HENS: The Huge Ensemble of Global Weather Extremes Using FourCastNet
Timothy Taylor, University of Colorado Boulder
FourCastNet (FCN), short for Fourier Forecasting Neural Network, is a deep-learning data-driven weather forecasting model developed by a team at NVIDIA. This model has the ability to make skillful weather forecasts comparable to state-of-the-art numerical weather prediction models at a fraction of the cost, allowing for new forays into large ensemble forecasting. Here, the plan to use such a Huge Ensemble to investigate the drivers of low-likelihood high-impact weather events is presented, along with some results of the specific metrics being used to evaluate the skill of FCN.
Abstract Author(s): Joey Taylor, William Collins, Ankur Mahesh