Researchers in Krasnoyarsk Develop AI Systems to Predict Floods and Wildfires

Scientists and students at Reshetnev Siberian State University of Science and Technology have developed intelligent systems designed to predict natural disasters. The tools can be used to forecast floods and forest fires, the university’s press service told IT Russia.

The team first analyzed water levels in the Nizhnyaya Tunguska and Podkamennaya Tunguska rivers in the Krasnoyarsk region to determine safe navigation periods during flood seasons. Using statistical analysis and machine learning, they built regression models that predict water levels based on historical data and weather forecasts. The algorithms help transport companies and emergency services better plan the start and end of navigation, improving safety.

Classification and Damage Forecasting

In parallel, the researchers processed data on five million fires recorded in Russia between 2012 and 2022. Using support vector machines, random forest, and gradient boosting methods, they developed models that classify fires involving injuries and fatalities and estimate potential damage. Classification accuracy exceeded 85%.

The team has already registered the development with Rospatent. Technical requirements have been defined for a university repository to store data on natural and man-made disasters. The platform is intended to serve as an open database for researchers, students, and emergency response specialists.

Earlier reports said researchers at Penza State University had developed an AI-based system for forecasting forest fires.