**1. Introduction**

18 Biodiversity Loss in a Changing Planet

Running, S.W. (2006). Is Global Warming Causing More, Larger Wildfires? *Science* 313,

Suarez, A. V.; Bolger, D. T. & Case, T. J. (1998). Effects of fragmentation and invasion on native ant communities in coastal southern California. *Ecology* 79: pp. 2041-2056. Sheffield, J.; Goteti, G.; Wen, F. & Wood, E.F. (2004). A simulated soil moisture based

Swetnam, T.W. & Betancourt, J.L. (1990). Fire-southern oscillation relations in the

Swetnam, T.W. & Betancourt, J.L. (1998). Mesoscale disturbance and ecological response to

Veblen, T.T.; Kitzberger, T. & Donnegan, J. (2000). Climatic and human influences on fire

Veblen, T.T.; Kitzberger, T.; Villalba, R. & Donnegan, J. (1999). Fire history in northern

Westerling, A.L.; Brown, T.J.; Gershunov, A.; Cayan, D.R. & Dettinger, M.D. (2003). Climate

Westerling, A.L. (2006a). *Climate and Forest Wildfire in the Western United States*. In Economics

Westerling, A.L.; Hidalgo, H.G.; Cayan, D.R. & Swetnam, T.W. (2006b.) Warming and earlier spring increases Western U.S. forest wildfire activity. *Science* 313:pp 940-943. White, M. D. & Greer, K. A. (2002). The effects of watershed urbanization on stream

California. Conservation Biology Institute. (16.12.2010) Available from

drought analysis for the United States. *Journal of Geophysical Research*

decadal climatic variability in the American Southwest. *Journal of Climate* 11:pp

regimes in ponderosa pine forests in the Colorado Front Range. *Ecological* 

Patagonia: The roles of humans and climatic variation. *Ecological Monographs* 69:pp

and wildfire in the Western United States. *Bulletin of the American Meteorological* 

hydrological characteristics and riparian vegetation of Los Peñasquitos Creek,

pp. 927.

3128-3147.

47-67.

109:D24108.doi:10.1029/2004JD005.82.

*Applications* 10:pp 1178-1195.

*Society* 84(5):pp 595-604.

Southwestern United States. *Science* 249:pp 1017-1020.

of Forest Disturbance, T. Holmes, Ed., Springer.

http://consbio.org/cbi/projects/ pp. 557-558.

There is a global consensus on the idea of the present loss of biodiversity is intimately linked with human development, and that the conservation and sustainable use of present biological diversity is paramount to current and future generations of all life on Earth (Duro et al. 2007).

The United Nation Convention on Biological Diversity (CBD, http://www. biodiv.org, last accessed May 2011) lays down that countries are responsible for conserving their biological diversity and for using their biological resources in a sustainable manner. It expands until 2020 with the global *Strategy Plan for Biodiversity 2011-2020 and the Aichi biodiversity targets* (http://www.cbd.int/2011-2020, last accessed May 2011) to promote effective implementation of the CDB and to stem biodiversity loss by 2020. It compels the contracting countries to develop scientific and technical capacities to provide the appropriate measures in order to prevent and halt the pace of biodiversity loss all around the world.

It was during 90s and 2000s when scientific community became conscious that habitat destruction is the most prominent driver of biodiversity loss (Dirzo and Raven 2003)and together with degradation and fragmentation represent the most important factors leading to worldwide species decline and extinction (Chhabra et al. 2006; Soule and Terborgh 1999).

To improve the current conservation efforts and draw new strategies around the commitments under the CBD, it is crucial that our progress is monitored (Pereira and Cooper 2006). Biodiversity monitoring should be focused on trends in the abundance and distribution of populations and habitat extent (Balmford et al. 2005) and be carried out at different scales, regional and global and even local (Pereira and Cooper 2006).

There are several biophysical features influence species distributions, population sizes and ranges like land cover, primary productivity, temporal vegetation dynamics, disturbance events or climate (Hansen et al. 2004). All of them could be used as biophysical predictors of biodiversity at different scales. Remote sensing has been shown to be effective in some extent to measure and mapping those indicators and it has become a powerful tool for ecological studies because it allows monitoring over significant areas (Kerr and Ostrovsky 2003).

Assessing Loss of Biodiversity in Europe

simultaneous monitoring of several species.

2004; McDermid et al. 2009; Thogmartin et al. 2004).

their potential habitats.

that the target species potentially occupy (Kerr and Ostrovsky 2003).

Through Remote Sensing: The Necessity of New Methodologies 21

Blondel's definition of habitat was adopted in different habitat classifications at European level like CORINE Biotopes (Devillers et al. 1991), Classification of Paleartic Habitats (Devillers et al. 1992), the database PHYSIS of the Institut Royal des Sciences Naturelles de Belgique, or in the EUNIS program - *Habitat* of the European Environment Agency (EEA). We can understand habitat monitoring as the repeated recording of the condition of habitats, habitat types or ecosystems of interest to identify or measure deviations from a fixed standard, target state or previous status (Hellawell 1991; Lengyel et al. 2008). Habitat monitoring has two attributes (Lengyel et al. 2008; Turner et al. 2003): i) it can cover large geographical areas, then it can be used to evaluate drivers of biodiversity change over different spatial and temporal scales; being specially interesting at regional scales; ii) it provides information on the status of characteristic species because many species are restricted to discrete habitats; then if the link between some key species and discrete habitat types has been previously established, habitat monitoring can be used as a proxy for

Nature resources management and biological conservation assessments require spatially explicit environmental data that come from remote sensing or derived thematic layers. Most of these studies assume that the selected geospatial data are an effective representation of the ecological target (such as habitat) and provide an appropriate source of information to the objectives (McDermid et al. 2009). For example, biodiversity has been frequently studied indirectly through associations with land cover, which represents that mapping land cover has been often used as a surrogate for habitats (Foody 2008). The suitability of these assumptions is a current scientific concern and a dynamic research issue (Glenn and Ripple

Some studies (McDermid et al. 2009) have evaluated the suitability of general-purpose land cover classifications and compared to other data sources like vegetation inventory or specificpurpose maps: they show the constraints of general-purpose remote sensing land cover maps for explain wildlife habitat patterns and recommend the use of specific-purpose databases based on remote sensing along with field measurements. Then, traditional or general-purpose land cover maps may not be appropriate proxies of habitats, as we will show after assessing the suitability of the CORINE Land Cover product (European Environment Agency, http://www.eea.europa.eu/publications/COR0-landcover, last accessed May 2011) in Europe on this target (Section 4). Multi-purpose land cover maps meeting the needs of a large number of users but they are not specifically designed to represent the habitat of one/some key specie/-s. Furthermore, land-cover classifications used for wildlife habitat mapping and modeling must be appropriate spatial and thematic resolution to identify reliably the habitats

Thus, the identification of habitats through remote sensing must be suited the characteristics of each habitat type, rather than follow general or standard images processing approaches. For example, binary and one-class classifiers have been used in the implementation of the European Union's Habitat Directive (Boyd et al. 2006; Foody 2008). Moreover, it should be based on *in situ* and ancillary measurements (Kerr and Ostrovsky 2003) and also on ecological expert knowledge that allow finding the relationship between key species and

There are some ongoing challenges with this issue "*habitat monitoring*": the identification of the habitats as ecological units and not simply as land covers and the assessment and quantification of habitats degradation and fragmentation. Currently, one of the main scientific challenges and one of the big issues are if we are able to identify proper and accurately habitats from remote sensing at landscape level: the mapping and monitoring of

Remote sensing technologies contribute to biodiversity monitoring both direct and indirectly and they have been intensively improved in the last two decades, especially since the beginning of 2000s when the very high spatial resolution sensors were launched. Medium spatial resolution images from sensors on satellites make especially available information related to biophysical factors. In this sense, Landsat TM and ETM+ sensors are widely used in ecological investigations and applications because they have several advantages)(Cohen and Goward 2004): 30m of spatial resolution that facilitates characterization of land cover and land cover change; measurements acquired in all major portions of the solar electromagnetic spectrum (visible, near-infrared, shortwave-infrared); more than 30 years of Earth imaging and a temporal resolution of 16 days makes possible a complete analysis of the dynamic of the ecosystem; moderate cost (actually, all Landsat data from the USGS\_ U.S. Geological Survey\_ archive are free since the end of 2008). Anyway, to some extent, indirect measures that rely on remote sensing of biophysical parameters are, especially for national-level analyses, not enough accurate when the aim is the analysis of some aspects of biodiversity.

In a direct way, hyperspatial and hyperspectral sensors potentially supply land elements like individual organisms, species assemblages or ecological communities. Finally, LIDAR and RADAR technologies make possible to map vegetation structure (Lefsky et al. 2002; Zhao et al. 2011). Direct measures of biodiversity are becoming feasible with this kind of sensors although processes are still expensive and time-consuming, at least to regional levels.

Then, through RS it is possible to estimate in some extent habitat loss and fragmentation and trends in natural populations. At global and regional level the keystone is how translating remote-sensing data products into real and accurate knowledge of habitats and species distributions and richness. Subsequently, at present it is recognized that remote sensing technologies are especially crucial for conservation-related science (Kerr and Ostrovsky 2003) but they are still challenging. In addition, what is finally missing are global and regional standards for developing methodologies so systematic monitoring can be carried out (Strand et al. 2007).

#### **2. Land-cover versus habitat data**

*Land cover* is the observed physical cover of the Earth's surface (bare rock, broadleaved forests, etc…) (Eurostat 2001). Land cover data are usually derived by using multispectral remotely sensed data and statistical clustering methods. Remotely-sensed land cover data have been used at different scales (local, regional and global) as: i) input variables in biosphere –atmosphere models simulating exchanges of energy and water between land surface and the atmosphere and in terrestrial ecosystem models simulating carbon dynamics at global scales; ii) input variables in terrestrial vegetation change assessments; iii) proxies of biodiversity distribution (DeFries 2008; Hansen et al. 2004; Thogmartin et al. 2004).

On the other hand, *habitat* is a three-dimensional spatial entity that comprises at least one interface between air, water and ground spaces, it includes both the physical environment and the communities of plants and animals that occupy it, it is a fractal entity in that its definition depends on the scale at which it is considered" (Blondel 1979). *Natural habitats*  means terrestrial or aquatic areas distinguished by geographic, abiotic and biotic features, whether entirely natural or semi-natural (EU Habitats Directive, 92/43/EC). The identification of one habitat implies a holistic perspective and involves not only the expression of the vegetation (land cover) but also the species and other biophysical parameters like topography, aspect, soil characteristics, climate or water quality.

Remote sensing technologies contribute to biodiversity monitoring both direct and indirectly and they have been intensively improved in the last two decades, especially since the beginning of 2000s when the very high spatial resolution sensors were launched. Medium spatial resolution images from sensors on satellites make especially available information related to biophysical factors. In this sense, Landsat TM and ETM+ sensors are widely used in ecological investigations and applications because they have several advantages)(Cohen and Goward 2004): 30m of spatial resolution that facilitates characterization of land cover and land cover change; measurements acquired in all major portions of the solar electromagnetic spectrum (visible, near-infrared, shortwave-infrared); more than 30 years of Earth imaging and a temporal resolution of 16 days makes possible a complete analysis of the dynamic of the ecosystem; moderate cost (actually, all Landsat data from the USGS\_ U.S. Geological Survey\_ archive are free since the end of 2008). Anyway, to some extent, indirect measures that rely on remote sensing of biophysical parameters are, especially for national-level analyses, not

In a direct way, hyperspatial and hyperspectral sensors potentially supply land elements like individual organisms, species assemblages or ecological communities. Finally, LIDAR and RADAR technologies make possible to map vegetation structure (Lefsky et al. 2002; Zhao et al. 2011). Direct measures of biodiversity are becoming feasible with this kind of sensors although

Then, through RS it is possible to estimate in some extent habitat loss and fragmentation and trends in natural populations. At global and regional level the keystone is how translating remote-sensing data products into real and accurate knowledge of habitats and species distributions and richness. Subsequently, at present it is recognized that remote sensing technologies are especially crucial for conservation-related science (Kerr and Ostrovsky 2003) but they are still challenging. In addition, what is finally missing are global and regional standards for developing methodologies so systematic monitoring can be carried

*Land cover* is the observed physical cover of the Earth's surface (bare rock, broadleaved forests, etc…) (Eurostat 2001). Land cover data are usually derived by using multispectral remotely sensed data and statistical clustering methods. Remotely-sensed land cover data have been used at different scales (local, regional and global) as: i) input variables in biosphere –atmosphere models simulating exchanges of energy and water between land surface and the atmosphere and in terrestrial ecosystem models simulating carbon dynamics at global scales; ii) input variables in terrestrial vegetation change assessments; iii) proxies of

On the other hand, *habitat* is a three-dimensional spatial entity that comprises at least one interface between air, water and ground spaces, it includes both the physical environment and the communities of plants and animals that occupy it, it is a fractal entity in that its definition depends on the scale at which it is considered" (Blondel 1979). *Natural habitats*  means terrestrial or aquatic areas distinguished by geographic, abiotic and biotic features, whether entirely natural or semi-natural (EU Habitats Directive, 92/43/EC). The identification of one habitat implies a holistic perspective and involves not only the expression of the vegetation (land cover) but also the species and other biophysical

biodiversity distribution (DeFries 2008; Hansen et al. 2004; Thogmartin et al. 2004).

parameters like topography, aspect, soil characteristics, climate or water quality.

enough accurate when the aim is the analysis of some aspects of biodiversity.

processes are still expensive and time-consuming, at least to regional levels.

out (Strand et al. 2007).

**2. Land-cover versus habitat data** 

Blondel's definition of habitat was adopted in different habitat classifications at European level like CORINE Biotopes (Devillers et al. 1991), Classification of Paleartic Habitats (Devillers et al. 1992), the database PHYSIS of the Institut Royal des Sciences Naturelles de Belgique, or in the EUNIS program - *Habitat* of the European Environment Agency (EEA).

We can understand habitat monitoring as the repeated recording of the condition of habitats, habitat types or ecosystems of interest to identify or measure deviations from a fixed standard, target state or previous status (Hellawell 1991; Lengyel et al. 2008). Habitat monitoring has two attributes (Lengyel et al. 2008; Turner et al. 2003): i) it can cover large geographical areas, then it can be used to evaluate drivers of biodiversity change over different spatial and temporal scales; being specially interesting at regional scales; ii) it provides information on the status of characteristic species because many species are restricted to discrete habitats; then if the link between some key species and discrete habitat types has been previously established, habitat monitoring can be used as a proxy for simultaneous monitoring of several species.

Nature resources management and biological conservation assessments require spatially explicit environmental data that come from remote sensing or derived thematic layers. Most of these studies assume that the selected geospatial data are an effective representation of the ecological target (such as habitat) and provide an appropriate source of information to the objectives (McDermid et al. 2009). For example, biodiversity has been frequently studied indirectly through associations with land cover, which represents that mapping land cover has been often used as a surrogate for habitats (Foody 2008). The suitability of these assumptions is a current scientific concern and a dynamic research issue (Glenn and Ripple 2004; McDermid et al. 2009; Thogmartin et al. 2004).

Some studies (McDermid et al. 2009) have evaluated the suitability of general-purpose land cover classifications and compared to other data sources like vegetation inventory or specificpurpose maps: they show the constraints of general-purpose remote sensing land cover maps for explain wildlife habitat patterns and recommend the use of specific-purpose databases based on remote sensing along with field measurements. Then, traditional or general-purpose land cover maps may not be appropriate proxies of habitats, as we will show after assessing the suitability of the CORINE Land Cover product (European Environment Agency, http://www.eea.europa.eu/publications/COR0-landcover, last accessed May 2011) in Europe on this target (Section 4). Multi-purpose land cover maps meeting the needs of a large number of users but they are not specifically designed to represent the habitat of one/some key specie/-s. Furthermore, land-cover classifications used for wildlife habitat mapping and modeling must be appropriate spatial and thematic resolution to identify reliably the habitats that the target species potentially occupy (Kerr and Ostrovsky 2003).

Thus, the identification of habitats through remote sensing must be suited the characteristics of each habitat type, rather than follow general or standard images processing approaches. For example, binary and one-class classifiers have been used in the implementation of the European Union's Habitat Directive (Boyd et al. 2006; Foody 2008). Moreover, it should be based on *in situ* and ancillary measurements (Kerr and Ostrovsky 2003) and also on ecological expert knowledge that allow finding the relationship between key species and their potential habitats.

There are some ongoing challenges with this issue "*habitat monitoring*": the identification of the habitats as ecological units and not simply as land covers and the assessment and quantification of habitats degradation and fragmentation. Currently, one of the main scientific challenges and one of the big issues are if we are able to identify proper and accurately habitats from remote sensing at landscape level: the mapping and monitoring of

Assessing Loss of Biodiversity in Europe

the future perspectives and an overall assessment.

Union (all except Bulgaria and Romania).

EIONET 2011)

Through Remote Sensing: The Necessity of New Methodologies 23

Through its 17th Article, Habitat Directive forces countries to monitor habitat changes every six years and to assess and report to the European Union on the conservation status of the habitats and wild flora and fauna species of Community interest: the mapping of the distribution area, the trends, the preservation of their structure and functions together with

Then, to meet the requirements of global and regional biodiversity targets such as the *Strategy Plan for Biodiversity 2011-2020 and the Aichi biodiversity targets, the 2020 EU Biodiversity Strategy* or the European *Natura 2000* Network, the development of more cost

At the moment, the first habitats reports were submitted in electronic format to the European Environmental Agency (www.eunis.eea.europa.eu, last accessed May 2011) (EEA) until March 2008, through an electronic platform on the Internet. This platform is managed by the EEA and the European Environment Information and Observation Network (EIONET) (http://bd.eionet.europa.eu/, last accessed May 2011). Currently, this information was supplied by 25 of the 27 countries that currently comprise the European

We have developed a map (Figure 1) about the distribution of habitats of Community interest derived from this information. The data were compiled, refined and standardized in

Fig. 1. Distribution of habitats of Community interest in Europe (Source: Developed from

and time effective monitoring strategies are mandatory (Bock et al. 2005).

the territory in terms of its habitats. We have to say not yet, at least not only with remote sensing technologies and with an adequate budget and an optimal time. We also need ancillary information, ecological expert knowledge, field work and other auxiliary tools like landscape ecology indices.
