The results presented demonstrate the potential of digital
photogrammetry for catchment-wide snow depth mapping.
The extensive validation using independent data sets acquired
simultaneously reveals an accuracy of approximately
30 cm (RMSE, NMAD), equivalent to 1 GSD of the input
images (Table 4). Due to the high radiometric resolution
of the images (12 bit) and the use of the near infrared band,
the images were not saturated over bright, snow-covered areas
and information could be acquired even in shadow. The
image correlations work even over very homogeneous areas.
Table 2 reveals almost the same correlation success with winter
images compared to summer images. The resulting snow
depth maps visualize the high spatial variability of snow
depth even within short distances of a few meters. Snow
traps for wind-blown snow, cornices and deposits from past
avalanche events can be identified easily by high snow depth
values of up to 15 m.
In this paper we applied six different methodologies to
map snow depth in high-alpine terrain. Table 6 lists the major
strengths and weaknesses of these methods based on the
experience of the authors. However, which method should be
applied in a specific case depends on many different factors
and should be evaluated with care.
We plan to acquire similar data sets at the end of upcoming
winters for an interannual comparison of snow depth. This
would also open the door for investigations into the representativeness
of snow depth measurements at given points,
for example at automated weather stations. Future comparisons
between snow depth maps generated by lidar and digital
photogrammetry will provide more detailed information
on the specific strengths and weaknesses of the two methods