| 摘要: | ;This research proposal aims to advance the scientific understanding and prediction capability of tropical cyclones (TCs) by integrating multi-sensor observations, cloud microphysics, aerosol–cloud interactions, and next-generation physics-informed artificial intelligence models. The primary objective is to develop a comprehensive, basin-resolved climatology of TC microphysical structures, such as raindrop size distributions, hydrometeor pathways, supercooled liquid water, and graupel processes, and quantify how aerosol regimes influence warm-rain and ice-phase evolution, rapid intensification (RI), rapid weakening (RW), and landfall transitions. These physical insights will be seamlessly coupled with physics-informed AI/ML frameworks to produce a new generation of TC forecasting tools capable of enhancing RI prediction, landfall intensity estimates, and high-resolution quantitative precipitation forecasting. The societal impact of this research is substantial. More accurate RI and rainfall predictions will strengthen early warning systems, reduce life-threatening flood risks, and support better decision-making for evacuations, emergency response, and disaster preparedness across Taiwan and the broader Indo-Pacific. Communities vulnerable to severe storms will benefit from earlier, more reliable hazard information, leading to reduced casualties and property losses. Economically, improved TC forecasting supports the protection of critical infrastructure such as energy facilities, transportation networks, agriculture, water-resource systems, and coastal industries. Enhanced prediction of extreme rainfall and landfall severity will enable more effective planning and mitigation strategies, lowering economic losses associated with typhoons and enhancing resilience in sectors vital to national development. Academically, this project will deliver significant scientific contributions, including a global TC microphysics–aerosol atlas, new microphysical and RSD parameterizations, and advanced physics-informed AI methodologies. The resulting datasets, models, and tools will serve as valuable resources for the atmospheric science and AI communities, fostering interdisciplinary collaboration. |
