Introduction

Russell A. Mittermeier¹, Patricio Robles Gil⁴, Michael Hoffmann², John Pilgrim², Thomas Brooks², Cristina G. Mittermeier¹, John Lamoreux³, Gustavo A.B. da Fonseca^{1, 2, 13}, Keith Alger¹, Frederick Boltz^{1, 14}, Katrina Brandon², Aaron Bruner², José Maria Cardoso da Silva⁵, Assheton Carter¹, Roberto Cavalcanti^{1, 15}, Don Church¹, Matthew Foster², Claude Gascon¹, Larry Gorenflo², Brian Gratwicke⁶, Marianne Guerin-McManus¹, Lee Hannah², David Knox², William R. Konstant^{1, 7}, Thomas Lacher², Penny Langhammer², Olivier Langrand¹, Nicholas Lapham¹, Dan Martin¹, Norman Myers⁸, Piotr Naskrecki^{2, 9} Michael Parr¹⁰, David Pearson¹¹, Glenn Prickett¹, Dick Rice², Anthony Rylands², Wes Sechrest^{2, 12}, Michael Leonard Smith², Simon Stuart¹², Jorgen Thomsen¹, Michael Totten¹ and Justin Ward¹

Global Priority Setting for Biodiversity Conservation
History of the Hotspots concept
Impact of Hotspots
An Updated Hotspots Analysis
Synthesis of the Updated Hotspots Data
Threats and Responses in the Hotspots
Hotspots as Priorities within Bioregional Classification
The Future: Hotspots and Species Distributions
Beyond Hotspots: Identifying Conservation Targets on the Ground
Coda: Hotspots as a Line in the Sand

Global Priority Setting for Biodiversity Conservation

Life on Earth faces a crisis of historical and planetary proportions. Unsustainable consumption in many northern countries and crushing poverty in the tropics are destroying wild nature. Expanding agriculture, industry, and urbanization are fragmenting and eliminating natural environments; accidental and deliberate introduction of exotic species is wreaking havoc on native communities; pollution is altering complex biogeochemical and climate cycles through the land, air, and water; and hunting, trade, and overfishing are decimating the last populations of large vertebrate species (Vitousek et al. 1997). Biodiversity is besieged.

Extinction is the gravest aspect of the biodiversity crisis: it is irreversible. While extinction is a natural process that is part of the history of this planet, the fossil record indicates that, in the absence of humans, the life span of a species averages one million years (May et al. 1995). Now, however, human impacts have elevated the rate of species extinction by at least a thousand, possibly several thousand times the natural background rate (Pimm et al. 1995). Mass extinctions of this magnitude have only occurred five times before in the history of our planet; the last, probably caused by a cataclysmic asteroid impact 65 million years ago, brought the end of the dinosaur age (Álvarez et al. 1980).

It is easy to imagine the disasters that humanity would face if the rate of other natural processes, such as the frequency of floods or disease transmission rates, increased a thousand-fold. The world as we know it would be devastated. But what exactly do we lose with the catastrophic extinction of other species? Foreclosing future resource-use options is perhaps the most obvious consequence. Scientists have recognized only a fraction (maybe less than 10%; perhaps even as little as 1%) of the species with which we share our planet, and know the biology of even fewer (Novotny et al. 2002). Thus, with species extinction we destroy a vast genetic storehouse (Myers 1983) that could one day be found to hold, for instance, a cure for AIDS. Current developments in the treatment of malaria, one of the world's biggest killers, use artemisinin-based compounds derived from the plant Artemisia annua (Sachs 2002). The biodiversity crisis could be compared with burning down the world's libraries without knowing the titles of 90% of the books or the content of most of the pages of the known books. Less tangibly, but no less importantly, species extinction inflicts a deep cultural, spiritual, and moral wound on humanity. All of the world's societies value species for their own sake, over and above any utilitarian purpose, and wildlife – especially the larger vertebrates and many plants – are an integral part of the fabric of all human cultures (Wilson 1984; Wilson and Kellert 1993).

The road to extinction is also perilous to people. For example, the destruction of montane forests causes frequent, massive landslides with dreadful human cost – witness the tragic mudslides that killed 10 000 people following Hurricane Mitch in Honduras in 1998 (Hellin et al. 1999) – while the Severe Acute Respiratory Syndrome (SARS) outbreak in East Asia has been directly linked to the trade in wildlife for human consumption (Guan et al. 2003). Other consequences of biodiversity loss are more subtle and cumulative, but equally significant, such as the progressive deterioration of the natural basis for sustainable economic growth. A number of high-profile studies (e.g., Costanza et al. 1997) have placed the annual economic value of ecosystem services such as climate and water regulation, pollination, and recreation in the tens of trillions of dollars – costs that society would have to bear if these generally free but unaccounted services were no longer accessible. A significant fraction of such services can be attributed directly to biodiversity. Balmford et al. (2002) concluded that conversion of natural ecosystems to anthropogenic landscapes roughly halves their economic value.

In order to stem the extinction crisis most effectively, we must prioritize where we should act first. To a large extent conservation is, and always will be, local. People care most about what is happening in their own backyards. We believe that all biodiversity is important and that all nations and communities, large or small, should do everything possible to conserve the biological riches on which they depend. However, some local efforts have planetary consequences and thus justify priority allocation of scarce financial resources.

The establishment of priorities for biodiversity conservation is a complex issue (Margules and Pressey 2000). The problem can best be framed by a question: In which areas would a given conservation dollar contribute the most towards slowing the current rate of extinction of global biodiversity? Species are distributed unevenly around the world (Gaston 2000), which means that mapping this variation is essential if we are to address the question. However, we can not simply measure the numbers of species living in particular areas. This is because several species-rich areas might hold a large fraction of the same species, meaning that the overall number that could be conserved within such areas might be rather small (Pressey and Nicholls 1989). Instead, we must measure not species richness but endemism: the degree to which species are only found in a given place. This can be thought of as a measure of irreplaceability – in essence, the number of geographic options one has for the conservation of the species found in a given area (Pressey et al. 1994). Since we can not conserve a species that is endemic to a given area anywhere except in that area, the area is wholly irreplaceable at a global scale.

A further problem concerns which species we should evaluate. We know that we can not map all species because we have not even named most of them. Quite fortuitously, vascular plants and vertebrate animals – the species we know best – tend to be large, and play prominent roles in structuring ecosystems (Terborgh 1988), although species that we know less about are also vital for ecosystem processes (Wilson 1987). Whether or not the distributions of plants and vertebrates are mirrored by the myriad of unknown terrestrial invertebrate species remains an open question, although some evidence suggests that they may be (Howard et al. 1998). Some taxa, such as tiger beetles, seem to exhibit excellent congruence with many other groups (Pearson and Cassola 1992; Carroll and Pearson 1998), while others show less clear patterns (Van Jaarsveld et al. 1998). These comparisons are also heavily dependent on scale (Reid 1998). At regional scales we often see much greater congruence than at fine scales (Pearson and Carroll 1999). Thanks to recent advances in bioinformatics, we will soon be able to use massive datasets on invertebrate species distributions to delimit the boundaries of biologically unique areas more precisely than ever (Meier and Dikow 2004). One of the largest gaps in our current knowledge remains in the aquatic realm, which is in critical need of effective conservation action. Distributions of marine and freshwater species remain largely unknown, although ongoing projects are addressing this issue.

Our ultimate goal is to keep nature intact, which means that we must stop anthropogenic species extinctions. To approach this goal, we must slow the rate of species extinction as much as possible (over and above simply conserving as many species as we can) with whatever conservation resources we have at our disposal, which requires incorporating threats (or vulnerability) and costs into priority setting. Like species, threats are hard to measure. The extent of habitat destruction is one useful metric, given the welldocumented relationship between the size of an area of habitat and the number of species it retains (Brooks et al. 1997). Other measures, such as human population density (Balmford et al. 2001), are also used.

Threats and costs are generally related to each other; the more threatened an area is, the more it will cost to conserve (Ando et al. 1997). However, the relationship is not always linear, as it depends in large part on the economic conditions of the country and immediate locale in which the priority area is located (Balmford et al. 2003). Some extremely threatened areas can still be conserved at low cost (e.g., much of Madagascar and some areas in Southeast Asia), often by addressing underlying poverty simultaneously with biodiversity conservation, while others tend to be quite expensive (e.g., the California Floristic Province and New Caledonia). However, because economic opportunity costs vary dramatically across the landscapes of hotspots and wilderness areas, there still exist areas of relatively low cost in all hotspots, offering great conservation opportunities, as well as areas of high cost in wilderness areas requiring immediate attention to threats (Chomitz et al. 2004).

We still face a paradox in determining how to incorporate threats, costs, and opportunities into conservation priorities. Intuitively, we want to conserve the most threatened areas first, to avoid losing them the soonest. But we also want to get the greatest return for our conservation dollar, which in theory would mean targeting the areas of lowest cost, greatest opportunity, and least threat first. This paradox can be resolved by the measurement of irreplaceability – or the degree of endemism (Mittermeier et al. 2003a). Thus, we identify those areas that hold species found nowhere else and that are guaranteed to lose species if the areas are not conserved. Among these, we rank our actions based on threats, with the most threatened biodiversity receiving the most urgent action. Wherever we have choices, or equal levels of endemism, we should select opportunities for attending to areas that are the least expensive to conserve (and often the least threatened). In effect, we need a dual conservation strategy that always prioritizes endemic-rich areas and ensures that we protect the most threatened places with species that we will otherwise lose, while preemptively protecting equally unique places that are not yet under extreme threat.

History of the Hotspots Concept

A seminal paper by Norman Myers (1988) first articulated the principles of irreplaceability and threat to inform terrestrial conservation priorities on a global scale. Myers identified ten tropical forest “hotspots” characterized both by exceptional levels of plant endemism and by uncommon rates of habitat loss, although without quantitative criteria as to what exactly constituted a hotspot. Subsequently, Myers (1990) added a further eight hotspots, including four Mediterraneantype ecosystems. Conservation International adopted Myers' hotspots as its institutional blueprint in 1989, making minor modifications and additions over the next seven years. In 1996, Conservation International made the decision to undertake a reassessment of the hotspots concept, including an examination of whether key areas had been overlooked. This was done in collaboration with Myers and took three years. A preliminary report (Mittermeier et al. 1998) was followed by an extensive global review (Mittermeier et al. 1999), a scientific analysis (Myers et al. 2000), and a detailed online publication (www.biodiversityhotspots.org). These efforts introduced quantitative thresholds for the designation of hotspots. To qualify as a hotspot, a region had to meet two strict criteria: it had to contain at least 1 500 species of vascular plants (>0.5% of the world's total) as endemics, and it had to have 30% or less of its original vegetation (extent of historical habitat cover) remaining. This analysis identified 25 hotspots, collectively holding as endemics no less than 44% of the world's plants and 35% of terrestrial vertebrates (mammals, birds, reptiles and amphibians) in an area that formerly covered only 11.8% of the Earth's land surface. However, the fulcrum around which these startling results were presented was that this land area had been reduced by 87.8% of its original extent, such that this amazing wealth of biodiversity was restricted to only 1.4% of Earth's land surface.

Concurrent with the development of the hotspots strategy was the recognition of the advantages to investing in the least threatened – and cheapest – highly biodiverse areas. In fact, Myers (1988) was the first to notice that three endemic-rich regions of tropical forest remained largely intact – he called these “good news areas.” Similarly, Mittermeier (1988) called attention to several high biodiversity tropical rainforest regions that were still in relatively intact condition. He later broadened the concept to address three regions – Amazonia, the Congo Forests of Central Africa, and the island of New Guinea – and referred to them as “major tropical wilderness areas” (McNeely et al. 1990; Mittermeier et al. 1998). While threatened to a much lesser extent than the hotspots, these areas are nevertheless under growing pressure from human activities.

Recently, the emphasis on biodiversity-rich wilderness has been reassessed against the background of all of Earth's wilderness areas, quantitatively defined as retaining at least 70% of their original habitat and holding human population densities of less than five people per square kilometer (Mittermeier et al. 2002, 2003b). This analysis found that while 44% of Earth's land area can still be considered wilderness, only five of these regions (covering just 6.1% of that area) are “high biodiversity wilderness areas” with more than 1 500 plant species as endemics. These are Amazonia, the Congo Forests of Central Africa, the island of New Guinea, the North American Deserts of the Southwestern United States and Northern Mexico, and the Miombo-Mopane Woodlands and Savannas of Southern Africa. Together, they hold 17% of the planet's plants and 8% of terrestrial vertebrates as endemics.

Based on these analyses, Conservation International uses a two-pronged strategy for global conservation prioritization, simultaneously focusing on the threatened and irreplaceable hotspots and on the high biodiversity wilderness areas, which are irreplaceable but still largely intact, and as such represent important conservation opportunities. The decision, at any given point in time, as to whether we should allocate particular resources to a hotspot or to a high biodiversity wilderness area depends on numerous factors, including donor interest, immediate political, economic or social opportunity or need, and other conservation benefits (e.g., protection of major ecosystem services). However, all of these areas – the hotspots and the high biodiversity wilderness areas combined – are on Conservation International's priority list, and the organization's programs employ a strategic mix of both.

A consistent concern for conservation practitioners is that data for aquatic species have yet to be synthesized at a global scale across many taxa and aquatic habitats. It was not until 2002 that the first comprehensive global assessment of conservation priorities for an aquatic system – coral reefs – was published (Roberts et al. 2002). This analysis identified 18 centers of endemism (across four assemblages – 1 700 coral reef fish, all 804 scleractinian coral species, three mollusk families, and 69 lobster species) and highlighted ten of these regions as hotspots facing high threats. Remarkably, the study found that eight of the ten reef hotspots (and 14 of the 18 centers of endemism) lie adjacent to terrestrial hotspots, raising an intriguing possibility that terrestrial hotspots may actually reflect aquatic ones rather well. The publication of Roberts et al. (2002) has attracted much-needed attention to marine hotspots, although data on these areas remains sparse compared with information on terrestrial systems (Lambshead 1993). Our lack of knowledge about freshwater systems is even more pronounced – where even a first look at global conservation prioritization has yet to be carried out. These areas constitute one of the world's most endemic-rich and threatened biomes (McAllister et al. 1997), making such an analysis most urgent.

Impact of Hotspots

The impact of the hotspots concept has been astounding. One measure of this is scientific. Searching the Web of Science for all citations including the word “hotspot” in the title yields numerous scientific papers. While many of these concern geology, astronomy, or genetics (and a few, behavioral ecology or remote sensing), nearly 100 citations use the word “hotspot” to refer to biodiversity conservation priorities. Analyzing these citations over time reveals a clear pattern of gradual increase following Myers’ (1988, 1990) original work, a rapid acceleration with the publication of Myers et al. (2000), and an increasing number of publications on marine hotspots following Roberts et al. (2002). In addition, the number of times that Myers et al. (2000) has been cited in the peer-reviewed scientific literature has shown a steady increase since its publication, and by January 2004 totaled 438 instances.

More importantly, the impact of the hotspots concept in terms of investment in conservation has been dramatic (Myers and Mittermeier 2003). As indicated above, Conservation International adopted hotspots as its central strategy in 1989. In the same year, the Chicagobased John D. and Catherine T. MacArthur Foundation adopted hotspots as its primary global investment strategy (Mittermeier et al. 1998). In 2000, the World Bank and the Global Environment Facility joined Conservation International to establish the Critical Ecosystem Partnership Fund, a conservation finance mechanism focused explicitly on the hotspots (Dalton 2000; www.cepf.net); the MacArthur Foundation and the Japanese Government have since joined the partnership, bringing the total amount available to $125 million. Conservation International's $100-million Global Conservation Fund, supported by the Gordon and Betty Moore Foundation, also uses hotspots (along with high biodiversity wilderness areas) to guide its investments. More than $750 million has been devoted to saving hotspots over the last 15 years, perhaps the largest financial investment in any single conservation strategy (Myers 2003). The hotspots concept has also entered the mainstream as a tool for forward-thinking private sector businesses that have adopted biodiversity conservation policies for their operations and supply chain systems. For example, Office Depot explicitly gives preference to pulp and paper vendors that protect natural forests in the biodiversity hotspots and high biodiversity wilderness areas.

Biodiversity conservation efforts in hotspots often require the ability to withstand and adapt to a rapidly changing socio-political climate. While it can be tempting to write off high-risk areas, experience demonstrates both the importance and the potential for operating and maintaining a conservation presence in hotspots that are undergoing political difficulties. Madagascar, one of the most important hotspots, was almost abandoned by conservationists in the early to mid-1980s, and again during a brief period of political strife in 2001 and 2002. Fortunately, several organizations persevered, notably the World Wildlife Fund, USAID, and the World Bank and, beginning in the early 1990s, Conservation International and the Wildlife Conservation Society. This resolve paved the road for a positive environment for the new President, Marc Ravalomanana, to give conservation a high priority in his government's development plans. At the Fifth World Parks Congress in September 2003, President Ravalomanana committed to tripling the country's protected area network over the next five years, and just five months after this pledge he announced the establishment of 14 new areas increasing coverage by 65%. At the time of his announcement, President Ravalomanana also requested the involvement of the international community in creating a $50-million trust fund for conservation over the next five years; seven months later, a total of $24 million in commitments has already been made. Liberia, one of the most important countries in the heavily impacted Guinean Forests of West Africa Hotspot, has, until very recently, been written off by most of the international conservation community. Nonetheless, Fauna and Flora International and Conservation International operated there through some of the worst periods of instability and violence. Largely because of their efforts, the Liberian Senate in 2003 enacted legislation expanding the country's protected areas network, and the stage is now set to make forest policy reform and conservation a priority for Liberia's reconstruction. Such cases provide excellent illustrations of the conservation return on investment produced by the hotspots strategy.

An Updated Hotspots Analysis

The hotspots analysis is in constant evolution. There are two major ways in which hotspots can change over time. The first is a real effect. Threats and their impacts change, meaning that some places may become more threatened, while others, if conservation efforts are successful, may eventually recover. The second is that our knowledge of biodiversity, threats, and costs is continually improving; new species are discovered, new populations are found, and higher-resolution land cover data is collected. Over the last few years, in concert with the information revolution and the emergence of the Internet, this data has become better compiled (Sugden and Pennisi 2000). Now, several years after the publication of the reassessment of the hotspots strategy (Mittermeier et al. 1999; Myers et al. 2000), it is time to revisit the hotspots themselves in light of new data regarding species distributions and changing conditions of the planet's ecosystems.

We should emphasize from the outset that the current effort is not a reworking of the entire hotspots concept. Rather, the aims of this analysis are to revisit the status of the existing hotspots, refine their boundaries, update the information associated with them and, most importantly, consider a number of potential hotspots that may qualify as additions to the existing list of 25. Consequently, the criteria for what qualifies as a hotspot remain unchanged. There continues to be much debate in the literature concerning total vascular plant diversity, with lower estimates ranging from 270 000 to 320 000 (May 1992; Prance et al. 2000), and higher estimates ranging up to 422 000 (Govaerts 2001; Bramwell 2002). For now, we have retained the lower figure of 300 000 used by Myers et al. (2000), given that the higher figure remains controversial (Thorne 2002).

When Myers et al. (2000) published the results of their analysis, they noted that a number of areas harbored exceptional plant endemism and were also under unusual threat, but were insufficiently documented to meet the hotspots criteria: the Ethiopian Highlands, the Angola Escarpment, southeastern China, Taiwan, and the forests of the Albertine Rift. The Queensland Wet Tropics in northeastern Australia were also mentioned as having a remarkably high species-to-area ratio, but insufficient endemic plant species to qualify as a hotspot. Additional data on Taiwan and the Queensland Wet Tropics now confirm that neither reaches the threshold of being a hotspot. However, because both are globally important and come so close to meeting the criteria, we include special treatment of them in this book (pp. 361 and 369, respectively). Furthermore, investigation of the definition of a hotspot for the rainforests of eastern Australia continues, and it is likely that an expanded interpretation of this region will result in confirmation of its hotspot status in the future. New data also demonstrates the hotspot status of the Ethiopian Highlands and Albertine Rift (discussed below). Finally, the Angola Escarpment and southeastern China remain little known, and thus it is still not possible to ascertain whether or not the regions qualify as hotspots.

One major finding of this updated analysis is that six previously overlooked areas qualify for hotspot status. These are the Madrean Pine-Oak Woodlands of northern Mexico and the southwestern United States, southern Africa's Maputaland-Pondoland-Albany region, the Horn of Africa, the Irano-Anatolian region, the Mountains of Central Asia, and Japan. In addition, one existing hotspot is divided into two. Mittermeier et al. (1998) first suggested the Himalaya and Indo-Burma regions as separate hotspots. In Mittermeier et al. (1999) and Myers et al. (2000) these were combined, but data is now sufficient to show that they contain quite distinctive biotas. That a number of these changes are in Asia is explained partly by the fact that biodiversity data for the continent has historically been less thoroughly synthesized than has data for the Americas and Africa, and partly because much of the data that does exist for key countries such as China and Japan has, at least until recently, been inaccessible to conservation scientists outside of these regions.

An important modification to the hotspots strategy presented here is the reconfiguration of several African hotspots. One problem we have always grappled with is that the combination of the East African Coastal Forests (Burgess and Clarke 2000) with Tanzania's Eastern Arc Mountains (Lovett and Wasser 1993) as a single hotspot is somewhat incongruous biogeographically (Myers et al. 1999). Furthermore, recognition by Myers et al. (2000) of the potentially high levels of endemism in the Ethiopian Highlands and Albertine Rift meant these regions needed to be re-evaluated with better data. It is now apparent that none of the montane areas – the Ethiopian Highlands, Albertine Rift, or Eastern Arc Mountains – qualify as hotspots on their own, because they do not meet the cut-off of 1 500 endemic plants. However, the classic work on African biogeography by Frank White (1983) provides a simple solution. The biogeographic affinities of these regions suggest that they are best considered as a single unit, the Afromontane Region, despite their fragmented geography. Thus, we identify this region as the Eastern Afromontane Hotspot, encompassing the Eastern Arc Mountains and Southern Rift, the Albertine Rift, the Ethiopian Highlands, and a few outliers. This leaves the Coastal Forests of Eastern Africa, running from southern Somalia south through Kenya, Tanzania, and Mozambique, as a unique hotspot in its own right.

The final change revealed in our reassessment of the hotspots is truly terrifying. Less than a decade ago, the islands of eastern Melanesia – the Bismarcks, Solomons, and Vanuatu – while known to be extremely endemic-rich, still held largely intact habitat. Since then, rampant logging and establishment of oil palm plantations have devastated these islands, leaving only 30% of their forests remaining, a situation mirroring the fate of Indonesia's forests a decade ago (Holmes 2000). Thus, the primary cause of the identification of the East Melanesian Islands Hotspot is a worsening threat over a very short period of time.

In revisiting the boundaries of the hotspots, we have tried to achieve a balance between what is scientifically defensible, and what is pragmatically acceptable. The hotspots are based on plant endemism, and so, as far as possible, our decision regarding where or whether to include a particular area or island within a hotspot is determined by the floristic affinities of the region in question. As before, the landmark publication Centers of Plant Diversity (Davis et al. 1994–1997) has been instrumental in guiding and influencing some of our decisions in this regard. However, in some cases, we have seen fit to deviate from this ideal, in order to accommodate tropical islands – many of which have very high proportions of threatened species – that might otherwise slip through the net of conservation priorities. For this reason, we have grouped certain islands with their closest-lying hotspots including, for example, Galápagos and Malpelo with Tumbes-Chocó-Magdalena, Juan Fernández with the Chilean Winter Rainfall-Valdivian Forests, the Azores and Cape Verde Islands (both part of the Macaronesian Islands along with the Canaries and Madeira) with the Mediterranean Basin; and Lord Howe and Norfolk islands with New Zealand. This is done solely for purposes of pragmatic convenience, and with full recognition that the floristic affiliations of these islands with their associated landmasses are tenuous at best.

Synthesis of the Updated Hotspots Data

In total, this updated analysis reveals the existence of 34 biodiversity hotspots, each holding at least 1 500 endemic plant species, and having lost at least 70% of its original habitat extent. Overall, the 34 hotspots once covered a land area of 23 490 101 km², 15.7% of the Earth's land surface, an area equivalent in size to Russia and Australia combined. Their individual areas spanned two orders of magnitude. Three of the regions historically covered more than two million square kilometers each (Indo-Burma, the Mediterranean Basin, and the Cerrado), and a further six, more than a million. The smallest, New Caledonia, covered only 18 972 km², and three others were smaller than 100 000 km². The average original size was 690 885 km², and the median size 385 316 km². This extent of habitat has now been reduced to 3 379 246 km², a mere 2.3% of the planet's land surface. Its size is slightly more than the country of India or a fraction less than the five largest American states combined (Alaska, Texas, California, Montana, and New Mexico = 3 392 950 km²). In all, 86% of the hotspots' habitat has already been lost, and only 14% remains. Table 1 details these statistics hotspot-byhotspot.

Table 1. Original extent, remaining habitat, and percentage of remaining habitat for each hotspot (determined using an equal-area projection) and its predominant biome type (Olson et al. 2001)

The distribution of the hotspots among biomes is greatly skewed towards tropical forests (Table 1). Of the 34 hotspots, 22 (65%) are predominantly tropical forest biomes, ranging from very wet hotspots (like the East Melanesian Islands) to sparsely wooded savanna and grassland (as in the Cerrado). Six hotspots (18%) primarily hold temperate forest, five (15%) Mediterraneantype ecosystems (two of which also have temperate forest elements), and one (3%) is desert.

Among them, the hotspots hold no less than 150 000 plant species as single-hotspot endemics (Table 3). That is 50% of the world's total. By far, the two hotspots with the most endemics are the Tropical Andes and Sundaland, with no less than 15 000 endemic plant species (Table 2). Two other hotspots – the Mediterranean Basin, and Madagascar and the Indian Ocean Islands – also exceed 10 000 endemic plant species; and five more exceed 5 000. The four hotspots richest in endemic plants are also the most speciose, with 20 000 or more plant species occurring in each. Plant numbers per hotspot are derived from specialist estimates rather than from species-by-species lists, which makes it impossible to calculate the total number of endemics or even the number of species occurring in hotspots (Table 3). This is because these estimates do not account for the plants that are shared between hotspots. In other words, if we were to attempt to produce such totals, we would underestimate overall plant endemism by failing to include species confined to multiple hotspots while inflating total hotspot richness by single occurrences counted more than once.

Table 2. Numbers of plant and vertebrate species endemic to (E) and occurring in (O) each of the 34 hotspots (percentages inparentheses)

Table 3. Numbers of plant and vertebrate species endemic to single hotspots, endemic to any hotspot(s), and occurring in any hotspot(s). The first row gives the total global number of species in each group, following Myers et al. (2000) for plants, and the World Wildlife Fund-U.S. database of terrestrial vertebrates in ecoregions and relevant IUCN assessments for terrestrialvertebrates

While data for most invertebrate groups remains sparse, we can produce much more accurate summary statistics for terrestrial vertebrates (Table 2). The growing accuracy of the figures provided for terrestrial vertebrates in the hotspots is largely due to major advances in the reliability of species distribution data. We have relied on two main sets of data here: the IUCN (the World Conservation Union) Red List partnership and data synthesized across terrestrial ecoregions by the World Wildlife Fund-U.S. (Olson et al. 2001). The new data presented for amphibians per hotspot derives entirely from the former, namely the groundbreaking work of the Global Amphibian Assessment. Mammal data is based on work initiated by the Global Mammal Assessment, conducted through the same partnership, and mapped to ecoregions. Distribution data for birds has always been the most advanced of the four terrestrial vertebrate groups, thanks to the pioneering research of BirdLife International, another member of the IUCN Red List partnership, and was likewise expanded to synthesize the distributions of non-threatened species across ecoregions. Reptile data remains poor (only crocodilians, turtles, and tortoises having been relatively well assessed), but a comprehensive online taxonomic reference exists (