Third Place (tie): Predicting Wildfire Rate of Spread Using Machine Learning and Remote Sensing

Wildfires have global economic and environmental costs. Accurate wildfire prediction is vital for effective fire management, with a focus on Rate of Spread (ROS), which measures a fire's speed away from its origin. Traditional models like FlamMap have limitations, particularly in accounting for real-world conditions such as changing winds. We advocate for physics-informed neural networks as an alternative to produce more accurate fire behavior predictions while reducing the need for extensive training data. The project's progress involved analyzing fire perimeters from NASA's VIIRS dataset to calculate ground truth rate of spread and utilizing Landfire.gov, NASA, and USGS data for input features. Next steps include training the neural network to predict rate of spread under natural conditions. In summary, we highlighted the need for more accurate wildfire prediction and showcased the potential of machine learning and remote sensing to enhance wildfire management.
A live presentation of this talk by Saad Lahrichi (University of Montana) was awarded Third Place (tie) in the Lightning Talks competition at the 2023 Energy Data Analytics Symposium: Accelerating Sustainability in the AI Era on Oct. 27, 2023.
Learn more about the Lightning Talks competition: nicholasinstit...
Learn about the 2023 Energy Data Analytics Symposium: nicholasinstit...
Learn about the Energy Data Analytics Lab at Duke's Nicholas Institute for Energy, Environment & Sustainability: nicholasinstitu...
Get email updates from the Nicholas Institute: nicholasinstitu...
Learn about the Energy Data Analytics Lab at Duke's Nicholas Institute for Energy, Environment & Sustainability: nicholasinstitu...
Get email updates from the Nicholas Institute: nicholasinstitu...