Artificial intelligence can now accurately forecast rain 2 hours earlier

DeepMind is working with the UK National Weather Service.

Weather prediction system based on artificial intelligence  of DeepMind has managed to make accurate forecasts rain with a period of one or two hours in advance instead of a few days as do the current numerical prediction models.

Numerical weather prediction models are currently capable of providing planetary-scale predictions several days in advance, but they fail to make predictions in short time frames of a few hours.

The challenge of forecasting rainfall and its usefulness

Knowing precisely if it will rain in an hour or two could help to better organize the day to day, and even has a direct impact on water management, agriculture, aviation, emergency planning and outdoor events .

This obstacle is the one that the DeepMind team , in collaboration with the UK National Weather Service, seeks to improve to offer accurate predictions in less than two hours.

The immediate forecasts are made possible by radar data and combined with machine learning. The adopted approach employs statistical, economic and cognitive measures in a generative model that makes predictions based on previous radars.

“Our model produces realistic and consistent predictions spatiotemporally on regions of up to 1,536 km x 1,280 kilometers and with delivery times of between 5 and 90 minutes in advance”, they explain in the text published in Nature .

The researchers say that the generative model has been ranked first in the evaluations of more than 50 meteorological experts for its “precision and usefulness in 89 percent of the cases compared to two competing methods.”