Applicability of Data Acquisition Characteristics to the Identification of Local Artefacts in Global Digital Elevation Models: Comparison of the Copernicus and TanDEM-X DEMs

Remote Sensing

David Moravec¹  Jiří Prošek¹  Jana Marešová¹  Vítězslav Moudrý¹  Kateřina Gdulová¹  Lukáš Gábor¹  Petra Pracná¹  Vojtěch Barták¹ 

[1] Department of Spatial Sciences, Faculty of Environmental Sciences, Czech University of Life Sciences Prague, Kamýcká 129, 16500 Praha-Suchdol, Czech Republic;

关键词: artefact;    Copernicus DEM;    digital elevation model;    DEM;    digital surface model;    DSM;   

DOI  :  10.3390/rs13193931

来源: DOAJ

【 摘 要 】

Several global digital elevation models (DEMs) have been developed in the last two decades. The most recent addition to the family of global DEMs is the TanDEM-X DEM. The original version of the TanDEM-X DEM is, however, a nonedited product (i.e., it contains local artefacts such as voids, spikes, and holes). Therefore, subsequent identification of local artefacts and their editing is necessary. In this study, we evaluated the accuracy of the original TanDEM-X DEM and its improved edited version, the Copernicus DEM, in three major European mountain ranges (the Alps, the Carpathians, and the Pyrenees) using a digital surface model derived from airborne laser scanning data as a reference. In addition, to evaluate the applicability of data acquisition characteristics (coverage map, consistency mask, and height error map) and terrain characteristics (slope, aspect, altitude) to the localization of problematic sites, we modeled their associations with the TanDEM-X DEM error. We revealed local occurrences of large errors in the TanDEM-X DEM that were typically found on steep ridges or in canyons, which were largely corrected in the Copernicus DEM. The editing procedure used for the Copernicus DEM construction was evidently successful as the RMSE for the TanDEM-X and Copernicus DEMs at the 90 m resolution improved from 45 m to 12 m, from 16 m to 6 m, and from 24 m to 9 m for the Alps, the Pyrenees, and the Carpathians, respectively. The Copernicus DEM at the 30 m resolution performed similarly well. The boosted regression trees showed that acquisition characteristics provided as auxiliary data are useful for locating problematic sites and explained 28–50% of deviance of the absolute vertical error. The absolute vertical error was strongly related to the height error map. Finally, up to 26% of cells in the Copernicus DEM were filled using DEMs from different time periods and, hence, users performing multitemporal analysis or requiring data from a specific time period in the mountain environment should be wary when using TanDEM-X and its derivations. We suggest that when filling problematic sites using alternative DEMs, more attention should be paid to the period of their collection to minimize the temporal displacement in the final products.

【 授权许可】

Unknown