Artificial intelligence is now capable of predicting hurricane formation and intensity. — AFP Relaxnews
Artificial intelligence is making it easier to predict the formation and, above all, the course of hurricanes and other tropical storms that hit the coasts of the US in particular. Extremely fast and already relatively effective systems are now showing promising results.
Major names in tech like the US firms Google and Nvidia, and China’s Huawei, are working on developing new AI-powered weather models. These are currently being tested during the hurricane season, which takes place from May to November, with peak activity generally observed from August to October, depending on the region.
These models have been designed to generate forecasts very quickly, in just a few minutes, unlike traditional models which can take hours to compute. Generally speaking, the results already seem to be promising, according to a Wired report.
With Hurricane Lee, for example, which recently hit Maine in the US and the provinces of New Brunswick and Nova Scotia in Canada, the storm’s trajectory was accurately predicted.
However, these systems are far from infallible. For example, they are not yet capable of predicting extreme events or of accurately quantifying future rainfall.
Unlike traditional weather models, based on past or real-time observations of factors such as temperature, wind and humidity, those based on AI operate on machine learning from decades of physical atmospheric data, from which they are trained to recognise patterns. This is just the beginning, and these models will be enriched over time with new data, in order to improve future versions.
In any case, AI is already producing extremely rapid forecasts, with projections made in a matter of minutes. Even if the results are still sometimes approximate, they can help meteorologists to anticipate certain disasters. – AFP Relaxnews
