Global digital elevation models -- fact or fiction?

Satellite altimetry is a familiar tool for monitoring the surface of the world's oceans, but data gathered over land have received comparatively little attention. Altimeter echoes returned from topographically varying surfaces can exhibit complex and disparate behaviour, and extensive reprocessing is necessary to determine, for example, which time point in the returned waveform corresponds to the mean range to surface. This paper uses the unique database of land returns with near-global coverage produced by the geodetic mission phase of ERS-1 to compare altimeterderived heights, obtained using a rulebased expert system, with existing ground-based Global Digital Elevation Models (GDEMs). One result is the identification of common error signatures in the GDEMs; in many areas, correlated height differences are seen between the altimeter data and the GDEMs, often apparent as clear geometrical patterns in difference maps. These errors can be significant in terms of height offset, persist over large areas, and are attributed to serious shortcomings in the GDEM representation of terrain. The ability to identify such erroneous ground truth data using altimetry facilitates the improvement of these global models, allowing these errors to be quantified and, potentially, corrected using altimeter-derived heights.