A time-varying parameter estimation approach using split-sample calibration based on dynamic programming

Abstract. Although the parameters of hydrological models are usually regarded as constant, temporal variations can occur in a changing environment. Thus, effectively estimating time-varying becomes significant challenge. Two methods, including split-sample calibration (SSC) and data assimilation, have been used to estimate parameters. However, SSC is unable consider parameter continuity, while assimilation assumes vary at every time step. This study proposed new method that combines (1) basic concept calibration, whereby assumed be stable for one sub-period, (2) continuity assumption; i.e. differences between consecutive steps small. Dynamic programming then determine optimal trajectory by considering two objective functions: maximization simulation accuracy continuity. The efficiency evaluated synthetic experiments, with simple 2-parameter monthly model second using more complex 15-parameter daily model. results show superior alone outperforms ensemble Kalman filter if proper sub-period length used. An application Wuding River basin indicates soil water capacity varies before after 1972, which interpreted according land use cover changes. A further Xun shows generally stationary on an annual scale but exhibit changes over seasonal scales. These demonstrate effective tool identifying