Geology Reference
In-Depth Information
The difference between the two parameters varies between
−4.6 and 3.2 °C and the correlation coefficient is 0.96.
The accuracy from using equation (7.53) was found to be
the same. These initial results confirmed the possibility of
using the 11
μ
m TIR channel for estimating IST using a
simple equation provided that no atmospheric influence
is “contaminating” the observed brightness temperature.
Usually the 11
μ
m channel is not affected by the atmos-
phere as much as the 12
μ
m channel.
Comparison between brightness temperature observa-
tions from 11 and 12
μ
m channels or the derived surface
temperature from each channel using any method pro-
vides an acceptable assessment of the effects of the
atmosphere. If the difference is smaller than 0.39 °C, then
clear atmosphere can be assumed [
Riggs et al.
, 1999]. In
the presence of atmospheric influences, a more accurate
estimate of the IST requires correction of the satellite
observations for attenuation of the observed radiation.
This is caused mainly by water vapor and partly by sev-
eral atmospheric constituents (e.g., carbon dioxide, nitro-
gen oxide, methane, ozone). Obviously, this issue becomes
more serious over the tropical ocean than polar ocean.
A commonly used technique that corrects for atmos-
pheric influences in calculating surface temperature is
called the “split window.” It was originally suggested by
Prabhakara et al.
[1974] and first used by
McMillin
[1975]
to obtain accurate estimates of land surface temperature.
The technique estimates surface temperature using the 11
and 12
μ
m channels because the 12
μ
m is more sensitive to
the water vapor than the 11
μ
m channel. Therefore, the
difference between them is a function of the absorptive
response of the water vapor content in the atmosphere.
The more water vapor contents the higher the difference
(or the ratio) between observations. The split‐window
technique incorporates a linear or a nonlinear equation
that combines the observations from the two channels to
determine the surface temperature
T
s
. Several studies
conducted in the early 1990s employed different forms of
the split‐window equation using TIR data from AVHRR
and other instruments [
Key and Haefliger
, 1992;
Lindsay
and Rothrock
, 1993;
Massom and Comiso
, 1994;
Yu et al
.,
1995]. The following form was used in
Key et al
. [1997]. It
provides better accuracy because it accounts for the
atmospheric effects as well as the sensor's scan angle:
between 0.5 and 1.5 K, depending on the accuracy of
determining the coefficients [
Key et al.,
1997].
The coefficients in equation (7.55) are usually determined
using measurements of
T
s
regressed against estimates from
using the RHS of the equation. Surface temperature
T
s
can
be obtained from surface measurements, and the
T
11
and
T
12
can be obtained from satellite measurements after correc-
tions for atmospheric influences. Alternatively, the coeffi-
cients can be obtained using modeled brightness temperature.
A commonly used radiative transfer model for this purpose
is LOWTRAN. This is a computer code for predicting
atmospheric radiance and transmittance for a given ice
surface temperature, composition, and the characteristics
of local atmospheric influences [
Kneizys et al.,
1989]. The
input atmospheric parameters are usually obtained from
measurements using a radiosonde or a drifting ice station.
The model also needs a value for local surface emissivity
at the point where brightness temperature is to be estimated.
The emissivity is often assumed, but it can be a source of
error in the output IST if the surface includes a highly
compositional mix of ice and metamorphozed snow.
Climatological emissivity data can be used though the data
are less commonly available in TIR than in microwave.
When IST is derived from TIR data it can be used to
estimate the thickness of thin ice. The principle idea is
that the ice surface temperature becomes a function of ice
thickness during the early stages of ice formation. When
the thickness exceeds 1 m, the relationship ceases and the
surface temperature becomes almost equal to air temper-
ature.
Groves and Stringer
[1991] made a first attempt to
explore this approach for estimating the ice thickness in
the Chukchi Sea polynya off the northwest coast of
Alaska during 1989. They used AVHRR TIR data to cal-
culate the IST and a simple expression developed by
Maykut
[1986] based on empirical observations to calcu-
late ice thickness from the IST:
kT T
CT T
si
(
)
f
s
(7.56)
h
i
(
)
t
s
a
where
k
si
is the thermal conductivity of sea ice,
T
s
is the
estimated surface temperature,
T
a
is the air temperature,
T
f
the seawater freezing temperature, and
C
t
is a bulk
transfer coefficient describing turbulent heat transfer
between the ice and atmosphere, which is determined to
be 50 cal/cm · d · °C. This equation produces results that
appear to be realistic for thickness of newly formed ice
that grows under steady oceanic and thermal conditions
with snow‐free surface. However, these conditions are not
usually encountered in nature. Auxiliary data such as air
temperature, surface emissivity and perhaps passive
microwave observations are usually needed to estimate
the ice thickness using theoretical predictive models.
)
TabT CT TdTT
s
(
)
(
)(sec
1
11
11
12
11
12
(7.55)
where
T
11
and
T
12
are the measured brightness tempera-
tures in the 11 and 12
μ
m channels, respectively,
θ
is the
sensor scan angle (at each pixel) that accounts for the
path length of the radiation in the atmosphere, and
a
,
b
,
c
, and
d
are regression coefficients. Accuracy of surface
temperature estimation from the above equation ranges
