**4. The CORINE land cover map as a proxy of biodiversity: difficulties and constraints**

The *CORINE land cover* project (EEA, 1999, http://www.eea.europa.eu/publications/COR0 landcover, last accessed May 2011) constitutes the first harmonized European land cover classification system, based on photo-interpretation of Landsat images. The minimum unit for inventory is 25 ha and the minimum width of units is 100m. Only area elements (polygons) are identified. Areas smaller than 25 ha are allowed in the national land cover database as additional thematic layers, but should be generalized in the European database. The CORINE land cover (CLC) nomenclature is hierarchical and distinguishes 44 classes at the third level, 15 classes at the second level and 5 classes at the first level. Third level is mandatory although additional national levels can be mapped but should be aggregated to level 3 for the European data integration. Any unclassified areas appear in the final version of the dataset (See CLC Legend in Appendix 1, Table 12).

Because of general land cover maps may not be suitable proxies of habitat maps, we have analyzed the spatial inconsistencies between a remote-sensed land cover map (CORINE Land Cover 2000) and some selected habitats of the map obtained from EIONET (Figure 1) which represents the spatial distribution of the natural and semi-natural habitats in the European Community. CORINE Land Cover is cross-comparable with habitats of Community interest (Council Directive 92/43/EEC) through the EUNIS system (www.eunis.eea.europa.eu, last accessed May 2011) (Tables 1 and 2). The comparative analysis was done using a 10 km2 UTM grid which is the base of the EIONET map. Spatial gaps and contradictions that arise between both sources, when land cover maps are used to assess biodiversity status, were evaluated through the analysis of coincidences at the cells of the grid between both databases.

The CORINE Land Cover (CLC) map was analyzed and compared at the third level in the European context and at the fifth level at the scale of Spain (Table 3 and 4). Following

a database using the *ETRS 1989 Lambert azimuthal equal-area* projection system (following INSPIRE Directive). All the spatial data of the habitats of Community interest, derived from each country, were harmonized and represented in a 10 km2 UTM grid (Universal Transverse Mercator Projection) following the recommendations of the European Commission (EuropeanCommission 2006). The output map follows the EUNIS (European Nature Information System) classification system and represents finally all the European

The EUNIS system constitutes a pan-European classification proposed by the EEA (www.eunis.eea.europa.eu, last accessed May 2011). It is developed and managed by the European Topic Centre for Nature Protection and Biodiversity (ETC/NPB in Paris), and covers the whole of the European land and sea area, i.e. the European mainland as far east as the Ural Mountains, including offshore islands (Cyprus; Iceland but not Greenland), and the archipelagos of the European Union Member States (Canary Islands, Madeira and the Azores), Anatolian Turkey and the Caucasus (Davies et al. 2004). It represents a common classification scheme for the whole of European Union, as it is compatible with the units of protection established in the strategy of *Natura 2000*-protected areas. It covers all types of habitats from natural to artificial, from terrestrial to freshwater and marine. EUNIS is also cross-comparable with CORINE Land Cover (Bock et al. 2005; Moss and Davies 2002)

**4. The CORINE land cover map as a proxy of biodiversity: difficulties and** 

of the dataset (See CLC Legend in Appendix 1, Table 12).

the grid between both databases.

The *CORINE land cover* project (EEA, 1999, http://www.eea.europa.eu/publications/COR0 landcover, last accessed May 2011) constitutes the first harmonized European land cover classification system, based on photo-interpretation of Landsat images. The minimum unit for inventory is 25 ha and the minimum width of units is 100m. Only area elements (polygons) are identified. Areas smaller than 25 ha are allowed in the national land cover database as additional thematic layers, but should be generalized in the European database. The CORINE land cover (CLC) nomenclature is hierarchical and distinguishes 44 classes at the third level, 15 classes at the second level and 5 classes at the first level. Third level is mandatory although additional national levels can be mapped but should be aggregated to level 3 for the European data integration. Any unclassified areas appear in the final version

Because of general land cover maps may not be suitable proxies of habitat maps, we have analyzed the spatial inconsistencies between a remote-sensed land cover map (CORINE Land Cover 2000) and some selected habitats of the map obtained from EIONET (Figure 1) which represents the spatial distribution of the natural and semi-natural habitats in the European Community. CORINE Land Cover is cross-comparable with habitats of Community interest (Council Directive 92/43/EEC) through the EUNIS system (www.eunis.eea.europa.eu, last accessed May 2011) (Tables 1 and 2). The comparative analysis was done using a 10 km2 UTM grid which is the base of the EIONET map. Spatial gaps and contradictions that arise between both sources, when land cover maps are used to assess biodiversity status, were evaluated through the analysis of coincidences at the cells of

The CORINE Land Cover (CLC) map was analyzed and compared at the third level in the European context and at the fifth level at the scale of Spain (Table 3 and 4). Following

habitats of Community interest.

(Appendix 1, Table11).

**constraints** 

Table 1. Correspondences between Corine Land Cover classification (3rd level) and habitats of Community interest

Table 2. Correspondences between Corine Land Cover classification (5th level) and habitats of Community interest

Assessing Loss of Biodiversity in Europe

in Sweden and Ireland.

Spain, Slovenia and Slovakia.

**gap**.

Through Remote Sensing: The Necessity of New Methodologies 27

similarities with a gap analysis (Jennings 2000; Scott et al. 1993), spatially explicit correlations was carried out and five thresholds representing the grade of inconsistency between both maps were set: i) less than 10% of coincidences represent a **total gap**; ii) coincidences between 10-30%, **very high gap**; iii) coincidences between 30-50%, **high gap**; iv) coincidences between 50-90%, **moderate gap**; v) coincidences upper 90% represent **no** 



At European level (Table 3), and by countries, some relevant habitats showed:

MODERATE GAP in Spain and Portugal; right correspondence in UK.



At European level, the type of habitats (among the evaluated set) with a worse representation on CLC map are coastal lagoons (1150\*), mires and bogs (7110\*, 7120, 7230), water courses (3260 and 3270), heaths (4020\* and 4030), *Molinia* and lowland hay meadows (6410 and 6510) and siliceous rocky slopes (8220). The different types of broadleaved forests show an acceptable representation, although a very important question is that CLC does not

Lithuania, Italy, Greece, France, Estonia, Spain, Slovenia, Slovakia and Belgium. - *Habitat 91E0\** (correspondence with CLC 311 class BROAD-LEAVED FORESTS): at European level it shows a moderate gap. By countries: HIGH GAP in Austria; right correspondence in Poland, Luxembourg, Hungary, Lithuania, Greece, France, Estonia

Netherlands, Lithuania y Latvia present a right correspondence.

Table 2. Correspondences between Corine Land Cover classification (5th level) and habitats

of Community interest

similarities with a gap analysis (Jennings 2000; Scott et al. 1993), spatially explicit correlations was carried out and five thresholds representing the grade of inconsistency between both maps were set: i) less than 10% of coincidences represent a **total gap**; ii) coincidences between 10-30%, **very high gap**; iii) coincidences between 30-50%, **high gap**; iv) coincidences between 50-90%, **moderate gap**; v) coincidences upper 90% represent **no gap**.

At European level (Table 3), and by countries, some relevant habitats showed:


At European level, the type of habitats (among the evaluated set) with a worse representation on CLC map are coastal lagoons (1150\*), mires and bogs (7110\*, 7120, 7230), water courses (3260 and 3270), heaths (4020\* and 4030), *Molinia* and lowland hay meadows (6410 and 6510) and siliceous rocky slopes (8220). The different types of broadleaved forests show an acceptable representation, although a very important question is that CLC does not


Table 3. Spatial correlations between Corine Land Cover cartography (3rd level) and CD 92/43/EEC habitats cartography in the EU Countries (units in percentage)

Assessing Loss of Biodiversity in Europe

*and sand plains*, or alluvial forests.

inconsistencies increase.

forests (91E0\*).

of *Ilex aquifolium* (9380).

Through Remote Sensing: The Necessity of New Methodologies 29

identify differences between forest compositions, i.e. the correspondence at the third level is with the broad CLC 311 class (broad-leaved forests) (Table1). Similar situation occurs with many other habitats like dunes whose correspondence is with CLC 331 class *Beaches, dunes* 

This constrains many possibilities in the use of CLC at the third level to monitor biodiversity. For instance, the CLC 311 class (Broad-leaved forests) could correspond to *Eucalyptus globulus* or any native broad-leaved forests as *Quercus robur*, both phenomena

Also, the finer the nomenclature detail, the worse the spatial correlations are. As the scale is finer and CLC is considered at 5th level (for example at Spain level) results get worse and




**HAB CLC** 

**7110\*** 0,74 **7120** 0,00 **7130** 6,58 **7230** 4,46 **8130** 16,30 **8220** 17,31 **9120** 98,44 **9180\*** 91,84 **91E0\*** 8,79 **9230** 77,98 **9330** 72,79 **9340** 58,21 **9380** 19,63


**Total gap**: less than 10% of coincidences **Very high gap**: coincidences between 10-30 % **High gap**: coincidences between 30-50% **Moderate gap**: coincidences between 50-90%

Table 4. Spatial correlations between Corine Land Cover cartography (5th level) and CD

**No gap**: coincidences upper 90%

92/43/EEC habitats cartography in Spain (units in percentage)

with very different implications for biodiversity (Pereira and Cooper 2006).


At Spain level (Table 4) and using a 10km grid results show:

woolands of *Quercus spp*. (9230, 9330, 9340).

**HAB CLC** 

**1130** 37,93 **1140** 25,64 **1150\*** 18,39 **1310** 31,82 **2120** 72,50 **2130\*** 71,88 **3150** 14,74 **3260** 8,75 **3270** 17,51 **4020\*** 60,90 **4030** 45,53 **5130** 100,00 **6170** 22,81 **6410** 70,00 **6510** 71,93

Table 3. Spatial correlations between Corine Land Cover cartography (3rd level) and CD

92/43/EEC habitats cartography in the EU Countries (units in percentage)

identify differences between forest compositions, i.e. the correspondence at the third level is with the broad CLC 311 class (broad-leaved forests) (Table1). Similar situation occurs with many other habitats like dunes whose correspondence is with CLC 331 class *Beaches, dunes and sand plains*, or alluvial forests.

This constrains many possibilities in the use of CLC at the third level to monitor biodiversity. For instance, the CLC 311 class (Broad-leaved forests) could correspond to *Eucalyptus globulus* or any native broad-leaved forests as *Quercus robur*, both phenomena with very different implications for biodiversity (Pereira and Cooper 2006).

Also, the finer the nomenclature detail, the worse the spatial correlations are. As the scale is finer and CLC is considered at 5th level (for example at Spain level) results get worse and inconsistencies increase.

At Spain level (Table 4) and using a 10km grid results show:



Table 4. Spatial correlations between Corine Land Cover cartography (5th level) and CD 92/43/EEC habitats cartography in Spain (units in percentage)

Assessing Loss of Biodiversity in Europe

extension of calcareous beech forest.

scale maps with fine thematic resolution.

detected at finer segmentation levels.

grasslands and semi-natural habitats.

**al. 2007; Franklin et al. 2002; Franklin et al. 2001)** 

Through Remote Sensing: The Necessity of New Methodologies 31

For example, for the Annex I habitat "Calcareous Beech Forest (code 9150)", first a rule was defined that selects the broadleaf forests from the CORINE land cover database, then a second rule was used to select the beech distribution map from the Atlas Florae Europaeae, and a third rule identified the calcareous soils from the European soil database. The combination of these three filters will form the decision rule that delimitates the spatial

The main advantage of this approach is the suggestion of using specific knowledge, implemented as a GIS decision rule, to identify individual habitat maps as they are described in Annex I of Habitats Directive. The approach use remote sensing data in an indirect way (through the use of CORINE land cover and other input variables) along with other suitable ancillary data. Results are appropriate at European scale in order to set

Bock et al. (2005) have proposed an object-oriented approach for EUNIS habitat mapping using remote sensing data at multiple scales with good results. The approach performs well when applied to high resolution satellite data (Landsat 30m) for the production of habitat maps at regional level with coarse thematic resolution; also it performs extremely well when applied to very high spatial resolution data (Quickbird 0,7m) for the production of local

The use of a multi-scale segmentation (implemented in the software package eCognition, Definiens, http://www.definiens.com) allow for the accurate classification of habitat types, which occur at different scales: for example, large-scale woodland habitat can be detected at coarser segmentation levels, while small-scale habitats such as woodland corridors can be

The main advantages of the object-oriented approaches to habitats mapping are: i) the ability to integrate ancillary data into the classification processes, related to shape, texture, context, etc.; ii) the option of developing knowledge-based rules in the classification process. Both questions make especially possible the accurate identification of habitats with similar spectral properties. Some results that show the advantage of these issues are (Bock et al. 2005): i) the effective separation of different grassland types like calcareous and mesotrophic grassland habitats to a high degree of accuracy through the use of geological data; ii) the use of multi-temporal remote sensing data to distinguish among arable lands, manage

**5.3 The use of binary classifications by decision trees (DT) (Boyd et al. 2006; Foody et** 

Some studies have shown binary classifications as one of the more appropriate methods to identify habitats in the territory. Binary classifications can be implemented by nonparametric Decision Trees (DT) algorithms. Some of these studies have focused on the mapping of one specific thematic class (Boyd et al. 2006; Foody et al. 2007) hypothesizing that non-parametric algorithm would be more suitable to habitats of conservation interest because of the scarce spatial distribution usually associated to them (the size of the training sample will be smaller). Other studies have combined that kind of techniques (binary classifications by DT) in hybrid approaches (Franklin et al. 2001). The hybrid approaches assumes that parametric algorithms like standard maximum likelihood (ML) are the best

guidelines for the strategic design of the Pan-European Ecological Network.

**5.2 Object oriented approaches (Bock et al. 2005; Díaz Varela et al. 2008)** 

It is relevant that at Spain level CLC map shows important inconsistencies with bogs and mires (7110\*, 7120, 7130, 7230), water courses (3260), alluvial forests (91E0\*) or coastal lagoons (1150\*).

Total, very high and high gaps should be considered as important inconsistencies which enhance the limited capacity of the CLC map for representing natural and semi-natural habitats and reveal the inappropriate use of the CLC map as a biodiversity proxy, both at European and regional level. In some cases gaps can be explained because of the CLC methodology which makes not possible to identify habitats with less than 25ha or linear features below 100m in width. Also, discrepancies among countries could be attributed to differences among the skills and expert knowledge of image interpreters.

Then, though theoretically possible (Groom et al. 2006; Hansen et al. 2004), the use of some components of the complex habitat entity, such land covers, as a surrogate parameter of a particular habitat is uncertain and it should be previously evaluated.
