variety
circle.
circle.
rectangle
rectangle
rectangle
rectangle
rectangle
rectangle 10
vegetation
radius 5
radius 6
5 by 5
6 by 6
7 by 7
8 by 8
9 by 9
by 10
1
374,0
337,3
502,9
462,9
428
396,9
369,2
344,4
2
282.5
295,7
200,9
223,7
256,5
274,5
288,0
297,6
3
81.1
103,6
22,7
36,2
51,6
67,5
83,2
98,7
4
13.6
22.2
1,22
2,9
5,7
9,4
14,1
19,7
5
1.6
3,5
0,09
0,32
0,84
1,7
2,9
6
0,07
0,27
0,03
0,08
0,2
variety
circle.
circle,
rectangle
rectangle
rectangle
rectangle
rectangle
rectangle 10
structure
radius 5
radius 6
5 by 5
6 by 6
7 by 7
8 by 8
9 by 9
by 10
1
363,2
327,4
487,8
449,4
415,4
385,1
357,7
333,2
2
263,4
272.9
197,0
223,4
243,5
258,4
269,5
276,8
3
96,4
117,0
37,2
52,6
68,0
83,2
97,9
111,9
4
26,6
38.5
4,2
8,5
14,0
20,5
27,8
35,8
5
5,3
9,2
0,28
0,82
1,87
3,5
5,5
8,0
6
0,71
1,66
0,05
0,15
0,34
0,7
1,24
7
0,11
0,26
0,02
0,05
0,1
0,17
8
0,05
0,03
Table 1: focaldiversitv (in some software "focalvariety ") of vegetation and structure of vegetation depending
on the moving window (4 m resolution) in hectares. The table shows, that the critical question for calling an
area "highly structured" is the determination of a filter size and the absolute values of a variety for each filter
size. For example, a rectangle moving window of 7 by 7 pixels results in 428 ha (out of 726 ha of this small
sample) with a variety of l, which means spatial homogeneity. This is clearly to define. But what is "highly
structured"? Is it values of 2 and more or values of 4 and more?
Table 1 shows how the results are depending on form and size of the moving window. This
simple fact is not always considered. Focaldiversity algorithms are implemented in most of the
relevant GIS-packages and increasingly (and sometimes unreflectingly) used in landscape
ecology. Dillworth et al. (1994) observed that geometric windows may have some
shortcomings when used to access spatial characteristics of nominally scaled data and suggest
the use of more complex 'geographic' windows. Meanwhile, sophisticated software packages
for landscape analysis such as FRAGSTATS (McGarigal and Marks 1994) exist for both
vector and raster data to analyse landscape and to quantify landscape structure. Nevertheless, it
is necessary to point at the importance of understanding the influence of the designation of
matrix elements, cell size, or size of the moving window.
Adding a qualitative aspect
The results of the focalvariety calculation of the layer structure are corrected in order to the
structural value of the classes. On the contrary to the vegetation layer, the survey consists of
qualitative terms such as "rich structured" or "low structured". This qualitative aspect has to be
added. The adding values are determined by experts. Values are recoded to an ordinal level. As
shown above, the absolute values of the resulting diversity depend on the size of the moving
window. Taking the original values would not only be pseudo-accurate but would lead to
wrong results. Nevertheless, recoding is one of the critical points in this method. This step can
influence or manipulate the spatial distribution of the resulting values more than anything else.
Therefore we need experts to establish rules for the reclassification of the resulting values due
to reality. We have to remember the size of the moving window (e.g. 7x7 rectangle at 4m pixel
size means 784 m 2 ). In this context, what means "high" and what means "low"? In this case
study it was tried to let some experts independently search for the answer and the mean value
was taken.
Correlation with the distribution of indicator species: Structural diversity as a measure
for habitat quality?