Surface elevation of terrain can be estimated from interferometric synthetic aperture radar (INSAR) observations from satellites. In this paper, we compare a satellite SAR estimate of surface elevation with a digital elevation map (DEM) compiled by the USGS from survey data. To compare the elevation obtained from SAR with a DEM, a new registration method is developed to register the SAR images with the DEM. To convert the DEM into the same format as the SAR data, a phase difference image is generated from the DEM according to SAR imaging geometry. The mean squared error between the simulated phase difference and the measured phase difference is minimised using a downhill simplex minimization algorithm. Registration parameters such as orientation, range and azimuth offsets are obtained in the minimization process. Elevation data generated from the Seasat SAR over an area west of Yellowstone National Park are compared with a DEM using the proposed registration algorithm. The root mean squared (rms) error between the computed elevation and the DEM is found to be 6·0 per cent of the total terrain variations in the image. Layover effects and other error sources are also explored in the paper.