Generative Artificial Intelligence for Earth System

Generative Artificial Intelligence (GenAI) is rapidly advancing, offering novel ways to exploit multi-disciplinary data and generate new knowledge for science. In Earth System Science (ESS), GenAI is emerging as a transformative technology, enabling a paradigm shift in understanding, predicting, and managing complex socio-environmental systems by cross-using diverse yet fragmented data sources (satellite and in-situ observations, models, experiments, texts).GenAI4Earth will go beyond the state of the art by designing, deploying, and operating trustworthy, reusable GenAI services within the EOSC ecosystem, advancing discovery on Earth–climate–environment–life interactions in co-design with user communities and research infrastructures at national and European levels. Aligned with GenAI4EU and Apply AI initiatives, the project builds on FAIR data, models, and workflows, integrating them into EOSC (AI4EOSC, EOSC Nodes such as Data Terra and NFDI) to foster standards, best practices, and confidence in AI-enabled dataspaces and foundation models.Concretely, GenAI4Earth will:•Develop GenAI tools to enhance FAIRness, machine-actionability, AI-readiness, and provenance of ESS data and services;•Implement AI-powered interfaces for seamless discovery, access, and cross-domain integration;•Demonstrate pilots in urban resilience, agro-environmental monitoring, and seismology through topical AI foundation models;•Promote responsible AI with reproducibility, explainability, transparency, and frugal computing;•Build capacity via training and engagement to ensure broad uptake of GenAI-enabled workflows.The consortium mobilises expertise across AI, data science, Earth systems, computing infrastructures, and ethics to co-design interoperable, reproducible, and impactful services.