A central and persistent problem concerning urban drainage models is that their performance and utility are strongly influenced by underdetermined parameters. Uncertainties in parameters and data affect the confidence in the modelled results. A well-calibrated model can generate useful information from observations to support informed decisions. However, traditional manual calibration is highly inefficient and does not ensure an accurate model. Automatic calibration offers a promising alternative, ideally a time-efficient approach to parameter estimation with no domain knowledge required. In this study, we propose an integration of PySWMM (a Python wrapper to EPA-SWMM) and Pymoo (a Python package for multi-objective optimisation) for automatic calibration. This integration is applied to the calibration and parametrisation of the groundwater and aquifer modules for an existing SWM model of a medium-sized catchment in Switzerland. A multi-objective evolutionary algorithm, NSGA-II (available in Pymoo), is used to calibrate twelve parameters in the groundwater modelling structure in SWMM using four objective functions (the Nash Sutcliffe Efficiency for two flow measurements in the main collector and the mean square error in two groundwater level measurements). Results show the performance of the automatically calibrated model highly depends on user-dependant choices such as the selection of the parameters to calibrate, their ranges, and the parameters of the optimisation algorithm. Nevertheless, this study demonstrates that PySWMM and Pymoo integration is a standardised approach to automatic calibration. Future work could explore how this integration could be used with alternative optimisation algorithms and to calibrate other parameters in SWMM, as well as allow an explorative approach for improving the automatic calibration of hydraulic-hydrological models.
Click here to watch recorded presentation on YouTube.