Data assimilation is central to improving environmental prediction by combining observations with models to produce accurate initial conditions and forecasts. This session highlights advances across the spectrum of data assimilation, including traditional physics-based methods and emerging machine learning (ML) and artificial intelligence (AI) approaches.

We invite contributions on variational and ensemble-based methods, hybrid techniques, and the assimilation of diverse observations (e.g., satellite, radar, and in situ data). We also welcome work on AI-enhanced data assimilation, including learned observation operators, surrogate models, data-driven error characterization, and the use of deep learning and generative models.

The session will explore challenges in integrating physics-based and AI-driven approaches, including uncertainty quantification, scalability, robustness, interpretability, and generalization. It aims to bring together researchers from academia, operations, and industry to share advances and discuss pathways for next-generation forecasting systems.

• H - Hydrology
• IN - Informatics
• NG - Nonlinear Geophysics
• OS - Ocean Sciences