Mapping Glacier Velocities at Spitsbergen Using ers tandem sar data

Mapping Glacier Velocities at Spitsbergen Using ERS Tandem SAR Data
Bjørn Wangensteen*, Dan Johan Weydahl**, Jon Ove Hagen*

*Department of Physical Geography, University of Oslo, PO Box 1042 Blindern, 0316 Oslo, Norway,

**Norwegian Defence Research Establishment, Division of Electronics, PO Box 25, 2007 Kjeller, Norway,

E-mail: [e-mail], [e-mail], [e-mail]

Abstract: The results presented here show how glacier velocities can be measured and calculated from ERS tandem INSAR data. A semi-automatic algorithm using existing and new implemented moduls in the GIS package Arc/INFO has been developed for the calculation of glacier velocity. The velocities are decomposed into the flow direction of the glacier using an external DEM. Velocity fields for the glaciers Isachsenfonna, Akademikerbreen and Nordbreen at Spitsbergen have been calculated.
INTRODUCTION

Surface velocities were firts calculated from intreferometric SAR data in 1993 [1]. When ERS-2 was launched in 1995 it was possible to get tandem scenes from ERS-1 and ERS-2 taken only one day appart. The scenes used in this study are from the time period of the tandem mode in 1995 and 1996. The short time interval between the acquisition of the scenes increases the possibility for detection of glacier movement due to less influence of factors like snowdrift and melting.

The method presented here was developed during the work with the first authors cand.scient thesis (equivalent to masters degree) in remote sensing. In the study several hight differentiated INSAR (DINSAR) scenes from the area between Kongsfjorden and Wodfjorden at Spitsbergen were used. Spitsbergen is the largest island of the artic archipelago Svalbard. Svalbard is situated in the North Atlantic between 74° and 81°N and 10° and 35°E. Svalbard is manly covered with ice caps and tundra vegetation. The INSAR scenes were processed at the Norwegian Defense Research Establishment (FFI) with SAR data from an ESA AO Tandem project [2]. The topographic influence in the INSAR scenes was removed during the processing at FFI using a DEM delivered by the Norwgian Polar Institute. These DEMs were also used in the decomposing of the glacier velocities.
METHOD

The method for calculating glacier velocities was developed using images of Isachsenfonna and parts of Holtedahlsfonna and Kronebreen (79°N, 13°E). These glaciers are situated close to Kongsfjorden and Ny Ålesund at the northwestern part of Spitsbergen. Since the topographical effect are removed, the DINSAR scenes only contain information about the movement between the acquisition of the two scenes. The distance between the satellites and a target could be described as a number of wavelengths and a phase. The DINSAR scenes show the phase difference between the two acquisitions. Since one only knows the phase difference, the number of wavelengths constitutes an ambiguity. One of the main problems with velocity calculations is therefore to transform the discrete fashion of the fringe pattern in the DINSAR scenes to continious values. This is called unwrapping. The DINSAR scene of Akademikerbreen is shown in figure 2. The method that has been developed is described in figure 1.

All of the steps have been done in the GIS package ARC/INFO. The phase difference is measured in intervals from 0 to 2 PI. This is represented by graylevels from 0 to 255 in the images. The phase difference increases from 0 to 2 PI, before one gets a discrete transition to 0 again. The discrete fringe transissions must be identified before the INSAR image can be unwrapped to continous values and eventually to glacier velocities.

The median filter was used twice. This removes a lot of the speckle noise that is characteristic for SAR and INSAR-images without any substansial smoothing of the phase transitions.The gradient filter is applied after the median filtering. It is a derivating filter that uses the Prewitt-operator to calculate the gradient in the directions of azimuth and range. The total gradient is then calculated. A value is added to increase the contrast in the image. There is still some noise in the image after the gradient filtering and it is therefore thresholded. The user can define the threshold after looking at the histogram of the images. All pixels with values less than the threshold is set to 0, and the others are kept as they are. A threshold around 150 has shown to be suitable. The selection of a threshold is a choice between the amount of noise reduction and the preservation of the edges between the fringes.

Figure 1: Flow scheme of the method for calculation of glacier velocities.

Figure 2: The DINSAR scene of the glacier Akademikerbreen acquiered 20. and 21. of October 1995.

The thresholded image is then vectorized with an excisting alogrithm in ARC/INFO. Points that are closer than a certain distance are connected with vectors. The resulting data have to be edited to some extent. This is done partially automatic and partially manuel. The method used in the unwrapping is quite simular to the edge segment linking described in [3].

It is important for the calculations to chose the right borders for the glacier. The border of the glacier is digitized manually after the median filtering to limit the amount of calculations that has to be done. It is best to use one of the amplitude images, the DEM and the DINSAR scene are more difficult to apply.

When the lines are edited, one can create the topology for the polygons that mark the transitions between the fringes. The polygons are numbered after their relative position. The outermost polygon gets the value 0, the next, 1, and so on. This information is used in the unwrapping of the INSAR image. This could ideally have been done automatic, but since some interpretation is needed, it is done manually. During the unwrapping of the DINSAR scene, half a wavelength (2.83 cm) is multiplied with the value of the polygons and added to the values in the DINSAR scene.

If the results from the unwrapping is divided by the sine of the look angle, i, (V_sat = U_W/ sin(i)) one gets the component of the glacier movement horisontally towards the satellite. This is often referred to as the line in sight component. For a glaciologist a velocity calculated towards a satellite is of limited use.

Since the Norwgian Polar Institute had terrain models available, the velocities were decomposed in the flow direction of the glacier. Surface parallel flow was assumed. The slope is calculated together with the aspect or slope direction. This is done automatically with a gradient filter. Before the terrain parameters could be calculated, substantial smoothing of the terrain model had to be done. In ground scale the size of the moving mean filters were several kilometers. The look angle was also calculated. It varies between 20 and 26 degrees from top to bottom in the image.

When these three parameters have been calculated, together with the glacier velocity towards the satellite, it is possible to calculate the velocity in the flow direction of the glacier (pers. com, Knut Eldhuset, FFI):

V_glac = U_w/(cos(s)cos(a)sin(i)+cos(i)sin(s))
U_w are the values in the unwrapped image, s is the slope of the glacier, a is the angle between the aspect of the slope and the range direction of the satellite and i is the look angle. Alle the parameters are calculated for every pixel at the glacier surface resulting in a complete velocity field for the whole glacier.
RESULTS

The described method has been used to calculate the surface velocities for three glaciers at Spitsbergen. Some data from the results is presented i table 1. Figure 3 shows the calculated velocity field of Akademikerbreen.

Isachsenfonna is situated east of Kongsfjorden. Iscachsenfonna is an ice cap with several outlets. The velocity is calculated for one of them. The maximum velocity is 42.3 cm/day from the images acquired in January 96. The average velocity is 22.5 cm/day. For the images acquired in April 96 the maximum velocity is 42.0 cm/day and the average velocity is 18.2 cm/day. In the transition between Holtedahlsfonna and Iscahsenfonna the calculated velocities are in agreement with GPS measurements done by the Norwegian Polar Institute. These shows movement of 10 cm/day.

Akademikerbreen is situated on Olav V Land on western Spistbergen. The maximum velocity is 40.9 cm/day and the average velocity is 7.25 cm/day. There is a significant

Table 1: Data from the velocity calculations .

Glacier	Location	Date	Vmax	Vmean
Isachsenfonna	79°N13°E	10/11.10.96	42,3	22.5
Isachsenfonna	79°N13°E	05/06.04.96	42,0	18.2
Akademikerbreen	79°N19°E	20/21.10.95	40,9	7.25
Nordbreen	80°N16°E	27/28.09.95	36,3	16.7
Nordbreen	80°N16°E	05/06.04.96	35,8	12.5
Nordbreen	80°N16°E	10/11.05.96	32,5	10.2

Figure 3: The velocity field of Akademikerbreen obtained from a INSAR scene from September 1995. The arrows indicate the orientaion of the map and the direction of glacier flow.

increas in the velocity in the narrow passage between two nunataks. In this passage the surface slope is also steeper than for the rest of the glacier.

Nordbreen is a smaller glacier in the nortwestern part of Spitsbergen, by Wijdeforden. This is an outlet of the ice cap Åsgårsfonna. The maximum velocities varies between 32.5 and 36.3 cm/day and the average velocities varies between 10.2 and 16.7 cm/day. The maximum velocities are found in the areas of the greatest slope.

The maps show velocity patterns like one would expect from glaciological theory. The velocity increases when surface slope increases, and the greatest changes in velocity is found along the edges of the glacier, since the edges excerts a drag on the glacier. Simular calculations of glacier velocity using external or interferometric generated DEMs have been done by [4], [5] and [6].

CONCLUSIONS

The results from this work show how glacier velocities at Spitsbergen can be calculated with terrain models and height differentiated tandem INSAR data. A semi-automatic algorithm is developed in order to calculate glacier velocities in the flow direction , resulting in maps of the velocity fields for the glaciers. Velocity fields for the three glaciers Isachsenfonna, Akademikerbreen and Nordbreen have been calculated.
ACKNOWLEDGEMENT

The SAR raw data are part of an ESA AO Tandem project carried out at FFI. The final part of this work was in part funded by the EC Environmental Climate Research Program through the project "The Response of Artic Ice Masses to Climate Change - ICEMASS", contract no. ENV4-CT97-0490.

REFERENCES
[1] Goldstein, R.M., Engelhardt, H., Kamb, B. and Frolich, R.M. "Satellite Radar Interferometry for Monitoring Ice Sheet Motion: Application to an Antartic Ice Stream", Science vol.262, pp. 1525-1530, 1993.
[2] Eldhuset, K., Aanvik, F., Aksnes, K., Amlien, J., Andersen, P.H., Hauge S., Isakson, E., Wahl, T. and Weydahl, D.J. "First results from INSAR processing on Svalbard" in Proceedings of ESA SP-406: Fringe'96, Zurich, Switzerland, 30 September - 2 October, 1996.
[3] Lin, Q., Vesecky, J.F. and Zebker, A. "Phase Unwrapping Through Fringe Line Detection in Synthetic Aperture Radar Interferometry", Applied Optics, vol.33 no.2 pp.201-208, 1994.
[4] Rignot, E., Jezek, K.C. and Sohn, H.G. "Ice Flow Dynamics of the Greenland Ice Sheet from SAR Interferometry", Geophysical Researsch Letter vol.22, no.5 pp.575-578, 1995.
[5] Kwok, R. and Fahnestock, M.A. "Ice Sheet Motion and Topography from Radar Interferometry", IEEE Transactions on Geoscience and Remote Sensing, vol.34, no.1,1996.
[6] Mohr, J.J., Reeh, N. and Madsen, S.N. "Three-dimensional Glacial Flow and Surface Elevation Measured with Radar Interferometry", Nature vol.391, pp.273-276, 1998.