BSC, a key player in the development of a digital twin to simulate future impacts of climate change on Earth

• ECMWF, ESA and EUMETSAT implement the Destination Earth initiative under the leadership of DG Connect and in collaboration with over 100 institutions throughout Europe.
• Visualizations created in Destination Earth allow us to look at our planet in ways we never thought before, as we can view the future for different climate scenarios.
• The high-resolution simulations of these future scenarios are only possible thanks to the computing power of the first pre-exascale supercomputers in Europe, such as MareNostrum 5 or LUMI.
• The Digital Twin for Climate Change Adaptation, in which BSC plays a crucial role, is implemented by a strong partnership through a contract procured by ECMWF and involves 12 institutions from 6 different European countries.
• This Climate Digital Twin is a pioneering effort to operationalise the production of global climate projections for the upcoming decades and provide globally consistent Earth system and impact sector information from global to local scales.

Framed within the European Green Deal, the European Union's ambitious Destination Earth Initiative (DestinE) aims to respond to the urgent need to enhance our capabilities to respond better and adapt to the environmental challenges posed by extreme events and climate change by developing a highly accurate digital replica of our planet. According to the initial schedule, the first phase will end in June with the demonstration at a scale of the prototypes of the first two digital twins: one twin named Climate Change Adaptation Digital Twin (aka Climate DT) and the other twin named Weather-Induced and Geophysical Extremes Digital Twin (aka Extremes DT). The Barcelona Supercomputing Center - Centro Nacional de Supercomputación (BSC-CNS) participated in these two twins and played a significant role in developing the Climate DT one during the initiative's first phase.

Under the leadership of Francisco Doblas-Reyes, ICREA Professor, director of the BSC’s Earth Sciences Department, and a world reference in climate modelling for adaptation, BSC is one of the key contributors to the development of the Climate DT, which will provide information and services to support climate adaptation policies and meet the needs of those sectors particularly vulnerable to climate change. The twin will offer a new capability to assess the impact of future socio-economic and policy decisions, which correspond to different emission scenarios, by performing cutting-edge numerical climate simulations at global scales, providing estimates of the corresponding climate hazards with unprecedented spatial resolution (5 to 10 km).

“The Climate DT represents a leap forward in the ability to produce climate change information for the next decades, not only because it uses a new generation of climate models that increase the realism of the phenomena represented through their enhanced resolutions, but also thanks to the ambitious software developments undertaken, led by the BSC. This digital twin allows users to be put at the centre of the production of climate information. The operational approach that we are implementing for the first time empowers users to formulate climate-relevant questions for the Twin to address them,” said Doblas-Reyes.

Collaborative European endeavour

The Climate DT is implemented through a genuinely collaborative European endeavour involving leading climate and supercomputer centres, national meteorological services and climate impact institutions, and it is coordinated by the Finish CSC (IT Center for Science). This partnership collaborates closely with teams at the European Centre for Medium-Range Weather Forecasts (ECMWF) and other European initiatives, fostering the development of a new generation of climate models underpinning the Climate DT (nextGEMS, EERIE, GLORIA, and WarmWorld).

Implementing the Climate DT entails fundamental computational development by integrating into the infrastructure of the DestinE’s Digital Twin Engine―the set of software facilities that underpin the digital twins―three high-resolution global climate models, a mathematical representation of the Earth's climate system used to simulate and predict its behaviour over time. The models are then deployed on the largest computers in Europe, those of the EuroHPC Joint Undertaking.

To give an idea of what such an endeavour represents, simulating 50 years of one of the climate change scenarios in DestinE at an unprecedented spatial resolution (in this case, by using 212 nodes of the EuroHPC JU supercomputer LUMI) requires 83 days of continuous execution on a pre-exascale platform, producing nearly 2,000 TB of data, i.e. 82,000 4K films. This wealth of data would have been impossible in the previous generation of climate models due to computing power limitations. In addition, even if this kind of simulation were possible, it would have taken years to complete.

During phase 2, DestinE will also make use of MareNostrum 5, one of the most potent supercomputers in Europe, thanks to strategic access granted via a special access call by EuroHPC JU. MareNostrum 5 is a pre-exascale machine, named for its computing capacity (maximum total performance of 314 petaflops), with computational power around 23 times greater than the previous generation (MareNostrum 4) and capable of performing up to 314,000 (European) billion calculations per second.

Multidisciplinary team

The development of the Climate DT requires multidisciplinary work, and the input of experts in different fields is necessary. The new twin involves fundamental computational development work, integrating three high-precision global climate models into the new DestinE digital twin infrastructure, developing groundbreaking solutions for real-time data access, and deploying the twin on the EuroHPC JU supercomputers.

During this first phase of DestinE, BSC has played (and will continue to play during the second phase) a fundamental role in deploying some of these climate models on the new EuroHPC JU supercomputers and ensuring that they run efficiently on these platforms. Moreover, BSC is also in charge of managing the vast amount of data produced by the models and processing them with innovative solutions to use them through intuitive applications for climate adaptation on energy and urban management, and providing the computing infrastructure to operate and automate the system.

Due to the extraordinary computational requirements of this new category of climate models, the contribution of researchers and engineers of the BSC’s Computational Earth Sciences (CES) group has been essential to enable the efficient use of the new generation of supercomputers, such as LUMI or the recently inaugurated MN5.

BSC’s researchers Mario Acosta and Miguel Castrillo, co-leaders of the CES group, highlight some primary outcomes achieved during the first phase. Acosta, in charge of the computational performance of the Climate DT models, commented: “Although the new computing capacity of MareNostrum 5 or LUMI supercomputers is crucial for making DestinE possible, our efforts have played a vital role in maximising the efficiency of these cutting-edge supercomputers. We have improved the performance and speed of the climate models by implementing optimisations in close collaboration with teams at MPI-M, AWI, and ECMWF.”

On the other hand, Castrillo is responsible for the Climate DT workflow. He stated, “Using a workflow tool to manage complex computational tasks developed at BSC, called Autosubmit, we have designed and implemented completely new workflow software to deploy and automate the Climate Digital Twin.” He added: “Importantly, the Autosubmit workflow manager has been critical to the operation, proving to be a rapidly deployable and reliable solution that allows us to configure, manage, curate and orchestrate the digital twin operation on multiple platforms under changing conditions.”

Pierre-Antoine Bretonnière, co-leader of the data team from BSC’s CES group and responsible for the data management, explained that “with the unprecedented amount of data generated in real-time by this project, new ways of generating, accessing and unifying the information needed to be developed, through the creation of new data, interfaces and transfer.”

Moreover, climate researchers at the BSC Earth System Services (ESS) and Climate Variability and Change (CVC) groups have substantially contributed to the execution and validation of the climate simulations, evaluating the scientific quality and accuracy of the results in close collaboration with other partner institutions, as well as the development of data processing tools that gave access to unprecedented data to illustrate the value of the new type of climate information for the urban and energy cases.

BSC will continue to play a relevant role, if not more, in the second phase of the Climate DT implementation. It will again be responsible for developing and performing simulations with one of the global climate models, which includes ECMWF’s IFS atmospheric model and the ocean and sea ice model NEMO. BSC will also maintain the whole software infrastructure, ensure the data management and provision, coordinate the development of new illustrations of the user value of the Climate DT, and show the power of artificial intelligence by introducing large language models.

About the Digital Twins of the Earth

A digital twin of the Earth system is an infrastructure that integrates observations, a cutting-edge numerical model, and subsystems for vulnerable sectors (e.g., energy resource management), all into a technological solution that favours its interaction. The twin allows the assessment of changes and their causes consistently across both local and global spatial scales and a range of time scales.

What makes Destination Earth unique is that climate data at an unprecedented spatial and time resolution will be made available from the Climate DT to the users as they are produced, allowing them to obtain information that was not possible before with such a level of detail, and will provide timely feedback on the output and further requirements. This is only possible with the latest EuroHPC JU supercomputers and a set of models and tools optimised to take advantage of the new possibilities.

This article was first published on 8 May by Barcelona Supercomputing Center.