Researchers enhance tool to predict wildfires
A newly-enhanced database is expected to help wildfire managers and scientists better predict where and when wildfires may occur by incorporating hundreds of additional factors which impact the ignition and spread of fire.
“There is a tremendous amount of interest in what enables wildfire ignitions and what can be done to prevent them,” said Erica Fleishman, an Oregon State University (OSU) professor. “This database increases the ability to access relevant information and contribute to wildfire preparedness and prevention.”
Revised database
The Fire Program Analysis Fire-Occurrence Database was developed in 2013 by the U.S. Forest Service (USFS) and has since been updated five times. It incorporates basic information such as ignition location, discovery date and final wildfire size.
The revised database now includes many new environmental and social factors, such as topography and vegetation, social vulnerability and economic justice metrics and practical attributes such as the distance from the ignition to the nearest road.
In addition to aiding on-the-ground firefighters and managers, the database could also help power companies evaluate short-term risk when deciding whether to implement a public safety power shutoff or land management agencies determine whether to reduce access to public lands or restrict campfires during certain times of year, Fleishman explained.
“There seem to be a lot of policies guided to some extent by intuition or emotions rather than by a large body of evidence,” she said. “These data present one way to increase the objective evidence to consider when making those decisions.”
The team, including Fleishman and led by Yavar Pourmohamad, a doctoral student at Boise State University, as well as Mojtaba Sadegh, an associate professor at Boise State University, added nearly 270 additional attributes. The database now includes information on 2.3 million fires in the U.S. from 1992 to 2020.
“This provides a considerably deeper understanding of the individual and compounded impact of these attributes on wildfire ignitions and size,” Pourmohamad said. “It also identifies the unequal effects of wildfires on distinct human populations and ecosystems, which can in turn inform efforts to reduce inequities.”
Artificial intelligence and machine learning
Information from the database can also be incorporated into artificial intelligence and machine learning models explaining drivers of past fires, project likelihoods or effects of future fires.
“It’s amazing what one can infer when they have the computational capacity and this much information,” Fleishman said. “We can ask a lot of questions informing different actions in different places to understand what is associated with wildfire ignitions and fire effects.”
A paper outlining the database was recently published in the journal Earth System Science Data.
Other coauthors of the paper are Eric Henderson and Sawyer Ball of Boise State University; John Abatzoglou of the University of California, Merced; Erin Belval, Karen Short, Matthew Reeves and Julia Olszewski of the USFSʼs Rocky Mountain Research Station; Nicholas Nauslar of the National Weather Service Storm Prediction Center; Philip Higuera of the University of Montana; Amir AghaKouchak of the University of California, Irvine and Jeffrey Prestemon of the USFS Southern Research Station.
The research was supported by the Joint Fire Science Program, a program of the USFS and U.S. Department of the Interior.
Sean Nealon is a news editor for OSU. This article was originally published by OSU on Aug. 2.
