An Automatic Calibration Framework for EPA-SWMM Parameters Based on NSGA-II and PySWMM A Case Study of Taipei City

The EPA’s Stormwater Management Model (EPA-SWMM) plays a vital role in simulating hydrological processes and has been extensively applied in rainfall-runoff analysis, as well as the planning and design of drainage systems in urban areas. Establishing and analyzing this hydrological model requires determining numerous physical parameter values. However, EPA-SWMM often faces challenges due to inadequate calibration of these parameters. To address this issue, this study integrates the nondominated sorting genetic algorithm II (NSGA-II) with PySWMM to automate parameter calibration. Calibration and validation of the EPA-SWMM model were performed using rainfall and runoff data from a monitoring catchment in Taipei City, Taiwan. The results reveal that calibrated parameters markedly improve EPA-SWMM accuracy compared to manual determination, with relative errors in peak flow generally within 25% and below 5% in certain events. However, it was observed that the accuracy of the outcomes is significantly influenced by baseflow in the watershed, as reflected by substantial relative errors in runoff volume during validation. These findings suggest the potential applicability of the proposed method to similar environmental modelling challenges, provided that baseflow separation and multi-parameter joint optimization are addressed in future work.