Urban areas are increasingly vulnerable to severe weather events, including localized intense precipitation, extreme heat, and urban windstorms. Developing effective Early Warning Systems (EWS) that are fit for purpose requires bridging high-resolution physical science, cutting-edge AI methodologies, and actionable co-produced risk communications and policies. This requires understanding both physical forecast performance and how warnings are operationalized to drive effective anticipatory action, particularly in vulnerable urban environments.

This session, organized as part of the WMO Urban-PREDICT initiative, explores recent advances in high-resolution urban weather forecasting, multi-hazard prediction, and the interdisciplinary co-design of early warnings. We welcome contributions covering:
• High-Resolution Physical Modelling: Sub-kilometer and hectometric-scale Numerical Weather Prediction for urban environments and their application to early warning.
• AI & Machine Learning Applications: Standalone AI or physics-informed AI prediction systems, fast emulators, data-driven downscaling, and novel observational datasets (e.g., radar, crowdsourced sensors, urban IoT), including pathways for operational deployment and data accessibility.
• Impact & Hazard Modelling: Translating meteorological variables into actionable hazard metrics for urban infrastructure, energy systems, and public health, considering the consequences for community response.
• Interdisciplinary Co-Production: Frameworks and case studies linking physical hazard forecasts to societal response, policy, and community-level EWS resilience.

This session will feature contributions from ongoing research under the Urban-PREDICT framework, including international intercomparison initiatives (e.g., RUMI high-resolution modelling over Hong Kong, HEAT-X AI heat downscaling) and local case study approaches to early warning and data availability (e.g., Hong Kong, Lusaka, and Montevideo). We encourage submissions across physical meteorology, machine learning, and social sciences.