OSA1.4 | Verification for AI- and physics-based weather prediction: new challenges, techniques and observations
Verification for AI- and physics-based weather prediction: new challenges, techniques and observations
Conveners: Estíbaliz Gascón, Bastien François, Sabrina Wahl

This session addresses recent advances and emerging challenges in the verification of numerical weather prediction (NWP), artificial intelligence–based weather prediction (AIWP), and climate modelling systems across a broad range of spatial and temporal scales. Contributions are welcome across the full verification spectrum, from methodological research to operational practice and user-oriented applications.

The scope encompasses established verification approaches for physical NWP models, as well as novel methodologies required for AIWP and hybrid systems, including their extension to subseasonal-to-seasonal and climate applications
• Use and interpretation of new and emerging observational datasets for verification, including non-traditional observations, impact-based data, and applications related to high-impact and user-oriented services such as warnings for hazardous weather.
• Advances in verification approaches tailored to different modelling systems (physical, artificial intelligence-based and hybrid models, and climate models), including suitable metrics, techniques, and effective communication of forecast skill and uncertainty.
• Methodological innovations such as spatial, temporal, and object-based verification, extremes verification, process-based evaluation, and probabilistic methods.
• Verification strategies adapted to high-resolution, convection permitting, ensemble, and variable-resolution modelling systems.
• Verification approaches aimed at supporting decision-making and end-user needs, including sector-specific verification, evaluation of risk-relevant events, and applications bridging weather and climate services.