WS
DE
a at
vere
tion
992
ons
tion
site
11S
2).
uU.
™~
Has the land cover \ No
of the polygon J
changed?
Yes
N N code?
NS gE DH C S fora
erit
No d
es 8 new date
( Where is the new T A
boundary line?
NAT A
>» LT Land Cover
: N ^
Has the entire Yes What is the new
& polygon changed? CORINE —= database
CORINE
Figure 2: Decision making process in updating of CLC database
Changes can be visualised by printing evolution maps and can be
summarised using statistical tools.
4.1 Identifying significant changes
Separate delineation of an object on both dates causes differences in
the geometry of corresponding polygons. Although the updating
procedure was performed through modifications in the original
(CLC100/1992) database, the iterative procedure of corrections and
update resulted sliver polygons in the change database. It was
necessary to:
Table 3
Area changes (ha) using different threshold cnteria
threshold] A =4ha A=10ha | A=25ha loss
code A/P =30 %
* : + - + - TES.
112 409 0 339 0| 225 0| 45| -
121 16 0 16 0 0 100| -
122 0 0 0 0 0 0 | -
131 63| 152 $7| :145\ 5711221 +10} 20
132 0 0 0 0 0 0 4 -
133 Ol : 79 gt 79 0l ..79 - 0
141 0 0 0 0 0l: 20 | -
142 49 0 49 0| 49 og 0. -
211 2376| 2176| 2343| 2030| 2050/16341 14| 25
221 265| 135} 269| 135| 256| 104 3} 23
222 0l- H3 0| 106 0| 106 - 6
231 1225| 2331| 1154| 2306| 982|2070| 20| 11
242 IS). 55 0| 46 0| 44] 100] 20
243 364 64] 321| 64| 270| 0| 26/100
311 1128| 463| 1105| 392|1057| 304| 6| 34
312 86 0 86 01 86; OF Of -
313 81 0 81 0| 81 oü "o -
321 0 0 0 0 gl. .0 | -
324 335| 945] 328| 939} 253| 957} 24| O
411 172]- 89} 172] 80/- 159} 65 8| 27
511 0| 44 0| 39 0| 38 -| 14
512 198| 134| 174| 130| 120| 119| 39| 11
e determine which of the changes were really significant, and
e eliminate small, insignificant polygons.
Significant changes should not by all means have the same area
threshold value as the minimum mapping unit of CLC database. Our
definition of significance is based upon the area and shape of
polygons. Analysing change polygons on the top of satellite images
at scale 1:50 000 we found that a change polygon as small as 3-4
hectares could be meaningful, if its shape tend to be not too
elongated. The ratio of area (A) and perimeter (P) of the polygon can
measure this property. We eliminated all the polygons that have
smaller area than 4 hectares and simultaneously an A/P ratio less
than 30 meters. Polygons not fulfilling these criteria were dissolved
to the neighbour with the longest common boundary. Table 3 shows
summarised changes by using different area threshold values in
eliminating insignificant polygons. Area increase (+) and decrease (-)
for all classes are shown in separate columns. The last column in
Table 3 includes area loss in case the threshold is 25 ha instead of 4
ha.
4.2 Spatial analysis
If we have the land cover change database, all changes can
potentially be visualised. However, regarding the great number of
transition types, a particular evolution map should answer specific
questions, like:
e Where have changes been between the two dates?
e What are the locations of changes (area increase and/or
decrease) of class “1” between the two dates?
e What are the locations where class “i” areas have developed to
class ^3?
Several maps were printed to visualise the above types of changes.
4.3 Statistical analysis
Statistics are specific summarised descriptors of variables, reflecting
different aspects of the same phenomenon. In our case area and the
number of polygons representing CLC classes are the main objects
of interest. Let us use the following notation:
Mi, - number of disappeared class “i” polygons from date T to T,
Mi; - number of new class i" polygons from date T; to T?
Maj ^ Mii t Mi», 2; Man; = 2 Mr, where
Mr - number of modifications in the whole database
International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 687
