Geology Reference
In-Depth Information
ice (>1 m). Thin ice between large floes can also be
crushed as a result of floe motion and turned into what is
known as brash ice. Unlike ridges and rubble ice, brash
ice thickness is not remarkably large.
Surface deformation may lead to mechanical thicken-
ing of the ice. This information is important for ship
navigation. Thick or deformed ice increases the mechani-
cal load on the ship, especially in the case of ridges
because they have more keel than sail as shown in
Figure 2.50. According to this figure if the sail of a ridge
is 2 m (a typical value), the keel could be 8.8 m. Ridges
vary in their length from a few meters to several hundred
meters. One of the largest observed ridges was in the
Beaufort Sea with a sail height of 12.8 m [
Melling et al
.,
1993]. Rubble fields do not have structure as ridges, yet
they contribute to an increase in ice thickness. In that
sense both ridges and rubble ice are important from a cli-
matic point of view because their large thickness affects
the total ice volume leaving the Arctic basin.
The most suitable remote sensing tool for detecting ice
surface deformation is SAR. That is partly because of its
fine spatial resolution but mainly because of its sensitivity
to surface roughness and structure. In fact, this desirable
feature has revolutionized many applications in the fields
of geology and Earth's surface morphology. All forms of
surface deformation at all scales are associated with some
degree of high backscatter in SAR images in both co‐ and
cross polarization. For example, backscatter coefficient
from pancake rough ice is typically > ‐8 dB. SAR can cap-
ture different surface roughness and deformation scales
through the different scattering mechanisms, ranging from
Bragg scatterings to double bouncing and random scatter-
ing (Section 7.6.2.3). Rafting can be identified visually in
radar images by its higher backscatter at the border of thin
ice floes (less than 0.3 m thick). The high backscatter in this
case is caused by the thickening of the overlapped ice edges,
not by surface roughness. This feature is used by ice analysts
in the Canadian Ice Service to identify the young ice types
associated with rafting, namely gray and gray‐white ice.
The interest in using remote sensing to identify and
gather statistics about ice surface deformation is centered
on ridges and to some extent brash ice between ice floes.
In addition to their high backscatter in co‐polarization
radar signal, ridges also return high backscatter in cross-
polarization because of the random scattering mecha-
nism (see Section 8.1.3). Their nearly linear geometrical
configuration is a key feature that facilitates their identi-
fication in the fine‐resolution SAR images either visually
or by applying an edge detection technique. However, a
difficulty may arise in quantifying the area of the ridge if
its width is smaller than the pixel size in the image (pixel
spacing of the resampled ScanSAR images used in ice
monitoring programs is typically 50-100 m). If only part
of the pixel is covered with a ridge, the entire pixel will be
Figure 9.1
Binary image showing ice floes as dark patches
with ridges as white features within a couple of floes. Rough
or brash ice appears as white areas. The image is roughly
12 km
×
12 km acquired by Radarsat‐2 HH on 2 May 2010.
bright. Therefore, areal information of the ridge cannot
be accurately extracted from SAR data.
Two approaches are commonly used to identify ridges
and other surface deformation forms in SAR images.
The first involves an appropriate image analysis tech-
nique to identify linear features that contrast with the
background in the image. The second involves inference
of the deformation from maps of the ice motion (when
convergence is identified). Pairs of SAR images acquired
within a short period (up to 3 days) should be used in
this approach. Each approach is discussed briefly in the
following.
In the image analysis approach, ridge pixels can be
identified based on their high backscatter or high gra-
dient of the backscatter with respect to the background.
Pixels of high backscatter or gradient can be identified
based on a simple threshold. A binary image resulting
from applying a threshold on the backscatter in a
Radarsat‐2 scene of sea ice in the Barrow Strait (close
to Somerset Island in the Canadian Arctic) is shown
in Figure 9.1. The pixel spacing is about 5 m (from the
quad‐polarimeteric fine beam mode). Note the bright
linear ridge within the ice floe at the top of the image.
Note also the presence of deformed ice between floes,
which appear white due to its high backscatter. This
approach can be used (yet with a limited success) to
generate statistics of the ridges and brash ice in terms
of area and spatial frequency. The threshold should be
carefully selected yet there is no guarantee for its
successful application.
