Weather and climate models continue to suffer from systematic biases due to missing or inadequately represented processes. Diagnosing and mitigating these biases remains critical for improving predictive skill. At the same time, emerging AI-based models offer new opportunities and challenges for representing Earth system processes. 

This session focuses on understanding and reducing biases across physics-based, AI-based, and hybrid models, with an emphasis on maintaining physical realism. We welcome studies that: 1) identify and diagnose biases in mean state, variability, and key processes, including assessments of physical behavior and consistency in AI-based or hybrid models; 2) use observations to constrain parameterized or learned representations; 3) evaluate both traditional physical parameterizations and ML-based parameterization approaches. 

Contributions from testbeds, model-intercomparison efforts, and operational systems are encouraged. The session aims to bring together researchers with expertise in processes who can help advance both physics- and AI-based modeling systems in a physically grounded manner.

• H - Hydrology
• NH - Natural Hazards
• OS - Ocean Sciences