Torlegärd et.al (1984), are used. The result is summarized 
in Table 1. 
Table 1. Karhunen-Loeve expansion of the average correlation 
funetions of the six ISPRS DEM test areas. 
Uppland 8 = 0.625 CP + 0.044 a, - 0.046 d; 
Bohuslün c = 0.698 à, - 0.099 3, - 0.002 d; 
Stockholm © = 0.522 à, * 0.128 q, - 0.055 à, 
Drivdalen 2 = 0.865 9, :719.287 d, + 0.001 dz 
Sühnstetten € = 0.317 §; + 0.503 3, - 0.053 à, 
Spitze 7 ='0.567:4, + 0.175 7 + 0,204 dz 
Variance (Ay /In) 85 % 15 % 2 % 
The variances in Table 1 are calculated using the 
eigenvalues Ak? and are measures of the significance of the 
base functions. They should be read in such a way that by 
using the first two base functions, 98 % of the correlation 
functions can be recovered. 
When comparing the different terrain types, some main 
differences of the coefficients of the first and most 
Significant base function d, can be noticed. The two German 
areas (Sühnstetten and Spitze) have here a rather small 
coefficient, while the Norwegian area (Drivdalen) has a 
fairly large coefficient. The three Swedish areas have a 
first coefficient somewhat between the steep Norwegian area 
and the smooth German areas. The two base functions d4 and 
q, are shown in Figure 1. 
- 686 -