Accuracy Assessment of DEMs in Different Topographic Complexity Based on an Optimum Number of GCP Formulation and Error Propagation Analysis

Abstract

One of the main concerns during digital elevation model (DEM) evaluation is the number of ground control points (GCP). Accordingly, in this paper, a new method is proposed for calculating the appropriate number of GCPs for DEM evaluation based on a confidence interval (CI) of the root-mean-square error (RMSE). This method was employed to determine the CI of the estimated vertical accuracy of Advanced Land Observing Satellite (ALOS) World 3D-30m (AW3D30), Shuttle Radar Topography Mission (SRTM), and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Digital Elevation Model (GDEM) Free 30 m resolution global DEMs in mountainous, hilly, flat, and urban regions of two study areas. To provide a more reliable estimation of errors, robust statistical methods, including median, normalized median absolute deviation (NMAD), and Huber’s μ and σ were also investigated. Furthermore, a new formulation was developed to analyze the propagation of the errors in the slope and aspect products of DEM. The results showed that, to evaluate the accuracy of AW3D30, ASTER GDEM, and SRTM with a CI of ±1 m and the probability of 99%, in the study area, a minimum number of 2,110, 1,483 and 750 GCPs are required, respectively. The results also showed that, in the flat, hilly, and mountainous study areas, AW3D30 was the most accurate DEM. However, SRTM fit better to the urban study area. Finally, the results of the error propagation analysis illustrated that the slope and aspect errors bore a striking relation to the surface gradient.