Diagnosing structural deficiencies of a hydrological model by time-varying parameters

Abstract

Time-varying parameters, which are often expressed in functional forms, have been widely used to improve hydrological modeling. Here, we assume that the time-varying parameters may be caused by deficiencies of the model structures. To justify this point, the SIMHYD model was selected as the base model, and the ensemble Kalman filter (EnKF) was used to identify the temporal variation of parameters. A synthetic experiment was designed to prove that model parameters can partly compensate for the structure deficiencies, while the parameters’ time-varying is unavoidable when the structure error has been involved. Two real case studies in the Tongtianhe and Xunhe basins were then conducted to explore the modifications of model structures based on the implications of time-varying parameters. The results are summarized as follows. (1) Model structural deficiencies can be diagnosed by testing whether parameters are time-varying. (2) Guided by the identified time-varying parameters, the energy-based snow module was added and the interception module was removed in the Tongtianhe basin, and the infiltration was modified to follow a power function of the soil moisture in the Xunhe basin. As a result, the model performance has been improved, and the parameter variations have been diminished. (3) The modifications of the SIMHYD model structure outperformed the models with time-varying parameters in functional forms. This study provides insights into the compensatory relationship between model structural deficiencies and parameters, and proposes a method to improve model structures by time-varying parameters.