Validation and Error Correction of Space-derived Digital Elevation Models Using ERS Satellite Altimeter Data

Abstract

The series of satellite missions flown over the past decade, together with the recently flown Shuttle Radar Topographic Mission (SRTM), has led to the creation of large databases of information, which are currently being transformed, into detailed Digital Elevation Models (DEMs) for relatively small areas, using both stereo and interferometric techniques. Whilst the high frequency information content means that these techniques can provide extremely detailed models of the topographic surface, distortions and vertical shifts can occur due to various combinations of instrument limitations, orbit mismodelling, shadowing and layover problems, incoherence effects, and the use of incorrect ground control points. Independent validation and assessment of these DEMs can be problematic over remote areas where the availability of public domain DEM data is extremely limited. Mosaicing of adjacent DEMs to create larger-scale coverage can also be complicated by the presence of distortions in the DEM data, even where the same technique has been used to generate the input models. This paper presents the results of a case study using land altimeter derived heights to assess and error correct stereo SAR and InSAR DEM data, as a first step towards creating an operational facility for validation of InSAR, stereo SAR and SRTM derived DEMs. The input DEM has been selected over varying terrain types to explore the potential and limitations of this technique. It is shown that a rule-based expert system can be used to both optimize the calculation of range to surface for different echo shapes, and to select the 'best' echoes for comparison. This allows the generation of an altimeter validation dataset of sufficient accuracy to enable the detailed assessment of stereo SAR and InSAR DEM data, over a variety of terrain types. The comparison between the Altimeter and DEM height datasets shows that both datasets agree at the 8-9m level. This level of agreement is obtained when comparison data over extreme terrain is excluded and the Altimeter dataset is optimized. This paper investigates the potential of altimetry for determining InSAR DEM errors.