-———
tion. The
y of the
Table 2
of the
ludes 44
ree level
surfaces,
reas, 4)
adapt the
national
purposes it is allowed to subdivide further any of level-3 elements of
the nomenclature. E.g. two level-4 categories were used in Ireland to
characterise pastures (2.3.1) of different quality (O'Sullivan, 1992).
Similarly, two different types of inland marshes (4.1.1) and peat
bogs (4.1.2) were defined in Estonia (Aaviksoo, 1997) The
European database however includes only level-3 categories.
Special features of the nomenclature are the categories of
"Heterogeneous agricultural areas". They are formed by objects, (e.g.
plots of arable land, areas of natural vegetation, etc.) which alone
would be smaller than the minimum mapping unit (25 hectares). E.g.
category 2.4.2 have béen introduced to characterise mixed
agricultural areas: mixtures of any two of the following cover types:
arable land, pastures, vineyards, fruit trees and berry plantations.
Class 2.43 is to characterise agricultural areas with significant
amount of natural formations (e.g. patches of forests, areas of scrub,
grasslands, wetlands or water bodies). These are very useful tools to
characterise a heterogeneous landscape at scale 1:100.000.
3. LAND COVER MAPPING AT SCALE 1:50.000
Standard CORINE Land Cover data are especially useful at
European level. To satisfy regional or local needs better, more details
are needed both in terms of geometry as well as in thematic content.
Several initiatives exist to extend CORINE Land Cover
methodology into working scale of 1:50.000 and even 1:20.000
(ETC/LC, 1997b). In the frames of the Phare programme an
experimental project has been executed at scale of 1:50.000 for 120
map sheets in four countries: Czech Republic, Hungary, Poland and
Slovak Republic. It was possible to use 4 hectares as minimum
mapping unit using integrated SPOT PAN and Landsat TM data.
International team of experts has extended the standard
nomenclature with level-4 classes representing the landscape
conditions of the above four countries (Feranec et al, 1995). The
number of level 4 classes was about twice of level 3 ones.
The experimental project proved the possibility of CORINE Land
Cover mapping at larger scale. One of the ongoing activities of
PTL/LC is to extend further the level-4 nomenclature including all
Phare countries. New version of the CORINE Land Cover
nomenclature at scale 1:50 000 for Phare countries has been finished
in May 1998 in the framework of PTL/LC activities. This
nomenclature includes 104 land cover classes and could be a base of
an all-European level-4 nomenclature. Main benefits of the extended
nomenclature and the 4 hectare minimum mapping unit compared to
the standard CORINE Land Cover mapping are:
e much more thematic detail in the "artificial surfaces" group,
which has the strongest impact on the environment,
e agricultural categories support better agrostatistics and the
needs of habitat mapping,
e more discrimination in forests and semi-natural vegetation and
in wetlands, which are important for nature conservation and
biotope mapping,
e decreased percentage of heterogeneous agricultural classes,
because of using smaller minimum mapping unit.
4. UPDATING AND CHANGE DETECTION
Updating is a central question of any databases including
features, which change in time. CORINE Land Cover database
can fulfil its aims only if the database is updated regularly. The
proposed average updating frequency of CORINE Land Cover
database is 10 years. This doesn’t mean that changes can not be
faster in certain areas (e.g. urbanization). Having land cover data
for more than one date, one has a possibility to analyze land
cover changes and to make predictions for the future.
4.1 Updating
CORINE Land Cover mapping is a human labour intensive
methodology, requiring skilled photointerpreters. Because of the
nature of nomenclature and the rules if interpretation, updating
also can not be automatic. Due to the fact however, that land
cover changes are generally slow, there is no need to repeat the
interpretation in the course of updating, only to recognise
changes what have happened from one date to the other. Having
a proper computer support, this process is evident for a
photointerpreter, familiar with the CORINE methodology.
Therefore updating is significantly cheaper than producing the
basic database.
The updating process is based on the computer-assisted photo-
interpretation (CAPI) technology, with simultaneous use of the
basic CORINE Land Cover map, the corresponding satellite
image map, and the new satellite image map. Most important
features of the necessary CAPI software are: raster background
handling capabilities, geographically linked multi-window
environment, ability to edit different databases in different
windows, building up and checking of the database topology and
general image processing capabilities. In addition to several
commercial GIS/IP processing software that supports this list,
JRC has developed the Co-Pilot (CORINE Photo-Interpretation
Land Cover Oriented Tool) software, which includes additional,
specific CORINE related features (Perdigao and Annoni, 1997).
The updating procedure usually reveals errors in the original
database that first should be corrected, in order to avoid detection
of false changes (PTL/LC, 1998; Biittner et al., 1998)
4.2 Evaluation of changes
Once we have produced the CORINE Land Cover database for
dates T1 and T2, change detection is an automatic procedure. The
change database includes polygons with attributes related to Ti
and T2. The change database can be visualised by printing
evolution maps and can be summarised using statistical tools.
The evolution matrix (contingency table) is the most detailed
statistical descriptor of summarised area changes between the
two dates. Its diagonal elements represent areas of no change,
while off-diagonal elements relate to area changes between T1
and T». Having 44 level-3 categories, the maximum size of
evolution matrix is 44*44. (In the practice the evolution matrix
includes lots of Os, because of impossible transitions between
several category pairs). Summary statistics for Ti and T», area
change for each category and total change can be derived from
the evolution matrix.
There are some other useful indicators of changes (ETC/LC,
19970):
e The normalised relative area change answers the question:
which are the classes with the largest area increase or
decrease per year?
e The relative occurrence change answers the question: which
are the most dynamic classes relative to their original
frequency of occurrences?
e The relative updating frequency (the ratio of the proportion
of modifications and the proportion of area of a given class)
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 637
