retrieval from ATM data (Abdalati and Krabill, 1999), which 
holds promise for highly detailed studies of individual glaciers. 
The method for velocity determination relies on the tracking of 
distinguishable elevation features over time. Because the 
scanner, unlike a profiler, measures over a swath rather than 
along a single line, many features can be identified 
unambiguously in a single-survey, and again, somewhere 
downstream in a second survey. This requires that the terrain 
be of sufficient topographic variability (about one meter rms 
roughness), and the surface topographic characteristics are 
sufficiently preserved in the time period that separates the 
surveys. Moreover, if the repeat survey time separation is on 
the order of a few days, motion must be sufficiently rapid that 
measurement errors are small when compared to the motion 
signal. Largely crevassed, fast-moving glaciers, such as those 
found in some of the Greenland ice sheet drainage basins, are 
well-suited for such purposes. 
In order to make comparisons between two sets of scanning 
surveys, the data must first be interpolated onto a consistent 
grid. We choose a 1 meter grid (roughly the size of a single 
laser shot) and perform an inverse distance weighting 
interpolation of the data within 7 meters of the grid point. 7 
meters is chosen because for a normally operating scanner on 
an aircraft flying at 150 m/s, it represents one half of the largest 
linear gap in the data that should occur. Once the data from the 
pair of surveys are interpolated onto a consistent grid, their 
elevation features are compared to one another to determine the 
offset between features. This comparison is done using the 
method and software of Scambos et al., (1992). In the earlier 
of the two “images”, a small sub-sample region, ranging from 
16 to 128 pixels (meters in our case) on a side, is identified, 
and its nearest match in the second image, within a pre-defined 
search area, is found. The offset is determined and is then 
converted to a velocity by dividing the distance by the time 
elapsed between surveys. 
Figure 2 shows an example of a segment of a pair of 1997 
surveys in the Jakobshavn Isbrae (Location A in Figure 3) that 
are separated by six days (May 13th — May 19th). Distinct 
elevation features, crevasses in particular, are clearly visible in 
both elevation images with an easily detectable offset between 
the first and second. Superimposed on the image are arrows 
showing the magnitude of the offset between the two, and the 
corresponding velocity scale. 
Some of the flight lines over the faster parts of the Jakobshavn 
area were repeated (twice) for velocity measurements again in 
1998, along with two regions on the east coast (Figure 3). 
Coincident with the altimetry measurements were thickness 
measurements made with the University of Kansas ice- 
penetrating radar (Chuah, et aL, 1996). In the cases where 
velocities and ice thicknesses can be successfully retrieved 
across a glacier, flux estimates can be made. Such estimates 
require sufficient swath width (in relation to flow rate) so that 
features can be matched across the width of the swath in the 
survey pairs. Retrieval of thickneses in steep-walled glaciers, 
however is particularly challenging because of clutter in the 
radar imagery due to reflections off the bedrock walls (S. 
Gogineni, personal communication). In the case of the 
Jakobshavn ice stream, thickness retrieval is further 
complicated by the extreme depth of the channel and the 
temperate character of the ice. 
  
  
  
  
May 13, 1997 
H Tr 
0H 3 
. H À 
ë | 
s nm ] 
?00rF 1 
ss E 
Gh 7 
zZ ] 
2 r À 
2n 
i MOL. 5 
S ; 
Q L 14 
z r N 
ob, i 4 1 i L i L j 
0 200 400 600 800 
East-West Distance (m) 
May 19, 1997 
  
200 
S 
CLRRIILEI IIR 0 
North-South Distance (m) 
    
  
  
Ww 
l4 
» 
= 
200 400 600 800 
East-West Distance (m) 
Figure 2. Interpolated elevation data from ATM measurements 
over a small section the Jakobshavn Ice Stream (location A in 
Fig. 3). Elevations range from a low of 88 m (blue) to a high 
of 113 m (red), and offsets of the features is apparent between 
the two images. The arrows represent the displacement over 
the intervening six days. Reprinted with some modification 
from Remote Sensing of Environment, Vol. 67, Calculation of 
ice velocities in Jakobshavn Isbrae area using airborne laser 
altimetry, W. Abdalati, and W.B. Krabill pp. 194-204 
Copyright (1999) with permission from Elsevier Science. 
  
eo proies 
The 1998 flight trajectories are shown in Figure 3 along with 
the regions for which ATM-derived velocities were studied. 
The figure shows the Jakobshavn region in the west, where 
repeat surveys were made in 1997 and 1998, as well as two 
areas in the east, Kangerdlugssuaq Glacier, and the ice margin 
at 65°N. 
4.2 Results and Discussion 
Velocity vectors for the areas of survey pairs are shown in 
Figure 3, and the vectors are overlaid on ERS Synthetic 
Aperture Radar imagery (© European Space Agency, 
Copyright 1992). In the vicinity of the Jakobshavn Isbrae, the 
flow character is very clear in the flow vectors. The boundaries 
between the ice streams and the more stagnant ice are easily 
distinguished by the changes in direction and magnitude of the 
  
   
   
      
Internationé 
flow vectors. The ma 
nearly 7 km per year. 
Similar views are give 
Glacier and the ice mai 
transects yield the best 
is because the region 
these flights were flow 
1998 AIM 
  
  
Figure 3. Maps of 1998 
in Case A, the Jakobsh 
imagery is from the Dig 
Ice Data Center a he U 
Copyright 1992), by M 
Calculation of ice velo 
Copyright (1999) with I