Space-borne digital elevation models (DEM) are considered as important proxy for canopy surface height and its changes in forests. Interferometric TanDEM-X DEMs were assessed regarding their accuracy in forests of Germany and Estonia. The interferometric synthetic aperture radar (InSAR) data for the new global TanDEM-X DEM 2020 coverage were acquired between 2017 and 2020. Each data acquisition was processed using the delta-phase approach for phase unwrapping and comprise an absolute height calibration. The results of the individual InSAR heights confirmed a substantial bias in forests. This was indicated by a mean error (ME) between – 5.74 and – 6.14 m associated with a root-mean-squared-error (RMSE) between 6.99 m and 7.40 m using airborne light detection and ranging (LiDAR) data as a reference. The bias was attributed to signal penetration, which was attempted to be compensated. The ME and RMSE improved substantially after the compensation to the range of – 0.54 to 0.84 m and 3.55 m to 4.52 m. Higher errors of the penetration depth compensated DEMs compared to the original DEMs were found in non-forested areas. This suggests to use the penetration compensation only in forests. The potential of the DEMs for estimating height changes was further assessed in a case study in Estonia. The canopy height change analysis in Estonia indicated an overall accuracy in terms of RMSE of 4.17 m and ME of – 0.93 m on pixel level comparing TanDEM-X and LiDAR height changes. The accuracy improved substantially at forest stand level to an RMSE of 2.84 m and an ME of – 1.48 m. Selective penetration compensation further improved the height change estimates to an RMSE of 2.14 m and an ME of – 0.83 m. Height loss induced by clearcutting was estimated with an ME of – 0.85 m and an RMSE of 3.3 m. Substantial regrowth resulted in an ME of – 0.46 m and an RMSE of 1.9 m. These results are relevant for exploiting multiple global acquisitions of TanDEM-X, in particular for estimating canopy height and its changes in European forests.
|Number of pages||17|
|Journal||PFG - Journal of Photogrammetry, Remote Sensing and Geoinformation Science|
|Early online date||1 Mar 2023|
|Publication status||Published - Apr 2023|