“Nowcast” (instead of forecast): A short-term prediction, say, of the next two hours.

Google-owned (NASDAQ: GOOGL), London-based AI Lab DeepMind collaborated with the Met Office to develop a nowcasting system for predicting, accurately, the likelihood of rain over the next 90 minutes. In comparison, existing forecasting methods can do so for periods from six hours to two weeks. (MIT Technology Review)

Deep-learning tool DGMR by DeepMind

DeepMind’s tool, DGMR (deep generative model of rainfall) , correctly predicted various rainfall parameters – such as location, extent, movement, and intensity – over 89% of the time.

That is quite an achievement, given that a host of weather-related factors such as temperature variation, cloud formations, and wind can affect how much water is held up in the sky, and when and where it’s going to fall.

To figure all that out, and then to issue a prediction valid for the next couple of hours, is a dramatic improvement over conventional methods, which are usually massive computer simulations that are useful for only longer-term assessments of rain.

However, accurate predictions over the very short term, as by DGMR, can prove immensely useful for industries such as aviation, outdoor events, and tourism. They are also critical for disaster management in the case of severe storms and floods that can ruin life and property.

“Extreme weather has catastrophic consequences, including loss of life and, as the effects of climate change suggest, these types of events are set to become more common,” Met Office partnerships and product innovation head Niall Robinson said. “As such, better short-term weather forecasts can help people stay safe and thrive.”

How it worked

DeepMind scientists fed weather-related radar measurements of clouds into a deep generative network that was able to create new radar snapshots that extrapolated the actual input data.

DeepMind senior scientist Shakir Mohamed said: “It’s very early days but this trial shows that AI could be a powerful tool, enabling forecasters to spend less time trawling through ever growing piles of prediction data and instead focus on better understanding the implications of their forecasts.”

“This will be integral for mitigating the adverse effects of climate change today, supporting adaptation to changing weather patterns and potentially saving lives.”

Related Story: AI Helps Researchers Understand Extreme Weather In The Midwest