Abstract: To investigate the patterns of rainfall-flood response and identify their key drivers in urbanized river network plain regions, this study focuses on the Wuchengxiyu area within the Taihu Lake basin. Based on hydrological observations from diverse urban river network units, four rainfall-flood response characteristic indices were established and a comparative analysis was then performed to discern variations in these responses across typical urban river network units, utilizing interpretable machine learning to probe into the multifactorial influences at play. The results reveal that regions characterized by sparse river networks, a high proportion of impervious surfaces, and being situated beyond urban flood control embankment systems experience the highest peak flood water levels for every rainfall intensity, along with the most intense rainfall-flood response processes, and the magnitude, scale, and dynamic characteristics of flood peaks are primarily controlled by rainfall features and pre-event river channel conditions. The impervious surface ratio significantly and positively impacts the rate at which floodwaters rise, while the ratio of the water surface area and the area-length ratio of the mainstream are pivotal in mitigating flood peaks and regulating flood flows. These findings contribute to a deeper scientific understanding of rainfall-flood response characteristics and their underlying mechanisms in plain river network regions, thus laying a solid foundation for flood risk mitigation and river system conservation.