equirements 
e concluded 
ility to cop 
lassification. 
with ERS-|. 
mparatively 
sifications is 
es of the e. 
bers. The re. 
he sampling 
fully trained 
nd exclusion 
ng. 
| ERS-1 SAR 
s. Biometrics 
egorical mar- 
of thematical 
ering and Re- 
g the accura- 
lata. Remote 
Calibrating 
ing confusion 
SPRS Annu- 
Calibration 
estimates for 
of Environ 
anutzung mit 
Dissertation 
at Stuttgart 
pper inagery 
Environment 
eme for esti- 
assifications. 
scheme for 
al data with 
echnometrics 
  
Automatic Approach to Detect Ice Sheet Margin 
Using ERS-1 Synthetic Aperture Radar Imagery 
Hong-Gyoo Sohn* and Kenneth C. Jezek 
Byrd Polar Research Center and 
* Department of Civil and Environmental Engineering and Geodetic Science 
The Ohio State University 
Columbus, OH 43210, USA 
Commission VII, Working Group 8 
KEY WORDS: SAR, SPOT, Aerial Photography, Change Detection 
ABSTRACT 
A portion of the Greenland ice sheet margin near the Jakobshavn glacier is mapped using an automatic approach 
applied to ERS-1 Synthetic Aperture Radar (SAR) imagery, visible imagery (SPOT), and a digitized aerial photograph. 
This process involves integration of an anisotropic diffusion algorithm, a local dynamic thresholding method, and a 
recursive edge following scheme. 
1. INTRODUCTION 
Polar ice sheet margins are sensitive indicators of 
changing climate. For example, melting at the margin of 
the ice sheet in Greenland will increase by 0.5 m water/a 
for each degree increase in summer temperature 
(Braithwaite and Olsesen, 1993). Based on repeated 
mapping of glacier systems in West Greenland, glacier 
margin advances and recessions of the ice sheet margin 
parallel prevailing climatic trends (Weidick et al., 1992). 
This suggests that consistent and long-term observations 
of the ice sheet margin change may provide a useful 
indicator of changing climate. 
An ice sheet margin detection algorithm is implemented 
to monitor and detect ice sheet margin fluctuations using 
visible and microwave imagery. A portion of the 
Greenland ice sheet margin near the Jakobshavn glacier 
(Fig. 1) is mapped using an automatic approach applied 
to ERS-1 SAR imagery, visible imagery (SPOT), and a 
digitized aerial photograph. 
2. STUDY AREA AND DATA SETS 
The aerial photograph was taken on July 10, 1985. The 
SPOT panchromatic image and the ERS-1 SAR image 
were obtained May 30, 1988 and August 20, 1992, 
respectively. The coverage area of the SAR image is 
about 100 x 100 km. The coverage area of SPOT and the 
aerial photograph is about 60 x 60 km and 37.5 x 37.5 
, Tespectively. 
627 
  
tn SL EE 
  
  
  
Liem 
Fig. 1. Map of Greenland and study area 
  
3. APPROACH 
Image tone and texture variations are used to map the ice 
sheet margin. Fig. 2 schematically depicts the approach 
for detecting ice sheet margin using the imagery. 
The ice sheet margin is detected by sequential application 
of several different algorithms. First, the original ERS-1 
SAR imagery is geocoded and terrain corrected using the 
available digital elevation model produced in Denmark at 
Kort- og Matrikelstyrelsen (KMS) (Tscherning et al., 
1992) (Fig. 3). The SPOT imagery and the digitized 
aerial photograph are also geocoded using ground control 
points from 1:50,000 scale orthophotos (Brecher, 
personal communication) (Fig. 4, 5). All imagery are 
resampled to 50 m resolution. After geocoding, the 
overlapping area is used for all subsequent algorithmic 
processing. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996