News

The fast development of DestinE’s Climate Change Adaptation Digital Twin

6 May 2024
A simulation of clouds and surface temperature from one of the prototype projections of the Climate DT performed on EuroHPC LIUMI with IFS-NEMO (at 4.4 km resolution for atmosphere and land and 1/12 degrees for ocean and sea ice)

The Climate Change Adaptation Digital Twin (Climate DT) developed in the Destination Earth initiative (DestinE) of the European Commission is a pioneering effort to build an operational and interactive capability for novel multi-decadal global climate projections with local granularity (at 5 to 10 km resolutions), providing information across scales that matter for decision-making in support of climate change adaptation. 

The achievements of the strong partnership that implements the Climate DT, led by the CSC – IT Center for Science, have been remarkable, given that work only started in October 2022 and access to the first pre-exascale supercomputer in Europe, LUMI was gained in April 2023. Here is a brief overview of what has been achieved during the first phase of DestinE (December 2021 – 2024 June) and the key objectives of the Climate DT teams for the second phase of the initiative (June 2024 – June 2026).

The Climate DT in a nutshell

While the previous multi-decadal climate projections were often research-focused efforts performed only every 7 to 10 years, the Climate DT will produce operational, quality-assured simulations at least once every year. This increased frequency will allow the integration of the latest advances in science and technology to better support decision-making and climate science.   

At the same time, the Climate DT is setting a unique capability to produce bespoke, cutting-edge numerical simulations addressing ‘what-if’ questions related to the impact of certain scenarios or policy decisions on the evolution of our planet. 

The Climate DT exploits and further develops the new generation of global storm-resolving and eddy-rich (or ‘km-scale’) models built through a cooperative model development approach supported by the European Horizon projects nextGEMS and EERIE as well as national projects such as WarmWorld in Germany and Gloria in Spain, involving dozens of leading climate and weather centres, supercomputing centers and academia throughout Europe. The three models used,  ICON, IFS-NEMO, and IFS-FESOM, are run at resolutions ranging between 5 and 10 km for the different components (atmosphere, land, ocean, and sea ice) exploiting the world-leading supercomputing facilities of the European High-Performance Computing Joint Undertaking (EuroHPC JU).

The Climate DT brings added value to the users compared to current climate modelling activities by providing globally consistent data with higher temporal (hourly) and spatial resolution (5 to 10 km), by delivering an updated data set more frequently and considering the needs of the users in terms of the output variables and simulation design.

The Climate DT is implemented by a strong partnership led by CSC, involving 13 leading climate institutions, supercomputing centres, national meteorological services, academia, and industrial partners, through a contract procured by ECMWF and funded by the European Commission in the framework of DestinE.

In less than two years, the Climate DT teams have succeeded in building the key elements of the system, deploying it on the EuroHPC LUMI supercomputer, demonstrating its key innovative features at scale, and performing the first-ever global multi-decadal climate change projections at 5 km resolution. This work was carried out in close collaboration with teams at ECMWF developing the Integrated Forecasting System – one of the models underpinning the Climate DT – and the Digital Twin Engine

The key achievements over the past twenty months (November 2022 – April 2024) can be summarised as follows.

A simulation of surface temperature from one of the prototype projections of the Climate DT performed on EuroHPC LUMI with IFS-NEMO (at 4.4 km resolution for atmosphere and land and 1/12 degrees for ocean and sea ice). Credit: ECMWF / Andreas Müller / LUMI

Key achievements in the first phase of DestinE 

Development and implementation of the end-to-end Climate DT workflow on LUMI 

Given the complexity of the task, developing and deploying the end-to-end Climate DT workflow on LUMI—the first pre-exascale supercomputer in Europe and one of the largest supercomputers in the world (no. 5 in the Top500)—in less than one year can be considered a real breakthrough.

The Climate DT workflow covers the full chain from global km-scale climate simulations to impact-sector applications. It thus makes the link between climate models and applications for the sectors most impacted by climate change via selected use cases. With this novel approach, the Climate DT moves towards a co-production of context-oriented climate information for impact sectors like renewable energy or water management, as opposed to traditional climate information systems, where users are rather recipients of the data produced.

The Climate DT workflow including three global storm-resolving and eddy-rich models, a standardisation of the climate model output, through the creation of a so-called Generic State Vector, the streaming of the GSV to different applications.
The Climate DT workflow including three global storm-resolving and eddy-rich models, a standardisation of the climate model output, through the creation of a so-called Generic State Vector, the streaming of the GSV to different applications. 

The raw output of the different models is transformed into a standardised Generic State Vector (or GSV). This is an important novelty of the Climate DT that ensures a unified model output format in terms of parameters, units, and grid, facilitating consistency and interoperability across models and datasets and enabling a step change in the usability of model output for applications.

The unified grid used for the Climate DT models is an equal-area hierarchical HEALPix grid, optimised for hierarchical data exploitation which is required in high-resolution visualisation, local data retrieval and AI/ML training. Credit: MPI

The applications included in the workflow are the quality assessment and uncertainty quantification framework (AQUA) as well as impact-sector applications (e.g., a hydrological model or computations of relevant indicators) that transform the climate data to actionable information on climate change impacts through a co-design approach. Data reduction tasks (so-called one-pass algorithms) are also important elements of this process. These ensure the efficient processing of climate datasets of unprecedented size at their native spatial and time resolution on the fly (as the model runs). This method ensures access to the information produced by the model while reducing computational and data overheads and improving responsiveness and the satisfaction of user requirements.

Near-surface wind speed, on a grid of 100 km resolution – corresponding to previous global climate simulations, and of 5 km resolution – as simulated in one of the prototype runs of the Climate DT. Use the slider to compare the different resolutions. Credit: Aleksander Lacima-Nadolnik & Katherine Grayson, BSC, Climate DT Team

Computation of climate indicators in the offshore wind energy use case: the raw model output (wind speed at 100m) is transformed into relevant indicators for wind energy, such as capacity factors, via data-reduction (one-pass) algorithms which compute the weekly distribution of the wind speed in each point of the globe (at native resolution, i.e. 5 or 10 km) from the hourly model output while the model runs; allowing to then derive the weekly capacity factor for any wind turbine.

Five applications for use cases have been integrated into the Climate DT workflow during  Phase I. These include:

  • An energy application focused on onshore and offshore wind energy supply and demand in the future climate
  • A hydro-river application for investigating freshwater availability and floods 
  • A hydrometeorology application for exploring statistics and characteristics of hydrometeorological extreme events, 
  • An urban application focused on urban heat waves and heat stress
  • A fire and carbon application for investigating the future risk of occurrence and emissions of wildfires. 

Other use case applications can access the Climate DT data from the DestinE data lake, implemented by Eumetsat, via the DestinE Core Service Platform implemented by ESA.

Software and data infrastructure as the basis for an operational Climate DT system

Another achievement of the Climate DT teams during Phase I, was developing the basis for the future operationalisation of the system. This included implementing the software infrastructure, workflow and data handling capabilities as part of the DestinE Digital Twin Engine, in close collaboration with ECMWF, and consisted of several elements.

The first important element was deploying the three climate models involved in the Climate DT on LUMI. The models need to be adapted and optimised to run efficiently on the first pre-exascale European HPC platform using heterogeneous architectures and both general-purpose and accelerated processors. This is a very important aspect and an absolute requirement to efficiently perform kilometre-scale global simulations (with a target of 5 to 10 km) on a multi-decadal timescale on HPC systems distributed across Europe.

Near surface winds over the Tibetan plateau in one of the prototype projections of the Climate DT with ICON at 5km resolution. Credit: MPI

ICON is now fully running its atmosphere component on the AMD GPU partition. IFS is also running partially on the AMD GPUs. This was a major technical success of Phase I. The models are now adapted for the other two pre-exascale computers, Leonardo and MareNostrum 5, which use NVIDIA GPUs.

A second element consisted of the development and deployment of the Autosubmit workflow manager, which allows to orchestrate the Climate DT components. ​​Autosubmit offers full traceability of the runs performed, with the possibility of repeating any step in case of infrastructure or communication failures. It includes a flexible way to describe the workflow, modern monitoring capabilities, and a notification system to inform the users promptly about the progress of the experiments, the data availability, and any incidence.

Thirdly, the Climate DT teams have defined and implemented a common data portfolio with data governance that follows WMO international standards. A monitoring framework containing a series of checks that verify that quality control is performed as the model runs was also developed, allowing data consumers to access the data reliably in streaming mode. This is possible because data formats are strictly defined by the use of a single approach to handle the model output, known as MultIO.
For more details on the Digital Twin Engine software architecture, visit the DestinE Digital Twin Engine repository.


BSC teams verifying the AQUA results. Credit: BSC

Finally, the AQUA framework was also included in the workflow and constitutes a unified basis for the evaluation and uncertainty quantification of the climate simulations that can be performed simultaneously as the models run, and also a posteriori. AQUA was developed in a modular and flexible manner in which new diagnostics can be added between simulations.

Production of the first set of global climate projections with local granularity on multi-decadal time scales 

The Climate DT infrastructure allows to both produce climate projections on a regular basis and perform bespoke simulations to answer specific ‘what-if’ questions for assessing the impacts of different scenarios and policy decisions, or for creating storylines of how a particular recent extreme weather event would unfold in an even warmer climate.

For the first prototype runs performed in phase 1, the Climate DT follows a modified version of the HighResMIP protocol, which entails running: (i) projections for the period 2020–2040 following the emission scenarios defined by ScenarioMIP, (ii) historical simulations starting in 1990, and (iii) control simulations with constant 1950 forcing, to evaluate model drift and performance. This experimental definition ensures continuity in the production of climate information with the current state-of-the-art research experiments that are widely used by decision-makers. 

The simulations are carried out at resolutions between 5 and 10 km for the different earth system components, producing consistent global climate information with local granularity. These simulations thus bridge the gap between large-scale climate projections and local climate impacts and avoid data gaps and possible inconsistencies that come with existing regional downscaling efforts.

Having the system ready for the production of the first-ever prototype projections at these km scales in December 2023, less than a year after gaining access to LUMI, was another major milestone during Phase I.

To complete part of these simulations on time for the official launch of the DestinE system in June 2024, DestinE benefited from a special reservation in LUMI from mid-February to mid-April, thanks to the support from the EuroHPC JU and CSC/LUMI.

The following simulations were carried out after two intense months for the Climate DT teams running the simulations in close collaboration with ​​ECMWF.

  • The first-ever multi-decadal climate projections at ~ 5 km across earth-system components, following the Shared Socioeconomic Pathway (SSP) 3-7.0 from ScenarioMIP (SPP3-7.0 scenario) with IFS-NEMO and ICON, streaming information to selected applications (e.g. wind energy and urban heat). ICON was run at 5 km across earth-system components and completed so far the period 2020–2030. IFS was run at 4.4 km for the atmosphere and land, coupled with NEMO at 1/12 degree for the ocean and sea ice, for the period 2020–2040.
  • Historical simulations for the past (starting in 1990) were initiated at about 10 km for ICON and IFS-NEMO, and completed 16 and respectively 12 years so far.

In the next months, control simulations with fixed radiative forcing conditions (starting in 1950) to diagnose model drifts will be run to complete an initial prototype contribution to the HighResMIP scenario. 

For all types of simulations, the ocean/sea-ice models were spun up using stand-alone ocean runs forced with the ERA5 reanalysis.

An innovative capability to explore how recent extreme events will look in the future  

Another notable innovation of the Climate DT is the production of storyline simulations using IFS-FESOM. This framework is designed to create global, kilometre-scale storylines of recent extreme weather events (2017 to today), enhancing the expression of climate change with respect to contemporaneous experiences to support more effective adaptation and mitigation strategies. These storyline simulations reconstruct the unfolding of recent extreme events such as heat waves, floods, and droughts across various climate scenarios (e.g., pre-industrial, present-day, +2ºC, and +3ºC) by nudging the atmospheric circulation in IFS-FESOM for different climate states to ERA5 reanalysis data for the period (1 January 2017 to today). 

The first simulations of this type were also produced on the LUMI supercomputer with IFS-FESOM, using resolutions of 9 km for the atmosphere and land, and 5 km on average for the ocean and sea ice. The observed weather from 2017–2020 was simulated for the climate conditions of the 1950s, 2020s, and 2050s, making the consequences of climate change more tangible.

Exploiting the value of EuroHPC systems for decision-making

The Climate DT harnesses the potential of the largest European HPC systems to perform its complex km-scale simulations. The EuroHPC JU contribution is critical to enable the Climate DT’s capabilities.

Indeed, the Climate DT simulations require extreme computing power and data handling capacities. For example, one 30-year simulation at 5km resolution needs 1.5 M GPU hours or 34 M core CPU hours. The Climate DT simulations are only possible thanks to the computing time awarded by the EuroHPC JU through a special access call on its three pre-exascale supercomputers: LUMI, MareNostrum 5, and Leonardo. During Phase I, the climate models, applications, and workflow were successfully deployed and optimised on LUMI (CPU & GPU partitions). Now they are being deployed on MareNostrum 5 and Leonardo. 

The Climate DT demonstrates thus the value of the EuroHPC systems for science-based, climate-related decision-making. At the same time, this effort supports preparing the HPC systems to handle highly complex Earth system simulations and data handling operations. The Climate DT thereby ensures the European supercomputing ecosystem serves climate adaptation efforts and supports critical advances in climate science.

Fostering innovation through European collaboration

Destination Earth highlights once again the value of European collaboration, bringing together expertise from different fields. The Climate DT contract team, led by CSC – IT Center for Science, includes 13 European organisations and excellence centers in climate modelling, high-performance computing, and impact assessments. 

The joint effort for the Climate DT also sets a perfect example of how European investments in research through Horizon projects (nextGEMS, EERIE), national funding (WarmWorld, Gloria) and European cooperation (ECMWF), combined with investments in digital technologies, supercomputing, and AI, through the Digital Europe Programme of the European Commission ensure Europe’s global leadership in climate modelling and the provision of climate information.

Phase 2: towards operationalisation

In the short term, a key objective of DestinE is the integration of its key components; the digital twins and the Digital Twin Engine implemented by ECMWF, the core Service Platform implemented by ESA and EUMETSAT’s Data Lake, ahead of the launch of DestinE by the European Commission on 10 June 2024 from the LUMI data center in Kaajani, Finland. 

Following the achievements of Phase I, during the next phase of DestinE (June 2024 – June 2026), all the teams involved in the Climate DT will continue working at full speed towards the consolidation and operationalisation of this ambitious system.

This implies operationalising the three next-generation storm- and eddy-resolving models (ICON, IFS-NEMO and IFS-FESOM); and the framework used to produce multi-decadal projections regularly, following a HighResMIP-like protocol, with a 5 km global grid, and bespoke climate simulations (for different scenarios or storylines of extreme events).

In addition to the operational framework, Phase II will include continuous development of the Climate DT system. The monitoring, evaluation, and uncertainty quantification capabilities will be further enhanced. Another objective is to further develop capabilities for special simulations, including storyline simulations for future periods of extremes as well as what-if scenario simulations, enabling a new level of interactivity. 

In Phase II the added value to users will be showcased through four use cases that are implemented within Climate DT in the form of impact sector applications included in the workflow. The selected use cases cover societally relevant climate change adaptation domains, including renewable energy management, disaster risk management (regarding wildfires and floods), agriculture, and water management, and involve key users. 

Another key objective for all the teams involved in the Climate DT is the integration of the current fast-evolving artificial intelligence and machine learning breakthroughs in the system. The Climate DT will benefit from these new techniques to, for example, support the user experience through the development of chatbots that make the Climate DT data more accessible to end-users, enabling access to the wealth of data for AI model training, and through the use of ML models for uncertainty quantification.

Further reading:

Climate Change Adaptation Digital Twin: a window to the future of our Planet

Destination Earth is a European Union-funded initiative launched in 2022, with the aim to build a digital replica of the Earth system by 2030. The initiative is being jointly implemented under the leadership of DG CNECT by three entrusted entities: the European Centre for Medium-Range Weather Forecasts (ECMWF), responsible for the creation of the first two ‘digital twins’ and the ‘Digital Twin Engine’, the European Space Agency (ESA) responsible for building the ‘Core Service Platform’, and the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT), responsible for the creation of the ‘Data Lake’.

The Climate DT, procured by ECMWF is developed through a contract led by CSC-IT Center forScience and includes Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research (AWI), Barcelona Supercomputing Center (BSC), Max Planck Institute for Meteorology (MPI-M), Institute of Atmospheric Sciences and Climate (CNR-ISAC), German Climate Computing Centre (DKRZ), National Meteorological Service of Germany (DWD), Finnish Meteorological Institute (FMI), Hewlett Packard Enterprise (HPE), Polytechnic University of Turin (POLITO), Catholic University of Louvain (UCL), Helmholtz Centre for Environmental Research (UFZ) and University of Helsinki (UH).   

We acknowledge the EuroHPC Joint Undertaking for awarding this project strategic access to the EuroHPC supercomputers LUMI, hosted by CSC (Finland), and the LUMI consortium, Marenostrum5, hosted by BSC (Spain) Leonardo, hosted by Cineca (Italy) and MeluXina, hosted by LuxProvide (Luxembourg) through a EuroHPC Special Access call. 

More information about Destination Earth is on the Destination Earth website and the EU Commission website.  

For more information about ECMWF’s role visit ecmwf.int/DestinE   

For any questions related to the role of ECMWF in Destination Earth, please use the following email links:   

General enquiries

Press and Communications enquiries