University of California-Irvine scientists will be exploring ways to use machine learning to predict the spread of California’s wildfires once they’ve ignited.
A team of scientists at UCI has developed a new computer technique for forecasting which blazes will burn out of control, knowledge that can help the allocation of scarce firefighting resources.
While the algorithm model initially focused on Alaska’s wildfires, one of the key researchers says an analysis of the Golden State is next in line.
“That is already starting,” Shane Coffield, the Alaska study’s lead author, told the Business Journal.
“That’s where I’m turning next—to focus on California.”
“What’s been the impact of the droughts, what’s the role of land management versus climate change in controlling our fires and how can we use that to improve fire protection,” Coffield added.
He and other UCI scientists will work on projects with other UC schools including UC Davis and Berkeley and results may be ready for publication “in the next year or so.”
Study Published
The initial, Alaska-focused machine-learning model can help forecast whether a blaze is going to be small, medium or large by the time it has run its course.
The researchers’ work was highlighted in a study published last month in the International Journal of Wildland Fire, just as California was heading into peak wildfire season. The team used Alaska as a study area for the project because the state has been plagued over the past decade by a rash of concurrent fires in its forests, threatening human life and vulnerable ecosystems.
Machine learning is the process by which a computer improves its own performance by continuously incorporating new data and experiences into an existing statistical model without being explicitly programmed.
“This idea of predicting fires from the time they ignite is really relevant, especially in the context of changing climate,” said Coffield, a UCI doctoral student in Earth System Science.
“We know it’s going to be getting a lot warmer and probably drier here, so fire frequencies will likely increase.”
Simultaneous Combustion
He and his colleagues applied their Alaska algorithm to a hypothetical situation in which dozens of fires break out simultaneously. The scenario has become common in recent years in parts of the western U.S. as climate change has resulted in hot and dry conditions on the ground that can put a region at high risk of ignition.
“Only a few of those fires are going to get really big and account for most of the burned area, so we have this new approach that’s focused on identifying specific ignitions that pose the greatest risk of getting out of control,” Coffield said. “We’re trying to ask the question of which fires go viral.”
He emphasized there are significant differences between the situation in Alaska and the one in California.
“The wind is really important here. We have the Santa Ana event that can really cause fires to get out of control,” Coffield said.
Decision Tree Algorithm
The UCI scientists used machine learning to identify patterns in historical wildfire data, such as what variables—for example weather, vegetation and topography—what combinations of them provide the best predictive information.
They chose a “decision tree” algorithm, which is a specific type of machine learning that scientists and fire managers can readily interpret. The researchers were able to predict the final size of a blaze 50% of the time.
One advantage of the new method is speed, Coffield said.
The algorithm “learns” with each new data point and can quickly figure out the critical thresholds for identifying large fires.
“The practical aspect really comes in terms of informing fire management, especially when a lot of fires break out in a short period of time, being able to make a smart prediction about which particular ignitions are going to grow really large and essentially inform triaging efforts,” Coffield said.
New Tools Needed
Faced with a climate change-induced jump in the number of wildfires expected each season; state, county and local firefighting authorities could benefit from some updated tools and techniques, said James Randerson, the study’s co-author and professor of Earth System Science.
“In places like Alaska, there’s a need to limit the area affected by fire, because if we keep having these unusual, high-fire years, more carbon will be lost from the landscape, exacerbating warming,” Randerson said.
“If we let the fires run away, we could be in a situation where there’s a lot of significant damage to both the climate system and ecosystems.”
