Speaker
Description
The European Commission’s Destination Earth (DestinE) initiative represents a paradigm shift in Earth system simulation, aiming to develop highly accurate digital twins of the Earth. At the core of this endeavour lies the Digital Twin Engine (DTE), an innovative software framework designed to connect the extreme data generation capabilities of High-Performance Computing (HPC) with the interactive, user-centric flexibility of cloud environments.
This keynote discusses the architectural challenges and solutions involved in designing the DTE to enable seamless data provisioning and sharing. A key focus will be the convergence of Numerical Weather Prediction (NWP), Climate Information, and Machine Learning (ML). We will examine how the DTE is evolving to support "AI-ready" datasets, particularly addressing the extensive data handling requirements for ML training, including the upcoming ERA6 reanalysis—the successor to ERA5 and a vital component for future AI model training. Additionally, we will outline the data throughput challenges related to operationalising ECMWF’s Artificial Intelligence Forecasting System (AIFS) and explain how we are developing scalable workflows to support the next generation of data-intensive prediction systems.