Speaker
operations numerical weather prediction presentation machine learning universityThe European Centre for Medium-Range Weather Forecasts is in the process of delivering a 10-year strategy for machine learning in the numerical weather and climate prediction. This strategy involves specific objectives regarding the prediction quality and performance of operational forecasts, as both ac- curacy and the timely delivery of weather forecasts are central to the work at ECMWF. Weather forecasting provides a unique opportunity to combine knowledge from multiple domains including geosciences, simulations, and statistics with data-driven machine learning models. This talk will present a general overview of different areas where machine learning can improve these objectives and then go into specific applications and their challenges. Specifically, I will discuss different post-processing techniques of forecasts that provide unique challenges regarding the amount of data, mathematical foundations and domain knowledge to approach this topic, which provides insights into bridging the gap between classically trained machine learning experts and decades of knowledge and conventions from subject matter ex- perts.
Visit Site Download
Screengrab