8 Satellite Microwave Radar Observations of Antarctic Sea Ice
8.4.2
The Future of Ice Classification
The advent of a global dataset from ERS and NSCAT scatterometers enables uninterrupted C- and Ku-band radar coverage of the Arctic and Antarctic sea-ice cover to be
generated. These synoptic-scale images can be exploited to address global ice classification in conjunction with ice dynamics issues. Recent work by Beaven and Gogineni
(this volume) fusing active and passive datasets indicates that radar in combination
with SSM/I passive microwave image data increases the amount of independent information and adds dimensionality to the problem of unambiguous identification of
classes of ice.
8.4.2.1
Scatterometer Ice Classification
Early attempts at one-dimensional classification of Antarctic ice have used simple
supervised clustering analyses on the basis of EScat backscatter statistics. Backscatter
coefficients at 40 0 incidence appear to characterize the large-scale deformation and surface roughness of the sea ice (Drinkwater et al. 1993a). Simple slicing of an EScat image
using the values given in Table 3 yields results shown in Fig. 8. Figure 8a is generated
for a 6 day period beginning on June 16,1992 (day 168), and Fig. 8b for a similar period starting at July 31,1996 (day 213). Each part of Fig. 8 indicates the track of R.Y.
Polarstern during WWGS' 92 together with the sea -ice sample sites marked by symbols.
Physical descriptions of individual samples of ice are provided by Drinkwater and Haas
(1994), together with shipborne scatterometer scattering coefficients (i.e., croShip)'
Figure 8a coincides temporally with the first four sea-ice samples acquired along
the Greenwich meridian (marked by"+" symbols). The classification routine identifies a large patch of pancake ice formation corresponding directly to the location
shown in Figure 3. Table 3 indicates that pixel values fall in the range -11 ~ croEScat < -6
dB. Later in the WWGS'92 experiment, as Polarstern traversed the Weddell Sea, large
expanses of undeformed level white ice were observed (-20.0 ~croEScat < -14.0 dB), and
the ship track was intentionally modified to avoid a large region of deformed ice, which
was identified in real time using on-board received AVHRR data. Level first-year ice
regions (FYS) are confirmed at sites marked by squares. The ship track bends around
a large patch of deformed, rough first -year ice which originated along the eastern coast
of the Weddell Sea and drifted progressively northwestwards from day 168 onwards.
The ice outflow region in the northwest Weddell Sea is characterized by rough firstyear and multiyear ice forms. In the classified image, the margin between level and
thicker, deformed first-year ice closely fits field observations of a boundary between
characteristics in first-year ice, where sample sites of thicker, deformed first-year ice
are marked by diamonds. At the time of imaging, the ship was preparing to exit the
sea ice, in the northwestern Weddell Sea MIZ, and circles symbols indicate mixed multiyear and brash ice samples in the location of orbit 5387 shown in Fig. 2 (Drinkwater
and Haas 1994). This region of high backscatter corresponds with brash and wavebroken pieces of deformed first-year sea ice and a small concentration of deep snowcovered old ice, and the algorithm classifies the region as pancakes and/or multiyear
ice.
8.4.2
The Future of Ice Classification
The advent of a global dataset from ERS and NSCAT scatterometers enables uninterrupted C- and Ku-band radar coverage of the Arctic and Antarctic sea-ice cover to be
generated. These synoptic-scale images can be exploited to address global ice classification in conjunction with ice dynamics issues. Recent work by Beaven and Gogineni
(this volume) fusing active and passive datasets indicates that radar in combination
with SSM/I passive microwave image data increases the amount of independent information and adds dimensionality to the problem of unambiguous identification of
classes of ice.
8.4.2.1
Scatterometer Ice Classification
Early attempts at one-dimensional classification of Antarctic ice have used simple
supervised clustering analyses on the basis of EScat backscatter statistics. Backscatter
coefficients at 40 0 incidence appear to characterize the large-scale deformation and surface roughness of the sea ice (Drinkwater et al. 1993a). Simple slicing of an EScat image
using the values given in Table 3 yields results shown in Fig. 8. Figure 8a is generated
for a 6 day period beginning on June 16,1992 (day 168), and Fig. 8b for a similar period starting at July 31,1996 (day 213). Each part of Fig. 8 indicates the track of R.Y.
Polarstern during WWGS' 92 together with the sea -ice sample sites marked by symbols.
Physical descriptions of individual samples of ice are provided by Drinkwater and Haas
(1994), together with shipborne scatterometer scattering coefficients (i.e., croShip)'
Figure 8a coincides temporally with the first four sea-ice samples acquired along
the Greenwich meridian (marked by"+" symbols). The classification routine identifies a large patch of pancake ice formation corresponding directly to the location
shown in Figure 3. Table 3 indicates that pixel values fall in the range -11 ~ croEScat < -6
dB. Later in the WWGS'92 experiment, as Polarstern traversed the Weddell Sea, large
expanses of undeformed level white ice were observed (-20.0 ~croEScat < -14.0 dB), and
the ship track was intentionally modified to avoid a large region of deformed ice, which
was identified in real time using on-board received AVHRR data. Level first-year ice
regions (FYS) are confirmed at sites marked by squares. The ship track bends around
a large patch of deformed, rough first -year ice which originated along the eastern coast
of the Weddell Sea and drifted progressively northwestwards from day 168 onwards.
The ice outflow region in the northwest Weddell Sea is characterized by rough firstyear and multiyear ice forms. In the classified image, the margin between level and
thicker, deformed first-year ice closely fits field observations of a boundary between
characteristics in first-year ice, where sample sites of thicker, deformed first-year ice
are marked by diamonds. At the time of imaging, the ship was preparing to exit the
sea ice, in the northwestern Weddell Sea MIZ, and circles symbols indicate mixed multiyear and brash ice samples in the location of orbit 5387 shown in Fig. 2 (Drinkwater
and Haas 1994). This region of high backscatter corresponds with brash and wavebroken pieces of deformed first-year sea ice and a small concentration of deep snowcovered old ice, and the algorithm classifies the region as pancakes and/or multiyear
ice.