162
M.R. DRINKWATER
contrasts the ranges and statistics of these data with matching summary statistics
inferred from the mean field scatterometer aOShip curves (Le., shaded pdf's) in Fig. 5.
Additional comparison data are included for open water backscatter coefficients, and
MIZ scattering from brash ice or pancakes.
In Fig. 7, the MIZ provides one of the largest ranges of backscatter coefficient, due
to the influence of floe size, wave environment, and variable air temperature.
Depending on the season and location, frazil ice growth and pancake formation may
be favored. However, brash ice also provides extremely high-values due to its similar characteristics in terms of sub resolution floes and floe packing. A good analogy
may be drawn between the Weddell and the Labrador Sea MIZ, which was observed
by SAR during the LIMEX experiments (Drinkwater 1989; Livingstone and Drinkwater 1991). In Fig. 7 SAR and EScat values overlap due to the broad range in scattering
conditions described in Sect. 8.4.1.1.
Field sampling of smooth first-year ice floes is inherently biased, since an icebreaking ship normally seeks the path of least resistance. Although this is generally
true, the ship's route during WWGS'92 was based largely upon large-scale information provided by AVHRR satellite images and did not always enable sampling of the
smoothest first-year ice. The benefit of SAR images is that they enable the very
smoothest ice forms to be easily found and selected. Samples shown in Fig. 7 show
that the measurement ranges plotted from the ship borne scatterometer and SAR do
not overlap particularly well. In other words, ship borne radar measurements of
smooth first-year ice indicate a higher range of backscatter coefficients than samples
selected from the SAR image data. Similarly, the ship navigated around multiyear ice
floes or heavily ridged ice regions visible in the AVHRR images. As a consequence,
the multiyear ice which was sampled is at the least deformed extreme of such old ice
floes. Haas et al. (1992) document the level nature of most of the multiyear ice floes
sampled. In particular, aO Ship values in Fig. 7 indicate that the field scatterometer measurements are biased toward the lowermost end of the multiyear and rough first -year
backscatter ranges.
Other than for smooth first-year ice areas, shipborne scatterometer measurements
are generally lower at 40 0 incidence than equivalent EScat image data extracted in
the same locations. Spatial resolution and the averaging over a number of resolution
cells influences mean EScat backscatter measurements. Intervening ice motion (typically 1-2 pixels over the 6-day Escat averaging window) and resulting mixtures of
backscatter elements combine to give a relatively higher aO than that of any individual ice component. In addition, the larger roughness elements dominate the scattering in the higher incidence angle range, yet ship borne scatterometer measurements
cannot easily measure the signature of such nonuniform or deformed ice floe surfaces.
Open water signatures are generated in Fig. 7 for comparison, using the CMOD4
algorithm (Stoffelan and Anderson 1993). As expected, the range of backscatter values under differing wind conditions brackets the entire range of sea-ice signatures.
Nevertheless, by skillfully using image context (i.e., knowledge regarding the shape
of leads within the ice pack) together with ice dynamics information from ice tracking' it is possible to distinguish open water areas within the pack. In any case, observing an open lead within the Weddell Sea in winter is rare, and for the most part new
leads freeze within a matter of hours after opening.
M.R. DRINKWATER
contrasts the ranges and statistics of these data with matching summary statistics
inferred from the mean field scatterometer aOShip curves (Le., shaded pdf's) in Fig. 5.
Additional comparison data are included for open water backscatter coefficients, and
MIZ scattering from brash ice or pancakes.
In Fig. 7, the MIZ provides one of the largest ranges of backscatter coefficient, due
to the influence of floe size, wave environment, and variable air temperature.
Depending on the season and location, frazil ice growth and pancake formation may
be favored. However, brash ice also provides extremely high-values due to its similar characteristics in terms of sub resolution floes and floe packing. A good analogy
may be drawn between the Weddell and the Labrador Sea MIZ, which was observed
by SAR during the LIMEX experiments (Drinkwater 1989; Livingstone and Drinkwater 1991). In Fig. 7 SAR and EScat values overlap due to the broad range in scattering
conditions described in Sect. 8.4.1.1.
Field sampling of smooth first-year ice floes is inherently biased, since an icebreaking ship normally seeks the path of least resistance. Although this is generally
true, the ship's route during WWGS'92 was based largely upon large-scale information provided by AVHRR satellite images and did not always enable sampling of the
smoothest first-year ice. The benefit of SAR images is that they enable the very
smoothest ice forms to be easily found and selected. Samples shown in Fig. 7 show
that the measurement ranges plotted from the ship borne scatterometer and SAR do
not overlap particularly well. In other words, ship borne radar measurements of
smooth first-year ice indicate a higher range of backscatter coefficients than samples
selected from the SAR image data. Similarly, the ship navigated around multiyear ice
floes or heavily ridged ice regions visible in the AVHRR images. As a consequence,
the multiyear ice which was sampled is at the least deformed extreme of such old ice
floes. Haas et al. (1992) document the level nature of most of the multiyear ice floes
sampled. In particular, aO Ship values in Fig. 7 indicate that the field scatterometer measurements are biased toward the lowermost end of the multiyear and rough first -year
backscatter ranges.
Other than for smooth first-year ice areas, shipborne scatterometer measurements
are generally lower at 40 0 incidence than equivalent EScat image data extracted in
the same locations. Spatial resolution and the averaging over a number of resolution
cells influences mean EScat backscatter measurements. Intervening ice motion (typically 1-2 pixels over the 6-day Escat averaging window) and resulting mixtures of
backscatter elements combine to give a relatively higher aO than that of any individual ice component. In addition, the larger roughness elements dominate the scattering in the higher incidence angle range, yet ship borne scatterometer measurements
cannot easily measure the signature of such nonuniform or deformed ice floe surfaces.
Open water signatures are generated in Fig. 7 for comparison, using the CMOD4
algorithm (Stoffelan and Anderson 1993). As expected, the range of backscatter values under differing wind conditions brackets the entire range of sea-ice signatures.
Nevertheless, by skillfully using image context (i.e., knowledge regarding the shape
of leads within the ice pack) together with ice dynamics information from ice tracking' it is possible to distinguish open water areas within the pack. In any case, observing an open lead within the Weddell Sea in winter is rare, and for the most part new
leads freeze within a matter of hours after opening.