Mac2006b

Mac2006b
Click
Here
GEOPHYSICAL RESEARCH LETTERS, VOL. 33, L13503, doi:10.1029/2006GL026576, 2006
for
Full
Article
Strong spatial variability of snow accumulation observed with
helicopter-borne GPR on two adjacent Alpine glaciers
H. Machguth,1 O. Eisen,2,3 F. Paul,1 and M. Hoelzle1
Received 12 April 2006; revised 23 May 2006; accepted 7 June 2006; published 13 July 2006.
[1] This study compares high-resolution helicopter-borne
radar measurements to extensive ground-based profiling of
the snow cover on Findel- and Adler Glacier, Switzerland.
The results demonstrate that derived accumulation values of
either method are well in accordance. The spatial distribution
of radar based snow depth allows a clear distinction of three
zones of different accumulation characteristics: (1) The lower
part of Findel Glacier shows a clear altitudinal trend while (2)
the upper part has no trend in altitude but high spatial
fluctuations in snow depth. (3) Adler Glacier’s accumulation
characteristics are similar to zone (2). However, despite their
close vicinity, accumulation on (3) is reduced by 40%
compared to (2). The observed strong spatial variability
emphasises the need for spatially continuous measurements
for studies involving accumulation on glaciers. Finally,
reasons for observed variations (e.g., preferential snow
deposition and snow redistribution) are discussed.
Citation: Machguth, H., O. Eisen, F. Paul, and M. Hoelzle
(2006), Strong spatial variability of snow accumulation observed
with helicopter-borne GPR on two adjacent Alpine glaciers,
Geophys. Res. Lett., 33, L13503, doi:10.1029/2006GL026576.
1. Introduction
[2] Utilisation of radio-echo sounding techniques to determine past and present accumulation rates has become a
standard method, especially in polar glaciology. On polar
ice sheets and polythermal glaciers this method is rewarding
because of the low absorption of electromagnetic energy in
cold firn. On temperated glaciers, application is usually
restricted to sounding of ice thicknesses with low frequencies (<100 MHz). High-resolution measurements are mostly
performed with so-called ground-penetrating radar systems
(GPR) [e.g., Richardson et al., 1997; Kohler et al., 1997],
operating directly from the surface. They are thus relatively
time consuming and the spatial coverage is limited to
accessible areas. Application of high-resolution airborne
radar, capable of mapping annual accumulation, is still rare
and mostly limited to fixed-wing aircraft [e.g.,
Kanagaratnam et al., 2004]. To cope with the difficulties
imposed by measurements in mountainous terrain and
valley glaciers, airborne radar is most suitable. Apart from
quasi-airborne measurements from an aerial tramway
[Yankielun et al., 2004], helicopter-borne radar provides the
most reasonable platform and has been applied for studies of
1
Geographisches Institut, Universität Zürich, Zurich, Switzerland.
Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie, ETH
Zürich, Zurich, Switzerland.
3
Alfred-Wegener-Institut für Polar- und Meeresforschung, Bremerhaven, Germany.
sea and river ice [Wadhams et al., 1987; Arcone and Delaney,
1987; Melcher et al., 2002], snow on ground [Marchand et
al., 2003] or glacier thickness measurements [Thorning
and Hansen, 1987; Damm, 2004]. However, on mountain
glaciers only few studies use helicopter-borne radar to
investigate the properties of the snow cover and none was
so far dedicated to the spatial accumulation distribution. For
instance, Arcone and Yankielun [2000] focus on intraglacial
features in the ablation zone of a glacier, whereas Arcone
[2002] investigates processing techniques and autonomously
derives physical properties of temperate firn.
[3] Today’s glacier mass-balance models are based on
sophisticated formulations of the energy fluxes while accumulation processes are treated in a very simple way: precipitation varies only with altitude and any processes of snow
redistribution are neglected [Hock, 2005]. Using a similar
model (described by Machguth et al. [2006]) we have
calculated the mass balance distribution for a glacierized
catchment in southwestern Switzerland, including Findel
Glacier (length 7.2 km, area 15.3 km2) and its neighbour
Adler Glacier (3.1 km, 2.0 km2) (Figure 1). (In this paper we
refer to ‘‘Findel Glacier’’. According to the maps of the Swiss
Federal Office for Topography (swisstopo) this is the official
name of the glacier. However, in most glacier inventories the
glacier is called Findelen Glacier.) The modelling for the
1971– 1990 time period resulted in a very positive mass
balance for Adler Glacier of 0.7 m water equivalent (m we)
while its larger neighbour’s mass balance was negative
( 0.25 m we). In fact, the shrinkage of both glaciers indicates
that a very positive mass balance is unrealistic. We assume
that, in reality, accumulation on Adler- and Findel Glacier
differs strongly. The model’s failure is most probably caused
by treating precipitation as a function of altitude alone and by
ignoring snow redistribution.
[4] In this study we combine high-resolution helicopterborne GPR measurements and extensive ground-based profiling of the snow cover to determine the spatial distribution
of accumulation and to validate our assumption of a strong
local variability in accumulation. In contrast to methods that
require measurements at two points in time (e.g., mapping of
elevation changes with laser altimetry, accumulation stakes
without snowpits), GPR has the strong advantage that
changes over time in surface elevation (e.g., melt, settlement
of the underlying snow cover and glacier movement after the
first measurement) do not affect the accuracy of the measurements. Furthermore, measuring only once requires less
logistical efforts.
2
Copyright 2006 by the American Geophysical Union.
0094-8276/06/2006GL026576$05.00
2. Methods
2.1. Radar System and Data Acquisition
[5] A commercial Noggin Plus 500 radar (Sensors &
Software Inc., Canada,) was operated with a shielded
L13503
1 of 5
L13503
MACHGUTH ET AL.: ALPINE ACCUMULATION VARIABILITY
L13503
Figure 1. Map of Findel and Adler Glacier (red square in inset on Swiss map) with radar profiles (black numbered circles)
and ground measurements. Color indicates radar-based snow depth. Reproduction of the background map with permission
from swisstopo (BA067878).
antenna (15 cm transmitter-receiver separation) at a centre
frequency of 500 MHz and 400 MHz bandwidth. The
helicopter – radar combination was developed by Airborne
Scan, Visp, Switzerland. The antenna was mounted to the
helicopter runner. Data acquisition was performed in a
constant-time triggering mode, with a time increment of
0.1 s between traces. Helicopter speed was about 6 m/s on
average, resulting in a mean trace increment of 0.6 m.
Helicopter altitude above ground was between 2.5 and
30 m, being 11 m on average, resulting in a footprint size
in the order of a few tens of m2. 1876 samples were recorded
per trace with a 0.2 ns sampling interval, resulting in a 375 ns
time window. Four-fold pre-storage stacking of traces at a
pulse repetition frequency of 25 kHz was applied. A GPS
antenna was mounted at the nose of the helicopter. For each
recorded trace the GPS-position was simultaneously stored.
Real-time GPS results in a position accuracy of <10 m.
[6] The radar flight was accomplished on 9 May 2005,
ground measurements for validation were conducted on
6 and 7 May 2005. During all four days air temperature at
the glacier terminus was slightly below the freezing point.
Below 2700 m a.s.l. the snow was wet and some water
drained off at the snow-ice interface. Above 3200 m the
entire snow pack was dry. In approximately 25 minutes of
flight 10.0 km of radar profiles were collected, thereof
1.9 km on Adler Glacier (Figure 1). On Findel Glacier radar
profiles reach from 2570 to 3560 m a.s.l., on Adler Glacier
they reach from 3240 to 3690 m a.s.l. Under the favourable
weather conditions the sites of snow pits and -probes were
visible from the helicopter allowing a consecutive visual
determination of the flight direction. According to GPS data,
the helicopter passed 26 snow pits and probes at a distance of
<5 m, four at 5 – 15 m and two at 15 – 30 m. Two snow probes
and one snow pit were missed by 60 – 130 m because they
could not be found again.
2.2. Radar Data Processing
[7] Post-recording processing and radar-data analyses
were carried out with ReflexWin (Sandmeier Scientific
Software, Germany). Processing steps include dewowing
(high-pass filter), background removal, application of a gain
function to mainly compensate for spherical spreading, and
additional filtering. The varying helicopter altitude above
ground required a static correction of each trace to the first
break of the surface reflection (time-zero correction). Due to
the relatively smooth surface topography and small layer
thickness compared to the helicopter altitude no migration
was necessary. Along most profiles one or more distinct
reflections of different magnitude are visible (see sample
radar profile (Figure 2 or Figure S1 in auxiliary material1).
Tracking of the uppermost strong continuous reflection
horizon resulted in continuous profiles of last winters snow
layer thickness (Figure 1). No interpretation was performed
where no distinct reflection horizon was visible. Density
measurements in the snow pits (see below) yielded a mean
density of 400 kg m 3. Based on the linear and quadratic
conversion formulas of Tiuri et al. [1984] and Kovacs et al.
[1995], respectively, conversion of the radar data from time
to depth domain is carried out with a mean wave speed in
snow of 2.2 108 ms 1. Using the same mean density of
400 kg m 3 the layer thickness is converted to water
equivalent.
2 of 5
1
Auxiliary material is available in the HTML.
L13503
MACHGUTH ET AL.: ALPINE ACCUMULATION VARIABILITY
L13503
selected and their mean value is used below for comparison
with the corresponding ground measurement.
3. Results
Figure 2. A section of the radar profile 2 at 3460 m a.s.l.
The horizontal axis corresponds to approximately 130 m.
The varying snow depth, the internal layering of the snow
pack as well as previous years’ firn layers with some
internal structures are clearly recognizable. Grey arrows
indicate the air-snow interface and black arrows indicate the
snow-firn interface.
2.3. Snow Pits and Snow Probes
[8] Snow pits and snow probes were used to measure
snow depth and density from the actual snow surface down
to the ice surface. Where the snow cover outlasted the
previous melting season (the remaining accumulation zone
of 2004 is located above 3300 m a.s.l. on Findel Glacier, for
Adler Glacier no data is available), snow pits were dug
down to the horizon of the previous autumn’s snow surface.
According to data from the meteo station Gornergrat (located 5 km west of Findel Glacier at 3100 m a.s.l.), the long
melt season of 2004 ended with heavy snow falls on 10
October. Consequently, the snow depth measured with
radar, snow pits and -probes was accumulated within the
time span of about 10 October 2004 to 7 May 2005.
Coordinates of snow pits and -probes were captured with
a hand-held GPS (position accuracy of 5 – 15 m). Within an
altitudinal range from 2590 to 3510 m a.s.l. 20 snow pits
have been sampled and snow depth and density as well as
the internal layering of the snow pack was determined in all
of them. This sample size was enhanced by 19 measurements of snow depth with snow probes. However, defining
previous autumn’s snow surface using snow probes is
sometimes difficult (e.g., misinterpretations due to ice layers
within the winter snow pack). Consequently, snow probes
were restricted to the ablation area and for every test site the
mean value of nine snow probes, sampled within a radius of
7 m, was calculated.
2.4. Data Merging
[9] Field measurements were only used to validate the
interpretation of the radar profiles, analyses were conducted
separately. Neither GPS data nor a map were used for the
interpretation of the radar profiles, thus the interpreter’s
knowledge about the field measurements could not influence his interpretation of the profiles. The data sets were
joined within a GIS software (ArcGIS 9.1). The twenty
closest radar traces to every ground measurement were
[10] The transition between winter snow and ice or winter
snow and snow having outlasted the previous summer is in
general clearly recognizable in the radar profiles. A total of
0.6 km radar profiles (6% of the total length) did not allow
any tracking due to lacking or disturbed layering. Most of
these zones are located within crevassed areas. Further
analysis of the data is based on almost 15000 radar traces,
representing the remaining 9.4 km of radar profiles. The
thickness of the winter’s snow layer varied in the ground
measurements (radar measurements) from 0.32 (0.44) to
4.4 (5.9) m. The specific density measured in the snow
pits is not correlated with altitude and varies from 360 to
470 kg m 3. The mean density is 400 kg m 3 with a
standard deviation of 30 kg m 3. Figure 3 shows the
agreement between radar profiles and all snow pits and -probes
where the horizontal distance to the radar track is less
than 30 m. The linear regression yields a correlation
coefficient of R2 = 0.84. Three data points must be
considered outliers, as discussed below. Excluding these
data points yields R2 = 0.96.
[11] According to the radar profiles three zones of different accumulation characteristics can be distinguished: the
lower part (profile 1) and the upper part of Findel glacier
(profiles 2, 3 and 4), as well as Adler Glacier (profile 5).
The lower part of Findel Glacier shows a clear correlation
between altitude and snow cover thickness (R2 = 0.81) and
the fluctuations in snow depth are small (Figure 4). On
Adler Glacier and the upper part of Findel Glacier accumulation has no altitudinal trend (R2 0.01). Fluctuations in
snow depth are very large. The correlation coefficient
Figure 3. Comparison of radar- and ground-based measurements of snow depth. The three outliers are marked
with a red circle.
3 of 5
L13503
MACHGUTH ET AL.: ALPINE ACCUMULATION VARIABILITY
Figure 4. Snow depth versus altitude for all profiles on
Findel and Adler Glacier.
calculated for the upper part of Findel Glacier is based on all
traces of the profiles 2, 3 and 4 but also represents well the
characteristics of every individual profile: none shows any
significant trend with altitude; and all show large fluctuations of snow depth. At altitudes where radar profiles exist
on both glaciers (3240 to 3560 m a.s.l.) the average
accumulation is 2.98 m snow (5704 traces) on Findel and
1.80 m snow (2164 traces) on Adler Glacier. The accumulation on Adler Glacier is 0.5 m we or 40% lower than on
Findel Glacier.
4. Discussion
[12] Our correlation between ground and radar measurements confirms the results of earlier studies with helicopterborne radar in non glacierized landscapes, having achieved
correlation coefficients of 0.82 –0.97 [e.g., Marchand et al.,
2003]. The three outliers mentioned above are presumably
caused by misinterpretations in the field. Snow-probe measurements at those three sites (see Figure S2 in auxiliary
material), located within a diameter of 100 m at 2850 m
a.s.l. are contradictory (0.5– 1.7 m snow). A massive ice
layer at 0.5 m depth was recorded in the snow profile closest
to them. We assume that this layer was interpreted as the ice
surface. The layer is visible in the radar profile as a long,
massive and smooth reflection horizon at a depth of 0.5 to
1 m. Normally, the snow-ice interface appears more uneven
because of previous surface melt and is similar to the
second, less distinct reflection horizon at about 1.8 m in
that profile. The example shows that in situ measurements
might present only partial truth. A correct interpretation of
the internal layering of the snow pack is essential for both
methods. Consequently, mutual comparison is important to
combine their strengths. In the ablation area snow pits allow
precise point measurements while the large footprint of
helicopter-borne GPR has the advantage of averaging out
the bumpiness of the ice surface. In the accumulation area,
both methods have to deal with the difficulty to correctly
identify last autumn’s snow horizon.
[13] The overall pattern of accumulation distribution on
Findel Glacier corresponds well to the observations based
on snow probes on other alpine glaciers (e.g., Vernagtferner
in Austria, Plattner et al. [2006]). On the lower part of
L13503
Findel Glacier the correlation between altitude and snow
depth can partly be explained by melt during long warm
weather periods in March and April 2005. The large
deviations in accumulation between Findel and Adler Glacier are very unlikely to be caused by enhanced melt in
March and April due to differing surface exposition: assuming that the entire difference in global radiation due to
exposition (20 W m 2) is available for melt and considering
the high snow albedo of approximately 0.8 for this period,
melt on Adler Glacier is estimated to be 0.05 m we higher
than on Findel. We therefore assume that the large deviations in accumulation are primarily due to spatial variability
of precipitation and redistribution of snow by wind. The
area of investigation receives high amounts of precipitation
under southerly wind directions. Reduced wind speed
leeward from ridges results in enhanced precipitation and
additional snow deposition from wind transport [e.g., Föhn
and Meister, 1983; Gauer, 2001]. Findel glacier is located
directly leeward of the main ridge and profits from this
effect, whereas Adler Glacier is farther away. In addition,
strongly reduced accumulation is also observed in crevassed
zones, probably caused by topography and microscale
turbulences. Within this study we have only assessed the
spatial and not the temporal variability of accumulation.
However, taking into account the concurrent shrinkage of
both glaciers, we assume that the observed deviations in
accumulation of both glaciers for the winter 2004/05 are not
exceptional.
5. Conclusions and Outlook
[14] The data of both methods used in this study are in
very good agreement. The observed distribution of the snow
cover confirms our assumption of strongly reduced accumulation on Adler Glacier. Our results emphasize that the
distribution of accumulation is not simply a function of
altitude, confirming other studies [e.g., Winstral et al.,
2002]. We suppose that the spatial variability of precipitation as well as the redistribution of snow are mainly
governing the accumulation distribution. The observed
variability of the mass balance for the accumulation period
(0.5 m we) is one order of magnitude higher than the error
of melt calculations for the entire ablation period with
energy balance models [compare Arnold et al., 1996;
Obleitner and Lehning, 2004]. The results underline that
major improvements in glacier mass balance modelling can
be achieved by focusing on the accumulation processes.
Helicopter-borne GPR is recommended as a reliable tool for
time-saving and accurate mapping of the snow cover. The
method allows to enhance the sparse database on accumulation distribution toward both spatial and temporal variability, providing a sound data basis for glacier monitoring
(e.g., yearly repeated measurement of winter balance), for
statistical analyses (e.g., digital elevation model attributes)
or for validation and calibration of physical modelling. Field
measurements remain essential for mutual validation and to
determine snow density.
[15] Acknowledgments. Simon Allen, Sabine Baumann, Xavier
Bodin, Esther Hegglin, Christian Huggel, Jeannette Nötzli, Theresa Tribaldos,
Michi Zemp and Michael Ziefle have helped us on field work. Their large
effort made this study possible and is gratefully acknowledged. We very much
appreciate the cooperation with Hubert Andereggen, Airborne Scan and Air
4 of 5
L13503
MACHGUTH ET AL.: ALPINE ACCUMULATION VARIABILITY
Zermatt. We would like to thank Michael Lehning and an anonymous reviewer
for their valuable comments. H.M. was funded by the grant 21-105214/1 of the
Schweizer Nationalfonds, O.E. was supported through an ‘‘Emmy Noether’’scholarship EI 672/1 of the Deutsche Forschungsgemeinschaft.
References
Arcone, S. (2002), Airborne-radar stratigraphy and electrical structure of
temperate firn: Bagley Ice Field, Alaska, U.S.A., J. Glaciol., 48, 317 –
334.
Arcone, S., and A. Delaney (1987), Airborne river-ice thickness profiling
with helicopter-borne UHF short-pulse radar, J. Glaciol., 33, 330 – 340.
Arcone, S. A., and N. E. Yankielun (2000), 1.4 GHz radar penetration and
evidence of drainage structures in temperate ice: Black Rapids Glacier,
Alaska, U.S.A., J. Glaciol., 46, 477 – 490.
Arnold, N., I. Willis, M. Sharp, K. Richards, and W. Lawson (1996), A
distributed surface energy balance model for a small valley glacier: I.
Development and testing for Haut Glacier d’Arolla, Valais, Switzerland,
J. Glaciol., 42, 77 – 89.
Damm, V. (2004), Ice thickness and bedrock map of Matusevich Glacier
drainage basin (Oates Coast), Terra Antarct., 11, 85 – 90.
Föhn, P. M. B., and R. Meister (1983), Distribution of snow drifts on ridge
slopes: Measurements and theoretical approximations, Ann. Glaciol., 4,
52 – 57.
Gauer, P. (2001), Numerical modelling of blowing and drifting snow in
Alpine terrain, J. Glaciol., 47, 97 – 110.
Hock, R. (2005), Glacier melt: A review of processes and their modelling,
Prog. Phys. Geogr., 29, 362 – 391.
Kanagaratnam, P., S. Gogineni, V. Ramasami, and D. Braaten (2004), A
wideband radar for high-resolution mapping of near-surface internal
layers in glacial ice, IEEE Trans. Geosci. Remote Sens., 42, 483 – 490.
Kohler, J., J. Moore, M. Kennett, R. Engeset, and H. Elvehøy (1997), Using
ground-penetrating radar to image previous years’ summer surfaces for
mass-balance measurements, Ann. Glaciol., 24, 355 – 360.
Kovacs, A., A. Gow, and R. Morey (1995), The in-situ dielectric constant
of polar firn revisited, Cold Reg. Sci. Technol., 23, 245 – 256.
Machguth, H., F. Paul, M. Hoelzle, and W. Haeberli (2006), Distributed
glacier mass-balance modelling as an important component of modern
multi-level glacier monitoring, Ann. Glaciol., in press.
L13503
Marchand, W. D., A. Killingtveit, P. Wilen, and P. Wikstrom (2003), Comparison of ground-based and airborne snow depth measurements with
georadar systems, case study, Nord. Hydrol., 34, 427 – 448.
Melcher, N. B., et al. (2002), River discharge measurements by using
helicopter-mounted radar, Geophys. Res. Lett., 29(22), 2084,
doi:10.1029/2002GL015525.
Obleitner, F., and M. Lehning (2004), Measurement and simulation of snow
and superimposed ice at the Kongsvegen glacier, Svalbard (Spitzbergen),
J. Geophys. Res., 109, D04106, doi:10.1029/2003JD003945.
Plattner, C., L. N. Braun, and A. Brenning (2006), Spatial variability of
snow accumulation on Vernagtferner, Austrian Alps, in winter 2003/
2004, Z. Gletscherkd. Glazialgeol., in press.
Richardson, C., E. Aarholt, S.-E. Hamram, P. Holmlund, and E. Isaksson
(1997), Spatial distribution of snow in western Dronning Maud Land,
East Antarctica, mapped by a ground-based snow radar, J. Geophys. Res.,
102, 20,343 – 20,353.
Thorning, L., and E. Hansen (1987), Electromagnetic reflection survey
1986 at the Inland Ice margin of Pakitsoq basin, central Greenland, Rapp.
Groenl. Geol. Unders., 135, 87 – 95.
Tiuri, M., A. Sihvola, E. Nyfors, and M. Hallikainen (1984), The complex
dielectric constant of snow at microwave frequencies, IEEE J. Oceanic
Eng., 9, 377 – 382.
Wadhams, P., M. A. Lange, and S. F. Ackley (1987), The ice thickness
distribution across the Atlantic sector of the Antarctic Ocean in midwinter, J. Geophys. Res., 92, 14,535 – 14,552.
Winstral, A., K. Elder, and R. Davis (2002), Spatial snow modeling of
wind-redistributed snow using terrain-based parameters, J. Hydrometeorol., 3, 524 – 538.
Yankielun, N., W. Rosenthal, and R. E. Davis (2004), Alpine snow depth
measurements from aerial FMCW radar, Cold Reg. Sci. Technol., 40,
123 – 134.
O. Eisen, Laboratory of Hydraulics, Hydrology, and Glaciology, ETH
Zürich, CH-8092 Zurich, Switzerland.
M. Hoelzle, H. Machguth, and F. Paul, Physical Geography Division,
Department of Geography, University of Zurich, Winterthurerstr. 190, CH8057 Zurich, Switzerland. ([email protected]; [email protected]
unizh.ch)
5 of 5
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement