Technical Note on Planimetric Misregistration Assessment

This study aims to compare two European DEMS: EU-DEM v1.1 and Copernicus DEM EEA-10 release 2021_1. As these two DEMs have different CRS and sampling distances, they should be resampled in order to be overlapped correctly. These reprojections offer the opportunity to assess different sampling methods (nearest neighbour, bi-linear and bicubic) as well as their effect on DEM difference statistics.This comparison exercise (study 1) is performed on 38 European countries, selecting for each country a tile of approximately 10 km x 10 km. These tiles have been chosen in a grid predefined by the DEMIX (DEM Intercomparison eXercise) group.A first global differences study aims to compute the difference between resampled instances of EU-DEM and EEA-10. A particular attention is given to the resampling methods and their impact on the difference statistics. At the end of study 1, the bilinear resampling method appears to be the best. This bi-linear interpolation having been chosen, the quality of DEM superimposition (planimetric accuracy) is measured performing a disparity analysis in study 2. This method is used in photogrammetry by searching for each point of Copernicus DEM EEA-10 a homologous point in EU-DEM. These pairings produce a field of error vectors which may depend on the land use / land cover but also on the source of the elevation values. Nominally, Copernicus DEM EEA-10 and EU-DEM have respectively been generated from TanDEM-X and SRTM data. However, most recent versions of these DEMs show areas filled with other DEMs. These other sources have an impact on statistics and on the quality of the planimetric superimposition of the two DEMs.