Images of and
were separately generated for vertically and horizontally
polarized
measurements; however, there were only 1/10 as many
horizontal measurements as there were vertical measurements for the study
region. This occurs because the instrument was primarily operated in a
dual-sided mode in which only the vertically polarized antennas on either side
of the spacecraft were used. The instrument was less frequently operated in a
single-side mode in which both polarizations were used. Hence, the intrinsic
resolution of the horizontally-polarized images is lower than the
vertically-polarized images and the horizontally-polarized images have
``holes'' due to missing coverage. For this reason, we first consider
enhanced resolution images of
and
generated from vertically-polarized
measurements.
Figure 6 shows an enhanced resolution estimate image generated using the SIRF
algorithm and the entire three-month SASS data set of vertically-polarized
measurements. The corresponding
estimate image is shown in Fig. 7.
The resolution of these images is
.
For reference, Fig. 1 illustrates typical
measurement cells used to
construct these images.
Perhaps the most striking feature of the image is the obvious boundary
between the light tropical rainforest in the upper half of the image and the
darker savanna woodlands below. Interdigitation and transition between the
woodlands and grass-shrublands are apparent in the lower right quarter of the
image. The Amazon river and many of its tributaries are clearly visible against
the lighter rainforest region. Within the rainforest region, there appear many
small regions of woodland. The large dark region at 12
south and
64
west in the lower left quarter of the rainforest is an area of
tropical tall grassland mixed with palms known as the Llanos de Mojos. The
smaller dark areas north of the Amazon in the upper center of the image are also
areas of tall grassland (very dark) mixed with scrubby trees and seasonal
evergreen forest (dark gray). The broad gray area left of center in the lower
portion of the image consists of woodland and thorn-forest. On the far right is
dry steppe-like vegetation, mixed with agricultural areas. The highest
backscatter occurs in a small region at the extreme southwest corner of the
landmass shown on the image corresponding to the Salar de Atacama, a salt
pan upland basin within the Atacama Desert. While not clearly evident in this
reproduction, coastal urban areas also have very high backscatter.
The remarkable correlation of the image a the vegetation map prompted us
to consider using the SASS-derived images in various vegetation discrimination
studies. Our results are described below. The
image shows only very limited
spatial detail which is generally obscured by noise artifacts; however, careful
inspection of the
images reveal some spatial features which correlate with
the key features evident in the
images. The
image provided only limited
additional information in the discrimination experiments described in later
sections.
We note, that when seen in greater detail as depicted in the image, the Amazon
rainforest region appears to have more small-scale spatial variation than
previously thought. Such inhomogeneity can contribute to the calibration
uncertainty when Amazon rainforest measurements are used to calibrate the
scatterometer. Use of these enhanced resolution
and
images to compensate
for the small heterogeneity of the rainforest region will result in improved
instrument calibration accuracy.