Weather forecasting is undergoing a paradigm shift, moving from the physics-based traditional numerical weather prediction (NWP) to machine learning (ML) based models. Operational weather prediction relies on using current atmospheric observations to provide initial conditions to NWP models (data assimilation), which then integrate the governing equations to make weather forecasts for the next 1-14 days. Advancing in numerical modeling and data assimilation have made slow but steady progress over the last four decades, leading to a “quiet revolution” in weather forecasting. Recently, advances in machine learning, availability of high-quality data, and advances in hardware (e.g. GPUs/TPUs) have set the stage for deep learning to tackle problems for weather and climate. In the last two years, several deep learning-based models for weather forecasting have been demonstrated with skill approaching or exceeding the best available NWP weather forecasts. These models include Graphcast, Fourcastnet, FuXi, Pangu-weather, ClimaX, Fengwu, and Stormer, each with vastly different training methods, machine learning architectures, and variables predicted. This talk explores the rise of ML-based weather prediction and discusses research at Argonne that is helping revolutionize how weather and climate are researched using machine learning.

ADD TO CALENDAR / View Recording