Nonlinear Runoff Modeling: Parameter Identification

A nonlinear functional rainfall‐runoff model is applied to an urban watershed (Curotte‐Papineau, Montréal) and the results are compared with those from the ILLUDAS model. Simulations are performed using a 5 minute time interval in order to better define the characteristics of the hydrographs. Particular emphasis is placed on the identification of the most appropriate set of model parameters, namely the memory of the catchment (U) and the order of expansion of the first‐ and second‐order kernels M1 and M2. Parameter identification based on calibration results alone can lead to unpredictable verification (prediction) results; while calibration results improve with increasing values of M1 and M2, verification results do not exhibit the same consistency. Consequently, a two‐way calibration‐verification analysis is recommended for obtaining the best set of model parameters. Further, results of the nonlinear functional runoff model compare favorably with those of the ILLUDAS model.