Information on snow depth and its spatial distribution in high-alpine catchments is crucial for many research topics and applications such as avalanche research, hydrology, snow and mountain ecology. Nowadays snow depth values are measured at point locations such as automatic weather stations and observers in the field and are interpolated to get an estimate of large scale snow depth. Several studies report a high spatial variability of snow depth and other snow pack parameters in mountainous regions, which cannot be reproduced by point measurements. Optical stereo data from different platforms offer a very attractive alternative to derive 3D surface data. Compared to laser scanning, photogrammetry is more economic and with the application of unmanned aerial systems (UAS) also more flexible (Bühler et al. 2016).
Recent developments of sensor- and software technology for digital photogrammetry enable the generation of high spatial resolution digital surface models (DSMs) from airborne digital imagery even in high alpine regions and over snow-covered surfaces.
For this project, digital imagery from the Leica ADS80 (2010-2013) and ADS100 (2014-today) opto-electronic scanners for a test site near Davos (145 km2) are used to derive high spatial resolution digital surface models by applying a highly automated workflow. The spatial resolution of the imagery is ~0.25 m (ADS80) and ~0.18 m (ADS100) and the radiometric resolution 12 bit (both). For the dense image matching the spectral bands in the near infrared (), red() and green() part of the electromagnetic spectrum are used.
Snow depth maps are calculated as the difference between the winter DSMs and a snow-free summer DSM, whereby trees/shrubs, buildings and other infrastructure are masked out. One of the main challenges to derive good quality snow depth maps is the absolute orientation of the ADS stereo imagery, especially in winter due to the limited numbers of usable ground control points in a snow covered landscape.
The accuracy of the derived snow depth maps of 2012 was estimated with reference data collected in a field campaign on the same day as the acquisition of the imagery in winter 2012. Collected reference data were: differential GNSS (dGNSS) height measurements (n=137), snow depths derived by terrestrial laser scanning (TLS) in a subregion, ground penetrating radar (GPR) measurements and hand measured snow depths.
The accuracy assessment applying dGNSS data results in a RMSE of 0.37m, NMAD of 0.28m and a mean of 0.21m. The assessment based on the TLS dataset results in a RMSE=0.33m and a NMAD=0.26m. The accuracy assessed with the hand measured snow depth plots results in a RMSE of 0.19 m and a NMAD of 0.18m (Bühler et al. 2015).
These results show the feasibility of the used method to map snow depth and its high spatial variability in high-alpine catchments. Until now the derived photogrammetric snow depth maps have already been used in different topics in other WSL/SLF groups like the investigation of the dependency of snow depth on elevation (Grünelwald et al. 2014 ), the parametrization of fractional snow covered area over complex topography (Helbig et al. 2015) and the scaling of precipitation input to distributed hydrological models (Vögeli et al. 2016) and is currently used for various further investigations at WSL/SLF.
Snow depth maps are available for the following dates: