The AI developed by Google Deepmind is capable of outperforming current medium-term weather forecasting models. baptism graphcastIt requires only a fraction of the computing power currently required to produce accurate results in record time.
You will also be interested in this
(on video) Have climate predictions for 2050 already come true? What might weather reports look like in 2050? In partnership with WMO, the organization…
If you find that the forecast weather reportweather report While not always reliable, especially in the medium term, artificial intelligence (AI) should help change your mind. According to a recent article published in the journal Sciencea team of GoogleGoogle DeepMind has developed a new 10-day climate forecasting program called graphcast,
Powered by AI, graphcast Almost always outperforms existing forecasting tools. Even better, this AI uses only a fraction of the computing power of existing systems (HRES). What matters is that AI graphcast It has captured decades of weather information as well as nearly forty years of data from satellites, weather stations, and radar. Current processes work differently. 10-day forecasts are established with huge databases, with huge computing power requiring hundreds of machines on supercomputers. This includes data related to thermodynamicsthermodynamicsOther parameters related to fluid dynamics and atmospheric conditions.
Example of a 10-day graphcast forecast showing humidity, surface temperature, and wind speed. © Graphcast, DeepMind
More accurate and less energy consuming than current models
Supercomputer requires only one machine of a single type graphcast Gives very accurate medium-term forecasts in less than a minute. Unlike existing models, by limiting the scope of analysis to the lowestAtmosphereAtmosphere Where the most visible meteorological phenomena are located, graphcast It proved to be even more accurate in over 99.7% of tests. It started in September with the European Center Weather ForecastWeather Forecast medium term (ECMWF), graphcast Nine days in advance, HRES was able to predict the exact trajectory of Hurricane Lee as it impacted Nova Scotia, three days less than HRES. graphcast Comes from open source and will continue to be used with the current HRES as well as ECMWF.