Steve Penny, Data Assimilation Lead, Sofar Ocean
Advances in integrating Artificial Intelligence / Machine Learning (AI/ML) with Data Assimilation (DA) to produce data-driven weather forecasts.
A number of exciting advances in Machine Learning (ML) weather models have occurred within the past year. An important next step is to determine how such models could be used to make forecasts - including incorporating observational data, producing accurate ensemble forecast statistics, and applying data assimilation (DA) methods. I will give an overview of the recent advances in ML weather models, discuss some of the deficiencies in their evaluation from a DA perspective, and describe some conditions from dynamical systems theory that we have found necessary in order for ML models to be useful in a DA and forecasting context.