Multi-temporal mapping of rock glacier displacements: insights from SfM, ALS and TLS high-resolution datasets on the Laurichard rock
Résumé
The deformation of rock glaciers is a climate-dependent process which affects many mountain debris slopes underlain with ice-rich permafrost (Haeberli et al., 2006). Under temperate conditions, like in the Alps, the typical magnitude of movement is in the order of a few dm/a to 1-2 m/a (Barsch, 1996). However, recent measurements of accelerating rock glaciers show that velocity can increase up to several m/a (Delaloye et al., 2012). This kinematics variability can locally lead to slope instabilities and trigger hazardous phenomena: in France, the collapse of the Bérard rock glacier in 2006 (Bodin et al., 2016) and the debris-flow that started at the Lou rock glacier in 2015 (Paulhe et al., 2015) can be considered as two events that exemplify the possible impacts of permafrost degradation for human societies living in mountain territories. At the limit between the Northern and the Southern-more Mediterranean-influenced-French Alps, the Laurichard rock glacier benefits from a long-term monitoring since the early 80's (Francou & Reynaud, 1992). Using 30-year long annual geodetic surveys of blocks, we detected inter-annual fluctuations of velocity (Bodin et al., 2009), which synchronicity with many series on other surveyed rock glaciers in the Alps suggests that the atmospheric warming, modulated by the nivological conditions, is controlling the rock glacier deformation rates (Delaloye et al., 2012). Nevertheless the knowledge on the processes that govern such behavior is still lacking physical basis: a better understanding requires to model the geomechanics of rock glacier, and for that accurate multi-temporal maps of surface displacements are fundamental. For that purpose, we produced very high-resolution 3D models of the Laurichard rock glacier, acquired either by terrestrial laserscanning (TLS) in 2005, 2006, 2011 and 2013, airborne laserscanning (ALS) in 2012 and terrestrial photogrammetry (or structure-from-motion, SfM) in 2013 and 2015. The density of the generated point clouds we generated varies from 2 pt/m² to 20 pt/m² and the accuracy of the coordinates after co-registration between the different datasets (2012 ALS dataset chosen as the reference) is in the order of 12-15 cm (standard error computed on stable areas). The main challenges for correct topographic mapping arose from the multi-scale variability: microtopographic features like plurimetric ridges and furrows shield some areas from the laser, whereas surface roughness of the pluri-decimetric coarse blocky cover may be difficult to cope with when comparing point-clouds between each other's. We employed 3D point-cloud processing and image correlation tools to compare high resolution hillshaded digital elevation models together and to extract various spatially-distributed measurements of surface movements. Both amplitude and direction of the displacements computed by images correlation were validated from comparison to in situ high precision geodetic (milimetric accuracy) measurements: at the various considered time scales and for 8 measured points, the mean difference between TLS/ALS /SfM-derived measurements and geodetics measurements is lower than 5 cm/a, for velocity that range between 25 and 150 cm/a. The spatially-distributed information provides rich insights into the deformation mechanisms of rock glaciers and open new challenging opportunities to move further into rheological laws and physical models. The ongoing next step of this work will consist in extracting 3D point-clouds from two fixed cameras surveying the Laurichard rock glacier, which will soon give access to higher resolutions, in both space and time dimensions.