Simulation of baseflow accounting for the effect of bank storage and its implication in baseflow separation

Most baseflow separation methods for measured streamflow discharge series are based on linear and non-linear solutions of the Dupuit–Boussinesq stream–aquifer model and do not take diverse geological, hydrological, and morphological factors into account. Therefore, a key issue in baseflow separation and drought flow estimation is to reveal baseflow variations related to these factors. This study analyzed the baseflow generated from the return of bank storage to the stream using a numerical groundwater model. Under different stream–aquifer hydrologic conditions in terms of river flood-wave shapes, hydraulic conductivities, stream–aquifer interconnection, recharge and evapotranspiration, and regional hydraulic gradients, baseflow was simulated in order to investigate the non-linearity in the baseflow recession process and to evaluate the baseflow separation methods for drought flow analysis. A comparison between flood recession derived from the Boussinesq equation and that from numerical models indicated that a linear aquifer system does not hold for bank storage effects. Analyses of numerically simulated baseflow discharge also demonstrate that values of power indices of the widely used storage–discharge functions which are derived from the single-valued power law are not constant but depend on hydraulic conductivities, stream–aquifer interconnection, and other surface hydrologic conditions. Bank storage due to the stream flood-stage fluctuation reduces groundwater discharge into the channel considerably in the flood stage rising period. Thus, neglecting the influence of bank storage on baseflow would result in a large error in baseflow separation for non-ideal stream–aquifer systems.