Ice Flow Dynamics of the Greenland Ice Sheet from SAR Interferometry

. Synthetic-aperture radar (SAR) interfero-grams produced from ESA's ERS-1 satellite, provide the first synoptic view of ice flow dynamics of the western sector of the Greenland Ice Sheet. Glacial motion is detected in the radar ranging direction at millimetric scales, across a complete sequence of snow accumulation and melting regimes, despite significant variations in their radar scattering properties. Ice flow evolves from a slow, regular motion at the higher elevations. At lower elevations, motion is strongly convoluted by meter-scale undulations in surface topography, which have a unique interferometric signature that enables a novel approach for retrieving flow direction. Inferred flow directions, combined with surface displacements in the radar ranging direction, yield ice velocity estimates that are within 6 % of in-situ measurements gathered along a 40 km survey line. Application of repeat-pass SAR interferometry to the entire Greenland Ice Sheet should its ice dynamics at an unprecedented level of spatial detail.


Introduction
Earth's great ice sheets are changing [Bentley, 1993;Van der Veen, 1991], and these changes can be relatively rapid. The western extent of the Greenland Ice Sheet has measurably varied on generational time scales [ Weidick, 1991] possibly associated with varying rates of surface melt along its western flank [Braithwaite, 1993].
The interior of Greenland is also changing. Variability in precipitation and accumulation patterns [Bromwich et al., 1993;Steffen et al., 1993] may be causing a regional ice sheet thickening, at a rate of 2-10 cm per year based on surface observations [Van der Veen, 1993; Kostecka and Whillans, 1988;Reeh and Gundestrup, 1985], or as much as 10-20 cm per year based on analysis of satellite data [Zwally et al., 1989]. Copyright  Changes in ice sheet dynamics are manifest through changes in ice sheet shape and motion. Satellite radar interferometry is one technique that shows promise for measuring very small variations in both parameters [Zebker et al., 1994;Goldstein et al., 1993]. Here, we report on the first SAR interferometry results obtained over the western flank of the Greenland Ice Sheet. Interferograms are presented along a swath that extends from the Jacobshavn Glacier through the ablation and soaked snow facies, to the percolation facies, and to the dry snow facies [Benson, 1962] (Fig. 1). The white, snow-covered surface is nearly featureless in optical and SAR-intensity imagery, but highly detailed information on ice motion is revealed in the SAR interferograms.

Observations
To obtain an ERS-1 SAR interferogram, two images acquired a few days apart along the same orbit of the spacecraft, are registered with sub-pixel accuracy, and the phase values of the radar signals are differenced, condensed and averaged together. The measured phase shifts, 5q5, are conditioned by the radar wavelength A (5.6 cm), the baseline separation, B, between the two slightly different positions of ERS-1 when the radar data for the two images were obtained, the surface topography, H, and the surface motion vector, V, as  (Fig. 2). They also help interpret the fringe pattern west of the camp as primarily caused by a 15 ø change in flow direction.
Along the survey line, we converted the surface displacements into ice velocities using the flow direction indicated by pairs of concentric circles both north and south of the survey line and linearly interpolated in between. We assume a surface flow parallel to the ice sheet surface. Surface slope is computed from the digital elevation model. The results are within 6 % on average of the in-situ measurements (Fig. 4). The difference between the two estimates reaches 15 % in the soul•hern section of the survey line. Surface slopes would have to be in error by more than 2 ø to explain those differences, but surface slopes usually do not exceed 1 ø. Flow directions would have to be in error by up to 10 ø, but in-situ

Conclusions
Interference fringes can be constructed across regions of dramatically different snow properties. It is also possible to estimate ice flow direction and ice velocity with a single interferometric observation in areas that contain km scale bumps and hollows, several 10's of rn in height. In other areas, surface measurements of at least two velocity vectors several tens of kilometers apart, are essential to bound the interpretation of the SAR interferometry. Once the validation is in hand, the detail on ice velocities derivable from spaceborne SAR interferometry is unprecedented by any other technique.