In a breakthrough that could transform wildfire preparedness, researchers at the UCLA Samueli School of Engineering and their collaborators have unveiled FuelVision an artificial intelligence system that uses satellite imagery to rapidly identify wildfire fuel sources with remarkable accuracy.
The system, detailed in the International Journal of Applied Earth Observation and Geoinformation demonstrated 77% mapping accuracy in validation tests using data from two of California’s most destructive wildfires in 2021 the Dixie and Caldor fires.
Unlike traditional models that depend heavily on expert input and can take longer to process, FuelVision is designed to work autonomously, generating wildfire fuel maps quickly using readily available satellite data.
“We’ve built a tool that lets anyone from local agencies to global researchers generate wildfire fuel maps using satellite data,” said Riyaaz Shaik lead author of the study and a research scientist at UCLA. “That helps make vital wildfire risk information accessible for faster, smarter response.”
Because the tool does not require extensive ground surveys, it can be adapted for use across diverse forested areas in the United States and beyond, making it a potentially scalable solution for global fire risk management.
How It Works
To develop FuelVision, the team trained the system with data from the U.S. Forest Service Forest Inventory and Analysis program. They also employed generative adversarial networks (GANs) a machine learning approach that generates synthetic training data and refines it through a feedback loop of creation and evaluation. This technique helped improve the model’s accuracy and reliability.
“FuelVision can help anticipate where fires might spread and how to prepare,” said Ertugrul Taciroglu corresponding author of the study and professor of civil and environmental engineering at UCLA. “It’s versatile, easily adaptable and can help agencies globally with both organizing emergency response and developing long-term fire mitigation strategies.”
The team plans to make FuelVision widely available in two forms Python-based interface for users with basic coding experience, and an on-demand fuel-map service for agencies and organizations seeking ready-made results.
With wildfires increasing in frequency and intensity due to climate change, FuelVision could prove to be a critical tool in building faster, smarter, and more accessible wildfire preparedness strategies worldwide.