Geology Reference
In-Depth Information
-160°
-140°
0.0
0.6
70°
Pond Fraction
Clouds
Alaska
80°
Figure 9.15 Reflectance from channel 1 of MODIS (left) and the derived pond fraction using equation (9.14)
(right). Data are for 13 June 2004 projected in 500 m EASE‐grid format [ Tschudi et al ., 2008, Fig. 3, with permission
from Elsevier]. (For color detail, please see color plate section).
9.3.2. Passive Microwave Observations
If it is greater than 4 K at a given point, then a dry snow
winter condition can be assumed. If it is less than −10 K,
then liquid water is assumed to be present in the snow-
pack and the algorithm classifies the day as a possible
snowmelt onset date. If the difference is between 4 and
−10 K, the algorithm determines if snowmelt onset has
occurred based on a 20 day time series of the difference.
In this case the maximum and the minimum differences
are determined within the 10 days prior to the potential
melt onset date, and the difference between these two
extremes is calculated. The same process is performed
for the 9 days after the potential melt date. The former
number is subtracted from the latter number, and if the
difference (Diff) is greater than 7.5 K (the threshold is
based on empirical data), the algorithm assigns melt
onset to that particular day. The process can be formu-
lated in the following equation:
Several approaches exist to determine the onset of
melt, advanced melt, and refreeze of Arctic sea ice from
satellite passive microwave data. Most of these approaches
utilize thresholds on brightness temperature or derived
parameters. Kunzi et al . [1982] found that the presence of
liquid water in the snow decreases the difference between
brightness temperatures ( T b ,18 h T b ,37 h ). Anderson [1997]
developed an algorithm, termed the horizontal range
(HR), to detect the onset of snowmelt by monitoring this
difference in temporal records from SMMR and SSM/I
channels. The study suggested that the difference changes
from positive in the case of dry snow in winter to zero
or negative at the onset of snowmelt.
Froster et al. (2001) supported this suggestion based on
the fact that the smaller wavelength of the 37 GHz chan-
nel experiences more scattering from the overlaying snow,
and therefore produces less brightness temperature.
However, it is worth noting that other studies found this
difference (in K) is negative for dry snow or ice surface in
winter ( Shokr et al ., 2009).
Instead of using a single threshold on ( T b ,18 h T b ,37 h ) Drobot
and Anderson [2001] improved the algorithm of Anderson
[1997] by using another scheme of monitoring the difference
between the  brightness temperature from the 18 GHz
SMMR (or  19 GHz SSM/I) and the 37 GHz. The method,
called advanced horizontal range algorithm (AHRA),
proceeds as follows. The difference Δ T b is calculated daily:
Diff max in
(
2
T
2
T
)
(
max in
1
T
1
T
)
b
b
b
b
(9.16)
where the prefixes max 2 and min 2 are the maximum
and  minimum of Δ T b over the subsequent 9 days and
min 2 and min 1 are the corresponding extremes over the
preceding 9 days.
Results from this method compared favorably with
other methods developed earlier to retrieve snowmelt
onset from passive and active microwave data. Drobot
and Anderson [2001] show an interesting map of the 20
year snowmelt onset date (1979-1998) in the Arctic
(Figure 9.16). Snowmelt begins at the ice edge between
TT T
b
(
)
(9.15)
b
,
18
h
b
,
37
h
Search WWH ::




Custom Search