Destination Earth

The twins monitor and simulate natural phenomena, hazards and human activity, and the goal for 2030 is a full digital replica of the Earth.¹ Destination Earth (DestinE) is funded by the European Union and run under the Commission's DG Connect by three entrusted entities.^5,8 ECMWF builds the Digital Twin Engine and the first two twins.¹ ESA runs the service platform that users enter through.¹ EUMETSAT runs the data lake, which joins the twins' output with Copernicus and other data.¹ The contribution agreements were signed in 2021 and the launch event was held in 2022.¹ ECMWF reported on 30 June 2026 that the twins and the engine had been operationalised in phase two.⁸ Phase three runs to June 2028.⁸

The Climate Change Adaptation twin is led by CSC, Finland's IT Center for Science, with ECMWF.³ Three climate models, ICON, IFS-FESOM and IFS-NEMO, cover the years 1990 to 2049 with hourly output, at about 5 km for the atmosphere and land and 5 to 10 km for the ocean and sea ice.³ Impact applications for wind energy, wildfire, hydrology and extreme rainfall run inside the same workflow.³ Storyline runs replay a recorded event, such as the 2024 Central European floods, under past, present and two-degree-warmer climates.^3,8

The Weather-induced Extremes twin has two parts.⁴ The global part is built on ECMWF's Integrated Forecasting System at 4.4 km.⁴ Each day it starts from the 9 km operational analysis (the best current estimate of the atmosphere, made from observations) and runs four days ahead on the LUMI supercomputer.⁴ The regional part, led by Météo-France on the ACCORD system, zooms to 500 to 750 m over Europe two days ahead.⁴ Flood, air-quality and renewable-energy models are being coupled to the regional part.⁴ ECMWF's December 2025 wrap-up says the global part runs daily and the regional one has been applied to hundreds of European events.⁵

The Digital Twin Engine is ECMWF's software underneath both twins. It deploys them on EuroHPC supercomputers;² the twins' workflows run on LUMI and MareNostrum5.⁵ Their output exceeds a petabyte a day, and the engine's Polytope service lets a user pull out only the slice of data they need.⁵ Machine-learning components for land, ocean, sea ice, waves and hydrology are built with Anemoi, an open-source framework from ECMWF and its member states.⁵ They are built on GPUs at LUMI, MareNostrum5, Leonardo and MeluXina.⁵ ECMWF says phase three will couple them into a European AI Earth-system model and make AI-ready datasets from twin output.⁸

The way in is the DestinE Platform run by ESA.¹ Registration is free and opens a set of services for data access, processing and visualisation.⁶ The twins' own data need "upgraded access", which public authorities, academia and research, and SMEs and start-ups apply for after registering.⁷ The platform says this access is for European actors and takes several business days to review.^6,7 Data then comes through Polytope or the Harmonised Data Access API.⁴ For the Extremes twin, Polytope serves native output for the last 15 start dates.⁴ The Climate twin has a user guide and example Jupyter notebooks.³

Outside groups can propose a service of their own, which DestinE stakeholders evaluate before it is integrated.⁶ The platform's October 2026 updates add UHEATS, an urban heat pilot service, and HiReNext, a high-resolution forecasting service for transmission system operators.^6,9 A flood forecasting service for Italy, FloodCast, is in beta testing.¹⁰