Research covers the observations, nowcasting, data assimilation, modeling and post-processing required to forecast weather-related hazards using coupled numerical weather, land surface, ocean and chemistry models, including modeling of floods, landslides, bushfires, air pollution etc. Research will focus on advances in the whole prediction chain needed to forecast the hazards, on prediction at convective-scale (<3km), on coupled modeling and on the use of ensembles to quantify probability and uncertainty. Specific activities will be carried out reviewing the use of existing and new observation sources; comparing new approaches to multi-scale coupled modeling and data assimilation systems, drawing on parallel activities in the Sub-seasonal-To-Seasonal project; developing ensemble perturbations for small scales and hazards; and meeting the product specifications identified by the Communications theme. The research will make use of a catalogue of hazardous weather case studies developed with the Predictability & Processes theme, together with datasets from recent and planned field experiments, re-analyses and re-forecasts, and will demonstrate and evaluate new techniques in Forecast Demonstration Projects.