Abstract
The global digital elevation model (DEM) produced by the TanDEM-X (TerraSAR-X add-on for digital elevation measurements) mission is an interferometric elevation model with unprecedented quality, accuracy, and coverage. It represents an unedited surface model as artifacts inherent to the interferometric synthetic aperture radar acquisition and processing technique are still present. The most prominent artifacts in the DEM are water bodies appearing with a rough surface due to low coherence. Additionally, outliers, voids, and larger data gaps may be present in this dataset. Therefore, DEM editing is crucial for many applications including hydrology or orthorectification of remote sensing data. Depending on the field of application, different techniques of quality enhancement are required. This article provides a comprehensive description of a semi-automatic framework specially developed for generating an edited version of the TanDEM-X dataset by shaping the high-resolution 12 m DEM with focus on water areas, outlier handling, and void filling. The default configuration parameters of the workflow can thereby be adapted interactively for challenging areas where appropriate. A quality assessment of the resulting edited DEM was done by statistical measures, visual methods, as well as by an artifact evaluation.
Highlights
D IGITAL ELEVATION models (DEMs) are numerical representations of the Earth surface
We present a sophisticated edge-preserving approach for DEM smoothing that is controlled by the height error map (HEM), a quality measure of the TanDEM-X DEM generation process
We define potential land areas if the following three thresholds are jointly fulfilled: the digital numbers (DN) of the amplitude images (AMP) layer must be above a threshold of 50 DN to identify brighter land areas, DEM values must be above the threshold of the mean geoid height +2 m as we assume land areas to be higher than the mean sea level plus a defined margin, and HEM values must be below 3 m as the height error values are in general lower for land areas than for noisy water areas
Summary
D IGITAL ELEVATION models (DEMs) are numerical representations of the Earth surface. They are an essential source of topographic information for various applications including hydrology, hydrodynamics, and flood inundation modeling [1]–[5], geomorphology and geomorphometry [6]–[9], as well as forest mapping [10]–[12]. Smooth water bodies appear as rough surfaces in the DEM due to two effects causing high noise: specular reflection leads to low backscatter and temporal decorrelation of the phase induces a loss of coherence. We present a sophisticated edge-preserving approach for DEM smoothing that is controlled by the height error map (HEM), a quality measure of the TanDEM-X DEM generation process. In addition to statistical measures, the latter was supplemented with an assessment of artifacts similar to the approach of Hirt [46]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
