Distributed hydrological models rely on a spatial discretization composed of homogeneous units representing different areas within the catchment. Hydrological Response Units (HRUs) typically form the basis of such a discretization. HRUs are generally obtained by intersecting raster or vector layers of land uses, soil types, geology and sub-catchments. Polylines maps representing ditches and river drainage networks can also be used. However this overlapping may result in a mesh with numerical and topological problems not highly representative of the terrain. Thus, a pre-processing is needed to improve the mesh in order to avoid negative effects on the performance of the hydrological model. This paper proposes computer-assisted mesh generation tools to obtain a more regular and physically meaningful mesh of HRUs suitable for hydrologic modeling. We combined existing tools with newly developed scripts implemented in GRASS GIS. The developed scripts address the following problems: (1) high heterogeneity in Digital Elevation Model derived properties within the HRUs, (2) correction of concave polygons or polygons with holes inside, (3) segmentation of very large polygons, and (4) bad estimations of units’ perimeter and distances among them. The improvement process was applied and tested using two small catchments in France. The improvement of the spatial discretization was further assessed by comparing the representation and arrangement of overland flow paths in the original and improved meshes. Overall, a more realistic physical representation was obtained with the improved meshes, which should enhance the computation of surface and sub-surface flows in a hydrologic model.