The reference image set is acquired in a short 
cuvely period, i.e., three months. Thus, the assumption of 
od. A insignificant variations for the area of interest during 
flange the period is reasonable. The images are geocoded 
dus and resampled to 12.5m*12.5m pixel spacing. To 
assure the geometrical quality for the images, a 
registration check was automatically performed by 
matching feature points. The RMSE is smaller than 
0.75 pixels, which corresponds a ground coverage of 
less than 10m. 
The reconstructed reference DTM is shown in figure 8. 
For consistency check , the predicted shoreline from 
the reference DTM and the real shoreline from image 
are compared in figure 9. In which, the error of the 
area is 1.409 km? ie., 5% of the sand bar area. 
The solid line depicts the real coastline and the dash 
line illustrates the predicted coastline from the 
reference DTM. 
  
  
Figure 8. The Reconstructed Reference DTM 
  
= Figure 7. Historical SPOT Image F,1986/09/28 The predicted shoreline from the reference DTM 
  
  
according to the tide elevation associated with the 
Table 1. Sampling Time and Associated Tide historical satellite image i.e. image F is then 
Elevations compared with the real shoreline delineated from 
Sampling Time Tide Elevation (M) image F. Figure 10 illustrates the changes of Wai- 
Reference Image San-Ting for the last eight years. In which, solid line 
ID Y-M-D H-M-S depicts the real coastline from image D and the dash 
A 1994-01-03 10-55-53 -0.5340 line illustrates the predicted coastline from the 
B 1994-01-11 11-00-11 0.6519 reference DTM. Quantitatively, the sand bar has lost 
C 1994-02-06 11-00-14 0.0156 its area of 37% during the last eight years. 
D 1994-03-04 11-00-18 -0.2760 
E 1994-04-17 10-56-00 0.0659 
Historical Image 
F 1986-09-28 10-45-00 -0.5690 
121 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996