River managers and scientists interested in hyporheic processes need adequate tools for characterizing hyporheic exchange flow (HEF) at local sites where only poor information on subsurface properties are available. This study evaluates a three-dimensional modelling approach, on the basis of detailed surface parameterization and a simplified subsurface structure, for comparison of potential HEF characteristics at three experimental reaches at the channel-unit scale. First, calibration is conducted to determine the best fit-of-heads given the model simplification, then the structure of residuals are used to evaluate the origin of the misfit, and finally, a sensitivity analysis is conducted to identify inter-site differences in HEF. Results show that such an approach can highlight potential magnitude differences in HEF characteristics between reaches. The sensitivity analysis is successful in delineating the small area of exchange that remains under conditions of high groundwater discharge. In this case, however, the calibrated model performs poorly in representing the exchange pattern at the sediment–water interface, thus suggesting that the approach is less adequate for a deterministic simulation of observed heads. The summary statistics are in the range of similar published models, for which the reported indicator is the root mean square error on heads normalized by the head drop over the reach. We recommend, however, that modellers use a more comparable indicator, such as a measure of the residuals normalized by a measure of observed vertical head differences. Overall, when subsurface data are unavailable or sparse, a three-dimensional groundwater model based on high-resolution topographic data combined with a sensitivity analysis appears as a useful tool for a preliminary characterization of HEF. Copyright © 2013 John Wiley & Sons, Ltd.