CEE Ph.D. Defense Announcement: Deterministic and Generative Machine Learning Approaches for Precipitation Estimation and Tropical Cyclone Detection

Abstract: Machine learning (ML) has become a powerful tool in hydrology and hydrometeorology, enabling innovative solutions to complex problems. Accurate precipitation estimates and natural disaster detections are crucial given the impact of extreme events. This dissertation explores deterministic and generative ML approaches with satellite data to address these challenges. Specifically, ML is applied to quantify precipitation rates at multiple temporal scales and to detect and segment tropical cyclones. The resulting outputs can potentially be used for data assimilation purposes thanks to their high spatial and temporal resolutions. This dissertation also includes a study about the extreme precipitation in the Western U.S. during the 2022–2023 winter.