JERUSALEM, Aug. 26 (Xinhua) -- Israeli researchers have developed an artificial intelligence (AI) method that significantly improves wildfire forecasting by adapting it to specific countries and regions, Tel Aviv University said in a statement on Tuesday.
The study, published in Nature's npj Natural Hazards, focused on the growing threat of extreme wildfires worldwide, which are becoming more frequent due to climate change.
These fires release large amounts of greenhouse gases, endanger communities and infrastructure, and cause serious damage to ecosystems.
Until now, most fire weather indexes, which measure wildfire risk, were designed in countries such as Australia, Canada, and the United States. While effective in their places of origin, they are often less accurate when applied to other regions.
The new method improves forecasts by adjusting fire indexes to local conditions such as climate, vegetation, land use, and ignition sources. This approach raised accuracy from about 70 percent to 86 percent.
In the study, the researchers first compared three leading fire indices across 160 countries and found the Canadian index to be the most reliable, with 70 percent accuracy.
Using a genetic algorithm, they then calibrated the index for each country, raising accuracy to 80 percent. Next, they developed a country-specific AI model and converted it into a simple decision tree that is easy to apply while maintaining 86 percent accuracy.
The researchers said that the new method can support emergency services, policymakers, and field teams by improving early warnings, guiding resource allocation, and reducing damage.
They also noted that the approach provides a foundation for future systems designed to prevent devastating wildfires in a warming world.
