The Destination Earth Digital Twin for Climate Change Adaptation

This invited talk presents the Climate Change Adaptation Digital Twin of Destination Earth, the European Commission initiative building high-resolution digital twins of the Earth, delivered by an ECMWF principal scientist on behalf of the Climate DT consortium led by CSC. Destination Earth is implemented by ECMWF, ESA, and EUMETSAT, combining digital twins, a digital twin engine deployed on EuroHPC supercomputers, a data lake, and a user-facing core platform. The Climate DT runs multi-decadal simulations with three earth system models (ICON, IFS-FESOM, and IFS-NEMO) at around 5 km resolution, including control runs, historical runs, CMIP-forced projections, and extreme event storylines that replay events such as the 2024 Central European floods under pre-industrial and plus-two-degree climates. More than 210 simulated years at 5 km and over 10 petabytes of data have been produced, with all models writing to a common HEALPix grid via on-the-fly I/O post-processing so downstream applications can consume and compare outputs uniformly. The speaker shows scaling results on LUMI, MareNostrum 5, and the JUPITER GPU system, where IFS reaches 51 forecast days per day at 1 km on 512 nodes, and stresses energy efficiency as the gating factor for further resolution increases, citing roughly 40 simulated days per megawatt hour as a practical threshold. Phase three will train AI-based climate emulators on the dataset to enable larger ensembles, and the software stack, including OpenIFS and the digital twin engine tools, is open source.