Forecasting Global Weather with Graph Neural Networks

Abstract:
Deep learning offers innovative approaches to modeling complex physical dynamics. This talk will focus on the application of a specific deep learning approach, graph neural networks, to the problem of forecasting global weather. A data-driven model was trained to step forward the current 3D atmospheric state by six hours, and multiple steps are chained together to produce skillful forecasts going out several days into the future. Our primary aim is to showcase a real-world use case in which a deep learning approach can efficiently approach or even surpass the predictive power of a traditional physical simulator.

Bio:
Ryan is a physicist working at the intersection of sensor data and data science. He is a Staff Data Scientist at KoBold Metals, where he uses his background in physics and data science to search for new deposits of battery metals. Previously Ryan was Chief Scientist at Descartes Labs and a cosmologist at the University of Chicago and Stanford University.