Analysis of Urban Flooding Impacts Based on Predicted Precipitation Uncertainty

This study analyzed the impact of rainfall prediction uncertainty on urban flooding, focusing on a heavy rainfall event (August 8, 2022) in the Dorim River Basin in the Seoul metropolitan area. Baseline data, observed by automatic weather station (AWS), was provided by the Korea Meteorological Administration. This study evaluated the rainfall prediction performance of two predictive rainfall datasets (LDAPS and MAPLE) using this data. In addition, the study conducted flood simulations based on estimated manhole overflow volumes. The results showed that the rainfall predicted by LDAPS exhibited an NSE of –0.482 and a PBIAS of 87.692, indicating a significant underestimation. Conversely, MAPLE demonstrated an NSE of 0.668 and a PBIAS of –4.176, suggesting an overestimation, but achieving a quantitatively superior performance. In the flood simulation, LDAPS-based predictions corresponded poorly with the AWS-based results with a 5.2% hit rate, whereas MAPLE achieved a hit rate of 91.9%, along with an additional 0.856 km² of flooded area. This study highlights that the uncertainty in predictive rainfall datasets significantly affects urban flood prediction accuracy, emphasizing the necessity of calibrating predictive rainfall data to improve flood prediction reliability.