A newly developed AI-based method can accurately predict wildfires

Wildfires have caused extreme fire damages across the globe, along with many deaths. It is significant to know when wildfires are spreading, and where, to prevent loss of life. Realizing this important information in advance is key. Forecasting wildfire danger can be a difficult task because of the complexity involving climate system, interactions with vegetation and socio-economic components.

Currently, available information for widespread fires only provides limited data and information. Also, they don’t provide lead times, the time between the initiation and completion, that allow useful regional details.

New method

The new tool created by researchers applies a deep learning algorithm to improve the prediction of wildfire danger in the western United States. Scientists from South Korea and the United States created a hybrid technique that combines artificial intelligence (AI) with weather forecasts to produce improved predictions of extreme fire dangers out to one week of accuracy. This allows for an increase in the efficiency for fire suppression and management.

The research has been published in the Journal of Advances in Modeling Earth Systems.

The research team incorporated AI with forecasts to predict with accuracy the future fires. “We tried numerous approaches to integrate machine learning with traditional weather forecast models to improve forecasts of wildfire risks. This study is a big step forward as it demonstrates the potential of such an effort for enhancing fire danger prediction without the need for extra computing power,” said Dr. Rackhun Son, lead author of the study and graduate from the Gwangju Institute of Science and Technology (GIST) in South Korea. He also currently works at The Max Planck Institute for Biogeochemistry in Germany.

“The fire danger forecasts could be improved further using constant development in both Earth System Models and recent AI developments,” he continued.

Although using data-driven AI methods shows excellent promise for assuming when there will be a wildfire, there also proves to be some challenges, such as explaining why and how inferences occur. This led the research team to labeling the AI a black box.

“But when AI was combined with computer models based on physical principles, we could diagnose what was going on inside this black box,” said Simon Wang, co-author and professor at Utah State University. “The AI-based predictions associated with extreme levels of fire danger are well grounded to strong winds and specific geographical characteristics, including high mountains and canyons in the Western United States that have been traditionally difficult to resolve with coarser models.” The machines learned to tolerate the extreme conditions, such as strong winds and high elevation.

Integrating AI with weather forecast models to create accurate wildfire forecast systems

Cost-effective and accurate

The team also described computational efficiency as an advantage of using this approach. The other method that is usually used, regional downscaling, according to the study, is more expensive and time-consuming.

The researchers did have to train the AI in its initial phases, but learned quickly how to use AI with forecast to predict wildfires. “Although comparable computational resources were required at the developing stage, once the training task for the AI was complete, i.e., performed once initially, it only took few seconds to use that component with the weather forecast model to produce forecasts for the rest of the season,” said Kyo-Sun Lim co-author and professor at Kyungpook National University, Korea.

The new AI-based method can create accurate forecasts in a short period of time while also being cost-effective.

In the future, the researchers hope to apply the AI to predicting other weather extremes around the world.

 

Original post: https://interestingengineering.com/science/a-newly-developed-ai-based-method-can-accurately-predict-wildfires

Leave a Reply

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