AI surpasses traditional weather forecasting in accuracy for the first time

Artificial intelligence (AI) has, for the first time, demonstrated a significantly higher accuracy in weather forecasting compared to meteorologists using traditional data collection and analysis systems.

According to a study published in the journal Science, the neural network GraphCast developed by Google DeepMind surpassed the prediction accuracy of the European Centre for Medium-Range Weather Forecasts.

Within the research framework, GraphCast delivered more precise forecasts for 90% of the 1380 parameters examined, including temperature, pressure, wind speed and direction, as well as humidity. Importantly, it outperformed its traditional counterpart notably in predicting extreme weather events. Additionally, the neural network achieved this at a much faster pace than conventional forecasting systems.

"The neural network predicts hundreds of weather variables for ten days with a resolution of 0.25° globally in less than one minute," as stated in the research.

Matthew Chantry, the machine learning coordinator at ECMWF, stated in an interview with the Financial Times that artificial intelligence in meteorology has progressed "much faster and much stronger than we expected two years ago." Furthermore, he emphasized that the neural network proved to be "a thousand times" more efficient than traditional systems in terms of energy consumption required for calculations.

These findings underscore a significant leap forward in utilizing artificial intelligence to enhance the accuracy and efficiency of weather predictions.