Extending the integrated monitoring of deep-seated landslide activity into the past using free and open-source photogrammetry

Deep-seated gravitational slope deformations (DSGSDs) pose serious threats to buildings and infrastructure in mountain regions. The understanding of past movement behavior are essential requirements for enhancing process knowledge and potential mitigation measures. In this context historical aerial imagery provides a unique possibility to assess and reconstruct the deformation history of DSGSDs. This study investigates the feasibility of 3D point clouds derived from historical aerial imagery using free and open-source (FOSS) photogrammetric tools for analyzing the long-term behavior of the Reissenschuh DSGSD in the Schmirn valley (Tyrol, Austria) and assessing related secondary processes as changes in creep velocity, rockfall or debris flows. For the photogrammetric analyses, scanned analogue and digital imagery of six acquisition flights, conducted in 1954, 1971/1973, 2007, 2010, and 2019, have been processed using the FOSS photogrammetric suite MicMac. Further point cloud processing was carried out in CloudCompare. An improved version of the image correlation approach (IMCORR) implemented in SAGA GIS was used for the area-wide assessment of slope deformation. For the georeferencing and scaling an airborne laser scanning (ALS) point cloud of 2008 provided by the Federal State of Tyrol (Austria) was used. In total five photogrammetric 3D point clouds covering the period from 1954 to 2019 were derived and analyzed in terms of displacement, velocity and acceleration. The accuracy assessment with computed Multiscale Model to Model Cloud Comparison (M3C2) distances between photogrammetric 3D point clouds and reference ALS 3D point cloud, showed an overall uncertainty of about ±1.2 m (95% quantile) for all 3D point clouds produced with scanned analogue aerial images (1954, 1971/1973 and 2007), whereas 3D point clouds produced with digital aerial imagery (2010, 2019) showed a distinctly lower uncertainty of about ±0.3 m (95% quantile). Also, digital elevation models (DEM) of difference (DoD) for each epoch were calculated. IMCORR and DoD results indicate significant displacements up to 40 meters in 65 years for the central part of the landslide. The historical datasets further indicate a change of spatio-temporal patterns of movement rates and a minor but overall acceleration of the landslide. The main challenges were the (i) gaps in the 3D point clouds on areas of steep, shadowed slopes and high vegetation, (ii) ground filtering on the photogrammetric point clouds for accurate calculation of digital terrain models (DTMs) and (iii) the quality of the scanned aerial imagery showing scratches, cuts, color irritations and linear artefacts. This research enabled the characterization of the spatio-temporal movement patterns of the Reissenschuh DSGSD over more than six decades. Further research will use the results as a reference for modelling the discussed multi-hazard processes.

This research was partly conducted within the project EMOD-SLAP funded by the Tyrolean Science Fund (TWF).