Effective intelligence contributes to short-term forecasts and storm forecasts and heavy rainfall, according to a study by the UK Meteorological Office and DeepMind, a London-based AI company.
Their project focused on “nowcasting” – pointing to the time, location and intensity of rainfall at higher altitudes up to two hours in the future – which are not well managed by computers used to predict big weather the next day or week.
This is especially important for applications ranging from alerts to emergency water hazards that could give emergency planners the impression that heavy rain is coming.
Results, published in Nature, point out that an AI-based approach called “modeling modeling” or DGM surpassed other destructive methods in a number of ways. It was first selected as the most accurate and effective by 89 percent of a team of 56 climate experts, who were blinded to the predictions.
“Improving the accuracy of short-term forecasting is a top priority,” said Niall Robinson, head of collaboration and innovation at Met Office, Britain’s meteorological and weather forecasting industry.
“Storm weather has serious consequences including damage to life and, as the effects of climate change show, these events should be common,” he added. “This study demonstrates that AI that can be presented as a powerful tool to help predict the future in terms of our understanding of how our climate is changing.”
DeepMind and Met Office researchers trained DGM to predict rainfall (rain and snow) by examining three years of UK maps. This shows the amount of rain that falls every 5 minutes at a stability of 1km.
AI analyzes the last 20 minutes of radar and predicts rain for the next 90 minutes © DeepMind
According to Shakir Mohamed, DeepMind’s chief researcher, the AI methods involved are different from the other two areas where DeepMind has been successful: play a game such as Go and inform form of protein molecules. The company was founded in 2010 and was acquired by Google in 2014.
DGM focuses on mental capacity as rainfall rather than simply achieving other results such as winning games or recognizing how proteins interact.
Robinson said the Met Office is considering how to use DeepMind research to better demonstrate how it works. “We need to carefully consider how the new equipment is used and maintained, the best environment for climate change experts, and how it fits into all other predictions we offer,” he said.
Looking ahead, Robinson said the Met Office “is exploring where we can agree on other research questions. [with DeepMind], as well as the use of AI to determine how climate change affects the process and how the weather is illuminated to create a better climate. “
The researchers said their aim was not to promote meteorologists and mechanics but to fulfill their mission.
“AI can be a powerful tool, helping predators not to spend too much time on continuous platforms but instead focus on understanding the meaning of what they are predicting,” Mohamed said. “This will help reduce the risk of climate change today, and contribute to climate change and save lives.”