Can AI Cut Tornado And Storm Deaths?

Elena Raklionskaya Pexels

Storm systems containing tornadoes tore across Missouri, Arkansas, Alabama, and parts of adjacent states. Some of the tornadoes had EF4 designations, which makes them among the most violent of their kind. A total of 80 tornadoes were registered, and 39 people died. One of the critical issues this destruction raises is whether weather forecasting can be improved, with a special eye toward forecasts for extremely local areas.

A major part of current forecasting tools is based on physics. The ECMWF states, “Atmospheric physics is a vital part of a weather forecast model and is often referred to as the physical parametrization.” While this is state-of-the-art, it does not rely on AI models, which have only existed for just over two years.

Google Weather Product

Google created one of the recently developed weather forecast services. GraphCast is described as “Our state-of-the-art model delivers 10-day weather predictions at unprecedented accuracy in under one minute.” Google claims it is a large advance in offering earlier warnings than current forecast tools offer. This is particularly true for wind storms, hurricanes, and weather that create flooding.

Nvidia Forecasts

AI Chip company Nvidia has released a tool. The company claims, “Up to 10,000 times more energy efficient than traditional high-resolution weather prediction, Earth-2’s CorrDiff model could improve disaster preparedness and save lives by enabling AI-powered kilometer-scale forecasts.”

There is an argument that if these systems work as well as the companies claim, violent weather would be signaled well before current tools.

More from ClimateCrisis 247

Similar Posts