Researchers at the Gwangju Institute of Science and Technology are developing an AI-based model that predicts extreme wildfire risk

GWANGJU, South Korea, October 21, 2022 /PRNewswire/ — Raging wildfires that are breaking out around the world have caused tremendous economic damage and loss of life. Knowing in advance when and where a widespread fire could strike can improve fire prevention and resource allocation. However, available forecasting systems provide only limited information. Additionally, they don’t provide long enough lead times to get useful regional details.

Scientists have now applied a deep learning algorithm to improve forecasting of US wildfire danger western United States. researchers out South Korea and The United States developed a hybrid method combining AI techniques and weather forecasting to produce improved forecasts of extreme fire hazards up to a week at finer scales (4km x 4km resolution) and increase their usefulness for firefighting and management.

“We have tried numerous approaches to integrate machine learning into traditional weather forecast models to improve wildfire risk predictions. This study is a major step forward as it demonstrates the potential of such an effort to improve fire hazard prediction without additional computing power.” says lead author Dr. Rackhun Son, a recent PhD Dr. from the Gwangju Institute of Science and Technology (GIST) in South Koreawho is currently working at the Max Planck Institute for Biogeochemistry in Germany. “Fire hazard predictions could be further improved through continued refinement of both Earth system models and recent AI developments,” he adds.

While data-driven AI methods have shown excellent ability to infer things, explaining why and how the inferences are arrived at still remains a challenge. This has led to AI being referred to as a black box. “But when AI was combined with computer models based on physical principles, we were able to diagnose what was going on in that black box.” says co-author Prof. Simon Wang von Utah State University, United States of America. “The AI-based predictions associated with extreme fire hazards are well grounded in strong winds and specific geographic features, including high mountains and canyons western United States that have traditionally been difficult to solve with coarser models.”

Computational efficiency is another major advantage of this method. Traditional methods of predicting fire risks at finer spatial resolutions, a process known as “regional downscaling,” are often computationally intensive, expensive, and time-consuming. “Although comparable computational resources were required in the development phase, once the training task for the AI ​​was complete, i.e. performed once, it took only a few seconds to use this component with the weather forecast model to produce forecasts for the rest of the season, “ says co-author Prof. Kyo Sun Lim at Kyungpook national university, Korea. Therefore, the newly developed AI-based method, with the ability to produce accurate high-resolution predictions in less time, was much more cost-effective compared to traditional prediction systems.

“In this study, AI is only tested for predicting fire hazards in the western United States. In the future, it could be applied to other types of weather extremes or to other parts of the world.” said co-author Dr. Philip J. Rasch of the Pacific Northwest National Laboratory and the University of Washington. “The flexibility of our AI method can help predict any weather-related feature.”

The research was published in the Journal of Advances in Modeling Earth Systems On September 22, 2022.


Original Paper Title: Deep Learning Provides Significant Improvements for County-Level Fire Weather Forecasting in the West United States

Diary: Journal of Advances in Modeling Earth Systems


About Gwangju Institute of Science and Technology (GIST)

Media contact:
Chang Sung Kang
82 62 715 6253
[email protected]



View original content to download multimedia: -extreme-wildfire-danger-301655174.html

SOURCE Gwangju Institute of Science and Technology (GIST)


Leave a Reply

Your email address will not be published. Required fields are marked *