8 Satellite Microwave Radar Observations of Antarctic Sea Ice
149
are distributed in a full resolution (i.e., 12.5 m pixel spacing), uncalibrated image form.
According to a study by Bally and Fellah (1995) for these three-look SAR images (i.e. ESA
SAR precision image products), the single pixel (12.5 m) 90% radiometric confidence
interval is bounded by a ±4.5 dB interval. Thus, the probability that the measured intensity lies within a ±4.5 dB error bound is 90%, whereas the radiometric accuracy and stability errors are within specifications of a fraction of a decibel (Laur et al. 1993).
To calculate the SAR backscattering coefficient (crOSAR) of a distributed target which
corresponds to an area of sea ice (N pixels in extent), averaging is performed. Block
averaging is generally first used to create a speckle-filtered image, with the new equivalent number of looks modified by the area of block averaging together with the target area (i.e., number of pixels averaged from target). The resulting confidence interval for an 8 x 8 block-averaged image with 100 m pixel spacing exceeds 90% for ±1 dB
bounds. With further target averaging of a sample box of over 250 pixels (i.e., 16 x 16
pixel box), a confidence level exceeding 90% may be achieved with error bounds of ±0.5
dB. Over 99% probability of errors less than ±0.5 dB may be reached with samples of
greater than 500 pixels (i.e., 23 x 23 pixel box). Thus, for most purposes, having already
used a box-filter (8 x 8 window) to reduce the image to 100 m pixel spacing, it is generally only necessary to derive samples from target box areas exceeding 8 x 8 pixels to
exceed the ±0.5 dB error bound at 99% confidence. All samples presented in this paper
are from target areas exceeding this minimum threshold area.
8.2.2
Antarctic SAR Receiving Stations
It is normally only feasible to acquire high-resolution SAR images within direct broadcast range or line-of-sight to a ground receiving station, due to the high data rates. Priorto the launch ofERS-1 in July 1991 there were no satellite receiving stations in Antarctica equipped to downlink or record SAR data. Since 1992, a chain of contiguous SAR
data reception areas has been developed to support the SAR data requirements of a
number of nations. Currently, McMurdo (US), O'Higgins (German), and Syowa (Japanese) are all equipped and capable of downlinking and recording SAR image data. Figure 1 illustrates that almost complete spatial coverage of the mean maximum seasonal
ice extent is possible when all three stations are operating together.
The first SAR station to open in Antarctica was the German Antarctic Receiving
Station (GARS) at the Chilean General Bernardo O'Higgins station, located on the
northern end of the Antarctic peninsula (at 63.32°S, 57.90 0 W). GARS has been collecting ERS-1 data since October 1991, but its data reception is limited to 1-2 month
"campaign -style" operating periods when it is manned. These periods typically coincide with German Antarctic experiments, or periods when the German icebreaker
R.V. Polarstern is operating in the Weddell or Bellingshausen Seas. To date the O'Higgins receiving station has enabled the most comprehensive SAR data records of
Antarctic ice to be collected.
Presently, SAR data are transported out of Antarctica by ship or aircraft (via the
Chilean Teniente Marsh Base), thereby resulting in delays before SAR image processing takes place. However, a high -speed satellite link is being developed which will
(in late 1997) enable McMurdo SAR data to be relayed back to the US for processing
at the Alaska SAR Facility (for further information see ASF URL: http://www.asf.alas-
149
are distributed in a full resolution (i.e., 12.5 m pixel spacing), uncalibrated image form.
According to a study by Bally and Fellah (1995) for these three-look SAR images (i.e. ESA
SAR precision image products), the single pixel (12.5 m) 90% radiometric confidence
interval is bounded by a ±4.5 dB interval. Thus, the probability that the measured intensity lies within a ±4.5 dB error bound is 90%, whereas the radiometric accuracy and stability errors are within specifications of a fraction of a decibel (Laur et al. 1993).
To calculate the SAR backscattering coefficient (crOSAR) of a distributed target which
corresponds to an area of sea ice (N pixels in extent), averaging is performed. Block
averaging is generally first used to create a speckle-filtered image, with the new equivalent number of looks modified by the area of block averaging together with the target area (i.e., number of pixels averaged from target). The resulting confidence interval for an 8 x 8 block-averaged image with 100 m pixel spacing exceeds 90% for ±1 dB
bounds. With further target averaging of a sample box of over 250 pixels (i.e., 16 x 16
pixel box), a confidence level exceeding 90% may be achieved with error bounds of ±0.5
dB. Over 99% probability of errors less than ±0.5 dB may be reached with samples of
greater than 500 pixels (i.e., 23 x 23 pixel box). Thus, for most purposes, having already
used a box-filter (8 x 8 window) to reduce the image to 100 m pixel spacing, it is generally only necessary to derive samples from target box areas exceeding 8 x 8 pixels to
exceed the ±0.5 dB error bound at 99% confidence. All samples presented in this paper
are from target areas exceeding this minimum threshold area.
8.2.2
Antarctic SAR Receiving Stations
It is normally only feasible to acquire high-resolution SAR images within direct broadcast range or line-of-sight to a ground receiving station, due to the high data rates. Priorto the launch ofERS-1 in July 1991 there were no satellite receiving stations in Antarctica equipped to downlink or record SAR data. Since 1992, a chain of contiguous SAR
data reception areas has been developed to support the SAR data requirements of a
number of nations. Currently, McMurdo (US), O'Higgins (German), and Syowa (Japanese) are all equipped and capable of downlinking and recording SAR image data. Figure 1 illustrates that almost complete spatial coverage of the mean maximum seasonal
ice extent is possible when all three stations are operating together.
The first SAR station to open in Antarctica was the German Antarctic Receiving
Station (GARS) at the Chilean General Bernardo O'Higgins station, located on the
northern end of the Antarctic peninsula (at 63.32°S, 57.90 0 W). GARS has been collecting ERS-1 data since October 1991, but its data reception is limited to 1-2 month
"campaign -style" operating periods when it is manned. These periods typically coincide with German Antarctic experiments, or periods when the German icebreaker
R.V. Polarstern is operating in the Weddell or Bellingshausen Seas. To date the O'Higgins receiving station has enabled the most comprehensive SAR data records of
Antarctic ice to be collected.
Presently, SAR data are transported out of Antarctica by ship or aircraft (via the
Chilean Teniente Marsh Base), thereby resulting in delays before SAR image processing takes place. However, a high -speed satellite link is being developed which will
(in late 1997) enable McMurdo SAR data to be relayed back to the US for processing
at the Alaska SAR Facility (for further information see ASF URL: http://www.asf.alas-