Can science be politicised?
DestinE's European ambition

Destination Earth. Ultimately, the desire to clone Earth to see how it behaves in real time. This idea, which is not at all far-fetched, comes from the European Commission and is currently being implemented by three European agencies: the European Space Agency (ESA), the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) and the European Centre for Medium-Range Weather Forecasts (ECMWF), which, through the Destination Earth programme – known as DestinE to those in the know – intend to develop an accurate digital twin of our planet. Sceptics will cry out that this is yet another example of stacking logic. Who isn't developing their digital twin these days, starting with bland avatars? For Loïc Rognant: “Many people are confused. It's true that there has been a proliferation of digital twins lately, especially in industry. In space, we create digital duplicates of physical objects to test the various stresses on equipment, for example, but the digital twin in the context of Destination Earth is very different. The aim is to scientifically monitor the evolution of climate change in real time using satellite data to obtain a global view.”

The approach is based on three main stages and is guided by the European Green Deal framework. The first stage is to model large-scale global environmental phenomena. The second is to be able to verify that these models remain representative of reality. The third stage, the ultimate goal, is to be able to carry out what is known as “What if...?” simulations based on reliable models. In short, it is a forward-looking approach. We project ourselves into the future in order to predict the evolution of climate change and its impacts, particularly in inhabited areas, so that informed, considered and effective decisions can be taken.

No one can criticise such a goal. However, it must be emphasised that, regardless of the quality of the scenarios presented, the transition from “what if” to effective action will not be easy, given the complexity of the administration/politics relationship and the very real difficulty of getting the whole world moving in the same direction. If it were simply a matter of being well informed in order to take relevant action, polar bears would no longer have white fur and we would not be where we are today. Translating scientifically proven results into action is hampered by capacity, which varies greatly from one context to another, by the financial envelopes allocated, which are subject to real internal institutional budgetary struggles, and finally by the political priorities of the moment, which tend to be short-term and geographically limited. Global issues struggle to make their way through the various decision-making layers...

The challenge of sovereign AI infrastructure

Under the leadership of the European Commission, the DestinE programme is being piloted by three institutions. The first agency involved is the European Centre for Medium-Range Weather Forecasts (ECMWF), which is responsible for scientific modelling and all the calculations needed to create reliable digital twins. The second institution is the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT). This institution is responsible for the ‘data lake’, i.e. the centralised repository that enables the storage of large volumes of data, which are then organised so that they can be processed efficiently. The third agency is the European Space Agency (ESA), which is responsible for the “core” platform, in other words, for developing the DestinE access platform, the access point where all users go to discover, use and analyse the system's resources.

As Loïc Rognant explains: “The ultimate goal is large-scale modelling. A global, continental, sub-continental or regional scope. The programme began in 2021 with the modelling component. A year later, the Data Lake component was launched, followed in 2023 by the access platform component, which Thales Alenia Space is pursuing."

This ambitious modelling effort relies on the use of high-performance computing (HPC). A European initiative, EuroHPC, has been set up around the European Union, 15 Member States and private partners to develop an ecosystem of huge supercomputers in Europe (known as ‘AI Factories’) in order to establish European digital sovereignty on a solid foundation. “The idea behind these infrastructures is to enable all European users to be close to one of these supercomputers in order to distribute computing capacity as efficiently as possible and allow as many people as possible to benefit from this digital resource.” The volume of data to be processed is growing exponentially and some AI calculations take several weeks on more traditional computers...

The first AI factories represent an estimated investment of €1.5 billion, half of which will be funded by the EU. These factories are intended to be hosted by leading technology and research entities located throughout Europe. The physical locations of the first ecosystems are in Finland (LUMI since mid-2022, then LUMI AIF at the CSC in Kajaani from the end of 2026), Spain (MareNostrum since the end of 2023, then BSC AIF at the Barcelona Supercomputing Centre from the end of 2026), Italy (Leonardo from mid-2023, then IT4LIA at CINECA, in the Bologna technology park), Luxembourg (Meluxina from mid-2021, then Meluxina-AI at LuxProvide, Bissen), the Czech Republic (Karolina at IT4Innovations NSC in Ostrava from mid-2021), in Bulgaria (Discoverer at Sofia Tech Park in Sofia from the end of 2021), in Slovenia (Vega at IZUM - Institute of Information Science in Maribor from mid-2021), in Portugal (Deucalion at FCT - Foundation for Science and Technology in Guimarães from the end of 2023), in France (Jules Verne, which will be hosted at the CEA in the Paris region from the end of 2025), in Sweden (MIMER, at Linköping University from mid-2026), in Germany (Jupiter at Forschungszentrum Jülich since the end of 2024 and HammerHAI, at the University of Stuttgart) and Greece (Daedalus, hosted at GRNET in Athens, will be operational from mid-2026, and Pharos, which will be hosted at GRNET in Athens from the end of 2026).

While no one disputes that powerful algorithms are necessary for Europe to remain sovereign and continue to exert influence in the digital sphere, the shift that has been made requires costly mega-infrastructures. Other options exist but have been less convincing to date.

The DestinE platform

Phase 1 of the Destination Earth programme's implementation ended with the launch of the first version of the platform on 10 June 2024 in Finland, during an official ceremony. The first users were able to access the first generation of the infrastructure, data and services. The system will continue to evolve until 2030. As Serge Rainjonneau explains: “Immediately after the launch, there were adjustments, particularly regarding access rights and cybersecurity. Some data is sensitive, even if we start with data that is initially open. When it is disseminated through digital twins, it becomes more sensitive, so we must remain cautious.” Loïc Rognant continues: “Typically, the platform allows us to accurately anticipate issues such as drought and the impact of global warming on agriculture. We will be able to estimate realistic consequences in terms of economics and population displacement. We cannot fail to control access. We therefore made the necessary adjustments in this regard after the launch. In September 2024, the second version of the platform replaced the first.”

As the platform is designed to inform policymakers, beyond access, another challenge lies in ensuring end-to-end reliability and integrity control. As part of the component managed by ESA, Thales Alenia Space is responsible for the cybersecurity of the entire platform and must guarantee the traceability of all data. Loïc Rognant uses this analogy: “As with food or medicine, you need to know where the satellite image comes from, which specific satellite it was taken from, which algorithm was used to process it, by whom, on which platform... This information is necessary to guarantee reliability, from the source to the end user. It's like an identity card for the data, really. If there is an anomaly, we want to know in which factory the data was produced, by what process, and so on, in order to trace the source of the problem. This is complicated to do in practice because algorithms change regularly, in line with updates.” To guarantee cybersecurity, Thales Alenia Space has proposed to the ESA that, in parallel with data protection, a process be put in place to seal the software that is regularly installed on the platform by users to prevent it from being intentionally altered or infected by malicious acts. Sealed software would thus become a black box and no one would be able to modify the algorithm inside it.

In terms of sources, DestinE's direct legacy is the history of Earth observation data from ESA, in particular data from Copernicus missions acquired since the early 2010s. With meteorological and climatological data, we can go back 30 years. EUMETSAT has been storing this data since the early 1990s and uses it to produce climatological trend models. Other complementary Earth observation data from commercial satellites (purchased by environmental agencies or ESA) are also available on the platform. A set of cartographic data also serves as a reference, sourced from agencies such as IGN, which collect information such as land registry and road data. Users will also be able to add their own data collections. These may include readings from drone sensors or stratospheric balloons, for example, or information collected by ordinary citizens engaged in participatory science (provided it is validated for integration into the platform's data). The idea is to build a large, verified and continuously updated database that measures the complexity and variety of environmental phenomena on the planet. The Earth is being scanned from every angle! The DestinE programme platform will also provide access to European Data Spaces, which contain a wide variety of information. This includes anonymised maps of diseases, for example, which will help to correlate the impact of a phenomenon with a disease.

What makes DestinE unique is that it brings together disciplines, uses, user profiles and a large volume of reliable archived data based on real measurements that can be used to build bridges. Until recently, a climatologist had nothing to say to an agricultural engineer, who in turn had nothing to say to a meteorologist. With its resolutely holistic approach, the DestinE platform aims to reveal the interconnections. The objective is to provide specific support to decision-makers through modelling to help them make the most informed decisions possible on a sound and reliable scientific basis.

But politicians still need to take the trouble to listen to science. At all levels, in all geographical areas and across all boundaries.