A coupled surface water and variably-saturated groundwater model is developed for a coastal salt marsh (Nueces Delta, Texas USA). To evaluate the relative impacts of common model simplifications, the new model is compared against a prior salinity transport model that neglects evaporation and groundwater coupling. Both marsh-scale (at coarse grid resolution) and bank-scale (at fine grid resolution) simulations are used to examine the flux mechanisms in the intertidal zone where soil-moisture evaporation and groundwater fluxes contribute to hypersalinity. New numerical methods provide the groundwater–surface water coupling mechanisms necessary for practical marsh-scale simulations using the two-dimensional (2D) Shallow Water Equations for surface water and the density-dependent 3D Richards equation for groundwater. An asynchronous surface–subsurface coupling scheme is shown to significantly reduce the computational cost of marsh-scale simulations. Comparison of marsh-scale and bank-scale results indicates that a coarse model grid can represent the mechanisms of surface/subsurface salinity flux in the intertidal zone, but likely misrepresents the inundation effects of topography smaller than the model grid. Due to relatively weak tidal amplitudes at the study site, surface-water evaporation is a stronger driver of hypersalinity than surface–subsurface exchange induced by soil-moisture evaporation. The combination of evaporation and groundwater exchange creates small hypersalinity “hot spots” that do not appear in the simulations neglecting these processes. However, including these processes has only a marginal impact on comparisons with field data previously collected in the marsh. Both types of simulations have good agreement for surface-water volumetric transport (as evidenced by water surface levels), but the remaining model-field disagreement in salinity appears to be dominated by biases in salinity transport for oscillatory (tidal, wind-driven) fluxes through narrow channels.