Previous attempts to map snow depth using scanned aerial
imagery were already made 50 years ago (Smith et al., 1967)
and the topic was investigated in detail by Cline (1993,
1994). However, their results suffer from image saturation
and insufficient reference data leading them to the conclusion
that photogrammetry has much potential but is not yet accurate
enough for large-scale snow depth mapping. Ledwith
and Lunden (2010) used scanned aerial imagery to derive
digital elevation models over glaciated and snow-covered areas
in Norway. They report a mean accuracy of 2.8m in comparison
with differential Global Navigation Satellite System
(dGNSS) transects, which is clearly too low for meaningful
snow depth mapping in alpine regions. Lee et al. (2008)
used a digital mapping camera (DMC) digital frame camera
to cover an area of approximately 2.3 km2 with a very high
mean ground sampling distance (GSD) of 0.08 m. The reported
mean differences compared to dGNSS measurements
are approximately 0.15 m, stressing the great potential of digital
photogrammetry for accurate snow depth mapping. However,
no snow depth mapping has been performed and compared
to different reference data sets, covering larger areas.
In this investigation we apply digital photogrammetry
based on high spatial resolution aerial imagery (0.25 m)
to calculate digital surface models (DSM) of winter and
summer terrain. Traditional photogrammetry using analogue
aerial imagery and 8 bit digital sensors faced problems over
snow-covered areas mainly due to saturation and the homogeneous
surface (Kraus, 2004). Modern digital sensors can
acquire data with 12 bit radiometric resolution to overcome
these limitations. We calculate spatially continuous snow
depth maps using the summer and winter DSMs for two test
sites near Davos, Switzerland (145 km2 in total). This technology
is much more economical for covering large areas
than ALS or TLS but still has an acceptable spatial resolution
to map small-scale spatial variability. To assess the accuracy
of our results, we compare the calculated snow depths to
hand measurements, dGNSS points, TLS measurements and
ground-penetrating radar (GPR) transects acquired simultaneously
with the aerial imagery.