June 5, 11:15am
Location: Santa Clara II

Augmenting Modeling Capabilities for Weather Forecasting and Climate Science using AI

As part of NVIDIA’s Earth-2 initiative to help address the challenges of simulating current and future climates accurately and with low latency, we present some of our latest results on AI-based modeling of atmospheric dynamics. NVIDIA’s FourCastNet, short for Fourier Forecasting Neural Network, is a global, data-driven machine learning model based on a neural operator architecture which provides accurate short-to-medium-range global weather forecasts at high-resolution. FourCastNet offers speedups of up to four orders of magnitude over numerical models and has comparable accuracy with the world’s premier operational Numerical Weather Prediction (NWP) models. AI models offer the opportunity to improve our understanding of extreme weather events by augmenting traditional models with unprecedented capabilities for large-ensemble simulation. We present NVIDIA’s continuing efforts for enabling earth-science research at an unprecedented computational scale.

Jaideep Pathak is a research scientist at NVIDIA corporation working on the applications of large-scale AI modeling for earth science research. He obtained his PhD in Physics from the University of Maryland, College Park and held a postdoctoral research fellowship at the Lawrence Berkeley National Laboratory with the National Energy Research Scientific Computing Center prior to joining NVIDIA.