As wildfires ravage communities across Los Angeles this week, residents and authorities face a daunting and nearly impossible challenge: convincing hundreds of thousands of people to leave their homes within hours or even minutes to avoid danger.

To do this, officials put years of wildfire evacuation research into practice. The field is small but growing, with recent research showing that the frequency of extreme fires has more than doubled since 2023. The increase was driven by horrific fires in the western United States, Canada and Russia.

"There's definitely been an increase in interest (in evacuation research) because of the frequency of wildfire burns," said Asad Ali, an engineering doctoral student at North Dakota State University whose work focuses on this area. “We’re seeing more publications, more articles.”

When evacuations go wrong, they really go wrong. In Los Angeles' Pacific Palisades neighborhood, panicked drivers stuck in traffic abandoned their cars in the middle of evacuation routes, preventing emergency crews from reaching the fire. Authorities used bulldozers to push the empty cars away.

To prevent this confusion, researchers tried to answer some basic but critical questions: Who responds to what warnings? When are people most likely to escape harm?

Many researchers get their ideas about evacuations from other types of disasters—from studies of residents' responses to floods, nuclear disasters, or volcanic eruptions, especially hurricanes.

But hurricanes and wildfires differ in some obvious and less obvious ways. Hurricanes are typically larger and affect entire regions, which can require many states and agencies to work together to help people travel greater distances. But hurricanes are also relatively predictable and slow-moving, often giving authorities more time to organize an escape and develop a phased evacuation strategy so everyone isn't on the road at once. Wildfires are unpredictable and require rapid communication.

People's decisions to stay or leave are also influenced by an uncomfortable fact: Residents who remain during a hurricane can do little to prevent disaster. But for those who stay in wildfires to defend their homes with hoses or water, the tactic sometimes works. "Psychologically, wildfire evacuations are very difficult," Assad said.

Research so far suggests that responses to wildfires, and whether people choose to stay, leave, or just wait a while, can be determined by a range of factors: whether residents have previously experienced wildfire warnings, and whether those warnings were followed by actual threats; How an emergency is communicated to them; and how the neighbors around them react.

A survey of about 500 California wildfire evacuees conducted in 2017 and 2018 found that some long-time residents who had experienced multiple wildfire events were less likely to evacuate, but the opposite was true for others. Overall, low-income people are less likely to flee, possibly due to limited transportation or accommodation. Authorities can use such surveys to create models that tell them when to instruct which people to evacuate.

One difficulty with current wildfire evacuation research is that researchers don’t necessarily classify wildfire events into the “extreme weather” category, said Kendra K. Levine, director of the Transportation Research Institute at the University of California, Berkeley. . For example, Santa Ana winds are not uncommon in Southern California. They happen every year. But combine winds with the region's historical droughts that may be linked to climate change, and wildfires start to look more like weather. "People are starting to accept" this relationship, Levin said, which has led to more interest and academic research from those who specialize in extreme weather.

Assad, the North Dakota researcher, said he has held meetings to discuss using data collected during this week's disaster in future studies. It's a glimmer of hope that the horrors experienced by Californians this week may yield important discoveries that help others avoid the worst-case scenarios in the future.