Remote Sensing of the Alpine Snow Cover:
A Review of Techniques and Accomplishments
from the Visible Wavelengths through the Microwave
Jeff Dozier 1
In regions where snow accumulation and ablation vary spatially, distributed snowmelt models are a
promising approach to improving runoff analysis and forecasts. Employment of distributed models
requires spatially distributed input data, hence the desire to remotely sense snow properties. Alpine
regions pose particular challenges for remote sensing. There is usually considerable variability in
snow depth and other properties at a fine spatial scale, and analysis of the remotely sensed signal
must account for geometric effects that the sloping terrain and range of elevations cause. In some
areas, persistent cloud cover hampers regular acquisition of data.
Many snow properties might be obtainable from remote sensing. Rather than try to cover all plausi-
ble variables, I concentrate on three.
The most elementary snow property is the presence or absence of a snow cover, so we must dis-
tinguish snow from other types of surfaces and from clouds. For snow mapping in the absence of
cloud cover, a combination of visible and near-infrared wavelengths gives excellent results. With
rich spectral data from multispectral or especially hyperspectral sensors that operate in the visible
and near-infrared parts of the electromagnetic spectrum, we can estimate the snow-covered portion
of a pixel and the snow's spectral albedo. In cloudy conditions, we can map snow with the micro-
wave part of the spectrum, using either active or passive sensors. The active microwave (radar) has
a finer spatial resolution that is necessary in mountainous areas.
The most essential property for hydrologic analysis is usually the snow water equivalence. For deep
mountain snowpacks, the active microwave is the spectral region where we might directly estimate
the water equivalence. The presence of liquid water in the snow makes the analysis of such data
difficult because water and ice have such different dielectric properties in the microwave. Topo-
graphic variability also complicates the analysis, because the signal is sensitive to the incidence
angle. Moreover, the models often incorporate snow properties that are difficult to measure, and they
are analytically formidable and difficult to invert.
Finally, while the effect of liquid water on the microwave signal hinders the retrieval of snow water
equivalence, it does allow estimation of the liquid water content in the near-surface layer to an
accuracy of about 2%.
1
Donald Bren School of Environmental Science and Management, University of California, Santa Barbara,
California 93106, USA
33
Previous Page
