AI Predicts Arctic Storms with Unprecedented Detail

Russian scientists have engineered an advanced AI algorithm designed to accurately predict extreme weather phenomena in the Russian Arctic, including storms and mesoscale vortices. This innovative neural network offers a resolution approximately five times higher than that of global climate models, providing much more detailed forecasts.

The AI was trained using eight decades of data from the ERA5 global weather archive, enabling it to generate environmental condition calculations with a resolution of 6 by 6 km. This represents a substantial improvement over previous methods that required vast computing resources for high-resolution predictions.

Comparisons with the WRF model show that while the quality of forecasts is comparable, the AI significantly reduces computation time. For instance, calculating annual wind fields takes the AI only 10 minutes, a stark contrast to the approximately 10 hours needed by the reference model. The AI demonstrates particular proficiency in predicting the movement of squall winds and wave heights in the Barents Sea, offering critical enhancements for maritime safety and storm warnings.