GraphCast, developed by Google, claims to provide 10-day weather forecasts with unprecedented accuracy in less than a minute.
It’s a small revolution in the world of meteorology. DeepMind, an artificial intelligence company that is actually a division of Google, has developed a program called GraphCast to predict weather and weather phenomena with great accuracy several days in advance.
And even brags about the ability to do so less than a minute…
Therefore, Google presents GraphCast as “the most accurate global 10-day weather forecast system in the world and can predict more extreme weather events in the future than was previously possible.”
Is it as effective as advertised? How revolutionary is this process?
How does this artificial intelligence program work?
According to a study published by these designers in scienceThis artificial intelligence program will be more accurate and reliable in weather forecasts than traditional tools used by specialists, with a success rate of up to 90% in cases subject to testing.
Weather forecasts are usually made from physical calculations and equations based on well-thought-out algorithms. But getting results takes a long time and is not always reliable.
AI will use data collected decades before ECMWF, The European Center for Medium-Range Weather Forecasts will use it to link it to the variables controlling the entire surface of the planet. to Each plot of land is demarcatedthe program predicts 5 variables Temperatures, pressure, direction and strength of winds accelerate it It will be coupled with six weather variants At 37 height levels: Humidity, wind and temperature.
It is rich in 36.7 million parameters, and will present its conclusions with events that should logically occur in light of the data thus exploited. This is a prediction based on previous data.
The results will not only be convincing, but will be long-lasting. With efficiency in seconds.
Thus many tests were carried out. For example, in September, the AI determined that Hurricane Lee was going to hit Canada 9 days before it happened. More traditional forecasts were able to predict it 6 days in advance. So the difference is three days.
The same applies to weather forecasts without major events. Thus the weather is predicted reliably 10 days in advance.
What will it be used for?
In addition to the classic forecasts that are used to know the weather for dressing or to adapt to the comfort of daily life, this type of artificial intelligence may be useful for starting solar panels or wind turbines.
Above all, to predict devastating weather events in advance to warn the population and thus save lives during violent adverse weather events.
Google makes discovery available
Google has decided to make the code of its artificial intelligence program available to forecasters and meteorologists so that everyone can benefit from it.
The scientists who designed this software explain that their approach cannot be a “replacement for traditional methods of weather forecasting, which have been developed for decades and rigorously tested in many real-world environments.” Especially since this artificial intelligence needs this data to function.
The software must be further improved to achieve highly accurate and reliable forecasts of hazardous weather events.