IEEE Geoscience and Remote Sensing Magazine - June 2017 - 62

Backscattering Coefficient σ 0 (dB)

Coherent
Component

Very Smooth Snow
Moderately Rough Snow
Very Rough Snow

Noncoherent Component
0

30
40
50
Incident Angle (°)

fIgURE 4. A plot of the backscattering coefficient as a function of the

incident angles for three different surface-roughness scales [38].

Normalized Magnitiude (dB)

-5

Strong Target

No Windowing
Hamming

-10
-15
-20

Weak Target

-25
-30
-35
-40
-45
-50
0.996 0.997 0.998 0.999 1 1.001 1.002 1.003 1.004
Normalized Beat Frequency (Hz/Hz)

fIgURE 5. The beat-frequency spectrum of two closely spaced targets with unequal amplitude. The first weak target is masked by the
sidelobes of the second strong target (blue line) unless windowing
is applied (red dashed line).

processing is computationally inexpensive, allowing a fast
data product turnaround time.
Typical snow thickness and annual accumulation can
range from less than 10 cm to approximately 100 cm on
sea ice and land ice. To accurately measure these quantities, FM-CW radars with resolution down to centimeter
scale are needed. The resolution of an FM-CW radar is
determined by the chirp bandwidth, B, and the windowing function used during Fourier-based frequency estimation. The minimum snow thickness, h min , that a radar
can resolve is given by
h min = n W

c
,
2B f r

(9)

where f r is the dielectric constant of snow and n W is the
factor accounting for the reduction in bandwidth due to
62

the use of a windowing function. Despite the reduction
in bandwidth, which implies a proportional reduction in
range resolution, the use of a windowing function is often
necessary during FM-CW radar data processing, because it
can significantly reduce sidelobes, revealing the returns of
weak targets. This is illustrated in Figure 5 where the beatfrequency spectrum of two closely spaced targets that are
three resolution bins apart is shown. The closer target (with
lower beat frequency) is a much weaker target than the primary target and is hidden under the sidelobes of the unwindowed primary target's response. The weak target becomes
discernable when a Hamming window is applied to reduce
the sidelobes. In snow-thickness measurement, the backscattered return from the snow-ice/ground interface is often
stronger than that from the air-snow interface due to the
greater dielectric contrast in the former. The control of sidelobe levels with FM-CW radars is therefore critical to map
the interfaces unambiguously and accurately measure snow
thickness. The effects of radar sidelobes on snow-depth retrieval on sea ice are discussed in further detail in [40].
Once the propagation time delay between the air-snow
and snow-ground interfaces is determined from the radar
data, the physical thickness of the snow is then retrieved
using (2) with an assumed dielectric constant of snow. Because the actual dielectric constant of snow depends upon
the snow density and impurity contents, as discussed in
[22] and [25], the retrieved snow thickness can have an
uncertainty of up to ±6.63% for a ±0.1 gcm−3 variation in
snow density. It is therefore recommended to use an appropriate reference snow density/dielectric constant for precise
snow-thickness retrieval.
Assuming a typical dielectric constant of 1.53 for dry
snow, the latest 2-18-GHz Snow Radar would be able to
discern snow thickness down to 1.14 cm, which is approximately a 3.5 times improvement from the first-generation
Snow Radar system with 4.5-GHz bandwidth. This makes
the measurement of thin snow (<10 cm) feasible.
IMpLEMENTATION AND EVOLUTION
Of THE SNOW RADAR
One of the earliest uses of FM-CW radars for snow-depth
measurement was reported in 1972 by Venier and Cross [41],
in which they demonstrated the mapping of the air-snow
and snow-ground interfaces with an 8-12-GHz system.
Since then, several different UWB FM-CW radars operating at S/C/X/Ku-bands have been proposed and developed
for snow measurement [17], [20], [42]-[44]. Although these
early radar systems showed promising results in snowthickness measurement, they can only be operated over a
short range due to the limitation in the frequency linearity
and sweep rate of the chirp signal over a wide bandwidth.
This restricted the use of these radars to ground-based applications with a very-limited spatial coverage. To manage
the coverage issue, the development of the first long-range,
fast-sweeping airborne UWB FM-CW Snow Radar began in
2003 at KU [17]. Since then, the Snow Radar has undergone
ieee Geoscience and remote sensing magazine

june 2017

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - June 2017