Deep Learning System Accurately Predicts Extreme Weather Events

Deep Learning System Accurately Predicts Extreme Weather Events

Rice University engineers have developed a deep learning computer system that accurately predicts extreme weather conditions using a outdated analog method. By examining the maps, Rice’s self-learning “capsule neural networks” are able to predict the five-day forecasts of extreme weather with 85% accuracy using minimum information about current weather conditions.  Despite continuous progression in the weather forecasts with improved and powerful computer-based numerical weather prediction (NWP), NWP failed to predict the heatwaves in France in 2003 and in Russia in 2010. Researchers are further developing the system to use it as an early warning system for weather forecasters and to learn more about the atmospheric conditions that lead to extreme weather.

Disclaimer: The above article has been aggregated by a computer program and summarised by an Steamdaily specialist. You can read the original article at rice
Close Menu