A typology of frameworks with particular reference to the global scale
It is important to recognise that in practice, the designation of protected areas has been and will continue to be influenced by diverse and complex environmental, socio-economic and political factors (e.g. Sutherland, 1998; Jepson & Whittaker, 2002b). But in the following account, in the interests of simplification, we focus exclusively on the scientific principles underpinning the designation of protected areas.
The concept that the world is a patchwork of regions differing in their biological makeup has profoundly influenced how humanity perceives and interacts with the natural world. The concept of biological regions emerged as a result of European expansion in the tropics from the 16th century onwards and the desire to understand the diversity, richness and patterns of nature for religious, intellectual and economic purposes (Lomolino et al., 2004). The realisation that nature differed in different parts of the world, and the subsequent spatial representations of these differences, played an important role in the planning and execution of the European colonial endeavour, the creation of national identities, and the development of international tourism among other things. Maps play an important role in the communication of biogeographic theory and are powerful tools within conservation biogeography (cf. Dalton, 2000; Myers & Mittermeier, 2003). But, maps are only a symbolic representation of nature (Demeritt, 2001), and the choice of criteria and variables used to create the maps will be strongly influenced by what the creators of the map value (Williams et al., 2000a).
In the 1960s there was widespread support within the conservation community for the goal of establishing a worldwide network of natural reserves encompassing representative areas of the world's ecosystems. Under the aegis of the IUCN, Dasmann (1972, 1973) and Udvardy (1975) put this ‘biogeographic representation principle’ into practice, by extending and combining earlier maps of faunal regions (Wallace) and vegetation zones (Clements) to create a nested hierarchy of biological regions (Table 1, Jepson & Whittaker, 2002a). Their work is foundational within conservation biogeography and provided the framework for a massive and rapid expansion of protected areas globally.
The Dasmann–Udvardy framework subdivides the globe into faunal regions (biotic realms) within which a biome classification system is applied, with biotic provinces delineated by subdividing a physiognomically defined climax vegetation type on the basis of a distinctive fauna. Areas with less than 65% of their species in common are delimited as separate faunal provinces, and this essentially zonal scheme is afforced by the recognition of azonal features such as high mountains and mountainous islands. As originally developed, the approach was confined to bird and mammal data and to a coarse scale of application, but it was subsequently refined for use at a finer scale, by using the same algorithm for smaller geographical units (Mackinnon & Wind, 1981; Mackinnon & Mackinnon, 1986a,b; Jepson & Whittaker, 2002a). This scheme provides a transparent methodology with a clearly defined purpose, and in principle the analysis can be repeated to assess the implications of, for instance, using updated distributional data, a different system of biogeographic regions (cf. Cox, 2001), or threshold similarity value.
Since the development of the biogeographic regions approach, the tools for biogeographic research have undergone a radical transformation, as computerised databases and advanced spatial analytical approaches have been developed. Whilst the Dasmann–Udvardy approach remains an important foundational scheme at a global scale (Table 1), it has in practical terms been superseded at this scale by other global schemes developed and promoted by major conservation NGOs, whilst within regions a far greater array of approaches have been developed both by NGOs and other actors from the scientific community (e.g. Redford et al., 2003). In Table 1 we identify the WWF Ecoregions scheme as a prominent global scheme based on the principle of biogeographic representation (e.g. Dinerstein et al., 1995). It incorporates data on biogeography, habitat type and elevation to identify biogeographic units at a finer scale than the original Dasmann–Udvardy framework. The approach aims to meet the goals of representation and also to draw up natural units within which ecological flows and linking processes are maintained. In this way, the approach combines both compositionalist and functionalist perspectives (as Callicott et al., 1999; Williams & Araújo, 2000, 2002). The WWF Ecoregions approach, specifically as applied within Indonesia, has been criticised by Jepson & Whittaker (2002a) for a variety of reasons, including data deficiencies and a lack of transparency as to the criteria involved in fine-scale designations. No schemes are entirely free of such problems, and Wikramanayake et al. (2002) provide a defence of the approach in an equally forthright reply. The scheme's proponents are of course driven by the heavy responsibility of getting on with the job of putting theory into practice, and effecting conservation actions on the ground. But, it would be naïve to think of any such map as being the final say in land-use planning and designation, as such battles have to be constantly re-fought and occasionally land abandonment provides new opportunities for ecological restoration and/or for inclusion of new areas into protected area networks (cf. Meir et al., 2004). Magnusson (2004) argues that as ecoregion approaches are being heavily promoted, more rigorous tests of their delineation are urgently needed. He was writing about an attempt to test the validity of Bailey's foundational ecoregions scheme, but the same case can be made for the WWF scheme. Magnusson suggests a simple approach involving looking for natural breaks in species distributions across ecoregion boundaries, and a more sophisticated approach to the problem is illustrated by Williams (1996) and Williams et al. (1999).
To improve further on the alternative representation frameworks available requires several components. First, the criteria of the scheme must be transparent and explicit. Second, the data on which the frameworks are built must be available and of a quality sufficient to the task, involving both biogeographical data and environmental data. Third, schemes once developed should be subject to further, independent tests of their efficiency and robustness (as e.g. Stoms, 1994; Williams et al., 1999). In practice, the broad framework for biogeographic representation is generally agreed at the global scale (although, see Cox, 2001) and it is arguably at the meso-scale that improvements in data quality might produce significant alterations in the maps and adjustments in conservation efforts.
An alternative approach to reliance on biogeographic data sensu stricto was formalised by Faith & Walker (1996a,b), who based their approach on the correspondence between increasing numbers of (and/or increasing difference between) environmental domains and increased biodiversity. They noted that sampling environmental pattern, or compositional variation within one indicator group, would predict compositional variation within other groups if they spanned the entire range of habitats or environments available. Hence, sampling environmental pattern (or assemblage pattern) itself would be an alternative to selecting areas using species distribution data directly. Whilst this approach has some merits, its effectiveness in achieving pre-determined goals has yet to be consistently demonstrated (compare Faith and Walker, 1996a,b; Ferrier & Watson, 1997; Araújo et al., 2001, 2004b). For example, Araújo et al. (2001) showed that European plants exhibited consistent non-random positive patterns of representation with conservation areas selected to maximise pattern variation among environmental domains, but that terrestrial vertebrates (especially reptiles and amphibians) were consistently under-represented with this approach. They argued that the degree of success of environmental-diversity strategies would be contingent on the contribution of historical biogeography shaping current distributions. According to this view, taxa with poor dispersal abilities (e.g. reptiles and amphibians) would be in clear disadvantage compared with those of good dispersal abilities (e.g. plants), which would more easily achieve conditions of quasi-equilibrium with current environmental conditions, and hence would be better predicted by environmental-based surrogates. Although evidence of the ability of environmental- and assemblage-based models to represent biodiversity at a rate higher than expected by chance is currently lacking (Araújo et al., 2003), the approach is still strongly advocated by its proponents (e.g. Faith, 2003; Faith et al., 2004). This is one example where conservation values, epistemology, theories of equilibrium in biogeography and numerical ecological techniques are intermingled in a complex but possibly important debate within Conservation Biogeography.
Hotspot approaches at the global scale contrast with biogeographical representation approaches by focusing on the richness and endemism of areas combined with a measure of the threat to biodiversity (Table 1). There are a number of prominent global schemes, including: the IUCN–WWF Centres of Plant Diversity, Birdlife International's Endemic Bird Areas (e.g. Long et al., 1996; Stattersfield et al., 1998), and Conservation International's (CI) hotspots (Myers et al., 2000). Applications of the hotspots approach are based, partly through necessity, on a limited array of taxa. For instance, the 25 CI hotspots are delimited solely on the basis of plant endemism and habitat conversion statistics. Myers et al. (2000) claim that the hotspots hold 44% of the world's plant species and 35% of the vertebrates in 12% of the land area of the earth. These taxa, however, represent a relatively small proportion of the species on earth, and it remains to be established what proportion of, for example, insects, are found within these areas. The areas delimited are also far too coarse, of themselves, to guide issues such as reserve placements. Rather they provide a global ‘cookie cutter’ of large areas deemed most deserving of conservation funding and attention. The approach has been remarkably successful in terms of fund raising, and if successful in implementation could make a significant contribution to reducing global biodiversity loss (Myers & Mittermeier, 2003).
The CI hotspots approach can be criticised on several grounds, including: (i) that it serves only a limited set of values (Jepson & Canney, 2001); (ii) that it sends out a powerful if entirely unintentional signal that biodiverse areas excluded from the list of hotspots do not matter in conservation terms (Bates & Demos, 2001); (iii) that it is a very coarse-resolution and simplistic analysis (Mace et al., 2000); and (iv) that there seems to be no underlying base map, and certainly nothing approaching an equal-area grid cell system is involved, so that the basis for delimiting the boundaries of the areas selected appears essentially arbitrary. Moreover, in examining the applicability of the approach to the marine realm, Hughes et al. (2002) find that first, richness peaks do not coincide with centres of high endemicity in Indo-Pacific corals and reef fishes, and second, that fish and coral endemicity patterns are at variance with one another. They therefore argue against marine prioritisation by ‘hotspots’, instead calling for a focus on preserving connectivity and genetic diversity of widely dispersed species combined with intensive protection of quite localised areas of high endemicity.
Criticisms have also been levelled at the way in which CI's 25 selected terrestrial hotspots have been ranked internally (Brummitt & Lughadha, 2003; Ovadia, 2003), with Brummitt & Lughadha correctly arguing that the use of unscaled species–area ratios to select the ‘hottest hotspots’ is inappropriate. Myers & Mittermeier (2003) sidestep the scientific criticism in their response to Brummitt & Lughadha (2003), pointing to the urgency of action for conservation, the $750 m raised in support funds for CI's hotspots scheme, and expressing frustration at the slowness of the scientific community to engage with their approach. The tension between the priorities of major NGOs and academic debate is understandable. The point of our listing the above criticisms of the WWF Ecoregions and CI hotspots programmes is not to undermine them as a basis for taking conservation action now, but to encourage an ongoing effort from conservation biogeographers to update, revise and test protected area planning frameworks at global and also regional scales. We see this as valuable because first, the broader goals of nature conservation are such that we need alternative sets of maps, capturing and representing different value sets, and second, because as our data, theory, and models improve, we can refine our planning frameworks to better serve these goals (cf. Mace et al., 2000; Williams et al., 2000a; Rodrigues et al., 2004). Moreover, as such large sums and efforts are involved, it is surely crucial that we provide the checks and balances to ensure that the resources are deployed effectively, and one part of this is about a continuing process of reviewing the scientific basis (Meir et al., 2004).
To complete our typology of global protected area approaches (Table 1), we have designated a third type of approach —‘important areas’— as being distinct from representation and hotspot approaches. As exemplified by Birdlife International's Important Bird Areas (IBAs) scheme, important areas are key sites for conservation, small enough to be conserved in their entirety and often already part of a protected-area network. The stated criteria for selection of IBA sites are that they do one or more of the following things: (i) they hold significant numbers of one or more globally threatened species; (ii) they are one of a set of sites that together hold a suite of restricted-range species or biome-restricted species; and (iii) they have exceptionally large numbers of migratory or congregatory species (http://www.birdlife.net/action/science/sites/ visited September 2004). The scheme provides a fine-scale network of sites, below the level of resolution of Birdlife's Endemic Bird Areas approach: for instance there are some 4000 IBA sites in Europe alone, in contrast to 218 EBAs globally.
In moving from the global to the regional scale of application, it is clear that there are numerous approaches to mapping and prioritising putative or existing protected areas (e.g. Pressey & Nicholls, 1989; Margules & Pressey, 2000; Williams et al., 2000b; Cabeza & Moilanen, 2001; Williams & Araújo, 2002; Dimitrakopoulos et al., 2004). Typically these approaches take grid-cell based species range data and use reiterative computer algorithms to select complementary sets of cells that achieve a predetermined goal. These approaches have the advantage of making goals, values and priorities explicit before area priorities are defined. They are quantitative, repeatable and ensure that efficient and effective solutions are obtained for a set of pre-selected goals and data (Williams et al., 2000a). Unfortunately, they are also very data-hungry and highly sensitive to data quality (Flather et al., 1997; Freitag et al., 1998; Araújo, 2004; Araújo et al., 2004b). Hence, quantitative ‘gap analyses’ have until very recently been restricted to problems of regional extent. The application of this approach at a global scale by Rodrigues et al. (2004) using data for several groups of terrestrial vertebrates thus represents an important development, supplementing the approaches identified in Table 1.