Event-based model calibration approaches for selecting representative distributed parameters in semi-urban watersheds

The objective of this study is to propose an event-based calibration approach for selecting representative semi-distributed hydrologic model parameters and to enhance peak flow prediction at multiple sites of a semi-urban catchment. The performance of three multi-site calibration approaches (multi-site simultaneous (MS-S), multi-site average objective function (MS-A) and multi-event multi-site (ME-MS)) and a benchmark at-catchment outlet (OU) calibration method, are compared in this study. Additional insightful contributions include assessing the nature of the spatio-temporal parameter variability among calibration events and developing an advanced event-based calibration approach to identify skillful model parameter-sets. This study used a SWMM5 hydrologic model in the Humber River Watershed located in Southern Ontario, Canada. For MS-S and OU calibration methods, the multi-objective calibration formulation is solved with the Pareto Archived Dynamically Dimensioned Search (PA-DDS) algorithm. For the MS-A and ME-MS methods, the single objective calibration formulation is solved with the Dynamically Dimensioned Search (DDS) algorithm. The results indicate that the MS-A calibration approach achieved better performance than other considered methods. Comparison between optimized model parameter sets showed that the DDS optimization in MS-A approach improved the model performance at multiple sites. The spatial and temporal variability analysis indicates a presence of uncertainty on sensitive parameters and most importantly on peak flow responses in an event-based calibration process. This finding implied the need to evaluate potential model parameters sets with a series of calibration steps as proposed herein. The proposed calibration and optimization formulation successfully identified representative model parameter set, which is more skillful than what is attainable when using simultaneous multi-site (MS-S), multi-event multi-site (MS-ME) or at basin outlet (OU) approach.