FUTURE EARTH OBSERVATION ARCHITECTURE FOR AI (ARCH-AI): NUMERICAL WEATHER PREDICTION (NWP) AND NUMERICAL OCEAN PREDICTION (NOP) - EXPRO PLUS

ARCH-AI addresses the rapid transition of numerical weather prediction (NWP), ocean prediction (NOP), and climate forecasting toward AI/ML-driven approaches. These methods enable faster forecasts, larger ensembles, and new probabilistic capabilities, but they fundamentally change what is required from Earth Observation.The key shift is from EO as an observational system to EO as an integralcomponent of predictive systems. This means observations are no longer just inputs; they must be optimised for AI training, inference, and continuous model evolution.Current European EO systems are strong but not designed for this paradigm. The ARCH-AI studies examines how the entire EO system-of-systems (space, ground, data, interfaces) must evolve to support AI-native and hybrid forecasting,while maintaining continuity, reliability, and European strategic autonomy.The ARCH-AI studies will deliver a set of coherent Earth Observation architecture options (3-5), spanning AI-native, hybrid, and observation-centric designs, each characterised in terms of performance, cost, risk, and dependencies. These are grounded in a reformulation of EO requirements derived directly from AI/ML needs, including data characteristics for training and inference, and trade-offs between temporal sampling, resolution, and long-termstability.The studies will define transition pathways from current systems, and translate architectures into programme and investment scenarios, including mission portfolios and options for maintaining European strategic autonomy. The study outputs will be used tohelp align future EO investments with AI-driven forecasting needs, ensuring that Europe's EO system evolves into an integrated, AI-ready prediction infrastructure that remains operationally relevant and competitive.