Israeli researchers develop new AI method to predict wildfires

A group of Israeli researchers have developed an improved method of predicting wildfires, using AI to create country-specific models.

“Published in the scientific journal Natural Hazards by Nature, the study shows that adapting weather-based fire risk indices to local conditions significantly boosts prediction accuracy,” Ynet News reported.

The study, part of a Nature Partner Journals (NPJ) series, found that creating regional-specific AI models resulted in far higher accuracy than a uniform, global model.

While Canada’s Fire Weather Index is the world’s most accurate index, with an accuracy of about 70% on average, it does not take into account regional differences in climate.

The team of researchers “developed country-specific artificial intelligence models, which they simplified into transparent decision trees using a technique called Knowledge Distillation,” Ynet reported. “This method achieved an impressive 86% accuracy while remaining simple enough for practical use.”

Dr. Assad Shmuel of Tel Aviv University, who co-led the study, spoke to Ynet about the benefits of the regionally adapted models.

“There’s no reason to assume a single index can predict wildfire risk across regions with different climates, vegetation and topography,” he said. “Our findings show that regional adaptations significantly improve the accuracy and reliability of fire risk forecasts.”

Other researchers who worked on the study include Tel Aviv University Professor Colin Price, University of Haifa (and Jönköping University) Professor Teddy Lazebnik, and Bar-Ilan University Assistant Professor Dr. Oren Glickman.

Israel, which has experienced a drought this year, saw massive wildfires outside of Jerusalem in late April and early May.

Although a number of suspected arsonists were arrested in connection with the fires, Israel’s Fire and Rescue services reportedly concluded that the fires occurred due to natural causes.