D.P. WINEBRENNER, D.G. LONG, B. HOLT
7.3
Enhanced-Resolution Scatterometer Observations of Freeze-Up
Spaceborne scatterometers are stable, well-calibrated instruments for measuring
backscattering cross sections, albeit at spatial resolutions coarser than the scales of cross
section (and geophysical) variability inherent in some geophysical targets. (For a discussion of relevant operational details of spaceborne scatterometry, including the matter of normalization to a standard incidence angle, see Long et al. 1993).
We are motivated to investigate the utility of scatterometers for observing seasonal
transitions on sea ice for at least two reasons. The first is a prospective advantage stemming from the shorter wavelength (higher frequency) at which many spaceborne scatterometers operate, relative to that of Radarsat or the ERS-l and -2 SARs and scatterometers. At the frequency (5.3 GHz) of those latter instruments, backscattering from
winter first-year sea is weak. Hence it is difficult to observe the melt onset transition
on first-year ice and to automate data interpretation in areas covered mostly by firstyear ice (see the previous section). However, the operating frequency of the Seasat scatterometer was 14.6 GHz, and that of the new NASA scatterometer (NSCAT) is 14.0 GHz.
Ground-based observations near those frequencies (Onstott 1992) indicate that winter
backscattering from first-year ice is considerably higher than at 5.3 GHz (though the
physical reason for this is not clear as of this writing). Dry snow would be expected, on
theoretical and observational grounds to be slightly less transparent at the higher frequencies than at 5.3 GHz, but a given thickness of wet snow would be expected to be a
more effective shield (Hallikainen and Winebrenner 1992). Therefore we would expect
a much stronger drop in 13- to 15-GHz backscattering cross sections at melt onset than
is observed at 5.3 GHz. This would make backscattering observations at the higher frequencies far more easily usable for melt onset observation in areas such as the Siberian shelves, where first-year ice predominates. This would significantly expand the geophysical utility of melt onset mapping based on backscattering.
The typical spatial resolutions of space borne scatterometers, however, and of the
Seasat, NSCAT and ERS-l scatterometers in particular, are nominally 50 km. As
Carsey (1985) and Wismann et al. (1996) have shown, even this resolution is of some
use on sea ice, but there is mixing and a loss of signature information because sea ice
characteristics and signatures vary on horizontal scales as short as meters. This is
not to say that information on all spatial scales down to meters is always useful for
all purposes - experience with low-resolution imagery of sea ice indicates that sea
ice imagery with 5-km resolution shows much more information than 50-km
imagery, especially if the 5-km imagery comes at an acceptable price in terms of spatial coverage, temporal resolution, and radiometric fidelity. As we discuss below, such
resolution is evidently available. Moreover, while the resolution of SAR provides a
way to usefully focus on a specific ice type (see the previous section), there remains
a strong practical motivation to distill the information in a high-resolution image
quickly into a smaller volume. We therefore have a second motivation, in addition to
the promise of shorter wavelength observations, to explore further the potential of
enhanced-resolution scatterometry - scatterometer observations may complement
or even, in some cases, substitute for SAR data. It seems likely that whatever is learned
in this effort will contribute insight useful in building better automated SAR, as well
as scatterometer, algorithms.
7.3
Enhanced-Resolution Scatterometer Observations of Freeze-Up
Spaceborne scatterometers are stable, well-calibrated instruments for measuring
backscattering cross sections, albeit at spatial resolutions coarser than the scales of cross
section (and geophysical) variability inherent in some geophysical targets. (For a discussion of relevant operational details of spaceborne scatterometry, including the matter of normalization to a standard incidence angle, see Long et al. 1993).
We are motivated to investigate the utility of scatterometers for observing seasonal
transitions on sea ice for at least two reasons. The first is a prospective advantage stemming from the shorter wavelength (higher frequency) at which many spaceborne scatterometers operate, relative to that of Radarsat or the ERS-l and -2 SARs and scatterometers. At the frequency (5.3 GHz) of those latter instruments, backscattering from
winter first-year sea is weak. Hence it is difficult to observe the melt onset transition
on first-year ice and to automate data interpretation in areas covered mostly by firstyear ice (see the previous section). However, the operating frequency of the Seasat scatterometer was 14.6 GHz, and that of the new NASA scatterometer (NSCAT) is 14.0 GHz.
Ground-based observations near those frequencies (Onstott 1992) indicate that winter
backscattering from first-year ice is considerably higher than at 5.3 GHz (though the
physical reason for this is not clear as of this writing). Dry snow would be expected, on
theoretical and observational grounds to be slightly less transparent at the higher frequencies than at 5.3 GHz, but a given thickness of wet snow would be expected to be a
more effective shield (Hallikainen and Winebrenner 1992). Therefore we would expect
a much stronger drop in 13- to 15-GHz backscattering cross sections at melt onset than
is observed at 5.3 GHz. This would make backscattering observations at the higher frequencies far more easily usable for melt onset observation in areas such as the Siberian shelves, where first-year ice predominates. This would significantly expand the geophysical utility of melt onset mapping based on backscattering.
The typical spatial resolutions of space borne scatterometers, however, and of the
Seasat, NSCAT and ERS-l scatterometers in particular, are nominally 50 km. As
Carsey (1985) and Wismann et al. (1996) have shown, even this resolution is of some
use on sea ice, but there is mixing and a loss of signature information because sea ice
characteristics and signatures vary on horizontal scales as short as meters. This is
not to say that information on all spatial scales down to meters is always useful for
all purposes - experience with low-resolution imagery of sea ice indicates that sea
ice imagery with 5-km resolution shows much more information than 50-km
imagery, especially if the 5-km imagery comes at an acceptable price in terms of spatial coverage, temporal resolution, and radiometric fidelity. As we discuss below, such
resolution is evidently available. Moreover, while the resolution of SAR provides a
way to usefully focus on a specific ice type (see the previous section), there remains
a strong practical motivation to distill the information in a high-resolution image
quickly into a smaller volume. We therefore have a second motivation, in addition to
the promise of shorter wavelength observations, to explore further the potential of
enhanced-resolution scatterometry - scatterometer observations may complement
or even, in some cases, substitute for SAR data. It seems likely that whatever is learned
in this effort will contribute insight useful in building better automated SAR, as well
as scatterometer, algorithms.