By Aryaman Bhatia, Science and Tech Editor, The Pawprint DAA
DeepMind, a Google-owned London AI lab, and the University of Exeter collaborated with the Met Office to develop the so-called nowcasting technology. Traditional techniques employ complicated equations and frequently forecast for periods ranging from six hours to two weeks. The AI system can produce more accurate short-term forecasts, particularly those for major storms and flooding. Climate change is making it more difficult to predict unfavorable weather conditions, as the frequency and intensity of heavy rain rises, leading academics to assume that both considerable material damage and mortality may result.
The algorithm learned how to detect typical patterns of rainfall using UK radar data from 2016 to 2018, was tested on maps from 2019, and was judged to be correct in 89 percent of cases by 50 Met Office meteorologists. According to the study, which was published in the journal Nature, “meteorologists considerably favored the [AI] approach over rival approaches.”
“It’s still early days, but our study indicates that AI may be a powerful tool, enabling forecasters to spend less time combing through ever-expanding mounds of prediction data and instead focus on better comprehending the consequences of their forecasts,” said DeepMind senior scientist Shakir Mohamed.
To combine effective A.I. approaches with weather forecasting, the team overlaid six sides of a cube onto planet Earth, then flattened the cube’s six faces, similar to an architectural paper model. Because of their particular significance in weather, the authors handled the polar faces differently as one approach to enhance prediction accuracy. The data-driven model would require more detail before it could compete with existing operational forecasts, according to the authors, but the idea shows promise as an alternative approach to generating weather forecasts, particularly with an increasing amount of previous forecasts and weather observations.
As previously said, as artificial intelligence advances, scientists will be able to forecast catastrophic weather disasters such as cyclones, tsunamis, and acid rain, all of which are fatal. However, the technology is still in its early stages and will take at least five more years of research before it can truly demonstrate its actual potential and anticipate events years in the future, allowing the human race to come up with answers to issues that haven’t even occurred yet!