> gray level 
Igorithm to 
  
gray level 
algorithm 
ice sheet 
traint used 
f image is 
  
  
  
  
  
  
  
ig. IZ. Final detected boundary using ERS-I SA 
imagery after removing unwanted edges. 
Fig. 13 shows the detected ice sheet boundary draped on 
Fig. 3. Fig. 14 and Fig. 15 represent detected ice sheet 
boundaries for the SPOT and the digitized aerial 
photograph draped over Fig. 4 and Fig. 5, respectively. 
The detected ice sheet boundary matches the visually- 
inspected ice sheet margin to within several pixels (150 
m). 
The three ice sheet boundaries produced from ERS-1 
SAR imagery, SPOT, and the digitized aerial photograph 
are compared (Fig. 16). The upper and lower parts of Fig. 
16 show that the ice sheet margin detected using the SAR 
imagery is several hundred meters outside of the ice sheet 
margin detected using the visible imagery. This is 
mainly caused by the uncertainty in the digital elevation 
model used to terrain correct the SAR imagery. 
In the middle part of Fig. 16, the ice sheet margin 
detected using the SAR imagery shows more variation 
than the visible imagery. This seems to be mainly due to 
different image signatures near the ice sheet margin. The 
SAR penetrates the surface snow and may include signals 
from the surface, volume and underlying rock. The 
visible imagery only maps the surface features. 
The ice sheet margins detected using different data sets 
show that over a 7 year period (1985-1992) the ice sheet 
margin north and south of the Jakobshavn glacier 
fluctuates by less than 350 m. The calving front 
fluctuates up to 2.8 km. 
  
Fig. 13. Detected ice sheet margin using ERS-1 SAR 
imagery (Fig. 10) draped over Fig. 3. 
  
   
Fig. 14. Detected ice sheet margin using SPOT imagery 
draped over Fig. 4. 
E 
   
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996