Researchers at Google DeepMind have developed an artificial intelligence model called GenCast, capable of making weather forecasts with greater accuracy and speed than traditional methods. The technology, announced in an article published in , promises to revolutionize the sector, offering more reliable forecasts, especially in extreme weather conditions.
GenCast was trained on climate data collected between 1979 and 2018, spanning 40 years of detailed analysis. This model uses machine learning to generate global forecasts for up to 15 days, analyzing more than 80 atmospheric and surface variables. The process is carried out in just eight minutes, marking a significant advance in terms of speed.
When compared to the traditional forecasting system of the European Center for Medium-Range Weather Forecasts (ENS), currently considered the most effective globally, GenCast proved to be superior in 97.2% of the 1,320 targets tested, according to the researchers.
Better response to extreme phenomena
In addition to more effectively predicting the weather in general, GenCast stands out in predicting extreme weather phenomena, such as tropical cyclones and sudden changes in weather conditions. It also offers more accurate estimates on wind energy production, a critical area for sustainable energy planning.
Accurate weather forecasts are essential for governments, organizations and citizens in their daily lives. From deciding when to carry an umbrella to planning strategies for dealing with natural disasters, climate information influences critical decisions. The Google DeepMind model promises not only to increase the accuracy of these predictions, but also to speed up their availability, allowing for more effective planning in various sectors.
A step forward in mid-range predictions
Traditional weather forecasting methods, known as deterministic, are based on numerical calculations that create possible future climate scenarios based on the current climate. However, these methods face limitations in terms of comprehensiveness and accuracy. GenCast overcomes these barriers by offering probabilistic forecasts that combine current and historical climate states to anticipate more accurate scenarios.
The researchers argue that GenCast represents a significant advance in meteorology, with the potential to transform the way governments and companies prepare for climate challenges. Its ability to accurately predict extreme phenomena could be an essential tool in natural disaster mitigation and energy planning.
Also read:
