Geoscience Reference
In-Depth Information
2.3 Remote Sensing of Snow Mass
The approaches discussed in Sects.
2.1
and
2.2
have been used to estimate snow clima-
tologies and detect climatic changes, but their continued reliance on intensive in situ
measurements leaves large uncertainties in some regions. This justifies continued devel-
opment of remote sensing products, which can provide global coverage for improved
estimates of snow mass where station density is insufficient.
Beginning with the work of Frappart et al. (
2006
) and continuing with Niu et al. (
2007
)
and others, the Gravity Recovery and Climate Experiment (GRACE) gravimetry mission
has been used to estimate snow mass based on observing changes in Earth's gravitational
field. GRACE responds directly to gravitational changes, suggesting that it should be well
suited to retrievals of deep snow or snow in forested areas where traditional remote sensing
has to see 'through' the trees. However, further modelling is required to control for other
changes in mass of the land surface associated with, for example, other forms of terrestrial
water storage. Additionally, GRACE is not suitable for high-resolution measurement with
Frappart et al. (
2006
)'s reported resolution being 660 km. Finally, it is not yet appropriate
for assessing long-term changes as the GRACE satellites were only launched in 2002.
As such, efforts for the remote sensing of snow mass have typically focussed on the passive
microwave regime, using frequencies near 19 and 37 GHz, for which there has been continuous
near-global coverage since the launch of the Scanning Multichannel Microwave Radiometer
(SMMR) on Nimbus-7 in late 1978. Many snow products typically utilise the Special Sensor
Microwave Imagers (SSM/I) (e.g., Tedesco et al.
2004a
) and/or the Advanced Microwave
Scanning Radiometer-Earth Observing System (AMSR-E) (Tedesco et al.
2004b
).
When observing a typical snowpack, the majority of radiation measured at these
wavelengths will have originated from the ground surface, with scattering within the snow
the dominant loss mechanism. This scattering is frequency dependent and increases with
the quantity of snow, allowing a determination of SWE from the difference between the
brightness temperatures in these two channels.
Figure
1
shows simulations of the brightness temperatures over a snowpack at 18.7 and
36.5 GHz horizontal polarisations viewed at 53. Snow is assumed to be a homogeneous single
layer with properties based on those typical of Colorado snowpacks of under 120 cm depth
discussed in Davenport et al. (
2012
). As the amount of snow increases up to 500 mm SWE, the
brightness temperature at both frequencies falls, but it falls more quickly at the higher frequency.
By considering the difference in brightness temperatures between the two frequencies,
the effect of absolute temperature change is reduced and this led to the simplest approach
to SWE retrieval, often called the Chang Algorithm, which was originally developed for
SMMR (Chang et al.
1987
), a general variant of which is as follows:
SWE
¼
AT
B
19H
T
B
37H
þ
B
ð
Þ ¼
A
ð
DT
B
;
H
þ
B
Þ
ð
1
Þ
where A and B are constants depending on the exact frequency of the channel and snow
properties, T
B
19H and T
B
37H are the recorded brightness temperatures at the available
channels nearest 19 and 37 GHz horizontal polarisation. Figure
1
shows this equation fit to
the first 100 mm SWE, and for this snow, the values are A = 2.54 mm
SWE
K
-1
and B = 3K.
Passive microwave measurements offer the advantage of being largely independent of
illumination conditions, precipitation or cloud cover, allowing night time measurements
when temperatures are likely to be lower and moisture within the snow is more likely to
have refrozen. However, the range of values which can be reliably sensed is limited at the
lower end by sensor precision, and, at higher values of SWE, the signal saturates (displayed
