Geology Reference
In-Depth Information
size in the near‐ and middle‐infrared regions (0.7-3.0
μ
m)
can be used to retrieve information about grain size.
Langlois et al
. [2012] provided a few references with prom-
ising results from various methods that link near‐infrared
reflectance to specific surface area of snow grains. While
snow grain size causes a decrease of albedo in the optical
region of the spectrum, its effect on the emission in the
TIR region exhibits very little dependence on grain size
(not shown in the figure). It is well known that liquid
water in snow decreases its albedo [
Grenfell and Maykut
,
1977]. Results from the model of
Wiscombe and Warren
[1980] along with measurements from another source
confirm this conclusion (Figure 8.26b). The spectral
albedo of snow decrease when it becomes wet, but there is
no further change when the snow refreezes.
Satellite remote sensing is the only means of providing
large‐scale near‐instantaneous coverage of albedo esti-
mates needed for climate models. In addition to an accu-
rate cloud masking, which is needed to calculate albedo,
the following factors should also be accounted for: (1)
atmospheric contribution to the observed radiance (or
estimated reflectance) at the TOA, (2) the viewing geom-
etry of the sensor, which differs between different pixels
from the same pass or same pixel from different passes,
(3) surface topography and large‐scale roughness, which
affect the reflection and scattering pattern for a given
solar and satellite zenith angle, (4) methods of combining
data with different viewing geometries from several passes
in order to perform the hemispheric integration of reflec-
tance to produce the required albedo. A few algorithms
to retrieve sea ice albedo under clear sky from AVHRR
observations have been developed [e.g.,
Linday and
Rothrock
, 1994;
De Abreu
, [1996]. Later,
Key et al.
[2001]
developed an algorithm to estimate albedo from data
collected under a cloudy sky. A comprehensive review of
albedo estimation from satellite data, including the steps
involved in deriving directional‐hemispherical broadband
clear‐sky albedo of sea ice from NOAA AVHRR observa-
tions, is presented in
Lubin and Masson
[2006]. Since
albedo is determined for a given solar zenith angle, it is
necessary to normalize the results (regardless of the
approach used) to a fixed solar zenith angle. This is usually
selected to be 60° or 70°. Normalization should facilitate
comparisons of data from different locations and times.
The seasonal cycle of clear‐sky hemispherically inte-
grated surface albedo of sea ice in the Arctic was esti-
mated in
Lindsay and Rothrock
[1994] from observations
of the TOA reflectance by AVHRR onboard NOAA‐10
and NOAA‐11 satellites. The data were acquired from
March through September 1989 when the sun was more
than 10° above the horizon. The study provided detailed
descriptions of calibration of observations, correction
for nonisotropic reflection and atmospheric interference,
conversion from narrowband to broadband albedo, and
normalization to a common solar zenith angle (70°).
Monthly averages of sea ice albedo in the central Arctic
were found to decrease from an average of 0.76 in April
to an average of 0.47 in August, with standard deviations
of 0.04 and 0.06, respectively. The data showed a rela-
tively strong correlation between ice surface albedo and
temperature in the spring (correlation coefficient ~0.77 in
March) and less correlation in the summer. A seasonal
cycle is demonstrated in the data yet with considerable
spread of data points.
A few attempts to estimate albedo from SAR data
have been undertaken using one of the following two
approaches. The first involves using the sea ice tracking
product from the Radarsat Geophysical Processor System
(RGPS) in the Alaska SAR Facility (ASF) to infer sea
ice albedo from ice thickness estimates. The RGPS tracks
the motion and geometrical distortion of individual
cells (initially 10 × 10 km
2
) in sequential SAR images
(once every 3 days) using Lagrangian representation
(Figure 10.52). As time advances, the deformation of the
cells is detected, showing possible convergence or diver-
gence of the ice sheet (more description of the RGPS ice
tracking is presented in section 10.7). The system then
estimates the thickness using an empirical relationship
between ice thickness and the freezing degree‐days since
the formation of the ice. With the parameterization of
albedo based on the ice thickness [
Ebert and Curry
, 1993]
and using the snow depth climatology, albedo can be esti-
mated from the ice thickness in each cell. Figure 8.27
shows a map of albedo in the early spring of 1997. The
data cover most of the Beaufort and Chukchi Seas. The
mean albedo is 0.79 with a standard deviation of 0.04.
Lower albedo values are found near the edge of the per-
ennial ice zone (marked by the ellipses in the figure). The
Radarsat‐1 image in the inset can be matched with the
map using the location of Barrow, Alaska (indicated by
the solid circle). It shows ice where there is no albedo data
in the map. The gaps in albedo data, mostly along the
coast in the southern regions, resulted because no ice
had existed in these regions on the initial day to track.
Therefore, it can be assumed that the ice in the corre-
sponding Radarsat‐1 image must have been formed after
that day. This is another shortcoming of the method; i.e.,
albedo can only be estimated if ice had already existed at
the initial day of the tracking. The biggest source of error
in this method is likely the assumed rate of snow accumu-
lation on new ice. However, the merit of the method
remains in its potential to estimate albedo independent
of cloud cover.
In the second approach, empirical expressions that
relate albedo to radar backscatter are used. The approach
is based on the hypothesis that both shortwave and micro-
wave interactions of snow‐covered sea ice are affected
by the same composition and thermodynamic processes
