We have assembled an elevation grid for the Greenland ice sheet using a combination of the best current digital elevation model (DEM) (Bamber and others, 2000a, 2001) and 44 Advanced Very High Resolution Radiometer satellite images acquired in spring 1997. The images are used to quantitatively enhance the representation of surface undulations through photoclinometry. Gridcell spacing of the new DEM is 625 m. To validate the new DEM, we compared profiles extracted from it and the Bamber and others DEM with airborne laser altimetry profiles collected in the 1990s by the Airborne Topographic Mapper (Krabill and others, 1995). The image-enhanced DEM has a greatly improved representation of decameter-relief surface features <15 km in lateral extent, and reduces the mean elevation error in regions having these features by 20–50%. Root-mean-squared errors are typically 7–15m in the Bamber DEM, and 4–10m after image enhancement. However, the photoclinometry process adds some noise. In very smooth portions of the ice sheet where decameter undulationsare absent, the photoclinometry process caused a slight increase in the rms error, from ~1 min the Bamber and others DEM to ∼2.5 min the image-enhanced DEM. The image-enhanced DEM will be useful for inferring accumulation-rate variations over the undulation field, or for improving maps of bedrock elevation through inversion of surface elevation, for example. We briefly explore the preliminary steps of this latter application.