Accuracy Assessment of Normalized Digital Surface Models from Aerial Images Regarding Tree Height Determination in Saxony, Germany

The data acquisition from normalized digital surface models (nDSM), based on regularly conducted aerial image surveys, represents an efficient opportunity for forest enterprises to ensure topicality as well as quality for tree height information. Two of such image data sets—L60 (60% along-track/30% across-track) and L80 (80/40%)—were each used to create an nDSM for a 23.5 km2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$^{2}$$\end{document} large study area in Saxony. The objective of the present study was to assess the accuracy of both nDSMs and to examine the influence of image overlap, tree type and terrain slope on the results of this accuracy assessment. Point clouds were initially extracted from the corresponding images using semi-global matching and were eventually used to create a digital surface model. An nDSM was produced by subtracting a digital terrain model to get absolute tree heights. The comparison between this nDSM and terrestrial measurements (n = 51) resulted in root mean square errors between 2.1 m and 2.2 m. Although opposing median values were found for the nDSMs based on L60 (−0.3 m) and L80 (0.5 m), no significance was found between the two models (Wilcoxon test, p = 0.16). However, large overlapping areas and lower terrain slopes have a positive effect on the completeness of the nDSM (L60: 92.9%, L80: 99.2%). Although the small number of measurements does not allow for more than the identification of a tendency, these results show the potential of image-based nDSMs for the regular update of tree height information in Saxony and beyond.