Speaker: Hannah Christensen (Oxford)
Title: Next Generation Earth System Prediction
Abstract: Earth System Prediction is changing before our eyes. The past few years have seen two major developments occurring in parallel. The first is the advent of higher and higher resolution climate simulations: we can now produce simulations many decades long with resolutions of just a few km in atmosphere and ocean. Meanwhile we are also observing the Machine Learning (ML) revolution, with purely statistical models now able to out-perform traditional dynamical prediction models for many tasks. In this presentation I will discuss how we can bring together high-resolution and ML to advance Earth System Prediction. A key focus will be on how ML tools can serve us as modelers. We will discuss how we can use ML tools to quantify and understand deficiencies in high-resolution models, highlighting regions that cannot be fixed by resolution alone. We will then explore how ML approaches can provide us with solutions to fix these remaining stubborn model biases.