ABSTRACT Urban drainage systems play an essential role in stormwater management. However, conventional evaluation methods typically focused on individual performance indicators and failed to account for the compound effects of multiple environmental factors. To address this gap, an integrated probabilistic‐hydrodynamic framework was proposed, combining dynamic hydraulic simulation (SWMM) with Bayesian network (BN) inference. A total of 150 scenario simulations were conducted, covering six rainfall return periods, five imperviousness levels, and five downstream water levels. Flooding volumes were classified into three severity levels using the 25th and 75th percentiles. Three complementary metrics were adopted: Risk Ratio (RR), Information Gain (IG), and Relative Contribution (RC). The framework was applied to a typical urban area in Yuanjiang City. The results indicated that rainfall was the most dominant factor (IG = 0.45 bits, 71%), followed by downstream water level (IG = 0.07 bits, 11%) and imperviousness (IG = 0.11 bits, 18%). Extreme rainfall events (50–100 years) exhibited the highest RR (2.12) and contributed 35% of the excess flood risk, whereas high water level contributed 32% and high imperviousness contributed 22%. Low‐severity floods were associated with low rainfall intensities and low water levels, while high‐severity floods required the simultaneous occurrence of extreme rainfall, high imperviousness, and elevated downstream water levels. The proposed framework facilitates a transition from single‐factor to multi‐factor assessment and provides a scientific basis for prioritizing drainage system improvements.
Song et al. (Mon,) studied this question.