Russian Scientists Develop AI Weather Model with 9-Month Forecasting Power

Created in Sevastopol, this neural network rivals top global systems in accuracy and can detect climate-shifting phenomena nearly two years in advance.

Russian researchers have unveiled a powerful new neural network capable of forecasting weather conditions up to nine months in advance. Developed at the Institute of Natural and Technical Systems in Sevastopol, the system delivers high-accuracy predictions of monthly temperatures and precipitation.

According to project lead Andrey Lubkov, PhD in physics and mathematics, the AI analyzes real-time global data from meteorological stations, satellites, and sensors. It’s also able to anticipate climate phenomena in the Pacific Ocean that influence global weather patterns—up to 22 months ahead of time.

The forecasts include monthly averages and assess risks of extreme weather events, particularly in Crimea’s river systems. Lubkov emphasized that the AI’s performance is on par with the best international models.

The system updates its forecasts at the start of each month and will be freely accessible to users. Further improvements to the model are underway, aiming to enhance its precision and usability.