Multi-ignition fires like California’s 2020 August Complex fire, shown here, can have a disproportionately devastating impact compared with single-ignition fires. (Credit: Mike McMillan, Pacific Southwest Forest Service, USDA)
Just weeks after the January 2025 Los Angeles wildfires became the costliest in U.S. history, with insured losses exceeding $37 billion and total economic damage estimates ranging from $95 billion to $164 billion, new research from Lawrence Livermore National Laboratory (LLNL) offers insights into one of the most dangerous and least understood wildfire phenomena: multi-ignition fire complexes.
In a study published in Science Advances, researchers at LLNL and the University of California, Irvine examined how separate fire ignitions merge into larger, more destructive events. Their findings reveal that multi-ignition fires account for just 7% of California’s total fires but contribute to 31% of the state’s burned area.
“Multi-ignition fires have a disproportionate effect on the burned area,” said LLNL atmospheric scientist Qi Tang, a co-author of the study in a release. “Although they are quite rare, their influence is large compared to single-ignition fires.”
The largest fire on record in California, the 2020 August Complex fire, exemplifies the phenomenon. It grew from the coalescence of 10 separate ignitions into a single massive blaze.
Fire-spawned thunderstorms
The LLNL research team used UC Irvine’s remote sensing data to track multi-ignition fires and capture behavioral data, then applied LLNL’s simulation framework to model pyrocumulonimbus events, intense thunderstorms triggered by wildfires themselves. The modeling taps the Department of Energy’s Energy Exascale Earth System Model (E3SM). That system can connect fire dynamics at kilometer-scale resolution to understand how wildfires ignite, spread, merge and interact with atmospheric conditions.
Pyrocumulonimbus clouds form when extreme fire heat creates powerful updrafts that lift hot air and moisture into the atmosphere, where it condenses into storm clouds. These fire-spawned storms can generate lightning strikes at locations distant from the original fire. In turn, that creates conditions for new ignitions that may later merge with existing blazes.
“Pyrocumulonimbus events often occur when there’s an enormous wildfire event, but not all wildfires can trigger them,” Tang said. “The distribution actually is very non-uniform around the world.”
California, Canada and Siberia are among the most likely locations for these fire-triggered thunderstorms. The research builds on related work published in September 2025 in Geophysical Research Letters, in which members of the same team successfully simulated the pyrocumulonimbus cloud generated by California’s 2020 Creek Fire, the first time such fire-induced storms had been realistically modeled within an Earth system model.
Award-winning infrastructure
The wildfire modeling research draws on climate simulation capabilities that have earned recognition in their own right. LLNL received an R&D 100 Award in 2017 for the Earth System Grid Federation (ESGF), the distributed data platform that underpins E3SM and serves as the backbone for international climate research, including UN Intergovernmental Panel on Climate Change assessments.
The laboratory has earned more than 180 R&D 100 awards since 1978. In 2023, the E3SM team also received the first-ever ACM Gordon Bell Prize for Climate Modeling for their work on SCREAM, a high-resolution atmospheric model component of E3SM.
From prediction to prevention
For firefighters, multi-ignition events pose acute dangers. Multiple fire fronts can trap crews if new flames spring up to surround them, and the dispersal of suppression resources across separate ignitions strains response capacity.
Tang said the modeling framework could help predict where pyrocumulonimbus events, and subsequent new ignitions, are most likely to occur, potentially enabling preventive action.
“We can help the community of firefighters know where the pyrocumulonimbus is more likely to occur, and that can lead to a prediction of where the fire triggers and the larger event would be,” Tang said in a release. “We might be able to do something to avoid multi-ignition events. That is one of our objectives.”
Additional observational data from a planned 2026 NASA field campaign will further refine the modeling. The LLNL team also aims to integrate their simulations with energy infrastructure planning to assess how fire events might affect the power grid.
The research was funded by LLNL’s Laboratory Directed Research and Development program, with additional support from the DOE Office of Science Biological and Environmental Research E3SM project.
