A standardized procedure for surveillance and monitoring European habitats and provision of spatial data
- First Online:
- Cite this article as:
- Bunce, R.G.H., Metzger, M.J., Jongman, R.H.G. et al. Landscape Ecol (2008) 23: 11. doi:10.1007/s10980-007-9173-8
Both science and policy require a practical, transmissible, and reproducible procedure for surveillance and monitoring of European habitats, which can produce statistics integrated at the landscape level. Over the last 30 years, landscape ecology has developed rapidly, and many studies now require spatial data on habitats. Without rigorous rules, changes from baseline records cannot be separated reliably from background noise. A procedure is described that satisfies these requirements and can provide consistent data for Europe, to support a range of policy initiatives and scientific projects. The methodology is based on classical plant life forms, used in biogeography since the nineteenth century, and on their statistical correlation with the primary environmental gradient. Further categories can therefore be identified for other continents to assist large scale comparisons and modelling. The model has been validated statistically and the recording procedure tested in the field throughout Europe. A total of 130 General Habitat Categories (GHCs) is defined. These are enhanced by recording environmental, site and management qualifiers to enable flexible database interrogation. The same categories are applied to areal, linear and point features to assist recording and subsequent interpretation at the landscape level. The distribution and change of landscape ecological parameters, such as connectivity and fragmentation, can then be derived and their significance interpreted.
KeywordsField recordingStratified samplingBiodiversityMonitoringSurveillanceRaunkiaer plant life formsGeneral habitat categories
When recording habitats and biodiversity at the landscape level, the difficulty has always been in reconciling the observed complexity of points, lines and patches with recognisable categories that can be consistently and repeatedly recorded in the field and then converted into national and regional estimates. It is therefore necessary to link the detailed records to a strategic framework, as described by Sheail and Bunce (2003). Monitoring and surveillance also have to be integrated spatially and temporally with other data sources. The primary goal of this paper is to describe a system that can lead to the production of a statistical profile of interdependent systems that make up European landscapes, and subsequently to enable the assessment of changes resulting from landscape ecological processes, such as fragmentation. The approach will enable the landscape ecological resources of the continent to be determined and, because it is based on plant life forms which are applicable throughout the world, further categories could therefore be developed for other continents.
In the final plenary session at the 2007 IALE World Congress (International Association for Landscape Ecology), the assessment of change in landscape ecological elements at the strategic level was identified as an important topic for future research. Many regional studies and some national inventories are provided in Bunce et al. (2007), but none at a continental scale. Surprisingly, within the Congress, the Symposium on Monitoring did not identify any new methodologies, probably because of regular communication within the IALE community.
Policy makers and land managers increasingly demand hard figures that detail the state of biodiversity and habitats, as well as the definition of historical trends. Arguments over the responsibility of man in driving global environmental change make the demand for incontrovertible evidence ever greater (Reid 2005). Such statistics are not only important for local and national policies, but may also be used to evaluate international conventions and commitments (e.g., the Goteborg Commitment by the European Union (EU) to halt biodiversity loss by 2010). However, there is a lack of consistent data to meet these requirements, especially at the supra-national level. Currently, reporting is based on national programs, without accepted protocols. As a result there are no consistent figures on habitats for Europe, because the available maps are derived from satellite imagery and are not at a sufficiently detailed level.
Throughout the world there are also many products at a strategic scale derived from satellite imagery, but usually with no link to in situ data. Regional landscape ecological studies are more common; e.g., Jones et al. (2001) provide a broad view of the relevance of assessing landscape ecological changes and give an example at the regional scale in the United Sates. They point out that, whilst there had been successful development of methods for broad scale assessment, a critical limitation was that field based methods had proved to be inconsistent. However, new data on land cover change are now available; e.g., the North American Landscape Characterisation program (NALC) contains an archive of Landsat Multispectral Scanner (MSS) images. Vogelman et al. (2001) also describe a comparable program. Taken together these two programs permit relatively fine scale assessments of landscape change across large areas, but they are not integrated with habitat records. Also, in the United States the Environmental Protection Agency (EPA, http://www.epa.gov/emap) is developing tools for monitoring, but there is no national coverage of habitats. In Australia, in various papers, Austin has explored a range of different sampling techniques and scales; e.g., Austin and Myers (1995); but has never applied them in a strategic, integrated project; although some of the conclusions were incorporated in the development of the present procedure. Some Australian habitats have national coverage; e.g., coastlines in the Coastal Water Mapping project (CWHM); but otherwise only regional specialist studies have been carried out, e.g., New (2000). A commentary on the situation in Australia, as reported in (http://www.environment.gov.au/soe/2006/publications/commentararie), stated that currently there was imperfect knowledge of the state and trends in biodiversity at any scale. Relevant figures were therefore derived from fragmented sources and expert opinion, as has been carried out in similar assessments in Europe.
Fundamental landscape ecological concepts, such as connectivity, isolation and dispersal, also require basic data on the spatial arrangement of habitats in landscapes. Changes in patterns can then be determined and the processes of change defined and interpreted. For example, Petit et al. (2004) used spatial data from habitats recorded in the UK Countryside Survey (Haines-Young et al. 2000) to assess changes in landscape ecological parameters, such as the adjacency of woodland elements. The definition of the landscape ecological characteristics of a particular area, or sample unit, also needs information about the habitats present, e.g., in landscape fragments, as well as associated species. Such information can enable the landscape ecology of an entire region to be understood, e.g., the long term studies of Bocage landscapes of Brittany (France) by Baudry (e.g., Baudry et al. 2000). Specific landscape ecological elements such as linear and point features may also be described (Hermy and De Blust 1997). Alternatively, data may be recorded from a series of samples and then used to build up landscape ecological descriptions based on statistically derived landscape units (Bunce et al. 1993).Whatever the objectives of a specific study might be, standardized categories would enable the results to be transferable. International modelling exercises would similarly benefit from common categories and protocols.
Although field recording has been central to ecology and landscape ecology since their inception, relatively little attention has been paid to the development of consistent recording procedures for monitoring habitats within landscape elements. Furthermore, the majority of the extensive literature on vegetation (e.g., Braun-Blanquet 1932) is not designed for long-term monitoring, although the individual records can be repeated, if the sites are re-locatable (e.g., Grabherr et al. 1994). Kirby et al. (2003) showed that consistent recording is essential for long-term monitoring of woodland vegetation. The data on point features collected thirty years before (using the standardized procedure of Bunce and Shaw 1973) was sufficiently accurate to detect changes in habitats, such as forest glades. However, studies of vegetation change are rarely integrated at the landscape level, although Sheail and Bunce (2003) describe how the principles of standardized recording and statistical sampling of vegetation were extended to the landscape level.
Landscape ecologists have been successful in the application of their results to spatial planning but have had limited impact in the development of strategic conservation policies, as described by Bunce and Jongman (2007). Many conservation agencies neither appreciate the need to sample landscape complexity nor consider it necessary to analyze the interrelationships between component elements. Conservation managers are also not familiar with standardized methods of recording and sampling, or the statistical procedures, and are inevitably usually concerned only with local issues. The present methodology was designed to provide categories that are at a level of detail for consistent recording of habitats, which can be linked to other measures of biodiversity. However, it is recognized that a major program of work would be needed to carry out integration with existing data. Common standards could also provide the basis for stimulating scientific enquiry into the characteristics and relationships between landscape ecological units in entire landscapes.
Whilst the development of the ecosystem concept was originally mainly based on vegetation, it is now widely recognized that habitats should be defined independently. This is partly because, in terms of significance for animal populations, vegetation structure is often more important than vegetation classes (cf Fox et al. 2003), but also because some widely recognized habitats are not directly linked to traditional vegetation associations (Rodwell et al. 2002). In the 1980s, habitat mapping progressively became a separate exercise from vegetation recording, because strategic conservation surveys could be carried out more rapidly and cost-efficiently without the involvement of vegetation experts. For example, Agger and Brandt (1988) monitored changes in small landscape patches (biotopes) on intensively farmed land, without using plant communities. In an examination of the development of the Countryside Survey in Great Britain (GB) Firbank et al. (2003) indicate that, although the project in 1978 initially concentrated on vegetation, by 2000 the reporting of status and change was integrated with habitats in landscape units, because these are more convenient for reporting and more readily understandable by policy makers. Nevertheless, whilst detailed vegetation records are not required for monitoring habitat extent, such data are essential in determining habitat quality and condition; i.e., conservation status (Haines-Young et al. 2000). Over the same period landscape ecologists were developing techniques for analysing changes in patterns, often utilizing detailed habitat maps but using different systems of classification and scales, according to individual objectives and landscape characteristics. For example, Bunce et al. (1993) analyzed the relationships between the composition of linear features and the surrounding land in GB and showed that in lowland landscapes the majority of biodiversity was restricted to such elements, whereas in the uplands it was dispersed more widely.
The initial objective of the BioHab project was to develop a framework for surveillance and monitoring of European habitats, using existing classifications. However, it did not prove possible to develop adequate field rules for these classifications that were sufficiently consistent for recording change. Accordingly, the project team combined basic scientific knowledge from the literature, practical knowledge from previous field experience, and trial surveys across Europe to develop General Habitat Categories (GHCs) based on plant life forms.
The present paper firstly summarizes the conceptual principles behind European habitat monitoring and the creation of consistent habitat categories. Secondly, the recording procedure is described, explaining the rules needed for field survey. Finally, field testing, and policy relevance are discussed.
Surveillance and monitoring
It is first useful to summarize several conceptual principles relevant for the present study, starting with the definitions adopted of two frequently used terms, surveillance and monitoring, because they are often used elsewhere in different ways. Surveillance is the act of surveying, i.e. the recording of features at a specific location in one time frame, i.e., taking stock. In contrast, monitoring involves repeated observation on a time-line such that change can be detected, i.e., assessing both stock and change.
For small areas (e.g., some nature reserves) it may be possible to survey the entire site, but in most cases the assessment of biodiversity or habitats must be based on samples. One of the main factors in deciding the characteristic of samples is that habitats often occur in patches of different sizes in contrasting landscapes. Sampling procedures must not be compromised by spatial heterogeneity or complexity. As sampling effort (i.e., the time taken to record information) is usually fixed, a choice has to be made between recording many small sample units or a smaller number of larger units. As discussed by Bunce et al. (1996) it costs more per unit area to sample many small units, although they may give statistically more precise estimates (Gallego 2002). On the other hand, Brandt et al. (2002) argue that larger sample units provide a more systematic inclusion of variations due to management. As there is no optimal sample unit size for all the habitats and landscapes at a continental scale; due to variation at landscape, patch and management scales; a 1 km square is a workable compromise, matching ease of survey, data content, and obtaining an adequate number of sample units for estimates of statistical probability . For some complex landscapes; e.g., Northern Ireland; sampling units of 0.25 km square may be more appropriate (Cooper and McCann 2002) and for aerial photographs larger units may be needed (Olschofsky et al. 2006). Using a standard size enables the direct comparisons to be made of relative heterogeneity. The 1 km square unit also enables internal spatial modelling of habitat patches and is suitable for scenario testing (Bunce et al. 1993).
The methodology is based on the principle that statistical inference requires samples (e.g., 1 km squares) to be drawn randomly from a defined population (e.g., Europe). Samples can be drawn from strata derived from the partitioning of the land surface by statistical analysis of climatic and topographic data from 1 km squares. The samples can then be analyzed to generate statistical estimates of the extent of required parameters for the region concerned. Bunce et al. (1996) described 32 classes for GB and Metzger et al. (2005) 84 strata for Europe. The former have been used for estimating habitat areas in the Countryside Survey of GB and the latter are appropriate for Europe (Jongman et al. 2006).
The majority of field habitat mapping projects involve surveillance and are not intended to monitor change. Monitoring requires more stringent procedures to ensure that differences recorded represent real change and not distortions due to differences between observers or recording technique, as described by Brandt et al. (2002). Further discussion of the details of the design of the monitoring procedure is given by Bunce et al. (2005)
Across Europe, there is much experience in applying such methodology in the detection of change; e.g., GB (Haines-Young et al. 2000), Northern Ireland (Cooper and McCann 2002), Denmark (Agger and Brandt 1988), and in interpreting changes from aerial photographs, e.g., Sweden (Skånes 1996). Strict rules have been developed for updating the initial information, including procedures for correcting errors in the baseline data. Investigators are therefore able to use the results to detect and evaluate alterations in a landscape context, e.g., changes in patterns of linear features or whether new forestry is planted on semi-natural vegetation or on arable land (Petit et al. 2001).
Consistent habitat definition
Monitoring European habitats requires definitions that can be applied consistently in the field across Europe (Brandt et al. 2002). Habitats are defined as: “An element of the land surface that can be consistently defined spatially in the field in order to define the principal environments in which organisms live” (Bunce et al. 2005). This definition includes water bodies and extends to the Mean High Water (MHW) at the coast. It therefore excludes marine systems. Existing European habitat classifications have been based on species, geographical location, vegetation classes and environmental factors (e.g., the EUNIS system, Davies and Moss 2002). Whilst these classifications have been successfully applied to produce general descriptions of the occurrence of classes in protected areas, they are not appropriate for monitoring, because definitions of many of the terms used; e.g., montane and sub-Mediterranean; are not provided.
The present recording procedure therefore adopted plant life forms, as described by Raunkiaer (1934) as the basis of the habitat categories. It is widely recognized (e.g., Woodward and Rochefort 1991) that, at a continental level, biomes need to be defined in terms of the physiognomy and life forms of the dominant species, because individual species are too limited to encompass widely dispersed geographical locations. Ecological behaviour of species can also vary within their distribution and vicarious species further preclude the use of individual species. A given species may also show plasticity, because of environmental and local factors such as grazing, so the overall height of the whole unit is used a measure of its status at a given time. Further advantages of using life forms are that they provide direct links between in situ data and dynamic global vegetation models (e.g., Sitch et al. 2003), but also with the patterns present on satellite images because of their relationship with vegetation structure.
Plant life forms (Raunkiaer 1934) are defined on the basis of the location of buds in the adverse season and separate grassland, shrub and forest species which can be used to develop rules for habitats that can be applied consistently in the field. Within the shrub and forest categories a further breakdown is made according to the way the leaves of the plants are retained in the adverse growth season. Raunkiaer demonstrated that the life form spectra in different regions were correlated with the main environmental gradient from the equator to the arctic: they are therefore widely used in global change modeling as indicators for projecting vegetation change (e.g., Sitch et al. 2003).
Life forms adopted for recording General Habitat Categories (GHCs), based on life forms as defined by Raunkiaer (1934)
1. Submerged hydrophytes
Plants that grow beneath the water. This category includes marine species and floating species which over-winter below the surface.
2. Emergent hydrophytes
Plants that grow in aquatic conditions with the main plant above water.
Plants that plants that grow in waterlogged conditions.
4. Leafy hemi-cryptophytes
Broad leaved herbaceous species, sometimes termed forbs.
5. Caespitose hemi-cryptophytes
Perennial monocotyledonous grasses and sedges.
Annual plants that survive the unfavorable season as seeds.
7. Succulent chamaephytes
Plants with succulent leaves.
Plants with buds below the soil surface.
Non-saxicolous bryophytes and lichens, including aquatic bryophytes,
10. Herbaceous chamaephytes
Plants with non-succulent leaves and non-shrubby form.
Shrubs and trees
11. Dwarf chamaephytes
Dwarf shrubs: below 0.05 m
12. Shrubby chamaephytes
Under shrubs: 0.05–0.3 m
13. Low phanerophytes
Low shrubs buds: 0.30–0.6 m.
14. Mid phanerophytes
Mid shrubs buds: 0.6–2.0 m
15. Tall phanerophytes
Tall shrubs buds: 2.0–5.0 m
16. Forest phanerophytes
Trees: over 5.0 m
Leaf retention divisions (to be used in conjunction with TRS)
Summer deciduous and/or spiny cushion
A major problem of using theoretical habitat classifications for monitoring is the proliferation of classes; e.g., Morillo Fernandez (2003) distinguished almost 1,000 classes and in EUNIS there are 350 classes at level three. Within the BioHab, as shown in Fig. 1, all possible feasible combinations of grouped pairs of life forms are included, to ensure complete coverage of Europe. The number is restricted by rules using percentages and prioritisation to exclude combinations which would include more than two life forms. For ‘Trees and Shrubs’ leaf retention divisions are also included, but not all of these are present in all height categories; e.g., there are no native Summer Deciduous trees over five metres in height in Europe. This procedure is arbitrary, but reproducible, and has restricted the number of GHCs to 130 in the Pan-European region, excluding Turkey. Other life forms, e.g., tall succulents, would have to be included for other continents. This restricted list acts as a lowest common denominator and enables decisions at the highest level to be made in the field, or to be derived from extant data (e.g., vegetation relevées). More detailed information (see below) is recorded in subsequent columns for the interpretation of change at the landscape level.
The determination of the GHC is based upon a series of five dichotomous divisions as shown in Fig. 1. These determine the set of life forms that can be used to identify the appropriate GHC. The first decision concerns whether the element is ‘Urban’, the second whether it is a ‘Crop’, the third whether it is ‘Sparsely Vegetated’, the fourth whether it is ‘Trees or Shrubs’, and the fifth whether it is ‘Wetland’ (Fig. 1). As discussed below, rules have then been added for further divisions in all super categories and habitat categories, including percentage criteria.
A matrix of environmental conditions was constructed for ease of recording, as described by Bunce et al. (2005), with moisture classes on the horizontal axis and soil factors on the vertical axis. Moisture classes suitable for application across the range of European habitats were adapted from Pyatt (1999). The soil factors are based on indicator values originally developed by Ellenberg et al. (1992) for Central Europe, but these are not available for all regions, so local experience on the ecological amplitude of indicator species may be needed. The overall balance of species should be used, not individual indicator plants. For example, in the Pannonian region the presence of some individuals of Melica ciliata is insufficient to assign the term ‘xeric’ to the element.
A provisional list of site qualifiers has been constructed (Bunce et al. 2005) and includes factors such as coastal attributes and rock types. Management qualifiers are grouped in convenient sections, e.g., ‘Forestry’ and ‘Recreation’, and are designed to give information on potential causes of change. This list will need further refinement and validation in a Pan-European field survey. Whilst the management qualifiers are more difficult to record consistently, Kirby et al. (2005) showed that if sufficiently well defined habitat categories are provided, then change can be reliably determined.
The recording procedure
The following section discusses the principal aspects of the recording process including practical mapping procedures. Standard data sheets and provisional lists of qualifiers are given in Bunce et al. (2005).
No continent-wide survey can be carried out without adequate field training for all surveyors to ensure that terms are fully understood and interpreted in the same way. For example, environmental terms are often used within a local context, e.g., ‘dry’ in Scotland may be ‘mesic’ compared with southern Italy. Surveyors across Europe therefore need to be familiar with predefined environmental categories. In the field, combined teams of two people, preferably consisting of a botanist and a cartographer, are needed to ensure that the necessary expertise is available.
The date for the recording of GHCs should be based on local phenology. The extent of the window needs to be set by region, using local information, and differs between environmental zones. The state of development of the vegetation at the recording date should therefore be relatively consistent between zones; thus in the Mediterranean region the recording period will be earlier than in central and northern Europe. Barr et al. (1993) showed that differences between dates of survey are a major source of noise in change statistics. Repeat visits for monitoring should therefore be carried out as close as possible to the date of the original visit, assuming that there is no shift in timings of the seasons.
Data quality control (i.e., supervision of surveyors) and assurance (i.e., independent checks of recording) are all essential to produce robust data. Barr et al. (1993) analyzed random checks of comparable categories to GHCs and showed a correspondence of 84%. Any future program would need to incorporate such checks, so that policy makers and scientists would have confidence in the results.
All major decisions are made in the field. At a later stage, it is possible to extract relevant data in the laboratory from available datasets (e.g., slope angles, and geology). Other more detailed data are added in the field, as described below. Brandt et al. (2002) emphasize that the quality of mapping is dependent on sufficiently accurate base maps. It is therefore preferable to carry out preparatory work on ecological interpretation and subsequent delineation of the major elements within the survey area from aerial photographs and related material, e.g., cadastral maps, preferably at a 1:10,000 scale. Surveyors therefore annotate the base map with labels attached to individual elements according to the rules. The boundaries of some elements may need to be adjusted or new parcels described which were not defined in the preparatory work, e.g., different categories of grassland can often not be seen on aerial photographs.
The procedure was initially developed for mapping 1 km square samples, but is also suitable for other scales, e.g.; Cooper and McCann (2002) used 0.25 km squares and Bloch-Petersen et al. (2006) applied the GHCs to small biotopes below about 200 m2. Within the 1 km square sample unit, the surveyor delineates all habitats with an area greater than 400 m2 (Minimal Mappable Element—MME). Figure 2 gives an example of a mapped 1 km square and a recording sheet (Bloch-Petersen et al. 2006). For each delineated unit the surveyor determines the GHC (Field 1) and the environmental, site, and management qualifiers, which are in sequential fields on the recording sheet (Fields 2–4). Next, all life forms with a cover of over 10% are recorded and individual plant species or crops with a cover of over 30% in the mapping unit (Field 5). Three further fields are provided for existing Pan-European habitat classifications (e.g., EUNIS (Davies and Moss (2002)), national habitat classifications (e.g. Morillo- Fernandez 2003) and phytosociological associations (e.g., Rodwell et al. 2002) depending upon the objectives of the project and the experience of the surveyors.
Although the MME has to occupy at least 400 m2 it can be a complex shape, so long as the shortest measurement is over 5 m, as in the GB Countryside Survey, and checked for Europe during BioHab. This contrasts with the 10,000 m2 of the CORINE land cover map and 2,500 m2 of the Biopress Project (Olschofsky et al. 2006). However, the finer detail of the MME is essential to express the landscape ecological characteristics of small scale landscapes; e.g., in Crete (Greece), Asturias (Spain), and Brittany (France). Bunce et al. (2005) provide detailed rules for mapping some elements, e.g., motorways will be mapped as areal elements, but may subsequently be allocated to linear features by database management for specific objectives (e.g., Haines-Young et al. 2000). The fundamental principle is that disaggregated data are collected, so that subsequent analyses can be sufficiently flexible to answer a variety of policy and landscape ecological objectives, e.g., loss of hedgerows and fragmentation of habitats.
Linear and point elements
Linear and point elements are often excluded from habitat surveys. However, many landscape ecological studies have shown that, especially in intensively managed agricultural landscapes, biodiversity has progressively become restricted to such features (e.g., Hermy and De Blust 1997). Whilst this process may have stabilized in Western Europe, it is likely to continue in Central Europe. Many cultural landscapes are rich in such features, largely as a result of management; e.g., terraces in Tuscany (Italy), walls in the Auvergne (France) and ponds in Cheshire (GB). It is therefore essential not only to assess the resources of linear and point elements in representative landscapes but also to monitor their patterns and change.
The same recording format as described in the previous section is used for linear and point elements, but on a separate sheet, in order to assist the recording process. The variation across different types of landscape can subsequently be integrated through the use of Geographical Information Systems (GIS), and the contribution to biodiversity of areal, linear and point features compared. In some projects, e.g., Cooper and McCann (2002) habitats only may be recorded, but data on other biota may also be collected in the same sites, e.g., vegetation and freshwater invertebrates (Haines-Young et al. (2000)) in order to present an integrated picture of biodiversity at the landscape scale.
Linear elements have a Minimal Mappable Length (MML) of 30 m. Those features that comprise only vegetation must be wider than 0.5 m, but less than 5 m wide in order to exclude narrow strips (e.g., lines of vegetation beside walls). Elements that are smaller than 400 m2 and shorter than 30 m can be recorded as points. Linear habitats often occur as complexes; e.g., a fence, a ditch and a hedge; in which case instructions are provided for mapping, so that a given combination is always recorded by single alpha-numeric code incorporating its detailed composition.
In some cultural landscapes the number of point features can be large, e.g., individual trees in parkland or hedgerows. Two guidelines are provided for recording such points. Firstly, the recorded point features should add to landscape diversity, usually because they represent a particular habitat which is generally absent from the surrounding area, e.g., rock outcrops or boulders in a grass field. Secondly, the recorded point features should also have an effect on the ecological functioning of landscapes, e.g., small water bodies which act as drinking places in grasslands, or weirs in watercourses, which hinder migration of fish. However, a given survey may decide to omit point features, in which case this should be documented on the separate general information sheet, which also includes information such as the date of survey and ownership details (Bunce et al. 2005).
Field testing and validation
The theoretical basis of the model is the correlation between the complexes of life forms and the environment. It is the substance of classical biogeography and can therefore be tested. The first such test was carried out in a valley in the Picos de Europa (Spain) which extends from evergreen forest at 200 m to rock and sub-alpine habitats at 2,500 m. Orthogonal regression, as described by Bunce et al. (1996), was used to calculate the correlation between Detrended Correspondence Analysis (DCA) scores of the mixtures of plant life forms recorded in 80 stratified random samples of 0.25 km square, drawn from eight environmental strata, using the mean altitude of each stratum as the independent variable. The correlation coefficient was 0.94 (6 df) and highly significant (P < 0.001) showing that the model is valid.
In the second test, the data used was for proportion of life forms in areal elements collected during the field excursions and workshops shown in Fig. 3. The results are only indicative because, although they include all environmental zones of Europe, they were not randomly stratified. The data were analyzed by Canonical Correspondence Analysis using the environmental zones as the independent variable (Metzger et al. 2005). The results confirm the hypothesis of Raunkiaer (1934) that life form spectra are correlated with the environment. However, these initial results indicate that there are several significant dimensions, e.g., from bare rock to habitats dominated by annual plants, and from grasslands to habitats with summer deciduous species. The axes from the analysis of the life forms were associated with the environmental zones of Europe, with Alpine North (i.e., Scandinavian mountains) and Mediterranean South (i.e., extreme southern Europe) being at the opposite ends of the primary gradient. Life form combinations are more important than the individual categories in expressing the overall environment, but also show modified patterns because of management by man. As with recording GHCs, individual species may diverge from the overall pattern, e.g., Koenigia islandica is an annual which grows in arctic environments dominated by chamaephytes.
Data collected from monitoring and surveillance of European habitats would provide direct support for European nature conservation policy. Such data would also have policy relevance to issues concerning the rural environment (e.g., agri-environment schemes). Policies on environmental issues can only be developed with knowledge of the stock and change of the environmental resources. Projects such as MIRABEL (Petit et al. 2001) have only been able to use expert judgment for assessing the distribution and extent in European habitats and the potential change caused by driving forces. The value of such studies would be greatly enhanced by actual habitat data. Mücher et al. (2005) have used the descriptions in Annex 1 of the Directive to derive rules which use existing databases to predict distribution of habitats. However, many of the descriptions do not contain enough detail for mapping, and reliable in situ data is lacking in many cases. In addition, many large scale European projects have no field validation of the results.
Inevitably, protected sites (e.g., in Europe the Natura 2000 sites) can only cover a limited proportion of the European land surface, and outside their boundaries there is little or no protection of habitats. Nevertheless, the non-designated ‘wider countryside’ contains a high proportion of the total wildlife resource, interacts with protected sites, and is also the domain that most people experience in everyday life, with recent pressures leading to major losses of biodiversity and changing landscape patterns. On the one hand there has been agricultural intensification and urbanization and, conversely, more isolated or less productive regions have become marginalized and abandoned (EEA 2005). Such changes will have major consequences on rural communities as well as habitats and biodiversity (Metzger et al. 2006). The BioHab procedure is designed to detect and report such change, with the ability to cover adequately the complexity of landscapes and spatial heterogeneity across Europe. It can thus provide European policy makers with statistical estimates of the stock and change in distribution of habitats in relation to environment and landscape ecology. The results need to be communicated using categories that will inform the public (e.g., figures on abandonment, marginalisation, and encroachment) and encourage further research. The data will also form a control against which to test the effectiveness of protection measures and could also be used to stimulate analysis of landscape ecological parameters at the European level. Many comparable processes are occurring throughout the world, as relevant abstracts in Bunce et al. (2007) show. The transferability of the categories described above, together with additional units for biomes not present in Europe, could help to assist international cooperation on landscape change and identify common driving forces.
A provisional list of life form categories outside Europe has already been prepared, and field work in Israel has already demonstrated how further categories can be added for deserts. However, habitats such as the tropical rain forest have complex structures with many levels of vegetation, which cannot be adequately represented by the vertical perspective. Further work is therefore needed to define appropriate additional categories and the necessary supporting rules.
A benefit of the sample approach is that detailed spatial and temporal data can be collected and can then be used in scenario studies or modelling exercises, as demonstrated in GB (Bunce et al. 1993; Parry et al. 1994). At a more detailed level, the GHCs provide a framework for placing extant figures onto a common basis, by screening available datasets, and then supplementing them by further survey, to produce data which could eventually lead to European estimates. Bloch-Petersen et al. (2006) have shown how the GHCs can be derived from existing studies. Recent work, in the GB Countryside Survey 2007 field program also indicates that there is direct correspondence of GHCs with existing disaggregated data on habitats.
In conclusion, this paper presents a procedure that has been based on experience, over the last thirty years, of recording and reporting habitats and spatial information at the landscape scale. It would enable integration between many European projects and would also enhance the understanding of landscape ecological change, as well as stimulating international collaboration.
The work presented in this paper was carried out as part of the EU Fifth Framework project BioHab (EVK2-CT-2002-20018). We thank K. Zaunberger and M. Sharman for their continued advice, support and interest. In addition, thanks are also given to the many other scientists who contributed valuable discussions and ideas, especially IALE members participating in the Ecoland Forum Working Group.