French scientists rely on artificial intelligence to anticipate periods of extreme drought

In the future, scientists say, droughts and heat waves will be more frequent. But it is still difficult to anticipate them today, hence the interest in artificial intelligence, which could make it possible to better predict extreme weather events. Researchers from the CNRS and ENS Lyon are thus trying to develop a tool capable of doing better than Météo France in terms of seasonal forecasts, that is to say forecasts at more than one month, which are today unreliable, too random.

>> Faced with the very worrying level of groundwater, the risk of drought is increasing again for this summer

Lack of data is a barrier

However, they come up against a lack of data, since these extreme events are certainly likely to multiply but remain rare for the moment. However, to train the algorithms, you have to present them with a lot of figures, a lot of information. This is called machine learning.

To meet this challenge, the French researchers took weather maps from the 2000s, including temperatures, pressures, soil humidity, etc. Using this data, they simulated 8,000 years of weather and more than 8,000 different summers.

Hope for electricity production

This type of artificial intelligence also aims to help us manage our renewable energies: as with the weather, it is complicated to anticipate electricity production. We saw it this winter with fears of a shortage. In addition, the share of renewable energies, by definition intermittent since they depend on the weather, will increase. Wind turbines, photovoltaic panels, hydroelectricity… Here too, such a tool could make it possible to predict weather conditions well in advance, and therefore adapt the energy production system.