A comparative test of photogrammetrically sampled digital elevation models

Abstract

A working group within the ISPRS has conducted an international comparative test of digital elevation models (DEM) based on a resolution from the ISPRS Congress in Hamburg. The object of the test was to study the relations among methods for data acquisition, interpolation, resulting accuracy and type of terrain. Six test areas were selected, having topographic structures varying from smooth rolling farmland to very steep mountainous terrain with forested valleys. Fifteen organisations have produced DEM's from aerial photographs in scales 1:4,000–1:30,000. The DEM's have then been used for the derivation of elevations in a set of check points, the location of which were unknown when the DEM was measured in the stereo-instrument. The elevations of these check points were then compared with their “true” values. The “true” errors found in the check points are presented and analysed. After elimination of blunders the remaining errors are composed of systematic and random parts. The systematic parts can originate from the reconstruction of the stereomodel, from the interpolation of the DEM, and from effects of the vegetation height. The size and distribution of the “true” errors are presented in Diagrammes and Illustrations. Errors of functions of the DEM (e.g. slope and curvature) are also studied.