Geology Reference
In-Depth Information
Chukchi, and East Siberian Seas) is noticed. The reduc-
tion spread rapidly starting from day 158 to cover the
entire ice region by day 179 (note the fast rate of surface
melt between days 158 and 161). In addition to the inter-
pretation of the reduction of backscatter in terms of pro-
gression of the snowmelt,
Forster et al.
, [2001] pointed
out that the occurrence of this progression at large scale
appears to be related to a synoptic weather system moving
through the area. However, no quantitative link was estab-
lished. The authors compared results from their NSCAT
algorithm against the passive microwave algorithm of
Anderson
[1997] and found that the NSCAT algorithm
predicts the snowmelt onset 1-10 days earlier than the
SSM/I‐based algorithm. In general, detection of snow-
melt onset is more robust in the case of MY ice than FY
ice because the contrast between the premelt and the melt‐
onset
σ
0
is much higher for MY ice.
In a more recent study,
Howell et al.
[2008a] used scat-
terometer data from QuikSCAT to estimate sea ice melt
onset, freeze, and melt duration within the Canadian
Arctic Archipelago for the period 2000-2007. They used
the same approach of setting a threshold on the backscat-
ter, but they treated snow on FY ice and MY ice differ-
ently. With some justification they further validate the
known fact that
σ
0
from snow‐covered FY ice would
increase when the snow becomes wet, a condition that
could occur when the snow‐ice interface temperature
reaches −5 °C. On the contrary,
σ
0
from snow‐covered
MY ice would decrease under the same condition (see the
example from QuikSCAT in Figure 7.46).
Howell et al.
[2008a] supported this argument using observations from
test sites and used this concept to identify the onset of
snowmelt. In their algorithm the melt onset is estimated
for FY ice and MY ice using a threshold of absolute
change in
σ
0
of greater than 2 dB from stable winter con-
ditions. This was based on satellite measurements of tem-
poral evolution of
σ
0
from numerous sites within the
Canadian Arctic Archipelago. The authors reported that
changing the threshold by ±1 dB results in a change in the
date of the snowmelt onset by only 1 or 2 days, which is
considered a small error.
Drinkwater and Liu
[2000] mapped the spatial pattern
of the timing of austral summer melt onset in Antarctic
sea ice between 1992 and 1997 using scatterometer data
from the C‐band (VV) active microwave instrument
onboard ERS‐1 and ERS‐2 and the Ku‐band (VV and
HH) NSCAT. The authors stated that ice surface melt
ponds had rarely been observed in the Antarctic because
the summer air temperature would rise only sporadically
above 0 °C. Antarctic sea ice typically retains a snow
cover year‐round. A comprehensive review of snow on
Antarctic sea ice is presented in
Massom et al.
[2001].
Drinkwater and Liu
[2000] also calculated the difference
in backscatter at each pixel between consecutive images
(a few days step). For each pixel, the melt onset date was
determined when the backscatter value decreased from
one image to the next by an amount > 0.5 dB and the total
decrease within the seven consecutive images window
was > 3 dB. The authors concluded that atmospherically
driven surface melt in the Antarctic is far less extensive in
time and space than in the Arctic. They also stated that
“atmospherically induced surface melting and resulting
albedo feedback is not the dominant sea‐ice removal
mechanism. Instead, a combination of ocean heat flux
and summer short‐wave radiation absorbed by the ocean
surface (in small lead fractions) rapidly disposes of most
of the sea‐ice cover (p. 1841).”
9.3.4. Airborne Photography
Aerial photography is another commonly used tool to
detect and quantify ice surface melt when it reaches the
melt pond phase. Digital and video cameras mounted on
aircrafts [
Rothrock and Thorndike
, 1984;
Tschudi et ai.
,
2001] or helicopters [
Holt and Digby
, 1985;
Perovich
et al.
, 2002;
Peng et al.
, 2010] with a motor drive to adjust
the viewing angle have been used to survey the field.
Depending on the camera's lenses, a photograph that
covers roughly 1 km
2
can be acquired from an altitude of
2000 m. Photographs can be digitally processed to parti-
tion the scene into solid ice surface and pond surface
(or leads). The advantage of using aerial photography
is its very fine resolution that allows capturing a wide
range of ponds on a spatial scale from a few meters to
hundreds of meters. This also facilitates the tracking of
pond evolution and the determination of their areal
fraction and density. The disadvantage is obviously the
very limited coverage and the high cost.
To enable quantification of the evolution of the ponds
throughout a melt season, a program of aerial photogra-
phy was carried out at the main site of the surface heat
budget of the Arctic Ocean (SHEBA) in the Beaufort Sea
between mid‐May through early October 1998 [
Perovich
et al.
, 2002]. Helicopter‐based digital and video cameras
were used to acquire images of the ice surface. The images
were processed to partition the contents into four surface
categories: snow cover and bare ice, ponded ice, newly
formed young ice, and open water. Partitioning was per-
formed based on selection of color thresholds derived
from the color distribution histograms of the images.
Melt pond has typically a bluish appearance, which
makes it distinct from the other three categories. The
thresholds for the four categories were subjectively
selected, but a quality check of the results was performed
to ensure that the sum of the area percentages for each
feature was nearly within 2% of the ideal 100%. Based on
the image processing, the authors concluded that melt
ponds started in early June and grew rapidly in size to
