Project results are expected to contribute to all of the following expected outcomes:

• Integrated AI/ML strategy across Copernicus Services, value chains and workflows;
• Improved quality, timeliness, reliability and resilience of Copernicus data, products and applications;
• Improved time-to-solution and energy-to-solution of Copernicus operational workflows;
• Transformed user experience through enhanced interactivity and on-demand capabilities for Copernicus services;
• Exchange of knowledge, benchmarking and best practices on using AI/ML in the context of Copernicus;
• Enhanced AI-readiness of Copernicus data, in particular open and free high-value labelled Copernicus data sets.

Scope:

The areas of R&I to address the above expected outcomes include:

• AI-supported retrieval algorithms on both passive and active sensing for existing and upcoming Copernicus missions;
• Fast, reliable, consistent, and as much as possible sensor agnostic identification of clouds and shadows in optical sensing;
• Multi-source multi-target AI models for automatic segmentation;
• Physics parameterization and parameter optimization to emulate poorly understood processes and increase the fidelity of numerical models;
• Fault and outlier detection in production and delivery workflows to ensure more robust services;
• Support to automated pre-processing and QA/QC of observations and data to reduce the risk of man-made errors and product deficiencies;
• Data fusion techniques towards added-value products;
• Data compression and mining methods to navigate big data efficiently, as the amount of data is becoming a limiting factor;
• Hybrid observation operator, ensemble data assimilation techniques, error calibration and uncertainty quantification towards improved (re-)analysis and forecast skill;
• Analysis-driven Earth system deep learning models to boost prediction skill and timeliness, including with Digital Twin Earth models. These methods have shown great promises when applied to reanalyses for example;
• Experimenting observation(-only)-driven forecasting to support time-critical service elements, circumventing analysis steps. These approaches could be particularly suited for observation-dense areas from which processes can be inferred from observations alone;
• Exploring the potential of large pre-trained foundation models and transfer learning at scale for Earth system modelling, including with publicly available training datasets from Copernicus;
• Downscaling and super resolution applications building on Copernicus data to refine products in space and time;
• Adaptive workflow optimizations;
• Enhanced interactive interfaces enabling on-demand product and service generation;
• Chatbots that can guide the user across a wide range of information sources within and across Copernicus services for enhanced user support and experience.

Proposals are expected to address as many of the above areas as possible.

During the last decade, artificial intelligence (AI), machine learning, big data volumes and computing capacities have developed at an unprecedented pace, and it is now evident that Copernicus needs to become even more proactive on the digital transition. AI and machine learning offer great opportunities across the Copernicus value chain and workflows to deeply transform its data, products, applications, services and user experience.

However, the scope and speed of developments also generate challenges, in particular regarding the necessary know-how that needs to be established, the software and hardware infrastructure that need to be developed, and the integration of machine learning and conventional tools within production workflows. These challenges need to be addressed within a comparably short period of time to keep up with evolving user requirements and to leverage emerging AI/ML developments. The project is expected to foster game changer and disruptive approaches in particular towards next generation Earth system (re-)analysis and prediction systems as well as foundation models, to promote integrated AI/ML strategies and intensive cooperation and knowledge transfer with and across Entrusted Entities to pave the way into the future of Copernicus. Given the QA/QC requirements on Copernicus products, explainable, trustworthy, open-source and responsible use of AI approaches are of particular interest, as AI mainly operates as a black box. In the context of recent EU policies, a robust framework is required to ensure the same stringent quality, reliability, and verifiability requirements of AI-generated products, as well as transparency and clearly labelled information to users. Benchmarking approaches to quantify the positive impact and improvements of AI/ML methods over time are particularly encouraged.

Collaboration with the EuroGEO initiative and the project(s) funded from the topic HORIZON-CL6-2025-03-GOVERNANCE-09: Delivering Earth Intelligence to accelerate the green and digital transition is encouraged.

The transfer of research results to operations should receive active attention during the project to strengthen the readiness for an operational deployment in the future. Appropriate involvement and/or interaction with the relevant Entrusted Entities of the Copernicus services and Destination Earth, the conditions for making available, for re-using and exploiting the results (including IPR) by the said entities must be addressed during the project implementation.

The possible participation of the JRC may consist in (1) ensuring access to relevant models, tools and datasets of the operational CEMS and CLMS, (2) providing a good understanding of existing operational workflows for CEMS/CLMS and advice regarding the operational feasibility of new developments and (3) testing of new developments/prototypes for CEMS/CLMS in a pre-operational setting.

In this topic, the integration of the gender dimension (sex and gender analysis) in research and innovation content should be addressed only if relevant in relation to the objectives of the research effort.