Abstract: Several physically-based models that couple 1D or 2D surface and 3D subsurface flow have recently been developed, but few studies have evaluated the errors directly associated with the different coupling schemes. In this paper we analyze the causes of mass balance error for a conventional and a modified sequential coupling scheme in worst-case scenario simulations of Hortonian runoff generation on a sloping plane catchment. The conventional scheme is noniterative, whereas for the modified scheme the surface-subsurface exchange fluxes are determined via a boundary condition switching procedure that is performed iteratively during resolution of the nonlinear subsurface flow equation. It is shown that the modified scheme generates much lower coupling mass balance errors than the conventional sequential scheme. While both coupling schemes are sensitive to time discretization, the iterative control of infiltration in the modified scheme greatly limits its sensitivity to temporal resolution. Little sensitivity to spatial discretization is observed for both schemes. For the modified scheme the different factors contributing to coupling error are isolated, and the error is observed to be highly correlated to the flood recession duration. More testing, under broader hydrologic contexts and including other coupling schemes, is recommended so that the findings from this first analysis of coupling errors can be extended to other surface water-groundwater models.