Hydrological extremes are becoming more frequent and severe under changing climate and land-use conditions, exposing fundamental limitations in both traditional process-based models and purely data-driven approaches. Physics-informed machine learning and emerging generative AI techniques provide a rigorous intermediate paradigm by embedding governing equations and physical constraints within flexible, data-adaptive architectures, thereby enhancing predictive skill, scalability, and robustness under nonstationarity. This session targets civil and water resources engineers and hydrologists concerned with floods, droughts, and compound extremes who seek operational, physics-consistent AI frameworks. Anticipated contributions include advances in physics-informed neural networks, generative models (e.g., diffusion and GAN-based downscaling), and hybrid neural–physical formulations for extreme-event characterization, multiscale uncertainty quantification, and real-time forecasting. We respectfully request co-location with hydrology, natural hazards, and AI/informatics sessions to foster high-level cross-disciplinary exchange among model developers, operational agencies, and engineering practitioners.

• A - Atmospheric Sciences
• GC - Global Environmental Change
• IN - Informatics
• NH - Natural Hazards