A critical parameter of ice sheets is their velocity field, which, together with ice thickness, allows the determination of discharge rates. Remote sensing, using moderate- to high-resolution satellite images, permits glacier movement to be measured on sequential images covering the same area; the velocities can be measured quickly and relatively inexpensively by tracking crevasses or other patterns that move with the ice. Especially important are velocities where the ice crosses the glaciers' grounding lines (locations along the coast where the ice is no longer ground supported and begins to float).
Landsat images are particularly useful because they provide synoptic views covering as much as 185 square km. Thus several fixed points in the scenes, needed for geometric corrections and coregistration of images, are likely to be found. On the other hand, Landsat images have disadvantages: the early Multispectral Scanner (MSS) images have moderate resolution (about 80 m), permitting tracking of only the larger patterns in the floating part of ice tongues or shelves. Thematic Mapper (TM) images have high resolution (about 30 m), but digital TM data are very expensive. Also, the long polar winter night reduces acquisition opportunities, and cloud cover may impede recognition of features. An alternative is ERS SAR images (European Remote Sensing Satellite, Synthetic Aperture Radar), which have 30-m resolution and similar viewing conditions regardless of season or cloud cover. Thus they permit the tracking of small crevasses and other patterns above or at the grounding line.
An extensive set of Landsat images covering Antarctica was acquired in the early to middle 1970s. Since 1984, new Landsat images of Antarctica's coastal regions have been obtained largely through a program sponsored by an international consortium of nations belonging to the Scientific Committee on Antarctic Research (SCAR). A period of 20 years between acquisitions of some of the Landsat images makes them an invaluable resource.
ERS-1 images have been available since mid-1991 in both ascending and descending orbits. They have repeat orbital cycles varying from 3 days to 35 days, and they cover 100 square km on the ground. We herewith acknowledge the support of the European Space Agency (ESA), which makes the images (and tapes) available at no cost to researchers of accepted projects.
Examination of the image pairs showed that many glaciers do not have suitable floating tongues. Tongues on coastlines where ice shelves are narrow or absent tend to be short, perhaps due to vigorous ocean-current and wind regimes. Also, short tongues having distinctive crevasse patterns may break off in a time frame shorter than that between image acquisitions. For these regions, only methods that employ high-resolution images that permit recognition of features near and above the grounding line can be used.
This report summarizes the results of velocity measurements of outlet glaciers, ice streams, and ice shelves around the Antarctic periphery. For some regions, where suitable images were available, the same area was measured repeatedly to validate the data or register changes in velocity with time. The results given here are a compendium of published papers and work in progress. The results constitute a data base that will be added to and amended as more velocity measurements become available.
We generally register Landsat 1, 2, and 3 images to Landsat 4 and 5 images, because the latter have more stable internal geometry and higher resolution than the earlier images. Several tests were made to compare the internal geometry of 3rd and 4th generations negatives with the original digital data. All of these tests, as well as several made between original and scanned images of transparencies, showed an insignificant degree of geometric error between products. These tests demonstrate that geometrical errors within the transparencies will contribute little to statistical variance between measurements. Loss of resolution and misidentification of features play a more important role in measurement error made with these images. Borgeson and others (1985) found that Landsat 5 images are accurate to about 0.4 pixels, meeting national Horizontal Map Accuracy standards for scales of 1:100,000 and smaller, and that Landsat 4 images are accurate to 0.8 pixel levels. Welch and others (1985) reported that Landsat 4 and 5 images meet accuracy standards for maps of 1:50,000 scale or smaller and are well suited to maps of 1:100,000 scale.
For ERS images, we obtain CCTs of the geocoded version (placed in Universal Polar Stereographic projection using the WGS 1984 ellipsoid). The pixel size is 12.5 m on the ground (resolution approximately 30m). The images are coregistered by either (1) matching fixed points such as nunataks (land masses projecting through the ice), or (2) using the furnished coordinates based on orbital parameters. We obtained the same results by both methods, increasing our confidence in the accuracy of the nominal image location, which is supposed to be less than 50 m (Roth and others, in press). For a more detailed error evaluation for Landsat images see Lucchitta and others (1993 and in press a), and for ERS-1 images see Lucchitta and others (1994 and in press b).
We use two methods to determine the glacial velocities: an interactive one in which we visually trace crevasse patterns (Lucchitta and others, 1993) and an autocorrelation program developed by Bindschadler and Scambos (1991) and Scambos and others (1992). First, we digitally co- register the images by using a minimum of three well-dispersed fixed points (such as nunataks or ice walls) to calculate a least-squares fit to a first-order polynomial equation. This insures that only a rotational/ translational correction is made and no new internal error is introduced during the geometric resampling. In the interactive technique, we then match and align the crevasse patterns displaced with time, and record the starting/ending image coordinates for each point. To obtain the distribution of average velocities over the length of the glacier tongues, we also use the distance from the location of each point on the earlier image to a base line drawn perpendicular to glacier movement and ideally lying on the grounding line; where the grounding line is complex, the base line may only approximate its position. Next, a digitized file is made, tracing the glacier ice movements and defining the glacier's baseline (or grounding line). This file is used to calculate the velocity and distance statistics by measuring the displacements along the curve that approximates the ices movement per given time interval. For each measured point, a displacement vector is plotted on the image, commonly the earlier one of the pair, to illustrate the relative velocities between glaciers and time intervals.
Because the velocity field may also change across the glacier tongues, we divide the wider glaciers into several longitudinal paths. Next we obtain an estimate of the spread of measured points by performing a regression analysis on the data. This includes an option to cull bad data points by inputting a variable for the standard deviation. If used, the mean absolute deviation of the points about this line is calculated and any points lying outside that distance are disregarded during the statistical analysis. Calculations are made for the entire glacier as well as for each individual path. The 95% confidence interval for the regression coefficient is calculated along with the correlation coefficient.
The files contained in this data base are the output ASCII files generated by this statistical software. Each file identifies the images used, their dates, and resolutions, the time interval between image acquisitions and the statistical variables used to make the calculations. This data is followed by a table of the distance and velocity values for each point and the statistics calculated per path. The measurement results are shown in graphs that display average velocities per given time interval versus the distance from the base line for all points in each field (not included in this data base).
In the auto-correlation method we use the same techniques for coregistration and graphic and statistical display. However, we may not divide the glaciers into segments and paths, but instead combine all velocities and show variations across the glacier by color contours (also not shown in this report).
Baerbel K. Lucchitta U.S. Geological Survey 2255 North Gemini Drive Flagstaff, AZ 86001-1698 Tel: (602) 556-7176 FAX: (602) 556-7014 email: blucchitta@iflag2.wr.usgs.gov
Janet M. Barrett U.S. Geological Survey 2255 North Gemini Drive Flagstaff, AZ 86001-1698 Tel: (602) 556-7328 FAX: (602) 556-7014 email: jbarrett@iflag2.wr.usgs.gov
Jo Ann Bowell U.S. Geological Survey 2255 North Gemini Drive Flagstaff, AZ 86001-1698 Tel: (602) 556-7114 FAX: (602) 556-7014 email: jbowell@iflag2.wr.usgs.gov
Jane G. Ferrigno Mail Stop 804, National Center U.S. Geological Survey 12201 Sunrise Valley Drive Reston, VA 22092 Tel: (703) 648-6360 email: jferrign@oemg.er.usgs.gov
Kevin F. Mullins U.S. Geological Survey 2255 North Gemini Drive Flagstaff, AZ 86001-1698 Tel: (602) 556-7182 FAX: (602) 556-7014 email: kmullins@iflag2.wr.usgs.gov
Christine E. Rosanova U.S. Geological Survey 2255 North Gemini Drive Flagstaff, AZ 86001-1698 Tel: (602) 556-7022 FAX: (602) 556-7014 email: crosanova@iflag2.wr.usgs.gov
Richard S. Williams, Jr. U.S. Geological Survey Quissett Campus Woods Hole, MA 02543 Tel: (508) 457-2347 FAX: (508) 457-2310 email: rwilliam@nobska.er.usgs.gov