Geology Reference
In-Depth Information
that the retrieval of any parameter of thin ice (<5 cm thick)
based on the polarization difference (e.g. ice concentra-
tion) will be in error if the data are associated with an event
of fresh snowfalls (the large polarization difference will
make the thin ice appear as open water).
In principle, the effect of snow depth on microwave emis-
sion cannot be correctly characterized without taking into
consideration the characteristics of the snowpack (e.g.,
snow wetness, stratification, grain size, etc.). That is why
past attempts to explore correlation between field measure-
ments of microwave emission and snow depth yielded con-
flicting information as they are not always accompanied
with full description of the snow conditions. For example,
Lohanick and Grenfell
[1986] found no correlation between
microwave emission and snow depth from measurements of
brightness temperature at 37 GHz conducted over cold FY
ice of thickness up to 30 cm near Tuktoyaktuk, Northwest
Territories. On the other hand,
Markus and Cavalieri
[1998]
presented correlation between brightness temperatures
from 37 GHz, obtained from SSM/I, with insitu snow
depth measurements from the Bellingshausen‐Amundsen
Sea in the Antarctic. The correlation coefficient over a
depth up to 80 cm was −0.64 and −0.66 for observations
from vertical and horizontal polarization, respectively.
These correlations were confirmed in later studies. These
results have led to the general acceptance, though with some
reservations, of the inverse relationship between microwave
brightness temperatures and snow depth (mostly linearly).
The relationship applies to both horizontal and vertical
polarizations. However, the rate of decrease of brightness
temperature is higher for higher frequencies because the
effect of scattering is also higher. This means that bright-
ness temperatures at 37 GHz are reduced more than bright-
ness temperatures at 19 GHz with increasing snow depth.
The implication is that the gradient ratio between 37 and
19 GHz channels GR
37
V
19
V
[defined in equation (8.11)] can
be used as an indicator of snow depth. This ratio is always
close to zero for bare ice surface. Scatterplots that depict
this relationship are presented in
Markus and Cavalirie
[1998]. A few forms of this relationship can be used to esti-
mate snow depth as presented in section10.6. However, one
has to bear in mind that accurate retrieval hinges upon
favorable properties of the snowpack such as grain size,
hoar layer structure, presence of slush, and ice layering.
These properties may alter the scattering pattern and there-
fore compromise the accuracy of the thickness retrieval.
In a modeling study by
Powell et al
. [2006], the authors
presented results of variation of GR
37
V
10
V
and GR
19
V
10
V
versus snow depth using the Microwave Emission Model
of Layered Snowpack (MEMLS) [
Mätzler and Wiesmann
,
1999]. Simulations were conducted using different snow-
pack densities, correlation lengths (a measure of grain size),
and snow depths but for a constant surface temperature
of 250 K. Results are shown in Figure 7.43. The decrease
of gradient ratio with snow depth is confirmed but only
up to a depth of approximately 30 cm for GR
37
V
10
V
and
100 cm for GR
19
V
10
V
. It means that the latter parameter
can be suitable for estimating snow depth over a wider
range but, of course, at a much coarser resolution. The
useful trend of the gradient ratio decreases with snow
depth and is valid only for snow with relatively large grains.
For very fine grains (correlation length is near zero), the
gradient ratio does not change with snow depth. This is
the case of the fresh dry snow. Therefore, the data provide
another manifestation of the transparency of fresh snow
to microwave radiation. Snow thickness cannot be esti-
mated from the gradient ratio of microwave observations
in this case. The plots in Figure 7.43 show also the higher
sensitivity of the gradient ratio to the snow grain size
(compared to the snow density) for the same snow depth.
In general, less negative gradient ratios correspond to less
snow depths, smaller grains, and/or higher densities.
Grain size is by far the most important parameter control-
ling microwave emission and radar scattering from dry snow.
When snow is exposed to relatively high temperature, smaller
grains tend to melt before larger ones. When wet snow
refreezes, the mean grain size increases not only because
smaller grains had already been lost but also because the
existing grains bond together through enhanced packing. In
addition, water that percolates through the snow and
refreezes can form ice lenses and ice glands. Ice lenses and
thick crust cause the microwave emissivity to drop to 0.74
compared to 0.94 at 31.4 GHz [
Yan et al.
, 2008]. As a rule of
thumb, for fine‐grained snow (relative to the wavelength of
the incident signal), absorption and scattering losses are typ-
ically small while for coarse‐grained snow the volume scat-
tering becomes significant. Depending on the incidence
angle of the receiver, scattering may or may not cause an
increase in the brightness temperature, but it almost cer-
tainly increases the radar backscatter. Due to the lack of
field measurements of snow grain size, most of the informa-
tion about the effect of this parameter on microwave signal
has been generated using a radiative transfer model of snow,
whether or not combined with sea ice, the ocean, and the
atmosphere. For example,
Fuhrhop et al.
[1998] found that
brightness temperature increases nonmonotonically with
snow grain size. Other studies that show the sensitivity of
brightness temperature to grain size include
Li et al.
[2006],
Andreadis et al.
[2008],
Durand et al.
[2009],
Kontu and
Pulliainen
[2010], and
Derksen et al
. [2012]. A main conclu-
sion from these studies is that most of the differences
between observed and measured brightness temperatures
from snow can be explained in terms of snow grain size.
The effect of snow grain size on radar backscattering
has been studied by
Du et al
. [2010] using a multiple scat-
tering model in which the snowpack is divided into multi-
ple layers while the interaction between scattering within
each layer is considered. A sample of the results, com-
pared to results from first‐order and second‐order scatter-
ing models, is presented in Figure 7.44. Backscatter from
