AI in Russia Learns to Forecast the Weather More Accurately
Researchers at Novosibirsk State Technical University are using neural networks to produce longer-range and more precise weather forecasts.
Russian scientists have developed a way to improve the accuracy of weather forecasting. The work was presented at Novosibirsk State Technical University, where specialists trained neural network algorithms to build meteorological forecasting models.
According to one of the project’s authors, Boris Malozyomov, who holds a PhD in engineering, neural networks are capable of processing large volumes of data and analyzing relationships between multiple factors such as temperature, pressure, precipitation, and others. This makes it possible to generate more accurate short-term and long-term weather forecasts.
Combining Multiple Forecasts Into One
Forecasts produced by each network are then combined to generate a final weather forecast. This approach makes it possible to account for the nuances of weather conditions in different regions, which is essential for everyday planning, agricultural operations, construction, and the efficient management of water, energy, and other resources.
Accuracy Up to 180 Hours
To generate forecasts, the neural network is fed data from weather stations, satellite observations, as well as geographic and historical indicators. The contribution of each individual neural network can also be adjusted to reduce errors.
Using this approach, researchers in Novosibirsk have managed to extend reliable forecast accuracy to up to 180 hours, compared with the 120-hour limit typical of other tools. The method can be applied not only to weather forecasting, but also to predicting financial trends, traffic congestion, and processes in other fields.
