This study focuses on the short groundwater residence time induced by the development of a dense rooting system in a forested area employed as a buffer zone planned for nitrogen removal. The buffer zone consists of a rectangular area of 0·7 ha sub-irrigated by two ditches and drained by an outflow ditch lying in the centre of the area. The pristine soil (silty clay loam) was characterized by a low hydraulic conductivity, but the agricultural practice first, and then the rooting system increased the permeability of the top soil. This was confirmed by the elevated discharge in the outflow ditch and by the elevated hydraulic conductivity values determined via slug tests. The groundwater residence time was monitored, once the rooting system was completely developed, via Rhodamine-WT injection in the inflow ditch. The breakthrough curves were simulated using a one-dimensional analytical model (CXTFIT) to calculate the residence time. In addition, a three-dimensional flow and transport model (MODFLOW and MT3DMS) was calibrated using piezometric heads, groundwater fluxes and Rhodamine-WT concentrations. The CXTFIT model overestimated the dispersivity and underestimated velocity respect to MT3DMS. This study shows that residence time can decrease respect to the value expected from the hydraulic conductivities of the pristine soils and that one-dimensional analytical models can give a false impression compared with soil parameters respect to three-dimensional flow and transport models. Given the importance of the residence time for biological processes in buffer zones, a greater effort is needed to improve measurement techniques on subsurface flow's heterogeneity. Copyright © 2013 John Wiley & Sons, Ltd.