Ian G. Cumming and Joe X. Zhang

Dept. of Electrical and Computer Engineering
The University of British Columbia
2356 Main Mall, Vancouver, B.C.
Canada V6T 1Z4

ianc@ee.ubc.ca, joez@ee.ubc.ca
http://www.ee.ubc.ca/sar/

ERS tandem mode differential SAR interferometry (InSAR) has been widely applied to the measurement of displacement or topography of the earth's surface. One interesting example is monitoring the flow of glaciers. In this paper, we resolve the InSAR-measured flow by assuming that the glacier's flow direction is parallel to the surface and to the medial moraine. Both ascending and descending pass data of the Lowell Glacier are processed to obtain the flow rate of the glacier surface. The combination of these two results, along with the assumptions above, can give the 3-D flow rate of the glacier.

Glaciers are important components of the hydrological cycle in mountainous areas. They are key indicators of climate change, and also are important natural resources that must be monitored and managed. An understanding of the process, which could lead to such change, is hindered by the inability to make accurate measurements of the state of the glaciers. Ground-based measurements are capable of sampling only limited locations at favourable times of the year and the costs of field logistics and personnel are not always affordable. Satellite remote sensing techniques offer the potential for dense and wide-scale monitoring of glaciers at many times of the year. Some research has been done to estimate the surface displacement of glaciers using differential space-borne InSAR techniques [Goldstein:93, Joughin:96, Vachon:96b, Rott:96, Cumming:96].

Interferometric Synthetic Aperture Radar (InSAR) is the process of extracting geophysical information from the phase differences between two or more SAR images. Since Graham [Graham:74] first demonstrated InSAR by using an airborne XTI SAR system in 1974, a lot of interesting research work has been done to examine the potential usefulness of InSAR systems. Major progress has been achieved to make digital elevation models (DEM) of the imaged area of the earth's surface by measuring the phase differences between two registered SAR images, taken with a shift in the viewing angle of the sensor. Accuracies of up to 10 m rms for height measurement over areas of relatively flat, stable terrain has been quoted [Massonnet:93, Askne:93].

In order to estimate the flow rates of glaciers, we need to obtain data with a suitable time interval between the two registered SAR images, then try to separate the phase due to the displacement between the data takes from that due to terrain topography. Sometimes this is not easy to accomplish accurately. Ideally, to get the phase due to the surface displacement only, the viewing angle shift (normal baseline) should strictly be zero and the displacement gradient should be small with respect to the radar wavelength. But this condition is rarely achieved in practice, as baselines are invariably non-zero. However, if the topography information is known, we can still derive the phase part due to surface motion [Vachon:96b].

The time interval between two successive SAR images which cover the same area of the earth's surface is 1, 3, 35 or 165 days for the ERS SARs, but the orbit is not controlled accurately enough to obtain the desired viewing angle for all pairs. Environmental conditions or glacier movements sometimes cause loss of coherence, which can even make registration impossible. So, in order to make accurate measurements of the flow rates of the glaciers, we choose C-band ERS1/ERS2 Tandem Mission data with one-day repeat overage. The potential accuracy of this method can be 4 cm/day rms for the glacier flow rate measurement [Vachon:96b].

In this paper, we select the Lowell Glacier near the border of British Columbia and the Yukon in Canada (N 60.3^o W 138.3^o) as our research site. We use both ascending and descending data to solve the 3-D flow rate of the glacier under the following assumptions:

• the glacier flow direction is parallel to its surface, and
• the glacier flow direction is aligned with the medial moraine.

The Lowell Glacier is a typical alpine glacier in the Yukon near the British Columbia border. We study an area around 5 km wide and 30 km long, running approximately from the west to the east extending to the terminus of the glacier.

Photo of the terminus of the Lowell Glacier taken from across the Alsek River (courtesy of Parks Canada).

Table 1 Summary of ERS-1/2 passes over the Lowell Glacier.

In order to achieve reasonable estimation accuracy, we require that the scene coherence magnitude be high enough (say, r > 0.4). So we selected the four highest-coherence pairs of data from the above table to continue our experiment. Figures 2 and 3 are representative interferogram maps for descending and ascending scenes.

Figure 2 Representative descending pass interferogram from October 22/23, 1995.

Figure 3 Representative ascending pass interferogram from February 16/17, 1996.

Because no DEM data is available on the Lowell Glacier, we use the 1:50,000 Canada topographic map (115B7-8/115C5) published in 1987. The elevation along the glacier centerline is shown in Figure 4. If the relative elevation accuracy read from the map is 10 m, a relative displacement error of 0.08 cm is made with a 100-m baseline.

Figure 4 Elevation along the glacier centerline.

At the toe of the glacier, we find a rock area and consider the displacement to be zero at this point. After registering the map data to the interferogram, the phase difference can be calculated by subtracting the phase due to topography from the interferogram. Then the phase change is converted to radar line of sight (LOS) displacement. Figures 5 and 6 show the LOS displacement dR along the centerline of the glacier for the descending and ascending passes respectively.

Figure 5 LOS displacement along the glacier centerline derived from the descending pass data

Figure 6 LOS displacement along the glacier centerline derived from the ascending pass data.

In this paper, we have shown that spaceborne SAR interferometry can be used to estimate the flow rate of the typical high-latitude alpine glacier. We measured a peak flow rate of 50 cm/day on the Lowell Glacier. The combination of ascending and descending passes can help understand the mechanism of the glacier surface movement, if both the ascending and descending pass scenes have high coherence, the glacier flow directions are within 70^o of the radar LOS and the assumption of constant flow is valid. We are now processing additional data sets to verify the measurement technique and further check the consistency of results.