A New Urban Waterlogging Simulation Method Based on Multi-Factor Correlation

Abstract

Waterlogging simulation is a key technology for solving urban waterlogging problems. The current waterlogging modeling process is relatively complex and requires high basic data, which is not conducive to rapid modeling and popularization. In this study, we evaluated the correlation between rainfall and waterlogging water using the following factors: terrain, evaporation, infiltration, pipe drainage capacity, and river flood water level. By quantifying the influence value of each factor on rainfall, we established a simplified model for fast calculation of waterlogging depth through input rainfall. Waterlogging data was collected from Guangzhou, China to set up the multi-factor correlation model, and verify the simulation results of the model. After the original rainfall is added/deducted, the added/loss value, the relationship between net rainfall, and maximum water depth is better than that between original rainfall and maximum water depth. Establishing a stable multi-factor correlation model for a waterlogging point requires at least three historical waterlogging event data for parameter calibration by sensitivity analysis. Comparing the simulation of four waterlogging points, the multi-factor correlation model (error = −13%) presented the least error in simulating the maximum water volume, followed by the Mike Urban model (error = −19%), and finally the SWMM model (error = 20%). Furthermore, the multi-factor correlation model and SWMM model required the least calculation time (less than 1 s), followed by the Mike Urban model (About half a minute). By analyzing the waterlogging data of Guangzhou, 42 waterlogging points with modeling conditions were screened out to further validate the multi-factor correlation model. Each waterlogging point was modeled based on the historical field, and the last rainstorm was used for model verification. The mean error of the comparison between the simulated maximum waterlogging and the measured maximum waterlogging was 3%, and the R2 value was 0.718. In summary, the multi-factor correlation model requires fewer basic data, has a simple modeling process and wide applicability, and makes it easy to realize the intelligent parameter adjustment, which is more suitable for the urgent requirements of current urban waterlogging prediction. The model results may prove accurate and provide scientific decision support for the prevention and control of urban waterlogging.