Researchers in the US have adapted a generative AI framework to work with satellite data to predict rather than detect wildfires, as reported by eeNews Europe.
The model developed at the University of Southern California (USC) uses satellite data to track the progression of a wildfire in real time.
This information is then fed into a generative AI framework that can accurately forecast the fire’s likely path, intensity, and growth rate.
Multiple blazes, fueled by a combination of wind, drought, and extreme heat, are currently raging across California.
The largest wildfire in the state this year, the Lake Fire, has already scorched over 38,000 acres in Santa Barbara County.
Various AI-based technologies have been developed to detect, monitor, and track wildfires.
The physics-informed approach used by the researchers allows for inferring the history of a wildfire from satellite measurements.
This provides the necessary information to initialize coupled atmosphere-wildfire models from a measured wildfire state.
The fire arrival time, which is the time the fire reaches a given spatial location, acts as a representation of the wildfire’s history.
A conditional Wasserstein Generative Adversarial Network (cWGAN), trained with simple simulation data, is employed to infer the fire arrival time from satellite active fire data.
The researchers conducted a comprehensive analysis of past wildfires, tracking how each fire started, spread, and was eventually contained.
They identified patterns influenced by various factors such as weather, vegetation, and terrain.
The cWGAN was tested on four California wildfires that occurred between 2020 and 2022.
Predictions for fire extent were compared against high-resolution airborne infrared measurements.
The predicted ignition times were then compared with reported ignition times, showing an average difference of 32 minutes, indicating high accuracy.
Bryan Shaddy, a doctoral student in the Department of Aerospace and Mechanical Engineering at the USC Viterbi School of Engineering and the study’s corresponding author, said: “This model represents an important step forward in our ability to combat wildfires.
“By offering more precise and timely data, our tool strengthens the efforts of firefighters and evacuation teams battling wildfires on the front lines.”
The success of the cWGAN, initially trained on simple simulated data, in its tests on real California wildfires is attributed to its use of actual wildfire data from satellite imagery.
Assad Oberai, Hughes Professor and Professor of Aerospace and Mechanical Engineering at USC Viterbi and co-author of the study, stated: “By studying how past fires behaved, we can create a model that anticipates how future fires might spread.”
Oberai and Shaddy were impressed by the performance of the cWGAN, which worked well even under complex real-world conditions such as varied terrain and multi-directional winds.
This demonstrates the potential for the model to be an invaluable tool in wildfire management and prevention efforts.
