Capturing the level of progress in vertical accuracy achieved by ASTER GDEM since the beginning: Turkish and Nigerian examples

Abstract

There is little information on the improvement achieved across all versions of ASTER GDEM in terms of vertical accuracy. This study assessed the vertical accuracies of all three versions in Turkey and Nigeria based on the mean error, root mean square error, mean absolute error, average relative error, residual dispersal patterns and combined error performance rankings. Although the improvements from GDEM v1 onward were obvious, they were rather marginal from GDEM v2 to GDEM v3 when five elevation classes were considered from coastal to high mountain. In the coastal, lowland and upland classes in Turkey, and lowland and upland classes in Nigeria, version 2 was superior to version 1. Compared to SRTM version 3.0, the last version produced relatively similar results, but no version outperformed it at any of the test sites. The same was also true in Nigeria. These results showed that a comprehensive vertical accuracy improvement was only partially realized through the successive development phases of the ASTER GDEM.