California burns, but AI could predict the wildfire’s next move

As California is battling wildfires, scientists created a novel AI model that could predict the spread of wildfires using satellite imagery.

Researchers at the University of Southern California (USC) in Los Angeles have created an artificial intelligence (AI) model that precisely predicts wildfire spread by integrating generative AI with satellite data.

Currently, California and much of the western United States are battling an increasingly severe wildfire season. ​​According to the California Fire Department's data, there are currently more than 30 active fire incidents in the state.

Multiple blazes, fueled by a hazardous mix of strong winds, prolonged drought, and extreme heat, are sweeping across the state. These conditions are exacerbating the fires, making them harder to control and placing additional strain on emergency services. Also, rising temperatures have significantly contributed to the intensity and frequency of wildfires.

Scientists believe that the model offers a potential breakthrough in wildfire management and emergency response. The model uses satellite data to monitor a wildfire’s progress in real time and then inputs this information into an advanced computer algorithm that can accurately predict the fire’s probable path, intensity, and rate of growth.

The researchers commenced their study by collecting historical wildfire data from high-resolution satellite images. After analyzing the behavior of past wildfires, they uncovered patterns and tracked how each fire ignited, spread, and was ultimately contained.

“Fuel like grass, shrubs, or trees ignites the fire, leading to complex chemical reactions that generate heat and wind currents. Factors such as topography and weather also influence fire behavior — fires don’t spread much in moist conditions but can move rapidly in dry conditions,” he said.

“These are highly complex, chaotic, and nonlinear processes. To model them accurately, you need to account for all these different factors. You need advanced computing.”

Researchers trained a generative AI-powered model called a conditional Wasserstein Generative Adversarial Network (cWGAN) to simulate the impact of factors such as terrain and potential fuel on wildfire behavior over time. The model was taught to identify patterns in satellite images that correspond to wildfire spread in its simulations.

“This model represents an important step forward in our ability to combat wildfires,” said Bryan Shaddy, a doctoral student and the study’s corresponding author, in a press release.

“By offering more precise and timely data, our tool strengthens the efforts of firefighters and evacuation teams battling wildfires on the front lines,” he concludes.