Abstract. The aerodynamic roughness length (z0) is an important parameter in the bulk approach for calculating turbulent fluxes and their contribution to ice melt. However, for heavily crevassed tidewater glaciers z0 estimations are rare or only generalized. This study used unmanned aerial vehicles (UAVs) to map inaccessible tidewater glacier front areas. The high-resolution images were used in a structure-from-motion photogrammetry approach to build digital elevation models (DEMs). These DEMs were applied to five different models (split across transect and raster methods) to estimate z0 values of the mapped area. The results point out that the range of z0 values across a glacier is large, with up to three (locally even four) orders of magnitude. The division of the mapped area into sub-grids (50 m x 50 m), each producing one z0 value, best accounts for the high spatial variability of z0 across the glacier. The z0 estimations from the transect method are in general higher (up to one order of magnitude) than the raster method estimations. Furthermore, wind direction (values parallel to the ice flow direction are larger than perpendicular) and the chosen sub-grid size turned out to have a large impact on the z0 values, again presenting a range of up to one order of magnitude each. On average, z0 values between 0.08 m and 0.88 m for a down-glacier wind direction were found. The UAV approach proved to be an ideal tool to provide distributed z0 estimations of crevassed glaciers, which can be incorporated by models to improve the prediction of turbulent heat fluxes and ice melt rates.