Geology Reference
In-Depth Information
on single‐channel SAR data as explained in sec-
tion  10.1.2, this approach remains to be imprecise. A
promise for accurate estimate of ice thickness from
SAR data resides in using multipolarization or polari-
metric data. Recent studies have been undertaken to
estimate thickness from dual‐polarization SAR data.
Nakamura et al . [2005] used airborne polarimetric SAR
observations (Pi‐SAR) of the Sea of Okhotsk, and
Nakamura et al . [2009] used ENVISAT ASAR data
over Lützow‐Holm Bay, East Antarctica. In both cases
the backscatter from the co‐polarization return HH
and VV were used. Similar to the ice thickness retrieval
from passive microwave, retrieval from SAR data is
based on empirical relations, in this case between the
co‐polarization ratio R vv hh
increases within the thickness range shown in the figures.
From the data in panel (c) in Figure  10.37 and 10.38,
Nakamura et  al . [2005] developed the following regres-
sion equations to estimate ice thickness:
H
202 09091
.
.
R vv hh
fortheXband
(10.92)
/
H
177 97692
.
.
R vv hh
fortheLband
(10.93)
/
and H is in centimeters. The plots in panel (c) show that
the L‐band offers a better correlation between H and
R vv / hh than the X‐band. This is truly reflected in less RMS
error between estimated and observed ice thickness in
panel (d). The estimated thickness shown in this panel
was obtained from equations (10.92) and (10.93). The
RMS errors from using the L‐band and X‐band data are
7.4 and 9.1 cm, respectively. The 95% confidence limits
for the ice thickness estimates using the X‐band and L‐
band were 68 ± 6 5.7 and 66.7 ± 6.0 cm, respectively.
Nakamura et al . [2009] confirmed the same correlation
using backscatter measurements from ASAR onboard
ENVISAT and ice thickness measurements along the
track of the icebreaker research vessel (RV) Shirase in the
East Antarctic during mid‐December 2004. The relation
proved to be valid for both fast ice and consolidated
pack ice. Data are presented in Figure 10.39. The sea ice
thickness and snow depth surveys were conducted using
a downward‐looking camera. Digital images were ana-
lyzed to quantify these two parameters by measuring the
cross section of broken ice pieces. Nakamura et al . [2009]
constructed also a logarithmic regression between the ice
thickness and  R vv / hh :
/ 0 0 and the ice thick-
ness. The rationale behind this correlation is explained
in Nakamura et al . [2005]. The backscatter from YI and
to some extent FY ice is affected by the subsurface
salinity and surface roughness. The salinity determines
the dielectric constant, which governs the amount of the
reflected signal while surface roughness determines
the scattering pattern. The contribution of the surface
roughness can be assumed to be the same for both co‐
polarization returns (
/
vv
hh
0 ). Consequently, any
change in R vv / hh should be attributed to changes in the
dielectric constant or equivalently the salinity of the top
layer of the ice sheet. The salinity decreases as the ice
thickens, causing a decrease in the vertically polarized
return more than the horizontally polarized return,
particularly in the case of YI (nevertheless, this claim
has not been confirmed in other studies).
Nakamura et al . [2005] explored the correlation between
radar backscatter observations from a dual‐frequency
airborne SAR (X‐band and L‐band) against in the situ
ice thickness sampling obtained from the Sea of Okhotsk
in 1997. The SAR system was fully polarimetric. Results
from the X‐band and L‐band SAR are shown in
Figures  10.37 and 10.38, respectively. Each figure has
four panels showing the relevant SAR image, two plots of
backscattering coefficients and co‐polarization ratio
against ice thickness, and a plot of observed versus esti-
mated ice thickness from equations (10.92) and (10.93).
The SAR image covers part of the Sea of Okhotsk off-
shore from Mombetsu, Hokkaido. It is a composite
image in which the RGB colors are assigned to HH, HV,
and VV backscatter, respectively (however, the image is
shown in gray tone). By comparing the SAR images in
the two figures, it is apparent that the L‐band image
shows more details of the surface, which can be a proxy
of the ice thickness. A remarkable observation from the
plots of ice thickness in both figures is that the thickness
is not correlated to the backscatter data, yet a strong
correlation is observed with the co‐polarization ratio
( R vv hh
0
and
hh
vv
HaRa
vv hh
/
(10.94)
1
/
0
H
exp
b Rb
vv hh
/
(10.95)
1
/
0
Sets of coefficients a 0 , a 1 , b 0 , and b 1 for pack ice and
fast ice are listed in Table  10.6. The equations can be
used to estimate the thickness of FY ice up to 2.5 m.
The exponential regression curves are shown also in
Figure 10.39.
Estimation of ice thickness from dual‐polarization
SAR as described above has not been extensively vali-
dated. According to the above‐mentioned rationale, this
approach may be valid as long as the surface salinity
follows a pattern of systematic decrease before being
stabilized. The presence of snow and other unfavorable
ice surface conditions such as frost flowers or surface
wetness may annul the relation between ice thickness
and the polarization ratio. More studies are expected to
be undertaken in the future to explore the potential of
0
0
/
). This ratio decreases as ice thickness H
/
vv
hh
Search WWH ::




Custom Search