Geology Reference
In-Depth Information
requires the incorporation of a module to exclude areas
with cloud cover from the calculations. Cloud filtering
is difficult during the dark polar season. Atmospheric
correction must also be applied before proceeding with
ice thickness retrieval. The ice surface must be dry and
snow‐free. A filter to exclude areas of thick ice (>50 cm
think) must also be introduced. In general, ice thickness
retrieval using TIR is a lengthy process and requires input
of a few meteorological as well as ice and snow physical
parameters. Since this is a physics‐based approach, it should
produce more accurate results if the assumptions for its
application are fulfilled.
2. Empirical approaches are used to estimate ice
thickness with an upper limit of 15-20 cm from passive
microwave sensors. This is possible from all operational
microwave frequencies in the range 18-89 GHz.
However, by using the L‐band sensor onboard the
SMOS satellite (1.4 GHz), a higher upper limit on ice
hickness (50-150 cm) can be reached. In all cases the
ice surface must be snow‐free and an open‐water filter
must be applied to reject ice‐free pixels. Atmospheric
correction must be applied only if observations from a
high‐frequency channel (85 or 89 GHz) are used. The
spatial resolution of the output thickness maps is rela-
tively coarse, which makes this approach not suitable
to retrieve thin ice thickness in narrow openings such
as polynyas and leads. On the other hand, the areal
coverage is wide (up to 2400 km), which is considered
to be an advantage. This approach is simple. However,
it is this simplicity that may render it not robust enough
because the results depend on the established coeffi-
cients in the empirical equations. These coefficients
vary between different data sets and depend on the
region, season, and sensor characteristics. Another
important limitation of this empirical approach is that
it can only be used to estimate ice thickness during the
ice growth phase but not during decay phase. It is pos-
sible to establish a relation between ice thickness and
emissivity (in the case of passive microwave) or die-
lectric constant (in the case of active microwave) as
ice grows because of the intervening process of brine
drainage during this phase. This possibility ceases,
however, during ice decay as the dominant processes do
not lend themselves to systematic thickness‐dependent
change of emissivity or dielectric constant. The pro-
cesses include, for example, surface flooding, snow wet-
ness, and snow metamorphism. The ice surface becomes
also highly heterogeneous within the sensor's footprint
during the premelting and melting periods.
For altimeter data (radar or laser), which use the
buoyancy law to infer ice thickness from its freeboard
measurements, a major source of error is the inaccurate
estimate of ice freeboard and the snow depth (in the case
of laser altimeter the freeboard includes the snow on ice).
Clouds and atmospheric constituents should also be
accounted for in the case of laser altimeter. Ice thickness
using altimeter data are generated at poor temporal and/or
spatial resolutions (a few tens of kilometers). However,
thickness maps of the entire polar region can be produced
from composite passes acquired over a few months. The
maximum thickness that can be estimated using altimeter
data is around 5 m. The advantage of this approach is the
wide range of the estimated thickness (though with signifi-
cant insensitivity to small thickness) and the wide coverage
of the thickness maps.
The above limitations indicate that no sensor can pro-
vide ice thickness data that simultaneously satisfy the
operational ice monitoring and the climate modeling
requirements. Requirements of the operational ice mon-
itoring entail mapping of relatively thick ice (>30 cm) at
a spatial scale of a few hundred meters with daily tem-
poral resolution in highly marine traffic areas. Climate
modeling and monitoring requirements, on the other
hand, entail mapping a wide range of ice thickness from
a few centimeters up to nearly 10 m thick at an accepta-
ble spatial resolution of a few kilometers or tens of
kilometers and temporal resolution of a few days or one
week. Regional or global maps are required for this
purpose. This product requires an integrated approach
of remote sensing data perhaps with the addition of
modeled meteorological data.
The material in this section is strictly relevant to
retrieval of ice thickness using airborne or space‐borne
observations. It is worth noting, however, that spatial
profiles of ice draft have been measured using submarine
sonar systems (ice thickness is 10%-15% of ice draft).
This approach was used in a few studies [
Rothrock et al
.
1999;
Wadhams and Davis
2000;
Winsor,
2001] to deter-
mine the interannual variability of Arctic ice thickness
using data from submarine cruises conducted between
1991 and 1997. Additionally, progress has been achieved
in modeling ice thickness at synoptic scale.
Zhang and
Rothrock
[2003] present the thickness and enthalpy dis-
tribution (TED) sea ice model that determines ice thick-
ness and explicitly simulates sea ice ridging. This is a
dynamic‐thermodynamic model that employs a three‐
layer thermodynamics and a viscous‐plastic rheology for
ice dynamics. The model estimated FY ice thickness in
the Arctic to be 1-3 m thick and in the Southern Ocean
to be 1-2 m thick. One of the interesting simulation
results is that the thickest ridged ice is 20 m in the Arctic
and 16 m in the Southern Ocean.
Miller et al.
[2005]
examined another model of sea ice (Los Alamos sea ice
model) to compare its output thickness/draft in the
Arctic Ocean with submarine observations during cruises
between 1987 and 1997. They found that the model
overestimated the thickness in the Beaufort Sea and
underestimated it near the North Pole. They pointed out
the need for a better representation of sea ice rheology
on the continuum scale.
