Retrieval of Sea Ice and Snow Parameters ...
537
19.3.1.1.2 Radar Discrimination of dry snow from snow-free ground is difficult, because
the dielectric properties of dry snow do not provide a strong contrast with those of air. Hence,
the backscattering coefficients of dry snow-covered terrain and snow-free terrain are close to
each other. Studies carried out in Sodankyla, northern Finland using a helicopter-borne radar
operating at 5.4 and 9.8 GHz, 4 linear polarizations, incidence angle 23 degrees (Hallikainen et
aI., 1994), suggest that the best radar channel to discriminate dry snow from snow-free ground
is 9.8 GHz, VV polarization. In the case of wet snow-free ground the contrast is 2.5 to 5 dB,
depending on land use category. For frozen snow-free ground the contrast is about 6 dB, Figure
19.7.
Ground-based studies in Switzerland show practically the same values for dry snow covered
terrain and snow-free terrain (Matzler and Schanda, 1984).
19.3.1.2 Wet snow
19.3.1.2.1 Radiometer Discrimination of wet snow from snow-free frozen terrain is not
straightforward, due to the similar dielectric properties of wet snow and frozen soil (Hallikainen
et aI., 1986). This is demonstrated in Figure 19.8.
Discrimination of wet snow from wet soil can be done best by using a low microwave frequency;
then the permittivity of wet snow is much smaller than that of wet soil. However, forest canopies
may mask the brightness temperature difference.
19.3.1.2.2 Radar Using ERS-1 SAR, the backscattering coefficient has been observed to
be about 3 dB smaller for wet snow than for other snow/soil categories, Figure 19.9 (Koskinen
et aI., 1994). This difference is rather independent of the biomass of boreal forests and can be
used to discriminate wet snow from other snow/soil categories. By averaging ERS-l SAR data
to eliminate speckle, classification accuracies of the order of 90% were achieved.
The physical basis for the discrimination is the following: wet snow absorbs incident radar
power on its way to the ground surface and again on its way into air. Absorption by dry snow
is negligible at presently used radar frequencies.
Similar results have been obtained by Rott and Nagler (1994) for Alpine test sites.
19.3.2 Snow water equivalent
Since wet snow absorbs microwave power, determination of the water equivalent of snow cover
seems to be possible only for dry snow.
19.3.2.1 Microwave radiometry
Most algorithms used to derive the water equivalent of dry snow cover rely on observations
made at frequencies 18 and 37 GHz. Experience about the SSM/I 85 GHz data is still limited,
although it seems to be useful. The algorithms are based on linear correlations observed between
the water equivalent and satellite-derived quantities.
The quantities are used in the algorithms include the spectral and polarization difference (SPD)
(Aschacher, 1989), spectral and temporal difference (STD) (Hallikainen and Jolma, 1992),
spectral gradient (SG) (Kiinzi et al., 1982) and spectral difference (SD) (Chang et al., 1987):
(19.15)
537
19.3.1.1.2 Radar Discrimination of dry snow from snow-free ground is difficult, because
the dielectric properties of dry snow do not provide a strong contrast with those of air. Hence,
the backscattering coefficients of dry snow-covered terrain and snow-free terrain are close to
each other. Studies carried out in Sodankyla, northern Finland using a helicopter-borne radar
operating at 5.4 and 9.8 GHz, 4 linear polarizations, incidence angle 23 degrees (Hallikainen et
aI., 1994), suggest that the best radar channel to discriminate dry snow from snow-free ground
is 9.8 GHz, VV polarization. In the case of wet snow-free ground the contrast is 2.5 to 5 dB,
depending on land use category. For frozen snow-free ground the contrast is about 6 dB, Figure
19.7.
Ground-based studies in Switzerland show practically the same values for dry snow covered
terrain and snow-free terrain (Matzler and Schanda, 1984).
19.3.1.2 Wet snow
19.3.1.2.1 Radiometer Discrimination of wet snow from snow-free frozen terrain is not
straightforward, due to the similar dielectric properties of wet snow and frozen soil (Hallikainen
et aI., 1986). This is demonstrated in Figure 19.8.
Discrimination of wet snow from wet soil can be done best by using a low microwave frequency;
then the permittivity of wet snow is much smaller than that of wet soil. However, forest canopies
may mask the brightness temperature difference.
19.3.1.2.2 Radar Using ERS-1 SAR, the backscattering coefficient has been observed to
be about 3 dB smaller for wet snow than for other snow/soil categories, Figure 19.9 (Koskinen
et aI., 1994). This difference is rather independent of the biomass of boreal forests and can be
used to discriminate wet snow from other snow/soil categories. By averaging ERS-l SAR data
to eliminate speckle, classification accuracies of the order of 90% were achieved.
The physical basis for the discrimination is the following: wet snow absorbs incident radar
power on its way to the ground surface and again on its way into air. Absorption by dry snow
is negligible at presently used radar frequencies.
Similar results have been obtained by Rott and Nagler (1994) for Alpine test sites.
19.3.2 Snow water equivalent
Since wet snow absorbs microwave power, determination of the water equivalent of snow cover
seems to be possible only for dry snow.
19.3.2.1 Microwave radiometry
Most algorithms used to derive the water equivalent of dry snow cover rely on observations
made at frequencies 18 and 37 GHz. Experience about the SSM/I 85 GHz data is still limited,
although it seems to be useful. The algorithms are based on linear correlations observed between
the water equivalent and satellite-derived quantities.
The quantities are used in the algorithms include the spectral and polarization difference (SPD)
(Aschacher, 1989), spectral and temporal difference (STD) (Hallikainen and Jolma, 1992),
spectral gradient (SG) (Kiinzi et al., 1982) and spectral difference (SD) (Chang et al., 1987):
(19.15)