Field Report Snow and Ice Processes AGF212

Transcription

1 Field Report 2013 Snow and Ice Processes AGF212 (picture) Names...

2 Contents 1 Estimation of ice thickness and snow distribution using Ground Penetrating Radar Introduction Instruments and methods Instruments Radar Theory for Snow and Ice Processing in RadExplorer Interpolation in Matlab Results Ice Thickness Investigations on the Marthabreen Glacier System Snow Thickness Investigations on the Marthabreen Glacier System Snow Thickness Investigations on the glacier system Tellbreen and Blekumbreen Snow Thickness Investigations on the glacier Fangenbreen Discussion Ice Thickness Investigations on the Marthabreen Glacier System Snow Thickness Investigations on the Marthabreen Glacier System Snow Thickness Investigations on the glacier system Blekumbreen and Tellbreen

3

4 1 Estimation of ice thickness and snow distribution using Ground Penetrating Radar by Philipp Franzen, Stefan Muckenhuber, Georg Seyerl Abstract Marthabreen, Fangenbreen, Blekumbreen and Tellbreen in central Svalbard have been analyzed regarding their snow and ice distributions using Ground Penetrating Radar (GPR) technique. The GPR measurements resulted in a mean snow depth of 1.58 m on Marthabreen and 1.30 m on the glacier system Bleckumbreen, Tellbreen. The ice thickness results of Marthabreen can be used as a reference for future investigations. 1.1 Introduction During mid march 2013 three different glacial sites in central Svalbard have been analyzed regarding their snow and ice distributions using Ground Penetrating Radar (GPR) technique (Daniels, 2000). Snow depth measurements have been taken on Marthabreen, Fangenbreen, Blekumbreen and Tellbreen. Whereas the latter two glaciers were considered as one glacial site due to spatial connection. Ice thickness investigations were only measured on Marthabreen. The snow accumulation on a glacier determines whether there is a loss or gain of mass, but if (englacial/subglacial) melt exceeds accumulation, the glacier will loose mass. It has a large effect on the albedo and influences the water supply of the surrounding region (Grabiec et al., 2011). The spatial distribution highly depends on local wind patterns, which are mainly driven by katabatic flows (Dadic et al., 2010). In addition to mass-balance investigation, the determination and monitoring of total glacial ice volume and ice-thickness distribution are important parameters for understanding the interactions between climate and the glacier system (Binder et al., 2009). As the ice thickness of Marthabreen has not been investigated before the data presented in this report can be used as a reference for future investigations. 1

5 1 Estimation of ice thickness and snow distribution using Ground Penetrating Radar 1.2 Instruments and methods The basic principle of GPR is the emission of an electromagnetic (EM) pulse from a transmitter and the subsequent detection of the returned wave by a receiver. By measuring the travel time, i.e. the time it takes for the radar waves to travel through the medium and bounce back to the surface, the depths of various internal reflectors can be estimated. Furthermore other parameters can be determined, namely the thermal properties of the medium and its water content Instruments The GPR used was a MALÅ ProEx system (MALÅ Geoscience, Malå Sweden), consisting of a control unit to save the data, a monitor as data acquisition platform to visualize the results simultaneously and an antenna to emit and detect the EM waves. The control unit and the monitor were attached to the snowmobile. The antenna was either carried in a sledge or simply towed to the snowmobile. Slow speed is favourable for keeping the vertical movement of the sledge at a minimum, so the aimed speed was below 20 km/h. To determine the snow distribution, a MALÅ 800 MHz Shielded Antenna was chosen, as higher frequencies yield a better resolution for depths of several meters. This antenna was placed in a sledge. To calibrate and verify the 800 MHz GPR measurements, snow depths were recorded manually around every 100 to 200 m by using snow probes with scales of 1 to 5 cm. To investigate the ice thickness a MALÅ 50 MHz Rough Terrain Antenna was used. The produced low frequency waves are mainly reflected by the bedrock and not by the ice itself. Also bigger rocks and melt water channels can be detected with it. Due to the flexible snake design, the antenna can easily be attached to the back of the snow mobile. Connected to the GPR was a Garmin hand held GPS receiver with an accuracy of around 4 m. This GPS was also used to mark the positions of our probing sites Radar Theory for Snow and Ice The basic principles for the GPR are the same as for detecting air planes overhead, but with GPR the antennas are moved over a surface to examine the medium below. A transmitting antenna is sending an EM wave into the subsurface. This wave penetrates into the ground until it hits a boundary where the electrical properties are suddenly changing (e.g. another material, different phase, warmer, wetter... ). At this point, a part of the wave will be scattered and reflected back and a part will continue to travel deeper down into the ground. When the reflected part reaches the surface again, it will be detected by the receiver. During this process the time t is measured and therefore, if the velocity v of the EM wave is known, the depth d of the boundaries can be estimated, by using x d = a a = v t T W T 2 (1.1) 2

6 1.2 Instruments and methods x is the distance between the emitting and receiving antenna and t T W T is the two way travel time (Bælum and Knudsen, 2006; Daniels, 2000). The general dependence of the phase velocity on electrical properties is: v = 1 ε µ (1.2) With ε being the dielectric permittivity and µ the magnetic permeability. For the frequency domain of radar waves, the permeability of most naturally occurring materials will be very close to the permeability of free space and therefore the dependence on ε is the predominant factor (Bælum and Knudsen, 2006) and the following equation can be used for dry snow: v = c εr ε = ε r ε 0 (1.3) c is the speed of light in vacuum and the relative permittivity ε r can be derived from an empirical relation including the change with snow density ρ s now (Kovacs et al., 1995; Dunse et al., 2009). ε r = ( ρ s now ) 2 (1.4) The snow density ρ s now was calculated from the mean of 12 snow pit measurements. In order to get an accurate error estimation, the standard deviation σ was calculated as well. Both the mean and the variance σ 2 were weighted by the depth d i of the different density measurements. 1 N ρ = N d i ρ i 363 kg/m 3 (1.5) i d i i σ = 1 N N d i (ρ i ρ) 27.2 kg/m 3 (1.6) i d i i This results in an average velocity of cm/ns with an error of around ± 0.4 cm/ns. A typical value for the two way travel time is 10 ns, which would lead to an error of ± 4 cm. This is small compared to the change in snow thickness and therefore the average snow density can be considered as a valid base for calculating the radar snow velocity for all three glaciers. The radar velocity for ice was set to a constant of 16.8 cm/ns, which is commonly used for cold ice (Bælum and Knudsen, 2006) Processing in RadExplorer The measured time can be displayed and post-processed by using MALÅ s software RadExplorer. By applying a series of filters and gains to remove the noise inherited from the device, the different boundary layers are getting more visible. At first, the mean of each trace is subtracted to get rid of the offset (DC Removal). As the antenna starts recording before the pulse 3

7 1 Estimation of ice thickness and snow distribution using Ground Penetrating Radar is sent out, the time for the first arrival is set to be zero (Zero Time Adjustment). This signal corresponds to the wave travelling along the surface. The next step applied was Background Removal. To remove the noise from the input, i.e. waves far away from the frequency, the Bandpass Filtering routine is used. The amplitude gets weaker, the deeper the signal penetrates into the subsurface. The Amplitude Correction gets rid of this problem and makes smaller events more visible. Now the snow-ice (800 MHz) or ice-bedrock (50 MHz) boundaries are visible. The different horizons were marked by using the routine Picking Horizons. With the velocities above, the two way travel time was converted into depth. Finally the data was exported to MatLab for further interpolation, following the scripts by Pfau, Interpolation in Matlab At the beginning of the interpolation accurate boundary conditions have to be chosen, i.e. the outline of the glacier. Within the program OziExplorer the outline of Blekuumbreen, Fangenbreen, Marthabreen and Tellbreen were drawn on a Svalbard map and this data was then exported to MatLab. The ice thickness along the outlines was set to zero and the snow depth was set to the mean of the respective glacier. Trying different methods of interpolations, the best result yielded the MatLab function»griddata(..., natural ). This is a natural neighbour interpolation, which means that the new estimate G at point (x,y ) is calculated by using the surrounding measured points f (x i,y i ) and weighting them with w i. At first the area has to be separated by drawing straight lines in the middle of two neighbouring points f (Voronoi tessellation). Then the same is done for G. The weighting w i now depends on how much of each of the surrounding areas is overlapping between the G and the f region or Voronoi cell. The resulting points G can then be derived by using: G (x,y ) = N w i f (x i,y i ) (1.7) i 1.3 Results The fact of different field advisers during the two weeks of fieldwork results in two teams working on two different Glacier Systems. The first Glacier system described below is Marthabreen, where ice- and snowdepth have been measured. The second working side consists of the glaciers Blekumbreen, Tellbreen and Fangenbreen, where only snowdepth measurements have been taken. Blekumbreen and Tellbreen were treated as one glacier system Ice Thickness Investigations on the Marthabreen Glacier System Due to mass balance calculations, initiated by the University of Oslo and Store Norske for the glacier system Marthabreen, an evaluation of the ice thickness was performed using a GPR. The glacier is about 8 km long and is located at the southeast of Nordenskiöld Land in central Svalbard. Marthabreen has one tributary called Propsbreen, flowing from the South West into the main trunk. We deliberately chose to study these two glaciers as one entity. 4

8 1.3 Results In figure 1.1 the topography of Marthabreen including the glacier margins can be seen. In addition the tracks of the profiles taken with the GPR are marked in black. Figure 1.1: Marthabreen topography including scooter cross section tracks, glacial ice borders, two crevasses and a meltwater channel. The crevasse (figure 1.2), meltwater channel one (figure 1.3) and meltwater channel two (figure 1.4) marked in the topographical map 1.1 above will be discussed in section in more details. Therefore we will use the filtered cross section plots of RadExplorer as described in section

9 1 Estimation of ice thickness and snow distribution using Ground Penetrating Radar Figure 1.2: 50 MHz GPR profile (1.2 km) showing glacier ice, the glacier bed and the crevasse from figure 1.1. As described in section the 18 measured cross sections and two center lines, using a 50 MHz antenna, were interpolated for the whole Marthabreen glacier system. The plot of the interpolation as a mercator projected map (figure 1.5) results in a plausible gradient between the as zero ice depth defined boundary conditions and the deeper center parts of the glacier. The maximum glacier ice depth of 158,83 m was reached at N, E, therefore its position is at the beginning of the glacier tongue facing north-westward. The minimum ice depth was measured at N, E with a value of m. However, this point of minimum ice depth is not the measurement point located closest to the boundary condition line of the glacier. Hence, the supposed outline of Mathabreen by analyzing map material is not as accurate as a GPR profile around the glacial border Snow Thickness Investigations on the Marthabreen Glacier System The process of manual picking of the snow/ice interface on the radagrams displayed in radarexplorer is not simple, due to the fact that this interface is particularly noisy/blurry. Hence, the results of the horizon picking can be very user specific. Therefore the mean GPR snow depth compared to the snow probe data is not as significant as its anomaly. Thus, all snow thickness interpolations are presented as its anomaly of the glacial mean snow depth. The snow probe data in figure 1.7 consists of 31 different probing sites. The maximum probing depth of 1.54 m was reached at the south border of the glacier, whereas the minimum depth of 0.64 m was measured at the northern part of the glacier tongue. Even though the GPR snow depth interpolation (figure 1.8) shows the same distribution of snow anomalies the maximum of 2.58 m is more than one meter higher than the probing maximum. The measured minimum GPR snow depth was 1.14 m. All GPR depths are calculated 6

10 1.3 Results Figure 1.3: Meltwater channel one Figure 1.4: Meltwater channel two Both figures above are 50 MHz GPR profiles showing glacier ice, the glacier bed, meltwater channel one and meltwater channel two from figure 1.1. using the velocity for the propagation of radar waves in snow as mentioned in section The figure 1.6 shows the snow depths measured by the GPR compared to the probing depths at the respective points on Marthabreen on the The green line displays perfect correlation. However, this is not the case for the measurements made during our fieldwork. The probing measurements are smaller than the values calculated from GPR data. On average the values differ by 46 cm with a standard deviation of 18 cm. One reason for these anomalies could be the low accuracy of our permittivity-density model. In addition, the snow probing gives only a single point where as GPR back-scatter is always received over a larger area (footprint). Further reasons could be the spatial variability of snow density, tilted probe when measuring, "wrong choice of summer layer" in either GPR data or probing data. The GPR Profile using a 800 MHz Antenna in figure 1.2 was taken in the accumulation area of the glacier. Unfortunately, due to failed GPS measurements as a result of a broken GPS cable we were not able to estimate the exact positions of the GPR profiles. Nevertheless, the different layers of snow, firn and ice can be seen and will be discussed in more detail in section Snow Thickness Investigations on the glacier system Tellbreen and Blekumbreen The glaciers Tellbreen and Blekumbreen are located approximately 10 km east of the settlement of Longyearbyen on Svalbard. Measured along the centre line, both glaciers together have an extension of about 8 km in east-west direction. The north-south extension varies mostly between 1 to 2 km. The glaciers are connected in their top area at an elevation of around 650 m. The glacier tongue of Blekumbreen is facing to the west and the one of Tellbreen to the 7

11 1 Estimation of ice thickness and snow distribution using Ground Penetrating Radar Figure 1.5: Marthabreen ice thickness interpolation east. (see also figure 1.10). The figure 1.10 shows the outline of the glacier system and the topography around it. On top of that the scooter-tracks are marked, along which the GPR measurements were taken. These measurements provided the basis for the interpolation of the snow depth on the glaciers (see section 1.2.4). This interpolation is resulting in figure It provides the anomaly of the snow depth from the glacier system compared to the mean snow depth which is approximately 1.30 m. The maximum snow depth of 3.24 m can be found at the coordinates N, E. This position is located at the northern glacier ice border of Tellbreen and is also shown in figure The minimum snow depth of 0.34 m can be found at the coordinates N, E, which is situated almost at the front of glacier tongue of Blekumbreen. Another result seen in figure 1.11 is that while moving to lower elevation levels, the snow depth decreases. This is true for both Tell- and Blekumbreen. Furthermore, a negative anomaly can be seen on the plateau at the top of both glaciers near the contact line. Mostly positive anomalies can be found close to the glacial ice borders. However, considering the interpolation map it appears that near the northern border, the anomaly is even larger than near the southern border. 8

12 1.3 Results Figure 1.6: Variances of GPR and probing snow depths The observations mentioned above can also be seen in the following plots. Figure 1.12 shows a center line cutting the whole glacier system, a cross section on Blekumbreen and the position of a crevasse on Tellbreen representing the maximum snow depth measured by the GPR. Respectively to these center lines, the snow depth is plotted in figure Although the center line is not taken exactly in east-west direction, it just causes a minor error which doesn t prevent a accurate analysis of the data. The snow depth along the center line on Blekumbreen varies approximately between 0.75 and 1.5 m, on Tellbreen from 0.9 m to 1.75 m. As mentioned before, the snow depth on the top of both glaciers is lower than in the surrounding area and decreases while moving further downward towards the glacier tongue. Figure 1.15 shows the cross section on Blekumbreen, which can also be seen in figure The graph shows the snow depth distribution between the south and north border of the glacier. As said before, the snow depth increases significantly while approaching the glacial ice 9

13 1 Estimation of ice thickness and snow distribution using Ground Penetrating Radar Marthabreen Probe data Marthabreen Snow Interpolation 2.00 Marthabreen outline Scooter track Marthabreen outline o N Anomaly of mean snowdepth [m] o N Anomaly of mean snowdepth [m] o E o E Figure 1.7: Snow Probe Data Snow mean depth = 0.93 m Figure 1.8: GPR Snow Interpolation Data Snow mean depth = 1.58 m borders near a slope. The same can be seen for cross sections on Tellbreen. Furthermore, there are probe measurements from the same locations as the GPR measurements included in the figure. A comparison of these two measurements shows that the absolute snow depth measured by the probe doesn t match with the GPR measurements. This can be traced throughout all measurements, whereas the reason for this uncertainty was already described in the section of Marthabreen and will be discussed in more detail in section 1.4. Nevertheless, we can rule out this uncertainty by comparing the anomaly, which is shown in figure The mean snow depth was only calculated out of the probing measurements and results in m. The maximum snow depth measured with the snow probe was 2.12 m at the position of N, E, located at the center part of Tellbreen. The minimum snow depth with 0.45 m was found at N, E situated on the lower part of Blekumbreen. The general shape of the figure shows the same structure as the interpolation map of the GPR-data (figure 1.11). This is easier to reproduce for Blekumbreen since there is more probing data available on this glacier than on Tellbreen Snow Thickness Investigations on the glacier Fangenbreen The glacier is about 2.5 km long, located in the northern part of Nordenskiöld Land in central Svalbard. The maximum elevation of the glacier is around 800 m in the west part, whereas the minimum elevation is around 500 m in the east part of the glacier as can be seen in figure The minimum snow depth of 0.64 m is located at N E which is also the lowest part of the glacier (figure 1.17). The maximum snow depth of 1.98 m was reached at N, E, therefore its position is at the highest measured point of the glacier. In comparison with the other glaciers the snow distribution seems to show an inverse behaviour, which means that the higher snow coverage can be found in the middle of the 10

14 1.4 Discussion Figure 1.9: 800 MHz GPR profile showing the snow, the firn and the ice layer of the glacier glacier. The GPR data was processed, but further work on interpolation was made impossible due to problems related to the malfunctionning GPS. 1.4 Discussion Ice Thickness Investigations on the Marthabreen Glacier System The process of differentiation between glacial features and dirt is a matter of experience. Therefore, the following interpretation of the most interesting 50 MHz GPR profiles as they were produced by RadExplorer is a try to give reasons for the travel time variations in the plots. Crevasse one (figure 1.2): The clearly visible open crevasse is about 75 meters deep and almost reaching down to the bed of the glacier. The challenging part of interpreting such a cross section can be demonstrated on the very right side of this profile, whereas we identified this formation as another dirt or stone layer on the bedrock due to its sharp contour and clear separation. Meltwater channel one (figure 1.3): The superficial meltwater channel ranging around 20 meters into the ice is located in the middle of the profile. Sharp contours on the upper side of the channel and three meltwater fingers pointing downward as a result of penetrating meltwater. Another reason for the sharp contours on the lower side could be a big amount of water in the channel which is disturbing the signal of the radar. Meltwater channel two (figure 1.4): 11

15 1 Estimation of ice thickness and snow distribution using Ground Penetrating Radar Figure 1.10: Scooter track and glacial ice borders of Tellbreen to the right and Bleckumbreen to the left This meltwater channel on the right side of the profile is about 75 meters deep. Nevertheless the difference to channel one can be seen in the upper and lower parts of it. For the upper part the flowing ice is already closing the channel Snow Thickness Investigations on the Marthabreen Glacier System For the presentation of the snow thickness we choose its anomaly of the glacial mean snow depth. One reason is the big uncertainty in the estimation of the snow depth by using the GPR data, another reason is the low amount of probing data points. The snow probe anomalies were calculated using the probe mean snow depth, whereas the snow interpolation anomalies were calculated using the snow interpolation mean. Nevertheless, the comparison of anomalies of mean snow depths shows the same distribution on the measured glacier parts for probe (figure 1.7) and GPR (figure 1.8) data. For the lower part of the glacier tongue in the ablation area the anomalies are negative. Thus the snow depth is rising moving south-east-ward towards higher terrain. The deepest snow is located at the border of the glacier, which can be explained by wind and avalanche induced convergence of snow Snow Thickness Investigations on the glacier system Blekumbreen and Tellbreen The glacier system of Tell- and Blekumbreen provides some special features regarding the snow depth distribution and its local peculiarities. 12

16 1.4 Discussion Figure 1.11: Interpolation map of the snow depth anomaly Mean snow depth = 1.30 m Snow depth measurements using a GPR system As expected, there is a larger snow cover on the upper part of the glaciers due to the precipitation increase with elevation. This can be seen in the interpolation map of the snow depth (figure 1.11) and also in the measurements taken along the center line. The little snow cover on the plateau between the glaciers is a result of the wind-exposed location. The snow is redistributed by the wind to the surrounding lower areas which is creating maxima in the snow distribution below the top of both glaciers (figure 1.11 and figure 1.13). There was weather data available from two year-round weather stations on Tellbreen. These stations, especially the one located at lower elevation, measured strong katabatic winds during the year implying that some snow was redistributed from the upper to the lower part of the glacier. This could explain the strong decline of the snow cover in the upper part, followed by the more constant snow depth profile in the lower parts of the glacier. (figure 1.13) The declining trend in the snow depth from the top to the bottom of Blekumbreen looks more linear than on Tellbreen. Unfortunately, we didn t have any weather station data from Blekumbreen for comparison. This could have proven useful to analyse the effect of the katabatic wind further, especially if these winds are really responsible for a significant transport of snow on the glacier. The increased values near to the ice borders of the glaciers, as seen in figure 1.11, especially on the upper part of Blekumbreen, can explained due to the slopes surrounding the glaciers. Avalanches and snow movement downward these slopes are responsible for this increase. The absolute snow depth on both glaciers does not vary a lot compared to each other. Due to 13

17 1 Estimation of ice thickness and snow distribution using Ground Penetrating Radar Figure 1.12: Position of center lines on Tell- (1) and Blekumbreen (2), the cross section (3) and the maximum snow depth (4) Figure 1.13: Snow depth along the center line on Tell and Blekumbreeen their orientation you could presume that the precipitation on Blekumbreen should be higher than on Tellbreen, because synoptic air masses transporting humid air to Svalbard are mostly approaching from westerly directions. Hence, causing a luv/lee-effect with Blekumbreen located on the luv side and Tellbreen on the lee side. Since there is no big difference to see in the interpolation map figure 1.11, it is likely that this effect gets ruled out by other effects as could possible by the redistribution by wind. Comparison with probing data The probing anomalies (figure 1.15) in accordance with the interpolation anomalies (figure 1.11) can be seen in the upper part (16 04 E to E) near the northern ice borders of Blekumbreen. However, due to GPS measurement uncertainties the differences in the lower part of Tellbreen might not represent real snow distribution. 14

18 1.4 References Figure 1.14: Cross section on Blekumbreen including probe measurements Comparison with old reports The interpolated snow distribution of Tellbreen in the current year shows the same structure as the one in However, there is a difference in the absolute snow depth. Two years ago the snow cover of the whole glacier added up to 1.4 m, except for slightly smaller values towards the glacier tongue. During our measurements we encountered less snow cover, especially in the center and the lower part of the glacier. Also the snow distributions of Blekumbreen are corresponding, particularly the high values of snow depth close to the northern glacial borders and the upper part of the glacier are in compliance with each other. As also seen on Tellbreen we are able to observe the same difference in the absolute snow height, whereas its amplitude is not that high. A possible reason for the observed differences in the absolute snow height may be the cold weather conditions throughout this winter. References Bælum, K. and N. T. Knudsen (2006). Mapping of the General Shape, Depth and Various Internal Structures of Tellbreen, a Glacier on Svalbard, by Means of GPR ( Ground Penetrating Radar ) Indeks. In: January. Binder, D. et al. (2009). Determination of total ice volume and ice-thickness distribution of two glaciers in the Hohe Tauern region, Eastern Alps, from GPR data. In: Annals of Glaciology 50.51, pp DOI: / Dadic, R. et al. (July 2010). Parameterization for wind-induced preferential deposition of snow. In: Hydrological Processes 24.14, pp ISSN: DOI: /hyp Daniels, J. J. (2000). Ground Penetrating Radar Fundamentals. In: pp

19 1 Estimation of ice thickness and snow distribution using Ground Penetrating Radar Tellbreen/Blekumbreen Probe data o N o E Marthabreen outline Anomaly of mean snowdepth [m] Figure 1.15: Map of probe measurements on Tell- and Blekumbreen Snow mean depth = 1.04 m Figure 1.16: Fangenbreen topography including scooter cross section tracks and glacial ice borders 16

20 1.4 References Fangenbreen Probe data o N o E Fangenbreen outline Anomaly of mean snowdepth [m] Figure 1.17: Map of probe measurements on Fangenbreen Snow mean depth = m Dunse, T. et al. (Mar. 2009). Recent fluctuations in the extent of the firn area of Austfonna, Svalbard, inferred from GPR. In: Annals of Glaciology 50.50, pp ISSN: DOI: / URL: xref?genre=article\&issn= \&volume=50\&issue=50\&spage=155. Grabiec, M. et al. (Dec. 2011). Snow distribution patterns on Svalbard glaciers derived from radio-echo soundings. In: Polish Polar Research 32.4, pp ISSN: DOI: /v Kovacs, A. et al. (May 1995). The in-situ dielectric constant of polar firn revisited. In: Cold Regions Science and Technology 23.3, pp ISSN: X. DOI: / X(94)00016-Q. URL: X Q (visited on 01/05/2013). Pfau, R. ( (2011). GPR Processing Intro. In: 17