sites
ined
; for
cant
ures
crub
tural
11],
nges
nges
.To
> Or
iod?
ears
itive
e of
into
le in
131]
Table 6
Relative frequency of changes (ordered)
code Nes Mas Ms, Man Mir
211 60 89 64 153 2,55
511 1 2 0 2 2,00
324 26 21 12 33 1,27
512 14 5 11 16 1,14
133 1 1 0 1 1,00
231 112 59 45 104 0,93
221 16 5 9 14 0,88
131 11 6 3 9 0,82
311 67 25 30 55 0,82
243 51 5 19 24 0,47
411 18 4 4 8 0,44
312 6 0 2 0,33
222 10 3 0 3 0,30
112 86 0 24 24 0,28
142 0 1 1 0,25
313 6 0 1 1 0,17
242 41 4 3 71017
121 19 0 1 1 0,05
141 1 0 0 0 0
321 5 0 0 0 0
122 3 0 0 0 0
132 2 0 0 0 0
classes, but the changes do not concern large areas in the case of
mineral extraction sites. Pastures [231] have a large area loss, but its
relative value is rather small. Significant relative area growth can be
observed in mixed forest [313], sport and leisure facilities [142] and
coniferous forest [312] categories.
Mii 7 Maj/Ni
It answers the question: Which are the most dynamic classes
relative to their original frequency of occurrences (Table 6).In
Table 6 we can see, that non-irrigated arable land [211] is the
most often modified class in the CORINE classification both in
absolute and in relative sense. Pastures were modified frequently,
but the relative occurrence change of this class is fairly low. We
can understand these, if consider the large contiguous areas of
arable land and the plenty of smaller pastures polygons. The
single river (511] polygon was updated at two locations,
therefore its relative occurrence change is high. The transitional
woodland-scrub class [324] shows also a high modification rate.
e Relative occurrence change:
e Relative updating frequency:
I m ( Man i/2*Mt ) / ( Au/Ar)
It answers the question: Which classes are likely updated more (I
> 1) or less (I < 1) then the average (Table 7).
Looking at Table 7, transitional woodland-scrub [324], mineral
extraction sites [131] (mainly gravel extraction in this case) and
lakes [512] show the highest values of relative updating
frequency. On the other hand, deciduous forests [311],
settlements [211] and arable land [211] are updated less than
then the average, because of their large relative area.
: eR E mm Uem rte sc Bi "gs
Table 7
Relative
code Ass
324 1397
131 429
512 1136
133 79
312 236
231 13704
142 136
221 2047
313 197
243 4771
411 1653
S11 414
222 741
311 19184
112 11632
211 86919
242 6885
121 1565
141 47
321 163
132 78
122 186
» »
v
I)
‘Un
1
1
1
1
1
0
0
0
0
©
N
po
5. CONCLUSIONS
A methodology has been implemented to update the CORINE
Land Cover - Hungary database according to the guidelines of
the Joint Research Centre (Perdigao and Annoni, 1997). The
procedure requires the simultaneous analysis of the following
data:
e original land cover database,
satellite images used in production and verification of the
original database,
satellite image(s) for the new date,
ancillary data (topographic maps).
The methodology is suitable to produce a land cover database also
for the past ("downdating"). Applying this approach, we downdated
a map-sheet (48 km * 32 km area) of the CLC100/1992 database for
the year of 1984. In the course of downdating many corrections were
made in the original (CLC100/92) database. The “new” database
(CLC100/84) has been constructed by using computer assisted
photointerpretation (CAPI) technology. Threshold values of
eliminating sliver polygons can strongly influence the final change
database. E.g. polygons of settlements [112] or water bodies [512]
are usually characterised by small real area changes, which can be
lost, if value of threshold is inappropriate.
Changes have been analysed both spatially (change maps) and
statistically. The following main observations can be made:
e We have found 229 cases of significant land cover changes,
affecting 4.41 % of the total area. 95.59 % of the area remained
unchanged.
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 689
