7 
s modem landscapes ; 
ipacts ; 4 : traditio- 
1 ,SM,Ti indicate 
cels and probability 
;ape component (based 
pixelsize of 30 m x 
ae text. 
lability of pure 
in the component 
AM 
BR 
AR 
S 
1 
76 
14 
3 
5 
76 
6 
8 
3 
84 
10 
3 
5 
76 
10 
7 
17 
62 
10 
7 
6 
62 
10 
12 
4 
85 
3 
* 7 
10 
63 
18 
4 
3 
89 
8 
0 
13 
63 
11 
5 
7 
56 
17 
14 
11 
72 
5 
9 
9 
55 
13 
12 
11 
57 
10 
10 
7 
63 
10 
15 
6 
63 
6 
17 
3 
71 
15 
3 
3 
55 
6 
10 
13 
65 
0 
0 
5 
91 
0 
0 
9 
48 
8 
5 
Iders, openfields and 
;t probability.. Low pro 
ves (with large AND 
?nsity of biotic 
. A further analysis 
its, based upon these 
nain groups : open 
field sizes) and 
d the compactness of 
se fields allowed the 
comparison of the compactness of the field shape 
with the pixel shape. Important to note is that clas 
sical geographical studies about the field patterns 
are always based upon cadastral fields and thus are 
not very significant for the fields seen by any re 
mote sensing system i.e. units having the same land 
use. Table 4 gives the results for six representati 
ve landscape types. The mean size of the fields va 
ries between 3-6 acres, but very large variations 
occur. Most fields in the sample areas are rectan-. 
cular, the compactness can be expressed easily by 
the ratio between their lenght and width. The varia 
tion of the width is generally larger than the 
length. Although the field width remains almost the 
same for all sites, it is also generally smaller than 
the pixel size. Important differences in compactness 
(L/B ratio) do occur. Most compact are the small 
block fields in the sites of Semmerzake (Se), 
Bazel (fig.3) and St.Joris (SJ). 
Table 4. Compactness and area of landuse fields in 
six typical landscapes (m = mean, s = standard 
deviation, CV = % coefficient of variation). 
Code Length 
Width 
L/B 
Area 
site L 
(in 
m) 
B 
(ir 
i m) 
(sq.m) 
m 
s 
CV 
m 
s 
CV 
m 
s 
CV 
m s 
CV 
Ze 29 
11 
39 
14 
7 
52 
3 
2 
67 
425 357 
84 
Wa 38 
28 
73 
13 
8 
65 
4 
4 
97 
425 348 
82 
Se 24 
10 
42 
13 
7 
52 
2 
1 
45 
345 291 
84 
JM 54 
25 
47 
11 
6 
53 
6 
4 
61 
563 371 
66 
Ba 22 
8 
36 
12 
4 
34 
2 
1 
49 
276 163 
59 
SJ 32 
13 
43 
18 
8 
45 
2 
1 
63 
618 421 
68 
Figure 3. Field size and shape in the test site of 
Bazel (L = length ; B = width) - Land van Waas. 
Nevertheless, the proportion of pure pixels in 
these sites remains poor because of their small 
size (especially width) and the high density of 
biotic screens as field edges. On the other hand 
there are the extremely elongated fields, as in 
the sites Waarschoot (Wa) and St.Jan (JM), which 
are both representative for the sandy parts of 
Flanders (fig.4). In these sites, the fields are 
larger and the landscape is more open. 
Figure 4. Field size and shape in the test sites of 
St.Jan-in-Eremo (JM) and Waarschoot (Wa) - 
Meetjesland. 
4.4. The field orientation 
Allongated fields occur frequently and they are 
grouped into blocks which can easily be recognized 
upon satellite imagery, but which may be different 
from the orientation of the individual fields. 
Even with large compact fields, the proportion of 
mixed pixels can be high, because the field pattern 
is largely discordant with the pixel pattern. 
Field boundaries which have an orientation which 
differ about 45 degrees from the flight path, will 
cause the largest proportionof mixed pixels (fig.5) 
Figure 5. Grientation of the field boundaries in 
the Zerkegem (Ze), Semmerzake (Se) and Waarschoot 
(Wa). Cumulated boundary lenght in meter per angle 
deviation from the NS direction.