developing computer models that could help communities plan ahead and save lives by more accurately predicting how people will respond to a disaster and how effective evacuation strategies might be.

“We’re living in a rapidly changing world, facing unprecedented events like the devastating fires in California. Wildfires in the West are becoming more intense and more frequent, and now we’re even seeing them on the East Coast,” Ashley Bosa, a post-doctoral research fellow at the institute, told TriplePundit. “The big questions are: “How do we help people build resilience to these acute and chronic stresses? How do we ensure they have the ability to adapt, be flexible, and feel empowered to protect themselves when these events happen?”

The computer models traditionally used for emergency preparation and response tend to assess communities based on factors like overall population size and basic evacuation routes, Bosa said. They’re significantly limited because they don’t account for the dynamics of how wildfires start from a specific ignition point, how quickly they spread, in which directions, and under what conditions.

They also tend to overlook the human component, she said. Specifically, the number of people in the affected area and how they will likely behave in response to the fire. Important information on traffic signals, road capacity and available evacuation routes is also underrepresented. 

“Most existing models incorporate only one or two of these elements, but rarely all three — fire behavior, human behavior and traffic flow — in an integrated way. They also tend to lack detailed social and behavioral data,” Bosa said. “To really understand how people will evacuate, we need to understand their intent. Are they prepared? Do they even plan to evacuate? All of that matters when modeling evacuation behavior.”

The new models take all of this information — behavioral science, the fire hazard and traffic conditions — into account when simulating potential disaster scenarios. “By combining all three, we’re creating a truly innovative and potentially transformative approach to evacuation modeling and planning for communities,” Bosa said.

At the community level, key information includes available evacuation routes, the use of contraflow — where one side of a highway is opened up to traffic moving in the opposite direction to increase the flow of evacuees — and congestion points where traffic bottlenecks occur. At the household level, the simulation considers preparation time, number and type of available vehicles, planned routes and final destinations.

Emergency managers’ decision-making — including when to issue alerts, orders and road closures — also feeds into the simulation.

“All of these individual decisions play a huge role in determining how accurate the evacuation model can be,” Bosa said. “One of the biggest challenges is that wildfires are incredibly site- and location-specific. Each one behaves differently based on terrain, weather and countless other factors.”

But the hardest part of this process is predicting household response, she said. That’s why the initial step in creating the new computer models is asking locals about their evacuation plans. These surveys help predict what people will do during an evacuation and are used to inform the simulations, Bosa said.

The research team is currently calibrating and using the model for wildfire evacuations in three communities across Idaho and Oregon, Bosa said. It was also used to assess tsunami evacuations in Oregon and wildfire evacuations in Greece.

“The goal is to develop these simulations as decision-making tools for communities,” Bosa said. “We want to hand them off so communities can use them to test different wildfire scenarios — adjusting variables like fire direction or speed — and see how long evacuation might take based on the actual survey responses.”

The survey results are also used to design evacuation workshops specific to each community, she said. The workshops help educate and engage residents by showing them why evacuations matter and taking them through the simulation results for their community. 

“If the data shows that many people would delay evacuation because they’re unsure what to do, for example, the workshops would focus on getting people prepared, so they can act immediately when a wildfire notification comes through,” Bosa said. “The overall goal … is to make sure communities are equipped to respond quickly and effectively, rather than wasting valuable time searching for information during a crisis.”