Vertical Accuracy Assessment of the AW3D30 Digital Surface Model Using GNSS Check Points North of Al-Marj, Libya

Abstract

Global digital elevation models are widely used, but their vertical performance varies with terrain and geographic setting. This study assessed the vertical accuracy of the ALOS World 3D–30 m digital surface model (AW3D30) north of Al-Marj, northeastern Libya, using 23 Global Navigation Satellite System (GNSS) checkpoints. Both datasets were referenced to a common coordinate system, and orthometric heights were referenced to the Earth Gravitational Model 1996. Vertical differences, computed as AW3D30 minus GNSS elevation, were summarised using mean error, mean absolute error, root mean square error, and sample standard deviation. Inverse distance weighting (IDW) was used for exploratory visualisation of broad spatial variation. AW3D30 was higher at all checkpoints, with differences ranging from 2.19 to 9.78 m. Mean error and mean absolute error were both 7.04 m, while root mean square error and sample standard deviation were 7.44 and 2.45 m, respectively. Larger differences occurred mainly among the higher sampled elevations, although the dataset could not confirm a continuous elevation difference relationship. The IDW surface suggested lower differences in the western and northwestern sectors, but uneven sampling and the absence of independent validation limited interpretation. Uncorrected AW3D30 elevations should therefore be used cautiously in applications requiring vertical accuracy better than the observed local performance.