Accurate Earth system prediction remains challenging during extreme events and high-impact hazards, when conventional data assimilation (DA) methods based on Gaussian and linear assumptions may be inadequate. In practice, these limitations can manifest in biases arising from skewed, bounded, or multimodal probability distributions of geophysical variables, leading to degraded analysis quality during rapidly evolving events. Nonlinear observation operators, complex remote sensing data, and interactions across coupled Earth system components can further complicate the analysis. Addressing these issues requires approaches ranging from advances in existing ensemble methods to more fundamental developments such as non-Gaussian and non-parametric filters and AI/ML-based methods. We welcome contributions spanning theory, algorithm development, and real-data applications, including advanced DA, coupled DA, AI/ML approaches for learning complex relationships or analysis increments, uncertainty quantification, and verification methods beyond root mean square error. Contributions demonstrating improved analyses or forecast skill using real observations are especially encouraged.

• A - Atmospheric Sciences
• H - Hydrology
• NH - Natural Hazards
• P - Planetary Sciences