AI is often criticised for the energy its data centers consume, which is driving more fossil fuel use and contributing to climate change. But it is also helping meteorologists make markedly more accurate forecasts, including for hurricanes, which are notoriously complex to predict and pose a growing threat in a warming world.
The US National Hurricane Center (NHC) began using Google DeepMind’s AI model last year and is working with the company on developing its successor. The work has come to the fore this year -- as an intense El Nino weather pattern led to a record number of hurricanes in the Pacific, and Hurricane Isaias struck the southeastern United States last month as a major storm.
Ferran Alet, a DeepMind research scientist who leads its cyclone work, explained that Google doesn’t have a hurricane-specific model -- rather it is the same model used for everyday weather, pushed to its limits. "We see extreme weather as almost like something we should strive for because it’s the most challenging thing for us and then everything will follow,” he told AFP.
Prior to AI, the NHC used several kinds of models to forecast tropical cyclones. "Dynamical” models use supercomputers to solve the equations of fluid motion and thermodynamics that govern the atmosphere. Statistical models rely on historical relationships between storm behaviour and variables such as location and date, and hybrids combine physics and statistics. AI models, on the other hand, learn from large datasets, like statistical models, but on a vastly bigger scale and with the ability to pick up far more complicated patterns.
A DeepMind paper published in Nature in August found the model gave forecasters on average an extra day’s worth of predictive accuracy, with three-day forecasts as good as what earlier models could provide for only the next two days.