Researchers used a deep learning model to map high-risk forest areas and prevent wildfires

Wildfire is a disaster and an uncontrolled fire that not only burns acres of forestry but also impact livings. As they are not limited to a particular continent or environment, predicting such disasters is very challenging. However, to this effort, researchers from Stanford University designed a new deep learning model that is able to map forest dryness to better predict wildfires. As predicting where a fire is likely to explode and how it might spread requires information, the new AI-powered model maps fuel moisture levels across 12 western states, including Colorado, Montana, Texas, Wyoming and the Pacific Coast.

As the model is still under development, it can detect areas at high-risk for forest fires where the landscape is unusually dry. According to Krishna Rao, a lead author and a Ph.D. student in earth system science, there is more testing needed for the model so that it can figure out fire management decisions and save lives and homes. As the model can see forest dryness in fine details, it can reveal greater risk areas to help fire management teams, he said.

To train the deep learning model, the researchers used three years of data for 239 sites across the American west starting in 2015, when SAR data from the European Space Agency's Sentinel-1 satellites became available. As noted by Futurity, they leveraged field data from the National Fuel Moisture Database and used it to estimate fuel moisture from two types of measurements collected by space-borne sensors. As the one involves measurements of visible light bouncing off Earth, the other, known as synthetic aperture radar (SAR), measures the return of microwave radar signals, which can infiltrate through leafy branches to the ground surface.

Alexandra Konings, an assistant professor of earth system science at Stanford University said that one of their big breakthroughs was to look at a newer set of satellites that are using much longer wavelengths, and allows the observations to be sensitive to water much deeper into the forest canopy and be directly representative of the fuel moisture content.

For decades, the report noted that scientists have predicted fuel moisture content indirectly, from informed but unproven presumptions about relationships between temperature, precipitation, water in dead plants, and the dryness of living ones. The new deep learning model could significantly improve fire studies, says Alexandra.

This new research and development in wildfire management come at a time when climate change has become a crucial concern for each country and is extending rapidly and causing more wildfires across the globe.

Wildfire risk relies heavily on the availability of ignition sources, the propensity of vegetation and litter fuel to ignite, and the ease of fire spread once the fuel has ignited. As noted by National Geographic, the wildfire season in the U.S. between June and September 2020 caused the severe calamity. Indeed, this summer was predicted as the hottest season on record, with drought conditions estimated in California through September. The COVID-19-induced crisis has also contributed to this catastrophe as it wrecked mitigation efforts, such as homeowner assistance programs and controlled burns, owing to concerns over social distancing and respiratory dangers.