Two-dimensional, ®nite element hydraulic models have been developed to simulate river ¯ood ¯ows at high spatial and temporal resolutions over river reach lengths of 1±60 km. Such models have been shown to be capable of simulating bulk ¯ow; however, model redesign to predict spatially distributed hydraulic variables has been constrained by lack of suitable topographic and hydraulic data. Here we begin this development process using a hypothetical river channel/¯oodplain domain where the topographic surface is parameterized using scaling information derived from a fractal analysis of a real ¯oodplain DTM. This is used to test the relative eect of the boundary friction calibration, numerical model grid resolution, topography sampling error and ¯oodplain relative height on model predictions of out¯ow discharge, inundation extent and local hydraulic variables. Simulations indicate that model calibration is the dominant factor aecting the above three quantities. Moreover, model sensitivity to spatially uniform change is shown to be simple for bulk ¯ow and inundation extent but spatially complex for local hydraulics. The study has a number of implications for model calibration and set-up procedures, as well as indicating the need to develop a new suite of analysis techniques for this class of model.