conclus

IMPORTANT LEGAL NOTICE: The information
on this site is subject to a
and a copyright
notice.

 

Some conclusions

The objective of the study
« From Land Cover to Landscape Diversity in the European Union »
is to assess the relevance of the CORINE Land Cover database as a source
of information for landscape characterisation.

Land cover is one of the
several components of landscape. Relief and transportation networks are
other important components for which databases are available at European
level. The latter has been partly integrated into one of the analyses and
further work should be undertaken on both these elements. 

For the present study, the
CORINE Land Cover information was aggregated in a limited number of indices
easily derived by automatic computation. The aim was then to evaluate the
feasibility of working with these indices, and to assess their relevance,
sensitivity, stability, redundancy and completeness when applied to the
question of landscape diversity. From the analyses, it can be concluded
that the problem is complex and that simple solutions can be misleading.

Feasibility:

No major problem was encountered
by the different teams in the study. Neither availability and compatibility
of data nor computation time were a limiting factor, even when dealing
with complete European coverage.

Relevance:

l
The working scale imposed by
CORINE Land Cover data has important consequences. The minimum polygon
size of  25ha is rather large when compared with the average European
agricultural field size. This results in an overall underestimation of
the diversity of those landscapes that are shaped by agriculture. l

l
Artefacts have to be identified
and carefully eliminated. In function of the local photo interpretation
rules and of the type of images available (Spot, Landsat) a more or less
detailed map has been produced. This could lead to an underestimation of
fragmentation for some regions.l

l
The analysis of changes requires
particular attention. Any variation in methodology (image resolution, photo
interpretation rules, nomenclature, dispersion of acquisition dates) will
affect the results. For instance, an error rate of image classification
/ photo interpretation of 10% at both dates could imply up to 20% non-existent
changes, leading to unreliable indices.l

Heterogeneous agricultural
areas are included in the CORINE Land Cover database to assess complexity
and diversity when mapping the European landscape from satellite data at
a scale of 1:100.000, with a minimum polygon size of 25 ha. These areas
should be handled accordingly when assessing landscape diversity in Europe.
However, the results presented did not take this particularity of the CORINE
Land Cover nomenclature into account, and a general underestimation of
landscape diversity resulted. A better understanding of the composition
of heterogeneous agricultural areas is needed for assessing rural landscape
diversity.

Stability:

The size of the observation
unit used for calculations is not neutral. The same indices computed on
administrative regions or square supports (3x3, 10x10, 25x25, 60x60 km²)
behave differently. It appears that the more detailed analysis should be
favoured.

The nomenclature has an important
effect on the indices. Confusion can result from merging at European level
the classical CORINE Land Cover nomenclature with similar but not equivalent
classification systems from Finland or Sweden. Also, the use of 18 or 23
classes leads to drastically different results in regions where the 5 additional
classes are well represented.

Redundancy:

The redundancy between indices
has been evaluated through their correlation coefficients. Graphical analysis
has confirmed the long known fact that the correlations observed depend
on the size of the unit of observation. Averaging on larger areas reduces
variability, leading to an apparent increase in correlation. The conclusion
should not be drawn from this that the indices are redundant. On the other
hand, the ad hoc scale of observation should be defined before any correlation
is evaluated.

Completeness

Aggregating the thematic
and topological information contained in the CORINE Land Cover database
by setting up a limited number of indices cannot pretend to be a comprehensive
approach for landscape diversity assessment. Nevertheless, the maps derived
show that the underlying information can be usefully represented by the
indices chosen. Further analysis of the conceptual problem of landscape
description is now needed.

Despite the various reservations
made above, three main conclusions can be drawn:

 

l
The automatic computation of
diversity indices using CORINE Land Cover data is feasible using reasonable
resources, and the derived European maps show relevant complementary information
on landscape.l

l
This study should help to orientate
the future work of updating CORINE Land Cover within the CLC2000 project.
The limitations and confusions encountered relate mainly to the handling
of heterogeneous agricultural areas and consequently the working scale.
Consideration should be given to this, so that the CORINE Land Cover data
can in future be used reliably as a homogeneous data set covering the whole
of Europe. The integration of CORINE Land Cover data with other geo-referenced
datasets such as elevation, linear and point features would improve the
assessment of landscape diversity.l

l
It appears that, besides CORINE
Land Cover, more detailed information will also be needed to answer the
wide range  of questions related to the description and ongoing analysis
of the European landscape. The availability of alternative, complementary
material, such as aerial orthophotographs, should be considered. l

[]
- [] - []

[
] - [
]