Table 4
Area (A) and number (N) of polygons
code Ns Ns As4 (ha) Asx(ha)
112 86 86 11632 12041
121 19 19 1565 1581
122 3 3 186 186
131 11 7 429 340
132 2 2 78 78
133 1 0 79 0
141 1 47 47
142 4 4 136 185
21i 60 51 86919 87119
221 16 21 2047 2177
222 10 9 741 628
231 112 114 13704 12598
242 41 40 6885 6845
243 51 53 4771 5071
311 67 63 19184 19849
312 6 7 236 322
313 6 7 197 278
321 5 5 163 163
324 26 17 1397 787
411 18 20 1653 1736
511 l 1 414 370
512 14 15 1136 1200
Total: 560 545 153600 153600
Ay - area of class "1" before updating (date T1)
Ap, - area of class "i^ after updating (date T2)
AT - total area of the database
Nj; - number of polygons of class "i" before updating (date T1)
Np - number of polygons of class "i" after updating (date T?)
Typical questions of statistics are:
e How large is the change in the number of polygons for class
“7
e How large is the area increase/decrease for class “i”?
These questions can be answered easily by comparing summary
statistics of CLC databases for both dates (Table 4).
Table 4 includes e.g. an area growth in non-irrigated arable land
[211] and broad-leaved forest [311] classes, while the number of
polygons in those categories have decreased The area of
discontinuous urban fabric [112] increased, while the polygon
number remained unchanged. The area loss was noticeable in
pastures [231], transitional woodland-scrub [324] or orchard [222]
categories, and the single occurrence of construction site [133]
disappeared.
e How large is the change in the area/number of polygons
between class “i” and class ^j" ?
The evolution matrix (contingency table) as the most detailed
statistical descriptor can answer these questions (Table 8). Diagonal
values in the evolution matrix indicate no-change, while off-diagonal
values show transitions between CLC classes during the eight years
Table 5
Normalised changes (ordered)
code | Asa-As4 | [Asz-As4]/As4 | rate of change
(percent) (percent/year)
133 -79 -100,00 -12,50
324 -610 -43,67 -5,46
131 -89 -20,65 -2,58
222 -113 -15,23 -1,90
511 „44 -10,62 -1,33
231 -1106 -8,07 -1,01
242 -40 -0,59 -0,07
141 0 0,00 0,00
321 0 0,00 0,00
132 0 0,00 0,00
122 0 0,00 0,00
211 200 0,23 0,03
121 16 1,02 0,13
311 665 3,47 0,43
112 409 3,51 0,44
411 83 5,01 0,63
512 64 5,60 0,70
243 299 6,28 0,78
221 130 6,37 0,80
142 49 35,69 4,46
312 85 35,97 4,50
313 81 41,34 517
period. Some categories, like road and rail network [122], dump sites
[132], green urban areas [141] or natural grassland [321] remained
unchanged. High values in the matrix (and in similar matrices for
percentage of area or number of polygons), show significant
transitions, €.g. between non-irrigated arable land [211] and pastures
[231], or broad-leaved forest [311] and transitional woodland-scrub
[324]. The matrix shows many area-exchanges between agricultural
land [211, 221, 222, 231, 243]. The non-irrigated arable land [211],
pastures [231] or broad-leaved forests [311] classes have been
transformed to several different categories.
e Which are the classes with the most significant area changes
per year?
e Which are the most dynamic classes?
e — Which classes are updated more than the average and less than
the average?
These questions help to make a reliable comparison between changes
derived for areas of different size, different time span or both. To
answer them, specific indicators were computed (ETC/LC, 1997).
. Relativeareachange: = C;=| Ai2z— Air |/ Air
It answers the question: Which are the most increasing or
decreasing classes in terms of area under the updating period?
This indicator can be normalised using the number of years
elapsed between T; and T? (Table 5). It shows 100% relative
area loss in case of class 133, which is the consequence of
transformation of the single construction site polygon into
settlement [112]. A significant relative area loss is observable in
transitional woodland- scrub [324] or mineral extraction site [131]
688 International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
It ansı
relative
Table
most à
absolu
but the
can un
arable
single
therefc
woodl:
e R
It ansv
> 1) or
extract
lakes
frequei
settlen
then th