The rapid advancement of Earth System prediction relies on a critical feedback loop: the better understanding of the physical processes, the innovation of predictive algorithms, and their successful integration into real-time operational forecasting models. While scientific research continuously uncovers novel physical mechanisms and develops advanced parameterizations across atmospheric, oceanic, wave, sea-ice, and terrestrial systems, the rapid rise of artificial intelligence (AI) and machine learning (ML) is transforming Earth System prediction. Operational centers worldwide are now actively rethinking their core infrastructures, navigating how to establish an optimal blend of classical physics-based models, pure data-driven architectures, and physics-informed hybrid systems.
Transitioning both cutting-edge physics and novel AI/ML models into a 24/7 operational production environment presents formidable hurdles. Ensuring that new algorithms run stably, satisfy stringent wall-clock limits, respect physical conservation laws, and deliver verifiable skill improvements across coupled domains requires extensive engineering. Operational centers must frequently retune parameterizations, re-engineer data assimilation interfaces, or develop bespoke hybrid coupling strategies. Conversely, operational implementation often exposes critical knowledge gaps—such as physical representation limitations or AI vulnerabilities during rare, extreme events—that demand further fundamental research. Due to the fast-paced nature of operational environments, these vital engineering strategies, diagnostic methodologies, and practical "tricks of the trade" are rarely documented in peer-reviewed literature.
This session provides a dedicated, collaborative platform for physical modelers, AI/ML researchers, software engineers, and operational practitioners to share their implementation experiences, failure modes, and practical solutions. By fostering an open dialogue on operationalizing conventional, hybrid, and data-driven Earth System Models, we aim to accelerate the Research-to-Operations (R2O) and Operations-to-Research (O2R) pathways and strengthen cross-agency collaboration across all scales.