Geoscience Reference
In-Depth Information
retrievals. The multilayer implementation of the HUT model (Lemmetyinen et al.
2010
)
simulated the brightness temperatures at 18.7 and 36.5 GHz vertical polarisation for each
snowpit with each of the layering structures described in Sect.
4.1
. The brightness tem-
perature difference
DT
B
;
V
¼
T
B
19V
T
B
37V
ð
12
Þ
was considered throughout, for consistency with the Globsnow product. This offers the
further advantage of being less sensitive to errors in ground or exposed vegetation
temperatures.
In addition to the HUT simulations, a Chang algorithm output based on Eq. (
1
) and
using the CLPX coefficients reported in Sect.
2.3
was produced for each snowpit.
In each of the CLPX mesoscale study areas (MSAs), the snowpits were assumed to be
representative of the actual snowpack, such that the mean DT
B
;
V
of all of the snowpits
within the MSA represents the DT
B
;
V
contribution of the snow within that MSA. The scene
brightness temperature has contributions from snow, open ground and vegetation over
snow.
DT
B
;
V
;
scene
¼
A
open
DT
B
;
V
;
open
þ
1
FF
ð
Þ
A
snow
DT
B
;
V
;
snow
ð
13
Þ
where A is the fractional area of the pixel that is either open or snow covered, and FF is the
forest fraction. With the assumption that DT
B
;
open
¼
0, the equation becomes
DT
B
;
V
;
scene
¼
1
F
ð Þ
A
snow
DT
B
;
V
;
snow
ð
14
Þ
Since the distribution of the snowpit properties is assumed to match the distribution of
the snow within the scene, then the brightness temperature difference of the snow should
be equivalent to the average brightness temperature difference of the N snowpits.
X
N
DT
B
;
V
;
scene
¼ð
1
FF
Þ
A
snow
N
DT
B
;
V
;
i
ð
15
Þ
i
where DT
B
;
V
;
i
is the brightness temperature difference between 18.7 and 36.5 GHz at
vertical polarisation for the i
th
snowpit.
Snow properties were assumed to be static throughout an Intensive Observation Period
(IOP) such that all measurements within each IOP could be used in the same analysis.
The fractional area of snow for each IOP and for each MSA was estimated by using the
8-day maximum extent snow cover map from MODIS, taken as the fraction of snow-
covered area divided by the total non-cloud-covered area. This offers the advantage of
minimising the effect of cloud cover, although can provide inconsistent results if signifi-
cant snowfall or melt occurs during the 8 days. The MODIS product is at 500 m spatial
resolution, so features 2,500 pixels within each 25-km passive microwave grid point.
Forest cover was estimated for each MSA using QuickSCAT data available from
Nilsson (
2003
) and the forest correction factor applied individually for each MSA. More
complex forest correction approaches exist, but are not adopted here.
Six sets of simulated scene DT
B
;
V
values were produced, for each of the layering
structures (1- to 5-layer plus the N-layer truth), and these were compared with SSM/I and
AMSR-E values, where all measurements within a day of each IOP period were recorded.
It should be noted that for IOP4, many snowpacks reported temperatures around the melt
point, suggesting the presence of liquid water, which acts to reduce the brightness tem-
perature difference through greater absorption and emission at both wavelengths. However,
