Impacts of parameter uncertainty on baseflow separation by a two‐parameter recursive digital filter

Abstract

Accurate baseflow quantification is an important step in the survey and evaluation of water resources, non‐point source pollution quantification, and optimization of water resource allocation. We selected a typical agricultural watershed with high precipitation rates in eastern China as the study area and investigated the impact of the combination of a recession constant (on an annual, monthly, or daily scale) and the maximum baseflow index (BFImax; empirical value vs. optimal value) on the accuracy and reliability of baseflow separated using a two‐parameter recursive digital filter (TPRDF). It was observed that the accuracy and reliability of the TPRDF‐separated baseflow varied with the time scale of the recession constant. Specifically, they decreased in the following order: baseflow estimated using daily recession constants (EBSD) (NSE = 0.797, R2 = 0.819) > baseflow estimated using monthly recession constants (EBSM; NSE = 0.779, R2 = 0.801) > baseflow estimated using annual recession constants (EBSY; NSE = 0.666, R2 = 0.675), where ‘NSE’ is the Nash–Sutcliffe efficiency coefficient. The empirical value setting of BFImax was another important reason for the uncertainties in the TPRDF‐separated baseflow. The NSE values of EBSY, EBSM, and EBSD increased from 0.666, 0.779, and 0.797–0.785, 0.859, and 0.876, respectively, after replacing the empirical values with numerically optimized values. The combination of BFImax with recession constants on various time scales significantly affected the annual total TPRDF‐separated baseflow. It also affected the overall variation characteristics of the baseflow time series, composition of different magnitude (frequency) components, and sensitivity of baseflow time series to rainfall events.