The Northwest can expect a widespread increase of days with cloud-to-ground lightning in the years to come—along with heightened wildfire risk—according to projections made with a unique machine-learning approach.
The new study in the journal Earth’s Future offers detailed projections of lightning across the Western US for the mid-21st century. The largest change in lightning is expected in parts of Idaho, Washington, and Oregon, with four to 12 more days of lightning per year in some parts of the Rocky Mountains.
Researchers also matched those lightning projections with future wildfire risk to calculate the changes in daily risk of lightning-caused fires. Although there is variability across the region in the projections, the trend was clear: a heightened risk of lightning-caused wildfires across 98% of Western lands susceptible to fire.
“The Northwest is emerging, in this study as well as in others, as the region where fire- and fire-related hazards are likely to increase substantially more than in other parts of the western US,” says Deepti Singh, an associate professor in the School of the Environment at the Washington State University Vancouver and coauthor of the paper.
The study adds urgency to the need to manage forests for wildfire risk and prepare at-risk communities for fires, as the planet continues to warm and wildfires grow in size and severity, the researchers say. Lightning already accounts for more than two-thirds of the acreage burned in wildfires across the West, but current global climate models are unable to directly simulate future lightning because they rely on geographic resolutions too coarse to capture the conditions that create it.
The machine-learning models developed in this study zoom in to create the most detailed picture yet of future lightning patterns and lightning-caused fire risk across the West.
“There are already a lot of studies that say future wildfire activity will increase in the Western US and that’s without even considering potential lightning increasing, which we’re showing is going to happen in many areas,” says Dmitri Kalashnikov, lead author of the paper.
“We’re also making projections for the near-term future—2031 to 2060. That period starts in just a few years, so it’s on our doorstep.”
To make these projections, Kalashnikov applied a machine learning technique known as a convolutional neural network. These neural network-based predictive models were tailor-made for each grid cell of 1 degree by 1 degree across the Western US. That’s an area of roughly 69 miles on each side, which is the typical spatial resolution of climate models. This approach allowed for targeted lightning projections at finer geographic scales than previous studies.
These neural-network based predictive models were the subject of a paper published in 2024, led by Kalashnikov and coauthored by Singh, among others.
In the current project, the team used data from three key meteorological variables conducive to lightning from the summers of 1995–2022 to train the network in each grid box to make mid-century projections.
“Instead of developing one model to predict lightning everywhere, we really went in on a finer scale to predict lightning at each 1-degree box,” Kalashnikov says.
The models identified days where cloud-to-ground lightning would be likely for each grid; researchers also quantified how many of these days are expected to be high fire-weather weather days, using the Fire Weather Index, a measure of wildfire risk based on weather and climate conditions. Critically, the authors found that most locations will experience an increased risk of lightning-caused fires due to increases in the Fire Weather Index, even in places where lightning occurrence might not increase.
An increase in lightning days does not result in a 1-to-1 increase in fire risk, however, because fire risk depends on other variables, such as temperature, rainfall or wind, and vegetation dryness. Across the Rockies, for example, the number of days with a high likelihood of lightning-caused fires is expected to grow by three or more days by the mid-21st century though the overall increase in lightning days is larger.
On the other hand, parts of Utah and Arizona showed a reduction in lightning days—but an increase in days of potential lightning-caused fires, due to higher wildfire risk in general.
The Southwest showed fewer projected increases in lightning days—and even declines in some areas—but the region is still expected to see a rise in days with a likelihood of wildfires ignited by lightning.
Kalashnikov led the project while completing his PhD at WSU and is now a post-doctoral fellow at the University of California, Merced. Coauthors included researchers from UC Merced, Colorado State University, Portland State University, and other institutions.
Source: Washington State University
