Geology Reference
In-Depth Information
Table 10.5
Coefficients of the empirical equation (10.81) that relates ice thickness to the polarization ratio from passive
microwave.
Reference
Sensor
Region
Freq.
α
β
γ
Tamura and Ohshima
[2011]
SSM/I
Arctic polynya
85 GHz
215.15
0.508
1.0395
SSM/I
Arctic polynya
37 GHz
88.40
1.023
1.113
Iwamoto et al
. [2013]
AMSR‐E
Polynya & MIZ
36 GHz
206
−5.4
1.02
AMSR‐E
Polynya
36 GHz
40
1.8
1.14
AMSR‐E
MIZ
36 GHz
300
−11.1
1.01
AMSR‐E
Polynya & MIZ
89 GHz
218
−3.0
1.03
AMSR‐E
Polynya
89 GHz
103
1.1
1.07
AMSR‐E
MIZ
89 GHz
298
−6.7
1.02
ice thickness less than 0.075 m. For the rest of the pixels
equation (10.80) is used. If the calculated thickness
from equation (10.80) is < 0.075 m, the thickness is set to
be 0.075. This means that the algorithm has discontinuity
at
h
≤ 0.075. The same equations, but with different coef-
ficients suitable for thin ice in the Antarctic Ocean, were
used in
Tamura et al.
[2007].
Alternatively, the ice thickness can be calculated using
the equation for the exponential curve fitting of the data,
which is given by
Martin et al
. [2004]:
OW filter should be applied to reject pixels that have domi-
nant OW coverage, namely with ice concentrations < 30%.
Those pixels were identified based on results from an ice
concentration algorithm. The thin ice thickness maps
from AVHRR and SSM/I in the figure generally agree
with two exceptions. The first is the higher ice thickness
estimate from SSM/I at the opening of Jones Sound into
the Baffin Bay. The second is the larger area of ice thick-
ness definition in the AVHRR data. This can be partly
explained by the finer resolution of the AVHRR, but it
can also be caused by the rejection of the low ice concen-
tration pixels from the calculations using SSM/I data. A
similar observation is made in
Iwamoto et al
. [2013] who
compared ice thickness retrieved from the TIR channels
on MODIS against retrievals from AMSR‐E data (using
the 89 GHz channel) in the Chukchi Polynya. While the
ice surface temperature map in Figure 10.31 shows clearly
the northern boundary of the polynya where the sudden
temperature change marks the ice arch (cold MYI and
thick FYI coexist upstream while thinner ice types exist
downstream of the arc), it does not show the southern
boundary between the polynya and the open water (below
latitude 77°N). On the other hand, ice thickness maps
show clearly the entire boundary of the polynya.
Although microwave data are not affected (in general)
by atmospheric contents such as cloud liquid water,
water vapor columns, and fog, the higher frequency
channels (85 or 89 GHz) are affected to some extent.
Therefore, any algorithm that uses one of these two
channels must incorporate a filter to reject pixels with
adverse atmospheric conditions. A standard method has
been used in previous studies [e.g
Martin et al
., 2004;
Tamura et al
., 2007] based on the examination of scat-
terplots of PR
36
and PR
89
of pixels under clear‐ and
cloudy‐sky conditions. An example of such a plot is pre-
sented in Figure 10.32, following
Iwamoto et al
. [2013].
The clear‐sky data are shown in black and the cloudy‐
sky and water‐vapor‐rich atmospheric data are shown in
red. Data from the two groups overlap but the second set
has smaller PR
89
. This means that the ice thickness tends
to be overestimated on days that have clouds or high
H
exp1
PR
(10.81)
This equation is applied with two sets of coefficients: one
for PR
85
from SSM/I (or PR
89
from AMSR‐E) and the
other for PR
37
from SSM/I (or PR
36
from AMSR‐E). The
coefficients are shown in Table 10.5 using regression of
data sets from the study of
Tamura and Ohshima
[2011]
and another study by
Iwamoto et al
. [2013]. The latter
was conducted on thin ice in the Chukchi Sea polynya,
and the least square fitting was conducted against ice
thickness derived from MODIS data using the same
technique of
Yu and Rothrock
[1996]. The resulting thick-
ness from equation (10.81) is in meter. As mentioned
above, the coefficients are specific for each region and for
each pair of TIR data from which thickness has been
retrieved. A new set of coefficient must be established
if this approach is to be adopted for a different region or
different sensors.
Tamura and Ohshima
[2011] presented a comparison
between thin ice thickness retrieved from SSM/I using
equations (10.79) and (10.80) and the corresponding esti-
mate using the TIR technique described above. Results
were obtained from data acquired over the North Water
polynya on 30 March 1998 (Figure 10.31). The AVHRR
data are mapped onto the SSM/I grid in the figure. The
surface temperature derived from AVHRR is included
as a validation tool for the ice thickness results. The
authors reported that the linear and exponential equa-
tions [(10.79)-(10.81)] produced very similar results. An
