EGU24-5142, updated on 08 Mar 2024
EGU General Assembly 2024
© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

Four nationwide Digital Surface Models from airborne historical stereo-images

Christian Ginzler, Livia Piermattei, Mauro Marty, and Lars T. Waser
Christian Ginzler et al.
  • Federal Institute for Forest, Snow and Landscape Research WSL, Remote Sensing, Birmensdorf, Switzerland (

Historical aerial images, captured by film cameras in the previous century, have emerged as valuable resources for quantifying Earth's surface and landscape changes over time. In the post-war period, historical aerial images were often acquired to create topographic maps, resulting in the acquisition of large-scale aerial photographs with stereo coverage. Using photogrammetric techniques on stereo-images enables extracting 3D information to reconstruct Digital Surface Models (DSMs), and orthoimages.

This study presents a highly automated photogrammetric approach for generating nationwide DSMs for Switzerland at 1 m resolution using aerial stereo-images acquired between 1979 and 2006. The 8-bit scanned images, with known exterior and interior orientation, were processed using BAE Systems' SocetSet (v5.6.0) with the "Next-Generation Automatic Terrain Extraction" (NGATE) package for DSM generation. The primary objective of the study is to derive four nationwide DSMs for the epochs 1979-1985, 1985-1991, 1991-1998, and 1998-2006. The study assesses DSM quality in terms of vertical accuracy and completeness of image matching across different land cover types, with a focus on forest dynamics and management research.

The elevation accuracy of the generated DSMs was assessed using two reference datasets. Firstly, the elevation differences between a nationwide reference Digital Terrain Model (DTM - swissAlti3d 2017 by Swisstopo) and the generated DSMs were calculated on points classified as "sealed surface". Secondly, elevation values of the DSMs were compared to approximately 500 independent geodetic points distributed across the country. Six study areas were chosen to assess completeness, and it was calculated as the percentage of successfully matched points to the potential total number of matched points within a predefined area. This assessment was conducted for six land cover classes based on the land cover/land-use statistics dataset from the Federal Office of Statistics.

Across the entire country, the median elevation accuracy of the DSMs on sealed points ranges between 0.28 to 0.53 m, with a Normalized Median Absolute Deviation (NMAD) of around 1 m (maximum 1.41 m) and an RMSE of a maximum of 3.90 m. The elevation differences between geodetic points and DSMs show higher accuracy, with a median value of a maximum of 0.05 m and an NMAD smaller than 1 m. Completeness results reveal mean completeness between 64 % to 98 % for the classes "glacial and perpetual snow" and "sealed surfaces," respectively and 93 % specifically for the “closed forest” class.

This work demonstrates the feasibility of generating accurate DSM time series (spanning four epochs) from historical scanned images for the entire Switzerland in a highly automated manner. The resulting DSMs will be available upon publication, providing an excellent opportunity to detect major surface changes, such as forest dynamics.

How to cite: Ginzler, C., Piermattei, L., Marty, M., and Waser, L. T.: Four nationwide Digital Surface Models from airborne historical stereo-images, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-5142,, 2024.