11 Snow and Ice
245
content, density, grain size and shape, temperature and stratification as well as snow
state and land cover. The sensitivity of the microwave radiation to a snow layer on the
ground makes it possible to monitor snow cover using passive microwave remote
sensing techniques to derive information on snow extent, snow depth, snow water
equivalent and snow state (wet/dry). Because the number of scatterers within a
snowpack is proportional to the thickness and density, SWE can be related to the
brightness temperature of the observed scene (Hallikainen and 10lma, 1986); deeper
snowpacks generally result in lower brightness temperatures.
The general approach used to derive SWE and snow depth from passive microwave
satellite data relates back to those presented by Rango et aI. (1979) and Kunzi et aI.
(1982) using empirical approaches and Chang et aI. (1987) using a theoretical basis
from radiative transfer calculations to estimate snow depth from SMMR data. As
discussed in Rott (1993), the most generally applied algorithms for deriving depth or
snow water equivalent (SWE) are based on the generalized relation given in Eq. 11.2
SWE = A + B «T Bifl) - T B(fl))/(fl-f1)) in mm, for SWE > 0
(11.2)
where A and B are the offset and slope of the regression of the brightness temperature
difference between a high scattering channel (fl, commonly 37GHz) and a low
scattering one ifl, commonly 18 or 19 Ghz) of vertical or horizontal polarization. No
single global algorithm will estimate snow depth or water equivalent under all snowpack and land cover conditions. The coefficients are generally determined for different climate and land covered regions and for different snow cover conditions; algorithms used in regions other than for which they were developed and tested usually
provide inaccurate estimates of snow cover. Also, accurate retrieval of information
on snow extent, depth, and water equivalent requires dry snow conditions, because
the presence of liquid water within the snowpack drastically alters the emissivity of
the snow, resulting in brightness temperatures significantly higher than if that snowpack were dry. Therefore, an early morning overpass (local time) is the preferred orbit
for retrieval of snow cover information to minimize wet snow conditions. It is also
recognized that knowledge of snowpack state is useful for hydrological applications.
Regular monitoring allows detection of the onset of melt or wet snow conditions
(Goodison and Walker, 1995).
The accuracy of the retrieval algorithms is a function of the quality of both the
satellite data and the snow cover measurements used in their development. Empirical
methods to develop algorithms involve the correlation of observed T B with coincident
conventional depth measurements (as from meteorological stations) or ground SWE
data (such as from areally representative snow courses). Goodison et aI. (1986) used
coincident airborne microwave data and airborne gamma data collected over the
Canadian prairie area, supplemented by special ground surveys, to derive their SWE
algorithm which has now been used for over 10 years in operational hydrological
forecast operations. Non-forested open prairie areas have generally shown the best
correlation between areal SWE and brightness temperature (e.g. Kunzi et aI., 1982;
Goodison et aI., 1986; Chang et aI., 1987). Hallikainen and 10lma (1986) and Hallikainen (1989) report on Finnish studies over various landscapes. Rott and Nagler
(1993) report on European algorithm development which incorporates information
245
content, density, grain size and shape, temperature and stratification as well as snow
state and land cover. The sensitivity of the microwave radiation to a snow layer on the
ground makes it possible to monitor snow cover using passive microwave remote
sensing techniques to derive information on snow extent, snow depth, snow water
equivalent and snow state (wet/dry). Because the number of scatterers within a
snowpack is proportional to the thickness and density, SWE can be related to the
brightness temperature of the observed scene (Hallikainen and 10lma, 1986); deeper
snowpacks generally result in lower brightness temperatures.
The general approach used to derive SWE and snow depth from passive microwave
satellite data relates back to those presented by Rango et aI. (1979) and Kunzi et aI.
(1982) using empirical approaches and Chang et aI. (1987) using a theoretical basis
from radiative transfer calculations to estimate snow depth from SMMR data. As
discussed in Rott (1993), the most generally applied algorithms for deriving depth or
snow water equivalent (SWE) are based on the generalized relation given in Eq. 11.2
SWE = A + B «T Bifl) - T B(fl))/(fl-f1)) in mm, for SWE > 0
(11.2)
where A and B are the offset and slope of the regression of the brightness temperature
difference between a high scattering channel (fl, commonly 37GHz) and a low
scattering one ifl, commonly 18 or 19 Ghz) of vertical or horizontal polarization. No
single global algorithm will estimate snow depth or water equivalent under all snowpack and land cover conditions. The coefficients are generally determined for different climate and land covered regions and for different snow cover conditions; algorithms used in regions other than for which they were developed and tested usually
provide inaccurate estimates of snow cover. Also, accurate retrieval of information
on snow extent, depth, and water equivalent requires dry snow conditions, because
the presence of liquid water within the snowpack drastically alters the emissivity of
the snow, resulting in brightness temperatures significantly higher than if that snowpack were dry. Therefore, an early morning overpass (local time) is the preferred orbit
for retrieval of snow cover information to minimize wet snow conditions. It is also
recognized that knowledge of snowpack state is useful for hydrological applications.
Regular monitoring allows detection of the onset of melt or wet snow conditions
(Goodison and Walker, 1995).
The accuracy of the retrieval algorithms is a function of the quality of both the
satellite data and the snow cover measurements used in their development. Empirical
methods to develop algorithms involve the correlation of observed T B with coincident
conventional depth measurements (as from meteorological stations) or ground SWE
data (such as from areally representative snow courses). Goodison et aI. (1986) used
coincident airborne microwave data and airborne gamma data collected over the
Canadian prairie area, supplemented by special ground surveys, to derive their SWE
algorithm which has now been used for over 10 years in operational hydrological
forecast operations. Non-forested open prairie areas have generally shown the best
correlation between areal SWE and brightness temperature (e.g. Kunzi et aI., 1982;
Goodison et aI., 1986; Chang et aI., 1987). Hallikainen and 10lma (1986) and Hallikainen (1989) report on Finnish studies over various landscapes. Rott and Nagler
(1993) report on European algorithm development which incorporates information