Modern natural hazards research increasingly leverages multi-source datasets, including from satellite, radar, in situ observations and relevant socioeconomic information, to improve the prediction of extreme events and to also capture impacts on human and general infrastructure systems. Recent advances in machine learning have demonstrated strong capabilities in modeling complex spatiotemporal dynamics using multimodal data. As result, natural hazards research, as in other fields, is transitioning from traditional approaches to more integrated, data-driven frameworks led by artificial intelligence, with growing efforts to address limitations through hybrid approaches that incorporate physical understanding.

These developments are particularly critical in the context of climate change, where non-stationary conditions and compound extremes require adaptive and different modeling strategies. The proposed session invites contributions that apply machine learning techniques not only to hazard forecasting, but also across the full disaster lifecycle which includes early warning, impact assessment, emergency response, as well as robust resilience planning.

• A - Atmospheric Sciences
• C - Cryosphere
• GC - Global Environmental Change
• H - Hydrology