Abstract Using decadal GCM simulations, the effects of a SVAT (Surface-Vegetation-Atmosphere-Transfer) and a “bucket” land surface parameterization on simulated hydrological persistence are contrasted. The bucket model promotes persistence, as measured by the one-month-lagged autocorrelations of precipitation and evaporation, relative to the SVAT model, particularly in the Tropics. An additional simulation shows that despite the many complexities incorporated into the SVAT model's evaporation formulation, it is the canopy interception reservoir alone that explains most of the difference. Water that evaporates from this reservoir bypasses bare soil and stomatal resistances. The interception reservoir thus acts as a short circuit in the soil and vegetation resistance network that decreases hydrological persistence by reducing the residence time of surface moisture.